A recent study in the Journal of the American Medical Association, entitled “Pharmaceutical Industry-Sponsored Meals and Physician Prescribing Patterns for Medicare Beneficiaries,” has received a ton of media attention lately, from the The Wall Street Journal to Bloomberg to NPR. But despite the hype, the study uses questionable methodology and draws biased conclusions.
The authors advance the perspective that the pharmaceutical industry is improperly influencing physician prescribing. The study authors do concede in their conclusions that “the findings represent an association, not a cause-and-effect relationship.” However, the authors make an inherent assumption that prescribing branded drugs is somehow inappropriate when cheaper alternatives (though not generic equivalents) are available. There is no evidence that this is the case; in fact, these drugs would have to have been covered by Medicare Part D plans during the analysis time period for the prescriptions to have been filled.
The relevance of the study must also be questioned. The four branded drugs evaluated in the study—Crestor, Bystolic, Benicar and Pristiq—were relatively lightly prescribed, representing 8.8%, 3.3%, 1.6% and 0.6% market share, respectively, of prescription volumes in their respective markets. It’s also worth noting that few physicians prescribing in these categories received meals related to these brands. Only 12%, 3%, 7% and 2% of prescribers in the four markets of interest, respectively, received meals. Crestor showed the most meal activity, but interestingly, it showed the lowest impact on prescribing, though this insight seemingly went unnoticed by the study authors.
Co-author R. Adams Dudley suggests that pharmaceutical companies would be better served to funnel meal money into R&D. In 2014, members of the Pharmaceutical Research Manufacturers of America spent an estimated $51.2 billion on R&D. Therefore, it’s unlikely that meal money is even a meaningful rounding error on this spend. At $20 per meal, these R&D expenditures could provide each of the 850,000 physicians in the U.S. with just over eight meals per day every day of the year!
The study authors also fail to discuss the relative impact of meals on prescribing versus other variables. For instance, cardiologists in the case of Crestor (a cardiovascular drug) are 1.45 times more likely to prescribe it than internists, while the impact of meals varies from 1.02 to 1.14 times. In the case of Pristiq (a depression drug), psychiatrists are 3.02 times more likely to prescribe it than internists, as compared to meals showing impact of 1.06 to 1.50. For all four drugs studied, a physician’s general tendency to prescribe brands across the board is more predictive of their likelihood to prescribe any of the four brands studied than meals they accepted. All of these data are available in eTable 3 of the study’s online supplement, but such comparisons are not made in the discussion section of the study.
If one takes the view that sales reps use meals as a way to gain access to prescribers in order to share information, it stands to reason that physicians who allow such access may be better informed on appropriate use cases for a particular drug and, thus, use it more often. In fact, in a 2012 study titled, "Can Access Limits of Sales Representatives to Physicians Affect Clinical Prescription Decisions," my colleagues found that physicians that significantly restricted access to sales representatives were four times slower than those allowing access to reduce prescribing of a drug for which a “black box” safety warning had been issued.
I hope this commentary provides a more balanced perspective on the study methodology and conclusions that can be drawn from it.