JBJS: Conflicts of Interest Associated with Favorable Research Outcomes
BY LAUREN UZDIENSKI, MARCH 19, 2007
More than 40% of investigators presenting data at the 2001 and 2002 Annual Meetings of AAOS reported conflicts of interest, and a new study in this month's JBJS shows that those conflicts were associated with positive study findings. Though investigators have been required by AAOS to disclose conflicts since 1985, the correlation suggests that disclosure may not be sufficient for assuring unbiased data.
Among 519 abstracts, Okike et al found that 212, or 40.8%, disclosed conflicts of interest. Physicians who specialized in adult knee reconstruction, adult hip reconstruction or spine were most likely to report a statistically significant conflict. The authors coded the abstracts by the type of conflict: research/institutional support; miscellaneous funding; royalties; stock options; or consultant/employee status. Research/institutional funding was the most common conflict, disclosed by 161 investigators, or 31%; 18.9% of the investigators claimed consultant/employee status; 12.3% received royalties; 9.8% disclosed miscellaneous funding; and 4.8% disclosed stock options.
Overall, positive results (where the experimental treatment was favored over the standard of care) were reported in 84% of the abstracts. However, when abstracts disclosing a conflict of interest were considered independently, the rate of positive results was shown to vary not only between authors who did and did not have conflicts of interest, but by the type of conflict. There was not a significant difference in positive outcomes between investigators with a research/institutional funding conflict and investigators with no conflict. However, 100% of studies whose investigators received stock options reported positive results, and results from investigators reporting other types of conflicts also skewed toward positive results.
Okike points out that correlation does not equate causation. He adds that a common explanation for the association is that industry is more likely to fund studies that are likely to succeed and suppress negative studies. But Okike's study suggests these reasons may not entirely explain why results from investigators with disclosed conflicts skew positively. Okike evaluated conflicts by type, as mentioned above, and investigators with research/institutional funding conflicts did not report results different from those with no conflicts. If studies likely to succeed were favored and negative data were being suppressed, the impact would theoretically occur equally across all investigators with conflicts.
At the same time, the paper isn't a call to end industry funding. According to Okike, one-quarter of all biomedical researchers receive industry funding; 70% of drug trials are funded by industry. Industry support encourages innovation and allows for research that may otherwise not have happened. As industry involvement is key to technological developments, steps must be taken to prevent bias in the data, keep communication and collaboration among scientists open and avoid putting patients at risk witih faulty data. To this end, Okike recommends penalties for failure to disclose conflicts; acceptance of support only when it does not come with restrictions; mandatory contracts with industry that prevent the suppression of negative results; independent statistical review; pooling of funding derived from industry sources; and the restriction of presentation and publication by individuals with a direct financial conflict.
Some research centers are making an effort to limit conflicts of interest: Stanford, among other institutions, has a policy prohibiting investigators from accepting gifts from industry. We'll be watching to see if other organizations follow their lead.
