Characterization of breast tumors on both estrogen receptor (ER) and progesterone receptor (PR) status suggests distinct biological and clinical profiles. We hypothesized that these tumor subtypes might also show specific differences in their relations with epidemiologic risk factors. Risk factors were assessed via a questionnaire mailed in January 1986 to 37,105 cancer-free women, ages 55-69 years: the Iowa Women's Health Study. To the end of 1992 (241,627 person-years of follow-up), 939 incident breast cancers were ascertained by the Iowa population-based Surveillance, Epidemiology, and End Results Cancer Registry. Joint ER and PR status was determined on a total of 610 (65%) tumors. Three patterns of association were seen in relation to epidemiologic risk factors. Endogenous hormone exposure variables--parity, age at first birth, age at menarche, body mass index, and body fat distribution as defined by waist-to-hip ratio--showed their expected pattern of associations only with PR+ breast cancers. In age- adjusted and polychotomous logistic regression analyses, both ER-PR- and ER+PR- breast cancers showed evidence of an inverted pattern of associations with several risk factors compared with that seen for ER+PR+ cancers [including parity (ER-PR-), waist-to-hip ratio (ER-PR-), body mass index (both), body mass index at age 18 years (ER-PR-), history of bilateral oophorectomy (ER+PR-), and oral contraceptive use (ER+PR-)]. Family history was not associated with ER+PR- cancers; only 8 (8%) of 99 patients with this subtype had a family history of breast cancer compared with 16% of all other types combined.(ABSTRACT TRUNCATED AT 250 WORDS)