C54-01 In Patients With Sepsis, a Hyperinflammatory Phenotype Is Associated With Gram-Negative Bacteraemia and Increased Mortality

Abstract Rationale Biological heterogeneity may account for the failure of pharmacological therapies in sepsis. Identification of sepsis phenotypes provides an opportunity for the alignment of interventions with disease biology and enriched trial enrolment. Hypoinflammatory and hyperinflammatory phenotypes have been described in sepsis cohorts and have been reported to be associated with bacteraemia 1, 2. The aim of this project was to investigate phenotypes in a sepsis cohort recruited in the United Kingdom. Methods TEST-IT was a multi-centre observational trial which recruited critically ill patients with suspected sepsis. We applied unsupervised latent class analysis to plasma biomarkers (IL-6, IL-8, sTNFR-1, ICAM-1, Protein C) and clinical data (PaO2/FiO2, highest bilirubin, highest creatinine, lowest platelets, lowest mean arterial pressure) to identify the optimal number of classes. Association with 28-day mortality was assessed using logistic regression adjusted for potential confounders (age, sex, PaO2/FiO2, suspected sepsis source, blood culture positivity, renal replacement therapy and vasopressors). Results A two-class model was the optimal fit; based on class-specific biomarker profiles, 197 (33%) classified as hyperinflammatory and 407 (67%) as hypoinflammatory (Figure 1). Bacteraemia was more common in the hyperinflammatory phenotype (13.3% vs 6.2%; OR 2.33, 95% CI 1.30-4.15; p = 0.003). Among monomicrobial blood culture isolates (n = 46), gram-negative infection was more common in the hyperinflammatory phenotype (OR 7.03, 95% CI 1.62-38.23; p = 0.006). Mortality at 28 days was significantly higher in the hyperinflammatory phenotype (38.6% vs 12.2%; unadjusted OR 4.54, 95% CI 3.00-6.87; p 0.001; adjusted OR 4.68, 95% CI 2.87-7.62; p 0.001). Figure 1. Mean z-scores for determinant variables plotted for each phenotype. The dashed horizontal line marks a z-score of 0. Positive values indicate above mean levels, and negative values below mean levels. Conclusions Two molecular phenotypes were found in this cohort of sepsis patients, with higher mortality, more bacteraemia, and more gram-negative infections in the hyperinflammatory phenotype. These findings support the concept that the pathogen is an important determinant of sepsis phenotypes. References: 1. Sinha P. et al. (2023). Lancet Respir Med; 11(11):965. 2. Chanderraj R. et al. (2024). Am J Respir Crit Care Med; 209:A4995. This abstract is funded by: TEST-IT was funded by Innovate UK. RRM is supported by the Medical Research Council. CSC received support from National Institutes of Health (R35HL177135).