Abstract Rationale Hyperoxemia worsens outcomes in critical illness, but its effects are heterogeneous. Subphenotypes of sepsis, hyperinflammatory and hypoinflammatory, have divergent metabolomic signatures, suggesting differences in oxidative stress and energy metabolism. Methemoglobin, the oxidized form of hemoglobin, is a marker of oxidative stress. Routinely measured at the University of Michigan, we leveraged it to test three hypotheses: (1) hyperoxemia confers greater mortality risk in the hyperinflammatory phenotype; (2) that the hyperinflammatory phenotype shows greater oxidative stress, with higher methemoglobin for any given oxygen level; and (3) that the harm of hyperoxemia is mediated through oxidative stress. Methods We studied 4,684 mechanically ventilated adults with sepsis admitted to the University of Michigan (2016-2020), classified as hypoinflammatory (3,379) or hyperinflammatory (1,305) using a validated clinical classifier. Exposures were the mean arterial oxygen (PaO2) and mean methemoglobin within the first 24 hours of ventilation. To test whether hyperoxemia confers greater mortality risk in the hyperinflammatory phenotype, we fit a multivariable logistic regression for 30-day mortality including PaO2, phenotype, and their interaction, adjusted for age, APACHE IV, Charlson index, sex, and ICU. To test whether the hyperinflammatory phenotype exhibits greater oxidative stress, we modeled mean methemoglobin as a function of phenotype, PaO2, and their interaction with the same covariates. To assess for mediation by oxidative stress, we added methemoglobin to the mortality model to detect attenuation of the PaO2 × phenotype interaction and improvement in model fit. Results Hyperoxemia increased mortality preferentially in the hyperinflammatory phenotype: each 10 mm Hg rise in PaO2 increased the odds of death by 7% (OR 1.07, 95% CI 1.03-1.12; Figure 1a). Across all PaO2, methemoglobin was higher in hyperinflammatory patients (difference 0.23% 95% CI 0.09-0.37) and rose more steeply with increasing PaO2 (0.036% CI 0.018-0.054 % per 10 mm Hg, Figure 1b). Methemoglobin independently predicted death (OR 1.22 95% CI 1.02-1.46) and attenuated the PaO2 × phenotype interaction (OR 1.01 95% CI: 1.00-1.02), improving model fit (p = 0.028), consistent with mediation through oxidative stress. Conclusions Patients with hyperinflammatory sepsis have greater oxidative stress and heightened vulnerability to oxygen-associated mortality. Methemoglobin predicts mortality and partially mediates the harm of hyperoxemia, implicating oxidative stress as a mechanistic driver of clinical outcomes. This abstract is funded by: Department of Veterans Affairs
Chanderraj et al. (Fri,) studied this question.