Currently, pathological staging according to the tumor-node-metastasis system remains the gold standard for the prediction of patient survival in colorectal cancer (CRC) but this classification system provides insufficient information and therefore additional prognostic markers are needed.
A genome-wide methylation analysis was done for two independent cohorts of non-metastatic CRC patients (screening cohort n = 578 and validation cohort n = 308). Initially, genome-wide differentially methylated CpG sites between 34 pairs of tumor and normal mucosa tissue samples from the same patients were identified. A variable screening for prognostic CpG sites was performed in the screening cohort using marginal testing based on the Cox model and subsequent adjustment of the p-values via independent hypothesis weighting (IHW) using the difference between tumor and normal mucosa tissue as auxiliary covariate. From the 1000 CpG sites with the smallest adjusted p-value, the 20 CpG sites with the smallest Brier Score for 3-year overall survival (in the screening cohort) were selected. Applying principal component analysis on these CpG sites, we derived a methylation-based classifier for the prognosis of non-metastatic CRC (ProMCol).
The ProMCol classifier was independently validated in the validation cohort, where it showed a significant reduction in the Brier score. Regarding the three year survival, the prediction error was reduced from 0.132 (calculated only with clinical variables), to 0.124 (combination of clinical variables with ProMCol classifier). An additional replication analysis showed that the ProMCol classifier was significantly associated with overall survival (OS) of non-metastatic CRC patients in the screening (HR = 0.22, 95%CI=0.13-0.35, p=6.2E-10) and the validation cohort (HR = 0.40, 95%CI=0.22-0.74, p=0.003), adjusted for standard clinical factors. Patients with a high methylation status, represented by higher values of the ProMCol classifier, showed a better prognosis for OS than patients with a low methylation status and lower ProMCol classifier values.
The usage of the ProMCol classifier could improve the prognostic accuracy for non-metastatic CRC patients.
Clinical trial identification
Legal entity responsible for the study
Barbara Burwinkel, German Cancer Research Center
German Research Council, German Federal Ministry of Education and Research
All authors have declared no conflicts of interest.