Innovative Statistical Approaches for Clinical Trials

The vast majority of applications of statistics in the design, monitoring, and analysis of clinical trials are based upon the classical, “frequentist” approach to statistical inference. An alternative that has seen increased use in medical research, particularly in the study of medical devices, is the Bayesian approach to statistical inference. While the frequentist collects data within a study to draw inferences regarding a true, but unknown, population, the Bayesian starts with an opinion regarding the population and modifies it based upon the data collected. In this module, we will describe the differences and similarities between the two approaches, and give examples of how Bayesian statistics may be useful in the design, monitoring, and analysis of clinical trials.

• Conditional probability and Bayes Theorem
• Bayesians versus Frequentists
• Prior and posterior distributions
• Eliciting prior distributions
• Bayesian applications to clinical trials, including safety/efficacy monitoring, and continual reassessment method (CRM)
• Examples