Until recently, the majority of newly recruited volunteer donors were typed for HLA-A and -B by serology onto the National Marrow Donor Program Registry. Quality control of this serological typing performed by contracted laboratories was carried out by retesting approximately 1% of each laboratory's test volume utilizing DNA-based techniques (SSOP). The criteria used for selection included samples presumed to be homozygotes, samples with split antigen specificities and samples with antigens considered to be difficult to define. Out of 1983 samples analyzed, 156 HLA-A (3.9%) and 265 HLA-B (6.7%) locus discrepancies were identified. Review of these discrepancies by both the serological and QC laboratory revealed that the majority of discrepancies were due to errors in serological typing. Serological discrepancies were categorized as follows: blank antigens identified (36.8%) and misassignments (63.2%). Misassignments were defined as either the incorrect assignment of antigens within a group ("wrong split"), or a complete misassignment. Antigens reported as blanks most frequently belonged to the A19 and A28 groups and to the B70, 46 and 40 groups. The most frequent misassignments within groups were the A19 and A10 groups, and the B40 and B15 groups. Other HLA-A misassignments included A2 vs A28 or A2 vs A69, while other HLA-B misassignments included B35 and B70. This QC analysis showed that serological typing of class I antigens for the purposes of NMDP registry typing is prone to a significant error rate. Careful evaluation and selection of contracted laboratories by the NMDP suggests methodological limitations rather than poor performance as the main cause of these observations.