Parallel Session #3a: Accelerated Evidence Generation: Innovative Clinical Trial Design & Execution
Statistical Methods
Friday, November 6 · 10:30 AM–12:00 PM
Session Chairs
Talks
Talk title to be announced
Rare disease drug development commonly faces challenges such as small patient populations, heterogeneous clinical manifestations, and slowly progressing disease courses, making crossover trial designs a potential alternative to traditional parallel group studies by offering improved statistical efficiency and reduced between-patient variability. This presentation investigates the use of 2-treatment, 2-period crossover designs in rare disease superiority trials using numerical analyses and case studies, with focus on statistical power and analysis methods. Our findings indicate that crossover designs can achieve comparable statistical power with substantially reduced sample sizes compared to parallel group designs. The presentation will also discuss practical considerations for implementation of crossover designs in rare disease trials, including handling of carryover effects and patient discontinuations.
When Prediction Meets Randomization: Regression, Machine Learning, or LLMs? Lessons from 125 Randomized Trials
Covariate adjustment is a standard tool for improving precision in randomized trials, but the value of increasingly flexible prediction methods remains unclear. This talk begins with an empirical comparison of regression and machine-learning-based adjustment across 50 completed trials. The results show that flexible machine learning methods are not automatically superior: simple regression adjustment with prognostic baseline covariates is often highly competitive. This finding motivates a new question: can large language models change the picture by extracting useful prognostic information from baseline covariates, trial descriptions, or other text-derived features? We present a unified framework for LLM-assisted covariate adjustment and empirical evidence from 125 randomized trials.
Investigation and Recommendation of Analysis Approaches for Adaptive Seamless Phase II/III Clinical Trials
Seamless Phase II/III clinical trials have emerged as a powerful strategy to accelerate drug development by integrating early-phase treatment selection and late-phase confirmatory evaluation within a unified protocol. Among the statistical methods for inference in such designs, closure principle-based approaches are widely adopted due to their strong control of the family-wise Type I error rate and broad regulatory acceptance. As a theoretically optimal alternative, the Uniformly Most Powerful Conditionally Unbiased (UMPCU) test explicitly accounts for data-dependent treatment selection and serves as a benchmark for evaluating the efficiency of other approaches. In this study, we conduct a comparison of different analysis procedures through extensive simulation studies, assessing statistical power, Type I error control, and practical implementation considerations, and provide concrete guidance for the design and analysis of adaptive seamless trials.
