A comparison of the published evidence for off-label uses of gabapentin with reports available to the manufacturer found significant discrepancies in outcome reporting and unpublished negative trials.

There is substantial evidence that selective and biased outcome reporting in published clinical trials exists – this is one of the reasons that clinical trial databases now available were set up. The authors of this paper report an analysis of the outcomes reported in published papers describing unlicensed (off-label) uses of gabapentin, using internal company documents made available as a result of US litigation. The off-label uses studied were migraine prophylaxis, treatment of bipolar disorders, neuropathic pain, and nociceptive pain, and the analysis determined the extent to which outcomes originally specified as the primary outcome actually appeared as such in a fully-published paper.

The authors obtained internal company documents describing 21 relevant clinical trials, 3 for migraine prophylaxis, 3 for bipolar disorder, 9 for neuropathic pain, and 6 for nociceptive pain.  Of these, 13 were published in some form: 10 as full-length papers, and one each as a letter to the editor, pooled as part of a non-systematic review, and as an abstract described as preliminary findings. One of the fully-published trials could not be included in the analysis as neither a study protocol nor an internal trial report was available in the company documents. The analysis thus included 20 trials of which 12 had been published in some form.

When the 8 unpublished trials were examined, in 6 the result for the primary outcome was not statistically significant, in one it was stated to be ‘positive’ but no P value was given, and in one the result was clearly positive with a P value <0.001.

Examination of the 12 published trials for which internal data were available found that a total of 21 primary outcomes had originally been specified for these trials. In only four was the originally specified primary outcome reported as such in the published version – one of these was published in abstract as ‘not significant’.  In the remaining 8 published trials, the reported primary outcome differed from that originally specified: of the original 21, 6 were not reported at all and 4 were reported as secondary. The discrepancies identified were: in 6 reports, a new primary outcome was introduced; in 2 there was no distinction between primary and secondary outcomes; in 2, the primary outcome was relegated to secondary; and in 5, one or more primary outcomes originally defined were not reported.

The authors conclude that they had identified selective outcome reporting in clinical trials of off-label uses for gabapentin. All the changes had the effect of increasing the drug’s apparent efficacy for these indications. They note that a change in outcome specification for a clinical trial may be valid, but that this should be done before the study finishes and should be reported in full with indications of why it was done and its effects on the statistical analysis. Overall, they conclude that practices such as those they identified threaten the validity of the evidence for efficacy of drugs in off-label indications, and comment that reviews of gabapentin in pain should be updated based on the new data.