It is unknown whether the type and grade of a primary endometrial carcinoma is reliably maintained in recurrence. All matched primary and recurrent endometrial carcinomas diagnosed from 2000 to 2010 at our institution were identified; 34 cases had available slides. Histologic classification was performed using modifications to the World Health Organization criteria. Immunohistochemical analysis for p53, p16, progesterone receptor (PR), and DNA mismatch-repair proteins (MMR) (MLH1, MSH2, MSH6, and PMS2) was performed. Endometrioid carcinoma recurrences were mostly local, whereas serous carcinoma recurrences were mostly peritoneal. Compared with endometrioid carcinoma patients, serous carcinoma patients were older, presented at high stage, and had shorter survival. Serous carcinomas were the most common recurrent endometrial carcinoma (18/34 cases). Overall, 21 cases (62%) displayed similar morphology when comparing primary and recurrent carcinomas, whereas 13 displayed discordant morphology. Seven of 13 endometrioid carcinomas (54%) had a morphologically discordant recurrence, compared with 3 of 14 serous carcinomas (21%), 1 of 4 morphologically ambiguous carcinomas (25%), and both mixed epithelial carcinomas. Serous and morphologically ambiguous carcinomas therefore demonstrated relative morphologic fidelity compared with endometrioid carcinomas. Four morphologically discordant cases demonstrated either pure clear cell carcinoma or clear cell features at recurrence. Seven of 23 matched pairs displayed discordant PR results, with 5 cases, including both endometrioid and serous carcinomas, showing diminished PR expression at recurrence. p53, p16, and DNA MMR staining results were generally concordant when evaluating matched pairs, with only occasional exceptions. Sixty-four percent of all pure endometrioid carcinomas and mixed epithelial carcinomas with an endometrioid component showed loss of expression of MLH1 and/or PMS2; no serous carcinoma demonstrated this abnormality. Clinical and immunohistochemical data supported the use of modifications to the World Health Organization diagnostic criteria. More importantly, the data suggest that when confronted with recurrent endometrial carcinoma, particularly a serous carcinoma, it is reasonable to base therapeutic decisions on the type of the primary tumor, especially if sampling or excising the recurrent tumor is problematic. However, in light of the PR results, sampling a recurrent endometrioid carcinoma may be worthwhile if hormonal therapy is planned. Recurrent endometrioid carcinoma may be enriched for tumors with DNA MMR abnormalities.