CEA monitoring among patients in multi-institutional adjuvant G.I. therapy protocols.

The Gastrointestinal Tumor Study Group (GITSG) has since 1975 included protocols for monitoring carcinoembryonic antigen (CEA) levels in its colorectal cancer adjuvant trials. Among the 563 patients on the colon cancer study (GI 6175) and the 207 patients on the rectal cancer study (GI 7175), one third had preoperative CEA determinations and more than 90% had some postoperative CEA monitoring. Colon cancer patients whose preoperative CEA was greater than 5 ng/ml had a greater probability of recurring than those whose values were lower (33% versus 18% recurrence with 21 months minimum follow-up; p < 0.05). The prognostic value of preoperative CEA was apparent only in patients with Dukes' C1 colon tumors. Preoperative CEA values were not of prognostic significance among the rectal adenocarcinoma patients. Although elevated levels of CEA after resection of either colon or rectum cancers were strongly associated with subsequent tumor recurrence, no single CEA value, arbitrarily defined as "elevated", provided an adequate screening test with both high sensitivity and high specificity. Postoperative CEA elevations were more strongly predictive of recurrence when part of a steadily rising trend. In the colon cancer study, the median monthly increase in CEA for disease-free patients was estimated to be zero, and for the relapsed patients 5.8%. The corresponding estimates for patients on the rectal cancer protocol were zero and 7.8%. Only 36 of the 344 disease-free patients on the colon protocol and 14 of the 94 disease-free patients on the rectal protocol (15%) exhibited a rate of increase of CEA as high as 3% per month over the entire period of observation. Two thirds of the relapsed patients on both studies showed a rate of increase this high or higher. The patterns of CEA rise in individual patients were quite varied, however, and monthly rates of increase as established in our study are not to be used as guidelines in patient management.