Skip to Main Content

Yale CMIPS Faculty Member Dr. Fan Li Receives PCORI Grant to Develop New Methods for Planning Cluster Randomized Trials

March 16, 2022
by Yazhini Ramesh

We as humans differ from one another in our backgrounds, genetics, and health conditions. For instance, most of us are aware that no two people are the same based on our genetic makeup and lived experiences. Yet clinical trials are often not designed to powerfully analyze how various individual differences like age, health history, and socioeconomic status impact the effect of specific interventions. Fan Li, PhD, Assistant Professor of Biostatistics and faculty member at the Center for Methods in Implementation and Prevention Science (CMIPS) at the Yale School of Public Health, has recently received an award from the Patient-Centered Outcomes Research Institute (PCORI) to develop new methods for planning cluster randomized trials that will incorporate such factors into relevant research and interventions.

Typically, individually randomized trials focus on interventions for a more homogenous pool of individuals. In contrast, pragmatic cluster randomized trials are conducted within diverse, real-world populations to help with clinical decision-making. Currently, methods are limited for designing cluster randomized trials to examine factors that affect participants differently across groups. “It’s not that researchers have not been interested in examining these factors,” explained Dr. Li. “It is just that the statistical methods in the field have not been really catching up, both from a design and analysis perspective.”

Through his research, Dr. Li aims to develop guidance for planning pragmatic cluster randomized trials to study the heterogeneous effect of treatments on pre-specified subgroups. The technical term, heterogeneity-of-treatment-effect analysis, helps determine “what works best, and for whom.” Even though heterogeneity-of-treatment-effect analysis is already recognized as a goal in individually randomized trials, little guidance is available for heterogeneity-of-treatment-effect analysis in pragmatic cluster randomized trials. “The question would be,” explained Dr. Li, “how can we plan these pragmatic trials so that we could have a better chance of detecting not only an average treatment effect but also a heterogeneous treatment effect between or across individuals that can be possibly explained by their baseline attributes, such as demographics, health conditions, and many other factors? The primary goal is to facilitate the design of these trials to ensure a more credible confirmatory heterogeneity-of-treatment-effect analysis.”

To achieve this goal, Dr. Li will propose new statistical methods and create open-source, user-friendly software to help researchers ensure sufficient statistical power for heterogeneity-of-treatment-effect analysis. In doing so, Dr. Li hopes that the “richness of data and the nature of pragmatic cluster randomized trial designs will motivate us to look into another generation of effect evidence, which is not only just for the population average, but more so for clinically meaningful differences across subgroups that can be useful for future health care decision making.”

As a member of the NIH Pragmatic Clinical Trials Collaboratory, Dr. Li will also work closely with its leaders and stakeholders to provide technical guidance on heterogeneity-of-treatment-effect analysis for cluster-randomized trials. Ultimately, these tools will allow scientists to identify the specificity of treatment effects across individuals and advance biomedical research to develop unique solutions based on individual differences.

Along with the award discussed above, Dr. Li is the Yale Subcontract Principal Investigator for four additional PCORI Methods awards. Three of these awards address emerging complexities in the design and analysis of cluster randomized trials, intentional incompleteness in data collection due to patient-centered considerations, and patient-centered outcome truncation by death. The most recent award aims to develop robust machine learning methods with time-varying treatment effects for longitudinal observational studies. “By bringing a more nuanced perspective to clinical trials through such methods,” CMIPS Director Dr. Donna Spiegelman noted, “we will be able to better consider patient-centered outcomes and create targeted treatments for diverse individuals, which is a huge step for implementation research and medicine in general.”

Submitted by William Tootle on March 16, 2022