Applying interpretable machine learning workflow to evaluate exposure–response relationships for large‐molecule oncology drugs

Abstract

The application of logistic regression (LR) and Cox Proportional Hazard (CoxPH) models are well‐established for evaluating exposure–response (E–R) relationship in large molecule oncology drugs. However, applying machine learning (ML) models on evaluating E–R relationships has not been widely explored. We developed a workflow to train regularized LR/CoxPH and tree‐based XGboost (XGB) models, and derive the odds ratios for best overall response and hazard ratios for overall survival, across exposure quantiles to evaluate the E–R relationship using clinical trial datasets. The E–R conclusions between LR/CoxPH and XGB models are overall consistent, and largely aligned with historical pharmacometric analyses findings. Overall, applying this interpretable ML workflow provides a promising alternative method to assess E–R relationships for impacting key dosing decisions in drug development.

Predicting survival and trial outcome in non-small cell lung cancer integrating tumor and blood markers kinetics with machine learning

“…Machine learning models, although increasingly used in pharmacological studies —including recently for TK-OS modeling and variable selection 22,53 —have yet rarely been rigorously compared to classical statistical models 24 . Here, such comparison revealed significantly better performance of the nonlinear random survival forest RSF model compared to the linear proportional hazards Cox model.…”

Realizing the promise of Project Optimus: Challenges and emerging opportunities for dose optimization in oncology drug development

“…Another valuable use of ABM is seen in optimizing the design and analysis of tumor biopsies by modeling spatiotemporal dynamics and leveraging advanced in vitro systems to emulate immune cell infiltration, which could be useful for finding early biomarkers of biological activity and treatment effectiveness 93,94 . Overall, we believe machine learning and artificial intelligence, and many other advanced analytical techniques, hold promise in the development of more accurate and personalized dose–response relationships in oncology drug development 95–97 . However, these techniques must be carefully validated and integrated with traditional pharmacological and clinical approaches to ensure their reliability and applicability in clinical practice.…”

Machine Learning‐Based Quantification of Patient Factors Impacting Remission in Patients With Ulcerative Colitis: Insights from Etrolizumab Phase III Clinical Trials

“…The process of E-R analysis involves making a causal inference, [9][10][11][12] which is a discipline of making quantitative statements regarding counterfactual (or "what-if ") outcomes, 13 as opposed to merely quantifying the observed outcomes. It is well known that in several diseases, the patients' disease status may have an impact on both their drug exposures 14 as well as on their outcomes, hence appropriate adjustment for such potential confounders is necessary in E-R analysis.…”

“…16 Although parametric modeling has been widely applied in E-R analyses, 17 recent works have shown that ML provides a promising avenue in situations where many confounders may exist and exerting nonlinear influences on exposure and/or response variables. 11,16,18 In this work, we apply an ML workflow 16 based on the explainability methodology of SHapley Additive exPlanations (SHAP) 19 and utilizing the underlying causal diagram to perform E-R analysis for the induction and maintenance phases of etrolizumab phase III clinical trials. Additionally, we aim to identify the set of key prognostic factors that contribute to the predicted outcomes beyond their impact on etrolizumab exposure and in what nonlinear functional forms they influence the predicted outcome.…”

“…In particular, in the setting of complex trial designs involving induction and maintenance phases, it is crucial to draw the causal diagram 15 to encapsulate the relationships among exposures, confounders, and clinical outcomes 16 . Although parametric modeling has been widely applied in E‐R analyses, 17 recent works have shown that ML provides a promising avenue in situations where many confounders may exist and exerting nonlinear influences on exposure and/or response variables 11,16,18 . In this work, we apply an ML workflow 16 based on the explainability methodology of SHapley Additive exPlanations (SHAP) 19 and utilizing the underlying causal diagram to perform E‐R analysis for the induction and maintenance phases of etrolizumab phase III clinical trials.…”