To understand Merck's approach, we need a crash course in biology. COX (cyclooxygenase) is an enzyme in the body that produces an inflammatory response -- swelling, fever and pain. Drugs that reduce COX activity are called NSAIDs, and traditionally include aspirin and ibuprofen. But COX is present throughout the body and has a host of other effects, some of which are beneficial. For example, it helps protect the lining of the G.I. (gastrointestinal) tract from ulceration and bleeding.

Approximately 1 in 1,200 patients taking traditional NSAIDs for at least two months will die from G.I.-related complications. A conservative estimate: 16,500 NSAID-related deaths occur every year in the United States, approximately equivalent to the number of deaths from AIDS.

The holy grail of anti-inflammatories would be a COX inhibitor without this side effect. This became a real possibility in 1990, when COX was shown to exist as two chemically distinct enzymes -- COX-1 and COX-2. COX-1 protects the G.I. tract, while COX-2 is responsible for the inflammatory response. The obvious answer to the G.I. complications: develop a drug that inhibits COX-2 but not COX-1.

Give Merck credit. It developed Vioxx, a drug with the same degree of anti-inflammatory and anti-pain response as NSAIDs, but with a significant reduction in G.I. complaints and complications.

But medicine rarely comes in such neat categories. As far back as 1984, Garret FitzGerald and his colleagues at the University of Pennsylvania, writing in the New England Journal of Medicine, showed that the COX enzyme might prevent arterial blood clot formation. FitzGerald suggested that this side effect might mean little in healthy persons but could be dangerous to patients with severe atherosclerosis. Which sounds reasonable -- a high level of an anti-clotting enzyme wouldn't be necessary in normal vessels, but would be necessary in damaged vessels in which clot formation was more likely.

In 1997, with Vioxx still in clinical trials, FitzGerald himself consulted with Merck. According to Merck's position statement, recently filed in a U.S. District Court in Louisiana, in response to a class-action lawsuit, FitzGerald warned the company that COX-2 had a possible anti-clotting benefit. He expressed concern that suppression of this mechanism by a "coxib" drug such as Vioxx might increase the risk of blocked arteries.

Merck responded to FitzGerald's research by re-analyzing all of its Vioxx clinical data, and included FitzGerald's potentially worrisome lab data in its FDA application. (This data was published in 1999 in the Journal of Pharmacology and Experimental Therapeutics.) Merck also declared that it had adopted the standard procedure for "facilitating rigorous scientific analysis, on an ongoing basis, of all competing hypotheses about potential CV risks or benefits from Vioxx."

Was this a sufficient evaluation of a potentially serious side effect of Vioxx? Or should more have been done? Given that it is standard practice to assess all serious potential risks of new drugs, FitzGerald's concern should have been directly addressed.

But rather than design a study focused on the C.V. risks of Vioxx, Merck created the VIGOR (Vioxx Gastrointestinal Outcomes Research) study in 2000. It compared the incidence of G.I. complications of Vioxx to naproxen, another conventional NSAID, whose most popular brand is Aleve. The goal was to prove that Vioxx was an equally effective but safer drug than an over-the-counter NSAID. Unfortunately, the study, published in the New England Journal of Medicine, revealed a twofold increase in C.V. risk for patients who took Vioxx. Of course, Merck acted surprised. What Merck didn't tell us is that it had stripped the deck of patients at high risk for heart disease.

In the VIGOR study, 80 percent of patients were women with an average age of 58, with Merck being well aware that women, on average, develop C.V. disease 10 years later than men. Only 4 percent of the total subjects had a prior history of C.V. disease and were felt to be at high risk for further C.V. events. It was no surprise, then, that 38 percent of the heart attacks in the group of patients who took Vioxx came from this small 4 percent subgroup.

Ironically, it was Merck's failure to eliminate these few high-risk patients that allowed the C.V. effects of Vioxx to be detected. According to a review of coxibs in the New England Journal of Medicine, published in 2001, if these few high-risk patients had been excluded from the VIGOR study, the difference between Vioxx and naproxen would not have been significant. Therefore Vioxx may have never been taken off the market.

But there's more.

Anyone who's gambled in a casino knows that the key to winning is to get lucky initially and quit before the odds catch up with you. Get in and get out before the odds get a chance to declare themselves.

To understand the risks involved with a coxib, it is necessary to conceptualize risk as a function of time. The risk of electrocution is immediate. You don't grab a hot wire and get electrocuted next year. It's now or never. Similarly, a fatal allergic reaction to a penicillin injection happens within minutes, not in six months. In assessing risk factors with a predictable time of occurrence, it is easy to determine the study period. But other risks can take years or decades to manifest themselves.

It took many years to recognize the detrimental effects of hormone replacement therapy, or the time lag between sun exposure and the subsequent development of a malignant melanoma. Such correlations require following large groups of patients for extended periods of time. A major source of our understanding of cholesterol levels and the development of C.V. disease came from large-population longitudinal studies such as the Framingham Heart Study, which began in 1948 and is still ongoing.

It is self-evident that risk factors can be present for many years before becoming clinically observable. Although elevated levels of cholesterol increase the risk of a heart attack, the rate of heart attacks in teenagers with high cholesterol is far less than that of middle-aged adults with similar cholesterol levels. The cholesterol effect is cumulative; atherosclerosis takes time to become manifest.

If we wanted to minimize or negate the apparent effect of cholesterol on heart disease, we would study the youngest patients we could, and for the shortest period of time. If we ran a nine-month study to compare the incidence of heart attacks in a group of 10-year-olds with high cholesterol vs. normal controls, it is highly unlikely that we would see a difference.

The Framingham study has been ongoing for more than 50 years. Merck's VIGOR study followed its patients for an average of nine months. Even this short period revealed a very disturbing trend.

During the first six weeks of the study, there was no significant difference between the two drugs. As the follow-up period continued, the patients taking naproxen clearly were having fewer heart attacks than those taking Vioxx. But between months 8 and 11, the naproxen rate was relatively constant, while the Vioxx rate appeared to be accelerating. It was at this critical point that Merck concluded its trial.

When asked why it concluded its VIGOR study at this key juncture, a Merck spokesperson replied with a terse e-mail: "VIGOR was an endpoint-driven study and it ended when it was designed to end."

Recent Stories