Methodological review showed that time-to-event outcomes are often inadequately handled in cluster randomized trials.

OBJECTIVES

To estimate the prevalence of time-to-event (TTE) outcomes in cluster randomized trials (CRTs) and to examine their statistical management.

STUDY DESIGN AND SETTING

We searched PubMed to identify primary reports of CRTs published in six major general medical journals (2013-2018). Nature of outcomes and, for TTE outcomes, statistical methods for sample size, analysis, and measures of intracluster correlation were extracted.

RESULTS

A TTE analysis was used in 17% of the CRTs (32/184) either as a primary or secondary outcome analysis, or in a sensitivity analysis. Among the five CRTs with a TTE primary outcome, two accounted for both intracluster correlation and the TTE nature of the outcome in sample size calculation; one reported a measure of intracluster correlation in the analysis. Among the 32 CRTs with a least one TTE analysis, 44% (14/32) accounted for clustering in all TTE analyses. We identified 12 additional CRTs in which there was at least one outcome not analyzed as TTE for which a TTE analysis might have been preferred.

CONCLUSION

TTE outcomes are not uncommon in CRTs but appropriate statistical methods are infrequently used. Our results suggest that further methodological development and explicit recommendations for TTE outcomes in CRTs are needed.