Evidence-based medicine attempts to rigorously apply the scientific method to clinically relevant decision making. This may be of particular importance in making decisions regarding pharmacotherapy. The evidence-based method requires that data be available for review and dissection in some detail.
When it comes to new drugs, industry-sponsored trials often represent the only available evidence regarding a particular agent. Complicating matters, the industry has come under fire repeatedly for altering the body of evidence for drugs in the ways noted above, in among other ways. In 2004, Pfizer agreed to plead guilty and pay a $430 million fine related to the off-label promotion of gabapentin (GBP).
Recently, Vedula et al studies documents which became available at trial in that case and in subsequent litigation to evaluate publication bias. They examined trials studying off-label indications. They identified 21 trials evaluating GBP for a variety of indications, including migraine (3 trials) and neuropathic pain (9 trials). Of these, 19 were randomized trials, 17 of which were parallel group and the other two crossover trials. Only 13 of the 21 trials were published, and only 10 of these, less than half, as full manuscripts. The authors eliminated one trial from analysis due to absence of available data, leaving 12 trials for the full analysis.
The results of the analysis were disconcerting. The outcome variables were often altered after the beginning of the trial. There were 21 primary outcomes identified in the 12 analyzed trials. Of these outcomes, nearly half were not properly reported; six were not reported at all, and another four were published as secondary, not primary outcomes. Further, 12 outcomes reported as primary outcomes were added after the trial began, and another five originally designated as secondary outcomes were not distinguished from primary outcomes in the final manuscript. The manipulation of secondary outcomes was even worse: in only 1 of the 12 published trials were all of the secondary outcomes defined in the protocol included in the final publication.
These changes in outcome variables affected the significance of the results. In five of the eight studies with changed primary outcome measures, statistically significant findings were reported despite the change. Four of these five were published as full-length articles. And all changes were more favorable regarding GBP than would have been the case with the initial study designs.
Thus, the authors conclude that the evidence provided by the sponsored trials for off label uses of GBP was altered significantly by publication changes. Fewer than half the studies were published as full length manuscripts and nearly half of the primary outcomes were not reported in a manner consistent with evidence-based medicine. All of the changes were ultimately favorable to the study drug. The authors make a strong statement about the implications of their findings: "We are concerned that the reporting practices observed in our analysis do not meet the ethical standards for clinical research or maintain the integrity of scientific knowledge."
In the interests of full disclosure, I note that I used to be on the Speakers' Bureau for both Parke-Davis and, later, Pfizer. Both bureaus operated in a most professional manner. And I am thus especially disappointed in the above findings. The pharmaceutical industry must do better.
Evidence-based medicine attempts to rigorously apply the scientific method to clinically relevant decision making. This may be of particular importance in making decisions regarding pharmacotherapy. The evidence-based method requires that data be available for review and dissection in some detail. The findings from a study which has endpoints changed mid-stream are potentially invalidated. And publication bias may present a treatment option in a more positive light than is justified by the evidence.