Objective: To assess the proportion of registered surgical trials with results published in journals with high-impact factors; compare the primary outcomes specified in trial registries with those reported in the published papers; and determine whether primary outcome-reporting bias favored significant outcomes.
Background: Outcome-reporting bias, that is, the selective reporting of a subset of the original registered outcome measures based on their results, has not yet been studied specifically for surgical trials.
Methods: We searched PubMed for reports of surgical randomized controlled trials (RCTs) indexed between 2007 and 2012 in 10 general medical journals and 10 surgical journals with the highest impact factors. For each article included, we obtained the trial registration information using a standardized data extraction form.
Results: Of the 327 evaluated surgical trials, registration was lacking for 109 (33%) published papers. Two (2%) of these papers were published in general medical journals and 107 (98%) in surgical journals. Twelve (6%) of the trials were still recruiting patients, and 48 (22%) were registered after completion of the study. A total of 152 trials were registered before the end of the trial with the primary outcome clearly specified. Among these papers, 49% (75 of 152) showed some evidence of discrepancies between outcomes registered and the outcomes published, most often related to omitting or introducing a primary outcome. These discrepancies favored statistically significant results in 28% of the papers.
Conclusions: The quality of registration of surgical trials published in surgical journals was inferior to those published in general medical journals. Comparison of the primary outcomes of surgical RCTs registered with their subsequent publication indicated that selective outcome reporting is prevalent and appears to be higher than in general medical trials.