**A comparison of two model-based approaches to investigating covariate effects on the dose-exposure relationship in a Phase III context**

Justin Wilkins & Michael Looby

Novartis Pharma AG, Basel, Switzerland

**Objectives:** Linear and nonlinear mixed-effects model-based approaches to investigating covariate effects on exposure were compared in order to determine whether the more straightforward linear approach was sufficient for addressing this question in a subject-rich but observation-poor context.

**Methods:** Sparse concentration-time data were simulated according to commonly-used Phase III pivotal study designs and pharmacokinetic sampling schemes. A one-compartment model with linear absorption and elimination, incorporating covariate effects (analogous to age and body weight) on CL/F and V/F, was used as the basis for simulation. The simulated covariate relationships exerted effects of between 5% and 50% on model parameters. Variability was included in model parameters – 40% in CL/F, 40% in V/F and 50% in KA. Each scenario was replicated 1000 times, after which each replicate was analyzed using two discrete approaches: nonlinear mixed-effects (NLME) analysis of all the simulated data in a given scenario using a compartmental model implemented in NONMEM VI, and linear mixed-effects (LME) analysis of observed peaks and pre-dose troughs, implemented in R 2.8.1. Relative efficiency of the approaches was evaluated in terms of the rate at which each approach in each scenario failed to select the included “true” covariate relationship at the 5% significance level, as well as in terms of bias and precision in the quantitative estimates of the magnitude of the covariate effect in each case.

**Results:** Although analysis is ongoing, preliminary results indicate that LME models provided an answer in a fraction of the computing time required for full NLME analyses, although the NLME approach generally provided more precise estimates of the magnitude of the effect. The complete analysis will be presented.

**Conclusions:** Although work in ongoing and definitive conclusions cannot yet be drawn, LME analyses are much faster than NLME analyses in this context, and seem to identify important covariate relationships appropriately, given the typical design scenarios studied. In the context of population pharmacokinetic analysis for the determination of covariate influences on the dose-exposure relationship in Phase III, the LME approach may be an acceptable alternative to a full compartmental analysis using NLME.