Comparison of conference abstracts and full-text publications of randomized controlled trials presented at four consecutive World Congresses of Pain: Reporting quality and agreement of results

Abstract

Background: Conference abstracts are a potential source of new and relevant information about randomized controlled trials (RCTs). However, their dependability is questionable. The objectives of this study were to quantify the agreement between results of RCTs reported in abstracts presented at the four most recent World Congresses on Pain (WCP) and their corresponding full publications, and to analyse the completeness of reporting in those abstracts. Methods: To identify RCTs, we screened all abstracts presented at four WCPs from 2008 to 2014. Two independent authors identified corresponding full-text reports published through August 2016. Data about the main outcomes in each abstract and full publication were extracted, including the outcome domains and numerical results reported. We reported discordance between abstracts and full texts. We evaluated abstracts against the CONSORT for Abstracts checklist. Results: Approximately half of the 614 included abstracts had been fully published. Among the 306 abstract/publication pairs, eight pairs were not evaluable, and in the remaining 298 we found some form of discordance in 31% of the cases; the majority of discordances were quantitative, i.e. numerical results were different in the two locations, but with the same direction of effect. In the abstract–publication pairs where the abstract presented only preliminary/interim results, 79% had some form of discordance, mostly quantitative. Conclusions: The reporting quality of the 614 abstracts was suboptimal; the median adherence across all domains for all abstracts was 26%. In conclusion, conference abstracts of pain research are often not necessarily dependable information. Authors should be required to report abstracts according to reporting guidelines. Significance: Abstracts of RCTs addressing pain are not often dependable information sources; half of them are not published, their reporting quality is suboptimal. When published, 30% of abstracts-full text pairs have discordant results, with 78% discordance when abstracts present preliminary results.