Exhaled breath-based clusters in children with post-COVID condition

Abstract

Pediatric post-COVID condition (PPCC) presents as a heterogeneous disease with a broad spectrum of symptoms. This study aimed to identify distinct phenotypes of PPCC through an unbiased cluster analysis of exhaled metabolites, with the goal of identifying biomarkers to stratify patients. Exhaled breath samples were collected from children with physician-diagnosed PPCC. An unsupervised clustering approach was applied to the exhaled breath metabolites, and the resulting clusters were compared with clinical variables. Sparse partial least squares-discriminant analysis (sPLS-DA) was applied to find most discriminative metabolites between clusters. A total of 54 children were included and categorized into two clusters. Compared to Cluster 1 (n= 38), Cluster 2 (n= 16) consisted predominantly of older girls (69%) with a median age of 16 years and exhibited more severe PPCC-related outcomes, including higher PROMIS fatigue scores. Six volatile organic compounds (VOCs) were identified as biomarkers that effectively differentiated the two clusters. These VOCs, previously reported in the literature, highlight metabolic and inflammatory disruptions and demonstrated high discriminatory performance (area under the receiver operating characteristic curve (AUROCC) = 1). This study found two distinct phenotypes of PPCC, and identified six discriminating VOCs, underscoring the potential of VOCs as non-invasive biomarkers for disease stratification in PPCC. While it could be a building block towards a better understanding of the metabolic disruptions underlying PPCC, further research with larger patient cohorts is necessary to elucidate the mechanisms driving these differences.