Early phase chemoprevention trials differ from standard therapeutic clinical trials because asymptomatic, healthy people are treated with a potentially toxic intervention for a prolonged period of time. Current subject selection protocols have relied upon epidemiological methods to identify high-risk individuals. Most available data provide risk estimates for various individual risk factors, but few have reported risk estimates for combinations of risk factors. Selection criteria for the large tamoxifen intervention trial (NSABP P1) were developed from the work of Gail et al. [1]. The Gail model takes into account non-genetic factors (e.g., nulliparity, age at menarche, preexisting pathological conditions) and genetic factors (family history). Using a lifetime risk of 10% of developing breast cancer as a standard to intervene, NSABP P1 uses the Gail algorithm to select pre- and postmenopausal women for a primary intervention trial. This approach has been criticized for being insufficiently selective (i.e., all women > or = 60 yrs), but appears to be the best available method to select subjects for a chemoprevention trial. Other approaches have been based on identification of very high-risk women with acknowledged pathologic conditions [lobular carcinoma in situ, ductal carcinoma in situ (DCIS)]. Attempting to use these proliferative lesions as pathologic endpoints for drug effect has not been attempted. DCIS as a risk factor for tamoxifen intervention was excluded because of controversies over its management and because of frequent difficulties in distinguishing microinvasive from non-invasive lesions. Women treated for early stage breast cancer (Stage I) may be subjects for early stage chemopreventive interventions.(ABSTRACT TRUNCATED AT 250 WORDS)