Abstract
BACKGROUND: To explore the diagnostic performance of interleukin (IL)-6 and IL-10 in discriminating Gram bacteria types and predicting disease severity in intensive care unit (ICU)-hospitalized pediatric sepsis patients. METHOD: We retrospectively collected Th1/Th2 cytokine profiles of 146 microbiologically documented sepsis patients. Patients were categorized into Gram-positive (G+) or Gram-negative (G-) sepsis groups, and cytokine levels were compared. Subgroup analysis was designed to eliminate the influence of other inflammatory responses on cytokine levels. RESULTS: After propensity score matching, 78 patients were matched and categorized according to Gram bacteria types. Compared with G+ sepsis, IL-6 and IL-10 were significantly elevated in G- sepsis (p < 0.05). Spearman test proved the linear correlation between IL-6 and IL-10 (r = 0.654, p < 0.001), and their combination indicators (ratio and differences) were effective in identifying G- sepsis. In the subgroup analysis, such cytokine elevation was significant regardless of primary infection site. However, for patients with progressively deteriorating organ function [new or progressive multiple organ dysfunction syndrome (NPMODS)], differences in IL-6 and IL-10 levels were less significant between G+ and G- sepsis. In the receiver operating characteristic (ROC) curves of the G- sepsis group, the area under the curve (AUC) value for IL-6 and IL-10 was 0.679 (95% CI 0.561-0.798) and 0.637 (95% CI 0.512-0.762), respectively. The optimal cutoff value for diagnosing G- sepsis was 76.77 pg/ml and 18.90 pg/ml, respectively. While for the NPMODS group, the AUC for IL-6 and IL-10 was 0.834 (95% CI 0.766-0.902) and 0.781 (95% CI 0.701-0.860), respectively. CONCLUSION: IL-6 and IL-10 are comparably effective in discriminating G+/G- sepsis in pediatric intensive care unit (PICU) patients. The deteriorated organ function observed in ICU patients reveals that complex inflammatory responses might have contributed to the cytokine pattern observed in severe sepsis patients, therefore confounding the discriminating efficacy of Th1/Th2 cytokines in predicting Gram bacteria types.