New Insights into VOC Incidence By Analyzing the Single Cell Quantification of HbF per RBC and Mchbs

Vaso-occlusive crisis (VOC), the main complication of sickle cell disease (SCD), is responsible for more than 50% of the associated deaths. The effect of hydroxyurea, the cornerstone treatment for SCD, on the occurrence of VOC is mainly mediated by increasing HbF levels, which is a protective factor by inhibiting hemoglobin S (HbS) polymerization. 1,2 However, HbF level is variable between patients, and its expression is heterogeneous among the red blood cell (RBC) population. 3 As a result, while an increase in HbF levels is associated with better survival, %HbF only partially explains patients' clinical manifestations. The protective effect of HbF is also related to the HbS content; the more HbS, the more HbF is needed to inhibit its polymerization. In this study, we wondered if the single-cell quantification of HbF that we developed could better explain the VOC incidence of SCD compared to %HbF. Patients and method We analyzed data from the FCDREP1 protocol, aiming to obtain a predictive score for VOC at 1 year. Inclusion criteria were SS and S- S-b 0 Thal SCD patients, more than 1 month after a vaso-occlusive crisis, 3 months after a transfusion, and on a stable dose of hydroxyurea (HU) for at least 3 months. After giving their consent (IRB no. 00003835) during a routine blood sampling visit, an additional sample was taken for HbF single cell quantification, providing HbF/cell distribution. VOC were collected prospectively for 1 year. HbF single cell quantification was measured by flow cytometry using a previously published protocol, 4 data are expressed as histograms representing % RBC in each HbF range in pg for the whole cluster. K-means clustering was performed on different combination of variables containing ranges of HbF/cell distribution versus %HbF, with MCHbS (MCH x %HbS) using UMAP for dimensionality reduction (into 2), after we applied Min-Max scaling to preprocess the features, with python 3 on sklearn and umap library. This clustering approach can group data points based on their similarity, providing insights into distinct patterns or subgroups within the dataset. Results One hundred two patients were included with a mean age of 38 ± 11 years, a sex ratio F/M of 0.56, 63% of patients were on HU. Patients had a total of 74 VOC with a mean of 0.71 ± 1.4 VOC per patient in the year. Analyses were performed finding 4 different clusters in different 2 groups. Group 1: clusters identified by the association of HbF/cell distribution and MCHbS (H