We evaluated the potential utility of occult circulating tumor DNA as a molecular marker of disease in subjects previously diagnosed with breast cancer. Using 24 microsatellite markers located at sites of frequent loss of heterozygosity (LOH) or allele imbalance in breast cancer, we analyzed DNA from 16 primary tumors (Stage IIA or more advanced) and 30 longitudinally collected plasma specimens. Clinical data at the time of plasma collection were obtained. All 16 tumors were characterized by an individual pattern of LOH. LOH was detected in 12 of 30 (40%) plasma samples, taken from 8 of 14 (57%) subjects. However, the number of LOH in plasma was small (n = 15), and the mean proportion of LOH was much lower than in the tumors (0.05 vs. 0.52). Although infrequent, 12 of 15 (80%) plasma LOH were concordant with abnormalities in the paired tumors, and the mean percent LOH was higher than in normal plasmas, suggesting that they were authentic tumor-derived abnormalities. We found, despite this, no association, between plasma LOH and tumor stage or clinical status at time of blood collection (i.e., LOH was as common in subjects with no evident disease as in those with evident disease). In addition, detection of LOH was not consistent between serial samples from 5 of 11 subjects (45%), despite stable clinical conditions. No association with clinical outcome was evident, although the sample size was small. Microsatellite instability in plasma was infrequent, nonconcordant with paired tumor and inconsistent in serial samples. This pilot study suggests that identifying tumor-specific LOH in the plasma of breast cancer subjects may not be useful for detecting occult metastases or for monitoring disease. Other detection techniques may be more promising, but circulating tumor DNA may not be a sufficiently accurate reflection of breast cancer clinical status or tumor activity.