The prognostic information provided by preoperative serum CEA and CA 19.9 antigen assay on the postoperative outcome of 150 patients with colorectal cancer was analysed. The influence of both markers was studied by Cox's proportional-hazard regression analysis. In the univariate analysis, patients whose initial CA 19.9 level was higher than 37 U/ml had a 4.32-fold greater risk of death due to the cancer (95% CI: 1.72-10.84) (p < 0.001) than patients with lower values. The 36-month survival rate posttreatment was lower for patients with CA 19.9 serum levels over 37 U/ml (61% versus 90%) (p < 0.001). Patients whose initial CEA level was higher than 5 ng/ml had a 2.9-fold greater risk of death (95% CI: 1.05-7.99) (p = 0.04) than patients with lower values. The 36-month survival rate posttreatment was lower for patients with CEA serum levels over 5 ng/ml (84% versus 76%) (p = 0.04). After adjustment for Dukes' stage, CEA, CA 19.9, tumor site, sex and age, only Dukes' stage and CA 19.9 continued to provide independent predictive information on survival. The risk of death increases by 1.008 for every 10 U/ml rise in the level of the marker (95% CI: 1.002-1.014) (p = 0.009). With respect to analysis of disease-free survival, only Dukes' stage provided independent predictive value. CA 19.9 is an independent prognostic factor of survival in colorectal cancer. The authors suggest including CA 19.9 in a future multifactorial analysis of survival.