The American Joint Committee on Cancer (AJCC) staging system, based on tumor extent, nodal involvement, and the presence of metastases (TNM classification), remains the reference tool for estimating colon cancer prognosis. However, it fails to account for the substantial variability in clinical outcome observed within a given stage, as it incorporates neither the molecular characteristics of the tumor nor the host immune response. Yet the density of cytotoxic and memory T lymphocytes infiltrating the tumor, measured by the Immunoscore (which combines CD3+ and CD8+ cell densities in the tumor center and at its invasive margin), has already been associated with survival. Previous studies, however, relied on retrospective series combining different stages, treatments, and follow-up durations, which limited their scope. The authors therefore sought to establish a stage III-specific prognostic classifier in a homogeneous population.
The study is based on 559 patients with resected stage III colon adenocarcinoma, drawn from the FOLFOX arm of a randomized phase 3 trial (NCCTG N0147). Cox models were built to predict disease-free survival, incorporating age, sex, T stage, number of involved lymph nodes, N stage, performance status, histological grade, laterality, KRAS/BRAF status, DNA mismatch repair, and the Immunoscore. Backward selection was used to retain the most informative variables. A recurrence occurred in 164 patients (29.3%).
Several factors were associated with poorer disease-free survival: T4 versus T3 tumors (hazard ratio 1.76), right- versus left-sided location (1.52), BRAF V600E mutation (1.74), KRAS mutation (1.66), and low versus high Immunoscore (1.69). In the final multivariate model, four variables stood out: the number of involved lymph nodes (43.1% of the contribution to risk), T stage (18%), BRAF/KRAS status (16.1%), and the Immunoscore (14.9%). Notably, in low-risk patients (T1-3 N1), the Immunoscore remained the only statistically significant factor. A bootstrap analysis internally validated these results. The authors further note that the prognostic value of tumor laterality, significant in univariate analysis, appears to be captured by the Immunoscore.
Based on the model, a nomogram was developed to estimate, on a patient-by-patient basis, the probability of being recurrence-free at three years. According to the authors, this tool allows a more precise and more personalized prediction than T and N stages alone, and can reclassify some patients toward higher or lower risk. They nevertheless acknowledge the need for external validation in an independent cohort of stage III patients treated with FOLFOX, and propose extending this approach to stage II, where risk stratification would guide the choice of adjuvant chemotherapy.