WASHINGTON (AP) — The diabetes pill Avandia has been used by millions of patients and racked up billions in sales in more than a decade on the market. This week the U.S Food and Drug Administration is trying to answer a seemingly simple question: Does it cause heart attacks?
It is the type of crucial safety question that the agency was established to answer. But the task of evaluating rare side effects of the drug that emerged after the agency approved it has dragged on for years without clear answers.
As the FDA began a two-day meeting Tuesday to discuss Avandia's future, it became clear that the agency's own scientists are deeply divided.
On Wednesday the FDA will ask a panel of outside advisers to untangle the mess of data that has stumped its own scientists for more than three years.
The group will vote on a range of recommendations, including possible withdrawal of the drug. The FDA will make a final decision in coming months.
The FDA has been down this road before. In 2007 a similar group of experts voted to keep Avandia on the market despite an analysis of dozens of studies suggesting it increased the risk of heart attack.
The agency is not required to follow the group's advice, though it often does.
New data on Avandia's risks, and pressure from public safety advocates, have prompted the agency to re-examine the drug's safety.
FDA Commissioner Margaret Hamburg opened the meeting by advising panelists to "follow the science wherever it leads and the rest will fall into place."
But it quickly became clear that the FDA's own staff have reached vastly different conclusions from the same science.
The FDA's first two presenters delivered conflicting opinions of a key study of Avandia which British-based GlaxoSmithKline has touted as proof of the drug's safety.
The so-called RECORD study is crucial to the debate because it is the only study designed specifically to look at heart safety of Avandia versus other diabetes treatments.
FDA drug reviewer Thomas Marciniak gave a stinging review of the study, which was sponsored by GlaxoSmithKline. Marciniak complained that an unusually high number of patients dropped out of the study and that several heart attack reports were excluded from the final results.
Part of the problem with the study was that doctors tracking patients — who were paid by GlaxoSmithKline — were aware of which patients were receiving Avandia versus other diabetes drugs. Often such trials are designed so doctors don't know who receives which pill.
Without the "blinded" structure, Marciniak said there may have been an incentive for doctors to underreport problems with the Glaxo's drug.
"I would have rejected this study design as completely biased," Marciniak told panelists.
Another FDA drug reviewer described the trial in even harsher terms.
"Everything we've seen up to this point shows you can't trust it," said David Graham. "And if you do trust it, you're engaging in the willful suspension of disbelief."
But following the criticism of RECORD, a senior level FDA official attempted to bolster support for the trial.
Ellis Unger, FDA's deputy director for drug evaluation, said that he found the RECORD results "largely free of bias, and very reassuring."
While some heart attacks were left out of the study, he pointed out that overall survival statistics were accurately reported and showed no disadvantage for Avandia.
Still, Unger acknowledged the difficult questions facing the FDA's advisers.
"There is no clear picture here, otherwise you'd be back at home," Unger told the panel. "That's why you guys are here: to read 1,000 pages and sort out what it all means."
Despite the reams of data presented at the meeting, the FDA's problem remains one of too little data.
While the original studies of Avandia showed the drug helped control blood sugar levels — the key measure for diabetes drugs — they were not large enough to detect all the potential side effects, such as heart attacks.
Since diabetics are already predisposed to heart risks it is extremely difficult to tell which heart attacks are drug-related and which are simply a result of the underlying disease.
Some scientists have tried to get an accurate picture of those risks by pooling hundreds of thousands of data points from various sources.
The most recent analysis — presented Tuesday by FDA's Graham, who wants the pill banned — estimated that as many as 100,000 heart-related problems may have been caused by Avandia.
Avandia was Glaxo's third best-selling drug in 2006 with U.S. revenue of $2.2 billion, according to health care statistics firm IMS Health. Safety concerns swirling around the drug have pummeled sales, which have fallen 75 percent to $520 million last year.