Conflict of interest 2: Clinical practice guidelines and Deans

.

My last blog post was about conflict of interest (COI) in medicine; there has been some recent literature on the subject, so I am providing a second posting on the topic.

In “Conflicts of interest in cardiovascular clinical practice guidelines”^[1], in the March 28, 2011 issue of Archives of Internal Medicine, Mendelson and colleagues reviewed the 17 most recent guidelines regarding the diagnosis and management of heart disease promulgated by the American College of Cardiology (ACC) and the American Heart Association (AHA), the most important issuers of such guidelines. These clinical practice guidelines (CPGs) are important, as not only do physicians use them as authoritative sources to inform their patient care, but increasingly payers – and courts -- are looking at whether they were followed to determine payment and bonuses, or culpability in malpractice cases. While thorough review of the published (and unpublished) evidence is required (indeed is a sine qua non) in the issuance of guidelines, these panels of experts are convened to allow their collective experience and knowledge fill the gaps where there is no evidence. The authors point to the 2009 Institute of Medicine (IOM, a group of distinguished physicians and medical scientists convened by the National Academies, which are private citizens gathered with federal support to advise on science and technology) report from its Committee on conflict of interest in medical research, education and practice, which states that there is insufficient study of COIs in guideline reporting, as the fact that both ACC and AHA have “recently placed restrictions on official participation in educational events and guidelines production”.

Using the concept of an “episode” to mean that 1 person participated 1 time in production of 1 guideline, there were a total of 651 episodes by 498 people in the production of the 17 guidelines. Overall, 56% of individuals (277) had some COI, and (coincidentally) a COI was present in 56% (365) of the total episodes, with a range among the 17 guidelines of 13% (2 of 15) to 87% (13 of 15). Interestingly, actual members of the guideline committees, as opposed to those who just reviewed the guidelines (“peer reviewers”) were more likely (63%, a level reaching statistical significance for those who care, p=.006) to have COIs, as were the chairs of the committees (or first authors) compared to committee members, 81% (p=.03). Only 6 of the 17 guideline committees reported whether the COIs were “modest” or “significant”; the authors found that of those 54% were modest and 29% significant. Perhaps more important, according to the authors, was that it was only a few companies, presumably those with the greatest economic stake in the outcomes, who were involved in most of the COIs. They also note that this is “a particular cause for concern given the fact that many of the newest ACC/AHA guideline recommendations are based more on expert opinion than on clinical trial data”.

Dr. Steven Nissen, commenting on the article in the same issue (“Can we trust cardiovascular practice guidelines?”^[2]), calls the “depth and breadth of industry relationships reported in this article…extraordinary,” in that they go beyond scientific collaboration to include significant stock ownership and serving as promotional speakers, and states that “no conceivable logic can defend [this] practice.” He continues “Such relationships are so antithetical to the academic mission that many medical schools have now forbid such relationships for their faculty. To allow such individuals to write CPGs defies logic.” I imagine his use of the word “extraordinary” is in the sense of “amazing”; since this is the first such study, we do not know if such COIs are indeed “extraordinary” in the sense of “well out of the ordinary”, or if they are, indeed, quite typical. Indeed, Mendelson et al cite a 2002 study by Choudry et al[3] in which the found that of the 52% of the authors of 100 CPGs for a variety of adult diseases who responded to their survey, 87% had “some type of relationship with the pharmaceutical industry” (and we would not imagine that the 48% who didn’t respond had fewer!) Dr. Nissen, who is from the Cleveland Clinic, reveals his own potential COIs, including research support from a number of companies and honoraria for speaking which he has turned directly over to charity. As covered in the New York Times (Duff Wilson, “Study finds conflict among panels’ doctors”, March 28, 2011), Dr. Nissen calls for “banning most of these conflicts rather than just disclosing them”.

But sometimes COIs are not even disclosed. The same issues of Archives of Internal Medicine contains a “research letter” titled “Failure by deans of academic medical centers to disclose outside income”^[4]. The authors, Freshwater and Freshwater, looked at holdings of deans of allopathic and osteopathic medical schools on the Morningstar directory of US corporate directors and executives, and EDGAR, the Securities and Exchange Commission database. They found that while there were only a few deans who served as directors of public companies involved in health care (9 of 161, with one serving as a director of 2 companies and one of 4), this information was not always disclosed on the schools’ websites, nor was the compensation received. This compensation for each directorship ranged from $11,250 to $386,439 with a mean of $217,454, with one dean who served on multiple companies receiving $640,038. “One dean’s official Web page disclosed the 2 directorships, but it did not disclose the compensation. Two medical schools’ Web sites disclosed 3 directorships, with one dean holding 1 and the other 2 directorships; however the Web sites underreported the deans’ compensation by 39%, 56%, and 80% compared with the compensation calculated from the companies’ EDGAR forms.” These deans are the same people who are expected to enforce the rules against COIs that Nissen refers to above. In a commentary[5], Lo (who was the chair of the 2009 IOM committee) et al, say that disclosure is not enough, and that “Some relationships need to be managed or even prohibited”. They do not go so far as to suggest that all should be banned, although I would note that 152 of 161 deans seem to do OK without serving as corporate directors.

So should all such conflicts, whether deans serving on boards of directors, guideline authors having financial relationships such as major stock holdings or speakers’ fees, or receiving grants from corporations, be disclosed or prohibited? I think that there is clear consensus that disclosure is the de minimis requirement. As I have noted before, without disclosure there is no way for the consumer of information in a research study to decide if their might be bias; this is at least as important for are those writing CPGs or presiding over our academic medical centers. But, as Nissen states explicitly and Lo hints at, disclosure is not sufficient. To extend the metaphor of the judge referred to by Howard Brody in his article Professional Medical Organizations and Commercial Conflicts of Interest: Ethical Issues and cited by me in The AAFP, Coca-Cola, and Ethics: Serving the public interest? August 20, 2010, knowing that a judge in a case you are party to owns large amounts of stock in the company you are suing is not sufficient to make you feel alright about her presiding in the case.

Mendelson et al note that “It has also been argued that relationships with industry may also bring a breadth of perspective and experience, especially if individuals have relationships with multiple different companies. Theoretically, these individuals may be less conflicted than those with fewer industry affiliations…”. Right. Taking graft from many sources makes you less beholden to any one of them. Unfortunately, if you buy this argument, then their study should disappoint you, as only a much smaller percentage of the CPG authors had more than one COI.

Come on. Prominent scholars take money from industry for the same reason that others do. They like the money. Perhaps they are eminent enough that their heads are so swelled that they can believe both that they are being paid solely for their wisdom, and not for any bias they might bring to the deliberations of academic groups like those writing CPGs or in the administration of medical schools, and that, in any case, they are so distinguished that there is no way they would demonstrate such bias. If so, they are deluded. No one else believes it, most especially the companies giving them money. These COIs need to be completely banned if we are to be able to trust their academic integrity and intellectual honesty.

---

[1] Mendelson TB, et al, “Conflicts of interest in cardiovascular clinical practice guidelines”, Arch Int Med 28Mar2011;171(6):577-85. (only abstract available on line without subscription).

[2] Nissen SE, “Can we trust cardiovascular practice guidelines?” Arch Int Med 28Mar2011;171(6):584-5.

[3] Choudhry NK et al, “Relationships between authors of clinical practice guidelines and the pharmaceutical industry”, JAMA 2002;287(5):612-17.

[4] Freshwater DM and Freshwater MF, “Failure by deans of academic medical centers to disclose outside income”, Arch Int Med 28Mar2011;1717(6):586-7.

[5] Lo B, Kelch RP, Grady D, “Illuminating physicians’ financial relationship with industry”, Arch Int Med 28Mar2011;1717(6): 587-8.

Statins and scientific integrity

.
“Statins” are a class of drugs that lower lipid levels, especially low-density lipoproteins (LDL, “bad” cholesterol), and raise high-density lipoproteins (HDL, “good” cholesterol). Since we know elevated LDL levels are associated with higher risk for a number of vascular diseases, mainly coronary heart disease but also stroke, it made sense that these drugs would be beneficial for preventing recurrences in people who had prior heart attacks and strokes (“secondary prevention”) as well as preventing a first attack in those who had the risk factor of elevated LDL (“primary prevention”). Indeed, studies have shown the former; that is, that the use of statins reduced the risk of both a second attack and dying. However, most of the studies that have been done, and showed benefit, studied both groups, people who had already had a coronary event and those with elevated LDL who had not, and combined the results. The first big study that presumed to show benefit of these drugs for primary prevention was the JUPITER study (“Justification for the Use of Statins in Primary Prevention” – all these studies have cute, if often tortured, eponyms). It specifically looked at the drug rosuvastatin, marked as Crestor ® by its manufacturer, AstraZeneca, who also happened to fund the JUPITER trial.

Two articles that just appeared in the Archives of Internal Medicine (V 170, #2, June 28, 2010) raise serious questions about the use of statins for primary prevention. But they also raise serious questions about research ethics, particularly when trials are industry funded, and the strategic manipulation (not falsification) of the data that is presented to make drug therapy look better – this can mean billions of dollars in sales for the manufacturer. They also point out the danger of looking at intermediate, or “surrogate”, outcomes (in this case, lowered cholesterol and LDL) rather than the ones of real interest to patients, which are, essentially, death (mortality) and the quality of life (morbidity). I have discussed this previously (“Quality and Chronic Disease Management,” Feb 24, 2009, and “Physician conflict of interest” Dec 8, 2008; see also the article by GY Gandhi, et. al., “Patient-important outcomes in registered diabetes trials”, JAMA. 2008 Jun 4;299(21):2543-9.)

The first article, “Statins and all-cause mortality in high-risk primary prevention”, by KK Ray, et. al. (Arch Int Med, 28 Jun 2010; 170(12):1024-31) is a “meta-analysis”. This is a type of study that looks at a group of previously-done studies to try to determine if there is a consistent conclusion that is stronger because it includes more patients than any one study, or weaker because the various studies contradict each other. The authors’ goal was to tease out the results for primary prevention from studies that had looked at the use of statins for both primary and secondary prevention. There were 11 studies that met their criteria, which had a total of 65,229 (= “a lot”) of participants. Their Conclusion (verbatim from the abstract): “This literature-based meta-analysis did not find evidence for the benefit of statin therapy on all-cause mortality in a high-risk primary prevention set-up.” This is not to say that the statins did not lower LDL; they did, and significantly, but in the population of people that had not yet had a coronary event or stroke, they did not prevent mortality. It is important to note that they did also, to a small degree, decrease the risk of having a heart attack, although not of dying. The reduction in risk was about 1.5%; that is, if 200 people are treated for 5 years, 3 will not have heart attacks (NNT=67 over 5 years). The previous studies, by mixing up those who did have a prior event and did have lower mortality (secondary prevention), with those who had not, showed, on net, a benefit. This study demonstrates that there is no reduction in mortality from using statins as primary prevention; whether the small reduction in MIs and their associated morbidity and cost is worth the administration of statins in high-risk patients is at least questionable, and may be an issue for each patient to decide with their doctor.

What about the JUPITER trial? This is the subject of the second article, “Cholesterol lowering, cardiovascular diseases, and the rosuvastatin-JUPITER controversy”, by M deLorgeril et.al. (Arch Int Med 28 Jun 2010;170(12):1032-36. In a scathing “critical reappraisal” the authors note a slew of flaws in both the conduct and reporting of the JUPITER study. These include ending the study early, after only 2 years, at a time when it appeared that there was mortality benefit from the patients treated with rosuvastatin but when the curves were beginning to come together (i.e., maybe with more time the apparent benefit to rosuvastatin treatment would have disappeared and there would have turned out to be no difference in mortality rate), publishing reports in which the end of that curve was truncated, to not show the coming together, and publishing “subgroup” analyses that seemed to show benefit (by gender) but not where they didn’t (patients with diabetes). Most surprisingly, JUPITER had what seemed to be an extraordinarily low rate of fatality from myocardial infarction. The authors note that, to be able to get meaningful data to look at, they had to do calculations from the data presented by the JUPITER authors. For example, to know how many people died from heart attacks (myocardial infarction, MI) they had to subtract “nonfatal myocardial infarction” from “any myocardial infarction. This is not falsification, but it is very unusual for authors to present the data in a way that readers are required to make such calculations in order to get very important information. The rate of fatal to non-fatal MI was extremely low compared to studies from the World Health Organization, that show it to be 40-50%; in JUPITER it was 8.8% in the placebo group and 29% in the rosuvastatin group. First of all, it seems, if this data is correct, the fatal to non-fatal MI rate was 3 times as high in the group treated with rosuvastatin as in the group that was not treated (surely not the a point the authors and sponsoring company wanted to emphasize). Moreover, everyone in the study was “…unexpectedly – and inexplicably – highly resistant to acute ischemia and infarction.” The authors of this article suggest that there are many inconsistencies and implausabilities in the JUPITER data.

They then go on to discuss at length the roles of the sponsor (the drug company) and conflict of interest in reporting the data both by the sponsor and the principal investigator, who is co-holder of a patent for a test that is used to show “risk” for coronary artery disease (called “C-reactive protein, or CRP). They refer to other industry-sponsored flawed studies, including those about rofecoxib (Vioxx ®) and gabapentin (Neurontin ®), which I have discussed previously (“The ‘Neurontin Legacy’”, Jan 22, 2009). A superb editorial by Lee Green, “Cholesterol-lowering therapy for primary prevention: still much we don’t know” (Arch Int Med 28 Jun 2010; 170(12):1007-8) summarizes these issues.

Think about this. We have a bunch of studies that seem to show that statins are effective for prevention of cardiovascular events (heart attack and stroke) and prevent death. Turns out that they do this for those who have already had a heart attack, but for those who haven’t, the reduction in heart attack is small and there is no reduction in mortality. The use of statins in the much larger group, people with elevated cholesterol who have not yet had a heart attack, means big money for the drugs’ manufacturers. Studies that mixed the two groups blurred the distinction.

Then we have a big study (JUPITER) that purports to show that statins ARE effective for primary prevention, but that study is funded by the drug company and is seriously flawed. We know elevated cholesterol is associated with heart attack and we presume lowering cholesterol would help prevent those heart attacks, but, amazing and very important, when the study is actually done, the data doesn’t show it. (In studies looking at another such indicator, homocysteine, it was shown that, while elevated homocysteine levels are associated with heart attack, and folate – a cheap drug – lowers homocysteine levels, this did not decrease the rate of MIs. Too bad for the makers of the expensive homocysteine level test.)

I assume that I do not have to review the conflict of interest in the study being funded by the drug company and having the principal investigator a patent holder of a test that is highly used for assessing MI risk. The obvious concern is not that this was a potential conflict in the JUPITER study, but that it in fact led to selective interpretation and presentation of data (and the answers to how they had such a low rate of fatal MIs is still not in). This certainly challenges the claim we sometimes hear that just because there is a potential conflict of interest, it doesn’t mean that there is bad science being done. Of course, intrinsically it doesn’t, but this is one more example – with Vioxx and Neurontin and others – to show that it often does. And it also makes it harder for those who would argue that accepting industry gifts (whether lunch or fancy vacations) is benign. All these issues make a great example for teaching students, about conflict of interest, surrogate measures, scientific integrity, and how, from big issues to small ones, self interest colors our perceptions (see J Dana and G Lowenstein, “A Social Science Perspective on Gifts to Physicians From Industry”, JAMA. 2003;290:252-255).
.