Drug Safety in the United States

This week's issue of JAMA has an interesting article and three letters (1, 2, 3) discussing the drug review and safety system in the U.S.

First off, Dr. Najman follows-up on Dr. Nissen's article on muraglitizar (Pragluva; BMS/Merck, withdrawn) . In particular, Najman is interested in mortality associated with new drugs, particularly drugs in the same class as muraglitizar. Najman makes an important point that is often overlooked in public debates about drug safety. Many drugs are tested in particular patient groups, for example, fibric acid deriviates, which are used by many diabetic patients, were not tested in diabetic patients, even though many diabetics with a particular lipid profile take these drugs. More disconcerting is that three large studies of fibric acid derivatives found increased mortality in each study, though none were statistically significant.

Drs. Parra, Beckey, and Thomas present a different argument for increased drug testing in additional patient populations. In particular, their letter takes issue with the presumption that demonstrating safety in high-risk populations is sufficient to presume safety in less-risky populations.

Dr. Nissen presents a strong defense of components of the existing drug review system and challenges the position that mortality and morbidity benefits should be demonstrated for all new classes of drugs. He points out that the first statin would not have been approved until 1994, if ever, had the statin been required to demonstrate the benefits Drs. Najman, Parra, Beckey, and Thomas are asking for. Rather, Nissen proposes:

a middle ground ... that new therapeutic classes should definitively demonstrate benefits for important and validated surrogate outcome measures.

I share Dr. Nissen's concern about significant delaying new drug approvals by requiring definitive evidence of mortality and morbidity benefits. I think Dr. Nissen is correct in demanding benefit on surrogate outcome mearues, but the phrase "important and validated" is difficult. There is a dispute in the scientific communtiy as to what constitutes a validated
surrogate outcome measure. For example, Iressa was approved to treat lung cancer on the basis of symptomatic improvement and increased progression free survival, but neither of those measures were indicative of an improvement in survival (indeed, Iressa did not improve survival for lung cancer patients), on the other hand there is some evidence that in colon cancer progression free survival is an excellent surrogate measure of survival. What seems to be needed is a process for reviewing surrogate outcome measures in light of the measure that one is attempting to accelerate.

A competing vission is proposed by Dr. Strom who argues for maintaining a system very similar to the current drug approval process, but with the caveat that drugs can only get conditional approval until there is a substantially larger safety database (on the order of 30,000-300,000). This proposal is theoretically quite enticing, but does face significant ethical problems. The first challenge is that it is not obvious that one can ethically randomize patients to receive a treatment that has received conditional approval or a placebo (whether active or passive), thus one of the major methods to assess safety appears to be ruled out. The second challenge is how to enforce any limitation on use or prescription of the drug during conditional approval (i.e. drug X is conditionally approved for condition Y in patient population Z) when physicians continue to have the power to prescribe drugs off-label. The third challenge is to understand how this process is different from merely adding post-marketing safety studies to the approval process.

One possible approach would be to have large insurers "enroll" patients into large cohort studies that can be used for nested case-control studies that obviate many of the challenges traditionally held against case-control and other potentially retrospective epidemiologic study designs, but that is a subject for another post.