Arteries Unplugged II: Great Expectations
The last paragraph of an AP story about Pfizer's withdrawal of torcetrapib awakened my inner statistician:
According to Pfizer spokesman Paul Fitzhenry, 82 patients taking the combination of torcetrapib died, compared to 51 deaths in the arm of the study where patients were taking Lipitor alone. Each arm of the study had 7,500 patients. Pfizer said that the study didn't raise any questions about Lipitor's safety.
As a social scientist who is somewhat competent with statistics, but who has no knowledge of clinical trials for pharmaceuticals, I found the above statistics distressing. That's because I honestly don't understand exactly why they present such a problem. Let me elaborate.
As a layperson, my assumption is that people participating in studies like the one above are already unhealthy to begin with, perhaps severely so. I can think of no reason why a person with healthy total cholesterol levels or good HDL-to-LDL ratios would ever be allowed to or asked to participate in such trials, but again, there is a lot I don't know.
So if these people already have serious health issues, then my prior expectation is that some of them will die during the course of drug trials. But exactly how many I don't know. And therein lies the problem. Of course some mortality rate would be expected in any population of 15,000 people over a significant period of time, even among healthy people. For people with serious health issues, it would have to be the case that the baseline mortality rate would be well above 1 per 1000 (0.10%), but how much higher I don't know. I am sure the data is out there somewhere, perhaps within the FDA, but I wouldn't know where or how to find it.
A second and related assumption is that even if you are testing a wonder drug (assuming such a thing even exists), the treatment won't work for everyone or at least not equally well. As in all things, individual responses to the same stimuli vary greatly. Moreover, it stands to reason that while some people will benefit from a drug, other people would die despite the medical treatment they get, not because of it.
A third assumption is that from a business and marketing standpoint, it looks bad if your products are associated with peoples' deaths and doubly bad if your business is products designed to improve health or prevent illness. But here we should be careful. There is ever a balance between risk and reward. And while it is axiomatic that the former must always be weighed against the latter, it is theoretically possible to kill off very beneficial drugs because unrealistic expectations about side effects. Clearly, consumers are ill-served by such a state of affairs.
What all of this brings to mind is the question of what exactly is the bad news. The problem for Pfizer appears to be that more people died when taking torcetrapib and Lipitor together than Lipitoralone- 31 more to be exact. There were two populations of subjects in the trials, 7500 each, making the mortality rates 1.09% for the combination group and 0.68% for the Lipitor-only group. As shown above, the differences in those rates is statistically significant and I understand why that fact, and that fact alone, is troubling.
What I would also like to know is the expected mortality rate for a study of this kind, a study involving subjects and drugs of these kinds. Clearly the Lipitor-only mortality rate of 0.68% must be well-below the threshold, otherwise there would be calls for it to come off the market as well. But there are none. And apparently the combination mortality rate of 1.09% is well over the expected mortality rate, and hence the rush to end the trials. Casino operations, I am told, make or lose money based on small differences in the percentage that the house is favored in games of chance. Pharma shouldn't be reduced to such a state.
If indeed the expected mortality rate is somewhere between say 0.80-1.00% then the Pfizer and other pharmaceutical companies have a very, very big problem. Essentially they can lose billions of dollars of sales and market capitalization and eliminate thousands of high-paying jobs because mortality rates didn't meet ambiguously-defined and possibly unrealistic expectations. What I mean to say here is that while there may be a number out there somewhere, not a single article I have read in the business press says what it is. Moreover, no matter what that percentage is, it must have some uncertainty around it, uncertainty that would be reflected in an allowable range within which observed mortality rates can fall and still be acceptable. It should not be the case that there is a magic number below which a drug is deemed "safe" and above which it is not.
That the business press neglects to mention this means that they are failing in their responsibility to place this event in its proper context. It wouldn't be the first time. Someone needs to clarify whether the aforementioned mortality rate for the combination therapy is as egregious as it has been portrayed or if the real issue is torcetrapib's inefficacy. This is an extremely important distinction. It's one thing to say a drug is killing people and another to say it doesn't work as had been hoped or expected. In both cases lots of jobs and market capitalization will be lost, but in the latter Pfizer may still emerge with its reputation intact and with good prospects for the 200+ other drugs under development. Both scenarios are bitter pills, to be sure, but one is a lot harder to swallow than the other.
See also: Arteries Unplugged I
Disclaimer: I own no shares in Pfizer or any other Pharma or biotech firm.
Update: Random John Reloaded has some sophisticated statistical questions of his own in a post entitled "Please ILLUMINATE me on these questions"
Meanwhile, former Pfizer VP Dr. Peter Rost is calling his former colleagues "Clueless". Though he surely has more than one axe to grind with the company, the fact that he once worked for the firm gives him a lot of credibility on the issue. That he wrote a book entitled "THE WHISTLEBLOWER, Confessions of a Healthcare Hitman" doubtlessly gives him much moral authority in this discussion as well. Though I live outside the US right now, I imagine Dr. Rost is getting a lot of calls for interviews right now.
Finally, KevinMD has a nice roundup of blogosphere reaction to the Pfizer crisis.
Comments
It doesn't matter what the untreated mortality is, because in effect the new baseline mortality is the lipitor group.
Posted by: richard | December 4, 2006 3:14 PM