• Shane Killeen, Panos Sourallous, Iain A Hunter, John E Hartley, and Helen L O Grady.
• From the University of Hull, Castle Hill Hospital, Hull, United Kingdom.
• Ann. Surg. 2014 Jan 1; 259 (1): 193-6.

ObjectiveTo determine the proportion of registered trials published in the surgical literature, to compare, in registered trials, the primary outcomes registered with those published and to determine whether outcome-reporting bias favored significant primary outcomes.BackgroundTrial protocol registration before patient enrolment for randomized controlled trials (RCTs) is a perquisite for many journals in attempt to decrease publication and selective reporting bias. Analysis of the medical literature demonstrates poor registration rates with discrepancies between reported and registered primary outcomes. This has not been evaluated in contemporary surgical journals.MethodsRCTs were identified for 2009 and 2010 from 10 high-impact factor surgical journals. One investigator identified all RCTs and extracted primary and secondary outcomes, dates of commencement and completion of study, funding source, and trial registration number. Trial registers were searched using the trial registration number for primary and secondary outcomes, dates of commencement and completion of study, and date of registration. Trial registration rates and registration adequacy were recorded. Register and published primary outcomes were then compared.ResultsA total of 246 papers were analyzed, among which 86 (34.9%) were not registered and 52 (21%) were inadequately registered. Of the 108 adequately registered trials, 32 (29%) had a discrepancy between the published primary outcome and that registered in trial register. In the 24 published studies where it was possible to assess, the discrepancy favored a statistically significant primary outcome in 22 (91.7%) whereas in 2 (8.3%) the discrepancy produced a statistically insignificant result.ConclusionsLess than half of all RCTs published in general surgical journals were adequately registered, and approximately 30% had discrepancies in the registered and published primary outcome with 90% of those assessable favoring a statistically positive result.

8 keywords...

hide…