Multivariate analysis of the prognostic value of CEA and CA 19-9 serum levels in colorectal cancer.

UNLABELLED: Carcinoembrionic Antigen (CEA) and carbohydrate antigen 19-9 (CA 19-9) are the most commonly used tumor-associated antigens in the management of patients with colorectal cancer. The aim of this study was to evaluate the prognostic value of preoperative serum levels of CEA and/or CA 19-9 and the classical prognostic factors (age, sex, tumor infiltration and staging) in 495 patients.

PATIENTS AND METHODS: The retrospective study was performed on frozen sera (stored at -70 degrees C) of patients with histologically proven colorectal cancer. Survival function estimates were calculated (Kaplan-Meier). The patients were separated into two groups according to the preoperative marker levels. Cut-off levels calculated at a specificity of 100% versus healthy individuals were used: < 4 ng/mL versus > or = 4 ng/mL for CEA and < 60 U/mL versus > or = 60 U/mL for CA 19-9. Survival curve differences were assessed using the log-rank-test. Mulivariate Cox's proportional hazard regression analysis was performed to examine the association between tumor marker levels and survival time. Classical prognostic factors such as age, sex, tumor infiltration, tumor stage (Dukes' classification) were included as covariants. The mantel-Haenszel method was used to assess the survival rate of patients with colorectal carcinoma and high versus low levels of tumor-associated antigens according to tumor stages.

RESULTS: The Dukes' stages (log-rank chi-square = 231.9; p < 0.0001) represent the best prognostic factor besides the preoperative values of CA 19-9 (log-rank chi-square = 162.5). CEA shows a log-rank chi-square of 71.4. Thus, CEA and CA 19-9 can be used to discriminate two groups of patients with significantly different survival times (p < 0.0001). The importance of different parameters in providing additional prognostic information was evaluated by multivariate analysis. Only items of statistically significant prognostic relevance (univariate analysis) were used for this analysis. Estimated relative risks of death adjusted for tumor stage were 5.5 considering Dukes' stage A versus Dukes' stage B/C and Dukes' stage B/C versus Dukes' stage D, respectively, and an increasing relative risk of 27.5 for Dukes' stage A versus Dukes' stage D (p < 0.001). The relative risk for preoperative CA 19-9 serum concentrations (> or = 60 U/mL versus < 60 U/mL) was 2.3 (p < 0.001) and for preoperative CEA concentrations (> or = 4 ng/mL versus < 4 ng/mL) 1.4 (p < 0.07). For CEA the 2-year survival rates in the group of patients with preoperative serum concentrations > 4 ng/mL versus < 4 ng/mL and Dukes' stage D were 16% versus 38%, in Dukes' stage B/C 73% versus 91% and in Dukes' stage A 100% versus 98%. For CA 19-9 the 2-year survival rates in the group of patients with preoperative serum concentrations > or = 60 U/mL versus < 60 U/mL and Dukes' stage D were 10% versus 39%, whilst in Dukes' stage B/C 58% versus 87%. In the group of patients with Dukes' stage A with serum levels > or = 60 U/mL a 2-years survival rate of 100% was found.

CONCLUSION: The postoperative Dukes' classification represents the best prognostic information besides the preoperative values of CA 19-9. The predictive information provided by preoperative CA 19-9 serum levels is independent from that obtained by the other factors investigated. Only Dukes' classification and CA 19-9 levels showed statistical significance (p < 0.001).