A while ago I posted about a study by Mathieu et al. in JAMA that appeared to suggest that registering clinical trials was not having the hoped-for benefits on the transparency of journal publications of trial results: http://brodyhooked.blogspot.com/2009/09/registering-clinical-trials-less-than.html. Another paper published shortly after that one by Robert Ewart and colleagues from Southern Illinois U.--
--adds to our level of concern that trial registries are not having the impact we had sought.
Ewart and friends looked at 158 reports of randomized clinical trials published in high-impact journals in 2008. They were able to analyze 110, including only trials that had been registered. Comparing the trial as described in the registry entry with the final report in the journal, they discovered that 31% of the time, a primary outcome had been changed, and in 70%, a secondary outcome had been changed, without these being mentioned in the publication.
Unlike the earlier paper, Ewart did look at trial sponsorship, but they found no difference related to industry vs. non-commercial sponsorship in whether an outcome had been altered.
Ewart and colleagues suggest that at least some of this problem may result from investigators' lack of familiarity with the registry process, but agree that a serious problem with the interpretation and integrity of trial results arises when changes are made but not explained in the methodology section.
I go back to the same point I made in the previous post--where are the journal editors or reviewers? What is the appropriate division of labor here? Is the reviewer supposed to be the one to check the trial registry, or is the journal editor? Do the instructions to reviewers make clear which is the case? Information is being made readily available now on the internet that is designed to allow us to have greater reassurance that what we read about a clinical trial is kosher--yet people responsible for the journals and their content are ignoring that information. It's ridiculous to say that each journal reader is responsible for going to the web to look up the trial on the registry. If a manuscript is submitted to a medical journal, and something has been altered without explanation from when the trial was first registered, then the editors should demand an explanation or else refuse to publish the paper. What is going to change, and when?