Stochastic gates for covariate selection in population pharmacokinetics modelling

Abstract

Covariate selection in population pharmacokinetics modelling is essential for understanding interindividual variability in drug response and optimizing dosing. Traditional stepwise covariate modelling is often time-consuming, compared to the new machine learning alternatives. This study investigates the use of Neural Networks with Stochastic Gates for automated covariate selection, aiming to efficiently identify relevant covariates while penalizing excessive covariate inclusion. On various synthetic datasets the approach demonstrated robustness in detecting important covariates, overcoming challenges such as high correlations, low covariate frequencies, high interindividual variability and complex covariate dependencies. In real clinical data from a monalizumab study, the method successfully identified covariates that matched those found by experts. However, for tixagevimab/cilgavimab, it identified a superset of covariates, indicating a potential need for further pruning. This machine learning-based method enhances the covariate pre-selection process in population pharmacokinetics model development, offering significant time savings and improving efficiency even under challenging scenarios. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=89 SRC="FIGDIR/small/656586v1_ufig1.gif" ALT="Figure 1"> View larger version (22K): org.highwire.dtl.DTLVardef@1f5180borg.highwire.dtl.DTLVardef@1fd97e7org.highwire.dtl.DTLVardef@1ffc3d0org.highwire.dtl.DTLVardef@90c207_HPS_FORMAT_FIGEXP M_FIG C_FIG Study HighlightsO_ST_ABSWHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?C_ST_ABSCovariate selection in population pharmacokinetics modelling is a crucial step in drug development and dosing determination. Traditionally, this process is conducted in a stepwise manner, which can be very time-consuming. Recently, fast machine learning (ML)-based methods for covariate selection have emerged in the literature, offering more efficient alternatives. WHAT QUESTION DID THIS STUDY ADDRESS?Can we use Neural Networks with Stochastic Gates, incorporating explicit penalization on the number of covariates, to select superior set of covariates compared to prior ML-based methods? WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?Neural Networks with Stochastic Gates provide a reliable approach in accurately detecting covariates because they can effectively eliminate irrelevant covariates even in cases with high inter-covariates correlation. HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?This approach accelerates pharmacokinetic model development by significantly reducing the time required for covariate evaluation. Additionally, it allows for the screening of a wider set of covariates, potentially leading to fewer falsely missed covariates and better-quality models, a task that would be infeasible with traditional stepwise covariate modelling.