D. V. Cherla1, B. Ibeche1, K. M. Mueck1, J. L. Holihan1, M. Moses1, J. R. Flores1, O. Olavarria1, C. Hannon1, T. C. Ko1, L. S. Kao1, M. K. Liang1 1University Of Texas Health Science Center At Houston,General Surgery,Houston, TX, USA
Introduction:
Conflicts of interest (COI) bias research by increasing the reporting of favorable results and until recently have relied on author self-disclosure. To increase transparency of COI reporting, the Centers for Medicare and Medicaid established the Open Payments Database (OPD) that reports payments between manufacturers, group purchasing organizations, and physicians and costs $180 million annually to maintain. Previously, we have shown that COI disclosed by authors as opposed to in OPD have a 30-50% discordance rate. We hypothesized that studies that fail to self-disclose all payments will express higher favorability towards their subject matter.
Methods:
PubMed was searched for medical and surgical articles published January 2014-June 2016 and published by US senior authors with an NPI number. Exclusion criteria included review articles, editorials, replies, and technical papers. COI were defined as payments received as honoraria or for research or consulting, compensation for serving as faculty or speaker, or company ownerships/partnerships. COI disclosed in the published manuscripts were compared to those reported in the OPD. Articles were divided into 3 categories: 1) full disclosure of COI 2) incomplete disclosure of COI and 3) no COI (Table 1). Independent, blinded reviewers read the abstracts and judged article favorability toward its subject matter. Favorability was analyzed as a categorical variable using Chi-square.
Results:
A total of 195 studies were selected for 240 rated topics. The number of subjects and percentage judged favorably were as follows per category: full disclosure of COI (58/64,90.6%), incomplete disclosure (96/123,78.0%), and no COI (37/57,64.9%). Unexpectedly, articles with full disclosure were more likely to show favorability towards their subject matters versus articles with incomplete disclosure (p=0.032). To subsequently adjust for self-perceived, self-reported “relevant COI,” we divided Incomplete Disclosure into Partial and No Disclosure and performed a subgroup analysis comparing Partial Disclosure to Full Disclosure and No Disclosure to No COI. On subgroup analysis, there were no differences between full disclosure and partial disclosure (56/67,83.6%) groups (p=0.230) or no disclosure (40/56,71.4%) and no COI groups (p=0.457).
Conclusion:
This is the first study to examine the impact of non-self-disclosed COI and demonstrates that they do not significantly affect reporting of favorable results. This study cannot ascertain if this is due to lack of relevance of non-self-disclosed COI. Further studies are needed to validate these findings, to evaluate reasons behind incomplete disclosure of COI, and to develop evidence-based guidelines for disclosure.
