Bayesian Methods in Solid Tumor Oncology: A Decision-Based Framework for Roles, Guardrails, and Responsible Use

Bayesian methods are increasingly used in solid tumor oncology trials to support dose finding and dose optimization, adaptive interim decision-making, subgroup estimation, information borrowing, and selected confirmatory applications. Their growing use reflects the complexity of modern oncology drug development, where trials often involve limited sample sizes, biomarker-defined populations, heterogeneous treatment effects, evolving standards of care, and consequential decisions under uncertainty. However, Bayesian methods may serve different roles across development settings, ranging from operational decision tools to supportive or primary analyese, and these roles are not always clearly distinguished in practice. This article proposes a decision-based framework for responsible use of Bayesian methods in solid tumor oncology. We organize Bayesian applications according to the development decisions they support and the evidentiary weight they carry, including dose finding and dose optimization decisions, interim go/no-go decisions, estimation in sparse or heterogeneous subpopulations, borrowing of external or nonconcurrent information, and confirmatory use. We further highlight the statistician’s responsibility to define the role of Bayesian analyses, align assumptions and decision criteria with the intended use, and communicate uncertainty in a way that supports credible decision-making. By approaching Bayesian methods as context-dependent tools, statisticians can support more disciplined, transparent, and fit-for-purpose use in oncology drug development.