A recent Cytel panel led by Vice President of Strategic Consulting Natalia Muhlemann evaluated the role that Bayesian methods played in development of a COVID-19 vaccine. The wide-ranging discussion covered the structure and utility of platform trials and the role of master protocols in infectious disease vaccines development, but also addressed the importance of adaptive Bayesian methods in the search for COVID-19 therapies.When people think of Bayesian methods, they often imagine a tactical design built to take advantage of acquired knowledge through the construction of informed priors. This notion of a Bayesian design might seem at odds with the challenges of the COVID-19 era, given that after the first spread of the pandemic, there is still no agreed upon disease model.
An ‘adaptive Bayesian design’ refers to a trial that utilizes Bayesian updating techniques to make decisions throughout the course of a trial. When working with small samples, no natural history of disease, and unreliable recruitment, an adaptive Bayesian design enables a trial that allows for the flexibility of interim looks. This encourages learning during a trial, and thereby making optimal use of all data for decision-making. Promising therapies are identified earlier, and patients are enrolled into trials that show these early results.
Given the amount of learning taking place during a COVID-19 trial, updates to dosages, enrichment strategies, trial arms, and so forth are a necessary part of research. The flexibility of adaptive Bayesian designs enable this to occur. Adaptive Bayesian designs are also better suited to handle the evolving standard of care in COVID-19 therapies.