Background: The purpose of this study was to develop a prognostic system applicable to patients with hepatic metastasis from colorectal cancer in whom extrahepatic disease was excluded by preoperative PET with [18F]fluoro-2-deoxy-D-glucose (FDG-PET). Data from two institutions were analyzed separately and together to improve general applicability of results. Study Design: Data were analyzed for 285 consecutive patients undergoing liver resection for colorectal metastases from 1995 to 2005 at 2 institutions routinely using preoperative FDG-PET with. Fifteen clinicopathologic variables of the primary and secondary tumors were examined to identify factors predictive of survival. Results: Outcomes were correlated with poorly differentiated tumor grade in both data sets. Because patients with poorly differentiated tumors comprised a small proportion (16%) of the population, patients with well-differentiated or moderately differentiated tumors were analyzed independently. In this subgroup, positive lymph node status in the primary colorectal tumor resection specimen was the only characteristic that predicted survival of patients in both institutions. Consequently, patients were sorted into three prognostic categories: poor tumor differentiation; well-differentiated or moderately differentiated tumors and node positive; and well-differentiated or moderately differentiated tumors and node negative. These groups had significantly different overall survival on Kaplan-Meier analysis (p = 0.0014). Conclusions: In patients with colorectal liver metastases staged with FDG-PET with overall survival can be predicted directly from data in the pathology report of the colorectal primary tumor. This study also indicates the need for new molecular tumor markers of prognosis to complement clinicopathologic markers if the goal of prediction of outcomes in individual patients is to be reached.