Women with gene mutations linked to high rates of breast cancer may actually lower their risk of the disease by smoking cigarettes, according to a study published Wednesday.

But researchers cautioned that smoking poses far greater risks from other tobacco-related diseases, including other cancers."This study is interesting scientifically, but it should not encourage anyone to smoke," said Jean-Sebastien Brunet, a researcher at the Women's College Hospital of the University of Toronto in Canada. He was lead author of a study published Wednesday in the Journal of the National Cancer Institute.

The study examined the breast cancer history of 372 women who had mutations of the BRCA1 or BRCA2 genes. By some estimates, about 80 percent of such women will develop breast cancer.

Half the women in the study were smokers.

Brunet said the incidence of breast cancer was 54 percent lower among heavy smokers than among nonsmokers. The effect, he said, was "dose-related" - the more a woman with a BRCA gene mutation smoked, the less likelihood of her developing breast cancer.

Smoking a pack a day for four years or four packs a day for a year reduced the risk by 35 percent, he said. The reduction was 54 percent for a woman who smoked more.

The study involved only women with the BRCA gene mutation. This mutation occurs, on average, in only one of every 250 women. Among some ethnic groups, the rate can be as high as one in 50. Between 5 percent and 10 percent of all women with breast cancer have a BRCA mutation.

Brunet said the study was scientifically valuable because it suggests that some action of smoking or of some of the 1,300 compounds in cigarette smoke may be protective against breast cancer. However, he said, "We don't have any idea what that compound is."

Some breast cancers have been linked to estrogen, the female hormone, and cigarette smoking is known to lower production of estrogen, said Brunet. Smoking also is linked to early menopause and to a decreased risk of endometrial cancer, Brunet said.

However, smoking significantly increases the risk of other, even more dangerous cancers, such as those of the lung, throat and pancreas.

Brunet said he and his 18 co-authors came to their conclusions reluctantly and only after putting the study results to rigorous statistical tests. Further, Brunet said that scientific referees on the journal also carefully scrutinized the data.

"We did everything we could to test the data, but we would really like for someone to replicate the study just to prove that our data set is correct," he said.