In the fall of 1999, the drug giant SmithKline Beecham secretly began a study to find out if its diabetes medicine, Avandia, was safer for the heart than a competing pill, Actos, made by Takeda.

But instead of publishing the results, the company spent the next 11 years trying to cover them up, according to documents recently obtained by The New York Times. The company did not post the results on its Web site or submit them to federal drug regulators, as is required in most cases by law.

“This was done for the U.S. business, way under the radar,” Dr. Martin I. Freed, a SmithKline executive, wrote in an e-mail message dated March 29, 2001, about the study results that was obtained by The Times. “Per Sr. Mgmt request, these data should not see the light of day to anyone outside of GSK,” the corporate successor to SmithKline.

The heart risks from Avandia first became public in May 2007, with a study from a cardiologist at the Cleveland Clinic who used data the company was forced by a lawsuit to post on its own Web site. In the ensuing months, GlaxoSmithKline officials conceded that they had known of the drug’s potential heart attack risks since at least 2005.

But the latest documents demonstrate that the company had data hinting at Avandia’s extensive heart problems almost as soon as the drug was introduced in 1999, and sought intensively to keep those risks from becoming public. In one document, the company sought to quantify the lost sales that would result if Avandia’s cardiovascular safety risk “intensifies.” The cost: $600 million from 2002 to 2004 alone, the document stated.

Mary Anne Rhyne, a GlaxoSmithKline spokeswoman, said that the company had not provided the results of its study because they “did not contribute any significant new information.”

The company said that Avandia was safe and that Dr. Freed no longer worked for GlaxoSmithKline.

A panel of experts will meet Tuesday and Wednesday to decide whether Avandia should still be sold and whether it is ethical to test Avandia directly against Actos.

Whether to withdraw Avandia is a question that has split the F.D.A., with some officials arguing that the drug is useful despite its risks and others insisting that it must be withdrawn.

According to the documents, Dr. John Jenkins, director of the agency’s office of new drugs, who has argued internally that Avandia should remain on the market, briefed the company extensively on the agency’s internal debate.

“It is clear the office of new drugs is trying to find minimal language that will satisfy the office of drug safety,” a top company official wrote in an e-mail message after he spoke with Dr. Jenkins, according to a sealed deposition obtained by The Times.

In the deposition, Dr. Rosemary Johann-Liang, a former supervisor in the drug safety office who left the F.D.A. after she was disciplined for recommending that Avandia’s heart warnings be strengthened, said of Dr. Jenkins’ conversations with GlaxoSmithKline, “This should not happen, and the fact that these kind of things happen, I mean, I think people have to make a determination about the leadership at the F.D.A.”

An F.D.A. spokeswoman said the agency would not comment on the contents of the deposition.

Members of Congress, where the Avandia case has led to legislative changes, said they were outraged at GlaxoSmithKline’s behavior.

“When drug companies withhold data regarding safety concerns about their medicines, they put patients at risk,” said Senator Max Baucus, Democrat of Montana, who is chairman of the Senate Finance Committee. Mr. Baucus and Senator Charles E. Grassley of Iowa, the committee’s ranking Republican, spent years investigating GlaxoSmithKline’s development of Avandia.

Besides the trial comparing Avandia with Actos, the company also conducted trials comparing Avandia with glyburide, a cheaper and older diabetes medicine.

When Rhona A. Berry, a company official, asked about publishing two of the trials, Dr. Freed responded in an e-mail message dated July 20, 2001, that referred to Avandia by the abbreviation of its generic name, rosiglitazone: “Rhona — Not a chance. These put Avandia in quite a negative light when folks look at the response of the RSG monotherapy arm,” the message said. “It is a difficult story to tell and we would hope that these do not see the light of day.”

Hiding the results of negative clinical trials was once widespread in the drug industry.

But after GlaxoSmithKline was found in 2004 to have hidden data that showed that its antidepressant, Paxil, led children and teenagers to have more suicidal thoughts and behaviors, the company settled a lawsuit by agreeing to publicly post data from all of its trials. In 2007, Congress mandated such disclosures. But the postings are often little more than cryptic references, so the issue is far from resolved.

With Avandia, GlaxoSmithKline has done more than hide trial data. An F.D.A. reviewer who closely examined a landmark Avandia clinical trial called “Record,” found at least a dozen instances in which patients taking Avandia suffered serious heart problems that were not counted in the trial’s tally of adverse events, mistakes that further obscured Avandia’s heart risks.

The company’s conduct of the Record trial has received sharp criticism from medical leaders for other reasons as well. To compare Avandia and Actos in 1999, researchers at SmithKline measured Actos’s effects in patients in the same way that they had conducted earlier trials of Avandia so that the results for the two drugs could be compared.

When the results of the study suggested that Avandia was more dangerous than Actos, the company decided against further comparisons.