New York Times editorial on FDA’s Avastin Decision
On July 26th, the New York Times posted an editorial relating to the FDA’s decision on the drug Avastin. Their summary entitled “When a Drug Fails” missed the mark on so many levels that even on vacation, our co-founder Steve Walker felt compelled to respond.
This editorial reflects a typical lack of understanding of why the FDA granted accelerated approval for Avastin as a first-line treatment for breast cancer in the first place. Because there are now multiple approved treatments for metastatic breast cancer, with most women going on to treatment with several of them after their first line treatment stops working, it has become impossible to measure a first line drug’s effect on overall survival (how long a patient lives) after beginning treatment. The first trial of Avastin in this setting (on which its accelerated approval was based) indicated a significant advantage in delaying progression of the disease. The problem is obvious. Given the uneven response of patients to various drugs, and the FDA’s crude clinical trial standards designed only to measure and compare the average response of populations (as opposed to how individual patients benefit or don’t benefit from a drug), the FDA made the right call with Avastin (by approving it) the first time around. Think about it. A patient is treated with Avastin and may, or may not, benefit individually in terms of delayed progression or extended life, because the trials are not designed to try to learn whether the individual patient benefited. The FDA’s statistical trials, by design (a problem that cannot be fixed – it is an inherent limitation of statistics-based trials) cannot measure individual benefit. It can only measure the time to progression and the time to death of an individual, which is then dumped into a pool of those measurements from hundreds of patients in each trial arm, and used to calculate an average time to progression and an average time to death for the two populations (those who got Avastin and those who didn’t). They then compare the average outcomes (actually the average patient in one arm to the average patient in the other arm). Seem simplistic? It is. Each patient then goes on to be treated with multiple other drugs (but each patient’s experience is unique – different drugs in different orders, with very variable outcomes, and mostly outside clinical trials where measurements relevant to the initial trial with Avastin may or may not be accurately measured recorded), then at some point, they die. The patient lives, say, 18 months after being treated with four different drugs. Why did she live that long? Which drugs worked to extend her life, which ones didn’t? The FDA’s statisticians have no idea because the trials they mandate are designed far too simplistically to measure it, and it is the FDA that makes and enforces the expectations for trial designs, despite their frequent assertions to the contrary. The media’s consistently incorrect take on this, including in this editorial, is not surprising only because the media almost always fails to properly explain what happened. The most recent trials of this drug actually confirmed that there is a progression-free survival advantage, just not as much of an advantage as the first trial indicated, which means the drug is having a positive effect – on average, and it is still impossible to statistically measure overall survival for a first-line breast cancer drug. If they ran another trial, the result would show yet another progression-free survival result. The real problem here is not that Avastin doesn’t work – it is that FDA has long refused (actually for about 50 years) to update its clinical trial and regulatory science. Randomized controlled trials and the simplistic statistics that drive them, invented in the 1950′s at the FDA and written into law by Congress in 1962, aren’t working anymore. Just a small amount of progress in a still incurable form of breast cancer makes them unworkable. It is very difficult to make progress against a disease when FDA’s regulation has become so obsolete, we can’t even measure progress. With so much in the balance – more than half a million lives a year in the US alone – can’t we do better? We actually can, but resistance to real change at FDA is so profound, even getting started is proving to be an epic struggle. They are not saints over there. They are government employees following very old rules and policies in very formulaic ways and failing on a grand scale. Maybe you should write an editorial about that.