Title
Year of Publication
Author
Number of Pages
Date Published
ISSN Number
Abstract
<p><strong>OBJECTIVE: </strong>Heterogeneity has hampered sepsis trials, and sub-phenotyping may assist with enrichment strategies. However, biomarker-based strategies are difficult to operationalize. Four sub-phenotypes defined by distinct temperature trajectories in the first 72 hours have been reported in adult sepsis. Given the distinct epidemiology of pediatric sepsis, the existence and relevance of temperature trajectory-defined sub-phenotypes in children is unknown. We aimed to classify septic children into de novo sub-phenotypes derived from temperature trajectories in the first 72 hours, and compare cytokine, immune function, and immunometabolic markers across subgroups.</p>
<p><strong>METHODS: </strong>This was a secondary analysis of a prospective cohort of 191 critically ill septic children recruited from a single academic pediatric intensive care unit. We performed group-based trajectory modeling using temperatures over the first 72 hours of sepsis to identify latent profiles. We then used mixed effects regression to determine if temperature trajectory-defined sub-phenotypes were associated with cytokine levels, immune function, and mitochondrial respiration.</p>
<p><strong>RESULTS: </strong>We identified four temperature trajectory-defined sub-phenotypes: hypothermic, normothermic, hyperthermic fast-resolvers, and hyperthermic slow-resolvers. Hypothermic patients were less often previously healthy and exhibited lower levels of pro- and anti-inflammatory cytokines and chemokines. Hospital mortality did not differ between hypothermic children (17%) and other sub-phenotypes (3 to 11%; p = 0.26).</p>
<p><strong>CONCLUSIONS: </strong>Critically ill septic children can be categorized into temperature trajectory-defined sub-phenotypes that parallel adult sepsis. Hypothermic children exhibit a blunted cytokine and chemokine profile. Group-based trajectory modeling has utility for identifying subtypes of clinical syndromes by incorporating readily available longitudinal data, rather than relying on inputs from a single timepoint.</p>
DOI
Alternate Title
PMID



