Posted March 2, 2005 at www.masterdocs.com/voting.htm and revised March 4 and March 10, 2005.
On February 25, 2005, the New York Times published an article by Gardiner Harris and Alex Berenson (10 Advisers Voting for Pain Pills' Sale Have Industry Ties) suggesting that financial ties to industry could have influenced voting behavior at the February, 2005 FDA Advisory Committee meeting on COX-2 pain drugs. The Times article was based on an analysis (CSPI Analysis) that was performed by the Center for Science in the Public Interest.
TMT has examined this issue and reached the following conclusions:
The following appear advisable:
Full public disclosure of all Committee member ties to industry.
Implementation of standard voting procedures across Advisory Committees.
Monitoring of impact of industry ties and question order on voting patterns.
Investigation into the basis for retaking of the Bextra vote.
It should be noted that these findings do not imply any improprieties on the part of the industry sponsors, the Committee members or the FDA. However, the findings do show that different groupings of Committee members show differences in voting responses:
Rheumatologists who have the daily task of managing patients with chronic pain tended to feel that all the drugs should remain available with appropriate restrictions on their use and with emphasis on selection of the best drug for an individual patient.
Epidemiologists and statisticians tended to focus more on identifying differences in population risks between individual drugs.
Members reported as having ties to industry tended to be more in favor of allowing marketing of the COX-2 drugs -- although unconscious bias cannot be excluded as a reason for this finding, the finding could also be based on sound analysis and detailed knowledge of individual drugs.
All groups on the Committee agreed on the following:
All three COX-2 drugs discussed increase the risk of cardiovascular events.
There is a need for effective management of risk so as to optimize the risk versus benefit profile in individual patients.
In widespread use, these drugs may be given to inappropriate patients, even when restrictive labeling and other risk management strategies are implemented.
A lack of long-term controlled data with all NSAIDs (whether COX-2 selective or not) presently prevents precise determinations of absolute and relative risk.
Members of the Committee put in months of preparation time before the meeting, and three days of hard work during the meeting. The public owes them a debt of gratitude. Although our analysis indicates that certain groupings of Committee members tended to be more or less favorably supposed to support marketing of the COX-2 drugs, there is no reason to suppose that such differences in attitudes were other than the honest opinions of the individual members. If one defines "bias" as "a partiality that prevents objective consideration of an issue or situation", one cannot reasonably attribute the Committee voting patterns to "bias". Rather, they may merely reflect differences in background, expertise and orientation -- one cannot say that one is correct and another is misguided.
There can be many different but legitimate points of view and it is useful to represent these different points of view when making important public health decisions.The present analysis focused on two variables (industry ties and the specialty of rheumatology). Other Committee groupings (such as clinicians compared with researchers) would be interesting. In addition, a more comprehensive statistical analysis factoring in all baseline variables would be of value.
The main message of this report is that different groups of well-intentioned people can have different perspectives that can importantly influence decision-making. A mix of perspectives provides a balanced view. However, Advisory Committee members should be selected with an awareness of the different voting patterns likely to be seen, and members themselves should be aware of group differences in voting patterns. In addition, there should be clear voting procedures applicable to all Advisory Committees.
Table 1 shows TMT’s analysis of voting patterns for marketing support for Celebrex, Bextra and Vioxx from the entire Committee:
There also appeared to be a more general relationship between the background/expertise of Committee members and voting behavior. The groups, in descending order of favorable response to COX-2 marketing were:
It is reasonable to conclude from the voting patterns of the full Committee that voting in favor of marketing of COX-2 drugs is strongly correlated with both financial ties to industry and to the specialty of rheumatology.
However, interpretation of the voting patterns is confounded by the fact that a more favorable voting pattern is associated with two variables (the specialty of rheumatology, and the presence of a financial tie to industry). The percentage of rheumatologists in the industry ties group was 3 of 8 (38%) versus 5 of 24 (21%) for the group without industry ties. As shown in Table 2, a larger percentage of Committee members with industrial ties voted to support marketing for all three COX-2 inhibitors regardless of whether they were rheumatologists (100% versus 60%) or non-rheumatologists (80% versus 11%).
TMT performed an additional analysis (Table 3) confined to the 75% of the Committee composed of non-rheumatologists. The relationship between a financial tie to industry and a favorable vote to support marketing persists in this analysis. Favorable Votes for Marketing were as follows:
Another issue of potential concern is the fact that the initial vote on the marketing of Bextra was disregarded, and voting was performed a second time. All 32 Bextra votes had been recorded before a decision was made to vote again. 31 votes were clearly stated at the initial vote. A few minutes later, Dr. Wood (the Chairman) announced that “Dr. Elashoff’s vote was not properly recorded because it was unclear what she said, apparently”. When she was asked to make a clear vote in response to this problem, she voted “No”.
Following this, the first Bextra 2B vote, as given to and announced by Dr. Wood was: “14 Yes, 5 Abstain, and 12 No” which did not provide a majority in favor of marketing Bextra. Note that, based on both the audio record and the official transcript, this tally is incorrect – the actual vote was 15 Yes, 10 Abstain, and 7 No.
The Committee then went on to discuss labeling restrictions for Bextra (Question 2C) and then Vioxx (Questions 3A and 3B). During these discussions, and prior to the 3B vote and repeat 2B vote, the following interaction between Dr. Wood and Dr. Nissen took place:
Dr. Nissen: You know, I’m, I’m disappointed in the abstentions. You know, we, we, we’ve all, we’ve all sat here and listened to the evidence, you know, we have …
Dr. Wood: Steve, I don’t think we should, should badger people into voting…
Dr. Nissen: Well, I, I, I actually I do want to ask people as we move forward to think about making a commitment one way or the other. Because, ah, you, what you have is a minority of us making a decision. And I think, ah, you know, it is, it is appropriate that people weigh in. So, one man’s opinion….
It is worth noting that Dr. Nissen asked many of the most probing questions during the meeting and showed no tendency to favor the Sponsors. He has also taken an exemplary position on divesting himself from financial conflicts of interest. Many people would consider that his position that Committee members should stick to Yes or No responses is also correct, and he made the point that his recommendations were intended to apply “as we move forward” – there is no reason to suppose that he thought that retaking of the Bextra vote would be done. Nevertheless, his “abstentions” comments, combined with the retaking of the Bextra vote, had profound consequences.
After voting on Questions 3A and 3B, the following discussion took place:
Dr. Wood: The hanging chads are... have raised their head. They want to go back. We can't agree on the vote apparently for 2B. That was... So, the Question for 2B was, em, "Does the overall risk versus benefit profile for valdecoxib support marketing in the US". Even though we announced the vote and everybody rushed out to file a story, em, <Dr. Wood laughs> it was premature. We're going to have to retake the vote because we're not sure what the vote was, apparently. So, so, em, I've forgotten which side we started on now. Who started? All right, Steve. So, <a Committee member laughs> let's go round again, and you'll vote. And let me remind everybody what we're voting here. We're voting for valdecoxib: Does the overall risk versus benefit profile for valdecoxib...
Questions to Dr. Wood: (inaudible).
Dr. Wood: Yes. We're.. Vald... We're going back to retake the vote for valdecoxib for question 2B, em, because there's some discrepancy apparently in the vote counting. Remember Florida? You thought I was kidding, right?
Questions to Dr. Wood: (inaudible).
Dr. Nissen: Where, where's Katherine Harris now that we need, need her?
Dr. Wood: Right. So we're going to go back to retake. Isn't that right? (obviously asking for confirmation from some person he was looking at). We're going back to 2B. We're going back to the question 2B. And we're taking the, the vote on 2B. So the question is: For valdecoxib, Bextra, does the overall risk versus benefit profile for valdecoxib support marketing in the US? A Yes would keep it on the market. A No would take it off the market. And Steve, were you the... which one was it?
Question from a Committee Member: Is it not on the tape recorder?
Abramson: Abramson, Yes.
<Voting proceeds to completion. The Committee votes in favor of Bextra marketing.>
<Further discussion and voting on Vioxx>
Dr. Wood: OK. And I’m now in a position to read you the votes for the, for t.. Question 2B and 3B, at least for now. Em, the, the, em vote for 2B, which was the vote on valdecoxib (for those of you who’ve forgotten already) em was 17 Yes, 2 Abstain and 13 No….
Thus, it appears as if the initial vote on the Bextra marketing was proper and did not support continued marketing. For reasons that are not clear, Dr. Wood was informed (presumably by some FDA person, and after the Vioxx 3B vote had given approval for Vioxx marketing) that the Bextra 2B vote should be retaken. Retaking of the Bextra 2B vote resulted in support for continued marketing. Dr. Nissen's exhortations to his colleagues not to abstain from voting occurred after the first Bextra 2B vote, and before the Vioxx 3B vote and Bextra 2B revote. If the first Bextra vote had been retained (as apparently would have been correct), Celebrex and Vioxx would have been the only drugs supported for marketing. This would have had the bizarre result that Vioxx (which was almost unanimously considered to have more adverse cardiovascular effects than Bextra) would have been supported for marketing whereas Bextra would not.
Another issue of concern is that these data suggest that the order in which questions are asked can affect voting patterns.
TMT has searched on the Internet for FDA rules on voting procedures but has not yet been able to find any such procedures. In 1992, the Institute of Medicine (IOM) evaluated Advisory Committees at FDA’s request and recommended that “the FDA develop uniform management guidelines for Advisory Committees applicable across all three centers and that it eliminate unnecessary differences in the management of committees” (www.nap.edu/books/0309048370/html/173.html). It is not clear if these recommendations have been implemented.
Implementation of standard voting procedures across Advisory Committees, evaluation of the impact of question order, and an investigation into the process by which the Bextra vote was retaken all appear advisable.
Note: The report has been revised to take into account the views expressed by several members of the Committee. Pfizer Inc. was given the opportunity to comment on this report but did not respond.
Table 1
Table 2
Table 3
Table 4
Voting Members of the COX-2 Advisory Committee
|
