OBJECTIVE: The aim of this study was to determine pathologic variables associated with disease-specific survival of node-positive patients with endometrial carcinoma treated with combination of surgery including pelvic and para-aortic lymphadenectomy and adjuvant chemotherapy. METHODS: Survival of 55 node-positive endometrial carcinoma patients prospectively treated with surgery and adjuvant chemotherapy between 1982 and 2002 at Hokkaido University Hospital was compared to various histopathologic variables. All patients underwent primary surgical treatment including pelvic and para-aortic lymphadenectomy followed by adjuvant chemotherapy consisting of intravenous cisplatin, doxorubicin, and cyclophosphamide. Survival analyses were performed by the Kaplan-Meier curves and the log-rank test. Independent prognostic factors were determined by multivariate Cox regression analysis using a forward stepwise selection. RESULTS: Among 303 consecutive endometrial cancer patients treated during the period of this study, 55 patients (18.2%), including 44 without peritoneal metastasis (FIGO stage IIIc) and 11 with peritoneal metastasis (FIGO stage IV), were found to have retroperitoneal lymph node metastasis. Multivariate Cox regression analysis revealed that peritoneal metastasis and lymph-vascular space invasion (LVSI) were independently related to poor survival in node-positive endometrial carcinoma. The estimated 5-year survival rate of stage IIIc patients with or without moderate/prominent LVSI was 50.9% and 93.3%, respectively with statistically significant difference (P=0.0024). The estimated 5-year survival rate of stage IV patients was 20.0%. Prognosis of stage IIIc patients could be stratified into three groups by the number of positive para-aortic node (PAN) with an estimated 5-year survival rate of 86.4% for no positive PAN (n = 23), 60.4% for one positive PAN (n = 13), and 20.0% for > or = 2 positive PAN (n = 8). The difference of survival rate between no or one positive PAN and > or = 2 positive PAN was statistically significant (P = 0.0007 for no positive PAN vs > or = 2 positive PAN, P = 0.0319 for one positive PAN vs > or = 2 positive PAN). Multivariate analysis including number of positive PAN groups showed that LVSI, number of positive PAN groups were independent prognostic factors for survival. Survival of patients with stage IIIc disease could be stratified into three groups by combination of LVSI and number of positive PAN groups with an estimated 5-year survival rate of 93.3% for no or one positive PAN group with nil or minimal LVSI, 62.6% for no or one positive PAN group with intermediate or prominent LVSI, and 20.0% for > or = 2 positive PAN groups irrespective of LVSI (P = 0.0002 for no or one positive PAN group with nil or minimal LVSI vs > or = 2 positive PAN groups, P = 0.0223 for no or one positive PAN group with nil or minimal LVSI vs no or one positive PAN group with intermediate or prominent LVSI, P = 0.0388 for no or one positive PAN group with intermediate or prominent LVSI vs > or = 2 positive PAN groups). CONCLUSIONS: LVSI and number of positive PAN groups were independent prognostic factors for stage IIIc endometrial cancer patients. Postoperative therapy and follow-up modality need to be individualized according to LVSI and the number of positive PAN for stage IIIc patients. New molecular markers to predict the prognosis of endometrial cancer patients preoperatively should be found for individualization of treatment. New chemotherapy regimen including taxane needs to be considered as an adjuvant therapy for patients with node-positive endometrial cancer.