When we last spoke about the Avastin-and-breast-cancer story here, the FDA had rescinded its provisional approval for that indication, and a number of people were shouting that here it was, health care rationing based on price, right in front of us. As I said at the time, I think that those worries were misplaced: the reason Avastin was approved for metastatic breast cancer was that it seemed to work (a little). But when the numbers were firmed up with more studies, it turned out that it didn’t. The whole point of a provisional approval is that it can be rolled back if things don’t work out they way that they looked at first.
Now Genentech is coming back to the FDA next month asking for approval again. Here’s an op-ed in the New York Times that I think does a good job of laying out the case against the whole idea:
Genentech presented progression-free survival as a surrogate for better quality of life, but the quality-of-life data were incomplete, sketchy and, in some cases, non-existent. The best that one Genentech spokesman could say was that “health-related quality of life was not worsened when Avastin was added.” Patients didn’t live longer, and they didn’t live better.
It was this lack of demonstrated clinical benefit, combined with the potentially severe side effects of the drug, that led the F.D.A. last year to reject the use of Avastin with Taxol or with the other chemotherapies for breast cancer.
In its appeal Genentech is changing its interpretation of its own data to pursue the case. Last year Genentech argued that the decrease in progression-free survival in its supplementary studies was not due to the pairing of Avastin with drugs other than Taxol. This year, however, in its brief supporting the appeal, Genentech argues that the degree of benefit may indeed vary with “the particular chemotherapy used with Avastin.” In other words, different chemotherapies suddenly do yield different results, with Taxol being superior. The same data now generate the opposite conclusion.
Another problem, as the piece says, is that the whole cancer drug approval process has a tendency to slip into ancedotal form: tearful patients testify that the drug saved their lives. But the plural of anecdote is still not data, and never will be. In oncology, there’s really not much way of being sure about any individual patient’s response. There are so many different types of cancer, and they occur in so many different kinds of people. The only way to say anything useful is in a well-designed clinical trial setting.
Now, that doesn’t mean that you just have to round up thousands of people with all kinds of cancer and let things rip. It’s perfectly acceptable – in fact, very useful – to screen the patients that go into the trials so that you’re sure that they, as far as can be told, all have the same sort of disease. But you have to do that up front to really trust the conclusions. Data-mining, running things in reverse, is tricky, and if you’re going to do it, it should be used to tell you how to run your next trial, not to argue for approval. Only when you’ve run these kinds of experiments can you say with any certainly that a cancer therapy is useful.
But that’s a hard sell, compared to someone who is convinced that they’re alive because of cancer drug X (or is convinced that a loved one would be alive, if they’d only been able to get it). If you’re trying to persuade a crowd (or a mob), that would be the way to go: Aristotle’s appeal to pathos. But keep in mind that Aristotle (and the rest of the Greeks) looked down on that technique, and they were right. Logos, used properly, is what we’re after here, mixed in with the ethos of a disinterested observer who’s trying to find the truth.
And this gets to the moral dilemma at the heart of the modern drug industry: are we trying to find drugs that work? Or are we trying to sell drugs, whether they work or not? Roche/Genentech has every right to make its case and to petition the FDA for whatever decision they want. But they (and every other drug company out there) owe the rest of us, and the rest of the world, something while they’re doing it: to present all the solid data they have, and to let the numbers speak for themselves. But if the numbers can’t persuade, then a company should go back and get some more before trying again.