A Three-Component Immune Framework Stratifies Treatment Response and Escalation in Pediatric Hemophagocytic Lymphohistiocytosis: A Prospective Single-Center Cohort Study.
In a prospective single-center cohort of 180 children with HLH, an integrated framework combining etiology, diagnostic and day-7 cytokine profiles, and lymphocyte subsets internally predicted mortality and treatment escalation with high discrimination.
Open original publication →What the AI sees
In a prospective single-center cohort of 180 children with HLH, an integrated framework combining etiology, diagnostic and day-7 cytokine profiles, and lymphocyte subsets internally predicted mortality and treatment escalation with high discrimination.
Research significance
The study provides observational evidence that etiology-specific immune phenotypes and the day-7 IFN-γ ratio are associated with mortality and escalation; if prospectively confirmed in external cohorts, these markers could support earlier risk-adapted treatment escalation, but the record does not show that biomarker-guided treatment improves outcomes.
Source abstract
BACKGROUND: Pediatric hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory syndrome that can be classified into three principal etiologies (primary HLH [pHLH], Epstein-Barr-virus-driven HLH [EBV-HLH], and secondary HLH [sHLH]) and is associated with disparate outcomes. Treatment escalation decisions remain largely empirical, guided by clinical presentation and conventional laboratory results rather than by an integrated biomarker framework. PROCEDURE: We prospectively enrolled 180 consecutive pediatric patients with HLH at Children's Hospital 1, Ho Chi Minh City, Vietnam (October 2022-October 2025): pHLH, n = 40; EBV-HLH, n = 46 (CD27 intact, n = 19; CD27 dim, n = 27); and sHLH, n = 94. A 15-analyte multiplex cytokine panel and six-marker lymphocyte subset flow cytometry were performed at diagnosis (T0) and Day 7 (T1). Logistic regression models were developed and internally validated using bootstrap optimism correction (1000 iterations), 5-fold stratified cross-validation, and a temporal split. RESULTS: Overall mortality was 31.1% (EBV-HLH 67.4%, pHLH 45.0%, and sHLH 7.4%). Three discrete cytokine and subset phenotypes emerged: EBV-HLH showed massive elevations in IFN-γ and sCD25, with profound CD4+ and double-negative T-cell depletion; pHLH showed elevated tumor necrosis factor-α with preserved subsets; sHLH had low-amplitude profiles with the highest CD4+ counts. Day-7 IFN-γ T1/T0 ratio (threshold 0.376) stratified mortality 6- to 27-fold within each etiology. Cytokine drivers of escalation were etiology specific (sCD25 in pHLH, sCD25+IL-18 in EBV-HLH, and IL-10/IL-6 in sHLH). The integrated three-component model achieved an apparent AUC of 0.963, a bias-corrected AUC of 0.947 (95% CI: 0.92-0.97), a 5-fold CV AUC of 0.942 ± 0.05, and a temporal validation AUC of 0.946-all three methods were convergent. CONCLUSIONS: Integrating etiology, cytokine phenotype, and lymphocyte subset profile provides biology-grounded prediction of mortality and treatment escalation in pediatric HLH. Multicenter external validation is the next essential step.