From the Stritch School of Medicine, Loyola University, Chicago, Illinois.
Corresponding author: Linda Brubaker, MD, MS, Loyola University Medical Center, Department of Obstetrics and Gynecology, 2160 South First Avenue, Building 103, Room 104D, Chicago, IL; e-mail: LBrubaker@lumc.edu.
Financial Disclosure The authors did not report any potential conflicts of interest.
OBJECTIVES: To describe the frequency, patterns, and associations of Obstetrics & Gynecology author conflict of interest disclosures over a 2-year period.
METHODS: All original research articles published in Obstetrics & Gynecology between July 2008 and June 2010 were individually reviewed to record conflict of interest disclosures. For each article, we assessed the number of authors, the proportion of authors with disclosures, and the assigned level of evidence. For each author, we assessed the number of disclosures. For individuals who participated as authors on more than a single article during this time period and reported disclosure at least once, we also assessed the consistency of the individual's conflict of interest disclosures.
RESULTS: Nineteen percent of articles contained at least one disclosure with a range of one to 73 disclosure per article (median 3). Nearly one third (31%) of articles with disclosures were randomized clinical trials. Articles contributing level I evidence were somewhat more likely to have disclosures (34% compared with 16% and 17% for those reporting levels II and III evidence, P=.004). The median study duration was shorter for articles with disclosures (36 compared with 54 months, P=.003). There were 2,017 unique authors with a range one to 25 per article (median 5). Three hundred twenty-five (16%) were “repeat” authors. Disclosure inconsistencies were noted in 12% of these repeat authors (40/325).
CONCLUSION: Conflict of interest is disclosed in approximately 20% of original research articles, is more common in level I evidence papers, and appears inconsistent for 12% of repeat authors, suggesting an opportunity to improve disclosure practices.
LEVEL OF EVIDENCE: III
Scientific dissemination of new knowledge is critical to advances in health care. However, ongoing concerns over conflict of interest disclosures are warranted, especially given multiple, well-documented instances of underreporting. In 2009, the Institute of Medicine's Committee on Conflict of Interest in Medical Research, Education and Practice published their recommendations for determination and disclosure of conflicts of interest, including in research. The Institute of Medicine defines conflict of interest as a situation in which an individual or institution has a secondary interest that creates a risk of undue influence on decisions or actions affecting a primary interest (eg, the conduct of objective and trustworthy medical research).1
Disclosure of conflict of interest in research is required by an increasing number of medical journals, although there is no single uniform journal policy and many individuals are confused about which circumstances constitute a conflict of interest. The prevailing philosophy is that authors should be exceedingly cautious and report any conflict of interest. Organizations such as the World Association of Medical Editors and the International Committee of Medical Journal Editors, which facilitate global communication and cooperation among peer-reviewed journal editors for the purpose of improving editorial standards and promoting professionalism, have released statements regarding conflict of interest, which are intended to guide journals. The International Committee of Medical Journal Editors has proposed a standardized disclosure form for authors.2 This is an important, albeit imperfect, improvement in the conflict of interest disclosure process for authors.
Given its central role in a scientific dissemination of obstetrics and gynecology research, the aim of this study was to describe the frequency, patterns, and associations of Obstetrics & Gynecology author conflict of interest disclosures over a 2-year period.
MATERIALS AND METHODS
Every issue of Obstetrics & Gynecology between July 2008 and June 2010 was individually reviewed to record disclosure-specific information from original research articles only. The conflict of interest policy during the most of this study period follows: “Each author represents and warrants that he or she has no financial affiliation (eg, employment, direct payments, stock holdings, retainers, consultantships, patient-licensing arrangements, or honoraria) or involvement within the last 3 years with any commercial organization with potential financial interest in the subject or materials discussed in the manuscript except as disclosed in an attachment to this document. Any other potential conflict of interest also is disclosed.” For 4 months (January to April 2010), the policy was briefly changed to “Each author must submit a completed ICMJE Uniform Disclosure Form for Potential Conflicts of Interest. Each author represents and warrants that the information disclosed on the form is complete, true, and correct.”
A database was constructed to record relevant information for all publications. For each article, the title, issue, volume, pages, level of evidence (I, II, or III), duration, and type of study (eg, randomized clinical trial, prospective cohort) were recorded. In addition, every author's name and reported conflict of interest were documented. Conflicts of interest were recorded in the following categories: employment, grants, gifts, honoraria, patents, royalties, educational presentations, stocks, and expenses for travel and research presentations. An “other” category was used for those rare conflicts that did not fall into the categories just named. We also assessed the consistency of authors' conflict of interest disclosures for those who were on more than one article during the targeted time period. To estimate the duration of research activities before enrollment, we arbitrarily added 6 months to level II studies and 12 months to level I studies to accommodate the planning and start-up activities of the study.
Conflicts of interest were summed across all categories and the frequencies and percentages of all study variables were computed with SPSS 18. The χ2 test for association was used in examining the relationship between level of evidence and the reporting of conflicts of interest. In addition, the Mann-Whitney U test was used to compare the median study duration times of articles reporting and not reporting disclosures.
During the study period, there were 439 original research articles. There were 2,017 unique authors with a range one to 25 per article (median 5). Sixteen percent (n=325) were “repeat” authors with 209 authors participating in two articles and 116 participating in more than three articles. Most articles (n=356 [81%]) did not include disclosures. Eighty-three articles (19%) included at least one disclosure with a range of one to 73 disclosures per article (median 3). No article reported management of reported conflict of interest.
There were 245 articles that were classified as reporting level II evidence. Those reporting levels I and III evidence were classified in 64 and 117 articles, respectively. As can be seen in Table 1, articles reporting level I evidence were somewhat more likely to report disclosures (χ2=11.12, P=.004). Nearly one third (31%) of articles with disclosures were randomized clinical trials. The median study duration was shorter for articles with disclosures (36 compared with 54 months, z=2.95, P=.003).
Three hundred twenty-five individuals were authors more than once during the study period. Of these, 40 authors reported at least one disclosure at least once and 12% of these had inconsistent numbers of disclosures with a common pattern of no disclosure in at least one article and at least one disclosure on another article within the study period.
Our finding that approximately one in five original research articles during the study period included a conflict of interest disclosure establishes a baseline for disclosure activity within our specialty. Longitudinal evaluation of this proportion over time is likely to provide testable hypotheses regarding the frequency and associations of conflict of interest disclosures in our field over time. We suspect the higher proportion of disclosures in articles that supply level I evidence is related to industry support of randomized trials, although we did not assess funding relationships in this initial study. The effect of industry-supported randomized trials on journal impact factors and revenue has been evaluated in a cohort study by Lundh et al.3 Journals that publish industry-supported trials experience an increase in journal impact factors and ensuing sales can provide substantial income. Medical journals should consider journal-level policies to provide for protection of potential biases resulting from these factors as well.
We also found inconsistencies in conflict of interest disclosure for some repeat authors. This may be explained by the timing of the conflicted circumstances over the study period, but it is possible that this inconsistency may also be related to variability in interpretation of the scope of disclosure requirements. Certainly, there is inconsistency in the disclosure requirement language itself with terms such as “potential” conflict of interest erroneously suggesting that there is a difference between a potential conflict and real conflict for purposes of authorship. Despite efforts at clarify describing disclosure requirements, variability in interpretation is likely.
Because conflict of interest is simply a set of circumstances involving a risk, disclosure of this set of circumstances does not confer an accusation of improper motivation on the part of an individual. Although it is likely that authors embrace conflict of interest disclosure requirements with varying levels of enthusiasm, Obstetrics & Gynecology is consistent with most medical journals in their requirement to disclose conflict of interests. This policy is intended to provide readers with information to judge the risk of bias in reporting scientific results. Disclosure of individual and institutional financial relationships is recognized as a critical but limited first step in the process of identifying and responding to conflict of interest. After this essential first step of disclosure, evaluation must occur to determine the level of bias.
Readers are given dichotomous information for conflict of interest disclosures. Although the circumstance requiring the conflict of interest disclosure may be dichotomous, the level of risk relates to the chance that the scientific decisions made under these conflicted circumstances would be unduly influenced and the level of harm or wrong that could follow. Currently, the journal does not assess this level of risk nor convey a risk assessment to readers. Therefore, a reader may see similar disclosures from authors in a conflict of interest circumstance that holds little if any risk of bias and an author with who is in a major conflict of interest circumstance with significant risk to the scientific process. Local institutional management to minimize this risk may be present, but this is not typically conveyed to readers. Each medical journal, including Obstetrics & Gynecology, has opportunities for continuous quality improvement in supporting authors attempting to comply with disclosure requirements and intent. Lo et al discussed areas for improvements in disclosure policies in a recent commentary.4 We did not include meta-analyses in our study. Roseman et al have reported that there is an opportunity to improve disclosures associated with original reports of included randomized controlled trials.5 Further research will be needed to determine the appropriate mechanisms for conveying a risk assessment based on the disclosure related to an original research article.
Conflict of interest reporting is assumed to improve trust regarding the scientific process, yet the effect of disclosures is not known. Further research needs to be done to clarify the concordance of assessment of bias assessment by journal editors and readers as well as disclosure compliance and trends over time.
1. Lo B, Field MJ, editors. Conflict of interest in medical research education, and practice. Washington, DC: The National Academies Press; 2009.
2. Ferris LE, Fletcher RH. Conflict of interest in peer-reviewed medical journals: the World Association of Medical Editors' position on a challenging problem. Acad Med 2010;85:739–41.
3. Lundh A, Barbateskovic M, Hrobjartsson A, Gotzsche PC. Conflicts of interest at medical journals: the influence of industry-supported randomised trials on journal impact factors and revenue—cohort study. PLoS Med 2010;7:e1000354.
4. Lo B, Kelch RP, Grady D. Invited commentary—illuminating physicians' financial relationships with industry. Arch Intern Med 2011;171:587–9.
5. Roseman M, Milette K, Bero LA, Coyne JC, Lexchin J, Turner EH, et al.. Reporting of conflicts of interest in meta-analyses of trials of pharmacological treatments. JAMA 2011;305:1008–17.