The goal of this study was to determine diagnostic accuracy of preoperative transvaginal sonography (TVS) to assess myometrial infiltration in patients with endometrial cancer and to determine the possibility of preoperatively selecting low-stage endometrial cancer patients at high risk of lymph node metastases. The depth of myometrial infiltration of endometrial cancer was assessed using TVS before or after curettage. Infiltration was classified as superficial if less than half of the myometrium was involved, otherwise it was classified as deep infiltration. Results were compared with the histology results of the definitive specimens. Patients were classified as high risk when they satisfied two of the following three criteria: 60 years of age or older; deep myometrial infiltration; and poorly differentiated or undifferentiated tumor. A total of 93 patients from 11 clinics were analyzed. The mean age was 66.1 years (SD +/- 11.4). The sonography and histology findings were in agreement in 69 of 93 patients. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), of "deep infiltration" by preoperative TVS were 79% (95% CI 0.65-0.93), 72% (95% CI 0.61-0.83), 61% (95% CI 0.46-0.75), and 86% (95% CI 0.76-0.96), respectively. Combining tumor grade and myometrial infiltration in the hysterectomy specimen and age, 30 of 81 patients were classified as high-risk patients. Sensitivity and PPV, specificity, and NPV for preoperative diagnosis of high risk were 80% (95% CI 0.65-0.94) and 88% (95% CI 0.79-0.97), respectively. Preoperative assessment of myometrial tumor infiltration using just TVS is only moderately reliable in endometrial cancer patients. If the results of TVS, however, are combined with the patient's age and the degree of tumor differentiation in curettings, it is possible to preoperatively select endometrial cancer patients with a high risk of pelvic lymph node metastases with sufficient reliability.