Utility or futility of prognostic scoring systems for colorectal liver metastases in an era of advanced multimodal therapy.

OBJECTIVES To assess the general applicability of prognostic scores for colorectal liver metastases (CRLM). METHODS Review of English language studies from 1980 to 2008 (Medline and Embase). Search keywords included "Colorectal neoplasms", "liver metastases", "liver resection", "prognostic scoring system". RESULTS Six scoring systems and fourteen prognostic factors within these studies were identified. No prognostic factor was common in all scoring methods. Five scores retained the number of metastases as a prognostic factor. Size of metastases and time between the onset of the primary tumor and the discovery of metastases were present in four scores. Three scores predicted 5-year survival using carcinoembryonic antigen (CEA) and R1 resection. Only two scores were assessed preoperatively. Successive scoring methods had improved predictive accuracy compared to earlier systems. However, their applicability in general populations remains debatable. An evaluation of the scores applicability to different patient populations demonstrated that the models were minimally effective in predicting disease-specific survival and recurrence, suggesting that stratification of patients by clinical and pathologic factors alone, may be clinically unreliable and not applicable for selection of patients for surgery. CONCLUSION The utility of prognostic models on general populations is inconsistent. Current clinicopathologic factors may be inadequate to determine disease prognosis in CRLM. Future attempts to develop prognostic scores should include additional biologic and clinical variables, and be validated in larger populations.