Association between blood eosinophil levels and prognosis in critically ill patients with different heart failure phenotypes

Abstract

Heart failure (HF) is associated with poor prognosis. While various immune markers have been linked to HF outcomes, the relationship between eosinophils (EOS) and prognosis across different HF phenotypes remains unclear and controversial. In this study, we analyzed 2,677 HF patients from the Medical Information Mart for Intensive Care III (MIMIC-III) database, categorizing them into reduced ejection fraction (≤ 50%) and preserved ejection fraction (> 50%) groups. Multiple logistic regression, Kaplan-Meier survival curve estimation, Cox proportional hazards model, and restricted cubic splines were employed to assess the association between admission EOS and mortality. Subgroup analyses were conducted based on age, gender, sepsis, atrial fibrillation, hypertension and rheumatic and immunological diseases and mechanical ventilation. A total of 954 patients with heart failure with reduced ejection fraction (HFrEF) and 1723 patients with heart failure with preserved ejection fraction (HFpEF) were enrolled, with a female proportion of 37.84% and 56.88%, respectively. The median age of patients was 74.46 (IQR: 63.34–83.26) years and the median EOS was 0.09 (IQR: 0.04–0.20). Among these patients, in-hospital mortality was 14.5% (389 patients), and the overall 90-day mortality rate reached 27.12% (726 patients). There were no significant differences in EOS levels between the HFrEF and HFpEF groups (0.1 vs. 0.09, P = 0.154). After adjusting for confounders, multivariate logistic regression showed that higher EOS quartiles were associated with reduced in-hospital mortality in HFrEF patients, with quartile 3 (Q3) (OR 0.54, 95% CI 0.31–0.96, P = 0.035) and Q4 (OR 0.37, 95% CI 0.19–0.71, P = 0.003) showing significant protective effects compared with the lowest EOS quartile (Q1), while Q2 did not reach statistical significance (P = 0.931). Subgroup analyses yielded similar findings. Multivariate Cox regression demonstrated a reduced hazard ratio in Q4 (HR 0.64, 95% CI 0.44–0.94, P = 0.023) compared with Q1 in HFrEF patients. However, no significant association was observed between EOS and prognosis in HFpEF. Our findings indicated that higher EOS was associated with improved in-hospital and 90-day survival in ICU patients with HFrEF, but not in those with HFpEF.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-90778-7.

Introduction

Heart failure (HF) is a complex clinical syndrome characterized by structural and/or functional cardiac abnormalities, resulting in elevated natriuretic peptide levels and/or pulmonary or systemic congestion^1,2. Despite a declining overall incidence, the global burden of HF continues to rise, affecting over 60 million patients worldwide^3–5. The persistent high mortality and hospitalization rates, particularly in resource-limited settings, underscore the urgent need for improved prevention and management strategies^6,7.

Clinicians categorize HF based on the left ventricular ejection fraction (LVEF), identifying three distinct phenotypes: HF with preserved ejection fraction (HFpEF, LVEF ≥ 50%), HF with mildly reduced ejection fraction (HFmrEF, LVEF 41–49%), and HF with reduced ejection fraction (HFrEF, LVEF ≤ 40%)⁸. These classifications reflect underlying pathophysiological differences and varying therapeutic responses⁹.

Mounting evidence suggest a pivotal role of immune system activation in HF pathogenesis^10–12. While the contributions of various immune cell subsets, including neutrophils, macrophages, monocytes, and regulatory T cells, have been extensively studied^13–16, the relationship between eosinophil (EOS) and HF remains poorly understood and controversial. EOS, though representing a smaller subset of blood cells, have traditionally been recognized for their role in parasite-killing and allergic reactions¹⁷. Although historically viewed as a risk factor in cardiovascular disease¹⁸, recent studies have revealed their potential beneficial effects in cardiac insufficiency¹⁹. The prognostic value of EOS in HF remains controversial, with recent studies yielding conflicting results. A prospective cohort study of 416 patients undergoing percutaneous coronary intervention found that elevated EOS counts correlated with increased HF incidence during 2-year follow-up²⁰. Conversely, analysis of a large CALIBRE cohort of 775,231 people demonstrated that low EOS counts predicted HF onset, although this association attenuated after six months²¹. Furthermore, several cohort studies focusing on acute decompensated HF admissions have reported an association between low EOS counts and increased mortality^22–24. These inconsistent findings underscore the critical need for further investigation into the prognostic value of EOS in HF. Notably, most previous studies have not differentiated between HF phenotypes based on ejection fraction, with one study focusing exclusively on HFrEF patients²³. This significant research gap hampers our understanding of how EOS influences different HF subtypes.

In this study, we conducted a retrospective analysis investigating the association between admission EOS and mortality outcomes among critically ill patients across different HF phenotypes.

Materials and methods

Data source

Data were extracted from the MIMIC-III database (version: v1.4), which contained comprehensive clinical data on patients admitted to the intensive care unit of Beth Israel Deaconess Medical Center from 2001 to 2012²⁵. The corresponding author (Bing Xiao) completed the Collaborative Institutional Training Initiative (CITI) program (Record ID: 57440109) and obtained access to the MIMIC-III database. As all data were de-identified, additional informed consent was not required.

Study population

We included adult patients (≥ 18 years) admitted to the ICU with a diagnosis of HF, identified using International Classification of Diseases, 9th Revision (ICD-9) codes (Figure S1, Table S1). Exclusion criteria were: (1) missing cardiac ultrasound data in the discharge summary; (2) missing EOS% data within ICU admission; (3) extremely abnormal values. Due to de-identification procedures, ages over 89 years were recorded as greater than 300 years; we adjusted these to 91 years. Due to the small number of patients with HFmrEF and their more typical characteristics of HFrEF⁹, we classified patients into reduced ejection fraction (LVEF < 50%) and preserved ejection fraction groups (LVEF ≥ 50%) based on echocardiographic findings. For patients with multiple ICU admissions, only the data from their first entry were analyzed. EOS was calculated by multiplying WBC counts by EOS%.

Data extraction and outcomes

Data were extracted using Structured Query Language (SQL) in PostgreSQL, with code adapted from the MIMIC Code Repository (https://github.com/MIT-LCP/mimic-code/). Extracted variables included: (1) Demographic information: age, gender, ethnicity, and weight; (2) Laboratory results of ICU admission: creatinine, blood urea nitrogen (BUN), platelet, EOS%, neutrophils, lymphocytes, white blood cell (WBC), sodium, potassium; (3) Vital signs on ICU admission: heart rate (HR), respiratory rate (RR), percutaneous arterial oxygen saturation (SpO₂); (4) Comorbidities: myocardial infarction, chronic lung disease, chronic kidney disease, atrial fibrillation, hypertension, obesity, rheumatic disease, asthma, diabetes and sepsis; (5) Interventions: angiotensin converting enzyme inhibitor (ACEI), Angiotensin II receptor blockers (ARB), β-blockers, steroids, continuous renal replacement therapy (CRRT) and mechanical ventilation. Variables with > 20% missing data were excluded; those with < 20% missing data were imputed using the mice package. The primary outcomes were in-hospital and 90-day mortality. Out-of-hospital mortality data were obtained through Social Security records linked to the MIMIC database.

Statistical analysis

Continuous variables were tested for normality using the Kolmogorov-Smirnov test and described as mean (standard deviation) or median (interquartile range). Differences between groups were compared using the Mann-Whitney U test or Student’s t-test, as appropriate. Categorical variables were expressed as frequencies or percentages (%) and compared using the Pearson’s chi-squared test or Fisher’s exact test.

Univariate and multivariate logistic regression analysis were used to assess the relationship between EOS and in-hospital mortality. EOS was entered as a categorical variable. Model 1 was adjusted for EOS alone, while Model 2 included variables with P values less than 0.2 in univariate analysis. The first quartile EOS served as the reference group across all models. Multicollinearity was assessed using variance inflation factor (VIF), with variables exceeding a VIF of 10 being excluded (Table S2). The nonlinear relationship between continuous EOS and mortality outcomes was explored using restricted cubic splines (RCS). The association between EOS and 90-day mortality was examined using Kaplan-Meier survival analysis, with group differences assessed by log-rank test. Hazard ratios (HRs) for EOS and mortality were calculated using Cox regression analysis. Subgroup analyses were performed stratified by gender, age (≤ 65 vs. > 65 years), atrial fibrillation, hypertension, sepsis, rheumatic and immunological diseases and mechanical ventilation. All statistical analyses were conducted using R software (version: 4.2.3), with statistical significance set at P < 0.05 (two-sided).

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki. The MIMIC-III database has been approved by Massachusetts Institute of Technology (Cambridge, MA) and Beth Israel Deaconess Medical Center (Boston, MA). As all protected health information in the MIMIC database was de-identified, so individual patient consent was not required, and we informed the Ethics Committee of the Second Hospital of Hebei Medical University of this study without a written report.

Results

Baseline characteristics

This study population included 2677 HF patients with a median age of 74.46 years (IQR: 63.34–83.26) and median EOS of 0.09 (IQR: 0.04–0.20). Among these patients, in-hospital mortality rate was 14.5% (389 patients), with overall 90-day mortality reaching 27.12% (726 patients). Table S3 compares baseline characteristics between survivors and non-survivors. Compared with survivors, non-survivors were characterized by older age, extended ICU stays (3.08 vs. 5.59, P < 0.001), and higher prevalence of sepsis, atrial fibrillation, but lower prevalence of chronic kidney disease. Laboratory tests and vital signs revealed that non-survivors exhibited significantly elevated WBC, neutrophils, creatinine, BUN, potassium, and RR, alongside decreased platelet, lymphocytes and SpO₂. Non-survivors showed higher rates of mechanical ventilation and steroid use but lower utilization of ACEI, ARB, an β-blockers, while CRRT use remained comparable between groups. EOS counts for non-survivors were slightly lower than those for survivors (0.08 vs. 0.09, P < 0.049).

Table S4 presented the clinical characteristics of HF patients stratified by ejection fraction. HFrEF patients were predominantly males (62%) and showed higher prevalence of myocardial infarction, ischemic heart disease, and chronic kidney disease, but lower rates of chronic lung disease, rheumatic and immune diseases, sepsis and obesity compared to HFpEF patients. Laboratory values revealed higher creatinine and lymphocytes in HFrEF patients. In addition, HFrEF patients had higher use of ACEI, β-blockers, and mechanical ventilation, but lower rates of steroid use. No significant differences were observed between groups in EOS levels (0.1 vs. 0.09, P = 0.154), in-hospital mortality (14.36% vs. 14.63%, P = 0.852) or 90-day mortality (28.72% vs. 26.23%, P = 0.166).

Quartile analysis by EOS demonstrated that HFrEF patients in Q1 had the highest neutrophil (86.50) and the lowest lymphocyte (7.00), with peak in-hospital and 90-day mortality rates (20.08% and 35.98%, respectively). Among HFpEF patients, Q4 was characterized by younger age (72.09, P = 0.011) and higher weight (80.15, P = 0.025), with highest rates of renal disease (35.42%, P = 0.005) and diabetes mellitus (40.28%, P = 0.035), but the lowest rates of hypertension (34.49%, P = 0.007). The Q4 group also had reduced steroid use and mechanical ventilation rates. No significant differences were observed in in-hospital or 90-day mortality across quartiles, while ICU stay duration paralleled that of HFrEF patients.

Association between EOS and in-hospital mortality

Multivariate logistic regression analysis (Table 1) showed that higher EOS quartiles were associated with reduced in-hospital mortality in HFrEF patients compared to the lowest EOS quartile (Q1). Specifically, Q3 (OR 0.54, 95%CI 0.31–0.96, P = 0.035) and Q4 (OR 0.37, 95%CI 0.19–0.71, P = 0.003) showed significant protective effects, while Q2 did not reach statistical significance (P = 0.931). No association was observed between EOS and in-hospital mortality in HFpEF patients (Table S8).

Table 1.

Relationship between different EOS% groups and in-hospital mortality in patients with HFrEF.

Variables	univariate		multivariable
	OR (95%CI)	P	OR (95%CI)	P
Eosinophils
Q1	1.00 (Reference)		1.00 (Reference)
Q2	0.88 (0.56 ~ 1.39)	0.591	0.98 (0.59 ~ 1.62)	0.931
Q3	0.49 (0.29 ~ 0.81)	0.006	0.54 (0.31 ~ 0.96)	0.035
Q4	0.36 (0.21 ~ 0.63)	< 0.001	0.37 (0.19 ~ 0.71)	0.003
Gender
Female	1.00 (Reference)
Male	0.80 (0.56 ~ 1.16)	0.242
Ethnicity
White	1.00 (Reference)
Other	1.24 (0.83 ~ 1.86)	0.289
Myocardial infarction
No	1.00 (Reference)
Yes	1.12 (0.78 ~ 1.61)	0.534
Pulmonary disease
No	1.00 (Reference)
Yes	0.80 (0.54 ~ 1.19)	0.268
Renal disease
No	1.00 (Reference)
Yes	0.84 (0.57 ~ 1.24)	0.381
Atrial fibrillation
No	1.00 (Reference)
Yes	1.24 (0.86 ~ 1.78)	0.247
Obesity
No	1.00 (Reference)
Yes	0.86 (0.40 ~ 1.85)	0.697
Hypertension
No	1.00 (Reference)
Yes	0.95 (0.65 ~ 1.37)	0.767
Ischemic heart disease
No	1.00 (Reference)
Yes	0.69 (0.37 ~ 1.29)	0.244
Rheumatic and immunological diseases
No	1.00 (Reference)
Yes	1.21 (0.66 ~ 2.22)	0.529
Sepsis
No	1.00 (Reference)		1.00 (Reference)
Yes	3.28 (2.18 ~ 4.95)	< 0.001	2.27 (1.41 ~ 3.67)	< 0.001
Diabetes
No	1.00 (Reference)
Yes	0.98 (0.68 ~ 1.43)	0.936
ACEI
No	1.00 (Reference)		1.00 (Reference)
Yes	0.63 (0.34 ~ 1.18)	0.148	1.19 (0.60 ~ 2.36)	0.618
ARB
No	1.00 (Reference)
Yes	0.00 (0.00 ~ Inf)	0.976
β-blockers
No	1.00 (Reference)		1.00 (Reference)
Yes	0.41 (0.27 ~ 0.62)	< 0.001	0.57 (0.37 ~ 0.90)	0.016
Steroids
No	1.00 (Reference)		1.00 (Reference)
Yes	2.01 (1.22 ~ 3.31)	0.006	1.41 (0.79 ~ 2.50)	0.241
Mechanical ventilation
No	1.00 (Reference)		1.00 (Reference)
Yes	2.24 (1.54 ~ 3.26)	< 0.001	2.33 (1.53 ~ 3.56)	< 0.001
CRRT
No	1.00 (Reference)
Yes	0.99 (0.46 ~ 2.15)	0.987
Age	1.02 (1.01 ~ 1.04)	0.003	1.03 (1.01 ~ 1.04)	0.002
Weight	1.00 (0.99 ~ 1.01)	0.655
WBC	1.04 (1.01 ~ 1.07)	0.011	1.03 (0.99 ~ 1.07)	0.109
Neutrophils	1.01 (1.00 ~ 1.03)	0.104	0.98 (0.96 ~ 1.01)	0.146
Lymphocytes	0.93 (0.90 ~ 0.96)	< 0.001	0.97 (0.93 ~ 1.00)	0.073
Creatinine	1.04 (0.93 ~ 1.16)	0.537
Platelet	1.00 (1.00 ~ 1.00)	0.165	1.00 (1.00 ~ 1.00)	0.512
BUN	1.01 (1.01 ~ 1.02)	< 0.001	1.01 (1.01 ~ 1.02)	0.018
Potassium	1.24 (0.90 ~ 1.71)	0.18	1.11 (0.78 ~ 1.60)	0.559
Sodium	0.97 (0.94 ~ 1.00)	0.079	0.97 (0.93 ~ 1.00)	0.078
HR	1.00 (0.98 ~ 1.01)	0.416
RR	1.04 (0.99 ~ 1.09)	0.141	1.03 (0.98 ~ 1.09)	0.259
SpO2	0.97 (0.95 ~ 0.98)	< 0.001	0.98 (0.96 ~ 0.99)	0.019

OR: Odds Ratio, CI: Confidence Interval, HFpEF: heart failure with preserved ejection fraction, BUN: blood urea nitrogen, WBC: white blood cell, HR: heart rate, RR: respiratory rate, SpO2: percutaneous oxygen saturation, ACEI: angiotensin converting enzyme inhibitor, ARB: Angiotensin II receptor blockers, CRRT: continuous renal replacement therapy

Association between EOS and 90-day mortality

Kaplan-Meier analysis (Fig. 1) demonstrated significant differences in 90-day mortality among EOS quartiles in HFrEF patients (log-rank test P = 0.002) with higher EOS% associated with improved 90-day survival.

Fig. 1.

Kaplan-Meier analysis results show cumulative mortality within 90 days for HFrEF patients by EOS group. EOS: Eosinophils.

Multivariate Cox regression (Table S7) showed a reduced HR in Q4 (HR 0.64, 95% CI 0.44–0.94, P = 0.023) compared with Q1. Although HR in Q2 and Q3 showed a downward trend, they did not reach statistical significance.

Subgroup analysis

Subgroup analyses (Fig. 2) were performed to examine the relationship between EOS and in-hospital mortality across different patient characteristics, including age, gender, hypertension, sepsis, mechanical ventilation, atrial fibrillation, and rheumatic and immunological diseases. The results were consistent with the overall findings, with no significant differences in EOS within any subgroup. Similar results were observed in subgroup analyses of 90-day mortality (Figure S2).

Fig. 2.

Subgroup analysis of the relationship between EOS and in-hospital mortality in patients with HFrEF. OR odds ratio; EOS: Eosinophils.

Restricted cubic spline

RCS analysis (Fig. 3) revealed a linear relationship between EOS levels and both in-hospital and 90-day mortality (P for non-linearity = 0.094 and 0.340, respectively). As EOS increased, the OR and HR for both outcomes decreased, eventually plateauing at higher EOS levels.

Fig. 3.

RCS curves of EOS versus in-hospital (A) and 90-day (B) mortality after adjustment for variables in patients with HFrEF. EOS: Eosinophils, RCS: restricted cubic spline.

Discussion

We conducted a study to investigate the relationship between admission EOS and short-term outcomes among critically ill patients with different HF phenotypes. Our findings demonstrated that EOS was a significant predictor of both in-hospital and 90-day mortality in ICU patients with HFrEF. Higher EOS was independently associated with improved survival outcomes in HFrEF patients, and this association persisted after adjustment for potential confounding factors. However, no significant association was observed between EOS and mortality outcomes in HFpEF patients. Despite significant advancements in understanding and treating HF, the patients prognosis remained unfavorable, with particularly high mortality rates among those requiring intensive care hospitalization²⁶. The early post-discharge period, typically within 2–3 months and termed the “vulnerable phase”, carried an elevated risk of adverse events, including readmission and death²⁷. Long-term follow-up studies have shown that while overall mortality rates were comparable between HF phenotypes, patients with HFrEF experienced significantly higher rates of cardiovascular disease and HF readmission compared to those with HFpEF^28–30. Given these challenges, identifying reliable predictors was critical for optimizing patient management and improving outcomes.

In recent years, inflammatory markers have emerged as easily measurable laboratory indicators with significant prognostic value in HF patients. A meta-analysis of 10 studies found that the neutrophil-to-lymphocyte ratio (NLR) predicted all-cause mortality in HF patients³¹. Additionally, a long-term follow-up study of 538 HF patients showed that C-reactive protein, red blood cell distribution width, and NLR were associated with all-cause mortality and improved the model’s predictive power³². While few studies have investigated the effect of EOS on prognosis across different HF phenotypes, and our study showed a protective effect of higher EOS on short-term outcomes in HFrEF patients, corroborating previous findings²³. Notably, we found no significant association between EOS and short-term prognosis in HFpEF patients. This observation aligns with a recent animal experiment showing unchanged EOS levels in HFpEF model mice, where EOS deficiency did not significantly affect cardiac function and structure³³, supporting our conclusion regarding the limited role of EOS in HFpEF development.

The precise mechanism underlying the association between EOS and HFrEF, and its absence in HFpEF populations, remains unclear. HFrEF and HFpEF represented two distinct phenotypes marked by different ventricular remodeling^9,34. HFpEF typically followed a progressive course, characterized by inflammation in the coronary microvasculature and drived ventricular dysfunction and remodeling³⁵. This manifested as concentric remodeling and diastolic dysfunction³⁶, frequently mediated by underlying comorbidities. In contrast, HFrEF might present as acute or chronic condition, predominantly resulting from massive myocardial injury⁹, with ischemic heart disease as the predominant cause, leading to eccentric remodeling and systolic dysfunction³⁷. Several studies have shown that EOS improved cardiac function during the acute phase following myocardial infarction. Post-infarction, EOS rapidly aggregated around the infarct site, releasing interleukin-4 (IL-4), the eosinophil cationic protein (ECP) homologue mEar1 and other cytokines. These factors reduced post-MI inflammatory cell aggregation and infiltration, thereby limiting cardiomyocyte apoptosis and fibroblast activation^38–40. Coincidentally, a recent study has confirmed that EOS played a protective role in abdominal aortic aneurysms by releasing IL-4 and mEar1, which promoted macrophage and Ly6^Clo monocyte polarization and suppressed NF-κB activation⁴¹. Although our study reveals a potential protective effect of higher EOS levels in HFrEF, it is important to note that, in hypereosinophilia syndrome, excessive proliferation and tissue infiltration of EOS could lead to endothelial cell damage⁴². Given the complex biological effects of EOS, further studies are needed to elucidate the mechanisms underlying their protective and pathogenic roles in cardiovascular diseases.

Our study confirmed that higher EOS was associated with decreased in-hospital and 90-day mortality in HFrEF patients, highlighting its potential clinical significance. Given that eosinophil measurements were routinely available in clinical practice, EOS could serve as a valuable risk stratification tool in managing HFrEF patients.

This study had several limitations. First, although echocardiographic data were available, the absence of brain natriuretic peptide data limited the comprehensive assessment of cardiac function. Second, the small number of HF patients with HFmrEF suggested that further research was needed to explore the role of EOS in such patients subgroup.

Conclusion

In summary, higher EOS was significantly associated with decreased in-hospital and 90-day mortality rates in ICU patients with HFrEF, but not HFpEF.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

XB contributed to the research design. GCJ, ZY and GAW contributed to data collection, data processing and graphing. WKK, YHT, ZSD, JS contributed data proofreading and formal analysis. GCJ contributed to the writing of the manuscript. XB and WLX contributed to review and to edit. All authors have read and approved the final manuscript.

Funding

The project was supported by the S.T Program of Hebei No.22377728D.

Data availability

The data utilized in this study were obtained from the MIMIC-III database (version 1.4), accessible at https://physionet.org/content/mimiciii/1.4/. The code for data extraction was publicly available on Github, accessible at https://github.com/MIT-LCP/mimic-code. Due to the restricted nature of the MIMIC database, researchers interested in accessing the data were required to complete an application process. However, upon reasonable request and confirmation of MIMIC database access, the corresponding author (Bing Xiao) can provide the specific dataset used in this study.

Declarations

Competing interests

The authors declare no competing interests.