Atypical proliferative (borderline) endometrioid tumors (APTs) and well-differentiated endometrioid carcinomas of the ovary constitute a spectrum of morphologically diverse proliferative tumors. There is currently no agreement on the criteria for distinguishing them. We report the clinicopathologic features of 56 proliferative endometrioid tumors focusing on the criteria for invasion, the clinical significance of microinvasion and cytologic atypia, and prognosis. Endometriomas, adenofibromas, adenosarcomas and moderately to poorly differentiated carcinomas were excluded, as were patients with concurrent endometrioid carcinoma of the endometrium. The tumors were classified as atypical proliferative tumor (APT) (33 tumors), APT with intraepithelial carcinoma (high-grade cytology in a tumor lacking stromal invasion) (three tumors), APT with microinvasion (invasion <5 mm) (five tumors), and invasive carcinoma (invasion > or = 5 mm) ( 15 tumors). All tumors were confined to the ovary (stage I). In 50 patients, the tumor involved one ovary, and in three patients, the tumors were bilateral. The predominant growth pattern was adenofibromatous in 29 tumors and glandular or papillary in 27 tumors. In 8 (24%) of 41 APTs, areas of benign adenofibroma were identified, and in 13 (87%) of 15 carcinomas, areas of associated APT were identified. Stromal invasion was manifested by confluent glandular growth in all 15 invasive carcinomas and all tumors with microinvasion. Destructive infiltrative growth was also present in 2 (13%) of 15 carcinomas. Confluent glandular growth was the most common manifestation of stromal invasion and therefore served as the best criterion for the diagnosis of carcinoma. Squamous differentiation was observed in 24 tumors, and mucinous differentiation was seen in 20 tumors and was most often seen in APTs. Endometriosis was present in 14 patients with APTs and one patient with carcinoma. Four patients had hyperplasia or atypical hyperplasia of the endometrium. One patient with an APT had a concurrent peritoneal serous neoplasm. Twenty-one patients had available clinical follow-up. Twenty (95%) of 21 patients, including six with invasive carcinoma, two with microinvasion, one with intraepithelial carcinoma, and 11 with APT were alive with no evidence of disease with a mean follow-up of 47 months. One patient with carcinoma had recurrent tumor after 46 months and was alive 40 months after resection of the recurrent tumor. In this large series of proliferative endometrioid tumors, all were stage I and only one patient had a recurrence. Most carcinomas contained evidence of a precursor APT, and in some APTs, an associated benign adenofibroma was identified. Microinvasion or intraepithelial carcinoma occurred in 19% of APTs. This finding likely reflects the various stages of endometrioid carcinogenesis in the ovary. For clinical management, we suggest that these tumors be divided into two categories-APTs and well-differentiated carcinoma-because based on the available data, cytologic atypia and microinvasion appear not to affect the prognosis.