Prospective serial ultrasound scanning (U/S) of the liver at 3-monthly intervals has been used to detect the presence of hepatic metastases from successfully resected Dukes' A, B and C grade colorectal carcinomas in a trial of the antimetastatic agent razoxane (ICRF 159). One hundred and twenty-six consecutive patients were randomly allocated to either adjuvant razoxane treatment (61) or to no further treatment (65). Twenty-six patients, 15 in the control group (23 per cent) and 11 in the razoxane treatment group (18 per cent) have developed hepatic metastases. All but three patients in the treated group and four in the controls were detected by U/S. In the control group four patients had definite metastases when metastases were first seen by ultrasound and three more, among the eight who initially had probable metastases, became definite metastases on ultrasound. In the treated group four patients had definite metastases and four more developed them amongst 12 who had probable metastases. The remaining eight of these 12 all regressed completely. This compared with only one complete regression (additionally on 5-FU for lung secondaries) amongst the eight controls who had probable metastases. The median time to development of liver metastases detected by U/S and other means was 59 weeks in the control group and 91 weeks in the razoxane treatment group. Ultrasound has proved to be a safe, sensitive and acceptable serial imaging technique for detecting colorectal liver metastases. Used prospectively, U/S has been valuable in monitoring the natural history of liver metastases and their behaviour during razoxane treatment.