PURPOSE: To compare the respective sensitivity of somatostatin receptor scintigraphy (SRS), computed tomography (CT), and magnetic resonance imaging (MRI) in the detection of liver metastases from well-differentiated gastroenteropancreatic endocrine tumor (WDGEP ET) patients. To define predictive factors for "high-sensitivity SRS." PATIENTS AND METHODS: Sixty-four patients with WDGEP ET underwent SRS with abdominal single-photon emission computed tomography (SPECT), spiral CT, and 1.5-T MRI within a 15-day interval, the order of which was randomized. Two readers analyzed images of each modality, blindly and independently. RESULTS: Hepatic metastases were present in 40 of the 64 patients and confirmed by pathology after liver biopsy or surgery in 32 and eight patients, respectively. SRS, CT, and MRI detected a total of 204, 325, and 394 metastases, respectively. The number of detected metastases was significantly higher with MRI than with CT (P = .02) and SRS (P < 10(-4)) and higher with CT than with SRS (P < 10(-4)). SRS was negative in seven patients with a positive CT and/or MRI. More lesions were detected in 10 patients by SPECT compared with static views. The median metastasis size was significantly correlated (P = .04) with the sensitivity of SRS. CONCLUSION: MRI seems to have an edge over CT and SRS for the detection of liver metastases from endocrine tumors. We recommend the systematic performance of liver MRI at WDGEP ET initial staging and before major therapeutic events. The low performance of SRS was mainly explained by the impact of the metastasis size on the detection capacity of SRS.