Since our first report in 1989, 26 transplants by means of umbilical cord blood have been reported. Furthermore, systematic studies of the feasibility of using banked placental blood for bone marrow reconstitution of unrelated recipients on a large scale are in progress worldwide. However, already by 1989, it was pointed out that the use of cord blood might be hampered by contamination of neonatal blood with maternal cells contributing unacceptably to graft-versus-host disease (GVHD). In the present study, we used the polymerase chain reaction (PCR) amplification of 2 minisatellite sequences (33.6 and MS 51) to address this question. The sensitivity of PCR amplification of minisatellite sequences is known to be of 1% to 0.1%. This sensitivity has been confirmed in the present study, in which a dilution analysis was performed for each experiment in which cell separation was performed. The inclusion of the dilution experiment in these analyses allowed us to estimate the relative amount of contaminating maternal cells, if any. Among 47 cases (31 whole blood analyses, 10 gradient separations, and 6 subpopulation separations), the coamplification of the 2 minisatellites sequences allowed the discrimination of maternal and neonate alleles in 42 cases (89%). In 1 case, we were able to detect a child-specific allele in a mother's whole blood sample, thus validating our approach to search for maternal cells in cord blood. In a single other case, we were able to detect a maternal-specific allele in the cord blood sample. This maternal specific allele was detected in the whole blood, polymorphonuclear cell, and lymphocyte fractions. Comparison of the signal intensity obtained with these 3 cord blood samples to the result of the dilution experiment performed in the same analysis led to an estimate of 1 to 5% maternal cells in the polymorphonuclear cell fraction and 0.1% to 1% maternal cells in the whole blood and lymphocyte cell fractions. In conclusion, our study indicates that maternal cells are very rarely present in the cord blood collected at birth because we detected them in only 1 of 47 cases. More importantly, when detected, they were present at very low level in the lymphocyte cell fraction (0.1% to 1%). However, although small, this amount of cells may result in GVHD in a susceptible recipient. Because the method we used allows the detection of maternal cells within cord blood from 10(4) nucleated cells, it would thus be of interest in a cord blood banking perspective.