The NYT has a piece today in which it describes the problems riddling Vytorin, a combination of ezetimibe and simvastatin, which is taken by 3 million around the world and makes 5 billion for its owners, Schering-Plough and Merck. The piece says that in spite of such widespread usage, there is apparently no clinical trial data that demonstrates that the second piece of the cocktail- ezetimibe- is efficacious. Much more concerning is the fact that there may be a link between ezetimibe and cancer, although the link looks fragile at best right now.
Statins have been the wonder drugs of our time, with Atorvastatin (Lipitor) being the best-selling drug in the world. Their efficacy in reducing heart attacks has been demonstrated in large-scale trials. Ezetimibe which blocks cholesterol absorption in the intestine and reduces LDL has much more tenuous effects. Recent large-scale studies found that while ezetimibe does reduce LDL, it's effect on the variables that actually matter are far less certain; there was no evidence that it actually reduces heart attacks. Yet it continues to be prescribed by thousands of doctors around the country, with patients shelling out considerable amounts for it.
However, the article raises the much bigger issue of knowing exactly when a drug is efficacious. According to the article, utility of a drug is usually gauged by "surrogate endpoints", that is, endpoints which indicate reductionist type effects rather than an actual increase in life-span or quality of lives. Take cancer for example. For most cancer drugs, tumor shrinkage is a convincing endpoint, not an actual increase in life-span. Or take cholesterol medications. The causal link between LDL cholesterol lowering and reduced heart attacks is apparently well-proven. Yet the body is complex enough for this causal link to be questioned; such questions arise most often in truly large sample studies, when the drug is on the market. Clearly the only true indication of side-effects or efficacy will come from such large-scale studies.
But there's a dilemma here as far as I can see it. The reason why surrogate endpoints are used seems clear to me; it's simply much more easy to look for such effects and ascribe them to drug action than effects like "increase in life-span" which can be controlled by multiple factors. Let's say two men who are the same age have been prescribed the same cancer medication. Both show shrinkage in tumors. Apparently the medicine works. But can this be translated into an observation about the difference in their life-spans, which can be attributed to so many different factors. Especially if the general health of one of the patients is worse than the other, then the cancer medicine can basically be an adjuvant and not the primary cause for his prolonged survival. How do we know that it's the cancer medicine and not his general health condition that actually increased his life-span? Naturally a reduced incidence of heart attacks is much easier to analyze than an increase in life-span. But even there it seems that so many factors can be responsible for a heart attack that it would be a problem trying to attribute specific effects to the medication, or especially the lack thereof of its effects.
The article also says that the FDA has become stringent about medications that address chronic problems like heart disease. The stringent bar set for a true study of efficacy seems to be about 10,000 patients over about four or five years. If every pharmaceutical company needs to do such a study to get approval, they might as well start digging their grave right away. Drug development is already so risky and expensive that putting a drug through 10,000 patients over five years and having the FDA almost certainly reject it after that would spell doom for all drug makers. Yet it is clearly unethical for doctors to keep on prescribing medication whose efficacy has not been demonstrated.
There does not seem to be an easy way out of this problem. To me right now it seems that, with all the compromises and problems it entails, the best bet might be to set such a stringent bar for medications for which good alternatives exist (and medications for chronic heart disease seem to fall into that category now) but relax the need for such large-scale trials for unmet and critical needs for which no good drugs exist. I am pretty sure the FDA is not going to set the bar for cancer so high.
So what's the way out for companies? To me the soundest scheme for now seems to be for doctors to provide information on the label saying that the drug did not show efficacy in a fair number of patients, and let the patient decide for himself or herself. It would be ridiculous in my opinion for the FDA to demand that the company withdraw the drug. Also, as a general thought, I think that both the FDA and public opinion need to get over their obsession of approving drugs only if they show efficacy in 100% or even 90% patients. What if a drug shows efficacy in 50% of patients? The rational thing would be for doctors and companies to explicitly say this on their label and let the consumer decide. That's the kind of thing that should happen in a liberal society.
Note: Over at The Pipeline, Derek has quite a few posts on this
Macrocycles, flexibility and biological activity: A tortuous pairing
6 hours ago in The Curious Wavefunction