op Food and Drug Administration officials admitted yesterday that they barred the agency's top expert from testifying at a public hearing about his conclusion that antidepressants cause children to become suicidal because they viewed his findings as alarmist and premature.

"It would have been entirely inappropriate to present as an F.D.A. conclusion an analysis of data that were not ripe," Dr. Robert Temple, the Food and Drug Administration's associate director of medical policy, said in an interview. "This is a very serious matter. If you get it wrong and over-discourage the use of these medicines, people could die."

Dr. Temple was seeking to quell a growing controversy into whether the agency's warnings on March 22 that antidepressant therapy could lead patients to become suicidal were sufficient.

"There is concern that we hid data," said Dr. Temple. "We did not hide data. It was there for all to see."

Recent studies have shown that children given antidepressants are more likely to become suicidal than those given placebos. But the studies have lead to different interpretations by psychiatrists. The refusal by drug companies to publish the studies has worsened the confusion. Internal agency documents obtained by The New York Times show that federal health officials are divided, too.

Dr. Andrew D. Mosholder, an agency epidemiologist, was the man charged with analyzing 22 studies involving 4,250 children and seven drugs. In a carefully argued, 33-page memorandum, he concluded that children given antidepressants were almost twice as likely as those given placebos to become suicidal.

He urged the agency to discourage doctors from prescribing to children all antidepressants except Prozac. Prozac is the only antidepressant proven effective in treating depressed children, and its studies showed no link with suicide, Dr. Mosholder wrote. Dr. Mosholder's conclusions mirrored those made by British health authorities.

But Dr. Mosholder's supervisors, Drs. Mark Avigan and Anne Trontell, wrote memorandums disagreeing, according to the documents.

"In particular, we disagree that the data are sufficiently robust to advocate preferential use" of Prozac in children, Dr. Trontell wrote.

Health officials convened a special advisory committee on Feb. 2 to offer guidance on how the agency should respond to the studies. As the agency's principal reviewer, Dr. Mosholder was scheduled to speak. He was removed from the agenda, Dr. Temple said.

Senator Charles E. Grassley, a Republican from Iowa and chairman of the Senate Finance Committee, said he was investigating whether the agency inappropriately suppressed crucial findings. Representative Joe L. Barton, a Republican from Texas who is chairman of the House Energy and Commerce Committee, said he was also investigating.

"It would have been very wrong for the F.D.A. to withhold any information it had about unintended consequences that might result from the use of antidepressants, especially for children and adolescents," Mr. Grassley said.

Dr. Temple was unapologetic about the decision.He said that Dr. Mosholder's underlying data were presented to the committee but that agency officials explained that Dr. Mosholder had failed to treat with the appropriate skepticism some reports of suicidal behavior.

The Food and Drug Administration has hired investigators at Columbia University to examine reports from about 400 of the 4,250 children and determine independently which should be classified as suicidal.

The controversy had its start when GlaxoSmithKline sought a six-month extension to its patents on Paxil. To help guide pediatricians' prescribing habits, federal law grants such extensions when companies test their medicines in children.

The company's studies in depressed children failed to show any positive effect. It was a disappointing finding, but patent extensions are granted even when studies fail. Reviewing the data, Dr. Mosholder noticed in October 2002 that they showed a disturbing number of problems listed under the category, "emotional liability." Suicidal thoughts and self-injurious behavior were among the things lumped into this category. Dr. Mosholder asked GlaxoSmithKline to provide more detail about these cases, according to one of the documents.

In May 2003, the company submitted a new report. It showed that children given Paxil were more likely to become suicidal than those given placebos. In June, the agency announced that doctors should avoid using Paxil in depressed children.

Three days later, according to documents and interviews, the agency asked for similar data from the makers of Prozac, Zoloft, Luvox, Celexa, Wellbutrin, Effexor, Serzone and Remeron.

Dr. Mosholder found that 108 of the 4,250 children had done something that merited being called suicidal. He divided the cases into those who were given drugs and those given placebos and found that children given drugs were almost twice as likely to become suicidal as those given placebo.

The agency declined to make Dr. Mosholder available for an interview.

Try The New York Times Electronic Edition Free for 1 Week!