Dynamic early biomarkers predict outcomes in pediatric Epstein–Barr virus‐associated hemophagocytic lymphohistiocytosis

ABSTRACT Importance Pediatric Epstein–Barr virus (EBV)‐associated hemophagocytic lymphohistiocytosis (HLH) is a life‐threatening disorder, and early identification of poor responders is critical to improving survival. However, no convenient clinical tools exist to predict individual prognosis using dynamic biomarkers during early treatment. Objective To investigate the predictive value of early dynamic changes in plasma biomarkers and to develop a prognostic model for children with EBV‐HLH. Methods This retrospective study enrolled 60 newly diagnosed pediatric EBV‐HLH patients. We analyzed the longitudinal changes in plasma EBV‐DNA (pEBV‐DNA), ferritin, and cytokine levels during the etoposide‐based induction first‐line therapy. Independent prognostic factors were identified using Cox multivariate regression and LASSO, followed by the construction of a prognostic nomogram. Model performance was evaluated through area under the curve (AUC), decision curve analysis, and internal cross‐validation. Results Multivariate analysis identified positive pEBV‐DNA at week 2 (w2) and low decreases in ferritin (ΔFerritin.w2) and interferon (IFN)‐γ (ΔIFN‐γ.w2) as independent predictors of adverse outcomes. A nomogram integrating these three variables demonstrated a superior AUC of 0.834 (vs. 0.677 for pEBV‐DNA.w2 alone, P = 0.003). The model successfully stratified patients into low‐ and high‐risk groups with significantly different 3‐year event‐free survival (84.8% vs. 33.3%, P < 0.001). Interpretation The proposed model, based on dynamic plasma biomarkers, provides a promising tool for the early risk stratification of pediatric EBV‐HLH. Composed of pEBV‐DNA, ΔFerritin, and ΔIFN‐γ at w2, this nomogram can effectively identify patients at high risk of first‐line treatment failure. Given the exploratory nature of this study, these findings require further validation in larger, independent cohorts.