It would be very useful if it were possible to identify, within an apparently normal population, those who will develop breast cancer. Although this can be achieved by genetic testing for mutations of BRCA1 and BRCA2, such susceptibility genes are involved in the aetiology of only about 5% of breast cancer cases (Lynch and Lynch, 1986). By contrast, events related to a woman's menstrual and reproductive history, which have been widely recognized as determinants of breast cancer incidence, are not impressive predictors of individual risks. Such poor predictive performance could result from these epidemiological variables being only manifestations of other underlying factors. Indeed the size of the effects of the menstrual and reproductive variables and the observation that breast cancer rates change over a woman's life have stimulated investigation of endogenous sex hormones as potential causes underlying these observed effects (Key and Pike, 1988;Toniolo et al, 1995;Key et al, 1996). Methodological issues related to hormone measurements, however, complicate the study of their relationship with breast cancer. Furthermore, there is no general consensus regarding which biological forms of these hormones are actually associated with the observed changes in incidence rates.The role of steroids in the aetiology of breast cancer is uncertain. Extensive analysis of the scientific literature shows a highly significant increase in blood oestradiol levels in post-menopausal women with breast cancer (Thomas et al, 1997). The position of oestrogens and premenopausal women is uncertain (Key and Pike, 1988; Thomas et al, 1997). Androgens have been studied less extensively than oestrogens and no meta-analysis has yet been undertaken.The first prospective study in this field was performed by Bulbrook and Hayward who showed that levels of androsterone (A) and aetiocholanolone (E) were abnormally low in women who subsequently developed breast cancer (Bulbrook and Hayward, 1967;Bulbrook et al, 1971). Most of the cases were premenopausal and the abnormality was present up to 11 years before the clinical diagnosis of the disease, suggesting that high levels of androgen metabolites might have a protective effect, at least in younger women. However, other mainly retrospective casecontrol studies have claimed that high androgen excretion is positively associated with breast cancer risk (e.g. Secreto and Zumoff, 1994). Such contrasting results could be reconciled if the effects of androgens depended on menopausal status. This would occur if the androgens acted as oestrogen inhibitors in premenopausal women and as oestrogen enhancers (or even oestrogens) in the postmenopausal (Secreto and Zumoff, 1994;Adams, 1998 Summary Between 1961 and 1967 a cohort of over 5000 women volunteered for a prospective study to determine the relationship between the urinary androgen metabolites, androsterone (A) and aetiocholanolone (E), and risk of breast cancer. During the first 10 years of the study the concentration of urinary A and E was determined in 1...