Defining and estimating effects in cluster randomized trials: A methods comparison

Abstract

Across research disciplines, cluster randomized trials (CRTs) are commonly implemented to evaluate interventions delivered to groups of participants, such as communities and clinics. Despite advances in the design and analysis of CRTs, several challenges remain. First, there are many possible ways to specify the causal effect of interest (eg, at the individual-level or at the cluster-level). Second, the theoretical and practical performance of common methods for CRT analysis remain poorly understood. Here, we present a general framework to formally define an array of causal effects in terms of summary measures of counterfactual outcomes. Next, we provide a comprehensive overview of CRT estimators, including the t-test, generalized estimating equations (GEE), augmented-GEE, and targeted maximum likelihood estimation (TMLE). Using finite sample simulations, we illustrate the practical performance of these estimators for different causal effects and when, as commonly occurs, there are limited numbers of clusters of different sizes. Finally, our application to data from the Preterm Birth Initiative (PTBi) study demonstrates the real-world impact of varying cluster sizes and targeting effects at the cluster-level or at the individual-level. Specifically, the relative effect of the PTBi intervention was 0.81 at the cluster-level, corresponding to a 19% reduction in outcome incidence, and was 0.66 at the individual-level, corresponding to a 34% reduction in outcome risk. Given its flexibility to estimate a variety of user-specified effects and ability to adaptively adjust for covariates for precision gains while maintaining Type-I error control, we conclude TMLE is a promising tool for CRT analysis.

Estimating Causal Mediation Effects in Multiple-Mediator Analyses with Clustered Data

“…Note that with clustered data, researchers have distinguished between different populations that can be of substantive interest (e.g., Benitez et al, 2023;Miratrix et al, 2021;Raudenbush & Bloom, 2015;Thoemmes & West, 2011), including the population of individuals and the population of clusters; the choice between the populations is usually driven by substantive contexts (e.g., . In this paper, we focus on the former, and define the causal estimands to be average effects among the population of individuals.…”

Statistical Analysis Plan for the Primary and Selected Secondary Endpoints in the ENHANCED-SPS Study

“…If there are sufficient data to support a by arm comparison of endline viral loads (Section 2.1), we will conduct the following analysis. Using targeted minimum loss-based estimation (TMLE), [1][2][3][4][5] we will compare the proportion of participants with viral suppression after 12 months of follow-up between arms. TMLE accounts for clustering and allows for flexible adjustment for the ways in which participants with measured outcomes could differ from participants missing outcomes.…”

“…All analyses will also account for dependence of participants within clinics. Specifically, for statistical inference, we will aggregate the influence curve to the clinic level, 4,5 and use the Student's t-distribution with 14-2=12 degrees of freedom as finite sample approximation to the standard normal distribution. 7 We test the null hypothesis that the ENHANCED-SPS intervention did not improve viral suppression, as compared to the standard-of-care, with a one-sided test at the 5% significance level.…”