The well-publicized hypothesis that consumption of methylxanthines (e.g., caffeine, theobromine, and theophylline) may increase the risk of breast cancer in women has not been confirmed by most epidemiologic studies, which have reported no effect or only a small increase in breast cancer risk with caffeine. Cohort studies among Seventh-Day Adventists' revealed no association between coffee and the risk of fatal breast cancer. A Norwegian cohort study' also reported no association between coffee consumption and subsequent diagnosis of breast cancer. Most previous casecontrol studies have failed to identify a positive association with breast cancer, whether they assessed only coffee or tea consumption'3' or methylxanthine intake.'-' A case-control study by Lawson8 reported a slight increased risk of breast cancer with the consumption of coffee and tea, but the increased risk was not consistent with varying levels of consumption. In general, however, total methylxanthine consumption has been assessed primarily through the consumption of coffee and tea, leaving unresolved the question of whether other forms of methylxanthines might influence the risk of breast cancer. Given that caffeine is one of the most widely consumed drugs, and given the broad availability of theobromine in Western diets, even a slight increase in risk of breast cancer would translate potentially into an excess of breast cancer morbidity. This report further explores the relationship between methylxanthines and breast cancer based on a large case-control study of 3,234 women, conducted in 18 contiguous counties in eastern New York State (United States). As previously described by Nasca et al,9 1,617 patients were selected from among female residents of the study area, ages 20 to 79, who were diagnosed with a primary, incident breast cancer between 1 April 1982 and 31 March 1984. Breast cancer patients were identified through the diagnostic index, tumor registry, and pathology files maintained by each hospital, as well as the New York State Cancer Registry. An equal number of controls were frequencymatched to the interviewed cases on year of birth and