In the present investigation we evaluated the variability of tumor marker levels in the follow-up of patients without evidence of disease after resection of primary breast cancer. CEA and CA15.3 were measured using commercially available methods in serial blood samples collected from 170 patients. The coefficient of variation among all samples from each patient, which accounts for the total variability (analytical variability+biological variability), was widely scattered (from 4 to 99% for CEA; from 4 to 52% for CA15.3). The critical difference was calculated using the formula designed by Fraser [CD = 2.77. (CVa2 + CVb2)1/2]. It ranged from 11 to 276 for CEA and from 11 to 144 for CA15.3. From the present findings we conclude that: 1) it is possible to identify individually tailored decision criteria to evaluate tumor marker variations in the follow-up of breast cancer patients; 2) in a considerable number of cases the non-tumor-related variability is too high to allow the early identification of minor tumor marker variations that are of clinical relevance.