JAMA Internal Medicine Bias – Now You See Me

Merlin the Magician would be proud of a recent study published in JAMA Internal Medicine attempting to prove that doctors who receive small gifts of food are more likely to prescribe a branded drug than generic drugs. The authors linked two national data sets in an attempt to quantity the association between industry payments and physician prescribing patterns.

The “researchers” utilized the most-prescribed brand-name drugs in four categories according to Medicare Part D information from 2013. They then cross-referenced those prescribing rates to the five months of data in the Open Payments database available for the year 2013. According to the authors, physicians who received only one meal promoting one of the four target drugs had higher rates of prescribing those medications.

The conclusion and relevance of the study in the abstract reads: Industry payments to physicians are associated with higher rates of prescribing brand-name statins. As the United States seeks to rein in the costs of prescription drugs and make them less expensive for patients, our findings are concerning.

However, once one digs a little deeper, it is clear that the investigation lauded by the press as evidence does not meet the scientific rigor that one would expect from a JAMA publication.

The Problems (Not an Exhaustive List)

First, the media reports all focused on the propensity of physicians to prescribe brand-name drugs when cheaper alternatives were available, not at all taking into consideration weather the patient may actually be better off on the branded drugs. In their main argument around statin use, many patients tolerate one statin and not another.

Second, when looking at the table of financial payments types and their association with brand name prescribing rates for statins, (the only table with actual numbers) food payments which was the source of many media stories did not meet statistical significance of .05 or 5%. The confidence intervals on meals actually skewed toward fewer prescriptions with a range of -2.56 to +.42. We recommend that the authors and the JAMA internal medicine editors review the Wikipedia definition of statistical significance, to help them understand that drawing inferences on data that does not meet the null hypothesis may not be in the best interest of their journal. But then again, pharmascolds hold to the believe to never let facts get in the way of a good narrative. Unfortunately, their allies in the media hold that same view.

Third, if one looks closely at Figures 1 and 2, out of almost 2,500 physicians analyzed, only a handful are creating the upward slopes between the covariates. The relationship is non-existent for physicians who receive less than $2,000, and for those who receive above $2,000, there are a few data points. Overall, the actual change listed at the end of their article was less than .1% increased prescribing per $1,000 increase in the open payments database.

Fourth, all the media mentions that the study looked at three therapeutic areas, lipids, hypertension and depression, yet their paper only discusses statins.  The editing and peer review process takes many months.  The authors would have had ample time to know what was actually being published.  So where is the “study showing even small meals influencing physicians.”

Fifth, we have made clear time and time again that the Open Payments data made available through the Affordable Care Act can be misused. Once again, we see opponents of marketing using the national registry as a stepping stone in their ultimate goal of eliminating any and all payments to physicians. As we have noted before, the data available through Open Payments has minimal pre-release vetting by the physicians, and can contain de-identified and disputed payments.

Sixth, some physicians may choose to attend industry events where information is provided about drugs they already prefer to prescribe. Such a choice would have little to no effect on their prescribing patterns. The authors of the study failed to take that into account.

The authors of the study recognized that there were several limitations, including equating five months of payment data to twelve months of prescription data. Comparing an entire year of prescribing patterns to five months of payment data should almost be considered malpractice: the five months of payment data is more than likely not representative of a full year.

Further, the study did not differentiate between new prescriptions and refills. By failing to take into account whether a prescription was a refill, the number of prescriptions written could very easily be inflated for each physician.

Media

The media absolutely missed it on this one headlines included:

Wall Street Journal

Free Meals Influence Doctors’ Drug Prescriptions, Study Suggests

“Doctors who received a single free meal from a drug company were more likely to prescribe the drug the company was promoting than doctors who received no such meals, according to a study.”

New York Times

Drug Company Lunches Have Big Payoffs

A free lunch may be all it takes to persuade a doctor to prescribe a brand-name drug instead of a cheaper generic, a new study suggests.

ProPublica

Feed Me, Pharma: More Evidence That Industry Meals Are Linked to Costlier Prescribing

A study published online Monday in JAMA Internal Medicine found significant evidence that doctors who received meals tied to specific drugs prescribed a higher proportion of those products than their peers. And the more meals they received, the greater share of those drugs they tended to prescribe relative to other medications in the same category.

To ProPublica, perhaps you should take that course on statistics as well as the editors.

Several publications took a second look:

Bloomberg actually did a good job laying out the alternative scenario in an article titled: Maybe Pharmaceutical Reps Actually Aren’t Bribing Doctors by Megan McArdle where she elegantly outlines how there are many more scenarios that the doctor being bribed as most the media coverage has focused on. In her analysis the physician may be doing this out learning new things about better drugs or perhaps he already writes the scripts.

Forbes in an editorial outlines how the choices of medications is important, and medical decisions are not made in a vacuum. The Way To Doctors’ Prescription Pads Is Through Their Stomachs So, once again, the pharmaceutical industry is under attack for coercing physicians to prescribe its drugs. This time the temptation involves food …

Industry Response

The Coalition for Healthcare Communications issued a statement in response to the “investigation,” stating,

Nothing new here.  Information is power.  The more information prescribers obtain, the more they are likely to use a drug safely and effectively. Dinner meetings are regulated by the FDA to ensure that the information is consistent with the basic information about the drug and is not false or misleading in any way. The fact that the doctors have dinner as part of the process does not change the facts of the presentation in any way. Education informs effective prescribing.

 The Pharmaceutical Research and Manufacturers of America also responded, noting that:

This study cherry-picks physician prescribing data for a subset of medicines to advance a false narrative. Manufacturers routinely engage with physicians to share drug safety and efficacy information, new indications for approved medicines and potential side effects of medicines. As the study says, the exchange of this critical information could impact physicians’ prescribing decisions in an effort to improve patient care.

Physicians’ prescribing patterns are dynamic and based on individual patients’ needs. According to a survey of physicians, 91 percent felt that a great deal of their prescribing was influenced by their clinical knowledge and experience. The survey also found that factors such as a patient’s particular situation, including drug interactions, side effects and contraindications; articles in peer-reviewed medical journals; and clinical practice guidelines, affected prescribing decisions a great deal.

Normally, an editor and peer reviewers would have had the authors pull back on many of the assumptions made by this article. One has to hope that this lack of checks and balances does not also apply to their clinical articles. Unfortunately, this study will be added to the lexicon and quoted by many opinion papers and journal editorials as fact, adding to the vicious cycle. We encourage the American Medical Association to do a full review on how this paper passed publication.