SC09: Trial Simulations in R: A Framework for Informing Modern Clinical Development

Clinical trial simulation, or simply trial simulation, is a computationally driven approach crucial for the design and analysis of clinical studies to predict outcomes, optimize protocols, and assess risks before actual patient enrollment. By modeling diverse trial scenarios - covering patient populations, treatment regimens, and clinical outcomes - trial simulations help evaluate potential trial success, identify sources of uncertainty, and inform decisions that improve both efficiency and cost-effectiveness. In an era of increasingly complex adaptive, Bayesian, and novel trial designs, simulation has become indispensable for determining operating characteristics where analytical derivation may not be feasible. Ultimately, simulation enhances the likelihood of obtaining meaningful results and supports regulatory interactions throughout drug development. As the power of trial simulations has been recognized within industry and organizations, the complexity of the trial designs, outcomes, and models has continued to grow in the trial simulations. Notably, the recent adaptive design draft guidance from the FDA emphasizes the critical role of simulations in evaluating type I error control, power, and other operating characteristics for complex innovative trial designs. Additionally, there has been a shift to generating patient-level data directly rather than summary statistics for trial simulations. This approach aligns more closely with real-world clinical processes, enables incorporation of key operational characteristics that influence trial performance, and facilitates patient-level adaptive designs such as response-adaptive randomization. It also allows statisticians to better evaluate model assumptions and visualize outcomes under varying data-generating mechanisms. This makes the results more realistic and relatable, which is particularly important in cross-disciplinary environments. Until recently, the lack of standardized software capable of handling the wide variety of statistical models and designs presented a major challenge. Many organizations depended on custom-built code that required specialized expertise and substantial validation and resulted in limited standardization across teams. This requires significant resources and hinders cross-industry collaboration. In this short course, we introduce a suite of R packages that provide a flexible and uniform framework for trial simulation aligned with regulatory expectations. These tools leverage functional programming principles in R to seamlessly integrate custom models, complex designs aligned with estimand frameworks, and computationally intensive simulation workflows required by FDA's Complex Innovative Trial Design program. In this evolving digital era, such standardized computational tools facilitate impactful transdisciplinary decision-making by enabling statisticians, clinicians, regulators, and data scientists to speak a common language when evaluating trial designs through reproducible, transparent simulations rather than fragmented custom code. In this short course, participants will: • Dive deep into the key components involved in trial simulations – objectives, scenarios, patient data generation models, trial design, and evaluation criteria. • Learn about a suite of R packages for simulation specializing in each of the trial simulation components with a heavy focus on flexible patient data generation and trial design. • Walk through a hands-on example based on a real industry project • Explore additional planned features and cross-industry collaborations. By the end of the course, participants will have a comprehensive understanding of modern trial simulation grounded in regulatory guidance, its applications across therapeutic areas and design types, and practical experience implementing flexible, reproducible simulations in R. The proposed framework empowers statisticians, sponsors, and regulatory reviewers to improve trial efficiency through transdisciplinary collaboration, reduce redundant programming efforts, and foster shared understanding.