Challenges in Small Sample Size Cluster Trials for Medication Adherence: Insights From <scp>TAKE</scp> ‐ <scp>IT</scp> ‐ <scp>TOO</scp>

Cluster randomized controlled trials (CRTs) are valuable for interventions involving the clinical care team but often require larger sample sizes due to within-cluster correlation and between-cluster variability. The TAKE-IT-TOO pilot study aimed to estimate the intraclass correlation (ICC) necessary for designing a full-scale CRT to improve medication adherence in pediatric kidney transplant recipients. We examined different approaches to summarizing electronically measured adherence and assessed how these choices impact ICC estimation and sample size requirements. In TAKE-IT-TOO, seven centers were randomized to either an adherence-promoting intervention or a healthy-living intervention. Medication adherence was tracked using electronic pillboxes over a 4-week run-in period followed by a 10-week intervention. Variance estimates were calculated using generalized linear mixed models (GLMM) and used to derive ICC values. We compared two methods of summarizing adherence data and estimated the number of clusters and cluster sizes needed to detect an odds ratio of 1.5 between intervention groups across a range of plausible ICC values. The analysis demonstrated that using repeated adherence measures within each participant offered advantages over relying on a single summary measure, improving statistical power. However, large standard errors around variance estimates made precise ICC estimation difficult. Despite this, we identified a plausible range of ICC values and corresponding sample size requirements for future trials. The TAKE-IT-TOO study underscores the challenges of conducting CRTs in small populations and highlights the value of repeated outcome measures for maximizing statistical efficiency in adherence research.