Assessing tools for management of noncolorectal nonneuroendocrine liver metastases: External validation of a prognostic model

Introduction Selection criteria and benefits for resection of noncolorectal, nonneuroendocrine liver metastases (NCNNELM) remain debated. A prognostic score was developed by the Association Française de Chirurgie (AFC) for patient selection, but not validated. We performed a geographic external validation of this score. Methods Patients with resected NCNNELM from six institutions (2000–2014) were assigned risk groups based on the AFC score. Discrimination was evaluated by visually inspecting separation of overall survival (OS) curves among risk categories. The slope of the continuous score on OS and hazard ratios for risk categories were examined. Results Of 165 patients, 53 (32.1%) were low‐risk, 85 (51.5%) intermediate‐risk, and 27 (16.4%) high‐risk. The OS curves did not separate among risk groups: 5‐year OS were 60.1% (low), 57.1% (intermediate), and 55.6% (high). The parameter estimate (0.02) indicated lower discrimination than in the AFC cohort. Hazard ratios of 1.05 (0.63 to 1.70) for low vs intermediate, 0.87 (0.46 to 1.64) for low vs high, and 0.83 (0.46 to 1.49) for intermediate vs. high, demonstrated lack of discrimination in OS among risk groups. Conclusion While long‐term survival is achievable, discrimination of the AFC score is not maintained in a geographic external cohort of resected NCNNELM. It is not generalizable to this external population.