Presentation to the IOM Committee on Conflict of Interest in Medical Research, Education and Practice

November 5, 2007

Peter Lurie, M.D., M.P.H.

Download slides in PDF format

Thank you for the opportunity to address the Committee on this important topic.  I have been asked to step back from the subject a little bit and discuss the way Public Citizen conceptualizes conflict of interest and how we think it should be addressed.

I will start by providing a basic framework for thinking about conflict of interest.  Then I will discuss four examples of issues we have worked on.  Finally I’ll apply the framework to those same examples.

The framework begins with the potential conflict.  One has three general options to address this: legal restrictions, policy restrictions or disclosure.  An alternative, somewhat overlapping, formulation of this framework might be ban, manage or disclose conflicts of interest.

The advantage of legal restrictions is that they tend to be straightforward and can be very effective in that they can eliminate the conflict of interest entirely.   In some circumstances, they will be particularly cost-effective.  However, the modern trend in conflict of interest seems to lean heavily toward disclosure as a remedy, sometimes the only remedy.  While disclosure certainly has its place (and may be the only option in some instances), if used exclusively it can amount to an evasion of responsibility.  In effect, disclosure transfers responsibility to the consumer of the information, who is then expected to understand what it means for an author, committee member or clinician to have an equity interest of $50,000 in some pharmaceutical company. 

Thus, the question before this committee is: What is the correct mixture of the three remedies in particular circumstances?

I am therefore concerned that the charge to this Committee belies an already agreed-upon approach, which is to reduce the emphasis on legal restrictions as solutions to conflict of interest.  The second of the three charges to the Committee is the most important.  It is to “Propose principles to inform the design of policies, guidelines, and other tools to identify and manage conflicts of interest in [medical research, education, and practice] without damaging constructive collaboration with industry.”[1]  I challenge this committee to think in terms of all three of the options that I laid out, and not just the so-called “management” of conflict of interest, which this charge seems to push toward.

Disclosure of Financial Conflicts of Interest by FDA Advisory Committee Members

In this example, I’d like to emphasize the way we conceptualized conflict of interest and the analytic tools we applied to the problem. As in the next example, we tried to go beyond simply describing the prevalence of conflicts of interest, which is what most of the research in this area does, to investigate the relationship between the conflict and actual behavior.

In 1997, the Food and Drug Administration Modernization Act required more extensive public disclosure of conflicts of interest by Food and Drug Administration (FDA) advisory committee members.  In September 2001, we realized that the FDA had not complied with the law and threatened the FDA with a lawsuit. As a consequence, in January 2002, the FDA issued a draft guidance that required more detailed disclosures.  In March 2007, the FDA made available a new draft guidance that takes an approach that goes beyond mere disclosure.  It recuses (dismisses for the meeting with the conflict) members with total conflicts that exceed $50,000, and members with conflicts below $50,000 can attend but not vote.  Just a couple of months ago, Congress passed a bill that requires a 5 percent annual reduction in the prevalence of conflicts of interest in the advisory committees.

This brief history sheds light on the frequently heard claim that it is impossible to find adequate numbers of unconflicted experts to serve on advisory panels (or to act as journal reviewers).  Every time the FDA reconsidered its policy, this argument was raised and yet the policy has become successively stronger, in effect negating the argument when it was previously raised. As a Lancet editorial stated, “Defenders of FDA policy say that it is difficult to find experts free of conflicts of interest.  But it is hard to believe that in a country with 125 medical schools – not to mention the pool of international experts – the FDA cannot find experts who do not have financial ties with companies whose products are under review.”[2]

The study[3] I am going to describe was conducted between the time of the two FDA draft guidances.  It examined the disclosed conflict rate at FDA drug advisory committee meetings between 2001 and 2004.  The study showed that conflicts were common: at least one person had a conflict in 73 percent of meetings and, overall, 28 percent of members had a conflict.  Importantly, there was no difference in the rate of disclosure of conflicts before and after the 2002 draft guidance, so the enhanced disclosure requirement had no apparent impact upon conflict rates, merely upon the detail in which these conflicts were disclosed. The recusal rate was only 1 percent, so many members served and voted despite large conflicts.

In this study we also tried to assess the link between the predictor variable (the financial conflict of interest) and a concrete outcome (in this case, the votes of advisory committee members).  We looked at this question in six different ways, divided into two categories.  The last two analyses, the Mantel-Haenszel and the Monte Carlo, looked at this relationship from the perspective of the individual.  The underlying premise is that a particular individual, having received funding from industry, might go on to vote in some particular way as a consequence. The analyses of “index conflict” (those with the company producing the drug under discussion at the meeting) were limited by small sample sizes.  In the Mantel-Haenszel analysis, we did find positive relationships between voting in favor of the drug and having “any conflict” (with the index drug’s manufacturer or a competitor) and, somewhat paradoxically, with competitor conflict.  The Monte Carlo simulation found a relationship only with “competitor conflict.”

The first four analyses took a different approach: instead of looking at the individual perspective, they consider a group-level perspective.  In other words, “Did the conflicts have an impact upon the overall voting behavior of the committee?”  In the first (“continuous outcome”), we asked whether there was a relationship between the fraction of people voting in favor of the index drug and the fraction of people who had conflicts of the three kinds.  In the second (“dichotomous outcome”), we tried to address the possibility that voting behavior is nonlinear – when members of a committee come to realize that there is a growing consensus in favor of one particular point of view, they start to drift towards that point of view.   In this scenario, voting behavior would be most likely to be influenced by a conflict when the vote was close to 50-50, so we used a dichotomous outcome variable – whether the vote was in favor/against the index drug.  (In fact, many advisory committees are unanimous, or close to unanimous, usually in favor of drug approval.)  None of these analyses yielded a statistically significant finding.

We then looked at whether or not excluding advisory committee members with the various conflicts would have led to a less favorable vote for the drug being considered.  In each case, we found that it did: in 64 percent of meetings for index conflict, 77 percent of meetings for competitor conflict and 72 percent of meetings for any conflict, the vote became less favorable toward the drug.  Finally, we looked at whether or not excluding people with conflicts of interest would have changed the overall vote outcome of the meeting. Regardless of the conflict type, we found no such meeting.

I would like to point out, however, that there has been at least one example where exclusions would have affected the vote outcome.[4],[5]  At the Cox-2 inhibitor advisory committee meeting, which occurred in 2005 and was therefore outside the data frame for our study, the committee examined Vioxx, which had already been taken off the market, and the other two Cox-2s, Celebrex and Bextra, and voted in favor of all three drugs.  However, of the members who had consulted with industry, 93 percent voted in favor of the drugs in the various votes, whereas only 56 percent of non-consultants did so.  Had those conflicted members been excluded, Vioxx would not have been recommended for a return to the market (the FDA didn’t allow it to return anyway) and Bextra would not have been recommended for continued marketing (it was later removed from the market despite the advisory committee vote).

Frequently, one hears that there are both financial and intellectual conflicts of interest; somehow this argument is offered as evidence to downplay the importance of the financial conflicts.  While intellectual conflicts are important, they can readily be distinguished from financial ones.  Financial conflicts of interest are extrinsic to the scientific endeavor, whereas intellectual conflict is the very way science moves forward.  Financial conflicts can occur at variable levels – some people have them, some people don’t – and they can be quantified, whereas intellectual conflicts are ubiquitous and not susceptible to quantification in the same way.  Moreover, in the context of debate on an advisory committee, for example, it is unlikely that the financial conflict information will naturally emerge, whereas it is likely that any relevant intellectual one will.  There are relatively straight-forward methods to alleviate financial conflicts, whereas it’s not nearly as clear how one should approach intellectual conflicts.  Finally, our legal system has long recognized the distinctions between the two.

Impact of campaign contributions on Congressional Voting Patterns

Our second study looked at tobacco industry campaign contributions and their relationship to voting behavior in Congress.[6]  As to the predictor variable, campaign contributions can be quite high – as much as $37,000 in a single Congressional session.

We then used logistic regression to look at the predictors of sponsorship of tobacco control legislation.  Receipt of money from the industry was the strongest predictor (for the highest tertile of contributions, odds ratio in the 102nd House of Representatives = 13.9 and odds ratio = 6.3 in the 103rd House of Representatives, compared to the lowest tertiles), much more so than whether or not legislators came from a tobacco state, or whether they were Democrats or Republicans.

 

Public Disclosure of Payments to Physicians

The previous two studies try to close the circle between predictor and outcome.  The two other studies I would like to mention simply describe the prevalence of various conflicts of interest, like most research in the conflict-of-interest area.

In the first, we looked at the disclosures of payments to physicians in Vermont and Minnesota, the first states to make such disclosures public.[7] Although there is significant underreporting, large volumes of gifts were disclosed: $1.01 million in Vermont over two years and $22.4 million in Minnesota over three years. 

There remain significant deficiencies in the reporting processes in these two states. As noted, there are high rates of underreporting, with companies that report millions of dollars one year, yet nothing the next.  The submissions are nonstandardized with much aggregation by both physician and gift, limiting the ability to analyze the data adequately.  Exemptions to the reporting requirements are common (e.g., samples and research studies) and public access is limited.  In Vermont, we had to file a lawsuit to get access to the data and a personal trip to Minneapolis was necessary to obtain the Minnesota data.

Relying on disclosure as the cure-all for conflicts of interest is particularly problematic when non-disclosure is common.  In a recent meta-analysis of statin clinical trials, published between 1999 and 2005, 37 percent had no conflict-of-interest disclosures.  In another striking example, an article in Neuropsychopharmacology reviewing the vagus nerve stimulation device had nine authors, one of whom worked for the manufacturer.  The remaining authors disclosed no conflicts of interest, even though all were consultants to the company.  The first author was actually the editor of the journal, and so he was in an especially good position to know that its policies required disclosure of such conflicts; he was forced to resign.[8],[9]

While disclosure as an approach to conflict of interest has the deficiencies outlined above, it still has a role to play and can be effective in certain circumstances.  In a controlled trial, members of the British Medical Association were sent the identical paper to review, but were randomized to receive or not receive conflict-of-interest declarations.[10]  Those who received the declarations rated the paper statistically significantly lower with respect to Importance, Relevance, Validity and Believability than those not receiving the declarations.

 

Medical Education Services Suppliers

In the second of our two studies describing the prevalence of certain conflicts of interest, we used data published in the industry magazine Medical Marketing and Media that described companies known as Medical Education and Services Suppliers (MESSs).[11]  The magazine identified 123 MESSs of whom 80 (65 percent) responded to the survey; only 42 of these (53 percent of respondents or 34 percent of known MESSs) provided financial data.  Nonetheless, these companies reported $643 million in revenue in 1999, including $115 million for organizing grand rounds, $114 million for medical symposia and $60 million in publications-related activities.  Seventy-six percent of the MESSs’ clients are drug companies.  These data emphasize the degree to which medical education and continuing medical education activities can be driven by corporate interests.

 

 

Applying the Framework

To return to the framework for conflict of interest, we can consider approaches to addressing conflict of interest to represent the tension between the prevention and the treatment of these conflicts.  Adopting a preventive approach tends to favor legal or regulatory approaches, while treatment or management approaches tend to lean heavily on disclosure.  I will now revisit each of the four projects I have described to show how we have approached each issue.

Regarding FDA advisory committees, it is notable that the approach has evolved from disclosure to regulation to legislation as discomfort rose over the preceding approach.  Thus, the FDA initially had a disclosure policy as the cornerstone of its approach, but there were very few recusals from the meetings.  The quality of these disclosures was improved by regulation.  Subsequently, an additional guidance was adopted that should result in considerably more recusals.  But now the Congress has stepped in and required incremental annual decreases in the conflict levels.

The issue of tobacco industry contributions to Congress is a totally different matter.  Here you have legal restrictions on campaign contributions that are in place, but there are many loopholes.  Policy restrictions are not really an option here, as the industry is unlikely to self-regulate.  There is much debate about this issue in Congress as we speak, but currently the Democrats are leaning toward disclosure rather than restrictions or prohibitions.  This is an example where a legal restriction would be more effective and easier to enforce.

Turning to research payment to physicians, I would agree that legal restrictions in this area are probably not desirable and they’re certainly not feasible.   A policy restriction, dating back to the late 1970s when it was proposed by Senator Gaylord Nelson, is to take all the money that would go toward research, put it in a single fund, and then establish a group of researchers whose job it is to conduct, analyze and report results in a completely unbiased fashion.  Politically, this idea does not seem to have much traction at the present time.  Instead what we have is disclosure by some journals, some of the time, and at some scientific presentations, some of the time.

The fourth and final example is non-research payments to physicians.  (The same general issues apply to MESSs, so these are not discussed separately.)  Here legal restrictions have limited political feasibility, although I will point out that in Minnesota there actually is a ban on any gift to physicians exceeding $50.  In the policy arena, we are starting to see a small number of universities and health care systems that are putting various policy restrictions in place, such as banning samples or restricting pharmaceutical company representative access to clinics.  In the disclosure category, there is the Kohl-Grassley bill, which would create a national database of pharmaceutical company payments to physicians.

In sum, the particular mix of approaches that one adopts is highly dependent on the particular conflict-of-interest problem one is addressing.

We therefore reach the following conclusions:

• There is limited research actually linking the conflicts with the outcomes of interest; this committee could play a helpful role in delineating how future research should be conducted
• Financial conflicts are of particular concern and merit specific attention; the existence of intellectual conflicts of interest should not be used to distract from the more-remediable problem of financial conflicts
• In general, committees/reviewers with low or no conflicts can be assembled if sufficient effort is made
• Disclosure is no substitute for prevention of conflicts when this is feasible and legal

---

[7] Ross J, Lackner J, Lurie P, Gross C, Wolfe SM, Krumholz H.   Pharmaceutical company payments to physicians: early experiences with disclosure laws in Vermontand Minnesota.  Journal of the American Medical Association. 2007;297:1216-23.

[8] Bero L, Oostvogel F, Baccheti P, Lee K. Factors associated with findings of published trials of drug-drug comparisons: why some statins appear more efficacious than others. PloS Medicine 2007’4:e184.

[9] Nemeroff CB, Mayberg HS, Krahl SE, McNamara J, Frazer A, Henry TR, et al. VNS therapy in treatment-resistant depression: clinical evidence and putative neurobiological mechanisms. Neuropsychopharmacology, 2006;31:1345-55.

[10] Schroter S, Morris J, Chaudhry S, Smith R, Barratt H. Does the type of competing interest statement affect readers’ perceptions of the credibility of research? Randomised trial. British Medical Journal. 2004;328:742-743.