Abstract Rationale Septic myocardial dysfunction manifests as either reduced or hyperdynamic left ventricular ejection fraction (LVEF); both extremes associate with poor outcomes. The hyperinflammatory molecular subphenotype also predicts high mortality in sepsis. We hypothesized that hyperdynamic LVEF might associate with hyperinflammatory subphenotype. Methods We analyzed critically ill adults with sepsis enrolled in a prospective molecular sepsis cohort (2011-present) who underwent transthoracic echocardiography within 72 hours of ICU admission. Patients with prior congestive heart failure (CHF) were excluded. We used latent class analysis (LCA) of 22 clinical and biomarker variables, as described previously, to assign a probability of hyperinflammatory subphenotype. We extracted LVEF from clinical echocardiograms as the primary exposure, truncated to 15% and 75% to limit extreme values, and categorized 28-day mortality as the main outcome. The mortality was modeled using multivariable logistic regression with restricted cubic splines (RCS, df = 3), adjusting for age, sex, race, smoking, diabetes, hypertension, admission body mass index, and cancer. Nonlinearity of the LVEF-mortality association was assessed by a likelihood ratio test comparing spline and linear models. For the LCA-available cohort, we used ordinary least squares regression with RCS (df = 3) on LVEF, adjusting for age and sex. To assess monotonicity, we performed a linear trend test and Spearman’s rank correlation. Results Of 1,339 patients meeting inclusion criteria, 1,144 had available LVEF data. After excluding CHF and missing data, 901 patients (78.8%) remained for the mortality analysis. The 28-day mortality demonstrated a significant U-shaped nonlinear association across the LVEF spectrum, with elevated risk at both low and high values (Figure 1A). The nonlinearity was significant (p = 0.0013) after adjustment for covariates. In the LCA-available cohort (n = 239), the relationship between LVEF and predicted probability of hyperinflammatory phenotype appeared visually monotonic and inverse on spline analysis (Figure 1B). Linear trend testing confirmed a significant negative association (β = -0.02 per 5% LVEF, p = 0.014), and Spearman’s correlation supported a weak but significant inverse relationship (ρ = -0.13, p = 0.046). Conclusions Both low and high ends of the LVEF spectrum were associated with excess mortality in sepsis but may represent distinct biological mechanisms. There was a weak negative correlation between LVEF and the probability of hyperinflammatory phenotype, suggesting cardio-inflammatory dysregulation in patients with hypodynamic LVEF. In contrast, hyperdynamic LVEF was not enriched for hyperinflammatory subphenotype. These findings highlight biological heterogeneity in septic myocardial dysfunction and prompt investigation of distinct mechanisms underlying mortality in hyperdynamic LVEF. This abstract is funded by: This work was supported by NIH grants R35HL161196 (Meyer), R01HL155159 (Reilly), R35HL177135 (Calfee), and TL1TR001880 (Meagher, ITMAT).
Hasegawa et al. (Fri,) studied this question.