Stepwise discriminant function analysis for rapid identification of acute promyelocytic leukemia from acute myeloid leukemia with multiparameter flow cytometry.

Abstract

Diagnosis of acute promyelocytic leukemia (APL) has been accelerated by multiparameter flow cytometry (MFC). However, diagnostic interpretation of MFC readouts for APL depends on individual experience and knowledge, which inevitably increases the risk of arbitrariness. We appraised the feasibility of using stepwise discriminant function analysis (SDFA) based on MFC to optimize the minimal variables needed to distinguish APL from other acute myeloid leukemia (AML) without complicated data interpretation. Samples from 327 patients with APL (n = 51) and non-APL AML (n = 276) were randomly allocated into training (243 AML) and test sets (84 AML) for SDFA. The discriminant functions from SDFA were examined by correct classification, and the final variables were validated by differential expression. Finally, additional 20 samples from patients with atypical APL and AML confusable with APL were also identified by SDFA method and morphological analysis. The weighed discriminant function reveals seven differentially expressed variables (CD2/CD9/CD11b/CD13/CD34/HLA-DR/CD117), which predict a molecular result for APL characterization with an accuracy that approaches 99% (99.6 and 98.8% for AML samples in training and test sets, respectively). Furthermore, the SDFA outperformed either single variable analysis or the more limited 3-component analysis (CD34/CD117/HLA-DR) via separate SDFA, and was also superior to morphological analysis in terms of diagnostic efficacy. The established SDFA based on MFC with seven variables can precisely and rapidly differentiate APL and non-APL AML, which may contribute to the urgent initiation of all-trans-retinoic acid-based APL therapy.