Data-driven phenotypic subclassification of older patients with AML. (Ai) Seven cell types were identified using the first five 8-color combinations of the EuroFlow panel for the characterization of AML, in a total of 15 healthy donors (HD). Each cell type was identified using the backbone markers present in all 8-color combinations (ie, CD34, CD45, CD117, and HLADR). (Aii) Expression percentage of 19 different markers (CD22, CD34, CD45, CD117, CD105, CD71, CD36, CD38, HLADR, CD13, CD15, CD33, CD64, CD35, CD14, CD300e, CD11b, CD16, and CD10). Markers were analyzed per healthy donor to create and define normal hematopoiesis. (Aiii) After obtained the reference of the normal hematopoiesis, principal component (PC) analysis was used to calculate the centroid for each cell type was calculated considering the mean values of PC1 and PC2. (Aiv) The percentage of expression for each marker was obtained in blasts of 252 patients with AML enrolled in the PETHEMA/FLUGAZA clinical trial (NCT02319135). (Av) After projecting each case onto the matrix generated from healthy adults, the distance between blasts and normal cell types was calculated, and the shortest distance identified the corresponding differentiation stage. (B) Pie chart showing the number of patients and relative distribution according to the 7 differentiation stages defined in healthy adults. Because of low numbers, patients were regrouped according into HPC-like, as well as granulocytic-, monocytic-, and erythroid-like, AML. (C) Concordance between the phenotypic and the FAB subclassification of patients with AML according to the differentiation stage of blasts. (D) OS of patients with HPC-like vs maturing AML subtypes treated with azacitidine. P values were determined by the 2-sided log-rank test. (E) Multivariate analysis of OS in the FLUGAZA cohort, including the data-driven phenotypic classification (HPC- vs maturing-like) and the Medical Research Council (MRC) risk stratification (favorable/intermediate vs adverse). (F) OS of patients from the TUH external validation cohort with HPC-like vs maturing AML subtypes treated with azacitidine. P values were determined by the 2-sided log-rank test. (G) Multivariate analysis of OS in the TUH external validation cohort, including the data-driven phenotypic classification (HPC- vs maturing-like), performance status (PS), age (>80 vs ≤80 years), concentration of white blood cells (>50 × 10⁹/L or ≤50 × 10⁹/L), and the MRC risk stratification (favorable/intermediate vs adverse). FAB, French-American-British; HR, hazard ratio; MRC, Medical Research Council; NRBC, nucleated red blood cell; TUH, Toulouse University Hospital; WBC, white blood count.