GOALS: Cell population data (CPD) are new morphologic parameters including volume, conductivity, and five light scattering characteristics used for leukocyte classification by an automated hematology analyzer, the UniCel DxH 800. We developed a discriminating CPD model to predict the leukemia lineage during routine complete blood cell count (CBC). PROCEDURES: We analyzed the CPD of 405 blood samples containing more than 10% blasts that were randomly divided into test and validation sets. With the test set, we produced a model for categorizing acute lymphoblastic leukemia (ALL) or acute promyelocytic leukemia (APL), using ranges of the CPD and regarding the remainder as non-APL acute myeloid leukemia. We verified these models against the validation set. RESULTS: In the test set, we formulated a 21-parameter model which identified 43 of 47 ALL cases (91.5% sensitivity) and ruled out 151 of 156 other leukemia cases (96.8% specificity), and a 13-parameter model which distinguished all 10 APL cases (100% sensitivity) and excluded 193 other leukemia cases (100% specificity). In the validation set, the ALL model showed 85.1% sensitivity and 94.2% specificity, and the APL model 100% sensitivity and 100% specificity. CONCLUSIONS: This study demonstrated a new solution for predicting blast lineage using the CPD on a CBC and leukocyte differential.
GOALS: Cell population data (CPD) are new morphologic parameters including volume, conductivity, and five light scattering characteristics used for leukocyte classification by an automated hematology analyzer, the UniCel DxH 800. We developed a discriminating CPD model to predict the leukemia lineage during routine complete blood cell count (CBC). PROCEDURES: We analyzed the CPD of 405 blood samples containing more than 10% blasts that were randomly divided into test and validation sets. With the test set, we produced a model for categorizing acute lymphoblastic leukemia (ALL) or acute promyelocytic leukemia (APL), using ranges of the CPD and regarding the remainder as non-APL acute myeloid leukemia. We verified these models against the validation set. RESULTS: In the test set, we formulated a 21-parameter model which identified 43 of 47 ALL cases (91.5% sensitivity) and ruled out 151 of 156 other leukemia cases (96.8% specificity), and a 13-parameter model which distinguished all 10 APL cases (100% sensitivity) and excluded 193 other leukemia cases (100% specificity). In the validation set, the ALL model showed 85.1% sensitivity and 94.2% specificity, and the APL model 100% sensitivity and 100% specificity. CONCLUSIONS: This study demonstrated a new solution for predicting blast lineage using the CPD on a CBC and leukocyte differential.
Entities:
Keywords:
ALL; AML; Acute leukemia; and Leukocytes morphology