Microglandular adenosis (MGA) of the breast is an uncommon, benign lesion that may mimic invasive carcinoma and has recently been recognized as having significant premalignant potential. When carcinomas arise in MGA, there is often a transition from ordinary MGA to atypical MGA (AMGA) to carcinoma. Nineteen cases of carcinoma arising in MGA are reported: 7 invasive carcinomas, 7 intraductal carcinomas (DCIS), and 5 with both invasive and intraductal carcinoma. A single case of AMGA without carcinoma is also reported. The 20 patients ranged in age from 36 to 81 years (mean 52). The most common clinical presentation was either a palpable mass (13 patients) or a mammographic abnormality (4 patients). All 20 cases contained AMGA, and in some cases AMGA was the predominant lesion. In 18 of the 19 cases with carcinoma, there was a clear transition from AMGA to the carcinoma. Twelve cases contained ordinary MGA, but in only 2 cases was MGA a prominent component of the lesion. In contrast to ordinary MGA, the glands of AMGA were more irregularly shaped, closely packed, and cytologically atypical and tended to lack secretions. A solid, occlusive proliferation of cells in the tubules was seen in 10 cases. All 12 examples of in situ carcinoma were either grade 2 or 3 and typically showed a solid proliferation of severely atypical cells within the glands; a cribrifrom pattern was also present in 1 case. The invasive carcinomas were morphologically diverse and included 2 with a basaloid morphology and 2 metaplastic carcinomas. Various immunostains were performed, and each lesion (AMGA, in situ, and invasive carcinoma) was separately assessed for immunoreactivity. As expected, S-100 was positive in the vast majority of AMGA and in situ carcinomas and in all 12 invasive carcinomas. S-100beta was also positive in the majority of cases although the staining was weaker. Laminin and type IV collagen highlighted the basement membrane around the AMGA and in situ carcinoma and are useful stains in difficult cases. Except for a single case, ER and PR were negative in all lesions. Cytokeratin 7 (CK 7) was positive, while cytokeratin 20 (CK 20) was negative in all cases. Immunostains for CK903 showed no reactivity in any of the invasive carcinomas, in situ carcinomas, or atypical MGA but was focally present in the associated MGA in 2 of the 8 cases studied. Immunostains for MIB-1 and p53 were semiquantitatively assessed and both were positive in AMGA but tended to show a more intense staining in the carcinomas. Five cases were also studied for immunoexpression of alpha-1 antitrypsin (AAT), alpha-1 antichymotrypsin (ACTP), lysozyme, and salivary gland amylase. All 5 invasive carcinomas were positive for ACTP, though the staining was very focal in about 10% of the cells in a basaloid carcinoma. The in situ carcinoma as well as the AMGA in 4 of the 5 cases were positive for ACTP. Three of the 5 invasive carcinomas were positive for AAT in 10% to 40% of the cells. The most intense positivity for AAT and ACTP was in cells with coarsely granular apocrine appearance evident in 2 of the 5 cases. Four of the 5 invasive carcinomas were positive for lysozyme in 10% to 50% of the cancer cells; the in situ carcinoma and the associated AMGA showed similar immunoreaction in each case. None of the 5 cases showed convincing positivity for salivary gland amylase. The MGA in all 5 cases was negative for AAT and ACTP; the MGA in 1 of the 5 cases was positive for lysozyme. This study confirms the potential of MGA to develop into an invasive carcinoma, more clearly defines the features of AMGA, highlights the importance of AMGA in the evolution of carcinoma from MGA, and expands our knowledge of the immunophenotype of AMGA and the carcinomas arising from it. The diagnostic criteria briefly noted previously for diagnosis of AMGA and carcinoma arising in MGA are expanded and formally proposed. Int J Surg Pathol 8(4):303-315, 2000