Do Bayesian Adaptive Trials Offer Advantages for CER?

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): We propose an innovative approach to designing and conducting randomized clinical trials (RCTs) for comparative effectiveness research (CER). Evidence from "pragmatic" RCTs can inform real-world decisions by comparing clinically relevant alternatives across a wide array of patient-centric outcomes in typical patient care settings serving diverse populations. Unfortunately, they typically increase the time and cost compared to the "explanatory" trials used to gain marketing approval. One example of a well designed and executed active comparator trial was the ALLHAT study, which cost over $130 million and took more than five years to complete. ALLHAT was very informative, yet raises the question of whether future trials of this nature could be designed more efficiently. Without alterations to how trials in the United States are conceived, designed, conducted, and analyzed, the nation risks spending large sums of money answering questions that are less impactful on clinical and policy decision making due to delayed reporting or incomplete comparisons. Bayesian adaptive RCT designs can reduce the sample size, time, and cost of obtaining decision-relevant information by formally incorporating external evidence and incorporating prospectively defined adaptations in order to focus the data collection on the primary aims. Despite successful use in regulatory trials, Bayesian adaptive methods have not yet been rigorously applied to pragmatic CER trials. We believe that exploring their value through a "proof of concept" will enhance scientific knowledge and help overcome the methodological inertia that tends to prevent the use of innovative approaches. Aim 1 is to demonstrate the applicability of Bayesian adaptive clinical trial design methods for CER by developing a series of 18 potential re-designs of the ALLHAT study. This aim includes a systematic review of evidence that was available prior to when ALLHAT was designed to inform the generation of "priors," which reflect probabilistic representations of what was known or hypothesized regarding antihypertensive treatments and includes designing adaptive features. Aim 2 is to evaluate the efficiency and performance of Bayesian adaptive designs through simulation using actual ALLHAT patient data. Aim 3 is to develop a clinically-motivated decision theoretic approach to Bayesian adaptive designs to address additional relevant research questions. The result will be a demonstration of the benefits - as well as some of the trade-offs - of a Bayesian adaptive approach to CER trial design and analysis. PUBLIC HEALTH RELEVANCE: CER is intended to help patients, prescribers, and other medical decision makers select the most appropriate treatments. Traditional clinical trials have strong internal validity, but may not be ideal for decision makers considering alternative treatments that were not "standard of care" when the clinical trial was designed. Bayesian adaptive clinical trial designs introduce efficiencies that retain internal validity while enabling active comparative trials to capture relevant and timely evidence to improve decision making that significantly impacts patients' health and safety.