OBJECTIVE: The endometrial origin of uterine carcinosarcoma has recently been well established. The current study investigates whether uterine carcinosarcomas can be included in protocols on high-risk endometrial cancer, given the similarities in biologic behavior of both entities. METHODS: Pathological and surgical notes of patients diagnosed with grade 3 endometrioid, carcinosarcoma, serous and clear cell endometrial cancer subtypes were retrospectively analyzed with special attention to the spread pattern of the different subtypes. Information on site of relapse and time to recurrence was obtained. RESULTS: We traced 146 patients of which 9 patients were ineligible. Histological subtypes of the remaining 137 patients were as follows: 50 (37%) grade 3 endometrioid carcinoma, 54 (39%) serous or clear cell carcinoma (non-endometrioid carcinoma), and 33 (24%) carcinosarcomas. Distribution of early stage disease (I and II) was 67, 46, and 78% for grade 3 endometrioid, non-endometrioid, and carcinosarcoma, respectively. Although we could not trace differences in hematogenic and transperitoneal spread among the three subtypes, non-endometrioid and carcinosarcomas were more likely to spread to pelvic and paraaortic lymph nodes (P < 0.01). Using univariate analysis, both stage (P < 0.006, Wald statistic) and histological type appear to determine the outcome, whereas lymphovascular space infiltration (P < 0.25) and age (P < 0.07) were not significantly different between the three histological subtypes. Cox Regression multivariate analysis on 127 women suffering from the three histological subtypes suggested that both stage III-IV disease (P < 0.00001) and histological type (carcinosarcoma) (P < 0.003) were of prognostic significance [hazard ratio (CI 95%) were, respectively, 3.8 (2.1-7.0) and 3.2 (1.7-5.9)]. Analyzing cases limited to stage I-II endometrial cancer, 24/28 (86%) grade 3 endometrioid, 18/24 (75%) non-endometrioid, and 11/25 (44%) carcinosarcomas survived, suggesting a worse outcome for endometrial carcinosarcoma when compared to the other subtypes (P < 0.008, Log Rank). A higher incidence of pulmonary metastases explained the worse outcome for early stage carcinosarcoma (P < 0.006), whereas the incidence of liver metastasis, transperitoneal spread, or recurrences in lymph nodes or vagina were comparable between the three pathologic subtypes. CONCLUSIONS: Although endometrial carcinosarcoma originates from epithelial cancer, the intrinsic more aggressive tumor biology suggests that this subtype should not be incorporated in studies on high-risk epithelial endometrial cancer.