For many years GlaxoSmithKline and its predecessor, SmithKline Beecham, marketed Paxil as an antidepressant that would reduce the risk of suicide in depressed patients. The results of the company’s clinical trials, presented to the Food and Drug Administration in 1989, suggested Paxil was far safer than a placebo.

Back then, the company reported that among nearly 3,000 patients treated with Paxil in the worldwide clinical trials, five committed suicide — a rate of about one in 600. By contrast, there were two suicides in a much smaller group of 554 patients randomly assigned to take a placebo pill — a rate of about one in 275, more than double that of the Paxil group.

But documents made public in the course of the lawsuit filed against Glaxo by Wendy Dolin, who alleges that paroxetine, a generic form of Paxil, contributed to the suicide of her husband, Stewart, suggest an alternate view.

Ms. Dolin’s lawyers argue that the documents show the two suicides in the placebo group occurred during the “wash out” or “run-in” period, when patients about to enter a new clinical trial are weaned from prior medications — before the new trial officially got started and before “randomization,” when trial participants are randomly assigned either to the placebo group or the drug group. That transition period is known to be an especially risky time for patients.

In early reports to the F.D.A., these suicides were marked with an asterisk and a footnote specifying they occurred during the run-in period. The asterisk did not appear in later company reports, according to documents obtained by Ms. Dolin’s lawyers.

Frances DeFranco, a Glaxo spokeswoman, said in an email that “the approach G.S.K. used in 1989 to 1991 in counting events that occurred during any phase of a clinical trial, including during the placebo run-in phase, was also used by F.D.A. in its safety review of Paxil in 1991.”

But the agency’s own experts said later that these placebo suicides should not have been included in the data.

In 1999, when the safety of Paxil was under scrutiny, an F.D.A. medical reviewer, Michael Seika, told the drug company as much, according to a Glaxo memo presented as part of the lawsuit. Dr. Seika had told company officials that if a patient were to die during the run-in, before randomization, “such a person should not be counted in our analyses.”

F.D.A. officials declined to comment on what difference, if any, this might have made in the drug’s approval.

In May 2006, after further analyses of a broader set of data, Glaxo added a warning to Paxil’s label and wrote a letter to health care providers cautioning that among adults of all ages with major depressive disorder, “the frequency of suicidal behavior was higher in patients treated with paroxetine compared with placebo” — a statistically significant 6.7 times higher.

A little over a year later, in June 2007, that warning was replaced by the standard black box warning label that the F.D.A. now requires all antidepressants to carry, which says the risk of suicidal behavior affects only those under age 25.

Glaxo officials testified that they had approached the F.D.A. four times about adding Paxil-specific information to the label, which might have included cautions about suicide in adult populations, but that the agency did not require any additional warnings.

“Drugs play an important role in so many people’s lives and can be lifesaving, but people need to know about the side effects,” Ms. Dolin said. “We want transparency.”