2.6 Covariates Model
The following variables were explored as covariates for their potential
to influence iberdomide PK parameters:
- Demographic factors: age, body weight, body surface area (BSA), body
mass index (BMI), sex, and race;
- Baseline laboratory test: total bilirubin (BILI), aspartate
aminotransferase (AST), alanine aminotransferase (ALT), Alkaline
Phosphatase (ALP) and Albumin (ALB);
- Renal function marker: creatinine clearance estimated by
Cockcroft-Gault formula, Estimated Glomerular Filtration Rate (eGFR)
by CKD EPI equation;
- Disease status: healthy subjects versus MM patients;
- Baseline disease: ECOG, ISS stage at study entry;
- Iberdomide combinations: monotherapy versus Iberdomide in combination
with DEX versus Iberdomide in combination with DEX and DARA
Data visualization was used to examine the relationship between
intrinsic or extrinsic factors and subject-level PK parameters. Initial
selection of covariates was guided by graphic inspection and biological
plausibility. Potential covariates were tested further in Monolix.
COSSAC (conditional sampling use for stepwise approach based on
correlation tests) was used for covariate search based on a previous
publication 11. The iterations of COSSAC alternated
between a forward selection and a backward selection, depending on the
results of the correlation tests. In brief, during the forward selection
step, covariate with the smallest correlation p-value (among the
remaining parameter-covariate relationships) was added to the model, or
the next smallest if the smallest has already been tried, and so until
no correlation p-values above a threshold remain. Acceptance/rejection
of the relationship was based BIC: The covariate was not retained if the
criteria (0.3) did not improve (with a threshold for the likelihood
ratio test). During the backward selection, among the covariates
presented in the model, covariate with the highest (less significant)
correlation p-value, or the next highest if the highest has already been
tried, was removed, until no correlation p-values below a threshold
remain. Acceptance/rejection of the relationship removal was based on
BIC: The covariate was not retained if the criteria (0.01) did not
improve (with a threshold for the likelihood ratio test). The algorithm
continued until no forward or backward selection was possible.