Vioxx, Avandia, Vytorin: absolute risk versus relative risk

There’s a fascinating discussion about Vioxx echoing around the blogosphere. (William Heisel has a nice summary here.) A few weeks ago Catherine DeAngelis, the editor of JAMA, spoke at the recent meeting of the Association of Health Care Journalists and said:

I believe Vioxx should not have been taken off the market. At least a million people, if told the truth about its side effects, would accept it because their lives could be lived with a lot less pain. There are a lot of people now whose lives are miserable because they have to live with arthritic pain.

At the same meeting, the editor of Reuters Health, Ivan Oransky, was critical of the way the Vioxx issue was reported by many journalists: “There was a doubling of heart attack risk, but the doubling was from a 1% risk to a 2% risk. If you lay that all out to readers, they can make much better decisions.”

I think DeAngelis and Oransky make valid points, but I think they miss a much more important point.

DeAngelis of course is right: there are probably a million people who even now would be willing to make the calculated risk and take Vioxx. The problem is that not only were those 1 million people never told about the risk, there were another 19 million people taking the drug as a first line analgesic who were extremely unlikely to derive any unique benefit from the drug and who never should have been started on the drug in the first place. In a carefully selected patient population the risk to benefit equation yields very different results than in an indiscriminate population 20 times the size.

Let’s assume Oransky’s and DeAngelis’s numbers are true. (It’s the general concept, not the actual numbers, that are important here.) If a million patients took Vioxx then the number of heart attacks doubled from 10,000 to 20,000. That seems like a lot, but if all these people were deriving significant pain relief from Vioxx, after having failed previous agents, then perhaps the excess risk could be deemed worthwhile.

But if 20 million people were taking Vioxx then the number of heart attacks jumps from 200,000 to 400,000, although only 1 million people are likely to derive any benefit from the drug. Suddenly the small increase in absolute risk is pretty frightening.

Imagine what would have happened if, following its approval, Vioxx had been used strictly as an analgesic of last resort in the few (say, DeAngelis’s million) appropriate people who were well schooled about the risks and benefits, and if a larger trial had been promptly initiated and brought to swift completion. What might have happened?

  • If the trial had been positive, Merck would have “lost” several years of high revenue for the drug in its early years but would have had a Lipitor-level blockbuster in the later years of the drug’s patent life.
  • If the trial had been negative, Merck would have “lost” some of the early revenue and might have lost additional revenue down the road, but it would have preserved its reputation and avoided years of lawsuits and investigations. (And unless you’re a masochist, it can’t have been too much fun working at Merck in the years since the Vioxx controversy started.)

The same general concepts hold true for both Avandia and Vytorin. In the case of Vytorin the safety issue has not been the focus of concern (though of course it briefly took center stage during the cancer scare frenzy). In all 3 cases, concerns were raised because the drugs were being taken by millions of people for whom there was no evidence or even reasonable case for benefit, due to the absolute lack of outcomes data. If Vytorin and Avandia had been used in a limited population of patients who had failed alternative therapies, and if a large convincing clinical trial had been swiftly initiated and completed, the controversies over these drugs would have been, at best, minimal and short-lived.

It’s all a bit like a high school prom. The companies acted like a bunch of crazy teenagers who drank too much and crashed their car into a tree. But in such a case the fault lies only in part with the kids. Where was the adult supervision? Where was the chaperone? Why did the FDA allow the companies to market these drugs so widely to primary care physicians and directly to consumers? Even more: where were the parents? Why were the academics and key opinion leaders sneaking the alcohol to the kids, instead of doing their job and keeping everybody safe?

Update (May 17, 2010): For an excellent followup to this post, please see William Heisel’s “Talking Risk: Vioxx, statistics and other complexities” on the Reporting on Health blog.


  1. Henry Greenspan, Ph.D. says

    Following up on Larry Husten’s post, some state Medicaid formularies (particularly Washington and Oregon) attempted to titrate the use of Vioxx to those patients who did not benefit from other NSAIDs, who had no known and significant CV risks, and who might derive a GI benefit. (Although assessing both CV risk and the Vioxx’s actual GI benefit are all dicey questions). Canada, where Vioxx was actually developed (Merck Frosst) , tried a similar approach.

    As usual, such schemes had some, but limited, success in impacting rx’ing. In the meantime, Merck itself made the decision to withdraw Vioxx, not FDA.

    Companies always know more about the actual benefit/risk equation of their products than any outside agency.. The bottom line is that Merck knew that continuing to sell Vioxx in a responsible manner would not be consistent either with the then undeniable science nor with their own marketing goals.

    In essence, by 2004 the jig was up.

  2. Michel Lièvre says

    DeAngelis comments rely on the assumtion that Vioxx was irreplaceable for at least one million people (in the USA, I suppose). Most studies sponsored by Merck compared Vioxx with high dose NSAIDs unopposed by a PPI. The population that would not have tolerated a classical NSAID + PPI combination falls probably to a handful. Another point is the distinction between severe gastrointestinal toxicity (Peptic Ulcer Bleeding) and pain. If it is right that a benefit in terms of gastric pain was obtained with Vioxx in a great number of patients, the benefit in terms of number of avoided PUBs was small, and, in patients at an elevated cardiovascular risk, in the same order as the MIs in excess. Here comes another point. The population that could derive a real benefit from Vioxx was a population at a low cardiovascular risk and a high risk of PUB. As the risk of both PUBs and cardiovascular events increases with age, this makes the target population elusive. Moreover, this target population could not be included in a clinicla trial, because it would have had a contraindication to classical NSAIDs (unless they were combined with PPIs).
    Finally, while one million people would have perhaps prefered to trade off a painless treatment of arthritis against a small risk of MI, those who would finally get the MI would probably have regreted their decision.

Speak Your Mind