In 81 healthy women, 26 pregnant women, 25 patients with fibrocystic disease and 144 breast cancer patients, the overall diagnostic sensitivity and specificity of the CA 15.3 test was 27 and 97%, respectively. The positive and negative predictive values were 93 and 43%. In 150 node-negative patients taking part in a chemoprevention trial CA 15.3 was assayed at baseline and every 4 months for a median follow-up of 24 months (range 4-48). In these patients, 5 had local recurrences, 1 had a regional recurrence, 9 had distant metastases and 3 developed cancer in the contralateral breast. Among the patients with recurrences, those with distant metastases showed the highest ratio of CA 15.3 increase (8/9); in local and regional recurrences, this ratio was lower (2/6). The patients with contralateral breast cancer had no significant increase in CA 15.3. Patients in whom metastases were detected showed an increase in CA 15.3 4-48 months before clinical or instrumental detection of the metastases.