This week's Food and Drug Administration

advisory committee hearings on Cox-2 arthritis drugs will feature detailed, dense depictions of data that look like a statistical version of the old quiz show

Who Do You Trust

?

The material filed with the FDA -- as well as research published recently in medical journals and presented at scientific conferences -- is demonstrating that different people can look at the same data and come to wildly varying conclusions.

The FDA is the tiebreaker. Its rulings on the Cox-2 drugs Vioxx, Celebrex and Bextra will govern not only the drugs' therapeutic future but also the legal and share-price implications for the drugs' manufacturers.

The FDA's advisory panel members will have to answer many questions in the name of consumers. For example, will they believe

Merck

(MRK) - Get Report

when it said it pulled Vioxx from the marketplace only after a company study showed that

long-term use increased cardiovascular risks?

Or will they believe critics who say Merck should have considered pulling Vioxx four years ago after a company-sponsored test in 2000 signaled that the drug posed risks.

What about the November report by several researchers in the

New England Journal of Medicine

that combined the results of several separate Vioxx studies to conclude that cardiovascular risks were apparent four years ago? Or do you believe Merck's response that this article mixed apples and oranges "in a scientifically inappropriate manner?"

And how will the FDA respond to data compiled by veteran FDA researcher

Dr. David J. Graham, who accused his agency of trying to suppress his findings that Vioxx caused an estimated 88,000 to 140,000 Americans to have heart attacks since the drug was introduced in 1999?

Will the advisory panel members support Graham's supervisors, who said the research was flawed but who placed his data on the agency's Web site anyway?

Will they back Merck, which said Graham's work was based on "speculation?" Will they agree with some of Graham's critics, who say his claims of excess Vioxx-related deaths and injuries are based on extrapolations rather than on corpses and patients?

Or will they defer to the British medical journal

The Lancet

, which recently published Graham's findings, giving Graham greater credibility?

Ironically, analysts say Graham's research makes

Pfizer's

(PFE) - Get Report

Celebrex look good by comparison.

More Than Vioxx

But Pfizer won't be able to dodge the spotlight. Its Cox-2 drugs Celebrex and Bextra each have been the subject of one high-profile medical study showing they increase patients' risk of

cardiovascular problems. However, most studies say the drugs don't exacerbate chances of heart attack and stroke.

Both drugs are also subject to the same types of conflicting opinions about data that have characterized the Vioxx debate. As one example, let's look at

an unpublished midstage clinical test in which Pfizer tried to determine if Celebrex could help patients with Alzheimer's Disease.

The consumer advocacy group Public Citizen, which wants the FDA to ban Celebrex and Bextra, recently said this research shows that Pfizer knew Celebrex posed a heart risk. Pfizer said the study, conducted between 1997 and 1999, was inconclusive.

Public Citizen argued its review of the data shows that patients taking Celebrex had a 3.6 times increased risk of "serious cardiovascular adverse events." Public Citizen said it analyzed "a composite of all cardiovascular events" rather than relying on the results for individual events. Such events included stroke, chest pains, heart failure, irregular heartbeats and heart attack.

Pfizer accused Public Citizen of "cherry-picking" data. "It is not possible to draw clinically meaningful cardiovascular conclusions from this study alone given the variations in patients' medical history, their advanced age and other significant study limitations," the company said. "This flawed use of data is misleading and unfair to a reasoned scientific discussion of these issues."

Researchers originally said the study's design posed problems: The sample size was small and the Celebrex group had more cardiovascular risks than the group receiving a placebo.

The researchers also provided some contradictory comments. In one sentence, their report said "individual cardiovascular events did not differ significantly between

Celebrex and placebo treatment groups." But in the next sentence, they said: "A statistically significant difference in adverse events was observed for certain cardiovascular-related body system terms."

These statistical skirmishes can play havoc with analysts' predictions.

"The reasoning in the Public Citizen letter is flawed," said Albert L. Rauch, a drug industry analyst for A.G. Edwards, in a Jan. 31 report to clients, reiterating a hold rating on Pfizer. "We believe Celebrex continues to be a viable product for the treatment of arthritis. Yet, pressure exerted by this strong patient advocacy group will serve to pressure the shares as debate continues on the safety of Celebrex."

Rauch took issue with Public Citizen for several reasons: The study was designed to test Celebrex's effect on Alzheimer's Disease patients, so controls weren't in place to assess cardiovascular risk. The length of the test -- each patient was tracked for 12 months -- was too short to accurately assess cardiovascular risk.

In addition, Rauch noted that many of the heart-related symptoms identified among test participants "can occur as part of the same disease process" so that adding these events together "may lead to double or triple counting" of the same events. As for Public Citizen's claim of excess deaths among Celebrex patients vs. placebo patients, Rauch said: "We found no significant differences between the groups." (Rauch doesn't own shares; his firm has a long position in Pfizer's stock.)

But Tim Anderson, of Prudential Equity Group, told clients on the same day that this news made him cautious. "That we now have two different datasets showing a cardiovascular risk is increasingly worrisome," said Anderson as he reaffirmed a neutral rating. "Yet we don't think withdrawing the product is in the offing." (His firm doesn't have an investment banking relationship; his research report says the analyst, a member of his team or a member of the analyst's household owns shares.)