Diagnostic and Prognostic Value of Ferritin and Soluble Interleukin-2 Receptor α in Critically Ill Children with...

Background/Objectives: Ferritin and soluble interleukin-2 receptor α (sIL-2Rα) are HLH-related biomarkers, but their comparative discrimination in pediatric critical illness is uncertain. We primarily compared their discrimination for adjudicated HLH in hyperferritinemic PICU children and secondarily assessed PICU mortality. Methods: This prospective single-center study enrolled 75 children aged 1 month-18 years with ferritin ≥500 ng/mL: 25 each with HLH, sepsis/septic shock, or other diseases. Both biomarkers were measured from the same admission sample. Research-use sIL-2Rα was reported in ng/mL without cross-calibration to the HLH-2004 U/mL scale; adjudicators were blinded to sIL-2Rα, whereas ferritin informed the reference process. Analyses used ROC, Firth regression, targeted post hoc paired AUC tests with Holm adjustment, bootstrap optimism correction, and cross-validation. Results: For adjudicated HLH, ferritin showed greater discrimination than sIL-2Rα (AUC, 0.898 vs. 0.690; ΔAUC, 0.208; Holm-adjusted p = 0.019). The exploratory 1614-ng/mL ferritin cutoff had 92% sensitivity and 76% specificity. Although sIL-2Rα remained associated with HLH after ferritin adjustment (OR, 1.089 per ng/mL; p = 0.002), adding it did not significantly improve discrimination beyond ferritin (p = 0.663). For secondary mortality, ferritin showed greater discrimination than sIL-2Rα (AUC, 0.754 vs. 0.495; Holm-adjusted p = 0.019). Adding ferritin to PRISM III increased the apparent AUC from 0.725 to 0.811, but the comparison was not statistically significant (p = 0.105); the bootstrap optimism-corrected AUC was 0.800. Conclusions: In this hyperferritinemic PICU cohort, ferritin showed greater discrimination for adjudicated HLH than the studied sIL-2Rα assay, and adding sIL-2Rα did not significantly improve discrimination. Findings should be interpreted in the context of the studied assay and reference process. The exploratory mortality model requires external validation.