Missing data were poorly reported and handled in randomized controlled trials with repeatedly measured continuous outcomes: a cross-sectional survey

Abstract

Background and ObjectivesMissing data are common in randomized controlled trials (RCTs) involving repeatedly measured continuous outcomes. Evidence on the reporting and handling of such outcome data is lacking, which has prevented further improvement in methods and reporting of RCTs. MethodsWe searched PubMed to identify RCTs published in the Core Clinical Journals in 2019 that reported a continuous primary outcome with repeated measures. A team of investigators conducted a study screening and collected data using pilot-tested, standardized questionnaires from a random sample of eligible RCTs. We thoroughly collected information about the reporting of missing data for the repeatedly measured continuous outcome and the methods used to handle the missing data. ResultsWe included 200 eligible trials, whose mean number of repeated measures for the continuous primary outcomes was 5.46 (SD = 3.4). Sixty-one (30.5%) trials explicitly reported missing data at both participant and outcome levels, 116 (58.0%) at the participant level only, and 2 (1.0%) at the outcome level only. Sixty (30.0%) trials reported missing data at the participant level by group and by time point, and 53 (26.5%) at the outcome level by group and by time point. Among 179 trials having reported missing data, 162 (90.5%) did not assess the balance of baseline characteristics, 143 (79.9%) did not assume missing mechanism; 65 (36.3%) used suboptimal methods for handling missing data (e.g., complete case analysis); 41 (22.9%) conducted sensitivity analyses, and 5 (11.9%) assumed alternative missing mechanisms for sensitivity analyses. ConclusionThe reporting of missing data for repeatedly measured continuous outcomes were inadequate and the use of statistical methods for handling missing data was far from optimal. Substantial efforts are warranted to improve the reporting and statistical handling of these outcome data.