The Challenge Of Quantifying CD34+ Myeloid Cells In Myelodysplastic Syndromes With Less Than 5% Bone Marrow Blasts. Reproducibility Among 6 Flow Cytometry Observers

In low-risk myelodysplastic syndromes (MDS), the morphological bone marrow (BM) blast cell count between 0 and 2%, and >2 - <5%, has demonstrated prognostic value and is critical for R-IPSS. Flow cytometry immunophenotyping (FCI) provides an accurate way for quantification of the immature BM cell compartment through the identification of CD34+ cells and may contribute to a better characterization of these MDS low-risk categories. However, there is a wide variety of FCI strategies to study the CD34+ cells, without a universal consensus on how many and which markers should be used to reach their best identification. There are some studies evaluating the correlation between FCI and the morphological blast cells count, but it is not clear if FCI allows a good concordance among several observers when the morphological blast count is <5%.

1-To explore the concordance among 6 FCI observers to quantify the CD34+ myeloid BM cells from patients diagnosed with MDS with <5% BM blasts 2- To study the correlation between FCI and morphology for detecting <5% BM blast cells. 3- To determine if the mophological threshold of 2% is reproducible by FCI.

FCI data files from 48 MDS BM samples with <5% blasts were simultaneously and independently evaluated by 6 FCI observers from 6 different Spanish hospitals. According to the WHO criteria patients were distributed as follows: 3 refractory cytopenia with unilineage dysplasia; 13 refractory anemia with ring sideroblasts; 25 refractory cytopenia with multilineage dysplasia; 1 unclassifiable MDS; 2 chronic myelomonocytic leukemia and 4 therapy-related myeloid neoplasms. Each participant contributed with 8 samples and all files were exchanged among them. All of them used the INFINICYT^TM software program for analysis according to their usual strategies. The morphological quantification of BM blast cells was provided by each centre and was blinded to the others. Each centre processed the samples for FCI according to their usual strategies in their clinical practice, without previous agreement on standardization of the protocols used. All centres used the stain-lyse-wash protocol but panels of monoclonal antibodies were different: combinations of 4, 6 and 8 fluorochrome–conjugated monoclonal antibodies were included in 28, 8 and 12 files respectively. Median number of events recorded per file was 157,200 (range 10,000-500,000). The combination CD34/CD45/CD117 was included in 38 files and 20 also associated HLA-DR. The fluorochrome attached to CD34 was PerCP-Cy5 in 26 samples. 8G12 was the clone used in 40 samples.

The degree of agreement among the 6 observers for quantification of CD34+ myeloid cells was evaluated using the intraclass correlation coefficient (ICC). The generalized kappa statistic for multiple rates (κ) calculated the concordance among observers after categorization of quantitative variables. Both the ICC and the generalized κ statistic were interpreted as follows: 0-0.2 poor; 0.3-0.4: fair; 0.5-0.6 moderate; 0.7-0.8 strong; >0.8 almost perfect agreement.

Finally, 47 samples could be evaluated by the FCI observers. The ICC showed a strong agreement among observers (0.720), and also a good concordance on the quantification of CD34+ cells at the critical level of 2% (k=0.587). Regarding the comparison between FCI and morphology, only one participant counted >5% CD34+ cells in a sample. However, the absolute quantification of BM blasts <5% by FCI showed poor agreement with morphology (ICC ranged from 0.106 to 0.458). Indeed, none of the FCI observers could reproduce the new morphological categories using the threshold of 2% BM blasts (k= 0.320).

In our study, FCI seems a reproducible tool to quantifying CD34+ cells in MDS patients with <5% BM blasts, despite of the great heterogeneity of the protocols used. A FCI threshold of 2% CD34+ cells was also reproducible among observers. However, the lack of a precise correlation between the morphological blast cell count and the number of CD34+cells by FCI, illustrates the importance of considering each value independently. Probably, homogenization of the FCI protocols will contribute to improve the correlation between the 2 techniques.

No relevant conflicts of interest to declare.