The original results showed that patients on Zyprexa, Lilly’s pill for schizophrenia, were 3.5 times as likely to experience high blood sugar levels as thosetaking a placebo, according to a February 2000 memo sent to top Lillyscientists. The memo is one of hundreds of internal Lilly documentsprovided to The New York Times by a lawyer in Alaska who representsmentally ill patients.

But the results that Lilly eventually provided to doctors until atleast late 2001 were very different. Those results indicated thatpatients taking Zyprexa were only slightly more likely to suffer highblood sugar as those taking a placebo, or an inactive pill.

Another Lilly report, from November 1999, shows that Lilly foundafter examining 70 clinical trials that 16 percent of patients takingZyprexa for a year gained more than 66 pounds.

The company did not publicly disclose that figure, instead focusingon data from a smaller group of clinical trials that showed about 30percent of patients gained 22 pounds.

Weight gain and high blood sugar are important risk factors for diabetes, and the question of whether Zyprexa causes diabetes has been a subject of scientific debate for several years.

Lilly says no link has ever been proven.

In response to questions about the difference between its first viewof the data and its subsequent public description, Lilly issued astatement yesterday saying that the later figures were accurate and theinformation in February 2000 was out of context.

In yesterday’s statement, the company said that after the February2000 memo, it re-examined its clinical trial results and found errorsin its “final, standard quality check of the data.”

But the February 2000 document, which is labeled “Confidential,”does not indicate that the figures it contains are preliminary. Infact, in a footnote, it explains that the data exclude patients “fromwhom there was a probable lab error.”

A separate document from November 1999 includes handwritten figuresidentical to those from the February document, with additional detailabout the increases in blood sugar that patients suffered.

The revised figures were shared with the Food and Drug Administration, Lilly said. It did not say whether it had ever disclosed the initial data to the F.D.A.

The F.D.A. did not respond to requests for comment yesterday.

The 2000 memo indicates that it was prepared as Lilly consideredchanging Zyprexa’s prescription label to provide doctors with moreinformation about the drug’s potential to raise blood-sugar levels.

The issue was crucially important to the sales prospects of Zyprexa,which was introduced in 1996. Psychiatrists were already increasinglyaware by 2000 that Zyprexa caused severe weight gain in many patients.

“In 1999, we already were thinking this drug causes weight gain —that’s clear — and there could be a lot of other metabolic consequencesof that,” Dr. David N. Osser, a psychiatry professor at Harvard University, said yesterday. “The weight gain itself is a known risk factor for diabetes.”

The February 2000 memo was prepared as background for a meeting ofLilly scientists to the possible changes for Zyprexa’s label.

According to the memo, Lilly scientists initially wanted to proposea relatively straightforward statement on the label that high bloodsugar had been observed in patients taking Zyprexa in clinical trials.That change was never made.

Lilly’s analysis in early 2000 came at a time when some doctors andregulatory agencies were beginning to question whether Zyprexa couldcause increases in blood sugar or diabetes. Although Lilly says that nolink between Zyprexa and diabetes has ever been proven, the AmericanDiabetes Association found in 2004 that Zyprexa was more likely tocause diabetes than other, similar drugs.

Zyprexa is by far Lilly’s best-selling product, with $4.2 billion insales in 2005, which represented 30 percent of Lilly’s overall revenue.Zyprexa’s active ingredient is a potent chemical that binds toreceptors in the brain to reduce the hallucinations and delusionsassociated with schizophrenia and acute bipolar disorder. About two million people worldwide took Zyprexa last year.

At the February 2000 meeting for which the memo was prepared, theagenda was to discuss Zyprexa’s tendency to cause high blood sugar,which is medically known as hyperglycemia.

According to the memo, Lilly had reviewed data from its clinicaltrials and found that “the incidence of treatment-emergenthyperglycemia in olanzapine group (3.6%) was higher than that in theplacebo group (1.05%).” Olanzapine is the generic name for Zyprexa.

But when Lilly subsequently discussed the clinical trial resultswith doctors, it used a different comparison. Lilly told doctors thatZyprexa had caused 3.1 percent of patients — not 3.6 percent — to havehigh-blood sugar. And it said that 2.5 percent of patients on theplacebo — not 1.05 percent — had high-blood sugar. As a result, therates of high blood sugar in the two groups seemed almost identical inthe revised data.