This file is the authoritative
register of covariate column names used in nlmixr2lib models. Every
covariate referenced inside a model’s model() block must
have an entry here. The register is seeded from a full audit of
inst/modeldb/ and extended whenever a new paper introduces
a covariate that isn’t yet registered.
covariateData[[name]]$source_name field.covariateData[[name]]$notes whether the values must be
transformed (e.g., SEXM -> SEXF inverts values; the
effect coefficient sign / reference category must be inverted as
well).Each entry has a Scope: field declaring whether it is
general (any model may use it without warning) or
specific (only the models listed under
Example models may use it; other usage is flagged by
checkModelConventions()). This prevents accidentally
reusing a covariate name whose meaning is tied to a particular
paper.
WT, AGE, SEXF,
CREAT, ADA_POS, CRP,
CRCL.Example models list (and consider
promoting to general).When adding or updating an entry, choose the most conservative scope:
if in doubt, start with specific and promote when a second
model legitimately ratifies the name.
Covariate column names should be ALL CAPS. Current non-all-caps
canonical names are dilution and nonECZTRA
(both scope: specific), preserved from their source files with “future
rename” notes. New entries should default to all caps.
- name: <CANONICAL_NAME>
description: <one-sentence definition>
units: <unit string, or "(binary)" / "(categorical)">
type: continuous | binary | categorical | count
scope: general | specific
reference_category: <the 0 group for binary/categorical, or NULL>
source_aliases:
- <ALIAS_NAME> (<transformation if any>) -- used in <model.R>
example_models:
- <model.R>
notes: <free text>(WT / ref_wt)^exponent. Reference weights observed:
70 kg (adults), 75 kg, 84.8 kg, 56 kg (Kloprogge 2014 quinine cohort
typical), 5 kg (infants), 25 kg (Wang 2012 Chinese pediatric epilepsy
cohort median).WEIG – weight column abbreviation used by Wang 2012
(Acta Pharmacol Sin 33:845-851); same biological quantity in kg, no
value transformation. Used in
Wang_2012_levetiracetam.R.Clegg_2024_nirsevimab.R,
Hu_2026_clesrovimab.R,
Zhu_2017_lebrikizumab.R,
Kovalenko_2020_dupilumab.R,
CarlssonPetri_2021_liraglutide.R,
Cirincione_2017_exenatide.R,
Grimm_2023_gantenerumab.R,
Grimm_2023_trontinemab.R,
Kyhl_2016_nalmefene.R,
Soehoel_2022_tralokinumab.R,
Xie_2019_agomelatine.R,
PK_2cmt_mAb_Davda_2014.R,
phenylalanine_charbonneau_2021.R,
Chua_2025_mirikizumab.R,
Jackson_2022_ixekizumab.R,
Kotani_2022_astegolimab.R,
Ma_2020_sarilumab_anc.R,
Ma_2020_sarilumab_das28crp.R,
Moein_2022_etrolizumab.R,
Tiraboschi_2025_amlitelimab.R,
Robbie_2012_palivizumab.R,
Bajaj_2017_nivolumab.R,
Quartino_2019_trastuzumab.R,
Wang_2020_ontamalimab.R,
Fau_2020_isatuximab.R,
Okada_2025_rocatinlimab.R,
Kunisawa_2014_olprinone.R,
Xu_2020_daratumumab.R (reference 78.6 kg; power exponents
0.451 on linear CL and 0.375 on V1),
Struemper_2017_belimumab.R (reference 67 kg; fixed
allometric exponents 0.75 on CL and Q, 1.00 on Vc, 0.8 on Vp;
baseline-only, source column BWT),
MedellinGaribay_2015_gentamicin.R (linear (not allometric)
weight scaling on both CL and Vc: CL = theta1 * BW + theta5 * (CRCL/75),
Vc = theta2 * BW; no reference weight used because the scaling is
linear, not divisive; source column BW; cohort mean 6.4 +/- 2.2 kg,
infants 1-24 months), Wahlby_2004_pefloxacin.R
(time-varying; reference 65 kg; linear-exp form on V with coefficient
0.014).Archary_2019_lamivudine.R,
Budha_2023_tislelizumab.R,
Chakraborty_2012_canakinumab.R,
Chen_2020_luspatercept.R,
Conrado_2014_alzheimer.R,
Diepstraten_2013_propofol.R,
Gandhi_2021_abatacept.R,
Goel_2016_Sonidegib.R, Hennig_2013_tobra.R,
Hong_2025_datopotamab.R,
Ide_2020_elotuzumab.R,
Koopman_2023_factorix.R,
Kuchimanchi_2024_dostarlimab.R,
Kunarajah_2017_doxorubicin.R,
Kyhl_2016_nalmefene.R,
Lahu_2010_roflumilast.R, Li_2006_meropenem.R,
Li_2017_cediranib.R, Li_2019_abatacept.R,
Lin_2024_casirivimab.R,
Martinez_2019_alirocumab.R,
Melhem_2022_dostarlimab.R,
Mulyukov_2018_ranibizumab.R,
NA_NA_tte_gompertz.R, NA_NA_tte_lognormal.R,
Retlich_2015_linagliptin.R,
Rosario_2015_vedolizumab.R,
Svensson_2016_bedaquiline.R,
Thakre_2022_risankizumab.R,
Wu_2024_inotuzumab.R,
Yassen_2025_asundexian.R,
Yu_2022_ofatumumab.R, Zhong_2026_abatacept.R,
Zhou_2021_belimumab.R,
Zhu_2017_lebrikizumab.R.AGE/40.LBW (lean body weight) – synonym; same biological
quantity (total body weight minus body fat). Hemophilia popPK literature
typically uses LBW (Hume or James formula) where mAb /
general literature uses LBM. Used in
Garmann_2017_BAY81_8973.R (reference 51.1 kg).Kyhl_2016_nalmefene.R
(reference 56.28 kg, exponent 0.626 on CL),
Garmann_2017_BAY81_8973.R (alias LBW;
reference 51.1 kg, exponents 0.610 on CL and 0.950 on Vc),
Schoemaker_2017_brivaracetam.R (alias LBW;
paediatric cohort, reference 50 kg adult typical value, fixed
theoretical allometric exponents 0.750 on CL/F and 1.00 on V/F).(FFM / ref)^exponent. Reference values observed: 40.69 kg
(Zhou 2021 belimumab pooled adult+pediatric SLE), 45 kg (Aguiar 2021,
Crohn’s disease cohort median).FFM is the
universal abbreviation.Zhou_2021_belimumab.R
(reference 40.69 kg; exponents 0.673 on CL and 0.891 on V1),
Aguiar_2021_ustekinumab.R (reference 45 kg; power exponents
0.598 on CL, 0.590 on Vc, 0.586 on Vp).LBM (lean body
mass) which is sometimes computed by the Boer or Hume formulae. When the
source paper reports the body-composition formula it used (e.g.,
Janmahasatian for FFM), record it in
covariateData[[FFM]]$notes. FFM is preferred over total
body weight when scaling monoclonal-antibody PK because mAb distribution
is largely confined to extracellular fluid; muscle / lean tissue tracks
extracellular volume better than total weight in heavier patients.(IBW / ref) for clearance / dose-rate normalisation.
Reference values observed: 60 kg (Holford 1992 tacrine adult Alzheimer’s
population mean IBW).IBW – standard abbreviation used directly in Holford
1992.Holford_1992_tacrine.R
(reference 60 kg; tacrine “clearance” relative to IBW = 60 enters as the
dose-rate normalisation factor (60 / IBW) driving the
tacrine effect-compartment input; the Holford-Peace Devine variant is
documented in covariateData[[IBW]]$notes),
VanWart_2004_garenoxacin.R (reference 64.2 kg; additive
linear-deviation effect 0.764 mL/min per (IBW - 64.2) kg on CL/F, plus
an in-model obesity flag derived as WT > 1.3 * IBW per
Van Wart 2004’s >130% IBW obesity definition),
Shoji_2011_pregabalin.R (reference 62 kg; power-of-ratio
scaling on CL/F (exponent 0.354) and V/F (exponent 0.819) per Shoji 2011
Table 3 final model; the IBW formula is the Devine-family variant cited
as reference [12] of the paper).general.
Per-model covariateData[[IBW]]$notes should record the
formula the source paper used. The Holford-Peace 1992 variant is: men
IBW (kg) = 52 + 0.75 * (height_cm - 152); women IBW (kg) = 49 + 0.67 *
(height_cm - 152). The classic Devine 1974 formula is: men 50 + 2.3 *
(height_in - 60); women 45.5 + 2.3 * (height_in - 60). Other variants
(Robinson 1983, Miller 1983, Hamwi 1964) exist; the per-paper choice
should be recorded so a user simulating against IBW can match the
source’s derivation. When the source dataset supplies IBW pre-computed,
the column name is typically IBW directly. When only
HT + SEXF are provided, the user must compute
IBW externally using the source-paper formula before passing it to the
model.(HT - ref) or with a power-style
scaling. Reference values observed: 167 cm (Naik 2016, vortioxetine
adult MDD/GAD population median); 165 cm (Zhang 2018 flurbiprofen and
Angeli 2016 healthy non-menopausal women).HGT – height (cm) abbreviation appearing in some NONMEM
control streams.HEIGHT – spelled-out form (used by Angeli 2016).Naik_2016_vortioxetine.R (reference 167 cm; linear-additive
effect 0.40 L/hr per (HT - 167) cm on CL/F, retained over weight and BMI
in stepwise selection because it produced the larger reduction in CL
IIV), Zhang_2018_flurbiprofen.R (reference 165 cm;
linear-multiplicative effect 1 + theta_height * (HT - 165)
on the effect-compartment equilibration rate Ke alongside a paired
linear-multiplicative WT effect),
Angeli_2016_iron_hepcidin.R (reference 165 cm; power-law
multiplier (HT / 165)^32.70 on the hepcidin post-menses
rebound parameter krel_hep; the very large exponent
reflects the narrow height range across the cohort, 158-173 cm at the
10th-90th percentile), Wu_2014_FEV1_asthma.R
(per-observation HT in cm enters directly into the exponential FEV1
disease-progression model:
FEV1 = exp(theta_age*AGE + theta_ht*HT - theta_int) + ...
with theta_ht = 0.0169 1/cm; paediatric population, ages
5-13 y).HT alongside one of those derived covariates should
document the dependency in covariateData[[HT]]$notes.
Promoted from specific to general on
2026-06-03 with the Angeli 2016 iron / hepcidin extraction (the third
adult-popPK model to register HT, satisfying the original promotion
condition documented when the canonical was introduced).(BSA / ref)^exponent.Yamada_2025_zolbetuximab.R (reference 1.70 m^2; exponents
1.06 on clearances and 0.968 on volumes).1 + e * (BMI - ref)) or a power form
((BMI / ref)^e). Document the reference value in
covariateData[[BMI]]$notes.Chua_2025_mirikizumab.R (reference 24.75 kg/m^2;
linear-deviation effect on logit of bioavailability),
NA_NA_lidocaine.R (DDMODEL00000281; binary stratification
at threshold 27.93 kg/m^2 adding +0.939 to the GX rate constant K30 in
the BMI > 27.93 cohort), Struemper_2017_belimumab.R
(kg/m^2, reference 24.7; power exponent -0.610 on Vc; baseline-only,
source column BBMI).WT / (height_m)^2; assume time-fixed at baseline
unless the source paper states otherwise.BMI (kg/m^2): BMIZ is
unitless and centred at 0 in the reference population, so the reference
value used in linear-deviation effects is 0 (not a population BMI in
kg/m^2). Time-varying when the source paper carries a per-visit z-score;
document baseline-vs-time-varying status in
covariateData[[BMIZ]]$notes.(1 + e * (BMIZ - 0)) so the reference
is 0 (population mean for the subject’s age/sex). Effect coefficients
are interpreted as fractional change per 1 z-score-unit deviation from
the reference.BMI – when a paper uses the column name
BMI for what is actually a z-score (e.g., the Harun 2019
NMTRAN control stream column BMI is documented in the
dataset header as “body-mass index z-score”). The canonical column is
BMIZ; the source-paper column name is recorded in
covariateData[[BMIZ]]$source_name.Harun_2019_cysticFibrosis.R (time-varying per-visit BMI
z-score; linear-deviation effect on baseline FEV1% predicted with
reference 0 and coefficient +0.0382 per z-score unit).BMI (raw kg/m^2
used in adult populations). Paediatric and adolescent studies routinely
report BMI as a z-score relative to a growth reference (WHO 2007 Growth
Reference for school-aged children, CDC 2000, etc.); document the
reference standard the source paper used in
covariateData[[BMIZ]]$notes. Specific scope until a second
paediatric model ratifies the name; at that point promote to
general.SEXM (values inverted: SEXF = 1 - SEXM;
effect coefficient sign and reference category both invert) – used in
CarlssonPetri_2021_liraglutide.R.SEX with "M"/"F" strings –
derive SEXF = as.integer(SEX == "F").SEX with 1=male / 2=female
numeric coding – derive SEXF = as.integer(SEX == 2). Used
in Netterberg_2017_docetaxel.R and
NA_NA_miridesap.R (DDMODEL00000262 source bundle; Sahota
2015 NONMEM convention).FEM (1 = female, 0 = male; same orientation as
canonical, no transformation) – used in
Guiastrennec_2016_gastric_emptying.R.Zhu_2017_lebrikizumab.R (canonical),
CarlssonPetri_2021_liraglutide.R (alias SEXM),
Bajaj_2017_nivolumab.R (male-indicator source; effect
applied as exp(coef * (1 - SEXF)) to preserve the paper’s
female-reference CL_REF / VC_REF), Fau_2020_isatuximab.R
(exponential effect on Vc; reference category 0 = male),
Netterberg_2017_docetaxel.R (multiplicative effect on
baseline ANC: BACOV *= (1 + theta * SEXF); source column
SEX with 1 = male, 2 = female encoding, decomposed via
SEXF = as.integer(SEX == 2)),
NA_NA_miridesap.R (DDMODEL00000262 / Sahota 2015;
multiplicative effect on baseline SAP via
SAP_BASE_ref * (1 + e_sexf_sap0 * SEXF) with
e_sexf_sap0 = -0.30; female baseline is ~30% lower than
male), Xu_2020_daratumumab.R (additive shift on V1
(1 + e_sexf_vc * SEXF) with
e_sexf_vc = -0.205: female V1 is 20.5% lower than male,
reference category 0 = male),
Guiastrennec_2016_gastric_emptying.R (multiplicative +40.7%
strengthening of the caloric-feedback slope SLPCAL on gastric emptying
in females; SLPCAL_eff = SLPCAL * (1 + 0.407 * SEXF)).SEXM, flag the sign/reference-category inversion to the
user.PREG directly.Birgersson_2019_artesunate.R (multiplicative effect on
dihydroartemisinin clearance; the published structural CLM = 190 L/h is
reported with the source-paper reference category PREG = 1, so the model
file applies the effect via
(1 + e_preg_cl_dha * (1 - PREG)) with
e_preg_cl_dha = -0.214 to preserve verbatim source values;
non-pregnant women have ~21% lower CLM relative to pregnant women).GA or
a trimester indicator, ratified separately when needed). The canonical
convention is reference category 0 (non-pregnant) following the broader
pharmacology default; source papers that use the pregnant cohort as the
reference (Birgersson 2019) preserve their published structural values
via a (1 - PREG) form on the effect coefficient. Ratified
canonically on 2026-05-07.PED – used in the Schoemaker 2018 LEV / BRV pediatric
extrapolation (DDMODEL00000239) as the pediatric-vs-adult indicator that
gates the Markov-amplitude term, the overdispersion IIV, and the four
pediatric offsets on log baseline rate / mixture / placebo / Emax /
EC50.CarlssonPetri_2021_liraglutide.R,
Schoemaker_2018_levetiracetam.R (DDMODEL00000239).ADOLESCENT; paper’s age cutoffs must be captured in
covariateData[[CHILD]]$notes.CarlssonPetri_2021_liraglutide.R.CHILD. Document age
cutoffs.(FM / ref)^exponent. Reference values observed: 19 kg
(Robarge 2017 efavirenz; cohort median calculated FM).FAT,
FATMASS.Robarge_2017_efavirenz.R (reference 19 kg; exponent 1 fixed
on V_p/F).FFM. Use
both in the same model when the paper scales different disposition
parameters by FFM vs FM (e.g., Robarge 2017 efavirenz: CL/F by FFM,
V_p/F by FM, motivated by lipophilic partitioning into adipose tissue).
When the source paper reports the body-composition formula it used
(Janmahasatian for FFM is the most common; direct DXA / BIA also
acceptable), record it in covariateData[[FM]]$notes.covariateData[[VISCERAL_ABDOMINAL_FAT]]$units when a
paper uses a different anatomical level (L3 / L4, L4 / L5) or a
different summary form (volumetric L1-L5 cm^3, MRI-derived).log(SI) = log(SI_ref) + theta * (VISCERAL_ABDOMINAL_FAT - ref).
Reference values observed: 141.8 cm^2 / CT-slice (Denti 2010 Table 1
pooled-population mean across 204 healthy adults aged 18-87).VAF – universal abbreviation in the IVGTT /
glucose-metabolism literature (Denti 2010, Basu 2003, Basu 2006). The
canonical column is VISCERAL_ABDOMINAL_FAT; the source
paper’s column name is recorded in
covariateData[[VISCERAL_ABDOMINAL_FAT]]$source_name.Denti_2010_glucoseMinimal.R (introduces canonical;
reference 141.8 cm^2 / CT-slice; linear-in-log effect coefficient
-0.00208 per cm^2 deviation on insulin sensitivity
lsi).covariateData[[VISCERAL_ABDOMINAL_FAT]]$notes. Distinct
from BODYFAT_PCT (whole-body adiposity fraction by DEXA)
and from FFM / LBM (fat-free mass / lean body
mass). The H3-name uses underscores to preserve readability; future
authors may abbreviate the source column as VAF in their
own datasets and declare source_name = "VAF".
Operator-ratified 2026-06-07 via the Denti 2010 extraction sidecar.covariateData[[BODYFAT_PCT]]$units when a paper uses a
fractional 0-1 scale.log(VD) = log(VD_ref) + theta * (BODYFAT_PCT - ref).
Reference values observed: 32.39 % (Denti 2010 Table 1 pooled-population
mean across 204 healthy adults).%TBF – paper symbol in the IVGTT / glucose-metabolism
literature (Denti 2010, Basu 2003). The leading % character
is dataset-incompatible (NMTRAN / R column-name rules); the canonical
column is BODYFAT_PCT and the paper’s symbol is recorded in
covariateData[[BODYFAT_PCT]]$source_name.PTBF – numeric-safe rendering of the paper symbol used
in some derived datasets.Denti_2010_glucoseMinimal.R (introduces canonical;
reference 32.39 %; linear-in-log effect coefficient -0.0101 per
percentage-point deviation on glucose distribution volume
lvd).%BF from future papers
can ratify under the same canonical with the assay method recorded
per-model in covariateData[[BODYFAT_PCT]]$notes. Distinct
from FFM (fat-free mass in kg) and from
VISCERAL_ABDOMINAL_FAT (regional visceral adiposity in cm^2
/ CT slice). The negative effect coefficient on glucose distribution
volume captures the physiological observation that glucose distributes
preferentially into lean rather than adipose tissue, so per-kg
distribution volume decreases as fat fraction rises. Operator-ratified
2026-06-07 via the Denti 2010 extraction sidecar.WT column. Operationally,
WT_BASE is equal to WT at the subject’s first
observation and remains constant for all subsequent records; the
within-subject delta is computed in model() as
(WT - WT_BASE).(WT_BASE - ref) in an exponential or linear-deviation
effect on a structural parameter, where ref is the cohort
median baseline weight.BWT (baseline weight) – Wahlby 2004 source-column
convention; used in Wahlby_2004_pefloxacin.R.Wahlby_2004_pefloxacin.R (reference 65 kg; saturating “up
to median WT” qualifier from Wahlby 2004 Methods plateaus the effect
above 65 kg, encoded as min(WT_BASE, 65) - 65).WT_BASE exists
conceptually as a per-subject snapshot of baseline weight that the
original modeller chose to enter separately from the time-varying weight
column. Promote to general if a second paper ratifies the
same baseline-weight column with consistent semantics. Distinct from
time-fixed adult-cohort weight (where the per-subject weight is just
constant by data design); use WT_BASE only when the source
paper deliberately separates the BCOV from the time-varying COV.
Ratified canonically alongside the Wahlby 2004 extraction.BBSA (baseline BSA) – Wahlby 2004 source-column
convention; used in Wahlby_2004_gentamicin.R.Wahlby_2004_gentamicin.R (replaces time-varying BSA in the
final-model V1 equation V1 = 8.63 * BSA_BASE * (ALB/34)^-0.41 because
BBSA was the better predictor than time-varying BSA, RSE 2.5% vs 110% on
delta-BSA).general if a second paper
ratifies the baseline-BSA column with consistent semantics. Ratified
canonically alongside the Wahlby 2004 extraction.GA_weeks / 4.35 + postnatal_months). Time-varying.Clegg_2024_nirsevimab.R,
Robbie_2012_palivizumab.R.PNA – used in Zhao 2018 (paper Methods ‘Population
pharmacokinetic-pharmacogenetic modelling’ and Table 2 report PNA in
DAYS; the canonical PNA carries months, so Zhao 2018’s
F_PNA = (PNA_days / 38)^0.472 is reparameterised inside
model() as F_PNA = (PNA_months / 1.249)^0.472
using the conversion PNA_months = PNA_days / 30.4375 and
reference 1.249 months = 38 days / 30.4375).Hu_2026_clesrovimab.R,
Zhao_2018_omeprazole.R (power effect on the formation
clearance of 5-hydroxy-omeprazole: (PNA / 1.249)^0.472; PNA
reference 1.249 months / 38 days from Zhao 2018 Table 2 cohort
median).Hu_2026_clesrovimab.R,
Clegg_2024_nirsevimab.R (folded into PAGE).WT), which is
time-varying.(1 + e * (WT_BIRTH - ref)) or power
scaling. Reference value observed: 2.59 kg (Voller 2017 newborn
cohort).BWEIGHT – used in
Voller_2017_phenobarbital.R (Voller 2017 source data column
for birth weight in kg).Bwb – used in Wu_2025_paracetamol.R (Wu
2025 source-paper symbol for birth weight in kg; same biological
quantity, no value transformation).Voller_2017_phenobarbital.R (linear-deviation effect on CL:
clbw = 1 + 0.369 * (WT_BIRTH - 2.59)),
Wu_2025_paracetamol.R (power-law scaling of PTNA CL_birth
and GFR_birth terms with reference 1.75 kg).GA (gestational age
at birth) when both are reported. The conventional clinical-PK
abbreviation BWT is intentionally NOT used as the canonical
name because it is already used across the codebase (Gandhi 2021, Li
2019, Chen 2022, Wojciechowski 2022, Lu 2019) as a source-name alias for
body weight (WT). The WT_BIRTH form keeps the
WT root consistent with the existing body-weight canonical
and avoids the BWT ambiguity.IF (AGE_DPF > 3) k12 = k12_3 * (1 + e_age_dpf_k12)) and
a per-day power-form effect on the elimination rate
(k25 = k25_3 * (1 + e_age_dpf_k25)^(AGE_DPF - 3)).
Reference age is 3 dpf (the youngest cohort in the study).AGE – van Wijk 2019 NONMEM column. Renamed to canonical
AGE_DPF because the human-PK canonical AGE
denotes subject age in years; the zebrafish dpf semantic is incompatible
and would silently corrupt any future model that mixed them. Same
orientation, no value transformation.vanWijk_2019_paracetamol.R.AGE (human age in years), PNA
(postnatal age in months), PAGE (postmenstrual age in
months), and GA (gestational age in weeks) – none of those
are appropriate for a non-mammalian organism whose developmental clock
is anchored at fertilization rather than birth. Integer values 3, 4, 5
in the van Wijk 2019 dataset, but treated as a continuous covariate in
the elimination-rate power form. Future zebrafish-PK or other
non-mammalian-developmental-age models should reuse this canonical only
when the covariate is indeed dpf-anchored; other developmental-time
conventions (e.g., somite-stage, hpf, dph) would warrant separate
canonicals. Ratified canonically on 2026-05-07.1 =
5-minute Apgar score was less than 5 (i.e., the neonate met the paper’s
perinatal-asphyxia criterion), 0 = 5-minute Apgar score was
5 or higher. Time-fixed per subject (an event at birth, not a
time-varying physiologic state). Used by neonatal popPK models that
report a structural-PK shift in subjects who experienced perinatal
asphyxia relative to those who did not.V = V_typical * (1 + e_asphyxia_vc * ASPHYXIA) with
e_asphyxia_vc = 0.13 (+13% V in subjects with 5-minute
Apgar < 5; p < 0.05).asphyxia in prose; the abstract does not report the source
NONMEM column label).Grasela_1985_phenobarbital.R (Grasela & Donn 1985;
preterm-neonate phenobarbital popPK; +13% on V when 5-minute Apgar <
5; no detected effect on CL).APGAR1_LT7) rather than reusing
ASPHYXIA with relaxed semantics. Ratified canonically
alongside the Grasela 1985 phenobarbital extraction.T_NUT_SUPP column rather than as a time-varying
value of MAL_NOURISH.MAL – used in Tikiso_2021_abacavir.R (the
dataset’s paper-defined indicator, 1 = malnourished, 0 = not
malnourished).NUT – used in
Catalan-Latorre_2018_taurine_rat.R (the dataset’s
paper-defined indicator, 1 = undernourished UN, 0 = well-nourished
WN).Tikiso_2021_abacavir.R
(gates the time-decaying malnutrition effect:
mal_decay = MAL_NOURISH * exp(-T_NUT_SUPP * log(2) / 12.2),
which then drives multiplicative shifts of +115% on F and
-64% on CL at the start of nutritional supplementation,
decaying with a 12.2-day half-life).
Tikiso_2021_abacavir.R (gates the time-decaying
malnutrition effect:
mal_decay = MAL_NOURISH * exp(-T_NUT_SUPP * log(2) / 12.2),
which then drives multiplicative shifts of +115% on F and
-64% on CL at the start of nutritional supplementation,
decaying with a 12.2-day half-life).Catalan-Latorre_2018_taurine_rat.R (static baseline
indicator – no T_NUT_SUPP pairing because there was no
nutritional rehabilitation phase in the preclinical Wistar-rat study;
MAL_NOURISH = 1 reduces the saturable tubular secretion
Vmax of taurine by 9.4% relative to well-nourished animals).covariateData[[MAL_NOURISH]]$notes must document the
criterion used. Pairs with T_NUT_SUPP (days on nutritional
supplementation) when the model uses a time-decaying recovery function;
otherwise MAL_NOURISH alone serves as a static baseline
indicator. Distinct from generic body-weight Z-scores (which are
continuous anthropometric metrics rather than a binarised malnutrition
indicator).exp(-T_NUT_SUPP * log(2) / T_half) whose half-life
T_half is an estimated parameter (12.2 days in Tikiso
2021).TNUTRI – used in Tikiso_2021_abacavir.R
(the dataset’s paper-defined column for days since start of nutritional
supplementation; same orientation as the canonical, 0 = start of
supplementation, increasing with time on supplementation).Tikiso_2021_abacavir.R
(paired with MAL_NOURISH; drives the recovery decay
exp(-T_NUT_SUPP * log(2) / 12.2) of the malnutrition effect
on F and CL).MAL_NOURISH == 0) the value is irrelevant because the
malnutrition effect is gated by MAL_NOURISH; supply 0 as a
default. For fully-recovered malnourished subjects, supply a large value
(e.g., >= 100 days, well beyond the 12.2-day Tikiso 2021
half-life) so the decay function reaches near zero and the effect
vanishes.DRINK_FAT – both indicators
cannot be 1 simultaneously.Guiastrennec_2016_gastric_emptying.R (selects the
OGTT-specific half-onset time T50OGTT = 15.7 min for the
gastric-emptying delay Hill function; water is recovered when DRINK_OGTT
= DRINK_FAT = 0 with the onset factor pinned to 1).DRINK_FAT: the two indicators jointly select the
appropriate gastric-emptying-delay T50 parameter (T50OGTT vs T50Fat) for
the Hill onset function. Ratified canonically alongside the Guiastrennec
2016 gastric-emptying / CCK / GBE extraction.DRINK_OGTT – both
indicators cannot be 1 simultaneously.Guiastrennec_2016_gastric_emptying.R (selects the
fat-specific half-onset time T50Fat = 23.1 min for the gastric-emptying
delay Hill function).DRINK_OGTT: the two indicators jointly
select the appropriate gastric-emptying-delay T50 parameter (T50OGTT vs
T50Fat) for the Hill onset function. Ratified canonically alongside the
Guiastrennec 2016 gastric-emptying / CCK / GBE extraction.(TPP^gamma) / (TPP^gamma + T50^gamma) whose
T50 (weeks-to-50%-recovery) and gamma (shape
parameter) are estimated. For pregnant visits (TPP = 0) the
sigmoid evaluates to 0 and the pregnancy-state PK parameter is
unchanged; for far-postpartum visits (TPP >> T50) the
sigmoid approaches 1 and the parameter reaches its asymptotic
non-pregnant value.TPP – de Kock 2017 NONMEM column for time after
delivery in weeks; same orientation, no transformation; assigned 0 for
samples collected during pregnancy.deKock_2017_sulfadoxinePyrimethamine.R (sigmoidal effect on
sulfadoxine CL with asymptotic fractional change -0.757, T50 = 6.35
weeks, gamma = 4.90; the sigmoid approaches its asymptote ~13 weeks
postpartum, consistent with the literature for return of GFR and renal
blood flow to prepregnant values within 6-12 weeks postpartum).PREG (pregnancy
status indicator) when the source paper models pregnancy as a step
contrast on one PK parameter and as a sigmoidal time-decay on another.
During pregnancy TPP = 0; after delivery
TPP > 0. Document the postpartum sampling window in
covariateData[[TPP]]$notes per model. Distinct from
T_NUT_SUPP (time on nutritional supplementation, days) and
GA (gestational age at birth, weeks) – those are different
timescale covariates anchored to different events. Ratified canonically
on 2026-05-18 alongside the de Kock 2017 sulfadoxine/pyrimethamine
extraction.1 = term birth (>= 37 weeks gestation),
0 = preterm birth (< 37 weeks gestation). Time-fixed per
subject. In Allegaert 2015 (paracetamol PK in young women) the indicator
is used to select between two typical-value clearances for the
sulphate-formation pathway.CL = TERM_BIRTH * theta_term + (1 - TERM_BIRTH) * theta_preterm
selection – so neither category is “the multiplicative reference”, and
both per-stratum clearances are estimated parameters.TERM (Allegaert 2015 NONMEM column; same orientation,
no transformation) – used in
Allegaert_2015_paracetamol.R.Allegaert_2015_paracetamol.R.GA (continuous
gestational age in weeks): TERM_BIRTH is the binarized
version with the conventional 37-week cutoff. Use GA when
the source paper carries gestational age as a continuous covariate; use
TERM_BIRTH only when the paper itself dichotomizes. Do not
derive TERM_BIRTH from GA programmatically
inside model() – the term-cutoff convention belongs in data
assembly, not the model file.1 = currently taking an oral
contraceptive (estrogen-progestin or progestin-only pill),
0 = not on hormonal contraception. Time-varying as women
cycle on/off contraception across study occasions.CL_glucuronide *= theta_CONMED_BIRTHCONTROL when
CONMED_BIRTHCONTROL == 1, with
theta_CONMED_BIRTHCONTROL = 1.46 (estrogen-driven UGT2B7
induction).BC (Allegaert 2015 NONMEM column; same orientation, no
transformation) – used in
Allegaert_2015_paracetamol.R.OC (Csajka 2005 NONMEM column; same orientation, no
transformation) – used in
Csajka_2005_ephedrine_caffeine.R.Allegaert_2015_paracetamol.R,
Csajka_2005_ephedrine_caffeine.R,
Angeli_2016_iron_hepcidin.R (multiplicative effect on iron
elimination kout_iron; women on oral contraception have
~18% lower iron elimination, consistent with the paper’s narrative that
contraception limits menstrual blood loss;
e_conmed_birthcontrol_kout_iron = -0.20).CONMED_BIRTHCONTROL_COMBINED,
CONMED_BIRTHCONTROL_PROGESTIN). The full-word canonical
name was chosen over a shorter BC_USE form for clarity in
source traces.1 = eoPE diagnosed before 34 weeks
gestation, 0 = not eoPE. Time-fixed per subject within the
gestational PK study window. Used by population PK models that compare
drug disposition in pregnant women with vs without early-onset
pre-eclampsia.CL_eoPE = CL * ThetaPE^DIS_EOPE with
ThetaPE = 0.617 (38% reduction in betamethasone CL when
eoPE is present); encoded in nlmixr2 as the log-additive shift
cl <- exp(lcl + etalcl + e_eope_cl * DIS_EOPE) with
e_eope_cl = log(0.617).PE – common abbreviation in obstetric pharmacology
papers when the cohort restriction is to early-onset PE only; used in
Schoenmakers_2025_betamethasone.R (paper notation:
eoPE / ThetaPE).Schoenmakers_2025_betamethasone.R (multiplicative effect on
CL, encoded via the log-additive form on lcl; reduces
apparent betamethasone clearance from 15.6 L/h to 9.6 L/h).PREECL
indicator that would pool early-onset, late-onset and postpartum
pre-eclampsia. The “early-onset” specifier corresponds to diagnosis
before 34 weeks gestation, the conventional clinical cutoff (Phipps 2019
Nat Rev Nephrol). Future papers that enrol mixed early-/late-onset
cohorts or that report PE status without the 34-week stratification
should register a separate canonical (e.g., PREECL for
any-onset PE, or LOPE for late-onset PE) rather than
reusing DIS_EOPE with relaxed semantics. Distinct from
PREG (pregnancy status indicator): DIS_EOPE is
a complication-of-pregnancy stratifier within a pregnant cohort, whereas
PREG discriminates pregnant-vs-non-pregnant subjects.
Ratified canonically on 2026-05-11 alongside the Schoenmakers 2025
betamethasone extraction.(SCORE_HIGHAM / tHIGHAM)^exponent on hepcidin
synthesis and elimination rate constants in Angeli 2016, with the
reference value tHIGHAM set near the population median
(96.6 in the Angeli 2016 HEPMEN cohort).HIGHAM – prior canonical name (pre-2026-06-19
standardization audit).HIGHAM – prior canonical name (pre-2026-06-19 SCORE_
family standardization).HiS – Angeli 2016 paper notation; same orientation, no
transformation.Angeli_2016_iron_hepcidin.R (per-subject mean of the three
most recent cycles’ Higham scores; enters ksynH and koutH via power-law
effects with exponents 0.66 and 0.83 respectively).MENSTRUAL_BLOOD_LOSS_ML
canonical). The threshold for menorrhagia is 100 (Higham 1990 BJOG
97:734); the Angeli 2016 cohort had a mean of 96.6 with SD 60.5 (Table
I), so most subjects sat near the menorrhagia threshold. Ratified
canonically on 2026-06-03 alongside the Angeli 2016 iron and hepcidin
extraction. Renamed from HIGHAM to
SCORE_HIGHAM on 2026-06-19 per the canonical-register
standardization audit (clinical-instrument SCORE_ family
normalization).t < DLOSS. Reference values observed: Angeli 2016
inclusion criterion required menses length between 3 and 5 days (Methods
p. 491); the per-subject value was fixed to the observed individual
menses length.dloss – Angeli 2016 paper notation; same orientation,
no transformation.Angeli_2016_iron_hepcidin.R (per-subject menses length in
days; bounds the loss phase during which kloss adds to both
kout_iron and kout_hep).TPP
(time postpartum, weeks; unbounded postpartum time after delivery) and
from a hypothetical CYCLE_LENGTH covariate (full menstrual
cycle length in days, which would warrant a separate canonical).
Ratified canonically on 2026-06-03 alongside the Angeli 2016 iron and
hepcidin extraction.1 = sample taken on the day of delivery;
0 = otherwise (pregnant non-delivery, or non-pregnant). In
Hirt 2007 the indicator is mutually exclusive with
PREG = 1: per the paper, on the day of delivery the coding
was PREG = 0 and DAY_DELIVERY = 1, so the two
indicators jointly partition women into non-pregnant (PREG = 0,
DAY_DELIVERY = 0), pregnant non-delivery (PREG = 1, DAY_DELIVERY = 0),
and delivery (PREG = 0, DAY_DELIVERY = 1) cohorts. Time-fixed per
sampling occasion.DEL – Hirt 2007 NONMEM symbol for the day-of-delivery
indicator; same orientation, no value transformation.Hirt_2007_nelfinavir.R
(multiplicative effect 1 + 1.92 * DAY_DELIVERY applied to
both maternal nelfinavir clearance CL_Nm_No and
distribution volume V; gates a body-weight power effect
(WT/73)^2.81 on CL_Nm_No within the delivery
cohort only).PREG (pregnant
non-delivery, 1 = pregnant; see entry above), TPP (time
postpartum, a continuous time-since-delivery covariate for postpartum
recovery), and TERM_BIRTH (term-vs-preterm birth at any
postnatal time). Use DAY_DELIVERY when the source paper
carries a contrasted day-of-delivery cohort against pregnant
non-delivery women in the same PK analysis (placental-transfer /
labour-PK studies). Specific scope until a second model ratifies the
name; promote to general when a future placental-transfer or labour-PK
paper uses the same indicator. Ratified canonically alongside the Hirt
2007 nelfinavir extraction.covariateData[[HR]]$notes per model.(HR / ref)^exponent or linear-deviation forms
(1 + e * (HR - ref)). Reference values observed: 158
beats/min (Ngamprasertwong 2016; population reference encoded in the
Table 2 equation CL = theta1 * (HR/158)^theta2).HR – same orientation as the canonical, no value
transformation; used in
Ngamprasertwong_2016_propofol_sheep.R (per-subject median
HR over the propofol-infusion observation window, treated as time-fixed
in line with the cohort-typical sheep hemodynamic state).Ngamprasertwong_2016_propofol_sheep.R (power effect on
maternal propofol clearance:
CL_indiv = theta1 * (HR/158)^theta2 with
theta2 = 0.764; clearance increases with heart rate,
plausibly reflecting heart-rate-driven increases in hepatic blood flow
that govern propofol’s high hepatic-extraction-ratio elimination).covariateData[[HR]]$notes. Distinct from
HR_BAND or HRV (not yet registered) which
would be a heart-rate-band stratifier or heart-rate variability metric,
respectively. Ratified canonically on 2026-05-23 alongside the
Ngamprasertwong 2016 propofol maternal-fetal sheep extraction.covariateData[[BODYTEMP]]$notes per model.(1 + e * (BODYTEMP - ref)) or
exponential forms exp(e * (BODYTEMP - ref)). Reference
values observed: 36.9 degC (Kloprogge 2013 lumefantrine; pooled-cohort
median in Ugandan pregnant + non-pregnant women with uncomplicated P.
falciparum malaria), 37.2 degC (Kloprogge 2014 quinine; pregnant Ugandan
women with uncomplicated P. falciparum malaria, cohort median at
admission).TEMP – common short form in malaria /
infectious-disease NONMEM control streams; used in
Kloprogge_2013_lumefantrine.R and
Kloprogge_2014_quinine.R (same orientation as the
canonical, no value transformation).Kloprogge_2013_lumefantrine.R (linear-deviation effect on
mean absorption transit time MTT:
MTT_indiv = TVMTT * (1 + e_bodytemp_mtt * (BODYTEMP - 36.9))
with e_bodytemp_mtt = 0.165 per degC; mean transit time
increases ~16.5% per degC over 36.0-39.8 degC, plausibly reflecting
reduced gut motility / prolonged absorption in feverish malaria
patients), Kloprogge_2014_quinine.R (exponential effect on
elimination clearance:
CL_indiv = TVCL * exp(e_bodytemp_cl * (BODYTEMP - 37.2))
with e_bodytemp_cl = -0.243 per degC, centered at the
cohort median; clearance decreases ~21.6% per degC increase in admission
body temperature over 36.0-38.9 degC, reflecting reduced metabolic
CYP3A4 activity during acute febrile malaria).covariateData[[BODYTEMP]]$notes. Units are degrees Celsius;
convert from Fahrenheit (degF) at data-assembly time, not inside
model(). Distinct from BODYTEMP_FEBRILE (not
yet registered) which would be a binary fever indicator if a future
paper dichotomises at the conventional 37.5 / 38.0 degC threshold.
Ratified canonically on 2026-05-16 alongside the Kloprogge 2013
lumefantrine extraction.covariateData[[PACO2]]$notes per model.exp(e_paco2_param * PACO2) or centered
exp(e_paco2_param * (PACO2 - ref)). Reference values
observed: 42.5 mmHg (Yano 1993 stable chronic airway obstruction cohort
mean; also the laboratory normal-range midpoint of 35-45 mmHg in
adults).PaCO2 – the standard physiological abbreviation; used
directly in Yano 1993 Table I and Eqs 7-8.Yano_1993_theophylline.R (log-linear effect on both CL and
Vd of theophylline:
CL = exp(lcl + e_hepimp_cl * HEPIMP + e_paco2_cl * PACO2 + etalcl)
with e_paco2_cl = 0.0233 per mmHg, and
Vd = exp(lvc + e_paco2_vc * PACO2 + etalvc) with
e_paco2_vc = 0.00934 per mmHg; both estimated by
forward-selection likelihood-ratio testing).covariateData[[CARDIAC_OUTPUT]]$notes per model.(param + e * (CARDIAC_OUTPUT - ref))
or power scaling (CARDIAC_OUTPUT / ref)^exponent. Reference
values observed: 5 L/min (Hendrickx 2006; population centering for the
WT=70 / CARDIAC_OUTPUT=5 typical-adult anchor in the Hendrickx covariate
equations).CO – common short form in inhaled-anesthetic and
hemodynamic popPK NONMEM control streams; same orientation as the
canonical, no value transformation.HR (heart rate, beats/min) – the two are physiologically
related (CARDIAC_OUTPUT = HR * stroke_volume) but they are
independent covariate columns and a paper may report either or both.
Distinct from BFR (blood flow rate through the
extracorporeal dialysis circuit, mL/min) which is
dialysis-circuit-specific. Distinct from ECMO_PUMP_SPEED
(ECMO centrifugal-pump rotational speed) which characterises an
externally-imposed cardiac-output augmentation rather than the patient’s
intrinsic cardiac output. The full-word CARDIAC_OUTPUT
canonical was chosen over the bare CO source-data
abbreviation per operator instruction (Hendrickx 2006 sidecar response,
task frompeople-744); CO is preserved as the
recognised source alias so paper NONMEM control streams using the bare
two-letter form translate without value transformation. Pre-registered
on 2026-06-10 ahead of the Hendrickx 2006 (BMC Anesthesiology 6:7, doi:10.1186/1471-2253-6-7) inhaled-anesthetic
extraction; the Hendrickx 2006 model files were not built because the
source data were generated by the Gas Man(R) physiological simulator
rather than measured in real patients (the paper fits empirical
3-compartment mammillary parameters to Gas Man-simulated
alveolar-concentration curves for four inhaled anesthetics, with no IIV
or residual error parameters reported), but the canonical is registered
so future real-patient hemodynamic / inhaled-anesthetic popPK papers can
use it without re-litigating the name.DBP_REL(t) = (dbp(t) - dbp0) / dbp0, then consumed as a
covariate by a downstream model. Distinct from a raw diastolic-BP value
(mmHg); this is the normalized deviation.exp(e_dbprel_haz * DBP_REL).DBPREL – Hansson 2013 e85 NONMEM column for the
relative dBP change; same orientation, no value transformation.Hansson_2013_sunitinib_os.R (time-varying relative dBP
change, simulated from Hansson_2013_sunitinib_dbp and
entered as an overall-survival hazard modulator: a greater relative dBP
increase lowers the hazard).RCFB1MAX (relative change from baseline in SUVmax driving
the Schindler 2016 OS hazard). The related raw-state PD output is the
dbp compartment in compartment-names.md; the
placebo-arm indicator paired with that model family is
PLACEBO.CL_renal = base + theta_URINE_FLOW * (URINE_FLOW - URINE_FLOW_ref)
with URINE_FLOW_ref = 100 mL/h in Allegaert 2015. A value
of 0 is a sentinel for “no urine collected during the
interval” (i.e., the urine pathway contribution is dropped); the
linear-effect term is gated by URINE_FLOW > 0 and not
extrapolated below the centering reference.UF (Allegaert 2015 NONMEM column; same orientation, no
transformation) – used in
Allegaert_2015_paracetamol.R.Allegaert_2015_paracetamol.R.URINE_FLOW == 0 sentinel-zero rule
reflects an Allegaert-specific convention rather than a
universally-agreed-upon parameterization. A second model that uses a
different effect form (e.g., direct
URINE_FLOW / URINE_FLOW_ref proportional scaling, no
zero-sentinel) should register its own canonical (e.g.,
URINE_FLOW_PROP) rather than reusing
URINE_FLOW with conflicting semantics. The full-word
canonical name was chosen over the bare UF source-data
abbreviation for clarity in source traces.URINE_FLOW
(instantaneous urine flow rate in mL/h measured per collection
interval).CL = CL_base + e_URINE_VOL_24H_cl * (URINE_VOL_24H / 100)
centered at 100 mL/24h (the anuria cutoff). Clinical cutoffs used in the
founding example: anuria < 100 mL/24h, oliguria 100-500 mL/24h,
preserved diuresis > 500 mL/24h.residual diuresis (Ulldemolins 2015 prose); typical
NONMEM column abbreviation DIUR. Same value orientation, no
transformation.Ulldemolins_2015_meropenem.R (critically ill adults with
septic shock and continuous renal replacement therapy; reference 100
mL/24h; additive linear effect 0.22 L/h per (URINE_VOL_24H / 100) on
meropenem total CL on top of the CRRT-mediated baseline 3.68 L/h).URINE_FLOW because (a)
URINE_VOL_24H is a 24-hour cumulative volume (mL/24h), not
an instantaneous rate (mL/h), and (b) the additive-linear effect form
base + slope * (URINE_VOL_24H / 100) is structurally
different from the centered-linear-effect-with-sentinel-zero form used
by URINE_FLOW. A future model that uses a proportional or
multiplicative effect form on a 24-hour volume should evaluate whether
the semantics still match this entry (proportional / multiplicative
forms are compatible); a centered-linear-with-sentinel form (analogous
to the URINE_FLOW convention) would diverge and should register a
sibling canonical. The full-word URINE_VOL_24H form was
chosen over the shorter URINE_24H per operator instruction
(sidecar request 001 of task frompeople-536). Ratified
canonically on 2026-06-27 alongside the Ulldemolins 2015 meropenem
extraction.1.73 x CrCl / BSA – or (b) a tracer-measured glomerular
filtration rate using an exogenous filtration marker (iohexol clearance,
inulin clearance, 99mTc-DTPA clearance, 51Cr-EDTA clearance), which is
the clinical gold standard for measured GFR. All variants enter popPK
models with the same operational role (a covariate on clearance) and the
same units; the per-model covariateData[[CRCL]]$description
and notes must state which assay the source paper
used.(CRCL / ref)^exponent or with linear centering
(1 + slope * (CRCL - ref)). Reference values observed: 76
mL/min/1.73 m^2 (Hamren 2008, iohexol-clearance-measured GFR), 80
mL/min/1.73 m^2 (Cirincione 2017, MDRD eGFR), 90 mL/min/1.73 m^2 (Li
2019, calculated GFR; Bajaj 2017, CKD-EPI eGFR), 100 mL/min/1.73 m^2 (Xu
2019, measured-CrCl BSA-normalized).eGFR – MDRD-estimated glomerular filtration rate; used
in Cirincione_2017_exenatide.R and
Kotani_2022_astegolimab.R.
Bajaj_2017_nivolumab.R uses the CKD-EPI variant.EGFR – all-caps variant.CRCL_BSA – BSA-normalized creatinine clearance
(measured CrCl / BSA x 1.73); used in
Xu_2019_sarilumab.R.1.73*CrCl/BSA – the formula form appearing in Xu 2019
Eq. for Vm.cGFR – calculated/estimated GFR, BSA-normalized; used
in Li_2019_abatacept.R.CLCR – source-paper column name; underlying assay form
varies. Used in Delattre_2010_amikacin.R (raw
Cockcroft-Gault, NOT BSA-normalized; median 55.5 mL/min in critically
ill septic adults) and in MedellinGaribay_2015_gentamicin.R
(Schwartz formula CLCR = K * length / SCr with K in {0.33,
0.45, 0.55}, BSA-normalized to mL/min/1.73 m^2). Document the assay form
per model in covariateData[[CRCL]]$description.CLiohexol – plasma iohexol clearance, the clinical
gold-standard tracer-measured glomerular filtration rate; used in
Hamren_2008_tesaglitazar.R (tesaglitazar / acyl glucuronide
interconversion popPK in subjects with varying renal function). Iohexol
is an exogenous contrast agent cleared exclusively by glomerular
filtration (no tubular secretion or reabsorption); its clearance is a
direct measurement of GFR.CLinulin – plasma inulin clearance, the historical gold
standard for measured GFR; analogous to iohexol clearance but requires
constant intravenous infusion of inulin.CL_DTPA – plasma 99mTc-DTPA clearance, a radiolabelled
tracer of glomerular filtration commonly used in clinical practice.CL_EDTA – plasma 51Cr-EDTA clearance, another
radiolabelled glomerular-filtration tracer.Cirincione_2017_exenatide.R (MDRD eGFR),
Xu_2019_sarilumab.R (measured CrCl BSA-normalized),
Kotani_2022_astegolimab.R (MDRD eGFR),
Li_2019_abatacept.R (cGFR),
Bajaj_2017_nivolumab.R (CKD-EPI eGFR, reference 90
mL/min/1.73 m^2), NA_NA_lidocaine.R (DDMODEL00000281;
binary stratification at threshold 52.7 mL/min adding -0.319 to the GX
rate constant K30 in the CRCL <= 52.7 cohort; the source
.ctl does not state the BSA-normalisation method),
Delattre_2010_amikacin.R (raw Cockcroft-Gault mL/min, NOT
BSA-normalized; reference 55.5 mL/min population median; additive linear
effect 1.42 L/h per (CRCL/55.5) on CL),
MedellinGaribay_2015_gentamicin.R (Schwartz BSA-normalized
CLCR; reference 75 mL/min/1.73 m^2 (population mean 76.7); additive
linear effect 0.06 L/h per (CRCL/75) on CL in infants 1-24 months),
Georges_2009_ceftazidime.R (raw MDRD-eGFR mL/min, NOT
BSA-normalized; mean 121 mL/min; additive linear effect 0.024 L/h per
mL/min on CL, no centering), Hamren_2008_tesaglitazar.R
(iohexol-clearance-measured GFR; reference 76 mL/min/1.73 m^2; linear
centered slope 0.99 %/(mL/min/1.73 m^2) on renal clearance of parent
tesaglitazar AND simple linear normalised scaling
(CRCL / 76) on the saturable Michaelis-Menten Vmax of
acyl-glucuronide renal elimination; founding tracer-GFR example).covariateData[[CRCL]]$description so future reviewers can
trace the source assay; the assay choice affects accuracy of the
absolute GFR value but does not change the covariate’s role as a linear
/ power scalar on renal clearance. Tracer-measured GFR is the clinical
gold standard (unaffected by muscle mass, malnutrition, and
tubular-secretion confounders that perturb creatinine-based estimates)
but is rarely available outside research and renal-impairment-cohort
studies; creatinine-based estimates dominate routine clinical popPK
datasets. Tracer-measured-GFR scope codified on 2026-06-17 alongside the
Hamren 2008 tesaglitazar extraction (per operator decision in sidecar
request 001).covariateData[[CREAT]]$units.(CREAT / ref)^exponent.CRE (umol/L, reference 70.73) – used in
Thakre_2022_risankizumab.R.SCR – common clinical-PK abbreviation; also Llanos-Paez
2020 source column for the patient’s individual serum creatinine.Thakre_2022_risankizumab.R,
Hennig_2013_tobra.R (umol/L; paired with
CREAT_REF for the SCR_mean / SCR ratio used in the Hennig
2013 renal-function factor), Llanos_2017_gentamicin.R
(umol/L; standardized per-patient against CREAT_REF rather
than a fixed cohort reference),
Llanos-Paez_2020_gentamicin.R (umol/L; used as the
patient’s SCR_i in the renal-function ratio
(CREAT_REF / CREAT)^0.58 on CL).CREAT chosen over the shorter
CRE/SCR as the NONMEM/clinical-PK convention
that is unambiguous. Per-model reference values must be documented in
covariateData[[CREAT]]$notes.CREAT to define a
renal-function factor on clearance.CREAT column. Document via
covariateData[[CREAT_REF]]$units.(CREAT_REF / CREAT)^exponent so that a patient with
measured SCR equal to the population-expected normal SCR has factor
1.SCR_mean – used in Hennig 2013 (Eq. 5:
f_SCR = (SCR_mean / SCR)^theta_SCR); also Llanos-Paez 2020
paper notation for the Ceriotti 2008 age/sex-matched physiological mean
SCR.Scrmean – Llanos-Paez 2017 paper notation; computed
from Ceriotti et al. 2008 age- and sex-stratified medians (Clin Chem
54:559-566, doi:10.1373/clinchem.2007.099648).SCR_standardised – Germovsek 2018 paper notation;
PMA-adjusted standardisation of raw SCR per the paper’s reference 28 (a
previously-developed Standing-style PMA stratification).CCR,adj – Ruhs 2012 paper notation; age- and
gender-adjusted reference creatinine derived from the paper’s reference
[23] (a Schwartz-style paediatric maturation adjustment); the main text
does not give the explicit formula.Hennig_2013_tobra.R,
Llanos_2017_gentamicin.R (umol/L; computed externally per
Ceriotti et al. 2008), Llanos-Paez_2020_gentamicin.R
(umol/L; ratio (CREAT_REF / CREAT)^0.58 multiplies the
maturation-scaled CL), Germovsek_2018_meropenem.R (umol/L;
ratio (CREAT_REF / CREAT)^0.40 multiplies the
maturation-scaled CL for renal meropenem clearance in neonates and young
infants; PMA-stratification reference per Germovsek 2018 Methods
reference 28), Ruhs_2012_methotrexate.R (mg/dL; age- and
gender-adjusted CCR,adj per the paper’s reference [23]; ratio
(CREAT_REF / CREAT)^0.314 multiplies the BSA-scaled MTX CL
in paediatric ALL patients).covariateData[[CREAT_REF]]$notes so that a user assembling
a virtual cohort can reproduce it. When no covariate data are available
to compute CREAT_REF, set CREAT_REF = CREAT so
the renal-function factor evaluates to 1 (matching the Hennig 2013
‘covariate set to 1 for missing data’ rule). Ratified canonically on
2026-05-08 alongside the Hennig 2013 tobramycin extraction.CREAT (the other commonly-reported renal-function marker)
because urea reabsorption is flow-dependent in the renal tubule, so BUN
is sensitive to volume status as well as GFR.covariateData[[BUN]]$units (1 mmol/L urea ~=
2.80 mg/dL BUN).(BUN / ref)^exponent. Reference values observed: 7 mg/dL
(Hall 2017 MARS hinge knot for ka; not a population median); 4.2 mmol/L
(Chen 2017 cohort median for the tacrolimus power-of-ratio CL
effect).Hall_2017_dapsone.R
(mg/dL; population median 13 mg/dL [range 7-28]; enters the MARS-based
covariate model on the absorption rate constant via the basis function
BF1 = max(0, BUN - 7), which interacts with a weight hinge
to drive Ka), Chen_2017_tacrolimus.R (mmol/L;
cohort median 4.2 mmol/L [range 1.7-10.4]; enters CL/F via standard
power-of-ratio scaling (BUN/4.2)^1.42 in low-dose oral
tacrolimus for Chinese myasthenia-gravis patients).general scope on
2026-06-03 alongside the Chen 2017 tacrolimus extraction, the second
model registering BUN. Hall 2017 enters BUN only through a
piecewise-linear MARS hinge
(max(0, BUN - 7) * max(0, 63.7 - WT)), not as a power
scaling – the per-model covariateData[[BUN]]$notes
documents this is part of a machine-learning-driven hinge model, not a
standard popPK covariate transform. Chen 2017 uses the conventional
centred power-of-ratio form (BUN/ref)^exponent referenced
to the cohort median; the positive exponent there is interpreted via a
urea-driven protein-carbamylation mechanism that reduces albumin binding
of the ~99%-protein-bound tacrolimus. Ratified canonically on 2026-05-18
alongside the Hall 2017 dapsone extraction.CRCL / CREAT / BUN: UACR
captures upstream glomerular albumin leak (kidney damage),
whereas CRCL / CREAT / BUN capture downstream filtration rate (kidney
function). Both axes are routinely reported together in
diabetic-kidney-disease (DKD) and CKD popPK / PD studies. Time-fixed at
baseline by default; document baseline-vs-time-varying status per model
via covariateData[[UACR]]$notes.covariateData[[UACR]]$units.(UACR - ref) or as a power scaling
(UACR / ref)^exponent. Reference values observed: 800 mg/g
(Goulooze 2022 FIDELIO-DKD centring value, close to the cohort median of
852 mg/g).UACR0 – Goulooze 2022 paper convention for baseline
UACR (the 0 suffix denotes the baseline / pre-randomisation
value).Goulooze_2022_finerenone.R (centred linear effects on Emax
and on the disease-progression slope TSLOPE, both centred at 800 mg/g;
coefficients 9.31e-5 g/mg on Emax and 1.14e-3 g/mg on TSLOPE per
Goulooze 2022 Table 1).(1 + e_sod_<param> * (SOD - ref)). Reference values
observed: 136 mmol/L (Thuo 2011 ciprofloxacin; cohort median in Kenyan
children with severe malnutrition, normal-range lower bound).Na+ / NA / SODIUM – common
source-paper printed forms; renamed to canonical SOD when
assembling input data. Used in Thuo_2011_ciprofloxacin.R
(the paper writes “Na+ (mmol/L)” in Table 1 and structural-model
equations).Thuo_2011_ciprofloxacin.R (linear centered-deviation
effects on apparent CL and apparent Vc:
1 + 0.0368*(SOD - 136) and
1 + 0.0291*(SOD - 136); reference 136 mmol/L is the cohort
median).covariateData[[SOD]]$notes. Distinct from any
“sodium content of dosed formulation” concept (e.g., sodium-rich oral
rehydration solution) – that would warrant a separate canonical
(DOSE_NA_MGML, etc.) if a future model retains it. Ratified
canonically on 2026-05-21 alongside the Thuo 2011 ciprofloxacin
extraction.All RRT-related canonicals follow the
RRT_<MODALITY>_<KIND> shape, where
MODALITY is the specific RRT type (HEMODIAL
for intermittent hemodialysis, CRRT for continuous /
extended modalities, PERIT_DIAL for peritoneal dialysis)
and KIND is STATUS for subject-level
treatment-status indicators (time-fixed within the analysis window) or
ACTIVE for per-time-point session gates (time-varying
within subject, 1 only during an active session). The 2026-06-19
canonical-register standardization audit renamed the prior
HEMODIAL, HEMODIALYSIS, and
CRRT_STATUS canonicals into this family to make the
modality + kind contrast explicit at the column name; pre-2026-06-19
names are preserved as source_aliases so existing data CSVs
continue to work for one release cycle.
HEMODIAL – prior canonical name (pre-2026-06-19
RRT_DIAL – used in Goti_2018_vancomycin.R
(binary indicator on CL and Vc in a 2-compartment vancomycin popPK
model). Goti 2018 Methods notes the indicator was created for the
routine-TDM cohort (n = 336 hemodialysis subjects of 1812 total) and
that all hemodialysis procedures were intermittent and used high-flux
membranes.Goti_2018_vancomycin.R
(multiplicative factors on CL and Vc:
0.7^RRT_HEMODIAL_STATUS on CL and
0.5^RRT_HEMODIAL_STATUS on Vc, so dialysis subjects have
30% lower CL and 50% lower central volume than non-dialysis
subjects).RRT_PERIT_DIAL_STATUS, RRT_CRRT_STATUS) if a
future paper retains them as covariates. Goti 2018 treats
RRT_HEMODIAL_STATUS as time-fixed per subject because
session-level dialysis timing was not reliably documented in the source
EHR data; a future paper that resolves drug clearance during versus
between dialysis sessions would use a time-varying form
(RRT_HEMODIAL_ACTIVE) or a separate per-session covariate.
When pairing RRT_HEMODIAL_STATUS with CRCL,
note that the Cockcroft-Gault CRCL of an anuric hemodialysis patient is
by convention very low or set per institution to a small floor value
(Goti 2018 truncated CRCL > 150 mL/min to 150 mL/min and corrected
SCr < 1 mg/dL in elderly subjects); residual renal function in
hemodialysis subjects is highly variable and the dialysis indicator
captures the bulk PK shift on top of the CRCL covariate. Ratified
canonically on 2026-05-16 alongside the Goti 2018 vancomycin extraction.
Renamed from HEMODIAL to RRT_HEMODIAL_STATUS
on 2026-06-19 per the canonical-register standardization audit (RRT
family normalization).CRRT_STATUS – prior canonical name (pre-2026-06-19
RRT_RRT – used in Shekar_2014_meropenem.R
(binary indicator selecting between the RRT-fixed-CL term and the
CrCL-driven non-RRT CL term in a 2-compartment meropenem popPK model).
Shekar 2014 Methods describes the RRT cohort as mixed CVVH (control RRT
subjects, true CRRT) and EDD-f (ECMO RRT subjects, extended daily
diafiltration; pharmacokinetically CRRT-like for slow-clearance solutes
such as meropenem) and the model treats the modalities as a single
binary covariate without distinguishing them.Shekar_2014_meropenem.R (piecewise CL:
TVCL = exp(lcl) * RRT_CRRT_STATUS + e_crcl_cl * CRCL_in_Lh * (1 - RRT_CRRT_STATUS),
with CRCL in raw Cockcroft-Gault mL/min converted to L/h inside
model(); 5/11 ECMO patients and 5/10 controls were on
RRT).RRT_HEMODIAL_STATUS (intermittent hemodialysis IHD only)
and from RRT_HEMODIAL_ACTIVE (per-time-point session gate
in within-subject time-varying dialysis-clearance models such as
Liesenfeld 2013 dabigatran). Anticipated as a future canonical
RRT_PERIT_DIAL_STATUS for peritoneal dialysis. Shekar 2014
ratification uses a mixed CVVH + EDD-f cohort because the source paper
treats them identically as a single binary RRT covariate; a future paper
that retains modality as a separate covariate (e.g. CVVH vs SLED vs
CVVHDF) would either reuse RRT_CRRT_STATUS with finer
per-modality columns layered on top, or warrant its own
modality-specific canonical (RRT_CVVH_STATUS,
RRT_SLED_STATUS, etc.). When pairing
RRT_CRRT_STATUS with CRCL, note that
Cockcroft-Gault CrCL is conventionally not defined / not reported for
RRT-dependent subjects; Shekar 2014 records CrCL only for non-RRT
subjects and the model formula switches off the CrCL term when
RRT_CRRT_STATUS = 1. Ratified canonically on 2026-05-18
alongside the Shekar 2014 meropenem extraction. Renamed from
CRRT_STATUS to RRT_CRRT_STATUS on 2026-06-19
per the canonical-register standardization audit (RRT family
normalization).ALB_gL = ALB_gdL * 10 on data ingestion. Models calibrated
to g/dL values must apply an inline conversion in model()
(e.g.,
alb_gdL <- ALB * 0.1 # SI g/L -> US-convention g/dL)
so the structural coefficients stay aligned with their original
calibration.(ALB / ref)^exponent.BALB (baseline albumin) – used in
Zhou_2021_belimumab.R. Maps directly to ALB;
baseline-vs-time-varying status documented in per-model notes.HSA (human serum albumin) – used in
Fauchet_2015_lopinavir_unbound.R where the column name
follows the paper’s protein-binding-equation notation distinguishing
serum albumin from alpha-1 acid glycoprotein. Maps directly to canonical
ALB; no value transformation. Reported in g/L; converted to
umol/L inside model() via molecular weight 66500 g/mol for
the K_HSA linear-binding term.Fasanmade_2009_infliximab.R (g/dL, reference 4.1),
Thakre_2022_risankizumab.R (g/L, reference 45),
Chua_2025_mirikizumab.R,
Moein_2022_etrolizumab.R,
Tiraboschi_2025_amlitelimab.R,
Yamada_2025_zolbetuximab.R,
Li_2019_abatacept.R (g/dL, reference 4.0; the Li 2019
Methods states ‘mg/dL’ which is a publication typo – see the model’s
covariateData[[ALB]]$notes),
Quartino_2019_trastuzumab.R (g/dL, reference 4; source
column ALBU; negative exponent -0.998 on linear CL),
Wang_2020_ontamalimab.R (g/L, reference 39),
Zhou_2021_belimumab.R (g/L, reference 40; baseline-only,
source column BALB), Okada_2025_rocatinlimab.R
(g/L, reference 44; source column ALBU; power exponent
-1.30 on linear CL), Xu_2020_daratumumab.R (g/L, reference
37.0; power exponent -1.149 on linear CL),
Struemper_2017_belimumab.R (g/L, reference 41;
baseline-only, source column BALB; power exponent -0.736 on
linear CL), Fauchet_2015_lopinavir_unbound.R (g/L; source
column HSA; enters the saturable-binding submodel via a
linear K_HSA * [ALB] * Cunbound term with K_HSA = 0.036 L/umol, not as a
power scaling on CL).covariateData[[ALB]]$units field is
load-bearing for model-file source-trace; effect-coefficient magnitude
is meaningless without the unit. Per-model conversion to US-convention
g/dL (where the structural coefficients were calibrated) is required in
model() via an inline alb_gdL <- ALB * 0.1
line - see the per-model notes for the conversion factor
used. Conversion factor: 1 g/dL = 10 g/L.TPRO_gL = TPRO_gdL * 10 on data ingestion. Models
calibrated to g/dL must apply an inline conversion in
model() (e.g., tpro_gdL <- TPRO * 0.1). 1
g/dL = 10 g/L.(TPRO / ref)^exponent. Reference value observed: 74 g/L
(Frey 2010 pooled-cohort median).PROT – Frey 2010 abbreviation in the final-model
equation.TP – common clinical-chemistry abbreviation.Frey_2010_tocilizumab.R (g/L, reference 74; exponent -1.1
on V1).ALB (serum
albumin, the largest single component of total protein). Frey 2010
retains both TPRO and ALB on V1 as separate
covariates with opposite signs (TPRO negative, ALB positive) and notes
there is no clear mechanistic explanation; the joint effect may reflect
serum-volume modifications. TPRO ratified canonically on
2026-04-28 alongside the Frey 2010 extraction.ALB / AAG / TPRO), which would
derive fu indirectly through a binding equation – FU is the
direct measurement.covariateData[[FU]]$units field documents whether the value
is a percent or a unitless fraction; effect-coefficient magnitudes are
meaningless without the unit.(1 + slope * (FU - ref)) or with power scaling
(FU / ref)^exponent. Reference values observed: 0.1%
(Hamren 2008 tesaglitazar cohort median imputed for three subjects with
missing fu measurements; range 0.06-0.2%).f_u – Hamren 2008 paper notation (italicised
fu).FRAC_UNBOUND – common NONMEM column-name form when the
value is reported as a unitless fraction.Hamren_2008_tesaglitazar.R (%, reference 0.1%; linear
centered effect 555 %/unit fu on the parent metabolic clearance CLmt –
per-subject fu measurements at day 42 by ultrafiltration drive the
realised CLmt range from approximately 1.5 to 3.0 L/h across the cohort
fu range 0.06-0.2%; founding example).fu paper-named parameter, not this covariate) or (b)
parameterise the effect through measured binding-protein concentrations
(ALB, AAG, TPRO). Use
FU only when the source paper provides individual measured
unbound fractions and uses them as a covariate. Ratified canonically on
2026-06-17 alongside the Hamren 2008 tesaglitazar extraction (per
operator decision in sidecar request 001).TPRO; used in popPK models of CSF-penetrating
drugs as a surrogate for blood-brain-barrier integrity (elevated CSF
protein indicates inflammation or barrier breakdown and typically
correlates with increased CNS penetration of small-molecule drugs).CSF_TPRO_gL = CSF_TPRO_mgdL * 0.01. 1 g/L = 100 mg/dL.(CSF_TPRO / ref)^exponent on a penetration
fraction. Reference value observed: 1.2 g/L (Germovsek 2018
typical-infant value; sick-neonate cohort median).CSF_protein – Germovsek 2018 paper notation.CSFPROT / CSF_PROT – compact column-name
forms common in NONMEM control streams.Germovsek_2018_meropenem.R (g/L, reference 1.2; additive on
the logit CSF barrier parameter with coefficient theta_CSFproteins =
-0.17 per g/L deviation from 1.2; ratified canonically on 2026-05-21
alongside the Germovsek 2018 meropenem extraction).TPRO (serum total
protein) – the two are biologically independent because the blood-brain
barrier prevents free equilibration of serum protein into CSF. Normal
CSF protein is approximately 0.15-0.45 g/L in healthy adults; sick
neonates and meningitis patients can reach several g/L. The covariate is
typically time-varying because CSF protein evolves over the course of
CNS inflammation; missing values are commonly imputed to the cohort
median when the source paper does not report a per-sample CSF protein
measurement.IGG_gL = IGG_mgdL * 0.01 on data ingestion. 1 g/L = 100
mg/dL. Models calibrated to mg/dL must apply an inline conversion in
model() (e.g., igg_mgdL <- IGG * 100).(IGG / ref)^exponent. Reference values observed: 14.8 g/L
(Zhou 2021), 9.65 g/L (Yang 2021).BIGG (baseline IgG) – used in
Zhou_2021_belimumab.R.IGGBL (baseline IgG) – used in
Yang_2021_cemiplimab.R.Zhou_2021_belimumab.R
(g/L, reference 14.8; baseline-only; exponent 0.293 on CL),
Yang_2021_cemiplimab.R (g/L, reference 9.65; small positive
exponent 0.184 on shared CL/Q), Struemper_2017_belimumab.R
(g/L, reference 13.7; baseline-only; exponent 0.347 on CL).covariateData[[IGG]]$units field is load-bearing (1 g/L ~=
100 mg/dL). Baseline-vs-time-varying status documented in
covariateData[[IGG]]$notes. Distinct from
lIgG0 / IgG-as-a-state in mechanistic FcRn-competition TMDD
models (e.g., Valenzuela_2025_nipocalimab.R), where IgG is
a dynamic state, not a baseline covariate; use IGG only
when the source paper treats IgG as a static (baseline) covariate
column.IGM_gL = IGM_mgdL * 0.01. 1 g/L = 100 mg/dL.(IGM / ref)^exponent. Reference values observed: 0.21 g/L
(Cheng 2026, pooled PID + SAD pediatric cohort median).Cheng_2026_immunoglobulin.R (g/L, reference 0.21;
baseline-only; power exponent 0.11 on baseline IgG (CBAS) – IgM enters
as a humoral-capacity proxy that informs the endogenous-IgG baseline
rather than directly modifying clearance).TBILI_umolL = TBILI_mgdL * 17.1 on data ingestion. 1 mg/dL
= 17.1 umol/L. Models calibrated to mg/dL must apply an inline
conversion in model() (e.g.,
tbili_mgdL <- TBILI / 17.1 # SI umol/L -> US-convention mg/dL).(TBILI / ref)^exponent.BIL (legacy NONMEM short label for total bilirubin) –
used in NA_NA_lidocaine.R (DDMODEL00000281; binarised at
threshold 0.53 mg/dL with
BIL_HIGH = as.integer(BIL > 0.53)).BILT (Urien 2005 capecitabine paper’s NONMEM short
label for “total bilirubin”) – used in
Urien_2005_capecitabine.R (umol/L, reference 8.8; power
scaling on the capecitabine non-transformation CL10 and on the 5’-DFUR
-> 5-FU rate constant K34).Yamada_2025_zolbetuximab.R (mg/dL, reference 0.38; small
positive exponent 0.0347 on V1), NA_NA_lidocaine.R (mg/dL,
source column BIL; binary effect at threshold 0.53 mg/dL on
the GX elimination rate constant K30),
Urien_2005_capecitabine.R (umol/L, reference 8.8; source
column BILT; positive exponent +0.32 on capecitabine
non-transformation CL10 and negative exponent -0.36 on the 5’-DFUR ->
5-FU rate constant K34).covariateData[[TBILI]]$units field
is load-bearing.TBILI: direct bilirubin is the
water-soluble glucuronide-conjugated fraction processed by hepatocytes
and excreted in bile, so a rise in DBIL specifically flags impaired
biliary excretion / cholestasis or intrahepatic shunting, whereas total
bilirubin also captures unconjugated (indirect) hyperbilirubinaemia from
haemolysis or Gilbert-type conjugation defects.DBIL_umolL = DBIL_mgdL * 17.1 on data ingestion. 1 mg/dL =
17.1 umol/L. Models calibrated to mg/dL must apply an inline conversion
in model() (e.g.,
dbil_mgdL <- DBIL / 17.1).(DBIL / ref)^exponent. Reference values observed: 2.6
umol/L (Chen 2015 voriconazole Chinese ICU cohort population
median).Chen_2015_voriconazole.R (umol/L, reference 2.6; negative
exponent -0.40 on CL: CL = TVCL * (DBIL / 2.6)^-0.40).TBILI because
direct vs total are not interchangeable: total = direct + indirect, and
the two fractions track different pathophysiologic processes. Scope kept
specific pending a second model that ratifies DBIL with
consistent semantics; promote to general once
corroborated.covariateData[[AST]]$units.(AST / ref)^exponent.SGOT (serum glutamic-oxaloacetic transaminase; the
legacy clinical-chemistry name for AST) – used in
Quartino_2019_trastuzumab.R.Lu_2014_trastuzumabemtansine.R (U/L, reference 27; small
positive exponent 0.071 on CL), Quartino_2019_trastuzumab.R
(IU/L, reference 24; source column SGOT; positive exponent
0.205 on linear CL).ALT and TBILI; register a separate
ALT canonical if a future paper requires it.
SGOT is the older lab-reporting name; values and units are
identical to AST.covariateData[[ALT]]$units.(ALT / ref)^exponent.SGPT (serum glutamic-pyruvic transaminase; the legacy
clinical-chemistry name for ALT, paralleling SGOT ->
AST) – used in NA_NA_lidocaine.R
(DDMODEL00000281; binarised at threshold 11 with
SGPT_HIGH = as.integer(SGPT > 11)).Nikanjam_2019_siltuximab.R (U/L, reference 19; small
negative exponent -0.096 on CL), Melhem_2022_dostarlimab.R
(U/L, reference 18; small negative exponent -0.0585 on CL,
time-varying), NA_NA_lidocaine.R (source column
SGPT; binary effects at threshold 11 on the GX rate
constant K30 and on the 2,6-xylidide rate constant K40).AST and TBILI. Ratified canonically
on 2026-04-24. SGPT is the older lab-reporting name; values
and units are identical to ALT.covariateData[[ALP]]$units.(ALP / ref)^exponent, with a linear-deviation form, or
binarized inline as alp_high <- (ALP > uln) for a
binary >ULN indicator.Gupta_2016_lenvatinib.R (binarized inline as
alp_high <- (ALP > 120); the source paper enters
ALP as a 0/1 NONMEM indicator with ALP = 1
when the ratio ALP/ULN > 1; multiplicative effect on CL/F:
0.883^alp_high).ALT
/ AST / GGT / TBILI. Ratified
canonically alongside the Gupta 2016 lenvatinib extraction.covariateData[[GGT]]$units.1 + theta * (GGT - ref) or power
scaling (GGT / ref)^exponent. Reference values observed: 33
U/L (Retlich 2015 popPK linagliptin median), 32.3 U/L (Retlich 2015
popPK/PD linagliptin median).Retlich_2015_linagliptin.R (U/L, reference 33;
linear-deviation effect on linagliptin CL with coefficient -0.0339 % per
U/L deviation. The PK/PD layer uses GGT (reference 32.3 U/L) as a
piecewise covariate on baseline DPP-4 activity BSL with a
linear-deviation effect below GGT = 175 U/L and a constant +21.3% effect
above the threshold).ALT
/ AST / ALP / TBILI. The
piecewise above/below-threshold form in Retlich 2015 reflects empirical
saturation of the GGT-vs-DPP-4-activity relationship at extreme values.
Ratified canonically alongside the Retlich 2015 linagliptin
extraction.covariateData[[LDH]]$units.(LDH / ref)^exponent or with an additive linear-on-log form
exp(coef * (log(LDH) - log(ref))) (algebraically equivalent
to (log(LDH) / log(ref))^coef). Reference values observed:
217 U/L (Sanghavi 2020).BLDH (baseline LDH) – used in
Sanghavi_2020_ipilimumab.R.Sanghavi_2020_ipilimumab.R (linear-on-log form on CL with
reference 217 U/L; coefficient 0.703), NA_NA_lidocaine.R
(DDMODEL00000281; binary stratification at threshold 195 U/L switching
the typical-value baseline of the 2,6-xylidide rate constant K40).(LDH/ref)^exponent form. Document the functional form in
covariateData[[LDH]]$notes.HEPIMP (with values 1 = mild / 0 = others)
– used in Lin_2024_casirivimab.R.Lin_2024_casirivimab.R
(NCI ODWG mild; multiplicative fractional change on CL),
Lu_2022_patritumab.R (NCI ODWG mild; paired with
HEPIMP_MOD_OR_MISSING; multiplicative fractional effect
0.706 on CLDXd for mild impairment vs the normal-hepatic-function
reference), Desai_2016_isavuconazole.R (Child-Pugh A;
log-additive shift
e_hepimp_mild_cl = log(1.55 / 2.54) = -0.494 on CL and
e_hepimp_mild_q = log(38.8 / 33.678) = +0.142 on Q vs the
healthy reference, paired with HEPIMP_MOD for the parallel
Child-Pugh B stratum).HEPIMP_MOD / HEPIMP_SEV rather than
overloading this entry.HEPIMP_MILD indicator paired with this column, so all-zero
corresponds to NCI ODWG group 1 = normal).HEPIMP_MOD_MISSING – prior canonical name
(pre-2026-06-19 OR-disambiguation standardization).HEPATIC / HEPATIC_MOD_MISSING – informal
NONMEM names for the composite group.Lu_2022_patritumab.R
(paired with HEPIMP_MILD; multiplicative fractional effect
0.532 on CLDXd; the composite group pools n = 6 moderate-impairment
patients with n = 6 missing/unknown patients per Lu 2022 Table S5).HEPIMP_MOD
canonical instead. Ratified canonically on 2026-04-28. Renamed from
HEPIMP_MOD_MISSING to HEPIMP_MOD_OR_MISSING on
2026-06-19 per the canonical-register standardization audit (operator
decision: insert OR so the composite-group meaning is
explicit in the name; the prior form could be misread as “moderate, with
missing-data flag” rather than “moderate OR missing pooled”).(B2M / ref)^exponent. Reference values observed: 3.90 mg/L
(Fau 2020 multiple-myeloma cohort median).B2M is the
universal abbreviation.Fau_2020_isatuximab.R
(mg/L, reference 3.90; exponent 0.343 on the steady-state linear
clearance CLinf).covariateData[[B2M]]$notes.(CYSC / ref)^exponent (Chung 2013) or
centred-linear scaling on the reciprocal
1 + e * (1/CYSC - ref_inv) (Viberg 2006). Reference values
observed: 0.91 mg/L (Chung 2013 vancomycin Korean adults with SCr <=
1.2 mg/dL; cohort median); 1.32 mg/L equivalent to 1/CYSC = 0.758
(mg/L)^-1 (Viberg 2006 cefuroxime adult patients with broad
renal-function range; population-typical).Cystatin C / cystatin – Chung 2013 paper
narrative and Table 2 footnote.CysC – Viberg 2006 paper narrative and Table 4
footnote.Chung_2013_vancomycin.R (mg/L, reference 0.91; power
exponent -0.780 on CL: CL_pop * (CYSC / 0.91)^-0.780),
Viberg_2006_cefuroxime.R (mg/L; centred-linear effect on
1/CYSC with coefficient 1.43 per (mg/L)^-1 and reference 0.758 (mg/L)^-1
on CL: CL_pop * (1 + 1.43 * (1/CYSC - 0.758))).CREAT (serum creatinine) – the two are commonly reported
alongside each other and can enter the same model as separate covariates
(as in Chung 2013, where CYSC explains 62% of CL variability vs SCr
13%). The functional form (power on CYSC vs centred-linear on 1/CYSC) is
paper-specific and lives in the model file; the canonical column is the
underlying biomarker concentration in mg/L.BHPTGRPN (categorical: 1 = normal, 2 = mild, 3 =
moderate, 4 = severe; 9999 = missing) – used in
Lu_2019_polatuzumab.R. Decompose:
HEPIMP = as.integer(BHPTGRPN > 1.5 & BHPTGRPN != 9999).HEP_IMP – retired canonical name; replaced by
HEPIMP for consistency with the HEPIMP_MILD
family.Lu_2019_polatuzumab.R
(multiplicative effect on FRAC_NS = 1.19, applied as
1.19^HEPIMP).HEPIMP_MOD canonical rather than overloading this
one.log(NASF / 4) for NASF >= 4 and zero contribution for
NASF < 4 (so NASF = 4 and any NASF < 4 reduce to the typical-value
reference). The cutoff of 4 distinguishes patients with a benign form of
NAFLD from those with biopsy-confirmed NASH (Pierre 2017 Methods
‘Covariate analysis’ and references 31 and 34).NASF – used in Pierre_2017_morphine.R
(Pierre 2017 Methods ‘Covariate analysis’).Pierre_2017_morphine.R
(linear effect on log(NASF / 4) for NASF >= 4 with
coefficient -0.628 on M3G clearance:
CL_M3G_i = CL_M3G_pop * (1 + e_nasf_cl_m3g * log(NASF / 4))
for NASF >= 4 and CL_M3G_i = CL_M3G_pop for NASF < 4;
higher NASF reduces M3G clearance via reduced biliary excretion and
increased basolateral efflux of M3G into systemic circulation).HEPIMP* (NCI ODWG
oncology-trial hepatic-impairment categories) and from continuous liver
enzymes (ALT, AST, ALP) – NASF is a biopsy-derived disease-severity
ordinal specific to NAFLD / NASH. Ratified canonically on 2026-05-18
alongside the Pierre 2017 morphine extraction.covariateData[[HEPIMP_SEV]]$notes. Two schemes
are commonly encountered:
Child-Pugh Class C – used in
vanderWalt_2013_dapagliflozin.R (covariate effects on
CLP_M15 and V2M; the paper dichotomizes severe hepatic impairment per
the Child-Pugh classification).vanderWalt_2013_dapagliflozin.R (Child-Pugh Class C;
multiplicative fractional effects -0.422 on the dapagliflozin -> D3OG
metabolic clearance and +1.33 on the D3OG central volume of
distribution; paper text “With severe HI (Child-Pugh Class C), CLP M15
decreased by 41% and V2M increased by 134%”).HEPIMP_MODSEV canonical rather than overloading this entry.
Companion to HEPIMP_MILD (mild only) and
HEPIMP_MODSEV (moderate + severe pooled); the SKILL.md
anticipates each severity level as its own canonical when the source
paper tests them as separate covariates.HEPIMP_MOD_OR_MISSING (which pools moderate cases with
missing-data cases, not with severe cases). The classification scheme
that defines the cut points is paper-specific and must be documented in
per-model covariateData[[HEPIMP_MODSEV]]$notes. Two schemes
are commonly encountered:
Child-Pugh Class B,C – used in
vanderWalt_2013_dapagliflozin.R (covariate effects on V3P
and CLM; the paper dichotomizes moderate-or-severe hepatic impairment
per the Child-Pugh classification).vanderWalt_2013_dapagliflozin.R (Child-Pugh Class B or C;
multiplicative fractional effects -0.600 on the dapagliflozin peripheral
volume of distribution V3P and -0.293 on the D3OG renal clearance CLM;
paper text “Moderate or severe HI (Child-Pugh Class B or C) decreased
CLM and the peripheral volume of distribution of dapagliflozin (V3P) by
29 and 60%, respectively”).HEPIMP_MILD (mild only) and HEPIMP_SEV (severe
only). Distinct from HEPIMP_MOD_OR_MISSING (which pools
moderate cases with subjects whose hepatic-function data are
missing/unknown, not with severe cases). The composite mod-or-sev
pooling is a different load-bearing convention than the mod-or-missing
pooling, so the two canonicals must remain separate.CRCL). The classification
scheme that defines “moderate” is paper-specific and must be documented
in per-model covariateData[[RENALIMP_MOD]]$notes. Two
schemes are commonly encountered:
RENAL (binary 1 = moderate / 0 = healthy or mild) –
used in Morris_2011_telapristone.R (Morris 2011 Methods
“Covariate Analysis”, paragraph after Equation for the proportional
covariate model).Morris_2011_telapristone.R (proportional fractional effect
on the telapristone absorption rate constant Ka; coefficient -0.744
yielding
Ka = Ka_typ * (1 + e_renalimp_mod_ka * RENALIMP_MOD), i.e.,
a 74% decrease in Ka in moderate-renal-impaired subjects vs the
reference healthy/mild-renal cohort).CRCL (continuous Cockcroft-Gault
clearance), RENALIMP_MILD and RENALIMP_SEV
(parallel canonicals reserved for future extractions following the
HEPIMP_* family precedent). The Morris 2011 Discussion
notes that the effect is “not directly attributed to renal impairment
(in terms of decrease glomerular filtration rate/changes in creatinine
clearance)” but “attributed to some other disease precluding/resulting
in renal impairment or other pathophysiologic states induced by renal
impairment (e.g., delayed gastric emptying as a result of diabetes)”, so
the binary indicator captures the cohort allocation rather than a
mechanistic GFR effect. Ratified canonically on 2026-06-09 alongside the
Morris 2011 telapristone extraction.covariateData[[CPK]]$units.(CPK / ref)^exponent. Reference values observed: 63 U/L
(Yang 2024 axatilimab; pooled-cohort median).BLCPK (baseline CPK) – informal usage in Yang
2024.Yang_2024_axatilimab.R
(baseline-only covariate on baseline NCMC concentration
BL_NCMC with power exponent 0.376; reference 63 U/L).AST / ALT (hepatic) and LDH
(general tissue turnover). Yang 2024 uses CPK alongside AST
and LDH as tracked safety biomarkers. Per-model
covariateData[[CPK]]$notes should document
baseline-vs-time-varying status and the clinical interpretation in the
source population (skeletal-muscle injury, macrophage-clearance
surrogate, or both). Distinct from any model state variable representing
CPK time-course dynamics – covariate column is the pre-dose laboratory
observation.RRT_HEMODIAL_ACTIVE)HEMODIALYSIS, and the 2026-06-19
audit then renamed it again to the current
RRT_HEMODIAL_ACTIVE (see that entry below for full
documentation). The DIAL form is retained ONLY as a
source_alias of RRT_HEMODIAL_ACTIVE so that
pre-2026-06-09 data CSVs continue to load.RRT_HEMODIAL_ACTIVE for the current canonical.DIAL = 1
– in the interdialytic period the value is sentinel and the
Michaels-equation term is gated off by DIAL.DFR and a hemodialyzer
mass-transfer-area coefficient. Values investigated in the ratification
source were 200, 300, and 400 mL/min (Liesenfeld 2013 Methods, Study
Design; Table 1).Liesenfeld_2013_dabigatran.R.DIAL (binary on/off
gate) and DFR (dialysate flow rate). Ratified canonically
on 2026-05-16 alongside the Liesenfeld 2013 dabigatran extraction.DIAL = 1.BFR. The ratification source fixed
DFR at 700 mL/min throughout (Liesenfeld 2013 Methods, Study Design) and
additionally simulated 500 mL/min (Methods, Simulations).Liesenfeld_2013_dabigatran.R.DIAL and
BFR. Ratified canonically on 2026-05-16 alongside the
Liesenfeld 2013 dabigatran extraction.covariateData[[ECMO_PUMP_SPEED]]$notes should document the
time resolution.(ECMO_PUMP_SPEED / ref)^exponent. The
reference value is paper-specific (median pump speed in the source
cohort): Yang 2017 uses 2350 RPM (cohort median; Table 1 / Results:
“median ECMO pump speeds of 2350 RPM”).Yang_2017_remifentanil.R (power effect on remifentanil CL:
(ECMO_PUMP_SPEED / 2350)^2.04; higher pump speed associated
with higher CL, hypothesised mechanism is increased spontaneous drug
degradation at high centrifugal-pump shear).(Q_CVVH / ref)^exponent with the
source paper additionally gating the effect to 1 when
Q_CVVH = 0 (no CVVH active). The reference value is
paper-specific (cohort median during CVVH operation): Ahsman 2010 uses
193 mL/min (cohort median; Table 1).QCVVH – used in Ahsman_2010_cefotaxime.R
(paper notation, subscript dropped).Ahsman_2010_cefotaxime.R (power effect on DACT metabolite
CL: (Q_CVVH / 193)^0.72 when CVVH active, 1
otherwise).ECMO_PUMP_SPEED
(the centrifugal-pump rotational speed setting cardiac-output
augmentation) and from BFR / DFR (blood-flow /
dialysate-flow rates characterising standalone hemodialysis circuits):
Q_CVVH is the slow-clearance flow through a CVVH filter integrated INTO
an ECMO circuit. Distinct from RRT_CRRT_STATUS (binary on /
off indicator for continuous RRT) which captures presence-of-therapy
rather than its rate. Future ECMO + CVVH popPK extractions that quantify
CVVH flow should reuse this canonical and extend the example list.
Ratified canonically alongside the Ahsman 2010 cefotaxime
extraction.(T_POST_ECMO / ref)^exponent with the
source paper additionally gating the effect to 1 when
T_POST_ECMO = 0 (before or during ECMO). The reference
value is paper-specific: Ahsman 2010 uses 100 h (Appendix).tEND – used in Ahsman_2010_cefotaxime.R
(paper notation, subscript dropped).Ahsman_2010_cefotaxime.R (power effect on parent CTX CL:
(T_POST_ECMO / 100)^0.16 when post-ECMO, 1 otherwise; power
effect on DACT metabolite CL: (T_POST_ECMO / 100)^0.53 when
post-ECMO, 1 otherwise).TPP (time postpartum) and T_ENTRY
(per-subject study-entry time) but anchored to ECMO decannulation.
Distinct from t_EC (time since ECMO START / beginning of
extracorporeal circulation, which the source paper tested but did not
retain) and from on-ECMO duration covariates. The gating-at-zero
convention is structural: during ECMO the recovery-time covariate is
undefined / not yet started, so the (0 / ref)^exponent expression would
collapse pathologically and the implementer must short-circuit the
multiplier to 1. Future ECMO popPK extractions that include a
post-decannulation recovery phase should reuse this canonical and extend
the example list. Ratified canonically alongside the Ahsman 2010
cefotaxime extraction.covariateData[[FILT_AGE_HI]]$notes.cl_dialysis * e_filt_age_hi_<arm>^FILT_AGE_HI so the
FILT_AGE_HI = 0 case recovers the fresh-filter baseline.ffCL_CVVHDF (a multiplicative
factor on the CVVHDF clearance arm) rather than naming a data-set
column. When ratified the column was assigned the canonical name
FILT_AGE_HI in the model file.Patel_2011_fluconazole.R (multiplicative factor
e_filt_age_hi_cl_renal^FILT_AGE_HI on the CVVHDF clearance
arm: e_filt_age_hi_cl_renal = 0.368, so > 48 h filters
give 36.8% of fresh-filter CVVHDF efficiency; bootstrap 95% CI
0.326-0.426 from Patel 2011 Table 2; Delta-OBJ = -11.46 vs base
model).RRT_CRRT_STATUS
(subject-level binary indicator that ANY continuous / extended
renal-replacement therapy is occurring) and from
RRT_HEMODIAL_ACTIVE (per-time-point on/off gate for the
dialysis-clearance term in within-subject time-varying
dialysis-clearance models). FILT_AGE_HI gates WITHIN the
CRRT-on subset by membrane age – meaningful only when RRT_CRRT_STATUS =
1 or RRT_HEMODIAL_ACTIVE = 1. The > 48 h threshold is paper-specific
(Patel 2011 Methods note that membrane fouling and clot formation in the
hemofilter pores progressively reduce dialysis efficiency, with the
inflexion typically observed near 48 h). Future extractions that retain
a finer filter-age structure (continuous filter_age_hours or multiple
thresholds) should reuse FILT_AGE_HI for the binary > 48 h case and
add a separate canonical for the continuous form. Scope kept specific
pending a second model that ratifies the same > 48 h threshold;
promote to general once corroborated. Ratified canonically on 2026-06-09
alongside the Patel 2011 fluconazole extraction.cl_hemodialysis term should gate
it by RRT_HEMODIAL_ACTIVE = 1 so that the interdialytic
clearance reduces to the intrinsic body clearance.HEMODIALYSIS – prior canonical name (2026-06-09 –
2026-06-19); the 2026-06-19 RRT family standardization renamed it to
RRT_HEMODIAL_ACTIVE to align with the sibling
RRT_HEMODIAL_STATUS / RRT_CRRT_STATUS
shape.DIAL – legacy form (used in Liesenfeld 2013, Jacobs
2016, Veinstein 2013 prior to the 2026-06-09 rename to the more explicit
HEMODIALYSIS).IHD_ON, HD_ACTIVE, RRT_ACTIVE
– variant abbreviations used in adjacent ESRD / CRRT popPK
literature.Liesenfeld_2013_dabigatran.R (Michaels-equation gate;
cl_total <- cl + RRT_HEMODIAL_ACTIVE * Michaels(BFR, DFR, KoA);
the dialysis arm is derived from blood flow rate, dialysate flow rate,
and a hemodialyzer mass-transfer-area coefficient).Jacobs_2016_colistin.R (additive on/off gate of fixed
device-level hemodialysis-clearance constants for CMS and colistin;
cl_tot <- cl + RRT_HEMODIAL_ACTIVE * cl_hd_cms).Veinstein_2013_gentamicin.R (additive arm with an
estimated primary ini() parameter;
cl_total <- cl + RRT_HEMODIAL_ACTIVE * cl_hemodialysis).RRT_HEMODIAL_STATUS
(the subject-level treatment-status indicator) –
RRT_HEMODIAL_ACTIVE is the per-time-point gate that turns
the dialysis-clearance term on and off WITHIN a single hemodialysis
subject’s record. Pair with BFR and DFR when
the dialysis clearance depends on flow rates; pair with a
filter-specific mass-transfer coefficient (estimated lkoa
in the model, not a covariate) when the Michaels parameterisation is
used. Pair with lcl_hemodialysis (canonical) when the
dialysis arm is a primary estimated structural parameter rather than a
derived expression. Ratified canonically on 2026-05-16 (as
DIAL), renamed to HEMODIALYSIS on 2026-06-09
alongside the Veinstein 2013 gentamicin extraction, then renamed again
to RRT_HEMODIAL_ACTIVE on 2026-06-19 per the
canonical-register standardization audit (RRT family normalization: the
RRT_<modality>_<kind> pattern makes the
modality + per-session-vs-subject-level distinction explicit at the
column name).covariateData[[HEPIMP_MOD]]$notes. Two schemes are commonly
encountered:
HEP2 – used in Desai_2016_isavuconazole.R
(paper’s liver-function index HEP2 for moderate Child-Pugh B; same
orientation as the canonical).Desai_2016_isavuconazole.R (Child-Pugh Class B;
log-additive shift
e_hepimp_mod_cl = log(1.326 / 2.54) = -0.650 on CL and
e_hepimp_mod_q = log(63.554 / 33.678) = +0.635 on Q vs the
healthy reference, paired with HEPIMP_MILD for the parallel
Child-Pugh A stratum; encodes the paper’s three-typical-value form
CL = theta_g * exp(eta_jCL) as a healthy-reference baseline
plus group-specific log shifts).HEPIMP_MILD (mild only) and
HEPIMP_SEV (severe only). Distinct from
HEPIMP_MOD_OR_MISSING (pools moderate cases with
missing-data cases) and HEPIMP_MODSEV (pools moderate with
severe). Anticipated by the HEPIMP_MILD entry’s Notes: “For
models that test moderate or severe as separate categories, register
additional canonicals HEPIMP_MOD / HEPIMP_SEV rather than overloading
this entry.” HEPIMP_SEV was registered separately; HEPIMP_MOD completes
the parallel.CRCL). The classification
scheme that defines “moderate” is paper-specific and must be documented
in per-model covariateData[[RENALIMP_MOD]]$notes. Two
schemes are commonly encountered:
RENAL (binary 1 = moderate / 0 = healthy or mild) –
used in Morris_2011_telapristone.R (Morris 2011 Methods
“Covariate Analysis”, paragraph after Equation for the proportional
covariate model).Morris_2011_telapristone.R (proportional fractional effect
on the telapristone absorption rate constant Ka; coefficient -0.744
yielding
Ka = Ka_typ * (1 + e_renalimp_mod_ka * RENALIMP_MOD), i.e.,
a 74% decrease in Ka in moderate-renal-impaired subjects vs the
reference healthy/mild-renal cohort).CRCL (continuous Cockcroft-Gault
clearance), RENALIMP_MILD and RENALIMP_SEV
(parallel canonicals reserved for future extractions following the
HEPIMP_* family precedent). The Morris 2011 Discussion
notes that the effect is “not directly attributed to renal impairment
(in terms of decrease glomerular filtration rate/changes in creatinine
clearance)” but “attributed to some other disease precluding/resulting
in renal impairment or other pathophysiologic states induced by renal
impairment (e.g., delayed gastric emptying as a result of diabetes)”, so
the binary indicator captures the cohort allocation rather than a
mechanistic GFR effect. Ratified canonically on 2026-06-09 alongside the
Morris 2011 telapristone extraction.CRCL column under Wahlby 2004’s extended covariate-model
decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The
within-subject delta is computed in model() as
(CRCL - CRCL_BASE).CRCL column and document per-model via
covariateData[[CRCL_BASE]]$units.(CRCL_BASE - ref) in a linear or linear-exp effect, where
ref is the cohort median baseline CRCL.BCLC (baseline creatinine clearance) – Wahlby 2004
source-column convention; used in Wahlby_2004_gentamicin.R
(raw mL/min) and Wahlby_2004_pefloxacin.R (concept tested
but not retained in final model).Wahlby_2004_gentamicin.R (raw mL/min; reference 71.7 mL/min
= BCLC median; coefficient 0.0098 per mL/min on CL alongside paired DCLC
effect 0.0068 per mL/min).general if a
second paper ratifies the baseline-CRCL column with consistent
semantics. Distinct from time-fixed adult-cohort CRCL (where per-subject
CRCL is just constant by data design); use CRCL_BASE only
when the source paper deliberately separates the BCOV from the
time-varying COV. Ratified canonically alongside the Wahlby 2004
extraction.TBILI column under Wahlby 2004’s extended covariate-model
decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The
within-subject delta is computed in model() as
(TBILI - TBILI_BASE).TBILI column and document per-model via
covariateData[[TBILI_BASE]]$units.(TBILI_BASE - ref) in a linear or linear-exp effect.BBIL (baseline total bilirubin) – Wahlby 2004
source-column convention; used in Wahlby_2004_pefloxacin.R
and Wahlby_2004_paclitaxel_myelosuppression.R.Wahlby_2004_pefloxacin.R (umol/L; centered at 25 umol/L;
coefficient -0.0068 per umol/L on CL via exp[e * (TBILI_BASE - 25)]),
Wahlby_2004_paclitaxel_myelosuppression.R (umol/L; used
together with TBILI as the within-subject delta TBILI - TBILI_BASE
entering the Slope-DBIL effect).general if a
second paper ratifies. Distinct from DBIL
(direct/conjugated bilirubin, a clinical-chemistry concept) – TBILI_BASE
is a methodologic baseline-snapshot column, not a separate biomarker.
Ratified canonically alongside the Wahlby 2004 extraction.ALP column under Wahlby 2004’s extended covariate-model
decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The
within-subject delta (or log-ratio) is computed in model()
from ALP and ALP_BASE.ALP column and document per-model via
covariateData[[ALP_BASE]]$units.log(ALP / ALP_BASE) or a linear delta
(ALP - ALP_BASE).BALKP (baseline alkaline phosphatase) – Wahlby 2004
source-column convention; used in
Wahlby_2004_voriconazole.R.Wahlby_2004_voriconazole.R (IU/L; population median 136
IU/L; paired with time-varying ALP to form log(ALP/ALP_BASE), the
paper’s log(DALKP) term, with coefficient 0.59 on CL inside Eq 7
Final-Model column).general if a
second paper ratifies. Ratified canonically alongside the Wahlby 2004
extraction.PERIT_DIAL = 1 selects the PD subgroup.DIA – used in Takama_2007_darbepoetin.R
(dialysis-modality indicator: DIA = 0 for hemodialysis,
DIA = 1 for peritoneal dialysis; maps directly onto
PERIT_DIAL with no value transformation). Takama 2007 Table
4 and Methods describe the covariate equation
V1 = theta_V1 * [1 + theta_V1_WT * (WT - 54) + theta_V1_DIA * DIA]
where theta_V1_DIA = 0.170 increments V1 by +17% in PD
subjects relative to HD subjects.Takama_2007_darbepoetin.R (+17% multiplicative-deviation
effect on central volume V1; chronic-dialysis cohort comparing 63 HD vs
68 PD Japanese adult patients).RRT_HEMODIAL_STATUS (intermittent-hemodialysis IHD
indicator; RRT_HEMODIAL_STATUS = 1 selects HD subjects) and
from RRT_CRRT_STATUS (continuous / extended RRT indicator).
In a chronic-dialysis cohort where every subject is on either HD or PD,
RRT_HEMODIAL_STATUS and PERIT_DIAL are
perfectly anti-correlated
(PERIT_DIAL = 1 - RRT_HEMODIAL_STATUS); the choice of which
to use as the covariate is dictated by which modality the source paper
treats as the reference category. Takama 2007 took HD as the reference
and reports a +17% effect on V1 for PD, so PERIT_DIAL is
the natural canonical for that paper (V1 increases when
PERIT_DIAL = 1). Anticipated in the
RRT_HEMODIAL_STATUS register-entry notes (“would warrant
its own canonical (RRT_PERIT_DIAL_STATUS,
RRT_CRRT_STATUS) if a future paper retains them as
covariates”) and ratified canonically on 2026-06-12 alongside the Takama
2007 darbepoetin alfa extraction. Future audit may rename this to
RRT_PERIT_DIAL_STATUS for consistency with the
RRT_covariateData[[HGB]]$units.(HGB / ref)^exponent.HGB is the
common NONMEM / clinical-PK abbreviation.Yamada_2025_zolbetuximab.R (g/L, reference 118; exponent
-0.374 on V1).covariateData[[HGB]]$units
field is load-bearing.HGB (which is the time-varying or
baseline-only hemoglobin value when entering a covariate effect on a
structural PK parameter): HGB_BL is a static per-subject
covariate that supplies the steady-state baseline for an Hb-state ODE,
typically together with a kin = kout * HGB_BL
synthesis-rate-from-baseline coupling so the Hb state starts at steady
state. Parallel to FERRITIN_BL (baseline ferritin) and
WT_BASE (baseline weight).covariateData[[HGB_BL]]$units.HBBASE – used in
Mulder_2025_ribavirin.R.Mulder_2025_ribavirin.R (mmol/L; initial condition for the
hb_state compartment; per-subject
kin = kout * HGB_BL so the Hb state is at steady state
pre-RBV; the source paper substitutes the cohort typical (median)
baseline for the one subject without an individual baseline value).kin per
subject from baseline rather than estimating a typical Hb0). Promote to
general if a second paper ratifies the same
baseline-Hb-as-initial-condition pattern. The unit varies widely across
papers (SI g/L, US g/dL, Dutch / European convention mmol/L; 1 g/dL = 10
g/L; 1 mmol/L of tetrameric haemoglobin ~ 16.114 g/L); the per-model
covariateData[[HGB_BL]]$units field is load-bearing.covariateData[[WBC]]$units.(WBC / ref)^exponent. Reference values observed: 10 x
10^9/L (Mould 2007, typical CLL Vmax normalization).WBC is the
universal clinical-PK abbreviation).Mould_2007_alemtuzumab.R (reference 10 x 10^9/L; exponent
0.194 on Vmax).covariateData[[WBC]]$notes.(NLR / ref)^exponent or exponential effects. Reference
values observed: 2.11 (Lin 2024, median in pooled COVID-19 +
non-infected cohort).NLR is the
universal abbreviation in clinical-PK and inflammation-biomarker
literature.Lin_2024_casirivimab.R
(time-varying; reference 2.11; small positive exponent +0.029 on
CL).covariateData[[NLR]]$notes. Although it derives from
WBC differential counts, register it as its own canonical
because the ratio (not the absolute counts) is what the model uses.covariateData[[HCT]]$units.(HCT / ref)^exponent. Reference values observed: 45 %
(Nestorov 2014, study-population median for severe hemophilia A
adults).HCT is the
universal NONMEM / clinical-PK abbreviation.Nestorov_2014_factorviii.R (reference 45 %, exponent -0.419
on V1).covariateData[[HCT]]$notes. Distinct from HGB
(mass concentration of hemoglobin); the two correlate but enter
different mechanistic relationships.HGB (mass concentration in plasma) and HCT
(volume fraction). Used in erythropoiesis / RBC-regeneration models as
the steady-state set point Base that drives the negative-feedback term
and seeds the steady-state initial conditions of the precursor
compartments.Base (Tetschke 2018
paper symbol).Tetschke_2018_erythropoiesis.R (reference 885.42 g;
Pottgiesser 2008 dataset of 29 healthy adult male volunteers).general if a second model registers this
quantity. Distinct from HGB (g/L or g/dL plasma
concentration) and HCT (RBC volume fraction):
THB_MASS is the absolute body-pool mass and is not
perturbed by short-term plasma-volume fluctuations (Pottgiesser 2008
Section 3.2 explicitly motivates the choice of mass over concentration).
Sex-dimorphic: typical value in adult males is meaningfully higher than
in adult females; document the sex composition of the population in
covariateData[[THB_MASS]]$notes.(NEUT - ref) or power scaling
(NEUT / ref)^exponent) or, in semi-mechanistic
myelosuppression models, as a per-subject initial-condition value for
the proliferation, transit, and circulating compartments.covariateData[[NEUT]]$units if the source paper uses a
different unit (e.g., 10^9 cells/L for
Ozawa_2007_docetaxel.R per the paper’s Table-3 reporting
unit).exp(coef * (NEUT - ref)), in power scaling
(NEUT / ref)^exponent, or as a direct per-subject
initial-condition assignment in semi-mechanistic Friberg-family models.
Reference values observed: 4133 cells/mm^3 (BAST PTTE 2017 simulated
cohort median; NA_NA_tte_gompertz.R Event 1 base hazard
model); 5 x 10^9/L (Ozawa 2007 typical Japanese cancer cohort, used as
the initial condition for the proliferation, transit, and circulation
compartments).BASE – per-subject baseline ANC supplied as a NONMEM
data column (used in Ozawa_2007_docetaxel.R; Appendix I
$INPUT).ANC – the absolute-neutrophil-count source-column
spelling (10^9/L); time-varying hazard modulator in
Hansson_2013_sunitinib_os.R (the overall-survival Weibull
hazard term exp(e_anc_haz * ANC + ...)). No value
transformation beyond the per-paper unit (document 10^9/L
in covariateData[["ANC"]]$units).NA_NA_tte_gompertz.R
(BAST PTTE 2017 / DDMODEL00000243 Event 1 hazard model; centred at NEUT
= 4133/mm^3; coefficient -1.56e-4 on the NONMEM rescaled scale,
equivalent to exp(-1.56e-4 * (NEUT - 4133)) on the hazard),
Ozawa_2007_docetaxel.R (Friberg-extension myelosuppression
PD; per-subject baseline ANC supplied via the NEUT column,
used as the initial condition for the proliferating, three transit, and
circulating compartments per the Methods text ‘Circ (t = 0) was fixed at
its observed value’), Hansson_2013_sunitinib_os.R
(time-varying ANC as ANC, simulated from the upstream
Hansson 2013 myelosuppression model and consumed as an OS-hazard
predictor).Ozawa_2007_docetaxel.R. The NEUT canonical units
are cells/mm^3, but the reporting unit 10^9 cells/L
(numerically NEUT_per_mm3 / 1000) is also common in oncology papers;
per-model covariateData[[NEUT]]$units documents the
per-paper unit. Distinct from WBC (total white blood cell
count, of which neutrophils are the largest fraction in healthy adults)
– NEUT is a specific differential-count subfraction. Also
distinct from NLR (neutrophil-to-lymphocyte ratio), which
is a derived ratio.ferritin(0) <- FERRITIN_BL; Bellanti 2015), and (2) as
a static per-subject covariate entering a power-law multiplier on
hepcidin turnover rate constants in iron-status / menstrual-cycle
turnover models (Angeli 2016). Distinct from a state-output ferritin
trajectory: FERRITIN_BL is a static per-subject covariate
(one value, supplied at simulation start); a state-output ferritin would
evolve over time per a disease ODE.covariateData[[FERRITIN_BL]]$units.Bellanti_2015_deferoxamine.R (ug/L; initial condition for
the ferritin compartment; n=27 transfusion-dependent beta-thalassaemia
major paediatric / adolescent cohort, median 2260, range 393-8500),
Angeli_2016_iron_hepcidin.R (ug/L; per-subject end-of-cycle
baseline used as a power-law multiplier
(FERRITIN_BL / 53)^exponent on hepcidin elimination
kout_hep (exponent -0.60) and on the hepcidin post-menses
rebound krel_hep (exponent -1.95); higher baseline ferritin
-> slower hepcidin elimination and a smaller rebound, consistent with
iron-regulatory feedback).specific to
general on 2026-06-03 with the Angeli 2016 iron / hepcidin
extraction (the second model to register FERRITIN_BL with
consistent baseline-ferritin semantics across two distinct mechanistic
uses).covariateData[[RBC]]$units.(RBC - ref) or power scaling
(RBC / ref)^exponent. Reference values observed: 4.40 x
10^12/L (Zhang 2012 healthy Chinese adult cohort median across the four
dosing groups in Table 1).RBC* – used in Zhang_2012_bivalirudin.R
(the asterisk in Zhang 2012 Table 1 denotes the per-subject baseline
demographic value).Zhang_2012_bivalirudin.R (linear-deviation effect on EC50:
EC50_i = theta_EC50 * exp(eta_EC50) * (1 + 1.70 * (RBC - 4.40))).HGB (hemoglobin mass concentration), HCT
(hematocrit volume fraction), and WBC (white blood cell
count). Document time-varying-vs-baseline-only status in
covariateData[[RBC]]$notes. RBC and HCT correlate via mean
corpuscular volume (HCT ~ RBC * MCV / 10); models that retain both
should record the dependency in notes.INR = INR_BASE + inrmax * (1 - (coag_s3 + coag_l3)/2) per
Xia 2024 supplement Section 1.1) so the simulated INR returns to the
subject-specific baseline when the drug is removed.covariateData[[INR_BASE]]$notes; the Xia 2024 simulation
uses the total-cohort mean of 1.13 (Table 1).INR_BASE, BL_INR, INRBASE –
pre-medication INR column in NONMEM data sets; document the
source-column name per-model in
covariateData[[INR_BASE]]$source_name.Xia_2024_warfarin.R
(additive baseline in the INR observation equation; cohort mean 1.13, SD
0.59 per Xia 2024 Table 1).INR variable). Healthy subjects with
no anticoagulation typically have INR around 1.0; the Hamberg / Xia 2024
model treats deviations from 1.0 as a subject-specific covariate rather
than an estimated parameter so the model returns to the observed
baseline when warfarin is withdrawn. Ratified canonically on 2026-05-16
alongside the Xia 2024 warfarin extraction.Yoshioka_2018_FXa_inhibitors_mbma. Must be > 0 because
downstream PD equations evaluate log(PTR).PTR, PT_RATIO, x – in
Yoshioka 2018 Eq. 1 / Eq. 2 the symbol x is used for the PT
ratio; document the source-column name per-model in
covariateData[[PTR]]$source_name.Yoshioka_2018_FXa_inhibitors_mbma.R (model-based
meta-analysis: per-arm population-mean PTR is input; outputs are per-arm
event probability of ischemic stroke/SE and major bleeding).INR_BASE (a
time-fixed baseline INR scalar used as an additive constant in warfarin
K-PD models). PTR is time-varying and must be supplied externally
(typically computed from an upstream popPK -> PT-ratio model for the
FXa inhibitor of interest, e.g., Girgis 2014 rivaroxaban, Leil 2014 /
Chang 2016 apixaban, Krekels 2016 / Koretsune 2015 edoxaban). Yoshioka
2018 corrects all PT measurements to RecombiplasTin reagent equivalence
per Gosselin 2016 before computing the ratio; downstream models that
consume PTR should document the reagent-correction convention they
assume. Scope: specific until a second model ratifies the canonical
name.covariateData[[VWF]]$units.
Some sources report VWF:Ag (antigen) versus
VWF:RCo (ristocetin cofactor activity); record which assay
was used in covariateData[[VWF]]$notes.(VWF / ref)^exponent. Reference values observed: 118 IU/dL
(Nestorov 2014, study-population median).VWF is the
universal abbreviation. Source papers may write vWF
(lowercase v) or specify the assay (VWF:Ag).Nestorov_2014_factorviii.R (reference 118 IU/dL, exponent
-0.343 on CL; VWF antigen).covariateData[[VWF]]$notes.FVIII-lowest (ever-lowest measurement) covariate. Distinct
from the model’s observed FVIII:C time profile after DDAVP – FVIIIRECENT
is the single pre-dose anchor value.(FVIIIRECENT / ref)^exponent. Reference value observed:
0.15 IU/mL (Schütte 2018, study-population median FVIII-recent in Table
1).FVIII-recent, FVIII_recent,
fviii_recent, FVIIIRECENT – the
recently-measured FVIII:C column. Document the source-column name
per-model in covariateData[[FVIIIRECENT]]$source_name.Schutte_2018_desmopressin.R (reference 0.15 IU/mL;
exponents +0.74 on baseline FVIII, -0.61 on V1, -0.73 on CL; Schütte
2018 Table 2 final covariate model).MISSING_FVIIIRECENT covariate, and simulation users are
expected to supply FVIIIRECENT for every simulated patient. Distinct
from FVIII-lowest (ever-lowest historical FVIII:C), which
Schütte 2018 tested but did NOT retain in the final covariate model.
Ratified canonically on 2026-05-30 alongside the Schütte 2018
desmopressin extraction.log10(DDIMER) / median(log10(DDIMER)) (Sherer 2012) or with
categorical strata (Sherer 2012 sensitivity analysis groups: <=150,
151-300, 301-900, >900 ng/mL). Reference values observed: 326 ng/mL
(Sherer 2012 cohort median; log10 approx 2.513).C^(D-dimer) – used in Sherer_2012_AAA.R
(the symbol in Sherer 2012 Methods equation page 2).Sherer_2012_AAA.R
(proportional log10-transformed covariate on the baseline AAA growth
rate beta1 (e_ddimer_b1 = 0.90 mm/year) and on
the first derivative of growth rate with size beta2
(e_ddimer_b2 = 0.37/year)).AAA_DIAM / median(AAA_DIAM) so the effect coefficients
represent the contribution at the cohort median. Reference value
observed: 32.7 mm (Sherer 2012 cohort median; q1 30.8, q3 36.0).Y(0) – used in Sherer_2012_AAA.R (the
symbol in Sherer 2012 Methods equation page 2; the baseline screening
ultrasound diameter).Sherer_2012_AAA.R
(proportional covariate on all three individual-level parameters:
e_aaadiam_b0 = 32.6 mm on baseline size beta0,
e_aaadiam_b1 = 2.03 mm/year on baseline growth rate beta1,
and e_aaadiam_b2 = 0.59/year on the first derivative of
growth rate with size beta2).Tiraboschi_2025_amlitelimab.R).EASI – prior canonical name (pre-2026-06-19
standardization audit).BEASI (baseline SCORE_EASI) – used in
Tiraboschi_2025_amlitelimab.R.Tiraboschi_2025_amlitelimab.R.BEASI), document in
covariateData[[SCORE_EASI]]$notes. Canonical name is
SCORE_EASI without the B prefix to match the
AGE / WT / ALB pattern where
baseline vs time-varying status is recorded in notes rather than the
column name.SCORE_MGADL = 0 by definition. Effect enters as a
baseline covariate on MG-ADL response parameters in gMG cohorts.MGADL – prior canonical name (pre-2026-06-19 SCORE_
family standardization).Valenzuela_2025_nipocalimab.R (reference 7 points;
power-form effect on IDecplacebo and on the slope between
MG-ADL change and IgG reduction).covariateData[[SCORE_MGADL]]$notes.
Canonical name is SCORE_MGADL without a BL
prefix to match the SCORE_EASI / AGE /
WT / ALB pattern.BCVA – prior canonical name (pre-2026-06-19
standardization audit).BVA (baseline visual acuity) – used in
Mulyukov_2018_ranibizumab.R.Mulyukov_2018_ranibizumab.R (baseline SCORE_BCVA used as
the center for the initial-condition draw
g0 = SCORE_BCVA + eta_g0).B prefix to match the SCORE_EASI /
AGE / WT / ALB pattern
(baseline-vs-time-varying status recorded in
covariateData[[SCORE_BCVA]]$notes). Scope is
specific until a second ophthalmology model ratifies the
name; at that point promote to general.covariateData[[PREV_AE_SCORE]]$notes)IF (PREV_AE_SCORE == 0) ...) or
via piecewise-FPS indicator decomposition; the natural reference is
PREV_AE_SCORE = 0 (no prior AE).PREVSCOR – used in
Girard_2012_pimasertib.R (CTCAE 0..3 ocular-AE score).Girard_2012_pimasertib.R (Markov-state covariate that
selects per-previous-score logit thresholds
b01/b11/b21 and
b02/b12/b22 and
per-previous-score emax levels; reset to 0 at TIME = 0 per
source IF (TIME.EQ.0) PREVSCOR=0).PREV_AE_SCORE = 0 at the first observation of every subject
and update each subsequent observation to the previous observation’s
sampled score – matching the NONMEM
IF (TIME.EQ.0) PREVSCOR=0 / PREVSCOR = DV
carry-forward idiom. Distinct from PAIN (continuous
baseline pain score) and from SCORE_MGADL /
SCORE_EASI (continuous severity scores not modelled as
Markov states).CDR_SOB – prior canonical name (pre-2026-06-19
standardization audit).CDR_bsl – used in Delor_2013_alzheimer.R
(baseline CDR-SOB at study entry).CDR – alternative bare-name often seen in ADNI /
CAMD-style NONMEM datasets.Delor_2013_alzheimer.R
(time-fixed baseline covariate; enters both the per-subject DOT power
form ((SCORE_CDR_SOB / 2)^e_cdr_sob_dot with
e_cdr_sob_dot = -0.072) and the per-subject
slow-progression mixture-logit additive form
(+ e_cdr_sob_slow * (SCORE_CDR_SOB - 1) with
e_cdr_sob_slow = -1.27)).SCORE_CDR_SOB
without a _BL suffix to match the SCORE_EASI /
SCORE_MGADL / SCORE_BCVA pattern
(baseline-vs-time-varying status recorded in
covariateData[[SCORE_CDR_SOB]]$notes). The CDR sum-of-boxes
form is distinct from the global CDR rating (CDR_GLOBAL, a
0 / 0.5 / 1 / 2 / 3 ordinal); the sum-of-boxes is preferred in
disease-progression modelling for its finer granularity. Ratified
canonically on 2026-05-16 alongside the Delor 2013 extraction.covariateData[[SCORE_ADAS_COG]]$units)ADAS_COG – prior canonical name (pre-2026-06-19
standardization audit).ADAS_bsl – used in Delor_2013_alzheimer.R
(baseline ADAS-cog total-11 at study entry).ADAS, ADAS_COG_11,
ADAS_COG_13 – alternative bare-name forms seen across ADNI
/ CAMD datasets.Delor_2013_alzheimer.R
(time-fixed baseline covariate; ADAS-cog total-11 form; enters the
per-subject DOT power form
(SCORE_ADAS_COG / 12.67)^e_adas_cog_dot with
e_adas_cog_dot = -0.0439).SCORE_ADAS_COG (no _BL suffix;
baseline-vs-time-varying recorded in notes). The total-11 vs total-13
form must be documented per-model in
covariateData[[SCORE_ADAS_COG]]$units because the same
numeric SCORE_ADAS_COG value has different clinical interpretation
across the two forms. Conrado 2014 uses ADAS-cog as the modelled
observation (response variable) rather than as a baseline covariate;
that model file therefore does not list SCORE_ADAS_COG in its
covariateData. Ratified canonically on 2026-05-16 alongside
the Delor 2013 extraction.MMSE – prior canonical name (pre-2026-06-19
standardization audit).MMSE_bsl – used in Delor_2013_alzheimer.R
(baseline SCORE_MMSE at study entry).Delor_2013_alzheimer.R
(time-fixed baseline covariate; modifies the per-subject
disease-progression acceleration parameter alpha via a power form
(SCORE_MMSE / 26)^e_mmse_alpha with
e_mmse_alpha = -2.01).SCORE_MMSE
(no _BL suffix; baseline-vs-time-varying recorded in
notes). SCORE_MMSE is the inverse-direction counterpart of SCORE_CDR_SOB
/ SCORE_ADAS_COG (SCORE_MMSE high = healthy; SCORE_CDR_SOB /
SCORE_ADAS_COG high = impaired); covariate-effect coefficient signs are
therefore typically opposite to those for SCORE_CDR_SOB /
SCORE_ADAS_COG. Ratified canonically on 2026-05-16 alongside the Delor
2013 extraction.FAQ – prior canonical name (pre-2026-06-19
standardization audit).FAQ_bsl – used in Delor_2013_alzheimer.R
(baseline SCORE_FAQ at study entry).Delor_2013_alzheimer.R
(time-fixed baseline covariate; enters the per-subject slow-progression
mixture-logit additive form + e_faq_slow * (SCORE_FAQ - 1)
with e_faq_slow = -0.341).SCORE_FAQ (no
_BL suffix; baseline-vs-time-varying recorded in notes).
Distinct from the cognitive scores (SCORE_CDR_SOB / SCORE_ADAS_COG /
SCORE_MMSE): SCORE_FAQ measures instrumental activities of daily living
rather than cognitive performance, and adds incremental information
about disease-stage severity in MCI cohorts. Ratified canonically on
2026-05-16 alongside the Delor 2013 extraction.RHPNM – prior canonical name (pre-2026-06-19
standardization audit).SCORE_RHPNM – used in
Delor_2013_alzheimer.R (baseline normalized hippocampal
volume; Delor 2013 derivation:
RHPNMbsl_i = HIPVbsl_i / HPNMbsl_i where
HPNMbsl_i = Age_i * (-26.6268 + EICVbsl_i * 0.0016 + 3340.4395)).Delor_2013_alzheimer.R
(time-fixed baseline covariate; enters the per-subject slow-progression
mixture-logit additive form
+ e_rhpnm_slow * (SCORE_RHPNM - 1) with
e_rhpnm_slow = 7.5, a strongly positive effect indicating
that less atrophic hippocampi (SCORE_RHPNM closer to 1) are associated
with a higher probability of being in the slow-progressing
subpopulation).HIPV canonical not yet registered; raw
HIPV is confounded with head size and age, hence the need for the
normalisation). The Delor 2013 paper notes that the same effect is only
marginally significant with unnormalised HIPV (P = 0.02) but strongly
significant with the normalised form (SCORE_RHPNM). The exact age / EICV
regression coefficients are paper-specific and any future model adopting
this canonical should re-derive the normalisation for its own population
or document why the Delor 2013 regression is reused. Ratified
canonically on 2026-05-16 alongside the Delor 2013 extraction.slope * (ACUTE_MED_DAYS - 5) above 5 (Fiedler-Kelly 2020).
The 5-day breakpoint reflects the clinical guideline for
medication-overuse headache.FiedlerKelly_2020_fremanezumab_em.R
and FiedlerKelly_2020_fremanezumab_cm.R.FiedlerKelly_2020_fremanezumab_em.R (slope 0.438 d/d,
episodic migraine), FiedlerKelly_2020_fremanezumab_cm.R
(slope 0.460 d/d, chronic migraine).general.e_amload2_vp_sap and
e_amload2_vp_sap + e_amload3_vp_sap respectively.AMLOAD – prior canonical name (pre-2026-06-19 DIS_
prefix standardization) and source-paper column in
Sahota_2015_miridesap.R.NA_NA_miridesap.R
(DDMODEL00000262; Sahota 2015 Eq. 2 multiplicative effect on SAP
peripheral volume V4: V4 = V4_ref * (1 + e_amload2_vp_sap *
I(DIS_AMYLOID_LOAD>=2) + e_amload3_vp_sap *
I(DIS_AMYLOID_LOAD>=3)); reported effects e_amload2_vp_sap = 6.39 /
e_amload3_vp_sap = 26.39 yielding ~7.4x V4 at moderate load and ~33.8x
at large load), Sahota_2015_miridesap.R (paper-only
extraction of the same Sahota 2015 final model with identical Eq. 2
effect on V4; values 6.39 / 26.39 taken from Table 2).DIS_AMYLOID_LIVER for the binary hepatic-involvement
modifier. Ratified canonically on 2026-05-15 alongside the
DDMODEL00000262 / Sahota 2015 extraction. Renamed from
AMLOAD to DIS_AMYLOID_LOAD on 2026-06-19 per
the canonical-register standardization audit (operator decision to apply
the DIS_<concept> prefix uniformly to disease-state
indicators and to spell out amyloidosis rather than the shortened
AM prefix that collides with chemistry symbols).AMLIVER – prior canonical name (pre-2026-06-19 DIS_
prefix standardization) and source-paper column in
Sahota_2015_miridesap.R.NA_NA_miridesap.R
(DDMODEL00000262; Sahota 2015 Eq. 2 multiplicative effect on SAP
intercompartmental clearance Q4: Q4 = Q4_ref * (1 + e_amliver_q4 *
DIS_AMYLOID_LIVER); reported effect 4.01, yielding ~5x Q4 in patients
with hepatic amyloid), Sahota_2015_miridesap.R (paper-only
extraction of the same Sahota 2015 final model with identical Eq. 2
effect on Q4; value 4.01 from Table 2).DIS_AMYLOID_LOAD grade so the model can express both
general amyloid burden and the specific hepatic-clearance modifier (the
SAP-CPHPC complex is cleared by the liver, motivating the
hepatic-amyloid-specific Q4 effect). Ratified canonically on 2026-05-15
alongside the DDMODEL00000262 / Sahota 2015 extraction. Renamed from
AMLIVER to DIS_AMYLOID_LIVER on 2026-06-19 per
the canonical-register standardization audit (operator decision to apply
the DIS_<concept> prefix uniformly to disease-state
indicators and to spell out amyloidosis rather than the shortened
AM prefix that collides with chemistry symbols).ORGF – used in Vet_2016_midazolam.R
(DDMODEL00000249 NMTRAN $INPUT column; values 0..>=4).
Renamed to canonical ORG_FAIL_COUNT when assembling input
data for the packaged model.Vet_2016_midazolam.R
(per-stratum typical CL values: ORG_FAIL_COUNT=0 fixed at 1.6 L/h for a
5 kg child with CRP=32 mg/L; ORG_FAIL_COUNT=1 -> 1.29 L/h;
ORG_FAIL_COUNT=2 -> 0.957 L/h; ORG_FAIL_COUNT=3 -> 0.842 L/h;
ORG_FAIL_COUNT>=4 -> 0.678 L/h).covariateData[["ORG_FAIL_COUNT"]]$notes. Decompose inside
model() into binary indicators
(orgf1 <- (ORG_FAIL_COUNT == 1),
orgf2 <- (ORG_FAIL_COUNT == 2),
orgf3 <- (ORG_FAIL_COUNT == 3),
orgf_ge4 <- (ORG_FAIL_COUNT >= 4)) and select
per-stratum CL with mutually-exclusive multiplicative-flag arithmetic.
Ratified canonically on 2026-05-06.(SAPS_II / ref)^exponent. Reference value observed: 50
points (Abboud 2009 typical-subject reference for the septic-shock
cohort, mean SAPS II = 64 +/- 23).SAPS II (with whitespace, as printed in the source
paper’s prose) – used in Abboud_2009_epinephrine.R.Abboud_2009_epinephrine.R (power exponent -0.67 on
epinephrine CL with reference 50; higher SAPS II is associated with
lower clearance), Polito_2016_fludrocortisone.R (power
exponents +0.019 on apparent oral fludrocortisone CL/F and +0.036 on the
absorption lag time Tlag, both with reference 53 = the cohort median
SAPS II among the 14 patients with detectable plasma concentrations;
higher SAPS II is associated with longer absorption lag and faster
apparent oral clearance).Example models rather than registering a new canonical.
Should not be confused with SAPS I or SAPS 3 (different scoring rules /
item sets) – register those under separate canonicals if a future paper
uses them. Ratified canonically on 2026-05-18 alongside the Abboud 2009
epinephrine extraction. ### RACHS1 (canonical for Risk
Adjustment for Congenital Heart Surgery 1 (RACHS-1)
category)RACHS-1 – the publication’s printed form with a hyphen,
not a valid R identifier; renamed to RACHS1 when assembling
input data.Oualha_2014_epinephrine.R (decomposed inside
model() into a binary indicator
rachs1_high <- (RACHS1 >= 3) that selects an additive
log-shift on SV*SVR_max from 0.44 to 0.26 for the high-risk pool).model() into mutually
exclusive binary indicators matching the source’s pooling (e.g.,
rachs1_high <- (RACHS1 >= 3)) and document the
pooling rule in covariateData[["RACHS1"]]$notes.Oualha_2014_epinephrine.R (enters Eq. 7 as the additive
offset in MAP = HR * SV*SVR + CVP; the vignette defaults to the cohort
median 11 mmHg when CVP is not supplied per subject).HIGHRISK / RISK (Thuo 2011 NONMEM
notation; paper writes “high risk” in prose and structural equations as
[1 + theta3 * (high risk)]) – used in
Thuo_2011_ciprofloxacin.R.Thuo_2011_ciprofloxacin.R (multiplicative fractional effect
on apparent CL: 1 + (-0.283) * MORTRISK_HIGH, i.e., a 28.3%
reduction in apparent oral clearance for high-risk children; the
standardised CL falls from 42.7 L/h/70 kg in low/intermediate-risk to
30.6 L/h/70 kg in high-risk; the paper attributes the contrast to
delayed gastric emptying / impaired gut absorption in critically ill
malnourished children).SAPS_II (continuous adult-ICU
severity score), RACHS1 (paediatric cardiac surgery risk
category), and ORG_FAIL_COUNT (integer organ-failure count)
– each is its own canonical with paper-specific scoring rules. Future
paediatric-severe-malnutrition popPK papers that retain the Berkley 2003
three-stratum score (or a close variant) can reuse this canonical;
per-model covariateData[[MORTRISK_HIGH]]$notes must
document the exact criteria the source paper used. Ratified canonically
on 2026-05-21 alongside the Thuo 2011 ciprofloxacin extraction.mechanical ventilation – printed phrase in Georges 2009
Table 1 (no/yes count column); renamed to canonical
MECH_VENT when assembling input data. Used in
Georges_2009_ceftazidime.R.Georges_2009_ceftazidime.R (selector effect on V1 via the
multiplicative log-effect form:
vc = exp(lvc + e_mech_vent_vc * MECH_VENT + etalvc), with
lvc = log(18.9) at MECH_VENT = 0 and
e_mech_vent_vc = log(9.02 / 18.9) recovering TVV1 = 9.02 L
at MECH_VENT = 1; Discussion attributes the V1 contraction to
PPV-induced hormonal fluid shifts).RRT_HEMODIAL_STATUS /
RRT_CRRT_STATUS (renal-replacement-therapy treatment-status
flags) and from SAPS_II (continuous severity score). When a
future critically-ill popPK paper reuses MECH_VENT with the same
definition (binary, time-fixed at ICU admission), extend the
Example models list rather than registering a new
canonical; promotion to general scope is appropriate after a second
corroborating paper. Per-model
covariateData[[MECH_VENT]]$notes must document whether the
indicator is time-fixed at admission or re-evaluated as ventilation
status changes during the ICU stay. Ratified canonically on 2026-06-10
alongside the Georges 2009 ceftazidime extraction.ICU_ADM_POLYTRAUMA + ICU_ADM_POSTSURG + ICU_ADM_MEDICAL = 1
for every subject). Time-fixed per subject (the admission category is
recorded once at ICU entry). Used as a categorical-stratum covariate on
PK distribution parameters that differ between trauma-induced fluid
shifts and elective surgical or medical patient profiles.Admission for polytrauma – printed row label in Georges
2009 Table 1 (count column); renamed to canonical
ICU_ADM_POLYTRAUMA when assembling input data. Used in
Georges_2009_ceftazidime.R.Georges_2009_ceftazidime.R (selector effect on V2: TVV2 =
theta6 = 57.1 L at ICU_ADM_POLYTRAUMA = 1; multiplicative log-effect
form encodes the shift relative to the ICU_ADM_MEDICAL reference stratum
as e_icu_adm_polytrauma_vp = log(57.1 / 13.6)).MORTRISK_HIGH (severity score), SAPS_II
(continuous severity), and ORG_FAIL_COUNT (organ-failure
count) – all of which describe critical-illness severity rather than
admission etiology. Future adult-ICU popPK papers that retain the same
three-stratum etiology partition can extend the
Example models list and the scope may be promoted to
general after a second corroborating paper. Ratified canonically on
2026-06-10 alongside the Georges 2009 ceftazidime extraction.ICU_ADM_POLYTRAUMA for the partition
definition). Time-fixed per subject.Admission for postsurgical – printed row label in
Georges 2009 Table 1 (count column); renamed to canonical
ICU_ADM_POSTSURG when assembling input data. Used in
Georges_2009_ceftazidime.R.Georges_2009_ceftazidime.R (selector effect on V2: TVV2 =
theta7 = 25.7 L at ICU_ADM_POSTSURG = 1; multiplicative log-effect form
encodes the shift relative to the ICU_ADM_MEDICAL reference stratum as
e_icu_adm_postsurg_vp = log(25.7 / 13.6)).ICU_ADM_POLYTRAUMA for
shared scope rationale and partition rules. Specific scope because the
postsurgical stratum definition is cohort-dependent and the partition
into the three Georges 2009 categories is not universal across ICU popPK
papers (some cohorts merge postsurgical with medical or split it into
elective vs urgent). Ratified canonically on 2026-06-10 alongside the
Georges 2009 ceftazidime extraction.ICU_ADM_POLYTRAUMA for the partition definition).
Time-fixed per subject. Commonly serves as the structural reference
category in models that compare polytrauma / postsurgical strata against
the medical baseline, because medical-reason patients in adult ICU
cohorts tend to have distribution volumes closer to non-critically-ill
individuals (per Georges 2009 Discussion).e_icu_adm_medical_<param> = 0 is implicit and only
the two non-reference effects are estimated.Admission for medical reason – printed row label in
Georges 2009 Table 1 (count column); renamed to canonical
ICU_ADM_MEDICAL when assembling input data. Used in
Georges_2009_ceftazidime.R.Georges_2009_ceftazidime.R (selector effect on V2: TVV2 =
theta8 = 13.6 L at ICU_ADM_MEDICAL = 1; this is the structural reference
of the multiplicative parameterisation – exp(lvp) recovers
13.6 L when both other indicators are zero).ICU_ADM_POLYTRAUMA for
shared scope rationale. Specific scope. The Georges 2009 Discussion
notes ‘Patients with a medical reason for admission presented a volume
of distribution in the same order of magnitude as those of healthy
subjects’ which motivates the choice of this stratum as the structural
reference for V2 in the model file. Ratified canonically on 2026-06-10
alongside the Georges 2009 ceftazidime extraction.covariateData[[HIE_POST]]$notes. Used as a binary state
indicator on PK parameters (most commonly CL).AED – used in
Ezzati_2014_dexmedetomidine_piglet.R (the paper defines
AED as “acute energy depletion”, computed by the time
integral of beta-nucleotide-triphosphate (NTP) peak-area changes during
hypoxia-ischemia; in the final model the AED severity score is used as a
binary 0 / 1 state flag for whether the subject experienced the HI
insult, NOT as a continuous covariate – 9 piglets had AED = 1 throughout
dosing and 1 control piglet had AED = 0). NOTE: the paper’s
AED column-name encodes a calculated NTP-integral severity
score that the model dichotomises to a 0 / 1 state flag; downstream
users assembling event tables should set HIE_POST = 1 from
the HI insult time onward for all HI-exposed subjects.Ezzati_2014_dexmedetomidine_piglet.R (multiplicative factor
on dexmedetomidine CL post-insult:
CL = CLstd * (Wt/70)^0.75 * (1 + Ftemp * (TEMP - 37)) * (1 + (FAED - 1) * HIE_POST)
with FAED = 0.558 – clearance is multiplied by 0.558 in the
post-HI state, i.e. CL is reduced TO 55.8% of the pre-insult value,
equivalent to a 44.2% reduction relative to the pre-insult typical
value).BURN_RECENT (acute hypermetabolic phase
indicator with different physiology and a time-windowed threshold) and
from DIS_* disease-state indicators (which describe chronic
conditions rather than acute insult states). Distinct from
ORG_FAIL_COUNT (integer organ-failure count) and
MORTRISK_HIGH (mortality risk score) – HIE_POST is a binary
state flag for a specific insult exposure, not a downstream severity or
outcome score. Future papers that report a continuous HI severity score
(e.g., AED integral, Sarnat staging, MRI biomarker) could either (a)
re-use HIE_POST as a binary derived from the continuous
score, or (b) register a sibling continuous canonical (e.g.,
HIE_SEVERITY) – operator decides per paper. Ratified
canonically on 2026-06-10 alongside the Ezzati 2014 dexmedetomidine
piglet extraction.(BGENE21 / ref)^exponent. Reference values observed: 32 in
Narwal 2013 (study-population median was 33), 12.04 in Zheng 2016
(median of the SLE phase IIb cohort, range 0.32-38.59).Narwal_2013_sifalimumab.R (reference 32, exponent 0.0558 on
CL), Zheng_2016_sifalimumab.R (reference 12.04, power
effect on CL with exponent 0.09).BGENE4,
IFN_SIG, …) to avoid conflating panel definitions.Almquist_2022_anifrolumab.R (binary high-IFN indicator on
CL).BGENE21
when the paper reports both. The high/low cut-off is paper-specific
(commonly the population median) and must be documented in
covariateData[[BGENE21_HIGH]]$notes for every model that
uses this covariate. Operator decision (2026-04-28): use
BGENE21_HIGH (not IFNGS_HIGH) so the link to
the existing BGENE21 register entry is explicit while the
binary nature stays visible in the column name.(EOS / ref)^exponent.BEOS (baseline EOS) – used in
Kotani_2022_astegolimab.R.Kotani_2022_astegolimab.R (reference 180 cells/uL,
baseline).B prefix to match the
SCORE_EASI / AGE / WT /
ALB pattern; baseline-vs-time-varying status is documented
in covariateData[[EOS]]$notes.(BLBCELL / ref)^exponent. Reference value observed: 200
cells/uL (Yu 2022, median of the pooled five-study cohort).Bcell0 – used in
Yu_2022_ofatumumab.R.BBCC (NHL Phase I/Ib/II convention; values in 10^6
cells/L = cells/uL) – used in Lu_2019_polatuzumab.R.Yu_2022_ofatumumab.R
(power effect on the maximum B-cell-lysis stimulatory effect Emax,
exponent 0.275, reference 200 cells/uL),
Lu_2019_polatuzumab.R (two distinct effects: power on
CL_INF with input floored at 1 cell/uL, and a thresholded power on CL_T
with the BLBCELL/121-cells/uL ratio floored at 1),
Lu_2017_polatuzumab_neuropathy.R (carries forward the Lu
2019 acMMAE PK-side effects via the inlined acMMAE popPK layer; not used
directly by the Lu 2017 TTE PD layer).covariateData[[BLBCELL]]$notes.covariateData[[BL_PARP_PBL]]$units).(BL_PARP_PBL / ref)^exponent. Reference value observed:
90.8 pmol/10^6 PBL (Wang 2015 rucaparib; population typical baseline E0
used as a stand-in for the unreported study-cohort median because the
paper reports only the typical E0 and the exponent value, not the
numeric BLB_median).BLB (Wang 2015’s notation for “baseline level in blood
/ PBL”) – used in Wang_2015_rucaparib.R.Wang_2015_rucaparib.R
(power effect on residual maximum-inhibition parameter Emin; exponent
0.620 with the form
Emin = TV(Emin) * (BL_PARP_PBL / BLB_median)^alpha; PBL
paired with a separate tumor-tissue PARP activity covariate that is not
yet a registered canonical because the units differ – pmol/mg protein
for tumor vs pmol/10^6 PBL for blood).covariateData[[BL_PARP_PBL]]$notes. The
paper does not publish the numeric study-cohort median of BLB used to
center the covariate; the model file uses 90.8 pmol/10^6 PBL (the
population typical baseline E0 reported in Wang 2015 Table 2) as a
defensible default reference and documents the assumption in the
vignette’s Assumptions and deviations section.covariateData[[CSF1]]$units.(CSF1 / ref)^exponent. Reference values observed: 549 pg/mL
(Yang 2024 axatilimab; pooled-cohort median).BLCSF1 (baseline CSF-1) and BL_CSF1
(model-parameter notation in Monolix / NONMEM control streams).Yang_2024_axatilimab.R
(baseline-only covariate on linear clearance CL with power
exponent 0.912 and on the model parameter BL_CSF1 with
power exponent 0.656; reference 549 pg/mL).covariateData[[CSF1]]$notes should document the assay used
and any LOQ-related imputation for samples below the assay’s limit of
detection.covariateData[[CRP]]$description and notes
must state the assay type (standard vs hs-CRP) and whether the column
carries a baseline-only or time-varying value, including the
paper-specific reference value used for power scaling.covariateData[[CRP]]$units).(CRP / ref)^exponent or exponential effects
exp(coef * (CRP - ref)). Reference values observed: 4.23
mg/L (Moein 2022, IBD standard assay), 4.31 mg/L (Moein 2022 Table 3
median), 5.21 mg/L (Thakre 2022, baseline hs-CRP), 7.41 mg/L (Chua 2025,
baseline standard assay), 14.2 mg/L (Xu 2019, baseline standard assay),
15.7 mg/L (Ma 2020, baseline standard assay), 0.837 mg/dL = 8.37 mg/L
(Wang 2020, IBD standard assay; the model carries the source unit
mg/dL).hsCRP – high-sensitivity CRP (mixed-case preserved from
earlier register drafts).HSCRP – all-caps variant.CRPHS – used in Thakre_2022_risankizumab.R
(baseline, high-sensitivity assay).BLCRP – baseline CRP; used in
Xu_2019_sarilumab.R and
Ma_2020_sarilumab_das28crp.R.Thakre_2022_risankizumab.R,
Xu_2019_sarilumab.R, Chua_2025_mirikizumab.R,
Moein_2022_etrolizumab.R,
Ma_2020_sarilumab_das28crp.R,
Wang_2020_ontamalimab.R (mg/dL, reference 0.837).hsCRP,
BLCRP, and standard-assay CRP canonicals were
merged on 2026-04-20 to a single general-scope CRP
canonical. Assay type (standard vs hs-CRP), baseline-vs-time-varying
status, and the paper-specific reference value all live in each model’s
covariateData[[CRP]]$description / notes. Only
aggregate values from hs-validated assays as CRP when the downstream
analysis relies on low-range sensitivity; for most inflammatory-disease
cohorts (IBD, RA/PsA), baseline CRP is well above the hs-sensitivity
range and the distinction is moot.covariateData[[AAG]]$units if a different unit is
used.AAG <= 1.34 and AAG > 1.34).
Reference values observed: 1.34 g/L (Kloft 2006 / Netterberg 2017,
cohort median in mixed adult-cancer cohort).AAG – used in Netterberg_2017_docetaxel.R
(per the bundle’s NM-TRAN $INPUT block; matching Kloft 2006).AGP1 – used in Ozawa_2007_docetaxel.R
(Appendix I $INPUT block); reported in mg/dL with conversion to
canonical g/L via AAG_g_per_L = AGP_mg_per_dL / 100.Netterberg_2017_docetaxel.R (piecewise-linear effects on
baseline ANC with separate low-AAG and high-AAG slopes around median
1.34 g/L; linear effect on the drug-effect slope SL via
(1 + theta * (AAG - 1.34))),
Ozawa_2007_docetaxel.R (multiplicative power-form effect on
the linear drug-effect slope:
SLOPE = theta_SLOPE * (AAG / 0.94)^e_aag_slope with
e_aag_slope = -1.38; reference value 0.94 g/L from the
published NONMEM control stream AGPm = 94 mg/dL).covariateData[[AAG]]$notes. Distinct from
CRP (a different acute-phase reactant with different
binding properties).covariateData[[IL6]]$notes whether the column
is baseline-only or time-varying.(IL6 / ref)^exponent or with the log-transformed form
(log(IL6 * 1000) / log(ref * 1000))^exponent that some
legacy NONMEM analyses adopt. Reference values observed: 20 pg/mL (Frey
2013 baseline; the formula (log(IL-6 * 1000)/9.9)^exponent
is the algebraic equivalent of
(log(IL-6) / log(20))^exponent after the constant-factor
rescaling that ties the reference log to 9.9 = log(20000)).BLIL6, bIL6, IL6_BASE –
baseline IL-6 (used in some NONMEM control streams; canonical drops the
BL prefix per the EOS /
SCORE_EASI / AGE convention with
baseline-vs-time-varying status documented in per-model notes).IL-6, IL_6 – punctuation variants seen in
publication tables and figures.Frey_2013_tocilizumab.R (baseline IL-6, reference 20 pg/mL;
log-transformed power effects on EC50, BASE, and the DMARD
background-effect parameter).covariateData[[IL6]]$notes. Frey 2013’s formula relies on
the relative log-IL-6 ratio rather than the linear concentration, so the
column units must be pg/mL exactly (not ng/mL) for the published
exponents to apply unchanged.covariateData[[HDLC]]$units (1 mmol/L ~=
38.67 mg/dL for cholesterol).(HDLC / ref)^exponent. Reference value observed: 54 mg/dL
(Frey 2010 pooled-cohort median).HDL-C – Frey 2010 spelling with hyphen.HDL_C – common alternative spelling.Frey_2010_tocilizumab.R (mg/dL, reference 54; small
negative exponent -0.2 on linear CL; the paper interprets the effect as
a body-size surrogate rather than a mechanism).HDLC which captures only the HDL fraction.covariateData[[TCHOL]]$units (1 mmol/L ~=
38.67 mg/dL for cholesterol).1 + theta * (TCHOL - ref) or power
scaling (TCHOL / ref)^exponent. Reference value observed: 3
mmol/L (Archary 2018, severely malnourished pediatric LPV cohort,
baseline mean 2.7-2.9 mmol/L).CHOL – Archary 2018 NONMEM column abbreviation; the
universal short form.TC – alternative abbreviation common in lipid-panel
literature.Archary_2018_lopinavir.R (mmol/L, reference 3; linear
effect on apparent CL/F: 1 + 0.207 * (TCHOL - 3); serves as
a surrogate for nutritional / hepatic-function recovery rather than a
mechanistic effect).HDLC (HDL fraction only) and from
CRP / ALB / AAG (separate
canonicals for related malnutrition / inflammation markers).covariateData[[TRIG]]$units (1 mmol/L ~=
88.5 mg/dL for triglyceride).(1 + theta * (TRIG - ref)) or power
scaling (TRIG / ref)^exponent. Reference value observed:
5.3 mmol/L (Archary 2019 lamivudine equation centring – the equation
centring in the source is reported as the cohort average; cohort median
in Archary 2019 Table 1 is 2.2-2.3 mmol/L, see model-file Errata).Archary_2019_lamivudine.R (mmol/L, reference 5.3;
linear-deviation effect on Vc/F with coefficient -0.13 per mmol/L
deviation from the reference; lower triglyceride implies higher apparent
central volume).covariateData[[TRIG]]$units is
load-bearing because the centring reference and slope are
unit-specific.covariateData[[LDLC]]$units (1 mmol/L ~=
38.67 mg/dL for cholesterol).(LDLC / ref)^exponent for the baseline-LDLC covariate role,
or with no reference (used directly as the PD response state) when it is
the modelled output. Reference values observed: 211 mg/dL (Pu 2021 HoFH
typical-patient definition).LDL-C – common spelling with hyphen.LDL_C – common alternative spelling.LDLBL (baseline LDL-C) – used in
Pu_2021_evinacumab.R (Pu 2021 NM-TRAN $INPUT column for
centred baseline LDL-C as a covariate on IC50).Pu_2021_evinacumab.R
(mg/dL, baseline reference 211 mg/dL; power exponent -1.17 on IC50,
where higher baseline LDL-C predicts a smaller IC50 and therefore
greater sensitivity to evinacumab; LDL-C is also the PD output state
initialised at the baseline value).HDLC (high-density lipoprotein cholesterol)
and from any total-cholesterol or non-HDL-C derivation. When LDL-C is
both the response variable AND a covariate (as in Pu 2021, where
baseline LDLC drives IC50 and the time-varying state is the modelled
PD), document the dual role in
covariateData[[LDLC]]$notes.covariateData[[ANGPTL3]]$units.(ANGPTL3 / ref)^exponent. Reference value observed: 0.08
mg/L (Pu 2021 typical-patient median).ANGBL (baseline ANGPTL3) – used in
Pu_2021_evinacumab.R (Pu 2021 NM-TRAN $INPUT column for
centred baseline ANGPTL3).ANGBL0 – alternative Pu 2021 raw column.Pu_2021_evinacumab.R
(mg/L, baseline reference 0.08 mg/L; power exponent +0.405 on Vmax,
where higher baseline target predicts a faster saturable elimination –
biologically consistent with evinacumab being co-cleared along with
bound ANGPTL3).covariateData[[ANGPTL3]]$notes.TVPARAM + theta * (FPCSK9 / ref) or
power-form (FPCSK9 / ref)^theta. Reference values observed:
72.9 ng/mL (Martinez 2019 time-varying median).Martinez_2019_alirocumab.R (time-varying; additive-linear
effect on Km with slope -0.541 per (FPCSK9/72.9), reference
72.9 ng/mL).FIL6R,
FTNF) rather than overloading FPCSK9.
Per-model covariateData[[FPCSK9]]$notes should state
whether the value is baseline-only or time-varying and how missing
values were imputed (Martinez 2019 used LOCF).covariateData[[SBCMA]]$units.(SBCMA / ref)^exponent. Reference value observed: 50 ng/mL
(Papathanasiou 2025 typical-patient definition).SBCMABL (baseline soluble BCMA) – used in
Papathanasiou_2025_belantamab.R.Papathanasiou_2025_belantamab.R (ng/mL, reference 50; power
exponents on initial CL +0.113, on ADC Vc +0.0401, on Imax +0.160).HER2_ECD already exists for HER2; an analogous
SCD20, SCD38 would follow the same pattern).
Multiple myeloma populations show sBCMA spanning roughly 2 to 2,000
ng/mL, so the (SBCMA/50)^exponent form should be evaluated with care
over the full clinical range.covariateData[[HBA1C]]$units when a paper reports IFCC
units (IFCC mmol/mol = 10.93 * NGSP% - 23.50).(HBA1C / ref)^exponent or linear-deviation form. Reference
value observed: 5.88% (Oniki 2018 NAFLD-risk dataset baseline
mean).HbA1c – used in Oniki_2018_nafld_risk.R
(dataset column for glycated hemoglobin in %).Oniki_2018_nafld_risk.R (%, reference 5.88; power exponent
-3.34 for (HBA1C / 5.88) on the (BMI50 - 17)
half-saturation offset of the sigmoidal logit-of-NAFLD function, Oniki
2018 Eq. 4 / Figure 2c).FPG (baseline fasting plasma
glucose), which is routinely reported alongside HbA1c in T2DM /
metabolic-syndrome populations. Distinct from GLU
(time-varying within-subject glucose regressor). Ratified canonically
alongside the Oniki 2018 NAFLD-risk extraction.Kin = TRACP5B_BL * Kout,
marker(0) <- TRACP5B_BL) and as a continuous covariate
on disease-progression and BMD-coupling parameters (power-model
standardisation to the cohort median).covariateData[[TRACP5B_BL]]$units). Some
clinical contexts report TRACP-5b in U/L; 1 U/L = 100 mU/dL only if the
assays are calibrated equivalently – per-model unit documentation is
load-bearing.(TRACP5B_BL / ref)^exponent. Reference value observed: 400
mU/dL (Mori 2018 cohort – mean baseline 401.1 mU/dL across N = 306
Japanese primary-osteoporosis patients; the paper says continuous
covariates were standardised to their cohort median but the numeric
median is not published, so the rounded cohort mean serves as the
documented reference).TRACP-5b – text-form of the marker name used in Mori
2018; transliterated to TRACP5B_BL for the per-subject
baseline canonical name.Mori_2018_zoledronicAcid.R (mU/dL, reference 400; power
exponents -1.534 on EKD50, -1.350 on Slope, -1.319 on T50; and -1.112 on
Scale in the active-treatment arm only – the Scale covariate is gated by
ON_TREATMENT).covariateData[[TRACP5B_BL]]$notes. Ratified canonically
alongside the Mori 2018 zoledronic-acid BMD extraction. Distinct from
any time-varying serum TRACP-5b state output (which would be a model
output rather than a covariate column).bmd(0) <- BMD_BL) and as the
per-subject deviation reference inside the BMD effect-compartment ODE
(d/dt(bmd) <- Ke0 * [Scale * (marker - Marker0) - (bmd - BMD_BL)]).covariateData[[BMD_BL]]$units). Some
clinical contexts report BMD as the dimensionless T-score; the canonical
column stores the raw areal density.BMD0 – in-text mathematical symbol used by Mori 2018
for the baseline BMD value entering the BMD ODE; the covariate column
name is BMD_BL.Mori_2018_zoledronicAcid.R (g/cm^2, used as the initial
condition bmd(0) <- BMD_BL and the deviation reference
inside the BMD ODE; per-subject median 0.677 g/cm^2 across the Mori 2018
cohort).BMD_FN_BL,
BMD_TH_BL) rather than overloading this column. Ratified
canonically alongside the Mori 2018 zoledronic-acid BMD extraction.
Companion bone biomarker to TRACP5B_BL.covariateData[[CAV]]$units).Emax * CAV / (EC50 + CAV)) or power (e.g.,
(CAV / CavMedian)^exponent) drug-effect terms. Set to 0 for
placebo periods.CAV, Cav, CAVG,
Cav,W (Svensson 2017 weekly-average bedaquiline plasma
concentration; same orientation as the canonical, in mg/L).METRIC_TASPO_C – Li 2015 (per-arm taspoglutide average
plasma concentration over weeks 2-4, in pmol/L; the model’s
source_name is “Cavg.2-4w (Li 2015 Section 3.2)”). This
descriptive column name maps onto CAV via this alias rather
than being a separate canonical, consistent with the MBMA usage already
documented in this entry’s Notes.FiedlerKelly_2020_fremanezumab_em.R,
FiedlerKelly_2020_fremanezumab_cm.R,
Schoemaker_2018_levetiracetam.R (DDMODEL00000239; LEV
plasma concentration in mg/L), Svensson_2017_bedaquiline.R
(weekly-average bedaquiline concentration in mg/L driving an Emax effect
on the mycobacterial-load half-life; EC50 = 1.42 mg/L, Emax fixed at
-100%; placebo subjects use CAV = 0),
Li_2015_taspoglutide_mbma.R (MBMA study-arm-level Cavg of
taspoglutide between weeks 2 and 4 of QW dosing, in pmol/L; 0 / 59.85 /
119.7 pmol/L for placebo / 10 mg / 20 mg arms; drives an additive Emax
response on body-weight change),
Lacy_2018_cabozantinib_tumor.R (individual predicted
daily-average cabozantinib plasma concentration in ng/mL from the
upstream Lacy 2018 popPK; drives a saturable Cavg/(EC50 + Cavg) Hill-1
term on a time-attenuating decay rate of tumor SOD; EC50 = 251 ng/mL;
per-cohort steady-state Cavg 375 / 750 / 1125 ng/mL for 20 / 40 / 60
mg/day starting doses; dose-hold periods set CAV = 0),
Lacy_2018_cabozantinib_dose_modification.R (same
upstream-popPK-derived Cavg in ng/mL; drives a log-linear effect with
coefficient theta_drug = 0.000807 per ng/mL on the active-dose log
hazard for repeated dose modifications; dose-hold periods set CAV = 0
and switch to the dose-hold baseline log hazard).covariateData[[CAV]]$notes should state how the Cav values
are derived (e.g., empirical-Bayes from a referenced population PK
model) and that the column is set to 0 for placebo periods. The
averaging window is also model-specific (per-dosing-interval Cav =
AUC_tau / tau in Schoemaker 2018 / Fiedler-Kelly 2020, but
weekly-rolling-mean Cav_W in Svensson 2017 – where the bedaquiline
once-daily loading + thrice-weekly maintenance schedule makes “per
dosing interval” ambiguous; weeks 2-4 Cavg in Li 2015 carried forward
for the entire 8-52 week follow-up); document the averaging convention
in each model’s covariateData[[CAV]]$notes. MBMA usage (Li
2015) treats CAV as a study-arm-level (not individual-level) exposure
metric – the meaning is the same (period-averaged plasma concentration
of the modelled drug) so a separate canonical is not warranted.AUCss [nM*h] = DOSE_EMPA_MGD * 1e6 / 450.91 / cl that
drives the Emax stimulation of FPG elimination in the indirect-response
PD model.DOSE – used in Baron_2016_empagliflozin.R
(Baron 2016 Methods; studied doses 1, 5, 10, 25, 50 mg QD per Table S2
with 10 and 25 mg dominating; 1129 / 40.9% and 1269 / 46.0% of patients
respectively).Baron_2016_empagliflozin.R (drives AUCss feeding Gmax *
AUCss / (AUC50 + AUCss) – AUC50 = 703 nM*h).DOSE_<DRUG>_<UNITS> auto-approve family (e.g.,
DOSE_PHT_MGKGD for phenytoin). Distinct from
DOSE_PHT_MGKGD (per-kg phenytoin daily-dose) –
empagliflozin is dosed in flat mg/day (no per-kg adjustment in label).
Future once-daily SGLT2-inhibitor extractions should register sibling
canonicals (e.g., DOSE_DAPA_MGD for dapagliflozin,
DOSE_CANA_MGD for canagliflozin) rather than reuse this
name. Ratified canonically on 2026-06-24 alongside the Baron 2016
empagliflozin extraction.F_powder = 1 - exp(-theta / DOSE_PHT_MGKGD). Has no
effect when paired with FORM_POWDER = 0 (tablet); a non-NA
non-zero placeholder must still be supplied.Dij – used in Yukawa_1990_phenytoin.R
(paper’s per-record daily-dose-per-weight regressor, mg/kg/d, in the
powder bioavailability equation 4 of Yukawa 1990).Yukawa_1990_phenytoin.R (powder bioavailability
F_powder = 1 - exp(-9.92 / DOSE_PHT_MGKGD); F approaches 1
below ~2 mg/kg/d and decreases monotonically as the daily dose
increases, reflecting the lower wettability of the Aleviatin brand
phenytoin powder formulation).DOSE_<DRUG>_MGKGD) rather than reuse this name – the
absolute coefficient (theta_BA2 = 9.92 in Yukawa 1990) is not
transferable across drugs. Computed as the total daily dose summed
across the 2-3 daily phenytoin doses (mg/d) divided by the patient’s
body weight at the dose record (kg). Ratified canonically on 2026-05-10
alongside the Yukawa 1990 phenytoin extraction.covariateData[[PRED_DOSE]]$units when a paper uses a
different unit (mg/kg/day) or a different glucocorticoid
(methylprednisolone, dexamethasone, hydrocortisone) – in the latter case
convert to prednisolone-equivalent mg/day before populating the column
and record the conversion factor in
covariateData[[PRED_DOSE]]$notes.(1 - Pred_max * PRED_DOSE / (Pred_50 + PRED_DOSE)) on
bioavailability (Storset 2014, Pred_max = 0.67,
Pred_50 = 35 mg/day, Hill = 1); threshold-form binary
multiplier (1 + e * (PRED_DOSE >= 20)) on intrinsic
clearance (ter Heine 2018, e = 0.31 for the >= 20 mg/day
high-dose contrast). Document the per-model functional form in
covariateData[[PRED_DOSE]]$notes.Prednisolone dose – used in
Storset_2014_tacrolimus.R (mg/day).Prednisolone dose (total daily dose, mg/day) – used in
TerHeine_2018_everolimus.R (mg/day; collapsed to a binary
high-dose indicator at the >= 20 mg/day threshold inside
model()).Storset_2014_tacrolimus.R (Emax-style fractional reduction
in tacrolimus oral bioavailability via prednisolone-driven induction of
intestinal CYP3A / P-glycoprotein; Storset 2014 Methods Equations 4 + 6
with Hill = 1), TerHeine_2018_everolimus.R (threshold-form
binary high-dose indicator at >= 20 mg/day driving a multiplicative
+31% increase in apparent intrinsic clearance for everolimus via
prednisolone-driven CYP3A4 induction; ter Heine 2018 Table 2 ‘Final
model’).PRED_CMAX_FREE
(free prednisolone Cmax co-medication exposure) –
PRED_CMAX_FREE is the modelled-from-data peak free
concentration, whereas PRED_DOSE is the administered
daily-dose level supplied directly from the dosing record. Both can
coexist in a future model that simultaneously tests dose-driven and
exposure-driven effects of prednisolone on a tacrolimus PK parameter.
Distinct from CONMED_STEROID (binary baseline / concomitant
corticosteroid use indicator) and PRICORT (binary prior
corticosteroid use indicator) – PRED_DOSE carries the daily
dose value, not just an on / off flag. Time-varying because tacrolimus /
everolimus PK depend on the conmed_steroid dose at the time of each
observation; the conmed_steroid taper schedule must be supplied as a
per-time-row covariate column. The corresponding methylprednisolone
single-dose induction-bolus indicator (Storset 2014 binary covariate,
not retained in the final model) would warrant a separate canonical
(e.g. MPRED_BOLUS) if a future model retains it.
Threshold-form binary indicators
(e.g. PRED_DOSE >= 20 mg/day in ter Heine 2018) are
derived inside model() from the continuous
PRED_DOSE column rather than as separate registered
canonicals so the underlying continuous dose value remains available for
sensitivity analyses. Scope promoted from specific to general on
2026-05-24 with the ter Heine 2018 everolimus extraction (second model
ratifying the canonical, this time as a CYP3A4-induction covariate on
intrinsic clearance rather than a CYP3A / P-gp-induction covariate on
oral bioavailability).(1 + e * (PRED_CMAX_FREE - ref)). Reference value
observed: 155.5 nmol/L (Bergmann 2014 study median).PredCmax,free / PREDCFR – used in
Bergmann_2014_tacrolimus.R (Bergmann 2014 Table 2 footnote;
162 nmol/L median per Table 1, 155.5 nmol/L centring value per Table 2
equation).Bergmann_2014_tacrolimus.R (linear deviation effect on
tacrolimus apparent central volume V1/F: every 1 nmol/L increase from
155.5 nmol/L decreases V1/F by 0.28%).PRED_AUC0_12_FREE for free AUC 0-12
h, or PRED_CMAX_TOTAL for total Cmax), register parallel
canonicals rather than overload PRED_CMAX_FREE. The
source-paper Cmax is derived from limited-sampling concentrations at 1 /
2 / 4 hours postdose per Bergmann 2014 Methods (validated against
earlier full-profile data from the same cohort). Distinct from
CAV (average concentration of the modelled drug) and
CP_MGL (instantaneous concentration of the modelled drug as
a time-varying PD driver) – PRED_CMAX_FREE is the maximum
concentration of a co-medication, used as a per-subject covariate.
Ratified canonically on 2026-05-08 alongside the Bergmann 2014
extraction.mg*min/mL); document per-model via
covariateData[[AUC_CARBO]]$units.CB (NONMEM $INPUT column in
DDMODEL00000217 / DDMODEL00000218) – used in
Zecchin_2016_tumorovarian.R and
Zecchin_2016_survival.R. The DDMORE bundles ship the
simulated datasets with the column re-labelled AUC0;
downstream consumers should map AUC0 ->
AUC_CARBO.Zecchin_2016_tumorovarian.R (Zecchin 2016 SLD model for
advanced ovarian cancer, DDMODEL00000217),
Zecchin_2016_survival.R (Zecchin 2016 OS model,
DDMODEL00000218; the OS model integrates the same SLD ODE inline, with
the prior IPP-fit subject-level KG/KD0/KD1/IBASE supplied via the
dataset).AUC_CISPLATIN,
AUC_OXALIPLATIN) when needed. The Zecchin 2016 SLD and OS
models use the value directly in the death-rate term
kd0 * AUC_CARBO * tumorSize, with an internal
/1000 numerical scaling carried verbatim from the source
$DES block.exp(coef * (AUC_BAST_FW - 3065.5)).AUC – verbatim NM-TRAN $INPUT column name
in DDMODEL00000243’s executable .mod files and in the bundle’s
Simulated_event_data.csv. Renamed to
AUC_BAST_FW in the canonical register so that future models
using a generic AUC column with different semantics will
not silently collide.NA_NA_tte_gompertz_ev2.R (BAST PTTE 2017 / DDMODEL00000243
Event 2 hazard model; centred at 3065.5 ug*h/L, coefficient 3.09e-4 on
the NONMEM rescaled scale, equivalent to
exp(3.09e-4 * (AUC_BAST_FW - 3065.5)) on the hazard).AUC_<DRUG>) with
explicit drug semantics rather than overload this name. The BAST guiding
document Section 2.2.2 defines this as “AUC of drug treatment given
within the first week (ug*h/L).”mg*h/L or the paper’s mol*day / 10^6 cells
scaling for the parent-plus-active-metabolite composite); document
per-model via covariateData[[AUC_GEM]]$units.G (NONMEM $INPUT column in DDMODEL00000217
/ DDMODEL00000218) – used in Zecchin_2016_tumorovarian.R
and Zecchin_2016_survival.R. The DDMORE bundles ship the
simulated datasets with the column re-labelled AUC1;
downstream consumers should map AUC1 ->
AUC_GEM.Zecchin_2016_tumorovarian.R (Zecchin 2016 SLD model for
advanced ovarian cancer, DDMODEL00000217),
Zecchin_2016_survival.R (Zecchin 2016 OS model,
DDMODEL00000218).kd1 * AUC_GEM * tumorSize, with an internal
/100 numerical scaling carried verbatim from the source
$DES block.mg*h/L (document per-model via
covariateData[[AUC_GCV]]$units if a different exposure unit
is reported).Emax * AUC_GCV / (EC50 + AUC_GCV) vanishes and the viral
load returns to the kin / kout steady-state baseline.AUC_0-12 – the printed variable name in Koloskoff 2025
(Methods Section 2.3, Eq. 1, and Table 1). Q24h dosing intervals are
entered as AUC_0-24 / 2 so all data live in the q12h
framework (Koloskoff 2025 Methods Section 2.1).Koloskoff_2025_ganciclovir.R (Koloskoff 2025 indirect viral
turnover model for CMV viral load in pediatric SOT / HSCT recipients;
AUC_GCV enters the ODE via
kout * (1 + Emax * AUC_GCV / (EC50 + AUC_GCV)) * viralLoad).AUC_CARBO,
AUC_GEM, AUC_BAST_FW, AUC_PAZO)
follow the same AUC_<DRUG> naming pattern; a future
PK/PD model that uses a different exposure metric for ganciclovir (e.g.,
trough concentration, instantaneous concentration) should register a
parallel canonical rather than overload AUC_GCV. Koloskoff
2025 Monte Carlo simulations are reported under AUC_0-24 (Tables 3 and
4) assuming AUC_0-24 = 2 x AUC_0-12 at steady state; nlmixr2 simulations
should set AUC_GCV to the q12h-interval value (i.e., AUC_0-24 / 2).ug*h/mL
(= mg*h/L). Document per-model via
covariateData[[AUC_PAZO]]$units.a = a0 * AUC_PAZO^e_auc_pazo_a and
c = c0 * AUC_PAZO^e_auc_pazo_c. Set to 0 in periods where
pazopanib is not administered; the model rate terms reduce to baseline
drug-off (a, c -> 0 when AUC_PAZO -> 0) under that
convention.AUC – used in
Ouerdani_2015_pazopanib_mouse.R and
Ouerdani_2015_pazopanib.R (Ouerdani 2015 Methods Equations
2-3; preclinical values 220.2, 656.8, 1140.8 ugh/mL for the 10, 30,
100 mg/kg mouse dose groups; clinical mean 771.6 ugh/mL for 800 mg
QD pazopanib in RCC patients, with per-subject values 629.4-802.4
ug*h/mL derived from an Emax fit to mean AUCs at the patient’s dose
history).Ouerdani_2015_pazopanib_mouse.R (preclinical TGI in CAKI-2
xenograft mice; AUC enters as c = c0 * AUC_PAZO^0.332 only
– the cytotoxic exponent e_auc_pazo_a was fixed to 0
because the in-mouse cytotoxic effect did not vary with exposure across
the 10-100 mg/kg range), Ouerdani_2015_pazopanib.R
(clinical TGI in RCC patients; AUC enters as both
a = a0 * AUC_PAZO^0.125 and
c = c0 * AUC_PAZO^0.142).AUC_SORAF for sorafenib,
AUC_SUNI for sunitinib, etc.) when needed. The Ouerdani
2015 paper reports the preclinical AUCs in ug*h/mL from a
separate preclinical PK study (cited as the FDA Pharmacology Review for
pazopanib NDA 022465); the clinical AUCs come from an Emax fit (Equation
derived from Methods) to pooled mean AUCs at varying daily doses (5 mg
to 2000 mg) across five prior pazopanib trials. Ratified canonically on
2026-05-12 alongside the Ouerdani 2015 pazopanib mouse and clinical
extractions.covariateData[[CLI]]$units).CL – used in the Hansson 2013 sunitinib biomarker / TGI
/ fatigue / OS PD-model family (DDMODEL00000197 and siblings, including
DDMODEL00000222 and the paper-derived
Hansson_2013_sunitinib_OS.R) and in
Schindler_2016_sunitinib.R (DDMODEL00000221) as the posthoc
CL column from the paper’s upstream 2-compartment popPK fit.Hansson_2013a_sunitinib.R (DDMODEL00000197; typical-value
reference 32.819 L/h, drawn from the bundle’s simulated dataset for
subject 1 – broadly consistent with Houk et al. 2010 typical sunitinib
CL), Hansson_2013b_sunitinib.R (DDMODEL00000198; tumor
growth inhibition variant; same per-subject CL column as
Hansson 2013a/c), Hansson_2013c_sunitinib.R
(DDMODEL00000222; uses a per-record CL column with
subject-specific values 30-43 L/h in the bundle’s three-subject
simulated dataset), Hansson_2013_sunitinib_OS.R (parametric
overall-survival Weibull TTE from Hansson 2013 e84 paper text; same
per-subject CL covariate fed in alongside DOSE for the
per-cycle exposure summary auc = DOSE / CLI),
Schindler_2016_sunitinib.R (DDMODEL00000221; per-subject
post-hoc CL fed in as the CL column, vignette uses 50 L/h
literature-typical sunitinib CL/F per Houk 2010),
Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85
paper-direct extraction; uses the same upstream Houk 2009 sunitinib
popPK as the source of CL / CLI per Methods),
Hansson_2013_sunitinib_dbp.R (Hansson 2013 e85;
indirect-response dBP sub-model),
Hansson_2013_sunitinib_hfs.R (Hansson 2013 e85; hand-foot
syndrome Markov + PO sub-model paralleling Hansson_2013c fatigue).CL column because
cl is the canonical nlmixr2 PK parameter name (a parameter,
not a data column). Each model’s covariateData[[CLI]]$notes
should cite the upstream popPK source (paper or DDMORE ID) and explain
how to populate the column for new simulations (typically: simulate the
upstream popPK first to obtain individual CL, or set every subject to
the typical-value CL for typical-trajectory simulations). Distinct from
DOSE (current administered dose level) – the two columns
jointly carry a per-cycle drug-exposure summary
(AUC = DOSE / CLI) for PD-only models that consume posthoc
PK from an upstream popPK fit instead of instantiating their own PK
ODE.covariateData[[CL_INDIV]]$units).model() as CL_INDIV in place of an estimated
cl <- exp(lcl + etalcl).CLI – used in
Friberg_2002_paclitaxel.R (NM-TRAN data column for
per-subject paclitaxel CL).Friberg_2002_paclitaxel.R,
Wahlby_2004_paclitaxel_myelosuppression.R (per-subject
paclitaxel CL EBE supplied as a covariate column following the Friberg
2002 convention; the Wahlby 2004 PD model layers BIL and DBIL covariate
effects on top of the Friberg-Karlsson chain).covariateData[[CL_INDIV]]$notes
should state which upstream popPK source the EBE values come from (e.g.,
Henningsson 2001 paclitaxel popPK, fixed in the DDMORE encoding) and
whether placebo periods are present. Companion volumes are registered as
VC_INDIV and VP_INDIV.covariateData[[CMAX_M1]]$units if a different concentration
unit is reported).(CMAX_M1 - CMAX_M1_REF) * theta (Girard 2012) or as a
power scaling depending on the source. Reference value observed: 0 ng/mL
(Girard 2012 sets MED17 = 0 as the centering
reference).CMAXM1 – used in
Girard_2012_pimasertib.R.Girard_2012_pimasertib.R (additive logit shift on the
cumulative-logit AE-score model:
theta_cmaxm1 * CMAXM1).CAV (average
dosing-interval concentration); both are derived exposure metrics fed
into downstream PD / safety models. Document the upstream PK model in
covariateData[[CMAX_M1]]$notes for any future user.CL_INDIV in sequential
PK->PD encodings.covariateData[[VC_INDIV]]$units).model() as VC_INDIV in place of an estimated
vc <- exp(lvc + etalvc).V1I – used in
Friberg_2002_paclitaxel.R (NM-TRAN data column for
per-subject paclitaxel V1).Friberg_2002_paclitaxel.R,
Wahlby_2004_paclitaxel_myelosuppression.R (per-subject
paclitaxel V1 EBE following the Friberg 2002 convention).CL_INDIV notes for the
broader convention.CL_INDIV
and VC_INDIV in sequential PK->PD encodings.covariateData[[VP_INDIV]]$units).model() as VP_INDIV in place of an estimated
vp <- exp(lvp + etalvp).V2I – used in
Friberg_2002_paclitaxel.R (NM-TRAN data column for
per-subject paclitaxel V2).Friberg_2002_paclitaxel.R,
Wahlby_2004_paclitaxel_myelosuppression.R (per-subject
paclitaxel V2 EBE following the Friberg 2002 convention).CL_INDIV notes for the
broader convention. For models requiring a second peripheral
compartment, register VP2_INDIV (and add a follow-on entry
to this register) when a second model legitimately needs it.covariateData[[BAS_SVEGFR3]]$units).model() as the initial condition
svegfr3(0) <- BAS_SVEGFR3 and inside the relative-change
driver bm = (svegfr3 - BAS_SVEGFR3) / BAS_SVEGFR3.BAS3 – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198), Hansson_2013c_sunitinib.R
(DDMODEL00000222), and Hansson_2013_sunitinib_OS.R as the
posthoc sVEGFR-3 baseline column from the paper’s upstream Hansson 2013a
biomarker indirect-response fit (DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198; tumor growth
inhibition with sVEGFR-3 driven shrinkage),
Hansson_2013c_sunitinib.R (DDMODEL00000222; bundle’s
three-subject simulated dataset reports BAS_SVEGFR3 values 42554-57365
pg/mL), Hansson_2013_sunitinib_OS.R (parametric
overall-survival Weibull TTE from Hansson 2013 e84 paper text; sVEGFR-3
dynamics simulated inline using the upstream biomarker-PD per-subject
parameters), Hansson_2013_sunitinib_myelosuppression.R
(Hansson 2013 e85 paper-direct extraction; simulates sVEGFR-3 turnover
in-model to drive the Emax effect on neutrophil proliferation),
Hansson_2013_sunitinib_hfs.R (Hansson 2013 e85; hand-foot
syndrome Markov + PO sub-model driven by sVEGFR-3 relative change
through an effect compartment).covariateData[[BAS_SVEGFR3]]$notes should cite the upstream
biomarker-PD source (paper or DDMORE ID) and explain how to populate the
column for new simulations (typically: simulate from the upstream
biomarker model to obtain individual posthoc baselines, or set every
subject to the typical-value baseline for typical-trajectory
simulations).kout3 = 1 / MRT_SVEGFR3 inside
model().covariateData[[MRT_SVEGFR3]]$units.MRT3 – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198), Hansson_2013c_sunitinib.R
(DDMODEL00000222), and Hansson_2013_sunitinib_OS.R as the
posthoc sVEGFR-3 MRT column from the paper’s upstream Hansson 2013a
biomarker indirect-response fit (DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198),
Hansson_2013c_sunitinib.R (DDMODEL00000222; bundle’s
three-subject simulated dataset reports MRT_SVEGFR3 values 313-408 h,
broadly consistent with the Hansson 2013a typical sVEGFR-3 MRT of 401
h), Hansson_2013_sunitinib_OS.R (parametric
overall-survival Weibull TTE; sVEGFR-3 dynamics simulated inline using
the upstream biomarker-PD per-subject MRT),
Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85
paper-direct extraction), Hansson_2013_sunitinib_hfs.R
(Hansson 2013 e85 paper-direct extraction).BAS_SVEGFR3. The downstream fatigue
model consumes the upstream MRT directly without re-fitting it.auc = DOSE / CLI in mgh/L, EC50_SVEGFR3
carries the same units; document per-model via
covariateData[[EC50_SVEGFR3]]$units).model() in the simple-Imax drug-effect term
eff3 = auc / (EC50_SVEGFR3 + auc).EC53 – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198), Hansson_2013c_sunitinib.R
(DDMODEL00000222), and Hansson_2013_sunitinib_OS.R as the
posthoc sVEGFR-3 EC50 column from the paper’s upstream Hansson 2013a
biomarker indirect-response fit (DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198),
Hansson_2013c_sunitinib.R (DDMODEL00000222; bundle’s
three-subject simulated dataset reports EC50_SVEGFR3 values 1.0-2.8
mgh/L, consistent with the Hansson 2013a typical sVEGFR-3 IC50
typical value of 1.0 mgh/L),
Hansson_2013_sunitinib_OS.R (parametric overall-survival
Weibull TTE; sVEGFR-3 dynamics simulated inline using the upstream
biomarker-PD per-subject EC50),
Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85
paper-direct extraction), Hansson_2013_sunitinib_hfs.R
(Hansson 2013 e85 paper-direct extraction).BAS_SVEGFR3. The downstream fatigue
model consumes the upstream EC50 directly without re-fitting it.covariateData[[BAS_SKIT]]$units).model() as the initial condition for treated and placebo
sKIT compartments and inside the relative-change driver
(skit_pla - skit_drug) / skit_pla (or the analogous
BAS_SKIT-denominated form when only one sKIT compartment is
simulated).SBAS – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198) as the posthoc sKIT baseline column from the paper’s
upstream Hansson 2013a biomarker indirect-response fit
(DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson
2013 e84 paper Table 2 reports a typical sKIT baseline of 39200 pg/mL
with ~50% CV, matching Hansson_2013a_sunitinib’s typical
value).covariateData[[BAS_SKIT]]$notes should
cite the upstream biomarker-PD source (paper or DDMORE ID) and explain
how to populate the column for new simulations (typically: simulate from
the upstream biomarker model to obtain individual posthoc baselines, or
set every subject to the typical-value baseline for typical-trajectory
simulations). Sister covariates: MRT_SKIT,
EC50_SKIT, SLOPE_SKIT (companions for the same
upstream biomarker fit).kout_skit = 1 / MRT_SKIT inside model().covariateData[[MRT_SKIT]]$units.SMRT – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198) as the posthoc sKIT MRT column from the paper’s
upstream Hansson 2013a biomarker indirect-response fit
(DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson
2013 e84 paper Table 2 reports a typical sKIT MRT of 101 days = 2424 h,
matching Hansson_2013a_sunitinib’s typical value of 2430
h).BAS_SKIT. The downstream
tumor-growth-inhibition model consumes the upstream MRT directly without
re-fitting it.auc = DOSE / CLI in mgh/L, EC50_SKIT
carries the same units; document per-model via
covariateData[[EC50_SKIT]]$units).model() in the simple-Imax drug-effect term
eff_skit = auc / (EC50_SKIT + auc).SEC5 – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198) as the posthoc sKIT EC50 column from the paper’s
upstream Hansson 2013a biomarker indirect-response fit
(DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson
2013 e84 paper Table 2 reports a typical (common across all four
biomarkers) IC50 of 1.0 mg*h/L, matching
Hansson_2013a_sunitinib’s shared typical value).BAS_SKIT. The downstream
tumor-growth-inhibition model consumes the upstream EC50 directly
without re-fitting it.dps_skit = BAS_SKIT * (1 + SLOPE_SKIT * t) and
kin_skit = dps_skit * kout_skit inside
model().covariateData[[SLOPE_SKIT]]$units).SLO – used in Hansson_2013b_sunitinib.R
(DDMODEL00000198) as the posthoc sKIT linear disease-progression slope
column from the paper’s upstream Hansson 2013a biomarker
indirect-response fit (DDMODEL00000197).Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson
2013 e84 paper Table 2 reports a typical disease-progression slope of
0.0261/month shared between VEGF and sKIT, which equals approximately
3.5e-5/h, matching Hansson_2013a_sunitinib’s typical
value).BAS_SKIT. Sign convention follows
the upstream biomarker fit – positive slope means the placebo /
natural-history sKIT trajectory drifts upward over time (capturing
disease progression).K_RBV
inside model() to reconstruct the individual ribavirin
concentration time-course analytically as
riba(t) = CSS_RBV * (1 - exp(-K_RBV * t)).covariateData[[CSS_RBV]]$units).model() in the analytical RBV-concentration expression
riba = CSS_RBV * (1 - exp(-K_RBV * t)) that drives the
inhibition term riba / (riba + ec50).css_mode (modal posterior estimate of individual
ribavirin Css from the upstream popPK fit, in ng/mL) – used in
Laouenan_2015_ribavirin.R (DDMODEL00000285). Rename
css_mode -> CSS_RBV before passing the
dataset to rxSolve.Laouenan_2015_ribavirin.R (DDMODEL00000285; the bundle’s
Simulated_Laouenant_2015_CPTPSP_hb_RBV.txt carries
css_mode values 2,400-4,000 ng/mL in the 15-subject
ANRS-CO20-CUPIC cohort).Css*(1-exp(-k*t)) parameterization; another drug
or another popPK parameterization would carry its own canonical. Each
model’s covariateData[[CSS_RBV]]$notes should cite the
upstream popPK source (paper or DDMORE ID) and explain how to populate
the column for new simulations (typically: simulate the upstream popPK
first to obtain individual Css and approach-rate, or set every subject
to the typical Css for typical-trajectory simulations). Companion
column: K_RBV. Distinct from CAV
(dosing-interval-averaged concentration used in Emax / EC50 PD models
with a single per-period exposure number) – CSS_RBV carries
the asymptotic steady-state value paired with the approach-rate
constant, supporting the full time-course reconstruction.CSS_RBV. Used together with CSS_RBV inside
model() to reconstruct the individual ribavirin
concentration time-course analytically as
riba(t) = CSS_RBV * (1 - exp(-K_RBV * t)).covariateData[[K_RBV]]$units).CSS_RBV.k_mode (modal posterior estimate of individual
ribavirin approach-to-Css rate constant from the upstream popPK fit, in
1/day) – used in Laouenan_2015_ribavirin.R
(DDMODEL00000285). Rename k_mode -> K_RBV
before passing the dataset to rxSolve.Laouenan_2015_ribavirin.R (DDMODEL00000285; the bundle’s
Simulated_Laouenant_2015_CPTPSP_hb_RBV.txt carries
k_mode values 0.013-0.47 day^-1 across the 15-subject
cohort, corresponding to approach-to-Css half-lives of 1.5-55
days).CSS_RBV. Distinct from any structural
kel PK parameter – K_RBV is an
apparent approach-to-Css rate from a lumped exponential
parameterization of the trough time-course, not the elimination-rate
constant of a one-compartment IV model (the lumped form absorbs
absorption, distribution, and elimination into a single first-order
rate). Companion column: CSS_RBV.CSS_DFO = 0 (or
a reduced value) over the affected interval – the paper’s
compliance-corrected effective concentration TCss_AV = SCss_AV * (1 -
CMPL) collapses to a CSS_DFO scaling in this implementation.covariateData[[CSS_DFO]]$units.DFO = SLP * CSS_DFO, where SLP has
units of 1/(ug/mL). Set to 0 to disable the chelation
effect (drug holidays, untreated baseline disease-progression
simulations). Reference values observed: ~3.5 ug/mL for 30 mg/kg/day,
~5.5 ug/mL for 45 mg/kg/day, ~7.5 ug/mL for 60 mg/kg/day at 45 kg body
weight on the 5-days-per-week 8-h SC infusion schedule (Bellanti 2015
Fig 3).SCssAV (Bellanti 2015 paper symbol; “simulated
steady-state concentration, average”) – the per-subject population-PRED
value before compliance correction.TCssAV (Bellanti 2015 paper symbol; “true steady-state
concentration, average”) – the post-compliance value
SCssAV * (1 - CMPL); collapse into a single time-varying
CSS_DFO column for nlmixr2lib by precomputing the (1 -
CMPL) reduction in the input data.Bellanti_2015_deferoxamine.R (ug/mL; time-varying input on
the linear DFO = SLP * CSS_DFO ferritin-degradation effect;
27 transfusion-dependent beta-thalassaemia major paediatric / adolescent
patients on 20-60 mg/kg/day DFO 5 days per week).slope * CssAV effect form used in this paper. Sibling
drug-specific Css canonical: CSS_RBV (ribavirin). The
companion K_RBV approach-to-Css rate constant is
not needed here because Bellanti 2015 treats CssAV as a
population-typical steady-state value rather than reconstructing the
rise-to-Css trajectory. For new simulations: a user supplies
CSS_DFO directly (either a constant typical-Css value, or a
time-varying column that switches to 0 during drug holidays); the
vignette walks through computing CssAV analytically from a desired dose
schedule via CssAV = (dose_per_week * F) / (CL_i * 168 h)
with CL_i = 19.3 * (WT/70)^0.75 L/h for allometric scaling
to paediatric / adolescent body weights. Ratified canonically on
2026-05-22 alongside the Bellanti 2015 extraction.covariateData[[CP_MGL]]$units if a different unit is
used.drug = SL * CP_MGL, where SL has units of
1/(mg/L)). Set to 0 for placebo periods or any time outside
the drug-exposure window. Reference values observed: docetaxel typical
Cmax ~3 mg/L after a 100 mg/m^2 1-hour IV infusion (Kloft 2006 /
Netterberg 2017 simulated dataset).CP (Kloft 2006 / Netterberg 2017 NM-TRAN $INPUT
convention for “predicted drug concentration”; values in mg/L) – used in
Netterberg_2017_docetaxel.R.Netterberg_2017_docetaxel.R (linear drug effect on the
proliferation rate of the Friberg myelosuppression chain:
(1 - SL * CP_MGL) with SL = 19.27 (mg/L)^-1
after Kloft 2006’s THETA(3)/808*1000 MW-808 conversion;
CP_MGL supplied per event row from an upstream docetaxel popPK
simulation), Puisset_2007_docetaxel.R (linear drug effect
Slope * CP_MGL on the proliferating compartment of the
Friberg myelosuppression chain; the upstream docetaxel PK in Puisset
2007 was Baille 1997 / Bruno 1996 individual Bayesian posthoc profiles
supplied externally and not encoded in the model file).CAV
(dosing-interval-averaged exposure used in Emax / EC50 PD models) –
CP_MGL is the instantaneous concentration, sampled at every
PD event time. When a future paper requires the same
time-varying-PK-as-PD-input pattern for a different drug, register a
drug-specific canonical (e.g., CP_PACL_MGL for paclitaxel)
rather than overloading this name; CP_MGL retains the
implicit “drug = the modeled drug under the PD analysis” semantics. When
a paper supplies the time-varying PK as separate per-subject
empirical-Bayes PK parameters (e.g., CL_INDIV,
VC_INDIV, VP_INDIV), use those columns in a
coupled PK-PD ODE model (see Friberg_2002_paclitaxel.R)
rather than reducing to CP_MGL. The choice between
PK-as-covariate (this canonical) and PK-as-EBE-parameters depends on
whether the source paper’s NM-TRAN dataset shipped Cp directly or
shipped the upstream individual PK parameters.1 - Rmax * CP_OXY_NGML / (CP_OXY_NGML + p50). Set to 0
outside the drug-exposure window or for non-XOI scenarios. Reference
values observed: mean daily concentration on 300 mg/day allopurinol is
approximately 10,000 ng/mL (Aksenov 2018, Eq. 13).[P]_PIN (oxypurinol) – the symbol used in Aksenov 2018
Eq. 9 for the production-inhibitor concentration when the inhibitor is
oxypurinol.Aksenov_2018_uricAcid.R (Hill-type production inhibition
with rmax_oxy = 0.84 and p50_oxy = 14000 ng/mL
per Aksenov 2018 Table 1).CP_FBX_NGML (febuxostat) and CP_LSN_NGML
(lesinurad). When the source paper supplies an upstream popPK for
oxypurinol (e.g., Wright et al. 2013, Anzai & Endou 2012), the user
simulates that PK to populate this column; otherwise a steady-state
value can be used. Ratified canonically on 2026-05-08 alongside the
Aksenov 2018 extraction.1 - Rmax * CP_FBX_NGML / (CP_FBX_NGML + p50). Set to 0
outside the drug-exposure window or for non-XOI scenarios. Reference
values observed: mean daily concentration on 40 mg/day febuxostat is
approximately 1000-2000 ng/mL (Aksenov 2018, Bhattaram & Gobburu
2017 regulatory review).[P]_PIN (febuxostat) – the symbol used in Aksenov 2018
Eq. 9 for the production-inhibitor concentration when the inhibitor is
febuxostat.Aksenov_2018_uricAcid.R (Hill-type production inhibition
with rmax_fbx = 1 (fixed) and
p50_fbx = 120 ng/mL for hyperuricemic subjects (or 87 ng/mL
for normouricemic subjects) per Aksenov 2018 Table 1).p50 parameter differs between hyperuricemic and
normouricemic populations in Aksenov 2018; Rmax is fixed at
1 per Bhattaram & Gobburu 2017. Distinct from
CP_OXY_NGML (oxypurinol) and CP_LSN_NGML
(lesinurad). Ratified canonically on 2026-05-08 alongside the Aksenov
2018 extraction.FE = FE0 + Fmax * CP_LSN_NGML / (CP_LSN_NGML + p50). Set to
0 outside the drug-exposure window. Reference values observed: peak
plasma concentration after single dose 200 mg lesinurad is approximately
6,000-9,000 ng/mL (Fleischmann et al. 2014; Shen et al. 2015).[P]_RIN (lesinurad) – the symbol used in Aksenov 2018
Eq. 10 for the reabsorption-inhibitor concentration when the inhibitor
is lesinurad.Aksenov_2018_uricAcid.R (Hill-type increase in fractional
excretion with fmax_lsn = 0.56 (fixed) and
p50_lsn = 23000 ng/mL for hyperuricemic subjects (or 11000
ng/mL for normouricemic subjects) per Aksenov 2018 Table 1).p50 parameter differs between hyperuricemic and
normouricemic populations in Aksenov 2018; Fmax was fixed
during estimation. Distinct from CP_OXY_NGML (oxypurinol)
and CP_FBX_NGML (febuxostat). Ratified canonically on
2026-05-08 alongside the Aksenov 2018 extraction.pain = pain_state - e_morph_pain * CP_MORPH_NGML + e_time_pain * time
(Valitalo 2017). Reference values observed: most individual predicted
concentrations in Valitalo 2017 were within 0-60 ng/mL (Figure 2a); the
IRT linear morphine-effect slope is 0.0091 (ng/mL)^-1, so a 20 ng/mL
morphine exposure reduces the latent pain by ~0.18 latent-variable
units.CP (Valitalo 2017 NM-TRAN $INPUT convention for
“morphine plasma concentration”; values in ng/mL) – used in
Valitalo_2017_morphine.R.Valitalo_2017_morphine.R (linear morphine
concentration-effect on the IRT latent pain variable; CP_MORPH_NGML
supplied per event row from an upstream morphine popPK simulation,
typically Knibbe_2009_morphine.R).CP_OXY_NGML /
CP_FBX_NGML / CP_LSN_NGML precedent
established with Aksenov 2018. Distinct from the broader
CP_MGL (mg/L PD-driver convention used in Netterberg 2017
docetaxel myelosuppression and similar) because the IRT PD models in
this family use ng/mL natively. When a future morphine PD analysis uses
mg/L, the conversion is
CP_MORPH_NGML = CP_MORPH_MGL * 1000. Ratified canonically
alongside the Valitalo 2017 morphine extraction (DDMODEL00000247).cl_b_eff = cl_b / (1 + CP_RIF_UM / ki) where ki is the
OATP1B inhibition constant in umol/L. Reference peak observed: a single
600 mg oral rifampicin dose in the Barnett 2018 cohort produces a
typical Cmax of approximately 29 umol/L in the
modellib(‘Barnett_2018_rifampicin’) typical-value simulation.CRIF (Barnett 2018 Eq. 4 and the analogous RSV
inhibition equation; values reported in umol/L).Barnett_2018_coproporphyrin_I.R (drives competitive OATP1B
inhibition of biliary CPI clearance:
cl_b_eff = cl_b / (1 + CP_RIF_UM / ki) with ki = 1.15
umol/L total / 0.13 umol/L unbound),
Barnett_2018_rosuvastatin.R (analogous form with ki = 2.23
umol/L total / 0.25 umol/L unbound),
Yoshida_2018_coproporphyrin_I_rifampin.R (drives
competitive OATP1B inhibition of the hepatic component of CPI clearance
in Yoshida 2018’s one-compartment fNH-parameterised model:
kdeg_eff = kdeg * (fnh + (1 - fnh) / (1 + CP_RIF_UM / kiu))
with kiu = 0.0203 umol/L unbound).CP_<drug>_<units> precedent
(CP_OXY_NGML, CP_FBX_NGML,
CP_LSN_NGML, CP_MORPH_NGML). The natural input
source is a coupled rifampicin popPK simulation; in the Barnett 2018
extraction package, users typically simulate
modellib('Barnett_2018_rifampicin') first (which returns
rifampicin Cc in umol/L after the in-model MW conversion) and feed its
central-compartment output as the CP_RIF_UM column on the CPI or RSV
event table. Distinct from the binary indicator CONMED_RIF
(which captures period-level effects like the V1 / V2 / Q binary
covariate shifts and does not carry magnitude information). Ratified
canonically on 2026-05-26 alongside the Barnett 2018 CPI / RSV
extractions. The Yoshida 2018 rifampin-CPI extraction (2026-05-30) used
the Simcyp v16r1 default single-dose rifampin model output for
portal-vein unbound concentration; that PBPK profile is not reproducible
from on-disk sources, and the paper itself documents a ~5x sensitivity
of the estimated Ki,u to the choice of perpetrator-PK model, so
downstream users must supply their own CP_RIF_UM profile and treat the
resulting CPI excursion as conditional on that choice.kdeg_eff = kdeg * (fnh + (1 - fnh) / (1 + CP_GDC_UM / kiu))
where kiu is the OATP1B unbound inhibition constant in umol/L (Yoshida
2018 estimated kiu = 0.00174 umol/L for GDC-0810).CGDC (Yoshida 2018 NM-TRAN convention for the
per-record GDC-0810 portal-vein unbound concentration; values in
umol/L). The Yoshida 2018 paper itself reports values from the in-house
Y. Chen et al. PBPK model (referenced as personal communication; not on
disk).Yoshida_2018_coproporphyrin_I_GDC0810.R (drives competitive
OATP1B inhibition of the hepatic component of CPI clearance in Yoshida
2018’s one-compartment fNH-parameterised model).CP_<drug>_<units> precedent
(CP_OXY_NGML, CP_FBX_NGML,
CP_LSN_NGML, CP_MORPH_NGML,
CP_RIF_UM). The natural input source in Yoshida 2018 was an
in-house PBPK model for GDC-0810 (Y. Chen et al., personal
communication); that source is not on disk and no GDC-0810 PK model
exists in the nlmixr2lib registry, so downstream users must supply
CP_GDC_UM externally. The paper notes (Discussion) that observed
GDC-0810 plasma AUC IIV was approximately 20%, so the IIV reported on
Ki,u (30.1% CV) partially includes per-subject variability in
portal-vein exposure rather than purely the intrinsic
inhibition-constant variability. Ratified canonically on 2026-05-30
alongside the Yoshida 2018 GDC-0810-CPI extraction.E = E0 + Emax * STIM_QUININE_MM^c / (RIC50^c + STIM_QUININE_MM^c)
(Sheng 2016, Methods, “Model for the drug effect”). Set to 0 for the
water (control) presentation. Reference values observed: Sheng 2016
presented seven concentrations 0 / 0.01 / 0.03 / 0.1 / 0.3 / 1 / 3 mM
(Table 1); the final-model RIC50 is 0.0423 mM.QUININE (Sheng 2016 NM-TRAN $INPUT column name for the
per-record applied stimulus concentration; values in mM).Sheng_2016_quinine_rat.R (sigmoid Emax on the logistic
mixing probability between a low-count generalized-Poisson distribution
and a right-truncated high-count generalized-Poisson distribution;
STIM_QUININE_MM is the only model covariate).CAV (systemic average drug plasma concentration over a
dosing interval) and the CP_* family (instantaneous plasma
concentration as time-varying PD driver) – STIM_QUININE_MM is the
applied stimulus concentration in solution that contacts the taste
receptors directly, with no PK absorption / distribution involved.
Future BATA-style extractions with a different bitter compound
(caffeine, denatonium, etc.) should register a parallel canonical
(e.g. STIM_CAFFEINE_MM) rather than overload this name; the
STIM_<drug>_<units> pattern mirrors the
established CP_<drug>_<units> precedent.
Ratified canonically alongside the Sheng 2016 quinine BATA
extraction.C – used in Simpson_2013_artesunate.R
(per-record applied well concentration in nM).Simpson_2013_artesunate.R (the only model covariate that
varies per record; drives the sigmoid Emax inhibition with the pfmdr1
genotype effects entering on EC50).STIM_<drug>_<units> family (siblings
STIM_CHLOROQUINE_NM, STIM_LUMEFANTRINE_NM,
STIM_MEFLOQUINE_NM, and the bitter-stimulus
STIM_QUININE_MM) – the applied stimulus / well
concentration that contacts the target directly, distinct from
CAV (systemic average plasma concentration), the
CP_<drug> plasma-PD-driver family, and the
CONC_<drug>_MGL in-vitro antibacterial family (which
is reported in mg/L). Ratified canonically alongside the Simpson 2013
antimalarial in-vitro extractions.C – used in Simpson_2013_chloroquine.R
(per-record applied well concentration in nM).Simpson_2013_chloroquine.R (the only model covariate that
varies per record; drives the sigmoid Emax inhibition with the pfmdr1
genotype effects entering on EC50).STIM_<drug>_<units> family (siblings
STIM_ARTESUNATE_NM, STIM_LUMEFANTRINE_NM,
STIM_MEFLOQUINE_NM, and the bitter-stimulus
STIM_QUININE_MM) – the applied stimulus / well
concentration that contacts the target directly, distinct from
CAV, the CP_<drug> family, and the
CONC_<drug>_MGL in-vitro antibacterial family.
Ratified canonically alongside the Simpson 2013 antimalarial in-vitro
extractions.C – used in Simpson_2013_lumefantrine.R
(per-record applied well concentration in nM).Simpson_2013_lumefantrine.R (the only model covariate that
varies per record; drives the sigmoid Emax inhibition with the pfmdr1
genotype effects entering on EC50).STIM_<drug>_<units> family (siblings
STIM_ARTESUNATE_NM, STIM_CHLOROQUINE_NM,
STIM_MEFLOQUINE_NM, and the bitter-stimulus
STIM_QUININE_MM) – the applied stimulus / well
concentration that contacts the target directly, distinct from
CAV, the CP_<drug> family, and the
CONC_<drug>_MGL in-vitro antibacterial family.
Ratified canonically alongside the Simpson 2013 antimalarial in-vitro
extractions.C – used in Simpson_2013_mefloquine.R
(per-record applied well concentration in nM).Simpson_2013_mefloquine.R (the only model covariate that
varies per record; drives the sigmoid Emax inhibition with the pfmdr1
genotype effects entering on EC50).STIM_<drug>_<units> family (siblings
STIM_ARTESUNATE_NM, STIM_CHLOROQUINE_NM,
STIM_LUMEFANTRINE_NM, and the bitter-stimulus
STIM_QUININE_MM) – the applied stimulus / well
concentration that contacts the target directly, distinct from
CAV, the CP_<drug> family, and the
CONC_<drug>_MGL in-vitro antibacterial family.
Ratified canonically alongside the Simpson 2013 antimalarial in-vitro
extractions.P(ROC) = C50^gamma / (C50^gamma + C^gamma) (Shin 2014
Methods). Reference values observed: typical C50 for return of
consciousness 0.37 vol % (mentally intact) and 0.19 vol % (mentally
disabled) in Shin 2014 Table 2; literature MAC-awake for sevoflurane is
0.6-0.78 vol % in healthy adults / children.DOSE – used in Shin_2014_sevoflurane.R
(Shin 2014 Appendix 1 $INPUT column name; the NONMEM column is labelled
DOSE but holds the per-record end-tidal concentration, not an
administered dose – the model uses it directly in
PROB = 1 - DOSE**GAM/(CE50**GAM + DOSE**GAM)).Shin_2014_sevoflurane.R (drives the sigmoid-Emax
probability of return of consciousness in pediatric dental-surgery
patients during emergence from sevoflurane / N2O general
anesthesia).ETISO, ETDES) rather than overload this
name; the ET<agent> pattern parallels the
CP_<drug>_<units> precedent for IV PD drivers.
Distinct from CP_* (systemic plasma concentration),
CAV (steady-state average plasma exposure), and
STIM_* (applied non-systemic stimulus concentration):
ETSEVO is the alveolar / circuit concentration that equilibrates with
brain tissue during inhalation anesthesia, sampled at the exhalation
peak.model() (typically conc <- CEFFECT followed
by a sigmoid Emax of conc); the user is responsible for
generating the CEFFECT trajectory upstream (TCI-controller simulation, a
registered propofol PK model chained with an external Keo effect-site
link, or observed plasma concentrations equilibrated with an effect-site
rate constant).covariateData[[CEFFECT]]$units. Reference
values observed: propofol effect-site Ce in
Koo_2012_propofol.R is reported in ug/mL.covariateData[[CEFFECT]]$notes when
the source paper uses a different column name.P(ROC) = C50^lambda / (C50^lambda + C^lambda). Set to 0 for
the drug-free reference (e.g., before infusion start or after full
washout). Reference values observed: propofol effect-site Ce range
approximately 4.4 +/- 1.1 ug/mL at LOC, 3.2 +/- 1.0 ug/mL at end of
surgery, and 1.1 +/- 0.3 ug/mL at ROC in Koo 2012’s adult population
(Results).Ce (Koo 2012 Methods; the per-record propofol
effect-site concentration predicted by the Schnider TCI controller,
ug/mL). Not stored as a single named NONMEM column in the source $TABLE
– the TCI controller computes it and feeds the NONMEM dataset
directly.Koo_2012_propofol.R
(drives the sigmoid Emax probability of return of consciousness in adult
eye / ENT patients during emergence from propofol-remifentanil TCI
anesthesia; CEFFECT carries propofol effect-site Ce in ug/mL).units field
tells the user which drug and which scale the trajectory must use.
Distinct from CP_* (systemic plasma concentration,
mass-balance from the modeled-drug central compartment),
CAV (dosing-interval-averaged steady-state plasma
exposure), STIM_* (applied in-vitro / sipper-tube stimulus
concentration), ETSEVO / ETISO /
ETDES (alveolar end-tidal volatile-anesthetic concentration
in vol %), and L_OPIOID_pM / L_ANTAGONIST_pM
(Mann 2022 multi-ligand competitive-binding effect-site slots in pM,
where the SLOT identity is the abstraction rather than the drug
identity). The CEFFECT canonical is the operator-ratified auto-approve
name for the “effect-site PD driver” family per the SKILL.md policy
precheck (effect-site PD driver -> CEFFECT). Future PD
extractions that consume an effect-site Ce as a PD driver should reuse
this canonical with the per-model units /
notes recording the drug and the scale; if a future
extraction requires simultaneous occupation of TWO separate effect-site
slots (a competitive / interaction model akin to Mann 2022’s two ligand
slots), register sibling canonicals (e.g., CEFFECT_AGONIST,
CEFFECT_ANTAGONIST) rather than overloading CEFFECT with
both. Ratified canonically on 2026-06-27 alongside the Koo 2012
propofol-ROC extraction (task frompeople-670).GLU (time-varying plasma glucose
regressor input for mechanistic glucose-kinetics models).covariateData[[FPG]]$units.1 + theta * (FPG - ref). Reference
values observed: 8.90 mmol/L (Retlich 2015 popPK/PD linagliptin median
fasting glucose at baseline).Retlich_2015_linagliptin.R (mmol/L, reference 8.90;
linear-deviation effect on baseline DPP-4 activity BSL with coefficient
1.46 % per mmol/L deviation).GLU which is a time-varying within-subject glucose
regressor for mechanistic glucose-kinetics models (Bizzotto 2016).
Ratified canonically alongside the Retlich 2015 linagliptin
extraction.covariateData[[DPP4_BL_RFU]]$notes since RFU values are not
directly comparable across assays).1 + theta * (DPP4_BL_RFU - ref).
Reference values observed: 12,497 RFU (Retlich 2015 popPK linagliptin
median baseline), 11,600 RFU (Retlich 2015 popPK/PD linagliptin median
baseline, applied to the individual-predicted BSL_i parameter on
EC50).Retlich_2015_linagliptin.R (RFU, reference 12,497;
linear-deviation effect on the central-compartment binding-site
concentration Bmax,C with coefficient 0.00332 % per RFU deviation –
captures the inter-individual correlation between baseline DPP-4 protein
concentration and the apparent saturable-binding amplitude).DPP4_BL_PMOL_MIN).
Ratified canonically alongside the Retlich 2015 linagliptin
extraction.GLU directly through a smoothing filter
into a site-of-action glucose variable.linear(GLU) so rxode2
linearly interpolates GLU between dataset rows.iglu (glucose at the current row time) – used in the
DDMORE bundle’s Simulated_glucoseKinetics.csv for
DDMODEL00000227. Rename iglu ->
GLU before passing to rxSolve.Bizzotto_2016_glucose.R (driving regressor for the
glucose-at-site-of-action delay), Lu_2014_sglt_qsp.R
(drives the rate of glucose entry into PCT1 by glomerular filtration in
the SGLT renal-glucose-reabsorption QSP model: filtered glucose load =
GFR * GLU mmol/h), Bosch_2025_glp1ra_hba1c.R (drives the
IGRH HbA1c sub-model: time-varying average daily glucose -> glycation
flux KG*GLU_mgdl and glucose-dependent RBC life span
(GLU_mgdl/149)^gamma_ls; internal conversion mmol/L -> mg/dL via
factor 18.016).GLU is
meaningful only for glucose-kinetics or glucose-PD models that take
plasma glucose as an exogenous regressor. The DDMORE bundle’s
hand-rolled piecewise-linear interpolation
(GL = (t-T1)/(TOBS-T1)*(GLU-GLU1)+GLU1 with bracketing
columns iglu / glun / td / tn) is replaced in nlmixr2 by
linear(GLU) declared in model(); the
bracketing columns are not required.INS directly through a smoothing filter
into a site-of-action insulin variable.linear(INS) so rxode2
linearly interpolates INS between dataset rows.iins (insulin at the current row time) – used in the
DDMORE bundle’s Simulated_glucoseKinetics.csv for
DDMODEL00000227. Rename iins ->
INS before passing to rxSolve.INSU – used in the DDMORE bundle’s
Simulated_ddmoremockdata2.txt for
DDMODEL00000228. Rename INSU ->
INS before passing to rxSolve.Bizzotto_2016_glucose.R (driving regressor for the
insulin-at-site-of-action delay), NA_NA_paracetamol.R
(DDMODEL00000228 OGTT model: drives the insulin-on-glucose-elimination
first-order effect compartment via
kie * (INS / 6.945 - effect_ins)),
Denti_2010_glucoseMinimal.R (Bergman minimal-model insulin
forcing function entering the insulin-action ODE as
p2 * si * (INS - INS_BL); units pmol/L; supplied via
linear(INS)).INS is
meaningful only for glucose-kinetics or insulin-PD models that take
plasma insulin as an exogenous regressor. For drugs that modify
circulating insulin as a downstream effect, use a different
mechanism-specific name. The DDMORE bundle’s hand-rolled
piecewise-linear interpolation
(I = (t-T1)/(TOBS-T1)*(INS-INS1)+INS1 with bracketing
columns iins / insn / td / tn) is replaced in nlmixr2 by
linear(INS) declared in model(); the
bracketing columns are not required.covariateData[[INS_BL]]$units). The example model rescales
via INS_BL / 6.945 to convert pmol/L to uU/mL.BASI (baseline insulin) – used in the DDMORE bundle’s
Simulated_ddmoremockdata2.txt for
DDMODEL00000228. Rename BASI ->
INS_BL before passing to rxSolve.NA_NA_paracetamol.R
(DDMODEL00000228 OGTT model: initialises the insulin-on-elimination
effect compartment effect_ins(0) = INS_BL / 6.945 and feeds
the steady-state baseline-glucose-production rate
gpro = gss * (kg + kgi * INS_BL / 6.945) * vg * 180 / 1000),
Hong_2013_glucose_insulin_HGC.R (anchors the dynamic-state
insulin(0) = ICss * VI initial condition in mU plus the
baseline insulin secretion ICss * CLI, no rescaling),
Hong_2013_glucose_insulin_MTT.R (same anchor as the HGC
companion), Denti_2010_glucoseMinimal.R (Bergman
minimal-model basal-insulin anchor; covariate effect on lsi
(-0.0282 per pmol/L) and lp2 (-0.0150 per pmol/L), and Ib
in the insulin-action ODE p2 * si * (INS - INS_BL); units
pmol/L despite the source paper’s apparent ‘pmol/ml’ table label, which
is documented as an apparent typo).INS (time-varying
regressor); INS_BL is a per-subject baseline-state anchor
used in initial conditions and steady-state derived quantities, not the
dynamic regressor itself. Specific scope because the conversion factor
(1/6.945) and the rescaled-units interpretation are paper-specific;
future extractions that report baseline insulin in mIU/L or pmol/L
directly without rescaling can ratify the same canonical and document
the per-model units / conversion in
covariateData[[INS_BL]]$units / notes.
Companion concept to FPG (baseline fasting plasma
glucose).CINH directly through
linear(CINH) and multiplies by the unbound fraction
fup inside model().linear(CINH) so rxode2
linearly interpolates CINH between dataset rows.Lu_2014_sglt_qsp.R
(drives the rate of unbound inhibitor entering PCT1 by glomerular
filtration: filtered drug load = GFR * fup * CINH nmol/h; set CINH = 0
for baseline / no-inhibitor simulations).CINH is
meaningful only for renal-glucose-reabsorption or other SGLT-mediated
models that take plasma SGLT-inhibitor exposure as an exogenous
regressor. The clinical reporting unit is ng/mL; convert by
1 ng/mL = (1000 / MW_drug) nmol/L (dapagliflozin MW = 409
g/mol so 1 ng/mL = 2.44 nmol/L; canagliflozin MW = 454 g/mol so 1 ng/mL
= 2.20 nmol/L). The Lu 2014 paper feeds the model an interpolated
dapagliflozin observed mean profile (DeFronzo et al. 2013) or a fitted
two-compartment canagliflozin PK profile (Devineni et al. 2013) – the
SGLT-mechanism model itself does not include an internal PK sub-model
for the inhibitor. Companion concept to GLU (the
plasma-glucose regressor that drives the same filtration arm).IU/mL are converted via
1 IU/mL = 2.42 ng/mL (Hayashi 2007 Methods). Document
per-model via covariateData[[IGE]]$units.(IGE / ref)^exponent. Reference value observed: 482.4 ng/mL
(Hayashi 2007 Japanese atopic-asthma cohort).IgE0 (baseline IgE concentration) – used in
Hayashi_2007_omalizumab.R.Hayashi_2007_omalizumab.R (ng/mL, reference 482.4; power
exponents -0.281 on apparent CL of free IgE and +0.657 on apparent IgE
production rate; also used as the initial value for the total-IgE state
at t = 0).X_TE, in
nmol or nmol/L) – IGE is the per-subject baseline column
used for covariate scaling and (when applicable) state initialization,
not the dynamic state itself. For models that use the alternative
reporting unit IU/mL, multiply by 2.42 before applying the
canonical-units (ng/mL) reference value, or document the per-model unit
choice in covariateData[[IGE]]$units so downstream tooling
can interpret the values correctly. Distinct from IGE_FREE
(the time-varying serum free IgE concentration used as
the PD-driver input in PD-only IgE-FEV1 models).IGE (per-subject baseline total IgE used for
power scaling in mechanism-based binding / turnover popPK models like
Hayashi 2007). For sequential PK-PD use, this column is populated from
an upstream omalizumab popPK/IgE-binding model
(e.g. modellib("Hayashi_2007_omalizumab") emits a
freeIgE observable in ng/mL); for standalone fits the
column is interpolated from observed free-IgE assays.freeIgE output convention).
Document per-model via
covariateData[[IGE_FREE]]$units.imax * IGE_FREE^hill / (ec50^hill + IGE_FREE^hill). Set to
the subject’s baseline total IgE at pretreatment / placebo records (per
Zhu 2023 Methods: free IgE equals total IgE before omalizumab treatment)
so the inhibition term evaluates at the subject’s natural free-IgE
level. Reference values observed: 16-960 ng/mL range across placebo +
omalizumab observations in pediatric Study IA05 (Zhu 2023 Results
paragraph).C_IgE,i(t) (Zhu 2023 paper symbol in the
structural-model equation, PDF page 2). The NM-TRAN dataset column name
is not disclosed in the paper text.Zhu_2023_omalizumab_pediatric.R (founding example; ng/mL,
per-event-record free IgE driving the IDR-Type-IV FEV1 percent predicted
response with imax = 0.0717,
ec50 = 39.4 ng/mL, hill = 9 fixed).IGE (baseline
total IgE for power-scaling). The free-IgE-vs-total-IgE distinction is
load-bearing for anti-IgE drugs: free IgE is the pharmacologically
active fraction and the relevant PD driver, while total IgE is the
assay-reported sum of free + drug-bound IgE that does NOT decline under
omalizumab treatment (Zhu 2023 Methods ‘Data Used in the Population
Pediatric IgE-FEV1 Model’). Sequential PK-PD compatibility:
Hayashi_2007_omalizumab.R emits a freeIgE
observable in ng/mL; downstream users simulate that model and populate
the IGE_FREE column in the PD dataset from the simulated
trajectory. For placebo subjects (no anti-IgE drug), the paper used the
subject’s average total IgE level across the steroid-stable period as
the IGE_FREE input, because free IgE equals total IgE in
the absence of anti-IgE drug (Zhu 2023 Methods).covariateData[[ESAD]]$units when the source paper uses a
different per-time unit (units/month, units/day) or converts darbepoetin
/ methoxy-polyethylene-glycol epoetin beta doses to epoetin
equivalents.Naik_2013_peginesatide.R (Naik 2013 eq 16; paper-reported
population median). The covariate effect is gated by an ESADF indicator
(1 if ESAD > 0, else 0) so that subjects with missing or unrecorded
prior ESA dose (encoded as ESAD = 0) carry no covariate adjustment to
HgbBL, matching Naik 2013’s “no effect of ESAD was incorporated for
subjects whose ESAD dose information was not available.”ESAD – used in Naik_2013_peginesatide.R
(Naik 2013 paper notation; prior epoetin alfa / darbepoetin alfa weekly
dose in units/week for CKD hemodialysis subjects enrolling on
peginesatide).Naik_2013_peginesatide.R (Naik 2013 eq 16; exponential
covariate on the baseline-hemoglobin parameter:
hgbbl = exp(lhgbbl + etalhgbbl + e_esad_lhgbbl * (ESAD - 7996) * ESADF)
with e_esad_lhgbbl = -4.49e-7 1/(units/week); effect is
small in magnitude but retained as the only PD-side statistically
significant covariate per backward-elimination at P < 0.005).(ESAD > 0)) handles the “no prior ESA dose
available” data-quality case used by Naik 2013; future papers that
distinguish prior-ESA-naive from prior-ESA-treated-with-unrecorded-dose
may register a parallel canonical (e.g., ESA_NAIVE).
Ratified canonically on 2026-05-22 alongside the Naik 2013 peginesatide
extraction.DOSE_IND – used in
Renard_2011_indacaterol.R (per-arm indacaterol dose,
ug/day; Renard 2011 Table 1).Renard_2011_indacaterol.R (per-arm once-daily inhaled
indacaterol dose driving the MBMA dose-response; dose range 18.75-600
ug/day across 11 trials, with the six discrete reported doses 18.75,
37.5, 75, 150, 300, and 600 ug).DOSE_PHT_MGKGD (phenytoin); future MBMA / dose-response
models for other drugs should register a sibling
DOSE_<DRUG> canonical rather than overloading this
name. Ratified canonically on 2026-05-27 alongside the Renard 2011
indacaterol extraction.Cc) and from the
CP_<DRUG> plasma-PD-driver family: this is an applied
experimental concentration in the in-vitro matrix.CRIF – used in
Clewe_2018_TB_MTP_GPDI_invitro.R (Clewe 2018 Materials and
methods; static rifampicin concentration).Clewe_2018_TB_MTP_GPDI_invitro.R (static rifampicin
concentration driving the multistate-TB kill effects; tested
concentrations 0.002, 0.008, 0.03, 0.125, 0.5, 8 mg/L per Figure
1).CONC_<DRUG>_MGL family
(siblings CONC_INH_MGL, CONC_EMB_MGL,
CONC_IPM_MGL, CONC_TOB_MGL) – use this family
for exogenous static or time-varying drug concentrations in in-vitro
time-kill / hollow-fiber PD models, as distinct from the
CP_<DRUG> plasma-concentration PD-driver family.
Ratified canonically on 2026-05-27 alongside the Clewe 2018
extraction.Cc and the CP_<DRUG> plasma-PD-driver
family.CINH – used in
Clewe_2018_TB_MTP_GPDI_invitro.R (Clewe 2018 Materials and
methods; static isoniazid concentration).Clewe_2018_TB_MTP_GPDI_invitro.R (static isoniazid
concentration driving the kill effects on the F and S sub-states and the
adaptive-resistance transition kon * CONC_INH_MGL; tested
concentrations 0.01, 0.039, 0.156, 0.625, 2.5, 10, 40 mg/L per Figure
1).CONC_<DRUG>_MGL family
(siblings CONC_RIF_MGL, CONC_EMB_MGL,
CONC_IPM_MGL, CONC_TOB_MGL). Ratified
canonically on 2026-05-27 alongside the Clewe 2018 extraction.Cc and the
CP_<DRUG> plasma-PD-driver family.CEMB – used in
Clewe_2018_TB_MTP_GPDI_invitro.R (Clewe 2018 Materials and
methods; static ethambutol concentration).Clewe_2018_TB_MTP_GPDI_invitro.R (static ethambutol
concentration driving the multistate-TB kill effects; tested
concentrations 0.0078, 0.031, 0.125, 0.5, 2, 8, 32 mg/L per Figure
1).CONC_<DRUG>_MGL family
(siblings CONC_RIF_MGL, CONC_INH_MGL,
CONC_IPM_MGL, CONC_TOB_MGL). Ratified
canonically on 2026-05-27 alongside the Clewe 2018 extraction.Cc) and from the
CP_<drug> plasma-PD-driver family.f(Bai) = alpha * log(CONC_BAI_UM + 1) (the
+1 shift encodes the control with CONC_BAI_UM = 0 -> f = 0 without an
undefined ln(0)); set to 0 for the control well. Reference values
observed: Xiang 2018 tested 0 (control), 10, 20, and 40 uM (Materials
and Methods; Figures 3-4).C_Bai (Xiang 2018 Eq 2) in prose; the model column is the
canonical CONC_BAI_UM.Xiang_2018_baicalein.R
(time-invariant baicalein concentration driving the log-linear
inhibition of LPS-stimulated TNF-alpha production in RAW264.7
macrophages, propagating downstream to IL-6, iNOS, and NO).CONC_<drug>_<units>
family; here the unit is uM (sibling concentration covariates such as
CONC_RIF_MGL are reported in mg/L, so the
<units> suffix is load-bearing). Distinct from the
STIM_<drug>_<units> antimalarial-well family
and the CP_<drug> plasma-PD-driver family. Ratified
canonically alongside the Xiang 2018 baicalein extraction.Cc and the
CP_<DRUG> plasma-PD-driver family.Landersdorfer_2018_imipenem_tobramycin.R
(externally-supplied time-varying unbound imipenem concentration driving
the Hill kill function; the HFIM used continuous infusion targeting the
5th-percentile 7.6, median 13.4, and 95th-percentile 23.3 mg/L unbound
concentrations from imipenem 4 g/day continuous infusion in critically
ill patients).CONC_<DRUG>_MGL family
(siblings CONC_RIF_MGL, CONC_INH_MGL,
CONC_EMB_MGL, CONC_TOB_MGL). Renamed from the
model’s earlier bare Cipm column on 2026-05-27 for
consistency with the CONC_<DRUG>_MGL family. Ratified
canonically on 2026-05-27 alongside the Landersdorfer 2018
extraction.Cc and the
CP_<DRUG> plasma-PD-driver family.tob_cut = 1.15 mg/L threshold; set to 0
for imipenem-monotherapy or control arms.Landersdorfer_2018_imipenem_tobramycin.R
(externally-supplied time-varying unbound tobramycin concentration
driving the Emax kill function and the 70-fold imipenem-KC50 synergy
reduction against population 3 when CONC_TOB_MGL >= 1.15 mg/L; HFIM
simulated the two-compartment unbound profile of 7 mg/kg q24h 0.5-h
infusions).CONC_<DRUG>_MGL family
(siblings CONC_RIF_MGL, CONC_INH_MGL,
CONC_EMB_MGL, CONC_IPM_MGL). Renamed from the
model’s earlier bare Ctob column on 2026-05-27 for
consistency with the CONC_<DRUG>_MGL family. Ratified
canonically on 2026-05-27 alongside the Landersdorfer 2018
extraction.tf1 = f1min + (f1max - f1min) * exp(-log(2)/d50 * DOSE_RIV_MGKG)
of Willmann 2021 supplement S2 NONMEM $PK block, with
f1max = 1.25, f1min = 0.59,
d50 = 14.4/82.48 = 0.1746 mg/kg, anchored to F1 = 1.0 at
DOSE_RIV_MGKG = 10/82.48 = 0.1213 mg/kg (the median adult
dose per the integrated adult popPK analysis in Willmann 2021 reference
19). Reference values observed: 0.1213 mg/kg (adult-anchor F1 = 1.0);
0.30 mg/kg (F1 = 0.791); 0.50 mg/kg (F1 = 0.681) – Willmann 2021 Results
p. 1199. Bodyweight-normalised single rivaroxaban doses ranged 0.1-0.5
mg/kg across the EINSTEIN-Jr pediatric program (Willmann 2021 Results
p. 1199 and Figure 1).DW (= DOSE/WGHT in supplement S2 $PK
block) – used in Willmann_2021_rivaroxaban.R (supplement S2
NONMEM $PK block
DW = DOSE/WGHT ; TF1 = F1MIN + (F1MAX - F1MIN) * EXP(-LOG(2)/D50 * DW)).Willmann_2021_rivaroxaban.R (input to the exponential-decay
relative-bioavailability function for the pediatric popPK model of
rivaroxaban;
tf1 = 0.59 + 0.66 * exp(-log(2)/0.1746 * DOSE_RIV_MGKG)
with f1 = tf1 * exp(etalfdepot) and
f(depot) <- f1).f1max, f1min, and d50
are intrinsically tied to the dose-dependency function carried over from
the integrated adult popPK analysis in Willmann 2021 reference 19 and
re-anchored to F1 = 1.0 at 10 mg/82.48 kg for the pediatric
extrapolation. Carrying rivaroxaban dose per kg as an explicit covariate
column (rather than back-computing from the rxode2 event-table amt and a
body-weight column) lets the bioavailability formula be evaluated on
every record including non-dose observation records. Mirrors the sibling
drug-self-dose canonicals DOSE_PHT_MGKGD (Yukawa 1990 daily
phenytoin dose per kg) and DOSE_RTV_MGKG (Zhang 2012
per-administration ritonavir dose per kg); pattern
DOSE_<DRUG>_MGKG (per-administration) vs
DOSE_<DRUG>_MGKGD (daily). Ratified canonically
alongside the Willmann 2021 rivaroxaban pediatric popPK extraction.km_apparent = km.km_apparent = km * (1 + CP_PRB_MGL / ki) where
ki is the probenecid competitive inhibition constant at the
renal tubular transporter in mg/L. Reference peak observed: the
Landersdorfer 2009 8-dose probenecid regimen (total 4.5 g over 70 h)
produces typical plasma probenecid concentrations in the range ~30 to 60
mg/L during the gemifloxacin sampling window (Landersdorfer 2009 Fig.
1C).[P] (Landersdorfer 2009 Methods / Table 1, “plasma
probenecid concentration” symbol entering the competitive-inhibition
formula km_apparent = km * (1 + [P] / kic)).Landersdorfer_2009_gemifloxacin.R (drives competitive
inhibition of the renal-tubular-secretion arm of gemifloxacin clearance:
cl_tubular_secretion = vmax * Cc / (km * (1 + CP_PRB_MGL / ki) + Cc)
with ki = 69.3 mg/L; gemifloxacin filtration clearance (fu * GFR ~ 2
L/h, fixed) and non-renal clearance are unaffected by this term but the
static-treatment-arm effect on non-renal clearance is encoded separately
via CONMED_PROBENECID).CP_<drug>_<units> precedent
(CP_OXY_NGML, CP_FBX_NGML,
CP_LSN_NGML, CP_MORPH_NGML,
CP_RIF_UM). Distinct from the binary indicator
CONMED_PROBENECID (which captures treatment-arm-level
static effects – e.g., on absorption or non-renal clearance – and does
not carry plasma-concentration magnitude information). The Landersdorfer
2009 paper does NOT report a probenecid PK sub-model parameter table on
disk (the Methods describe the structural form – 1-compartment + lag +
parallel first-order + mixed-order elimination – but no parameter
table), so users simulating the with-probenecid arm must supply
CP_PRB_MGL from an external probenecid popPK source (e.g.,
literature digitisation of Landersdorfer 2009 Fig. 1C, or an unrelated
published probenecid popPK model). Ratified canonically on 2026-06-07
alongside the Landersdorfer 2009 gemifloxacin extraction.AGONIST_CODE covariate, supplied as a per-record exogenous
covariate to the static sigmoidal Emax concentration-response model of
perfusion pressure in the isolated, perfused male Wistar rat tail
artery. Polymorphic in agonist identity: a single CONC_AGONIST_M column
carries phenylephrine concentrations on phenylephrine records,
arg-vasopressin concentrations on arg-vasopressin records, mastoparan-7
concentrations on mastoparan-7 records, and Bay K8644 concentrations on
Bay K8644 records, with the per-agonist EC50 / Emax pair selected inside
model() by AGONIST_CODE. The model has no PK
and no time dynamics; CONC_AGONIST_M is the dose-response x-axis.PP = emax * CONC_AGONIST_M^hill / (ec50^hill + CONC_AGONIST_M^hill);
CONC_AGONIST_M = 0 yields PP = 0 (no agonist contraction).CONC_AGONIST_M.Grzesk_2016_m3M3FBS.R
(Grzesk 2016 Table I CRC data: per-agonist EC50 values 7.50e-8 M (PHE),
1.84e-8 M (AVP), 4.48e-8 M (mastoparan-7), 1.96e-6 M (Bay K8644);
m-3M3FBS-shifted EC50 values 6.45e-8, 1.42e-8, 2.55e-8, 2.05e-6 M
respectively).STIM_<drug>_<units> and
CONC_<DRUG>_<UNITS> in-vitro-drug-concentration
families (where the drug identity is fixed in the canonical name and the
unit suffix is mandatory) because Grzesk 2016 deliberately tests four
mechanistically distinct vasoactive agonists in a shared experimental
scaffold with a shared concentration axis – baking the agonist identity
into the canonical name would force four parallel columns rather than
the paper’s one-column-plus-selector encoding. M (mol/L) chosen as the
unit because Grzesk 2016 Table I reports EC50 values in mol/L and the
four agonists span 1e-8 to 1e-6 M (a uniform molar axis is the only
practical choice); the _M suffix preserves the
CONC_<...>_<UNITS> family’s unit-in-name
discipline. Future ex-vivo / in-vitro CRC papers from the Grzesk-group
series (Biomed Rep 2 / 2014; Mol Med Report 5 / 2012; Exp Ther Med 4 /
2012; etc.) that test multiple agonists on the same vascular-reactivity
scaffold should reuse this canonical with an extended
AGONIST_CODE mapping. Ratified canonically alongside the
Grzesk 2016 extraction.APR = Imax * CP_REM_NGML^hill / (IC50^hill + CP_REM_NGML^hill)
(Weber 1993). Set to 0 outside the drug-exposure window (the inhibition
term then collapses to 0). Reference values observed: mean Cmax was 4-6
ng/mL after 200 mg oral, 23-27 ng/mL after 300 mg oral, 47-83 ng/mL
after 600-800 mg oral, with substantial intersubject variability (Weber
1993 Table 1).Cp (Weber 1993 Methods ‘Concentration-effect modelling’
Hill equation; values in ng/mL) – the observed plasma remikiren
concentration measured by HPLC with fluorescence or coulometric
electrochemical detection or by a protein-binding assay (Weber 1993
Methods, ‘Remikiren concentrations’ section).Weber_1993_remikiren.R
(PD-only Hill / Emax model for inhibition of the angiotensin I
production rate in hypertensive patients; CP_REM_NGML is the observed or
simulated plasma concentration of remikiren supplied per event
row).CP_<drug>_<units> precedent
(CP_OXY_NGML, CP_FBX_NGML,
CP_LSN_NGML, CP_MORPH_NGML,
CP_RIF_UM). Weber 1993 fit the PD model directly to
observed plasma concentrations because no structural population PK model
was estimated – the paper characterised remikiren PK with
model-independent NCA only. Users wishing to drive the PD model from a
simulated PK source must supply their own concentration trajectory (the
published single-dose Weber-group papers from the same series report
compartmental PK in healthy volunteers but are not on disk in
nlmixr2lib). Ratified canonically on 2026-06-10 alongside the Weber 1993
remikiren PD extraction.QTcF = e0 + slope * (CP_GLASDEGIB_NGML / 1000) (Fostvedt
2021; the in-model / 1000 converts the canonical ng/mL
covariate to the paper’s reported microgram-per-mL slope scaling). Set
to 0 outside the drug-exposure window (the concentration-slope term then
collapses to 0). Reference values observed: geometric-mean steady-state
Cmax was 1137 ng/mL at 100 mg QD (therapeutic dose) and 2445 ng/mL at
200 mg QD (a 2-fold supratherapeutic exposure used in Fostvedt 2021’s
bootstrap predictions; see Fostvedt 2021 Methods 2.3 and Table 4).CONC (Fostvedt 2021 Methods 2.2 and 2.3
‘concentration-QTc modelling’ equations; values in ng/mL but model fit
with the slope reported per microgram-per-mL after a
CONC / 1000 rescaling internal to the estimation – “The
estimates are reported on the microgram scale as the scaling helped the
estimation procedure”, Fostvedt 2021 Table 3 footnote).Fostvedt_2021_glasdegib_QTcF.R,
Fostvedt_2021_glasdegib_QTcS.R (PD-only linear
mixed-effects E-R model for QTcF / QTcS prolongation in patients with
cancer; CP_GLASDEGIB_NGML is the observed or simulated plasma
concentration supplied per event row).CP_<drug>_<units> precedent
(CP_OXY_NGML, CP_FBX_NGML,
CP_LSN_NGML, CP_MORPH_NGML,
CP_RIF_UM, CP_REM_NGML). Fostvedt 2021 fit the
PD model directly to observed plasma concentrations because no
structural population PK model is published in the source paper – the
geometric-mean Cmax values used for the parametric-bootstrap QTcF
predictions (Fostvedt 2021 Section 3.4; cited to reference [15] of the
paper) come from a separate Pfizer phase 2 study in patients with
hematologic malignancies, not from a parameterised popPK model included
in this paper. Users wishing to drive the PD model from a simulated PK
source must supply their own concentration trajectory; no glasdegib
popPK model exists in the nlmixr2lib registry. Ratified canonically on
2026-06-24 alongside the Fostvedt 2021 glasdegib QTc extraction.These columns are specific to count / Markov / time-to-event PD models that fit per-record event counts (e.g., daily or monthly seizure counts) with optional dependence on the previous-period count. Register names retain their source-paper conventions where those names are unambiguous and readable.
Q2 – used in the Schoemaker 2018 LEV / BRV pediatric
extrapolation (DDMODEL00000239) as the treatment-phase gating multiplier
on the combined placebo + drug-effect log-rate term
(LE = LS0 + Q2 * LTRTE).Schoemaker_2018_levetiracetam.R (DDMODEL00000239).Q2 collides with the canonical PK parameter q2
(inter-compartmental clearance to peripheral2) – q2 could
not be used as a covariate column without confusing source-trace
lookups. Useful for any phase-gated PD model where placebo and drug
effects are constrained to a specific study period; analogous to a
“treatment-on” indicator.<param> * PDV / (ES50 + PDV) (Markov-Hill) or as a
linear coefficient <param> * PDV in additive-mean
Markov models (Velez de Mendizabal 2013 equation 5).PDV – used in the
Schoemaker 2018 LEV/BRV pediatric extrapolation (DDMODEL00000239), in
the Ahn 2010 pregabalin Markov seizure-count model that the Schoemaker
2018 publication cites as the precursor structure, and in the Velez de
Mendizabal 2013 MS CEL-count NB nested MAK2 model.Schoemaker_2018_levetiracetam.R (DDMODEL00000239),
VelezdeMendizabal_2013_multipleSclerosis.R (additive-mean
Markov form, equation 5).Schoemaker_2018_levetiracetam.R where the model gates the
Markov contribution on CHILD = 1 so the sentinel is
multiplied by zero and is harmless; (b) the natural zero value PDV = 0,
used in VelezdeMendizabal_2013_multipleSclerosis.R for the
first per-subject observation, which makes the Markov contribution
exactly zero without any indicator gating. Pick whichever convention
matches the source paper’s encoding for the model at hand. The companion
second-order Markov-state covariate is [[PPDV]] (the observed count two
periods prior); see the PPDV entry for the second-order form.<param> * PPDV in additive-mean
Markov-count models (Velez de Mendizabal 2013 equation 5).PPDV – used in the
Velez de Mendizabal 2013 MS CEL-count NB nested MAK2 model (paper
notation “PPDV” = “previous-previous DV”, paralleling PDV = “previous
DV”).VelezdeMendizabal_2013_multipleSclerosis.R (equation 5
second-order Markov term; theta_PPDV = 0.150, materially smaller than
theta_PDV = 0.447 per the source paper’s decreasing-pattern observation
across Markov orders).VelezdeMendizabal_2013_multipleSclerosis.R. The Velez de
Mendizabal 2013 paper also explored a third-order Markov term [[PPPDV]]
– “previous-previous-previous DV”, coefficient theta_PPPDV – but the
third-order fit improvement was not statistically significant, and the
third-order column is not registered until / unless a future model
retains it.LAMB = exp(LE) * NDAYS).NDAYS – used in the
Schoemaker 2018 LEV/BRV pediatric extrapolation (DDMODEL00000239).Schoemaker_2018_levetiracetam.R (DDMODEL00000239).MOMENT – used in the
Valitalo 2017 IRT morphine PD model (DDMODEL00000247).Valitalo_2017_morphine.R (DDMODEL00000247; selects between
the three baseline pain typical values presuct /
suct / aftsuct and the matching 3x3 correlated
etas).ITEM – used in the
Valitalo 2017 IRT morphine PD model (DDMODEL00000247).Valitalo_2017_morphine.R (DDMODEL00000247; switches the IRT
graded-response discrimination / difficulty parameters per row).ITEM_<study>) when the codings collide. The COMFORT-B
“muscle tension” item is omitted from the Valitalo 2017 coding because
it could not be assessed from video recordings.OBSTYPE – used in the
Valitalo 2017 IRT morphine PD model (DDMODEL00000247).Valitalo_2017_morphine.R (DDMODEL00000247; selects between
diff_vas_video vs diff_vas_bedside,
discr_vas_video vs discr_vas_bedside, and
addSd_vas_video vs addSd_vas_bedside).Canonical pattern: RACE_<GROUP>.
Use one indicator per race/ethnicity group the source models. Reference
category is the implicit 0 = all other groups; document explicitly which
groups are in the reference. When the source uses composite groups
(e.g., “Black or Other”), name them accordingly
(RACE_BLACK_OTHER) and list the components in
notes. The base RACE_<GROUP> indicators
are scope: general; composite groupings are scope: specific because the
grouping is tied to the study’s analysis plan.
BLACK – used in Hu_2026_clesrovimab.R,
Robbie_2012_palivizumab.R.Zhu_2017_lebrikizumab.R (canonical form),
Robbie_2012_palivizumab.R.(1 - RACE_WHITE).RACE (with values
1 = White / 0 = non-White) – used in
Lin_2024_casirivimab.R. Source column name
RACE is generic; the canonical name is intentionally
explicit because some other models use RACE for a different
dichotomy.RACE (Caucasian-vs-non-Caucasian dichotomy as named in
Hu 2014 Table 2) – used in Hu_2014_bapineuzumab.R. Same
canonical column name and 1 = White / 0 = non-White encoding; the
typical-value reference is the Caucasian subgroup, so the model
implements the 15% non-Caucasian effect on
(1 - RACE_WHITE).Lin_2024_casirivimab.R
(multiplicative fractional change on CL relative to non-White
reference), Hu_2014_bapineuzumab.R (multiplicative 15%
increase in CL for non-Caucasian relative to Caucasian reference).RACE_BLACK / RACE_ASIAN / etc.; do
NOT combine RACE_WHITE with the decomposed indicators in
the same model. The model’s typical-value reference category (which
subgroup gets the unmodified lcl / lvc) varies
between papers – Lin 2024 uses non-White as the reference, Hu 2014 uses
Caucasian (White) as the reference; both share the same canonical column
encoding.BLACK_OTH – used in
Clegg_2024_nirsevimab.R.Clegg_2024_nirsevimab.R.RACE_BLACK
because the composite is not interchangeable.ASIAN – used in
Hu_2026_clesrovimab.R,
Robbie_2012_palivizumab.R,
Fau_2020_isatuximab.R. RAAS
(race-Asian-vs-other indicator as named in Bajaj 2017 Table 1) – used in
Bajaj_2017_nivolumab.R. RACEN (race-numeric
indicator as named in Lu 2019 / Shi 2020 NONMEM control stream; ASIAN =
1 if RACEN == 1) – used in Lu_2019_polatuzumab.R.Zhu_2017_lebrikizumab.R (canonical form),
Robbie_2012_palivizumab.R,
Bajaj_2017_nivolumab.R, Fau_2020_isatuximab.R,
Lu_2019_polatuzumab.R (multiplicative factor
e_asian_vc = 0.929 on acMMAE Vc, i.e., 7.1% lower V1 in
Asian patients; verbatim Shi 2020 (PMID 32770353) ethnicity-sensitivity
re-quote of the Lu 2019 popPK Asian-race covariate).RACE_ASIAN_AMIND_MULTI (Clegg 2024 grouping
that includes Multiracial; pooled against a different reference), from
RACE_ASIAN_OTH (within-Asian-population sub-indicator), and
from RACE_BLACK_OTH (different composite).Frey_2013_tocilizumab.R (multiplicative fractional effect
on the DAS28 first-order loss rate Kout:
Kout * (1 - 0.25 * RACE_ASIAN_AMIND_OTH) – Kout is 25%
lower in the Asian/AmInd/Other composite group relative to the
White+Black reference).RACE_ASIAN, RACE_OTHER, etc. indicators in the
same model; the composite indicator is mutually exclusive with the
decomposition. Ratified canonically on 2026-04-29 in support of the Frey
2013 tocilizumab DAS28 PKPD model. Frey 2013 uses TWO distinct race
covariates: this RACE_ASIAN_AMIND_OTH indicator on Kout
(DAS28-PD-side; pools Asian + AmInd + Other vs White+Black) AND the
within-Asian RACE_ASIAN_OTH indicator on CL (PK-side;
isolates the “Other Asian” subgroup within the Asian-only cohort).ASIAN_AMIND_MULTI –
used in Clegg_2024_nirsevimab.R.Clegg_2024_nirsevimab.R.covariateData[[RACE_ASIAN_OTH]]$notes for every model that
uses this covariate.Brown_2017_osimertinib.R (paper’s “Asian (not Japanese or
Chinese)” composite indicator with linear effect on apparent clearance
of the AZ5104 metabolite; reference category Caucasian).RACE_ASIAN_AMIND_MULTI (a 4-way composite of Asian +
American Indian + Multiple Races), RACE_ASIAN_AMIND_OTH (a
3-way Asian + AmInd + Other composite against a White+Black reference,
used in Frey 2013), and RACE_BLACK_OTH (different
composite). RACE_ASIAN_OTH is a within-Asian-population
sub-indicator, not a multi-race composite. Operator decision
(2026-04-28): kept separate from RACE_ASIAN because the
paper’s “Other Asian” category is its own grouping, not an alias of
“Asian (any)”. Brown 2017 uses Caucasian (not Chinese) as the dominant
reference.RACE_ASIAN but
specifically restricted to the East Asian subgroup most-relevant to ICH
E5 ethnic-sensitivity / Asian-region bridging analyses.RACE_NEAS – prior canonical name (pre-2026-06-19
readability standardization).RAC4 – used in Zhou_2021_belimumab.R (Zhou
2021 Table 2 footnote d).Zhou_2021_belimumab.R
(multiplicative factor 1.07 on V1).RACE_ASIAN (which can include South / Southeast Asian
populations) because Zhou 2021 specifically tested whether
Chinese/Japanese/Korean patients had different PK from the rest of the
dataset; the analysis explicitly compared RAC4 (North East
Asian) against alternative race definitions and chose RAC4
by AIC. Renamed from RACE_NEAS to
RACE_ASIAN_NORTHEAST on 2026-06-19 per the
canonical-register standardization audit (operator decision: spell out
“Northeast” rather than the opaque NEAS abbreviation;
consistent with RACE_<region> rather than
RACE_<abbreviation>).MULTIRACIAL – used in
Hu_2026_clesrovimab.R.Hu_2026_clesrovimab.R.OTHER – used in
Robbie_2012_palivizumab.R.Zhu_2017_lebrikizumab.R,
Robbie_2012_palivizumab.R.RACE_BLACK on the same model so the two indicators are
mutually exclusive and the implicit White reference category is
reproduced cleanly.Fisher_2008_fosamprenavir.R (multiplicative CL/F factor
1.06 vs the White reference; the composite pools Race = 3 (Asian), 4
(Hispanic), 5 (American Indian), 6 (Other) into a single indicator
alongside the separate RACE_BLACK indicator).RACE_ASIAN, RACE_HISPANIC,
RACE_AMIND, RACE_OTHER indicators in the same
model (do not combine the composite indicator with its constituent
decompositions). Distinct from RACE_ASIAN_AMIND_OTH (Frey
2013 grouping; explicitly excludes Hispanic) and from
RACE_BLACK_OTH (Clegg 2024 grouping; includes Black in the
composite). Ratified canonically alongside the Fisher 2008 fosamprenavir
extraction.HISPANIC – used in
Robbie_2012_palivizumab.R.Robbie_2012_palivizumab.R (fractional effect on CL;
additional effect on Vc).RACE_<GROUP> indicator-decomposition pattern.JAPANESE_HV – used in
Wang_2017_benralizumab.R (Japanese healthy-volunteer cohort
indicator; the healthy-volunteer vs. asthma-patient distinction is
captured separately, not in this covariate).Wade_2015_certolizumab.R (multiplicative fractional effect
on V/F; Wade 2015 breaks Japanese [RACE.EQ.8] out separately from
RACE_ASIAN), Wang_2017_benralizumab.R (multiplicative
factor 1.34 on Vc).RACE_NEAS (North
East Asian composite, includes Chinese, Japanese, and Korean) and from
RACE_ASIAN. Use RACE_JAPANESE only when the
source paper breaks out Japanese heritage as its own indicator; do not
aggregate with other Asian groups when the paper keeps them separate.
Ratified canonically on 2026-04-26.CHINESE,
RACE.EQ.X).Brown_2017_osimertinib.R (linear additive effect
(1 + 0.17 * RACE_CHINESE) on apparent clearance of the
AZ5104 metabolite; reference category Caucasian).RACE_NEAS (North
East Asian composite, includes Chinese, Japanese, and Korean) and from
RACE_ASIAN. Use RACE_CHINESE only when the
source paper breaks out Chinese heritage as its own indicator alongside
RACE_JAPANESE and RACE_ASIAN_OTH; do not
aggregate with other Asian groups when the paper keeps them separate.
Parallels the established RACE_JAPANESE entry. Ratified
canonically on 2026-05-09.RACA = 3.3 – the FIDELIO-DKD popPK analysis dataset
encoded race / ethnicity as a numeric race-and-ethnicity composite
(RACA); RACA = 3.3 flags Korean subjects, with
all other races at non-3.3 values. Used in
vandenBerg_2021_finerenone.R.vandenBerg_2021_finerenone.R (multiplicative effect
1.29^RACE_KOREAN on Vc/F; 2.4% of the FIDELIO-DKD cohort
were Korean, and the paper Discussion notes the effect “may be a
spurious finding based on limited data”),
Schmid_2017_nintedanib.R (multiplicative effect 0.781 on
nintedanib relative bioavailability F1; reference category Caucasian /
Black / other Asian).RACE_ASIAN_NORTHEAST (composite of Chinese + Japanese +
Korean) and from RACE_ASIAN. Use RACE_KOREAN
only when the source paper isolates Korean heritage as its own effect
distinct from other Asian subgroups; do not aggregate with Chinese /
Japanese when the paper keeps them separate. Parallels the established
RACE_JAPANESE and RACE_CHINESE entries.Papuan indicator paired with the
f multiplicative covariate term in Brussee 2016 Table 2 –
used in Brussee_2016_arginine.R (values 1 = Papuan, 0 =
non-Papuan; same orientation as the canonical).Brussee_2016_arginine.R (multiplicative factor
f = 1.9 on L-arginine CL:
CL_i = CL * (WT/60)^k1 * f^RACE_PAPUAN – Papuan patients
have ~1.9x higher arginine clearance than the non-Papuan
reference).RACE_<GROUP> indicators. Papuans are an indigenous
Melanesian / Oceanic population genetically and culturally distinct from
RACE_ASIAN (mainland-Indonesian transmigrants are typically
Austronesian Asian by international population-genetics classification).
The Brussee 2016 cohort was enrolled exclusively at the Mitra Masyarakat
Hospital in Timika, Papua, so a paired REGION_INDONESIA
indicator is not needed for that model – future models that mix Papuan
and non-Papuan Indonesian sites may need such a region indicator
separately. Ratified canonically on 2026-06-09 (general scope; the
Brussee 2016 L-arginine PKPD model is the second example following Yeo
2008 arginine, which also used the same indigenous Papuan vs non-Papuan
stratification).RACE_KOREAN to capture the
Korean subgroup separately).RACE_IND_CHI_TWN = as.integer(ETHNIC %in% c("Indian", "Chinese", "Taiwanese")).Schmid_2017_nintedanib.R (multiplicative effect 1.33 on
nintedanib relative bioavailability F1; reference category Caucasian /
Black / other Asian).RACE_CHINESE is canonical; RACE_INDIAN was
added alongside this entry; RACE_TAIWANESE is not yet
canonical but can be added when a future paper requires it).KOREAN,
RACE.EQ.K).Schmid_2017_nintedanib.R (multiplicative effect 0.781 on
nintedanib relative bioavailability F1; reference category Caucasian /
Black / other Asian), vandenBerg_2021_finerenone.R
(multiplicative effect 1.29^RACE_KOREAN on Vc/F; 2.4% of
the FIDELIO-DKD cohort were Korean, and the paper Discussion notes the
effect “may be a spurious finding based on limited data”).RACE_CHINESE / RACE_JAPANESE entries –
individual East Asian nationality indicators when the source paper
distinguishes them as distinct subgroups rather than pooling under
RACE_ASIAN_NORTHEAST. Korean is the third NE-Asian
individual nationality canonical alongside Chinese and Japanese;
together they exhaust the RACE_ASIAN_NORTHEAST composite
(Chinese + Japanese + Korean) so a model that uses all three individual
indicators can omit RACE_ASIAN_NORTHEAST without
information loss. Distinct from RACE_ASIAN_NORTHEAST
(composite) and from RACE_ASIAN_OTH
(within-Asian-population sub-indicator).INDIAN,
RACE.EQ.I).Schmid_2017_nintedanib.R (multiplicative effect 1.90 on
BIBF 1202 relative bioavailability F2; paired with a non-Indian-Asian
indicator at 1.20; reference category Caucasian / Black).RACE_IND_CHI_TWN and RACE_KOREAN for the
Schmid 2017 nintedanib extraction (request-002 Q1, option B). Parallels
RACE_CHINESE / RACE_JAPANESE /
RACE_KOREAN – individual South / East Asian nationality
indicators when the source paper distinguishes them as distinct
subgroups. DISTINCT from
RACE_ASIAN_AMIND_OTH and
RACE_ASIAN_AMIND_MULTI, both of which use “AmInd” / “AMIND”
to mean American Indian / Alaska Native (a North American indigenous
population), not South Asian Indian. This is a recurring source of
confusion in race-canonical lookups; reviewers should verify the meaning
whenever a paper mentions “Indian” patients in a popPK cohort.Geographical study-site region indicators. Distinct from race /
ethnicity (RACE_*), which describe subject ancestry; these
describe the geographical location of the clinical trial site that
enrolled the subject. Used in multi-regional studies (typically those
including bridging analyses for Japan or East Asia) to capture
region-specific clinical-practice or unmeasured-environment effects on
PK that remain after accounting for body weight, race, and laboratory
covariates. Encoded as a set of mutually exclusive binary indicators
with US as the implicit reference category (all indicators = 0). When a
paper groups some non-US regions with US (e.g., Hong 2025 groups US and
Japan as the DXd CL reference), the model code uses only the indicators
that distinguish the non-reference groups; the data column for the
grouped region (e.g., REGION_JAPAN) is still recorded so
the same dataset can serve other parameters that do separate that
group.
COUNTRY_JPN – retired canonical; used in
Yin_2021_trastuzumabDeruxtecan.R as an enrollment-country
(not study-site-region) indicator. Some papers report
country-of-enrollment rather than site region; both map to
REGION_JAPAN when the binary contrast is Japan
vs. non-Japan.Hong_2025_datopotamab.R (multiplicative effect 1 + (-0.219)
= 0.781 on Dato-DXd linear clearance),
Yin_2021_trastuzumabDeruxtecan.R (multiplicative effect
0.903 on CL_intact and 0.738 on V2_intact when REGION_JAPAN = 1; Yin
2021 retained Japan enrollment-country over Japanese race because the
two were highly confounded, correlation -0.81).RACE_JAPANESE
(subject ancestry). A subject of Japanese ancestry enrolled at a US site
has RACE_JAPANESE = 1 but REGION_JAPAN = 0.
Some papers (e.g., Yin 2021) report enrollment country rather than
study-site region; both are encoded as REGION_JAPAN when
the binary contrast is Japan vs. non-Japan. Paired with
REGION_EUROPE and REGION_ROW in multi-regional
studies (e.g., Hong 2025); REGION_JAPAN = 0 for a US-only
cohort.Hong_2025_datopotamab.R (multiplicative effect 1 + 0.240 =
1.240 on DXd clearance versus US/Japan reference),
Naik_2016_vortioxetine.R (additive intercept-shift form:
TVCL_EU = 39 L/hr is the typical CL/F when
REGION_EUROPE = 1, versus USA reference TVCL = 51
L/hr).REGION_JAPAN and
REGION_ROW to encode multi-regional study membership;
subjects with all three indicators = 0 are in the “US” reference
group.REGION_RoW – mixed-case variant in some source
publications (e.g., Naik 2016).Hong_2025_datopotamab.R (multiplicative effect 1 + 0.196 =
1.196 on DXd clearance versus US/Japan reference),
Naik_2016_vortioxetine.R (additive intercept-shift form:
TVCL_RoW = 38 L/hr is the typical CL/F when REGION_ROW = 1,
versus USA reference TVCL = 51 L/hr; the RoW group for Naik 2016 spans
study sites in Canada, Australia, and Asia).covariateData[[REGION_ROW]]$notes per model.PROT == 4 – protocol-number alias used in Holford 1992
(the France cohort is protocol 970-04, encoded as PROT = 4;
the US cohort is protocol 970-01, encoded as PROT = 1).
Derive REGION_FRANCE = as.integer(PROT == 4).Holford_1992_tacrine.R
(multiplicative scale factors on baseline ADAS-cog
(1 + e_region_france_s0 * REGION_FRANCE) with
e_region_france_s0 = 0.08; on placebo potency
(1 + e_region_france_betap * REGION_FRANCE) with
e_region_france_betap = 0.76; on placebo elimination
half-time (1 + e_region_france_t12elp * REGION_FRANCE) with
e_region_france_t12elp = 1.78 – so the France cohort has 8
percent higher baseline ADAS-cog, 76 percent larger placebo response,
and 2.78x longer placebo wash-out half-time relative to the US
cohort).RACE_FRENCH
(subject ancestry, no canonical at present). Holford 1992 introduces
this as a behavioural-response covariate rather than a PK exposure
covariate; the underlying mechanism is hypothesised cultural /
clinical-trial-conduct differences, not pharmacology. Pair with
REGION_JAPAN, REGION_EUROPE,
REGION_ROW etc. when the same model needs to encode
multiple geographic contrasts. Specific scope until a second model
ratifies the name; at that point promote to general.
Ratified canonically on 2026-05-23 alongside the Holford 1992 tacrine
extraction.deKock_2017_sulfadoxinePyrimethamine.R (multiplicative
-20.2% effect on apparent pyrimethamine clearance; +57.6% scaling on
observed pyrimethamine concentrations; +21.2% scaling on observed
sulfadoxine concentrations; the on-observation scalings capture residual
site-specific differences in apparent bioavailability or
dried-blood-spot sample handling).REGION_SUDAN and
REGION_ZAMBIA for a 4-country sub-Saharan African IPTp
trial design with Mali as the implicit reference. Ratified canonically
on 2026-05-18.deKock_2017_sulfadoxinePyrimethamine.R (+15.5% scaling on
observed sulfadoxine concentrations; +33.2% scaling on observed
pyrimethamine concentrations).REGION_MOZAMBIQUE and REGION_ZAMBIA. Ratified
canonically on 2026-05-18.deKock_2017_sulfadoxinePyrimethamine.R (-24.8% scaling on
observed sulfadoxine concentrations; -5.4% scaling on observed
pyrimethamine concentrations).REGION_MOZAMBIQUE and REGION_SUDAN. Ratified
canonically on 2026-05-18.Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.44
shift on pre-ART CD4 z-score intercept; Ukraine reference).REGION_RUSSIA, REGION_SWITZERLAND,
REGION_UK, REGION_SPAIN,
REGION_GERMANY, REGION_ITALY for the EPPICC
8-country pediatric HIV cohort with Ukraine as the implicit reference.
Distinct from a Polish race/ethnicity indicator. Ratified canonically on
2026-05-22.Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.69
shift on pre-ART CD4 z-score intercept; Ukraine reference).REGION_POLAND, REGION_SWITZERLAND,
REGION_UK, REGION_SPAIN,
REGION_GERMANY, REGION_ITALY for the EPPICC
8-country pediatric HIV cohort with Ukraine as the implicit reference.
Ratified canonically on 2026-05-22.Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.02
shift on pre-ART CD4 z-score intercept; Ukraine reference).Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive -17.5
shift on pre-ART CD4 z-score intercept with Ukraine reference; magnitude
implausibly large for a z-score effect and anchored on a UK cohort of
only 2 subjects – reproduced verbatim per the published table and
flagged in the model file and vignette as a small-sample artifact).Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +2.89
shift on pre-ART CD4 z-score intercept; Ukraine reference).Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.34
shift on pre-ART CD4 z-score intercept; Ukraine reference).Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive -3.63
shift on pre-ART CD4 z-score intercept; Ukraine reference; small-sample
subgroup with n = 2).CLD_PREM – prior canonical name used in
Robbie_2012_palivizumab.R prior to the 2026-06-19 DIS_
prefix standardization.CLD – used in
Robbie_2012_palivizumab.R.BPD – bronchopulmonary-dysplasia shorthand.Robbie_2012_palivizumab.R (fractional +20% effect on
CL).CLD_PREM to
DIS_CLD_PREM on 2026-06-19 per the canonical-register
standardization audit (operator decision to apply the
DIS_<concept> prefix uniformly to disease-state
indicators).covariateData[[DIS_DIAB]]$source_name / notes
fields rather than maintaining a parallel canonical.DIAB – prior canonical name (pre-2026-06-19 DIS_ prefix
standardization) used in Chen_2022_guselkumab.R,
Yao_2018_guselkumab.R,
Overgaard_2019_semaglutide.R.T2DM – prior parallel canonical (merged into
DIS_DIAB on 2026-06-19) used in
NA_NA_paracetamol.R (DDMODEL00000228),
Lu_2014_sglt_qsp.R,
Sharma_2018_naltrexone_bupropion.R.Diabetes – used in Sherer_2012_AAA.R
(Sherer 2012 Methods page 2 symbol “Diabetes”).T2D – used in
Guiastrennec_2016_gastric_emptying.R (matched-cohort flag,
1 = T2D patient vs 0 = matched nondiabetic control).ZDF – used in
Han_2018_methionineMetabolismCycle.R (Zucker Diabetic Fatty
rat as T2DM animal model; ZDF/Gmi fa/fa coded as DIS_DIAB = 1 vs ZDF/Gmi
fa/? non-diabetic littermate control coded as DIS_DIAB = 0).Chen_2022_guselkumab.R (multiplicative effect on CL/F:
1.15^DIS_DIAB, +15% in patients with diabetes).Yao_2018_guselkumab.R (multiplicative effect on CL/F:
1.12^DIS_DIAB, +12% in patients with diabetes).Overgaard_2019_semaglutide.R (multiplicative effects
1.12 on CL and 0.544 on ka per Overgaard 2019 Table 4;
T2D-vs-normoglycaemic stratification).Sherer_2012_AAA.R (additive shift on the first
derivative of AAA growth rate with size beta2:
e_diab_b2 = -0.32/year for diabetics; cohort prevalence
14%).Sharma_2018_naltrexone_bupropion.R (Type-2 diabetes
mellitus stratifier on Emax, kout, kpro, baseline BW; from Sharma 2018
study 6, NB-304).Lu_2014_sglt_qsp.R (multiplicative +17.6% shift on the
typical-value Vmax2 of SGLT2 in the renal-glucose-reabsorption QSP
model: Vmax2_DIS_DIAB = 110 mmol/h vs Vmax2_healthy = 93.5 mmol/h per Lu
2014 Table 2 calibration).NA_NA_paracetamol.R (DDMODEL00000228).Guiastrennec_2016_gastric_emptying.R (multiplicative
-81.1% depression of POTcarbC, the carbohydrate potency on CCK release;
all other parameters are common across cohorts).Han_2018_methionineMetabolismCycle.R (preclinical rat
MMC model; multiplicative-on-log-scale T2DM effects on five rate
constants K_SH (+16 %), K_HM (-92 %), K_HC (-95 %), K_HP (-86 %), K_PH
(-99 %) derived from Han 2018 Table 1 ZDF/control ratios).DIS_UC. Type 1 vs Type 2 mellitus are pooled onto the
single DIS_DIAB column; if a paper distinguishes them,
record “T2DM” or “Type 2” in the per-model
covariateData[[DIS_DIAB]]$source_name notes, and if a
future paper genuinely needs a binary Type-2-vs-Type-1 split (e.g., a
popPK/PD analysis stratifying by HbA1c level), register
DIS_DIAB_TYPE2 at that point. Diabetic patients tend to
have higher inflammation and altered IgG turnover, which can manifest as
modest changes in monoclonal-antibody clearance. In vascular populations
(Sherer 2012) diabetes is associated with slower AAA growth, possibly
via aberrant monocyte-matrix interactions (Golledge 2008 mechanism cited
in Sherer 2012 Discussion). Renamed from DIAB to
DIS_DIAB and merged the prior T2DM canonical
into it on 2026-06-19 per the canonical-register standardization audit
(operator decision: apply the DIS_<concept> prefix
uniformly to disease-state indicators; pool T1/T2 at the column level
and record the Type-2 detail per-model rather than maintaining a
parallel T2DM canonical).HYPERT – prior canonical name used in
Girard_2012_pimasertib.R prior to the 2026-06-19 DIS_
prefix standardization.MHHY (medical history of hypertension) – used in
Girard_2012_pimasertib.R.Girard_2012_pimasertib.R (additive shift on the
cumulative-logit AE-score model: theta_mhhy * DIS_HYPERT;
+0.539 logit units in patients with prior hypertension).DIS_DIAB
(diabetes-mellitus comorbidity); both are baseline binary
medical-history flags collected from clinical-history forms. Captures
any prior or current hypertension diagnosis, regardless of treatment
status; if a future model needs to separate treated vs untreated
hypertension, register a refinement (DIS_HYPERT_TREATED).
Renamed from HYPERT to DIS_HYPERT on
2026-06-19 per the canonical-register standardization audit (operator
decision to apply the DIS_<concept> prefix uniformly
to disease-state indicators).(1 + e_pod_param * (POD - ref_pod)),
sometimes with an upper cap (e.g., values > 180 fixed to 180 days
when the residual time-varying effect plateaus).POD – used in Bergmann_2014_tacrolimus.R
(capped at 180 days; centred at 22.7 days, the dataset mean).Bergmann_2014_tacrolimus.R (linear deviation from POD =
22.7 days on tacrolimus CL/F; coefficient -0.0021 per day implies a
0.21% per-day decrease in apparent oral clearance with a 180-day
plateau).POD
is the conventional NONMEM $INPUT column name. When the
source paper reports a different name (DPT for “days
post-transplant”, TX_DAY, T_POSTOP), record
the alias here. Distinct from TIME (rxode2 time clock) and
from OCC (integer-valued occasion / period indicator for
IOV). When a paper uses POD jointly with an IOV occasion
column, both can coexist in the dataset: POD enters the
typical-value covariate equation, OCC multiplexes the IOV
etas. The 180-day cap in Bergmann 2014 is data-driven (most observations
are within the first 90 days post-transplant, so the linear effect is
identifiable only over that window) – document any per-model cap in
covariateData[[POD]]$notes. Ratified canonically on
2026-05-08 alongside the Bergmann 2014 extraction.CL(TTD) = CL_pop - theta_D * exp(-theta_rate * TTD), so the
decay term vanishes as TTD -> infinity (asymptotic CL far from death)
and reaches its peak drop theta_D as TTD -> 0 (day of
death).TTD – used in Franken_2015_morphine.R
(Franken 2015 NONMEM column for time-to-death in days; paper Eq. 3 and
Table 2).Franken_2015_morphine.R (Franken 2015 Clin Pharmacokinet;
first-order exponential decay term on morphine clearance with theta_D =
17.6 L/h and theta_rate = 0.13 /day; CL drops from 47.5 to 29.9 L/h as
TTD goes from infinity to 0).POD (post-operative day, monotonically increasing from a
surgery date) – TTD counts down to death rather than up from a
procedure. Ratified canonically on 2026-05-16 alongside the Franken 2015
morphine extraction.first day post-transplant – used in
Storset_2014_tacrolimus.R.Storset_2014_tacrolimus.R (multiplicative ~2.68-fold
increase in oral bioavailability on day 1:
fdepot *= 2.68^POSTTX_DAY1; Storset 2014 Table 2 final
theory-based model retains the day-1 factor with subject-level IIV of
57% CV on the day-1 multiplier).POD (post-operative day) canonical above –
POSTTX_DAY1 is a derived binary indicator (operationally
POSTTX_DAY1 = as.integer(POD < 1) when both columns are
present in the dataset). The two coexist in the same model when the
source paper uses POD-based continuous effects on some parameters and a
separate binary day-1 effect on others (Storset 2014 retains the binary
day-1 factor explicitly because the day-1 oral bioavailability is
~2.68-fold higher than the rest of the post-transplant period and is not
well captured by a continuous POD effect). Storset 2014 Discussion
attributes the day-1 oral-bioavailability spike to candidate mechanisms
including methylprednisolone-bolus inhibition of intestinal CYP3A /
P-glycoprotein, surgery-related inflammation, anaesthesia / opioid
effects on gut motility, and reduced food intake – but no single
mechanism was identifiable in the data. The 2.68-fold factor was
retained because it produced a 209-point OFV decrease and was crucial
for predicting concentrations measured on the first post-transplant day.
In Storset 2014 the day-1 effect carries its own subject-level eta (BSV
57% CV on the day-1 factor); only subjects with day-1 observations
contribute to that eta. Ratified canonically on 2026-05-08.BW(t) = (BWPREOP + PFA/1000) * (1 - (1 - fbw) * t^hill_bw / (tbw50^hill_bw + t^hill_bw)),
with the /1000 conversion applying the 1 mL ~ 1 g
fluid-density-1 convention so PFA in mL maps onto a kg increment of body
weight. Future surgical / critical-care popPK papers may use PFA
differently (additive on Vc, multiplicative on CL); the per-paper
parameterisation should be recorded in
covariateData[[PFA]]$notes.PFA – used in Oualha_2018_enoxaparin.R
(Oualha 2018 BJCP; total intra-operative fluid volume in mL summed over
the entire liver-transplant procedure; cohort median 2634 mL, range
1008-6520 mL).Oualha_2018_enoxaparin.R (Oualha 2018 paediatric
liver-transplant cohort; PFA in mL adds to BWPREOP after conversion to
L, defining a transient post-operative body-weight curve BW(t) that
drives the time-varying allometric scaling of V; the Hill / fBW / tBW50
parameters of the BW(t) curve are jointly estimated with the enoxaparin
PK).general. The 1 mL ~ 1 g (density-1) convention is
conventional and adequate for resuscitation fluids (crystalloids and
blood products); a more precise mass-balance accounting could account
for the slightly higher density of packed RBCs (~1.08 g/mL) or 5%
albumin (~1.02 g/mL), but the difference is well within the IIV captured
by lfbw / ltbw50. Distinct from
URINE_FLOW (instantaneous urine flow rate, time-varying
within an observation interval) and from WT (the pre- or
post-operative body weight scalar). Founding example: Oualha 2018
enoxaparin (paediatric liver transplantation, intra-operative fluid
resuscitation typical of major paediatric surgery). When a future model
needs a separate accounting of crystalloid vs colloid vs blood-product
volumes, register sibling canonicals (PFA_CRYST,
PFA_COLLOID, PFA_BLOOD_PROD) rather than
overloading this entry.Hepatic trans_CL – used in
Nanga_2019_tacrolimus_metaanalysis.R (Table 3
covariate-effect label; encoded as a multiplicative power coefficient
theta^TX_LIVER on CL/F).Nanga_2019_tacrolimus_metaanalysis.R (multiplicative effect
on apparent oral clearance: cl_typ *= 0.38^TX_LIVER, so
liver-graft recipients have CL/F 62% lower than non-liver recipients at
the same body weight and post-transplant day; Nanga 2019 Table 3
‘Hepatic trans_CL’ = 0.38), Perrottet_2009_ganciclovir.R
(Perrottet 2009 pools lung and liver under a single CL slope; the model
applies a shared multiplicative effect via
(TX_LUNG + TX_LIVER): pooled lung-or-liver
theta_lung/liver = 1.17 vs
theta_kidney = 1.68, so lung-or-liver recipients have a CL
slope 30% lower than kidney recipients at the same GFR and sex;
Perrottet 2009 Table 3).TX_KIDNEY, TX_HEART, TX_LUNG)
rather than overloading this entry. Distinct from
POSTTX_DAY1 (first-24-hour-post-transplant indicator,
time-varying) and from POD (continuous post-transplant
day); all three can coexist when a model parameterises both organ type
and time-after-transplantation effects. Ratified canonically on
2026-05-18 alongside the Nanga 2019 tacrolimus meta-analysis
extraction.donor / Donor – common NONMEM
$INPUT form (used in
Andrews_2017_tacrolimus.R; Andrews 2017 codes the column as
a binary living-vs-deceased indicator).Andrews_2017_tacrolimus.R (multiplicative effect on
apparent oral clearance: living-donor recipients have CL/F 26% lower
than deceased-donor recipients of the same body weight, CYP3A5 genotype,
eGFR, and hematocrit; Andrews 2017 Table 2 final-model coefficient:
theta_donor_living = 0.74 for the living-donor cohort
relative to the deceased-donor reference, equivalent to deceased-donor
recipients having ~35% higher CL/F than living-donor recipients as
reported in Section 3.4).0 = living donor (not
0 = deceased donor) to match the “1 = the perturbed / index
condition” convention used elsewhere in the register (e.g.,
RRT_HEMODIAL_STATUS, POSTTX_DAY1,
TX_LIVER); the index condition is “received a
deceased-donor graft” because it carries the more variable graft quality
and is the condition that perturbs clearance upward in published
kidney-tacrolimus models. Source papers that encode the column as “1 =
living donor” (e.g., Andrews 2017 itself parameterises the equation with
a multiplier on living-donor recipients) should still record the column
under the canonical orientation: set
DONOR_DECEASED = 1 - source_living_indicator and document
the value inversion in
covariateData[[DONOR_DECEASED]]$notes. Distinct from the
donor-genotype canonical CYP3A5_EXPR_DONOR: donor-source
(deceased vs living) is a logistic / graft-procurement covariate, while
donor-genotype is a pharmacogenetic covariate; both can coexist in the
same dataset when the source paper genotypes both recipients and donors.
Ratified canonically on 2026-05-25 alongside the Andrews 2017 tacrolimus
extraction.GAST – used in
Yamada_2025_zolbetuximab.R.Yamada_2025_zolbetuximab.R (fractional effects on CLss,
CLT, V1).GAST on 2026-04-20
to follow the PRIOR_TNF / PRICORT naming
pattern for prior-treatment and surgical-history indicators. Applicable
to any PK model where gastrointestinal anatomy affects absorption,
first-pass, or protein turnover; not inherently oncology-specific. No
distinction between partial vs total gastrectomy unless the source paper
separates them.graft_type – pooled categorical column with levels K /
H / Lu / Li; decompose into binary
TX_HEART = as.integer(graft_type == "H") and document the
decomposition in covariateData[[TX_HEART]]$notes. Used in
Perrottet_2009_ganciclovir.R.Perrottet_2009_ganciclovir.R (multiplicative effect on the
CL-vs-GFR slope: theta_heart = 0.86 vs
theta_kidney = 1.68, so heart-transplant recipients have a
CL slope 49% lower than kidney recipients at the same GFR and sex;
Perrottet 2009 Table 3).TX_LIVER,
TX_LUNG, and the implicit kidney-reference cohort. Ratified
canonically on 2026-06-08 alongside the Perrottet 2009 ganciclovir
extraction.graft_type – pooled categorical column with levels K /
H / Lu / Li; decompose into binary
TX_LUNG = as.integer(graft_type == "Lu") and document the
decomposition in covariateData[[TX_LUNG]]$notes. Used in
Perrottet_2009_ganciclovir.R.Perrottet_2009_ganciclovir.R (Perrottet 2009 pools lung and
liver under a single CL slope; the model applies a shared multiplicative
effect via (TX_LUNG + TX_LIVER): pooled lung-or-liver
theta_lung/liver = 1.17 vs
theta_kidney = 1.68, so lung-or-liver recipients have a CL
slope 30% lower than kidney recipients at the same GFR and sex;
Perrottet 2009 Table 3).TX_LIVER,
TX_HEART, and the implicit kidney-reference cohort. Future
models that estimate distinct lung-vs-liver effects (rather than
pooling) should keep this canonical for lung and continue to use
TX_LIVER for liver. Ratified canonically on 2026-06-08
alongside the Perrottet 2009 ganciclovir extraction.covariateData[[INTRAOP]]$notes. Used as a
categorical covariate to capture composite physiological changes during
active surgery (hemodynamic effects of general anaesthesia, blood loss,
fluid resuscitation, surgical stress, temperature management) that are
jointly unidentifiable individually in typical perioperative PK
datasets.Stricker_2013_aminocaproicAcid.R (Stricker 2013 Methods
“Full covariate model” and Table 4: the intra-operative period is
defined as “the time immediately after the post-loading dose PK sample
through the end of the surgery”; CL and V1 are estimated separately for
pre/postoperative vs intraoperative periods, with multiplicative shifts
of 0.89 on CL and 0.80 on V1 during the intra-operative window).Stricker_2013_aminocaproicAcid.R (multiplicative effect on
CL and V1: cl = cl_postop * 0.89^INTRAOP,
vc = vc_postop * 0.80^INTRAOP; reference category is
pre/postoperative).POD
(post-operative day – continuous days since surgery, rising
monotonically from 0) and POSTTX_DAY1 (binary
first-24-hours-post-transplant indicator). INTRAOP captures the
active-surgery window itself, whereas POD and POSTTX_DAY1 capture the
post-surgical recovery time course. The three canonicals can coexist in
a single dataset that pools perioperative and post-discharge PK samples.
The intra-operative effect is typically interpreted as a composite of
multiple confounded physiological perturbations and is not directly
attributable to any single mechanism (e.g., Stricker 2013 Discussion:
“the model results may be simply reflecting the net effect of a possible
increased CL due to blood loss and decreased CL due to other confounding
intra-operative factors”). Pairs with a future INTRAOP_BL
(per-subject calculated intra-operative blood loss) canonical when a
source paper reports a continuous blood-loss column alongside or instead
of the binary indicator – that is a separate canonical to be proposed
when needed.UC – used in Hua_2015_anrukinzumab.R.Hua_2015_anrukinzumab.R (multiplicative fractional increase
in CL, +72.8%, on top of weight and albumin effects).DISEXT_EP /
DISEXT_OTHER, which operate within a UC-only
cohort (disease extension). Start as scope: specific; promote to general
if a second paper pools UC with a non-UC reference.sAsthma – used in
Hua_2015_anrukinzumab.R.Hua_2015_anrukinzumab.R (multiplicative fractional change
in SC bioavailability, -30.9%).JIA – used in Gandhi_2021_abatacept.R and
Zhong_2026_abatacept.R.Gandhi_2021_abatacept.R (additive coefficient on logit-F:
pJIA patients have markedly higher SC bioavailability than RA
reference), Zhong_2026_abatacept.R (additive coefficient
+3.08 on logit-F transferred verbatim from a previous internal JIA PPK
model that matches Gandhi 2021’s published value).CHILD and ADOLESCENT, which are pure age-band
indicators independent of indication. Scope: specific; promote to
general if a third paper pools pJIA with a non-pJIA reference and the
reference category remains adult RA.POP
or STUDY categorical alongside
DIS_HEALTHY.Yang_2024_axatilimab.R
(multiplicative effect on baseline NCMC:
BL_NCMC x exp(1.22 x DIS_CANCER + 0.618 x DIS_HEALTHY);
reference category cGVHD when both indicators are 0),
Bonate_2004_apomine.R (log-additive shifts on baseline CL/F
and Vc/F for the advanced-solid-tumor cohort vs the healthy adult-male
reference: e_cancer_cl = log(10.2 / 40.7) = -1.384 and
e_cancer_vc = log(7.11 / 12.3) = -0.548; reference category
0 = healthy adult male volunteer).DIS_HEALTHY
to decompose a three-level “participant population” categorical (cGVHD
reference, advanced solid tumor, healthy volunteer) into two orthogonal
binary indicators (Yang 2024 form), or as a single binary stratifier
between an oncology cohort and a healthy-volunteer reference (Bonate
2004 form). Scope: specific because the disease-pooling reference
category is paper-defined (Yang 2024 reference is patients with cGVHD;
Bonate 2004 reference is healthy adult males). Ratified canonically on
2026-04-28; extended to the Bonate 2004 apomine extraction on
2026-06-04.DIS_CANCER, which is restricted to advanced/metastatic
solid tumors in adults; DIS_CANCER_PED is the pediatric
variant in the DIS_CANCER* family and explicitly covers
leukemia-dominant pediatric cohorts.ONCOLOGY – Llanos-Paez 2020 NONMEM column with the same
orientation (1 = oncology, 0 = nononcology); maps directly to
DIS_CANCER_PED.Llanos-Paez_2020_gentamicin.R (multiplicative cohort shifts
on V1 (-0.154) and Q (-0.321) relative to the nononcology baseline; CL
has no oncology effect), deAlwis_1998_ondansetron.R (cohort
indicator used together with DIS_HEALTHY and
AGE thresholds to switch the per-subject
proportional-residual-error magnitude across five paper-defined
sub-populations; DIS_CANCER_PED = 1 routes to the
paediatric-chemotherapy stratum (Table 1 group 4, propSd 0.178),
DIS_CANCER_PED = 0 paired with DIS_HEALTHY = 0 routes to the
paediatric-general-anaesthesia stratum (Table 1 group 5, propSd
0.145); reference complement is the paediatric general-anaesthesia
cohort (study 4) plus all non-paediatric subjects).DIS_CANCER_PED rather than
DIS_CANCER whenever the source paper’s “oncology” cohort
includes hematologic malignancies (leukemia / lymphoma) or pediatric
blastomas, because DIS_CANCER is canonically restricted to
advanced/metastatic solid tumors. Reference-category complement is
paper-defined (Llanos-Paez 2020 complement is the pooled pediatric
non-oncology admissions cohort; de Alwis 1998 complement is the
paediatric general-anaesthesia sub-cohort plus all non-paediatric
subjects). Scope: specific because the complement is paper-defined.
Covariate-effect parameters drop the DIS_ prefix per the
existing DIS_CANCER -> e_cancer_*
convention (Yang 2024); use
e_cancer_ped_<param>.HV, HEALTHY, DIS_HV).Nikanjam_2019_siltuximab.R (multiplicative effects: 0.77 on
CL, 0.83 on Vss; reference category is the pooled non-healthy oncology
cohort), Okada_2025_rocatinlimab.R (multiplicative shift
1 - 0.532 on Vmax when 1; reference complement is the
pooled atopic-dermatitis + ulcerative-colitis + plaque-psoriasis patient
cohort), Yang_2024_axatilimab.R (multiplicative effect on
baseline NCMC:
BL_NCMC x exp(1.22 x DIS_CANCER + 0.618 x DIS_HEALTHY);
reference category cGVHD), Goel_2016_Sonidegib.R
(multiplicative power-form effect on CL/F:
2.96^DIS_HEALTHY; reference category is the pooled
cancer-patient cohort across X2101 / X1101 / A2201),
Brown_2017_osimertinib.R (linear factor
(1 + 0.44 x DIS_HEALTHY) on apparent osimertinib clearance
and (1 + 1.25 x DIS_HEALTHY) on apparent AZ5104 clearance;
reference category is the pooled NSCLC cohort across AURA / AURA2),
Lu_2015_vismodegib.R (additive-on-log-scale shift on ka via
exp(0.671 * DIS_HEALTHY) and on F via
exp(0.881 * DIS_HEALTHY) gated on the Phase I formulation
indicator; reference category is the pooled cancer-patient cohort across
SHH3925g / SHH4610g / SHH4476g), Gupta_2016_lenvatinib.R
(multiplicative power-form effect on CL/F:
1.15^DIS_HEALTHY; reference category is the pooled
solid-tumor / thyroid-cancer patient cohort across 15 phase 1-3 studies;
healthy subjects show +15 percent CL/F vs cancer patients),
Bienczak_2025_ligelizumab.R (log-additive
exp(-0.087 * DIS_HEALTHY) on apparent ligelizumab CL/F;
reference category is the pooled chronic-spontaneous-urticaria patient
cohort across C2201 / C2202 / C2302 / C2303),
Lu_2015_tacrolimus.R (multiplicative factor
1 / 0.562 = 1.78 on CL/F at DIS_HEALTHY = 1; reference
category is the adult Chinese orthotopic liver-transplant recipient
cohort; healthy-volunteer CL/F is 1.78x patient CL/F at ALT = 0. Paper
Eq. 10 uses the reverse-coded SubPop indicator – the model
file re-expresses it as DIS_HEALTHY = 1 - SubPop),
Li_2017_CC292.R (binary stratifier on the proportional
residual error magnitude: HNP cohort uses propSd = sqrt(0.234) and
patient cohort uses propSd = sqrt(0.659); reference category 0 is the
pooled relapsed/refractory B-cell-malignancy patient cohort from
AVL-292-003; source column HNP),
Yoneyama_2017_emicizumab.R (exponential effects
exp(-0.232 * DIS_HEALTHY) on CL/F and
exp(-0.175 * DIS_HEALTHY) on Vd/F; reference category is
the pooled Japanese male adult/adolescent severe-hemophilia-A patient
cohort across the Japanese MAD phase I and phase I/II extension; paper
Eq. 2 uses the reverse-coded PATIENT indicator and the
model file re-expresses it as DIS_HEALTHY = 1 - PATIENT, shifting the
structural typicals to the patient state),
Kleideiter_2017_cebranopadol.R (multiplicative effect on
bioavailability: f_disease *= 0.837 for healthy volunteers
relative to the LBP/OA chronic-pain reference; paired with
DIS_DPN and DIS_BUNIONECTOMY to form the
four-level disease-status stratification),
Kleideiter_2018_cebranopadol.R (multiplicative power-form
effect on bioavailability F: factor 0.837, i.e. about -16% F for healthy
adults vs the pooled nociceptive-pain (LBP and OA) reference cohort;
sibling indicators DIS_DPN and DIS_BUN carry
the diabetic-polyneuropathy and bunionectomy effects in the same
four-level disease-status encoding; Kleideiter 2018 Table 13),
Taubert_2018_finafloxacin.R (log-additive effects on the
canonical lcl_renal + lcl_nonren decomposition:
exp(+0.985 * DIS_HEALTHY) on CL_renal and
exp(+0.068 * DIS_HEALTHY) on CL_nonren; reference category
is the cUTI patient cohort (Trial III), with healthy effects rederived
from paper Table 3 CL_total = 20.9 L/h * patient effect (1 - 0.29) and
FER1 = 0.40 / FER2 = 0.21; source column PATIENT in the
paper, re-expressed as DIS_HEALTHY = 1 - PATIENT),
Goggin_2004_emfilermin.R (log-additive
exp(+0.4325 * DIS_HEALTHY) on apparent CL/F, where +0.4325
= -log(0.649); the patient reference category is the IVF-ET
premenopausal cohort with recurrent implantation failure (Study 3, n =
39) and the model file shifts lcl to the IVF-ET state lcl = log(57 *
0.649) = log(37.0) so the +0.4325 shift restores the
healthy-postmenopausal-women-typical 57 L/h at DIS_HEALTHY = 1; source
column TYPE in the paper, re-expressed as DIS_HEALTHY = 1 -
TYPE), Klunder_2017_upadacitinib.R (paired healthy/RA
structural means lcl_h / lcl_ra and
lvc_h / lvc_ra gated by
DIS_HEALTHY, with cohort-specific log-normal IIV on CL/F
and Vc/F; reference category 0 is the adult RA cohort. The CL/F means
encode the paper’s verbatim Table 3 contrast
lcl_ra = log(39.7 * 0.76); Vc/F means are identical because
Klunder 2017 reports no disease-state effect on typical Vc/F, only on
its ISV), Bulitta_2010_ceftazidime.R (log-additive effects
on CL and on V1 / V2 / V3 of the 3-compartment ceftazidime IV model:
exp(log(1 / 1.17) * DIS_HEALTHY) on CL and
exp(log(1 / 1.01) * DIS_HEALTHY) shared across V1, V2, and
V3; reference category 0 is the cystic-fibrosis patient cohort, and
Bulitta 2010 Table 3 reports FCYFCL = 1.17 and FCYFVSS = 1.01 with the
healthy-volunteer cohort as the paper’s structural reference; the model
file re-expresses the paper’s CF-vs-HV scale factors onto the canonical
DIS_HEALTHY orientation so the typical-value parameters equal the Table
3 CF column), Desai_2016_isavuconazole.R (multiplicative
fractional effect on peripheral volume V_p:
1 + e_dis_healthy_vp * DIS_HEALTHY with e_dis_healthy_vp =
-0.3765, so healthy V_p baseline is about 38% lower than the patient
reference at BMI 24.80 kg/m^2; reference category is the pooled
SECURE-trial invasive-aspergillosis / other-filamentous-fungi patient
cohort; source column SP (1 = patient, 0 = healthy)
re-expressed as DIS_HEALTHY = 1 - SP; assignment of Table 5 theta_4 =
417 L to patients and theta_11 = 260 L to healthy subjects was inferred
from the Discussion typical-value report V_p ~390 L (patients) / ~292 L
(healthy)), deAlwis_1998_ondansetron.R (cohort indicator
used together with DIS_CANCER_PED and AGE
thresholds to switch the per-subject proportional-residual-error
magnitude across five paper-defined sub-populations: DIS_HEALTHY = 1
routes the subject into one of three volunteer strata selected by AGE
(young < 45 y -> propSd 0.125, elderly 45-75 y -> propSd 0.133,
aged >= 75 y -> propSd 0.169), DIS_HEALTHY = 0 routes the subject
into one of two paediatric strata selected by DIS_CANCER_PED
(chemotherapy -> propSd 0.178, anaesthesia -> propSd 0.145);
reference complement under DIS_HEALTHY = 0 is the pooled
paediatric-patient cohort from studies 3 and 4),
Iida_2008_nicorandil.R (multiplicative log-additive effects
on apparent CL, V1, Q, V2 of the 2-compartment IV nicorandil model:
exp(log(1 / 1.94) * DIS_HEALTHY) on CL,
exp(log(1 / 1.39) * DIS_HEALTHY) on V1,
exp(log(1 / 0.519) * DIS_HEALTHY) on Q,
exp(log(1 / 4.06) * DIS_HEALTHY) on V2; reference category
0 is the acute heart failure (AHF) patient cohort, and Iida 2008 Table 2
reports POP_CL, POP_V1, POP_Q, POP_V2 with multiplicative FCL = 1.94,
FV1 = 1.39, FQ = 0.519, FV2 = 4.06 carrying the AHF-vs-healthy contrast
with the healthy-volunteer cohort as the paper’s structural reference;
the model file re-expresses the paper’s HV-vs-AHF scale factors onto the
canonical DIS_HEALTHY orientation so the typical-value parameters equal
the AHF cohort), Shoji_2011_pregabalin.R (binary stratifier
on the combined proportional + additive residual error magnitudes:
healthy cohort uses propSd = 0.220 and addSd = 0.0239 ug/mL; patient
cohort uses propSd = 0.285 and addSd = 0.236 ug/mL; reference complement
under DIS_HEALTHY = 0 is the pooled post-herpetic neuralgia (PT01-PT04)
and diabetic peripheral neuropathy (PT05) patient cohort; Shoji 2011
Table 3 final model rows for residual variability),
Farrell_2013_conestatAlfa.R (cohort indicator gating the
endogenous-baseline structural mean and IIV between healthy volunteers
and patients with hereditary angioedema: DIS_HEALTHY = 1 selects
lrbase_hv = log(0.901) with etalrbase_hv
(12.7% CV) and DIS_HEALTHY = 0 selects
lrbase_hae = log(0.176) with etalrbase_hae
(54.4% CV); reference complement is the pooled adolescent / adult
HAE-patient cohort from Studies C1 1101-01 / 1202-01 / 1203-01 / 1205-01
/ 1304-01), PerezRuixo_2006_tipifarnib.R (ratio-form
multiplicative effects on five structural parameters: CL ratio 1.21, V2
ratio 0.55, Q4 ratio 8.83, V4 ratio 2.66, and Ka ratio 2.31 applied as
ratio^DIS_HEALTHY; reference category is the pooled
adult-cancer cohort across 14 phase 1, 2, and 3 studies; healthy
subjects show +21 percent CL and 2.31-fold higher Ka than cancer
patients, while V2 is 0.55-fold smaller),
Yin_2020_pexidartinib.R (multiplicative effect on CL/F:
1.26x when DIS_HEALTHY = 1 corresponding to a 21 percent lower
steady-state AUC0-24 in healthy subjects vs patients per Yin 2020
Results; the residual-error proportional SD also switches by
DIS_HEALTHY: 0.297 for patients vs 0.196 for healthy subjects; reference
category 0 is the pooled TGCT / other solid-tumour patient cohort from
Studies PLX108-01 and ENLIVEN).DIS_HV (healthy-volunteer) was
renamed on 2026-05-11 because “volunteer” terminology is discouraged for
clinical-trial participants. Ratified canonically on 2026-04-24.DIS_HEALTHY and DIS_BUNIONECTOMY carrying the
healthy-volunteer and bunionectomy effects respectively).DIS – Kleideiter 2017 (paper Table 13 categorical
disease status decomposed to binary DIS_DPN).Kleideiter_2017_cebranopadol.R (multiplicative effect on
bioavailability: f_disease *= 1.132 for DPN patients
relative to the LBP/OA nociceptive-pain reference; the value 1.132
reflects the 2018 erratum correction in which Table 13 rows 27-28 for
bunionectomy and DPN were swapped),
Kleideiter_2018_cebranopadol.R (DPN-vs-LBP/OA
bioavailability ratio 1.132 applied as ratio^DIS_DPN; one
level of the four-level disease-status stratification {LBP/OA reference,
healthy, DPN, bunionectomy}, Kleideiter 2018 Table 13
erratum-corrected).LBP/OA reference plus other strata) and DPN disease status
is retained as a covariate. Pairs with DIS_HEALTHY and
DIS_BUNIONECTOMY in the Kleideiter four-level
disease-status stratification (LBP/OA reference, healthy, DPN,
bunionectomy). Distinct from the existing DIS_DIAB
canonical (binary type-1-or-type-2 diabetes-mellitus comorbidity
indicator; pre-2026-06-19 names DIAB and T2DM
were merged into DIS_DIAB) because DIS_DPN
flags the specific painful-polyneuropathy complication of diabetes used
as a chronic-pain clinical-trial enrollment category (cebranopadol phase
IIa trials 10 and 12; phase II trial 14). Distinct from the
DIS_HEALTHY canonical because both indicators may be 0
simultaneously in a pooled chronic-pain analysis where DIS_HEALTHY = 0
means ‘patient’ but DIS_DPN = 0 may still mean ‘non-DPN
patient’ (e.g., LBP / OA or bunionectomy patients). Scope: specific
because the disease-pooling reference category is paper-defined.
Ratified canonically on 2026-05-25 alongside the Kleideiter 2017
cebranopadol extraction.DIS_HEALTHY and DIS_DPN
carrying the healthy-volunteer and DPN effects respectively).DIS – Kleideiter 2017 (paper Table 13 categorical
disease status decomposed to binary DIS_BUNIONECTOMY).Kleideiter_2017_cebranopadol.R (multiplicative effect on
bioavailability: f_disease *= 1.801 for bunionectomy
patients relative to the LBP/OA nociceptive-pain reference; the value
1.801 reflects the 2018 erratum correction in which Table 13 rows 27-28
for bunionectomy and DPN were swapped; in the corrected Table 14
bunionectomy patients had 80% higher exposure than the reference, the
largest disease-status effect in the analysis),
Kleideiter_2018_cebranopadol.R (bunionectomy-vs-LBP/OA
bioavailability ratio 1.801 applied as
ratio^DIS_BUNIONECTOMY; one level of the four-level
disease-status stratification {LBP/OA reference, healthy, DPN,
bunionectomy}, Kleideiter 2018 Table 13 erratum-corrected).DIS_HEALTHY and DIS_DPN in the Kleideiter
four-level disease-status stratification. Distinct from POD
(continuous post-operative day) and POSTTX_DAY1
(first-24h-post-transplant indicator) – those describe surgical-recovery
time windows rather than the bunionectomy-cohort enrollment label
itself. The shorter form DIS_BUN is deliberately avoided to
prevent confusion with the common BUN (blood urea nitrogen) laboratory
abbreviation. Scope: specific because the disease-pooling reference
category is paper-defined. Ratified canonically on 2026-05-25 alongside
the Kleideiter 2017 cebranopadol extraction.CD,
CASTLEMAN).Nikanjam_2019_siltuximab.R (multiplicative +24% effect on
CL; no Vss effect).DISTYPN – used in Pu_2021_evinacumab.R (Pu
2021 NM-TRAN $INPUT column for HoFH-vs-HV disease type, 1 = HoFH).Pu_2021_evinacumab.R
(multiplicative exp(theta * DIS_HOFH) factor on Vmax with
theta = -0.289, i.e. HoFH patients show ~25% lower target-mediated Vmax
than the HV reference; biologically consistent with the LDLR-pathway
disruption in HoFH altering ANGPTL3 catabolic kinetics).Diep_2026_donidalorsen.R (linear
(1 + theta * DIS_HAE) multiplicative effects on apparent
central volume Vc/F (theta = +0.426, +42.6%), apparent
intercompartmental clearance Q/F (theta = -0.261, -26.1%), baseline
plasma prekallikrein BL (theta = -0.132, -13.2%), and donidalorsen IC50
on PKK production (theta = +0.770, +77.0%) for patients with HAE vs
healthy volunteers).DIS_HAE_C1INH_T1) can be added without conflicting with
this pooled indicator. Scope: specific because the complement reference
category is paper-defined.Zuo_2016_UDCA.R
(multiplicative scaling on liver-to-biliary rate constants when DIS_PBC
= 1: K_LB,0 scaled by 0.10 – 90% reduction; K_LB,1 scaled by 0.30 – 70%
reduction; K_LB,2 scaled by 0.10 – 90% reduction; reproduces the Zuo
2016 Figure 3 PBC simulation).DIS_HEPATIMP (hepatic-impairment
severity categorical), DBIL (direct bilirubin biomarker),
and ALP (cholestasis biomarker), which describe
pathophysiology rather than the disease label itself.SPOP (Wojciechowski
2022 study-population indicator with the same orientation: 1 = DMD
pediatric patient, 0 = healthy adult volunteer).Wojciechowski_2022_domagrozumab.R (additive
1 + theta shift on baseline myostatin and on the joint
kdeg/kint axis; theta_BASE = -0.641, theta_kdegkint = -0.900),
Hajjar_2018_DMD_6MWT.R (gates the latent-disease
coefficients ALPHA and BETA – both zero for DIS_DMD = 0 – and scales the
6MWT production rate KIN by the multiplier KCOV = 0.63 for DMD subjects;
reference cohort is the Henricson 2012 healthy-boy controls).SMM,
SMOLDMM).Nikanjam_2019_siltuximab.R (multiplicative -23% effect on
Vss; no CL effect).MM – prior canonical name (pre-2026-06-19 DIS_ prefix
standardization) and source-paper column in
Ogasawara_2020_durvalumab.R. The short MM
token also collides with the SI unit millimolar (used in
STIM_*_MM in-vitro stimulus concentrations and
RIC50 in mM) and with the melanoma abbreviation in some
oncology datasets (e.g., TUMTP == "MM" in Aoyama 2012 where
MM means malignant melanoma); the DIS_MM form
disambiguates.Ogasawara_2020_durvalumab.R (multiplicative factor 0.820 on
Vc per Ogasawara 2020 Table 3 footnote c; active (non-smoldering)
multiple myeloma cohort from studies MEDI4736-MM-002 and -MM-005).MM to
DIS_MM on 2026-06-19 per the canonical-register
standardization audit (operator decision to apply the
DIS_<concept> prefix uniformly to disease-state
indicators; the bare MM token clashed with the millimolar
SI unit and with the TUMTP “MM = malignant melanoma” usage). Distinct
from DIS_SMM (smoldering multiple myeloma) and
MM_NIGG (the immunoglobulin-subtype stratifier within
multiple myeloma cohorts).PNH,
DPNH).Lin_2024_pozelimab.R
(additive-fractional +34.07% effect on Vc; no CL or Vp effect; reference
category pools healthy volunteers and CHAPLE patients).MDSAML – prior canonical name (pre-2026-06-19 DIS_
prefix standardization) and combined indicator used directly in source
analyses; MDSAML is the column name in
Ogasawara_2020_durvalumab.R.Ogasawara_2020_durvalumab.R (multiplicative factor 1.26 on
CL; reference group is the union of MM and NHL subjects),
vanIersel_2018_posaconazole.R (multiplicative effect on
relative bioavailability F1:
fdepot *= (1 + (-0.165) * DIS_MDS_AML) – 16.5% lower F1 in
AML/MDS patients relative to the non-AML/MDS reference (healthy
volunteers and HSCT recipients pooled); van Iersel 2018 Table 2
final-model ‘AML/MDS on F1’ = -0.165).DIS_MDS_AML as a combined
MDS+AML indicator when the source paper collapses the two diagnoses into
one covariate. If a future paper separates MDS and AML as distinct
indicators, register DIS_MDS and DIS_AML
separately. Scope: specific because the reference category is
paper-defined. Ratified canonically on 2026-04-26. Renamed from
MDSAML to DIS_MDS_AML on 2026-06-19 per the
canonical-register standardization audit (operator decision to apply the
DIS_<concept> prefix uniformly to disease-state
indicators and to insert the underscore so the combined indicator
matches the sibling DIS_MDS / DIS_AML
shape).DIS_MDS and DIS_CMML indicators that decompose
the hematologic-malignancy cohort).DISEASE_abb == "AML" – used in
Xu_2023_MBG453.R (the Monolix supplement Appendix S2
encodes disease as the categorical column DISEASE_abb with
categories {AML, CMML, MDS, Solid_Tumor} and reference
Solid_Tumor; the canonical column carries the binary
as.integer(DISEASE_abb == "AML")).Xu_2023_MBG453.R
(exponential effect on CL: exp(-0.0146 * DIS_AML); not
statistically significant in the full covariate model but retained
because Xu 2023 used the full-covariate-model approach).DIS_AML (rather than the
combined DIS_MDS_AML) when the source paper separates AML
from MDS as distinct indicators. Scope: specific because the
disease-pooling reference category is paper-defined.DIS_AML and DIS_CMML indicators).DISEASE_abb == "MDS" – used in
Xu_2023_MBG453.R (Monolix supplement Appendix S2
categorical column; reference category Solid_Tumor).Xu_2023_MBG453.R
(exponential effect on CL: exp(-0.149 * DIS_MDS);
statistically significant, p = 0.021 – patients with MDS have ~14% lower
CL than the solid-tumor reference).DIS_MDS (rather than the
combined DIS_MDS_AML) when the source paper separates MDS
from AML as distinct indicators. Scope: specific because the
disease-pooling reference category is paper-defined.DIS_AML and DIS_MDS indicators that decompose
the hematologic-malignancy cohort).DISEASE_abb == "CMML" – used in
Xu_2023_MBG453.R (Monolix supplement Appendix S2
categorical column; reference category Solid_Tumor).Xu_2023_MBG453.R
(exponential effect on CL: exp(-0.0411 * DIS_CMML); not
statistically significant in the full covariate model but retained
because Xu 2023 used the full-covariate-model approach).ALL – used in Wu_2024_inotuzumab.R (Wu
2024 calls it the “ALL effect” and notes it bundles disease type with
the corresponding bioanalytical assay difference).Wu_2024_inotuzumab.R
(additive fractional-change effects on CL1 (-0.767) and CL2 (-0.362),
and gates the BLSTABL and AGE effects on kdes; for kdes itself a -0.924
fractional change for BCP-ALL).SAD,
IMD, DIS).Cheng_2026_immunoglobulin.R (multiplicative
theta^DIS_SAD factors on CL (0.542) and on baseline IgG
(CBAS, 0.541); reference category PID).DIS_SAD specifically partitions hypogammaglobulinaemia by
its underlying mechanism (genetic vs. acquired). Scope: specific because
the SAD cohort composition is paper-defined (in Cheng 2026, 75%
post-rituximab and 25% post-CAR-T cell therapy). Ratified canonically on
2026-04-28.AD,
STATUS, DISGRP).PerezRuixo_2025_posdinemab.R (acts on baseline free
p217+tau in CSF, R0; healthy R0 = 0.793 pmol/L vs AD R0 = 5.995 pmol/L,
a 656% relative increase, no PK-parameter effects).COPD – used in Lahu_2010_roflumilast.R
(paper text covariate symbol in equation 6 and 7).Lahu_2010_roflumilast.R (linear additive effects on
roflumilast parent CL (-39.4%) and V1 (+184%) and on roflumilast N-oxide
CL (-7.9%) and Vd (-21.4%); reference category 0 = pooled phase I
healthy volunteers, 1 = pooled phase II/III moderate-to-severe COPD
patient), Facius_2018_roflumilast.R (linear additive phase
II-III patient effects on KA (-73.3%), parent CL (-55.2%), N-oxide CL
(-24.4%), and N-oxide central V3 (-20.7%) on the joint Lahu 2010 base
model re-estimated on the OPTIMIZE + REACT phase III COPD dataset;
reference category 0 = the implicit phase I healthy-volunteer cohort
that backed the upstream Lahu 2010 fixed structural parameters).OBESE,
MO, COHORT).deHoogd_2017_morphine.R (selects per-cohort proportional
residual error magnitudes for each of three observed species – morphine,
M3G, M6G – after a pooled-cohort fit of 20 morbidly obese surgical
patients and 20 healthy volunteers).BMI (which is
the continuous body-mass-index covariate used for parameter scaling) –
DIS_OBESE_MORBID is the binary cohort-membership flag and
does not encode a specific BMI threshold for general use; the threshold
is paper-defined. Scope: specific because the complement reference
category is paper-defined. Ratified canonically on 2026-05-11.DIS indicator (e.g.,
Okada 2025: DIS=1 for psoriasis, DIS=0 for
healthy, DIS=2 for UC, DIS=3 for AD),
decomposed into a binary DIS_PSORIASIS indicator at
ingestion.Okada_2025_rocatinlimab.R (multiplicative shift
1 - 0.372 on linear CL when 1; reference complement is the
pooled atopic dermatitis + ulcerative colitis + healthy-volunteer
cohort).POPULATION /
STUDY categorical alongside other DIS_*
indicators.Li_2018_PF04236921.R
(log-shift e_ra_cl = log(0.00588 / 0.00546) = 0.0741 on
linear CL, log-shifts on baseline CRP / IC50 / logit-Imax; one of three
orthogonal indicators (DIS_RA / DIS_CD / DIS_SLE) decomposing the
four-level HV / RA / CD / SLE cohort with HV as the reference
category).DIS_PJIA (polyarticular juvenile idiopathic
arthritis, a pediatric-cohort sibling indicator). Scope: specific
because the disease-pooling reference category is paper-defined.
Ratified canonically on 2026-06-01 alongside the Li 2018 PF-04236921
extraction.IBD_CD, which is a pooled-UC+CD discriminator
with UC as the reference category; DIS_CD is used when the
complement group is a heterogeneous non-IBD cohort rather than UC
specifically.POPULATION /
STUDY categorical alongside other DIS_*
indicators.Li_2018_PF04236921.R
(log-shift e_cd_cl = log(0.00946 / 0.00546) = 0.5499 on
linear CL – a 73 percent higher typical CL in CD vs the HV reference,
consistent with the paper’s reported 60 percent higher CL in CD when
other covariates are held at reference; also log-shifts on baseline CRP,
IC50, and logit-Imax).IBD_CD instead.
Scope: specific because the disease-pooling reference category is
paper-defined. Ratified canonically on 2026-06-01 alongside the Li 2018
PF-04236921 extraction.POPULATION /
STUDY categorical alongside other DIS_*
indicators.Li_2018_PF04236921.R
(log-shift e_sle_cl = log(0.00643 / 0.00546) = 0.1632 on
linear CL, log-shifts on baseline CRP / IC50 / logit-Imax, and a Hill
coefficient effect e_sle_gamma = log(1.55) = 0.4383
shifting gamma from 1 in the HV/RA/CD reference to 1.55 in SLE).BGENE21 / BGENE21_HIGH (continuous / binary
IFN-21-gene scores within an SLE cohort) – those operate within an
SLE-only population whereas DIS_SLE is the across-cohort
disease-state flag. Scope: specific because the disease-pooling
reference category is paper-defined. Ratified canonically on 2026-06-01
alongside the Li 2018 PF-04236921 extraction.covariateData[[DIS_INFECT_CSSSI_SEV]]$notes should document
the underlying definition when the source paper provides one.SOI (severity-of-infection indicator; same orientation,
1 = severe / 0 = not severe) – used in
Lodise_2018_iclaprim.R (ASSIST-1 / ASSIST-2 phase 3 cSSSI
trials).Lodise_2018_iclaprim.R
(additive-linear shift on inter-compartmental clearance Q:
q_typ = exp(lq) + e_infect_csssi_sev_q * DIS_INFECT_CSSSI_SEV
with e_infect_csssi_sev_q = +13.5 L/h, so severe-cSSSI
patients have Q rise from 1.85 L/h to 15.35 L/h relative to non-severe
patients).DIS_INFECT_PNEUM_SEV, DIS_INFECT_HABP_SEV)
rather than overloading this entry. Distinct from
DIS_SASTHMA and other disease-state indicators (which
contrast a disease cohort with a non-disease reference) –
DIS_INFECT_CSSSI_SEV operates within an
already-cSSSI cohort. The covariate-effect parameter naming drops the
DIS_ prefix per the existing DIS_CANCER ->
e_cancer_* / DIS_CANCER_PED ->
e_cancer_ped_* convention; here that gives
e_infect_csssi_sev_<param>. Ratified canonically on
2026-05-30 alongside the Lodise 2018 iclaprim extraction.CARRAGEENAN – used in
VasquezBahena_2009_lumiracoxib_rat.R (1 = groups II-IX, 100
uL of 1% carrageenan suspension into the right hind paw; 0 = group I,
100 uL of 0.9% saline solution).VasquezBahena_2009_lumiracoxib_rat.R (switches the COX-2
synthesis-rate model: CARRAGEENAN = 0 selects the constant saline
synthesis rate ks_cox2_saline and CARRAGEENAN = 1 selects
the time-variant gamma function
ks_cox2(t) = A * t^alpha * exp(-beta * t) driving the
carrageenan-induced inflammation profile).LT paw withdrawal latency,
seconds) is the typical observable when this indicator is in use, but
other readouts (paw oedema, mechanical-allodynia von Frey threshold) are
equally valid. Distinct from disease-state indicators
(DIS_*) because the inflammatory state is experimentally
induced at a defined time, not a chronic patient condition.MEN – used in Shin_2014_sevoflurane.R
(Shin 2014 Appendix 1 NONMEM $INPUT column; same
orientation: MEN = 1 mentally disabled, MEN = 0 mentally intact).Shin_2014_sevoflurane.R (stratifies both C50 and the Hill
coefficient of the sigmoid-Emax probability of return of consciousness
during emergence from sevoflurane anesthesia: typical-value C50 0.37 vol
% vs 0.19 vol % and gamma 16.4 vs 4.53 for intact vs disabled).covariateData[[MENT_DISABLED]]$notes should
document the diagnostic basis when the source paper provides one.
Distinct from DIS_* disease-state indicators (which name a
specific disease) and from cognitive-score covariates
(SCORE_ADAS_COG, SCORE_MMSE,
SCORE_CDR_SOB, SCORE_FAQ) which are continuous
measurements; MENT_DISABLED is a binary developmental /
cognitive-impairment cohort indicator. Future emergence-from-anesthesia
or perioperative PK/PD models that use a similar binary
cognitive-impairment indicator may extend this entry; promote to general
scope once a second model ratifies the name with the same diagnostic
criterion family. Ratified canonically alongside the Shin 2014
sevoflurane emergence-PD extraction.Harun_2019_cysticFibrosis.R model which is single-cohort
(all CF) and therefore carries no DIS_CF column;
DIS_CF is the canonical for popPK / PD analyses that pool a
CF patient cohort with a non-CF reference (typically healthy volunteers)
and retain the CF-vs-non-CF contrast as a covariate.CF, CYS, PATIENT, or via a
STUDY / GROUP categorical).Bulitta_2007_piperacillin.R (multiplicative power-form
effects on volumes via fcyf_vss^DIS_CF with
fcyf_vss = 0.926, i.e. about a 7.4% reduction in V1 and V2
for CF patients vs healthy adults of the same lean body mass; CL carries
the boundary-estimated fcyf_cl^DIS_CF with
fcyf_cl = 1.00, retained from the published model for
traceability though it has no numerical effect at the published point
estimate; reference category is the pooled adult healthy-volunteer
cohort), Bulitta_2011_cefpirome.R (log-scale effects on the
canonical lcl_renal + lcl_nonren decomposition and on the central /
shallow-peripheral / deep-peripheral volumes:
e_cf_cl_renal = log(1.07) = +0.06766 on CL_R,
e_cf_cl_nonren = log(1.13) = +0.12222 on CL_NR, and
e_cf_vc_vp_vp2 = log(0.98) = -0.02020 applied uniformly to
V1, V2, V3; typical values anchored to the HV column of paper Table 3
with FCYF effects from paper Table 4 LBM-allometric row).LBM rather than WT as
the size descriptor when the source paper found LBM to reduce
unexplained BSV more than total body weight, which is common in CF
cohorts whose body composition differs substantially from healthy
adults. Ratified canonically on 2026-06-07 alongside the Bulitta 2007
piperacillin extraction.covariateData[[DIS_SEPSIS]]$notes whether the indicator is
fixed at PK-sampling onset or updated dynamically over the observation
period. Used as a binary co-condition indicator on PK parameters
(clearance, volume) when a study population pools septic and non-septic
subjects, typically in ICU / burn-ICU / critically ill cohorts.SEPS – used in Han_2013_fluconazole.R (Han
2013 Methods / Table 1 covariate with values 0 = no sepsis, 1 = sepsis;
same orientation as the canonical).Han_2013_fluconazole.R
(additive shift on the non-RRT CL arm:
CL_norrt = theta1 + (CLCR/120)*theta4 + TBI*theta7 + SEPS*theta9
with theta9 = -0.369 L/h – septic patients have about 0.37
L/h lower CL in the non-RRT arm), Jeon_2014_piperacillin.R
(additive shift on central volume V1:
TVV1 = exp(lvc) + e_sepsis_vc * DIS_SEPSIS with
e_sepsis_vc = 14.8 L – septic patients have V1 = 25.3 +
14.8 = 40.1 L versus 25.3 L in non-septic patients, consistent with
capillary leakage and interstitial edema in sepsis).DIS_UC, DIS_HAE) – sepsis is
an acute / transient co-condition rather than a chronic indication
label. Distinct from CRP / IL6 /
SAPS_II, which are biomarker / severity-score columns that
may co-vary with clinical sepsis but are conceptually distinct from a
binary clinical-sepsis diagnostic flag. Ratified canonically on
2026-06-09 alongside the Han 2013 fluconazole extraction.covariateData[[DIS_EDEMA]]$notes whether the indicator
is fixed at PK-sampling onset or updated dynamically. Used as a binary
co-condition indicator on PK volume (most commonly) when third-space /
extravascular fluid expansion is expected to alter drug
distribution.EDEM – used in Han_2013_fluconazole.R (Han
2013 Methods / Table 1 covariate with values 0 = no edema, 1 = clinical
edema by the paper’s puffy-face and leg-pitting criterion; same
orientation as the canonical).Han_2013_fluconazole.R
(additive shift on V:
V = (WT/65)*theta5 + EDEM*theta6 + TBI*theta8 with
theta6 = 13.6 L – edematous patients have about 13.6 L
larger V).DIS_HAE (hereditary angioedema disease-state indicator) –
DIS_EDEMA is a generic clinical-sign flag (e.g.,
burn-related capillary leak, congestive heart failure, hypoalbuminemia)
rather than a disease classification. Distinct from a continuous
fluid-balance / weight-gain column. Ratified canonically on 2026-06-09
alongside the Han 2013 fluconazole extraction.covariateData[[DIS_BURN_RECENT]]$notes. Used as a piecewise
covariate on PK parameters (CL and / or V) that change during the acute
hypermetabolic phase.TBI – used in Han_2013_fluconazole.R (Han
2013 Methods / Tables 2 and 3 covariate with values 1 = within 30 days
of burn injury, 0 = at least 30 days postburn; same orientation as the
canonical, with the recency threshold = 30 days documented per-model).
NOTE: TBI in this register is a recoded binary covariate;
it does NOT denote traumatic brain injury (the common medical
abbreviation) and is NOT a continuous time-from-burn-injury value.Han_2013_fluconazole.R
(recency threshold = 30 days; additive shifts on both CL and V:
CL_norrt += TBI*theta7 with theta7 = 0.504 L/h
and V += TBI*theta8 with theta8 = 9.61 L –
recent-postburn patients have about 0.5 L/h higher CL and 9.6 L larger V
relative to the resolved-phase reference).TPP for time-postpartum) would warrant its own canonical
(suggested name: POSTBURN_DAYS);
DIS_BURN_RECENT is reserved for binary phase indicators.
The paper’s source-column name TBI is preserved as a
documented alias even though it collides with the medical abbreviation
for traumatic brain injury – this is the author’s chosen column name and
the canonical avoids the collision. Ratified canonically on 2026-06-09
alongside the Han 2013 fluconazole extraction.idiopathic scoliosis /
idiopathic kyphoscoliosis diagnosis labels in Stricker 2015
Table 1, decomposed from a 3-level diagnosis categorical {craniofacial,
idiopathic scoliosis, non-idiopathic scoliosis} into binary
DIS_SCOL_IDIO with value 1 for the idiopathic-scoliosis cohort.Stricker_2015_aminocaproic_acid.R (multiplicative
power-form effect on CL: cl *= 1.10^DIS_SCOL_IDIO;
idiopathic-scoliosis adolescents have ~10% higher CL than the
craniofacial-cohort reference at the same body weight and post-natal
age, after allometric scaling; Stricker 2015 Table 5 ‘Impact of
idiopathic spines’ = 1.10; effect retained for magnitude / precision
reporting even though the authors note it is not clinically
meaningful).DIS_SCOL_NONIDIO in the Stricker 2015
three-level surgical-cohort stratification {craniofacial reference,
idiopathic scoliosis, non-idiopathic scoliosis}; both = 0 selects the
craniofacial reference. Distinct from ADOLESCENT (pure
age-cohort indicator independent of aetiology) and from
CHILD (broader paediatric age-cohort indicator). Ratified
canonically on 2026-06-12 alongside the Stricker 2015
epsilon-aminocaproic acid extraction.Stricker_2015_aminocaproic_acid.R (multiplicative
power-form effect on CL: cl *= 0.97^DIS_SCOL_NONIDIO;
non-idiopathic-scoliosis adolescents have ~3% lower CL than the
craniofacial-cohort reference at the same body weight and post-natal
age, after allometric scaling; Stricker 2015 Table 5 ‘Impact of
non-idiopathic spines’ = 0.97; the authors describe this effect as not
clinically meaningful but retain it in the final model for precision
estimation).DIS_SCOL_IDIO in the Stricker 2015 three-level
surgical-cohort stratification {craniofacial reference, idiopathic
scoliosis, non-idiopathic scoliosis}; both = 0 selects the craniofacial
reference. Ratified canonically on 2026-06-12 alongside the Stricker
2015 epsilon-aminocaproic acid extraction.chf – used in Thomson_1989_lisinopril.R
(Thomson 1989 Table 2 cardiac-failure indicator; treated as a
multiplicative factor 0.645 raised to chf on CL/F so the
unaffected reference group keeps factor 1).Thomson_1989_lisinopril.R (multiplicative power-form effect
on CL/F: e_chf_cl^DIS_CHF with
e_chf_cl = 0.645; ~35% lower apparent clearance with
compensated cardiac failure).CONMED_SPIRON (which captures the diuretic /
aldosterone-antagonist effect typically given alongside CHF therapy) and
from haemodynamic / autonomic covariates such as HR and the
blood-pressure outputs. For papers that decompose CHF severity into NYHA
classes, prefer a per-class encoding (e.g., DIS_CHF_NYHA3,
DIS_CHF_NYHA4) when the strata are retained in the final
model. The covariate-effect parameter form is
e_chf_<param> (drops the DIS_ prefix per
the existing DIS_CANCER -> e_cancer_<param>
precedent). Ratified canonically on 2026-06-10 alongside the Thomson
1989 lisinopril extraction.covariateData[[FEV1]]$notes if non-default – ATS / ERS
post-bronchodilator is the typical convention).(theta + e_fev1_param * (FEV1 - ref)) or power forms
(FEV1 / ref)^exponent. Reference values observed: 1.6 L
(Tortorici 2017, population median across the RAPID-RCT/RAPID-OLE A1-PI
augmentation cohort).FEV1
directly.Tortorici_2017_a1pi.R
(linear-deviation effect on the lung-density decline rate:
theta5 * (FEV1 - 1.6) with
theta5 = +0.56 (g/L/year per L FEV1); lower-FEV1 patients
have steeper natural decline rates independent of A1-PI exposure).Harun_2019_cysticFibrosis.R rather than a covariate
column). Use FEV1 only when the source paper supplies the
absolute-volume value; if the source supplies a percent-predicted value
as a covariate, the canonical for that surface is
FEV1_PCTPRED. Scope: specific until a second model ratifies
the absolute-litre semantics; promote to general at that point. Ratified
canonically on 2026-05-09 alongside the Tortorici 2017 extraction.(FEV1_PCTPRED / ref)^exponent or linear-deviation
forms (1 + e * (FEV1_PCTPRED - ref)). Reference values
observed: 62.1 % (Ting 2014, population median across the combined
three-study cystic-fibrosis cohort).FEV1% predicted – used in
Ting_2014_tobramycin_inhaled.R (paper Table 1 / Table 2).
Free-text label, not a typical NMTRAN column header; downstream data
sets should use the canonical column name
FEV1_PCTPRED.Ting_2014_tobramycin_inhaled.R (power-form effect on
apparent central volume of distribution:
(FEV1_PCTPRED / 62.1)^-0.303; lower lung-function patients
have larger apparent Vd/F, consistent with the paper’s hypothesis that
worsening lung disease increases central-airway aerosol
deposition).FEV1 (canonical for the unstandardised volume) and from
FEV1 as an outcome variable in disease-progression models
(e.g. Harun_2019_cysticFibrosis.R, where percent-predicted
FEV1 is the dependent variable rather than a covariate). The reference
equation used to derive the percent-predicted value is paper-specific;
document any non-default reference standard (Hankinson 1999 NHANES III,
GLI 2012, Wang 1993, etc.) in
covariateData[[FEV1_PCTPRED]]$notes per model. Scope:
specific until a second model ratifies the percent-predicted semantics;
promote to general at that point. Ratified canonically on 2026-05-12
alongside the Ting 2014 extraction.Cc rather than A1PI.covariateData[[A1PI]]$units). Conversion:
1 umol/L A1-PI ~= 5.2 mg/dL (using MW ~52 kDa). The 11 umol/L “putative
protective threshold” used clinically corresponds to ~57 mg/dL.(A1PI / ref)^exponent on the post-treatment
exposure intercept and slope. Reference values observed: 5.5 umol/L
(Tortorici 2017, approximate median pre-treatment A1-PI among RAPID-RCT
placebo-randomised patients, used as the normalisation denominator for
power-form covariate effects).Cbase – used in
Tortorici 2017’s published equation 6 to denote the baseline
pre-treatment A1-PI value; the column name in the modelled dataset would
be A1PI.Tortorici_2017_a1pi.R
(two power-form effects: (A1PI/5.5)^theta5 with
theta5 = +0.73 on the placebo-arm post-treatment exposure
intercept, and (A1PI/5.5)^theta4 with
theta4 = -0.12 on the dose-rate slope; together they encode
the modest dose-exposure dependence on each subject’s endogenous A1-PI
level).covariateData[[A1PI]]$notes per model. Specific scope
because the canonical is tied to AATD augmentation-therapy modelling;
future PK / PD analyses of A1-PI (or AAT) in non-AATD contexts
(acute-phase response, smoking-induced inflammation) should ratify
general scope at that time. Ratified canonically on 2026-05-09 alongside
the Tortorici 2017 extraction.ATS5C – used in
Harun_2019_cysticFibrosis.R (Harun 2019 NMTRAN
$INPUT column, “presence of air trapping at age 5”; values
0 = absent, 1 = present).Harun_2019_cysticFibrosis.R (linear-deviation effect on
baseline FEV1% predicted at age 5:
(1 + e_at_baseline * AIR_TRAP_5Y) with coefficient -0.0417,
i.e., subjects with severe air trapping at age 5 have a baseline FEV1%
predicted approximately 4.17% lower than those without).AIR_TRAP_8Y,
AIR_TRAP_PRAGMA, etc.) rather than overload this name.
Ratified canonically on 2026-05-08 alongside the Harun 2019
extraction.HOSPRA – used in
Harun_2019_cysticFibrosis.R (Harun 2019 NMTRAN
$INPUT column, “hospitalisation at the time of FEV1%
predicted measurement; 0=no, 1=yes”).Harun_2019_cysticFibrosis.R (linear-deviation effects on
the disease-progression maximum drop and on the half-effect age:
(1 + e_hpe_dmax * HOSPRA) with coefficient -0.22 on the
maximum FEV1% drop and (1 + e_hpe_t50max * HOSPRA) with
coefficient -0.235 on the age at which 50% of the maximum drop occurs;
hospitalised visits accelerate both the magnitude and the onset of FEV1%
decline).(1 + e * (LNPC - ref)). Reference
values observed: 5.88 log(parasites/uL) (Birgersson 2019, population
median in the pooled pregnant + non-pregnant Burkina Faso cohort).PARA (raw asexual parasite count per microlitre) is
provided alongside LNPC in the Birgersson 2019 NONMEM
dataset but the model uses LNPC = log(PARA) as the active
covariate.Birgersson_2019_artesunate.R (linear-deviation effect on
relative bioavailability F1:
F1LNPC = 1 + e_lnpc_f * (LNPC - 5.88); positive coefficient
e_lnpc_f = +0.138 per unit increase in log-parasite-count,
reflecting increased oral artesunate bioavailability with higher
parasite burden).LNPC but should document
their own cohort-specific reference value in
covariateData[[LNPC]]$notes. Distinct from
PARA (raw parasitaemia in parasites/uL), which is the
companion canonical for models that apply the log transform inside
model() rather than pre-computing it at dataset-assembly
time, and that use a time-varying (last-observation-carried-forward)
parasitaemia trajectory rather than the admission-only fixed value.
Ratified canonically on 2026-05-07.covariateData[[PARA]]$notes per model.(1 + e * log10(PARA)) or
(1 + e * (log10(PARA) - ref)) when the source paper applies
the log transform inside the model. Reference values observed: implicit
log10(PARA) = 0 (i.e. PARA = 1 parasite/uL) in Kloprogge 2014 (the
linear form anchors F at the typical population estimate when PARA =
1).PARA – used in Kloprogge_2014_quinine.R
(raw parasitaemia in parasites/uL, time-varying via
last-observation-carried-forward across the 7-day quinine treatment
course), in Kloprogge_2018_lumefantrine.R (admission-only /
time-fixed parasitaemia, centered at the model-building geometric-mean
15,800 parasites/uL = log10 4.2), and in
Tarning_2012_dihydroartemisinin.R (admission-only /
time-fixed parasitaemia, centered at the typical-patient log10(PARA) =
3.98 reported in the Table 4 footnote of Tarning 2012 AAC).Kloprogge_2014_quinine.R (linear effect on relative
bioavailability with the log10 transform applied inside
model():
F1PARA = 1 + e_para_f * log10(max(PARA, 1)) with
e_para_f = +0.389 per log10 parasitaemia; gated by
max(PARA, 1) so that PARA values below 1 parasite/uL
(effectively zero / below detection) collapse to no covariate effect,
matching the source paper’s “effect only during the acute phase when
parasitaemia was above the limit of detection” wording).
Kloprogge_2014_quinine.R – linear effect on relative
bioavailability with the log10 transform applied inside
model():
F1PARA = 1 + e_para_f * log10(max(PARA, 1)) with
e_para_f = +0.389 per log10 parasitaemia; gated by
max(PARA, 1) so that PARA values below 1 parasite/uL
(effectively zero / below detection) collapse to no covariate effect,
matching the source paper’s “effect only during the acute phase when
parasitaemia was above the limit of detection” wording.Kloprogge_2018_lumefantrine.R – exponential effect on
relative bioavailability with the log10 transform applied inside
model(), centered on log10(15,800) = 4.2:
F1PARA = exp(e_lnpc_f * (log10(max(PARA, 1)) - 4.2)) with
e_lnpc_f = -0.643 per log10 parasitaemia (higher
pre-treatment parasitaemia is associated with lower relative
bioavailability, consistent with reduced visceral blood flow in more
severe disease). Same max(PARA, 1) gating convention as
Kloprogge 2014.Tarning_2012_dihydroartemisinin.R – linear-deviation
effect on relative bioavailability with the log10 transform applied
inside model(), centered on log10(PARA) = 3.98 (the
pooled-cohort median in the Thai-Myanmar-border
dihydroartemisinin-piperaquine combination trial):
fpara = 1 + e_para_f * (log10(max(PARA, 1)) - 3.98) with
e_para_f = +0.278 per log10 unit (Results: ‘27.8% linear
increase per unit logarithmic parasitemia’). Higher pre-treatment
parasitaemia is associated with higher relative bioavailability of
dihydroartemisinin (Tarning 2012 Discussion: ‘…consistent with a
disease-related decrease in first-pass metabolism possibly compounded by
reduced hepatic blood flow…’). Same max(PARA, 1) gating
convention as the Kloprogge models; admission-only / time-fixed.LNPC – both
canonicals describe the same biological quantity (asexual Plasmodium
parasite count) but differ in the transform location: LNPC
is pre-transformed to natural log at dataset-assembly time and used as
(LNPC - ref), whereas PARA is the raw count
with the log transform applied inside model(). Use
PARA when the source paper reports the covariate
coefficient on the log10 scale (per log10 parasitaemia) and / or when
parasitaemia is time-varying. Use LNPC when the source
dataset pre-computes the natural log at admission. Scope: specific
because the gating-at-1 convention and the log10-inside-model() idiom
reflect the Mahidol-Oxford malaria popPK lab’s specific implementation;
a future malaria PK model that uses a different gating threshold (e.g.,
LOQ-aware, or log base-e inside the model) or a different reference
value should document its own convention in
covariateData[[PARA]]$notes. Ratified canonically on
2026-05-21 alongside the Kloprogge 2014 quinine extraction; extended to
Kloprogge 2018 lumefantrine on 2026-05-22 (same log10-inside-model()
idiom, distinct centering value: 4.2 = log10(15,800) for Kloprogge 2018
vs 0 = log10(1) for Kloprogge 2014); extended to Tarning 2012
dihydroartemisinin on 2026-06-01 (same log10-inside-model() idiom with
the source paper’s typical-patient centering value 3.98 =
log10(pooled-cohort-median ~ 9,550 parasites/uL)).SERPOS – used in
Lin_2024_casirivimab.R.Lin_2024_casirivimab.R
(multiplicative fractional change on CL).covariateData[[SARS_SEROPOS]]$notes.OXYSUP_HIGH indicator).OXYSTAT1 – used in
Lin_2024_casirivimab.R.Lin_2024_casirivimab.R
(multiplicative fractional change on CL; +10.6%).OXYSUP_HIGH indicator. Register a
separate OXYSUP_VENT canonical if a future analysis splits
mechanical ventilation from high-flow oxygen.OXYSTAT2 – used in
Lin_2024_casirivimab.R.Lin_2024_casirivimab.R
(multiplicative fractional change on CL; +38.0%).OXYSUP_LOW. In Lin 2024 the rare mechanical-ventilation
cases were pooled into the high-flow indicator (n = 24 across the
7598-subject dataset).HIV – used in Jonsson_2011_ethambutol.R
(DDMODEL00000220 NMTRAN $INPUT column with values 0 = HIV
negative, 1 = HIV positive; same orientation as the canonical).Jonsson_2011_ethambutol.R (multiplicative
1 + e_hiv_pos_f * HIV_POS shift on bioavailability;
HIV-positive patients exhibit a 15.5% reduction in ethambutol
bioavailability versus HIV-negative reference),
Jonsson_2011_ethambutol_ddmore.R (multiplicative
1 + e_hiv_pos_f * HIV_POS shift on bioavailability;
HIV-positive patients exhibit a 15.4% reduction in ethambutol
bioavailability versus HIV-negative reference)._POS suffix
convention used by ADA_POS, SARS_SEROPOS, and
other serostatus / antibody-positivity indicators. Distinct from a
primary disease-state indicator like DIS_HIV (not yet
registered) – HIV_POS is a comorbidity flag in
non-HIV-primary indications where HIV-vs-non-HIV is tested as a PK
covariate. Ratified canonically on 2026-05-06.TB – used in Bisaso_2014_albumin.R (paper
text and Table 1 stratification column with values 0 = HIV only, 1 = HIV
+ TB co-infection; same orientation as the canonical).Bisaso_2014_albumin.R
(multiplicative additive shift on baseline albumin secretion rate Q0:
Q0 = exp(lq0) * (1 + e_tb_pos_q0 * TB_POS) with
e_tb_pos_q0 = -0.308; TB-positive subjects have ~30.8%
lower Q0 than the HIV-only reference – equivalent to the paper text’s
“44.2% lower” framing relative to the TB-HIV cohort)._POS suffix
convention used by HIV_POS, ADA_POS,
SARS_SEROPOS, and other serostatus / disease-state
indicators. Distinct from any TB-treatment-regimen indicator
(e.g. CONMED_RIF_LPVR4 for concomitant rifampicin) –
TB_POS is the active-disease flag; the medication exposure
is a separate concept. In Bisaso 2014 all 158 TB-positive subjects were
also on rifampicin-based anti-TB therapy, so the two are confounded in
that single cohort; the canonical preserves the conceptual distinction
for future studies that decouple them. Ratified canonically on
2026-05-20 alongside the Bisaso 2014 albumin extraction.HCV – used in
Majekodunmi_2017_HIV_HCV_CD4_recovery.R (paper Methods
‘Definitions’ and Table 2 covariate ‘C:Coinf’ for HIV/HCV coinfected vs
HIV monoinfected; same orientation as the canonical).Majekodunmi_2017_HIV_HCV_CD4_recovery.R (multiplicative
fractional reduction on the CD4 z-score recovery-rate constant c:
c = (c_pop + etac) * (1 + e_hcv_pos_c * HCV_POS) with
e_hcv_pos_c = -0.77; HCV-coinfected children recover at 23%
of the HIV-monoinfected rate – 0.357 /year versus 1.55 /year
typical)._POS suffix
convention used by HIV_POS, TB_POS,
ADA_POS, SARS_SEROPOS, and other serostatus /
disease-state indicators. Distinct from any anti-HCV treatment-regimen
indicator (e.g., pegylated interferon + ribavirin in Majekodunmi 2017’s
coinfected subset of 10 children) and from any HCV-genotype indicator
(1/2/3/4 distribution reported in Majekodunmi 2017 Table 1 but not used
as a covariate). Distinct from a primary disease-state indicator like
DIS_HCV (not yet registered) – HCV_POS is the
coinfection / comorbidity flag in non-HCV-primary indications. Ratified
canonically on 2026-05-22 alongside the Majekodunmi 2017 CD4 recovery
extraction.HCV_GT1B.Wang_2018_daclatasvir_asunaprevir.R (switches the
daclatasvir and asunaprevir antiviral IC50 and the daclatasvir
resistance coefficient between GT1A and GT1B values via fixed scaling
factors: IC50,DCV 0.041 -> 0.0074 ug/L (SCL = 0.18), IC50,ASV 2.45
-> 0.74 ug/L (SCL = 0.30), Kr,DCV 0.43 -> 0.13 per day; Kr,ASV is
the same for both subtypes).HCV_POS (the HCV coinfection / comorbidity flag in
non-HCV-primary indications) – HCV_GT1B is a
within-HCV-population subtype indicator for an HCV-primary antiviral
analysis. Future HCV antiviral models that distinguish additional
genotypes (GT2/3/4) should register sibling subtype canonicals rather
than overload this binary 1B-vs-1A indicator. Ratified canonically
alongside the Wang 2018 daclatasvir/asunaprevir extraction.Svensson_2017_bedaquiline.R (multiplicative
(1 + e_xdr_hl * DIS_TB_XDR) shift on the half-life of
mycobacterial load decline; pre-XDR/XDR patients have a 28.1% longer MBL
half-life than MDR/susceptible patients, translating to 2-4 weeks longer
median time-to-sputum-culture-conversion per Svensson 2017 Table
3).DIS_TB_XDR_STRICT for XDR-only,
DIS_TB_RIFRES for rifampicin-monoresistant) rather than
overloading this name. Distinct from HIV_POS (a comorbidity
indicator on TB-coinfected patients, not a drug-resistance stratum).
Ratified canonically on 2026-05-21 alongside the Svensson 2017
bedaquiline extraction.(TTP_MGIT_BASE / ref)^exponent. Reference value observed:
6.8 days (Svensson 2017 Table 1 cohort median, n = 191).mTTP0 (Svensson 2017 paper symbol, also written
mTTP_{0,i}) – used in
Svensson_2017_bedaquiline.R.Svensson_2017_bedaquiline.R (power-form covariate on the
starting mycobacterial load
mbl0_i = exp(lmbl0) * (TTP_MGIT_BASE / 6.8)^e_ttp_mbl0 with
e_ttp_mbl0 = -3.69; a four-fold longer median TSCC is
predicted in patients with the lowest baseline TTP versus the highest,
per Svensson 2017 Discussion paragraph 3).covariateData[[TTP_MGIT_BASE]]$notes. Distinct from
LNPC (log-transformed admission Plasmodium parasitaemia, a
different pathogen-quantification method for malaria) and from
SARS_VLOAD (SARS-CoV-2 RT-qPCR log10 copies/mL, a different
infectious-disease baseline biomarker). MGIT samples without a positive
signal within 42 days are classified as negative in the source paper;
for the per-subject baseline mean, only the positive triplicate values
are averaged (a subject with all-negative baseline samples has missing
TTP_MGIT_BASE). Ratified canonically on 2026-05-21 alongside the
Svensson 2017 bedaquiline extraction.covariateData[[CD4_ABS]]$notes.
Used as a continuous disease-severity / immunosuppression-depth
biomarker in HIV / AIDS popPK models and as a covariate on clearance for
drugs whose disposition is altered by advanced HIV
immunosuppression.CL = exp(lcl) + e_crcl_cl * CRCL + e_cd4_abs_cl * CD4_ABS;
coefficient in L/h per cell/mm^3) or as a centred linear / power term
(CD4_ABS / ref)^exponent against a population reference
(typical reference 50 or 200 cells/mm^3 in HIV cohorts). Document the
form and reference value in each model’s
covariateData[[CD4_ABS]]$notes.CD4 – bare-name source-paper column; common in
clinical-PK papers (McLachlan 1996 prints “CD4 cell count” in the
abstract and “CD4+ T-lymphocyte count” in the Methods).CD4_COUNT – explicit-suffix variant.CD4_T_COUNT – verbatim “CD4+ T-lymphocyte count” form
used in some clinical-PK papers.McLachlan_1996_fluconazole.R (raw additive linear slope on
the CL covariate model: 0.00068 L/h per cell/mm^3; cohort mean 69
cells/mm^3 with 97 of 109 covariate-evaluable subjects below 200
cells/mm^3 per Table 3; entered as e_cd4_abs_cl * CD4_ABS
in the additive intercept-plus-slopes CL regression alongside
e_crcl_cl * CRCL).Majekodunmi_2017_HIV_HCV_CD4_recovery.R), which is a
paediatric-recovery standardised metric on the natural-z-score scale and
is not directly interchangeable with the absolute count. Distinct from
CD4 percentage (CD4_PCT, the proportion of total lymphocytes that are
CD4+; commonly reported in paediatric HIV studies and not yet
registered) – register CD4_PCT separately if a future paper retains it.
Distinct from HIV_POS (the binary serostatus indicator),
which captures HIV-positive-vs-negative rather than the depth of
immunosuppression within an HIV-positive cohort. The two can be paired
in a single model when an HIV-positive subset is further stratified by
CD4 count, but most HIV-popPK papers focus on either the comorbidity
flag (mixed HIV+/HIV- cohort) or the absolute count (HIV+-only cohort),
not both. Ratified canonically on 2026-06-10 alongside the McLachlan
1996 fluconazole extraction.(SUVmax(t = 168 h) - SUVmax(0)) / SUVmax(0), i.e., the most
negative (largest reduction) relative change in
[18F]FDG-PET standardized uptake value at one week of
sunitinib therapy. The OS Weibull hazard is
lambh * alphh * t^(alphh - 1) * exp(theta_pred * RCFB1MAX)
with alphh = 1 (degenerates to constant baseline
hazard).RCFB1MAX
(greater week-1 SUVmax suppression) reduces the OS hazard (Schindler
2016 reports a positive theta_pred = 5.36, so
exp(5.36 * RCFB1MAX) is < 1 for negative RCFB1MAX).RCFB1MAX (Schindler 2016 NONMEM $ERROR
block intermediate; “max relative change in SUVmax from baseline at week
1 across lesions”). Computed inline in the source .mod from
the on-the-fly SUVmax states at TIME = 168 h and reused in subsequent
records.Schindler_2016_sunitinib.R (DDMODEL00000221)..mod RCFB1MAX is a
record-loop state, not a true subject-level covariate – it is captured
at FLAG = 1 / TIME = 168 h from the running SUVmax compartment values
and reused on every subsequent record. nlmixr2 / rxode2 do not have an
idiomatic equivalent of NONMEM’s record-loop persistent state, so the
model file consumes RCFB1MAX as a per-subject input
covariate; reproducing the source’s behavior requires a two-stage
simulation (run the SUVmax + SLD ODEs first, compute
RCFB1MAX per subject from the t = 168 h SUVmax values, then
run the OS / dropout TTE arms with RCFB1MAX bound). The
vignette virtual cohort follows this pattern.covariateData[[TUMSZ]]$units and notes).(TUMSZ / ref)^exponent for continuous effects, or with a
paper-specific threshold for categorical-stratum indicators (e.g.,
Gibiansky 2014 splits BSIZ at 1750 mm^2 into low- vs high-burden
strata). Reference values observed: 41 mm (Zhou 2025); 63 mm (Budha
2023); 90 mm (Lu 2014, source reference 9 cm converted to mm); 1750 mm^2
threshold (Gibiansky 2014, SPPD).LDIAM (Zhou 2025; pediatric lymphoma “linear diameter”
of target lesions in mm).TMBD (originally in cm;
TUMSZ_mm = TMBD_cm * 10) – used in
Lu_2014_trastuzumabemtansine.R.BSIZ (Gibiansky 2014; baseline tumor size as the sum of
products of perpendicular diameters of target lesions, mm^2; used as the
categorical indicator (BSIZ <= 1750) in the obinutuzumab
popPK model rather than as a continuous power covariate).Budha_2023_tislelizumab.R (reference 63 mm),
Lu_2014_trastuzumabemtansine.R (reference 90 mm; source
column TMBD in cm, values converted to mm on ingestion),
Zhou_2025_brentuximab.R (reference 41 mm; source column
LDIAM is the sum of linear diameters of target lesions; effect on ADC
clearance only), Gibiansky_2014_obinutuzumab.R (SPPD in
mm^2; used as a categorical indicator (TUMSZ <= 1750) on
the time-dependent clearance decay rate kdes, not as a continuous power
covariate), Hansson_2013b_sunitinib.R (DDMODEL00000198;
observed baseline tumor SLD used as the per-subject IC of the tumor-size
ODE via the IPP-style proportional baseline-residual construction
tumor(0) = TUMSZ * (1 + etaibase * propSd); the source .mod
reads OBASE from DV at TIME=0/FLAG=4, but nlmixr2 / rxode2 cannot
replicate the in-record assignment idiom and consumes the observed
baseline as a covariate instead).(TUMSZ / ref)^exp is numerically invariant. For SPPD
constructs the natural unit is mm^2 (a product of two perpendicular
diameters in mm); record that in the per-model
covariateData[[TUMSZ]]$units field and do NOT cross-mix mm
and mm^2 within a single ingest. When a source paper specifically
reports the RECIST 1.1 “sum of longest diameters” of target lesions, use
the more specific TUM_SLD canonical instead –
TUMSZ remains the pooled-tumor-burden register.TUMSZ canonical; use TUM_SLD when the source
paper explicitly reports “sum of longest diameters” (or “sum of
lesions”) as the tumor-burden metric, distinct from the pooled “sum of
diameters / SPPD / sum of linear diameters” mixture covered by
TUMSZ.(TUM_SLD / ref)^exponent. Reference values observed: 70.0
mm (de Vries Schultink 2020 zenocutuzumab population median).SoL / “sum of lesions” (de Vries Schultink 2020
zenocutuzumab) – same construct, mm.deVriesSchultink_2020_zenocutuzumab.R (reference 70.0 mm;
power exponent 0.447 on Vmax of the parallel non-linear /
Michaelis-Menten clearance).TUMSZ (pooled
tumor-size canonical covering RECIST sum-of-diameters / SPPD /
sum-of-linear-diameters); TUM_SLD is the precise RECIST 1.1
sum-of-longest-diameters metric. Ratified canonically on 2026-04-29
alongside the pilot bispecific extraction (de Vries Schultink 2020
zenocutuzumab). When the source paper reports tumor burden in cm,
convert to mm (the canonical unit) on data ingestion and scale the
per-model reference accordingly so (TUM_SLD / ref)^exp is
numerically invariant. Also used as the per-subject initial-condition
input for tumour-growth / angiogenesis-inhibition (TGI) ODE models where
the source paper sets the SLD state at time zero from the observed
baseline SLD rather than estimating a typical-value baseline (e.g.,
Ouerdani_2015_pazopanib.R uses
tumorSize(0) <- TUM_SLD).volume = (length * width^2) / 2 (the standard
ellipsoid-approximation formula used in xenograft pharmacology).
Per-subject (per-animal) baseline measurement at randomisation into a
dosing group. Used both as a covariate stratifier (size at
randomisation) and as the per-subject initial-condition input for TGI
ODE models where the source paper sets the tumor-volume state at time
zero from the observed measurement rather than estimating a
typical-value baseline.mm^3tumorSize(0) <- TUM_VOL); not
a covariate effect coefficient.P0 (Ouerdani 2015 mouse model symbol for the observed
initial tumour volume; mm^3 in the CAKI-2 xenograft cohort; range
100-250 mm^3 at randomisation per the paper’s preclinical Methods).Ouerdani_2015_pazopanib_mouse.R (Ouerdani 2015 preclinical
TGI in CAKI-2 xenograft mice; per-subject TUM_VOL
initialises the tumorSize state and is held constant per
individual across the 24-day dosing window).TUM_VOL whenever a
preclinical paper supplies per-animal caliper tumor volumes as the
per-subject initial state of a tumor-volume ODE; use
TUM_SLD for clinical RECIST sum-of-longest-diameters (a
length, not a volume) and TUMSZ for the pooled “baseline
tumor burden as a covariate on PK” use case in clinical models. Ratified
canonically on 2026-05-12 alongside the Ouerdani 2015 pazopanib mouse
extraction.TUMTP_CHL – prior canonical name (pre-2026-06-19
standardization audit).TUMTP (categorical column with levels like
cHL, GC, …) – decompose into
TUMTP_HODGKIN_CLASSICAL = as.integer(TUMTP == "cHL").DIS (Zhou 2025; integer code with DIS == 1
flagging HL) – decompose into
TUMTP_HODGKIN_CLASSICAL = as.integer(DIS == 1). Zhou 2025
calls the complement “non-HL”; in the Zhou 2025 cohort the non-HL group
is exclusively sALCL.Budha_2023_tislelizumab.R,
Zhou_2025_brentuximab.R (effects on ADC Q2, MMAE central
volume VM, and the ADC->MMAE conversion-decay rate ALFM; the Zhou
2025 paper anchors typical-value parameters to HL patients so the model
uses (1 - TUMTP_HODGKIN_CLASSICAL) as the on-effect
indicator with reference category 1 = HL).TUMTP_GASTRIC in
Budha 2023; a patient can have at most one of the indicators set to 1
(the remaining tumor types collapse into the reference 0 group). The
reference category is the off-encoded value (0) by definition; when a
source paper anchors typical-value parameters to the HL group rather
than the non-HL group (as Zhou 2025 does), encode the effect as
coef^(1 - TUMTP_HODGKIN_CLASSICAL) so the canonical column
meaning (1 = cHL/HL) is preserved while the paper’s reference (HL) still
receives multiplier 1.(HER2_ECD / ref)^exponent. Reference 25 ng/mL used in Lu
2014; reference 8.23 ng/mL (population median) used in Bruno 2005, with
a 200 ng/mL plateau cap.ECD – used in
Lu_2014_trastuzumabemtansine.R and
Bruno_2005_trastuzumab.R.Lu_2014_trastuzumabemtansine.R (reference 25 ng/mL;
exponent 0.035 on CL), Bruno_2005_trastuzumab.R (reference
8.23 ng/mL; power exponents 0.041 on CL and 0.105 on V; HER2_ECD capped
at 200 ng/mL inside model() to reflect the Bruno 2005 plateau
observation), deVriesSchultink_2018_trastuzumab_LVEF.R
(inherits the Bruno 2005 trastuzumab popPK as a deterministic forcing
function with the same 8.23 ng/mL reference and 200 ng/mL plateau cap;
PK IIV is omitted per the source paper’s ‘fixed effect parameters’
clause, so HER2_ECD only enters via the typical-value PK covariate
equations).EGFR_ECD) rather than reusing this one.
Disambiguated from the covariate-columns register by the explicit
HER2_ prefix.exp(coef * TRAST_BL). TRAST_BL = 0
corresponds to no detectable residual trastuzumab (reference condition
used in Lu 2014).TBL – used in
Lu_2014_trastuzumabemtansine.R.Lu_2014_trastuzumabemtansine.R (linear-on-log coefficient
-0.002 per ug/mL on CL).EC50_i = EC50_pop * (TROPONIN_T_MAX / 18)^(-1.16), where 18
ng/L is the cohort-median TROPONIN_T_MAX. A TROPONIN_T_MAX value at the
cohort median leaves EC50 unchanged from its population value.TRPmax – used in
deVriesSchultink_2018_trastuzumab_LVEF.R.deVriesSchultink_2018_trastuzumab_LVEF.R (power covariate
effect on EC50 with median-centered normalisation, exponent -1.16,
explaining 15.1% of inter-individual variability in EC50; per de Vries
Schultink 2018 Table 2 and the EC50 covariate equation in Results).deVriesSchultink_2018_anthracycline_troponinT.R). Treated
as a time-fixed baseline covariate for the LVEF model: even though the
troponin T peak occurs during anthracycline treatment, by the time the
LVEF observations start (typically a few weeks after the last
anthracycline dose), the peak is a determined historical scalar per
subject. Specific scope until a second cardiotoxicity /
cardiac-biomarker model legitimately reuses this indicator. Values are
positive; sub-LLOQ peak values are imputed at LLOQ/2 (= 1.5 ng/L for the
Roche Modular E hs-TnT assay used in de Vries Schultink 2018).TUMTP_HODGKIN_CLASSICAL).TUMTP_GC – prior canonical name (pre-2026-06-19
standardization audit).TUMTP (categorical column) – decompose into
TUMTP_GASTRIC = as.integer(TUMTP == "GC").TTYPE (Quartino 2019; categorical column with levels
MBC, EBC, HV, AGC,
Others) – decompose into
TUMTP_GASTRIC = as.integer(TTYPE == "AGC").TTYPE4 (Wang 2024; level 4 of a five-level tumor-type
factor labelled “GCGEJ” in the source) – decompose into
TUMTP_GASTRIC = as.integer(TTYPE4 == 1).Budha_2023_tislelizumab.R,
Quartino_2019_trastuzumab.R (advanced gastric cancer;
per-group typical-value switch on linear CL and Vc rather than an
exponential multiplier), Wang_2024_sugemalimab.R (gastric +
GEJ adenocarcinoma pooled; exponential coefficient log(1.13) on CL and
log(1.14) on Vc).RACE_<GROUP>
indicator-decomposition pattern. New oncology tumor types should be
added as additional TUMTP_<GROUP> entries so the
reference set stays explicit. “Advanced gastric cancer” (AGC), “gastric
cancer” (GC), and “GC or adenocarcinoma of the gastroesophageal
junction” (GCGEJ) are pooled onto a single TUMTP_GASTRIC
indicator; document the per-paper stage-of-disease and GEJ-inclusion
detail in covariateData[[TUMTP_GASTRIC]]$notes. ESCC
(squamous histology) is captured by the separate TUMTP_ESCC
indicator and is not pooled here.TUMTP_<GROUP> indicators defined in
the same model.TUMTP_OTH – prior canonical name (pre-2026-06-19
standardization audit).TTYPE (Quartino 2019) – decompose into
TUMTP_OTHER = as.integer(TTYPE == "Others").PAT2 (Sathe 2024) – integer-coded tumor type column
with levels 1 (mTNBC), 2 (mUC or HR+/HER2- mBC), 4 (Other epithelial);
the source NONMEM control stream collapses PAT2 = 1 and PAT2 = 2 into
the reference (no effect) and applies the deviation only when PAT2 = 4.
Decompose into TUMTP_OTHER = as.integer(PAT2 == 4).Quartino_2019_trastuzumab.R (per-group typical-value switch
on linear CL; NSCLC plus a small residual group of prostate, ovarian,
and other histologies), Sathe_2024_sacituzumab.R
(multiplicative effect on tAB CL: -13.4% when TUMTP_OTHER = 1; “Other”
pool = small-cell and non-small-cell lung cancer, colorectal cancer,
esophageal cancer, pancreatic ductal adenocarcinoma, etc., n = 184;
reference = pooled mTNBC + mUC + HR+/HER2- mBC, n = 345),
Lacy_2018_cabozantinib.R (multiplicative fractional effect
on CL/F = +0.178 and on Vc/F = -0.186 for “other malignancies” relative
to the healthy-volunteer reference; n = 40 / 1534 = 2.6% of the cohort,
all from Study XL184-001 mixed-malignancy first-in-human cohort),
Wang_2024_sugemalimab.R (multiplicative effect
exp(e_oth_cl * TUMTP_OTHER) on CL and
exp(e_oth_vc * TUMTP_OTHER) on Vc, with the “Other” pool =
the residual tumor-type bucket complementing TUMTP_LYMPH / TUMTP_GASTRIC
/ TUMTP_ESCC in the same model).TUMTP_OTHER columns are not
interchangeable. Document the exact per-paper composition (e.g., “NSCLC
+ prostate + ovarian + other, n = 107 in Quartino 2019”; “small-cell +
non-small-cell lung + CRC + esophageal + pancreatic ductal
adenocarcinoma, n = 184 in Sathe 2024”; “advanced mixed malignancies
enrolled in the FIH Study XL184-001, n = 40 in Lacy 2018”) in
covariateData[[TUMTP_OTHER]]$notes. A given subject can
have at most one of the TUMTP_<GROUP> indicators
(including TUMTP_OTHER) set to 1; all-zero means the
reference group.(SPDL1 / ref)^exponent. Reference value observed: 173.8
pg/mL (study-population median in Ogasawara 2020).SPDL1 is the
standard abbreviation used directly in source analyses.Ogasawara_2020_durvalumab.R (power effect on CL, exponent
0.0617, reference 173.8 pg/mL; time-varying; values below LOD imputed as
LOD/2 = 33.55 pg/mL), Quartino_2019_trastuzumab.R
(per-group typical-value switch on linear CL; NSCLC plus a small
residual group of prostate, ovarian, and other histologies),
Wang_2024_sugemalimab.R (heterogeneous solid-tumor residual
group of n = 174; exponential coefficient log(0.885) on CL and
log(0.926) on Vc; NSCLC is the reference group, not part of
TUMTP_OTHER).TTYPE3 (Wang 2024; level 3 of a five-level tumor-type
factor labelled “Other” in the source) – decompose into
TUMTP_OTHER = as.integer(TTYPE3 == 1).Quartino_2019_trastuzumab.R (per-group typical-value switch
on linear CL; NSCLC plus a small residual group of prostate, ovarian,
and other histologies), Wang_2024_sugemalimab.R
(heterogeneous solid-tumor residual group of n = 174; exponential
coefficient log(0.885) on CL and log(0.926) on Vc; NSCLC is the
reference group, not part of TUMTP_OTHER),
Ogasawara_2020_durvalumab.R (power effect on CL, exponent
0.0617, reference 173.8 pg/mL; time-varying; values below LOD imputed as
LOD/2 = 33.55 pg/mL).TUMTP_OTHER columns are not
interchangeable. Document the exact per-paper composition (e.g., “NSCLC
+ prostate + ovarian + other, n = 107 in Quartino 2019”; “miscellaneous
solid tumors excluding NSCLC, lymphoma, GCGEJ, and ESCC, n = 174 in Wang
2024”) in covariateData[[TUMTP_OTHER]]$notes. A given
subject can have at most one of the TUMTP_<GROUP>
indicators (including TUMTP_OTHER) set to 1; all-zero means
the reference group.covariateData[[MCPROT]]$units.exp(theta * MCPROT) (i.e., MCPROT enters
un-log-transformed). Reference values observed: 0 g/dL (Ide 2020
Vmax,REF reference) and 2.0 g/dL (Ide 2020 figure-1 reference
patient).TMCPROT (time-varying serum M-protein concentration) –
used in Ide_2020_elotuzumab.R. NONMEM column with
imputation sentinel -99 for missing observations, replaced
by population median 2.1 g/dL via
IF(TMCPROT.EQ.-99) TMCPROT = 2.1.Ide_2020_elotuzumab.R
(g/dL, time-varying; entered un-log-transformed as
exp(0.277 * MCPROT) on Vmax of the Michaelis-Menten
target-mediated elimination from the central compartment).MM_NIGG (which is the immunoglobulin subtype, an MM-disease
stratifier that is time-fixed), SBCMA (soluble BCMA, a
different MM tumor-burden biomarker for BCMA-targeting drugs), and
B2M (beta-2-microglobulin, a
renal-function-and-MM-disease-burden marker). The 1 g/dL = 10 g/L
conversion lets future SI-convention papers register the same canonical
with their own unit string.SPDL1 (soluble serum PD-L1 in pg/mL); PDL1_TUM
is a tissue biomarker, SPDL1 is a circulating
biomarker.exp(theta * PDL1_TUM) (uncentred multiplicative) or
1 + theta * (PDL1_TUM - ref) (centred linear deviation).
Reference values observed: 15 (Struemper 2025 population median across
PD-1 inhibitor arms); 70 (Struemper 2025 INTR@PID imputed value when the 22C3 assay was
unavailable).PD-L1,
PDL1, TPS, PDL1_TPS – paper
notation varies; map to canonical PDL1_TUM with no value
transformation.Struemper_2025_tumorsize_OS_nsclc.R (exponential effect on
the tumor shrinkage rate ks of a Stein bi-exponential TGI
model; coefficient 0.00902 per percent; applied only to PD-1
inhibitor-containing treatment arms via a derived has_pd1
indicator that gates the effect when TRT is one of the
eight pembrolizumab- or dostarlimab-containing categories).covariateData[[PDL1_TUM]]$notes. Tumor
Proportion Score (TPS) is the percent of TUMOR cells stained; Combined
Positive Score (CPS) and Immune Cell (IC) score are distinct percent
metrics that should be registered separately if a paper relies on them.
PDL1_TUM is a tissue-level biomarker that may coexist in a model with
the serum-level SPDL1. Scope: general because PDL1 IHC is a
widely reused oncology covariate. Founding example: Struemper 2025
(exp(PDL1_TUM * 0.00902) on ks, paper Table 3
footnote b).LMET is the
common NONMEM / clinical-PK abbreviation used directly by the source
papers.Quartino_2019_trastuzumab.R (exponential effect on linear
CL; +16.4% typical CL when LMET = 1, per Quartino 2019 Table 1 theta12 =
0.152).LMET a commonly tested covariate in oncology mAb population
PK analyses. Scope: general so future oncology papers can reuse the
canonical column. Time-fixed baseline indicator; if a source paper
treats it as time-varying (progression during treatment), document in
covariateData[[LMET]]$notes.MET – used in Bruno_2005_trastuzumab.R
(Bruno 2005 Methods: “MET=1 if number of metastatic sites=4 or greater;
otherwise MET=0”). When a source paper supplies the raw integer count
column (NMET, N_METS, etc.) rather than the
pre-binarised indicator, derive
MET_GE4 = as.integer(NMET >= 4).Bruno_2005_trastuzumab.R (multiplicative effect on linear
CL: typical CL multiplied by (1 + 0.221 * MET_GE4),
i.e. +22.1% CL in MET_GE4 = 1 patients; reference MET_GE4 = 0 per Bruno
2005 Table 3).MET_GEN canonical rather
than overloading this entry. A MET_GE4 = 1 patient may also
have LMET = 1; the two columns are not mutually exclusive
(liver-only metastases would have LMET = 1,
MET_GE4 = 0).TUM_SLD
(continuous sum of longest diameters, mm), which captures tumor burden
magnitude; NTARGET_GE3 captures lesion multiplicity /
spread.NTARGET – raw integer count of target lesions per
RECIST 1.1 (paper notation in Struemper 2025). When a source paper
supplies the integer count column, derive
NTARGET_GE3 = as.integer(NTARGET >= 3).Struemper_2025_tumorsize_OS_nsclc.R (multiplicative effect
on the typical-value baseline tumor size TVTSb:
TVTSb * (1 + 0.288 * NTARGET_GE3); +28.8% TVTSb for
patients with three or more target lesions vs one or two).NTARGET_GEN
canonical rather than overloading this entry. Scope: general so future
oncology popPK / popPD papers using the same dichotomisation can reuse
this column. Founding example: Struemper 2025.PS / BPS – used in
Bajaj_2017_nivolumab.R (BPS = “baseline performance
status”; the one study using Karnofsky Performance Status was mapped to
ECOG via Oken 1982 before binarization) and
Zhang_2019_nivolumab.R (paper’s binary collapse PS=0
vs. PS>0).ECOG_1 – alternative explicit form; equivalent to
ECOG_GE1 when ECOG only takes values 0, 1, 2 in the
analysis dataset (the typical oncology case).ECOG_PS_GT0 – retired name used in earlier register
drafts; semantically identical (>= 1 equals
> 0 for integer ECOG scores).ECOG101 (categorical 0/1/2 score with thresholding
IF(ECOG101.GT.0.5)) – used in
Ide_2020_elotuzumab.R. Decompose:
ECOG_GE1 = as.integer(ECOG101 >= 1).Bajaj_2017_nivolumab.R
(exponential effect on CL with coefficient 0.172),
Zhang_2019_nivolumab.R (exponential effect exp(0.181) on
baseline CL; additive effect -0.138 on the time-varying-CL Emax
parameter), Ide_2020_elotuzumab.R (multiplicative effect on
CL = 1.03; paired with ECOG_GE2 for separate ECOG=1 vs
ECOG>=2 effects), Netterberg_2017_docetaxel.R
(multiplicative effect on baseline ANC of the Friberg myelosuppression
chain: BACOV *= (1 + theta * ECOG_GE1) with theta = 0.130;
source column PERF with ordinal ECOG 0/1/2 values,
binarized via ECOG_GE1 = as.integer(PERF >= 1) per Kloft
2006).ECOG_GE1 = as.integer(ECOG >= 1).
Zhang 2019 uses ECOG_GE1 on both baseline CL and the
time-varying Emax parameter (unlike Bajaj 2017, which uses it on CL
only); document the structural role in each model’s
covariateData[[ECOG_GE1]]$notes. When a paper retains
separate effects for ECOG = 1 vs ECOG >= 2 (Ide 2020), pair this
column with ECOG_GE2 and supply both indicators in the
event dataset.ECOG_GE1.ECOG_GE1 and
ECOG_GE2, both indicators = 0 corresponds to ECOG = 0 and
(ECOG_GE1 = 1, ECOG_GE2 = 0) corresponds to
ECOG = 1).ECOG101 (categorical 0/1/2 score with thresholding
IF(ECOG101.GT.1.5)) – used in
Ide_2020_elotuzumab.R. Decompose:
ECOG_GE2 = as.integer(ECOG101 >= 2).Ide_2020_elotuzumab.R
(multiplicative effect on CL = 1.15; paired with ECOG_GE1
to retain separate ECOG = 1 vs ECOG >= 2 effects).ECOG_GE1. Use only
when the source paper reports a separate effect for ECOG >= 2 in
addition to ECOG_GE1; otherwise ECOG_GE1 alone is
sufficient. The paired (ECOG_GE1, ECOG_GE2)
decomposition reproduces a three-level (ECOG = 0,
ECOG = 1, ECOG >= 2) ordinal effect with
two binaries.TUMTP (categorical column with levels including
melanoma, NSCLC, SCLC,
CRC, HCC, RCC) – decompose into
TUMTP_SCLC = as.integer(TUMTP == "SCLC").Sanghavi_2020_ipilimumab.R (exponential coefficient -0.124
on CL).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC
decomposition pattern. SCLC is the only retained tumor-type indicator in
the Sanghavi 2020 final model after backward elimination; the other
tumor types collapse into the reference (melanoma) group.TUMTP (categorical column with levels including
melanoma, NSCLC, other) –
decompose into
TUMTP_NSCLC = as.integer(TUMTP == "NSCLC").Ahamadi_2017_pembrolizumab.R (proportional change of +14.5%
on CL for NSCLC patients relative to melanoma; the “other” cancer type
cohort – 1.01% of the population – is pooled into the melanoma reference
per the paper’s model description),
Aoyama_2012_sepantronium.R.TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Scope: general because
NSCLC is a high-frequency tumor-type contrast (with melanoma or “other”
reference) across PD-1 / PD-L1 / chemotherapy popPK analyses, and is
likely to recur in future extractions. Ratified canonically on
2026-05-17 alongside the Ahamadi 2017 pembrolizumab extraction.TUMTP_NSCLC (which distinguishes
NSCLC patients overall from other tumor types).TUMTP_NSCLC_NONADENO = as.integer(NSCLC_histology == "non-adenocarcinoma").Schmid_2017_nintedanib.R (multiplicative effect 1.36 on
BIBF 1202 ka2 absorption rate; reference category adenocarcinoma NSCLC
pooled with IPF and unknown-histology NSCLC).TUMTP_<type>
decomposition pattern (auto-approved family). Within-NSCLC histology
distinctions arise when a popPK / PD model in NSCLC differentiates
adenocarcinoma from squamous-cell carcinoma; the paired adenocarcinoma
reference can be made explicit by adding a
TUMTP_NSCLC_ADENO sibling in a future extraction if needed.
Ratified canonically on 2026-06-27 alongside the Schmid 2017 nintedanib
extraction.TUMTP_HRPC is the
canonical column for both wordings (no separate TUMTP_CRPC
is registered; if a future paper distinguishes hormone-naive metastatic
prostate cancer from CRPC, register a more specific canonical at that
time).TUMTP (categorical column with levels including
HRPC, NSCLC, MM) – decompose into
TUMTP_HRPC = as.integer(TUMTP == "HRPC").STUDY / CANCER_TYPE factor columns where
one level is HRPC or CRPC – decompose
identically.Aoyama_2012_sepantronium.R (proportional change of -4.5% on
CL for HRPC patients relative to the NSCLC reference; ratio THETA_HRPC =
0.955 in the paper’s power form), Lacy_2018_cabozantinib.R
(multiplicative fractional effect on CL/F = -0.009 and on Vc/F = -0.241
for CRPC patients relative to the healthy-volunteer reference; Lacy 2018
enrolled 823 CRPC patients across Studies XL184-203, XL184-306, and
XL184-307, the largest cancer-type cohort in the integrated popPK
analysis).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC / TUMTP_NSCLC decomposition
pattern. Scope: general because prostate-cancer cohorts (under either
HRPC or CRPC nomenclature) recur across small-molecule and
targeted-therapy popPK analyses. The “hormone-refractory” wording
reflects the 2010s convention; modern papers using CRPC map onto the
same canonical. Ratified canonically on 2026-05-20 alongside the Aoyama
2012 sepantronium extraction.POP (Lacy 2018 NONMEM categorical population indicator
with levels HV / RCC / CRPC / MTC / GB / Other) – decompose into
TUMTP_MTC = as.integer(POP == "MTC").Lacy_2018_cabozantinib.R (multiplicative fractional effect
on CL/F = +0.928 and on Vc/F = -0.07 for MTC patients relative to the
healthy-volunteer reference; the +93% CL/F increase in MTC is the
load-bearing finding of Lacy 2018 and explains why the MTC capsule label
dose is 140 mg/day while the RCC tablet label dose is only 60
mg/day).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC / TUMTP_NSCLC decomposition
pattern. Scope: general because MTC is the approved indication for
Cometriq (cabozantinib capsule) and Caprelsa (vandetanib), and future
TKI popPK papers in MTC populations are likely to recur. Ratified
canonically on 2026-05-25 alongside the Lacy 2018 cabozantinib
extraction.POP (Lacy 2018 NONMEM categorical population indicator
with levels HV / RCC / CRPC / MTC / GB / Other) – decompose into
TUMTP_RCC = as.integer(POP == "RCC").Lacy_2018_cabozantinib.R (multiplicative fractional effect
on CL/F = -0.129 and on Vc/F = -0.63 for RCC patients relative to the
healthy-volunteer reference; Lacy 2018 enrolled 282 RCC patients in
Study XL184-308 dosed at 60 mg/day cabozantinib tablet).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC / TUMTP_NSCLC decomposition
pattern. Scope: general because RCC cohorts recur frequently across TKI,
anti-PD-1, and anti-VEGF popPK analyses. Ratified canonically on
2026-05-25 alongside the Lacy 2018 cabozantinib extraction.TUMTP (categorical column with levels including
MM, NSCLC, HRPC) – decompose into
TUMTP_MEL = as.integer(TUMTP == "MM"). NOTE: in Aoyama 2012
the abbreviation MM denotes malignant (unresectable)
melanoma, NOT multiple myeloma; do not confuse with the canonical
MM register entry (which is specifically active multiple
myeloma).STUDY / CANCER_TYPE factor columns where
one level is melanoma or MM (melanoma sense) –
decompose identically.Aoyama_2012_sepantronium.R (proportional change of +24% on
CL for melanoma patients relative to the NSCLC reference; ratio THETA_MM
= 1.24 in the paper’s power form).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC / TUMTP_NSCLC decomposition
pattern. The canonical name uses MEL (not MM)
to disambiguate from the existing MM register entry for
multiple myeloma. Scope: general because melanoma cohorts recur across
PD-1 / PD-L1 / BRAF-inhibitor / small-molecule popPK analyses. Ratified
canonically on 2026-05-20 alongside the Aoyama 2012 sepantronium
extraction.TUMTP_* indicators; or leukemia as
the implicit reference in the Akbar 2025 cohort).TTYPE1 (Wang 2024) – decompose into
TUMTP_LYMPH = as.integer(TTYPE1 == 1). The Wang 2024 source
paper uses a multi-level TTYPE factor with levels 1 =
lymphoma, 2 = lung cancer (reference), 3 = other, 4 = GCGEJ, 5 =
ESCC.TUMTP_LYMPH = as.integer(cancer_type == "Lymphoma").Wang_2024_sugemalimab.R (exponential coefficient log(0.877)
on baseline CL and log(0.879) on Vc),
Akbar_2025_voriconazole.R (additive-fractional +1.91%
effect on CL relative to leukemia reference; 95% CI spans zero).TUMTP_HODGKIN_CLASSICAL (which is specifically classical
Hodgkin lymphoma). Wang 2024 pools two lymphoma histologies (extranodal
NK/T-cell lymphoma from CS1001-201 / NCT03595657 and classical Hodgkin
lymphoma from CS1001-202 / NCT03505996) into a single lymphoma
indicator; the indicator therefore captures a generic
“hematologic-vs-solid-tumor” contrast rather than a histology-specific
effect. Akbar 2025 uses a single “Lymphoma” category alongside leukemia,
sarcoma, breast cancer, myeloma, and glioma in a heterogeneous-cancer
TDM cohort. When a future paper studies a single lymphoma histology
distinct from cHL, register a more specific canonical (e.g.,
TUMTP_ENKTL, TUMTP_NHL) rather than
overloading this one. Document the per-paper histology composition in
covariateData[[TUMTP_LYMPH]]$notes. Promoted from
Scope: specific to Scope: general on
2026-05-09 alongside the Akbar 2025 voriconazole extraction so that any
heterogeneous-cancer-cohort PK analysis can use this canonical name
without scope-violation.TUMTP_BC – prior canonical name (pre-2026-06-19
standardization audit).TUMTP (categorical column with levels including
BC, NSCLC, CRC) – decompose into
TUMTP_BREAST = as.integer(TUMTP == "BC").Lu_2022_patritumab.R
(multiplicative fractional effect 0.811 on CLlin of DXd-conjugated
antibody for breast-cancer patients vs the NSCLC reference; CRC effect
was tested and found insignificant so CRC is pooled into the
reference).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Registers the
breast-cancer arm of an oncology-cohort tumor-type contrast; pair with
sister TUMTP_<GROUP> indicators (e.g.,
TUMTP_NSCLC, TUMTP_CRC) when a future paper
retains separate effects for additional tumor types beyond the implicit
reference. Ratified canonically on 2026-04-28.TUMTP_* indicators are also 0).TTYPE5 (Wang 2024) – decompose into
TUMTP_ESCC = as.integer(TTYPE5 == 1).Wang_2024_sugemalimab.R (exponential coefficient log(0.99)
on baseline CL and log(1.08) on Vc).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Distinct from
gastroesophageal-junction adenocarcinoma (which is captured by the
broader TUMTP_GASTRIC indicator that pools GC and GEJ
adenocarcinomas) – ESCC is a squamous-cell histology, not
adenocarcinoma. Document the per-paper histology composition in
covariateData[[TUMTP_ESCC]]$notes.PCALCL – used in Suri_2018_brentuximab.R.
Suri 2018 reports the effect as a power-form multiplier
cl_adc *= 0.728^TUMTP_PCALCL (pcALCL CL ~27% lower than
non-pcALCL).Suri_2018_brentuximab.R (effect on ADC clearance
only).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Distinct from
TUMTP_LYMPH (heterogeneous lymphoma pool) and
TUMTP_HODGKIN_CLASSICAL (classical Hodgkin lymphoma).
pcALCL is one of two histologies pooled into the broader CTCL category
in Suri 2018 (alongside mycosis fungoides); the model singles out pcALCL
because Suri 2018 backward elimination retained pcALCL as a separate
effect on ADC clearance after exploring the broader CTCL contrast.
Ratified canonically on 2026-04-28.TUMTP_* indicators all =
0).TUMTP_SARC = as.integer(cancer_type == "Sarcoma").
Used in Akbar_2025_voriconazole.R.Akbar_2025_voriconazole.R (additive-fractional +18.5%
effect on CL relative to leukemia reference).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Akbar 2025 pools
soft-tissue and bone sarcoma histologies into a single sarcoma category.
Scope: specific because the reference category (leukemia in Akbar 2025)
is paper-defined. Ratified canonically on 2026-05-09.TUMTP_* indicators all =
0).TUMTP_MYELO = as.integer(cancer_type == "Myeloma").
Used in Akbar_2025_voriconazole.R.Akbar_2025_voriconazole.R (additive-fractional -2.33%
effect on CL relative to leukemia reference; the 95% CI spans
zero).MM entry
(which is reserved for multiple-myeloma-as-primary-indication PK
studies; the MM definition lacks a complete schema and
predates the TUMTP_* convention) and from DIS_SMM
(smoldering multiple myeloma, an asymptomatic plasma-cell disorder). Use
TUMTP_MYELO when the source paper pools multiple myeloma
alongside other tumor types in a heterogeneous oncology cohort and
treats cancer type as a many-level categorical covariate.
Scope: specific because the reference category (leukemia in Akbar 2025)
is paper-defined. Ratified canonically on 2026-05-09.TUMTP_* indicators all = 0; in
Lacy 2018 the reference is healthy volunteer when paired with the other
TUMTP_* indicators = 0).TUMTP_GLIO = as.integer(cancer_type == "Glioma"). Used
in Akbar_2025_voriconazole.R.POP (Lacy 2018 NONMEM categorical population indicator)
– decompose into TUMTP_GLIO = as.integer(POP == "GB") (Lacy
2018 enrolled glioblastoma multiforme patients in Study XL184-201 dosed
at 140 mg/day cabozantinib capsule).Akbar_2025_voriconazole.R (additive-fractional +8.81%
effect on CL relative to leukemia reference; the 95% CI spans zero),
Lacy_2018_cabozantinib.R (multiplicative fractional effect
on CL/F = +0.216 and on Vc/F = -0.569 for GB patients relative to the
healthy-volunteer reference; n = 39 GB patients in Study
XL184-201).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Akbar 2025 reports a
single “glioma” category without further subdivision by histology or
grade. Scope: specific because the reference category (leukemia in Akbar
2025; healthy volunteer in Lacy 2018) is paper-defined. Ratified
canonically on 2026-05-09.TUMTP_LEUK = as.integer(cancer_type == "Leukemia").
Implicit reference category in Akbar_2025_voriconazole.R
(so the model file does not consume this column directly; it is
registered for future heterogeneous-cancer-cohort analyses).Akbar_2025_voriconazole.R.DIS_AML, DIS_BCPALL, DIS_CMML,
DIS_MDS_AML entries – those are for leukemia-only or
leukemia-vs-leukemia contrasts; TUMTP_LEUK is for
heterogeneous-cancer pooled cohorts where leukemia is one of several
tumor types and the analysis treats cancer type as a
many-level categorical. Akbar 2025 had leukemia as 56.8% of the cohort
and used it as the reference category. Scope: specific because the
reference category in any source paper is paper-defined. Ratified
canonically on 2026-05-09.TUMTP_DLBCL and TUMTP_MCL all = 0; the
residual indolent-B-cell-lymphoma pool in this paper is dominated by
follicular lymphoma).DIS (Gibiansky 2014; integer code with
DIS == 2 flagging BCL) – decompose into
TUMTP_BCL = as.integer(DIS == 2).Gibiansky_2014_obinutuzumab.R (effect on time-dependent
clearance decay rate kdes via the composite
(TUMTP_BCL + TUMTP_DLBCL + TUMTP_MCL) (any-NHL effect;
ratio 2.08) and on both time-dependent CL_T and steady-state CL_inf via
the composite (TUMTP_BCL + TUMTP_DLBCL) (shared BCL/DLBCL
effect; ratio 0.834 in the reverse direction, i.e., 16.6% lower CL than
CLL)).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Distinct from
TUMTP_LYMPH (a broader heterogeneous lymphoma pool that
lumps cHL with NHL histologies) – TUMTP_BCL is specifically
B-cell lymphoma and pairs with sibling TUMTP_DLBCL and
TUMTP_MCL for histology-specific contrasts within the NHL
family. When a future paper studies follicular lymphoma in isolation
(rather than pooled into BCL), register a more specific canonical (e.g.,
TUMTP_FL) rather than overloading this one. Ratified
canonically on 2026-05-11.TUMTP_BCL and TUMTP_MCL all = 0).DIS (Gibiansky 2014; integer code with
DIS == 3 flagging DLBCL) – decompose into
TUMTP_DLBCL = as.integer(DIS == 3).Gibiansky_2014_obinutuzumab.R (effect on kdes via the
any-NHL composite indicator; effect on CL_T and CL_inf via the shared
BCL/DLBCL composite indicator),
Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on
the grade >= 2 peripheral neuropathy hazard via
exp(theta_DLBCL * TUMTP_DLBCL) with
theta_DLBCL = -0.0697, SE 0.365; the median hazard ratio of
DLBCL vs FL is 0.931 per Figure 3).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Distinct from
TUMTP_LYMPH (broader lymphoma pool) and
TUMTP_PCALCL (primary cutaneous anaplastic large-cell
lymphoma; a CD30+ T-cell-lineage entity unrelated to DLBCL). DLBCL is
the most common high-grade B-cell-NHL subtype; the Gibiansky 2014 cohort
had only 30 DLBCL patients (4.4%), so a single estimated effect on CL is
shared with BCL (a much larger pooled group; see TUMTP_BCL
notes). Ratified canonically on 2026-05-11.TUMTP_BCL and TUMTP_DLBCL all = 0).DIS (Gibiansky 2014; integer code with
DIS == 4 flagging MCL) – decompose into
TUMTP_MCL = as.integer(DIS == 4).Gibiansky_2014_obinutuzumab.R (effect on kdes via the
any-NHL composite indicator; separate effect on CL_T and CL_inf via the
standalone MCL indicator (ratio 1.75, i.e., 75% higher CL than
CLL)).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Distinct from
TUMTP_LYMPH (broader lymphoma pool). The Gibiansky 2014
cohort had only 20 MCL patients (2.9%); the paper reports the highest
obinutuzumab CL among the four B-cell-malignancy histologies for MCL,
consistent with the highest CD20 expression density on MCL B-cells
relative to the other histologies. Ratified canonically on
2026-05-11.TUMTP_DLBCL = 0 and
TUMTP_OTHER_NHL = 0; the Lu 2017 covariate model uses FL
itself as the reference category, so the indicator does not appear
explicitly in model() – both TUMTP_DLBCL and
TUMTP_OTHER_NHL are 0 for FL patients).FL (Lu
2017) – decompose into
TUMTP_FL = as.integer(tumor_histology == "FL").Lu_2017_polatuzumab_neuropathy.R (FL is the implicit
reference category; the model file declares TUMTP_FL in
covariatesDataExcluded to record FL as the reference
without referencing it in model()).TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC /
TUMTP_SCLC decomposition pattern. Distinct from
TUMTP_BCL (a broader pooled-residual
indolent-B-cell-lymphoma indicator in Gibiansky 2014 that lumps FL with
other indolent histologies); use TUMTP_FL when the source
paper names follicular lymphoma in isolation. Ratified canonically on
2026-06-24 alongside the Lu 2017 polatuzumab vedotin TTE
extraction.TUMTP_DLBCL = 0).FL, DLBCL} (Lu 2017 “others” category) –
decompose into
TUMTP_OTHER_NHL = as.integer(!tumor_histology %in% c("FL", "DLBCL")).Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on
hazard of grade >= 2 peripheral neuropathy via the
exp(theta_otherNonFL * TUMTP_OTHER_NHL) proportional-hazard
term with theta_otherNonFL = 0.688, SE 0.758; the median
hazard ratio of “other non-FL” vs FL is 1.98 per Figure 3 with very wide
uncertainty given the small “others” sub-cohort).TUMTP_BCL
(Gibiansky 2014 four-level B-cell-lymphoma decomposition: CLL, BCL,
DLBCL, MCL – where BCL is the indolent-B-cell-lymphoma residual
including FL). When a future paper retains a different pool of non-FL
non-DLBCL histologies, document the per-paper histology composition in
covariateData[[TUMTP_OTHER_NHL]]$notes rather than
overloading the canonical. Ratified canonically on 2026-06-24 alongside
the Lu 2017 polatuzumab vedotin TTE extraction.LINE (categorical column with levels 1L,
2L, 3L+, …) – decompose into
LINE_1L = as.integer(LINE == "1L").RRFN (relapsed/refractory flag; treatment-naive
corresponds to RRFN == 0) – used in Lu_2019_polatuzumab.R.
Decompose: LINE_1L = as.integer(RRFN == 0).Sanghavi_2020_ipilimumab.R (exponential coefficient -0.0949
on CL), Lu_2019_polatuzumab.R (multiplicative effects on V1
= 1.20, kdes = 3.38, CL_T = 3.53, FRAC_NS = 0.756; the same pooled-trial
NHL cohort mixes 415 R/R and 45 first-line patients).exp(-0.0949 * LINE_1L) and Lu reports
theta^LINE_1L with theta < or > 1 depending on the
parameter); both reduce to the same canonical 0/1 column. If a future
paper requires finer resolution (separate effects for 2L vs 3L+), add a
parallel LINE_2L canonical rather than overloading this
one.NIVO_REGIMEN (categorical column with levels
none, 0.3 mg/kg Q3W, 1 mg/kg Q2W,
1 mg/kg Q3W, 3 mg/kg Q2W,
3 mg/kg Q3W) – decompose into
NIVO_1Q3W = as.integer(NIVO_REGIMEN == "1 mg/kg Q3W").Sanghavi_2020_ipilimumab.R (exponential coefficient 0.0950
on ipilimumab CL).NIVO_3Q2W in the
Sanghavi 2020 final model; both decomposed indicators are 0 for
ipilimumab monotherapy. Other nivolumab regimens (0.3 mg/kg Q3W, 1 mg/kg
Q2W, 3 mg/kg Q3W) were tested but not retained in the final model and
collapse into the reference 0 group.NIVO_REGIMEN (categorical column) – decompose into
NIVO_3Q2W = as.integer(NIVO_REGIMEN == "3 mg/kg Q2W").Sanghavi_2020_ipilimumab.R (exponential coefficient 0.191
on ipilimumab CL).NIVO_1Q3W; same
reference grouping convention.COMBO – used in
Sanghavi_2020_ipilimumab.R. Equivalently derivable from
NIVO_REGIMEN as
COMBO_NIVO = as.integer(NIVO_REGIMEN != "none").Sanghavi_2020_ipilimumab.R (additive effect -0.202 on the
Emax parameter of the time-varying CL function).NIVO_1Q3W / NIVO_3Q2W indicators on baseline
CL: COMBO_NIVO aggregates across all nivolumab regimens and
acts on the time-varying-CL Emax parameter, whereas the per-regimen
indicators act on baseline (time-zero) CL.(BLSTABL / <ref>)^exponent. Reference value observed:
0.352 x 10^9 counts (Wu 2024 Table 3, BCP-ALL median).BLSTABL – used in
Wu_2024_inotuzumab.R.Wu_2024_inotuzumab.R
(power exponent -0.0484 on kdes for BCP-ALL patients only; the effect is
gated off for B-cell NHL by multiplying the exponent by
DIS_BCPALL).BLSTPB (percentage of blasts
in peripheral blood, used by the predecessor Garrett 2019 adult model).
Not applicable for B-cell NHL patients in pooled BCP-ALL + NHL analyses
(Wu 2024 retains the effect only in BCP-ALL patients via the DIS_BCPALL
gate). When supplying BLSTABL for an NHL subject, set the value to the
BCP-ALL reference (0.352) so the gated power term evaluates to 1
numerically. Scope: specific because the covariate is most meaningful in
B-cell-leukemia population PK analyses; promote to general if a second
paper retains it.COMBO (categorical: 0 = single agent, 1 = + rituximab,
2 = + obinutuzumab) – used in Lu_2019_polatuzumab.R.
Decompose: COMBO_RG = as.integer(COMBO == 1 | COMBO == 2).
The Lu 2019 NONMEM separately defines
RTX = as.integer(COMBO == 1) and
GA101 = as.integer(COMBO == 2) and applies effects as
theta^(RTX + GA101); because RTX and GA101 are mutually
exclusive, RTX + GA101 takes values {0, 1} and the effect collapses to
theta^COMBO_RG.Lu_2019_polatuzumab.R
(multiplicative effects on CL_INF = 0.844, kdes = 0.932, FRAC_NS =
0.709).COMBO_R and
COMBO_G separately rather than overloading this
canonical.COMB – used in Hwang_2022_tremelimumab.R
($INPUT NM-TRAN data item; control-stream switch
IF(COMB.EQ.0) selects monotherapy parameters and
IF(COMB.EQ.1) selects combination-therapy parameters).Hwang_2022_tremelimumab.R (selects between monotherapy and
combination-with-durvalumab values of the time-varying-CL Tmax and
lambda parameters).COMBO_NIVO but for
durvalumab rather than nivolumab co-administration. Acts on the
time-varying-CL component (Tmax and lambda); baseline CL is shared
between monotherapy and combination groups in Hwang 2022.exp(theta * (COMBO_LEN_DEX - 1)) so that COMBO_LEN_DEX = 1
yields factor 1 (paper’s reference) and COMBO_LEN_DEX = 0 activates the
effect.LENDEX (1 = with Ld, 0 = without Ld; in Ide 2020
derived from STUDY != 204011 because study 204011 was the
Ld-free elotuzumab-monotherapy cohort) – used in
Ide_2020_elotuzumab.R.COMBO_LD (retired canonical name;
renamed to COMBO_LEN_DEX on 2026-04-27 for clarity).Ide_2020_elotuzumab.R
(multiplicative effects: CLLd = 0.74 on nonspecific CL, encoded as
exp(log(0.74) * (COMBO_LEN_DEX - 1)); KINTLd = 10.1 on the
second-order target-mediated elimination rate from the peripheral
compartment, encoded as
exp(log(10.1) * (COMBO_LEN_DEX - 1))).COMBO_BELAMAF (which pools Ld with bortezomib-dex and
pomalidomide-dex into a single broader “any-combination” belantamab
indicator); COMBO_LEN_DEX is the per-backbone Ld-only flag.
If a future paper distinguishes “Ld-only” from a broader
“any-IMiD-plus-dex” backbone with separate effects, register a parallel
canonical (e.g., COMBO_PD for pomalidomide-dex,
COMBO_VD for bortezomib-dex). Sign of the exponential
coefficient is paper-dependent: Ide 2020 reports
CLLd = 0.74 so the Ld-coadministration arm has 26% lower
nonspecific CL than the Ld-free arm, but KINTLd = 10.1 so
the Ld arm has 10x higher second-order target-mediated elimination –
both are mechanistically interpretable (dexamethasone suppresses
non-specific catabolic clearance; lenalidomide-activated NK cells
increase target-cell-binding-mediated elimination).COMBO (when the source dataset uses a generic
combination flag for belantamab mafodotin pooled regimens) – used in
Papathanasiou_2025_belantamab.R.Papathanasiou_2025_belantamab.R (multiplicative factor
theta = 1.44 on the Imax parameter of the time-varying CL function –
combination therapy increases the steady-state CL reduction from 33.2 %
to 44.0 %).COMBO_BELAMAF_BORDEX, etc.) rather than overloading this
aggregate.KG / 1000 * tumorSize
in the SLD ODE so the column units are (1/day) * 1000 as
published in the source NONMEM run. Document per-model via
covariateData[[KG]]$units.Zecchin_2016_survival.R).KG (NONMEM $INPUT column in
DDMODEL00000218; identical column shipped in the bundle’s
Simulated_OS.csv).Zecchin_2016_survival.R (Zecchin 2016 OS model,
DDMODEL00000218).modellib('Zecchin_2016_tumorovarian') / DDMODEL00000217).
When this OS model is used standalone, the user must supply
KG per subject – typically by first fitting the SLD model
and extracting the per-subject empirical-Bayes posterior. The internal
/1000 scaling is preserved verbatim from the source
$DES block to maintain numerical equivalence with the
published estimates.tumorSize(0) <- IBASE * 1000 in the Zecchin 2016 model:
source convention multiplies by 1000 to convert the internal value to
mm) and to scale the time-varying tumour-size ratio
(mmbas <- IBASE * 1000) inside the OS hazard.IBASE * 1000 = mm so
the column itself is in metres (1 m = 1000 mm). Document per-model via
covariateData[[IBASE]]$units.(tumorSize - mmbas) / mmbas.IBASE (NONMEM $INPUT column in
DDMODEL00000218; identical column shipped in the bundle’s
Simulated_OS.csv with values typically in the 0.04-0.50 m range).Zecchin_2016_survival.R (Zecchin 2016 OS model,
DDMODEL00000218).TUM_SLD column. TUM_SLD carries the
measured baseline tumour size in mm (used to compute the
time-fixed NSLD0 = TUM_SLD / 70 covariate term in the
Zecchin 2016 OS hazard), whereas IBASE carries the
empirical-Bayes fitted baseline from the upstream SLD model
(used to initialise the integrated SLD trajectory and to define the
time-varying TSR(t) reference). The two are correlated but not equal
because the upstream IPP fit smooths measurement noise away from the
observed SLD0. Specific scope because the column is the empirical-Bayes
output of a specific upstream model fit and the internal
*1000 unit-conversion is tied to the source NONMEM coding
convention.NEW_LESION flips
to 1 it stays 1 for subsequent observation times in that subject (a
step-function flag, not a transient pulse).NWLS – prior canonical name (pre-2026-06-19 readability
standardization) and NONMEM $INPUT column in
DDMODEL00000218; the bundle’s Simulated_OS.csv re-labels the same column
NWLSCOV. Downstream consumers should map
NWLSCOV -> NEW_LESION.Zecchin_2016_survival.R (Zecchin 2016 OS model,
DDMODEL00000218; multiplicative effect on the Weibull hazard via
exp(e_nwls_haz * NEW_LESION) with
e_nwls_haz = 1.23 per Output_real_OS.lst FINAL TH5 / Table
2 of Zecchin 2016).NEW_LESION directly (no
time-gating), which is faithful to the simulated dataset shipped in the
DDMORE bundle. The source Output_real_OS.lst (the listing
on the original real dataset) gates the indicator with an additional
TNWLS (lesion-appearance-time) column not shipped in the
bundle’s simulated dataset; the two encodings are functionally
equivalent when the dataset’s NEW_LESION column is
constructed as a 0/1 step that flips at the lesion-appearance time. The
bundle’s simulated dataset uses the simpler step-function form, and that
is the form the nlmixr2lib model expects. Renamed from NWLS
to NEW_LESION on 2026-06-19 per the canonical-register
standardization audit (operator decision: readability over abbreviation,
since NWLS is opaque and not a universally-recognized
clinical-trial term; no DIS_ prefix because this is a RECIST-imaging
event indicator rather than a disease-state indicator).Ig_type – used in Fau_2020_isatuximab.R.
Values 0 / 1 with the same orientation as the canonical (1 = non-IgG
MM).Fau_2020_isatuximab.R
(exponential effect on the steady-state linear CL CLinf with coefficient
-0.751, and on the time-varying-CL half-time KCL with coefficient
-0.931), Xu_2020_daratumumab.R (additive shift
(1 + 0.806 * (1 - MM_NIGG)) on linear CL — Xu 2020
parameterises with non-IgG MM as reference, so an IgG MM patient
receives an 80.6% higher linear CL than a non-IgG MM patient; canonical
column semantics 1 = non-IgG / 0 = IgG are preserved).DIS_SMM = smoldering MM);
applies only after a multiple-myeloma diagnosis is established. Scope:
specific because the comparison is a within-MM stratifier rather than a
cross-population indicator. Reference category at the model level (which
value of MM_NIGG corresponds to TVCL = base) varies between papers: Fau
2020 anchors to 0 (IgG MM) and Xu 2020 anchors to 1 (non-IgG MM); the
canonical column orientation (1 = non-IgG) is fixed across papers and
the per-model covariateData[[MM_NIGG]]$reference_category
field records which anchor each model uses.Mazzocco_2015_temozolomide.R (exponential effect on the
temozolomide tumour-cell-death rate constant gamma:
gamma = gamma0 * exp(beta_p53 * TUM_TP53_MUT) with
beta_p53 = log(0.143 / 0.254) = -0.574; TP53-mutant LGG
tumours are 44% less sensitive to TMZ than TP53-wild-type tumours).SNP_<GENE>_<RSID> family, which
encodes inherited host germline genotypes affecting drug PK;
TUM_TP53_MUT encodes a tumour-cell mutation and only makes
mechanistic sense for drugs whose effect depends on a functional p53
pathway in the tumour. The Mazzocco 2015 cohort assayed p53 status by
IHC overexpression; future extractions that use direct TP53 sequencing
should still record their values under this canonical and document the
assay method in covariateData[[TUM_TP53_MUT]]$notes.
Ratified canonically on 2026-05-17 alongside the Mazzocco 2015
temozolomide extraction.Mazzocco_2015_temozolomide.R (exponential effect on the
damaged-quiescent-to-proliferative repair rate constant
kQpP:
kQpP = kQpP0 * exp(beta_1p19q * TUM_1P19Q_CODEL) with
beta_1p19q = log(0.00807 / 0.00947) = -0.160;
1p/19q-codeleted LGG tumours have 15% lower kQpP than non-codeleted
tumours, consistent with longer reported duration of response in
codeleted patients).CL ~ (1 - theta_PS * WHO_PS) or
CL ~ exp(theta_PS * WHO_PS)) rather than binarizing into
ECOG_GE1 / ECOG_GE2. Time-fixed per subject in
the typical baseline-PS application; document time-varying use in
per-model notes.covariateData[[WHO_PS]]$notes)WHO_PS = 0 (Leger 2004
reference patient implicit; the typical CL formula
(1 - 0.12 * WHO_PS) evaluates to 1 at
WHO_PS = 0).PS – WHO performance status integer; used in
Leger_2004_topotecan.R and historically in Mould 2002 and
Gallo 2000 topotecan popPK papers (cited by Leger 2004) using the same
linear-ordinal form.ECOG – equivalent ECOG performance status integer.PERF – common NONMEM column header for the ordinal
performance-status integer.Leger_2004_topotecan.R
(linear-ordinal effect (1 - 0.12 * WHO_PS) on the
renal-plus-non-renal CL formula
(theta1 + theta2 * CrCl_Lh) * (1 - e_who_ps_cl * WHO_PS);
PS distribution in cohort 0 / 1 / 2 / 3 = 79 / 98 / 11 / 2
patients).ECOG_GE1 /
ECOG_GE2 (binary indicators used when the source paper
collapses the ordinal score). Use WHO_PS when the paper
retains the integer score in a linear / exponential / power covariate
form. If a future paper provides the ordinal integer column but
binarizes the effect downstream, derive
ECOG_GE1 = as.integer(WHO_PS >= 1) and
ECOG_GE2 = as.integer(WHO_PS >= 2); do not duplicate the
same patient-level integer under two columns. WHO and ECOG PS are
operationally identical and cross-walked one-to-one per Oken 1982 (Am J
Clin Oncol 5:649-655); the canonical name WHO_PS is chosen
because the founding example (Leger 2004) names it as WHO performance
status. Scope general because ordinal-linear PS is the dominant
historical NONMEM covariate-model form for oncology popPK papers and is
likely to recur in future extractions.OXGEM = 0; the combination-vs-monotherapy
reference is encoded jointly with OXGEM.GEMOX – used in
Jiang_2008_gemcitabine.R.Jiang_2008_gemcitabine.R (multiplicative factor 0.65 on
apparent dFdU central volume V_C,dFdU/F when GEMOX = 1; Jiang 2008 page
330 covariate equation: 0.65^GEMOX).OXGEM as the second
arm of a 3-level oxaliplatin-coadministration-sequence categorical
decomposition (gem-alone / gem-then-ox / ox-then-gem). The two
indicators are mutually exclusive: GEMOX + OXGEM <= 1.
Follows the NIVO_1Q3W / NIVO_3Q2W
decomposition pattern. Scope: specific because the sequence-specific
effect is tied to Jiang’s gemcitabine + oxaliplatin study design;
promote to general if a second paper retains the same pair of
indicators.GEMOX = 0.OXGEM – used in
Jiang_2008_gemcitabine.R.Jiang_2008_gemcitabine.R (multiplicative factor 0.54 on
apparent dFdU central volume V_C,dFdU/F when OXGEM = 1; Jiang 2008 page
330 covariate equation: 0.54^OXGEM).GEMOX (see that
entry for the joint encoding). The Jiang 2008 cohort distribution is
gem-alone n=31, gem-then-ox n=38, ox-then-gem n=25; the larger
V_C,dFdU/F reduction observed for ox-then-gem (factor 0.54) vs
gem-then-ox (factor 0.65) is hypothesised by the authors to reflect
order-dependent tissue-uptake effects of dFdU.albumin_observed / ULN_albumin).(ALBR / <ref>)^exponent. Reference 0.78 used in Xu
2019 (corresponds to a median serum albumin of 38 g/L at a typical ULN
of ~48.7 g/L).Xu_2019_sarilumab.R.covariateData[[SCORE_CALPRO]]$units.(SCORE_CALPRO / ref)^exponent. Reference 700 mg/kg used in
Rosario 2015 (overall population median).CALPRO – prior canonical name (pre-2026-06-19 SCORE_
family standardization).Rosario_2015_vedolizumab.R (reference 700 mg/kg; exponent
+0.0310 on linear clearance CLL).covariateData[[SCORE_CALPRO]]$notes.(SCORE_CDAI / ref)^exponent. Reference 300 used in Rosario
2015 (typical moderate-CD score).CDAI – prior canonical name (pre-2026-06-19 SCORE_
family standardization).Rosario_2015_vedolizumab.R (reference 300; exponent -0.0515
on CLL gated by IBD_CD so the effect applies only to CD
patients).SCORE_PMAYO in pooled UC+CD populations: each patient has
exactly one disease-activity score (SCORE_CDAI for CD, partial Mayo for
UC). Gate via the IBD_CD indicator when pooling.IBD_CD indicator when pooling UC+CD.(SCORE_PMAYO / ref)^exponent. Reference 6 used in Rosario
2015 (typical moderate-UC score).PMAYO – prior canonical name (pre-2026-06-19 SCORE_
family standardization).Rosario_2015_vedolizumab.R (reference 6; exponent +0.0408
on CLL gated by (1 - IBD_CD) so the effect applies only to
UC patients).SCORE_CDAI in pooled UC+CD populations.SCORE_PMAYO) excludes it. Time-fixed per subject.covariateData[[SCORE_MAYO_E]]$reference_category.MAYO_E – prior canonical name (pre-2026-06-19
standardization audit).MPRE – used in Faelens_2021_infliximab.R
(NONMEM column for “Mayo endoscopic score pre-induction”). The Faelens
2021 dataset additionally codes a “missing” sentinel
MPRE = -99; treat this as out-of-domain when applying the
model and document per-model.Faelens_2021_infliximab.R (categorical effect on KE:
separate typical KE for Mayo 1 / Mayo 2 / Mayo 3; reference category
Mayo 2).SCORE_PMAYO, 0-9). The endoscopic
subscore alone is the core inclusion criterion in many UC
induction-therapy popPK datasets (typically Mayo 2 or 3 =
moderate-to-severe disease). Mutually compatible with
SCORE_PMAYO in datasets that report both.Aguiar_2021_ustekinumab.R (Aguiar 2021 Table 3; covariate
on baseline fecal calprotectin FC0: 213 mg/kg with ulcers vs 102 mg/kg
without).DIAGNOSIS, DX (categorical
"UC"/"CD") – derive
IBD_CD = as.integer(DX == "CD").Rosario_2015_vedolizumab.R (two typical-CLL switch between
UC and CD; gates SCORE_PMAYO and SCORE_CDAI
effects; multiplicative +1% effect on Vc).covariateData[[CONMED_AD]]$notes.1 for
CONMED_AD = 1 and 1 + e_sl_conmed_ad_off for
CONMED_AD = 0. The non-standard “most-common-as-reference”
convention is preserved from the source for traceability; future
Alzheimer’s models that adopt the more common “off-treatment as
reference” convention should register a successor canonical
(CONMED_AD_TREATED) rather than overloading this one.COMED2 – used in Conrado_2014_alzheimer.R
(DDMORE Foundation Model Repository entry DDMODEL00000290). The
2 suffix in the source distinguishes this binary from
upstream COMED and PRIMCOMED columns
(free-text concomitant-medication entries) in the same NONMEM input
dataset; the binary COMED2 flag is what enters the
model.Conrado_2014_alzheimer.R (multiplicative factor on the
typical-value disease-progression slope:
slope_factor = 1 + (1 - CONMED_AD) * e_sl_conmed_ad_off
with e_sl_conmed_ad_off = -0.302, i.e. ~30% smaller
progression slope for the off-treatment reference cohort)..mod does not
specify the symptomatic-medication class beyond the binary flag; the
Conrado 2014 publication context (CAMD ADAS-Cog disease-progression
dataset, 2014) makes cholinesterase-inhibitor / memantine the dominant
interpretation. Treating CONMED_AD = 1 as the reference
category is unusual relative to the rest of the CONMED_*
family (CONMED_PARA, CONMED_NSAID,
CONMED_AZA, etc.) which all use 0 = not-on as reference;
the inversion is preserved here only because the source paper’s
coefficient was estimated with the “most common = 1” convention.
Ratified canonically on 2026-05-06 alongside the Conrado 2014 DDMORE
extraction.CO_AED – used in Yukawa_1990_phenytoin.R
(paper’s CO indicator inverted: source CO = 1
if PHT alone, theta_co otherwise; canonical
CONMED_AED = 1 - CO_indicator so 0 is the PHT-monotherapy
reference).Yukawa_1990_phenytoin.R (multiplicative
^CONMED_AED factor on Vmax (1.08) and Km (1.32) for chronic
phenytoin patients on at least one of phenobarbital, carbamazepine,
valproate, primidone, clonazepam, sultiame, ethotoin, ethosuximide,
acetazolamide, or diazepam),
Yukawa_2002_clonazepam_pediatric.R (paired with
CONMED_AED_GE2 to recover the paper’s 3-tier
drug-interaction factor on clonazepam CL/F).covariateData[[CONMED_AED]]$notes. Distinct from
drug-specific concomitant-AED indicators (e.g., a future
CONMED_PB for concomitant phenobarbital alone) which would
warrant separate canonicals when a paper distinguishes effects by AED
class. When a paper distinguishes AED-count tiers (monotherapy vs +1 AED
vs +>=2 AEDs) rather than the pooled mono-vs-anything contrast, pair
CONMED_AED with the sibling [[CONMED_AED_GE2]] binary; the
>=2-AEDs tier is CONMED_AED = 1 AND
CONMED_AED_GE2 = 1, and data assemblers must enforce that
CONMED_AED_GE2 = 1 implies CONMED_AED = 1.
Follows the CONMED_* family pattern
(CONMED_AZA, CONMED_NSAID, etc.). Ratified
canonically on 2026-05-10 alongside the Yukawa 1990 phenytoin
extraction.CONMED_AED_GE2 = 0 with
CONMED_AED = 1) is “1 concomitant AED”; the absolute
reference (CONMED_AED_GE2 = 0 with
CONMED_AED = 0) is “monotherapy”.CONMED_AED_GE2 – used in
Yukawa_2002_clonazepam_pediatric.R (paper’s Polytherapy (B)
bucket per Table I and Table II groupings: any record with >=2 AEDs
alongside clonazepam).Yukawa_2002_clonazepam_pediatric.R (paired with
[[CONMED_AED]] inside model() to recover the paper’s 3-tier
drug-interaction factor: monotherapy DIF = 1, +1 AED DIF = 1.18, +>=2
AEDs DIF = 2.12 * TBW^(-0.119)).CONMED_AED_GE2 = 1 MUST imply CONMED_AED = 1.
A record asserting CONMED_AED_GE2 = 1 and
CONMED_AED = 0 is malformed and the model() body will
silently treat it as the >=2 tier (since the indicator decomposition
uses CONMED_AED * (1 - CONMED_AED_GE2) for the +1 tier and
CONMED_AED_GE2 for the >=2 tier); data assemblers should
validate the pair before passing to rxSolve(). Distinct
from drug-specific CONMED_<drug> indicators
(CONMED_CBZ, CONMED_VPA,
CONMED_PB) which capture a specific AED rather than a count
tier; when a future paper distinguishes BOTH the count AND the specific
drugs, the two families coexist. Ratified canonically on 2026-06-21
alongside the Yukawa 2002 clonazepam pediatric extraction.ABIRATER – used in
Yoshida_2021_ipatasertib.R (Yoshida 2021 NONMEM control
stream variable name; same per-subject 0/1 orientation).Yoshida_2021_ipatasertib.R (linear-additive effects:
-18.5% on apparent ipatasertib parent CL/F when abiraterone
is present (Table 3 theta_CLI,Abi = -0.185); on the apparent M1
bioavailability the abiraterone effect is +61.5% but ONLY
at multiple-dose state (Table 4 theta_FM1,Abi = 0.615 applied as
(1 + 0.615 * MULTI_DOSE * CONMED_ABI) so that the same
single-dose effect was zero per the source NONMEM control stream)).CONMED_AZOLE (CYP3A4
inhibitor used for fungal indications) and from
CONMED_STEROID (the prednisone / prednisolone component
that is always co-administered with abiraterone in mCRPC); when a model
explicitly separates abiraterone vs the steroid backbone, use both
covariates. Yoshida 2021 also reports that the precise mechanism of the
abiraterone-on-ipatasertib effect is unknown (the standard CYP3A4 /
CYP2C8 / CYP2D6 routes were ruled out by in-vitro and DDI studies), so
the indicator captures the empirical exposure shift rather than a
defined enzyme-inhibition mechanism. Ratified canonically on 2026-05-30
alongside the Yoshida 2021 ipatasertib extraction.CM1 – used in Xia_2024_warfarin.R (Xia
2024 Figure 1 / Table 2 covariate-screening variable labelling:
CM1 = 1 indicates combined amiodarone, 0 no
combination).Xia_2024_warfarin.R
(piecewise multiplicative effect on warfarin EC50:
ec50 *= (1 + e_amio_ec50 * CONMED_AMIO) with
e_amio_ec50 = -0.602, i.e. amiodarone reduces EC50 by ~60%
in the Han Chinese cohort).covariateData[[CONMED_AMIO]]$notes. Ratified canonically on
2026-05-16 alongside the Xia 2024 warfarin extraction.AMINO – used in
Rosario_2015_vedolizumab.R.Rosario_2015_vedolizumab.R (power-form on CLL:
CLL * 1.02^CONMED_AMINO).CONMED_AMINO rather than CONMED_5ASA unless
the source paper explicitly restricts the indicator to 5-ASA
monotherapy.DOX – used in Zhou_2025_brentuximab.R (the
NONMEM dataset uses the doxorubicin-administration flag as the
AVD-coadministration indicator because doxorubicin is given on the same
days as the other AVD agents in this regimen).Zhou_2025_brentuximab.R (power-form effect on ADC
clearance: CL * 2.12^CONMED_AVD – ADC clearance is
~2.1-fold higher under A+AVD vs single-agent BV).CONMED_CHEMO
(which is nivolumab + platinum-based chemotherapy). The A+AVD regimen is
the standard chemotherapy backbone for frontline classical Hodgkin
lymphoma; promote to general scope if a second BV paper reports the same
A+AVD covariate effect with a comparable encoding.AZA – used in
Rosario_2015_vedolizumab.R.Rosario_2015_vedolizumab.R (power-form on CLL:
CLL * 0.998^CONMED_AZA; effect ~= null).AZOLE – used in
Kirubakaran_2022_tacrolimus.R.Kirubakaran_2022_tacrolimus.R (state-dependent typical
CL/F: 21.1 L/h without azole and 4.2 L/h with azole, an 80% reduction;
also a state-dependent BSV magnitude on CL/F: 61% CV without azole vs
89.5% CV with azole).covariateData[[CONMED_AZOLE]]$notes must document (1) which
azoles are pooled into the indicator, (2) any post-cessation lag
(Kirubakaran 2022 carries CONMED_AZOLE = 1 for one week
after azole discontinuation to allow tacrolimus apparent clearance to
stabilize given itraconazole’s long half-life), and (3) whether the
indicator is a baseline-only proxy or a true time-varying flag.CBZ – used in
Schoemaker_2017_brivaracetam.R (paper covariate
CBZ for carbamazepine coadministration) and
Hashimoto_1994_zonisamide.R (paper covariate
CBZ for carbamazepine coadministration on zonisamide
Vmax).Schoemaker_2017_brivaracetam.R (multiplicative effect on
apparent oral clearance: cl *= (1 + 0.479 * CONMED_CBZ);
+47.9% relative to no-CBZ reference, corresponding to ~32% lower
brivaracetam exposure, Schoemaker 2017 Table 1),
Hashimoto_1994_zonisamide.R (multiplicative power-form
effect on Vmax of Michaelis-Menten zonisamide elimination:
vmax *= 1.13^CONMED_CBZ; +13% relative to no-CBZ reference,
Hashimoto 1994 Table II theta2).CONMED_CBZ, [[CONMED_VPA]] rather than collapsing into the
class-level indicator. Hashimoto 1994 (zonisamide) tested phenytoin and
valproate coadministration on zonisamide PK but found no significant
effect (page 325, “data not shown”) and retained only CBZ in the final
model, consistent with the AED-specific decomposition rationale.CsA – used in de Winter 2009 (binary indicator on the
MPAG-to-gallbladder transport rate constant k57 to encode cyclosporine
inhibition of MRP2-mediated biliary efflux); the canonical column is
CONMED_CSA with the same value semantics.deWinter_2009_mycophenolic_acid.R (multiplicative
power-form effect on the fMPAG-to-gallbladder rate constant k57:
k57 <- exp(lk57) * e_csa_k57^CONMED_CSA with
e_csa_k57 = 0.002, so cyclosporine cotreatment reduces k57
from 0.0796 1/h to 0.000159 1/h, suppressing EHC by ~99.8% relative to
the tacrolimus reference; Table 2 / Eq. 9 of de Winter 2009).CONMED_CSA = 1 and CONMED_TAC (if registered
later for tacrolimus-as-perpetrator effects) would be mutually exclusive
at the patient level. Cyclosporine and tacrolimus differ in their effect
on MPA pharmacokinetics: cyclosporine inhibits MRP2 (decreasing EHC of
MPAG -> MPA) while tacrolimus does not, so co-administered MPA
exposure is typically lower under cyclosporine than under tacrolimus for
the same MMF dose. Data assemblers can derive
CONMED_CSA = as.integer(cni_drug == "cyclosporine") from a
raw CNI_DRUG text column.CHEMO – used in
Zhang_2019_nivolumab.R.MONOTR – used in
Kuchimanchi_2024_dostarlimab.R (the paper’s
structural-equation indicator for monotherapy; the canonical column
carries the inverse value, i.e. CONMED_CHEMO = 1 - MONOTR,
so the canonical column is 1 when the patient is on
combo-chemotherapy).Zhang_2019_nivolumab.R
(exponential effect on baseline CL: exp(-0.104) ~= 0.90
fold, i.e. ~9.7% lower CL relative to monotherapy),
Kuchimanchi_2024_dostarlimab.R (multiplicative effect on
baseline CL: 1 - 0.0779 = 0.922, i.e. 7.79% lower CL on
dostarlimab + carboplatin/paclitaxel relative to dostarlimab
monotherapy).exp(theta * CONMED_CHEMO) (exponential) and
Kuchimanchi 2024 uses (1 + theta * CONMED_CHEMO)
(multiplicative); these are different parameterisations of the same
underlying study-design indicator and the canonical column meaning is
unchanged. Document the per-model functional form in
covariateData[[CONMED_CHEMO]]$notes.CONMED_GLASDEGIB = 1 in the only on-disk source (Lin 2020),
where the decitabine arm carries both indicators = 1 and the LDAC arms
have CONMED_DECITABINE = 0).Lin_2020_glasdegib_decitabine.R (BRIGHT AML 1003 Phase 1b
Arm B: glasdegib + decitabine, n = 7).Lin_2020_glasdegib_decitabine.R (paired with
CONMED_GLASDEGIB to identify the three trial arms;
multiplicative hazard reduction
1 - 0.618 * (CONMED_GLASDEGIB * CONMED_DECITABINE) on the
exponential overall-survival hazard).CONMED_<INN>
family.CONMED_DOXORUBICIN and
CONMED_EPIRUBICIN are used together as a pair of
complementary indicators (as in de Vries Schultink 2018), both being 0
means no anthracycline.ANTH_TYPE = 'doxorubicin' – used in
deVriesSchultink_2018_anthracycline_troponinT.R (the source
paper records anthracycline as a categorical with levels {‘doxorubicin’,
‘epirubicin’}; the canonical register splits this into a pair of binary
indicators following the CONMED_<drug> precedent
rather than the categorical ANTH_TYPE).deVriesSchultink_2018_anthracycline_troponinT.R (used
jointly with CONMED_EPIRUBICIN to switch the K-PD SLOPE
parameter for troponin T between doxorubicin (reference) and epirubicin;
epirubicin = 0.524-fold of doxorubicin effect, per de Vries Schultink
2018 Table 2).CONMED_PLDH (PEGylated liposomal doxorubicin)
and from CONMED_AVD (the doxorubicin component of the AVD
backbone in Hodgkin lymphoma); use those canonicals for
product-formulation or regimen-backbone semantics. When the source
dataset’s anthracycline column is categorical (ANTH_TYPE),
derive the canonical indicators as
CONMED_DOXORUBICIN = (ANTH_TYPE == 'doxorubicin'),
CONMED_EPIRUBICIN = (ANTH_TYPE == 'epirubicin').CONMED_EPIRUBICIN
with CONMED_DOXORUBICIN captures the
cardiotoxicity-relevant choice within the anthracycline class;
epirubicin is less cardiotoxic per mg than doxorubicin (lifetime
cumulative dose threshold approximately 950 mg/m^2 vs 550 mg/m^2 for
doxorubicin).CONMED_DOXORUBICIN and
CONMED_EPIRUBICIN are used together (as in de Vries
Schultink 2018), both being 0 means no anthracycline.ANTH_TYPE = 'epirubicin' – used in
deVriesSchultink_2018_anthracycline_troponinT.R. See the
matching entry for CONMED_DOXORUBICIN.deVriesSchultink_2018_anthracycline_troponinT.R (used
jointly with CONMED_DOXORUBICIN to switch the K-PD SLOPE
parameter; epirubicin = 0.524-fold of the doxorubicin reference effect
on troponin T).CONMED_DOXORUBICIN and
CONMED_EPIRUBICIN) are not mutually exclusive in principle
(a subject receiving both anthracyclines could have both set to 1), but
in practice the de Vries Schultink 2018 cohort assigned exactly one
anthracycline per subject.EFV – used in Tikiso_2021_abacavir.R (the
dataset’s paper-defined indicator, 1 = on EFV-based ART, 0 = on standard
LPV/r 4:1).Tikiso_2021_abacavir.R
(multiplicative effect on apparent oral clearance:
cl *= (1 + 0.120 * CONMED_EFV); +12.0% relative to the
LPV/r 4:1 reference in HIV-infected children on combination ART),
Svensson_2013_bedaquiline.R (multiplicative factor on
apparent CL_BDQ and CL_M2:
cl_eff = cl_base * 2.07^CONMED_EFV and on apparent CL_M3:
cl_eff = cl_base * 1.12^CONMED_EFV; healthy-volunteer DDI
study with no-co-medication reference and 1-week onset lag from start of
600 mg once-nightly EFV), Hoglund_2015_lumefantrine.R
(multiplicative linear-deviation effect on lumefantrine apparent CL/F:
cl *= (1 + 0.726 * CONMED_EFV); +72.6% relative to the
no-ART reference in HIV-infected Ugandan adults, Hoglund 2015 Table 2
attributing the effect to CYP3A4 induction),
Hoglund_2015_artemether.R (multiplicative linear-deviation
effect on artemether relative bioavailability F:
fdepot_typ *= (1 + (-0.715) * CONMED_EFV); -71.5% relative
to the no-ART reference, Hoglund 2015 Table 3),
Kay_2020_lumefantrine.R (TWO independent linear-deviation
effects, on lumefantrine apparent CL/F:
cl *= (1 + 0.982 * CONMED_EFV) (+98.2%, CYP3A4 induction)
and on the first-order absorption rate constant:
ka *= (1 + 0.484 * CONMED_EFV) (+48.4%); relative to the
no-ART reference in HIV-infected Ugandan children with malaria (ASTMH
2020 poster 2167 Table 1)), Kay_2022_lumefantrine.R (TWO
independent linear-deviation effects: on apparent oral lumefantrine CL/F
cl *= (1 + 0.982 * CONMED_EFV) (+98.2 %, CYP3A4 induction)
and on the first-order absorption rate constant
ka *= (1 + 0.484 * CONMED_EFV) (+48.4 %); HIV-uninfected
children are the reference (CONMED_EFV = 0) in a pediatric Ugandan
malaria cohort on six-dose Coartem Dispersible, Kay 2022 Table 2 theta_7
and theta_10).MED (with the value convention inverted: source codes 1
= absence of EIAEDs and 0 = presence; canonical inverts this so 1 =
presence and 0 = absence) – used in
Rodrigues_2017_oxcarbazepine.R.Rodrigues_2017_oxcarbazepine.R (exponential effect on MHD
apparent clearance:
cl_mhd *= exp(e_eiaed_cl_mhd * (1 - CONMED_EIAED)); CL_MHD
is 29.3% higher with EIAEDs vs without, encoded as +0.257 on the
absence-indicator in the source paper).covariateData[[CONMED_EIAED]]$notes should list the
specific EIAEDs counted as “EIAED = 1” since inclusion criteria vary
across antiepileptic-drug studies. Rodrigues 2017 counts carbamazepine,
phenobarbital, and phenytoin as EIAEDs; other AEDs in the dataset
(vigabatrin, clobazam, valproic acid, clonazepam, lamotrigine, diazepam,
ethosuccimide, progabide) are not. The source paper uses an “absence of
EIAED” indicator (MED = 1 if no EIAED, MED = 0
if EIAED present); the canonical column inverts this so that the 1 group
is “on EIAED”, matching the convention for other CONMED_*
indicators.CONMED_PDE5I and CONMED_ERA_PDE5I).PAHCOMED category 1 (“ERA only”) – decomposed from the
categorical PAHCOMED source column with levels {naive, ERA,
PDE5, ERA+PDE5} in Krause_2017_selexipag.R.Krause_2017_selexipag.R (multiplicative effect on the
ACT-333679 elimination rate constant:
km *= (1 + 0.15 * CONMED_ERA); +15% relative to the
PAH-comedication-naive reference, Krause 2017 Table 1).CONMED_PDE5I
and CONMED_ERA_PDE5I to decompose a four-level
PAH-comedication categorical (naive / ERA-only / PDE5-only /
ERA-and-PDE5) into three orthogonal mutually-exclusive binary
indicators, with the PAH-comedication-naive group as the reference (all
three indicators = 0). The four-level categorical encoding preserves the
source paper’s stratum-specific coefficient set (Krause 2017 reports a
separate categorical b-coefficient for each non-reference stratum rather
than independent class-level effects). Specific scope because the
indicator’s semantics are tied to the GRIPHON study’s PAH-comedication
taxonomy.CONMED_ERA and CONMED_PDE5I).PAHCOMED category 3 (“ERA and PDE5 inh.”) – decomposed
from the categorical PAHCOMED source column with levels
{naive, ERA, PDE5, ERA+PDE5} in
Krause_2017_selexipag.R.Krause_2017_selexipag.R (multiplicative effect on the
ACT-333679 elimination rate constant:
km *= (1 + 0.37 * CONMED_ERA_PDE5I); +37% relative to the
PAH-comedication-naive reference, Krause 2017 Table 1; the combined
stratum has its own categorical coefficient distinct from the sum of the
ERA-only and PDE5I-only effects).CONMED_ERA
and CONMED_PDE5I to decompose a four-level PAH-comedication
categorical (naive / ERA-only / PDE5-only / ERA-and-PDE5) into three
orthogonal mutually-exclusive binary indicators. The standalone
combined-stratum coefficient (rather than a product of class-level
indicator effects) preserves Krause 2017’s full categorical-effect
parameterisation. Specific scope because the indicator’s semantics are
tied to the GRIPHON study’s PAH-comedication taxonomy.Kuchimanchi_2018_evolocumab.R (multiplicative effect 1.20
on Vmax: Vmax * 1.20^CONMED_EZE; labeled “Statin +
ezetimibe exponent” in Kuchimanchi 2018 Table 3 because ~99% of
ezetimibe users in the dataset were also on a conmed_statin, so the
effect effectively captures combination therapy),
Kuchimanchi_2018_evolocumab_ldlc.R (same PK-layer effect
plus a multiplicative exponent 0.768 on baseline LDL-C in the
exposure-response layer, Table 4),
Kakara_2014_atorvastatin.R,
Kakara_2014_pitavastatin.R,
Kakara_2014_rosuvastatin.R (additive +0.109 contribution to
the indirect-response Imax inhibition fraction INH on the LDL-C
synthesis rate Kin:
INH = Imax * DOSE / (ID50 + DOSE) + 0.109 * CONMED_EZE;
Kakara 2014 Table 2 INH_EZT).Lin_2020_glasdegib_treatment.R the reference is the LDAC
alone arm (BRIGHT AML 1003 Phase 2). In
Lin_2020_glasdegib_decitabine.R the reference is also LDAC
alone, paired with CONMED_DECITABINE to identify the
three-arm structure of the exploratory pooled analysis.Lin_2020_glasdegib_treatment.R and
Lin_2020_glasdegib_decitabine.R (BRIGHT AML 1003,
NCT01546038).Lin_2020_glasdegib_treatment.R (binary indicator on the
exponential overall-survival hazard; the published equation uses the
complementary indicator LDAC_alone = 1 - CONMED_GLASDEGIB
so the model file derives
ldac_alone <- 1 - CONMED_GLASDEGIB inside
model() to preserve the paper’s parameter values verbatim),
Lin_2020_glasdegib_decitabine.R (paired with
CONMED_DECITABINE in the three-arm exploratory
analysis).CONMED_<INN> family. Ratified canonically on
2026-06-24 alongside the Lin 2020 BRIGHT AML 1003 overall-survival
extraction. The Lin 2020 treatment-response equation uses LDAC_alone as
the binary covariate (1 = LDAC alone, 0 = glasdegib + LDAC) with
parameter theta_ldac_alone = 1.376; the canonical-column
form (1 = glasdegib + LDAC) preserves the published value via the
in-model 1 - CONMED_GLASDEGIB mapping rather than refitting
the parameter against a re-encoded covariate.H2 – used in Goel_2016_Sonidegib.R (Goel
2016 dataset; defined as “significant” CONMED_H2RA use, i.e. duration of
CONMED_H2RA use >= 80% of the PK assessment phase).Goel_2016_Sonidegib.R
(multiplicative effect on F: 0.996^CONMED_H2RA, no
clinically meaningful effect; reported alongside CONMED_PPI
for completeness).covariateData[[CONMED_H2RA]]$notes must document the
operational definition (Goel 2016: >= 80% of PK assessment phase).
Distinct from CONMED_PPI.IFNB1A – used in
Savic_2017_cladribine.R.Savic_2017_cladribine.R (multiplicative effect on
cladribine non-renal clearance:
cl_nonrenal *= (1 + e_ifn_clnr * CONMED_IFNB1A) with
e_ifn_clnr = 0.21, i.e. a 21% increase in non-renal CL when
coadministered with IFN beta-1a).CONMED_IFNB1B,
CONMED_IFNALPHA) if a different interferon species is
intended.IFNa – used in Han_2016_bevacizumab.R (Han
2016 Table 3 categorical indicator on bevacizumab CL; the 102
IFN-alpha-coadministered patients were enrolled in renal cell carcinoma
study BO17705 per Han 2016 Table 1 and Table 2 row ‘Concomitant
treatment’).Han_2016_bevacizumab.R
(multiplicative effect on bevacizumab CL:
cl *= exp(e_ifna_lcl * CONMED_IFNALPHA) with
e_ifna_lcl = log(0.844) = -0.170, i.e. CL is 15.6% lower
when IFN-alpha is coadministered).CONMED_IFNB1A (interferon beta-1a) under the
CONMED_<INN> family; the existing CONMED_IFNB1A entry
explicitly anticipated this CONMED_IFNALPHA name for the
alpha-interferon species. Per-paper alpha-subtype detail (IFN-alpha-2a
vs -2b vs pegylated) is documented in
covariateData[[CONMED_IFNALPHA]]$notes where available;
pooled here because the source paper (Han 2016) does not separately
enumerate subtypes. Distinct from
CONMED_IFNB1A/CONMED_IFNB1B (beta-interferon
species). Ratified canonically on 2026-06-25 alongside the Han 2016
bevacizumab extraction.IMM – used in Frymoyer_2017_infliximab.R
(pooled purine-analogue-or-MTX indicator per Table 1 footnote of
Frymoyer 2017: “Concomitant immunomodulation refers to purine-analogue
or methotrexate.”).Frymoyer_2017_infliximab.R (power-of-coefficient effect on
CL: CL_typ * 0.863^CONMED_IMMUNOMOD, i.e. -13.7% when on an
immunomodulator).CONMED_AZA, CONMED_MP, and
CONMED_MTX, which model the effect of each immunomodulator
separately. Use CONMED_IMMUNOMOD only when the source paper
itself pools the three under a single binary; use the per-drug
indicators when the source paper estimates separate effects. Future IBD
popPK extractions that pool thiopurines + MTX into a single indicator
should reuse this canonical and extend the example list. Ratified
canonically on 2026-05-20 alongside the Frymoyer 2017 infliximab
pediatric Crohn’s disease extraction.IPI1Q6W – used in
Zhang_2019_nivolumab.R.Zhang_2019_nivolumab.R
(exponential effect on baseline CL: exp(0.159) ~= 1.17 fold
increase relative to monotherapy).CONMED_IPI_3Q3W.
See the CONMED_IPI_3Q3W note for how the other ipilimumab schedules
collapse into the reference group.IPI3Q3W – used in
Zhang_2019_nivolumab.R.Zhang_2019_nivolumab.R
(exponential effect on baseline CL: exp(0.227) ~= 1.25 fold
increase relative to monotherapy).CONMED_IPI_1Q6W;
both indicators can coexist in one population, but a single subject has
at most one set to 1 in the Zhang 2019 cohort. The remaining ipilimumab
schedules (1 mg/kg q3w x 4 induction, 1 mg/kg q12w) had no statistically
significant effect on nivolumab CL and were therefore folded into the
reference (0) group along with monotherapy, leaving only IPI3Q3W and
IPI1Q6W as named non-reference indicators.IPICO – used in
Zhang_2019_nivolumab.R.Zhang_2019_nivolumab.R
(additive effect on the time-varying-CL Emax parameter: Emax += -0.0668
when CONMED_IPI_ANY = 1).covariateData[[CONMED_METFORMIN]]$notes.Retlich_2015_linagliptin.R (1 = study 4 add-on-to-metformin
cohort; multiplicative effect on linagliptin relative bioavailability F:
+69% F for metformin co-administration vs the monotherapy reference; the
effect is attributed to a metformin – linagliptin drug-drug interaction
consistent with a separately published DDI study, Graefe-Mody
2009).CONMED_*
concomitant-medication pattern (AZA / MP / MTX / AMINO / NSAID / PARA /
AD / RITUX / AED / CHEMO / EIAED / EFV / AZOLE). Metformin is a
widely-co-prescribed first-line T2DM oral antidiabetic; future
T2DM-popPK / -DDI extractions should reuse this canonical. Ratified
canonically alongside the Retlich 2015 linagliptin extraction.MP – used in
Rosario_2015_vedolizumab.R.Rosario_2015_vedolizumab.R (power-form on CLL:
CLL * 1.04^CONMED_MP).CONMED_AZA for a given subject.SU – used in Baron_2016_empagliflozin.R
(Baron 2016 Methods and Table S2 ‘background sulfonylurea’ cohort;
multiplicative effects on the model-predicted baseline FPG and on Gmax
in Table S3).Baron_2016_empagliflozin.R (multiplicative effect on BFPG:
bfpg_su^CONMED_SULFONYLUREA = 1.01^...; on Gmax:
gmax_su^CONMED_SULFONYLUREA = 1.27^... – sulfonylurea
co-treatment increases empagliflozin’s maximal FPG-lowering effect,
attributed by the authors to possible improvement in beta-cell
function).CONMED_AZOLE, CONMED_NSAID,
CONMED_STATIN, CONMED_AED, etc. Future T2DM
popPK / PD extractions that need to distinguish individual sulfonylureas
(e.g., for a DDI sub-analysis on a CYP2C9-metabolized member) should
register a per-INN sibling canonical (CONMED_GLYBURIDE,
CONMED_GLIPIZIDE, …) and keep
CONMED_SULFONYLUREA for the class-pooled summary. Ratified
canonically on 2026-06-24 alongside the Baron 2016 empagliflozin
extraction.PIO – used in Baron_2016_empagliflozin.R
(Baron 2016 Methods and Study 7 background pioglitazone cohort;
multiplicative effects on the model-predicted baseline FPG and on Gmax
in Table S3).Baron_2016_empagliflozin.R (multiplicative effect on BFPG:
bfpg_pio^CONMED_PIOGLITAZONE = 0.999^...; on Gmax:
gmax_pio^CONMED_PIOGLITAZONE = 1.02^... – both effects are
not statistically significant in the source analysis but are retained
per the full-covariate-modelling approach).CONMED_<INN> pattern (CONMED_METFORMIN, CONMED_LPV,
CONMED_ATAZANAVIR, CONMED_RIF, CONMED_PROBENECID, etc.). Distinct from a
putative future CONMED_TZD (class indicator) – the
thiazolidinedione class additionally contains rosiglitazone (largely
withdrawn) and lobeglitazone (Korean market); pool a TZD class indicator
only if a future paper aggregates them. Ratified canonically on
2026-06-24 alongside the Baron 2016 empagliflozin extraction.MTX – used in
Rosario_2015_vedolizumab.R.Rosario_2015_vedolizumab.R (power-form on CLL:
CLL * 0.983^CONMED_MTX).NSAID – used in Li_2019_abatacept.R.Li_2019_abatacept.R
(exponential effect on CL: CL * exp(0.0640 * CONMED_NSAID);
~6.6% higher CL, not clinically relevant per Li 2019).CONMED_*
concomitant-medication pattern established for IBD models (AZA / MP /
MTX / AMINO).N – in-equation indicator used by Jullien 2006 (final
covariate submodel: CL/F = ... * 1.34^N with N = 1 if
nevirapine combined with lopinavir).NVP – standard 3-letter HIV abbreviation in NONMEM
control streams from pediatric-ART cohorts; same orientation, no value
transformation.Jullien_2006_lopinavir.R (exponential effect on CL/F:
cl *= exp(log(1.34) * CONMED_NVP); +34% CL/F when
nevirapine is coadministered), Hoglund_2015_lumefantrine.R
(multiplicative linear-deviation effect on lumefantrine relative
bioavailability F:
fdepot_typ *= (1 + (-0.248) * CONMED_NVP); -24.8% relative
to the no-ART reference in HIV-infected Ugandan adults, Hoglund 2015
Table 2), Hoglund_2015_artemether.R (two independent
linear-deviation effects: on artemether relative bioavailability F
fdepot_typ *= (1 + (-0.663) * CONMED_NVP) (-66.3 %) and on
apparent dihydroartemisinin CL/F
cl_dha *= (1 + (-0.445) * CONMED_NVP) (-44.5 %); Hoglund
2015 Table 3), Svensson_2014_bedaquiline_nvp.R (power-form
effects on apparent CL of bedaquiline and its M2 metabolite at full
induction: cl *= 0.915 ^ CONMED_NVP and
cl_m2 *= 1.05 ^ CONMED_NVP; Svensson 2014 Supplementary
Table S1b ‘EFF NVP on BDQ CL = 0.915’ and ‘EFF NVP on M2 CL = 1.05’),
Kay_2022_lumefantrine.R (TWO independent linear-deviation
effects: on apparent oral lumefantrine CL/F
cl *= (1 + 0.0191 * CONMED_NVP) (+1.91 %; not statistically
significant, 95% CI -0.324 to 0.362) and on the first-order absorption
rate constant ka *= (1 + (-0.0589) * CONMED_NVP) (-5.89 %;
not statistically significant, 95% CI -0.207 to 0.0891); HIV-uninfected
children are the reference (CONMED_NVP = 0) in a pediatric Ugandan
malaria cohort, Kay 2022 Table 2 theta_9 and theta_12),
Kay_2020_lumefantrine.R (TWO independent linear-deviation
effects, on lumefantrine apparent CL/F:
cl *= (1 + 0.0191 * CONMED_NVP) (+1.9%) and on the
first-order absorption rate constant:
ka *= (1 + (-0.0589) * CONMED_NVP) (-5.9%); both retained
in the final model despite 95% CIs spanning zero (ASTMH 2020 poster 2167
Table 1)).CONMED_EFV (efavirenz) entry, applying to
the second commonly co-administered NNRTI. Future ART popPK models that
test an NVP-vs-comparator contrast should extend the example list when
the comparator matches.PCM – used in Plan_2012_pain.R (DDMORE
Foundation Model Repository entry DDMODEL00000194).Plan_2012_pain.R
(additive shift on the placebo-effect logit of the typical pain-score
lambda: phl = logit(TVLAM) + 0.364 * CONMED_PARA; mean pain
score ~9% higher on the 0-10 Likert scale during paracetamol use).CONMED_NSAID –
paracetamol is not classed as an NSAID (no anti-inflammatory mechanism,
distinct AE profile). Time-varying use is permitted; the daily Likert
measurements in Plan 2012 carry PCM as a per-observation flag. Follows
the CONMED_* concomitant-medication pattern established for
IBD models (AZA / MP / MTX / AMINO) and CONMED_NSAID (Li
2019). Ratified canonically alongside the Plan 2012 DDMORE
extraction.PB – used in
Schoemaker_2017_brivaracetam.R (paper covariate
PB for phenobarbital or primidone coadministration; the
source pools primidone with phenobarbital because primidone is
metabolised to phenobarbital).Schoemaker_2017_brivaracetam.R (multiplicative effect on
apparent oral clearance: cl *= (1 + 0.408 * CONMED_PB);
+40.8% relative to no-PB reference, corresponding to ~29% lower
brivaracetam exposure, Schoemaker 2017 Table 1).covariateData[[CONMED_PB]]$notes
should document whether primidone is pooled with phenobarbital
(Schoemaker 2017 pools them because primidone is metabolised to
phenobarbital). Distinct from the broader [[CONMED_EIAED]] (any
enzyme-inducing AED) and [[CONMED_AED]] (any concomitant AED). When a
paper distinguishes individual AEDs separately, use the drug-specific
canonicals [[CONMED_CBZ]], CONMED_PB, [[CONMED_VPA]] rather
than collapsing into the class-level indicator. Ratified canonically on
2026-05-20 alongside the Schoemaker 2017 brivaracetam paediatric
extraction.PLDH – used in Schmitt_2018_vinflunine.R
(Schmitt 2018 NONMEM dataset; 1 = vinflunine + PEGylated liposomal
doxorubicin combination cohort).Schmitt_2018_vinflunine.R (power-form effect on vinflunine
clearance: cl *= 0.865^CONMED_PLDH – vinflunine apparent CL
is reduced to 86.5% of single-agent value when coadministered with PLDH,
Schmitt 2018 Table 2 and explicit CL formula on p.1607).CONMED_AVD (brentuximab vedotin + AVD chemotherapy
backbone, which includes free doxorubicin rather than the PEGylated
liposomal formulation) and from PRIOR_ANTHRACYCLINE_DOSE
(cumulative prior anthracycline exposure, not concomitant). Ratified
canonically on 2026-05-25 alongside the Schmitt 2018 vinflunine
extraction.CONMED_ERA and CONMED_ERA_PDE5I).PAHCOMED category 2 (“PDE5 inh.”) – decomposed from the
categorical PAHCOMED source column with levels {naive, ERA,
PDE5, ERA+PDE5} in Krause_2017_selexipag.R.Krause_2017_selexipag.R (multiplicative effect on the
ACT-333679 elimination rate constant:
km *= (1 + 0.07 * CONMED_PDE5I); +7% relative to the
PAH-comedication-naive reference, Krause 2017 Table 1. The PDE5-only
coefficient is statistically not significant (p = 0.19) but retained in
the final model so the four-level PAH-comedication categorical is
preserved end-to-end).CONMED_ERA
and CONMED_ERA_PDE5I to decompose a four-level
PAH-comedication categorical (naive / ERA-only / PDE5-only /
ERA-and-PDE5) into three orthogonal mutually-exclusive binary
indicators. Specific scope because the indicator’s semantics are tied to
the GRIPHON study’s PAH-comedication taxonomy.CONMED_PPI – used in Goel_2016_Sonidegib.R
(Goel 2016 dataset; defined as “significant” CONMED_PPI use,
i.e. duration of CONMED_PPI use >= 80% of the PK assessment
phase).Goel_2016_Sonidegib.R
(multiplicative effect on F: 0.696^CONMED_PPI, ~30% lower
bioavailability under CONMED_PPI coadministration).covariateData[[CONMED_PPI]]$notes must document the
operational definition (e.g., Goel 2016 requires CONMED_PPI use covering
>= 80% of the PK assessment window; other studies may use a simpler
ever-vs-never indicator or a per-record time-varying flag). Distinct
from CONMED_H2RA (H2-receptor antagonist) which acts on
gastric pH via a different mechanism.Xie_2000_m3g_rat.R (the paper distinguishes a
probenecid-treatment arm from a control arm by experimental day; the
canonical column is 1 during the day-2 probenecid co-infusion and 0
otherwise, including on day 1 of the probenecid arm before the
probenecid loading dose).Xie_2000_m3g_rat.R
(multiplicative exponential effect on the unbound BBB influx clearance
CL_u,in:
cluin *= exp(e_conmed_probenecid_cluin * CONMED_PROBENECID)
with e_conmed_probenecid_cluin = log(0.17 / 0.11) = 0.4353,
i.e. a 1.55-fold increase in CL_u,in into rat brain ECF under probenecid
co-administration; CL_u,out and the intercompartmental brain clearance
Q_br are not statistically affected by probenecid in the paper’s model
selection and so are not paired with this canonical),
Landersdorfer_2009_gemifloxacin.R (binary treatment-arm
covariate that captures the static / non-mechanism-specific probenecid
effects on gemifloxacin absorption rate (Ka 0.839 -> 0.897 1/h;
+6.9%), absorption lag (Tlag 0.223 -> 0.129 h; -42%), and non-renal
clearance (CL_NR 25.2 -> 21.0 L/h; -16.7%); the mechanism-specific
competitive inhibition of renal tubular secretion is encoded separately
via the time-varying CP_PRB_MGL covariate),
Hamren_2008_tesaglitazar.R (multiplicative effect on parent
CLrt and metabolite Vmax under probenecid co-administration:
(1 - 0.75 * CONMED_PROBENECID) reduces both by 75% during
the three-dose probenecid window in four healthy-control subjects; the
shared estimate is encoded as two canonical-named parameters
e_probenecid_cl_renal and
e_probenecid_vmax_gluc to keep each effect’s naming
canonical), Stocker_2012_oxypurinol.R (multiplicative
linear-deviation effect on apparent CL/Fm:
cl *= (1 + 0.383 * CONMED_PROBENECID), i.e. +38.3% apparent
oxypurinol clearance with concomitant probenecid in adults with gout,
consistent with the paper’s mechanism narrative of probenecid inhibiting
renal tubular reabsorption of oxypurinol; Stocker 2012 Table 3
theta8).covariateData[[CONMED_PROBENECID]]$notes must document the
dose / regimen of probenecid used (Xie 2000: 70 umol/kg IV loading dose
plus 70 umol/kg/h constant infusion in male Sprague-Dawley rats), the
parameter the indicator modifies, and any per-subject vs per-record
time-varying convention.Padoin_1998_cephalexin_rat.R (the paper’s final model
specification is Ka_j = Ka2 and CL_j = CL2
when oral quinapril is coadministered with oral cephalexin; the
canonical column is 1 for group 5 (cephalexin GT + quinapril GT) and 0
for groups 1, 2, 3, and 4).Padoin_1998_cephalexin_rat.R (multiplicative exponential
effect on both absorption rate Ka and elimination clearance CL:
ka <- exp(lka + e_conmed_qprl_oral_ka * CONMED_QPRL_ORAL + etalka)
with e_conmed_qprl_oral_ka = log(0.177 / 0.249) = -0.3413
(~29% lower Ka under oral quinapril coadministration, paper Table 4);
cl <- exp(lcl + e_conmed_qprl_oral_cl * CONMED_QPRL_ORAL + etalcl)
with e_conmed_qprl_oral_cl = log(0.640 / 0.810) = -0.2356
(~21% lower CL under oral quinapril coadministration, paper Table 4).
The CL effect was found significant only in the oral-cephalexin +
oral-quinapril group (paper found no DDI on CL when cephalexin was given
intra-arterially, attributed to higher cephalexin renal concentrations
outcompeting quinapril at the carrier)…..covariateData[[CONMED_QPRL_ORAL]]$notes must document the
dose / regimen of quinapril used (Padoin 1998: 0.8 mg/kg single oral
dose via gastric tube, 15 min before cephalexin, in male Wistar rats),
the parameters the indicator modifies, and any per-subject vs per-record
convention. The paper’s full specification distinguishes intra-arterial
vs oral quinapril (no DDI on cephalexin CL was observed for either
intra-arterial cephalexin with intra-arterial quinapril or
intra-arterial cephalexin with oral quinapril); the canonical
CONMED_QPRL_ORAL = 0 collapses all three non-DDI conditions
(no quinapril, intra-arterial quinapril, oral quinapril paired with
intra-arterial cephalexin) into the reference category because the model
predicts the same CL and Ka values for each. Future popPK extractions of
beta-lactam / ACE-inhibitor DDIs that test oral quinapril should reuse
this canonical; ACE-inhibitor coadministration with a different
perpetrator drug (e.g., enalapril, lisinopril) should register a
separate canonical with the perpetrator’s INN in the name.RIF – used in Svensson_2014_bedaquiline.R
(paper-defined indicator switching on at day 3 of 600 mg daily
rifampicin co-administration; the 3-day lag was selected by NONMEM
objective-function search over 1-8 day candidates and chosen as the best
fit per the paper’s Methods). Also used in
Barnett_2018_coproporphyrin_I.R and
Barnett_2018_rosuvastatin.R as the binary period-level
covariate that captures Barnett 2018 Table 1’s reductions of Vcpi (CPI)
and V1 / V2 / Q (RSV) during the rifampicin phase of the three-occasion
crossover study. Also used in Gatti_1996_dapsone.R (Gatti
1996 used the source-paper symbol R; the indicator switches
on for the 7 of 53 patients on rifampin co-administration for at least 2
weeks before blood sampling, so the chronic CYP3A4-induction effect was
at equilibrium).Svensson_2014_bedaquiline.R (multiplicative factor on
apparent CL_BDQ and CL_M2:
cl_eff = cl_base * 4.78^CONMED_RIF; 4.78-fold induction of
apparent clearance at chronic full induction),
Barnett_2018_coproporphyrin_I.R (multiplicative factor on
Vcpi:
vc <- exp(lvc + etalvc) * (1 + e_rif_vc * CONMED_RIF)
with e_rif_vc = -0.4841 capturing the Vcpi 6.59 L -> 3.4 L reduction
in the acute rifampicin phase), Barnett_2018_rosuvastatin.R
(multiplicative factors on V1, Q, and V2 with the same encoding form,
capturing V1 430 -> 2.98 L, Q 45.3 -> 5.03 L/h, V2 865 -> 128 L
reductions during the acute rifampicin phase),
Gatti_1996_dapsone.R (multiplicative factor SHARED between
apparent CL/F and V/F:
cl = exp(lcl + etalcl) * (1 + e_rif_cl_vc * CONMED_RIF),
vc = exp(lvc) * (1 + e_rif_cl_vc * CONMED_RIF) with
e_rif_cl_vc = 0.696; 69.6% increase in both CL/F and V/F
driven primarily by a first-pass / bioavailability effect, with the
shared theta enforced by likelihood-ratio test against a non-shared
parameterisation), Zhang_2012_lopinavir_ritonavir.R
(multiplicative factors on LPV CL/F (+71.0%), RTV CL/F (+36.0%), LPV F
(-20.0%), and RTV F (-45.0% at the 100 mg reference RTV dose) captured
by the same (1 + e_rif_* * CONMED_RIF) encoding; the
CONMED_RIF flag also gates the secondary dose-dependent boost on RTV F
((1 + e_dose_f_rtv * (DOSE - 100))) so the dose-effect is
identifiable only within the rifampicin-coadministered arm),
Zhang_2013_lopinavir_ritonavir.R (per-population
multiplicative shifts on apparent CL of LPV adults +58% / children +48%
and RTV adults +34% / children +22%, and per-population log-additive
shifts on relative bioavailability of LPV and RTV with separate
adult-only and child-only RIF effects).CONMED_RIF_LPVR4
(which is a compound indicator for rifampicin + super-boosted LPV/r 4:4
used by Tikiso 2021 pediatric abacavir popPK and carries a non-CL
effect). The acute-vs-chronic semantic distinction is per-paper –
document the exact window (rifampicin-dose time and effect duration; lag
days to full induction) in
covariateData[[CONMED_RIF]]$notes so downstream simulations
can construct the correct event-table indicator. Future popPK
extractions that hold rifampicin co-administration as a binary indicator
should reuse this canonical; if a future model needs a paper-specific
time-decaying induction trajectory rather than a step indicator,
register a separate canonical (CONMED_RIF_INDUCTION_TIME).
Specific scope because the effect-window definition is paper-specific.
Ratified canonically on 2026-05-21 alongside the Svensson 2014
bedaquiline chronic-induction extraction; broadened on 2026-05-26 to
cover acute OATP1B-inhibition use alongside the Barnett 2018 CPI / RSV
extractions; example list extended on 2026-05-31 to include the Gatti
1996 dapsone chronic-induction shared-CL/V-effect case.RPT – used in Svensson_2014_bedaquiline.R
(paper-defined indicator switching on at day 3 of 600 mg daily
rifapentine co-administration; same lag as the sibling
CONMED_RIF arm of the same paper).Svensson_2014_bedaquiline.R (multiplicative factor on
apparent CL_BDQ and CL_M2:
cl_eff = cl_base * 3.96^CONMED_RPT; 3.96-fold induction of
apparent clearance at full induction).CONMED_RIF. Rifapentine and rifampicin are both
CYP3A4-inducing rifamycins with similar mechanisms; the per-paper
magnitude differs (rifapentine often slightly less induction than
rifampicin per the on-disk Svensson 2014 fit). Like
CONMED_RIF, this is a step indicator at full induction with
the paper-specific lag documented in
covariateData[[CONMED_RPT]]$notes. Specific scope until a
second model ratifies the canonical; ratified on 2026-05-21 alongside
the Svensson 2014 extraction.RITUX – used in Wu_2024_inotuzumab.R.Wu_2024_inotuzumab.R
(additive fractional change on CL1:
CL1 * (1 + (-0.132) * CONMED_RITUX) ~= 13% lower CL1 with
concomitant rituximab).covariateData[[CONMED_RITUX]]$notes
(CONMED_RITUX = 1 - source$RITUX) rather than registering a
second canonical. Ratified canonically on 2026-04-26.notes field should document the per-study
cumulative-exposure threshold if one is applied (e.g., FIDELIO-DKD used
“long-term use” = “SGLT2 inhibitor present for at least 50% of the
on-finerenone treatment period”). Distinct from the CINH
plasma SGLT-inhibitor concentration regressor used in
renal-glucose-reabsorption QSP models – that canonical carries
continuous time-varying plasma exposure of the SGLT2 inhibitor itself,
whereas CONMED_SGLT2I is a binary drug-drug-interaction
indicator for the SGLT2i as a coadministered drug.SGLT / SGLT_INH0 /
GLYCSI_CATN = 2 – used in
vandenBerg_2021_finerenone.R (the FIDELIO-DKD popPK NONMEM
control stream’s three-level SGLT categorical with
2 flagging long-term SGLT2i use).vandenBerg_2021_finerenone.R (multiplicative effect
1.10^CONMED_SGLT2I on apparent CL/F AND on relative
bioavailability F1 inversely; long-term SGLT2 inhibitor coadministration
in the FIDELIO-DKD analysis reduces steady-state finerenone AUC by
~17.1% per Table 2 / Figure 3 of the source paper).CONMED_<concept> concomitant-medication pattern.
SGLT2 inhibitors are a class of anti-diabetic agents with cardio-renal
protective effects; concomitant use is increasingly common in chronic
kidney disease and type 2 diabetes populations, where they overlap
therapeutically with mineralocorticoid receptor antagonists (e.g.,
finerenone). The mechanism by which SGLT2i coadministration modestly
alters finerenone PK is hypothesised by van den Berg 2021 to relate to
SGLT2i-mediated partial reversal of uremic-toxin-induced CYP3A4
inhibition (Discussion paragraph 4.1 ‘The small effect of SGLT2i use on
clearance might also be attributable to this phenomenon, as at least
pre-clinical data indicate that SGLT2i use may partly reverse this
pathophysiology’). Distinct from DIS_DIAB
(diabetes-mellitus comorbidity) and from CINH (continuous
plasma SGLT2-inhibitor concentration regressor).HASPDR – used in Xu_2023_MBG453.R (Monolix
supplement Appendix S2; the source describes the column as “this patient
HAS received PDR001 [spartalizumab, anti PD-1 mAb]”).Xu_2023_MBG453.R
(exponential effect on CL: exp(0.0194 * CONMED_SPART); not
statistically significant in the full covariate model but retained
because Xu 2023 used the full-covariate-model approach).COMBO_NIVO
(ipilimumab + nivolumab) and COMBO_DURVA (durvalumab
combinations) but for spartalizumab. Promote to general scope if a
second paper reports a spartalizumab-coadministration covariate with
comparable encoding.SPI – used in Zhou_2010_digoxin.R (Zhou
2010 Table 1 and Table 7 column label).Zhou_2010_digoxin.R
(multiplicative linear-deviation form on Cl/F:
cl *= (1 - 0.412 * CONMED_SPIRON), i.e. ~41% lower Cl/F
with concomitant spironolactone in the older Chinese CHF cohort; Zhou
2010 Table 7), MarquesMinana_2010_vancomycin.R
(multiplicative linear-deviation form on weight-normalised V_d:
vc *= (1 - 0.344 * CONMED_SPIRON), i.e. ~34% smaller V_d
with concomitant spironolactone in the 70-neonate Marques-Minana 2010
cohort; Table 3 theta4).Martinez_2019_alirocumab.R (additive effect on linear
clearance CLL:
CLL = TVCLL + COV1*(WT-82.9) + COV2*CONMED_STATIN; +0.00644
L/h when conmed_statin is coadministered).covariateData[[CONMED_STATIN]]$notes must document which
statins and dose thresholds are included in the “CONMED_STATIN = 1”
category, since inclusion criteria vary by study. Martinez 2019 codes
CONMED_STATIN = 1 for coadministration of rosuvastatin (< 20 mg/day),
atorvastatin (< 40 mg/day), or simvastatin (any dose); other
conmed_statin regimens are coded as 0.Kuchimanchi_2018_evolocumab.R (multiplicative effect 1.13
on Vmax: Vmax * 1.13^CONMED_STATIN_MONO),
Budha_2015_rg7652.R (time-varying within-subject
multiplicative log-effect on the indirect-response Ksyn anchor for
LDL-C:
Ksyn = Kdeg * LDLC * exp(-0.648 * CONMED_STATIN_MONO(t));
same parameter dynamically reproduces both the ~45% lower pre-dose
baseline LDL-C in atorvastatin-pretreated cohorts and the LDL rebound
after statin cessation, Budha 2015 Table II ‘Statin on screening LDLc’ =
-0.648), Kuchimanchi_2018_evolocumab_ldlc.R (same PK-layer
effect plus an exponent 0.797 on baseline LDL-C and 0.937 on Emax in the
exposure-response layer, Table 4).CONMED_EZE: a subject on
conmed_statin+ezetimibe has CONMED_STATIN_MONO = 0 and
CONMED_EZE = 1; a subject on conmed_statin alone has
CONMED_STATIN_MONO = 1 and CONMED_EZE = 0; a
subject on no lipid-lowering therapy has both 0. Per-model temporal
grain (time-fixed per subject vs time-varying within subject) is
documented in covariateData[[CONMED_STATIN_MONO]]$notes per
model. Future popPK/PD models that adopt a broader “any conmed_statin”
definition should register a separate CONMED_STATIN or
CONMED_STATIN canonical rather than reusing this name.BSTEROID – used in
Narwal_2013_sifalimumab.R and
Zheng_2016_sifalimumab.R (time-fixed baseline-use
form).STEROID – used in
VelezdeMendizabal_2013_multipleSclerosis.R (time-varying
per-monthly-record acute-administration form).Narwal_2013_sifalimumab.R (time-fixed multiplicative on CL:
CL * (1 + 0.195 * CONMED_STEROID)),
Zheng_2016_sifalimumab.R (time-fixed multiplicative on CL
(1 + 0.11 * CONMED_STEROID) and on V1
(1 - 0.09 * CONMED_STEROID) in the SLE phase IIb cohort,
which was ~85% conmed_steroid-treated at baseline),
VelezdeMendizabal_2013_multipleSclerosis.R (time-varying
per-monthly-record switch of the first-order Markov coefficient from
theta_pdv to theta_pdv_s when a corticosteroid course was given for a
clinical MS relapse that month).PRICORT, which is
strictly a prior (pre-study) indicator. CONMED_STEROID
covers both concurrent chronic corticosteroid use at / from study
baseline and per-record acute corticosteroid pulses; the per-model
covariateData[[CONMED_STEROID]]$notes field documents the
temporal grain (time-fixed vs time-varying) the source paper used. When
a future paper needs CONMED_STEROID and
PRICORT jointly, both can coexist on the same subject. The
name STEROID_BL was used as an alias in earlier register
drafts and is retired; use CONMED_STEROID for all future
models.steroid sparing centre – used in
Passey_2011_tacrolimus.R (Passey 2011 Methods: “Centres
were designated as using a steroid sparing immunosuppressive regimen if
they administered steroids for <= 7 days post transplant”).Passey_2011_tacrolimus.R (power-of-binary-indicator
multiplicative factor on apparent oral clearance:
e_steroid_spare_cl ^ CONMED_STEROID_SPARING with
e_steroid_spare_cl = 0.70; steroid-sparing patients have
30% lower apparent oral tacrolimus CL/F than continuous-steroid
patients; Passey 2011 Discussion attributes the effect to reduced CYP3A
induction in the absence of ongoing corticosteroid therapy).CONMED_STEROID
(which captures concurrent / baseline corticosteroid USE;
CONMED_STEROID_SPARING captures the protocol-level decision
to MINIMIZE corticosteroid use). The two coexist in the same dataset
when needed: a steroid-sparing patient still has
CONMED_STEROID = 1 during days 1-7 post-transplant (the
short-duration administration window) and
CONMED_STEROID = 0 thereafter. The Passey 2011 binary
indicator collapses both phases into a single time-invariant per-subject
attribute via the centre-level assignment; document the per-model
temporal interpretation in
covariateData[[CONMED_STEROID_SPARING]]$notes. Distinct
from PRICORT (pre-study corticosteroid history) and from
HCT_COND_RIC (reduced-intensity conditioning regimen for
HSC transplantation, which is a different protocol axis). Future models
that distinguish the specific duration of steroid administration (e.g. 7
days vs 14 days vs 30 days) should register companion canonicals rather
than overloading CONMED_STEROID_SPARING. Ratified
canonically on 2026-05-20 alongside the Passey 2011 tacrolimus
extraction.VPA – used in
Schoemaker_2017_brivaracetam.R (paper covariate
VPA for valproate coadministration).Schoemaker_2017_brivaracetam.R (multiplicative effect on
apparent oral clearance: cl *= (1 - 0.101 * CONMED_VPA);
-10.1% relative to no-VPA reference, corresponding to ~11% higher
brivaracetam exposure, Schoemaker 2017 Table 1).CONMED_VPA rather than collapsing into the class-level
indicator. Ratified canonically on 2026-05-20 alongside the Schoemaker
2017 brivaracetam paediatric extraction.SLDB – used in
Bender_2009_pregabalin_rat_binary.R (Bender 2009 binary
sildenafil-presence indicator).Bender_2009_pregabalin_rat_binary.R (proportional reduction
in pregabalin CL when sildenafil is coadministered:
cl *= (1 - e_sild_cl * CONMED_SILDENAFIL) with
e_sild_cl = 0.302, i.e. 30.2% lower CL on the sildenafil
occasion, Bender 2009 Table IV).CONMED_<drug> binary
concomitant-presence family. Per-occasion in the Bender 2009 crossover
(each rat received saline on one occasion and sildenafil on the other,
separated by a >= 3-day washout). Companion to
CONMED_SILDENAFIL_NMETAB_CC, the continuous
N-desmethyl-sildenafil-metabolite-concentration covariate used in the
saturable-inhibition variant of the same model. Ratified canonically
alongside the Bender 2009 pregabalin rat extraction.SLDM – used in
Bender_2009_pregabalin_rat_smetab.R (Bender 2009 measured
N-methyl-sildenafil-metabolite plasma concentration; paper notation
[SLDM]).Bender_2009_pregabalin_rat_smetab.R (saturable inhibition
of pregabalin CL:
cl *= (1 - CONMED_SILDENAFIL_NMETAB_CC / (e_sldm_cl + CONMED_SILDENAFIL_NMETAB_CC))
with the IC50 e_sldm_cl = 1350 ng/mL, Bender 2009 Table
IV).CONMED_<drug>_CC time-varying-concentration
family (the _CC suffix marks a dynamic concentration
covariate, as distinct from the CONMED_<drug>
binary-presence indicator). Continuous-covariate counterpart to
CONMED_SILDENAFIL (the binary-presence variant of the same
Bender 2009 drug-drug-interaction analysis). Ratified canonically
alongside the Bender 2009 pregabalin rat extraction.nal naltrexone, bup bupropion).
0 = the named drug is not part of the regimen at this time; positive
value = current total daily dose. Drives the combined dose- and
time-dependent Emax drug effect in the naltrexone/bupropion
fixed-dose-combination PD model.DOSE_NAL_MGD (NAL, Sharma 2018 Eq. 4 daily
naltrexone dose in mg) – earlier
DOSE_<drug>_<unit> form used in
Sharma_2018_naltrexone_bupropion.R before the
CONMED_<drug>_DOSE rename. Maps to
CONMED_NAL_DOSE.DOSE_BUP_MGD (BUP, Sharma 2018 Eq. 4 daily
bupropion dose in mg) – earlier
DOSE_<drug>_<unit> form used in
Sharma_2018_naltrexone_bupropion.R before the rename. Maps
to CONMED_BUP_DOSE.Sharma_2018_naltrexone_bupropion.R (each daily-dose
covariate enters the combined dose- and time-dependent Emax drug effect
on body-weight kout:
drug_term = CONMED_NAL_DOSE / (ed50nal + CONMED_NAL_DOSE) + CONMED_BUP_DOSE / (ed50bup + CONMED_BUP_DOSE),
Sharma 2018 Eq. 4).CONMED_<drug>_DOSE daily-dose family (siblings
CONMED_ATV_DOSE, CONMED_INH_DOSE, etc.); the
CONMED_<drug>_DOSE shape replaces the earlier
DOSE_<drug> /
DOSE_<drug>_<unit> names (which conflated a
covariate with a dose-amount column) per the naming audit. Companion to
DIS_DIAB (the diabetes covariate in the same model;
pre-2026-06-19 canonical name T2DM). Ratified canonically
alongside the Sharma 2018 naltrexone/bupropion extraction.Ma_2020_sarilumab_das28crp.R (multiplicative on DAS28-CRP
Kout: Kout * theta^PRICORT),
Ma_2020_sarilumab_anc.R (power-form on Emax:
Emax * 0.819^PRICORT).param * theta^PRICORT in both DAS28-CRP
and ANC PD models. Generally applicable clinical-history indicator.(1 + theta * (PRIOR_ANTHRACYCLINE_DOSE - ref))) on a
baseline parameter (e.g., baseline cardiac troponin I before the next
anthracycline cycle). Reference values observed: 90 mg/m^2 (Kunarajah
2017, cohort median).PCAMT – Kunarajah 2017 NM-TRAN convention (“Prior
Cumulative Anthracyclines aMounT”; doxorubicin-equivalent mg/m^2).Kunarajah_2017_doxorubicin.R (linear shift on baseline
cardiac troponin I:
bl_cTnI * (1 + 0.00308 * (PRIOR_ANTHRACYCLINE_DOSE - 90)) –
~0.31% increase in baseline cTnI per 1 mg/m^2 of prior cumulative
anthracycline exposure).PRIOR_ANTICANCER
(a binary modality indicator, 1 = any prior anticancer therapy) –
PRIOR_ANTHRACYCLINE_DOSE carries the actual cumulative
dose, restricted to the anthracycline drug class (doxorubicin,
daunorubicin, epirubicin, idarubicin), and is the column needed when the
source paper’s effect is dose-response in the prior-exposure regime
rather than presence / absence. When a paper records anthracycline
exposure as anthracycline-class-by-class doses and the model effect
aggregates them, sum to a single doxorubicin-equivalent value before
populating this column (use the published bone-marrow / cardiotoxicity
isoeffective conversion factors). When a paper distinguishes the type of
anthracycline (e.g., doxorubicin vs daunorubicin separately), register
parallel canonicals (PRIOR_DOXORUBICIN_DOSE,
PRIOR_DAUNORUBICIN_DOSE) rather than overloading this name.
Scope: specific because the column meaning is intrinsically tied to
anthracycline-class chemotherapy exposure; promote to general if a
second paper ratifies the same definition.LINE_1L, which is
specifically a treatment-line indicator restricted to systemic drug
therapy lines. PRIOR_ANTICANCER captures the full clinical
concept of prior cancer treatment exposure as used in
cytotoxic-chemotherapy myelosuppression analyses, where any prior
anticancer modality may have depleted the bone-marrow proliferating pool
and therefore affects baseline ANC.PC (Kloft 2006 / Netterberg 2017 NM-TRAN convention for
“previous anticancer therapy”; values 0 = naive, 1 = had prior
anticancer therapy) – used in
Netterberg_2017_docetaxel.R.Netterberg_2017_docetaxel.R (multiplicative effect on
baseline ANC: BACOV *= (1 + theta * PRIOR_ANTICANCER) with
theta = -0.147; prior-anticancer patients have ~14.7% lower baseline ANC
than treatment-naive patients).LINE_1L (which is
the inverse semantics for systemic-drug therapy lines only) and
PRIOR_TNF / PRIOR_BIO (which are
modality-specific to anti-TNF / biologic exposure in
inflammatory-disease cohorts). Use PRIOR_ANTICANCER when
the source paper’s covariate counts any anticancer modality (including
radiotherapy and surgery) as prior exposure. When a future paper
restricts the indicator to cytotoxic chemotherapy alone, use
LINE_1L (with values inverted: paper’s
PRIOR_CHEMO = 1 - LINE_1L). When a paper distinguishes
prior chemotherapy from prior radiotherapy, register a parallel
PRIOR_RADIATION canonical.Lu_2017_polatuzumab_neuropathy.R.Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on
the grade >= 2 peripheral neuropathy hazard via
exp(theta_priorRadioTx * PRIOR_RADIATION) with
theta_priorRadioTx = -7.94e-3, SE 0.319 – effectively no
detectable effect given the large SE).PRIOR_ANTICANCER
(the broader any-modality indicator) and PRIOR_VINCA /
PRIOR_PLATIN (chemotherapy-class-specific siblings). Use
this canonical when the source paper distinguishes radiotherapy as a
separate covariate from prior chemotherapy / hormonal / immunotherapy
exposure. Pre-authorized as a parallel-sibling canonical in the
PRIOR_ANTICANCER notes. Ratified canonically on 2026-06-24
alongside the Lu 2017 polatuzumab vedotin TTE extraction.Lu_2017_polatuzumab_neuropathy.R.Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on
the grade >= 2 peripheral neuropathy hazard via
exp(theta_priorVinca * PRIOR_VINCA) with
theta_priorVinca = -0.102, SE 0.469 – not detectable given
the large SE; the paper hypothesized prior vinca-alkaloid exposure could
sensitize patients to antimicrotubule-induced PN but did not find a
clinically meaningful effect).PRIOR_ANTICANCER (broader any-modality
indicator), PRIOR_RADIATION, PRIOR_PLATIN, and
PRIOR_TAXANE (parallel chemotherapy-class siblings).
Ratified canonically on 2026-06-24 alongside the Lu 2017 polatuzumab
vedotin TTE extraction.Lu_2017_polatuzumab_neuropathy.R.Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on
the grade >= 2 peripheral neuropathy hazard via
exp(theta_Platin * PRIOR_PLATIN) with
theta_Platin = 0.159, SE 0.345 – not detectable given the
large SE; the paper hypothesized prior platinum exposure could sensitize
patients to subsequent antimicrotubule-induced PN given that platinum
agents themselves cause sensory PN, but did not find a clinically
meaningful effect).PRIOR_ANTICANCER (broader any-modality
indicator), PRIOR_RADIATION, PRIOR_VINCA, and
PRIOR_TAXANE (parallel chemotherapy-class siblings).
Distinct from AUC_CARBO (the continuous per-cycle
carboplatin-AUC drug-exposure covariate, which is time-varying and used
as a direct drug-exposure driver in tumor-dynamics models,
e.g. Zecchin_2016_survival.R). Ratified canonically on
2026-06-24 alongside the Lu 2017 polatuzumab vedotin TTE
extraction.PRIOR_TNF: includes anti-TNF
agents plus anti-integrin, anti-IL-12/23, anti-IL-17, anti-IL-23,
anti-IL-6, etc.), 0 = biologic-naive.bio-naive (Aguiar 2021 source-paper variable, with the
indicator inverted: paper’s bio-naive = 1 - PRIOR_BIO).
Effect coefficients in the source paper apply to bio-naive;
in model() derive
bio_naive <- 1 - PRIOR_BIO to preserve the paper’s
reported coefficient.Aguiar_2021_ustekinumab.R (Aguiar 2021 Table 2 footnote a;
multiplicative effect on CL: factor
(1 - 0.227 * (1 - PRIOR_BIO)), so bio-naive patients have
~23% lower CL than previously-exposed patients).PRIOR_TNF (a
strict subset). Use PRIOR_BIO when the source paper’s
covariate counts any biologic as prior exposure (anti-TNF,
anti-integrin, anti-IL-12/23, anti-IL-17, anti-IL-23, anti-IL-6, etc.);
use PRIOR_TNF when the source paper specifically tested
anti-TNF exposure. When the source paper uses the inverted “bio-naive”
indicator (1 = naive), document the inversion in
covariateData[[PRIOR_BIO]]$notes and apply
1 - PRIOR_BIO in model() so the canonical
column stores 1 = previously exposed.PRIORTNF (all caps, no underscore) – acceptable
alternative spelling.Moein_2022_etrolizumab.R (multiplicative fractional effect
on CL, +4.9%).IPI (Ahamadi 2017; categorical with levels
IPI-naive, IPI-treated, missing)
– decompose into
PRIOR_IPI = as.integer(IPI == "IPI-treated") and treat the
missing category like naive unless the source paper retains a separate
“missing” coefficient.Ahamadi_2017_pembrolizumab.R (proportional changes on CL of
+14.0% and on Vc of +7.36% for IPI-treated relative to IPI-naive;
“missing” 26.4% of cohort is pooled with naive in the canonical encoding
because Table 3 reports only the naive-vs-treated coefficient).PRIOR_ANTICANCER
(any modality), PRIOR_BIO (any biologic),
PRIOR_TNF (anti-TNF biologic). Use PRIOR_IPI
when the source paper specifically tested prior ipilimumab exposure as a
covariate; this is a common covariate in advanced-melanoma popPK
analyses where ipilimumab was the standard-of-care immune-checkpoint
inhibitor preceding PD-1 / PD-L1 entrants. Ratified canonically on
2026-05-17 alongside the Ahamadi 2017 pembrolizumab extraction.NNI – used in Hirt_2006_nelfinavir.R
(paper’s pooled “nonnucleosidic inhibitor” indicator: 1 if EFV or NVP, 0
otherwise; pooling justified by the paper’s finding that the two drugs’
inducer effects on M8 elimination were not significantly different and
the two were never administered simultaneously in the cohort).Hirt_2006_nelfinavir.R
(additive-form multiplicative effect on M8 apparent elimination rate
kel_m8: kel_m8 *= (1 + 0.91 * CONMED_NNRTI_IND); KM0 is
~1.9-fold higher under EFV or NVP co-administration, consistent with
CYP3A4 induction of M8 elimination).CONMED_EIAED (any enzyme-inducing AED) and
CONMED_AZOLE (any azole antifungal). Distinct from
CONMED_EFV (efavirenz alone, founded by Tikiso 2021): the
pooled CONMED_NNRTI_IND indicator is used when the source
paper did not separately identify EFV vs NVP effects. A future paper
that needs to distinguish them should use CONMED_EFV (and a
future sibling CONMED_NVP) rather than this pooled
indicator. The “inducer” qualifier excludes non-inducing NNRTIs
(rilpivirine, doravirine, etravirine); record the per-paper list of
pooled drugs in covariateData[[CONMED_NNRTI_IND]]$notes.
Specific scope until a second model ratifies the convention; at that
point promote to general. Ratified canonically on 2026-06-09 alongside
the Hirt 2006 nelfinavir extraction.PSEU_jk,
sample-level).PSEU – used in VanWart_2004_garenoxacin.R
(paper covariate PSEU_jk, sample-level indicator that the
kth PK sample from the jth patient was collected during concomitant
pseudoephedrine administration).VanWart_2004_garenoxacin.R (multiplicative effect on the
structural CL/F covariate term:
cl_structural *= (1 - 0.144 * CONMED_PSEUDOEPHEDRINE);
-14.4% relative to the no-pseudoephedrine reference, Van Wart 2004 Table
4).CONMED_PROBENECID
(anionic OAT inhibitor, mechanistically a perpetrator on anion-secreted
substrates) – pseudoephedrine is the cationic / mixed-substrate
competitor rather than a transporter inhibitor.F_inotrope = theta8 = 0.708 when
CONMED_INOTROPE == 1 and 1 otherwise; encoded in nlmixr2 as
f_inotrope <- e_inotrope_cl ^ CONMED_INOTROPE with
e_inotrope_cl = 0.708.inotropic agents / inotropes – used in
Zhao_2014_ciprofloxacin.R (Zhao 2014 Table 2 ‘Comedication’
row and Table 4 F_inotrope notation); same orientation as
the canonical, no value transformation.Zhao_2014_ciprofloxacin.R (multiplicative selection on CL:
0.708 ^ CONMED_INOTROPE – 29.2% CL reduction in patients on
inotropes).covariateData[[CONMED_INOTROPE]]$notes for each model (Zhao
2014 does not enumerate the exact agents; the cohort is critical-care
neonates so the practical mix is dominated by dopamine / dobutamine /
epinephrine / norepinephrine / milrinone). Distinct from
extracorporeal-support modalities (ECMO_PUMP_SPEED, etc.) –
inotropes are a pharmacological intervention rather than a
mechanical-support modality. Distinct from heart rate (HR)
– inotropes may cause both reduced GFR and altered HR but the indicator
captures the therapy itself rather than its hemodynamic consequence. If
a future paper retains a finer-grained taxonomy (epinephrine-only vs
dopamine-only vs vasopressor-only), register CONMED_EPI /
CONMED_DOBU / CONMED_VASOPRESS as siblings
rather than overload CONMED_INOTROPE with conflicting
semantics. Ratified canonically on 2026-06-09 alongside the Zhao 2014
ciprofloxacin extraction.stanozolol indicator (no specific NONMEM column name
reported in Ni 2013) – used in Ni_2013_ciclosporin.R.Ni_2013_ciclosporin.R
(multiplicative power-of-binary effect on CL/F:
cl *= e_conmed_stanozolol_cl ^ CONMED_STANOZOLOL with
e_conmed_stanozolol_cl = 0.83, i.e. ~17% lower CL/F when
stanozolol is coadministered relative to the no-stanozolol reference; Ni
2013 Table 2 ‘F comedication’).CONMED_ANDROGEN) – defer that decision
until a second paper appears. Ratified canonically on 2026-06-07
alongside the Ni 2013 ciclosporin paediatric aplastic-anemia extraction
(sidecar request-001, response Option A).covariateData[[CONMED_BUDESONIDE]]$notes).Wu_2014_FEV1_asthma.R (the CAMP-trial randomized-arm
assignment with three levels {placebo, nedocromil, budesonide};
CONMED_BUDESONIDE = as.integer(arm == "budesonide");
nedocromil is pooled into the reference because the paper reports no
significant nedocromil effect on FEV1).Wu_2014_FEV1_asthma.R
(additive shift on the linear-scale FEV1 typical-value:
FEV1 = exp(theta_age*AGE + theta_ht*HT - theta_int) + e_conmed_bud_fev1 * CONMED_BUDESONIDE
with e_conmed_bud_fev1 = +0.103 L (paper Table 2) –
budesonide-arm subjects have ~0.10 L higher FEV1 relative to the
placebo+nedocromil reference, with an IIV SD of 0.129 L on the
drug-effect term itself; the high shrinkage 67.3% reported in the paper
reflects that only ~30% of the cohort carries the indicator).CONMED_<INN>
convention (full INN spelled out, matching
CONMED_METFORMIN, CONMED_PROBENECID,
CONMED_SILDENAFIL, CONMED_SPIRON). The
contrast is between an active treatment arm and a reference arm in a
randomized trial, encoded as a per-subject time-fixed binary. Specific
scope because the budesonide-vs-{placebo, nedocromil} head-to-head is
tied to the CAMP-trial design; future inhaled-corticosteroid-vs-placebo
popPD models can extend this entry’s example_models list, while a
contrast between a different ICS (e.g., fluticasone, mometasone) and
budesonide should register a sibling canonical rather than reusing
CONMED_BUDESONIDE. Distinct from TRT
(Novakovic 2017 categorical 3-level cladribine-arm covariate) and
TRT_PHASE (active-treatment-phase indicator);
CONMED_BUDESONIDE is a per-subject randomized-arm indicator
rather than a time-varying treatment-phase flag. Ratified canonically on
2026-06-12 alongside the Wu 2014 FEV1-asthma extraction.NNRTI – used in
Kappelhoff_2005_indinavir.R (Kappelhoff 2005 Table 2
footnote: “NNRTI is 1 when efavirenz or nevirapine are administered and
0 when they are not”). Kappelhoff 2005 specifically pooled efavirenz and
nevirapine because “inclusion of separate effects for each drug did not
improve goodness-of-fit.”Kappelhoff_2005_indinavir.R (multiplicative power-form
effect on CL/F: cl *= 1.41^CONMED_NNRTI; concomitant NNRTI
use increases indinavir apparent CL by 41%).CONMED_EFV (efavirenz only) – use CONMED_EFV
when the source paper estimates the efavirenz effect alone and
CONMED_NNRTI when the source paper pools the class.
Per-model covariateData[[CONMED_NNRTI]]$notes should list
which specific NNRTIs are counted as “1” in the dataset (Kappelhoff
2005: efavirenz and nevirapine; future extractions including rilpivirine
/ doravirine / etravirine / delavirdine should document each one).
Ratified canonically on 2026-06-10 alongside the Kappelhoff 2005
indinavir extraction.ETSEVO (continuous end-tidal sevoflurane concentration in
vol %), which is the quantitative driver used in
emergence-from-anaesthesia models such as Shin 2014;
CONMED_SEVO is the per-subject on/off flag.SEV – used in
Kleijn_2011_sugammadex_rocuronium.R (Kleijn 2011 Table 4
covariate-effect rows “No sevoflurane: keoSEV = 1” and “Sevoflurane:
keoSEV = 1 + theta”).Kleijn_2011_sugammadex_rocuronium.R (multiplicative effect
on the rocuronium effect-compartment equilibration rate
keo *= (1 + e_conmed_sevo_keo * CONMED_SEVO) with
e_conmed_sevo_keo = -0.567 and on the rocuronium NMB EC50
ec50 *= (1 + e_conmed_sevo_ec50 * CONMED_SEVO) with
e_conmed_sevo_ec50 = -0.395; sevoflurane simultaneously
slows the central-to-effect-compartment equilibration and lowers the
rocuronium concentration required for half-maximal neuromuscular
blockade, reproducing the well-known sevoflurane potentiation of
aminosteroid NMBAs).ETSEVO (end-tidal
sevoflurane volume percent, a continuous PD driver). Use
CONMED_SEVO when the source paper encodes sevoflurane
maintenance as a per-subject on/off indicator without a continuous
end-tidal trajectory; use ETSEVO when the per-record
end-tidal concentration is available and the model uses it directly.
Follows the CONMED_* family pattern
(CONMED_VPA, CONMED_RIF,
CONMED_AMIO). Future perioperative models that use a
similar binary indicator for isoflurane (CONMED_ISO) or
desflurane (CONMED_DES) should register parallel canonicals
rather than overloading this name. Ratified canonically alongside the
Kleijn 2011 sugammadex-rocuronium extraction.DOPA – used in Fuchs_2014_gentamicin.R
(paper covariate name in Table 2; same orientation as the canonical, 1 =
on dopamine, 0 = otherwise).Fuchs_2014_gentamicin.R (multiplicative linear effect on
CL: 1 + (-0.120) * CONMED_DOPA, i.e. 12% lower CL on
dopamine-coadministered neonates relative to no-dopamine reference;
Fuchs 2014 Table 2 theta_CL_DOPA = -0.120, SE 22%).notes.covariateData[[RHEUMATOID_FACTOR]]$units.(log(RHEUMATOID_FACTOR) / log(ref))^exponent (or,
equivalently, the source-paper form
(LRF / log(ref))^exponent where
LRF = log(RHEUMATOID_FACTOR)). Reference value observed:
110 U/mL (Frey 2010, corresponding to LRF = 4.7 in the paper’s
final-model equation).LRF – log-transformed RF (natural log of the value in
U/mL); Frey 2010 fits the covariate on the log scale and reports the
reference as LRF = 4.7 (i.e.,
log(110) ~= 4.7). The canonical column carries the raw RF
concentration in U/mL; the log transform is applied inside
model().RF – universal NONMEM/clinical-PK abbreviation;
rejected as the canonical name on 2026-04-28 because the bare two-letter
abbreviation is uncommon in published popPK papers and could be confused
with other shortenings.Frey_2010_tocilizumab.R (U/mL, reference 110 U/mL == LRF =
4.7; small positive exponent +0.1 on linear CL applied to
log(RHEUMATOID_FACTOR)).(BLPHYVAS / <ref>)^exponent. Reference 66 used in Ma
2020.Ma_2020_sarilumab_das28crp.R.(BLHAQ / <ref>)^exponent. Reference 1.75 used in Ma
2020.Ma_2020_sarilumab_das28crp.R.BLPHYVAS (the physician’s
global assessment of disease activity). Both baseline and time-varying
usages are covered; document per-model in
covariateData[[PAIN]]$notes whether the column is
baseline-only.(PAIN / ref)^exponent. Reference value observed: 60 in Frey
2013 (approximate dataset median across OPTION/TOWARD).Frey_2013_tocilizumab.R (baseline; power effect on the
indirect-response BASE parameter with reference 60 and exponent
0.062).(PAIN/60)^exp return 0. Frey 2013 documents an
explicit 0.010 floor on PAIN in its Table 2 covariate range; the model
file applies the same floor inside model() so simulation
under PAIN = 0 returns a finite BASE rather than collapsing to zero.
Canonical name follows the EOS / SCORE_EASI
convention (no BL prefix); baseline-vs-time-varying status
is per-model.notes.((SWOL_28JOINT + 1)/(<ref> + 1))^exponent to
avoid the zero-count edge case. Reference value observed: 16 in Li 2019
(approximate dataset median of the popPK cohort).SWOL – used in Li_2019_abatacept.R (Li
2019 Methods abbreviation).Li_2019_abatacept.R
(power effect on CL with exponent 0.0965; not clinically relevant per Li
2019)._28JOINT suffix
distinguishes this from the 66/68-joint swollen count used in some
earlier RA scales – register a separate canonical
(SWOL_66JOINT or similar) if a future paper uses a
different joint-count scale. Canonical name drops the BL
prefix to match the SCORE_EASI / AGE /
WT / ALB convention where
baseline-vs-time-varying status is documented in
covariateData notes rather than the column name.*allele genotypes, etc.); the per-model
covariateData[[CYP2D6]]$units, description,
and notes document which proxy is in force and the
population-median reference value used inside the model. Time-invariant
in all known examples (germline genotype or one-time probe-substrate
measurement).ng/L in Ter Heine 2014 where the value is the
dextromethorphan-probe model-based individual CYP2D6 clearance).covariateData[[CYP2D6]]$notes.CYP2D6 – used directly in
TerHeine_2014_tamoxifen.R.TerHeine_2014_tamoxifen.R (power-law effect on the
tamoxifen -> endoxifen formation clearance:
(CYP2D6 / 1560)^e_CYP2D6_cl_endx).CYP2D6 column
with an enumerated notes-documented encoding
field, OR introduce companion canonicals
(CYP2D6_PHENO_GROUP for the categorical PM/IM/EM/UM
grouping, CYP2D6_ACTSCORE for the AS sum) when a future
model needs the categorical form. The current general-scope continuous
canonical is sufficient for the Ter Heine 2014 use case but will likely
need refinement as more CYP2D6-aware models are added. The companion
binary CYP2D6_PM entry below covers source papers that
report only a poor-metabolizer-versus-not-poor-metabolizer
dichotomy.2D6PM – used in
Knights_2015_aripiprazole.R (Knights 2015 Eq. 2 binary
2D6PM = 1 if CYP2D6 poor metabolizer, = 0
otherwise; in the supplement NONMEM control stream the dataset column
CYP2D6EM is recoded as PM = (CYP2D6EM == 0) –
i.e., the source dataset’s CYP2D6EM extensive-metabolizer
indicator is value-inverted to produce the model’s PM indicator).P1 (PM) / “poor metabolizer” – Sherwin 2012 (paper
Table 2 mixture-model subpopulation P1; Table 1 phenotype assignment for
genotyped subjects). Sherwin 2012 uses CYP2D6_PM paired
with CYP2D6_EM to encode all three PM / IM / EM phenotypes;
IM is the implicit reference (both indicators = 0).Knights_2015_aripiprazole.R (proportional-shift effect on
apparent oral clearance:
CL/F = TVCL * (1 + e_2d6pm_cl * CYP2D6_PM) with
e_2d6pm_cl = -0.478; CYP2D6 poor metabolizers have 47.8%
lower apparent oral clearance than non-poor-metabolizers; Knights 2015
Equation 2 and Figure 1B), Sherwin_2012_risperidone.R
(phenotype-indicator covariate gating subpopulation-specific apparent
oral clearance and metabolite-formation fraction in a one-compartment
risperidone + (+/-)-9-hydroxyrisperidone mixture model: PM CL/F = 9.38
L/h, IM CL/F = 29.2 L/h, EM CL/F = 37.4 L/h at the 70 kg allometric
reference; KF (fraction metabolized) PM = 0.16, EM = 0.13, IM = 1 fixed;
Sherwin 2012 Table 2).CYP2C9_EM precedent (which encodes the
extensive-metabolizer-equals-1 phenotype) – here the 1
indicates the poor-metabolizer end of the phenotype spectrum because the
Knights 2015 source paper explicitly defines its binary as
2D6PM (PM = 1) and reports the coefficient sign relative to
that orientation; preserving the PM-equals-1 orientation reproduces the
paper’s reported coefficient sign and typical-value parameters directly.
Distinct from the continuous CYP2D6 activity-score
canonical above: use CYP2D6_PM when the source paper
reports a categorical phenotype (PM / IM / EM); use CYP2D6
when the source reports a probe-derived continuous activity number. For
three-level PM / IM / EM phenotype encoding (Sherwin 2012 mixture
model), pair CYP2D6_PM with the companion
CYP2D6_EM below following the
SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM
two-binary-indicator pattern with IM as the implicit reference (both
indicators = 0); this is consistent with the CYP2C9_EM register entry’s
anticipated CYP2C9_PM companion. The Knights 2015 source
dataset records CYP2D6EM (extensive-metabolizer indicator,
1 = EM, 0 = not-EM); the model’s PM indicator is derived by value
inversion (CYP2D6_PM = 1 - CYP2D6EM after handling the
supplement’s -99 missing-value sentinel as
CYP2D6_PM = 0). Ratified canonically on 2026-05-21
alongside the Knights 2015 aripiprazole extraction; scope retained as
general on 2026-05-24 alongside the Sherwin 2012 risperidone extraction
(second model using the same indicator-binary encoding, demonstrating
the canonical generalizes beyond the Knights 2015 PM-vs-non-PM dichotomy
into the three-level PM / IM / EM paired-indicator pattern with
CYP2D6_EM).CYP2D6_EM is paired with
CYP2D6_PM (Sherwin 2012 three-level encoding), the
reference category specifically corresponds to the
intermediate-metabolizer (IM) stratum: both CYP2D6_PM = 0
and CYP2D6_EM = 0 together indicate an IM subject.P2 (EM) / “extensive metabolizer” – Sherwin 2012 (paper
Table 2 mixture-model subpopulation P2; Table 1 phenotype assignment for
genotyped subjects).Sherwin_2012_risperidone.R (phenotype-indicator covariate
gating subpopulation-specific apparent oral clearance and
metabolite-formation fraction in a one-compartment risperidone +
(+/-)-9-hydroxyrisperidone mixture model; paired with
CYP2D6_PM with IM as the implicit reference; Sherwin 2012
Table 2).CYP2D6_PM
canonical above; the paired-indicator pattern follows
SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM (two
binary indicators encoding a three-level categorical with an implicit
reference). When a source paper reports only a binary PM-vs-non-PM
dichotomy (e.g., Knights 2015), CYP2D6_PM alone is
sufficient and CYP2D6_EM is omitted from the model’s
covariateData. When a source distinguishes all three PM /
IM / EM phenotypes (Sherwin 2012), include both indicators with IM as
the implicit reference. UM (ultrarapid metabolizer) – when present in a
source cohort – must be handled by registering an additional paired
CYP2D6_UM indicator on the same pattern; this is not yet
needed for any existing model and is deferred until a UM-aware paper is
extracted. Distinct from the continuous CYP2D6
activity-score canonical above and from CYP2D6_PM: use
CYP2D6_EM when the source separately identifies the EM
stratum (vs lumping IM + EM into a single non-PM group). Ratified
canonically on 2026-05-24 alongside the Sherwin 2012 risperidone
extraction.CYP2D6_STAR10_HOM flags the T/T homozygous-mutant
group). Time-fixed per subject (germline genotype). Distinct from the
broader CYP2D6_PM / CYP2D6_EM phenotype
canonicals above: CYP2D6_PM / CYP2D6_EM
summarise the overall metabolic phenotype across all assayed CYP2D6
alleles, while CYP2D6_STAR10_HET /
CYP2D6_STAR10_HOM resolve just the rs1065852 SNP and are
the right canonical when the source paper genotypes only the 10
variant.CYP2D6_STAR10_HOM = 0). The reference
group is the C/C wild-type stratum; CYP2D6_STAR10_HOM flags
the T/T homozygous-mutant stratum.CYP2D6*10 C/T – used in
Pei_2016_iloperidone.R (Pei 2016 Results ‘Population
pharmacokinetic (PPK) models’ encodes CYP2D6*10 C/C, C/T, T/T as
integers 1, 2, 3; the C/T = 2 stratum maps to
CYP2D6_STAR10_HET = 1).Pei_2016_iloperidone.R
(multiplicative effect on the iloperidone -> M2 (P-95) formation rate
constant K24: K24_CT = 0.693 * K24_typical relative to the C/C wild-type
reference; Pei 2016 Table 3 and Results ‘Population pharmacokinetic
(PPK) models’ equation
(K24)_i = theta * 0.00649 * exp(eta_i) with theta = 0.693
for C/T).CYP3A5_STAR1_HET /
CYP3A5_STAR1_HOM and SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM precedents (paired binary indicators for
a three-level germline genotype with an implicit wild-type reference).
The _STAR10_ token names the variant-allele orientation
(the *10 allele is the reduced-function variant of CYP2D6 that causes
lower CYP2D6 metabolic activity); the C/C wild-type stratum is the
implicit reference because the source-paper coefficient signs are
reported relative to that reference. Pei 2016 pooled C/C and C/T as the
reference for the K23 (M1 formation) effect (only T/T was distinguished
on K23), so a model using these two canonicals will route the K23 effect
through CYP2D6_STAR10_HOM alone while routing the K24
effect through both CYP2D6_STAR10_HET and
CYP2D6_STAR10_HOM – this asymmetric per-parameter use is
normal and is what the paired-indicator pattern is designed to express.
Ratified canonically on 2026-06-03 alongside the Pei 2016 iloperidone
extraction.CYP2D6_STAR10_HET flags the C/T heterozygous group).
Time-fixed per subject (germline genotype).CYP2D6_STAR10_HET = 0). The reference
group is the C/C wild-type stratum; CYP2D6_STAR10_HET flags
the C/T heterozygous stratum.CYP2D6*10 T/T – used in
Pei_2016_iloperidone.R (Pei 2016 Results ‘Population
pharmacokinetic (PPK) models’ encodes CYP2D6*10 C/C, C/T, T/T as
integers 1, 2, 3; the T/T = 3 stratum maps to
CYP2D6_STAR10_HOM = 1).Pei_2016_iloperidone.R
(multiplicative effects on two iloperidone metabolite-formation rate
constants: K23 (M1 formation) is multiplied by 1.34 in T/T subjects
relative to the C/C + C/T pooled reference, and K24 (M2 formation) is
multiplied by 0.492 in T/T subjects relative to the C/C wild-type
reference; Pei 2016 Table 3 and Results ‘Population pharmacokinetic
(PPK) models’ equations
(K23)_i = theta * 0.00451 * exp(eta_i) with theta = 1.34
for T/T and (K24)_i = theta * 0.00649 * exp(eta_i) with
theta = 0.492 for T/T).CYP2D6_STAR10_HET
(see that entry’s Notes for the three-level decomposition rationale, the
variant-orientation convention, and the asymmetric K23-vs-K24 use the
paired indicators support in the Pei 2016 model). Allele frequencies for
the CYP2D6*10 (rs1065852) variant in Chinese populations are 48-70% per
the source paper’s Introduction; the Pei 2016 cohort (n = 70 Chinese
schizophrenia patients) observed C/C 15.7%, C/T 60.0%, T/T 24.3%.
Ratified canonically on 2026-06-03 alongside the Pei 2016 iloperidone
extraction.CYP2D6 above. Some sources measure
CYP3A4 alone via probe substrate; others (Ter Heine 2014) report a
combined CYP3A4/5 activity because the chosen probe (dextromethorphan
N-demethylation) cannot distinguish CYP3A4 from CYP3A5. The per-model
notes field documents whether the value is CYP3A4-only or
the CYP3A4 + CYP3A5 combined score.ng/L in Ter Heine 2014 where the value is the
dextromethorphan-probe model-based individual CYP3A4/5 clearance).CYP3A4 – used directly in
TerHeine_2014_tamoxifen.R (the column carries combined
CYP3A4 + CYP3A5 activity per the source paper).CYP3A4/5 – long form sometimes used in source
manuscripts; standardize the column name to CYP3A4 and
document the combined-isoform semantics in per-model
notes.TerHeine_2014_tamoxifen.R (power-law effect on the
tamoxifen -> endoxifen formation clearance:
(CYP3A4 / 44.7)^e_CYP3A4_cl_endx).CYP1A2,
CYP2A6, CYP2B6, CYP2C8,
CYP2C9, CYP2C19, CYP2E1,
CYP3A5) using the same continuous-individual-activity-score
pattern, so future popPK models that report CYP-probe-derived covariates
can drop straight into the existing convention rather than
re-deliberating the encoding each time. Coordinate with the
consolidation TODO on CYP2D6 so categorical-vs-continuous
encoding is handled uniformly across all CYPs.X – used in Bergmann_2014_tacrolimus.R
(Bergmann 2014 Table 2 footnote: X = 1 for 1/1 and
1/3; X = 0 for 3/3).CYP3A5 expresser – used in
Storset_2014_tacrolimus.R (Storset 2014 Table 2 final
theory-based model; *1/*1 and *1/*3 pooled as
expressers because Storset 2014 had only n = 3 1/1
subjects).Bergmann_2014_tacrolimus.R (multiplicative effect on
tacrolimus CL/F: theta_CYP3A5 ^ CYP3A5_EXPR, with
theta_CYP3A5 = 1.60; expressers have 60% higher apparent
oral clearance than nonexpressers),
Storset_2014_tacrolimus.R (multiplicative effects on
apparent plasma clearance: cl *= 1.30^CYP3A5_EXPR; and on
oral bioavailability: fdepot *= 0.82^CYP3A5_EXPR; Storset
2014 Table 2 final theory-based model).SNP_<GENE>_<RSID> (which encodes “mutant allele
presence” – 1 = at least one variant allele). For CYP3A5 the 3
allele (rs776746 G) is the variant that abolishes function, so a literal
“mutant-allele-presence” indicator (1 = any G allele) would group
1/3 heterozygotes with the 3/3 nonexpressers, which is
the opposite of the clinically meaningful
expresser-vs-nonexpresser dichotomy used by every CYP3A5-aware popPK
model. The CYP3A5_EXPR canonical preserves the
expresser-equals-1 orientation directly. Future CYP3A5 papers using a
3/3 indicator (rather than 1 carrier) should still record
their values under CYP3A5_EXPR and document the value
inversion in notes
(CYP3A5_EXPR = 1 - source_indicator); registering a
parallel CYP3A5_NONEXPR is discouraged. The canonical name
follows the <gene>_<phenotype> rather than the
<gene>_<rsid> pattern because the column
captures derived metabolic phenotype rather than raw genotype. Distinct
from CYP3A4 (continuous individual-activity score for
CYP3A4 / CYP3A4 + CYP3A5 combined): the binary CYP3A5_EXPR
is the right fit for source papers that report only the rs776746
genotype, while the continuous CYP3A4 is for sources that
report a probe-substrate-derived activity number. In
solid-organ-transplant popPK / pharmacogenetics studies that genotype
both the recipient and the donated graft separately (e.g., Moes 2016
liver-transplant tacrolimus), the recipient genotype goes into
CYP3A5_EXPR and the donor genotype into the sibling
canonical CYP3A5_EXPR_DONOR below. Ratified canonically on
2026-05-08 alongside the Bergmann 2014 extraction.CYP3A5_EXPR.Donor CYP3A5*3 – used in
Moes_2016_tacrolimus.R (Moes 2016 Methods; donor and
recipient genotypes pooled with the four-level combination indicator
C1 / C2 / C3 / C4 defined in Methods, where
CYP3A5_EXPR_DONOR = 1 corresponds to graft from a
*1-carrying donor).CYP3A5 donor – used in
Ji_2018_tacrolimus.R (Ji 2018 derives the donor-side
genotype from the four-level combinational CYP3A5 group: REDE/RNDE ->
1; REDN/RNDN -> 0).Moes_2016_tacrolimus.R
(categorical donor + recipient CYP3A5 combination effect on apparent
oral CL/F: reference C1 = both nonexpressers; C2 = recipient
*1 carrier + donor nonexpresser = +33%; C3 = recipient
nonexpresser + donor *1 carrier = +33%; C4 = both
*1 carriers = +71%; Moes 2016 Table 4 final model – the C2
/ C3 / C4 levels are reconstructed inside model() from the
two binary inputs CYP3A5_EXPR and
CYP3A5_EXPR_DONOR), Ji_2018_tacrolimus.R
(combinational categorical effect on tacrolimus CL/F that depends on
both recipient and donor CYP3A5 status: multiplier 2.314 when the
recipient is an expresser and the donor is an expresser (REDE), 1.523
when the recipient is an expresser and the donor is a nonexpresser
(REDN), and 1.0 otherwise (RNDE / RNDN reference, which Ji 2018 merged
because the two estimated effects were similar)).CYP3A5_EXPR for the recipient,
CYP3A5_EXPR_DONOR for the donor) rather than as a single
four-level combination column, so the underlying recipient / donor
biology is explicit in the dataset. Models that fit a categorical
four-level combination effect (paper-Moes-style C1 / C2 / C3 / C4)
reconstruct the levels inside model() from the two binary
inputs, so the source-paper’s per-level coefficients remain the
estimated quantities. The intestinal-CYP3A5 contribution (recipient
genotype) and the hepatic-CYP3A5 contribution (donor genotype) act on
different anatomic compartments of tacrolimus’s first-pass metabolism,
which is why both donor and recipient genotypes are independently
informative in liver-transplant tacrolimus PK. Ratified canonically on
2026-05-20 alongside the Moes 2016 tacrolimus extraction.CYP3A5_STAR1_HOM flags
the homozygous-expresser group). Time-fixed per subject (germline
genotype).CYP3A5_STAR1_HOM = 0). The
reference group is the homozygous 3/3 nonexpresser stratum;
CYP3A5_STAR1_HOM flags the homozygous-expresser
stratum.CYP3A5*1/*3 – used directly in
Passey_2011_tacrolimus.R (Passey 2011 Table 3 reports a
separate multiplicative factor of 1.70 for the heterozygote stratum vs
the 3/3 reference).Passey_2011_tacrolimus.R (power-of-binary-indicator
multiplicative factor on apparent oral clearance:
e_cyp3a5_het_cl ^ CYP3A5_STAR1_HET with
e_cyp3a5_het_cl = 1.70; 1/3 heterozygotes have
~70% higher apparent oral CL/F than 3/3 nonexpressers; paired
with CYP3A5_STAR1_HOM and used jointly).SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM precedent (paired binary indicators for a
three-level genotype) rather than overloading CYP3A5_EXPR
(which pools 1/1 and 1/3 into a single expresser
indicator and is the right canonical when a source paper does the same
pooling). Use the paired CYP3A5_STAR1_HET +
CYP3A5_STAR1_HOM binaries when the source paper assigns a
distinct typical-value covariate effect to each of the three CYP3A5
genotype strata (3/3, 1/3, 1/1). Passey 2011
motivated the three-level decomposition because the cohort (n = 681) had
enough 1/1 homozygotes (72, 11%) to identify a distinct
typical-value factor for that stratum, whereas earlier popPK papers
(Bergmann 2014, Storset 2014) had too few 1/1 subjects (or the
equivalent expresser-pooled treatment) to distinguish 1/1 from
1/3. The _STAR1_ token (rather than
_STAR3_) names the functional-allele-presence orientation
consistent with the parent CYP3A5_EXPR canonical’s
expresser-equals-1 convention. Ratified canonically on 2026-05-20
alongside the Passey 2011 tacrolimus extraction.CYP3A5_STAR1_HET flags the
heterozygous group). Time-fixed per subject (germline genotype).CYP3A5_STAR1_HET = 0). The
reference group is the homozygous 3/3 nonexpresser stratum;
CYP3A5_STAR1_HET flags the heterozygous-expresser
stratum.CYP3A5*1/*1 – used directly in
Passey_2011_tacrolimus.R (Passey 2011 Table 3 reports a
separate multiplicative factor of 2.00 for the homozygote-expresser
stratum vs the 3/3 reference).Passey_2011_tacrolimus.R (power-of-binary-indicator
multiplicative factor on apparent oral clearance:
e_cyp3a5_hom_cl ^ CYP3A5_STAR1_HOM with
e_cyp3a5_hom_cl = 2.00; 1/1 homozygotes have 100%
higher apparent oral CL/F than 3/3 nonexpressers; paired with
CYP3A5_STAR1_HET and used jointly).CYP3A5_STAR1_HET
(see that entry’s Notes for the three-level decomposition rationale).
Ratified canonically on 2026-05-20 alongside the Passey 2011 tacrolimus
extraction.CYP2C9_PM_IM). Time-fixed per subject (germline
genotype-derived phenotype).CYP2C9_PM_IM).CYP2C9 phenotype – Jeong 2022 (paper Table 1 phenotype
classification: *1/*1 = EM; *1/*3,
*1/*13 = IM).Jeong_2022_torsemide.R
(linear-deviation effect on apparent clearance and inter-compartmental
clearance: CL/F = tvCL/F * (1 + 0.510 * CYP2C9_EM) and
Q/F = tvQ/F * (1 + 0.365 * CYP2C9_EM); CYP2C9 extensive
metabolizers have 51% higher apparent clearance and 36.5% higher
apparent inter-compartmental clearance than intermediate metabolizers;
Jeong 2022 Table 4 final Pop-PK model),
Kleideiter_2017_cebranopadol.R (additive log shifts on CL:
e_cyp2c9em_lcl = log(82.4 / 74.3) = 0.1037 and
e_cyp2c9pmim_lcl = log(58.7 / 74.3) = -0.2353; reference
category is unknown phenotype with both indicators = 0, the most common
stratum; paired with CYP2C9_PM_IM),
Kleideiter_2018_cebranopadol.R (multiplicative effect on CL
applied as e_em_cl^CYP2C9_EM with
e_em_cl = 82.4 / 74.3 = 1.109; EM subjects have about +11%
CL vs the model’s CYP2C9 reference, which in this paper is the ‘unknown
phenotype’ pool rather than IM/PM – only 38.3% of the analysis cohort
had a known CYP2C9 phenotype, so the 0-level here pools unknown subjects
together with PIM subjects whose effect is carried separately by the
sibling CYP2C9_PIM canonical; Kleideiter 2018 Table
13).CYP3A5_EXPR
precedent above (functional-allele-carrier = 1) rather than the
SNP_<GENE>_<RSID> mutant-presence pattern,
because (a) clinical CYP2C9 phenotype is reported as EM / IM / PM and
the canonical name should mirror the clinically meaningful axis, and (b)
Jeong 2022 chose IM as the model reference, so the EM-equals-1
orientation preserves the paper’s reported coefficient signs and
typical-value parameters directly. The canonical name pools
reduced-function alleles (2, 3, *13, etc.) into the
0 category because the population-PK literature typically
does not separately resolve them; future papers that distinguish IM from
PM should propose a paired companion (CYP2C9_PM) so the
three-level EM / IM / PM phenotype can be encoded with two binary
indicators on the SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM pattern. Distinct from
CYP2D6 / CYP3A4 (continuous-activity-score
canonicals): use CYP2C9_EM when the source paper reports a
discrete phenotype label; use a continuous CYP2C9 canonical
(not yet ratified; see TODO on CYP3A4) when a future paper
reports a probe-derived activity number. Ratified canonically on
2026-05-17 alongside the Jeong 2022 torsemide extraction.CYP2C9_EM
carries the 1 = EM indicator; subjects with unknown CYP2C9 status appear
with both CYP2C9_PM_IM = 0 and
CYP2C9_EM = 0.CYP2C9_EM, both indicators
= 0 indicates the unknown-phenotype stratum and
CYP2C9_EM = 1 with CYP2C9_PM_IM = 0 indicates
the EM stratum.CYP2C9 – Kleideiter 2017 (paper Table 13 row “CYP2C9
poor and intermediate metabolizers 58.7 L/h”).CYP2C9_PIM – used in
Kleideiter_2018_cebranopadol.R (the erratum-corrected
re-extraction; same pooled poor-or-intermediate-metabolizer semantics,
abbreviated PIM).Kleideiter_2017_cebranopadol.R (additive log shift on CL:
e_cyp2c9pmim_lcl = log(58.7 / 74.3) = -0.2353; reduced
apparent clearance vs the unknown-phenotype reference; paired with
CYP2C9_EM to form the three-level stratification),
Kleideiter_2018_cebranopadol.R (multiplicative CL ratio
58.7 / 74.3 = 0.790 applied as ratio^CYP2C9_PIM; paired
with CYP2C9_EM, both 0 = the unknown-phenotype reference,
Kleideiter 2018 Table 13).CYP2C9_EM for the
three-level “unknown / EM / PM-IM” stratification used in the Kleideiter
cebranopadol model, where the unknown-phenotype reference (both
indicators = 0) is the most common category (only 38.3% of the analysis
cohort had a known CYP2C9 phenotype). PM and IM are pooled because the
source covariate analysis does not separately resolve them; downstream
papers that distinguish PM from IM should register a paired
CYP2C9_PM canonical and split this group, encoding the
three-level EM / IM / PM phenotype with two binary indicators on the
SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM pattern.
Distinct from the CYP2C9_EM canonical’s Jeong-2022 use
(where 0 = IM/PM, no unknown subjects, and the 0-level directly carries
the PM/IM phenotype): in the Kleideiter cohort the 0-level of both
CYP2C9_EM and CYP2C9_PM_IM pools all subjects
not assigned that specific phenotype label (including the ‘unknown’
fraction), so per-model covariateData notes must document
the reference complement. Ratified canonically on 2026-05-25 alongside
the Kleideiter 2017 cebranopadol extraction.CYP2B6_SM and
CYP2B6_USM). Time-fixed per subject (germline
genotype-derived phenotype). EM (516GG | 983TT) is the reference
category when all three of CYP2B6_IM,
CYP2B6_SM, and CYP2B6_USM are 0.metabolizer status IM / IM – Bienczak 2016
(paper Methods ‘Covariate effects’ paragraph 2, Results ‘Population
pharmacokinetics’ paragraph 3, and Table 3 final estimates; categorical
metabolizer indicator entered as multiplicative effect on intrinsic
clearance CLint with EM reference).Bienczak_2016_nevirapine.R (multiplicative log-additive
effect on CLint:
cl_meta <- exp(e_cyp2b6_im_cl * CYP2B6_IM + e_cyp2b6_sm_cl * CYP2B6_SM + e_cyp2b6_usm_cl * CYP2B6_USM),
with e_cyp2b6_im_cl = log(1 - 0.17) = -0.186, giving the
17% lower CLint reported in Bienczak 2016 Table 3 / Results ‘Population
pharmacokinetics’ paragraph 3).CYP2B6_SM
and CYP2B6_USM together encode the four-level EM / IM / SM
/ USM phenotype with three binary columns (EM = all three zero); follows
the dummy-coding pattern used elsewhere in the register for multi-level
categoricals (e.g. RACE_*, HEPIMP_MILD /
HEPIMP_SEV / HEPIMP_MODSEV). The IM grouping
pools the two distinct genotypes (516GT | 983TT and 516GG | 983TC) into
a single indicator because Bienczak 2016 Results ‘Population
pharmacokinetics’ paragraph 3 reports that ‘Using six rather than four
516G>T | 983T>C SNP-vector metabolizer groups reduced OFV by only
5 points (df = 2, P = 0.08) and was therefore not used.’ Distinct from
the continuous CYP3A4 activity-score canonical (which
captures probe-substrate-derived activity rather than SNP-vector
phenotype) and from CYP3A5_EXPR (binary expresser indicator
built on a single rs776746 genotype). Ratified canonically on 2026-05-21
alongside the Bienczak 2016 nevirapine extraction.CYP2B6_IM and CYP2B6_USM). Time-fixed per
subject (germline genotype-derived phenotype). EM (516GG | 983TT) is the
reference category when all three of CYP2B6_IM,
CYP2B6_SM, and CYP2B6_USM are 0.metabolizer status SM / SM – Bienczak 2016
(paper Methods ‘Covariate effects’ paragraph 2, Results ‘Population
pharmacokinetics’ paragraph 3, and Table 3 final estimates; categorical
metabolizer indicator entered as multiplicative effect on intrinsic
clearance CLint with EM reference).Bienczak_2016_nevirapine.R (multiplicative log-additive
effect on CLint:
cl_meta <- exp(e_cyp2b6_im_cl * CYP2B6_IM + e_cyp2b6_sm_cl * CYP2B6_SM + e_cyp2b6_usm_cl * CYP2B6_USM),
with e_cyp2b6_sm_cl = log(1 - 0.50) = -0.693, giving the
50% lower CLint reported in Bienczak 2016 Table 3 / Results ‘Population
pharmacokinetics’ paragraph 3).CYP2B6_IM
and CYP2B6_USM together encode the four-level EM / IM / SM
/ USM phenotype with three binary columns (EM = all three zero). See
CYP2B6_IM Notes for the SNP-vector pooling rationale and
the link back to the dummy-coding pattern used elsewhere in the
register. Ratified canonically on 2026-05-21 alongside the Bienczak 2016
nevirapine extraction.CYP2B6_IM and
CYP2B6_SM). Time-fixed per subject (germline
genotype-derived phenotype). EM (516GG | 983TT) is the reference
category when all three of CYP2B6_IM,
CYP2B6_SM, and CYP2B6_USM are 0.metabolizer status USM / USM – Bienczak
2016 (paper Methods ‘Covariate effects’ paragraph 2, Results ‘Population
pharmacokinetics’ paragraph 3, and Table 3 final estimates; categorical
metabolizer indicator entered as multiplicative effect on intrinsic
clearance CLint with EM reference).Bienczak_2016_nevirapine.R (multiplicative log-additive
effect on CLint:
cl_meta <- exp(e_cyp2b6_im_cl * CYP2B6_IM + e_cyp2b6_sm_cl * CYP2B6_SM + e_cyp2b6_usm_cl * CYP2B6_USM),
with e_cyp2b6_usm_cl = log(1 - 0.68) = -1.139, giving the
68% lower CLint reported in Bienczak 2016 Table 3 / Results ‘Population
pharmacokinetics’ paragraph 3).CYP2B6_IM
and CYP2B6_SM together encode the four-level EM / IM / SM /
USM phenotype with three binary columns (EM = all three zero). Bienczak
2016 is the first study to quantify the USM phenotype on nevirapine
clearance (cohort prevalence 0.6%; Table 2 row 4); the rs28399499
(983T>C) loss-of-function allele is essentially absent from
European-ancestry populations but reaches appreciable frequency in
sub-Saharan African cohorts, so models on European or East Asian
populations may report only EM / IM / SM (with CYP2B6_USM
identically zero across the dataset). See CYP2B6_IM Notes
for the SNP-vector pooling rationale. Ratified canonically on 2026-05-21
alongside the Bienczak 2016 nevirapine extraction.CYP3A4 continuous-activity-score
canonical above: CONMED_CYP3A4_INH captures
concomitant-medication exposure (a drug-drug-interaction indicator), not
intrinsic enzyme activity. Use this canonical when the source paper
enters CYP3A4-inhibitor coadministration into the popPK model as a
binary indicator, regardless of which inhibitor strengths (strong /
moderate / weak) the paper pools into the 1 category.CYP3A4_INH – prior canonical name (pre-2026-06-19
CONMED_ prefix standardization).CYP3AI,
CYP3A4I, CYP3AINH, or a free-text
concomitant-medication indicator. Document the source-column name
per-model in
covariateData[[CONMED_CYP3A4_INH]]$source_name.Yassen_2025_asundexian.R (proportional-shift effect on
CL/F: (1 + e_cyp3a4_inh_cl * CONMED_CYP3A4_INH) with
e_cyp3a4_inh_cl = -0.0531; the asundexian dataset pools
weak + moderate CYP3A4 inhibitors into the
CONMED_CYP3A4_INH = 1 category because strong inhibitors
were a Phase II exclusion criterion).covariateData[[CONMED_CYP3A4_INH]]$notes must document
which inhibitor strengths (strong / moderate / weak) and which specific
drug examples are pooled into the CONMED_CYP3A4_INH = 1
category, since inclusion criteria vary by study. Future models that
need stratified encoding (separate strong / moderate / weak indicators)
should register companion canonicals
(e.g. CONMED_CYP3A4_INH_STRONG,
CONMED_CYP3A4_INH_MOD, CONMED_CYP3A4_INH_WEAK)
rather than overloading CONMED_CYP3A4_INH. The
complementary CYP3A4-inducer indicator follows the same pattern as
CONMED_CYP3A4_IND. Ratified canonically on 2026-05-08
alongside the Yassen 2025 asundexian extraction. Renamed from
CYP3A4_INH to CONMED_CYP3A4_INH on 2026-06-19
per the canonical-register standardization audit (operator decision: the
indicator captures a concomitant medication, so it belongs in the
CONMED_<concept> family alongside
CONMED_PROBENECID, CONMED_AZOLE, etc.).CONMED_CYP3A4_INH that
stratifies by cumulative exposure intensity (rather than by inhibitor
strength). Distinct from CONMED_CYP3A4_INH_LO which
captures the complementary “lower-exposure / unclassified” category. The
two indicators are mutually exclusive per subject (a subject is either
in CONMED_CYP3A4_INH_HI = 1, CONMED_CYP3A4_INH_LO = 1, or both = 0);
when both are 0 the subject was either never on a CYP3A4 inhibitor or
was on one only briefly.CONMED_CYP3A4_INH_LO = 1).CYPINH = 4 (strong, >=50%) or
CYPINH = 6 (moderate, >=50%) or CYPINH = 8
(weak, >=50%) in the FIDELIO-DKD popPK NONMEM control stream’s
9-level CYPINH categorical variable; used in
vandenBerg_2021_finerenone.R.vandenBerg_2021_finerenone.R (multiplicative effect
0.951^CONMED_CYP3A4_INH_HI on CL/F AND on F1 inversely, per
paper’s NONMEM control stream).CONMED_CYP3A4_INH (any inhibitor coadministration, no
exposure-intensity stratification). The pooled strong/moderate/weak
grouping at the >=50% exposure threshold matches the FIDELIO-DKD
analysis’s covariate-engineering choice (the study did not separate
strong from moderate from weak inhibitors but did separate
high-cumulative-exposure from lower-exposure). The accompanying
inhibitor-strength-stratified canonicals
CONMED_CYP3A4_INH_STRONG /
CONMED_CYP3A4_INH_MOD / CONMED_CYP3A4_INH_WEAK
mentioned in the CONMED_CYP3A4_INH Notes follow a different
stratification axis (by strength, ignoring cumulative duration) and
would coexist with the HI / LO duration-based pair without
conflict.CONMED_CYP3A4_INH_HI
that captures all CYP3A4-inhibitor exposure other than the
high-cumulative-exposure-and-known-strength category. The two indicators
are mutually exclusive per subject.CONMED_CYP3A4_INH_HI = 1).CYPINH IN (1, 2, 3, 5, 7) in the FIDELIO-DKD popPK
NONMEM control stream’s 9-level CYPINH categorical variable
– corresponding to unclassified <50% (1), unclassified >=50% (2),
strong <50% (3), moderate <50% (5), weak <50% (7). Used in
vandenBerg_2021_finerenone.R.vandenBerg_2021_finerenone.R (multiplicative effect
0.996^CONMED_CYP3A4_INH_LO on CL/F AND on F1 inversely, per
paper’s NONMEM control stream).CONMED_CYP3A4_INH_HI. The unclassified-inhibitor category
pools any concomitant medication the source dataset’s
medication-classification logic could not assign to a strong / moderate
/ weak tier (typically due to ambiguous drug-coding); these subjects
would otherwise be lost to analysis if the encoding required a strength
tier. Combining unclassified + <50%-exposure into a single
“lower-impact” indicator was the FIDELIO-DKD analysis’s choice; the
effect size (0.996) is much smaller than the high-exposure category
(0.951).CONMED_CYP3A4_INH; both capture
concomitant-medication exposure (a drug-drug-interaction indicator), not
intrinsic enzyme activity (distinct from the CYP3A4
continuous-activity-score canonical above). Use this canonical when the
source paper enters CYP3A4-inducer coadministration into the popPK model
as a binary indicator, regardless of which inducer strengths (strong /
moderate / weak) the paper pools into the 1 category.CYP3A4_IND – prior canonical name (pre-2026-06-19
CONMED_ prefix standardization).CYP3AIND,
CYP3A4IND, INDU, INDUCER, or a
free-text concomitant-medication indicator. Document the source-column
name per-model in
covariateData[[CONMED_CYP3A4_IND]]$source_name.Gupta_2016_lenvatinib.R (multiplicative power-form effect
on CL/F: 1.30^CONMED_CYP3A4_IND with
e_cyp3a4_ind_cl = log(1.30) ~ 0.262; the Gupta dataset
pools any concomitant CYP3A4 inducer reported in the per-subject
medication log into the CONMED_CYP3A4_IND = 1 category,
with n = 19 (2.4%) of the 779-subject pooled cohort flagged
positive).covariateData[[CONMED_CYP3A4_IND]]$notes must document
which inducer strengths (strong / moderate / weak) and which specific
drug examples are pooled into the CONMED_CYP3A4_IND = 1
category, since inclusion criteria vary by study. Future models that
need stratified encoding (separate strong / moderate / weak indicators)
should register companion canonicals
(e.g. CONMED_CYP3A4_IND_STRONG,
CONMED_CYP3A4_IND_MOD, CONMED_CYP3A4_IND_WEAK)
rather than overloading CONMED_CYP3A4_IND. Sibling
canonical to CONMED_CYP3A4_INH. Ratified canonically
alongside the Gupta 2016 lenvatinib extraction. Renamed from
CYP3A4_IND to CONMED_CYP3A4_IND on 2026-06-19
per the canonical-register standardization audit (operator decision: the
indicator captures a concomitant medication, so it belongs in the
CONMED_<concept> family).covariateData[[APOE4_COUNT]]$notes.APOE4C – used directly in
Conrado_2014_alzheimer.R. The “C” suffix in the source
distinguishes the cleaned continuous APOE-epsilon4 count column (0 / 1 /
2 with unknown recoded to the population mean) from the
upstream raw APOE4 column (0 = non-carrier, 1 =
heterozygous, 2 = homozygous, 3 = unknown).Conrado_2014_alzheimer.R (centring 0.72; multiplicative
effect on baseline ADAS-Cog and on disease-progression slope:
factor = 1 + e * (APOE4_COUNT - 0.72) with
e_blapoe4 = 0.0372 on baseline and
e_slapoe4 = 0.195 on slope).APOE4_CARRIER) rather than overloading
APOE4_COUNT. The unknown category (often
recorded as APOE4 = 3 in CDISC datasets) is conventionally
recoded by the source paper to the population-mean count to avoid
dropping subjects; document the recoding rule used per-model. Ratified
canonically on 2026-05-06 alongside the Conrado 2014 DDMORE
extraction.0 category
because the Horita 2018 source paper found no significant differences in
t1/2, CL/F, or AUC0-8 between rapid and intermediate genotypes and
combined them as the “nonslow” group; this pooling is the standard
convention in the NAT2-aware antituberculosis-isoniazid popPK
literature._SLOW naming,
and the source-paper-reported coefficient signs map back to
typical-value parameters via (1 - NAT2_SLOW) selection
inside model() where needed.NAT2 (categorical with values "slow" /
"intermediate" / "rapid" or 0 /
1 / 2): derive
NAT2_SLOW = as.integer(NAT2 == "slow") (or
as.integer(NAT2 == 0) depending on the source’s level
coding); the intermediate and rapid levels collapse to NAT2_SLOW =
0.NAT2_SS (slow-vs-not-slow indicator already in source
datasets) – same orientation as the canonical, no transformation.ACETYL_SLOW (slow-acetylator indicator) – same
orientation as the canonical, no transformation.Horita_2018_isoniazid.R (selects between two typical-value
clearances via
lcl_slow * NAT2_SLOW + lcl_nonslow * (1 - NAT2_SLOW) and
pairs each typical value with its own IIV variance; reproduces the
source paper’s separate CL/F slow = 4.44 L/h and
CL/F nonslow = 8.08 L/h typical-value estimates with
separate omegas 0.105 and 0.230 respectively).NAT2_RAPID) so the three-level phenotype can be
encoded with two binary indicators on the SLCO1B1_HAP15_HET
/ SLCO1B1_HAP15_HOM and CYP3A5_STAR1_HET /
CYP3A5_STAR1_HOM patterns. The _SLOW
orientation (slow = 1) follows the clinically meaningful axis (slow
acetylators are the at-risk group for isoniazid hepatotoxicity and the
higher-AUC group for treatment outcomes), paralleling the
CYP2D6_PM = 1 orientation for the poor-metabolizer end of
the CYP2D6 phenotype spectrum. Distinct from any genotype-string column
(which carries the raw allele information); NAT2_SLOW
captures the derived metabolic phenotype only. Ratified canonically on
2026-05-26 alongside the Horita 2018 isoniazid extraction.FCGR3A (genotype string, e.g., "V/V" /
"V/F" / "F/F"): derive
FCGR3A_VV = as.integer(FCGR3A == "V/V").rs396991 (raw allele coding, often "AA" /
"AC" / "CC" or "GG" /
"GT" / "TT" depending on assay strand): map V
allele -> 1, F allele -> 0 with the assay-specific allele
convention; derive
FCGR3A_VV = as.integer(genotype is V-homozygous).Aguiar_2021_ustekinumab.R (Aguiar 2021 Table 2 footnote;
logit-scale effect on subcutaneous bioavailability F: 88.8% in V/V vs
71.0% in V/F + F/F).covariateData[[FCGR3A_VV]]$notes. Future models that use a
recessive (F/F vs V/* combined) or codominant (additive 0/1/2) coding
should register a separate canonical (e.g., FCGR3A_FF,
FCGR3A_VCT for V-allele count) rather than overloading
FCGR3A_VV.DsbAL (three-level column 0 = G/G, 1 = G/T, 2 = T/T) –
used in Oniki_2018_bmi.R; derive
DSBAL_TT = as.integer(DsbAL == 2). G/G and G/T are pooled
because the T/T state is the functional minor-allele genotype most
strongly associated with elevated BMI.Oniki_2018_bmi.R
(additive +1.5 kg/m^2 shift on the typical BMI for T/T carriers vs the
pooled G/G-or-G/T reference, Oniki 2018 Eq. 1).DSBAL_GT indicator on the
SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM
precedent. Ratified canonically alongside the Oniki 2018 BMI
extraction.PNPLA3_GG to encode the three-level rs738409 genotype with
two binary indicators (C/C is the reference when both are 0). Time-fixed
per subject (germline genotype).PNPLA3_GG is also 0).PNPLA3 (three-level column 0 = C/C, 1 = C/G, 2 = G/G) –
used in Oniki_2018_nafld_risk.R; derive
PNPLA3_CG = as.integer(PNPLA3 == 1).Oniki_2018_nafld_risk.R (multiplicative factor 0.761 on the
(BMI50 - 17) half-saturation offset of the logit-of-NAFLD sigmoid for
C/G heterozygotes vs the C/C reference, Oniki 2018 Eq. 4 / Figure 2c;
closer to 1 than the G/G factor, consistent with an additive allele-dose
effect).PNPLA3_GG (homozygote) following the
SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM
paired-binary precedent for a three-level genotype where each stratum
carries a distinct typical-value effect. Ratified canonically alongside
the Oniki 2018 NAFLD-risk extraction.PNPLA3_CG to encode the three-level rs738409 genotype with
two binary indicators (C/C is the reference when both are 0). Time-fixed
per subject (germline genotype).PNPLA3_CG is also 0).PNPLA3 (three-level column 0 = C/C, 1 = C/G, 2 = G/G) –
used in Oniki_2018_nafld_risk.R; derive
PNPLA3_GG = as.integer(PNPLA3 == 2).Oniki_2018_nafld_risk.R (multiplicative factor 0.592 on the
(BMI50 - 17) half-saturation offset of the logit-of-NAFLD sigmoid for
G/G homozygotes vs the C/C reference, Oniki 2018 Eq. 4 / Figure
2c).PNPLA3_CG; see PNPLA3_CG notes
for the joint three-level usage and reference category. Ratified
canonically alongside the Oniki 2018 NAFLD-risk extraction.UGT2B7 211GG / UGT2B7-211GG /
211GG – used in Yu_2017_mycophenolic_acid.R
(paper Table 5 column headers; one of three binary group-membership
indicators reconstructed from the paper’s ordinal column “UGT2B7
genotype”).Yu_2017_mycophenolic_acid.R (one of three UGT2B7-211 binary
indicators that reconstruct the paper’s ordinal column
UGT2B7 genotype in {GT=1, GG=2, TT=3}; the model() block
computes
ugt2b7_211_code = UGT2B7_211GT * 1 + UGT2B7_211GG * 2 + UGT2B7_211TT * 3
and applies V1/F = 14.7 + 7.72 * ugt2b7_211_code per Yu 2017 Table 3 /
page 1574 final-model formula).UGT2B7 211GT / UGT2B7-211GT /
211GT – used in Yu_2017_mycophenolic_acid.R
(paper Table 5).Yu_2017_mycophenolic_acid.R (one of three UGT2B7-211 binary
indicators; see UGT2B7_211GG notes for the paper’s
ordinal-code reconstruction).UGT2B7_211GG and UGT2B7_211TT; see
UGT2B7_211GG notes for variant biology and allele-frequency
context. Ratified canonically on 2026-06-03 alongside the Yu 2017
mycophenolic acid extraction.UGT2B7 211TT / UGT2B7-211TT /
211TT – used in Yu_2017_mycophenolic_acid.R
(paper Table 5).Yu_2017_mycophenolic_acid.R (one of three UGT2B7-211 binary
indicators; see UGT2B7_211GG notes for the paper’s
ordinal-code reconstruction).UGT2B7_211GG and UGT2B7_211GT; see
UGT2B7_211GG notes for variant biology and allele-frequency
context. Ratified canonically on 2026-06-03 alongside the Yu 2017
mycophenolic acid extraction.ADA (semantically “ever positive”) – used in
Zhu_2017_lebrikizumab.R. When translating from a paper that
uses ADA as “ever positive,” verify the time-frame matches
ADA_POS semantics before renaming.ADA (time-varying positivity, primary covariate in Xu
2019) – used in Xu_2019_sarilumab.R.NAB (neutralizing antibody positive – used in
Petrov_2024_romiplostim.R). Strictly a subset of total
ADA-positive (ADA antibodies that neutralize the drug’s biological
effect). Document per-model when the source assay measured NAB only and
the canonical column thus excludes binding-only ADA.ATAPOSNEW (ADA-positive in the newer/updated-assay
cohort) – used in Suri_2018_brentuximab.R as the
modern-assay arm of a two-era ADA decomposition.ADA_POSNEW (retired intermediate name;
renamed to ADA_POS on 2026-04-29 for consistency across
single- and multi-assay models).Clegg_2024_nirsevimab.R,
Hu_2026_clesrovimab.R,
Petrov_2024_romiplostim.R,
Suri_2018_brentuximab.R (multi-assay; paired with
ADA_POSOLD and ADA_MISSING;
cl *= (1 + 0.125 * ADA_POS)),
Xu_2019_sarilumab.R.ADA_POS represents modern/current-assay positivity;
companion indicators ADA_POSOLD and
ADA_MISSING capture historical-assay-positive and
missing-result sub-groups respectively. All three are mutually
exclusive; reference is ADA-negative (all three = 0).ADA_POS and
ADA_MISSING.ATAPOSOLD – used in
Suri_2018_brentuximab.R.Suri_2018_brentuximab.R (multiplicative additive effect on
ADC clearance: cl *= (1 + 0.177 * ADA_POSOLD)).ADA_POS
(multi-assay form); see that entry’s Notes for the decomposition
rationale. The “newer” vs “older” assay split is paper-specific (Suri
2018 newer assay: sensitivity 23.573 ng/mL, drug tolerance 25 ug/mL;
older assay: sensitivity 4 ng/mL, drug tolerance 3,125 ng/mL).
Time-varying once positive. Ratified canonically on 2026-04-28.ADA_POS and
ADA_POSOLD.ATAMISSING – used in
Suri_2018_brentuximab.R.Suri_2018_brentuximab.R (multiplicative additive effect on
ADC clearance: cl *= (1 + 0.192 * ADA_MISSING)).e_adam_adc_cl indicates ADA-missing
patients are not exchangeable with the ADA-negative reference –
interpret with caution given the missingness mechanism is not random.
Ratified canonically on 2026-04-28.ADA_TITRE, with ADA_TITRE = 1 for
ADA-negative so log_e(1) = 0 cancels a log-linear effect)
and the American-spelling linear-titer convention
(ADA_TITER, with ADA_TITER = 0 for
ADA-negative). The per-model
covariateData[[ADA_TITER]]$description and
notes must state which zero-encoding convention is in force
so the covariate column cannot be misinterpreted.covariateData[[ADA_TITER]]$units.ADA_TITRE – British spelling (reciprocal-dilution
convention; 1 for negative).ADA titre – British spelling long form.ADAT – used in Moein_2022_etrolizumab.R
(American linear-titer convention; 0 for negative).Jackson_2022_ixekizumab.R (reciprocal-dilution reference
convention with ADA_TITER = 1 for negatives and
(1 + coef * log_e(ADA_TITER)) on CL),
Moein_2022_etrolizumab.R (linear-titer convention with
ADA_TITER = 0 for negatives and
exp(theta * ADA_TITER) on CL, per-unit-titer theta =
0.0365), Robbie_2012_palivizumab.R (reciprocal-dilution
values 0/10/20/40/>=80 with category-specific multiplicative effects
per titer bin; 0 = ADA negative reference).ADA_TITRE
(British, 1 = negative) and ADA_TITER
(American, 0 = negative) canonicals were merged on
2026-04-20 into a single general-scope ADA_TITER. The
zero-encoding convention is the load-bearing semantic and must be
documented per-model. Distinct from ADA_POS (binary
presence/absence); when the paper reports both, the final model usually
keeps only one. Imputation rules (LOCF / NOCB / baseline-as-negative)
should be documented per-model.RIC (Dunlap 2025 NM-TRAN convention; binary 0 / 1) –
used directly in Dunlap_2025_tacrolimus.R.Dunlap_2025_tacrolimus.R (Dunlap 2025 Table 2
reduced-covariate-model column; exponential effect on apparent oral
clearance: cl *= 0.63 ^ HCT_COND_RIC, so RIC recipients
have ~37% lower apparent oral tacrolimus clearance than MAC
recipients).covariateData[[HCT_COND_RIC]]$notes. When a future paper
distinguishes a third intensity tier (non-myeloablative, NMA) as a
separate covariate level rather than pooling NMA into RIC, register a
parallel canonical (e.g. HCT_COND_NMA) instead of
overloading HCT_COND_RIC. Scope: specific because the
column is meaningful only for allo-HCT recipients. Ratified canonically
on 2026-05-09 alongside the Dunlap 2025 tacrolimus extraction.DISEXT_EP and
DISEXT_OTHER,
DISEXT_EP = 0 AND DISEXT_OTHER = 0 corresponds to the
left-sided-colitis reference group; in papers that use a single binary
indicator for extensive colitis, the reference is pooled non-extensive
(left-sided + any other extension).EXTCOL – used in Faelens_2021_infliximab.R
(binary 0/1 for extensive colitis at baseline; no separate “other”
category).DISEXT column in the source
(levels: left-sided colitis, extensive/pancolitis, other):
DISEXT_EP = as.integer(DISEXT == "extensive/pancolitis").Moein_2022_etrolizumab.R (paired with
DISEXT_OTHER; multiplicative effect on CL, +8.2%
vs. left-sided colitis), Faelens_2021_infliximab.R
(single-binary encoding; multiplicative fold-change on V of 1.25 when
DISEXT_EP = 1).DISEXT_OTHER when the source paper decomposes a three-level
disease-extension categorical (left-sided / extensive-pancolitis /
other) into two indicators; the pairing is paper-specific and not
required. Promoted from scope: specific to scope: general on 2026-04-27
because the binary “extensive colitis vs not” semantics generalize
across UC popPK papers regardless of whether the original dataset
additionally distinguished an “other” disease-extension category.DISEXT_EP = 0).DISEXT column:
DISEXT_OTHER = as.integer(DISEXT == "other").Moein_2022_etrolizumab.R (multiplicative effect on CL, +18%
vs. left-sided colitis; large uncertainty due to 2% prevalence).DISEXT_EP; together
they encode the three-level disease-extension categorical.PTAX – used in vanHasselt_2015_eribulin.R
(van Hasselt 2015 paper notation; binary 0 / 1 for prior docetaxel
pretreatment).vanHasselt_2015_eribulin.R (multiplicative effect on
baseline serum PSA:
psa0 = exp(lpsa0 + etalpsa0) * e_prior_taxane_psa0^PRIOR_TAXANE
with e_prior_taxane_psa0 = 3.23 – prior-taxane patients
have ~3.2x higher baseline PSA than taxane-naive patients, consistent
with more advanced disease at study entry).PRIOR_TAXANE_DAYS
when a continuous duration-of-pretreatment is also relevant (e.g., van
Hasselt 2015 uses both: PTAX on PSA0 and NTRT on KD). Distinct from
CONMED_* (which is concomitant medication during the study,
not pretreatment history) and from generic PRIOR_*
chemotherapy indicators (which would warrant a separate canonical when a
paper differentiates by drug class rather than collapsing to taxanes).
Scope: specific because the population semantics (CRPC) and the “any
prior taxane” pooling are tied to van Hasselt 2015; future papers that
distinguish per-drug pretreatment (docetaxel vs paclitaxel vs
cabazitaxel separately) should register parallel canonicals rather than
overloading PRIOR_TAXANE.PRIOR_TAXANE = 0). Time-invariant
within a subject (records the pretreatment history at baseline).(1 + PRIOR_TAXANE_DAYS / 720) ^ e_prior_taxane_days_<param>.
The +1 inside the bracket makes the covariate effect
collapse to a multiplier of 1 for taxane-naive patients
(PRIOR_TAXANE_DAYS = 0) regardless of the estimated exponent.NTRT – used in vanHasselt_2015_eribulin.R
(van Hasselt 2015 paper notation; cumulative number of days of prior
taxane treatment).vanHasselt_2015_eribulin.R (van Hasselt 2015 Eq. 4 power
covariate on drug PSA inhibition rate KD0:
kd0 = exp(lkd0 + etalkd0) * (1 + PRIOR_TAXANE_DAYS / 720)^e_prior_taxane_days_kd0
with e_prior_taxane_days_kd0 = -4.00 – KD0 decreases with
longer prior-taxane exposure, encoding cross-resistance between
docetaxel and eribulin via the shared microtubule-inhibition
mechanism).PRIOR_TAXANE
(binary). The paper also considered an alternative continuous
parameterisation in cycles of prior taxane (NCYCL, median
30 cycles) which was deemed slightly less informative (dOFV = -8 for
NCYCL vs -10 for NTRT) and was not retained in the final model. If a
future model needs the cycle-count form, register a parallel canonical
(e.g. PRIOR_TAXANE_CYCLES) rather than overloading this
one. Scope: specific because the 720-day normalisation reference is tied
to the van Hasselt 2015 study population (post-docetaxel mCRPC
patients).T_DIAG_T1D
canonical if registered in the future).(T_DIAG_DIAB / 2)^theta on the model-predicted
baseline FPG, on Gmax (the maximal SGLT2-inhibition-driven
FPG-reduction), and on kHbA1cout. Reference value 2 years per Baron 2016
Tables S3 and S4.DUR – used in Baron_2016_empagliflozin.R
(Baron 2016 Methods column “duration of diabetes”; PK/PD dataset
categorisation reported as <1 year / 1-5 years / >5 years in Table
S2 with 58.5% of patients in the >5 years group).Baron_2016_empagliflozin.R (power-form effects centred at 2
years on BFPG: (T_DIAG_DIAB / 2)^0.0512, on Gmax:
(T_DIAG_DIAB / 2)^0.0117, on kHbA1cout:
(T_DIAG_DIAB / 2)^-0.577).DIS_DIAB indicator – DIS_DIAB
records the comorbidity at baseline (1/0) while T_DIAG_DIAB
quantifies how long that diabetes has been present. The covariate is a
time-since-event under the canonical T_<event>
family. Distinct from T_NUT_SUPP (time on nutritional
supplementation) and T_POST_ECMO (time after ECMO
decannulation) which are different T_<event>
siblings. For a baseline-only continuous power-form effect, a 0-year
subject (newly diagnosed at study entry) yields
(0/2)^theta = 0 – supply at least a small floor value
(e.g. 0.1 year) for newly-diagnosed subjects in simulation; the floor
convention should be documented in the per-model
covariateData[[T_DIAG_DIAB]]$notes. Ratified canonically on
2026-06-24 alongside the Baron 2016 empagliflozin extraction.Lu_2017_polatuzumab_neuropathy.R.Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on
the grade >= 2 peripheral neuropathy hazard via
exp(theta_baselinePN * BL_PN_GR1) with
theta_baselinePN = -0.222, SE 0.324 – not detectable given
the large SE; the paper reports a sensitivity analysis in which this
indicator was replaced by a broader “history of prior PN” indicator with
similarly inconclusive results).PREV_AE_SCORE (the time-varying
ordinal Markov-state AE-score covariate from Girard 2012 pimasertib,
which conditions a current-observation outcome on the previous
observation’s grade); BL_PN_GR1 is a time-fixed baseline
indicator that does not update during the analysis window. Distinct from
DIS_DPN (painful diabetic peripheral polyneuropathy
disease-state indicator from chronic-pain cohorts); DIS_DPN
flags a diabetes-related primary diagnosis used as a stratification
covariate, while BL_PN_GR1 flags a pre-existing low-grade
adverse-event status from any prior cause in an oncology trial. When a
future paper uses a different baseline-PN grade threshold (e.g., grade
<= 2 permitted) or a different AE category, register a parallel
canonical (e.g., BL_PN_GR2, BL_FATIGUE_GR1)
rather than overloading this one. Ratified canonically on 2026-06-24
alongside the Lu 2017 polatuzumab vedotin TTE extraction.covariateData[[PCSK9]]$units if a different unit –
typically nM – is used in a given model; conversion uses a PCSK9
molecular weight of ~72 kDa, so 1 nM ~= 72 ng/mL).(PCSK9 / ref)^exponent. Reference values observed: 425
ng/mL (= 5.9 nM) in Kuchimanchi_2018_evolocumab.R
(population median).Kuchimanchi_2018_evolocumab.R (power exponent 0.194 on
Vmax: Vmax * (PCSK9/425)^0.194).Canonical pattern:
SNP_<GENE>_<RSID>. Use one binary
indicator per SNP genotype that the source paper tests as a model
covariate. The SNP_ prefix makes the category unambiguous;
the gene symbol disambiguates rsIDs grouped by gene; the rsID provides a
globally unique identifier. Encoding follows the most common
pharmacogenomic convention (also used by Papachristos 2020):
1 = mutant allele present (heterozygous or homozygous
mutant); 0 = homozygous wild-type. When a paper uses a
different encoding (e.g., per-allele dosage 0/1/2, dominant
model with mutant homozygotes only, or recessive model), document the
encoding explicitly in covariateData[[<COL>]]$notes
and consider registering a separate canonical name. SNP indicators
default to scope: specific because the parameter on which they act and
the encoded reference category are tied to the source paper’s analysis
plan; promote to general when a second paper ratifies identical
semantics.
GENECAT – Roberts 2016 (paper Results ‘Population
Pharmacokinetic Analysis’ section: the individual-Ka equation Ka_i =
Ka_pop * exp(theta_1 * GENECAT), with GENECAT = 1 for AG / AA and 0 for
GG).Roberts_2016_topotecan.R (exponential effect on oral
topotecan absorption rate constant Ka:
Ka = Ka_pop * exp(1.06 * SNP_ABCG2_RS4148157); AG/AA
carriers have Ka approximately 2.89x higher than GG homozygotes,
corresponding to an observed Cmax approximately 1.7x higher).ABCG2 421A/A – used in
Ueshima_2018_apixaban.R (Ueshima 2018 Methods Eq. of the
final-model CL/F: dichotomous parameter ABCG2 equals 1 for 421A/A and 0
for 421C/C or 421C/A).Ueshima_2018_apixaban.R (multiplicative factor on the
non-renal arm of apparent oral clearance:
cl_nonren = exp(lcl) * e_abcg2_homvar_cl_nonren^SNP_ABCG2_RS2231142_HOM
with e_abcg2_homvar_cl_nonren = 0.341; 421A/A homozygotes
have non-renal CL/F reduced by 65.9% relative to 421C/C or 421C/A; paper
Table 4 theta6 = 0.341).SNP_ABCG2_RS4148157 in two ways: (a) a different SNP
(rs2231142 is the well-characterised coding Q141K variant in exon 5,
whereas rs4148157 is an intronic variant in intron 11 that is in strong
linkage disequilibrium with rs2231142 and used by Roberts 2016 as a
surrogate marker), (b) a different genetic model (recessive 421A/A-only
here vs dominant any-A-allele in Roberts 2016 rs4148157). When a future
paper uses the same recessive-model encoding for rs2231142 the scope may
be promoted from specific to general; when a future paper uses a
dominant-model encoding (any-A-allele = 1) for rs2231142, register a
paired companion canonical SNP_ABCG2_RS2231142_CARRIER
following the SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM precedent so the two genetic-model
orientations remain separate canonicals. The Q141K variant impairs ABCG2
plasma-membrane localisation and function; it is the most commonly
studied ABCG2 pharmacogenetic SNP in popPK literature (substrates
include apixaban, rosuvastatin, sulfasalazine, topotecan, methotrexate).
Ratified canonically on 2026-05-30 alongside the Ueshima 2018 apixaban
extraction.cat – Papachristos 2020 (the paper writes the indicator
as cat in the CL covariate equation; no formal column name
is given in the published narrative).Papachristos_2020_bevacizumab_pk.R,
Papachristos_2020_bevacizumab_qss.R,
Papachristos_2020_bevacizumab_pkpd.R (multiplicative effect
on bevacizumab CL: CL * exp(-0.423 * SNP_ICAM1_RS1799969)
in the PK and PK/PD models;
CL * exp(-0.33 * SNP_ICAM1_RS1799969) in the binding QSS
model – mutant carriers have lower CL and higher trough levels).Notes Table 1 of
the paper). The biological mechanism by which the ICAM1 mutant
slows bevacizumab clearance is unknown; the association is empirical and
may be specific to mCRC.cat1 – Papachristos 2020 (used as the first categorical
indicator in the inter-compartmental clearance equation of the PK model;
no formal column name in the narrative).Papachristos_2020_bevacizumab_pk.R (multiplicative effect
on bevacizumab Q: Q * exp(0.378 * SNP_VEGFA_RS1570360) –
mutant carriers have higher inter-compartmental clearance).cat2 – Papachristos 2020 PK model (second categorical
indicator on Q).cat – Papachristos 2020 binding QSS model (effect on
K_ss and BM0) and PK/PD model (effect on Q).Papachristos_2020_bevacizumab_pk.R (effect on Q: -0.429),
Papachristos_2020_bevacizumab_qss.R (effect on K_ss: +1.22,
on BM0: -0.851), Papachristos_2020_bevacizumab_pkpd.R
(effect on Q: -0.414).SLCO1B1 rs11045819 genotype – Hennig 2015 (paper text;
the source NONMEM control stream is in the unrecovered AAC supplement,
so the formal column name is not on disk).Hennig_2015_rifabutin.R (multiplicative effect on rifabutin
bioavailability F: F * (1 + 0.304 * SNP_SLCO1B1_RS11045819)
– AC carriers have ~30% higher rifabutin F than CC reference; dOFV =
-6.5).ABCB13435 – used in Bisaso_2014_albumin.R
(paper text “ABCB1c.3435C>T mutation” and Figure 3 stratification
“ABCB13435==0 stands for ABCB1c.3435CC while ABCB13435==1 stands for
ABCB1c.3435CT and ABCB1c.3435TT”; same orientation as the
canonical).Bisaso_2014_albumin.R
(multiplicative additive shift on baseline albumin secretion rate Q0:
Q0 = exp(lq0) * (1 + e_snp_abcb1_rs1045642_q0 * SNP_ABCB1_RS1045642)
with e_snp_abcb1_rs1045642_q0 = 0.167; T-carriers have
16.7% higher Q0 than CC wild-type, equivalent to the paper text’s “16%
higher” framing).ABCB1_HAP_TTT
(which is the multi-SNP haplotype across rs1128503 / rs2032582 /
rs1045642 jointly – a different concept even though rs1045642 is one of
the three contributing SNPs); use ABCB1_HAP_TTT when the
source paper reports a phased haplotype, and
SNP_ABCB1_RS1045642 when the source paper reports the
single c.3435C>T SNP alone. Heterozygote and homozygote T-carriers
are pooled in Bisaso 2014 because the TT cohort was n = 1; future
extractions that estimate separate het / hom effects should register
paired SNP_ABCB1_RS1045642_HET and
SNP_ABCB1_RS1045642_HOM indicators following the
SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM
precedent. ABCB1 c.3435C>T is associated with altered P-glycoprotein
expression and has been linked in the literature to predisposition to
ART and rifampicin-based anti-TB drug-induced liver injury (Yimer 2011,
cited in Bisaso 2014 Discussion). Ratified canonically on 2026-05-20
alongside the Bisaso 2014 albumin extraction.*15 reduced-function haplotype (the cis
combination of the 388A>G / rs2306283 and 521T>C / rs4149056
variants; encodes the OATP1B1 N130D + V174A double mutant). 1 = subject
carries exactly one 15 allele (heterozygous: *1a/*15 or
*1b/*15), 0 = otherwise (the union of 15-noncarriers
and *15-homozygotes; the paired indicator SLCO1B1_HAP15_HOM
flags the homozygous group). Time-fixed per subject (germline
haplotype).SLCO1B1_HAP15_HOM flags the 15/15 group.HT – Ide 2009 (paper text Eq. for
Frel = 1 * theta1^HT * theta2^HM where HT = 1
for heterozygotes *1a/*15 and *1b/*15).OATP1B1 phenotype – Jeong 2022 (paper Section 3.2 +
Table 1 phenotype-to-haplotype mapping: ET = 1a/1a +
1a/1b + 1b/1b, IT = 1a/15 + 1b/15,
PT = 15/15; IT subjects map to SLCO1B1_HAP15_HET = 1).Ide_2009_pravastatin.R
(multiplicative effect on relative bioavailability Frel:
Frel = 1.50^SLCO1B1_HAP15_HET * 1.95^SLCO1B1_HAP15_HOM –
15 heterozygotes have 50% higher Frel than 15-noncarriers; dOFV
= 32.2 in backward elimination, p < 0.001),
Jeong_2022_torsemide.R (linear-deviation effect on apparent
central volume V/F:
V/F = tvV/F * (1 + (-0.410) * SLCO1B1_HAP15_HET + (-0.646) * SLCO1B1_HAP15_HOM)
– 15 heterozygotes have 41% lower V/F than 15-noncarriers;
Jeong 2022 Table 4).SLCO1B1_HAP15_HOM
to encode a three-level haplotype categorical (noncarrier / heterozygote
/ homozygote) with *15-noncarrier as the implicit reference
(both indicators = 0). Distinct from the SNP-level canonical
SNP_SLCO1B1_RS11045819 (which encodes only the C>A
variant at a different position; rs11045819 = P155T) and from the 15
component SNPs rs2306283 (388A>G) and rs4149056 (521T>C)
individually: future Ide-style extractions that pool 5 (521T>C
only) with *15 should still record their values under this canonical and
document the pooling rule in
covariateData[[SLCO1B1_HAP15_HET]]$notes. Distribution in
the Ide 2009 cohort of 57 healthy Japanese male volunteers (Table I): 28
noncarriers, 23 heterozygotes, 6 homozygotes; in the Jeong 2022 cohort
of 112 healthy Korean male volunteers (Table 1): 86 ET noncarriers
(76.8%), 23 IT heterozygotes (20.5%), 3 PT homozygotes (2.7%). Ratified
canonically on 2026-05-12 alongside the Ide 2009 extraction; scope
promoted from specific to general on 2026-05-17 alongside the Jeong 2022
torsemide extraction (second model using the same haplotype encoding
across an unrelated drug class, demonstrating the canonical generalizes
beyond statins).*15 reduced-function haplotype. 1 =
subject carries two 15 alleles (homozygous: *15/*15), 0
= otherwise (the union of 15-noncarriers and *15-heterozygotes; the
paired indicator SLCO1B1_HAP15_HET flags the heterozygous
group). Time-fixed per subject (germline haplotype).SLCO1B1_HAP15_HET flags the *15-heterozygote group.HM – Ide 2009 (paper text Eq. for
Frel = 1 * theta1^HT * theta2^HM where HM = 1
for homozygotes *15/*15).OATP1B1 phenotype – Jeong 2022 (paper Section 3.2 +
Table 1 phenotype-to-haplotype mapping: PT = 15/15 corresponds
to SLCO1B1_HAP15_HOM = 1).Ide_2009_pravastatin.R
(multiplicative effect on relative bioavailability Frel:
Frel = 1.50^SLCO1B1_HAP15_HET * 1.95^SLCO1B1_HAP15_HOM –
15 homozygotes have 95% higher Frel than 15-noncarriers; dOFV =
33.7 in backward elimination, p < 0.001),
Jeong_2022_torsemide.R (linear-deviation effect on apparent
central volume V/F:
V/F = tvV/F * (1 + (-0.410) * SLCO1B1_HAP15_HET + (-0.646) * SLCO1B1_HAP15_HOM)
– 15 homozygotes have 64.6% lower V/F than 15-noncarriers;
Jeong 2022 Table 4).SLCO1B1_HAP15_HET
to encode a three-level haplotype categorical (noncarrier / heterozygote
/ homozygote) with *15-noncarrier as the implicit reference
(both indicators = 0). See SLCO1B1_HAP15_HET Notes for the
broader context; population distribution in Ide 2009 was 6 of 57 (10.5%)
homozygotes and in Jeong 2022 was 3 of 112 (2.7%) homozygotes. Ratified
canonically on 2026-05-12 alongside the Ide 2009 extraction; scope
promoted from specific to general on 2026-05-17 alongside the Jeong 2022
torsemide extraction (second model using the same haplotype encoding
across an unrelated drug class).CYP2C9_S2_COUNT and CYP2C9_S3_COUNT to encode
the three loss-of-function-allele dosage form used by Hamberg-family
warfarin models, where the subject’s CL is the sum of per-allele CL
contributions across the two CYP2C9 alleles. The three count columns sum
to 2 for each subject.CYP2C9 (genotype string such as "*1/*1",
"*1/*3"): derive
CYP2C9_S1_COUNT = (length of *1 matches in the genotype string).
The model Xia_2024_warfarin.R carries the source
CYP2C9 genotype string mapped to the three count
columns.Xia_2024_warfarin.R
(per-allele CL contributions:
cl = CYP2C9_S1_COUNT * 0.174 + CYP2C9_S2_COUNT * 0.0879 + CYP2C9_S3_COUNT * 0.0422,
times an age effect).CYP2C9_S5_COUNT etc.) rather
than overloading the existing three counts. Distinct from the
categorical phenotype canonicals CYP3A5_EXPR (binary
expresser) and from continuous-activity scores like CYP3A4
– the count form preserves loss-of-function-allele dosage exactly.
Ratified canonically on 2026-05-16 alongside the Xia 2024 warfarin
extraction.CYP2C9_S1_COUNT and
CYP2C9_S3_COUNT; the three counts sum to 2 for each
subject. The *2 allele (rs1799853, R144C) encodes a reduced-function
CYP2C9 isoform.CYP2C9 (genotype string such as "*1/*2",
"*2/*2", "*2/*3"): derive
CYP2C9_S2_COUNT = (length of *2 matches in the genotype string).Xia_2024_warfarin.R
(per-allele CL contributions; the Xia 2024 Han cohort had no *2
carriers, but the model retains the term for general use across CYP2C9
papers).CYP2C9_S1_COUNT for the
broader rationale. Ratified canonically on 2026-05-16 alongside the Xia
2024 warfarin extraction.CYP2C9_S1_COUNT and
CYP2C9_S2_COUNT; the three counts sum to 2 for each
subject. The *3 allele (rs1057910, I359L) encodes a strongly
reduced-function CYP2C9 isoform and is the dominant CYP2C9
pharmacogenomic risk variant in East-Asian populations (warfarin /
phenytoin sensitivity).CYP2C9 (genotype string such as "*1/*3",
"*3/*3"): derive
CYP2C9_S3_COUNT = (length of *3 matches in the genotype string).Xia_2024_warfarin.R
(per-allele CL contributions; 5.7% of the Han cohort were 1/3
heterozygous per Xia 2024 Table 1).CYP2C9_S1_COUNT for the
broader rationale. Ratified canonically on 2026-05-16 alongside the Xia
2024 warfarin extraction.2 - VKORC1_1639G_COUNT.ec50_typ = VKORC1_1639G_COUNT * ec50_per_G + (2 - VKORC1_1639G_COUNT) * ec50_per_A.
Distribution in the Xia 2024 Han cohort (Table 1): AA 80.3%, GA 18.7%,
GG 0.9% (G allele frequency ~10%, consistent with East-Asian
populations).VKORC1 (genotype string such as "AA",
"GA", "GG" or "1639AA" etc.):
derive
VKORC1_1639G_COUNT = (count of G in the two-letter genotype).Xia_2024_warfarin.R
(per-allele EC50 contributions, re-estimated for the Han Chinese cohort:
4.3 mg/L per G allele, 1.14 mg/L per A allele).model() into mutually-exclusive heterozygous
(count == 1) and homozygous (count == 2)
indicators, each multiplied by an independently estimated CL/F shift
(-0.5 L/h for GT, -1.3 L/h for TT relative to the GG reference). In the
Olagunju 2018 efavirenz model the count is summed together with
SNP_CYP2B6_RS28399499_C_COUNT to derive a composite
metaboliser status (n_variant == 0 fast, == 1
intermediate, >= 2 slow), and the per-group CL/F is
encoded as log-ratio multiplicative shifts on the fast-metaboliser
reference.X_516GT / X_516TT – Schipani 2011 (paper
Table 2 / final-model equation
TVCL = theta0 + theta_BW * (BW - 72.5) + theta_516GT * X_516GT + theta_516TT * X_516TT + theta_983TC * X_983TC;
the two indicators are mutually exclusive across the three 516
genotypes, so the canonical count column reconstructs them
deterministically via
X_516GT = as.integer(SNP_CYP2B6_RS3745274_T_COUNT == 1) and
X_516TT = as.integer(SNP_CYP2B6_RS3745274_T_COUNT == 2)).CYP2B6 516G>T (rs3745274) – Olagunju 2018 (paper
Methods ‘Sample Collection …’ paragraph 1; genotype reported per-allele
and used in combination with rs28399499 to define a composite CYP2B6
metaboliser status).Schipani_2011_nevirapine.R (additive linear shift on CL/F:
cl = exp(lcl) + e_516gt_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 1) + e_516tt_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 2) + ...;
516TT homozygotes have approximately 37% lower CL/F than the GG
reference; Schipani 2011 Table 2),
Olagunju_2018_efavirenz.R (composite-metaboliser-status
encoding: variant alleles from rs3745274 and rs28399499 are summed to
classify subjects as fast / intermediate / slow, with per-group CL/F =
18.0 / 16.1 / 6.24 L/h reported in Olagunju 2018 Table 2).SNP_CYP2B6_RS28399499_C_COUNT) and is
consistently associated with reduced metabolism of nevirapine,
efavirenz, bupropion, and methadone. The count-form encoding is
preferred over paired binary HET / HOM indicators because (a) it follows
the established _COUNT precedent used for
CYP2C9_S{1,2,3}_COUNT and VKORC1_1639G_COUNT,
(b) a single count column captures the underlying genotype without
redundancy, and (c) the model code can deterministically derive either a
linear per-allele effect (* count) or a non-additive HET /
HOM decomposition (* (count == 1),
* (count == 2)) depending on the source paper’s
parameterization. Ratified canonically on 2026-05-21 alongside the
Schipani 2011 nevirapine extraction; scope promoted to general on
2026-05-26 alongside the Olagunju 2018 efavirenz extraction, which uses
the same count column under a composite-metaboliser-status encoding
combining rs3745274 + rs28399499.model() into a heterozygous indicator
(count == 1) multiplied by an estimated CL/F shift (-1.4
L/h relative to the TT reference); the homozygous indicator
(count == 2) is not estimated because no 983CC homozygotes
have been reported in any published cohort. In the Olagunju 2018
efavirenz model the count is summed together with
SNP_CYP2B6_RS3745274_T_COUNT to define a composite CYP2B6
metaboliser status (slow / intermediate / fast); a hypothetical 983CC
subject is still classified as slow (each variant allele contributes to
the composite count) although that substratum is not present in the
fitted cohort.X_983TC – Schipani 2011 (paper Table 2 / final-model
equation; the heterozygous indicator is mutually exclusive with the
TT-homozygous reference, so the canonical count column reconstructs it
via
X_983TC = as.integer(SNP_CYP2B6_RS28399499_C_COUNT == 1)).CYP2B6 983T>C (rs28399499) – Olagunju 2018 (paper
Methods ‘Sample Collection …’ paragraph 1; combined with rs3745274 to
define a composite metaboliser status).Schipani_2011_nevirapine.R (additive linear shift on CL/F:
cl = exp(lcl) + e_983tc_cl * (SNP_CYP2B6_RS28399499_C_COUNT == 1) + ...;
983TC heterozygotes have approximately 40% lower CL/F than the TT
reference; Schipani 2011 Table 2),
Olagunju_2018_efavirenz.R (composite-metaboliser-status
encoding: variant alleles from rs3745274 and rs28399499 are summed to
classify subjects as fast / intermediate / slow; Olagunju 2018 Table
2).SNP_CYP2B6_RS3745274_T_COUNT. Ratified
canonically on 2026-05-21 alongside the Schipani 2011 nevirapine
extraction; scope promoted to general on 2026-05-26 alongside the
Olagunju 2018 efavirenz extraction, which uses the same count column
under a composite-metaboliser-status encoding combining rs3745274 +
rs28399499.CYP2C19_S2_HET /
CYP2C19_S2_HOM indicators (or, where the source uses the
per-allele count form, a CYP2C19_S2_COUNT column following
the CYP2C9_S2_COUNT precedent).CYP2C19*2 – Danielak 2017 (paper Methods ‘Determination
of genetic polymorphisms’ and Table 2 final-model
Effect of CYP2C19*2 on FM (COV) row; PCR-RFLP genotyping
for rs4244285). The Danielak 2017 cohort had no 2/2
homozygous-poor-metabolizers, so heterozygous 1/2 carriers were
pooled into the binary CYP2C19_S2_CARRIER = 1 group with no information
loss.Danielak_2017_clopidogrel.R (linear-deviation effect on the
fraction of clopidogrel metabolised to the active thiol H4:
fm = TVFM * (1 + e_cyp2c19_s2_fm * CYP2C19_S2_CARRIER) with
e_cyp2c19_s2_fm = -0.45; carriers convert 45% less of the
absorbed clopidogrel to the active H4 metabolite, giving a 36.7% lower
predicted AUC of H4 vs non-carriers; Danielak 2017 Table 2 final-model
and Results page 1628).CYP2C19_S17_CARRIER indicator following the same pattern).
The continuous-individual-activity-score consolidation TODO logged on
CYP2D6 line 3321 also applies prospectively to CYP2C19, but
the binary carrier indicator remains the standard discrete encoding used
by most published clopidogrel popPK / PD models. Ratified canonically on
2026-05-20 alongside the Danielak 2017 clopidogrel extraction.*1/*2, *2/*17); 0 = any other CYP2C19
phenotype (EM, UM, PM, or RM). Time-fixed per subject (germline
genotype-derived phenotype). Paired with CYP2C19_PM to
encode the three-level EM/UM (reference, both indicators 0) / IM
(CYP2C19_IM = 1) / PM (CYP2C19_PM = 1)
phenotype with two binary indicators.CYP2C19_PM = 0 is the extensive / ultrarapid-metabolizer
phenotype pool used in Zhao 2018.CYP2C19 IM / IM – Zhao 2018 (paper Table 2
reports F_CYP2C19 IM = 0.449 relative to the EM/UM
reference; genotypes pooled into IM: *1/*2,
*2/*17).Zhao_2018_omeprazole.R
(power-of-binary-indicator multiplicative factor on CLOMZ-M1 formation
clearance: e_cyp2c19_im_kmet_5oh ^ CYP2C19_IM with
e_cyp2c19_im_kmet_5oh = 0.449; IM subjects have ~55% lower
formation clearance of 5-hydroxy-omeprazole than the EM/UM reference;
Zhao 2018 Table 2).CYP2B6_IM /
CYP2B6_SM / CYP2B6_USM three-binary precedent
for multi-level metabolizer phenotypes. The Zhao 2018 cohort pooled
extensive (EM, *1/*1) and ultrarapid (UM,
*1/*17, *17/*17) metabolizers into a single
reference because the typical-value clearance of 5-hydroxy-omeprazole
formation was indistinguishable between the two strata in n = 38 EM/UM
subjects (Zhao 2018 Methods ‘Population pharmacokinetic-pharmacogenetic
modelling’). Future extractions that fit a separate UM coefficient
should register a paired CYP2C19_UM companion indicator
following this pattern. Distinct from CYP2C19_S2_CARRIER
(binary *2-allele carrier indicator used by Danielak 2017
clopidogrel) – CYP2C19_S2_CARRIER pools heterozygous and
homozygous *2 carriers into a single 0/1 contrast, while
CYP2C19_IM resolves the heterozygous *2 (IM)
stratum separately from the homozygous *2/*2 (PM) stratum
that CYP2C19_PM flags. Ratified canonically on 2026-05-25
alongside the Zhao 2018 omeprazole extraction.*2/*2); 0
= any other CYP2C19 phenotype (EM, UM, IM, or RM). Time-fixed per
subject (germline genotype-derived phenotype). Paired with
CYP2C19_IM to encode the three-level EM/UM (reference) / IM
/ PM phenotype with two binary indicators.CYP2C19_IM = 0 is the extensive / ultrarapid-metabolizer
phenotype pool used in Zhao 2018.CYP2C19 PM / PM – Zhao 2018 (paper Table 2
reports F_CYP2C19 PM = 0.125 relative to the EM/UM
reference; genotype pooled into PM: *2/*2).Zhao_2018_omeprazole.R
(power-of-binary-indicator multiplicative factor on CLOMZ-M1 formation
clearance: e_cyp2c19_pm_kmet_5oh ^ CYP2C19_PM with
e_cyp2c19_pm_kmet_5oh = 0.125; PM subjects have 87.5% lower
formation clearance of 5-hydroxy-omeprazole than the EM/UM reference;
Zhao 2018 Table 2).CYP2C19_IM. See
CYP2C19_IM Notes for the three-level decomposition
rationale. The PM phenotype indicator carries a separate typical-value
coefficient because the Zhao 2018 cohort observed *2/*2
poor metabolizers (n = 2) had substantially lower 5-OH-omeprazole
formation clearance than the heterozygous *1/*2 and
*2/*17 intermediate metabolizers (12.5% vs 44.9% of EM/UM
reference). Ratified canonically on 2026-05-25 alongside the Zhao 2018
omeprazole extraction.ABCB1_C3435T_MUT flags the homozygous variant
group). Time-fixed per subject (germline genotype).ABCB1_C3435T_MUT = 0). The reference group
is the homozygous wild-type C/C stratum; ABCB1_C3435T_MUT
flags the homozygous-variant T/T stratum.ABCB1 C3435T C/T / C/T – Zhao 2018 (paper
Table 2 reports a multiplicative scaling factor of 1.86 on Ka for the
heterozygote stratum relative to the C/C reference).Zhao_2018_omeprazole.R
(power-of-binary-indicator multiplicative factor on absorption rate
constant Ka: e_abcb1_c3435t_het_ka ^ ABCB1_C3435T_HET with
e_abcb1_c3435t_het_ka = 1.86; C/T heterozygotes have an
absorption rate constant approximately 86% higher than the C/C wild-type
reference; paired with ABCB1_C3435T_MUT and used
jointly).CYP3A5_STAR1_HET /
CYP3A5_STAR1_HOM and SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM paired-binary precedent for a three-level
genotype where each stratum carries a distinct typical-value covariate
effect. Distinct from the ABCB1_HAP_TTT haplotype canonical
(which jointly tests the cis combination of rs1128503 / rs2032582 /
rs1045642 SNPs as a single haplotype block; used in de Wit 2016
everolimus). Use ABCB1_C3435T_HET +
ABCB1_C3435T_MUT when the source paper fits a distinct
typical-value covariate effect to the single rs1045642 SNP without
phasing it into a haplotype, and when both the heterozygous and
homozygous-variant strata are large enough to identify independent
effects (Zhao 2018 cohort: n = 22 heterozygotes, 43.1%, and n = 4
homozygous variant, 7.8%, with n = 25 wild-type, 49.0%, as the
reference). Mechanistically the C3435T variant has been associated with
altered P-gp expression and substrate efflux in some studies (often via
linkage with functional variants in the same haplotype block), though
directionality of the effect on substrate exposure varies across
substrates and tissues. Ratified canonically on 2026-05-25 alongside the
Zhao 2018 omeprazole extraction.ABCB1_C3435T_HET flags the heterozygous group). Time-fixed
per subject (germline genotype).ABCB1_C3435T_HET = 0). The reference group
is the homozygous wild-type C/C stratum; ABCB1_C3435T_HET
flags the heterozygous C/T stratum.ABCB1 C3435T T/T / T/T – Zhao 2018 (paper
Table 2 reports a multiplicative scaling factor of 6.93 on Ka for the
homozygous-variant stratum relative to the C/C reference).Zhao_2018_omeprazole.R
(power-of-binary-indicator multiplicative factor on absorption rate
constant Ka: e_abcb1_c3435t_mut_ka ^ ABCB1_C3435T_MUT with
e_abcb1_c3435t_mut_ka = 6.93; T/T homozygotes have an
absorption rate constant approximately 6.93-fold higher than the C/C
wild-type reference; paired with ABCB1_C3435T_HET and used
jointly).ABCB1_C3435T_HET.
See ABCB1_C3435T_HET Notes for the three-level
decomposition rationale, the distinction from the
ABCB1_HAP_TTT haplotype canonical, and the Zhao 2018 cohort
distribution. Ratified canonically on 2026-05-25 alongside the Zhao 2018
omeprazole extraction.TTT haplotype across the rs1128503
(1236C>T, exon 12, synonymous Gly412Gly) / rs2032582 (2677G>T/A,
exon 21, Ala893Ser/Thr) / rs1045642 (3435C>T, exon 26, synonymous
Ile1145Ile) SNP block. 1 = subject carries at least one TTT
haplotype (heterozygous or homozygous; pooled because the homozygote
frequency was < 0.1 in the de Wit 2016 cohort); 0 = no
TTT haplotype. Time-fixed per subject (germline
haplotype).TTT
haplotype).ABCB1 TTT haplotype – de Wit 2016 (paper text Methods
‘Pharmacogenetic analysis’ and Table 2 final-model
theta TTT on F row; haplotypes phased in gPLINK with
certainty > 0.97).deWit_2016_everolimus.R (multiplicative effect on apparent
bioavailability F: F = 1 * 0.792^ABCB1_HAP_TTT – carriers
have 20.8% lower F than non-carriers; dOFV = 9.6 in backward
elimination, P < 0.01).TTT haplotype of
ABCB1 (P-glycoprotein, MDR1 efflux transporter) is associated
with enhanced P-gp efflux activity and reduced everolimus
bioavailability (de Wit 2016 Discussion paragraph 5); de Wit cites prior
evidence that the same TTT haplotype also reduces exposure
/ efficacy of other P-gp substrates [refs 20-22 in the paper], although
directionally inconsistent results have been reported [refs 7, 23, 24].
Het and hom carriers were pooled in de Wit 2016 because the homozygote
frequency was < 0.1 (Methods ‘Pharmacogenetic analysis’); future
extractions that estimate separate het / hom effects should register
paired ABCB1_HAP_TTT_HET and ABCB1_HAP_TTT_HOM
indicators following the SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM precedent above. Distinct from any
individual ABCB1 SNP indicator (rs1128503, rs2032582, rs1045642 alone)
because the haplotype is the cis combination tested jointly. Ratified
canonically on 2026-05-16 alongside the de Wit 2016 everolimus
extraction.*2 (inactive) variant allele (rs671 G>A;
Glu487Lys, aldehyde dehydrogenase 2 mitochondrial isoform). 1 = subject
carries at least one ALDH22 allele (heterozygous 1/2 or
homozygous 2/2); 0 = ALDH21/*1 wild-type. Time-fixed per
subject (germline genotype).ALDH2 (the source NONMEM indicator used in
Nemoto_2017_ethanol.R; 1 for 1/2 carriers, 0 for
1/1 wild-type; the Nemoto 2017 cohort had no 2/2
homozygotes so the indicator is effectively heterozygote-vs-wild-type in
that cohort).Nemoto_2017_ethanol.R
(additive shift on Vd/F: -20.4 L when ALDH2_S2_CARRIER = 1 vs
ALDH21/1 reference; Nemoto 2017 Table II final model).ALDH2_S2_HET and ALDH2_S2_HOM indicators
following the SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM precedent. Ratified canonically on
2026-05-18 alongside the Nemoto 2017 ethanol extraction.*2 (high-activity) variant allele (rs1229984
G>A; Arg47His, alcohol dehydrogenase 1B class I beta-subunit). 1 =
subject is homozygous 2/2; 0 = otherwise (the default reference
covers ADH1B2/1 heterozygotes; ADH1B1/1 wild-type
subjects fall outside the published parameterization of Nemoto 2017 and
are conventionally assigned the same reference value as heterozygotes).
Time-fixed per subject (germline genotype).ADH1B (the source NONMEM indicator used in
Nemoto_2017_ethanol.R; 1 for 2/2 homozygotes, 0
for 2/1 heterozygotes; Nemoto 2017 Table II structural model
encodes the Vmax conditional via two separate THETAs).Nemoto_2017_ethanol.R
(additive shift on Vmax: +176 mg/h for 2/2 homozygotes relative
to the 2/1 reference Vmax of 7790 mg/h; Nemoto 2017 Table II
final model).ADH1B_S2_HET as a companion indicator following
the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM
precedent. Ratified canonically on 2026-05-18 alongside the Nemoto 2017
ethanol extraction.X1 – used in Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (per-isolate Genotype-2
indicator).Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (multiplicative proportional
effect on the sigmoid-Emax EC50,
1 + e_pfmdr1_86y_ec50 * PFMDR1_86Y, drug-specific magnitude
per Simpson 2013 Table 3 Genotype-2 percent-change column).ABCB1_* canonicals (ABCB1 / MDR1 is
the human P-glycoprotein efflux-transporter gene; pfmdr1 is the
orthologous P. falciparum multidrug-resistance transporter). Member of
the four-member mutually-exclusive PFMDR1_* Simpson 2013
genotype set (PFMDR1_86Y, PFMDR1_1042D,
PFMDR1_CN2, PFMDR1_CN3PLUS), with single-copy
WT 86N/1042N as the all-zero reference. Future pfmdr1-genotype
antimalarial extractions should reuse this set or register sibling
parasite-genome canonicals. Ratified canonically alongside the Simpson
2013 antimalarial in-vitro extractions.X2 – used in Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (per-isolate Genotype-3
indicator).Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (multiplicative proportional
effect on the sigmoid-Emax EC50,
1 + e_pfmdr1_1042d_ec50 * PFMDR1_1042D, drug-specific
magnitude per Simpson 2013 Table 3 Genotype-3 percent-change
column).ABCB1_* canonicals (see PFMDR1_86Y notes).
Member of the four-member mutually-exclusive PFMDR1_*
Simpson 2013 genotype set with single-copy WT as the all-zero reference.
Ratified canonically alongside the Simpson 2013 antimalarial in-vitro
extractions.X3 – used in Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (per-isolate Genotype-4
indicator).Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (multiplicative proportional
effect on the sigmoid-Emax EC50,
1 + e_pfmdr1_cn2_ec50 * PFMDR1_CN2, drug-specific magnitude
per Simpson 2013 Table 3 Genotype-4 percent-change column).ABCB1_* canonicals (see PFMDR1_86Y notes).
Member of the four-member mutually-exclusive PFMDR1_*
Simpson 2013 genotype set with single-copy WT as the all-zero reference;
PFMDR1_CN2 (two copies) and PFMDR1_CN3PLUS
(three-or-more copies) jointly encode the copy-number amplification
axis. Ratified canonically alongside the Simpson 2013 antimalarial
in-vitro extractions.X4 – used in Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (per-isolate Genotype-5
indicator).Simpson_2013_artesunate.R,
Simpson_2013_chloroquine.R,
Simpson_2013_lumefantrine.R,
Simpson_2013_mefloquine.R (multiplicative proportional
effect on the sigmoid-Emax EC50,
1 + e_pfmdr1_cn3plus_ec50 * PFMDR1_CN3PLUS, drug-specific
magnitude per Simpson 2013 Table 3 Genotype-5 percent-change column; for
artesunate and mefloquine the triple+ copy group carries the largest
EC50 shift, +127% and +188% respectively).ABCB1_* canonicals (see PFMDR1_86Y notes).
Member of the four-member mutually-exclusive PFMDR1_*
Simpson 2013 genotype set with single-copy WT as the all-zero reference;
paired with PFMDR1_CN2 on the copy-number amplification
axis. Ratified canonically alongside the Simpson 2013 antimalarial
in-vitro extractions.*allele genotypes, etc.);
the per-model covariateData[[CYP2C19]]$units,
description, and notes document which proxy is
in force, the orientation (higher value = higher or lower
activity, which depends on the probe), and the population-mean reference
value used inside the model. Time-invariant in all known examples
(germline-genotype-determined enzyme expression measured once via
probe-substrate assay).covariateData[[CYP2C19]]$notes.CYP2C19 – used directly in
Areberg_2006_escitalopram.R (Areberg 2006 Equation 2
variable name; the value is the urinary S/R-mephenytoin ratio defined in
the Materials and Methods ‘CYP2C19 Phenotyping’ section).Areberg_2006_escitalopram.R (linear-deviation effect on
apparent oral clearance:
CL/F = 19 - 17 * (CYP2C19 - 0.769); CYP2C19 = urinary
S/R-mephenytoin ratio, population mean = 0.769; one unit of additional
ratio = 17 L/h lower clearance, consistent with the biology
that higher S/R ratio = lower CYP2C19 activity = lower escitalopram
metabolic clearance; Areberg 2006 Equations 1-2 and Table 3).CYP2D6 (line 4055) and CYP3A4 (line 4089)
canonicals, ratified here per the prospective registration TODO on the
CYP3A4 line. Orientation depends on the
probe. Two patterns coexist in the CYP2C19-aware popPK
literature:
covariateData[[CYP2C19]]$notes must document which
orientation is in force alongside the population-mean reference value
and the column’s units, mirroring the per-model documentation already
used for the CYP2D6 and CYP3A4 continuous
canonicals. Distinct from the binary genotype-derived canonicals
CYP2C19_S2_CARRIER (loss-of-function *2 allele carrier),
CYP2C19_IM (intermediate-metabolizer phenotype), and
CYP2C19_PM (poor-metabolizer phenotype); use the binary
indicators when the source paper reports only a discrete phenotype
label, and use CYP2C19 (continuous) when the source reports
a probe-derived continuous number. Ratified canonically on 2026-06-09
alongside the Areberg 2006 escitalopram extraction (founding example
uses the urinary S/R-mephenytoin ratio form).CYP2C19_PM (strict homozygous-PM only) and from
CYP2C19_IM (intermediate-metabolizer-only); CYP2C19_NON_EM
is a paper-defined composite of PM + IM that the source paper used as a
single multiplicative covariate on clearance.PM (Wahlby 2004 / Walsh 2004 voriconazole source-column
convention) – used in Wahlby_2004_voriconazole.R. Note that
the column name PM in this paper denotes the composite
non-EM group, not the strict CYP2C19_PM phenotype; users converting from
the source paper’s data should not collapse PM to canonical
CYP2C19_PM without re-checking the genotype-to-indicator
mapping.Wahlby_2004_voriconazole.R (multiplicative effect on CL:
(1 - 0.46 * CYP2C19_NON_EM), so a non-EM subject has 46
percent lower CL than a homozygous-EM subject).CYP2C19_PM and
CYP2C19_IM canonicals instead. The Wahlby 2004 / Walsh 2004
voriconazole encoding has been retained because the source paper does
not provide separate per-genotype coefficient estimates. Ratified
canonically alongside the Wahlby 2004 extraction.ABCB1 (rs3842) – Mukonzo 2009 (paper Table 1 ‘ABCB1’
section: position c.4036 A/G, rs number
rs3842, allele New SNP, protein
3' UTR, observed SNP frequency 16.8%; Table 3 final-model
‘Effect of ABCB1 (rs 3842)’ row).Mukonzo_2009_efavirenz.R (linear-deviation effect on
relative bioavailability Frel:
f(depot) = (1 + e_rs3842_fdepot * SNP_ABCB1_RS3842) * exp(etalfdepot)
with e_rs3842_fdepot = 0.257; mutant carriers (any G
allele) have 25.7% higher Frel than AA wild-type carriers; Mukonzo 2009
Table 3 and Results paragraph ‘Pharmacokinetic modelling’).SNP_ABCB1_RS1045642 (a different SNP at c.3435C>T in
exon 26 that affects P-glycoprotein expression / function) and from
ABCB1_HAP_TTT (the multi-SNP haplotype across rs1128503 /
rs2032582 / rs1045642 jointly); use SNP_ABCB1_RS3842 only
when the source paper genotypes the c.4036A>G variant in the 3’ UTR.
Mukonzo 2009 selected rs3842 from a panel of 13 ABCB1 SNPs as a
“previously uncharacterised polymorphism” predicted to have potential
regulatory function by bioinformatics tools (Mukonzo 2009 Table 1
‘Relevance’ = ‘Undetermined’ at extraction time); the modeled +25.7%
effect on relative bioavailability was the strongest single-SNP signal
among the 13 ABCB1 candidates in the Mukonzo 2009 backward-elimination
step. Heterozygote and homozygote G-carriers are pooled in Mukonzo 2009
because the paper explicitly groups them (Results: “Mutant homozygote
and heterozygote individuals for ABCB1 rs3842 exhibited 26% greater
efavirenz bioavailability than wild-type carriers”); future extractions
that estimate separate het / hom effects should register paired
SNP_ABCB1_RS3842_HET and SNP_ABCB1_RS3842_HOM
indicators following the SLCO1B1_HAP15_HET /
SLCO1B1_HAP15_HOM precedent. The Mukonzo 2009 Discussion
notes that the rs3842 minor-allele frequency is >20% in Asians,
Europeans, and Sub-Saharan Africans according to dbSNP, so the canonical
may have broad applicability once additional papers ratify the same
encoding. Ratified canonically on 2026-06-13 alongside the Mukonzo 2009
efavirenz extraction.model() into a homozygous-mutant-only indicator
(count == 2) multiplied by a -0.199 multiplicative shift on
CL/F. Heterozygotes are pooled with wild-type and receive no shift,
consistent with the source paper’s reported effect for “homozygous
CYP2B6*11” only (Mukonzo 2009 Results paragraph ‘Pharmacokinetic
modelling’).CYP2B6*11 – Mukonzo 2009 (paper Table 1 SNP-table
column ‘Allele’ and Table 3 ’Effect of CYP2B6*11’ row; rs35303484 is the
defining SNP at c.136 A->G with p.M46V protein change).Mukonzo_2009_efavirenz.R (homozygous-mutant-only
multiplicative shift on CL/F:
cl = exp(lcl + etalcl) * (1 + e_2b6_6_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 2)) * (1 + e_2b6_11_cl * (SNP_CYP2B6_RS35303484_G_COUNT == 2))
with e_2b6_11_cl = -0.199; homozygous *11 carriers have
19.9% lower CL/F than the wild-type / heterozygous reference; Mukonzo
2009 Table 3).SNP_CYP2B6_RS3745274_T_COUNT and
SNP_CYP2B6_RS28399499_C_COUNT: a single count column
captures the underlying genotype without redundancy and lets the model
code derive either a linear per-allele effect (* count) or
a non-additive HET / HOM decomposition (* (count == 1),
* (count == 2)) depending on the source paper’s
parameterization. Ratified canonically on 2026-06-13 alongside the
Mukonzo 2009 efavirenz extraction.Smoking (case-insensitive) – used in
Ma_2020_sarilumab_anc.R.Ma_2020_sarilumab_anc.R (power-form on baseline ANC:
BASE * 1.15^SMOKE).SMOKE_CURRENT +
SMOKE_NEVER indicators below instead – the 3-level encoding
cannot be reduced to a single SMOKE column without losing
information.SMOKE_NEVER
to encode a 3-level smoking-status categorical with former smoker as the
implicit reference (both indicators = 0).SMOKE_NEVER = 0). The pairing follows the
RACE_<GROUP> convention for paired indicators.Smoking status = Current / SMOK = 2
(case-insensitive) – derived from a 3-level smoking-status column.Hwang_2023_monalizumab.R (proportional-shift effect on V1:
(1 + 0.0484)^SMOKE_CURRENT; reference category former
smoker).SMOKE_NEVER (paired indicator) and SMOKE
(binary current-vs-non-smoker encoding when the source paper does not
split former vs never).SMOKE_CURRENT to encode a 3-level smoking-status
categorical with former smoker as the implicit reference (both
indicators = 0).SMOKE_CURRENT = 0). The pairing follows the
RACE_<GROUP> convention for paired indicators.Smoking status = Never / SMOK = 0
(case-insensitive) – derived from a 3-level smoking-status column.Hwang_2023_monalizumab.R (proportional-shift effect on V1:
(1 - 0.141)^SMOKE_NEVER; reference category former
smoker).SMOKE_CURRENT (paired indicator) and SMOKE
(binary current-vs-non-smoker encoding when the source paper does not
split former vs never).All ROUTE_<TARGET> canonicals follow the same
shape: a binary indicator where 1 = subject received the
<TARGET> administration route and 0 = subject
received the reference (non-target) route. The reference
category is route-pair-specific and is documented per-canonical in the
Reference category: field (e.g., ROUTE_IV
references SC; ROUTE_IP references non-IP;
ROUTE_NGT references oral). The 2026-06-19
canonical-register standardization audit reviewed renaming this family
to a more explicit
ROUTE_<TARGET>_VS_<REFERENCE> shape but the
operator declined the rename and instead set the following documentation
discipline:
Policy (2026-06-19 audit): every
ROUTE_* canonical MUST document its reference category
explicitly in the Reference category: field of its H3
block. The reference is what ROUTE_<TARGET> = 0 means
in the source dataset and is paper-specific. New ROUTE_*
canonicals registered after 2026-06-19 must follow this rule; existing
entries already comply. The discipline eliminates the
silent-mis-encoding risk (where a model file might assume the wrong
reference category) by making the reference explicit in the register
rather than relying on the column name alone.
IV – used in
Zierhut_2008_osteoprotegerin.R (DDMODEL00000233
dataObj column flagging IV vs SC cohort, switching the PK
observation residual SD between CcpropSdIV and
CcpropSdSC).Wang_2021_pertuzumab.R
(per-subject FeDeriCa arm indicator P+H IV vs PH FDC SC, switching the
proportional residual SD between CcpropSdIv and
CcpropSdSc).Yu_2022_ofatumumab.R (exponential effect on R0, CL, Q,
ksyninf).Zierhut_2008_osteoprotegerin.R (per-subject indicator
switching the PK observation residual SD between the IV cohort
(CcpropSdIV) and the SC cohort
(CcpropSdSC)).Wang_2021_pertuzumab.R (per-subject indicator switching
the proportional residual SD between the IV (CcpropSdIv =
0.175) and SC (CcpropSdSc = 0.155) cohorts of the FeDeriCa
popPK).Fiedler-Kelly_2019_fremanezumab.R (per-subject
indicator switching both the central volume of distribution (Vc,IV =
2.98 L FIXED vs Vc,SC = 1.88 L) and the residual-error structure (IV:
proportional-only with SD = sqrt(0.0467) = 0.21610; SC: combined
additive sqrt(0.204) + proportional sqrt(0.0531)) in the pooled phase
1/2b/3 fremanezumab popPK).Fanta_2007_ciclosporin.R (per-occasion indicator
carrying a 23% lower typical CL on oral occasions than on IV occasions
(exp(e_route_iv_cl * ROUTE_IV) with
e_route_iv_cl = log(1/0.77) = 0.2614) and switching the
residual-error structure between IV (propSdIv = 0.089 +
addSdIv = 1.5 ug/L) and oral (propSdPo = 0.20)
arms; the comparator non-IV route is oral microemulsion ciclosporin, not
SC).Koolen_2010_docetaxel.R (per-dose-record indicator
switching the docetaxel central volume of distribution between the IV
value (V_central_iv = 9.8 L, polysorbate-80-micelle-sequestered) and the
oral value (V_central_po = 44.0 L) per Koolen 2010 Results ‘Volume of
distribution’ section; the reference category is oral here, not SC,
because the source paper pools IV and oral docetaxel cohorts).cmt column that
names the target compartment. When simulating, set
ROUTE_IV = 1 for IV cohorts and dose into the central
compartment; set ROUTE_IV = 0 for SC cohorts and dose into
the depot. Scope: specific because the set of parameters that differ by
route is paper-specific (Yu 2022 carries route-specific exponential
effects on disposition parameters; Zierhut 2008, Wang 2021, and
Fiedler-Kelly 2019 carry route-specific PK observation residual SDs;
Fiedler-Kelly 2019 additionally carries a route-specific central volume
of distribution).Johnson_2011_olanzapine_rat.R (per-dose-record
indicator selecting the IP bioavailability FIP = 0.636 with
87% CV log-normal IIV; ROUTE_IP = 0 selects F = 1 for SC and IV. The
encoding f(central) <- exp(ROUTE_IP * (lfip + etalfip))
collapses to 1 when ROUTE_IP = 0 because exp(0) = 1, so subjects dosed
via SC or IV inherit complete bioavailability without IIV on F).cmt column that names the target compartment (Johnson 2011
doses all routes directly into central because the
absorption rate constant was not estimable from the available data).
When simulating IP doses, set ROUTE_IP = 1 on the dose
record(s); set ROUTE_IP = 0 for SC and IV dose records.
Scope: specific because the IP-vs-other contrast and which parameter it
modifies (here, bioavailability) is paper-specific; complementary to
ROUTE_IV (IV-vs-SC indicator) — a future tri-route study
could use both indicators jointly.NGT – used in Denti_2018_levofloxacin.R
(per-dose-record nasogastric-tube indicator).Denti_2018_levofloxacin.R (multiplicative effect on the
absorption lag time: (1 + e_route_ngt_tlag * ROUTE_NGT)
with e_route_ngt_tlag = -0.856, so NGT delivery shortens
T_lag to ~14.4% of its oral value; in the cohort 90/109 (82.6%) children
were dosed by crushed tablet via NGT, Denti 2018 Table 2).cmt column that names the target compartment. When
simulating NGT doses, set ROUTE_NGT = 1 on the dose
record(s); set ROUTE_NGT = 0 for oral dose records. Scope:
specific because the NGT-vs-oral contrast and which absorption parameter
it modifies are paper-specific. Distinct from the ROUTE_IV
(IV-vs-SC) and ROUTE_IP (IP-vs-non-IP) parenteral-route
indicators – ROUTE_NGT is an enteral-delivery-method
indicator within oral administration. Ratified canonically alongside the
Denti 2018 levofloxacin extraction.Yu_2022_ofatumumab.R
(exponential effect on k_e(P) and R0),
Diep_2026_donidalorsen.R (Phoenix linear-effect
(1 + e_device_ai_ka * DEVICE_AI) on the typical SC
absorption rate constant with theta = +0.262 -> multiplier 1.262 for
autoinjector vs vial-and-syringe reference; characterized in the ISIS
721744-CS9 single-dose bioequivalence cohort).ROUTE_IV instead. Scope: specific because
the AI / vial / PFS contrast and which parameters it affects depend on
the study’s device-comparison design.Diep_2022_eplontersen.R (additive log-shift
e_injsite_arm_ka = log(ka_arm / ka_ab) on the typical
absorption rate constant: ka_arm = 0.217 1/h vs ka_ab = 0.282 1/h, ~30%
higher ka for abdomen; INJSITE_ARM = 1 selects the arm typical value),
Diep_2026_donidalorsen.R (Phoenix linear-effect
(1 + e_injsite_arm_ka * INJSITE_ARM) on the typical
absorption rate constant with theta = -0.338 -> multiplier 0.662 for
arm; the paper’s reference category is “abdomen or thigh” rather than
“abdomen” alone, but is consistent with the canonical reference because
abdomen is the universal SC reference site and the thigh effect is
pooled into the reference category by the Diep 2026 model).INJSITE_THIGH (thigh-vs-abdomen indicator anticipated for
future SC-route models with thigh-specific absorption).
Per-administration rather than per-subject – a subject in a multi-dose
simulation can switch SC injection sites between doses; supply the
indicator on each dose record. Distinct from ROUTE_IV (IV
vs SC route, not within-SC site) and from DEVICE_AI
(autoinjector vs prefilled syringe, device rather than anatomical
site).Yu_2022_ofatumumab.R
(exponential effect on Emax of B cell lysis).Yu_2022_ofatumumab.R
(exponential effect on B cell elimination rate kout).BID – used in
Girard_2012_pimasertib.R.Girard_2012_pimasertib.R (additive shift on the
cumulative-logit AE-score model: theta_bid * REGI_BID;
-0.399 logit units for BID vs QD).REGI_TID, REGI_QW) following the same
pattern. Distinct from DOSE (dose level in mg) and from
total-daily-dose aggregates: a 60 mg/day cohort can include either a 60
mg QD subgroup or a 30 mg BID subgroup, and both share the same
DOSE = 60 while differing in REGI_BID.Kuchimanchi_2018_evolocumab_ldlc.R).
Document the per-paper comparator in
covariateData[[REGI_QM]]$notes.REG – Kuchimanchi 2018 notation for the binary
indicator i in the exposure-response formula
Eff = Emax * AUC / (EC50 * REG^i + AUC). The linear-scale
multiplier paired with this indicator (paper estimate 2.30) maps to the
canonical parameter reg_qm.Kuchimanchi_2018_evolocumab_ldlc.R (founding example;
exposure-response Emax on LDL-C distinguishes 420 mg SC QM from 140 mg
SC Q2W via ec50_eff = ec50 * reg_qm^REGI_QM).REGI_BID
(within-QD-vs-BID schedules) under the REGI_* family
pattern; future regimen-comparison models that contrast different
intervals (e.g., Q4W vs Q8W) should register a sibling indicator
(REGI_Q4W, REGI_Q8W) following the same
pattern. Distinct from DOSE (dose level in mg) and from
regimen-implied total-AUC aggregates: a steady-state exposure-matched
140 mg Q2W and 420 mg QM pair share the same AUC over 4 weeks but differ
in REGI_QM. Pairs with the canonical parameter
lreg_qm / reg_qm (the EC50 multiplier
exponentiated by this indicator). Use only in static Emax-on-AUC
exposure-response models; dynamic ODE-based PK / PD layers do not need
this indicator.Dorlo_2017_miltefosine.R (sets
tlowf <- 7 * MIL_REGIMEN + 1 * (1 - MIL_REGIMEN) inside
model(); tlowf is then used in the
(t > 0) * (t <= tlowf) predicate that gates the
typical 74.3% reduction in relative bioavailability during the initial
absorption window).REGI_BID
(within-monotherapy QD vs BID schedule) and from the
CONMED_* family (the LAmB co-administration in the
combination arm is captured implicitly via the regimen indicator rather
than via an explicit concomitant-medication binary, because the
LAmB-driven effect modelled here is on the absorption-window duration
rather than on miltefosine clearance or distribution).FED (per-dose-record fed-vs-fasted indicator): MEAL_FLAG is
a time-varying intra-day flag that switches on at meal onset, stays on
through the meal duration, and switches off afterward. Used by
enterohepatic-recirculation / gallbladder-contraction models that need
to scale post-prandial transport rate constants for the duration of a
meal.Zuo_2016_UDCA.R
(multiplies the biliary-to-intestine rate constants K_BI,0/1/2 by E_meal
= 35.33 during meal windows to simulate gallbladder contraction; two
meals modelled per day, lunch at +4 h and dinner at +10 h after the
morning dose, each 1 hour long).SNACK_FLAG
for studies that distinguish meal and snack effects. Distinct from
FED and FED_HIGHFAT (per-dose-record
meal-state indicators tied to a single dosing event); MEAL_FLAG is
decoupled from any specific dose record and instead drives ongoing
physiology over a multi-hour window.MEAL_FLAG by
enterohepatic-recirculation / gallbladder-contraction models that scale
post-prandial transport rate constants with a smaller magnitude for
snacks than for full meals.Zuo_2016_UDCA.R
(multiplies the biliary-to-intestine rate constants K_BI,0/1/2 by
E_snack = 9.53 during snack windows; one snack modelled per day at +7 h
after the morning dose in the Xiang 2011 single-dose study, 0.5 hour
long).MEAL_FLAG. Time-varying.f(<depot>) <- FRACABS); the
user supplies the per-dose F value derived from the paper’s regression /
lookup table.Zuo_2016_UDCA.R
(dose-dependent fractional absorption derived from a log-dose linear
regression: F = 0.66 at 150 mg and F = 0.31 at 1000 mg, with R^2 = 0.99
over 200-2000 mg per the paper; combined with the UDCA molecular-weight
conversion in the bioavailability hook
f(stomach_udca) <- FRACABS / mw_udca).lfdepot / f(depot) in models
where bioavailability is an estimable PK parameter rather than a
supplied covariate. When the paper’s F regression is itself a function
of DOSE, the user can derive FRACABS from the DOSE column
upstream of rxSolve rather than carrying the regression
inside model().Kyhl_2016_nalmefene.R,
Goel_2016_Sonidegib.R (the Goel 2016 healthy-fasted F
effect is applied via
e_healthy_fast_f ^ (DIS_HEALTHY * (1 - FED));
cancer-patient records have FED = 1, healthy-fasted-arm records have FED
= 0, high-fat-meal arm records have FED = 1 and the additional
FED_HIGHFAT = 1 indicator), Niebecker_2015_edoxaban.R (the
Niebecker 2015 “study 6” indicator is encoded via FED – study 6 was the
only Fed-state phase 1 study in the pooled analysis; all 12 other phase
1 studies and the Hokusai-VTE phase 3 study were overnight-fast. Two
multiplicative effects: ka * (1 + (-0.690) * FED) for the -69%
absorption slowdown and CLnr/F * (1 + 0.204 * FED) for the +20.4%
non-renal CL with food).FED indicator for studies that specifically test the
high-fat-meal food effect on bioavailability or absorption rate (a
common solubility-limited absorption phenotype).Fatmeal – used in Goel_2016_Sonidegib.R
(covariate on F).FATM – used in Goel_2016_Sonidegib.R
(covariate on Ka; same indicator as Fatmeal in that
paper).Goel_2016_Sonidegib.R
(multiplicative effect on F: 5.74^FED_HIGHFAT – ~5.7-fold
higher F under high-fat meal vs 2 h post-light-meal reference;
multiplicative effect on Ka: 1.01^FED_HIGHFAT – no
meaningful effect).FED (binary
fed-vs-fasted): FED_HIGHFAT carries the specific “high-fat
meal” semantic (typically >= 800 kcal, >= 50% calories from fat
per FDA guidance). Document the per-protocol meal definition in
covariateData[[FED_HIGHFAT]]$notes. When a paper reports a
high-fat-meal arm and a separate fasted-healthy arm (e.g., Goel 2016),
use FED_HIGHFAT for the high-fat semantic and the existing
FED + DIS_HEALTHY indicators for the composite
healthy-fasted effect (Goel 2016 applies the e_healthy_fast_f effect via
(DIS_HEALTHY * (1 - FED))); the retired
HV_FAST composite indicator was deleted on 2026-05-11.FMDD – used in Goel_2016_Sonidegib.R (Goel
2016 covariate on F).Goel_2016_Sonidegib.R
(multiplicative effect on F: 1.16^MULTI_DOSE_PT – ~16%
higher apparent F during the multiple-dose phase relative to first dose,
attributed in the paper to occasional non-fasting compliance),
Fang_2010_etanercept.R (multiplicative effect on F:
0.674^MULTI_DOSE_PT – ~33% lower apparent F during the
multiple-dose phase relative to the single-dose reference; Fang 2010
attributes the reduction to partitioning of the rhTNFR-Fc fusion protein
into local subcutaneous adipose tissue with repeated injection. In Fang
2010 the multi-dose cohort is the AS-patient arm and the single-dose
cohort is the healthy-volunteer arm, so MULTI_DOSE_PT is effectively
subject-level: source column M),
vanIersel_2018_posaconazole.R (multiplicative effect on
apparent clearance: cl *= (1 + 0.750 * MULTI_DOSE_PT) – 75%
higher CL in multiple-dose records relative to the single-dose
reference; van Iersel 2018 Table 2 final-model ‘Dosing regimen on CL’ =
0.750).FED and FED_HIGHFAT (which are
per-record meal-state indicators) and from REGI_BID
(regimen indicator). Future models that need a generic “occasion
boundary” effect should consider the existing ooc<n>
IOV pattern instead.TABLET – earlier name used by
Kyhl_2016_nalmefene.R and
Tikiso_2021_abacavir.R; renamed to FORM_TABLET
for consistency with the FORM_* family
(FORM_CAPSULE, future FORM_SUSPENSION,
etc.).Kyhl_2016_nalmefene.R
(additive shift on residual error: tablet vs solution),
Tikiso_2021_abacavir.R (multiplicative effect on the
absorption mean transit time MTT: tablet (abacavir + lamivudine
fixed-dose-combination tablet) vs liquid solution;
mtt *= (1 + 0.249 * FORM_TABLET), i.e. 24.9% slower
absorption for the FDC tablet relative to the abacavir liquid
reference), Kleideiter_2017_cebranopadol.R (tablet is the
typical-value reference, FORM_TABLET = 1 leaves the formulation effects
on ka, klag, and bioavailability at zero;
paired with FORM_CAPSULE and the derived
is_solution = (1 - FORM_TABLET) * (1 - FORM_CAPSULE) to
encode a three-level formulation stratification),
Fisher_2008_fosamprenavir.R (relative bioavailability of
tablet vs fed-suspension reference: F_tab = 1.09; encoded as
log(F) <- log(F_tab) * FORM_TABLET + log(F_food,sus) * (1 - FORM_TABLET) * (1 - FED)
so tablet bioavailability does not additionally depend on food, while
fasted suspension carries the F_food,sus reduction).FORM_FDC canonical (Wilkins 2008
antitubercular fixed-dose-combination of multiple drugs, contrasted
against single-drug tablets) because here the comparator is a non-tablet
liquid / solution rather than a separate tablet product. Future
formulation-comparison models should either extend this entry’s example
list when the comparator is a liquid / solution, or register a sibling
canonical when contrasting two tablet products.FORM_DOX_DORYX_MPC = 0) for Hopkins 2017 doxycycline.
Document the comparator and the reference-category bioavailability
per-model in covariateData[[FORM_CAPSULE]]$notes.fdepot fixed to 1 for the solution arm; tablet in
Gupta 2016 with F fixed to 1 for the tablet arm. Salem 2014 anchors F =
1 on the capsule arm (FORM_CAPSULE = 1): the capsule is the structural F
= 1 reference and the liquid (suspension or solution) arm carries the
estimated age-dependent relative bioavailability. The IIV (when present)
is gated to the arm that carries the estimated F; document the
orientation in covariateData[[FORM_CAPSULE]]$notes.PREP – used in Hennig_2006_itraconazole.R
(Clin Pharmacokinet 2006;45(11):1099-1114; PREP = 1 = capsule, PREP = 0
= oral solution) and in Hennig_2007_itraconazole.R (Br J
Clin Pharmacol 2007;63(4):438-450; DOI 10.1111/j.1365-2125.2006.02778.x;
same orientation, capsule typical absorption parameters as the published
reference).CAPSULE – earlier name used in the
Hennig_2006_itraconazole.R and
Hennig_2007_itraconazole.R model files before the
FORM_* rename.Hennig_2006_itraconazole.R (Hennig 2006 Table II final
estimates: capsule ka 0.09 h^-1 vs solution ka
0.96 h^-1 and capsule relative bioavailability 0.55 vs solution 1; the
etalfdepot IIV applies only to the capsule arm),
Hennig_2007_itraconazole.R (selects between
lka_cap and lka_sol typical-value absorption
rate constants and applies
f(depot) <- (1 - FORM_CAPSULE) + FORM_CAPSULE * fdepot
so the relative bioavailability F_rel = 0.817 is applied
only to the capsule arm), Gupta_2016_lenvatinib.R (relative
bioavailability of capsule vs tablet is 0.896; F1 fixed to 1 for the
tablet reference; the etalfcap IIV (30.2% CV) applies only
to the capsule arm), Lacy_2018_cabozantinib.R
(multiplicative fractional effect of capsule (vs tablet reference) on Ka
= -0.579 (57.9% slower absorption for capsule) and on overall
bioavailability F = -0.144 (14.4% lower exposure for capsule); tablet F
fixed at 1 as the reference; comparator capsule = Cometriq 140 mg
approved for MTC), Kleideiter_2017_cebranopadol.R (paired
with FORM_TABLET to encode the three-level cebranopadol
formulation stratification: tablet reference, oral solution,
liquid-filled capsule; multiplicative effects on ka (log
shift log(2.09 / 0.864) = 0.883), klag (log
shift log(0.077 / 0.087) = -0.122), and bioavailability
(factor 1.174) for capsules relative to the tablet reference),
Kleideiter_2018_cebranopadol.R (three-level formulation
factor {tablet, oral solution, liquid-filled capsule}
encoded into two binary indicators FORM_CAPSULE and the
sibling new canonical FORM_SOLUTION; tablet is the
reference when both indicators are 0; capsule multiplicative effects per
Kleideiter 2018 Table 13 are 2.09 / 0.864 = 2.419 on Ka, 0.077 / 0.087 =
0.885 on klag, and 1.174 on bioavailability F),
Hopkins_2017_doxycycline.R (three-level Doryx formulation
factor
{Doryx tablet (delayed-release), Doryx MPC (modified-acid-resistance delayed-release), Doryx capsule (conventional-release)}
encoded into two binary indicators FORM_CAPSULE and the
sibling canonical FORM_DOX_DORYX_MPC; Doryx tablet is the
reference when both indicators are 0; Doryx-capsule relative
bioavailability vs the Doryx-tablet reference is 0.978 per Hopkins 2017
Table 3 F1CAP, shared 0.115 h absorption lag with
FORM_DOX_DORYX_MPC per Table 3 ALAG1, and the Doryx-capsule
food effect on KTR matches the Doryx-tablet -20.9% reduction per Table 3
COVFED rather than the Doryx-MPC -54.9% reduction),
Salem_2014_efavirenz.R (capsule is the structural F = 1
reference; the oral suspension and oral solution arms are pooled as
‘liquid’ since Salem 2014 found no difference in F between them, and the
liquid arm carries an Emax age-dependent relative bioavailability with
mature asymptote 0.79 (Salem 2014 Table 2 TVF, RSE 12.5%) and TM50,F =
10.6 months (Salem 2014 Table 2 TM50,F, RSE 38.7%); IIV on the mature
liquid F is 39.9% CV (etaltvf_liq) and is gated to the liquid arm via
fdepot <- FORM_CAPSULE * 1 + (1 - FORM_CAPSULE) * f_liquid).FORM_TABLET (Kyhl
2016 / Tikiso 2021 tablet vs solution) under the FORM_*
family. Future formulation-comparison models that need a capsule
indicator should reuse this canonical, extending the example list and
documenting the comparator in per-model notes.covariateData[[FORM_SOLUTION]]$notes.FORM_CAPSULE
also at 0 in the reference state).FORM (three-level formulation column {tablet, oral
solution, liquid-filled capsule}) – used in
Kleideiter_2018_cebranopadol.R; the oral-solution level
maps to FORM_SOLUTION = 1 and is paired with
FORM_CAPSULE so that both = 0 selects the tablet
reference.Kleideiter_2018_cebranopadol.R (third member of the
three-level cebranopadol formulation stratification {tablet reference,
oral solution, liquid-filled capsule}, encoded into the two binary
indicators FORM_SOLUTION and FORM_CAPSULE;
oral-solution multiplicative effects relative to the tablet reference
per Kleideiter 2018 Table 13 are 2.43 / 0.864 = 2.813 on Ka, 0.077 /
0.087 = 0.885 on klag, and 1.045 on bioavailability F),
Othman_2013_ABT_102.R (two-level oral-solution vs
solid-dispersion comparator for the TRPV1 antagonist ABT-102;
solution-vs-solid-dispersion-reference effects per Othman 2013
PK/PD-model Results paragraph 1 are 0.40 multiplicative on
bioavailability F and 0.5 multiplicative on absorption lag time),
PerezRuixo_2006_tipifarnib.R (oral-solution-vs-solid
(capsule or tablet) absorption-effect indicator; ratio-form
multiplicative effects on D1 (0.348), Ka (2.07), and tlag (0.183)
applied as ratio^FORM_SOLUTION; reference category is the
solid oral form – Perez-Ruixo 2006 found the absorption profile
statistically indistinguishable between capsule and tablet, so the two
solid forms are pooled as the reference, and the solution shows ~3-fold
faster zero-order release, 2-fold higher Ka, and a 5.5-fold shorter lag
time relative to the solid reference).FORM_* family alongside
FORM_TABLET (tablet vs solution) and
FORM_CAPSULE (capsule vs per-paper comparator); register
prose for those entries already anticipates this name. Where a model
carries all three formulations, FORM_CAPSULE +
FORM_SOLUTION form the two-indicator encoding with tablet
as the all-zero reference. Distinct from FORM_ASV_LIQUID (a
drug-specific suspension/solution-vs-capsule/tablet indicator for
asunaprevir). Ratified canonically alongside the Kleideiter 2018
cebranopadol extraction.FORM_ASV_LIQUID.Wang_2018_daclatasvir_asunaprevir.R (switches the
structural zero-order absorption fraction FK between fk_cap_asv = 0.184
(capsule/tablet) and fk_sol_asv = 0.334 (suspension/solution); both
estimated with a shared 65.0% inter-arm-variability CV encoded as a
single eta on the logit of FK, Wang 2018 Table 3).FORM_TABLET /
FORM_CAPSULE / FORM_SOLUTION indicators
because the contrast here lumps both liquid forms (suspension and
solution) against both solid forms (capsule and tablet) for a single
named drug (ASV). Per-dose-occasion in principle (a participant could
receive both formulations across occasions), but in the Wang 2018 trials
each subject received a single ASV formulation. Ratified canonically
alongside the Wang 2018 daclatasvir/asunaprevir extraction.Li_2019_abatacept.R
(additive shift on the logit-scale bioavailability:
logit_F = logit_F_TV + FORM_ABA_PHASE2 * (-1.16); Li 2019
Table 2 reports the Phase-2 formulation effect as -1.16 logit-scale
units, mapping to an absolute F of ~0.56 vs the Phase-3 reference
~0.81).FORM_<drug>_PHASE2 canonical rather than reuse this
entry. Distinct from the generic FORM_TABLET /
FORM_CAPSULE / FORM_SUSPENSION family (oral
solid-dosage manipulations) because the contrast here is two SC
injectable formulations with different excipient pH. Ratified
canonically on 2026-05-28 per the naming-audit D20 review.Yin_2020_pexidartinib.R (multiplicative effect on the depot
bioavailability: F1 = 1.0 when FORM_PEX_PHASE1 = 0 (Phase 3 reference)
and F1 = 0.855 * exp(etalfdepot) when FORM_PEX_PHASE1 = 1, where the
0.855 typical-value anchor is held fixed in the source model and the IIV
etalfdepot ~ N(0, 0.101) is gated by FORM_PEX_PHASE1 so it has no effect
on Phase 3 subjects).FORM_<drug>_PHASE1
canonical rather than reuse this entry (sibling to
FORM_ABA_PHASE2 for abatacept). Distinct from the generic
FORM_TABLET / FORM_CAPSULE /
FORM_SUSPENSION family (oral solid-dosage manipulations)
because the contrast here is two oral formulations differentiated by
clinical-development phase rather than by dosage form. For typical
simulations of the approved 800 mg/day TGCT regimen used in ENLIVEN, set
FORM_PEX_PHASE1 = 0 for every subject.covariateData[[FORM_SUSPENSION]]$notes.FORM – used in Svensson_2018_bedaquiline.R
(paper’s narrative “whole vs suspended” with the suspended formulation
as the 1 level in the analytical control stream).Valle_2005_exemestane.R (3x3 Latin-square crossover with
the extemporaneous suspension as the 1 level).Svensson_2018_bedaquiline.R (multiplicative effect on the
typical mean absorption time MAT:
mat_typ = exp(lmat) * (1 + e_susp_mat * FORM_SUSPENSION)
with e_susp_mat = +0.23 (suspended-tablet MAT is 23% longer
than whole-tablet MAT; Svensson 2018 Table 2, 95% CI 2.1-48%, P = 0.03);
relative bioavailability F is held identical between formulations
because the paper found no statistically significant difference (95%
nonparametric CI 94-108% within the 80-125% bioequivalence criteria)),
Valle_2005_exemestane.R (multiplicative effects on
absorption rate ka (suspension/SCT-fasting ratio 7.6/2.35 = 3.234x;
suspension absorbs ~3.2x faster than the SCT swallowed whole) and on
apparent bioavailability F (suspension/SCT-fasting ratio = 1.2x);
intrinsic V/F is shared across formulations; the paper’s per-treatment
V/F values 1360 vs 1120 L collapse to a constant intrinsic V at
V/F_SCT_fasting / F_suspension = 1133 ~= 1120 L).FORM_TABLET (Kyhl 2016 / Tikiso 2021 tablet vs liquid
solution), FORM_CAPSULE, and FORM_POWDER.
Future extemporaneous-suspension bioequivalence extractions should reuse
this canonical and extend the example list, documenting the per-paper
comparator solid-oral form. Ratified canonically on 2026-05-16 alongside
the Svensson 2018 bedaquiline extraction.FORM_SUSPENSION (Svensson 2018 / Valle 2005
extemporaneously-compounded suspension contrasted with a whole-swallowed
solid) and from FORM_GRANULE (Ayyoub 2016
reconstitution-from-granules contrasted with a tablet) because here all
subjects receive the same manufactured suspension product and
the contrast is whether the suspension was diluted or not before
swallowing.ka in
Willmann 2021).FORM,DILU (composite: FORM = 2 for
ready-to-use suspension AND DILU = 1 for undiluted) – used
in Willmann_2021_rivaroxaban.R (supplement S2 NONMEM $PK
block
IF(FORM.EQ.2.AND.DILU.EQ.1) FD = 2 ; IF(FD.EQ.2) KA = THETA(6)).Willmann_2021_rivaroxaban.R (piecewise selection of the
typical first-order absorption rate constant Ka:
ka = exp(lka) * (1 - FORM_UNDILUTED_SUSP) + exp(lka_susp_undiluted) * FORM_UNDILUTED_SUSP;
the undiluted suspension Ka = 0.226 1/h (95% CI 0.154-0.297) is ~3.5x
slower than the comparator pool Ka = 0.799 1/h (95% CI 0.655-0.944); the
relative bioavailability F is unaffected by preparation state because
absorption rate does not enter the area under the curve at the dose
levels studied per Willmann 2021 Discussion p. 1201 ‘A lower rate of
absorption of the undiluted ready-to-use suspension was found in this
model … but the extent of absorption is not affected by ka and, thus,
AUC(0-24h),ss is independent of the formulation’).FORM_* family alongside
FORM_SUSPENSION (extemporaneous suspension),
FORM_GRANULE (granules for reconstitution), and the
tablet/capsule/solution siblings, capturing a third type of
liquid-administration contrast (preparation state of the same
manufactured product). Future ready-to-use-suspension
dilution-preparation extractions should reuse this canonical with the
per-paper comparator pool documented in
covariateData[[FORM_UNDILUTED_SUSP]]$notes. Promote to
general scope when a second paper ratifies the same encoding. Ratified
canonically alongside the Willmann 2021 rivaroxaban pediatric popPK
extraction.covariateData[[FORM_GRANULE]]$notes). Per-subject
(regimen-fixed) categorical indicator typical of pediatric-development
popPK pooled analyses.FORM / FORMULATION – used in
Ayyoub_2016_pyronaridine.R (paper’s
formulation indicator with the granule formulation as the 1
level; matches the canonical encoding directly without value
transformation).Ayyoub_2016_pyronaridine.R (additive-multiplier effect on
the typical absorption rate constant Ka:
ka_typ = exp(lka) * (1 + e_form_granule_ka * FORM_GRANULE)
with e_form_granule_ka = +1.63 (granule Ka is 2.63x the
tablet Ka, i.e. 47.1 1/day vs 17.9 1/day; Ayyoub 2016 Table 2, %RSE
37.8); CL/F, V2/F, V3/F, and Q/F are held identical between formulations
because backward elimination (P < 0.001) retained formulation only on
Ka).FORM_SUSPENSION (Svensson 2018: tablets
crushed-and-suspended) and from FORM_TABLET /
FORM_SOLUTION / FORM_POWDER (different sibling
formulation contrasts). Future pediatric-granule popPK extractions (a
common pattern in antimalarial, antiretroviral, and antitubercular
pediatric trials – e.g. Pyramax granules, Coartem dispersible,
lopinavir-ritonavir pellets) should reuse this canonical and extend the
example list. Promote to general scope when a second paper ratifies the
same encoding. Ratified canonically alongside the Ayyoub 2016
pyronaridine extraction.covariateData[[FORM_POWDER]]$notes).FORM_POWDER – used in
Yukawa_1990_phenytoin.R (paper’s BA indicator
inverted: source BA = 1 if tablet, 0 if powder; canonical
FORM_POWDER = 1 - BA_indicator so 0 is the tablet
reference).Yukawa_1990_phenytoin.R (Yukawa 1990 Model 2 dose-dependent
powder bioavailability
F_powder = 1 - exp(-9.92 / DOSE_PHT_MGKGD); tablet F fixed
at 1), Retlich_2015_linagliptin.R (multiplicative shift on
the linagliptin first-order absorption rate constant Ka:
powder-in-bottle Ka = 0.933 1/h vs tablet formulation 2 reference Ka =
0.441 1/h; the tablet formulation 1 comparator is captured by the
sibling canonical FORM_LINAG_TAB1),
Hong_2011_atazanavir.R (multiplicative linear-deviation
effect on relative bioavailability:
fdepot *= (1 + e_form_powder_frel * FORM_POWDER) with
e_form_powder_frel = -0.355, i.e. the pediatric atazanavir
oral powder has 35.5% lower bioavailability than the capsule reference;
Hong 2011 Table 4).FORM_TABLET (Kyhl 2016 / Tikiso 2021 tablet vs solution)
and FORM_CAPSULE (Hennig 2006 / Hennig 2007 capsule vs
solution) under the FORM_* family. Future
powder-formulation models should reuse this canonical, extending the
example list and documenting the per-paper comparator. Ratified
canonically on 2026-05-10 alongside the Yukawa 1990 phenytoin
extraction.FORM_APNECUT – prior canonical name (pre-2026-06-19
standardization audit).AP – used in Suda_2008_theophylline.R
(Suda 2008 NONMEM AP indicator; same orientation, no
transformation; AP = 1 if Apnecut, 0 if theophylline alcohol).Suda_2008_theophylline.R (multiplicative effect on CL/F per
Suda 2008 final model page 638:
cl <- exp(lcl + etalcl) * (WT / 1)^1.08 * (1 + e_form_apnecut_cl * FORM_THEO_APNECUT)
with e_form_apnecut_cl = -0.282; the APC formulation has
approximately 0.71x the CL/F of the TA reference, equivalent to
approximately 1.41x higher dose-normalised exposure consistent with the
higher trough concentrations the authors observed clinically that
motivated the analysis).FORM_POWDER/FORM_SYRUP/FORM_TABLET/FORM_CAPSULE
because the contrast here is between two specific liquid drug products
rather than between dosage-form categories. Suda 2008 attributes the
CL/F difference primarily to absorption (per Discussion, page 641: HPLC
content analysis confirmed both products were within label, so the
formulation effect on apparent oral clearance is interpreted as a
bioavailability difference rather than an actual clearance difference)
but the parameter is encoded as an effect on CL/F to match the published
equation. Mirrors the sibling drug-product-version FORM_*
entries (FORM_SAR_DP2 sarilumab, FORM_ISA_P2F2
isatuximab, FORM_LINAG_TAB1 linagliptin,
FORM_VISMO_PHASEI vismodegib, FORM_TAC_IR
tacrolimus) under the FORM_* family. Set to 0 to simulate
the TA reference; in the absence of TA-specific dosing in a downstream
cohort, set FORM_THEO_APNECUT = 1 to represent Apnecut. Ratified
canonically on 2026-05-24 alongside the Suda 2008 theophylline
extraction.covariateData[[FORM_SYRUP]]$notes).
Per-dose-occasion indicator in principle; in paediatric cohorts often
time-fixed per subject by clinical convention.syrup formulation – used in
Nanga_2019_tacrolimus_metaanalysis.R (Table 3
covariate-effect label; relative bioavailability of syrup vs capsule =
0.53).FORM (dry syrup vs tablet) – used in
Sarashina_2005_epinastine.R (Table 4 theta_9 multiplier on
CL/F: dry-syrup / tablet ratio 1.06; paediatric arm received dry syrup
only, adult arm received tablet or dry syrup).Nanga_2019_tacrolimus_metaanalysis.R (multiplicative effect
on bioavailability per Eq. 4:
f(depot) <- 0.53^FORM_SYRUP, so capsule users have F = 1
and syrup users have F = 0.53; Nanga 2019 Table 3 ‘Bioavailability for
syrup formulation’ = 0.53 with 95% bootstrap CI 0.31 - 0.75),
Sarashina_2005_epinastine.R (multiplicative ratio 1.06 on
CL/F for dry syrup vs tablet; paediatric atopic-dermatitis patients all
received dry syrup, healthy adults received either tablet or dry syrup;
tablet is the reference (FORM_SYRUP = 0) and dry syrup is FORM_SYRUP =
1).FORM_SUSPENSION (Svensson 2018: tablets
extemporaneously suspended in water at bedside immediately before
swallowing – same tablet swallowed two different ways, where the
manipulation affects MAT rather than F) because here the contrast is
between two distinct drug products (commercial capsules vs paediatric
syrup / suspension). Distinct from FORM_CAPSULE,
FORM_TABLET, and FORM_POWDER, which compare
those solid-oral forms against a liquid solution or against each other.
Future paediatric-syrup / oral-suspension formulation comparisons should
reuse this canonical; if a future model contrasts syrup against tablet
(rather than capsule), extend the per-model notes rather than
registering a sibling canonical. Ratified canonically on 2026-05-18
alongside the Nanga 2019 tacrolimus meta-analysis extraction.Ahn_2014_parathyroidHormone.R (multiplicative effect on the
depot bioavailability per Ahn 2014 Table 2 ‘Relative F1’ = 1.98:
fdepot = exp(lfdepot + e_form_caco3_fdepot * FORM_CACO3)
with lfdepot fixed at log(1) for the thermal-water
reference and e_form_caco3_fdepot = log(1.98) for the CaCO3
arm; 95% CI on relative F1 1.06-2.90).FORM_TABLET) – the
0-level here is calcium-containing spring water naturally rich in
dissolved calcium, not a manufactured aqueous solution of the modelled
drug. Future calcium-bioavailability studies that compare CaCO3 against
a different calcium source (citrate, gluconate, dietary-calcium
baseline) should reuse this canonical with the reference clearly
documented in per-model notes, or register a sibling canonical if the
comparator is meaningfully different. Sits in the FORM_*
family alongside FORM_FDC (Wilkins 2008 rifampicin
co-formulation), FORM_SYRUP (Nanga 2019 tacrolimus
paediatric syrup), and FORM_TABLET (Kyhl 2016 nalmefene /
Tikiso 2021 abacavir tablet vs solution). Ratified canonically on
2026-05-21 alongside the Ahn 2014 parathyroid-hormone calcium-absorption
extraction.covariateData[[FORM_FDC]]$notes. Reference values observed:
1 = FDC (Wilkins 2008 rifampicin + isoniazid + pyrazinamide +/-
ethambutol antitubercular FDC); 0 = single-agent metformin tablet (Choi
2018 metformin vs metformin-containing antidiabetic FDC).FDC – used in Wilkins_2008_rifampicin.R
(DDMODEL00000280 NMTRAN $INPUT column; values 0 / 1 with
the same orientation as the canonical, 1 = FDC).formulation (lower-case as printed in the paper) – used
in Choi_2018_metformin.R (Choi 2018 Methods ‘Covariate
analysis’ equation: formulation = 0 single-agent reference, formulation
= 1 FDC test).Wilkins_2008_rifampicin.R (multiplicative
(1 + e_fdc0_mtt * (1 - FORM_FDC)) shift on MTT and
(1 + e_fdc0_cl * (1 - FORM_FDC)) shift on CL; SDC subjects
(FDC = 0) had 104% longer MTT and 23.6% higher CL than the FDC = 1
reference per Wilkins 2008 final estimates. Antitubercular
co-formulation: rifampicin + isoniazid + pyrazinamide +/-
ethambutol).Choi_2018_metformin.R (multiplicative effects on
first-order absorption rate Ka and on relative bioavailability F:
ka *= 0.83^FORM_FDC (Ka shrinks to 83.0% of the
single-agent value for FDC) and f_rel *= 0.94^FORM_FDC
(relative bioavailability shrinks to 94.0% of the single-agent value for
FDC) per Choi 2018 Table 3 final estimates. Antidiabetic co-formulation:
the FDC drug-product is metformin co-formulated with an unspecified
second antidiabetic agent – typical Korean metformin FDCs co-formulate
with sitagliptin / glimepiride / vildagliptin / dapagliflozin; the paper
does not name the specific co-formulant. Reference category 0 =
single-agent metformin tablet).covariateData[[FORM_FDC]]$notes. The mechanism is
co-formulation-driven perturbation of absorption (excipient
interactions, dissolution rate, gastric residence) rather than
drug-product manufacturing of the single drug – distinct from
FORM_TABLET (Kyhl 2016 nalmefene tablet vs solution) and
the rest of the FORM_* (drug-product-version) family
because the FDC-vs-single-drug-tablet contrast compares two tablet
products that both contain the modelled drug, not a tablet vs a
non-tablet. When the source paper’s typical-value parameters are
anchored to the FDC arm, set the per-paper note accordingly; when
anchored to the single-drug-tablet arm, set the per-paper note
accordingly. Both orientations are valid and the canonical column
semantics (1 = FDC, 0 = single-drug tablet) preserve a consistent input
column across simulations.Kyhl_2016_nalmefene.R.IMMUNOASSAY
below.IMMUNOASSAY / IA / ASSAY –
common NONMEM $INPUT forms.Andrews_2017_tacrolimus.R (per-sample binary indicator
switching the additive + proportional residual error between immunoassay
(9% of samples; pre-2013 microparticle / chemiluminescence immunoassay)
and LC-MS/MS (91% of samples; lower LLOQ 1.0 ng/mL vs 1.5 ng/mL) – both
magnitudes estimated jointly in the final model per Andrews 2017 Section
2.3 and Table 2).RIA_ASSAY canonical (use that one when the paper identifies
the immunoassay as radioimmunoassay; use this IMMUNOASSAY
for MEIA / CMIA / EMIT / ELISA / generic immunoassay). The two siblings
could in principle be unified into a single IMMUNOASSAY
canonical with per-model notes documenting the immunoassay subtype, but
the existing RIA_ASSAY registration is preserved for
backwards compatibility with Kyhl_2016_nalmefene.R. Per-row
time-varying when the dataset spans the historical introduction of
LC-MS/MS at the analytical lab (e.g., Andrews 2017’s 2009-2016 sampling
window crosses the lab’s switch from immunoassay to LC-MS/MS; the
per-sample assay method is recorded). When the dataset is fully
LC-MS/MS, set IMMUNOASSAY = 0 for every row and the
immunoassay residual-error parameters become non-identifiable – consider
dropping them from the model. Ratified canonically on 2026-05-25
alongside the Andrews 2017 tacrolimus extraction.METH1 – used in Abrantes_2017_moroctocog.R
(Abrantes 2017 Table 2 footnote g: OSA-central indicator on the
bioavailability multiplier and on the proportional residual-error
magnitude).Abrantes_2017_moroctocog.R (multiplicative effects: F
multiplier (1 - 0.390 * ASSAY_OSA) so OSA-assayed
observations have F reduced by 39.0% relative to the CSA reference;
proportional residual SD multiplier (1 + 0.403 * ASSAY_OSA)
so OSA-assayed observations have 40.3% larger residual error
magnitude).ASSAY_OSA_LOCAL to
capture an additional inter-laboratory shift for OSA performed at
non-central laboratories. Ratified canonically on 2026-06-21 alongside
the Abrantes 2017 moroctocog extraction.ASSAY_OSA: when
ASSAY_OSA_LOCAL = 1, ASSAY_OSA is also 1 (the
local-lab samples are a subset of OSA-assayed samples). Used to capture
an additional inter-laboratory variability term on the bioavailability
multiplier.METH2 – used in Abrantes_2017_moroctocog.R
(Abrantes 2017 Table 2 footnote g: OSA-local indicator on the
bioavailability multiplier; only one study in the Abrantes 2017 13-study
pool, B1831003, used a local laboratory).Abrantes_2017_moroctocog.R (multiplicative effect on F:
(1 - 0.146 * ASSAY_OSA_LOCAL) so OSA-local-laboratory
observations have an additional 14.6% F reduction beyond the
central-laboratory OSA effect captured by ASSAY_OSA).ASSAY_OSA. When the pooled dataset contains no
local-laboratory observations, set ASSAY_OSA_LOCAL = 0 for
every row and the corresponding F multiplier collapses to 1. Ratified
canonically on 2026-06-21 alongside the Abrantes 2017 moroctocog
extraction.FORM_NS0 – prior canonical name (pre-2026-06-19
FORM_Zhu_2017_lebrikizumab.R.FORM_NS0 to
FORM_LEB_NS0 on 2026-06-19 per the canonical-register
standardization audit (operator decision: insert the drug stem so the
formulation indicator is namespaced to the specific drug, matching the
FORM_<DRUG>_<FEATURE> pattern used by
FORM_DOX_DORYX_MPC, FORM_ITR_SUBA, etc.).FORM_CHO_PHASE2 – prior canonical name (pre-2026-06-19
FORM_Zhu_2017_lebrikizumab.R.FORM_CHO_PHASE2 to
FORM_LEB_CHO_PHASE2 on 2026-06-19 per the
canonical-register standardization audit (operator decision: insert the
drug stem so the formulation indicator is namespaced to lebrikizumab
specifically, matching the
FORM_<DRUG>_<FEATURE> pattern).COMB (used by Yamada
2025 Table 1 with the EOX level coded as the non-reference category;
renamed to CONMED_EOX to preserve the semantic meaning of
the 1-level).Yamada_2025_zolbetuximab.R (fractional effect on V1).COMB_CAPOX, COMB_FOLFOX, …) with a single
reference group.Passey_2011_tacrolimus.R (multiplicative effect on apparent
oral CL: cl_typ * e_ccb_cl^CONMED_CCB with
e_ccb_cl = 0.812, i.e. CCB coadministration reduces
tacrolimus CL/F by ~19% in adult kidney transplant recipients; Passey
2011 final model Table 4 row “CCB”).atv atorvastatin,
flv fluvastatin, lov lovastatin,
prv pravastatin, rsv rosuvastatin,
smv simvastatin, ezt ezetimibe,
inh isoniazid). 0 = the named drug is not part of the
regimen for this study arm / subject; positive value = total daily dose.
Captures the dose-response amplitude in MBMA or co-administered-drug
PK/PD models where each drug arm contributes its own dose-effect
curve.CONMED_INH_DOSE (paper-specific unit, documented
in covariateData[[CONMED_INH_DOSE]]$units).DOSE_<drug>
legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename,
Chen 2017 TB mouse pre-rename); aliases of the canonical
CONMED_<drug>_DOSE form per the 2026-05-28 naming
audit rename.Vargo_2014_statins_ezetimibe_mbma.R (each
CONMED_<statin>_DOSE drives the corresponding
statin’s Hill / Emax dose-response curve; per-arm dose level in mg/day,
default 0 outside the named statin arm; ezetimibe and statin arms
combine via the sub-additive gamma_int term),
Chen_2017_TB_MTP_GPDI_mouse.R (CONMED_INH_DOSE
drives the isoniazid CL adjustment:
cl_inh = cl_inh_lowdose * (1 - slope_inh * (CONMED_INH_DOSE - 12.5))).CONMED_<drug>_DOSE
shape replaces the earlier DOSE_<drug> /
DOSE_<drug>_<unit> names (which conflated a
covariate with a dose-amount column). New co-medication dose-effect MBMA
models should reuse this family and add the appropriate
<drug> INN abbreviation. The units field is
per-paper. Ratified 2026-05-28 per the naming audit.BID_XR_FLV legacy form
(used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the
canonical FORM_FLV_BID_XR form per the 2026-05-28 naming
audit rename.Vargo_2014_statins_ezetimibe_mbma.R (multiplies the
fluvastatin ED50 by 0.645 when set to 1; Vargo 2014 Table 3 row
“ED50,fluvastatin (b.i.d.|XR) / ED50,fluvastatin”; same ratio used for
either regimen because the paper found b.i.d. and XR ED50 estimates were
similar).FORM_LOV_BID_XR (lovastatin)
under the FORM_<drug>_BID_XR family pattern. Ratified
2026-05-28 per the naming audit.BID_XR_LOV legacy form
(used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the
canonical FORM_LOV_BID_XR form per the 2026-05-28 naming
audit rename.Vargo_2014_statins_ezetimibe_mbma.R (multiplies the
lovastatin ED50 by 0.59 when set to 1; Vargo 2014 Table 3 row
“ED50,lovastatin (b.i.d.|XR) / ED50,lovastatin”).FORM_FLV_BID_XR (fluvastatin)
under the FORM_<drug>_BID_XR family pattern. Ratified
2026-05-28 per the naming audit.ACS legacy form (used
in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the
canonical DIS_ACS form per the 2026-05-28 naming audit
rename.Vargo_2014_statins_ezetimibe_mbma.R (additive shift of
-0.117 on the statin Emax in ACS arms; greater statin LDL-C lowering in
ACS patients than in the non-ACS reference).DIS_HEFH (HeFH cohort indicator) and the broader
DIS_<indication> family. Ratified 2026-05-28 per the
naming audit.HEFH legacy form (used
in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the
canonical DIS_HEFH form per the 2026-05-28 naming audit
rename.Vargo_2014_statins_ezetimibe_mbma.R (additive shift of
+0.127 on the statin Emax in HeFH arms; smaller statin LDL-C lowering in
HeFH patients than in the non-HeFH reference; biologically consistent
with the LDLR-pathway disruption in HeFH),
Kuchimanchi_2018_evolocumab_ldlc.R (multiplicative exponent
1.28 on baseline LDL-C in the evolocumab Emax-on-AUC exposure-response
layer, Table 4; HeFH patients have higher baseline LDL-C than the
non-HeFH reference).DIS_HOFH (homozygous form, Pu_2021_evinacumab) and the
broader DIS_<indication> family. Ratified 2026-05-28
per the naming audit.rif rifampicin, inh isoniazid,
emb ethambutol, str streptomycin,
cab capreomycin / a second antibiotic, col
colistin, mer meropenem, gen gentamicin,
cip ciprofloxacin). Used as the driving exposure for the
corresponding drug-effect term in combination PK/PD or in vitro
time-kill models. 0 = the named drug is not part of the regimen for this
subject / arm.covariateData[[CONMED_<drug>_CC]]$units.Ccol /
Cmer / Cgen / Ccip (Mohamed 2016,
Sadouki 2025) and bare EMB / INH /
RIF / STR / CAB (Clewe 2018, Khan
2015) – legacy forms used before the 2026-05-28 naming-audit rename. All
map to the canonical CONMED_<drug>_CC form.Clewe_2018_rifampicin.R (in vitro time-kill of rifampicin +
isoniazid + ethambutol against M. tuberculosis; each drug exposure is a
fixed concentration driving the Hill-Emax effect on the F / S / N
bacterial states), Khan_2015_ciprofloxacin.R (in vitro
streptomycin + capreomycin time-kill),
Mohamed_2016_colistin_meropenem.R (in vitro colistin +
meropenem time-kill against WT and meropenem-resistant P. aeruginosa),
Sadouki_2025_meropenem_gentamicin_ciprofloxacin.R
(combination PD with meropenem / gentamicin / ciprofloxacin time-varying
exposures).CONMED_<drug>_DOSE (registered above). The
_CC suffix distinguishes the dynamic concentration
covariate from the dose-level covariate; both share the
<drug> INN abbreviation. Ratified 2026-05-28 per the
naming audit.CONMED_<drug>_CC concentration
covariate is then unused / 0).MER_PRESENT,
GEN_PRESENT, CIP_PRESENT (Sadouki 2025) – the
legacy _PRESENT suffix is dropped because binary semantics
are conveyed by the type field; renamed 2026-05-28 per the naming audit.
GENT (Cohen-Wolkowiez 2014) maps to CONMED_GEN
(1 if gentamicin was given concurrently with piperacillin-tazobactam, 0
otherwise).Sadouki_2025_meropenem_gentamicin_ciprofloxacin.R (used
together to compute a combination-of-three indicator that triggers a
categorical -1 shift on the BETA parameter when all three antibiotics
are co-administered), CohenWolkowiez_2014_tazobactam.R
(CONMED_GEN enters the tazobactam clearance covariate
equation as a multiplicative power-form factor
1.52^CONMED_GEN; in the source paper the gentamicin
coadministration covariate was found to be confounded by postmenstrual /
postnatal age differences between infants who received gentamicin and
those who did not, so the effect should be interpreted as a marker of an
older / sicker subset rather than a mechanistic drug-drug
interaction).CONMED_<drug>_CC (above). For a given drug,
CONMED_<drug> is the binary “coadministered?”
indicator and CONMED_<drug>_CC is the time-varying
plasma concentration. Scope promoted to general on
2026-06-01 alongside the Cohen-Wolkowiez 2014 piperacillin-tazobactam
extraction; the underlying semantic (binary coadministration of a named
antibiotic) is the same whether the indicator gates a PD interaction
term or scales a PK covariate equation, so a single canonical name is
preferred over splitting into PD-only and PK-only variants. Ratified
2026-05-28 per the naming audit; scope updated 2026-06-01.covariateData[[BACT]]$notes.Mohamed_2016_colistin_meropenem.R (1 = ARU552
meropenem-resistant clinical isolate, 2 = ATCC 27853 wild-type; hard
switch on 25+ strain-specific structural parameters).Sadouki_2025_meropenem_gentamicin_ciprofloxacin.R (gates a
low-inoculum-specific shift on the bacterial-kill amplitude; the paper
found the low-inoculum arm had a measurably different kill rate from the
standard-inoculum arm).If – used in Boucher_2018_naproxen_mbma.R
(Boucher 2018 Eqs 2-3).Boucher_2018_naproxen_mbma.R (shifts both baseline WOMAC
pain e0 and emax; 12 of 18 trials were flare
designs).checkModelConventions() recognises it).In – used in Boucher_2018_naproxen_mbma.R
(Boucher 2018 Eqs 3-4).Boucher_2018_naproxen_mbma.R (shifts emax and
shortens et50; all 18 included trials had both a naproxen
and a placebo arm).checkModelConventions() recognises it).QBL = baseline (off-HD / off-CRRT) renal Q (L/h); the
typical patient’s residual renal clearance.QEFF = effective on-HD/CRRT renal Q (L/h); the apparent
dialysis-augmented clearance during a session (including the effective
renal Q plus the dialyser/filter clearance, depending on the paper’s
parameterisation).Leuppi-Taegtmeyer_2019_colistin.R (colistin /
colistimethate sodium popPK in CRRT recipients; QBL drives the off-CRRT
typical-value CL, QEFF drives the on-CRRT augmented CL during the
session window).CHD_PCT legacy form
(used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the
canonical DIS_CHD_PERCENT form per the 2026-05-28 naming
audit rename.Vargo_2014_statins_ezetimibe_mbma.R (linear coefficient
e_chd_emax_statin = -0.000649 per percentage point on
Emax_statin, i.e. a 24% CHD arm reduces Emax_statin by
0.000649 * 24 = 0.016; the paper’s typical-patient
definition uses 24% CHD).FORM_P2F2 – prior canonical name (pre-2026-06-19
standardization audit).Drug_mat – used in Fau_2020_isatuximab.R.
Values 0 / 1 with the same orientation as the canonical (1 = P2F2 /
commercial-bound material).Fau_2020_isatuximab.R
(exponential effect on Vc with coefficient -0.137; P2F2 patients had
~13% lower Vc than P1F1).FORM_POWDER; this indicator switches between the two tablet
formulations.FORM_POWDER = 1 for the powder).Retlich_2015_linagliptin.R (multiplicative shift on the
linagliptin first-order absorption rate constant Ka; typical Ka = 0.441
1/h for tablet formulation 2 (reference), 0.795 1/h for tablet
formulation 1, 0.933 1/h for the powder formulation).FORM_SAR_DP2 (sarilumab) and FORM_ISA_P2F2
(isatuximab) entries under the FORM_* family. Set to 0 for
routine marketed-formulation simulation. Ratified canonically alongside
the Retlich 2015 linagliptin extraction.FORM_DP2 – prior canonical name (pre-2026-06-19
standardization audit).DP2 – used in Xu_2019_sarilumab.R.Xu_2019_sarilumab.R.vanIersel_2018_posaconazole.R (multiplicative effect on
relative bioavailability F1:
fdepot *= (1 + e_formposaab_fdepot * FORM_POSA_AB) with
e_formposaab_fdepot = +0.247, i.e. the prototype A/B
formulations have 24.7% higher F1 than the later C/D reference; van
Iersel 2018 Table 2 final-model ‘Tablet formulation A/B on F1’ = 0.247,
RSE 21.5%).FORM_<drug>_<contrast> family alongside
FORM_ASV_LIQUID (asunaprevir), FORM_SAR_DP2
(sarilumab), FORM_VISMO_PHASEI (vismodegib),
FORM_LINAG_TAB1 (linagliptin), FORM_ABA_PHASE2
(abatacept). Set to 0 for routine commercial-formulation simulation
(tablet D). Ratified canonically alongside the van Iersel 2018
posaconazole extraction.ka reference in Eq. 4).Lu_2015_vismodegib.R
(multiplicative shifts on Ka and relative bioavailability F: typical Ka
= 9.025 1/day for the Phase II reference in patients, with
exp(-0.602) = 0.55x for the Phase I formulation and
exp(0.671) = 1.96x for healthy volunteers; F = 1 for the
Phase II reference, F = 0.346 for the Phase I formulation in patients
and 0.836 in HV).FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2
(isatuximab), and FORM_LINAG_TAB1 (linagliptin) entries
under the FORM_* family of drug-product-version indicators.
Set to 0 for routine commercial-formulation simulation. Ratified
canonically alongside the Lu 2015 vismodegib extraction.FORM_CAPSULE).
Per-subject (regimen-fixed) categorical indicator used in popPK analyses
that pool the three Doryx formulations and test formulation as a
covariate on relative bioavailability, absorption lag, and the transit
absorption rate’s food effect.FORM_DORYX_MPC – prior canonical name (pre-2026-06-19
standardization audit).FMPC – used in Hopkins_2017_doxycycline.R
(Hopkins 2017 Methods ‘Base model’ paragraph: “Structural covariates … a
formulation effect (for Doryx MPC, FMPC; for Doryx capsule, FCAP) on
relative bioavailability (RELF)”).Hopkins_2017_doxycycline.R (multiplicative effects per
Hopkins 2017 Table 3 final model: relative bioavailability
F1MPC = 0.863 vs the Doryx-tablet reference (Theta 8);
shared absorption lag ALAG1 = 0.115 h with the Doryx
capsule (Theta 12); strengthened food effect on the transit rate
constant KTR: fed state reduces KTR by 54.9 % for Doryx MPC
(COVFED2 = -0.549, Theta 11) vs 20.9 % for Doryx tablet /
capsule (COVFED = -0.209, Theta 10)).FORM_CAPSULE (the Doryx capsule
indicator) so that both indicators = 0 selects the Doryx-tablet
reference; the two indicators are mutually exclusive at any given dose
record. Mirrors the existing FORM_SAR_DP2 (sarilumab),
FORM_ISA_P2F2 (isatuximab), FORM_LINAG_TAB1
(linagliptin), FORM_VISMO_PHASEI (vismodegib),
FORM_TAC_IR (tacrolimus IR vs PR), and
FORM_THEO_APNECUT (theophylline) entries under the
FORM_* family of drug-product-version indicators. Set to 0
for routine Doryx-tablet simulation, 1 for Doryx MPC simulation.
Ratified canonically on 2026-06-02 alongside the Hopkins 2017
doxycycline extraction.FORM_SUBA – prior canonical name (pre-2026-06-19
standardization audit).DRUG – used in
Abuhelwa_2015_itraconazole.R (Appendix S1 NONMEM control
stream:
IF (DRUG.EQ.0) THEN DRUGF = 1 (Sporanox) ELSE DRUGF = (1 + THETA(8)) (SUBA-itraconazole)).Abuhelwa_2015_itraconazole.R (multiplicative effects per
Abuhelwa 2015 Table 3: SUBA-itraconazole has +73% relative
bioavailability vs Sporanox (F = 1 + 0.729 x FORM_ITR_SUBA)
and 21.3% less F variability
(ETASCALE = 1 + (-0.213) x FORM_ITR_SUBA = 0.787 scales the
shared FVAR random effect)).FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2
(isatuximab), FORM_LINAG_TAB1 (linagliptin), and
FORM_VISMO_PHASEI (vismodegib) entries under the
FORM_* family of drug-product-version /
formulation-comparison indicators. Distinct from
FORM_CAPSULE (Hennig 2006 / 2007 capsule-vs-solution
contrast for the same itraconazole molecule) because here both arms are
capsule formulations and the contrast is between two capsule drug
products. Ratified canonically alongside the Abuhelwa 2015 itraconazole
extraction.study – used in Woillard_2011_tacrolimus.R
(Woillard 2011 Methods: “study factor (assumed to be similar to drug
formulation) … study = 1 for the Prograf cohort, study = 0 for the
Advagraf cohort”). The source paper’s study and
formulation factors are aliased; the canonical preserves
the source paper’s orientation (Prograf = 1 = IR, Advagraf = 0 = PR),
which also matches the standard convention of treating the older
immediate-release formulation as the non-reference level.FORMULATION (with 1 = PR-T, opposite polarity) – used
in Lu_2019_tacrolimus_industry_meta.R. Lu 2019 writes the
Ka covariate equation as
Ka = 0.375 * (1 - (1 - theta_form) * FORMULATION) with
FORMULATION = 1 for prolonged-release; the model derives
form_pr = 1 - FORM_TAC_IR inside model() to
match the paper’s published equation while keeping the canonical column
oriented Prograf = 1.Woillard_2011_tacrolimus.R (multiplicative effects per
Woillard 2011 Table 4: Ktr = theta1 * theta2^FORM_TAC_IR
with theta1 = 3.34/h and theta2 = 1.53
(Prograf ~53 % faster absorption than Advagraf);
Vc/F = theta6 * theta7^FORM_TAC_IR with
theta6 = 486 L and theta7 = 0.29 (Prograf Vc/F
is 29 % of the Advagraf reference)),
Lu_2019_tacrolimus_industry_meta.R (linear effect on Ka per
Lu 2019 Table 3 ‘Prolonged-release tacrolimus on Ka’ = 0.499, encoded as
Ka(PR-T) = Ka(IR-T) * 0.499 via
form_pr = 1 - FORM_TAC_IR; 50 % slower absorption for PR-T
vs IR-T).general after
Lu 2019 corroborated the Prograf-vs-Advagraf contrast in a second,
larger-cohort population (Woillard 2011 n = 173 + 174, Lu 2019 n = 408
across 8 Astellas Phase II studies). Mirrors the existing
FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2
(isatuximab), FORM_LINAG_TAB1 (linagliptin), and
FORM_VISMO_PHASEI (vismodegib) entries under the
FORM_* family of drug-product-version indicators. The
Woillard 2011 paper notes the formulation effect partially confounds
with time-post-transplant (Prograf cohort sampled within the first 6
months post-transplant, Advagraf cohort > 12 months post-transplant);
Lu 2019 has no such confounding because most studies are within-subject
IR-T-to-PR-T conversions. Future tacrolimus models that include Envarsus
XR (modified-release once-daily granules) or LCP-Tacro
(life-cycle-pharma melt-extrusion tablets) should register a sibling
canonical (e.g., FORM_TAC_ENVARSUS) rather than overloading
FORM_TAC_IR. Ratified canonically alongside the Woillard
2011 tacrolimus extraction.PROD – used in Abrantes_2017_moroctocog.R
(Abrantes 2017 Table 2 footnote g and Methods: F multiplier
1.38^PROD so Xyntha effectively delivers 1.38x the labelled
dose in CSA-reference units).Abrantes_2017_moroctocog.R (multiplicative effect on F:
(1 + 0.38 * FORM_XYNTHA) – equivalent to
1.38^FORM_XYNTHA when binary; the model anchors F = 1 for
the CSA-reference Refacto/Refacto AF products and applies the +38.0%
multiplier when the product administered is Xyntha).FORM_<drug>_<variant>
family pattern (FORM_LEB_NS0 lebrikizumab cell-line,
FORM_SAR_DP2 sarilumab drug-product version,
FORM_DOX_DORYX_MPC doxycycline Doryx-MPC). Per-observation
(per-row) when a subject could switch between products across trials
(e.g., Abrantes 2017 study B1831066 enrolled both Refacto and Refacto AF
in the same protocol); subject-level otherwise. The 1.38 OSA-vs-CSA
potency ratio for B-domain-deleted FVIII is documented in Mikaelsson
1998 and Hubbard 2013 (cited in Abrantes 2017 reference 13). Ratified
canonically on 2026-06-21 alongside the Abrantes 2017 moroctocog
extraction.Soehoel_2022_tralokinumab.R.DILUTION. Kept as alias here to
match existing file.Soehoel_2022_tralokinumab.R.NON_ECZTRA or
STUDY_NON_ECZTRA.Clegg_2024_nirsevimab.R.(COHDOSE / ref)^exponent. Reference values observed: 1
mg/kg in Narwal 2013 (human IgG1), 3 mg/kg in Niloy 2026 (preclinical
mouse small molecule).DOSE – used in Narwal_2013_sifalimumab.R
(the paper’s Eq. 3 variable name; renamed to COHDOSE here
to avoid colliding with the rxode2/nlmixr2 event-column convention where
DOSE or AMT carries the administered
dose).Dose_i – used in
Niloy_2026_MTMSATrp_mouse.R (Niloy 2026 Methods Eq. 2
notation; the per-mouse mg/kg cohort dose).Narwal_2013_sifalimumab.R (reference 1 mg/kg, exponent
0.0542 on CL), Niloy_2026_MTMSATrp_mouse.R (preclinical
mouse; reference 3 mg/kg, exponent -0.30 on CL describing empirical
dose-dependent CL over 0.3-10 mg/kg).(COHDOSE / ref)^exponent on apparent
clearance and the shared escalating-cohort design (each subject on one
dose for the full study) carry across drug class, species, and study
size. For fixed-dose simulations in a weight-based-dosing paper, set
COHDOSE = nominal_dose_mg / WT per subject. When the
subject receives a per-kg dose label directly (preclinical mg/kg dosing
or human mg/kg cohort labels like the MI-CP152 0.3 / 1 / 3 / 10 mg/kg
arms), COHDOSE is the mg/kg label.AUC = DOSE / CLI) from a posthoc-CL covariate without
instantiating a PK ODE. The column is set to 0 during off-treatment
periods (drug holidays, placebo arm) so the derived exposure becomes
0.covariateData[[DOSE]]$units if a different dose unit is
used).(DOSE / ref)^exponent for use case (a), or directly inside
derived-exposure expressions for use case (b). Reference values
observed: 600 mg in Zheng_2016_sifalimumab.R (middle of the
200/600/1200 mg phase IIb dose range).Dose – used in Zheng_2016_sifalimumab.R
and Castro-Surez_2020_nimotuzumab.R.DOS – used in the Hansson 2013 sunitinib biomarker /
TGI / fatigue PD-model family (DDMODEL00000197 and siblings) as a
per-record sunitinib dose column.Zheng_2016_sifalimumab.R (power effect on V1 with exponent
0.06), Castro-Surez_2020_nimotuzumab.R (binary-indicator
usage (DOSE == 50) applying a 53 % decrease in V1 for the
50 mg cohort), Hansson_2013a_sunitinib.R (DDMODEL00000197;
time-varying record-level dose feeding AUC = DOSE / CLI),
Hansson_2013b_sunitinib.R (DDMODEL00000198; same
time-varying AUC = DOSE / CLI form for the tumor growth
inhibition model), Schindler_2016_sunitinib.R
(DDMODEL00000221; same AUC = DOSE / CLI form, with the
daily-dose column toggling between 50 mg/day on-cycle and 0 on
dose-holiday records), Schindler_2017_imatinib.R
(time-varying daily imatinib dose in mg/day feeding the size and density
drug-effect terms as Kdrug,S * (DOSE / 400) * exp(-k * t)
and Kdrug,D * (DOSE / 400), with 400 mg/day as the
reference normalisation), Girard_2012_pimasertib.R (linear
coefficient on the dropout-hazard log-rate:
exp(beta * DOSE) Weibull multiplier; per-subject daily
dose, observed range 1-255 mg).DOSE_70MG (binary
indicator for a specific dose group in a trinary-dose design) and from
the rxode2/nlmixr2 event column amt (which carries the
administered dose at dose events). For use case (a), the values are
typically time-fixed per subject; for use case (b), the values are
time-varying with on/off cycling – for sunitinib 4-weeks-on /
2-weeks-off cycling, set DOSE = nominal_daily_mg (e.g., 50)
during on-cycles and 0 during off-cycles or for the placebo arm.
Per-model covariateData[[DOSE]]$notes should state which
use case applies.EXPS = CD * 104.5 / CRL as an exposure surrogate driving an
Emax-style symptomatic effect on disease progression, not as a
power-form covariate.CD – column name used
in the DDMODEL00000223 input dataset
(Simulated_Novakovic_2016_multiplesclerosis_cladribine_irt.csv).Novakovic_2017_cladribine.R.DOSE (per-subject
assigned dose level, time-fixed) and COHDOSE (mg/kg cohort
label, time-fixed). CD is the cumulative dose
accrued at each timepoint, supplied as a time-varying covariate column
rather than via dosing events because the Novakovic 2017 model does not
carry an explicit cladribine-PK compartment. Scope: specific because the
constant 104.5 inside the exposure-surrogate equation is hard-coded for
cladribine in the source.covariateData[[TRT]]$notes, because the
same TRT column header carries different cohort semantics
in different studies.covariateData[[TRT]]$notes).TRT, ARM,
TRTARM, STUDY_ARM – common column-name
variants for the per-subject treatment-arm integer indicator; all map
directly to canonical TRT with no value transformation.
Per-model integer coding is documented inline.Novakovic_2017_cladribine.R.
Novakovic_2017_cladribine.R (DDMODEL00000223): 0 =
placebo, 1 = cladribine 3.5 mg/kg cumulative-dose cohort, 2 = cladribine
5.25 mg/kg cumulative-dose cohort. Reference = 0 (placebo). Gates the
symptomatic and protective drug-effect terms via
TRT >= 1 && t > 0; the categorical level (1
vs 2) is informational because the dose-response is driven by the
time-varying CD covariate and the per-cohort dosing
schedule, not by TRT itself.Struemper_2025_tumorsize_OS_nsclc.R: 1 = PEMBRO
(pembrolizumab, INTR@PID
LUNG 037, n = 152), 2 = FELAD (feladilimab, INDUCE-1, n = 52), 3 = CHEMO
(docetaxel, Entree Lung Part 2, n = 34), 4 = FELAD+CHEMO (pooled
INDUCE-1 + Entree Lung Part 2, n = 78), 5 = IO-COMBO (feladilimab + IO;
pooled INDUCE-1 + INDUCE-2, n = 23), 6 = DOSTAR (dostarlimab, GARNET, n
= 67), 7 = DOSTAR+CHEMO (PERLA, n = 121), 8 = PEMBRO+CHEMO (PERLA, n =
122), 9 = COBO100MG+DOSTAR cohort B (AMBER, n = 14), 10 =
COBO300MG+DOSTAR cohort B (AMBER, n = 41), 11 = COBO900MG+DOSTAR cohort
B (AMBER, n = 29), 12 = COBO300MG+DOSTAR cohort D (AMBER, n = 53).
Reference = 1 (PEMBRO; the largest single-agent PD-1 cohort and the
model’s canonical default). Selects per-arm TVKG and
TVKS via a 12-branch if-else cascade and gates the
PDL1_TUM effect on ks via a derived
has_pd1 indicator equal to 1 when TRT is 1 or
in {6, 7, 8, 9, 10, 11, 12}.TRT
canonical with different integer codings: Novakovic 2017 (3 levels,
placebo reference) and Struemper 2025 (12 levels, pembrolizumab
reference). The column header is shared because both source NONMEM
control streams use TRT; the integer-to-cohort mapping is
intentionally per-model. Scope upgraded from specific to
general on 2026-06-22 when Struemper 2025 added a second
consuming model. Future models adding new TRT codings
extend the Example-models list rather than registering a new canonical,
so long as the column header is TRT; a model that uses a
fundamentally different semantics (e.g., a generic on-treatment 0/1 flag
distinct from per-arm cohort indexing) should register a new canonical
name (e.g., ON_TREATMENT) rather than reusing
TRT.antibody type – Zhang 2022 Results section 3.4
covariate-effect prose (“antibody type was determined as the covariate
significantly affecting the model”); the source NMTRAN column name is
not separately reported.Zhang_2022_ormutivimab.R (additive typical-value shifts on
the linear-scale Emax and ET50 of the vaccine-induced RVNA Emax model:
Emax_tv = exp(lEmax) + e_drug_ormu_Emax * DRUG_ORMU with
e_drug_ormu_Emax = +0.143 IU/mL and
ET50_tv = exp(lET50) + e_drug_ormu_ET50 * DRUG_ORMU with
e_drug_ormu_ET50 = -3.8 day, yielding a higher and faster
vaccine-induced antibody peak in the Ormutivimab arms relative to the
HRIG comparator).FORM_* family
(within-product formulation-version contrasts of a single drug) because
the contrast here is between two biologically distinct products: HRIG is
a polyclonal plasma-derived immunoglobulin, while Ormutivimab is a
recombinant monoclonal antibody (CHO-cell-produced; the first rhRIG
approved in China). Specific scope because the head-to-head
HRIG-vs-rHRIG comparator design is tied to the Zhang 2022 phase II
rabies-vaccine study. Future head-to-head rhRIG-vs-HRIG popPD models
(e.g., SII Rabishield or Twinrab against HRIG) should register a sibling
canonical (DRUG_SIIRMAB, DRUG_TWINRAB) rather
than overloading DRUG_ORMU; cross-product comparisons that
need both indicators in the same dataset can carry them as independent
binaries with HRIG as the shared reference. Ratified canonically
alongside the Zhang 2022 ormutivimab extraction.Shoji_2017_fosdagrocorat_p1np.R,
Shoji_2017_fosdagrocorat_oc.R (binary multiplier switching
the active KDE and the active sigmoid-Emax inhibition parameters between
fosdagrocorat and prednisone:
lkde_active = lkde_fos * (1 - DRUG_PRED) + lkde_pred * DRUG_PRED,
with analogous switching of logitimax_active /
imax_active and ledk50_active; the rebound
parameters RBmax, T50 and the response-side
parameters Kd, BL, SLP are shared
between the drugs and do not multiply by DRUG_PRED).DRUG_ORMU (rhRIG vs
HRIG) and follows the same “head-to-head drug-arm” pattern: the contrast
is between two structurally distinct active comparators in a randomized
trial. Specific scope because the fosdagrocorat-vs-prednisone
head-to-head is tied to the Shoji 2017 P1NP / OC analyses; future
DAGR-vs-prednisone (or DAGR-vs-other glucocorticoid) popPK/PD models can
extend this entry’s example_models list, while a contrast
between a different test drug and prednisone (e.g., methylprednisolone
vs prednisone) should register a sibling canonical rather than reusing
DRUG_PRED. Ratified canonically alongside the Shoji 2017
P1NP / OC extractions.Kotani_2022_astegolimab.R.70 mg vs {210 mg, 490 mg} combined reference.AMT) – used in Othman_2014_daclizumab.R and
the Diao 2016 family.I_Dose50mg – subject-level indicator derived from
study-arm randomization, used in
Jorga_2000_tolcapone_fluctuators.R for the 50 mg t.i.d.
tolcapone fluctuator arm.Othman_2014_daclizumab.R (record-level SC),
Diao_2016_daclizumab_cd25.R,
Diao_2016_daclizumab_cd56bright.R,
Diao_2016_daclizumab_treg.R,
Jorga_2000_tolcapone_fluctuators.R (subject-level oral
t.i.d.).e_dose_50mg_f = 0.57/0.84 - 1 = -0.321 scales
bioavailability only on 50 mg SC dose records. For clinical-range
simulation (150 mg SC Q4W Phase III regimen) leave
DOSE_50MG = 0. The Diao 2016 PK/PD models inherit the
Othman 2014 PK backbone verbatim. Jorga 2000 uses the indicator
subject-level on the central and peripheral volumes of distribution:
(1 + e_dose_50mg_vc_vp * DOSE_50MG) with
e_dose_50mg_vc_vp = -0.45 (V is 55% of the 200 mg reference
at the 50 mg t.i.d. arm); paired with DOSE_400MG in the
same fluctuator model so the 200 mg arm is the joint reference (both
indicators = 0).DOSE_50MG = 0
to select the 200 mg t.i.d. reference).I_Dose400mg – subject-level indicator derived from
study-arm randomization, used in
Jorga_2000_tolcapone_fluctuators.R for the 400 mg t.i.d.
tolcapone fluctuator arm.Jorga_2000_tolcapone_fluctuators.R (subject-level oral
t.i.d.).DOSE_50MG and
DOSE_70MG; member of the DOSE_<N>MG
family of dose-level indicators. Jorga 2000 uses the indicator
subject-level on the central and peripheral volumes of distribution:
(1 + e_dose_400mg_vc_vp * DOSE_400MG) with
e_dose_400mg_vc_vp = +0.40 (V is 140% of the 200 mg
reference at the 400 mg t.i.d. arm). The dose-dependent V was an
empirical finding in the fluctuator cohort that the authors could not
confirm in the nonfluctuator cohort (only 200 and 400 mg arms were
enrolled there); the effect plausibly reflects a few high-V outliers in
the small-volume cohort rather than a true mechanistic dose-V
relationship (Jorga 2000 Discussion). Ratified canonically alongside the
Jorga 2000 tolcapone extraction.DOSE_260MG = 0 to select the no-AT2221 reference for ATB200
linear CL).Hajjar_2018_cipaglucosidase.R as a categorical
multiplier on ATB200 linear CL
(e_dose130mg_cl ^ DOSE_130MG).Hajjar_2018_cipaglucosidase.R (per-dose-record oral
co-medication indicator).DOSE_50MG,
DOSE_70MG, DOSE_260MG, and
DOSE_400MG; member of the DOSE_<N>MG
family of dose-level indicators. In Hajjar 2018 the indicator marks the
AT2221 (miglustat) co-administration dose paired with the same-occasion
20 mg/kg ATB200 IV infusion; mutual exclusivity with
DOSE_260MG is enforced by the study design (a single dose
occasion uses at most one AT2221 dose level). Ratified canonically
alongside the Hajjar 2018 ATB200 / AT2221 extraction.DOSE_130MG = 0 to select the no-AT2221 reference for ATB200
linear CL).Hajjar_2018_cipaglucosidase.R as a categorical
multiplier on ATB200 linear CL
(e_dose260mg_cl ^ DOSE_260MG).Hajjar_2018_cipaglucosidase.R (per-dose-record oral
co-medication indicator).DOSE_50MG,
DOSE_70MG, DOSE_130MG, and
DOSE_400MG; member of the DOSE_<N>MG
family of dose-level indicators. In Hajjar 2018 the indicator marks the
AT2221 (miglustat) co-administration dose paired with the same-occasion
20 mg/kg ATB200 IV infusion; mutual exclusivity with
DOSE_130MG is enforced by the study design. Ratified
canonically alongside the Hajjar 2018 ATB200 / AT2221 extraction.STUDY5 = 0 selects the pooled “other” residual
error).DVID = "study1" (character-valued study identifier;
STUDY1 = as.integer(DVID == "study1")) – legacy form
previously used in Cirincione_2017_exenatide.R.Cirincione_2017_exenatide.R.STUDY5. When both
are 0, the subject is in the pooled “other studies” residual-error
group.DVID = "study5" (character-valued study identifier;
STUDY5 = as.integer(DVID == "study5")) – legacy form
previously used in Cirincione_2017_exenatide.R.Cirincione_2017_exenatide.R.STUDY1. When both
are 0, the subject is in the pooled “other studies” residual-error
group.Cirincione_2017_exenatide_er.R.STUDY_MD indicator selects between them. For the phase III
external validation cohort (multi-dose weekly 2 mg ER for 24 weeks), use
STUDY_MD = 1 so the multi-dose f_rel and residual
magnitudes apply. Specific scope because the indicator is tied to the
AAPS J 2017 ER exenatide combined analysis (single-dose phase II +
multi-dose phase II) – a future pooled analysis with additional study
cohorts would justify its own canonical or a promotion to general.PKU-015 -> 1, PKU-004 ->
0).Qi_2014_sapropterin.R.STUDY_PKU015 indicator selects between them. Specific scope
because the indicator is tied to the BioMarin sapropterin
clinical-development program (PKU-004 = phase 3b extension, PKU-015 =
phase 3b pediatric).PHASE2 – prior canonical name (pre-2026-06-19
paper-specific-disambiguation standardization). Generic single-token
name collided with the parallel PHASE1 canonical from
Valenzuela 2025 that picks the opposite reference category.MORAb-003-001 -> 0, MORAb-003-002 ->
1).Farrell_2012_farletuzumab.R.STUDY_FARLETUZUMAB_PHASE2 indicator selects
between them. Renamed from generic PHASE2 to paper-specific
STUDY_FARLETUZUMAB_PHASE2 on 2026-06-19 per the
canonical-register standardization audit (operator decision: study-phase
indicators must be paper-specific because two papers can pick opposite
reference categories - Farrell 2012 picks Phase II as 1-level while
Valenzuela 2025 picks Phase I as 1-level - and a generic “PHASE2”
canonical conflates the two encodings).PHASE1 – prior canonical name (pre-2026-06-19
paper-specific-disambiguation standardization). Generic single-token
name collided with the parallel PHASE2 canonical from
Farrell 2012 that picks the opposite reference category.NCT03772587 Vivacity-MG -> 0).Valenzuela_2025_nipocalimab.R.STUDY_FARLETUZUMAB_PHASE2 – the reference category is
inverted (Valenzuela 2025 picks Phase 1 as the 1-level). Renamed from
generic PHASE1 to paper-specific
STUDY_NIPOCALIMAB_PHASE1 on 2026-06-19 per the
canonical-register standardization audit (operator decision: study-phase
indicators must be paper-specific because two papers can pick opposite
reference categories, and a generic “PHASE1” canonical conflates the
encodings).vanIersel_2018_posaconazole.R (switches the log-additive
residual SD on Cc between expSd_p1 = 0.42 (phase 1) and
expSd_p3 = 0.322 (phase 3), per van Iersel 2018 Table 2
final-model ‘SD (phase 1 studies)’ and ‘SD (phase 3 study)’).STUDY_<DRUG>_PHASE<N> family alongside
STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025) and
STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012); distinct from
those entries because the reference category (phase 1) and 1-level
(phase 3) are paper-specific to van Iersel 2018. Subject-level
(time-fixed); set once from the trial identifier on each subject record.
Ratified canonically alongside the van Iersel 2018 posaconazole
extraction.C2201 -> 1, else -> 0).Bienczak_2025_ligelizumab.R (Table S6: study C2201 on CL/F
= 0.176, log-additive; cl *= exp(0.176) for C2201
subjects).STUDY – the paper-text label for the binary study
identifier; derived per subject from the trial-of-origin column in the
source dataset.Harrold_2020_radiation_neutropenia.R (selects between the
two OS parameter sets in Table III: lambda_ANC -2.15 vs -0.229,
lambda_BC -0.347 vs 0.300, k_e0 0.668 vs 0.156 1/day).STUDY_<id> canonical family.INDR – used in Zhou_2021_belimumab.R (Zhou
2021 Table 2 footnote: study indicator).Zhou_2021_belimumab.R
(multiplicative factors 1.63 on CL and 1.26 on V1 when STUDY_LBSL =
1).STUDY1
/ STUDY_FARLETUZUMAB_PHASE2 / ELISA /
STUDY_NIPOCALIMAB_PHASE1 (per-study switches) but specific
to the belimumab program. Subject-level (time-fixed); set from the trial
identifier on each subject record.STDY_VORI – prior canonical name (pre-2026-06-19 typo
correction); a missing-vowel abbreviation of
STUDY_VORI.STUDY / STDY identifier column.Friberg_2012_voriconazole.R (drives several effects: -0.382
Study-1 pediatric modifier on Km and Vmax,1; non-adult uplift on Q
(+0.637); adolescent ka modifier; non-adult CL IIV scaling (+1.70); F1
IIV magnitude switching between adult and non-adult; per-study
residual-error switching across the four levels Study 1, Study 2,
Studies 3+4, Study 5).STUDY1
/ STUDY5 / STUDY_PKU015 precedent because the
Friberg 2012 analysis uses five distinct studies and four of them carry
distinct typical-value or residual-error coefficients (Studies 3 and 4
share one residual-error magnitude). Encoding as a single integer column
avoids registering five paired binary indicators; the model file derives
(STUDY_VORI == 1) style indicators inline. Renamed from
STDY_VORI to STUDY_VORI on 2026-06-19 per the
canonical-register standardization audit (operator decision: typo
correction - STDY was a missing-vowel abbreviation of
STUDY).Friberg_2012_voriconazole.R (combines with
STDY_VORI == 5 to switch the adult residual-error magnitude
between IV-only expSdStdy5Iv = 0.0912 and oral
sqrt(0.0912^2 + 0.132^2) = 0.160 per Table 3 footnote on
the residual-error structure W).SAMPLE_INTENSIVE (a generic per-observation switch between
estimated residual-error magnitudes); the contrast here is dosing route
(oral vs IV) within the same subject’s crossover protocol rather than
sampling design. Specific scope because the route-vs-residual-error
switch is paper-specific to the Friberg 2012 voriconazole analysis.Macpherson_2015_rosuvastatin.R (Table 2 final model: 39.4%
CV intensive vs 59.5% CV sparse residual error; switched per observation
as
propSd <- propSdIntensive * SAMPLE_INTENSIVE + propSdSparse * (1 - SAMPLE_INTENSIVE)).STUDY1
/ STUDY5 / PHASE1 / ELISA
(per-record switches that select between estimated residual-error
magnitudes) but the contrast is sampling design rather than study cohort
or bioanalytical assay. The indicator is generally applicable: any
pediatric / dense-vs-sparse pooled popPK design that estimates two
residual errors can carry it. Within-subject variation is permitted (a
single subject can have both intensive and sparse observations, as in
the Macpherson 2015 CHARON PK-pilot cohort where 12 subjects had a Day-0
intensive profile followed by 2 years of sparse troughs).Valenzuela_2025_nipocalimab.R.NCT03772587 -> 1, else -> 0).Valenzuela_2025_nipocalimab.R.FRIgG0_M281_004 = 0.777 vs. 1 in other studies). Distinct
from the disease-state indicator implied by gMG – it is
specifically the Vivacity-MG study flag because the IgG baseline factor
was only estimated for that study.STUDY_ABA2_HLA88).Cohort = ABA2 7/8 -> 1, else
-> 0). Takahashi 2023 Supplemental Table 4 names the corresponding
theta thetaCohort_CL / thetaCohort_Vc.Takahashi_2023_abatacept.R (multiplicative
Ratio factors on CL = 0.70 and on Vc = 0.99 vs the RA/JIA
reference; values from Takahashi 2023 Supplemental Table 4).STUDY_ABA2_HLA88 to
reproduce the three-level cohort categorical (RA/JIA, ABA2 7/8, ABA2
8/8) the paper reports as the only retained categorical PK covariate. At
most one of STUDY_ABA2_HLA78 and
STUDY_ABA2_HLA88 is 1 per subject; both 0 reproduces the
RA/JIA reference. Scope: specific because the contrast is tied to the
ABA2-vs-RA/JIA pooling design.STUDY_ABA2_HLA78).Cohort = ABA2 8/8 -> 1, else
-> 0). Takahashi 2023 Supplemental Table 4 names the corresponding
theta thetaCohort_CL / thetaCohort_Vc.Takahashi_2023_abatacept.R (multiplicative
Ratio factors on CL = 0.91 and on Vc = 1.32 vs the RA/JIA
reference; values from Takahashi 2023 Supplemental Table 4).STUDY_ABA2_HLA78. At
most one of the two indicators is 1 per subject; both 0 reproduces the
RA/JIA reference. Scope: specific.gamma_eGFP for STUDY_d2eGFP =
0 vs gamma_d2eGFP for STUDY_d2eGFP = 1) and its IIV
variance differ.Experiment 1 /
eGFP -> 0, Experiment 2 / d2eGFP -> 1; see
code.zip project/models/experiments_transfection_ribo.m in
the Frohlich 2018 Zenodo deposit doi:10.5281/zenodo.1228899).Frohlich_2018_mRNA_translation.R.STUDY_FARLETUZUMAB_PHASE2,
STUDY_NIPOCALIMAB_PHASE1), but the contrast is between two
reporter mRNA constructs measured under the same physical / imaging
protocol rather than between trial phases.NCT01119833 -> 1, else -> 0); the source
STUD column in the NONMEM dataset described in Tammara 2017
Table 1 footnote b.Tammara_2017_rivipansel.R (Table 1: additive effect 0.234
on CL via 1 + 0.234 * STUDY_RIV201; selects the
cohort-specific additive and proportional residual SDs in
model()).STUDY_RIV201 = 1 for every subject. Subject-level
/ time-fixed; set once from the trial identifier on each subject
record.STUDY – the Wagh 2021 NONMEM dataset’s study identifier
column (1 / 2 in the source; mapped to 0 / 1 in this binary
indicator).Wagh_2021_spectinamide_1810_mouse.R (Table 3:
study-specific coefficient for K_kill_max = 1.15, RSE 4.0%; applied
multiplicatively as
killmax = killmax_base * (1 + STUDY_WAGH_2 * (1.15 - 1))).Simulated_Lid_B04_ddmore.csv). Values
1-4 (or higher) flag distinct study-protocol dose / regimen tiers; the
NA_NA_lidocaine.R model binarises as
DLVL > 2 to switch the typical-value baselines for both
the GX elimination rate constant K30 and the lidocaine apparent central
volume V1 between a “low” (DLVL <= 2) and a “high” (DLVL > 2)
regimen.DLVL <= 2 is the reference (THETA(4) for K30 and
THETA(14) for V1 in the source .ctl);
DLVL > 2 selects the higher-exposure regimen (THETA(5)
and THETA(15) respectively).DLVL – the column
header used in the DDMORE bundle’s .ctl $INPUT
and the Simulated_Lid_B04_ddmore.csv data file.NA_NA_lidocaine.R
(DDMODEL00000281; binary derivation
DLVL_HIGH = as.integer(DLVL > 2) on K30 base and V1
base).> 2 reproduces the
source .ctl line IF(DLVL.GT.2)P1=0. If a
future model needs a different dose-level binarisation or a continuous
treatment, register a distinct canonical name rather than overloading
DLVL.NA_NA_lidocaine.R model the value
S1A2 == 3 selects the “CYP1A2 inducer present” sub-cohort
(lidocaine N-deethylation to MEGX is CYP1A2-mediated, so the modifier
acts on the GX elimination rate constant K30 in the source’s
parameterisation). Other integer codes (0, 1, 2) are pooled into the
reference.S1A2 != 3 (i.e.,
values 0, 1, 2) – pooled into the reference.S1A2 – the column
header used in the DDMORE bundle’s .ctl $INPUT
and the simulated dataset, with sibling columns D1A2 and
H1A2 carried in the data file but dropped via
=DROP in the source .ctl.NA_NA_lidocaine.R
(DDMODEL00000281; binary derivation
S1A2_IND = as.integer(S1A2 == 3) on the GX elimination rate
constant K30).S1A2 are paper-specific to the lidocaine BAST.dat study and
the linked publication is not on disk for this extraction. The exact
biological meaning of each integer level (0/1/2/3) is not fully
reconstructable from the bundle alone – the natural interpretation,
given the column name encodes “CYP1A2” and the model attaches a sizeable
positive K30 modifier of +0.853 to the level-3 cohort, is a
CYP1A2-induction or smoking / inducer co-medication indicator. Sibling
columns D1A2 (donor / inhibitor?) and H1A2
(host / inhibitor?) are dropped in the source .ctl so only
the level-3 indicator is structurally identifiable from the surviving
model code. If a future model needs a richer encoding of CYP1A2
modulation, register a separate canonical (e.g.,
CYP1A2_IND) rather than overloading S1A2.Jansson_2008_eflornithine_rat.R.FED_HIGHFAT because Zvada 2010 isolates four
operationally-distinct meal compositions (A/B/C/D) each with its own
bioavailability effect estimate and because the paper’s Discussion
explicitly attributes part of meal A’s effect to eggs rather than total
fat.Zvada_2010_rifapentine.R (additive contribution to TVF = 1
* (1 + sum of per-meal fractional changes); RXF_A = +0.857, RSE 20.1%;
Zvada 2010 Table 3).MEAL_B, MEAL_C, MEAL_D - at most
one of the four indicators can be 1 on a given dosing record; the fasted
reference is all four = 0. Future rifapentine / antimycobacterial
food-effect studies that reuse the same meal compositions (uncommon -
meal compositions are typically protocol-specific) can extend this
entry’s example list. Distinct from MEAL_FLAG /
SNACK_FLAG (time-varying intra-day flags) because
MEAL_A is a per-dose-record indicator that selects between
five mutually exclusive meal states.Zvada_2010_rifapentine.R (additive contribution to TVF;
RXF_B = +0.327, RSE 40.4%; Zvada 2010 Table 3).MEAL_A, MEAL_C, MEAL_D. See
MEAL_A notes for the per-dose-record indicator
pattern.Zvada_2010_rifapentine.R (additive contribution to TVF;
RXF_C = +0.457, RSE 29.3%; Zvada 2010 Table 3).MEAL_A, MEAL_B, MEAL_D. Meal C
and meal A have similar total fat content but the Zvada 2010 Discussion
attributes the substantially smaller effect of meal C (+45.7%) versus
meal A (+85.7%) to the absence of eggs in C - supporting the authors’
egg-specific bioavailability hypothesis.Zvada_2010_rifapentine.R (additive contribution to TVF;
RXF_D = +0.489, RSE 30.7%; Zvada 2010 Table 3).MEAL_A, MEAL_B, MEAL_C. Zvada
2010 Discussion notes that monosodium glutamate (MSG) in the chicken
noodle soup may contribute to the effect by accelerating gastric
emptying (per Boutry et al., Am J Physiol Endocrinol Metab 2011, on MSG
and gastric emptying).ATORVASTATIN atorvastatin (spelled out to
avoid an ATV collision with atazanavir; see
CONMED_ATAZANAVIR), flv fluvastatin,
lov lovastatin, prv pravastatin,
rsv rosuvastatin, smv simvastatin,
ezt ezetimibe, inh isoniazid). 0 = the named
drug is not part of the regimen for this study arm / subject; positive
value = total daily dose. Captures the dose-response amplitude in MBMA
or co-administered-drug PK/PD models where each drug arm contributes its
own dose-effect curve.CONMED_INH_DOSE (paper-specific unit, documented
in covariateData[[CONMED_INH_DOSE]]$units).DOSE_<drug>
legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename,
Chen 2017 TB mouse pre-rename); aliases of the canonical
CONMED_<drug>_DOSE form per the 2026-05-28 naming
audit rename. CONMED_ATV_DOSE (legacy abbreviated form used
briefly between the 2026-05-28 naming audit and the atazanavir-collision
rename to CONMED_ATORVASTATIN_DOSE).Vargo_2014_statins_ezetimibe_mbma.R (each
CONMED_<statin>_DOSE drives the corresponding
statin’s Hill / Emax dose-response curve; per-arm dose level in mg/day,
default 0 outside the named statin arm; ezetimibe and statin arms
combine via the sub-additive gamma_int term),
Chen_2017_TB_MTP_GPDI_mouse.R (CONMED_INH_DOSE
drives the isoniazid CL adjustment:
cl_inh = cl_inh_lowdose * (1 - slope_inh * (CONMED_INH_DOSE - 12.5))).CONMED_<drug>_DOSE
shape replaces the earlier DOSE_<drug> /
DOSE_<drug>_<unit> names (which conflated a
covariate with a dose-amount column). The atorvastatin slot is spelled
out as CONMED_ATORVASTATIN_DOSE rather than
CONMED_ATV_DOSE to keep the ATV abbreviation
reserved for the binary atazanavir coadministration indicator
(CONMED_ATAZANAVIR, source alias ATV) used in
HIV-PK extractions like ArabAlameddine 2012 raltegravir. New
co-medication dose-effect MBMA models should reuse this family and add
the appropriate <drug> INN abbreviation; spell out
drug names whenever the natural 3-4 letter abbreviation would collide
with another CONMED_ entry. The units field is
per-paper.Trt (Lee 2011 paper notation:
Trt = (0, 1, ...), where the source paper notes the
indicator could in principle take more than two levels but only the
binary placebo-vs-active contrast is used in the actual analyses) – used
in Lee_2011_parkinson_progression.R.Lee_2011_parkinson_progression.R (additive shifts on the
placebo disease-progression slope and on the placebo short-term
symptomatic effect:
slope = beta0 + beta1 * ON_TREATMENT + etaslope and
symeff = gamma0 + gamma1 * ON_TREATMENT + etasymeff;
default ini() parameter values are from the TEMPO study
rasagiline arm, so ON_TREATMENT = 1 predicts the
rasagiline-active disease trajectory and ON_TREATMENT = 0
predicts the placebo trajectory).TRT (which is the
Novakovic 2017 cladribine-cohort categorical indicator with three levels
and a paper-specific 1-vs-2 cohort labelling) and from
DRUG_ORMU / DRUG_PRED / DRUG_OBI
(which are head-to-head drug-product comparator indicators carrying the
contrast between two named active drugs in the same trial). Use
ON_TREATMENT only when the structural model has no
drug-specific PK / exposure and the active arm enters as a pooled on/off
switch. When extracting a head-to-head comparator paper (drug A vs drug
B) register a sibling DRUG_<X> canonical instead.
When the source paper differentiates multiple active dose levels with
distinct parameters, register a categorical indicator with
paper-specific level labelling (e.g., TRT for the
cladribine cohorts) rather than overloading ON_TREATMENT.
Per-model covariateData[[ON_TREATMENT]]$notes should
document which drug the active arm corresponds to in that file’s default
ini() values, the trial design (parallel-group vs
cross-over), and any time-varying interpretation. Ratified canonically
alongside the Lee 2011 Parkinson’s disease-progression extraction (the
existing TRT canonical’s Notes already directed future
generic on-treatment indicators to register this canonical rather than
reusing TRT).Hansson_2013_sunitinib_dbp.R (switches the typical baseline
diastolic blood pressure dbp0 from 71.8 mmHg in the active
arm to 77.6 mmHg in the placebo run-in arm via
exp(ldbp0 + e_placebo_dbp0 * PLACEBO); the sunitinib AUC
drug effect on kin is modelled separately and
exposure-driven).ON_TREATMENT (the generic pooled active-vs-placebo on/off
switch used when the active arm REPLACES exposure in the structural
model, reference 0 = placebo): PLACEBO has the opposite
polarity (1 = placebo) and is used alongside, not instead of, an
exposure-driven drug effect, marking the placebo subpopulation for an
arm-specific structural parameter. If a model needs the pooled on/off
drug switch, use ON_TREATMENT (= 1 - PLACEBO)
instead.SEMESTER1 – used in Gonzalez-Sales 2015 testosterone
circadian model with the same encoding (winter or spring = 1, summer or
fall = 0).GonzalezSales_2015_testosterone.R (+8.09% multiplicative
effect on Base; Table III).SEMESTER1 / SEMESTER2 covariate
(SEMESTER1 = 1 if winter or spring;
SEMESTER2 = 1 if fall or winter), keeping only
SEMESTER1 in the final model. The canonical column omits
the 1 suffix because the second indicator was not retained;
the source-paper naming is recorded as a source alias. Calendar
boundaries are reported verbatim in the paper Methods (winter: Dec 21 -
Mar 19; spring: Mar 20 - Jun 19; summer: Jun 20 - Sep 21; fall: Sep 22 -
Dec 20). Scope: specific until a second endogenous-rhythm model adopts
the same encoding; at that point promote to general.
Distinct from SEASON2 (RSV-second-exposure indicator) which
has different semantics.CEN – Wahlby 2004 source-column convention; used in
Wahlby_2004_pefloxacin.R.Wahlby_2004_pefloxacin.R (multiplicative effect on CL via
exp[0.19 * CEN]; CEN = 1 increases CL by ~21 percent relative to CEN =
0).FORM – used in Garrett_2014_axitinib.R
(Garrett 2014 Methods ‘Statistical analysis’ and Results ‘Full model and
final model’ describe the formulation as a categorical {Form IV, Form
XLI} with Form XLI = 1 as the test polymorph and Form IV = 0 as the
typical-value reference).Garrett_2014_axitinib.R (linear-proportional effect on
bioavailability per Garrett 2014 Table 3 final estimates:
f = exp(lfdepot) * (1 + e_fast_f * (1 - FED)) * (1 + e_xli_f * FORM_AXI_XLI)
with e_xli_f = -0.150 (15% reduction in F for Form XLI
relative to Form IV in the fed state); Form XLI has no retained effect
on ka, CL, Vc, Q, or Vp).FORM_TABLET, FORM_CAPSULE,
FORM_SOLUTION, and the rest of the dosage-form
FORM_* family. Mirrors the existing
FORM_SAR_DP2 (sarilumab drug-product 2),
FORM_ISA_P2F2 (isatuximab P2F2 drug material),
FORM_LINAG_TAB1 (linagliptin tablet 1),
FORM_VISMO_PHASEI (vismodegib Phase I capsule),
FORM_ITR_SUBA (SUBA-itraconazole vs Sporanox capsule),
FORM_TAC_IR (tacrolimus IR vs prolonged release),
FORM_DOX_DORYX_MPC (Doryx MPC tablet), and
FORM_THEO_APNECUT (theophylline Apnecut) entries under the
FORM_* family of drug-specific drug-product-version /
polymorph indicators. Set to 0 to simulate the Form IV early-phase
reference; set to 1 to simulate the Form XLI marketed product (the
routine clinical-use default). Ratified canonically on 2026-06-16
alongside the Garrett 2014 axitinib extraction.STUD – used in Abrantes_2017_moroctocog.R
(Abrantes 2017 Table 2 footnote b: study indicator on CL).Abrantes_2017_moroctocog.R (multiplicative effect on CL:
(1 - 0.347 * STUDY_B1831090) so B1831090 subjects have
~34.7% lower CL than the other 12 studies; Abrantes 2017 Table 2 also
reports a 95% CI of -40.7% to -24.2%).1, 2, …, N identify the occasion
to which each observation belongs (typically a dosing visit, study
period, or sampling occasion). Time-varying within subject; constant
within an occasion.OCC is
decomposed inside model() into mutually-exclusive binary
indicators, e.g., oc1 <- (OCC == 1),
oc2 <- (OCC == 2), …, that are then multiplied against
per-occasion eta* slots.OCC – used in Jonsson_2011_ethambutol.R
(DDMODEL00000220 NMTRAN $INPUT column; values 1..4).Jonsson_2011_ethambutol.R (4-occasion IOV on log-CL;
cl <- exp(lcl + etalcl + oc1 * etalcl_oc1 + oc2 * etalcl_oc2 + oc3 * etalcl_oc3 + oc4 * etalcl_oc4) * (WT/50)^0.75,
where each etalcl_oc<k> is a separate
~ fix(0.127) after the first to encode NONMEM
$OMEGA BLOCK(1) SAME),
Aregbe_2012_alvespimycin.R (5-occasion BOV on Q2 and V1),
Oosten_2016_fentanyl.R (10-occasion IOV on transdermal Ka;
only OCC >= 1 records carry IOV, sc / non-transdermal
records pass OCC = 0 so all indicators zero out),
Jonsson_2011_ethambutol_ddmore.R (4-occasion IOV on log-CL;
cl <- exp(lcl + etalcl + oc1 * etalcl_oc1 + oc2 * etalcl_oc2 + oc3 * etalcl_oc3 + oc4 * etalcl_oc4) * (WT/50)^0.75,
where each etalcl_oc<k> is a separate
~ fix(0.127) after the first to encode NONMEM
$OMEGA BLOCK(1) SAME).OCC is the recommended
canonical for new IOV-using models – the binary ooc1..oocN
indicators below remain canonical for legacy / pre-existing models that
ship the data already-decomposed. Ratified canonically on
2026-05-06.Xie_2019_agomelatine.R.OCC canonical above and decompose into binary indicators
inside model().Sassen_2017_crisantaspase.R (multiplicative shift on
CL:
cl <- exp(lcl + etalcl) * (WT/70)^0.75 * (1 + e_month1_cl * MONTH1)
with e_month1_cl = 0.14, encoding the 14% higher CL in the
first month of treatment relative to subsequent months).OCC and ooc<n> (which decompose
multi-occasion sampling for IOV), from DAY14 (which uses a
14-day cutoff for malnutrition-recovery contrasts), and from
CYCLE (which is a dose-number counter, not a single binary
landmark). Data assemblers can derive
MONTH1 = as.integer(time_post_treatment_start_days < 30)
for a regularly-sampled multi-month study. Ratified canonically on
2026-05-20 alongside the Sassen 2017 extraction.Archary_2019_abacavir.R (multiplicative additive shift
on CL/F:
cl <- exp(lcl + etalcl) * (WT/7)^0.75 * (1 + e_day14_cl * DAY14)
with e_day14_cl = 0.760, encoding the day-1 typical CL/F =
3.33 -> day-14 CL/F = 5.86 L/h per 7 kg step).Archary_2019_lamivudine.R (multiplicative additive
shift on ka:
ka <- exp(lka) * (1 + e_day14_ka * DAY14) * exp(etalka)
with e_day14_ka = 0.133, encoding the day-1 typical ka =
0.30 -> day-14 ka = 0.34 /h step).DAY28). Distinct from OCC and
ooc<n> (which decompose multi-occasion sampling for
IOV), from TRT_PHASE (which gates active-vs-baseline
study-phase contributions), and from EARLY_ART (which is a
between-subject randomization-arm indicator, not a within-subject
landmark). Data assemblers can derive
DAY14 = as.integer(time_post_treatment_start_days >= 14)
for a regularly-sampled multi-day study. Ratified canonically on
2026-05-08.CYCLE^Fm (Fm typically negative) to
capture cycle-over-cycle decline in a derived quantity such as
ADC-to-payload conversion fraction (Li 2017 brentuximab vedotin), or
with a piecewise indicator CYCLE == 1 vs CYCLE >= 2 to
capture a step change in PK between the first and subsequent
administrations (Hong 2025 datopotamab deruxtecan; Huynh 2026
VRC07-523LS).CYCLE – used in
Li_2017_brentuximab.R,
Hong_2025_datopotamab.R, Lu_2022_patritumab.R,
and Huynh_2026_VRC07523LS.R with the same canonical
name.Li_2017_brentuximab.R (exponent on the fraction of ADC
that converts to MMAE by proteolytic degradation, Fm = -0.261, to
reflect tumor-burden reduction across successive treatment cycles).Hong_2025_datopotamab.R (cycle-1 vs cycle-2+ piecewise
scaling Factor1 = 0.696 on the DAR equation that drives DXd formation
rate from total Dato-DXd elimination).Lu_2022_patritumab.R (cycle-1 vs cycle-2+ piecewise
scaling Factor1 = theta = 0.648 on the payload-to-intact-drug ratio PIR
that scales DXd release rate from intact ADC).CYCLE^Fm is undefined at 0; the piecewise form requires
CYCLE to be a positive integer at every observation row).
Distinct from ooc<n> binary-occasion indicators:
CYCLE is an integer count, not a mutually-exclusive set of
indicator columns. Data-assembly helper: set
CYCLE = floor((TIME - TIME_FIRST_DOSE) / cycle_length_days) + 1
for a fixed-interval dosing regimen.TIME column carries the integration-time variable (with
TIME = 0 the integration origin, typically well before any
subject’s disease onset); T_ENTRY records, per subject, the
integration-time value at which the subject’s first observation falls.
Used in models whose pre-study and post-study time semantics differ –
e.g., a disease-progression activation function defined on global
disease time plus a placebo-effect term defined on
time-since-study-entry need access to both reference clocks within the
same model() block.units$time field declares)T_ENTRY is a
per-subject time-offset covariate that anchors the
time-since-study-entry clock used by post-entry-only model terms
(placebo / learning effects, study-design transient drops).Delor_2013_alzheimer.R
(placebo-term clock:
t_pl_raw <- time - T_ENTRY; post_entry <- t_pl_raw > 0; placebo_term <- pl_indiv * (1 - exp(-kpl_indiv * t_pl_raw * post_entry)) * post_entry).T_ENTRY
per subject in the simulation event table; a reasonable construction is
T_ENTRY = DOT_individual + a few years of established disease
so the patient is observed during disease progression (see the Delor
2013 vignette for the construction used to reproduce Figures 2-4).
Ratified canonically on 2026-05-16 alongside the Delor 2013 extraction.
## Mixture / latent-class indicators$MIXTURE block (Tsuji 2017 Equation 8 and
Methods, “Mixture model” subsection).MIXTURE – NONMEM
$MIXTURE block class index in the Tsuji 2017 estimation run
(component 1 = PDI, component 2 = PDS); the binary indicator is
MIX_PDI = as.integer(MIXTURE == 1).Tsuji_2017_linezolid.R
(gates the drug-effect term that enters the platelet ODE chain:
MIX_PDI = 1 activates SLOPE * Cc on the
formation rate, MIX_PDI = 0 activates
SMAX * Cc / (SC50 + Cc) on the circulating-compartment
elimination rate).MIX_PDI = 1 is the estimated population mixture fraction
FPOP_inhibit = 0.969 (Tsuji 2017 Table 2; 95% CI
0.867-1.00, 78/80 patients in the source dataset). For typical-value
simulation set MIX_PDI = 1 (dominant clinical phenotype,
slower 2-week nadir; Figure 5 left panel) or MIX_PDI = 0
(rare immune-mediated-like phenotype, faster 2-day nadir; Figure 5 right
panel). For population simulation, draw
MIX_PDI ~ Bernoulli(0.969) per subject. Scope: specific
because the binary semantics are tied to Tsuji 2017’s two-mechanism
platelet model; mixture indicators from future papers that share the
same biological dichotomy may extend this entry, but mixture indicators
from unrelated dichotomies should register a new canonical name (e.g.,
MIX_FAST_ELIM for a fast/slow eliminator mixture).$MIXTURE NONMEM block describing a bimodal
apparent-clearance distribution. 1 = subject classified to the high-CL /
fast-eliminator subpopulation; 0 = subject classified to the low-CL /
slow-eliminator subpopulation. Not a measured clinical covariate – the
mixture assignment is a posterior latent-class index from the NONMEM
$MIXTURE block. One assignment per subject,
time-fixed.MIXTURE – NONMEM $MIXTURE block class
index in the Morris 2011 estimation run (component 1 = high-CL with P =
0.251, component 2 = low-CL with P = 0.749); the binary indicator is
MIX_FAST_ELIM = as.integer(MIXTURE == 1).Morris_2011_telapristone.R (selects the telapristone CL/F
typical value between the high-CL anchor
exp(lcl_pop1) = 11.6 L/h and the low-CL anchor
exp(lcl_pop2) = 3.34 L/h; the shared IIV
etalcl ~ 0.200 is applied multiplicatively to whichever
typical the mixture-class selects),
Wilkins_2011_isoniazid.R (gates the typical apparent CL/F
via the power-form covariate effect e_mix_fast_elim_cl on
CL/F:
cl = exp(lcl + e_mix_fast_elim_cl * MIX_FAST_ELIM + etalcl) * (WT/70)^0.75 * (1 + e_hiv_pos_cl * HIV_POS)
with lcl = log(9.70) (slow-eliminator reference) and
e_mix_fast_elim_cl = log(21.6/9.70) approx 0.8005).MIX_PDI
register entry’s Notes block (“mixture indicators from unrelated
dichotomies should register a new canonical name (e.g.,
MIX_FAST_ELIM for a fast/slow eliminator mixture)”).
Population probability of MIX_FAST_ELIM = 1 is recorded
per-model in covariateData[[MIX_FAST_ELIM]]$notes; in
Morris 2011 it is 0.251 (Table II, “Probability” row, RSE 61.0%). The
Morris 2011 Discussion attributes the bimodal CL distribution to
polymorphic CYP3A5 (functional CYP3A5 present in 10-40% of Caucasians,
~50% of African Americans, ~33% of Asians) but does not test the
genotype hypothesis directly; the mixture indicator is therefore a
latent-class label, not a CYP3A5-genotype indicator. For typical-value
simulation set MIX_FAST_ELIM = 1 (fast eliminator; lower
steady-state exposure, ~12 h elimination half-life per Morris 2011
Results) or MIX_FAST_ELIM = 0 (slow eliminator; higher
steady-state exposure, ~35 h half-life). For population simulation, draw
MIX_FAST_ELIM ~ Bernoulli(0.251) per subject. Scope:
general because the fast/slow-eliminator dichotomy is a recurring popPK
concept across drugs metabolized by polymorphic enzymes
(CYP2C92/3, CYP2C192, CYP3A53, NAT2 slow/fast
acetylators); future popPK papers fitting a $MIXTURE on CL
should re-use this canonical rather than introducing per-drug-specific
MIX_<drug>_FAST names. The reference category (= 0 =
slow eliminator) is chosen so the binary numerically matches the paper’s
mixture-indicator orientation. Ratified canonically on 2026-06-09
alongside the Morris 2011 telapristone extraction.$MIXTURE block describing the observed bimodal
distribution of baseline tumor sizes (Schindler 2017 Methods, “Maximum
transaxial diameter, actual volumes, and ellipsoidal volume models”
subsection: “Semiparametric distributions and mixture models were
investigated to describe the observed bimodal distribution in baseline
MTD, Vactual, and Vellipsoid”). One assignment per subject is shared
across the three size models (MTD, Vactual, Vellipsoid) and across the
subject’s up-to-two lesions; density (D0) is not
mixture-indexed.MIXTURE – NONMEM
$MIXTURE block class index in the Schindler 2017 estimation
run (component 1 = larger-baseline subpop with P = 0.348, component 2 =
smaller-baseline subpop); the binary indicator is
MIX_LARGE_BASE = as.integer(MIXTURE == 1).Schindler_2017_imatinib.R (gates the per-lesion typical
baselines S0_<metric>_l<k>_typ between
subpopulation-1 and subpopulation-2 anchors, and selects between
subpopulation-specific IIV etas etalS0_<metric>_pop1
and etalS0_<metric>_pop2 whose log-scale variances
differ across mixture classes).MIX_LARGE_BASE = 1 to reproduce the
larger-baseline phenotype (dominates Figure 2 right panel
typical-individual trajectory) or MIX_LARGE_BASE = 0 to
reproduce the smaller-baseline phenotype. For population simulation,
draw MIX_LARGE_BASE ~ Bernoulli(0.348) per subject. Scope:
specific because the binary semantics are tied to Schindler 2017’s
two-class baseline-tumor-size mixture; future tumor-burden mixture
models that share the same large-vs-small baseline dichotomy may extend
this entry, while mixture indicators from unrelated dichotomies (e.g.,
fast / slow tumor-growth phenotypes) should register a new canonical
name (MIX_FAST_GROW, etc.). Ratified canonically on
2026-05-18 alongside the Schindler 2017 imatinib extraction.$MIX block (paper Methods, “Basic pharmacokinetic model”
subsection: “Subpopulations were estimated using the $MIX function in
the control stream”).$MIX class assignment
– NONMEM $MIX block class index in the Kappelhoff 2005
estimation run (component 1 = P1 small-RUV with P = 0.648, component 2 =
P2 large-RUV with P = 0.352); the binary indicator is
MIX_LARGE_RUV = as.integer($MIX == 2).Kappelhoff_2005_ritonavir.R (gates the active additive
residual SD inside model():
add_sd_eff <- addSd_p1 * (1 - MIX_LARGE_RUV) + addSd_p2 * MIX_LARGE_RUV;
the proportional component is shared).MIX_LARGE_RUV = 0 (dominant subpopulation).
For population simulation, draw
MIX_LARGE_RUV ~ Bernoulli(0.352) per subject. Scope:
specific because the binary semantics are tied to Kappelhoff 2005’s
two-class additive-RUV mixture; future popPK models that share the same
large-vs-small additive-RUV dichotomy may extend this entry, while
mixture indicators from unrelated dichotomies should register a new
canonical name (e.g., MIX_LARGE_PROPRUV for a mixture on
the proportional component instead). Ratified canonically alongside the
Kappelhoff 2005 ritonavir extraction.$MIXTURE block (Bonate 2004 Methods “Base model
development”, equation (1.4), and Results paragraph 1).MIXTURE – NONMEM
$MIXTURE block class index in the Bonate 2004 estimation
run (component 1 = lagged-absorption Group 1 with P = 0.970, component 2
= no-lag Group 2 with P = 0.030); the binary indicator is
MIX_LAGGED_ABS = as.integer(MIXTURE == 1).Bonate_2004_apomine.R
(gates ka between two subpopulation-specific typical values and gates
the lag time on the depot compartment between Group 1 =
exp(ltlag) and Group 2 = 0; the shared etalka
log-normal IIV scales whichever group’s ka is active for the subject),
Frymoyer_2013_mycophenolic_acid.R (gates only the lag time
on the depot compartment between Group 2 =
exp(ltlag) = 1.96 h and Group 1 = 0; ka, F and disposition
parameters are shared between the two classes per Frymoyer 2013
“Allowing other absorption parameters (ka and F) to vary for the two
subpopulations by using an extended mixture model did not improve the
overall model or predictions”).MIX_LAGGED_ABS = 1 is the estimated mixture fraction
P(Group 1) = 1 / (1 + exp(P1)) = 0.970 (Bonate 2004 Table 3
with P1 = -3.47). For typical-value simulation set
MIX_LAGGED_ABS = 1 to reproduce the dominant 97 % phenotype
(lagged-absorption with ka = 1.77 /h and lag = 0.821 h); set
MIX_LAGGED_ABS = 0 to reproduce the rare 3 % no-lag
phenotype (ka = 0.361 /h, no lag). For population simulation, draw
MIX_LAGGED_ABS ~ Bernoulli(0.970) per subject. Scope:
specific because the binary semantics are tied to Bonate 2004’s
two-class absorption mixture; future popPK papers that share the same
lagged-vs-no-lag absorption-mixture dichotomy may extend this entry,
while mixture indicators from unrelated dichotomies (e.g., fast vs slow
elimination subpopulations) should register a new canonical name
(MIX_FAST_ELIM, etc.). Ratified canonically on 2026-06-04
alongside the Bonate 2004 apomine extraction.Vp_i = TVV3 * theta_Study2 with
theta_Study2 = 23.5); 0 = subject classified to the
typical-Vp subpopulation. Not a measured clinical covariate – the
mixture assignment was identified by Bonate 2004 from a bimodal
empirical-Bayesian Vp histogram and back-fitted as a dichotomous
study-membership multiplier (Bonate 2004 Results paragraph 1 and
Discussion).STUDY2 – the paper
modelled the high-Vp subgroup as a dichotomous indicator on Study 2
membership (the 4 healthy-male multiple-dose subjects in Study 2); the
binary indicator is MIX_HIGH_VP = as.integer(STUDY == 2)
for the source cohort.Bonate_2004_apomine.R
(multiplicative log-additive effect on Vp:
vp = exp(lvp + e_high_vp_vp * MIX_HIGH_VP + etalvp) with
e_high_vp_vp = log(23.5) = 3.157).MIX_HIGH_VP = 1 is 4 / 38 = 0.105 in the Bonate 2004
model-development set. For typical-value forward simulations set
MIX_HIGH_VP = 0 (the recommended default); the paper notes
that simulations with and without the multiplier showed minimal
differences in concentration-time profiles. Set
MIX_HIGH_VP = 1 only to reproduce the Bonate 2004 Study 2
healthy-male multiple-dose subgroup specifically. Scope: specific
because the binary semantics are tied to Bonate 2004’s anomalous Study 2
high-Vp subgroup; future popPK papers that retain a similar dichotomous
high-Vp subgroup multiplier may extend this entry, while
peripheral-volume mixture indicators from unrelated dichotomies should
register a new canonical name. Ratified canonically on 2026-06-04
alongside the Bonate 2004 apomine extraction.$MIXTURE NONMEM block describing a bimodal
apparent-clearance distribution. 1 = subject classified to the high-CL /
fast-eliminator subpopulation; 0 = subject classified to the low-CL /
slow-eliminator subpopulation. Not a measured clinical covariate – the
mixture assignment is a posterior latent-class index from the NONMEM
$MIXTURE block. One assignment per subject,
time-fixed.MIXTURE – NONMEM $MIXTURE block class
index in the Morris 2011 estimation run (component 1 = high-CL with P =
0.251, component 2 = low-CL with P = 0.749); the binary indicator is
MIX_FAST_ELIM = as.integer(MIXTURE == 1).Morris_2011_telapristone.R (selects the telapristone CL/F
typical value between the high-CL anchor
exp(lcl_pop1) = 11.6 L/h and the low-CL anchor
exp(lcl_pop2) = 3.34 L/h; the shared IIV
etalcl ~ 0.200 is applied multiplicatively to whichever
typical the mixture-class selects),
Wilkins_2011_isoniazid.R (gates the typical apparent CL/F
via the power-form covariate effect e_mix_fast_elim_cl on
CL/F:
cl = exp(lcl + e_mix_fast_elim_cl * MIX_FAST_ELIM + etalcl) * (WT/70)^0.75 * (1 + e_hiv_pos_cl * HIV_POS)
with lcl = log(9.70) (slow-eliminator reference) and
e_mix_fast_elim_cl = log(21.6/9.70) approx 0.8005).MIX_PDI
register entry’s Notes block (“mixture indicators from unrelated
dichotomies should register a new canonical name (e.g.,
MIX_FAST_ELIM for a fast/slow eliminator mixture)”).
Population probability of MIX_FAST_ELIM = 1 is recorded
per-model in covariateData[[MIX_FAST_ELIM]]$notes; in
Morris 2011 it is 0.251 (Table II, “Probability” row, RSE 61.0%). The
Morris 2011 Discussion attributes the bimodal CL distribution to
polymorphic CYP3A5 (functional CYP3A5 present in 10-40% of Caucasians,
~50% of African Americans, ~33% of Asians) but does not test the
genotype hypothesis directly; the mixture indicator is therefore a
latent-class label, not a CYP3A5-genotype indicator. For typical-value
simulation set MIX_FAST_ELIM = 1 (fast eliminator; lower
steady-state exposure, ~12 h elimination half-life per Morris 2011
Results) or MIX_FAST_ELIM = 0 (slow eliminator; higher
steady-state exposure, ~35 h half-life). For population simulation, draw
MIX_FAST_ELIM ~ Bernoulli(0.251) per subject. Scope:
general because the fast/slow-eliminator dichotomy is a recurring popPK
concept across drugs metabolized by polymorphic enzymes
(CYP2C92/3, CYP2C192, CYP3A53, NAT2 slow/fast
acetylators); future popPK papers fitting a $MIXTURE on CL
should re-use this canonical rather than introducing per-drug-specific
MIX_<drug>_FAST names. The reference category (= 0 =
slow eliminator) is chosen so the binary numerically matches the paper’s
mixture-indicator orientation. Ratified canonically on 2026-06-09
alongside the Morris 2011 telapristone extraction.$MIXTURE block (Parra-Guillen 2013 Methods ‘Data
Evaluation, Biological Assumptions and Mathematical Model’ subsection
bullet (e); Discussion ‘A challenge when using a mixture model’
paragraph). Time-fixed per subject.MIXTURE – NONMEM
$MIXTURE block class index in the Parra-Guillen 2013
estimation run (component 1 = responder / cure subpop with P(1) = 0.844
FIX, component 2 = relapser subpop); the binary indicator is
MIX_VAC_RELAPSE = as.integer(MIXTURE == 2) (orientation
flipped from the NONMEM convention so the named state corresponds to the
deviation from the dominant cure phenotype, following the Schindler 2017
/ Tsuji 2017 MIX_* naming pattern).ParraGuillen_2013_cyaaE7.R,
ParraGuillen_2013_il12.R (gates the SVAC degradation rate
inside Eq. 3: k2 <- MIX_VAC_RELAPSE * k1 so
MIX_VAC_RELAPSE = 0 freezes the inhibitory signal
indefinitely (cure) and MIX_VAC_RELAPSE = 1 makes it decay
at rate k1 (relapse). The same indicator is reused across both drug
files because the biological dichotomy is identical (responders vs
relapsers to immunotherapy) and the paper fixed P(1) = 0.844 from
CyaA-E7 when re-fitting IL-12).MIX_VAC_RELAPSE = 0 (dominant cure phenotype, complete
tumor regression; Figure 3 ‘light grey’ individual predictions) or
MIX_VAC_RELAPSE = 1 (relapser phenotype, tumor regrowth;
Figure 3 ‘dark grey’ individual predictions). For population simulation,
draw
MIX_VAC_RELAPSE ~ Bernoulli(1 - 0.844) = Bernoulli(0.156)
per subject. Discussion p. 803 notes the mixture is empirical: ‘no
statistical differences on tumour size at the moment of vaccine
administration were found between both populations (p > 0.05)’. The
cure semantics (‘a sub-population of mice able to trigger only a
temporal tumour response was postulated to describe the relapse observed
in a few mice … the vaccine is able to trigger a permanent immune
response (probably mediated by memory T-cells) in the cured mouse
population, but no in all the mice’) are tied to immune-memory
immunotherapy specifically and may not generalise to
cytotoxic-chemotherapy relapse mixtures, so scope is
specific. Future immunotherapy / vaccine / oncolytic-virus
papers that share the same cure-vs-relapse dichotomy at the level of a
vaccine-elicited inhibitory signal degradation rate may extend this
entry; mixture indicators from unrelated dichotomies (e.g.,
fast-acquired-resistance vs slow-acquired-resistance to cytotoxic
chemotherapy) should register a new canonical name. Ratified canonically
on 2026-06-03 alongside the Parra-Guillen 2013 CyaA-E7 + IL-12
extraction.MIXTURE – NONMEM
$MIXTURE block class index in the Svensson 2012 estimation
run (component 1 = fast with P = 0.827, component 2 = slow with P =
0.173); the binary indicator is
MIX_SLOW_ELIM_NVP = as.integer(MIXTURE == 2).Svensson_2012_nevirapine.R (gates the typical apparent oral
clearance between subpopulation-specific anchors:
cl_typ <- exp(lcl_fast) * (1 - MIX_SLOW_ELIM_NVP) + exp(lcl_slow) * MIX_SLOW_ELIM_NVP,
with lcl_fast = log(3.12) and
lcl_slow = log(1.45); the single 24.9 % BSV on CL/F is
shared across the two mixture sub-populations).MIX_SLOW_ELIM_NVP = 0 (the recommended default) to
reproduce the majority fast-eliminator phenotype; set
MIX_SLOW_ELIM_NVP = 1 to reproduce the rarer slow phenotype
that the paper associates with CYP2B6 516TT homozygotes. For population
simulation, draw MIX_SLOW_ELIM_NVP ~ Bernoulli(0.173) per
subject. Scope: specific because the drug-specific _NVP
suffix scopes the binary semantics to nevirapine CL/F mixtures; future
fast/slow CYP2B6-driven elimination mixtures for other antiretrovirals
(e.g., efavirenz, which shares CYP2B6 as the major elimination enzyme)
should register sibling canonicals (MIX_SLOW_ELIM_EFV,
etc.) rather than reuse this entry. The Bienczak 2016 and Schipani 2011
nevirapine extractions in the registry use CYP2B6 genotype indicators
directly (CYP2B6_IM / CYP2B6_SM /
CYP2B6_USM / SNP_CYP2B6_RS3745274_T_COUNT) and
are NOT mixture models – they assume the user supplies known genotype;
MIX_SLOW_ELIM_NVP is the orthogonal pathway for the
Svensson 2012 paper which estimates the phenotype distribution without
genotype data. Ratified canonically on 2026-06-14 alongside the Svensson
2012 nevirapine extraction.(1 + g_IL6(t)) to the IL-6
production rate); 0 = no elevated IL-6 production in this cycle (the
surge function is zeroed and IL-6 follows its baseline turnover only).
Not a measured clinical covariate – the gate implements the NONMEM
$MIXTURE block frequency that the Netterberg 2018 model
estimated from the data. Time-fixed within a single chemotherapy cycle;
redrawn independently between cycles 1 and 4 in the paper’s cohort.MIXTURE – NONMEM
$MIXTURE block class index in the Netterberg 2018
estimation run; the binary indicator is
MIX_ELEV_IL6 = as.integer(elevation-active component). The
paper reports 16 subpopulations across the 2 cycles x 2 biomarkers
indicator combinations, but the marginal IL-6-elevation probability per
cycle is a single value (Table 2 Pelevation,IL-6 = 63.4 %, RSE 10
%).Netterberg_2018_breast_cancer_FN_biomarkers.R,
Netterberg_2018_breast_cancer_FN_tte_preFN.R,
Netterberg_2018_breast_cancer_FN_tte_atFN.R (gates the IL-6
surge function:
g_il6 <- MIX_ELEV_IL6 * sa_il6_i / (((t - pt_il6_i) / sw_il6_i)^4 + 1),
so MIX_ELEV_IL6 = 0 zeroes the surge contribution while
MIX_ELEV_IL6 = 1 activates the Netterberg 2018 Equation 1
Lorentzian surge centred at PT_IL-6).MIX_ELEV_IL6 = 1 (the dominant
63.4 % per-cycle phenotype); for population simulation, draw
MIX_ELEV_IL6 ~ Bernoulli(0.634) per (subject, cycle). The
companion MIX_ELEV_CRP indicator gates the CRP surge
function in the same model. Ratified canonically on 2026-06-27 alongside
the Netterberg 2018 breast-cancer febrile-neutropenia extraction.MIX_ELEV_IL6 = 1). Not a measured clinical
covariate – the gate implements the NONMEM $MIXTURE block
frequency that the Netterberg 2018 model estimated from the data.
Time-fixed within a single chemotherapy cycle; redrawn independently
between cycles in the paper’s cohort.MIXTURE – NONMEM
$MIXTURE block class index in the Netterberg 2018
estimation run; the binary indicator is
MIX_ELEV_CRP = as.integer(elevation-active component). Per
Netterberg 2018 Results, the per-cycle marginal probability is 44.3 %
(Table 2 Pelevation,CRP, RSE 20 %), lower than the IL-6 marginal because
the IL-6 -> CRP linear regulation contributes additional CRP
elevations even when the dedicated CRP surge is off.Netterberg_2018_breast_cancer_FN_biomarkers.R,
Netterberg_2018_breast_cancer_FN_tte_preFN.R,
Netterberg_2018_breast_cancer_FN_tte_atFN.R (gates the CRP
surge function:
g_crp <- MIX_ELEV_CRP * sa_crp_i / (((t - pt_crp_i) / sw_crp_i)^4 + 1),
so MIX_ELEV_CRP = 0 zeroes the dedicated CRP-surge
contribution while MIX_ELEV_CRP = 1 activates the
Netterberg 2018 Equation 1 Lorentzian surge centred at PT_CRP = PT_IL-6
+ PT_CRP+).MIX_ELEV_CRP = 1
to reproduce a CRP-elevation cycle; for population simulation, draw
MIX_ELEV_CRP ~ Bernoulli(0.443) per (subject, cycle). The
paper observes that the actual fraction of CRP elevations is
higher than Pelevation,CRP because of the IL-6 RCFB
coupling through Slope; the canonical refers strictly to the dedicated
surge function and not the coupled-elevation path. Ratified canonically
on 2026-06-27 alongside the Netterberg 2018 breast-cancer
febrile-neutropenia extraction.covariateData[[SEIZURE_ACUTE]]$notes).A (values inverted: source A = 1 means
non-seizing control and A = 0 means seizing, per Clinckers
2008 Methods ‘Both A and B were set to 1 in control animals; to 0 and 1,
respectively, in seizing animals’; the canonical encoding flips the
polarity via SEIZURE_ACUTE = 1 - A) – used in
Clinckers_2008_MHD_rat.R.Clinckers_2008_MHD_rat.R (mutually-exclusive selection of
the seizure-specific biophase volume V3b in place of the control V3a
inside the V3 expression of the one-compartment-plus-biophase model for
MHD in male Wistar rats;
v3 = exp(lv3a + etalv3a) * (1 - SEIZURE_ACUTE) * (1 - EFFLUX_INHIB) + exp(lv3b) * SEIZURE_ACUTE * (1 - EFFLUX_INHIB) + exp(lv3c + etalv3c) * (1 - SEIZURE_ACUTE) * EFFLUX_INHIB).CONMED_AED, CONMED_EIAED,
PDV previous-period seizure count) describe distinct
concepts and should not be conflated. Future preclinical studies that
test a similar seizure-vs-non-seizure contrast should extend the example
list; promote to general scope once a second model legitimately ratifies
the name. Mutually exclusive with EFFLUX_INHIB in Clinckers
2008 (each rat is allocated to exactly one of {control, seizure,
efflux-inhibition}); the model() encoding uses
(1 - SEIZURE_ACUTE) * (1 - EFFLUX_INHIB) as the control
multiplier, so any data record that asserts both flags as 1 would zero
out the V3a term – data assemblers should preserve mutual exclusivity
unless a model is explicitly designed for the cross-condition case.
Ratified canonically on 2026-05-16 alongside the Clinckers 2008
extraction.T_ENTRY is a
per-subject time-offset covariate that anchors the
time-since-study-entry clock used by post-entry-only model terms
(placebo / learning effects, study-design transient drops).Delor_2013_alzheimer.R,
Clinckers_2008_MHD_rat.R (mutually-exclusive selection of
the verapamil-specific biophase volume V3c in place of the control V3a
inside the V3 expression:
v3 = v3a*(1-SEIZURE_ACUTE)*(1-EFFLUX_INHIB) + v3b*SEIZURE_ACUTE*(1-EFFLUX_INHIB) + v3c*(1-SEIZURE_ACUTE)*EFFLUX_INHIB).T_ENTRY
per subject in the simulation event table; a reasonable construction is
T_ENTRY = DOT_individual + a few years of established disease
so the patient is observed during disease progression (see the Delor
2013 vignette for the construction used to reproduce Figures 2-4).
Ratified canonically on 2026-05-16 alongside the Delor 2013
extraction.factive(HUMAN_SERUM_PCT) that scales the
total static drug concentration to the effective (free) concentration.
Categorical experimental level, not a continuous interpolatable
axis.HS – Garonzik 2016 (paper Methods + Table 2, “% Human
Serum”; the model column was renamed from the short HS to
the spelled-out canonical HUMAN_SERUM_PCT on 2026-05-27 to
align with the register’s naming standards and avoid an ambiguous
two-letter abbreviation).Garonzik_2016_daptomycin.R (five experimental levels {0,
10, 30, 50, 70} percent; drives factive multiplying the static
daptomycin concentration to the effective concentration DAP_EF, Garonzik
2016 Eq 2; factive estimates 0.346 / 0.284 / 0.239 / 0.252 at 10 / 30 /
50 / 70 percent and 1 at 0 percent by construction).DIS_BUNIONECTOMY entry’s avoidance of
DIS_BUN). Future in-vitro protein-binding-versus-serum
experiments should extend the example list. Ratified canonically on
2026-05-27 alongside the Garonzik 2016 daptomycin extraction.model().M3M3FBS_PRESENT.Grzesk_2016_m3M3FBS.R
(binary selector that swaps the per-agonist control (EC50, Emax) pair
for the +m-3M3FBS pair in the sigmoidal Emax CRC; the agonist identity
is selected by the companion AGONIST_CODE).CONMED_* (clinical-trial-level concomitant medication
indicators) because m-3M3FBS is an experimental ex-vivo pretreatment,
not a clinically administered comedication. Ratified canonically
alongside the Grzesk 2016 extraction.HCV_GT1B.HS – Garonzik 2016 (paper Methods + Table 2, “% Human
Serum”; the model column was renamed from the short HS to
the spelled-out canonical HUMAN_SERUM_PCT on 2026-05-27 to
align with the register’s naming standards and avoid an ambiguous
two-letter abbreviation).Garonzik_2016_daptomycin.R,
Wang_2018_daclatasvir_asunaprevir.R (multiplicative effect
on IC50 of both drugs via the fixed scaling factors SCL_IC50_DCV = 0.18
and SCL_IC50_ASV = 0.30; the encoding is
ic50_dcv_t0 = exp(lic50_dcv_gt1a + etalic50_dcv) * scl_ic50_dcv^HCV_GT1B
and
ic50_asv_t0 = exp(lic50_asv_gt1a + etalic50_asv) * scl_ic50_asv^HCV_GT1B.
Also switches the DCV resistance coefficient Kr_DCV between 0.43 /day
for GT1A and 0.13 /day for GT1B; Kr_ASV is the same for both
subtypes).DIS_BUNIONECTOMY entry’s avoidance of
DIS_BUN). Future in-vitro protein-binding-versus-serum
experiments should extend the example list. Ratified canonically on
2026-05-27 alongside the Garonzik 2016 daptomycin extraction.1 = bite from a snake of the family Elapidae (front-fanged
elapids – cobras, kraits, mambas, sea snakes, Australian terrestrial
elapids such as taipans / brown snakes / death adders); 0 =
bite from a snake of the family Viperidae (true vipers and pit vipers,
including Bothrops, Crotalus,
Daboia / Vipera russelli,
Vipera aspis / berus / ammodytes, Bitis,
Hypnale, Cerastes). The covariate is a
per-bite-event property of the dose source (the snake), not a property
of the patient; in a dataset of envenomed patients each subject’s row(s)
carry one value across the entire follow-up.Mann_2022_respiratory_physiology.R,
Sanhajariya_2018_snake_venom.R (Sanhajariya 2018 Table A1
covariate model:
f(central) <- exp(lfdepot + e_snakefamily_elapid_fdepot * SNAKEFAMILY_ELAPID + etalfdepot)
with lfdepot = log(1) fixed and
e_snakefamily_elapid_fdepot = log(0.569)).Syvanen_2011_verapamil_rat.R (the paper distinguishes a
tariquidar-treatment arm from a vehicle arm; the canonical column is 1
for the 21 tariquidar-treated rats and 0 for the 21 vehicle-treated
rats).Syvanen_2011_verapamil_rat.R (multiplicative
theta^COVARIATE form on three structural parameters:
vp = exp(lvp + e_conmed_tariquidar_vp * CONMED_TARIQUIDAR)
with e_conmed_tariquidar_vp = log(1.20) = 0.1823 (20%
increase in plasma peripheral 1 volume);
vbr1 = ... * exp(e_conmed_tariquidar_vbr1 * CONMED_TARIQUIDAR + ...)
with e_conmed_tariquidar_vbr1 = log(2.41) = 0.8796
(2.41-fold increase in fast-exchange brain compartment volume);
qin = exp(lqin + e_conmed_tariquidar_qin * CONMED_TARIQUIDAR)
with e_conmed_tariquidar_qin = log(12.0) = 2.4849
(12.0-fold increase in BBB influx clearance). The paper screened
tariquidar as a covariate on Qout as well but did not
retain it.), Syvanen_2012_quinidine_rat.R (multiplicative
exponential effects on CL, Q_out, Q_in, and f2:
e_conmed_tariquidar_cl = log(0.835) = -0.180;
e_conmed_tariquidar_qout = log(0.366) = -1.005;
e_conmed_tariquidar_qin = log(2.65) = 0.975;
e_conmed_tariquidar_f2 = log(5.66) = 1.733. Combined Q_in /
Q_out shift raises the pseudo-equilibrium brain ECF : plasma ratio
7.2-fold and the combined f2 shift raises the total-brain : plasma ratio
about 40-fold per paper Results p91),
Westerhout_2013_quinidine.R (P-gp split into separate
passive CL_X,p and P-gp-mediated CL_X,P-gp
clearances per the paper Appendix mass-balance equations; the P-gp
components are switched off when CONMED_TARIQUIDAR = 1 via
the multiplier pgp_active = 1 - CONMED_TARIQUIDAR and
contribute additively (to efflux clearances CL_X-PL) or
subtractively (to influx clearances CL_PL-X) when
CONMED_TARIQUIDAR = 0; 1.9-fold P-gp effect on systemic
CL_E from Table 4 SBPK combined column, with full
bidirectional P-gp components on plasma <-> brain_deep / brain_ecf
/ brain_csf_lv exchanges and P-gp absent at the CM in this final
model).covariateData[[CONMED_TARIQUIDAR]]$notes must
document the dose / regimen of tariquidar used (Syvanen 2011: 15 mg/kg
IV bolus in 3 mL/kg of 5% glucose in saline, administered 20-30 min
before the radiotracer) and any per-subject vs per-record time-varying
convention. Analogous to [[CONMED_PROBENECID]] in the Xie 2000 rat
BBB-transport paper – both are experimentally-administered transporter
inhibitors used to dissect a tracer’s brain-vs-plasma distribution,
encoded as a binary co-administration indicator regardless of the
inhibitor’s clinical use case. Future preclinical-PET extractions that
test tariquidar (or a related Pgp inhibitor like elacridar / zosuquidar
with comparable mechanism) on a different probe-substrate should reuse
this canonical when the inhibitor is tariquidar; structurally distinct
P-gp inhibitors should register a separate canonical with the
inhibitor’s INN in the name (e.g., CONMED_ELACRIDAR).
Ratified canonically on 2026-06-03 alongside the Syvanen 2011
(R)-[11C]verapamil rat PET extraction.L_OPIOID_pM. In a composed Mann 2022
+ Laffont 2024 / 2025 chain, the upstream antagonist PK layer
(Laffont_2024_naloxone or
Laffont_2024_nalmefene) is post-processed in the vignette
by (a) converting time to minutes, (b) convolving plasma concentration
with the Mann 2022 ke0 = 0.001774 1/s effect-site equilibration (carried
into Laffont 2024 Supp Table S3 unchanged for both nalmefene and
naloxone), and (c) converting ng/mL to pM via the antagonist’s free-base
molecular weight (naloxone 327.37 g/mol, nalmefene 339.43 g/mol); the
resulting per-subject time series is supplied as this covariate.Mann_2022_mu_receptor_binding.R.L_OPIOID_pM.
Ratified canonically on 2026-05-29 alongside the Mann 2022
translational-model extraction.RL_op output of
Mann_2022_mu_receptor_binding.R; in standalone
physiology-only use, the operator supplies CAR_OPIOID as a time-varying
data column.RL_op /
CAR (binding-model output name).Mann_2022_respiratory_physiology.R.CAR_KAPPA). Ratified canonically on 2026-05-29
alongside the Mann 2022 translational-model extraction.Q_Scale = 1 + 1 / (1 + exp((1.6 - Q_TOTAL_LPM / 4.87) / 0.05)),
clamped to [1, 2]. Q_Scale scales the effective central volume of
distribution down (vc_eff = vc / Q_Scale) so that
hyperperfusion-driven concentration of opioid in the central / biophase
compartment is captured during overdose-induced chemoreflex
hyperperfusion. Without this feedback the standalone PK model
under-estimates effect-site concentration in shock conditions and
produces PaO2 troughs too shallow to reach the cardiac-arrest
threshold.q_total /
Q_total / Q (physiology layer state
name).Mann_2022_fentanyl_iv.R,
Mann_2022_carfentanil_iv.R.Q_total = qb + qt output of
Mann_2022_respiratory_physiology.R; in standalone PK-only
use the operator supplies Q_TOTAL_LPM = 4.87 (or whatever fixed baseline
appropriate). Registered canonically on 2026-06-07 alongside the FDA
shock-state PK amplification fix.Mann_2022_respiratory_physiology.R.severe HF or LF – narrative composite used in
Fattinger_1991_quinidine.R (Table 1 row ‘CL_nonrenal for
patients with severe HF or LF’; severe HF defined as low output or
pulmonary oedema, n = 2; severe LF defined as serum bilirubin > 30
umol/L AND prothrombin time < 60% of normal, n = 3; pooled because
the two effects on non-renal CL were of similar magnitude and per-group
counts were small).Fattinger_1991_quinidine.R (multiplicative reduction of the
non-renal CL arm of total apparent CL from 12.6 L/h to 6.8 L/h when set
to 1; log-multiplicative effect
e_dis_hf_or_lf_sev_cl_nonren = log(6.8/12.6) = log(0.5397);
Table 1 reduction in objective function 10.8, P < 0.005).specific because the
pooled HF-or-LF semantic is paper-specific (Fattinger 1991 pools the two
severe end-organ-failure axes because the cohort had n = 2 + 3 subjects
in those groups and similar effect sizes). A future model that retains
severe HF and severe LF as separate covariates (with enough subjects in
each to estimate them independently) should use the existing
HEPIMP_SEV canonical for the severe-hepatic axis and
register a parallel CARDIMP_SEV canonical for the
severe-cardiac axis rather than reuse this pooled indicator. The
DIS_HF_OR_LF_SEV pooling preserves the load-bearing convention of the
source paper without forcing later users to artificially separate the
two axes when the published evidence base lumps them. Ratified
canonically on 2026-06-04 alongside the Fattinger 1991 quinidine
extraction.Fattinger_1991_quinidine.R (structural switch driving
dur(central) between
exp(ldur_qs + etaldur_qs) = 1.37 h * exp(eta) and
exp(ldur_qbs) = 6.0 h, and f(central) between
1 (QS reference) and exp(lfdepot) = 1.36 (QBS); the IIV on
QS absorption duration applies only when FORM_QUIN_SR = 0 per Methods
page 282).FORM_* family alongside
FORM_TAC_IR (tacrolimus immediate-release vs
prolonged-release), FORM_LINAG_TAB1 (linagliptin tablet 1),
FORM_VISMO_PHASEI (vismodegib Phase I dry-blend capsule),
FORM_THEO_APNECUT (Apnecut vs theophylline-alcohol),
FORM_ASV_LIQUID (asunaprevir liquid), and
FORM_ABA_PHASE2 (abatacept SC Phase-2). Doses must be
entered in mg of quinidine BASE (apply the Windholz 1983 stoichiometric
factors of 0.829 mg base per mg quinidine sulphate and 0.663 mg base per
mg quinidine bisulphate before passing to the model); the f(central)
term then captures only the formulation-driven absorption difference,
not the stoichiometric salt-vs-base conversion. Ratified canonically on
2026-06-04 alongside the Fattinger 1991 quinidine extraction.Mann_2022_mu_receptor_binding.R::ini()): 1 = alfentanil; 2
= buprenorphine; 3 = butyryl fentanyl; 4 = carfentanil; 5 =
fluorobutyryl fentanyl; 6 = fentanyl; 7 = fluoroisobutyryl fentanyl; 8 =
furanyl fentanyl; 9 = isobutyryl fentanyl; 10 = naloxone; 11 =
remifentanil; 12 = sufentanil; 13 = nalmefene (added by task 131 from
Laffont 2024 Supp Table S3 scaling approach over Cassel 2005).
Out-of-range values cause every dispatch indicator to evaluate to 0,
which zeros the agonist slot rates rather than throwing an error.Mann_2022_mu_receptor_binding.R.OPIOID_ID. Mann 2022 itself
uses only naloxone (10) as the antagonist; Laffont 2024 / Laffont 2025
add nalmefene (13) as the second commercially-relevant intranasal
antagonist option. The wider 1..13 range is retained so downstream tasks
can flip slots without code changes (e.g., partial-agonist buprenorphine
2 in the antagonist slot to model receptor occupancy
without full agonist effect).Mann_2022_mu_receptor_binding.R.OPIOID_ID (Mann 2022 Table S2 rows 1..12 verbatim; Laffont
2024 Supp Table S3 nalmefene at row 13). Out-of-range values zero the
antagonist slot rates safely. Ratified canonically on 2026-05-29
alongside the Mann 2022 translational-model extraction; 13th-ligand
extension committed by task 131 on the same day.Mann_2022_mu_receptor_binding.R consumes this column
directly. In a composed Mann 2022 chain, the upstream IV-opioid PK layer
(Mann_2022_fentanyl_iv.R or
Mann_2022_carfentanil_iv.R) exposes its effect-site
Ce_pM as this covariate; in standalone use, the operator
supplies it as a time-varying data column on the subject’s records.Ce_pM (upstream PK
output name when composed in-chain).Mann_2022_mu_receptor_binding.R.L_OPIOID_nM) rather than reusing this name with a
different unit, because the binding model’s downstream math depends on
the unit match. Ratified canonically on 2026-05-29 alongside the Mann
2022 translational-model extraction.L_OPIOID_pM. In a composed Mann 2022
+ Laffont 2024 / 2025 chain, the upstream antagonist PK layer
(Laffont_2024_naloxone or
Laffont_2024_nalmefene) is post-processed in the vignette
by (a) converting time to minutes, (b) convolving plasma concentration
with the Mann 2022 ke0 = 0.001774 1/s effect-site equilibration (carried
into Laffont 2024 Supp Table S3 unchanged for both nalmefene and
naloxone), and (c) converting ng/mL to pM via the antagonist’s free-base
molecular weight (naloxone 327.37 g/mol, nalmefene 339.43 g/mol); the
resulting per-subject time series is supplied as this covariate.Mann_2022_mu_receptor_binding.R.L_OPIOID_pM.
Ratified canonically on 2026-05-29 alongside the Mann 2022
translational-model extraction.RL_op output of
Mann_2022_mu_receptor_binding.R; in standalone
physiology-only use, the operator supplies CAR_OPIOID as a time-varying
data column.RL_op /
CAR (binding-model output name).Mann_2022_respiratory_physiology.R.CAR_KAPPA). Ratified canonically on 2026-05-29
alongside the Mann 2022 translational-model extraction.Mann_2022_respiratory_physiology.R.Grzesk_2016_m3M3FBS.R::ini()
parameter names): 1 = phenylephrine (PHE; alpha1-adrenergic receptor
agonist); 2 = arg-vasopressin (AVP; V1 vasopressin receptor agonist); 3
= mastoparan-7 (heterotrimeric G-protein direct activator); 4 = Bay
K8644 (L-type voltage-gated calcium channel agonist). Out-of-range
values cause every dispatch indicator to evaluate to 0, which zeros the
sigmoidal Emax expression rather than throwing an error.M3M3FBS_PRESENT selects within each agonist between the
control and +m-3M3FBS sub-pairs).AGONIST_CODE.Grzesk_2016_m3M3FBS.R
(Grzesk 2016 Table I + Table II Phase 2: per-agonist EC50 (M/L) and
maximal-perfusion-pressure Emax (mmHg) pairs; companion
M3M3FBS_PRESENT flips between control and
m-3M3FBS-pretreatment sub-pairs).OPIOID_ID /
ANTAGONIST_ID (Mann 2022 opioid-binding panel) because the
agonists here are not within a single receptor-binding family – PHE,
AVP, mastoparan-7, and Bay K8644 hit four mechanistically distinct
contractile pathways (alpha1, V1, G-protein direct, L-type Ca channel),
and the model uses agonist-specific (EC50, Emax) pairs rather than a
shared binding-kinetic parameter set. Ratified canonically alongside the
Grzesk 2016 extraction.Schmid_2017_nintedanib.R (combined with
STUDY_NSCLC_NIN_PH2 to drive the ka phase-II trial-group
effect = 2.20x; also captures the ka2 BIBF 1202 metabolite phase-II
trial-group effect = 0.756x).Schmid_2017_nintedanib.R (combined with
STUDY_LUMELUNG2 to drive the F1 trial-group effect = 1.30x;
combined with STUDY_TOMORROW to drive the ka phase-II
trial-group effect = 2.20x and the ka2 BIBF 1202 metabolite phase-II
trial-group effect = 0.756x).Schmid_2017_nintedanib.R (acts as the reference cohort for
the F1 trial-effect and as part of the ka / ka2 reference cohort).STUDY_NSCLC_NIN_PH2 for
the F1 trial-group effect.Schmid_2017_nintedanib.R (combined with
STUDY_NSCLC_NIN_PH2 to drive the F1 trial-group effect =
1.30x; acts as the reference cohort for the ka / ka2 trial-effect).D x Q – used in
BuchwalderCsajka_1999_angiotensin.R (Buchwalder-Csajka 1999
Table 1 row 5 column header, where D is the raw angiotensin
dose in ug and Q is the molar-weight conversion
factor).BuchwalderCsajka_1999_angiotensin.R (drives the algebraic
Emax dose-response
E = Emax * DOSE_AGT_UG / (DOSE_AGT_UG + ED50) separately
for SBP and DBP).AMT / EVID = 1 dosing column
because the model is purely algebraic (no PK, no ODEs); the dose enters
the model() block as a covariate symbol rather than via a dosing event.
Specific scope because the column’s semantics (Q-corrected angiotensin
dose) are tied to angiotensin-challenge protocols; future
angiotensin-challenge extractions may reuse this canonical. Ratified
canonically on 2026-06-10 alongside the Buchwalder-Csajka 1999
extraction.nmol*h/L (document per-model
via covariateData[[AUC_EMPA]]$units if a different exposure
unit is reported).emax_i * AUC_EMPA / (auc50 + AUC_EMPA) on the HbA1c
drug-effect term. Set AUC_EMPA = 0 to recover the placebo arm
(drug-effect term vanishes; HbA1c follows baseline + placebo drift
only). Reference values observed: AUC50 = 498 nmol*h/L (Johnston 2019
Table 2 typical-value AUC at half-maximal effect; corresponds
approximately to the simulated median AUCss for the 2.5 mg QD arm).AUCss (or AUCSS,i) – the printed variable
name in Johnston 2019 Equation 1 (the individual steady-state AUC
entering the Emax expression).Johnston_2019_empagliflozin.R (M-EASE-2 exposure-response
PD model for HbA1c in adults with type 1 diabetes; AUC_EMPA drives the
direct Emax HbA1c-reduction term).AUC_CARBO,
AUC_GEM, AUC_BAST_FW, AUC_PAZO,
AUC_GCV, AUC_RTV) follow the same
AUC_<DRUG> naming pattern. A future PK/PD model that
uses a different empagliflozin exposure metric (trough concentration,
q12h-interval AUC) should register a parallel canonical rather than
overload AUC_EMPA. Ratified canonically alongside the
Johnston 2019 M-EASE-2 extraction.INS_BL (baseline fasting plasma insulin concentration, a
measured biomarker in pmol/L) – INSDOSE_BL is the
administered dose level, not the resulting plasma level.U/kg/day (international units
of insulin per kilogram body weight per day). Document per-model via
covariateData[[INSDOSE_BL]]$units if a paper reports total
daily units without body-weight normalisation.(INSDOSE_BL / ref)^e_insdose_bl_<param>. Reference
values observed: 0.660 U/kg/day (Johnston 2019 reference-patient
description).IDB – “Insulin daily dose at Baseline” abbreviation
used in Johnston 2019 Table 2.Johnston_2019_empagliflozin.R (M-EASE-2 covariate effect on
the typical baseline HbA1c and on Emax in T1DM patients with background
insulin therapy; effect exponents 0.0141 on baseline HbA1c and 0.0552 on
Emax, both with CIs that cross zero – the covariate was retained as part
of the full-random-effects covariate model rather than dropped).covariateData[[INSDOSE_BL]]$notes. A model that uses
time-varying insulin dose (rather than baseline only) should register a
parallel canonical (e.g., INSDOSE). Companion concept to
INS_BL (plasma insulin concentration) and to
HBA1C (glycemic control). Ratified canonically alongside
the Johnston 2019 M-EASE-2 extraction.INSDT – “Insulin Dose Type” abbreviation used in
Johnston 2019 Table 2 (the two-level categorical column decomposed into
the binary CSII indicator with MDI as the reference category).Johnston_2019_empagliflozin.R (M-EASE-2 covariate
multiplier on baseline HbA1c, Emax, and placebo rate; CSII shifts
placebo drift upward by a factor of 1.47 vs MDI – a moderate effect on
the placebo-adjusted HbA1c change).<COLUMN>_<LEVEL> decomposition pattern used for
REGI_BID / TUMTP_GLIO. A future paper that
retains the same INSDT covariate but with more than two levels (e.g.,
CSII vs MDI vs continuous glucose monitoring + open-loop algorithm)
would register sibling canonicals (e.g., INSDT_CGM) rather
than overload this entry. Ratified canonically alongside the Johnston
2019 M-EASE-2 extraction.LSCeff(t) = 1 - LSCI * Kdiet * (exp(-Kdiet * t) - exp(-Kred * t)) / (Kred - Kdiet)
applied multiplicatively to EI. Subjects in the same study arm typically
share the same LSCI value; the column is time-fixed at study entry (a
single LSCI per subject per arm).LSCI – used directly in
Bosch_2024_glp1ra_bodyweight.R (Bosch 2024 supplement S10
THETA7..THETA16 are per-study fitted LSCI values).Bosch_2024_glp1ra_bodyweight.R (introduces the canonical;
per-study fitted values 0..0.548 across diet and GLP-1RA studies, with
-0.0288 reported for the Blundell 2017 arm).covariateData[[LSCI]]$notes so users can match the
source-paper LSCI to the simulated arm. Ratified canonically on
2026-06-22 alongside the Bosch 2024 extraction.IFLAG – used in
Bosch_2024_glp1ra_bodyweight.R (Bosch 2024 supplement S10
IFLAG > 0 gates the activity effect).Bosch_2024_glp1ra_bodyweight.R (introduces the canonical;
gates the body-weight-dependent activity effect on PAE for STEP-family
arms in the Bosch 2024 dataset).covariateData[[WM_IBT]]$notes so users can distinguish
“weekly clinic counselling + diary review” intensity (STEP 3) from
“minimal-counselling -500 kcal/d” intensity (STEP 5 / STEP 8). Ratified
canonically on 2026-06-22 alongside the Bosch 2024 extraction.