The effect of a conflict of interest disclosure form using closed questions on the number of positive conflicts of interest declared - a controlled study

Objective. While declarations of conflicts of interest (COI) have become an integral part of medical articles, COIs are often not declared completely and accurately. One of several possible reasons for deficient COI declarations is the lack of standardized and comprehensive COI forms. In 2010, the International Committee of Medical Journal Editors (ICMJE) introduced a COI form using clear definitions and closed questions. Deutsches Ärzteblatt (DA), the journal of the German Medical Association, adapted this form and implemented it in early 2011. However, it is unclear whether changing COI forms leads to more positive COI statements. Material and Methods. In a controlled pre-post design, positive COI statements were analyzed at three German medical journals: one had changed its COI form (DA), two had not: Deutsche Medizinische Wochenschrift (DMW) and Nervenarzt (Ner), both of whom used open questions in their forms. At the levels of both authors and articles, respectively, the proportion of positive COI declarations in orignal and review articles was recorded for volumes 2010 (before implementation of the new COI form at DA) and 2012 (after). The change in positive COI disclosures at the journals was compared. Chi-square tests were used to compare the figures by journal in 2010 versus 2012 and among DA, DMW, and Ner. Results and Discussion. While positive COI statements more than doubled at DA, there was no meaningful change in either of the control journals: In 2010, 19.1% [95% CI: 15.4-23.2] of all DA-authors submitted positive COI declarations, relative to 39.6% [35.0-44.5] in 2012, a factor of 2.1. At the level of articles, positive COI statements increased from 32.3% [23.7-42.0] to 70.1% [60.5-78.6] (factor 2.2). At DMW, positive declarations rose by a factor of 1.3 to 12.1% [9.7-14.8] in 2012 at author level and by a factor of 1.3 to 19.4 [14.2-25.7] for articles. At Ner, figures fell: to 19.9% for authors [16.9-23.4] and 30.7% for articles [24.0-38.1] (both by a factor of 0.8). P-values for the comparison of positive COI statements between 2010 and 2012 were low at DA (p for both author and article level comparisons 0.05). Although this is not a randomized controlled study, the findings support the hypothesis that the steep increase in positive COI statements at DA from 2010 to 2012 is the result of its new COI form: Relative to two journals that had not modified their COI forms the effect size o