Three years ago, Dr. Steven E. Nissen, a cardiologist at the Cleveland Clinic, conducted a landmark study that suggested that the best-selling diabetes drug Avandia raised the risk of heart attacks. The study led to a congressional inquiry, stringent safety warnings, a sharp drop in the drug's sales and a plunge in the share price of GlaxoSmithKline, Avandia's maker.

The battle between Nissen and GlaxoSmithKline was waged from afar in news releases and published papers. But on May 10, 2007, 11 days before Nissen's study was published in The New England Journal of Medicine, he and four company executives met face to face in a private meeting whose details have not been disclosed until now.

Fearing he would face pressure and criticism from executives, Nissen secretly recorded the meeting—which is legal in Ohio as long as one party to the conversation is aware of the taping. On a recent day in his sunny office at the Cleveland Clinic, Nissen shared the contents of the recording with The New York Times.

What was said at the 2007 meeting raises questions about science and ethics that have suddenly become keenly relevant. A congressional investigation released Saturday concluded that GlaxoSmithKline had threatened scientists who tried to point out Avandia's risks, and internal memorandums from the Food and Drug Administration show that some government health officials want Avandia withdrawn. The drug is still being taken by hundreds of thousands of patients, and sales last year were $1.19 billion.

So the battle over Avandia has begun anew, and issues raised in the meeting between the four executives and Nissen are likely to be raised again. For instance, during the meeting, company executives repeatedly promised to begin a crucial analysis of the safety of Avandia “;within days.”; Nearly three years later, such a study has not been published in a medical journal, although the company has posted results on its Web site.

Also during the meeting, Dr. Ronald L. Krall, GlaxoSmithKline's chief medical officer, predicted almost exactly the results of another crucial study of Avandia that was two months from publication and whose results, according to scientific protocols and the company itself, should have been kept secret from the company. In an interview, Nissen said the recording showed that the executives hoped to persuade him not to publish his study by suggesting that they had contradictory information they would share with him in a joint study.

“;In retrospect, it seems clear that neither statement was true,”; Nissen said. “;They did not have contradictory data, and they never intended to cooperate in any analyses.”;

In response to questions, GlaxoSmithKline said that its executives went to see Nissen to discuss a possible research collaboration and to review the safety information surrounding Avandia. “;GSK was not aware that Dr. Nissen secretly recorded the May 10 meeting and is disturbed to learn that he did so without advising G.S.K.,”; the company said in an e-mail message.

Krall, in an e-mail message on Monday, said he did not recall discussing potential results of the Avandia study in the meeting, but added that he did not learn the results “;until many days after the meeting with Dr. Nissen.”;

The recording and events surrounding it offer a rare window into an unusual confrontation between a prominent cardiologist and one of the world's biggest drug makers. Since industry is the source for much of the money and studies available in academic medicine, executives normally drive the agenda in such meetings.

But GlaxoSmithKline was sued in 2004 by Eliot Spitzer, who was New York's attorney general at the time, over the company's failure to publicize studies that helped to reveal that antidepressants could lead children and teenagers to engage in suicidal behavior. The company settled the lawsuit by agreeing to post all of its clinical trial data. These Internet postings became the grist for Nissen's analysis.

During the meeting with Nissen, the four executives spoke as if they did not know the results of Nissen's still-unpublished study. And Nissen did not mention that he had already sent it to the journal, the recording shows.

But a week before the meeting, the Congressional investigators said, GlaxoSmithKline had been secretly and inappropriately faxed a copy of Nissen's manuscript by a journal reviewer who also worked as a consultant to GlaxoSmithKline.

The company confirmed that the faxed copy of Nissen's study had been read by its executives, “;some of whom attended the meeting with Dr. Nissen.”; Revealing this knowledge to Nissen was not the point of the meeting, the company said.

This game of cat-and-mouse became, in retrospect, almost comical. At one point, the executives said that studies showed that Avandia had no impact on cardiac death rates. Nissen's study had concluded that the risk was increased by 64 percent.

“;I get different numbers on this, but we'll talk about it later,”; Nissen said.

“;Hmm!”; one executive responded. “;I would like to hear more about that.”;

Another executive said, “;That would be interesting to see.”;

“;This is one that is important to us,”; Krall concluded.

At another point, Nevine Zariffa, a company statistician, asked what Nissen was studying. “;So you've done a summary type meta-analysis looking at”; heart attacks and heart-related deaths? she asked, precisely describing Nissen's study.

“;That's right,”; Nissen answered, then asked rhetorically, “;Now, what am I going to do? What is my responsibility? I mean, answer the question for me. Do I sit on it? Drug's being used by how many people?”;

He was told that he should do an analysis that included information about when patients in trials suffered problems. Such detailed analyses are more reliable than the kind of study Nissen performed.

Nissen lost patience regarding the increased risk of heart attacks, or myocardial ischemia, in Avandia patients. “;I hope you guys understand how much trouble GSK is in here,”; he said. “;You've got a bunch of people who are incredibly vulnerable to myocardial ischemia, and you've had evidence that you're provoking ischemia in those people, and that is of grave public health consequences.”;

Krall asked Nissen if his opinion of Avandia would change if the Record trial—a large study then under way to assess Avandia's risks to the heart—showed little risk. Krall said he did not know the results of Record.

“;Let's suppose Record was done tomorrow and the hazard ratio was 1.12. What does ... ?”; Krall said.

“;I'd pull the drug,”; Nissen answered quickly.

The interim results of Record were hastily published in The New England Journal of Medicine two months later and showed that patients given Avandia experienced 11 percent more heart problems than those given other treatments, for a hazard ratio of 1.11. But the trial was so poorly designed and conducted that investigators could not rule out the possibility that the differences between the groups were a result of chance.

“;Dr. Krall did not know what the point estimates of the Record interim analysis would be on May 10”; because the company did not officially get access to the data until four days later, GlaxoSmithKline said.

Even as Krall and the other executives tried to corral Nissen, executives were strategizing over how to blunt the impact of his study if it were published. But according to internal documents, the company's own scientists had concluded that Nissen's work was sound.

In one internal e-mail message, the congressional investigation found that a company statistician stated that “;there is no statistical reason for disregarding the findings”; of Nissen's study. In another, Dr. Moncef Slaoui, head of research at GlaxoSmithKline, wrote that federal drug regulators, Nissen and the company's own researchers all seemed to agree that studies of the drug showed that it substantially increased the risks of death and heart attacks, also known as ischemic events:

“;FDA, Nissen and GSK all come to comparable conclusions regarding increased risk for ischemic events, ranging from 30 percent to 43 percent!”; Slaoui wrote.

In public comments after Nissen's study was published, company executives insisted that the study had been “;based on incomplete evidence and a methodology that the author admits has significant limitations.”;

Indeed, this battle was foreshadowed in the May 10 meeting when executives asked Nissen why he would publish his study if a more detailed look at the data—called a patient-level analysis—would provide a more reliable result.

“;But suppose we did this patient-level analysis and it looked very different from what you have?”; Krall asked.

“;But there's no way it can,”; Nissen soon said. “;Come on, guys. You already did your patient-level analysis for 42 trials. You're about to add in two trials that went the wrong way. What do you think's going to happen?”;

Krall said the two sides disagreed on the numbers.

“;And, you know, frankly, the last thing we want to do is get into a public debate about whose analysis is right. Now, sometimes that happens…”; Krall said.

“;No, public debates are just fine,”; Nissen interjected. “;In fact, in science, the best way I know of to get to the truth is you just get it all out there and you let the chips fall where they may.”;

One of the executives responded: “;And I supposed the science is the issue. And that's why we think this patient-level approach is the right one.”;

“;It is the right approach,”; Nissen said. “;Now I'm going to be equally blunt: you should have done this a long time ago.”;

That approach still has not been published in a medical journal.