Bayesian Borrowing: Some Philosophical Thoughts on Statistical Inference скачать в хорошем качестве

Bayesian Borrowing: Some Philosophical Thoughts on Statistical Inference

We are pleased to announce an exciting seminar on clinical trial innovations as part of the ISBS Webinar Series. The seminar, titled "Bayesian Borrowing: Some Philosophical Thoughts on Statistical Inference" will be held on Nov 7, 2025, 11am-12pm EDT, featuring distinguished speaker Dr. Carl-Fredrik Burman from stats innovation team in AstraZeneca. Abstract: There are situations where it is not feasible to power a trial to show positive efficacy in e.g. a paediatric population, a subgroup or for an (ultra-)rare disease. This presentation explores two Bayesian models for borrowing. First, we examine Bayesian dynamic borrowing of historical control data to augment limited trial populations. Second, we consider models that link disjunct subpopulations, for example, borrowing across biomarker-defined strata. We see a rational for borrowing in some situations, but stress that the methods rely on strong assumptions that cannot be verified by pure statistical methods alone. Scientific and clinical judgment, along with critical discussions of model choices and robustness, is essential for responsible implementation. To assess the proper role of Bayesian borrowing in confirmatory trials, it is valuable to reconsider the underlying principles of statistical inference and regulatory decision-making. This talk will illustrate key concepts with examples and highlight practical and philosophical considerations for stakeholders.