Women who drink are at greater risk of contracting breast cancer than teetotallers, according to a pioneering Oxford scientist and his research team.

In a study published yesterday, Sir Richard Doll, who first proved the links between smoking and lung cancer, revealed that a woman's risk from the disease increases by six per cent with every alcoholic beverage she drinks per day.

A woman who drinks less than half a unit a day -- or the equivalent per week -- would be less likely to get breast cancer.

Study co-author Prof Valerie Beral, of Cancer Research UK's epidemiology unit, based at the Radcliffe Infirmary, in Woodstock Road, said: "This research tells us there is a definite link between alcohol and breast cancer, and the evidence suggests that the more a woman drinks, the greater her risk.

"The impact of drinking on breast cancer is small compared to child-bearing factors, but women are drinking far more than they used to and if this pattern continues it is bound to have an impact on the rates of the disease in the future."

The worldwide study showed that although smoking causes a third of all cancers, it has no links to breast cancer. But four per cent of all cases of the disease are alcohol-related, including 2,000 cases a year in the UK.

The researchers expressed concern that women's alcohol consumption had already increased in the last decade, and the proportion of women aged 16 to 24 who drank three units a day had doubled from nine per cent to 18 per cent.

The researchers warned that if this increase continued, breast cancer rates were likely to rise.

Prof Beral said: "We are not intending to shock people with our results. Drinking is actually good for some things, like heart disease, so the message is to drink in moderation."

In the past, researchers have found it difficult to separate the effects of tobacco and alcohol on breast cancer, because the more women drink, the more they tend to smoke.

But the Oxford research combined results from more than 50 studies, enabling the scientists to make the most accurate estimates ever.