Serum cytokine panels in pediatric clinical practice

Abstract

Background: Cytokines are soluble signaling proteins that regulate inflammation and coordinate immune responses. Serum cytokine panels are increasingly used in medical practice, yet our understanding of cytokines as biomarkers for disease remains limited.

Objective: We sought to analyze real-world single-center use of a multiplexed cytokine panel, correlate its results with diagnosis and severity, and explore its use in pediatric practice.

Methods: A multiplexed cytokine panel, able to return same-day results, was implemented in April 2020 at the Children's Hospital of Philadelphia (Philadelphia, Pa) and its performance was validated for clinical use. Coded patient data were collected using the REDCap database, and correlations between cytokine levels and outcomes of interest were analyzed retrospectively.

Results: Cytokine levels correlate with acuity of care, with patients admitted to the pediatric intensive care unit having the highest cytokine values. Patients with familial hemophagocytic lymphohistiocytosis (fHLH) showed prominent peaks in IFN-γ, IL-10, and TNF, whereas patients with sepsis exhibited high IL-6 and IL-8 with relatively modest IFN-γ. Cytokine release syndrome (CRS) after chimeric antigen receptor T-cell therapy often demonstrated pan-panel positivity at peak levels, with a similar pattern as that of fHLH. A ratio of [IFN-γ] + [IL-10]/[IL-6] + [IL-8] levels was able to distinguish fHLH and CRS from severe sepsis.

Conclusions: Cytokine levels correlate with severity of illness and can help differentiate between syndromes that present similarly, including fHLH and CRS compared with sepsis. Cytokine panels can be used as biomarkers to inform diagnosis and management decisions, but significant work remains to dissect complex clinical patterns of disease.

Keywords: Cytokines; cytokine release syndrome; hemophagocytic lymphohistiocytosis; inflammation.

FIG 1.

A, Underlying disease category associated with each cytokine panel was extracted from the medical record. B, All cytokine profiles in this data set were then combined into a single figure to exhibit dynamic ranges of each cytokine (n = 465). Cytokine values lower than the normal threshold are set to 1 for visualization purposes only. KD, Kawasaki disease

FIG 2.

A, UMAP analysis of all cytokine panels and colored by admission status. B-D, All cytokine panels (n = 465) were then stratified by admission status: outpatient (n = 106; Fig 2, B), inpatient (n = 196; Fig 2, C), and pediatric ICU (n = 218; Fig 2, D). Cytokines that were significantly higher in ICU patients compared with inpatients include IFN-γ (P < .005), IL-1β (P < .05), IL-6 (P < .0001), IL-8 (P < .0001), IL-13 (P < .05), and TNF (P < .00001). E, IL-6, IL-8, IL-10, and IFN-γ were separately analyzed and stratified by admission and severity. ICU defined as patients admitted to the ICU without significant intervention, and severe ICU as those admitted to the ICU with significant interventions (vasopressors, supplemental oxygen, positive pressure ventilation, intubation/mechanical ventilation, or increased settings in patients on baseline respiratory support). Increased levels of clinical severity were associated with increasing cytokine levels, including IL-6 (P < .001), IL-8 (P < .001), IL-10 (P < .01), and IFN-γ (P < .001) (by 1-way ANOVA). Data in Fig 2, B–E, are represented as violin plots with median, 25%, and 75% quartiles indicated by dotted lines. UMAP, Uniform manifold approximation and projection. **P < .01, ***P < .001

FIG 3.

Cytokine profiles of patients diagnosed with SARS-CoV-2–associated disease were extracted from our data set and stratified into COVID-19 (n = 49) and MIS-C (n = 96). If a patient had multiple cytokine panels, the maximum cytokine panel was used. Compared with patients with COVID-19, those with MIS-C exhibited higher IFN-γ (P < .05), IL-10 (P < .05), IL-13 (P < .05), and TNF (P < .005). Data are represented as violin plots with median, 25%, and 75% quartiles indicated by dotted lines.

FIG 4.

The initial data set was enriched with patients with diagnoses of fHLH and severe sepsis. A, Cytokine panels of patients with HLH (n = 11). B, Heat map of correlation coefficients for cytokines and ferritin in patients with HLH. C, Cytokine panels of patients with severe sepsis (n = 19). D, Ferritin levels of patients with HLH and severe sepsis (P value is nonsignificant). E, For each patient, a ratio of [IFN-γ] + [IL-10]/[IL-8] + [IL-6] was calculated, with a high ratio favoring HLH and a low ratio favoring sepsis. F, ROC curve for the cytokine ratio in Fig 4, E (C statistic, 0.9761). Data in Fig 4, A, C, and E, are represented as violin plots, with median, 25%, and 75% quartiles indicated by dotted lines. Underlying genetic and microbial diagnoses are listed in the corresponding Results section. ROC, Receiver-operating characteristic.

FIG 5.

The initial data set was enriched with patients with a diagnosis of CRS post-CART. A, Peak cytokine profile for every patient with CRS (n = 39). B-D, Cytokine panels were stratified by admission status: outpatient (n = 3; Fig 5, B), inpatient (n = 19; Fig 5, C), and ICU (n = 17; Fig 5, D). Comparing non-ICU inpatients to ICU patients, the following cytokines were significantly higher in ICU patients with CRS: IFN-γ (P < .005), IL-1β (P < .05), IL-4 (P < .05), IL-6 (P < .005), IL-8 (P < .005), IL-12-p70 (P < .05), IL-13 (P < .05), and TNF (P < .005). Data in Fig 5, A–D, are represented as violin plots with median, 25%, and 75% quartiles indicated by dotted lines. E, For each patient, a ratio of [IFN-γ] + [IL-10]/[IL-8] + [IL-6] was calculated, with a high ratio favoring CRS and a low ratio favoring sepsis. F, ROC curve for the cytokine ratio in Fig 5, E (C statistic, 0.946). ROC, Receiver-operating characteristic.

FIG 6.

A and B, Medical documentation was evaluated to determine whether cytokine panel results changed/informed diagnosis (Fig 6, A) and changed/informed management (Fig 6, B). C, If management was informed, the medical record was reviewed for how management was directed. Definitions can be found in the corresponding Methods section. JAK, Janus kinase; PLEX, plasma exchange.