A biannual external quality assurance (EQA) scheme for flow cytometric typing of the HLA-B27 antigen is operational in The Netherlands and Belgium since 1995. We report here on the results of the first seven send-outs to which 36 to 47 laboratories participated. With the send-out, four specimens from blood bank donors, who had been typed for HLA Class I antigens by complement-dependent cytotoxicity, were distributed. Subtyping of the HLA-B27 allele was performed by PCR-SSP. Ten samples were HLA-B27(pos) (all HLA-B*2705) and 18 were HLA-B27(neg). For flow cytometry, the most widely monoclonal antibody (MoAb) used was FD705, followed by GS145.2 and ABC-m3. The majority of laboratories used more than 1 anti-HLA-B27 MoAb for typing. The HLA-B27(pos) samples were correctly classified as positive by the large majority of participants (median 95%; range 85% to 100% per send out); some participants considered further typing necessary and misclassification as negative was only sporadically seen. The classification of HLA-B27(neg) samples as negative was less straightforward. Ten samples were correctly classified as such by 97% (82% to 100%) of the participants, whereas 64% (range 53% to 70%) of the participants classified the remaining eight samples as HLA-B27(neg). There was no significant prevalence of a particular HLA-B allele among these eight "poor concordancy" samples as compared to the ten "good concordancy" samples. Inspection of the reactivity patterns of the individual MoAb with HLA-B27(neg) samples revealed that ABC-m3 showed very little cross-reactivity apart from its well-known cross-reactivity with HLA-B7, whereas the cross-reactivity patterns of GS145.2 and FD705 were more extensive. The small sample size (n = 18) and the distribution of HLA-B alleles other than HLA-B27 did not allow assignment of specificities to these cross-reactions. Finally, we showed that standardized interpretation of the combined results of two anti-HLA-B27 MoAb reduced the frequency of false-positive conclusions on HLA-B27(neg) samples. In this series, the lowest frequency of false-positive assignments was observed with the combination of the FD705 and ABC-m3 MoAb.