Altogether, 243 female breast cancer patients with localised disease diagnosed between 1991 and 1995 in the Tampere University Hospital area were followed prospectively after primary treatment until the first relapse. In the follow-up period, the serum tumour marker Ca 15-3 was analysed every 6 months to ascertain the validity of this marker in detecting the first relapse. The sensitivity and specificity of the test were analysed in different metastatic situations. During the 5 years of follow-up, 59 (24%) relapses were discovered. Ca 15-3 was elevated in 21/59 (36%) of the relapsed cases at least once. The 59 patients were subjected to 199 tests, of which 25 (13%) were positive. Among the 184 patients without recurrence, there were 6 (3%) with a positive Ca 15-3 level. The test failed to detect locoregional relapse or contralateral breast cancer. It was elevated in approximately half of bone-only metastases and in all of the liver-only metastases. In the pulmonary-only recurrences, the marker value was not elevated. We conclude that the Ca 15-3 tumour marker test is specific, but not sensitive enough to indicate the first relapse earlier than other methods. The positive predictive value especially remained poor in patients with a relatively good prognosis. Our results confirm that the test is not suitable alone for breast cancer follow-up.