PS56: From Data to Decisions: Causal Machine Learning for Drug Development and Regulatory Decision-Making

Machine learning is increasingly being used in pharmaceutical development, but many important decisions require an understanding of treatment effects rather than prediction alone. This session will highlight how flexible, principled modeling approaches can support clinical development by addressing population heterogeneity, complex longitudinal data, and multiple sources of clinical and patient-reported information. The presentations will illustrate applications across key stages of development, including interim decision-making when trial populations may evolve over time, estimation of treatment effects when outcomes are collected through a mixture of onsite and remote assessments, and dose optimization that incorporates both clinician- and patient-reported outcomes. Together, these examples demonstrate how machine learning can help translate diverse clinical trial data into more informed decisions about trial continuation, treatment benefit, and dose selection.