Serum CA-15.3 and CEA levels were longitudinally determined in 307 patients with breast carcinoma during postsurgical follow-up and/or therapy. Of 120 patients with no apparent disease, the specificity of marker levels fluctuating within the normal range (true-negative) was 98% for CA-15.3 alone (P = 0.004) and about 88% for CEA alone or for the tests combined. However, the false-negative levels in patients with progressive cancer reduced the predictive value of the tandem to around 76%, i.e. normal levels of both markers correctly predicted uneventful postsurgical course in only three fourths of the patients. Of 187 patients with active disease, the sensitivity of raised or increasing marker levels was around 70% for CA-15.3 alone or CEA alone, and 82% for the tests combined (P = 0.006). The 11% false-positive rate of CEA in patients with no apparent disease decreased the predictive value of a positive test from 98% for CA-15.3 alone (P = 0.006) to 91% for CEA alone or the tandem. Serum CA-15.3 or CEA paralleled the site of relapse: at least one marker was found elevated in 60% of patients with locoregional disease or with metastases to the lungs or bones exclusively, and in 90% of those with metastases to the lungs and bones or to the liver. A concurrent decrease of both marker levels reflected response to therapy while an increase of at least one marker level reflected treatment failure. It may be concluded that the marker tandem was better than either marker alone for follow-up aimed at detection of relapse, and that the tests were approximately 80% accurate for follow-up and/or monitoring therapy.