PS52: From Statistical Innovation to Regulatory Implementation: Advanced Approaches for External Data Integration in Clinical Trials

The integration of external data sources in clinical trials has become increasingly critical for addressing challenges in drug development, particularly for rare diseases, pediatric populations, and innovative therapies. Traditional randomized controlled trials face mounting difficulties in achieving adequate power and enrollment, especially when studying conditions with limited patient populations or ethical constraints that preclude placebo controls. External data sources, including historical controls, natural history studies, patient registries, and other real-world evidence, offer promising solutions to enhance trial efficiency while maintaining scientific rigor. However, the successful implementation of external data integration requires sophisticated statistical methodologies to address potential biases, population heterogeneity, and data quality concerns, alongside clear regulatory frameworks that ensure appropriate evidentiary standards. This session presents a comprehensive view of current methodological advances and regulatory perspectives on external data utilization, progressing from foundational statistical approaches through adaptive implementation strategies to regulatory applications in specialized therapeutic areas. The first presentation examines propensity score weighting and power prior approaches for historical data borrowing in bivariate regression analysis. Multiple endpoints and historical data borrowing are simultaneously incorporated to enhance efficiency and speed up new drug development processes. The methodology employs a two-step approach: first using propensity scores as weights in bivariate regression likelihood to make borrowed populations comparable to target study populations, then applying dynamic borrowing approaches such as power priors based on observed treatment effect consistency. This approach enables borrowing of historical data from both the same endpoint and other endpoints to potentially increase analytical efficiency, with simulation studies and data examples demonstrating the method's performance and applications. The second presentation introduces a novel adaptive Bayesian borrowing (ABB) method featuring a prospectively specified strategy. The approach dynamically borrows information from historical controls when the current control mean falls within a predefined congruence range, and switches to a no-borrowing scenario when substantial divergence is detected. This process effectively mitigates risks arising from conflicts or misspecification between historical and current control data. Additionally, the ABB method supports adaptive sample size adjustments to accommodate uncertainty in endpoint variability, further enhancing study robustness. The last presentation focuses on regulatory considerations for external control arms in cell and gene therapy development. Cell and gene therapies present unique challenges for traditional randomized controlled trial designs due to ethical considerations, small patient populations, and the irreversible nature of many interventions. This presentation discusses FDA's current thinking on external control arm use in cell and gene therapy development programs, including acceptable data sources, methodological requirements, and evidentiary standards. Key topics include evaluation of natural history studies, patient registries, and real-world data as external controls; considerations for population matching and bias mitigation; and integration of external control evidence within regulatory decision-making frameworks. Case studies illustrate successful applications and common pitfalls in external control implementation for cell and gene therapies, while addressing ongoing initiatives to develop guidance and best practices for external control use in these rapidly evolving therapeutic areas.