Friday, February 13, 2009

Ninety-four Percent of Industry Cancer Trials Not Published?

Thanks to my friends Rick Bukata and Jerry Hoffman and their "Primary Care Medical Abstracts" for calling my attention to this paper.

Ramsey and Scoggins of the Fred Hutchinson Cancer Center in Seattle asked a simple question--if you look for clinical trials of cancer treatments (almost all of which are drug therapies) on the major trials registry, ClinicalTrials.gov, and then look to see if they were published in a journal indexed by PubMed, how many show up? They did a download of eligible trials from the registry in September, 2007 and checked out PubMed in November of the same year.

One thing they found was that 21% of trials registered three years or more previously had been published while only 11.9% of more recent trials had been. So if they had waited longer, they might well have found a higher number of trials published.

The next thing they found was that overall, only 17.6% of trials had been published. So even if waiting longer had caused this number to double, that's still an awful lot of data that is not being captured by the published literature.

The point I want to focus on is the breakdown by type of sponsor. While no one was terrific given the 17.6 % overall publication percentage, the worst statistics were for industry sponsored trials, with only 5.9% being published. Moreover, 64.5% of all trials that were published were reported to have been positive, suggesting as one would expect a substantial publication bias toward positive studies. Of the industry sponsored studies, fully 75% that were published were positive.

Conclusion: An industry sponsored trial that is negative is far more likely never to be published than a negative study of another sponsor's; and the total number of studies conducted by industry that remain unpublished in the cancer area is staggering.

Of course, if a study turns out to be flawed, or never enrolls enough subjects, or fails for many other possible reasons, then there are justifications for why it may never be published. The problem is imagining that these bad things happen selectively to industry-funded trials--especially when the industry, as I documented in HOOKED and on the blog, devotes considerable funds and efforts to asssuring that these unfortunate things do not happen to their trials.

The authors are concerned more with the way that oncology is forced to depend on a very biased and incomplete literature, than they are with industry-sponsored studies specifically. And that concern is obviously a highly legitimate concern. Nonetheless this study provides substantial statistical backup for a phenomenon that we had feared all along was occurring--systematic suppression of negative studies by industry.

Ramsey S, Scoggins J. Commentary: practicing on the tip of an information iceberg? Evidence of underpublication of registered clinical trials in oncology. The Oncologist 13:925-929, 2008.

1 comment:

Marilyn Mann said...

Perhaps you have already posted on this, but the same phenomenon has been found in trials of antidepressants.

Turner et al., Selective Publication of Antidepressant Trials and Its Influence on Apparent Efficacy, NEJM (2008);358:252-260

http://content.nejm.org/cgi/content/abstract/358/3/252

Also of interest:

Sanjay J. Mathew and Dennis S. Charney, Commentary: Publication Bias and the Efficacy of Antidepressants, Am J Psychiatry 166:140-145, February 2009