Histopathologic-Based Prognostic Factors of Colorectal Cancers Are Associated With the State of the Local Immune Reaction

PURPOSE: The prognosis of patients with colorectal cancer has sometimes proved uncertain; thus, the prognostic significance of immune criteria was compared with that of the tumor extension criteria using the American Joint Committee on Cancer/International Union Against Cancer-TNM (AJCC/UICC-TNM) staging system. PATIENTS AND METHODS: We studied the intratumoral immune infiltrates in the center of the tumor and in the invasive margin of 599 specimens of stage I to IV colorectal cancers from two independent cohorts. We analyzed these findings in relation to the degree of tumor extension and to the frequency of recurrence. RESULTS: Growth of the primary tumor and metastatic spread were associated with decreased intratumoral immune T-cell densities. Sixty percent of patients with high densities of CD8(+) cytotoxic T-lymphocyte infiltrate presented with stage Tis/T1 tumor, whereas no patients with low densities presented with such early-stage tumor. In patients who did not relapse, the density of CD8 infiltrates was inversely correlated with T stage. In contrast, in patients whose tumor recurred, the number of CD8 cells was low regardless of the T stage of the tumor. Univariate analysis showed that the immune score was significantly associated with differences in disease-free, disease-specific, and overall survival (hazard ratio HR, 0.64, 0.60, and 0.70, respectively; P < .005). Time-dependent receiver operating characteristic curve analysis illustrated the predictive accuracy of the immune parameters (c-index = 65.3%, time-dependent c-index Cτ = 66.5%). A final stepwise model for Cox multivariate analysis supports the advantage of the immune score (HR, 0.64; P < .001; Cτ = 67.9%) compared with histopathologic features in predicting recurrence as well as survival. CONCLUSION: Assessment of CD8(+) cytotoxic T lymphocytes in combined tumor regions provides an indicator of tumor recurrence beyond that predicted by AJCC/UICC-TNM staging.