Objectives To evaluate the accuracy of different preoperative modalities for staging of endometrial cancer to restrict extensive surgery to patients at high risk of metastatic disease. Setting Aarhus University Hospital. Population 156 women referred in 2006-2011 because of atypical endometrial hyperplasia (G0) or endometrial cancer. Methods Patients were offered preoperative transvaginal ultrasonography (TVS), magnetic resonance imaging (MRI), and hysteroscopic-directed biopsies from the uterine tumor and cervix. Final pathology of the removed uterus was the reference standard. Patients were divided into low risk (<50% myometrial invasion, and grades 0, 1, 2, and no cervical invasion) or high risk (all others). Main outcome measures Accuracy, sensitivity, specificity, positive/negative predictive value. Results Patients were aged 32-88 years, with a mean body mass index of 29. At final pathology 81% had cancer and 19% G0 or no residual tumor; 54% were high risk. Hysteroscopy-directed biopsies had a higher accuracy (92%) than endometrial biopsy (58%) for differentiating G0 from cancer (p < 0.001); grade 3 tumor identification had similar accuracy (93 vs. 92%). Deep myometrial invasion was estimated with higher accuracy by MRI (82%) than TVS (74%) (p < 0.02). For cervical involvement, hysteroscopy-directed biopsies had higher accuracy (94%) than MRI (84%,) and TVS (80%) (p < 0.02). Accuracy for identifying high-risk women was highest (83%) using a combination of MRI and hysteroscopic-directed biopsies, compared with TVS and endometrial biopsy (72%) (p < 0.05). Conclusion Preoperative staging with MRI and hysteroscopy-directed biopsy can identify eight of 10 women with high risk of lymph node metastases and spare eight of 10 low-risk women extended surgery.