Arena Pharmaceuticals Inc.'s experimental diet pill may cause cancer and works only by a "slim margin," regulators said, sending shares down the most ever.

The medicine, lorcaserin, was linked to "a number of malignant tumor types" in studies of rats, Food and Drug Administration staff said in a report released Tuesday. The drug helped people lose 3 percent more weight than a placebo, short of the recommended 5 percent, the agency said.

Arena and partner Eisai Co. are racing against Mountain View's Vivus Inc. and Orexigen Therapeutics Inc. to introduce new treatments for the one-third of Americans who are obese. Lorcaserin is similar to fenfluramine, a component in the diet pill fen-phen that was withdrawn from the U.S. market in 1997 because of links to heart-valve damage.

"The big surprise to us here is the cancer thing," said Jon Lecroy, an analyst at Hapoalim Securities in New York, in a telephone interview Tuesday. "Knowing the drug doesn't work that well, I knew the safety concerns would be looked at extremely closely."

Arena fell $2.72, nearly 40 percent, to close at $4.13 on the Nasdaq Stock Market. The shares were down 42 percent at their lowest point during the day, their biggest decline since the stock began trading in July 2000. Before Tuesday, shares in the San Diego company had almost doubled in price so far this year.

An FDA advisory panel will review the staff findings Thursday in Adelphi, Md., ahead of the agency's scheduled decision on approval next month.

Lorcaserin is Arena's first product and has the proposed brand name Lorqess. Peak global sales may reach $800 million in 2015 if the drug is approved, Lecroy said. Eisai bought rights to sell the drug on July 1 and plans to give about a third of the revenue to Arena.

According to FDA recommendations released in 2007, medical products designed for weight management should help at least 35 percent of people lose at least 5 percent of their weight and double the proportion of the placebo-treated group, or they should work 5 percent better than placebo. Lorcaserin only barely satisfied the first criteria, and didn't meet the second, the agency said.

"Lorcaserin did not satisfy the guidance's mean efficacy criterion," said Eric Colman, deputy director of the FDA's Division of Metabolism and Endocrinology Products, in a memo to the advisory panel included in the staff report. The proposed 10-milligram, twice-daily dose "did, by a slim margin, satisfy the categorical efficacy criterion."