Although positron emission tomography (PET) detects occult metastatic disease in approximately 20% of patients with isolated hepatic colorectal metastases, it is associated with false negative results in up to 16%. We hypothesized that patients with a poorer prognosis (as defined by clinical risk score [CRS]) would have a higher yield from PET. All patients with colorectal liver metastases who were imaged by means of PET between 1998 and 2002 were identified from a prospective PET database. All patients were assigned a CRS, with one point added for each of five preoperative factors (disease-free interval <1 year, tumor size >5 cm, tumor number >1, carcinoembryonic antigen >200, and node-positive primary lesion). A total of 85 PET scans were reviewed. In half the patients (53%), PET provided no additional information over conventional imaging. Occult extrahepatic disease was detected or questionable findings seen on conventional imaging were confirmed in 20% of PET scans, whereas PET readings were inaccurate in 27%. PET findings were correlated with CRS in a subset of 63 patients presenting with a first occurrence of hepatic colorectal metastases. Among patients with a CRS of 0, no patient had extrahepatic disease detected by PET and 57% had false positive readings, whereas among patients with a CRS of 1 or more, 14% were found to have additional disease that was detected only by PET, and there were no false positive readings (P<0.001, Fisher's exact test). Patients with isolated hepatic colorectal metastases and a CRS of 0 should undergo conventional imaging alone prior to surgical exploration.