While many of us just recovered from the news that Elsevier was paid to produce fake Journals to promote pharmaceutical products, another news item has appeared about “conflicts of interests in scientific publications”
This news is based on a new journal article from researchers from the University of Michigan’s Comprehensive Cancer Center in Ann Arbor, published in an early online edition of Cancer 
As mentioned in my previous post about the Elsevier “Fake Journals”, pharma-sponsored trials rarely produce results that are unfavorable to the companies’ products [e.g. see 3 for an overview, and many papers of Lisa Bero]. Concerned by these findings, the main medical journals now require researchers to disclose their potential conflicts of interest (COI).
The present study  analyzes the frequency of self-reported conflicts of interest (COI), source of study funding, and (their relationship with) other characteristics in original clinical cancer research (thus no reviews or basic research) published in 8 medical Journals in 2006. The 8 journals are high-impact clinical journals, 5 are oncology journals (Journal of Clinical Oncology, the Journal of the National Cancer Institute, Lancet Oncology,Clinical Cancer Research, Cancer) and 3 are core general medical journals (New England Journal of Medicine,JAMA, the Journal of the American Medical Association, Lancet).
In these medical journals 1534 original oncology studies were found. Twenty-nine percent of the oncology articles reported COI: 17% declared industrial funding and the remaining 12% of the studies had authors who were an employee of industry at the time of publication, or were funded by industry.
The study was thoroughly done: 2 students independently coded the articles and 2 other coders, blinded for the initial coding, assessed all randomized trials (within those 1543 papers) for the outcomes. They graded the authors’ subjective interpretations as positive (in favor of the intervention), neutral, or negative (in favor of the control arm). Overall survival was assessed quantitatively.
The main results:
- Conflicts of interest varied by discipline (P<.001). Studies that had a corresponding author from a medical oncology department or division were most likely to have conflicts (45%), and studies from diagnostic radiology were least likely to have conflicts (4%)
- Likewise the cancer type mattered, especially with regard to likelihood of industrial funding (P = .001). Studies on the male reproductive system and lung cancers scored highest and studies on neurological cancers scored lowest as to the likelihood of funding. (however there is some contradiction because gynecologic departments have a high score and gynecologic cancers have a relatively low score, conf. figures 1 and 2)
- Continental origin was also an important variable (P<.001). COI were observed in 33% of the North American studies, 27% of the European studies, 5% of the Asian studies, and 40% of the studies from other locations.
- COI was most likely in articles with male first and senior authors (P<.001).
- Industry funded studies were more likely to focus on treatment (P<.001), and less on epidemiology, prevention, risk factors for incidence, screening, or diagnostic methods.
- The randomized trials (n=124) that assessed survival were more likely to report positive survival outcomes when a COI was present (P=.04). (see below)
The paper has received a lot of media attention, initiated by the press release of the University of Michigan Health System itself. The data however are less shocking then they may seem. The main finding is that “conflicts of interest characterize a substantial minority of the clinically oriented cancer research published in high-impact medical journals”. This and the characteristics of the papers with COI (see above) add to earlier papers that report on the occurrence of COI in published articles, including papers in the field of clinical oncology.”
Some outcomes are not very surprising, such as that pharmaceutical industries and funding will be most involved in intervention studies in medical oncology studies (not so much in radiology or diagnostics).
In itself, COI does not mean that the results cannot be trusted or that they are plain wrong. Credibility could be questioned if only positive results are published or if the results are represented more positive then they really are.
Indeed, Jagsi et al show that “randomized trials with a COI were more likely to report positive survival outcomes (P=.04)”. However, the likelihood that the author interpretation was positive or more positive than the objective effect on overall survival wasn’t influenced by COI. And differences in industrial funding didn’t influence any of the blinded outcomes assessed. Also in this study, the non-neutral findings are emphasized.
On the other hand, authors had to rely on the information given, i.e. not all conflicts of interests may have been reported. Another issue is that not all known COIs are disclosed to the public (i.e. medicalnewstoday)
The following conclusion of the lead author Reshma Jagsi seems most relevant:
“Given the frequency we observed for conflicts of interest and the fact that conflicts were associated with study outcomes, I would suggest that merely disclosing conflicts is probably not enough. It’s becoming increasingly clear that we need to look more at how we can disentangle cancer research from industries”
- Jagsi, R., Sheets, N., Jankovic, A., Motomura, A., Amarnath, S., & Ubel, P. (2009). Frequency, nature, effects, and correlates of conflicts of interest in published clinical cancer research Cancer DOI: 10.1002/cncr.24315
- University of Michigan Health System (2009, May 13). 29 Percent Of Cancer Studies Report Conflict Of Interest. ScienceDaily. Retrieved May 14, 2009, from http://www.sciencedaily.com /releases/2009/05/090511090846.htm
- Smith R. Medical Journals Are an Extension of the Marketing Arm of Pharmaceutical Companies. PLoS Med. 2005 May; 2(5): e138. Published online 2005 May 17. doi: 10.1371/journal.pmed.0020138.
You may also want to read:
Hattip: @sciencebase, Reinout Rietveld (via NRC-next)