Multivariate analysis of a tissue CEA, TPA, and CA 19.9 quantitative study in colorectal cancer patients. A...

Tissue CEA, TPA, and CA 19.9 concentrations from samples of surgical specimens were measured in 47 evaluable colorectal cancer patients (median follow-up, 20 months, 13 recurrences) and correlated with individual patient follow-up status. The quantitative method appeared to be sensitive, easily reproducible, and standardizable. The tissue marker concentration was analyzed by means of the multivariate discriminant analysis, to evaluate the risk of relapse in each patient; the tumor CEA (CEA T) showed the best discriminant capacity (P = .005). The relative Fisher function provided a reliable prognostic patient index, independently of other recognized prognostic factors (Dukes' stage and cellular differentiation grade). The Cox model showed a statistical significance analyzing the tumor (T) and healthy mucosa (M) CEA values (P = .001 and P = .006, respectively). The combination of these two variables allowed for identification of three classes of patients according to CEA T and M threshold values of 216 and 85 ng/mg of protein, respectively, and different disease-free curves were obtained for each group. The two-year disease-free rate was 81 percent for patients with low values of both CEA T and M, and 21.4 percent for the group with both values above these thresholds (P = .0008). In the third class (CEA T or M higher than the reported cut-off levels), the two-year disease-free rate was 65.9 percent.