A Food and Drug Administration panel split on whether it thought the agency should approve liraglutide, a proposed diabetes drug from Novo Nordisk Inc. (NVO).

The panel voted 6 to 6 with one abstention on whether data on a certain type of thyroid tumors seen in rodent studies permitted marketing of the product. However, the panel said a small number of a common type of thyroid tumor found in humans in clinical studies were likely found because study investigators were looking for thyroid problems, and 12 panel members said that alone shouldn't stop marketing of the drug. One person abstained from voting.

Mary Parks, the director of FDA's division of metabolism and endocrinology products, said the agency faced a "difficult task" in deciding whether to approve the product given the panel's divergent votes.

Earlier, most panel members said the product wasn't linked to an excess in heart attacks and strokes, a key concern the FDA has with diabetes drugs.

Specifically, the panel voted 8 to 5 on a question that asked whether the liraglutide data was enough to rule out an "unacceptable excess cardiovascular risk relative to competitors."

Liraglutide is Novo Nordisk's most important drug in development. It belongs to the same class of drugs as Byetta, which is sold by Eli Lilly & Co. (LLY) and Amylin Pharmaceuticals Inc. (AMLN).

The FDA said it saw a low cardiovascular event rate in clinical studies of liraglutide but raised concerns about a small number of patients who developed a type of thyroid cancer during clinical studies as well as tumors that were seen in animal studies. The agency said it didn't have data to "dismiss" the relevance of the rodent studies relative to humans.

Shares of Novo Nordisk fell 4.3% to $47.92 during the regular NYSE session Thursday and were at $47.20 in recent late trading.

-By Jennifer Corbett Dooren, Dow Jones Newswires; 202-862-9294; jennifer.corbett@dowjones.com