FDA Logo--links to FDA home page
U.S. Food and Drug Administration
HHS Log--links to Department of Health and Human Services website

FDA Home Page | Search FDA Site | FDA A-Z Index | Contact FDA | FDA Centennial

horizontal rule

Comment on "Financial Conflict of Interest Disclosure and Voting Patterns at Food and Drug Administration Drug Advisory Committee Meetings"

The article by Lurie et al. on "Financial Conflict of Interest Disclosure and Voting Patterns at Food and Drug Administration Drug Advisory Committee Meetings" draws on 221 meetings of FDA advisory committees. The authors analyze voting based on 76 product-specific meetings that involved yes or no votes on individual drugs, and find that none of the 76 voting outcomes would have changed had voters with conflicts of interest been excluded. Using other measures of the relationship between voting and conflicts of interest, they also find a weak tendency for financial ties to pharmaceutical companies to lead to voting results slightly more favorable to the drug being voted on. This note identifies an important result that was not highlighted in the original article: Advisory committee members and voting consultants with financial ties to pharmaceutical companies tend to vote against the financial interest of those companies. This result suggests that fears that disclosed conflicts of interest are leading to tainted, unreliable recommendations are unfounded.

To see how this result comes about, we start with the three categories of conflicts of interest used in the original analysis: (1) an "index conflict," which is a reported financial tie to the product's sponsor; (2) a "competitor conflict," which is a reported financial tie to a competing pharmaceutical company; and (3) "any conflict," which is a reported financial tie to either the sponsor or a competitor. Because competitor conflicts are the most common type, they dominate the "any conflict" category.1

Lurie et al. analyze the effects of conflicts of interest by comparing votes of advisory committee members and consultants who report conflicts of interest with votes of those who do not report conflicts of interest. The comparisons include individual level Mantel-Haenszel relative risk and Monte Carlo simulations of meeting votes The simulations compare actual votes to votes based on the assumption of no difference between voters with and voters without reported conflicts of interest. In addition to the individual-level voting analysis, Lurie et al. estimate the effects of waivers on actual votes and voting margins.

Lurie et al. interpret all yes votes on drugs as favoring the financial interests of pharmaceutical companies and all no votes as opposing their interests. But that interpretation only holds for the sponsoring companies; for competitors, yes votes on the index drug oppose their interests and no votes favor their interests. As an alternative to interpreting all yes votes as favoring pharmaceutical companies and all no votes as opposing them, we suggest interpreting votes based on their effects on the financial interest of the pharmaceutical companies. If a voter has a financial tie to the sponsoring company, we interpret a yes vote as favoring the company; if a voter has a financial tie to a competitor, we interpret a no vote as favoring the competing company. Rather than asking whether having a financial tie to any pharmaceutical company tends to increase votes in support of a drug (a notion inconsistent with conventional interpretations of conflict of interest), we ask whether having a financial interest tends to increase votes in favor of that interest.2

We can re-interpret Lurie et al.'s results using the statistics reported in their Table 4.3 The re-interpretation of votes by members and consultants with ties to the sponsoring company (index conflicts) is similar to the interpretation in the original analysis in that a yes vote on the drug is a vote for the financial interest of the company to which the voter has reported a financial tie. The relative risk of those with index conflicts is 0.74 (95% CI, 0.39-1.39), indicating that they are less likely to vote for the pharmaceutical company's interest than are participants without any financial ties , although the result is not statistically significant.4 The Monte Carlo simulation result is consistent with the relative risk; the predicted average of 8.2 votes (95% CI, 4.5-10.5) for the financial interest of the company exceeds the actual average of 6 favorable votes; the participation of voters with reported financial ties apparently reduces the votes for the product being considered by the committee, although this result is also not statistically significant. The only result consistent with the conflict-of-interest hypothesis is that the voting margin for the index drug becomes less favorable more often (14 meetings) than it became more favorable (7 meetings) when voters with index conflicts were excluded.

For members and consultants with ties to competing companies (competitor conflicts), a yes vote on a drug can be interpreted as a vote against the financial interest of the company to which the voter has reported a financial tie. The analysis reported in Table 4 shows a relative risk of 1.20 for those voters (95% CI, 1.12-1.28) indicating that they are more likely to vote against the interest of the company with which they have an affiliation than are participants without any financial ties.5 The Monte Carlo simulation predicts an average of 29.5 (95% CI, 23.5-34.5) votes against the company's financial interest, with actual average votes against its financial interest of 36. The analysis of voting margins provides additional evidence inconsistent with the conventional conflict-of-interest hypothesis. The voting margin for the index drug becomes less favorable more often (27 meetings) than it becomes more favorable (5 meetings) when voters with competitor conflicts are excluded.

For index conflicts and competitor conflicts, then, advisory committee members and voting consultants with conflicts of interest more frequently vote against the interests of the pharmaceutical companies than do advisory committee members and voting consultants without conflicts of interest.

Lurie et al.'s finding that, if voters with conflicts of interest had been excluded, not a single outcome of 76 FDA advisory committee meetings would have changed provides powerful evidence against the charge that the participation of voters with reported conflicts of interest taints recommendations of advisory committees. Although excluding voters with index conflicts reduces the favorable vote margin for the index drug more often than not, excluding voters with competitor conflicts more often reduces the favorable vote margin. The additional result that (compared with members and consultants without disclosed ties) advisory committee members and consultants with financial ties to pharmaceutical companies tend to vote against the financial interest of those companies provides valuable further evidence against the charge that the financial interests of voters taint committee votes.

Reference

Lurie, Peter, Cristina M. Almeida, Nicholas Stine, Alexander R. Stine, and Sidney M. Wolfe, "Financial Conflict of Interest Disclosure and Voting Patterns at Food and Drug Administration Drug Advisory Committee Meetings." JAMA , April 26, 2006 ; 295:1921-1928.

Footnotes

1Table 3 in the original article shows that before January 31, 2002 , only 36% of reported conflicts were specified as index or competitor. After January 31, 2002 , the draft guidance took effect and 97% of waivers disclosed at advisory committee meetings specified the type of conflict. This note does not attempt to re-interpret the original article's results for the "any conflict" category because our interpretation precludes combining the "yes" votes of index and competitor conflicts of interest.

2 In other words, we test for a relationship not between conflict of interest and yes votes on a drug but between conflicts of interest and votes for the interest of a pharmaceutical company.

3 We emphasize that we do not re-calculate any of the Lurie et al.'s statistics. We accept their numbers but give them a different reading; we believe our reading to be more consistent with conventional notions of conflict of interest.

4If there is no difference between those with and those without financial ties, the relative risk is 1.0. A relative risk less than 1.0 indicates that participants with financial ties to a company are less likely to vote in the company's interest.

5Because a vote of yes for this group represents a vote against a company's interest, a relative risk greater than 1.0 indicates that participants with financial ties to a company are less likely to vote in the company's interest.

horizontal rule