The article, Evidence b(i)ased medicine—selective reporting from studies sponsored by pharmaceutical industry: review of studies in new drug applications , compares the studies submitted to the Swedish regulatory agency for approval of 5 SSRI's, with the articles that were eventually published based on those studies.
They note 3 problematic areas.
1. Multiple Publication-individual studies yielded multiple publications, without acknowledging that they came from the same study. Also, subsets of data from multiple studies were pooled to create new publications.
2. Selective Publication-studies that made the drug look good were published more often than studies that showed no difference from placebo.
3. Selective Reporting-studies ignored the results of intention to treat analyses in favor of per protocol analyses, which make the drug look better.
In case you've forgotten, in an intention to treat analysis, subjects who dropped out of a study, for whatever reason, are analyzed as thought they had completed the study, with the last available result carried forward.
In a per protocol analysis, subjects who drop out are ignored in the final analysis. Since many subjects drop out due to adverse effects, or lack of effect, if you don't include those subjects, your results will look better (the subjects who tolerate the drug well and do better on it are more likely to remain in the trial).
Please check out the study and comment (I already wrote that, but it bears repeating).
On another, related note, Ann Wolpert, MIT's director of libraries since 1996, died last month. The reason this is relevant is because she was a big believer in open access, and beginning in 2000, developed DSpace,
"an open-source digital archive for research that has been adopted by more than 1,000 institutions worldwide. DSpace ensure(s) that there (is) a common, permanent platform for library material. Then, in 2009, she conceived the MIT Faculty Open Access Policy, where professors give the Institute permission to disseminate journal articles for open access through DSpace@MIT. In turn, those articles can be republished on open Web sites like the Public Library of Science or PLOS. It was the first policy of its kind in the United States, and has been imitated by other universities around the globe." (Jason Pontin, MIT Technology Review Editor)
“(Wolpert) believed in open access, but it went deeper than that...Her central insight was that in the age of the Internet, a great research library could serve not only as a window into scholarly output for given members of university and research communities, but also as a window for the world at large into the scholarly enterprise. That was a great and thrilling idea...” (Hal Abelson, Class of 1922 Professor of Computer Science and Engineering at MIT and founding director of both Creative Commons and the Free Software Foundation (source))
POLL, our free online journal club that uses open access articles, was created in the spirit of Wolpert's work. Please make use of it.