U.S. health advisers said GlaxoSmithKline Plc's diabetes drug Avandia raised a heart attack concern compared with diabetes drugs in other classes and a rival drug in the same class.
A final vote on whether to recommend Avandia's withdrawal from the market is still to come.
Avandia's U.S. sales were just 1.5 percent of Glaxo's 2009 revenues, but investors fear the company could face more lawsuits if the drug is pulled from the market.
The early votes taken Wednesday were on Avandia's risks versus other classes of diabetes drugs and a rival drug Actos, a medicine in the same class as Avandia, made by Takeda Pharmaceutical Co.
The Food and Drug Administration convened the 33-member panel of outside experts to help the agency settle a three-year safety dispute over Avandia.
Eighteen members of the panel saw a heart attack concern versus other classes of diabetes drugs. Twenty-one members saw a heart attack concern versus Actos.
It is difficult to predict how early votes may affect the final vote.
In 2007, an advisory panel voted 20-3 that Avandia may increase the chances of a heart attack for some patients, but by 22-1 recommended that the drug stay on the market.
The current 33-member panel of scientific experts heard two days of sharply conflicting opinions and reviewed hundreds of pages of data on whether Avandia causes heart attacks.
Glaxo shares were trading up 0.2 percent in New York having closed up 0.4 percent in London.
The advisory panel is due later Wednesday to vote on recommendations ranging from keeping the drug on the market with no warnings, to urging a withdrawal. Restrictions on use or beefing up warnings are among other choices.
Because the FDA asked panelists to select only one of the five options, there may be no clear majority. However, any vote short of withdrawal would back continued Avandia sales in some form.
The agency will make the final call in the coming months but usually follows the advice of its panels.
The debate on Avandia's safety has raged since warnings were placed on the drug in 2007 saying some research linked the drug to a higher heart attack risk but the data is inconclusive.
(Reporting by Lisa Richwine and Susan Heavey; Editing by Tim Dobbyn)