GENEVIEVE PHAM-KANTER
I. INTRODUCTION
In medicine, there is currently robust debate about the ways in which physician-industry relationships can best be managed. On the one hand, clinicians and physician researchers have experience that can contribute to the greater public good if they lend their expertise to help develop clinical therapies. On the other hand, financial ties between physicians and industry can create incentives for physicians to, either intentionally or unintentionally, behave in ways that are not in the best interest of their patients. In addition, the perception of conflicts of interest created by these financial ties, even if these ties do not affect behavior, can lead to diminished public trust in physicians and in medicine (Grande et al. 2012; Perry, Cox, and Cox 2014).
Medical centers and policy makers thus face an empirical challenge in distinguishing the conditions under which physician-industry relationships can be harmful rather than helpful. In addition, to the extent that the same financial relationship can have both good and bad consequences, institutions and regulators must determine the criteria to be used in evaluating and weighting these opposing effects.
A. FDA Advisory Committees
At the U.S. Food and Drug Administration (FDA), industry financial interests held by external experts who serve on FDA advisory committees have been, since at least the 1990s, a source of policy concern (Glodé 2002; Institute of Medicine 2002). Although the agency itself issues final decisions on the approval of pharmaceuticals that can be sold and marketed in the United States, it relies heavily on its advisory committees for guidance. These advisory committees are composed of biomedical researchers, clinicians, pharmacologists, and biostatisticians who can provide independent expert evaluation of the scientific evidence related to the safety and efficacy of products under review. At product approval meetings, committee members listen to hours of testimony and review thousands of pages of technical reports on the product being considered for approval. At the end of these meetings, members vote on questions that elicit members’ broad assessments of the safety and efficacy of the product under review. For example, members might be asked to vote yes or no on the question: “Should the sponsor [manufacturer] conduct further controlled clinical trials to assess whether and how drug X should be used in patients with condition Y?” Members are also asked to vote for or against a recommendation for product approval. The approval recommendations of advisory committees are typically (although not always) adopted by the FDA.
Advisory committee members typically serve two-year terms, and during their service, receive special government employee appointments. Because committee members are not regular employees of the FDA—many members’ main appointments are at academic medical centers—the financial conflict-of-interest rules governing the conduct of federal employees are not as stringently applied to them. More precisely, although the level and type of financial interest some members hold would, according to the published rules, disqualify them from participating and voting in meetings, they are often able to obtain waivers to participate despite their financial interests. These waivers are justified on the basis of the “essential expertise” provided by these members or that the benefits of a member’s participation are deemed by the FDA to outweigh the risks (Food and Drug Administration 2008).
B. FDA Advisory Committees as Research Focus
Because of the gravity of the decisions made by FDA advisory committees, the possibility of committee members holding financial interests that might affect the impartial scientific review of prescription products has far-reaching health and public trust implications. For this reason, conflicts of interest among advisory committee members are not merely a scholarly curiosity but a matter with potentially large health and social welfare consequences.
In addition, because of how the waiver system operates, FDA advisory committees are also a particularly good site for the study of how financial ties might work to influence physician and researcher behavior. In other contexts, individuals with known financial interests in firms that are potentially affected by a decision in which these individuals are involved would be disqualified or would recuse themselves (for example, in presiding over legal proceedings). In the case of FDA advisory committees, however, individuals with relevant financial interests can often participate and vote in approval decisions. Moreover, because the process for obtaining a waiver involves a well-defined process of disclosure, it is possible to obtain detailed information about these financial interests. We can thus, for instance, differentiate among various kinds of financial ties. Being able to draw these kinds of distinctions is important because we might think that, for example, the effect of an ownership stake in a firm whose product is under review is very different from the effect of employment at a university that receives research grants of the same dollar value. Because of the waiver disclosure process, it is also possible to distinguish individuals who have a financial tie to a single firm—and whether that tie is to the product sponsor or a competitor—from those who have financial ties to many different firms.
Finally, in studying FDA advisory committees, we have behavioral measures that are good first-order signals of bias associated with financial interests. For example, it is possible to examine members’ voting patterns in committee and see if they are associated with members’ financial interests to determine if committee members vote in a way that benefits the firms to which these members have financial ties. Note that, based on this association alone, we cannot make definitive causal claims; that is, we cannot claim that members vote in favor of a firm because they have a financial interest in that firm. We can say, however, that if we observe an association between financial ties and voting, then individuals who have financial ties to a firm systematically vote in a way that is favorable to that firm.
This chapter reports findings from an analysis of the association between the financial ties of FDA advisory committee members and their voting. The findings reported here relate to a facet of conflicts of interest that has not been empirically well studied—differences between exclusive (single) financial ties and nonexclusive (multiple) financial ties. In the medical literature, there appears to be some consensus that having many financial ties is unambiguously undesirable and worse than having a single financial tie (Henry et al. 2005; Lurie et al. 2006; Campbell et al. 2007). Having many financial relationships suggests that a physician is primarily driven by financial motives, or worse, is a hired gun, willing to change his testimony to suit his financial underwriter. In the context of FDA advisory committees, a member who has many financial ties may be more likely to favor industry interests—because he or she is financially dependent on them—rather than the public interest.
II. PREVIOUS LITERATURE
Previously, there have been two quantitative studies that have looked at the relationship between the financial interests of FDA advisory committee members and member voting behavior. In 2009, the FDA commissioned the Eastern Research Group (ERG) to examine the relationship between the financial ties of committee members and member voting in drug or device approval recommendations. That report found no statistically significant relationship between individuals’ financial ties to the firm sponsoring the drug or device under review and these individuals’ likelihood of voting in favor of approval (Eastern Research Group 2009). These results were similar to those from an earlier published study looking solely at drug approvals; in that study, Lurie et al. (2006) found no relationship between individuals’ financial ties to the firm sponsoring the drug under review and these individuals’ voting in favor of the drug’s approval.
One difficulty with these two studies, however, is their small sample sizes. In the Lurie et al. study, as the authors themselves note, although approximately 200 meetings were in the original sample, only eleven meetings could be analyzed for the relationship between voting and financial interest in the drug sponsor. In the ERG report, at most thirty-nine meetings could be analyzed for sponsor relationships. Small sample sizes mean that underlying associations may go undetected. Statistical theory tells us that it can be very difficult to statistically detect an effect in small samples even if there is a relationship in the underlying population.
A second difficulty with the previous studies is that the reported estimates between financial ties and drug approval voting were the result of simple two-variable relationships. In other words, the estimates were obtained first by looking at the link between voting and financial tie to a sponsor, and then separately, examining the relationship between voting and financial tie to a competitor. This type of analysis, which does not control for both key variables at the same time, is problematic because the lack of association between voting and financial tie to the sponsor does not also account for the possibility of someone having ties to both the sponsor and its competitors. In particular, it is entirely plausible, as discussed in more detail below, that having financial interests in many different firms might cancel each other out in terms of preferential voting. Economic theory predicts that a single firm may benefit from a financial tie, but if there are multiple competing firms, no single firm has an advantage over the others. It is therefore possible that none of the firms will preferentially benefit. By failing to control for the possibility of multiple relationships (which account for approximately 14 percent of conflicts in our sample), the estimates from these previous studies could lead to a misleading conclusion of no association between voting and financial interests in the sponsoring firm, especially if there are small sample sizes.
III. THEORETICAL ACCOUNTS OF MULTIPLE PHYSICIAN-INDUSTRY TIES
Comprehensively accounting for multiple relationships and exclusive relationships is important for clarifying the kinds of mechanisms through which conflicts of interest might operate. A common presumption, as noted previously, is that having more financial ties is worse than having fewer financial ties. This could be because, with multiple ties the physician is more beholden to commercial interests, which could create more bias, or it could be because the many ties are a signal of more worrisome but unobservable motives, which themselves are the source of bias in decision making. Although this account is the dominant narrative in the conflicts-of-interest literature, it is not the only theory available to us for interpreting the behavior of physician researchers and their financial ties to industry. An alternative account based in economic theory is that if multiple firms are establishing financial ties to the same researcher, no single firm will have a particular advantage. Each individual firm may seek initially to generate preferential treatment through its tie to a physician researcher. If, however, other firms see one firm developing a financial relationship, they too will want to develop financial ties to the same researcher. In competing to create individual advantage through financial ties, firms can enter into an arms race in which all firms are expending money to generate advantages over other firms. In the end, no individual firm “wins” in terms of holding an advantage because all firms are using the same strategy. Informally speaking, their expenditures cancel out.
A third theory related to multiple financial ties is that physicians who are very good at what they do are sought out by many different firms. These types of researchers are well-respected in their fields and may be highly influential. It may come as no surprise that high-status physician researchers, because of their ability to influence others, are pursued by many firms; but the crucial factor in this third narrative is not the influence of researchers with multiple ties but rather the researchers’ expertise. Because these researchers are extremely proficient in their specialty, they are of course attractive to many firms that would like to use their expertise in drug development. At the same time, however, these researchers are also expert at detecting problems with scientific evidence and product claims. These researchers will consequently take a more skeptical stance toward evidence presented to them, regardless of the source. Their exceptional expertise and skepticism mean that they will be less likely to be influenced by any particular firm.
These three theoretical accounts are developed in more detail elsewhere (Pham-Kanter 2014), but for now, the most important point to note is that these accounts generate three different empirical implications. In the first account, where physician researchers are hired guns, committee members who have multiple, nonexclusive financial ties should tend to vote in ways that are favorable to industry because they financially benefit from any and all firms. In this context, “favorable to industry” refers to whether individuals vote in favor of firms whose products are under review. For example, an individual who votes to recommend product approval would be considered to vote in a way favorable to the sponsoring firm and to industry. In the second account, where firms are in an arms race to gain a voting advantage, individuals with multiple financial ties should not vote differently from individuals who have no financial ties. The financial ties of competing firms in effect cancel each other out so no single firm—and in particular, the sponsor—is able to generate a voting advantage. In the third account, where physician researchers with multiple ties are highly skilled experts, individuals with many financial relationships should behave more skeptically than other members toward the scientific evidence presented in committee. Those individuals with multiple ties would therefore be less likely to vote in favor of the sponsor—for example, less likely to vote in favor of drug approval—than individuals with no financial ties.
Because these three accounts have different predictions for the effects of multiple financial interests on voting, they can be empirically distinguished from each other. In particular, if the experts are simply pro-industry hired guns, we should observe a positive association between having multiple financial ties and voting in favor of the sponsor. If the arms race mechanism is dominant, we should observe no association between having multiple ties and voting in favor of the sponsor. If the expert selection mechanism is important (selecting for top-performing but more skeptical experts), we should observe a negative relationship between having multiple ties and voting in favor of the sponsor.
IV. DATA AND METHODS
To analyze the first-order validity of these alternative accounts, the frequency with which members voted in favor of the drug sponsor was modeled. Predictor variables in this model included: (1) an indicator of whether a member had an exclusive tie to the sponsor; (2) an indicator of whether a member had an exclusive tie to a competitor; and (3) an indicator of whether a member had ties to both the sponsor and at least one competitor. The base comparison group was therefore the set of committee members who had no financial ties to either the sponsor or to competitors. This model was estimated using data collected on the financial interests and voting behavior of all individuals who attended and voted in meetings of advisory committees convened by the FDA Center for Drug Evaluation and Research between 1997 and 2011. The final sample consisted of fifteen years of meetings held for sixteen committees; 1,168 questions and 15,739 question-votes from 379 meetings were analyzed. The full data panel included 1,379 unique persons who cast at least one vote during the fifteen years. Because this data set is ten times larger than previous data sets, this study significantly improves the degree to which a link between voting and financial interests can be detected. Data collection protocols and statistical models and methods are reported in detail elsewhere (Pham-Kanter 2014).
V. RESULTS
A. Voting Behavior with Multiple Ties Versus Exclusive Ties
In brief, the study found that contrary to conventional wisdom, individuals with multiple, nonexclusive financial relationships do not vote differently from committee members who have no financial ties. Whereas individuals with an exclusive relationship to the sponsor have an increased odds of voting in favor of the sponsor when compared to individuals with no financial ties to any firms (odds ratio = 1.49, p = 0.032; reference group = 1.00), individuals with nonexclusive ties to multiple firms are not more likely to vote in favor of the sponsor (odds ratio = 1.16, p = 0.481). In other words, if one compares the voting behavior of individuals with exclusive ties to the sponsoring firm with the voting behavior of individuals with no financial ties, one observes that members with sponsor-only ties are more likely to vote in favor of the sponsor. When one looks at the voting behavior of individuals with ties to multiple firms (i.e., those with ties to the sponsor and to at least one competitor) and compares this behavior to the voting behavior of individuals with no financial ties, one observes that those with multiple nonexclusive ties do not vote differently from those who have no financial ties at all. This finding of no difference in voting between those with multiple ties and those with no financial ties supports the second of the three narratives presented—multiple financial ties reflect multiple firms competing in an arms race in ultimately futile efforts to secure a voting advantage.
Similar results are observed when the sample is restricted to only nonunanimous votes (about 60% of the original sample). Unanimous votes occur when there is overwhelming evidence in favor of the efficacy of a product or unambiguous and alarming safety evidence against a product. In these cases, committee members are in general agreement regarding the efficacy or safety of a product based on the evidence presented. In contrast, nonunanimous votes occur when safety or efficacy evidence is more ambiguous. In these cases, there is room for more subjectivity and for nonscientific factors to affect voting. By analyzing a sample that includes only nonunanimous votes, one can focus on situations in which the scientific evidence is more mixed and there is more room for discretion and bias. Interestingly, when the original model is estimated using this restricted sample, one finds that—in terms of point estimates—individuals who have multiple ties have reduced odds of voting in favor of the sponsor compared to members who have no financial ties (odds ratio = 0.79, p = 0.369; reference group = 1.00). That is, members with multiple ties were less likely to vote in favor of the sponsor than members who had no financial ties at all, although the difference was not statistically significant. These results are thus statistically supportive of the arms race account where multiple ties cancel out. At the same time, the point estimates are suggestive of the third account, in which individuals with many different ties are expert researchers who bring a more skeptical eye to the evidence than other committee members.
In general, these empirical findings suggest that, contrary to the conventional narrative described in the conflicts-of-interest literature, individuals who have many financial ties do not appear to vote differently from members with no financial ties. They do not consistently vote in pro-industry ways, and under some conditions, such as when the scientific evidence is more ambiguous, tend to react more skeptically, voting more frequently against the sponsor than individuals with no ties do.
B. Voting Behavior and Types of Financial Ties
The findings reported so far are estimates of overall associations between financial ties and voting, aggregating over all types of payments. Because the type of payment may matter (e.g., a research grant may have a different effect than an ownership or royalty financial interest), a more elaborate model was estimated using indicators of financial ties specific to the payment type. In particular, the sponsor-only ties, competitor-only ties, and multiple ties were also characterized by the type of financial relationship: (1) research (investigator or grant/contract recipient); (2) employer grant or contract; (3) ownership interest such as equity or bond holdings as well as income from royalties or licenses; (4) consulting; (5) member of scientific or other advisory board or steering committee; (6) blinded endpoint reviewer or member of data safety monitoring board; or (7) paid speaker.
Separating payments in this way, one finds interesting patterns with sponsor-exclusive versus multiple ties for some categories. For example, individuals with ownership interest (e.g., stocks, income from royalties or licenses) only in the sponsoring firm had greater odds of voting in favor of the sponsor than those with no financial ties, but those with ownership interests in multiple firms voted no differently from those with no financial ties. Here, again, there is empirical support for the second account—multiple payments reflect payments arms races and result in no net voting effect.
When one looks at speaking payments rather than ownership interests, one finds that individuals who have sponsor-only financial interests have greater odds of voting in favor of the sponsor, whereas those with multiple financial interests have lower odds of voting in favor of the sponsor. Thus, when one focuses on financial interests related to paid speaking, one finds that those with nonexclusive ties are less likely to vote in favor of the sponsor than those with no financial ties. This is evidence in favor of the third account, where individuals with multiple ties are more skeptical than other members of the advisory committee. It would appear that in this case multiple financial ties are paradoxically more innocuous than single, exclusive financial ties.
It is important, however, to emphasize that the estimates reported here are of overall effects of multiple financial relationships across many individuals and different committees. There are likely to be specific individuals with strong pro-industry preferences among those researchers with multiple industry ties. In addition, we know that committees vary a great deal in their average levels of conflictedness (Pham-Kanter 2014) and in the degree to which financial ties are associated with bias (results available upon request). Thus, the negative consequences of multiple ties may be more manifest in some committees and product markets than others. More detailed theorizing and further investigation of the factors underlying heterogeneities and of the conditions under which multiple financial relationships can be detrimental will be important.
The findings reported here also suggest that policy makers might do well to focus on the exclusivity of financial ties. Contrary to popular belief, a single financial tie need not be a signal of weak industry bias. The results described in this chapter suggest quite the opposite—a single exclusive tie could be associated with substantial bias.
This analysis shows that the number of financial ties held by a physician researcher can be a misleading measure of conflictedness and that exclusivity appears to be an important neglected dimension of conflictedness. Overall, this chapter argues for a more granular and nuanced approach to managing financial conflicts of interest in FDA advisory committees and, more broadly, in medical activities.
NOTE
Author Genevieve Pham-Kanter has no conflicts of interest to disclose. This research was supported by the Edmond J. Safra Philanthropic Foundation through a grant to the Edmond J. Safra Center for Ethics at Harvard University. Pham-Kanter had full access to all of the data in the study and takes responsibility for the data and the accuracy of the data analysis. Laquesha Sanders, Igor Gorlach, Magdalina Gugucheva, Kenneth Oshita, and John Barnes provided valuable research assistance.
Campbell, Eric G., Russell L. Gruen, James Mountford, Lawrence G. Miller, Paul D. Cleary, and David Blumenthal. 2007. “A National Survey of Physician-Industry Relationships.” New England Journal of Medicine 356:1742–50.
Eastern Research Group. 2009. “Financial Conflict-of-Interest Disclosure and Voting Patterns at FDA Advisory Committee Meetings.” Unpublished report.
Glodé, Elizabeth R. 2002. “Advising Under the Influence?: Conflicts of Interest Among FDA Advisory Committee Members.” Food and Drug Law Journal 57:293–322.
Grande, David, Judy A. Shea, and Katrina Armstrong. 2012. “Pharmaceutical Industry Gifts to Physicians: Patient Beliefs and Trust in Physicians and the Health Care System.” Journal of General Internal Medicine 27(3):274–79.
Henry, David, Evan Doran, Ian Kerridge, Suzanne Hill, Paul M. McNeill, and Richard Day. 2005. “Ties That Bind: Multiple Relationships Between Clinical Researchers and the Pharmaceutical Industry.” Archives of Internal Medicine 165(21):2493–96.
Institute of Medicine. 1992. Food and Drug Administration Advisory Committees. Washington, D.C.: National Academy Press.
Lurie, Peter, Cristina M. Almeida, Nicholas Stine, Alexander R. Stine, and Sidney M. Wolfe. 2006. “Financial Conflict of Interest Disclosure and Voting Patterns at Food and Drug Administration Drug Advisory Committee Meetings.” Journal of the American Medical Association 295(16):1921–28.
Perry, Joshua, Dena Cox, and Anthony D. Cox. 2014. “Trust and Transparency: Patient Perceptions of Physicians’ Financial Relationships with Pharmaceutical Companies.” Journal of Law, Medicine & Ethics 42(2):475–91.
Pham-Kanter, Genevieve. 2014. “Revisiting Financial Conflicts of Interest in FDA Advisory Committee Meetings.” Milbank Quarterly 92(3):446–70.
