Pfizer, the pharmaceutical giant, is betting big on the safety of its cox-2 painkiller, Celebrex.

The company has agreed to finance a study with an estimated cost of more than $100 million to examine the safety of Celebrex in arthritis patients prone to heart attack and stroke.

To lead the study, Pfizer has turned to an early critic of the cox-2 painkillers, Dr. Steven E. Nissen, a cardiologist at the Cleveland Clinic.

Dr. Nissen announced yesterday that the clinic's cardiovascular research center would coordinate the study of about 20,000 patients. He called it the first large-scale trial to compare the safety of a cox-2 painkiller with older nonsteroidal anti-inflammatory drugs in high-risk patients.

"This could show that celecoxib is the best drug or the worst drug," said Dr. Nissen, using the generic name for Celebrex. He said the study was a big gamble for Pfizer.

The trial could provide answers for millions of patients who have various forms of arthritis. Along with their doctors, many of those patients have struggled to weigh the risks and benefits of painkillers since September 2004, when Merck voluntarily withdrew its cox-2 drug, Vioxx, after a study showed that it increased the risk of heart attacks and strokes.

Since then, a Pfizer cox-2 drug, Bextra, has also been withdrawn.

"We're certainly very pleased that the Cleveland Clinic is heading the study and it is about to get under way, because clearly this is an important question and a very important patient population that the study is going to look at," said a Pfizer spokeswoman, Mariann Caprino.

Ms. Caprino said that it would be premature to discuss the potential costs of the study, but Dr. Nissen said that studies of this size cost at least $100 million.

Worldwide sales of Celebrex, the only remaining cox-2 drug marketed in the United States, have declined 45 percent since last year, but the drug still had revenue of $1.26 billion in the first nine months of this year, according to company figures.

The study could increase Celebrex sales by raising consumer confidence in the drug. Other companies that are developing new-generation cox-2's, including Merck, have expressed hope that the class will regain some of its luster after the risks and benefits are sorted through.

Pfizer first said in October 2004, just weeks after the cox-2 controversy arose, that it planned to undertake a long-term study to prove Celebrex's safety.

The idea was raised again at a meeting of a Food and Drug Administration panel in February that reviewed painkiller safety. Dr. Robert Temple, the director of medical programs for the F.D.A., and several panel members suggested a study to compare a cox-2 with other nonsteroidal anti-inflammatory drugs.

"It may be that at the end of this, one or the other drug will be superior," said Dr. Alastair Wood of Vanderbilt University, a professor of medicine and pharmacology who headed the panel and now recommends the painkiller naproxen instead of Celebrex. "That's exactly what we need to know."

Dr. Nissen, who served on the F.D.A. panel, said he began talks with Pfizer last summer to coordinate the trial, which will involve study sites around the world. The study will not include countries of the European Union, where authorities recommend against use of Celebrex by high-risk heart patients.

Patients in the study will have some form of arthritis along with histories including a previous heart attack, stroke, angina, coronary bypass surgery or stent procedure, Dr. Nissen said. Others will have diabetes, a known risk factor for cardiovascular disease.

"All the other trials that I'm aware of are really in quite low-risk patients, and so they just aren't comparable," he said.

Enrollment will begin next year and patients will be followed for about two years. One-third of the patients will be given Celebrex; one-third Motrin, or ibuprofen; and one-third will receive Aleve or Naprosyn, both brand names for naproxen.

The study will end when about 700 of the patients have died or had a heart attack or stroke, the number expected in such a high-risk group even without painkiller treatment.