THE VALUE OF SERUM CARCINOEMBRYONIC ANTIGEN IN PREDICTING RECURRENT DISEASE FOLLOWING CURATIVE RESECTION OF COLORECTAL-CANCER

PURPOSE: This study was designed to determine the natural history of patients with normal and abnormal levels of CEA. METHODS: Patients undergoing potentially curative resection of colorectal tumors (Dukes Stage A-C) entered a prospective, randomized trial comparing two follow-up regimens (to be reported separately) had CEA levels measured every 3 months for two years; then every G months for the next three years. In the study protocol, a rise in CEA was not an indication for investigation to determine recurrence unless there was also other evidence of recurrent disease. RESULTS: Three hundred eleven patients were followed for a median of 4.5 (range, 2-5) years. Recurrent disease developed in 98 (32 percent) patients, 57 of whom had an elevated CEA (sensitivity 58 percent), with a median lead time of six (range, 1-30) months from first abnormal CEA to diagnosis of recurrent disease by other means. The specificity, positive predictive value, and negative predictive value of CEA as an indicator of subsequent recurrent disease was 93 percent, 79 percent, and 83 percent, respectively. The sensitivity of CEA for predicting hepatic metastases was 80 percent, with a median lead time of eight (range, 1-30) months, compared with only 46 percent for sites of recurrent disease other than the liver. CONCLUSIONS: CEA was the first indicator of recurrent disease in 58 percent of all patients and in 80 percent of patients with liver metastases. The diagnosis of recurrent disease may be made several months earlier by investigating the first abnormal CEA level, although any benefit in