In a year when the Food and Drug Administration has tackled a number of controversial issues, such as silicone breast implants, Prozac, and the AIDS drug DDI, few decisions seem as puzzling as the agency's advisory committee ruling late last month clearing the sleeping pill Halcion.

In the previous eight months, Halcion, the world's most popular sleeping pill, had been banned in Britain because of inaccuracies in its safety data. The FDA had been flooded with reports of serious side effects. A string of lawsuits had been filed against Halcion's manufacturer, numerous press accounts had painted Halcion as dangerous, and sales of the drug, which once amounted to a quarter of a billion dollars a year, had plummeted.

Yet after a day of relatively low-key debate and discussion, a special FDA panel concluded that there was no evidence that the drug was any riskier than other sleeping pills and reaffirmed the product as safe and effective.

How can a drug be banned in one country and linked with thousands of side effects and still be considered safe and effective for the 7 million Americans who take it?

The answer lies largely in how closely side effects of these drugs are linked to their dose. The answer also lies in the unprecedented effort taken by the FDA to re-analyze all of the existing clinical data on Halcion -- some of it more than 20 years old -- in order to get a more accurate glimpse of the drug's true safety profile.

"Because of the questions about the original data, because of the questions about possible inaccuracies, the FDA sat down and assembled from scratch a new data base, looking for what the incidence of adverse reactions was," said FDA Commissioner David Kessler. "I think that's what brought this issue back to science, that's what made all the difference."

Halcion -- or, as it is known to chemists, triazolam -- is part of the benzodiazepine family of drugs that are widely used as sedatives. It differs from the others, however, in one key respect. While some benzodiazepines can linger in the body for days after being taken, triazolam is cleared through the liver in a matter of hours.

This is the drug's great selling point: because it has such a short half life in the body, it cuts down on the next-day drowsiness or hangover associated with longer-acting sleeping pills. But this is also the drug's great weakness. Traditional sleeping pills are difficult to consistently overprescribe because anyone who takes too many pills in the evening will have so much left in their system the next morning they will keep on sleeping. But Halcion clears the body so quickly that there is no immediate and visible way of knowing when a patient has taken too much.

In fact during the early years after Halcion was approved, it was consistently recommended at what are now known to be doses much higher than are either safe or necessary.

The key question considered by the committee was whether this difference in how quickly Halcion acts in the body meant that the drug itself was qualitatively different or more dangerous than other sleeping pills. For example, some of its critics claim that Halcion binds to certain key receptors in the brain differently, suggesting that there is a biological mechanism that may cause it to pose greater risks than other benzodiazepines.

This, the committee rejected. Halcion, it said, was not qualitatively different than any other drug in its class. This conclusion was partly reached by a discussion of specific research relating to the biological action of the Halcion molecule. It also was strongly reinforced by the new data base compiled by the FDA showing that Halcion -- at appropriate doses -- was not linked to increased side effects.

What the agency did was to re-analyze all of the information from the data from 116 separate clinical studies involving 4,870 patients on triazolam, 1,377 on one of its main competitors (flurazepam), and 2,266 on placebo -- an inactive compound.

The agency discovered two things: In the first analysis, triazolam was in fact associated with more adverse reactions than flurazepam. Overall, 12.4 percent of the people had a reaction to triazolam, compared to 9.6 percent for its competitor. The most striking differences: Triazolam was 2.5 times more likely than flurazepam to cause anxiety, 2.1 times more likely to result in irritability and twice as likely to cause dizziness and lightheadedness. This, agency scientists said, provides some of the explanation for why so many more complaints are received about triazolam than other sleeping pills.

However, they pointed out, the data base included many people who were receiving triazolam at doses of .5 milligrams and above, well above what the FDA now believes to be the ideal and safest dose. When the data was re-analyzed comparing just the "appropriately" dosed Halcion patients with those on flurazepam, the rate of adverse reactions associated with triazolam actually dropped below that of its competitor (7 percent to 7.3 percent) and was virtually identical to the experience of those subjects given a placebo (6.9 percent).

The only specific symptom for which Halcion continued to do worse than flurazepam after the dose adjustment was anxiety, which still showed up three times as often among Halcion patients. The committee was unsure why this happened. One theory, advanced by Tufts University researcher David Greenblatt, was that because flurazepam has such a longer half life, those taking that drug were still mildly sedated the following day. In other words, the problem wasn't that triazolam caused anxiety, it was that flurazepam patients had less than normal anxiety because of the drug's lingering effects. Evidence for this argument was strengthened by the anxiety levels among those taking the placebo, which were only slightly lower (2.1 percent to 2.8 percent) than triazolam.

This FDA data base helps to explain the committee's final conclusion that the way to solve the Halcion problem was not to ban the drug but to include stronger warnings about its appropriate use. At the right dose, they concluded, Halcion doesn't seem to be any riskier than alternative medications.

It also helps to explain how Britain could have banned Halcion. Without the evidence of the FDA re-analysis, U.S. drug officials say, the British had no way of knowing that the higher incidence of side effects related to triazolam were largely related to dose and not to uniquely dangerous properties of the drug itself.

FDA officials are quick to point out that their data base is not the final word on the safety of the drug. In some cases, they said, there were too few patients treated at the "appropriate" dose to give them a great deal of confidence in their statistical results. Complete answers, they said, probably await the results of a 10,000-patient study now being conducted by Upjohn, Halcion's manufacturer, scheduled to be available in about three years.

Many panel members and witnesses at the FDA hearing also seemed eager to reduce public expectations about how a sleeping pill can and should be used.

"There is very little doubt in my mind that for a long period of time this drug was used at dosage levels that were higher than necessary," said Tom Roth, head of the sleeping disorders division at Henry Ford Hospital in Detroit. "What brought this about was the fact that people have this view that the proper functioning of a sleeping pill is to knock you out for eight hours. It is not. It is only to improve sleep . . . The use of sleeping medication is symptomatic management. It is not a treatment of root causes."