Two to 3 percent of children treated with antidepressants had suicidal thoughts or behavior as a result of the drugs, officials said yesterday based on the Food and Drug Administration's most comprehensive analysis of past clinical trials of the widely used drugs.

In a sharp departure from a decade-old position, agency officials said the increase in suicidal tendencies was not a result of the children's underlying depression but was caused by the medications themselves.

"Out of 100 patients treated, we may expect to see two to three patients [who] will experience increases in suicidality due to short-term treatment," said Tarek A. Hammad, the FDA analyst who conducted the latest review. The increase, he said, "is beyond the suicidality as a result of the disease being treated."

Senior FDA officials acknowledged that the results of the new analysis were substantially the same as the conclusions more than seven months ago of an agency scientist who also found an increase in suicidal tendencies in children taking the medications. At the time, the agency declined to make the conclusions of staff scientist Andrew D. Mosholder public, citing a lack of confidence in his results.

Hammad spoke at a meeting of an FDA expert advisory board that began a two-day hearing yesterday on the controversy over the reported risks of prescribing the widely used antidepressants to young people.

Robert J. Temple, the FDA's associate director for medical policy, defended the decision to hold off presenting Mosholder's controversial conclusions in March.

"Obviously people can disagree," he said. At a news briefing that included Mosholder and Hammad, Temple added: "We are cognizant that taking these drugs might increase suicidal ideation. But depression alone is a cause of potential suicide, so scaring people needlessly and overdoing it is worrisome also."

Parents and other critics of the agency had accused senior officials of turning a blind eye to the worrisome data. But Temple, Hammad and Mosholder agreed that the studies lent themselves to differing interpretations, in part because they were not originally designed to answer questions about suicidal tendencies. While individual clinical trials sometimes contradicted each other, the FDA officials said a broad pattern of evidence had emerged that, on average, children taking the medications had about twice the risk of developing suicidal behavior or thoughts as children receiving sugar pills, or placebos.

"It's very consistent," Temple said, adding that regulators would examine today whether the drugs' labeling should be modified and whether some drugs for children should be recommended to physicians ahead of others.

The controversy over the drugs is the subject of several congressional inquiries. Physicians and families have complained that pharmaceutical companies kept secret the results of more than half the trials that found the drugs were no more effective than placebos.

Prozac is so far the only medicine approved by the FDA specifically to treat children's depression, but physicians routinely prescribe other drugs of the same type, called selective serotonin reuptake inhibitors.

Experts had previously said that studies of Prozac found no increase in suicidal tendencies, but its unique status came into question yesterday. Hammad testified that a recent government-sponsored study, which researchers had described as convincing evidence of Prozac's effectiveness, found that it carried the same risk of triggering suicidal behavior as other drugs.

In an exchange with the head of the government advisory panel meeting at the Holiday Inn in Bethesda, the lead investigator of the new Prozac study also said it might not have been considered a success by the strict criteria of the FDA. The drug proved superior to placebos in three out of four criteria, said John March of Duke University, but "by the technical definition of the FDA, it would require the study be considered negative."

FDA officials declined to say whether they would have ruled the trial positive or negative, saying only that a number of factors are usually taken into account.

Prozac is so far the only antidepressant that has been shown to be more effective than placebos in two separate studies -- the FDA's standard for approval -- but supporters of the medications and senior FDA officials have said that does not mean the others are ineffective.

In March, the FDA officials issued an advisory to physicians reminding them to be vigilant to the possibility of suicide in children and adults taking the drugs. The data presented yesterday suggested that the increased suicidal tendencies seen in the trials may be unique to children.

Pharmaceutical company representatives offered conflicting recommendations to the panel.

Steven J. Romano of Pfizer Inc., which makes Zoloft, said the agency should warn patients selectively about those medications that showed a higher risk in the studies. The FDA advisory in March applied to the entire group of medications.

"Approaching this as a class effect may potentially jeopardize potentially beneficial treatments for children with major depressive disorder," he told the panel.

Joseph S. Camardo of Wyeth Pharmaceuticals immediately disagreed. Wyeth's drug, Effexor, showed the highest risks of suicidal behavior and thoughts, but Camardo said all the drug studies were small and used varying criteria. While agreeing that physicians should be informed of the results of all trials, he said the FDA should not issue warnings about some medications and not others.

Until the new Prozac study came out, Camardo pointed out, everyone had assumed there was no problem with that drug. "Everyone thought Prozac had a risk below 1," he said, referring to a statistical benchmark. "There is uncertainty in the results of these trials."

David Fassler of the American Psychiatric Association cautioned that the FDA should not scare patients away from treatment. "Half of all kids who suffer from depression will attempt suicide at least once, and at least 7 percent will die as a result," he said.

The advisory panel is expected to issue recommendations today on how the agency should respond to the new analysis. The agency is not bound to follow the recommendations but frequently does.