• Jingchun Fan, Fujian Song, and Max O Bachmann.
• Epidemiology and Evidence Based Medicine, School of Public Health, Gansu University of Chinese Medicine, Lanzhou City, Gansu Province, P.R. China; Public Health and Health Services Research, Norwich Medical School, University of East Anglia, Norwich, Norfolk, UK.
• J Clin Epidemiol. 2019 Apr 1; 108: 17-25.

ObjectivesThe aim of the article was to assess the appropriateness and rationales of subgroup analyses planned in protocols of randomized controlled trials and reported in subsequent corresponding trial publications.Study Design And SettingWe searched PubMed to identify trial protocols published in journals during 2006-2017. From a total of 3,774 initially identified records, we included a random sample of 479 protocols and identified 280 trial publications corresponding to the included protocols.ResultsSubgroup analyses were specified in 19% of the protocols and reported in 21% of the trial publications. Of the 94 protocols with planned subgroup analyses, 32% mentioned testing for interaction, and only three considered statistical power. Subgroup analyses were not prespecified in 56% of the 59 trial publications with subgroup analyses. Subgroup analyses were stated as prespecified in nine trial publications, without support evidence from the corresponding protocols. Subgroup analyses were often reported insufficiently for assessing the consistency of subgroup effects across studies. Justifications for subgroup analyses were provided in only four trial protocols and seven trial publications.ConclusionInappropriate specification and reporting of subgroup analyses remain problematic in protocols and reports of randomized controlled trials. Justifications or rationales for subgroup analyses were only rarely provided in trial protocols and reports.Copyright © 2018 The Authors. Published by Elsevier Inc. All rights reserved.

10 keywords...

hide…