Bayesian Methods in Clinical Trials: Improving Accuracy and Reducing Bias in Medical Data

Abstract

Bayesian statistics has been playing a pivotal role in advancing medical science by permitting healthcare companies, regulators, patient groups or committees, and the broader society to quantify how well one can believe (probabilities) regarding the safety and efficacy of new treatments, interventions, or medical procedures. Unlike the classical framework, the Bayesian framework provides a coherent way to incorporate expert opinion or historical information when designing a new trial or analyzing data from a trial. Especially, the unique advantage of the Bayesian framework is underscored when one has an opportunity to sum up that body of prior information and incorporate it in a formal way into the design/analysis of a new trial with rich sets of quality external data.

In recent years, there has been a noticeable increase in regulatory submissions using the Bayesian approach for confirmatory clinical trials. The flexibility in adopting the Bayesian framework allows companies to obtain valuable results thereby facilitating regulatory decisions about a potential new treatment. Furthermore, the significant step anticipated to be taken by EMEA toward the full use of the Bayesian paradigm Broadly, the patient and his/her physician would choose a treatment subjectively on the basis of patient characteristics. For this reason the regulatory methodology should not be limited to the randomized clinical trial and should be less prescriptive on the design of the trial, permitting sponsors to design trials tailored specifically to the clinical question. On the other hand, the regulators’ concern – and the point of the discussion – is to demonstrate that the potential drug is effective and safe on a group basis. The sponsor’s approach must be scientifically valid, to ensure accurate decisions on the basis of often limited data and to avoid clinical and ethical errors.