Numerologists might point out that a drug named VIOXX was more than likely to suffer a reverse in 2004 (20-O-4). Indeed, the non-US trade name of the drug, CEOXX, perhaps might also have served as a warning for Ray Gilmartin, the head of Merck. Frippery aside, Merck's highly publicized withdrawal of rofecoxib (Vioxx) in recent weeks—and growing concerns over other cycloxygenase 2 (COX2) inhibitors (e.g., Pfizer's celecoxib (Celebrex) and valdecoxib (Bextra)—are prompting increasingly strident calls for revisions to postmarketing surveillance and to the FDA's supervision of the process.

Biotech companies should take note. Although most of the prominent drug withdrawals have involved pharma in recent years, the problem is equally relevant to smaller companies that are foolish enough to get involved in drugs for chronic conditions with large markets and large patient populations (e.g., insomnia, diabetes, daytime sleepiness), rather than orphan indications and cancer. In assessing a potential deal on a biotech phase 2 product, for instance, post-Vioxx pharma would undoubtedly discount for the costs and uncertainties of its postmarketing process, thereby driving down still further the value of early drug development.

One reaction within industry has been that COX2 inhibitors are exceptions and that the approval and postmarketing process works relatively efficiently to provide safe and effective drugs to patients. But it is not just Vioxx. According to the US General Accounting Office, the number of approved drugs recalled increased from 1.56% for 1993–1996 to 5.35% for 1997–2001. It is possible of course that the same numbers indicate that adverse-effect reporting improved between these two periods, or that the thresholds for adverse reactions have been lowered, or even that public pressure for drug safety is higher. But, at the moment at least, the favored interpretation is that the user-fee-accelerated approval process at FDA is more leaky than its predecessor.

At the same time, the US Department of Health and Human Services has reported that fewer than half of the postmarketing studies that manufacturers made a commitment to undertake as a condition of approval have been completed, and many have not even been initiated. There are also allegations that certain companies are dallying in reporting, ignoring or even concealing data on adverse events. Of course, no drug company sets out to kill patients, but faced with the loss of a blockbuster, there might be the tiniest temptation to downplay bad news about a drug's toxicity, especially if close attention to statistics is required to notice the adverse events.

A solution put forward by the Journal of the American Medical Association in December is for an organizational separation of the clinical approval and postmarketing arms of the FDA. As others have already pointed out, this would undermine the whole basis of current drug regulation, which strikes a balance between risk on the one hand and benefit on the other. If the two hands are separated, not only will each not know what the other is doing, they may not care much. The need to learn to work with, in effect, two distinct agencies born of disharmony and with incompatible remits could only work against biotech companies and play into the hands of those in pharma with greater regulatory contact.

Calls for draconian postmarketing requirements are thus entirely misplaced. Postmarketing surveillance is simply the wrong experiment to undertake. It is far too late and far too expensive. And it would be addressing a problem that in all likelihood is a temporary one.

Vioxx was a drug developed in an awkward age in the history of pharmaceutical development. The principle advantage of Vioxx over existing nonsteroidal anti-inflammatory drugs (NSAIDs) was that, because of its selective inhibition of COX-2, it could treat the symptoms of osteoarthritis without apparently causing the concomitant gastrointestinal inflammation. It sprang from a single discovery—of two cyclooxygenase systems. This, in its time, was a valuable insight from the leading edge of science. But the drug's subsequent development, although relatively recent, did not benefit fully from the possibility of skimming through the catalog of the human genome sequence or from the use of broadly based banks of wet assays for likely cross-reactivities. In other words, in Vioxx's case, a little knowledge was initially a highly valuable, but subsequently a somewhat dangerous, thing.

In such circumstances of partial information, could the FDA and other regulators justify scientifically all the label extensions that they have allowed for Vioxx? In a relatively narrow indication where the risk/benefit ratio of one drug versus another (or no treatment) can be readily assessed through controlled trials, the original approval of Vioxx was certainly valid. But label extensions that remodel the drug as a general treatment for pain might only be warranted with the benefit of a broader knowledge of the drug's behavior.

In order to reduce the risk in the postmarketing period for a drug, extensive phase 4 safety nets will simply never suffice. There will always be a broader phase 4 study to be performed. This is mere backside-covering of the first degree on the part of both the regulators and the companies involved, and not much better than a loincloth in the protection of pharmaceutical propriety.

The right experiments are those that can be performed during preclinical development—the major contribution of the biotech sector—experiments that lead to an underpinning understanding of the molecular, cellular and whole organism basis of diseases and their treatment. Such studies could direct preapproval development, and should be used also as an alert system that triggers specific and very directed phase 4 work. That directed work would need to be properly resourced at the regulatory level, perhaps with the introduction of compulsory guillotines for companies to establish structured postmarketing studies and punitive damages for noncompliance.