Summary.We have compared the sensitivity and specificity of four PCR methods of RHD gene detection using different sets of primers located in the regions of highest divergence between the RHD and RHCE genes, notably exon 10 (method I), exon 7 (method II), exon 4 (method III) and intron 4 (method IV). Methods I-III were the most sensitive and gave a detectable signal with D-pos/D-neg mixtures containing only 0 . 001% D-positive cells. Moreover, method II could detect the equivalent DNA amount present in only three nucleated cells in the assay without hybridization of PCR products, whereas the sensitivity of the other methods was 10-50 times less. Investigation of D variants indicated that false-negative results were obtained with method II (D IVb variant), method III (D VI and DFR variants) and method IV (D VI variants), but not method I. Weak D (D u ) was correctly detected as D-positive by all methods, but most cases of Rh null appeared as false-positives, as they carry normal RH genes that are not phenotypically expressed. Some false-positive results were obtained with method I in a few Caucasian DNA samples serotyped as RhD-neg but carrying a C-or E-allele, whereas a high incidence of false-positives was found among non-Caucasian Rh-negative samples by all methods. In the Caucasian population, however, we found a full correlation between the predicted genotype and observed phenotype at birth of 92 infants. Although we routinely use the four methods for RHD genotyping, a PCR strategy based on at least two methods is recommended.