Interlaboratory Variation in CD34+ Cell Quantification in Low Risk Myelodysplastic Syndromes

Background: Accurate bone marrow (BM) blast count is essential for establishing prognosis in lower-risk myelodysplastic syndromes (MDS). Flow cytometry immunophenotyping (FCI) helps to evaluate the most immature BM cell compartment, but it is not known whether the CD34+ myeloid cell count in MDS are comparable among different observers. We investigated the FCI approaches used for enumeration of CD34 + myeloid cells in MDS in the routine practice in 6 clinical Spanish centers, and then explored if data were reproducible among participants.

Methods: In this prospective study, participants exchanged 47 FCI files from MDS BM samples with <5% blasts, and quantified the CD34+ myeloid cells in a blind manner. No agreement regarding cytometers, number, reagents, fluorochromes and events recorded were previously established.

Results: The mAb combinations selected to identify BM myeloid CD34+ cells was highly variable: CD45 was included in 45/47 (95.7%) files, CD34/CD45/CD117 in 37 (78.7%), and HLA-DR in 19 files (40.4%). The FCI assays used in every centre were quite different, but a very good concordance among the 6 FCI observers for CD34+ cell enumeration was observed with the intraclass correlation coefficient (0.720). Considering the specific analysis of each sample, 2 groups were distinguished. Twenty-five cases (53.2%) exhibited the highest concordance, with an interlaboratory coefficient of variation (CV) <30%, or a difference among observers <0.5%. In the remaining 22 cases, the concordance among participants was low. Comparison of the characteristics of high and low agreement assays showed no statistical differences. In order to determine whether the lower concordance obtained among observers after the analysis of these 22 samples could be improved, we designed a protocol with specific guidelines with gating strategies for identification of the CD34+ myeloid cells in several consecutive steps. The strategy was designed for samples with the CD34/CD45/CD117 mAb combination (n=15), with or without HLADR. Using this strategy, the interlaboratory CV improved in 7/15 cases (46.66%), but no improvement was obtained in the remaining 8 cases. Finally, the percentage of CD34+ myeloid cells was reproducible in 32/47 cases (68%) diagnosed with low risk MDS. Conclusions: Despite different methodology, accuracy among observers in CD34 cell quantification is feasible. Selection of appropriate protocols and targeted training would help to enhance the reproducibility of the technique among different observers. However, the intrinsic heterogeneity of these cells in MDS might justify the difficulty for their identification in some cases.