If I had not gotten off onto the COI tangent in recent posts, the big news of the week would probably have been the publication in the New England Journal of the paper by Vedula et al. on selective outcome reporting for gabapentin.
The quick once-over is that in order to reveal how research sponsored by pharmaceutical companies sometimes undergoes selective publication or suppression of data, to favor marketing the company's drug, you need both the publication itself (the "after") and some snapshot of the actual research data in its more-closely-raw state (the "before"). (Or, to show that the data was entirely suppressed, you need the "before" that is not followed by any "after," despite enough time having elapsed to allow for publication.) One way to get the "before" is to get the trials submitted to the FDA as part of successful new drug applications. (See paper by Rising et al. mentioned below.) Another way is to get hold of confidential in-house company documents that are released as part of the discovery process in a lawsuit.
The latter method was used in this instance. Pfizer, after acquiring Parke-Davis, was stuck with lawsuits against the latter company around its marketing of gabapentin (Neurontin), especially its off-label uses for which legally the company was not supposed to market at all. The major of-label diagnoses for which gabapentin was commonly recommended include various pain syndromes and bipolar disorder; its major labelled use is as an anti-epilepsy drug. Vedula and colleagues (two of whom worked with plaintiff's lawyers in the relevant suits) obtained reports of company sponsored trials investigating the off-label uses and compared those internal reports with published results for those trials that were eventually published.
What they discovered will be no news to anyone here--more evidence of the routine manipulation and suppression of data to make the favored drug look better. The authors particularly looked at the primary outcomes reported in the eventual publication vs. in the original trial results. In no less than 2/3 of the published trials, there had been a change in the primary outcomes. Secondary outcomes that were statistically significant were substituted for primary outcomes that were not; or else new primary outcomes appeared out of nowhere in the process of publication. (Without going into a lot of detail, this basically is not kosher and increases greatly the likelihood that you'll report as a true scientific outcome something that actually occurred solely by chance.) The trials that were not published were primarily those that showed gabapentin in an unfavorable light.
Lest anyone think that this was a fluke due to the company's goal of off-label marketing, Vedula et al. cite the earlier paper by Rising et al., which used FDA documents to cover a much wider class of drugs, and showed very similar results.
Now for the warning note. Vedula et al. very appropriately balance their claims of undue commercial bias by admitting the presence of apparently non-commercial bias. Specifically, they cite the paper by Chan et al. looking at 48 trials funded by the Canadian Institutes of Health Research, and showed that in 40 percent of those trials there were differences between primary outcomes as reported in the actual trial and in the final publication. The bottom line seems to be that trial registries are urgently needed for both commercial and non-commercial research, even though the quantitative problems in commercial research appear to be more substantial.
Vedula SS, Bero L, Scherer RW, Dickersin K. Outcome reporting in industry-sponsored trials of gabapentin for off-label use. New England Journal of Medicine 361:1963-71, Nov. 12, 2009.
Rising K, Bacchetti P, Bero L. Reporting bias in drug trials submitted to the Food and Drug Administration: review of publication and presentation. PLoS Medicine 5(11):e217, 2008; http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.0050217
Chan A-W, Krleza-Jeric K, Schmid I, Altman DG. Outome reporting bias in randomized trials funded by the Canadian Institutes of Health Research. CMAJ 171: 735-40, 2004; http://www.cmaj.ca/cgi/content/full/171/7/735