Advertisement

Discrepancies between published and registered primary outcomes in analgesic trials: ACTTION systematic review

      The National Institutes of Health released the trial registry ClinicalTrials.gov to the public in 2000 to increase public reporting and transparency of clinical trials. By 2005, the International Committee of Medical Journal Editors required trial registration for publication. The systematic review we conducted examined whether registered primary outcomes (POs) in analgesic treatment trials correspond with published POs. Trials with publications (n=95) were selected from the Repository of Registered Clinical Trials (RReACT) database, which includes all postherpetic neuralgia, painful diabetic peripheral neuropathy, and fibromyalgia clinical trials registered at ClinicalTrials.gov as of 1/1/2012. POs were identical in 23% of trials; there were discrepancies between registered and published POs in 55% (22% failed to register or publish POs). Common discrepancies were inconsistencies between, or failure to report, the timing of PO assessment in the registry and publication (42%), POs that were more vaguely described in the registry than the publication (38%), or instances where a registered PO was “demoted” to a secondary outcome or omitted from the publication without explanation (16%). When POs were identical, the PO was statistically significant in 81% of the trials. However, when registered and published POs differed, the ultimately published PO was frequently statistically significant (65%), whereas when registered and published POs differed and the publication “demoted” the registered PO, only 1 registered PO was reported as statistically significant, suggesting bias toward publishing statistically significant POs. At best, PO discrepancies may be attributable to carelessness (e.g., failing to report PO assessment timing) or to difficulty uploading registry information. At worst, discrepancies could indicate investigator impropriety (e.g., registering imprecise POs [“pain”], then publishing whichever pain assessment produced statistically significant results). Improvements in registry reporting are needed, as well as greater attention by investigators, journal editors, and reviewers to differences between registered and published POs.