BACKGROUND:Tumour metabolic response to chemotherapy is increasingly recognized as a prognostic indicator for colorectal cancer liver metastases (CRCLM). However, its clinical role and the underlying biological mechanism of its prognostic ability are unclear. This study compares metabolic to pathologic response for CRCLM, and correlates metabolic response to tumour expression of six key biomarkers. METHODS:Thirty-seven patients who had positron emission tomography imaging before and after pre-operative chemotherapy prior to liver resection for CRCLM were included. Metabolic response was assessed according to the positron emission tomography response criteria in solid tumours (PERCIST) and correlated to recurrence-free and overall survival. PERCIST was compared to tumour regression grading, computed tomography (CT) response, tumour necrosis and mucin and immunohistochemical expression of Ki-67, hypoxia inducible factor 1α, vascular endothelial growth factor, p53, p16 and vimentin. Area under the receiver operating characteristic curve (AUC), Kaplan-Meier survival, Spearman's correlation (rs ) and multivariate Cox regression analyses were used. RESULTS:PERCIST correlated significantly to 2-year mortality (AUC = 0.162, P < 0.01) and 2-year recurrence (AUC = 0.284, P = 0.03). Metabolically responsive tumours conferred a better overall survival (P = 0.01) and recurrence-free survival (P = 0.03). Tumour regression grading did not stratify for outcome. Metabolic response was significantly correlated to Ki-67 and p16 expression (rs = 0.559 and rs = -0.549, respectively). Multivariate analysis revealed only PERCIST to be correlated to death and recurrence. CONCLUSION:Pre-operative PERCIST assessment of CRCLM was more prognostic than pathologic and CT response assessment. Metabolic non-response correlated with tumour proliferation and loss of tumour suppression.