Context: The most-read section of a research article is the abstract, and therefore it is especially important that the abstract be accurate.
Objective: To test the hypothesis that providing authors with specific instructions about abstract accuracy will result in improved accuracy.
Design: Randomized controlled trial of an educational intervention specifying 3 types of common defects in abstracts of articles that had been reviewed and were being returned to the authors with an invitation to revise.
Mean outcome measure: Proportion of abstracts containing 1 or more of the following defects: inconsistency in data between abstract and body of manuscript (text, tables, and figures), data or other information given in abstract but not in body, and/or conclusions not justified by information in the abstract.
Results: Of 250 manuscripts randomized, 13 were never revised and 34 were lost to follow-up, leaving a final comparison between 89 in the intervention group and 114 in the control group. Abstracts were defective in 25 (28%) and 30 (26%) cases, respectively (P=.78). Among 55 defective abstracts, 28 (51%) had inconsistencies, 16 (29%) contained data not present in the body, 8 (15%) had both types of defects, and 3 (5%) contained unjustified conclusions.
Conclusions: Defects in abstracts, particularly inconsistencies between abstract and body and the presentation of data in abstract but not in body, occur frequently. Specific instructions to authors who are revising their manuscripts are ineffective in lowering this rate. Journals should include in their editing processes specific and detailed attention to abstracts.