Identification of red cell distribution width as a key predictor of leukemic transformation in polycythemia vera and essential thrombocythemia: A machine learning approach involving 10,560 cases
Machine learning predicts leukemic transformation risk in polycythemia vera patients, suggesting RDW as a significant prognostic marker.
Key Points
This research aims to identify predictive factors for leukemic transformation in patients with polycythemia vera and essential thrombocythemia using machine learning.
Used a large discovery cohort from electronic medical records of Maccabi Healthcare Services.
Applied machine learning to establish a predictive model based on baseline clinical and laboratory parameters.
Utilized ROC curve analysis to determine optimal cutoff points for predicting leukemic transformation.
Identified higher RDW as a significant risk factor for leukemic transformation.
Validated predictive models showed robust risk stratification with AUCs ranging from 0.75 to 0.81 over 10-20 years.
Confirmed the independent predictive value of RDW and other risk factors, particularly for female patients.