Predicting development of pediatric chronic immune thrombocytopenia at disease onset using a statistical risk model

Abstract

Abstract Immune thrombocytopenia (ITP) is associated with a variable and unpredictable clinical course in children, including a spectrum of bleeding and systemic symptoms in the months following diagnosis. Although many children will have spontaneous resolution of the disease prior to 1 year, up to 30% will go on to develop chronic disease. The known predictors for developing chronic ITP are limited, making clinical management and guidance during this early course of disease highly challenging. In addition, the pathophysiology of immune dysregulation in ITP is complex, with multiple variables likely contributing to the development of chronic disease. We aimed to create a statistical model to predict the development of chronic ITP. Using a retrospective training cohort of 611 children with ITP from 2 institutions and 2 validation cohorts comprised of 161 children, we developed and validated a multivariable logistic regression model and found that age; sex; immunoglobulin G (IgG), IgA, and IgM levels; presenting platelet count; presenting lymphocyte count; known secondary cause at diagnosis; and direct antiglobulin test positivity were useful in predicting chronic ITP. The external validations demonstrated consistent discriminative performance and clinical utility. The model is available for use at https://opal.shinyapps.io/citp-rm/. A chronicity prediction tool for use at the time of ITP diagnosis will better equip hematologists to counsel patients and families and engage in appropriate treatment strategies for individual patients earlier in their course.