Women who took at least two alcoholic drinks a day were three times more likely than abstainers to die before age 65, a study found.

The difference was nearly two-fold for men in the study, which was based on a national sample of people who died in 1986.For women, 40.7 percent of so-called heavier drinkers died before age 65 versus 13.2 percent of abstainers. For men, the figures were 42.3 percent versus 22.4 percent.

Even people classified as moderate or light drinkers showed higher death rates than the abstainers.

The numbers cannot be interpreted as indicating risks of early death because of the way the study was done, cautioned co-author Darryl Bertolucci.

The study began with people who were already dead, and it worked backward to classify them according to their drinking habits, rather than starting with people of known habits and following them to find their risk of early death.

Nonetheless, the study "does show a relationship that needs to be looked into," he said.

Bertolucci, a statistician at the National Institute on Alcohol Abuse and Alcoholism, presented the results Monday at the annual meeting of the American Public Health Association.

The study used data on the drinking habits during adulthood of 8,303 men and 7,158 women, ages 25 or older, obtained several months after their deaths from next of kin or other knowledgeable sources.

Those who had fewer than 12 drinks a year were classified as abstainers. Light drinkers were those who had averaged up to three drinks a week, moderate drinkers were those taking four to 13 drinks a week, and heavier drinkers were those who took two or more drinks a day.

Among men, 33 percent of moderate drinkers and 32 percent of light drinkers died before age 65. For women, the figures were 29 percent and 26 percent respectively.

As expected, cirrhosis of the liver was a more frequent cause of death in men for heavier drinkers than lighter drinkers or abstainers. Cancers of the trachea, bronchus, lung and esophagus increased with greater drinking for both sexes.