Analyses of repeatedly measured continuous outcomes in randomized controlled trials needed substantial improvements

Abstract

ObjectivesSystematic understanding is lacking regarding how current trials handle repeated measure data and the extent to which appropriate statistical methods are used for such data set. This study investigated the current practice of analyzing the repeated measure data among randomized controlled trials (RCTs). Study Design and SettingWe searched the Core Clinical Journals indexed in PubMed for RCTs published in 2019 and included a continuous primary outcome with repeated measures. We randomly sampled RCTs from the eligible trials. Team of methods trained investigators screened studies for eligibility and collected data using the pilot-tested, standardized questionnaires. We thoroughly documented statistical analyses of the continuous primary outcome with repeated measures and particularly recorded how statistically advanced methods were used to handle these repeated measures. ResultsIn total, 200 trials were included. Of these trials, the mean number of repeated measures for the continuous primary outcome was 5.46 (SD = 3.4); 58 (29.0%) trials did not specify the time point of primary outcome in the method; 113 (56.5%) trials did not use statistically advanced methods for handling repeated measure data in the primary analyses. Among187 trials included the baseline values, 88 (47.1%) trials did not adjust for outcome value at baseline. Among 87 trials using statistically advanced methods, 49 (56.3%) did not specify correlation structure for model. ConclusionsThe statistical analyses of repeatedly measured continuous outcomes in RCTs need substantial improvements. Careful planning of the primary outcome and the use of statistically advanced methods for analyzing data are warranted.