The unexpected lesson of the Avandia debacle.
Also in Slate: Eliot Spitzer writes about the FDA's failed efforts to regulate drug companies.
This week, a Food and Drug Administration panel decided against yanking the popular drug Avandia—which was supposed to prevent heart attacks in diabetics. Avandia's problems were first suspected widely in 2007, when two Cleveland Clinic cardiologists reviewed 42 preliminary studies and speculated that the drug might actually boost cardiac problems by almost 40 percent. As a result, the FDA mandated a "black box" warning about Avandia's risks, the Veterans Administration removed the drug from its formulary, and the advocacy group Public Citizen petitioned to ban it.
However, just three weeks after the Cleveland Clinic report's release, GlaxoSmithKline researchers reported midterm results from their massive randomized trial of Avandia's effects on heart health, known as the RECORD study. "There was no evidence of any increase in death from either cardiovascular causes or all causes," they wrote in the New England Journal of Medicine. The final results confirming Avandia's safety were published in the Lancet last year. Shortly afterward, GSK asked the FDA to remove much of the prior "black box" warning, though the FDA ultimately refused.
But all along, several FDA officials had viewed GSK's data less charitably. Last month, a 135-page agency memo discovered almost a dozen patients whose cardiac complications were omitted—possibly by honest mistakes, perhaps not—from the study results by company-paid researchers. As a result, FDA investigator Thomas Marciniak concluded, the RECORD trial "cannot provide any reassurances" about Avandia. Furthermore, FDA researchers scanned tens of thousands of Medicare Part D claims data from 2006-09 and found that Avandia users have more strokes, heart attacks, and deaths than users of Actos, a competing drug.
There are a few notable points here. First, the FDA's investigations of Avandia's possible risks appear deep, scientifically rigorous, and dogged. (See their complete reports here.) Second, the data around Avandia are so mind-bendingly complex and open to interpretation that the American Heart Association called for more studies while the American Diabetes Association advised against taking Avandia. Even given the new GSK disclosures, the FDA's Avandia panel sent a very mixed message yesterday, since roughly the same proportion of the 33 members voted to ban the drug outright as voted to continue its sales without any major new restrictions.
Predictably, commentators are using the Avandia affair to push for more regulation of industry-sponsored drug trials. In an odd editorial move, the head of the New England Journal of Medicine rewrote the lyrics of a classic 1972 Jim Croce song to argue for more oversight; USA Today accused the FDA of "disturbing lack of urgency."
This leads us to the real danger of the Avandia affair. In our efforts to better police clinical research, we may inadvertently create more and more clinical trials asking largely pointless questions.
Few people talk about the broader problem with the Avandia studies: The goal in diabetes care is to normalize blood-sugar levels, which theoretically prevents kidney and retinal damage (so-called "microvascular" damage) and also stops heart attacks and strokes (so-called "macrovascular" damage). But treating high sugar levels doesn't really seem to help prevent heart attacks.
In the late 1990s, British studies using generic metformin, sulfonylureas, and insulin to lower blood-sugar levels showed no significant reductions in heart attacks and cardiac deaths among Type 2 diabetics. In 2008, the ACCORD trial showed that normalizing blood-sugar levels in diabetics actually led to 35 percent more fatal heart attacks.
From the start, the now-controversial RECORD trial was designed to show that Avandia was "non-inferior"—that is, no better and hopefully no worse at preventing cardiac problems—compared with existing treatment using existing generic drugs. But that aim was pointless for most patients: The generics themselves have never been convincingly shown to stop heart attacks.
Let's assume that the FDA tightened regulations so that all of the current drama around the RECORD trial was eliminated. In the end, if the results confirmed "non-inferiority," Avandia would have been a me-too drug, no better than cheaper alternatives. If it was worse—as it appears to be—diabetics are no worse off than they were with their generics.
Ultimately, the main problem with trials like RECORD isn't that they may collect biased results. It's that they can cost hundreds of millions of dollars, monopolize the best minds from major academic medical centers, and don't do anything useful for patients. Drug companies are not charities, and they expect a return on their massive investments. They design clinical trials with the goal of selling their product, not to advance the health and knowledge of the global population. They don't ask the questions about drugs for which we most need answers.
Because federal agencies like the National Institutes of Health rarely shoulder the costs of large drug trials, the pharma companies are the only ones throwing in money. They are the sugar daddies for many cubs at academic centers and health centers. Pharma companies pay lavishly and, in return, buy the right to design trials with very narrow aims.
Consider another recent example: the controversy over the 2008 AstraZeneca-funded trial of the cholesterol-lowering drug Crestor, known as JUPITER. That study concluded that even adults with normal cholesterol levels (but other risk factors like high blood pressure or obesity) should take Crestor if they had an abnormally high "C-reactive protein," or CRP. The kicker is that CRP can be measured only by a "high-sensitivity" blood test patented by the study's lead author. Last month, the Archives of Internal Medicine published several articles accusing the study authors of misclassifying data and succumbing to commercial biases to promote the drug's benefits.
As with the Avandia trial, this fixation on supposed data-manipulation misses the larger point. The Crestor trial was specifically designed, for better or for worse, to push an agenda (in this case, widespread CRP testing). Most observers believe the study should have included a comparison group of high-risk people with hypertension and obesity but normal CRP levels; this would have allowed them to compare responses to Crestor and evaluate the utility of the CRP test. The public's chance to learn the answer has been lost.
Had it been conducted perfectly, the Avandia trial still wouldn't have helped people with Type 2 diabetes avoid any long-term problems. More regulations on research are likely to drive up research costs, which will incentivize pharma companies to narrow their studies' goals further. That isn't in the best interests of our health.
The alternative is for the government to invest directly in more useful clinical trials. Currently, the FDA approves drugs for sale without any cost-benefit analysis—it looks only at safety and efficacy—and Medicare later foots the bill. The FDA effectively acts as the nation's largest pharmaceutical purchaser; its approval of Avandia in 1999 meant Medicare soon would spend billions more. Why not, then, invest $100 million up front to decide that's a good buy?
Darshak Sanghavi is Slate's health care columnist. He is chief of pediatric cardiology and associate professor of pediatrics at the University of Massachusetts Medical School as well as the author of A Map of the Child: A Pediatrician's Tour of the Body. Follow him on Twitter.
Photograph of Avandia by Joe Raedle/Getty Images.