Abstract
BACKGROUND: Lower respiratory infections are a significant cause of morbidity and mortality in children. The aim of this study was to determine whether cytokine levels measured in plasma at the time of admission to the hospital can predict disease etiology or severity. METHODS: Blood was collected from pediatric inpatients, and cytokine levels were determined by cytokine multiplex analyses. Plasma cytokine concentrations were then analyzed using logistic regression and machine learning approaches to determine if we could accurately predict if a child would require longer-term hospitalization (≥5 days), intensive care, or exhibit hypoxemia (SpO(2) < 90%). RESULTS: A total of 159 patients were enrolled, and 59 cytokines were assessed in relation to the type of infection and severity. The most prevalent viral infections were human rhinovirus/enterovirus (hRV/EV; 24.4%), respiratory syncytial virus (RSV; 21.8%), and influenza virus (16.7%). Several cytokines (CHI3L1, IL-1Rα, IL-6, G-CSF, MCP-1, and MIP-1α) were elevated in severe pneumonia cases, regardless of disease etiology. Predictors of duration in RSV cases were distinct from other causes, with a predominance of type-2 immune response. Cytokines such as chitinase-3-like-1 (CHI3L1), pentraxin-3, osteopontin, and IL-20 correlated with severity across multiple groups. Plasma levels of IL-6, MMP-2 and LIGHT could be employed to separate viral vs. community acquired pneumonia (CAP). In influenza cases, longer-term hospitalization and ICU admission could be predicted based on two cytokines, CHI3L1 and sTNFR1. RSV severity was closely correlated with levels of MIP-1α, IL-26, G-CSF, and IFNβ. CONCLUSIONS: This study highlights the heterogeneity of immune responses to severe pneumonia and provides new groupings of cytokines which may distinguish between viral and non-viral pneumonia.