The usefulness of the CEA as an indicator of recurrence and a guide to selected second-look surgery was evaluated from a retrospective analysis of 358 patients with colorectal cancer and from a prospective experience with 16 patients all of whom had been admitted for second-look surgery because of postoperative elevations of CEA only. Our previous experience had shown that after curative resection the CEA usually returned to normal levels (less than 5 ng/ml) within one month, but became elevated at time of clinically obvious recurrence being very high in patients with liver metastases, but only moderately elevated or normal in patients with local recurrence. All 16 patients had previously had curative resection of colorectal cancer; 13 in the rectum or rectosigmoid and three in the right colon. There were 13 Dukes' C and three Dukes' B cancers. All had been followed clinically and by CEA testing at three monthly intervals and were considered free of disease (NED) at time of CEA elevation. The median disease free interval was 13 months (range 4-57 months) and the median CEA prompting admission for second-look operation was 21 ng/ml (range 10-56 ng/ml). The sites of recurrence were liver in six, lung in two and localized disease in six. Two patients had negative exploration for recurrence and were found to have cholelithiasis only (one of these later died of metastases). Resection for cure was done in seven and palliative resection or biopsy only was done in nine patients. At this time, four patients are NED (12-37 months), five are living with disease (10-16 months) and seven have died of disease (2-12 months). The CEA test provides a method of early detection of recurrence and may permit surgical retrieval in selected patients and earlier initiation of palliation in other patients. The longterm effects in patient salvage remain to be defined.