The usefulness of a urine-alcohol determination in the evaluation of abstinence in alcoholic liver disease has been investigated in 181 patients. Alcohol was tested in morning urine samples collected on a follow-up visit in 103 patients, and the results were compared with those in 78 patients, where three samples were collected within the same week; one in the morning and two in the evening. Although the percentage of urine samples containing alcohol measured in a morning sample was similar to the patients' self-report (31% and 34%, respectively), urine analysis identified an additional 7% of patients who denied alcohol intake. Alternatively, serial urine-alcohol determinations were significantly more effective than patient reports (54% and 35.9%, respectively, P less than 0.01), particularly when urine was collected in the evenings. This difference was due mainly to the reluctance of women to admit drinking (7.4% of positive self reports vs 51% of women with alcohol positive urine samples, P less than 0.001). We conclude that serial measurements of alcohol in urine were useful tests which should be used to complement personal interview in the control of abstinence in patients with alcoholic liver disease.