[Alcohol, breast cancer and causal inference in epidemiology]

In order to demonstrate how possible causal relationships are critically evaluated in epidemiologic research, literature on the association between alcohol and breast cancer is reviewed and discussed. A cause can be defined as a factor which, in combination with other factors, known and unknown, is sufficient to produce an effect. Since the hypothesis-generating study was published in 1977, a total of 34 positive and 16 negative studies have been published. Methodological problems, such as chance, bias and confounding, cannot be considered as plausible explanations for the above majority of positive findings. The question of causality was then evaluated using the guidelines developed by Bradford Hill in 1965. Among these, the strength of the association, consistency, temporality, biological gradient and biological plausibility, are the most important. In spite of the relatively weak association and somewhat inconsistent results, it is concluded that alcohol consumption should be considered as a cause of breast cancer. It is estimated that in Norway, between 24 and 180 cases of breast cancer may be attributed to alcohol consumption. Future research should focus on the question of effect-modification and on the possible implications of different patterns of alcohol consumption.