AI-driven dynamic grouping for adaptive clinical trials: Rethinking randomization in precision medicine.

Abstract

Integrating artificial intelligence into biomedical human subjects research is transforming traditional experimental paradigms. This perspective introduces the concept of "dynamic grouping," wherein artificial intelligence (AI) systems continuously reassign participants across experimental conditions based on real-time biomarker data and clinical response patterns. Unlike traditional biomedical research designs that rely on fixed treatment and control groups, dynamic grouping allows participant assignments to evolve throughout the study. We examine the ethical implications, methodological challenges, and research opportunities associated with this paradigm, particularly in clinical trials, precision medicine, and digital therapeutics. To support this analysis, we present three computational simulations that quantify its impact: (i) a heterogeneity simulation demonstrating how patient variability affects the advantage of dynamic grouping, (ii) a statistical power analysis showing potential sample size reductions in adaptive designs, and (iii) a clinical outcome distribution analysis highlighting how dynamic grouping reduces negative treatment outcomes and optimizes patient responses. Our findings suggest that dynamic grouping can improve treatment effectiveness, enhance resource allocation, and increase statistical efficiency, although it also raises new challenges for causal inference, informed consent, and regulatory oversight. As AI continues to reshape medical research, adapting ethical and methodological frameworks will be essential for its responsible implementation.