Estimates of under-reported clinical trial data are overblown, according to the US Food and Drug Administration and the National Institutes of Health. The agencies are responding to questions from Congress spurred by a recent British Medical Journal paper, which found that only 22 percent of trial sponsors in 2009 followed the law and reported trial conclusions within a year of trial completion, reported Nature.
The FDA argues that the BMJ study was flawed by including some trials completed before the 2007 law took effect, failing to exclude some exempted trials or trials of unapproved products, and including trials whose data had not yet been checked by the NIH. ClinicalTrials.gov, the public database the study’s authors drew on, also does not make clear if a trial’s deadline has been extended. Based on these data, and information not accessible to the public, the FDA is pursuing 15 cases of late reporting.
Andrew Prayle, one of the paper’s authors, argued that the different findings are based on the FDA-only data. “There remain two key questions,” Prayle told Nature. “First, can they account for all of the 78 percent of trials which had not reported? We felt that this was unlikely. And second, why is it so hard to establish if a trial falls under the regulations?”
Whether ClinicalTrials.gov should house data from trials on unapproved drugs and medical devices will be clarified later this year in new rules from the FDA and NIH.
For further discussion on clinical trial reporting, including how much of the data should be made available, see this month’s feature, “Data Diving.”
Correction (May 14, 2012): This story has been updated from its original version to correctly reflect that the FDA and the NIH do not deny that problems exist with the reporting of clinical trial results, but merely argue that the issue may have been overstated. Furthermore, while the FDA is currently following up on 15 instances of late reporting, the agency never stated that these were the only 15 cases of late reporting. The Scientist regrets the error.