Orexigen Contrave Study Results "Unreliable" "Misleading"

http://www.forbes.com/sites/matthewherper/2015/03/05/top-fda-official-says-orexigen-data-unreliable-likely-false/


Top FDA Official Says Orexigen Study Result 'Unreliable,' 'Misleading'


The study results showing that Orexigen’s obesity drug Contrave reduced the risk of heart attacks and cardiovascular death and sent shares of the tiny La Jolla, Calif., biotechnology company soaring 30% were probably “unreliable,” “misleading,” and “likely false,” according to a top Food and Drug Administration official. If Orexigen cannot find a way to set things right, it could face fines, civil penalties, or even the withdrawal of Contrave from the market.

John Jenkins is the director of the Office of New Drugs. He had a key role in negotiating the specifics of a big heart safety study of Contrave, as well as the safety of the guidance used to design big heart trials that are required of diabetes and obesity drugs. I interviewed him earlier today about the release of the Contrave data, which Orexigen released on Tuesday via a patent and a filing with the Securities and Exchange Commission over the protests of the FDA, the researchers leading the clinical trial, and its marketing partner, Takeda, and about Orexigen’s earlier failure to keep the data confidential to even its own executives.

The idea behind these heart trials, which the FDA began requiring of obesity and diabetes drugs in 2008, is that a small, firewalled group of employees of the drug company is given early access to the data in order to show it to the FDA to show that the drug doesn’t cause a big increase in cardiovascular events. Takeda followed this protocol with its diabetes drug, alogliptin. Jenkins says Sanofi actually withdrew a marketing application because it was afraid to compromise a trial of one of its diabetes drugs by letting interim data slip out.

Says Jenkins:

The paradigm has always been that the interim analysis data must be kept very confidential so that it doesn’t become available for business purposes within the company.

And that’s intended to:

(1) Make sure we don’t do anything to compromise the integrity of completing the trial so we get the definitive answer on cardiovascular risk and

(2) It’s based on the knowledge that when you’re looking at a sample of only 25% of the data any estimates you get of the treatment effect of the drug are highly unreliable and can lead to false conclusions about either the safety or the efficacy of the drug.

So our two concerns right now are:

(1) Making sure that we can get the definitive answer about cardiovascular risk for Contrave, meaning that the required studies can be enrolled and completed in a timely manner, and

(2) We’re concerned that physicians and patients not make healthcare decisions based on data that are highly unreliable. I characterized this earlier as trying to understand who is going to win a football game at the end of the first quarter. We have lots of examples where interim analyses can give very misleading results compared to what the eventual outcome of the trial may be.

Yes, he spoke in numerated bullet points, saying “number one” and “number two.” When I asked him for examples, he referred me to a presentation that University of Washington statistician Thomas Fleming, the head of the Data Monitoring Committee that conducts interim analyses of the Contrave study made at an FDA meeting last August.

“Step back and think for a second,” Jenkins says. ”We required this study because we’re concerned that Contrave may cause adverse cardiovascular events because of its effect on blood pressure and heart rate. So the likelihood that that drug is going to have an early benefit is highly unlikely. So people need to be very cautious about making medical decisions based on these data, and we’re very concerned that investigators and patients may be unwilling to be in a trial based on these data when they are likely false readings of the actual effect of the drug.”

But the p values (a measure of statistical significance) released by Orexigen were very low. That usually means the result shouldn’t have happened by chance. Doesn’t that at least mean that it’s unlikely that Contrave causes cardiovascular harm, and mean that the trial will probably be positive? “I think those are highly unreliable findings,”

Jenkins responded. “I am not a statistician, but I can tell you that Dr. Fleming, who is a statistician, and the statisticians at the agency, and other people who are expert in this area will tell you the only thing you can really conclude with confidence from this trial is that excess cardiovascular risk is not two or greater. You have a 95% confidence that the excess risk is not two or greater. You also have 95% confidence that the actual point estimate of the effect of this drug [on cardiovascular events] is somewhere below two. So the finding of .59 at the interim is highly unreliable independent of the p value.”

So what can the FDA do about this? It told Orexigen when Contrave was approved that it would need to do a second big study, because Orexigen had not kept the data fire walled, instead letting over 100 people, including people outside the company and Orexigen’s CEO, learn about the results, according to FDA documents. Now, because of the release of data via a press release, some experts question whether doctors or patients will be willing to participate in that second trial. What if it can’t be completed?

Jenkins said he wouldn’t engage in “a hypothetical” and referred me to the FDA’s guidance. I asked him to explain what the guidance means in a generic case, not specifically related to Orexigen.

“Congress passed a law in 2007, FDAAA,” Jenkins said. ”They gave us the authority to require these trials. If companies are not meeting their obligations there are fines, there are civil money penalties, there’s a possibility for seizure, and there’s even a possibility for initiating withdrawal procedures.”

Documents released when the FDA approved Contrave say that the FDA had two problems with the Contrave heart trial. One was that Orexigen had allowed its data to leak out to far too many people to trust that it wouldn’t change the final result of the trial. The second was that too many patients (more than two thirds) had dropped off study drug. Both were concerns, Jenkins said. But the drop-outs were the result of something FDA had built into the study: because it wasn’t clear that Contrave was safe, patients were not to stay on treatment (or placebo) unless they were losing weight. ”The main reason we no longer have confidence in this trial to be the definitive answer to this question is the unblinding,” he says.

He used this issue to come back to his main point to doctors, patients, and (though he never mentions them) investors: Don’t trust the data that Orexigen released. “It points out the paradox of people rushing to believe the interim point estimate, because going into this trial all the priors were about cardiovascular harm. That shows the paradox of believing the interim analysis suggesting benefit.”