A56-30 Premorbid Conditions Are Associated With Phenotype Classification in Patients With E. Coli Bacteremia

Rationale

The two established phenotypes of sepsis—Hyperinflammatory and Hypoinflammatory—describe markedly different circulating molecular responses to infection. Enterobacterales infections are associated with Hyperinflammatory sepsis. E. coli, the most prevalent Enterobacterales pathogen, exhibits a sharply bimodal distribution of phenotype probabilities, enabling high-certainty classification. To reduce organism-driven heterogeneity and address gaps in understanding host determinants of phenotype assignment, we analyzed patients with E. coli bacteremia to identify premorbid comorbidities associated with phenotype classification.

Methods

We performed a retrospective cohort study of mechanically ventilated adults with sepsis and E. coli bacteremia at the University of Michigan (2014-2020). Phenotypes were assigned using a validated clinical classifier that yields a probability of Hyperinflammatory phenotype. Primary analyses were restricted to high-confidence phenotype assignments (Hyper probability ≥0.75 versus Hypoinflammatory probability ≤0.25). Comorbidities present on admission were derived from ICD codes coalesced into previously validated standardized groups. In single-variable analyses, chi-square tests were used to compare the prevalence of premorbid conditions between phenotypes. In multivariable analyses, logistic regression was used to evaluate independent associations between premorbid conditions and Hyperinflammatory phenotype, adjusting for age, sex, race, and BMI. Sensitivity analyses repeated these models using a less restrictive phenotype cutoff (Hyperinflammatory probability ≥0.50).

Results

We identified 215 ICU admissions with E. coli bacteremia; 154 met high-confidence phenotype criteria (Hyper n = 73; Hypo n = 81). Phenotype probabilities were well separated (Hyper 0.92 [0.85-0.97] versus Hypo 0.07 [0.01-0.15], p < 0.001). In single-variable analyses, no comorbidity differed significantly by phenotype. In multivariable analysis, premorbid cirrhosis was independently associated with Hyperinflammatory phenotype (adjusted OR 3.36, 95% CI 1.20-9.41; p = 0.021). Higher BMI was independently associated with lower odds of Hyperinflammatory phenotype (adjusted OR 0.94 per kg/m², 95% CI 0.89-0.99; p = 0.025) (Figure). Using a 0.50 cutoff (Hyper n = 109; Hypo n = 106), the cirrhosis association attenuated (adjusted OR 2.08, 95% CI 0.96-4.53; p = 0.064), while BMI remained inversely associated (adjusted OR 0.96, 95% CI 0.92-1.00; p = 0.030).

Conclusion

Premorbid cirrhosis was associated with Hyperinflammatory phenotype in mechanically ventilated patients with sepsis due to E. coli bacteremia, with the strongest signal observed in high-confidence phenotype assignments. Higher BMI was independently associated with lower odds of Hyperinflammatory phenotype. Together, these findings demonstrate that baseline host biology meaningfully shapes sepsis host-response phenotypes in gram-negative bloodstream infection. Future work should determine how premorbid hepatic and metabolic conditions influence these phenotypes.

This abstract is funded by: None