PS03: Advanced Covariate Adjustment in Randomized Clinical Trials: from Linear, Survival, to ML-Enhanced Methods

The importance of covariate adjustment in randomized clinical trials (RCTs) has been emphasized by recent FDA guidance "Adjusting for Covariates in Randomized Clinical Trials for Drugs and Biological Products" (2023), which recommends pre-specified and methodologically sound adjustment to improve the efficiency, interpretability, and regulatory reliability of trial outcomes. Covariate adjustment has the potential to enhance statistical power, control chance imbalances, and reduce required sample sizes, making it a reliable high-quality analysis approach for successful regulatory submissions. In a landscape shaped by rapid advancements in statistical methodology and evolving regulatory standards, it is essential for trial practitioners to master the latest strategies for covariate adjustment. This session brings together leading experts to present actionable guidance and highlight breakthrough advances that are setting new benchmarks for analytic rigorous causal inference, regulatory alignment, and real-world clinical impact. Presentations and discussions will cover linear regression, survival models, and integration with machine learning (ML)-enhanced methods. Special emphasis will be placed on the rapidly growing capabilities of ML, which enable flexible, data-driven covariate adjustment strategies that overcome model limitations, adapt to complex prognostic information, and further increase analytic precision and efficiency. Two speakers from the industry will share their recent innovative research on covariate adjustment: • Speaker 1: Prognostic Covariate Adjustment in Non-Ideal Conditions: Limitations and Doubly Robust Alternatives. The speaker will present cases when prognostic covariate adjustment may or may not help, illustrating how ML can be leveraged in constructing and validating prognostic scores within regulatory frameworks. • Speaker 2: Covariate Adjustment for Survival Endpoints in Randomized Clinical Trials. The speaker will Introduce a flexible, model-agnostic augmentation approach for time-to-event data, highlighting how ML techniques can drive significant efficiency gains in survival analysis and provide practical pathways for implementation in RCTs. In addition, discussants from FDA, academia and industry will share their insights on this important topic.