Decreased Left Ventricular Ejection Fraction is Associated with Increased Mortality in Sepsis: A Retrospective Cohort Study

Septic cardiomyopathy (SCM) is characterized by transient, non-ischaemic cardiac dysfunction during an episode of sepsis. This clinical entity is both common, and poorly understood, and has been the subject of debates around diagnosis and management.1, 2 The reported incidence of SCM varies widely, from 10% to 70%.3 While early studies suggested that SCM was associated with lower mortality,1, 3 more recent studies have shown an association with significantly increased mortality.4 The uncertainty regarding the prognostic impact of SCM may be due in part to the small and heterogeneous sample of prior studies, and lack of a standardized definition.3, 5-7 To address this question, we conducted the largest-yet retrospective study on the prognostic relevance of SCM with the aim to (i) assess the prevalence of SCM in a large cohort of patients admitted with sepsis or septic shock, and (ii) assess the prognostic implications of new-onset SCM, specifically the presence of an association with increased mortality in this population. In this retrospective cohort study, we evaluated patients admitted to the University of Colorado Hospital with a diagnosis of sepsis or septic shock between 1 January 2015 and 28 February 2022. We automatically abstracted clinical data from the local electronic medical record (EMR) (Epic, May 2023). Data included age, sex, body mass index (BMI), lactate, troponin, and demographic information, as well as ejection fraction. Troponin I values were converted to high-sensitivity troponin values by multiplication by 1000. Mortality data were obtained from the local EMR and corroborated by the Colorado Department of Public Health and Environment, the state mortality database. We included patients that underwent at least one transthoracic echocardiogram (TTE) prior to the index hospitalization (baseline) and at least one TTE during the index hospitalization. We defined SCM as a new-onset left ventricular ejection fraction (LVEF, two-dimensional calculation) of ≤50% and a ≥10% decline in LVEF compared to baseline.1, 2 Patients with known heart failure with reduced ejection fraction, a previously documented LVEF of ≤50%, or a diagnosis of acute coronary syndrome or myocarditis were excluded. The pre-specified analysis included a primary outcome of in-hospital mortality and secondary outcomes 30-day mortality, 1- and 2-year mortality, and intensive care unit (ICU) and hospital length of stay. Summary statistics were tabulated using Pearson's chi-squared, Wilson's rank sum, and Fisher's exact tests. We used multivariable logistic regression to estimate associations between SCM status and categorical outcomes of in-hospital mortality, 1-year mortality, 2-year mortality. Multivariable linear regression was used to assess the relationship between presence of SCM and hospital and ICU length of stay. We constructed a Kaplan–Meier curve following a Cox-proportional hazards model to estimate the risk of death over a follow-up period, with right-censoring occurring on 20 May 2022. Model covariates of age, sex, and BMI were selected a priori based on data availability and clinical relevance. We screened 10 384 patients with a diagnosis of sepsis or septic shock between 1 January 2015 and 28 February 2022. Of these, 4515 patients met inclusion criteria. The majority (n = 2386, 53%) were male, and the median age was 65 (interquartile range 54–74) years. Of the 4515 patients included in this analysis, 544 (12%) had SCM (Table 1). In-hospital mortality was 24% in the entire screened population, 16% among those that met inclusion criteria, 17% among SCM patients, and 12% among non-SCM patients (p < 0.001). One-year and 2-year mortality were increased among SCM patients (Table 2). Median length of stay was 9 days for both groups (p > 0.9). In a Kaplan–Meier survival analysis, there was significantly decreased survival among patients with SCM (Figure 1). A Cox-proportional hazards model including covariates of age, sex, and BMI showed a hazard ratio for mortality of 1.26 (95% confidence interval [CI] 1.11–1.456, p < 0.001) among SCM patients (Table 2). Linear regression models including the same covariates showed significant increases in inpatient mortality (odds ratio [OR] 1.58, 95% CI 1.18–2.09, p = 0.002), 1-year mortality (OR 1.25, 95% CI 1.01–1.55, p = 0.039), and 2-year mortality (OR 1.26, 95% CI 1.02–1.55, p = 0.031) among patients with SCM. There was no significant association between SCM and length of stay. Among survivors, those with SCM had persistently lower LVEF measurements on subsequent echocardiograms (Figure 2). This large retrospective analysis from a single centre demonstrated a significant association between new-onset SCM and mortality risk both during the index hospitalization and over subsequent years of follow-up. Mortality rates for patients without SCM were similar to those previously reported for patients requiring ICU admission for sepsis or septic shock (45% or 50%).6, 8 The finding that SCM is associated with increased inpatient mortality corroborates those of Hanumanthu and colleagues,4 who reported an OR of 4.46 for patients with SCM in a smaller cohort of 359 patients, using a broader definition of SCM. The definition for SCM remains an area of debate. Early studies identified what appeared to be a protective effect of a depressed LVEF in the setting of sepsis, and recent reviews and meta-analyses have not shown an association between LVEF and mortality among septic patients.2, 3, 5 However, these studies have largely been limited by small sample sizes and varying definitions of SCM. The higher inpatient mortality in the overall cohort than among patients included in the final analysis may reflect increased mortality among patients with pre-existing cardiac dysfunction. In the current analysis, we used a conservative ejection fraction cut-off and excluded patients with existing heart failure and those without a pre-morbid TTE result. Importantly, there is likely a spectrum of sepsis-induced cardiomyopathy phenotypes, and a reduced LVEF may identify only the most severe cases, given the systemic vasoplegia and reduced afterload associated with sepsis and septic shock. Additionally, patients with impairments in myocardial relaxation with diastolic dysfunction, either separate from or in combination with systolic impairment, likely constitute a discrete subgroup not evaluated in this study, and warrant further investigation. This is an important distinction and reference point for future intervention trials – newer echocardiographic techniques including global longitudinal strain may have greater sensitivity for clinically relevant SCM,1 and prospective studies focusing on the diagnosis and management of SCM are needed. Another area of discussion in SCM research is the inclusion of systolic function recovery. Some definitions of SCM require that there be an improvement in ejection fraction, typically a 10% increase, within 2 weeks. In clinical practice, the diagnosis of SCM is often not made until weeks after the initial echocardiogram. This delay in diagnosis is important because of the potential to over-estimate survival by excluding patients who did not get a repeat echocardiogram due to morbidity/mortality within a 2-week timeframe. For this reason, we chose not to include an improvement in ejection fraction in the inclusion criteria of this study. Strengths of our study include a large sample size as well as a specific definition that allows for closer characterization of the phenomenon of SCM. We also used a state-level database to confirm mortality. Furthermore, we add to the existing literature by characterizing the long-term morbidity and mortality associated with SCM, whereas previous studies have been limited to evaluating inpatient mortality. Weaknesses of this study include its retrospective, single-centre design, and lack of available data on patients' comorbidities and echocardiographic parameters other than LVEF. Furthermore, because mortality was assessed using a state-level database, deaths that occurred outside Colorado may have been missed. While these caveats limit the generalizability of our findings, the results do suggest that SCM is a marker of a poor prognosis in sepsis. Future prospective studies should focus on capturing more granular data about septic patients including comorbidities, aetiologic organisms, and therapeutic strategies employed, in addition to ejection fraction, to better adjust for potential confounders. In a large single-centre cohort, we found that septic cardiomyopathy was significantly associated with increased mortality. These data contribute to the growing body of evidence suggesting that SCM is associated with worse clinical outcomes and identifies a high-risk population. While the results should be interpreted with caution, they highlight the need for further large, prospective studies assessing diagnostic criteria, incidence, mechanism, clinical relevance, predictive biomarkers and echocardiographic criteria to inform management strategies for this highly morbid condition. Conflict of interest: none declared.