A cross-sectional study assessing visual abstracts of randomized trials revealed inadequate reporting and high prevalence of spin

Abstract

Objectives

Visual abstracts (VAs) lack study-specific reporting guidelines and are increasingly used as stand-alone sources in medical research dissemination although not designed for this purpose. Therefore, our objectives were to describe 1) completeness of reporting in VAs and corresponding written abstracts (WAs) of randomized controlled trials (RCTs), and 2) the extent and type of spin (ie, any reporting pattern that could distort result interpretation and mislead readers) in VAs and WAs of RCTs with a statistically nonsignificant primary outcome.

Study Design and Setting

We conducted a cross-sectional study evaluating VAs and WAs of RCTs published between January 1, 2021, and March 3, 2023. We searched MEDLINE via PubMed for reports of RCTs published in the 15 highest impact factor journals from six medical fields (among which 34 journals producing VAs of RCTs were identified). One reviewer identified primary reports of RCTs published with a VA and randomly selected a maximum of 10 reports from each journal to avoid overrepresentation. The completeness of reporting assessment was based on the Consolidated Standards of Reporting Trials extension for abstracts. Spin was explored using a standardized spin classification for RCTs with statistically nonsignificant primary outcome results. Both assessments were conducted in duplicate, with discussion until consensus in case of discrepancy.

Results

A random sample of 253 reports from 34 journals was identified. The information provided in VAs was frequently incomplete: primary outcome identification, primary outcome results, and harms were respectively described or displayed in only 47% (n = 116/247), 30% (n = 75/247), and 35% (n = 88/253). Reporting was slightly better for some items in WAs, although still unsatisfactory. Among trials with nonsignificant primary outcome results (n = 101), 57% (n = 58) of the VAs and 55% (n = 56) of the WAs exhibited at least 1 type of spin. Posthoc analyses showed VAs produced by journal editors of high-impact general medical journals were more complete and more accurate than those produced by specialty journals or authors.

Conclusion

The information conveyed in VAs was frequently incomplete and inaccurate, highlighting the urgent need to refer to appropriate specific reporting guidelines to avoid misinterpretation by readers.