WASHINGTON — A federal drug official on Friday dealt a severe blow to the popular diabetes drug Avandia, issuing a scathing review of a major clinical trial that its manufacturer has been using to argue that the drug was safe.

The reviewer, Dr. Thomas Marciniak of the Food and Drug Administration, found a dozen instances in which patients taking Avandia appeared to suffer serious heart problems that were not counted in the study’s tally of adverse events.

Such repeated mistakes “should not be found even as single occurrences” and “suggest serious flaws with trial conduct,” Dr. Marciniak wrote.

The detailed report could prove crucial next week, when a panel of experts will meet to consider whether to recommend to the F.D.A. that the manufacturer, GlaxoSmithKline, withdraw Avandia from the market or restrict its sale.

The panel’s decision will have broad consequences for the company, the F.D.A. and perhaps even the entire process by which medical products are approved. The agency almost never does clinical trials on its own, depending on drug companies to conduct them appropriately.

Avandia, which helps patients get better control of blood sugar levels, has already come under intense criticism. It has been shown to increase the risks of bone fractures and to cause swelling that can lead to heart failure and eye problems.

And a number of studies, including some by GlaxoSmithKline, suggest that it could increase the risks of heart attacks, strokes and death. But the company has relied heavily on the major clinical trial, nicknamed Record, to demonstrate that those risks are exaggerated.

Dr. Marciniak’s review of the Record study calls that assertion into question. He found one case in which a seizure patient was hospitalized for bleeding in the brain, but all mention of the episode was deleted from records. Another patient was hospitalized for 67 days after a severe stroke, but the study record showed no sign of a cardiovascular problem.

Still another died after being hospitalized for a serious heart problem, but the death was listed as arising from an unknown cause and not heart-related.

Correctly interpreted, he concluded, the study actually supports critics’ contentions that Avandia may cause heart attacks and strokes. “One does not have to be a mathematician or to perform calculations,” he wrote, to come to the conclusion that a combined look at all the trials of Avandia would demonstrate that it causes heart attacks.

Mary Anne Rhyne, a spokeswoman for GlaxoSmithKline, said the study demonstrated that Avandia is safe and added, “The Record study was conducted according to good clinical practices and the data are reliable.”

Dr. Marciniak’s review is part of a reassessment of Avandia’s safety by F.D.A. medical officers to educate the panel, who will meet Tuesday and Wednesday in Gaithersburg, Md., to advise the agency whether Avandia should be withdrawn from the market — after millions of prescriptions, billions of dollars in sales and 11 years of wide use.

The agency’s initial review of the Record trial, in 2007, found no problems. But independent researchers who have examined the health records of thousands of patients taking Avandia have found strong hints that the drug is more dangerous than GlaxoSmithKline has claimed. So Dr. Marciniak took an unusually close look at the Record study, demanding to see records that the agency rarely examines.

Only then did he turn up the study’s many problems. If given more time, he wrote, he would probably have found still more.

Debate about Avandia’s safety and how to handle reports of its dangers has split the food and drug agency and led to fierce recriminations, staff departures and questions from Congress.

It was the F.D.A.’s delay in issuing stronger warnings about Avandia that led Congress in 2007 to give the agency greater powers over drug makers.

Within the F.D.A., some officials insist that the evidence is mixed and others say it is strong enough to merit the drug’s withdrawal. An advisory meeting in 2007 concluded that Avandia did increase heart attack risks but that it should stay on the market.

In response, Avandia’s sales fell sharply, to $1.2 billion last year from a peak of $3.2 billion in 2006. But while endocrinologists have advised against its use, Avandia remains popular — with nearly two million prescriptions last year. If the drug is withdrawn, GlaxoSmithKline — already facing lawsuits claiming Avandia caused injuries — would see its liability soar.

In that 2007 advisory meeting, the Record trial — specifically devised to assess Avandia’s heart risks — was a significant source of reassurance for Avandia’s supporters. GlaxoSmithKline insisted the trial had proved that the drug was safe. In a news conference on Thursday, Dr. Janet Woodcock, director of the F.D.A.’s drug center, called the Record trial one of the “most germane because it’s a study of cardiovascular outcomes.”

Conducted at 338 sites in 23 countries, the Record study involved 4,447 patients with Type 2 diabetes who were recruited between April 2001 and April 2003 and followed for nearly six years. Half of the patients received two older diabetes medicines while the other half were given Avandia in combination with an older diabetes medicine.

The study was paid for by GlaxoSmithKline and conducted in part by Quintiles, a clinical trials company. A spokesman for Quintiles declined to comment.