Thirteen years ago, Garret FitzGerald, chair of the department of pharmacology at the University of Pennsylvania, warned that new painkillers called COX-2 inhibitors might increase the risk of heart attacks and strokes.

He was right. Since then, COX-2 inhibitors, notably Vioxx, have been yanked from the market or failed to win approval. Surviving brands have had to add a stern label warning about cardiovascular risks.

Recently, FitzGerald's lab published two studies in mice that aim to "close the loop" on the debate about exactly how COX-2 inhibitors harm the cardiovascular system.

FitzGerald answered health reporter Marie McCullough's questions about the research and its implications:

Question: First, give us a little biochemistry. The body produces enzymes called cyclo-oxygenase, or COX. What are the two types of COX, and what does each one do?

Answer: COXs come in two flavors, COX-1 and COX-2, and both produce fats called prostaglandins, which have contrasting effects in the body. The ones made by COX-1 tend to stop us bleeding and protect the lining of our gut. The ones made by COX-2 cause pain and inflammation.

Q: COX-2 inhibitors are part of a huge class of painkillers called NSAIDS (nonsteroidal anti-inflammatory drugs) that includes aspirin and ibuprofen. What are the safety concerns with older NSAIDS, and why were the newer COX-2 inhibitors expected to be better?

A: A problem with the older drugs was that they caused tummy upsets, in some cases serious bleeding ulcers. This was thought largely to result from the older drugs inhibiting COX-1 along with COX-2. Avoiding COX-1 inhibition was thought to protect the gut while blocking COX-2 alone was expected to preserve most of the clinical benefit. This expectation was upheld in clinical trials for Vioxx (rofecoxib) and another such drug which never made it to the market, Prexige (lumiracoxib).

Q: From the start, your lab raised a red flag about COX-2 inhibition because of its effect on a hormonelike substance called prostacyclin. What did you find?

A: Before any of these drugs were on the market, we had a chance to examine how Vioxx and Celebrex (celecoxib) acted in humans. To our surprise, we found that both drugs suppressed prostacyclin, as reflected by a breakdown product that we measured in urine. As we knew from test-tube experiments that prostacyclin had heart-protective properties, we suggested that this might spell trouble later on.

Q: Is prostacyclin also involved in pain and swelling?

A: Ironically, prostacyclin, formed by COX-1 or COX-2, is one of the fats that cause pain — for example, in an inflamed joint — as well as affording heart protection when it is formed by COX-2 in the lining of blood vessels. So blocking formation of prostacyclin by a drug like Celebrex is both part of how it is effective as well as how it confers cardiovascular risk.

Q: Your prostacyclin finding was controversial, and some researchers argued this would not increase the cardiovascular risk of COX-2 inhibitors. What were their arguments?

A: Well, some doubted that the marker in urine used in our first studies reflected production of prostacyclin by blood vessels. Others suggested that even if prostacyclin was blocked, other protective properties of blood vessels, especially production of nitric oxide (NO), would step in and compensate. Finally, some held that all this resulted from the drugs causing a rise in blood pressure and that this had nothing to do with inhibition of COX-2.

Q: Based on clinical trials done since then, what is the magnitude of the cardiovascular risk from chronic use of a COX-2 inhibitor?

A: We now have eight placebo-controlled trials — the gold standard of clinical evidence — showing that three distinct COX-2 inhibitors — Vioxx, Celebrex, and Bextra (valdecoxib) — increase cardiovascular risk. Estimates of the magnitude of the risk are not that precise, but it is probably around 1 to 2 percent of patients treated. This approximates the cardiovascular risk of being a diabetic or a cigarette smoker.

Q: Do older, nonprescription NSAIDS such as naproxen and ibuprofen also raise cardiovascular risk?

A: We don't have any gold-standard evidence for the older drugs. Less compelling evidence — observational studies — suggests that some of them, particularly Voltaren (diclofenac) and Mobic (meloxicam), do confer a risk, and indeed these drugs are as selective for inhibiting COX-2 as Celebrex. The evidence for risk from Advil (ibuprofen) is much shakier and indeed this drug inhibits both COXs together. The evidence for a cardiovascular risk from Naprosyn (naproxen) is least of all amongst the NSAIDs, but as it favors blockade of COX-1, it may be the one most likely to cause tummy problems.

Q: You've recently published two studies, done in specially bred mice, that confirm and expand on the prostacyclin work. How does this "close the loop" on debate about heart risks?

A: Our paper in Science Translational Medicine fills in the missing bits of the jigsaw. Here we removed COX-2 just from blood vessels in mice and made them susceptible to both blood clotting and a rise in blood pressure; this coincided with suppression in urine of the same breakdown product of prostacyclin that we measured in our first studies in humans of Vioxx and Celebrex. We also made the unexpected discovery that instead of NO (nitric oxide) stepping in to substitute for prostacyclin, it was itself suppressed, amplifying the consequence of COX-2 blockade. …

Q:When you blocked a protein called FLAP in the mice, the artery-hardening effect of COX-2 inhibition was reduced. You speculate that FLAP-inhibiting drugs, now being developed, could protect the heart from NSAID side effects. Shouldn't we avoid giving drugs to treat side effects of drugs?

A: Yes, the fewer drugs we have to take the better. However, if further studies in humans support our observations, FLAP inhibitors may offer promise to patients who must take NSAIDs for relief of chronic pain. Furthermore, hardening of the arteries takes months and years to develop, and even after stopping Vioxx, studies have shown that it takes some time for the increased risk to decline. Perhaps FLAP inhibitors can hasten resolution of that risk.

Q: Looking back on the COX-2 saga, do you have thoughts on how the U.S. could improve the way drugs are approved and monitored for safety?

A: The FDA is now inclined to move away from a binary approach to drug approval and withdrawal to a more conditional and gradual approach at both ends of the process. Even when a drug is approved, we have a lot to learn about how it works in the real world, for good or ill. On the other hand, withdrawal of a drug like Vioxx removes a therapy that has been useful to many but risky to some. We hope to work with many colleagues in the years ahead to do the science to predict safe and effective responses to NSAIDs, preserving the value of these drugs while managing their risk.