BMJ Study: In the Age of Transparency, COI Disclosures Still Lacking
The famous (or infamous) NEJM meta-analysis of rosiglitazone (Avandia; GSK) studies by Nissen and Wolski—who concluded that the drug increases the risk of MI—initiated a flood of articles in the medical literature, most of which were opinion pieces.* As logic would dictate, the authors of these articles either 1) favored the continued use of Avandia (denying that the drug increases MI risk); 2) recommended discontinuation of Avandia (stating that the drug increases MI risk); or 3) were neutral on the subject (stating that there is insufficient evidence to make a recommendation).
Given these widely differing opinions, physicians at the Mayo Clinic recently assessed these response articles to determine whether the opinions offered might be associated with the authors' relationships to pharma generally and the manufacturer of Avandia specifically. Their findings were reported yesterday in the BMJ.
What is most surprising in the Mayo assessment is the relative lack of disclosure among the 202 articles discovered. (Note: The authors excluded the many follow-up articles written by Nissen.) Only 108 (53%) provided a conflict of interest (COI) statement, despite the fact that all were published within the last 3 years (ie, during the burgeoning Age of Transparency). While 9 of 10 original research articles provided a COI statement, only 43% of the 91 letters, editorials, or commentaries and 59% of the 101 reviews, meta-analyses, or guidelines offered COI statements.
Consequently the Mayo physicians were obliged to search the Interweb for the remaining authors' potential COIs. They then determined that 90 (45%) articles were written by authors with COIs, but that only 69 (77%) of these articles provided a COI statement. In 3 articles, the authors declared no COI, but a web search revealed otherwise.
Among the 180 unique authors that were counted, the largest percentage (44%) were neutral on the subject of whether Avandia increases the risk of MI; 17% claimed no increased MI risk, and 36% concluded that Avandia increases the risk of MI. The COI breakdown is tabulated below.
|
COI |
No MI Risk |
Neutral |
MI Risk |
|
General, % |
94 |
38 |
28 |
|
Avandia, % |
87 |
24 |
20 |
|
Actos, % |
65 |
30 |
22 |
The Mayo docs concluded that authors who came down on the side of no increased MI risk were significantly more likely to have relationships with pharma in general, the manufacturer of Avandia (GSK), and the manufacturer of Avandia's main competitor, pioglitazone (Actos; Takeda).
A similar pattern was discovered when considering whether authors recommended the continued use of Avandia.
|
COI |
Favor Use |
Neutral |
Disfavor Use |
|
General, % |
88 |
40 |
26 |
|
Avandia, % |
81 |
26 |
24 |
|
Actos, % |
73 |
26 |
26 |
It is important to note that the possible influence of a pharma relationship wasn't limited to opinions on Avandia. Of the 29 articles that "strongly recommended" using Actos over Avandia, 86% were written by individuals who had a COI with the manufacturer of Actos.
Although the BMJ study is intriguing (in its narrowly focused sort of way), a big problem with it, in my mind, is that the authors failed to acknowledge the distinction between the COIs of pharma employees and those of academics.*** The former is expected (and expected to be highly influential), while the latter is the primary concern when it comes to preserving the integrity of the medical literature and mitigating any undue influence that pharma might have on medical practice. Unfortunately the BMJ authors neglected to make the distinction and stratify their assessments accordingly.
* The interim analysis of Home et al, which concluded no increased risk of MI with rosiglitazone use, also instigated its share of published responses, albeit it to a lesser extent than the meta-analysis by Nissen and Wolski.
** A direct competitor of Avandia. Actos actually has a cardiopotective effect, according to a randomized, controlled trial by Dormandy et al (2005).
*** Another big problem: Implicit within the study is that a relationship with pharma is necessarily a negative influence, rather than an indicator of greater knowledge of a drug's activity and associated clinical data.
