Endocrine ductal carcinoma in situ (E-DCIS), first characterized by Cross et al. in 1985, is an uncommon entity, and there is little information on its pathobiologic features and natural history in the literature. This report describes the largest series of 34 cases: 14 cases were pure in situ (group A), and 20 were accompanied by an invasive component (group B). All except three patients were over the age of 60 years, with the mean being 69.5 years for group A and 72.6 years for group B. Except for six patients in group A who had nipple discharge, all had a breast mass. On follow-up, one of five group A patients developed local recurrence 5 years after mastectomy, and two of seven group B patients developed another invasive primary in the contralateral breast. Histologically, E-DCIS showed expansile intraductal growths forming solid sheets and festoons traversed by delicate fibrovascular septa. Accumulation of basophilic mucin might be found within the growth and the fibrovascular septa. There were variable degrees of stromal sclerosis. In some cases, the solid intraductal cellular proliferations were focally punctuated by microglandular spaces and rosettes. Comedo necrosis was absent. Intraductal papillomas were found in the immediate vicinity of the tumors in 18 cases and invariably showed pagetoid involvement by E-DCIS. Pagetoid spread into the adjacent ducts and ductules was also a common feature (17 cases). The tumor cells were polygonal, oval, or spindly, often with eccentrically placed, bland-looking, ovoid nuclei and abundant eosinophilic granular cytoplasm. Intracellular mucin was commonly demonstrable. Immunostaining for myoepithelium using muscle-specific actin antibody confirmed the in situ nature of the E-DCIS component. The majority of tumor cells showed strong staining with the neuroendocrine markers chromogranin, synaptophysin, and neuron-specific enolase (monoclonal). Immunostaining also dramatically highlighted the pagetoid spread into the papillomas and ductules by outlining the tumor cells between the negatively stained residual ductal epithelium and myoepithelium. All cases were immunoreactive for estrogen and progesterone receptor, but not p53 and c-erbB2. The Ki-67 index was < 5%. Ultrastructural studies on four cases showed many dense-core neurosecretory granules and larger mucigen granules. In group B cases, the invasive component, which comprised 5-95% of the tumor, included colloid carcinoma, 12; "carcinoid" tumor, 3; mixed "carcinoid"/colloid carcinoma, 4; and small cell neuroendocrine carcinoma, 1. Neuroendocrine markers were also consistently demonstrable in the invasive component. In conclusion, E-DCIS is predominantly a disease of older women that is frequently accompanied by papillomas in the vicinity and may present as nipple discharge (an uncommon presentation in the usual forms of DCIS). It can mimic epitheliosis histologically, but the pagetoid spread is a helpful clue to its neoplastic nature. The bland nuclear morphology, lack of necrosis, and biologic marker profile suggest that E-DCIS is a form of low-grade DCIS despite its solid growth pattern. The invasive carcinomas associated with E-DCIS are also neuroendocrine programmed rather than the usual types of ductal carcinomas, suggesting that E-DCIS represents a biologically distinctive category of DCIS.