ADA: Avandia Critic Meets Skeptical Diabetologists

CHICAGO, June 26 -- When the spotlight here focused on rosiglitazone (Avandia) and its harshest critic, Steven E. Nissen, M.D., of the Cleveland Clinic, the cardiologist offered a mild apology to a room full of diabetologists.

"I know I've made life more difficult for many of you in the last few months," said Dr. Nissen at the American Diabetes Association meeting.

Dr. Nissen's meta-analysis in the New England Journal of Medicine, indicating that rosiglitazone was associated with a significantly increased risk of myocardial infarction, led to a deluge of phone calls from concerned patients taking the agent.

So in a hastily scheduled symposium, the ADA pitted Dr. Nissen against Philip Home, M.D., of the University of Newcastle in Newcastle, England, and director of the steering committee for the RECORD (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes). Unlike Dr. Nissen's damning meta-analysis, an interim analysis of RECORD, also published in NEJM, was inconclusive.

John Buse, M.D., of the University of North Carolina at Chapel Hill, president-elect of the ADA, was also on the symposium panel. Some diabetologists in the room were restive.

One audience member called the publication of the meta-analysis "irresponsible," and accused Dr. Nissen of sensationalizing what should have been a matter for clinicians, drug manufacturers and regulatory authorities to hash out among themselves.

Dr. Nissen argued that publishing negative results about a blockbuster drug in a major peer-reviewed medical journal was permissible in the face of a broken drug-approval system.

He asserted that the FDA rushed approval of both rosiglitazone and pioglitazone (Actos), another drug in the thiazolidinedione class, following reports in the late 1990s of the hepatotoxicty of troglitazone (Rezulin). In its haste, said Dr. Nissen, the agency ignored "a strong signal of excess myocardial infarction."

Initial studies, he said, showed an ischemic heart disease rate of 1.24% for rosiglitazone, compared with 0.69% for comparators, which translated into a relative risk for rosiglitazone of 1.8.

Dr. Nissen also noted that an FDA reviewer said that a long-term study of the cardiovascular effects of thiazolidinedione was needed, and that Dr. Buse had expressed concerns about the effects at an ADA scientific session at the time.

But Dr. Nissen and co-author Kathy Wolski, M.P.H., also of the Cleveland Clinic, came under criticism from some of the panel members and from the audience for their use of a meta-analysis.

In an interview prior to the symposium, Joel Zonszein, M.D., a diabetologist at the Montefiore Medical Center in New York, did not find special fault with the meta-analysis itself, but in the way it was presented in the NEJM.

"The publication was basically endorsed by the New England Journal of Medicine, and the editorial was written not by an endocrinologist or by a cardiologist, but by a statistician," Dr. Zonszein said, "and I feel that has done a lot of damage in general: A, to our patients, B, to the studies that we have, and C, makes our life more difficult. We have to wait for good prospective studies to find out whether these drugs work or don't work."

At the debate, Dr. Nissen answered such criticisms by replying that there had been many opportunities down the line for the FDA or others to step in and set up an orderly review process, with open access to information and notification of physicians, patients, and others of their findings through appropriate channels.

But such a review process never occurred, he said, and it wasn't until GlaxoSmithKline, as part of a settlement with New York State, published on a Web site all trial results of rosiglitazone studies that the cardiovascular events data came to light.

"There were alternate routes for this information to come before the medical community, and all those routes failed," Dr. Nissen said.

Dr. Home discussed the unplanned interim analysis from the RECORD trial, which appeared in the online version of NEJM on June 5. In that study, a 4,400-patient open-label, cardiovascular safety trial of rosiglitazone, the investigator found no significant increase in MI or cardiac death, but suggested an increased risk of heart failure, a previously documented adverse effect.

The RECORD trialists found that there were 49 MIs in the rosiglitazone arm versus 40 in a control arm (P=0.34) less than four years into a planned six-year trial.

According to Dr. Home, there is no evidence of increases in either all-cause mortality or cardiovascular mortality, and that the data do not allow a conclusion as to whether treatment with rosiglitazone results in a higher rate of myocardial infarction than therapy with either metformin (Glucophage) or a sulfonylurea.

Dr. Nissen jabbed away at RECORD, however, stating that it was poorly designed, underpowered, and would have crossed the boundary of futility had a futility analysis been performed.

"It's a matter of statistics," he said in a interview. "We know exactly how many events you need to get an upper confidence interval of less than 1.2. At the end of the study they will have an insufficient number of heart attacks. In fact, to determine whether there's an upper confidence interval of less than 1.2, they will have less than a 1% chance of having enough events to do that. To determine whether there is an upper confidence interval of 1.4 they will have an 11% chance of having enough events."

He continued, "The trial postulated an 11% event rate and it saw a 2.5% event rate. I've never seen a worse mismatch between actual event rates and postulated event rates. Even for the primary endpoint of death plus cardiovascular hospitalization it has less than 50% power."

It was unclear in the end whether the debate changed any minds, although a survey of the panelists and informal interviews with physicians at the meeting suggested that few, if any, would be inclined to start new patients on rosiglitazone.

Asked whether he would continue prescribing rosiglitazone for his patients who have good glycemic control on the drug, Dr. Zonszein said that he would.

"We see a lot of patients with type 2 diabetes who are very obese and very insulin resistant," he said, "and since the first thiazolidinedione, Rezulin, came around, I've seen these medications as good agents to use in that specific population."