Blood Eosinophil Stability Predicts Clinical Outcomes in Hospitalized Patients with Acute Exacerbations of COPD

Introduction

Chronic obstructive pulmonary disease (COPD) is a highly heterogeneous lung condition characterized by persistent respiratory symptoms and airflow limitation due to airway and alveolar abnormalities.¹ COPD is the third leading cause of death worldwide, responsible for approximately 3 million deaths annually,² with nearly 90% of these occurring in low- and middle-income countries.³ Acute exacerbation of COPD (AECOPD) worsens respiratory symptoms, accelerates disease progression, and increases comorbidity and mortality.⁴ Studies have shown that 22–40% of COPD patients experience at least one acute exacerbation, with 9–16% experiencing multiple exacerbations.⁵ AECOPD also represents the largest component of the socioeconomic burden associated with COPD.

Blood eosinophils (EOS) are considered a potential clinical biomarker for identifying specific AECOPD phenotypes,⁶ guiding glucocorticoid therapy,⁷ and predicting exacerbation risk.⁸ However, existing research on the ability of EOS to predict the prognosis of AECOPD exacerbations remains controversial.^9–11 Previous studies typically used a single EOS measurement at admission to predict clinical outcomes, which may not fully capture the dynamic changes in EOS during exacerbations.¹² Multiple studies have shown that persistent high EOS levels during stable periods are associated with reduced mortality.^13,14 However, EOS levels can fluctuate during exacerbations due to disease progression and medical interventions. This suggests that short-term EOS stability may be a more accurate predictor of outcomes than a single measurement.¹⁵ Our previous work found that elevated EOS at admission was associated with shorter hospital stays in AECOPD patients.⁷ Still, the relationship between EOS changes during hospitalization and clinical outcomes remains unclear. Only one small sample size cohort study involving 530 patients has suggested a potential link between EOS changes and AECOPD outcomes, but more evidence is needed to confirm these findings.¹⁶

In this large-sample study, we examined changes in EOS levels during AECOPD and their impact on clinical outcomes, such as mortality, ICU admission, and AECOPD readmission within three years after discharge. These findings could improve risk stratification and AECOPD management in clinical practice.

Materials and Methods

Study Design and Patients

This retrospective study included patients admitted to our Hospital for AECOPD from March 2013 to March 2022. Inclusion criteria were as follows: (1) age ≥40 years; (2) primary discharge diagnosis of acute exacerbation of COPD [International Classification of Diseases (ICD)-10 code J44]; (3) a hospital stay of at least 24 hours; (4) two EOS measurements obtained during hospitalization—one within 24 hours of admission and the other at the time of discharge. Exclusion criteria were as follows: (1) comorbid asthma, allergic diseases, autoimmune diseases, hematologic diseases, or malignancies; (2) hospital stay ≥40 days; (3) for patients with multiple hospitalizations during the study period, only the first hospitalization was included. This study is a retrospective cohort study, and the protocol was approved by the Ethics Committee of Hospital.

Data Collection

Data were retrieved from a validated electronic medical record (EMR) platform established by the Beijing Institute of Respiratory Medicine and Beijing Chao-Yang Hospital.¹⁷ This respiratory-focused big data platform integrates structured records from over 4 million cases since 2013, with rigorous quality control and linkage to citywide hospital discharge databases.^18,19 Patients in this study were admitted to the Department of Respiratory and Critical Care Medicine, the Emergency Department, or the General Medicine Department. Information included demographic characteristics, smoking status, exacerbation in the previous year, comorbidity, laboratory tests, treatment and clinical outcomes. Clinical outcomes assessed during the hospitalization included in-hospital death, use of invasive mechanical ventilation (IMV), transfer to the intensive care unit (ICU), length of stay (LOS), and total hospitalization costs. AECOPD-related readmissions were tracked for up to three years following discharge. Adverse hospital outcomes were defined as transfer to ICU, use of IMV, in-hospital mortality, and LOS >14 days.

EOS stability during hospitalization was classified into four groups based on the initial and final EOS counts: (1) persistent high group: both initial and final EOS counts ≥ 2%; (2) persistent low group: both initial and final EOS counts < 2%; (3) decreased group: initial EOS count ≥ 2% but final EOS count < 2%; (4) increased group: initial EOS count <2% but final EOS count ≥ 2%. This classification method follows previous studies that evaluated eosinophil stability across multiple time points.^12,20

Statistical Analysis

Categorical variables were expressed as counts and percentages, compared using chi-square or Fisher’s exact tests. Continuous variables with non-normal distributions were reported as medians and interquartile ranges (IQR) and analyzed using the Kruskal–Wallis test. Patients with missing data were excluded. Univariate regression identified variables for multivariable logistic regression, which analyzed the association between EOS stability and adverse hospital outcomes, using the persistent low EOS group as a reference to calculate odds ratios (OR) and 95% confidence intervals (CI). Receiver operating characteristic (ROC) curves and area under the curve (AUC) values assessed predictive model performance, compared using the DeLong test. Nomograms were developed based on the models. Multivariate Cox regression predicted AECOPD readmission risk within three years, with subgroup analyses by age, gender, smoking, Charlson comorbidity index, and prior-year exacerbations. Statistical analyses were performed using R 4.3.1, with significance set at p < 0.05.

Results

Study Population

As shown in Figure 1, 7039 patients aged ≥40 years with a primary discharge diagnosis of AECOPD were identified. After exclusions, 2105 patients were included in the final analysis. The number of patients enrolled each year is shown in Table S1.

Figure 1 Study flow.

Clinical Characteristics of AECOPD Patients with Different EOS Stability Groups

The numbers of patients in the persistent high, decreased, increased, and persistent low EOS groups were 586 (27.8%), 154 (7.3%), 593 (28.2%), and 772 (36.7%), respectively. Overall, 35.1% had an initial EOS count ≥2%, while 64.9% had a initial EOS count <2%. Stable EOS levels (persistent high/low) were observed in 64.5% of patients, while 35.5% showed unstable levels. The decreased group’s median EOS dropped from 3.05 (2.30–4.10) % to 0.80 (0.10–1.40) %, while the increased group rose from 0.50 (0.00–1.20) % to 3.00 (2.40–4.10) %, was shown in Table 1.

Table 1 Baseline Characteristics of Patients with AECOPD Stratified by EOS Stability

The persistent high EOS group had a higher proportion of males (78.8%) and a lower median age (P<0.05). Cardiovascular disease prevalence was lowest in this group (70.5%, P=0.03), with no differences in smoking status or prior exacerbations. White blood cell, neutrophil counts, and CRP were significantly lower in the persistent high group compared to increased and persistent low groups, while lymphocyte counts were higher (P<0.001). Respiratory failure was less frequent in the persistent high group (P=0.03), and systemic corticosteroid use was lower. Systemic corticosteroid use varied significantly among groups, with the highest proportion observed in the decreased group (46.1%) and the lowest in the persistent high group (8.9%) (P < 0.001).

Clinical Outcomes and Prognostic of AECOPD Patients with Different EOS Stability Groups

The persistent high EOS group consistently had the lowest rates in these categories, including a 2.9% ICU transfer rate, 1.7% IMV use, 27.8% LOS > 14 days, and 0.2% in-hospital mortality. Total hospital costs were highest in the persistent low EOS group compared to the persistent high EOS group (P< 0.001). Adverse hospital outcomes were most frequent in the decreased EOS group (41.6%), followed by the persistent low EOS group (40.5%) (P<0.001). No significant differences were found in post-discharge AECOPD readmission within three years among the groups (P = 0.21), although the persistent low EOS group had a lower readmission rate (34.5%) compared to the other groups, as shown in Table 2.

Table 2 Comparison of Clinical Outcomes Among AECOPD Patients with Different EOS Stability

Predictive Factors for Adverse Hospital Outcomes in AECOPD Patients

Multivariate logistic regression analysis (Table 3) shows that the persistent high EOS group had lower odds of adverse hospital outcomes, compared to the persistent low EOS group (adjusted OR: 0.77, 95% CI: 0.59–0.99, P = 0.040). No significant differences were found for the decreased or increased EOS groups. Age (adjusted OR: 1.01, 95% CI: 1.00–1.02, P = 0.05) and hemoglobin (adjusted OR: 1.01, 95% CI: 1.00–1.02, P = 0.082) slightly increased the risk. Current smokers had lower odds than non-smokers (adjusted OR: 0.65, 95% CI: 0.48–0.88, P = 0.006). Risk factors included a history of exacerbations (adjusted OR: 1.78, 95% CI: 1.39–2.30, P = 0.006), Charlson index ≥2 (adjusted OR: 2.04, 95% CI: 1.61–2.30, P < 0.001), respiratory failure (adjusted OR: 1.76, 95% CI: 1.44–2.15, P < 0.001), and corticosteroid use (adjusted OR: 1.34, 95% CI: 1.10–1.70, P = 0.005). Higher lymphocyte counts were protective (adjusted OR: 0.84, 95% CI: 0.73–0.97, P = 0.022).

Table 3 The Predicting Factors for Adverse Hospital Outcomes Among Patients with AECOPD

Predictive Models for Adverse Hospital Outcomes: AUC and ROC Analysis

Table S2 illustrates the predictive performance of models using EOS stability for adverse hospital outcomes in AECOPD patients. Model 1, including EOS stability, age, and smoking status, had an AUC of 0.59 (95% CI: 0.57–0.62). Adding exacerbation history and Charlson index in Model 2 increased the AUC to 0.64 (95% CI: 0.62–0.67). Model 3, which included corticosteroid use and respiratory failure, improved the AUC to 0.66 (95% CI: 0.64–0.69). Model 4, incorporating WBC, lymphocyte count, and hemoglobin, achieved the highest AUC of 0.67 (95% CI: 0.65–0.70). The ROC curve for Model 4 showed moderate discrimination, with the DeLong test confirming stepwise improvement. A nomogram derived from ten predictors (Figure S1) performed well, with Hosmer‒Lemeshow test P = 0.938.

The Risk of AECOPD Readmission in Different EOS Stability Groups

Table S3 and Figure 2 show that after adjusting for baseline characteristics (age, smoking status, exacerbations in the previous year, Charlson comorbidity index), systemic corticosteroid, respiratory failure, and laboratory tests, both the persistent high (adjusted HR: 1.38, 95% CI: 1.13–1.68) and increased EOS groups (adjusted HR: 1.25, 95% CI: 1.05–1.50) were associated with a higher risk of AECOPD rehospitalization, while no significant difference was observed in the decreased EOS group (adjusted HR: 1.08, 95% CI: 0.81–1.45).

Figure 2 Cox proportional hazards analysis curve for AECOPD readmission by eosinophil stability groups.

Subgroup Analysis

Table S4 shows the subgroup analysis of persistent high EOS and AECOPD readmission risk within three years. Patients aged ≥65 (HR: 1.46, 95% CI: 1.17–1.81, P = 0.001), males (HR: 1.47, 95% CI: 1.17–1.85, P <0.001), former smokers (HR: 1.54, 95% CI: 1.17–2.02, P = 0.001), those with a Charlson index <2 (HR: 1.31, 95% CI: 1.03–1.65, P = 0.03), or a history of exacerbations (HR: 1.29, 95% CI: 1.04–1.63, P = 0.02) had higher readmission risks. Persistent high EOS was associated with increased readmission risk, regardless of corticosteroid use. No significant interaction effects were observed.

Discussion

This large-sample retrospective cohort study based on EMR details the changes in EOS levels and their association with clinical outcomes in hospitalized patients with AECOPD. The results showed that more than 30% of patients did not maintain stable EOS levels during hospitalization, exhibiting either increasing or decreasing trends. Patients with persistent high EOS exhibited better hospital outcomes, including lower ICU transfer rates, reduced mortality, fewer instances of IMV, and lower total costs. However, our observations revealed that the risk of readmission due to AECOPD within three years after discharge was significantly higher in the persistent high and increased EOS groups, while there were no significant differences in the decreased EOS group compared to the persistent low EOS group.

Given the significant heterogeneity in COPD, with various phenotypes and pathophysiological mechanisms,²¹ precise clinical phenotyping can help clinicians assess risk and implement personalized treatment plans. Bafadhel et al classified acute exacerbation events into four phenotypes based on etiology and inflammatory markers: bacterial, viral, eosinophilic, and paucigranulocytic.²² Increased EOS levels identify the eosinophilic phenotype and are a key, treatable trait in COPD.²³ The Global Initiative for Chronic Obstructive Lung Disease (GOLD) recommends using EOS levels as a biomarker to guide corticosteroid therapy in acute exacerbations.²⁴ In our study, we selected an EOS threshold of ≥2%, and the results showed that 35.1% of AECOPD patients had an initial EOS exceeding 2% upon admission, which is consistent with prior research.^25,26

However, the high EOS phenotype may not be stable and can fluctuate due to various factors. Accurate assessment of EOS stability is crucial for clinical evaluation and personalized corticosteroid treatment. In stable COPD patients, EOS levels can be influenced by factors like allergies, obesity, medications, and smoking, leading to fluctuations.²⁷ The PROMISE-COPD and ECLIPSE studies indicated significant variability in EOS throughout the stable COPD course.^12,28 Previous studies have explored the long-term stability of EOS in COPD patients and its association with outcomes.^14,29 Our prior research also suggested that patients with consistently high EOS levels at admission had reduced all-cause mortality after discharge.³⁰ However, during a single acute exacerbation, EOS levels can fluctuate with disease progression and treatment effects. Current research on EOS stability in AECOPD is limited. A similar study¹⁶ selected a threshold of 300 cells/μL for eosinophil level, which is relatively high in Chinese populations.³¹ As a result, the eosinophil distribution in their cohort was more concentrated in the lower range, with 71.9% of patients in the persistent low EOS group and only 10.9% in the persistent high EOS group. In contrast, our study adopted a 2% blood eosinophil percentage as the cutoff to differentiate eosinophilic and non-eosinophilic groups, based on previous research demonstrating its sensitivity in detecting eosinophilic airway inflammation and its relevance to eosinophil-driven exacerbations.^22,32 This relative threshold allowed us to capture fluctuations in eosinophil levels and resulted in a more balanced distribution across the EOS trajectory groups.⁷

Previous studies have shown that patients with high EOS levels at admission tend to have better clinical outcomes,^32,33 a trend that was even more pronounced in the persistent high EOS group in our study. We also found that these patients had lower rates of adverse hospital outcomes and reduced hospitalization costs. Consistent with previous studies, which demonstrated that AECOPD patients with lung infections tend to have elevated peripheral neutrophil counts and reduced EOS and lymphocyte levels,^34,35 our study observed that patients with EOS counts below 2% exhibited higher WBC, lower lymphocyte counts, and elevated CRP levels, indicating increased susceptibility to infection. Additionally, these patients had a higher incidence of respiratory failure and more frequent use of systemic corticosteroids, reflecting greater disease severity, as noted in other studies.^33,36 Interestingly, patients in the decreasing group—those with elevated EOS at admission but decreased at discharge—appeared to diverge from the typical eosinophilic phenotype and demonstrated worse in-hospital outcomes. This suggests that a subgroup of patients with initially high EOS but poor response to treatment may exist, and this warrants further investigation. Further multivariate logistic regression analysis of adverse hospital outcomes revealed that, compared to the persistent low group, only the persistent high EOS group emerged as a protective factor. This finding could explain why many earlier studies, which focused solely on a single EOS level at admission without accounting for EOS variability,⁹ failed to show consistent associations between eosinophil levels and clinical outcomes. A comprehensive assessment of the risk of adverse outcomes during hospitalization in AECOPD patients based on eosinophil stability is crucial. By integrating previously identified risk factors (baseline characteristics, severity assessments, and hospitalization details),³⁷ we constructed a nomogram to predict adverse hospital outcomes. This model exhibits high predictive power and precision, providing a robust tool for evaluating patient risk and informing personalized treatment strategies.

The role of eosinophils in predicting readmission after discharge in AECOPD patients remains debated. Some studies link high eosinophil levels with increased rehospitalization risk³⁸, while others find no strong association.¹⁰ Many previous studies have not accounted for eosinophil levels at discharge, which may overlook critical insights. In our cohort, patients with elevated eosinophils at discharge-both in the persistent and increasing eosinophil groups-demonstrated a heightened risk of rehospitalization. Interestingly, further subgroup analysis of the persistent high eosinophil group showed that these patients had an elevated risk of exacerbation, independent of corticosteroid use during hospitalization. This finding indicates that eosinophil-driven inflammation may not have been fully controlled, leaving patients in a pro-inflammatory state post-discharge.³⁹

In our cohort, systemic corticosteroid use varied across eosinophil stability groups, with a higher proportion in the persistent low group. Although current guidelines suggest that eosinophilic AECOPD patients may benefit more from corticosteroids,^7,40 in clinical practice in China, treatment decisions are often based on physician judgment and symptom severity rather than blood eosinophil counts.^33,36 Regarding the observed eosinophil increase at discharge in some patients, this may reflect either the intrinsic eosinophilic inflammation phenotype of COPD or a rebound effect after corticosteroid withdrawal. The CORTICO-COP trial, for example, demonstrated a rebound increase in blood eosinophils between days 3 and 5 following the initiation of corticosteroid treatment.¹⁵ Additionally, Citgez et al reported that corticosteroid use did not significantly impact eosinophil categorization stability during subsequent AECOPD episodes. To more precisely delineate the impact of treatment on eosinophil dynamics, future studies incorporating prospective designs, standardized treatment protocols, and serial eosinophil monitoring are warranted.

Beyond eosinophils, the association between systemic inflammation and disease severity in AECOPD patients has also been explored through other hematologic biomarkers. One of the most studied markers is the neutrophil-to-lymphocyte ratio (NLR), which has been proposed as a reliable indicator of systemic inflammatory burden. Shao et al⁴¹ reported that NLR measured at admission was significantly associated with adverse clinical outcomes and the risk of readmission in patients with AECOPD. However, their analysis was based solely on static, single-time-point measurements, which may overlook important dynamic inflammatory changes during hospitalization. Jiang et al⁴² subsequently highlighted not only the prognostic relevance of NLR stability, suggesting that the trajectory of inflammatory markers may better reflect disease progression than absolute values. Nevertheless, their outcome assessment was limited to in-hospital events, without consideration of longer-term outcomes such as post-discharge readmission. In contrast, our study focused on the stability of EOS levels during hospitalization and developed a four-group trajectory model based on changes between admission and discharge (persistent high, persistent low, increased, decreased). Our work incorporates a larger sample, longer follow-up, broader outcome assessment, and a dynamic modeling approach—together underscoring the clinical relevance of inflammatory phenotype stability in AECOPD risk stratification.

This study has limitations. First, being a single-center study, its representativeness is limited. Second, as a retrospective study, pre-hospital treatment records for acute exacerbations are often missing, precluding the assessment of pre-hospital corticosteroid treatment. Finally, although the changes in EOS percentage among the increase and decrease groups were ≥0.3%, there is no established consensus on the minimal clinically meaningful change or the threshold for defining eosinophil stability. Further research is warranted to evaluate the prognostic value of more detailed eosinophil trajectories using repeated measurements and standardized definitions.

Conclusions

Our findings reveal that the stability of blood eosinophil levels during hospitalization holds prognostic significance in AECOPD. Early-phase eosinophil changes may guide acute treatment decisions, while discharge eosinophil levels could help predict long-term outcomes such as readmission risk. These findings support incorporating eosinophil dynamics into individualized AECOPD management strategies.