Conflict of interest comes in two varieties. The first kind of COI has been treated extensively. COI type 1 is most relevant to medical literature as a scientific endeavor: science tends toward truth if its methodologies and data are transparent and free of bias. Various blog posts, most notably at the Incidental Economist group blog, have made the point that we are not there yet. Data is not fully open; methodologies are underspecified. We know that pharmaceutically funded studies are more likely to produce positive findings and demonstrate fewer harms (see this Cochrane review); but there are plenty of other biases affecting the scientific literature. and it makes sense to try and track them down for the sake of scientific rigor.
By COI type 2, on the other hand, I mean that which questions are found to be worthwhile, prestigious, fundable, popularizable in the lay press, and - in short - “important” tend to align with the drive to innovate, market, and sell new pharmaceutical products and medical devices. The “front-page” randomized controlled trial in the New England Journal of Medicine is often devoted to trials of statistical but dubious clinical significance. A recent instance was the IMPROVE-IT trial, showing that adding a non-statin to a statin slightly improves the rate of a composite cardiac outcome a number of years after starting the combination, compared to a statin alone, in patients who have had acute coronary syndrome.
By COI type 2, on the other hand, I mean that which questions are found to be worthwhile, prestigious, fundable, popularizable in the lay press, and - in short - “important” tend to align with the drive to innovate, market, and sell new pharmaceutical products and medical devices. The “front-page” randomized controlled trial in the New England Journal of Medicine is often devoted to trials of statistical but dubious clinical significance. A recent instance was the IMPROVE-IT trial, showing that adding a non-statin to a statin slightly improves the rate of a composite cardiac outcome a number of years after starting the combination, compared to a statin alone, in patients who have had acute coronary syndrome.
This trial follows contemporary methodological standards, with inclusion, exclusion, and outcome criteria exhaustively specified in an appendix, and appropriate statistical analysis of the composite outcome. But the wide-eyed enthusiasm of the accompanying editorial (“IMPROVE-IT is a landmark study in that it is the first clinical trial to show a benefit of adding a nonstatin lipid-modifying agent to statin therapy”) is misplaced, for the simple reason that the “benefit” is clinically negligible (2% difference in a composite outcome over 7 years of follow up), applicable to only the most carefully selected patients (scroll through the eligibility criteria) who have suffered acute coronary syndrome. The last paragraph is particularly egregious in its cheerleading: “These data help emphasize the primacy of LDL cholesterol lowering as a strategy to prevent coronary heart disease. “
Note the absence of any modifier on the “prevent.” Note too the last sentence, whose excited tone, to the regular reader of academic journals, is like a mariachi serenade at a harpsichord concert: “Perhaps the LDL hypothesis should now be considered the “LDL principle.”” (The “perhaps” is delicious, when the editorial title trumpets “proof.”)
One could go for the easy mark and adduce conflict of interest type 1 - there is pharma involvement at every level of this study. To be honest, however, the paper appears well done for what it is, and one is hard pressed to figure out how a 2% difference in a composite outcome would emerge from some hidden bias that is pharma-specific. (Sure, the inclusion and exclusion criteria could explain such a bias quite easily, but such details would be expected of any top-tier randomized controlled trial.)
Simply put, this study doesn’t matter to many patients. It does not add any new treatment; doesn’t provide discernibly better quality of care; and illuminates no new questions. It appears to be a study whose very existence is justified by the existence of a lipid-lowering agent that needs distribution and marketing.
Where are the NEJM randomized controlled trials on disparity-reducing interventions? The patient-relevant outcomes research and cost-effectiveness studies? Such studies do exist in the literature but typically in “second tier” journals that devote more space to social determinants of health; pride of place in the top tier is given to RCTs of the IMPROVE-IT type: not because pharma dictates journal content directly, but because the entire apparatus of biomedical literature is biased towards such studies. This is COI type 2. No amount of study-level transparency or open-data availability will fix it. Throwing out pharma sandwiches won’t work, and neither will calls to divorce moral reasoning from our scientific judgment because it’s “impeding our thinking.”
Since COI type 2 involves systematic bias, we aren’t going to get rid of it overnight (or maybe ever). The first step, however, is to recognize it for what it is. Perhaps I can make a resolution for myself and suggest it to others: when I read the NEJM, and wonder quizzically why the top RCT is thought of sufficient importance to devote marquee space in a top journal, to not accept it for what it is, but - on the contrary - to set about performing, advocating for, and critiquing the research, clinical care, and policy I would *like* to see in our top journals.
Policies to disclose and limit conflict of interest are widespread. Do we need them for COI type 2 as well?
Note the absence of any modifier on the “prevent.” Note too the last sentence, whose excited tone, to the regular reader of academic journals, is like a mariachi serenade at a harpsichord concert: “Perhaps the LDL hypothesis should now be considered the “LDL principle.”” (The “perhaps” is delicious, when the editorial title trumpets “proof.”)
One could go for the easy mark and adduce conflict of interest type 1 - there is pharma involvement at every level of this study. To be honest, however, the paper appears well done for what it is, and one is hard pressed to figure out how a 2% difference in a composite outcome would emerge from some hidden bias that is pharma-specific. (Sure, the inclusion and exclusion criteria could explain such a bias quite easily, but such details would be expected of any top-tier randomized controlled trial.)
Simply put, this study doesn’t matter to many patients. It does not add any new treatment; doesn’t provide discernibly better quality of care; and illuminates no new questions. It appears to be a study whose very existence is justified by the existence of a lipid-lowering agent that needs distribution and marketing.
Where are the NEJM randomized controlled trials on disparity-reducing interventions? The patient-relevant outcomes research and cost-effectiveness studies? Such studies do exist in the literature but typically in “second tier” journals that devote more space to social determinants of health; pride of place in the top tier is given to RCTs of the IMPROVE-IT type: not because pharma dictates journal content directly, but because the entire apparatus of biomedical literature is biased towards such studies. This is COI type 2. No amount of study-level transparency or open-data availability will fix it. Throwing out pharma sandwiches won’t work, and neither will calls to divorce moral reasoning from our scientific judgment because it’s “impeding our thinking.”
Since COI type 2 involves systematic bias, we aren’t going to get rid of it overnight (or maybe ever). The first step, however, is to recognize it for what it is. Perhaps I can make a resolution for myself and suggest it to others: when I read the NEJM, and wonder quizzically why the top RCT is thought of sufficient importance to devote marquee space in a top journal, to not accept it for what it is, but - on the contrary - to set about performing, advocating for, and critiquing the research, clinical care, and policy I would *like* to see in our top journals.
Policies to disclose and limit conflict of interest are widespread. Do we need them for COI type 2 as well?
RSS Feed
