Blog | Thursday, July 11, 2013

Why the FDA should lift warnings on Avandia, even if no one will use it


In June the FDA conducted a two-day hearing on Avandia to review of controversial RECORD trial. I have blogged about Avandia extensively. In case you are not up to speed on what's happening, Matthew Herper at Forbes has one of the best summaries I have read called "Battleground FDA: How Tomorrow's Avandia Panel Could Help Shape The Future Of Diabetes."

The FDA's panel eased restrictions on the drug, but few doctors will likely ever prescribe this medication. Since rosiglitazone is off patent, it could be generic, but it is unlikely that any generic company will invest in bringing it to market. For the same reason, it is unlikely the GSK will start marketing the medication again. However, I believe there are important reasons why Avandia should be exonerated, even if few additional patients end up taking it.

1. The un-blinded RECORD analysis is enough reassurance
The re-analysis of RECORD was consistent with the original findings that Avandia did not lead to cardiovascular death or heart attacks. In addition, the Duke group that did the re-analysis did a pretty darn thorough job. Thus, the central issue of the discussion was whether an un-blinded, large randomized trial was enough to prove safety and negate the results of a meta-analysis.

The gold standard for science is the double-blinded, randomized control trial. Double-blinded means that when the researchers are analyzing the data, they don't know which group is the treatment group and which is the control, and patients don't know whether they are taking the experimental drug or a placebo. The reason for blinding is that it reduces potential bias. When patients are enrolled in a study, if they think they are taking a medication that will help them, it will help them, even if the drug is a placebo. The more they think the drug is helping the more it is likely to help.

So, if you did a headache study and gave half the patients Tylenol and half the patients nothing, the patients taking Tylenol would surely do better, but you couldn't prove that Tylenol relieves headaches because your results might have just been due to the placebo effect. In other words, in a study of efficacy, not blinding patients to a treatment may bias a study in favor of that treatment.

In the RECORD study, the investigators were blinded to the results (single blinded), but patients knew whether or not they were taking Avandia. There were several reasons the investigators did this, one being that in Europe (RECORD was an international study), Avandia was not indicated with insulin. However, having open label treatments in RECORD should not affect the results, primarily because record is a safety study. If anything, if you are patient in a safety study and you know you are taking the drug they are studying (must be something wrong with it if they are studying safety), you would be more likely to have adverse events. This is known as the nocebo effect.

It is the reason why if you read the all the information they give you at the pharmacy when picking up a prescription, you are much more likely to think you have one of the millions of side effects listed. In other words, in a study of safety, not blinding patients to a treatment may bias a study in favor of the control. However, in spite of this bias, the RECORD study still showed no harmful effects of Avandia. Thus, if anything, the un-blinded design of the RECORD study should lead the FDA to be even more assured about Avandia's safety.

2. The TZD class is good for treatment of diabetes.
When the Nissen publication was released and the Avandia storm began, patients got worried. Some patients simply stopped taking Avandia altogether, without substitution. Many doctors switched their patients to Actos, the other thiazolidinedione (TZD) available. Merck got lucky and launched Januvia around this time, so when physicians were looking to add a medication to metformin, due to TZD fear, and presumed safety of a new class of medications, they started prescribing Januvia. In other words, while Actos ended up getting most of the TZD scripts, fewer doctors were writing for TZDs.

This is a shame because the TZDs are the only class of medications to show that they keep a diabetic's sugar under control for many years. Older drugs like metformin and sulfonylureas usually fail after three to six years. TZDs also don't cause hypoglycemia. They also have positive benefits on lipids (raising good cholesterol and lower tirglycerides).

The TZDs are not perfect drugs. They increase fluid retention and can lead to congestive heart failure in patients who have decreased heart function. There is a small risk of osteoporosis, and with Actos there may be an increase in bladder cancer. Yet, while not perfect, most of these risks are manageable. Heart attacks are not, and so due the Avandia fiasco, the TZD class is likely underutilized. Pioglitazone is available without restrictions and is now generic. Thus, exoneration of Avandia would likely (and appropriately) increase prescriptions of these medications.

3. Other options may be less safe than once thought.
There are two reasons why in 2010, the FDA chose to severely restrict Avandia. The first was due to concerns of missing data and a non-independent analysis in the RECORD trial (which we now know was unfounded). Second was the notion that Actos seemed just as good as Avandia, but didn't have the risks, and therefore many saw no reason to keep Avandia on the market.

However, what we now know is that Actos may be associated with bladder cancer. (In fact the FDA knew about this in 2010, but brushed it under the rug at the July meeting--see page 213 of the original transcript). While pioglitazone may have a slight edge over rosiglitazone in a number of areas, in patients with a higher risk of bladder cancer (smokers), I might use Avandia over Actos.

Also, the new agents (DPP4s like Januvia and GLP-1s like Victoza) were new at the time. We now know that these agents may have risks of their own including pancreatitis, and the FDA is even looking into pancreatic cancer as highlighted by a recent New York Times article. Now there are pros and cons to all agents. A physician's job is to weigh the risks and benefits of each medication and choose the most appropriate product. However, this is made particularly difficult for diabetes when the specter of heart attack looms over the TZD class. Lifting the heart attack risk from Avandia will truly give diabetics more options (even if they don't take Avandia).

4. Clearing Avandia may help restore public confidence in the FDA, and possibly pharma.
The problem with the whole Avandia fiasco was it decreased the public's confidence in both the FDA and the pharmaceutical industry. The heavily marketed Vioxx was a lesson in caution. However, certain doctors and journalists made a name for themselves looking for the next Vioxx, and Avandia was a great target.

The FDA knew about the risk seen in their own meta-analysis but wisely chose to wait on any restrictions or warnings until other trials, including RECORD, were complete. We now know that this was the correct decision. However, some inside the FDA were not happen with this decision and made sure that the data got to the public. Hysteria ensued.

While industry watchdogs are necessary, they can sometimes to more harm than good. Ask any physician, and they will tell you that, in general, most patients are afraid to take any medications. This was not nearly as true 10 years ago. Today, patients don't trust drug companies or the FDA. The Avandia fiasco played a major role in this. The FDA can now right this wrong.

Matthew Mintz, MD, is a Fellow of the American College of Physicians. He is board certified in internal medicine and has been practicing for more than a decade. He is also an Associate Professor of Medicine at an academic medical center on the East Coast. His time is split between teaching medical students and residents, and caring for patients. This post originally appeared at Dr. Mintz' Blog. Conflict-of-interest disclosures are available here.