Abstract Introduction The Human Immune Dysregulation Evaluation Framework (Hi-DEF) uses whole blood gene expression to quantify myeloid and lymphoid immune dysregulation in critical illness (Moore et al. Nature Medicine). Because this framework is based on multiple prior endotypes in critical illness, we hypothesized that Hi-DEF would overlap significantly with the previously-identified protein-based hyper- and hypo-inflammatory Latent Class Analysis (LCA) sub-phenotypes. Methods We assessed the overlap between LCA and Hi-DEF across four cohorts: (1) the Stanford ICU biobank (N = 124), (2) the MESSI cohort (N = 149) an ICU sepsis cohort from the University of Pennsylvania, (3) the CAF-PINT trial (N = 268), a clinical trial evaluating glycemic control in the pediatric ICU, and (4) publicly available data from the UCSF EARLI cohorts (GSE236892; N = 189). We calculated Hi-DEF scores using the existing transcriptomic signature. LCA sub-phenotypes were assigned by providers at each site using de novo LCA (EARLI and CAF-PINT) or the IL-6 (Stanford) or IL-8 (MESSI) 3-biomarker parsimonious classifiers. The association of myeloid and/or lymphoid dysregulation and LCA sub-phenotype was assessed using Fisher’s exact test. Across all cohorts, a composite Hi-DEF score was generated by combining myeloid and lymphoid scores using principal component analysis, generating a best-fit line. The association of the combination with LCA status was assessed through area under the receiver operating characteristic (AUROC) and logistic regression meta-analysis. Results Across all cohorts, LCA was associated with increased myeloid and lymphoid dysregulation (framework plot for EARLI pictured in Figure 1A, results similar across sites). Across all cohorts (N = 735), dysregulation on either axis was highly associated with hyper-inflammatory LCA assignment (OR 5.5, 95% CI 3.3-9.6, p = 7.9e-15), with the with the majority of hyper-inflammatory patients identified as system-wide dysregulation by Hi-DEF (Figure 1B). This led us to evaluate the association of a composite HI-DEF inflammatory score using a combination of myeloid and lymphoid dysregulation. This composite HI-DEF score was associated with LCA status across all sites with AUROC for discriminating hyper-inflammatory status ranging from 0.69-0.86 and summary log(OR) of 0.84 (95% CI 0.46 - 1.22, p 0.0001, Figure 1C). Conclusion The Hi-DEF immune framework is associated with LCA sub-phenotype across four independent ICU cohorts. These results highlight the similar biologic information captured by LCA and gene expression-based sub-phenotyping schemas. This abstract is funded by: K12TR004930
Moore et al. (Fri,) studied this question.