Construction of a forest plot prediction model based on Lasso regression for Epstein-Barr virus associated hemophagocytic lymphohistiocytosis in children

Abstract Objective To identify risk factors for the progression of Epstein-Barr virus(EBV) infection to Epstein-Barr virus-associated hemophagocytic lymphohistiocytosis (EBV-HLH) and guide clinical intervention by analyzing the clinical data and laboratory examination between infectious mononucleosis and EBV-HLH infection using a forest plot prediction model. Methods Clinical data and laboratory tests of children with "Epstein-Barr virus infection and hemophagocytic lymphohistiocytosis " who were hospitalized in Children's Hospital of Soochow University from January 2019 to December 2024 were collected. A total of 1358 children of infectious mononucleosis associated with EBV (IM group) and 86 children of hemophagocytic lymphohistiocytosis associated with EBV (EBV-HLH group) were included. The differences between the groups were retrospectively analyzed and regression analysis was performed. The proximity matching method was selected for 1:4 matching between the EBV-HLH group and the IM group. The forest plot prediction model was established based on Lasso regression to analyze the clinical differences between the IM group and the EBV-HLH group. Results Lasso regression model screening identified hemoglobin (HB), ferritin (FER), fibrinogen (FIB) and CD3 + CD4 + as hexhibiting good predictive value for EBV-HLH, with areas under the receiver operating characteristic (ROC) curve of 0.904, 0.973, 0.866 and 0.783, and specificities of 0.799, 0.965, 0.802 and 0.892, respectively. The prediction model constructed using HB, FER, FIB, and CD3 + CD4 + showed excellent predictive accuracy. With an optimal cut-off value of F = 56.95, the model achieved a sensitivity of 95.30% and a specificity of 99.70%. Conclusions The early diagnosis of EBV-HLH lacks specific indicators. In this study, a predictive model for EBV-HLH was established using LASSO regression, incorporating four key parameters (HB, FER, FIB, and CD3 + CD4 + T-cell subsets). This model may serve as a screening tool for the early diagnosis of EBV-HLH and provide a diagnostic basis for clinical practice.​