Improving the design and statistical analysis of cluster-randomized trials on tropical infectious diseases

PROJECT SUMMARY This Pathway to Independence Award application is submitted by a statistician committed to improving the design and analysis of tropical infectious disease cluster-randomized trials (CRTs). Worldwide, hundreds of CRTs are carried out annually to evaluate the effect of new interventions against infectious diseases, especially in tropical developing countries experiencing dengue, Ebola, malaria, and other infectious disease outbreaks. The scientific rigor of these CRTs relies on valid statistical analysis methods that adequately address the complexity in the CRT designs. However, the emergence of CRTs with complex and novel designs has outpaced the development of causal inference methods for data analysis. This gap represents a key barrier to providing valid sample size calculation, efficient estimation, and correct interpretation of the intervention effect estimates. The overarching goal of this research is to surmount this barrier by developing valid, robust, and efficient statistical methods. Specifically, the applicant will address the statistical challenges of three CRT designs: (1) covariate-adaptive randomization, which has been extensively used for reducing baseline imbalance, (2) the test-negative design, which has been increasingly popular in recent years for achieving cost-efficiency, and (3) the multi-arm stepped-wedge design, which has the potential to improve flexibility and efficiency for future CRTs. In the K99 phase, the applicant will extend the empirical process theory to handle covariate-adaptive randomization in CRTs and provide both theoretical and computation evaluations of current statistical models. During the first year of the R00 phase, the applicant will focus on test- negative designs in CRTs and eliminate the bias from differential healthcare-seeking behavior by characterizing the underlying causal graph and performing inference on self-nondiagnosable symptoms. Finally, the applicant will develop an optimal design that can simultaneously handle treatment roll-out, multiple interventions, and various outcome types. The applicant will accomplish the research aims under the mentorship of established researchers in infectious disease, statistics, and biostatistics to assure his transition to a tenure-track faculty position in the R00 phase and his emergence as a leading infectious disease biostatistician. At the University of Pennsylvania, the applicant enjoys rich internal resources of courses, seminars, computational equipment, collaborations, and intellectual interactions with prestigious researchers; furthermore, the applicant has access to external training opportunities including summer institutes, national conferences, and hands-on learning in trial conduct in Kenya. These training activities will propel the research career of the application, thereby supporting his achieving academic independence and ultimately leading a research team to advance the research of infectious diseases.