Methods for Adaptive Randomized Trials in Infectious Disease Research

Harvard University D/B/A Harvard School Of Public Health, Boston MA

Investigators

Abstract

Project Summary/Abstract In building a proactive response to future epidemics and pandemics, interim analyses can help investigators evaluate countermeasures efficiently even under unpredictable outbreak conditions. At a pre-planned interim analysis, a clinical trial can save time and resources by stopping early when overwhelming evidence exists in favor of or against an experimental intervention. When investigators can no longer collect further data for the foreseeable future due to waning transmission, an unplanned interim analysis can help determine whether a trial should stop early given the existing observations or pause until more infections emerge in a subsequent outbreak. At the same time, study power may be lost if investigators conduct interim analyses without modifying the trial to account for the early stopping opportunities. Statistical challenges also arise when conducting interim analyses in cluster randomized trials (CRTs), which are well-suited for infectious disease research because of their desirable statistical and logistical features. This proposal seeks to address these methodological gaps by developing meth- ods for (1) calculating the sample size of a CRT with pre-planned interim analyses and (2) maintaining statistical validity when conducting unplanned interim analyses, which will help bolster the public health community's ability to mount a timely response to epidemics and pandemics. Traditional interim analysis methods depend on the independence of additional data between analyses, which does not necessarily hold for CRTs because outcomes are correlated and recruitment can occur at both the cluster and individual levels. Aim 1 assesses when this independent increments property holds and proposes a Fisher information-based approach to determine the maximum and expected number of clusters or cluster size for a CRT with pre-planned interim analyses and continuous, binary, or event rate outcomes. Motivated by the need to extend a trial across multiple outbreaks to accommodate the unpredictable nature of an epidemic or pandemic, Aim 2 develops methods for adaptively redesigning an ongoing trial to incorporate unplanned interim analyses in randomized trials with binary or time-to-event outcomes while maintaining a desired Type I error rate and power. The proposed research will be conducted at the Harvard T.H. Chan School of Public Health and Harvard Med- ical School, two highly collaborative, innovative, and impactful research environments. Working with professors from the Department of Biostatistics, Department of Population Medicine, and Center for Biostatistics in AIDS Research, the trainee will learn how to conduct research that tackles pressing problems in public health and statistics. During this fellowship, the trainee will cultivate his statistical and epidemiological knowledge through additional coursework; develop his written and oral communication skills by publishing manuscripts in high-impact peer-reviewed journals and presenting at workshops and conferences; and grow his teaching skills by working as a teaching fellow and instructor to high school students interested in biostatistics.