The role of estrogen as a promoter agent of sporadic breast cancer has been considered by assaying, in benign breast disease (BBD) and in situ carcinomas (CIS), 2 markers, the estrogen receptor alpha (ERalpha) and cathepsin D (cath-D) involved in estrogen action on mammary tissue. ERalpha and cath-D were assayed by quantitative immunohistochemistry using an image analyzer in 170 lesions of varying histological risk (94 BBD and 76 CIS), and in "normal" glands close to these lesions. The ERalpha level increased significantly in proliferative BBD with atypia (P < .001), in non-high-grade CIS (P < .001), and in adjacent "normal" glands. ERalpha level was decreased in high-grade ductal CIS (DCIS) and also in adjacent "normal" glands. Cath-D level increased in ductal proliferative BBD (P < or = .01) and in high-grade DCIS (P < or = .003), but not in the other lesions. After menopause, ERalpha level was increased (P = .012) but not cath-D level. According to Mac Neman test, the high-grade DCIS were predominantly ERalpha negative and cath-D positive (P = .0017), and the other CIS were predominantly ERalpha positive and cath-D negative (P = .0002). The 2 markers are overexpressed early in premalignant lesions, but independently. This dissociation suggests a branched model of mammary carcinogenesis involving 1 estrogen-independent pathway with high cath-D and low ERalpha levels (including high-grade DCIS) and 1 estrogen-dependent pathway, with high ERalpha level (including proliferative BBD with atypia and low-grade DCIS). We propose that ERalpha-negative breast cancers may develop directly from high-grade DCIS and that ERalpha assay in preinvasive lesions should be considered in prevention trials with antiestrogens.