Background: To obtain a prognostic stratification model for resected gastric cancer patients.
Patients and methods: Clinicopathological and molecular data (expression of Cdx2, Apc, β-catenin, E-cadherin, Fhit, p53, and human epidermal growth factor receptor-2 (Her2); HER2 and TOPO2A gene copy number; PIK3CA mutations; microsatellite instability) were correlated to cancer-specific/overall survival (CSS/OS) using a Cox model. Individual patient probability (IPP) was estimated by logistic equation. A continuous score to identify risk-classes was derived according to the model ratios.
Results: Two-hundred eight patients were studied (median follow-up 20 months). At multivariate analysis, sex, stage, margins, location, nodes, Apc, and Fhit were independent predictors for CSS; the same factors (and age and Her2, except Fhit) predicted OS. Multivariate model predicted IPP with high prognostic accuracy (0.90 for CSS; 0.91 for OS). A two-class model significantly separated low- and high-risk patients for CSS (23.4% and 85.6%, P < 0.0001) and OS (21.4% and 82.0%, P < 0.0001). A three-class model differentiated low-, intermediate-, and high-risk patients for CSS (6.3%, 35.3%, and 88.0%, P < 0.0001) and OS (6.1%, 34.6%, and 86.5%, P < 0.0001).
Conclusions: A risk classification system comprising the immunohistochemical expression of three proteins (Apc, Fhit, and Her2) and five clinicopathological parameters (stage, resected nodes, margins, location, and sex) accurately separates the resected gastric cancer patients into three classes of risk.