Articles provided insufficient information to conduct an appropriate retrospective assessment of the pragmatic/explanatory features of medicine trials with the PRECIS-2 tool.

Abstract

To assess whether, in the retrospective assessment of the pragmatic/explanatory features of pragmatic randomized controlled trials (pRCTs), the nine PRECIS-2 domain scores using the information provided in articles were modified after using the information reported in other publicly available sources. This is a cross-sectional study of participant-level pRCTs published in July 2018 to December 2019 in the four highest-impact general medicine journals. The articles described the main results of pRCTs assessing medicines in one or more arms that were not in the pre-licensing phases. The information reported in trial full protocols, published protocols, and other publications, registries, and trial websites were assessed and scored, and compared with that previously obtained after reviewing the information reported in the articles. Out of 76 articles on pRCTs, 13 (17%) were included in the analysis. All were two-arm trials, assessing medicines only (n = 7), medicine vs device (n = 2), medicine vs surgery (n = 1), or medicine vs placebo (n = 3). Seven were open-label trials, and six had any type of masking. All except one had the full protocol available and/or published protocol; seven had other types of publication available. The assessment of the nine PRECIS-2 domains with the information reported in the 13 articles was changed in all trials after using the information included in other additional available sources. Between one (n = 1 article) and six (n = 2) domains were modified in each pRCT. The domains that most commonly changed were "organization" (n = 12), "recruitment" (n = 11), and "follow-up" (n = 8). "Primary outcome" and "primary analysis" were not modified in any trial. Eight percent of all domains could not be assessed due to inadequate or lack of information in seven articles; those were "recruitment" (n = 3), "organization" (n = 3), "setting" (n = 2), and "flexibility:adherence" (n = 1). Articles describing the trial main results are usually insufficient for the appropriate retrospective assessment of the pragmatic/explanatory features of a pRCT by authors not involved in the conduct of the trial. To address this issue, editors should require the submission of the original full protocol and final full protocol with the history of amendments to be published as supplementary material to the article.