Abstract
This study estimates the prevalence of conflict of interest disclosures in medical journal articles published in 2016 and associations between disclosures and median journal impact factor and article Altmetric scores.
Conflict of interest disclosures are an indicator of risk of bias in biomedical research and are associated with reporting of statistically significant results in primary studies. Although prevalence data are lacking, some suggest conflict of interest is widespread, unavoidable, and even desirable. Our aim was to estimate the prevalence of conflict of interest disclosure, extending beyond previous studies of specific populations, settings, or outcomes and to determine whether the presence of a published conflict of interest disclosure was associated with attention in the scientific literature and media.
Methods
We conducted a cross-sectional study of all articles published between January 1, 2016, and December 31, 2016, in journals conforming to the International Committee of Medical Journal Editors (ICMJE) policies and indexed in PubMed. Articles were randomly sampled to reach a prespecified sample size of 1000. We included primary research articles, commentaries, editorials, narrative reviews, systematic reviews, and meta-analyses without language restrictions.
Disclosures were classified as positive when at least 1 author reported a conflict of interest of any type, excluding current study funding or industry employment; negative if all authors stated they had no conflicts; and missing if there was no disclosure statement. Article focus was classified as drugs, devices or surgery, or neither. Journal impact factors were taken from the Journal Citation Reports 2016. Altmetric score was used to estimate media attention (mainstream media, social media, blogs, and other sources). All articles with a positive disclosure were examined by 2 investigators (Q.G. and A.G.D.).
The primary outcome was prevalence of positive disclosures with 95% CIs calculated using the Clopper-Pearson exact method. We used χ2 tests for differences in the prevalence of disclosed conflicts of interest by article type and focus, and Kruskal-Wallis tests for differences in the median journal impact factors and Altmetric scores across categories of articles. The threshold for significance was a 2-sided P value less than .05, adjusting for multiple comparisons using a Bonferroni correction. Analyses were conducted using SPSS version 24 (IBM) and Matlab version R2016a (MathWorks).
Results
After randomly sampling 1650 articles from 332 journals, 1002 articles from 269 journals were included. Overall, 22.9% (95% CI, 20.3%-25.6%) included a positive conflict of interest disclosure, 63.6% (95% CI, 60.5%-66.6%) a negative disclosure, and 13.6% (95% CI, 11.5%-15.9%) had no statement. Prevalence of conflict of interest disclosures differed by article type and focus (Table 1). Among commentaries, editorials, and narrative reviews, 31.4% (95% CI, 26.1%-37.1%) included a positive disclosure compared with 19.8% (95% CI, 16.9%-23.0%) in primary research articles (P < .001). Among primary research articles, 26.6% (95% CI, 21.4%-32.3%) of those focused on drugs, devices, or surgical procedures included a positive disclosure compared with 15.4% (95% CI, 12.1%-19.3%) of those that did not (P < .001). Articles with positive conflict of interest disclosures were published in journals with a higher median impact factor (6.0; interquartile range [IQR], 3.3-19.9) compared with other articles (2.7 [IQR, 1.4-5.0], P < .001), and received higher median Altmetric scores (3.7 [IQR, 0.5-36.4] vs 0.5 [IQR, 0.0-3.1], P < .001; Table 2). Patterns were similar by article type and focus. Although striking, the results for systematic reviews and meta-analyses were not significant, likely due to the small sample size (n = 30).
Table 1. Prevalence of Author Conflict of Interest Disclosures by Type of Article.
| Author Conflict of Interest Disclosure | ||||||
|---|---|---|---|---|---|---|
| Yes | No | Missing | ||||
| No./Total | % (95% CI)a | No./Total | % (95% CI)a | No./Total | % (95% CI)a | |
| All articles (N = 1002) | 229 | 22.9 (20.3-25.6) | 637 | 63.6 (60.5-66.6) | 136 | 13.6 (11.5-15.9) |
| Primary research articles (n = 682) | 135 | 19.8 (16.9-23.0) | 462 | 67.7 (64.1-71.2) | 85 | 12.5 (10.1-15.2) |
| Drug-focused | 39/124 | 31.5 (23.4-40.4) | 69/124 | 55.6 (46.5-64.6) | 16/124 | 12.9 (7.6-20.1) |
| Device-focused | 27/121 | 22.3 (15.2-30.8) | 75/121 | 62.0 (52.7-70.7) | 19/121 | 15.7 (9.7-23.4) |
| Both | 5/22 | 22.7 (7.8-45.4) | 16/22 | 72.7 (49.8-89.3) | 1/22 | 4.5 (0.1-22.8) |
| Neither | 64/415 | 15.4 (12.1-19.3) | 302/415 | 72.8 (68.3-76.8) | 49/415 | 11.8 (8.9-15.3) |
| Commentaries, editorials, and narrative reviews (n = 290) | 91 | 31.4 (26.1-37.1) | 150 | 51.7 (45.8-57.6) | 49 | 16.9 (12.8-21.7) |
| Systematic reviews and meta-analyses (n = 30) | 3 | 10.0 (2.1-26.5) | 25 | 83.3 (65.3-94.4) | 2 | 6.7 (0.8-22.1) |
The 95% CIs were calculated using the Clopper-Pearson exact method.
Table 2. Journal Impact Factor and Altmetric Scores by Article Type and Disclosed Conflict of Interest (COI)a.
| Article Type | Median (Interquartile Range) | |||
|---|---|---|---|---|
| Journal Impact Factor | Altmetric Score | |||
| Positive COI Disclosure | Negative COI Disclosure or No Statement | Positive COI Disclosure | Negative COI Disclosure or No Statement | |
| All articles (N = 1002) | 6.0 (3.3-19.9) | 2.7 (1.4-5.0) | 3.7 (0.5-36.4) | 0.5 (0.0-3.1) |
| Primary research articles (n = 682)b | 5.3 (2.9-13.5) | 2.4 (1.2-4.0) | 3.6 (0.5-19.0) | 0.5 (0.0-2.4) |
| Drug or device-focused (n = 267) | 4.9 (2.7-8.1) | 2.4 (1.3-4.0) | 3.0 (0.8-17.9) | 0.2 (0.0-1.2) |
| Any drug focus (n = 146) | 7.3 (3.6-19.9) | 2.5 (0.8-3.7) | 8.0 (1.1-46.6) | 0.4 (0.0-1.2) |
| Any device focus (n = 143) | 3.7 (1.1-6.1) | 1.6 (1.1-3.4) | 1.6 (0.5-10.3) | 0.2 (0.0-1.2) |
| Neither drug nor device (n = 415) | 5.7 (2.9-44.4) | 2.7 (1.5-4.7) | 4.4 (0.3-29.7) | 0.8 (0.0-3.3) |
| Commentaries, editorials, and narrative reviews (n = 290) | 7.7 (3.9-20.8) | 3.8 (1.9-7.5) | 3.7 (0.5-40.8) | 1.0 (0.0-8.4) |
| Systematic reviews and meta-analyses (n = 30) | 20.8 (9.3-38.5) | 3.2 (1.9-5.9) | 93.1 (25.0-234.5) | 3.0 (0.0-16.7) |
P < .001 for all comparisons except among systematic reviews and meta-analyses for which the P values are .04 for median journal impact factor and P = .11 for median Altmetric score.
The subgroups of primary research are not mutually exclusive.
Discussion
Among articles expected to conform to ICMJE disclosure standards, 22.9% included a positive conflict of interest disclosure. This cross-sectional analysis of articles published in 2016 did not associate disclosures with bias, nor did it account for undisclosed conflicts of interest. These estimates confirm that disclosures are more concentrated in certain types of studies. The highest prevalence of a positive disclosure was among articles most likely to contain subjective opinions. Articles with a positive disclosure were published in journals with higher impact factors and received more media attention and, thus, are likely to receive more attention from both the research community and the public. This may distort the perception of how common conflicts of interest are and amplify the effect of their results.
To enhance transparency, journals should enforce requirements that manuscripts include both disclosures and funding statements. Adoption of a standardized nomenclature for the types of conflicts of interest would enable the sharing of author conflicts in a public registry and may provide a clearer analysis of their potential influence on research integrity.
Section Editor: Jody W. Zylke, MD, Deputy Editor.
References
- 1.Ahn R, Woodbridge A, Abraham A, et al. . Financial ties of principal investigators and randomized controlled trial outcomes: cross sectional study. BMJ. 2017;356:i6770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Friedman LS, Richter ED. Relationship between conflicts of interest and research results. J Gen Intern Med. 2004;19(1):51-56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Drazen JM. Revisiting the commercial-academic interface. N Engl J Med. 2015;372(19):1853-1854. [DOI] [PubMed] [Google Scholar]
- 4.Steinbrook R, Kassirer JP, Angell M. Justifying conflicts of interest in medical journals: a very bad idea. BMJ. 2015;350:h2942. [DOI] [PubMed] [Google Scholar]
- 5.Dunn AG, Coiera E, Mandl KD, Bourgeois FT. Conflict of interest disclosure in biomedical research: a review of current practices, biases, and the role of public registries in improving transparency. Res Integr Peer Rev. 2016;1(1):1-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bekelman JE, Li Y, Gross CP. Scope and impact of financial conflicts of interest in biomedical research: a systematic review. JAMA. 2003;289(4):454-465. [DOI] [PubMed] [Google Scholar]