Signal Over Sprawl: Adaptive Randomization To Focus Early Oncology Development

By Matthew Confeld, PharmD, PhD, Director, Clinical Research Methodology

Early-phase oncology programs increasingly target molecules with potential efficacy across multiple indications, but the conventional "all-comers" Phase I approach often works against this promise—broad, unfocused enrollment produces heterogeneous datasets that dilute efficacy signals, complicate indication selection, and run counter to FDA Project Optimus expectations for thoughtful dose optimization.

This infographic outlines a more disciplined path forward: using structured portfolio review to prioritize the 3-5 most promising indications, then applying Bayesian Adaptive Randomization (BARD) to sharpen dosing decisions. By combining Bayesian Optimized Interval methodology with backfilling and covariate-balanced randomization, trials can generate cleaner, more comparable data while minimizing the number of participants needed. For sponsors navigating indication selection and dose optimization under Project Optimus, this resource offers a practical framework for replacing scattershot Phase I designs with a data-driven, adaptive approach that preserves signal clarity, shortens trial duration, and protects patient resources throughout early oncology development.