Chair: Kaiyuan Hua, PhD (BeOne Medicines)
Co-Chair: Xiaofei Wang, PhD (Duke University)
Title: Empower Pharmaceutical Developments with Innovative Adaptive Clinical Trials
Abstract: Innovative adaptive clinical trial designs have drawn increasing interest from pharmaceutical companies, academics, and regulatory agencies with the potential to accelerate the development of new therapies and to provide more robust support for regulatory decision-making. Such designs offer the advantages of improved efficiency and robustness over traditional approaches by enabling selection of patient subpopulations, endpoints, and treatment arms and other adaptive modifications, such as patient enrichment and sample size re-estimation based on interim trial data. These adaptive features are particularly valuable in addressing the challenges of complex diseases and rapidly evolving therapeutic landscapes in the era of precision medicine. Regulatory agencies, such as the FDA and EMA, have reinforced the support for these innovative designs through recent guidance that promotes innovative adaptive and multi-stage designs that ensure flexibility, reproducibility, and patient safety. This session will feature leading academic and industry experts to discuss recent advancements and practical implementation of adaptive designs in clinical trials.
Speaker: Cyrus Mehta, PhD (Cytel)
Title: Flexible Adaptive Procedures for Testing Multiple Treatments, Endpoints or Populations in Confirmatory Clinical Trials
Abstract: The statistical methodology for the classical two-arm group sequential design has advanced vastly over the past three decades to incorporate, adaptive design changes, multiple treatments and multiple endpoints, while nevertheless preserving strong control of the family wise error rate. The graph-based approach to multiple testing is an intuitive method that enables a clinical trial study team to represent clearly, through a directed graph, its priorities for hierarchical testing of multiple hypotheses, and for propagating the available type-1 error from rejected or dropped hypotheses to hypotheses yet to be tested. Although originally developed for single stage non-adaptive designs, we show how it may be extended to two-stage designs that permit early identification of efficacious treatments, adaptive sample size re-estimation, dropping of hypotheses, and changes in the hierarchical testing strategy at the end of stage one. We will present the statistical methodology for controlling the family wise error rate in the presence of these adaptive changes and will generate the operating characteristics of different underlying scenarios and adaptive decision rules through a large simulation experiment.
Speaker: Kaiyuan Hua, PhD (BeOne Medicines)
Title: Error Rate Control for Group-Sequential Study Design After Adaptive Treatment or Population Selection
Abstract: We propose a strategy for managing the issue of multiplicity in clinical trials with adaptive selection followed by group-sequential testing. The approach employs a two-stage design and addresses trials with multiple hypotheses. The first stage adaptively selects a subset of hypotheses for further testing, while the second stage monitors the remaining hypotheses based on group-sequential procedures. We provide a rigorous framework for controlling the overall Type-1 error rate across both stages, utilizing group-sequential p-values and the closed testing principle to ensure statistical validity in the adaptive setting. Through a simulation study based on an oncology trial example, we demonstrate the effectiveness of the proposed method in controlling Type-1 error rate while maintaining sufficient power.
Speaker: Anastasia Ivanova, PhD (University of North Carolina at Chapel Hill)
Title: Optimal two-stage biomarker-stratified designs with enrichment
Abstract: We consider biomarker-stratified trials with a pre-defined targeted patient group, the biomarker-positive subgroup. Most of the biomarker-stratified trials test the treatment effect in the biomarker-positive subgroup and in the overall population. It can be desirable to evaluate the treatment effect in the biomarker negative subgroup as well. We discuss optimality criteria for these designs and describe optimal two-stage designs. We give as an example PrecISE, a biomarker-stratified clinical trial in patients with severe asthma, where the optimal two-stage design was used.
Speaker: Peijin Wang (Duke University School of Medicine)
Title: A Similarity-Based Bayesian Adaptive Design for Hybrid Controlled Trials
Abstract: As real-world data have become increasingly accessible, hybrid controlled trials incorporating external controls (ECs) have emerged as promising alternatives to traditional randomized controlled trials (RCTs), particularly in rare diseases where recruiting sufficient concurrent controls (CCs) is challenging. However, fixed hybrid designs rely on the assumption that ECs remain comparable to the trial population, and violations of this assumption can leave studies underpowered. To address this limitation, we propose a Bayesian adaptive hybrid trial design that dynamically updates the treatment–control allocation ratio during interim analyses using a model-free similarity metric. The design increases reliance on ECs when similarity is high and shifts toward a conventional RCT when similarity deteriorates. This generalizable framework mitigates the risk of underpowered studies and can be paired with any Bayesian dynamic borrowing method. Extensive simulation studies are conducted to evaluate the ethical and operating characteristics of the proposed design against existing hybrid approaches.
