Janet R Wade (1), Rik Schoemaker (1), Lene Alifrangis (2)
(1) Occams, The Netherlands, (2) Symphogen A/S, Denmark
Introduction: Sym004 is a mixture of two synergistic full-length anti-EGFR antibodies (futuximab & modotuximab) that bind to two separate non-overlapping epitopes on the EGFR and inhibit the sustained growth of cancer cells. A population PK analysis of Sym004 on preliminary data has been performed previously [1].
Objectives:
1. Update the Sym004 population PK model that describes the available PK data from four trials.
2. Build full and reduced covariate models where specific focus will lie on the identification of covariates that could help describe interindividual variability in Sym004 PK. The full covariate model [2] can be considered exploratory in nature. The reduced covariate model will contain only statistically significant and clinically relevant covariates. The implications of different methods used to estimate the standard errors (SEs) of the covariate effects will be explored.
Methods: Sparse and richly sampled Sym004 PK data points (5341) from four completed phase 1 and 2 trials (Sym004-01, Sym004-02, Sym004-05 and Sym004-06) were included in the analysis. The majority of the 330 patients had metastatic colorectal cancer (mCRC) (n=247) and the remaining had various types of advanced solid tumours.
A population PK analysis was performed using NONMEM version 7.3.0. A suitable base structural model that accounted for the observed non-linearity in Sym004 PK was developed comprising of a two-compartment model with both linear and non-linear Michaelis-Menten-(MM) type elimination to describe target mediated drug disposition. An appropriate statistical model that describes the variability in the data was developed.
A full covariate model was developed which contained all pre-defined plausible covariate influences regardless of statistical significance and effect size. The development of the full covariate model required simplification of the statistical model originally defined for the base model due to computational difficulties. The full covariate model was then reduced by backwards deletion, retaining only statistically significant and clinically relevant covariate effects.
The SEs of the model parameters were obtained by MATRIX=RSR, MATRIX=S, bootstrap, and sampling importance resampling (SIR) [3].
The covariates in the analysis included age, weight, sex, race, glomerular filtration rate, albumin, total bilirubin, alanine transaminase, tumour size at baseline, visit-specific tumour size, tumour type, ECOG, time since previous treatment with anti-EGFR, previous treatment with cetuximab, previous treatment with panitumumab, and previous treatment with bevacizumab.
Results: The Sym004 base population PK model included the influence of weight on clearance (CL), maximum capacity of the MM elimination (Vmax), and the volumes of the central and peripheral compartments (V1 and V2, respectively).
The full covariate model contained all predefined plausible covariates regardless of statistical significance; 16, 11, 2 and 1 covariates were included on CL, Vmax, V1 and V2, respectively.
The reduced covariate model included the influences of weight on CL, Vmax, V1 and V2, the influence of sex and albumin on CL, and baseline tumour size of Vmax. Both the full and reduced covariate models described the data well and as expected, the covariate models reduced the IIVs as compared to the base model.
The estimated standard errors of the covariate effects in the full and reduced covariate models varied with the method used to obtain the SE’s. The MATRIX=S method yielded inflated SE estimates compared to other methods.
Conclusions: The full covariate model included all covariates and thus allowed the assessment of which covariates are important for potential dosing modifications, which can be said to have no clinical relevance, and for which covariate effects more information may be needed. In all these three situations the magnitude of the estimated SEs for the covariate effects provides useful information. The reduced covariate model for Sym004 contained well estimated statistically significant influences of weight, albumin, sex and baseline tumour size.
References:
[1] Wade JR, Schoemaker R, Alifrangis L. Population PK analysis of Sym004 and the influence of variations in base model structure on covariate model building. PAGE 26 (2017) Abstr 7119 [www.page-meeting.org/?abstract=7119].
[2] Gastonguay MR. Full Covariate Models as an Alternative to Methods Relying on Statistical Significance for Inferences about Covariate Effects: A Review of Methodology and 42 Case Studies. PAGE 20 (2011) Abstr 2229 [www.page-meeting.org/?abstract=2229].
[3] Dosne AG, Bergstrand M, Harling K, Karlsson MO. Improving the estimation of parameter uncertainty distributions in nonlinear mixed effects models using sampling importance resampling. J Pharmacokinet Pharmacodyn. 2016;43(6):583-596.
Reference: PAGE 28 (2019) Abstr 8841 [www.page-meeting.org/?abstract=8841]
Poster: Methodology - Covariate/Variability Models