Predictive model for survival after liver resection for noncolorectal liver metastases in the modern era: a Japanese multicenter analysis

Background Survival benefit of liver resection for noncolorectal liver metastases (NCRLM) remains to be defined. Methods This multicenter, retrospective cohort analysis included consecutive patients with NCRLM whose primary tumor and all metastases were treated with curative intent between 2000 and 2013. The primary endpoint was 5‐year overall survival. Clinicopathological factors that affected prognoses were identified using multivariate Cox regression analyses and were included in a predictive model. Results Data for 205 patients were analyzed. The three most common primary tumor sites were stomach (39%), pancreas (13%), and urinary tract (10%), with adenocarcinomas the main pathology (52%). R0 resection was achieved in 85%, and the overall survival at 5 years was 41%. In the multivariate analysis, synchronous liver metastases, R1/2 resection, and adenocarcinomas and other carcinomas (with gastrointestinal stromal tumors, neuroendocrine tumors G1/G2, and sarcomas set as the reference group) were independent negative indicators of overall survival. A predictive model effectively stratified the NCRLM patients into low‐, intermediate‐, and high‐risk groups with overall 5‐year survival rates of 63%, 38%, and 21%, respectively (P