Gene Expression Profiles Accurately Predict Outcome Following Liver Resection in Patients with Metastatic Colorectal Cancer

Purpose

The aim of this study was to build a molecular prognostic model based on gene signatures for patients with completely resected hepatic metastases from colorectal cancer (MCRC).

Methods

Using the Illumina HumanHT-12 gene chip, RNA samples from the liver metastases of 96 patients who underwent R0 liver resection were analyzed. Patients were randomly assigned to a training (n = 60) and test (n = 36) set. The genes associated with disease-specific survival (DSS) and liver-recurrence-free survival (LRFS) were identified by Cox-regression and selected to construct a molecular risk score (MRS) using the supervised principle component method on the training set. The MRS was then evaluated in the independent test set.

Results

Nineteen and 115 genes were selected to construct the MRS for DSS and LRFS, respectively. Each MRS was validated in the test set; 3-year DSS/LRFS rates were 42/32% and 79/80% for patients with high and low MRS, respectively (p = 0.007 for DSS and p = 0.046 for LRFS). In a multivariate model controlling for a previously validated clinical risk score (CRS), the MRS remained a significant predictor of DSS (p = 0.001) and LRFS (p = 0.03). When CRS and MRS were combined, the patients were discriminated better with 3-year DSS/LRFS rates of 90/89% in the low risk group (both risk scores low) vs 42/26% in the high risk group (both risk scores high), respectively (p = 0.002/0.004 for DSS/LRFS).

Conclusion

MRS based on gene expression profiling has high prognostic value and is independent of CRS. This finding provides a potential strategy for better risk-stratification of patients with liver MCRC.