Introduction

Exacerbation of chronic obstructive pulmonary disease (ECOPD) drives 20% of COPD-related hospitalizations.1–4 ECOPD demonstrates marked heterogeneity in clinical presentations and immune-inflammatory signatures, contributing to elevated hospitalization, readmission rates and mortality risks. Hospitalized exacerbations demand vigilant monitoring for clinical deterioration, particularly for invasive mechanical ventilation (IMV) requirement, intensive care unit (ICU) transfer, and mortality prevention.3,5–10

Our previous study on asthma exacerbations identified three distinct phenotypes associated with disease progression and adverse in-hospital outcomes, highlighting the significant heterogeneity present even during exacerbations.11 Similar to asthma, COPD is a heterogeneous airway disease. In stable COPD, cluster analyses have delineated distinct phenotypes, characterized by differing clinical features and progression trajectories.5 This underscores the necessity for the implementation of phenotype-specific management strategies for COPD. For example, Burgel et al identified four COPD phenotypes, marked by variations in airflow limitation, symptoms, and comorbidities, aspects not encompassed by traditional GOLD classifications, thereby underscoring the imperative for a multidimensional assessment.5 Castaldi et al characterized four COPD phenotypes among smokers based on baseline lung function and genetic variants; however, the absence of biomarker measurements, comorbidities, and follow-up outcomes constrained a thorough evaluation of their long-term clinical significance.12 Furthermore, Gregory et al conducted a prospective cohort study integrating multi-omics biomarkers and comorbidities to assess subtypes over time, thus deepening our understanding of their progression and clinical significance.13 Collectively, these studies have substantially deepened our understanding of stable COPD, offering crucial insights for clinical practice.5,12,13 However, to our knowledge, a notable gap exists: few studies have categorized ECOPD patients into distinct phenotypes. Moreover, there is a lack of validation regarding these phenotypes and their associated immunoinflammatory mechanisms in the context of adverse outcomes during the index hospitalization.

This study proposes that patients with ECOPD can be categorized into distinct clinical and immunoinflammatory phenotypes. The objectives are: (1) to identify these phenotypes; (2) to investigate their influence on adverse outcomes during the index hospitalization; (3) to identify the associated immunoinflammatory factors and elucidate how these factors mediate the relationship between the identified phenotypes and adverse in-hospital outcomes.

Methods Study Design and Patients

This study included two prospective observational cohort studies with a total of 29,844 inpatients. Cohort 1 (20,890 inpatients, training set) was conducted from December 2018 to October 2022, and Cohort 2 (8,954 inpatients, validation set) was conducted from November 2022 to January 2023 (Figure S1). Adult patients (age≥18 years) hospitalized for ECOPD, as defined by the Global Initiative for Obstructive Lung Diseases (GOLD), with a hospital stay longer than 24 hours, were prospectively recruited at West China Hospital, Sichuan University, China.14,15 An ECOPD was defined according to the GOLD definition.15 Patients with clinically and radiologically proven pneumonia and those who received any dose of systemic corticosteroids prior to admission were excluded. The patients were followed-up until discharge or death to assess prognosis. Cohort 1 (training set) was used for performing cluster analysis and cohort 2 (validation set) for validating the clusters. To assess whether the cluster analysis in training set had reproducibility, the same algorithm of cluster analysis in training set was used for the validation set. As this was a “real-world” study, patients management was guided by the consulting physicians based on GOLD. The study was approved by the Institutional Review Board (IRB) of West China Hospital, Sichuan University (Chengdu, China) (No. 2023–1882). All patients provided written informed consent. Additional details of the management of ECOPD are available in Methods section of the Supplementary Material.

Data Collection and Clinical Assessments at Baseline

Patients underwent a standardized assessment according to the GOLD within 24 hours of admission. This assessment collected comprehensive information, including demographic data, smoking status, body mass index (BMI), comorbidity, the COPD-specific co-morbidity test (COTE-index), oxygen therapy requirements, treatment history, GOLD staging, and spirometric data from clinical records prior to admission. Multimorbidity was defined as the co-existence of two or more long-term health conditions.16–18 Additional details of multidimensional assessment and data collection are available in Methods section of the Supplementary Material.

Laboratory Testing

Blood samples were collected within 24 hours of admission before in-hospital treatments for laboratory tests, including complete blood count, hematology, arterial blood gas analysis, inflammatory markers (C-reactive protein [CRP] and procalcitonin [PCT]), biochemistry, coagulation and electrolyte measurements. All measurements were completed within 2 hours of blood sampling.

In-Hospital Management of ECOPD

As a real-world study, ECOPD management adhered to GOLD under attending physicians’ discretion. All patients received standardized evaluations including comprehensive medical history, physical exams, and systematic assessments (Table S1). In-hospital therapies were delayed until after baseline blood sampling to avoid confounding effects on eosinophil quantification.

Primary and Secondary in-Hospital Outcomes

To examine the associations between the identified clusters and in-hospital outcomes, all patients were followed up until hospital discharge or death. The analysis focused on events occurring during the hospital stay. The primary outcome was a composite outcome, defined as the occurrence of any of the following events during hospitalization: in-hospital all-cause death, ICU admission, or IMV. Secondary outcomes: individual components of the composite outcome. Additional details of definition of the outcomes are available in Methods section of the Supplementary Material.

Statistical Analyses and Cluster Analyses Variable Selection

A comprehensive dataset of 130 variables was established. Variables with >10% missing data were excluded. For the remaining 86 variables, missing values were imputed using multilevel generalized linear models via multiple imputation. Dimensionality reduction was performed through exploratory factor analysis (continuous variables) and multiple correspondence analysis (categorical variables) to identify PCA candidates. Twelve variables were ultimately selected for PCA based on factor loadings (absolute value > 0.5), correlation coefficient < 0.6, and clinical relevance (Table S2), including: sex, age, BMI, smoking status, the presence of anemia and peripheral vascular disease, COTE-index, neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), level of serum albumin, prior-year severe exacerbations, and level of CRP.19

Cluster Analysis

Cluster identification followed a two-stage approach using PCA-derived components, as described in our previous studies.11,20 Methodological details are provided in the Methods section of Supplementary Materials.

Other Analysis

Categorical variables are presented as counts (%). Inter-cluster comparisons employed: ANOVA/Kruskal-Wallis tests for continuous variables (normality assessed via Shapiro-Wilk); χ²-test/Fisher’s exact tests for categorical variables, with Bonferroni-adjusted post hoc analyses. A decision-tree classifier (10-fold cross-validation) quantified cluster predictability, reporting misclassification rates.21 Multivariable logistic regression adjusted for demographics (sex, age), clinical characteristics (exacerbation severity, COPD grade, BMI, mean arterial pressure [MAP], COTE index), treatment factors (non-invasive ventilation, pre-/in-hospital therapies), and smoking status when assessing phenotype-outcome associations. Statistical significance was set at α=0.05.

Guided by our causal inference framework (Figure S2), we implemented a four-phase mediation analysis to quantify how inflammatory mediators explain phenotype-outcome associations.22–25 The pure indirect effect (PIE) was estimated through: 1) phenotype-mediator regressions, 2) mediator-outcome logistic models (covariate-adjusted), and 3) counterfactual decomposition using the mediation R package (v4.3.1). Age, sex, and smoking status were included as adjusted variables throughout. Methodological specifications are detailed in the Supplementary Methods.

Results Subject Characteristics

The training cohort included 20,890 ECOPD patients (validation: n=8,954; Tables 1, S3 and S4). Patients had a median age of 72.3 years (IQR: 65.5–79.5) and BMI of 22.0 kg/m² (IQR: 19.49, 24.46), with 19.89% being female and 85.65% current/ex-smokers. Clinically, 41.97% experienced ≥1 severe exacerbation in the prior year, while prevalent comorbidities included anemia (57.29%), hypertension (40.15%), cardiovascular disease (CVD; 36.38%), and diabetes (19.81%). Frailty and coagulopathy were observed in 18.34% and 18.48% of patients, respectively. Critical hospitalization outcomes comprised 2.04% (426) mortality, 7.59% (1,587) invasive ventilation events, 7.71% (1611) ICU admissions, and length of stay (median=10 days, IQR: 7–15).

Table 1 Demographic and Clinical Characteristics of ECOPD Patients by Cluster Analysis in the Training Cohort

Cluster Profile Cluster 1: Female Eosinophilic Phenotype

Cluster 1 comprised 4,944 patients (23.67%), characterized by female predominance (65.31%) and higher blood eosinophil counts (BEC; median 0.11×109/L [IQR: 0.05–0.20]) compared to Cluster 3 (0.05×109/L [IQR:0–0.15]; P<0.001). This cluster showed the fewer acute exacerbations (AEs) (22.11%) in the preceding year, in-hospital mortality (0.24%), and the composite outcome incidence (2.95%).

Cluster 2: Male Eosinophilic Phenotype

Cluster 2 (n=10,814, 51.77%) demonstrated male predominance (99.5%) with elevated eosinophilic markers: median eosinophil count 0.13×109/L [IQR:0.06–0.23] and eosinophil–basophil ratio (EBR) 5.25 [IQR:2.75–9], both significantly higher than Cluster 3 (P<0.001).

Cluster 3: Geriatric Multimorbidity-Associated Neutrophilic Systemic Inflammatory Phenotype

Cluster 3 comprised 5,132 patients (24.57%) and was characterized by older age (median 76.20 years [IQR: 68.8–83.0]), the highest COTE-index score, and extensive comorbidities, including hypertension (43.49%), anemia (82.36%), cognitive impairment (29.9%), frailty (40.4%), and CVD (51.21%). This phenotype exhibited more neutrophilic inflammation (7.28×109/L [IQR: 4.20–10.99] neutrophils vs 3.92×109/L [2.90–5.20] in Cluster 1; P<0.001) and systemic inflammation markers: CRP (44.50 mg/L [11.40–104.00]), PCT (0.15 ng/mL [0.06–0.54]), and D-Dimer (1.78 μg/mL [0.76–4.37] vs 0.62μg/mL [0.32–1.34] in Cluster 2). Cachexia prevalence (BMI < 18.5 kg/m²) was nearly 3-fold higher than Cluster 1 (28.47% vs 9.91%; P<0.001). Cluster 3 demonstrated the worst outcomes with 84.49% prior-year severe exacerbation, 22.31% in-hospital composite outcome rate, and 6.97% mortality.

In-Hospital Outcomes and Cluster Prediction

Cluster 3 exhibited the most severe in-hospital outcomes, with a composite outcome rate of 22.31% (n=1,145) versus 2.95% in Cluster 1, alongside significantly higher mortality (6.97%, n=358) and prolonged median hospitalization duration (13 days, IQR: 9–20). This cluster also demonstrated elevated critical care needs: 18.70% required IMV and 18.99% required ICU admission. Logistic regression analyses revealed substantially increased risks in Cluster 3 compared to Clusters 1–2, with adjusted odds ratios of 8.18 (95% CI: 6.72–10.14; P<0.001) for the composite outcome, 17.07 (95% CI: 9.45–30.66; P<0.001) for mortality, 7.59 (95% CI: 6.08–9.51; P<0.001) for IMV, and 7.72 (95% CI: 6.22–9.82; P<0.001) for ICU admission (Figure 1). A decision-tree model (sex, smoking status, prior severe exacerbation, neutrophil-to-lymphocyte ratio [NLR]) achieved 85.3% classification accuracy, comparable to models using all 12 variables (Figure 2). Detailed risk factor analyses are presented in Tables S5–S11.

Figure 1 Forest plot illustrating the three identified clusters in relation to in-hospital outcomes.

Notes: This figure presents the correlations of the three clusters with death, ICU admission, invasive ventilation, and a composite outcome, using Cluster 1 as the reference. (A) Cluster 2 exhibited a significantly increased risk of in-hospital death, ICU admission, invasive ventilation, and the composite outcome compared to Cluster 1. The models are as follows: Model I = crude; Model II = adjusted for sex and age; Model III = Model II plus further adjustments for exacerbation severity, COPD grades, smoking status, BMI, and MAP; Model IV = Model III plus the COTE-index and prehospital and in-hospital treatments collected in our study. (B) Cluster 3 showed a significantly increased risk of in-hospital death, ICU admission, invasive ventilation, and the composite outcome compared to Cluster 1, with the models adjusted in the same manner as for Cluster 2.

Figure 2 Decision tree analysis and classification of predicted clinical cluster assignments.

Abbreviations: SE, Severe exacerbation in the preceding year; NLR, Neutrophil-to-lymphocyte ratio.

Notes: This figure illustrates a decision tree analysis and the classification of predicted clinical cluster assignments for patients. Patient assignments to the three clusters were based on four variables.

Mediation Analysis Reveals Phenotype-Specific Immunoinflammatory Mechanisms in ECOPD

Mediation analysis investigated immunoinflammatory mechanisms linking ECOPD phenotypes to adverse outcomes (Figures S3–S6). Key mediators (eosinophils, neutrophils, CRP, and serum albumin) differentially modulated phenotype-outcome associations across clusters. In Cluster 2, eosinophils mediated 17.2–19.6% of adverse outcome risks: 17.2% for mortality, 19.2% for ICU admission, 19.6% for IMV, and 17.3% for composite outcomes. In contrast, composite outcome of Cluster 3 was predominantly driven by neutrophil-mediated pathways (33.6% risk elevation), systemic inflammation (CRP-mediated effects; 20.31%), and hypoalbuminemia (32.1%). Neutrophils mediated 34.7% of the effect of Cluster 3 on mortality (PIE = 0.72, 95% CI: 0.46–0.94), 31.6% on ICU admission (PIE = 0.11, 95% CI: 0.02–0.21), and 32.2% on invasive mechanical ventilation (PIE = 0.12, 95% CI: 0.03–0.21).

Discussion

Our study is the first to employ cluster analysis for stratifying ECOPD patients into distinct clinical-inflammatory phenotypes at hospitalization and validate their association with adverse in-hospital outcomes. In a real-world cohort of 20,890 patients, we identified three phenotypes, with the “Geriatric Multimorbidity-Associated Neutrophilic Systemic Inflammatory Phenotype” (Cluster 3) demonstrating poor outcomes: 6.97% mortality, 22.31% the composite adverse outcome (IMV, ICU admission, or death), and 18.99% ICU admission rates. These figures fall within the ranges reposted in prior ECOPD studies, which document mortality of 1.0–11.8% and ICU admission rates of 17.4–24.4%, underscoring the limitations of existing etiology-based classifications (eg, viral vs bacterial) that fail to adequately stratify risk in multidimensional clinical-inflammatory profiles.26–29

Our study reveals that blood eosinophils significantly moderate the influence of ECOPD phenotypes on several in-hospital outcomes, including a diminished risk of the composite outcome, IMV, ICU admission, and in-hospital mortality. Mechanically, elevated eosinophils during exacerbations likely indicate the persistence of Type 2 inflammation.30 In our study, the BEC threshold for defining eosinophilic ECOPD (≥0.11×109/L) is lower than the established stable-phase cutoff (≥0.3×109/L) associated with ICS efficacy. This threshold discrepancy likely reflects acute-phase eosinophil depletion, evidenced by Bafadhel et al’s study of median BEC declines at 0.11×109/L (IQR: 0.10–0.13) during ECOPD.31 Whether such changes can adequately predict ICS responsiveness during hospitalization remains unproven, necessitating prospective trials to correlate acute-phase BEC with post-ICS or targeted eosinophil therapy outcomes. Mechanistic studies are essential to verify whether eosinophil-rich exacerbations display elevated glucocorticoid receptor-α expression in the airway.31–33 Our findings support the GOLD guideline’s recommendation to consider ICS for stable COPD patients with BEC of ≥100 cells/μL, particularly when this aligns with their history of exacerbations.14 Future research should verify if acute-phase thresholds for BEC need to be adjusted downward relative to stable-phase criteria, ensuring a balance between therapeutic benefits and the risks of excessive ICS and targeted eosinophil therapy use in non-eosinophilic phenotypes.

Besides, our additional analysis revealed significant sex-based differences in eosinophil counts, with male patients demonstrating higher median blood eosinophil levels (0.12×109/L) compared to females (0.08×109/L; P<0.001), a finding consistent with prior studies.34–36 Experimental models indicate estrogen suppresses eosinophilopoiesis through ERα-mediated inhibition of IL-5/IL-13 signaling in bone marrow progenitors, while clinical data show premenopausal women exhibit 32–41% lower eosinophil counts than age-matched men, potentially reflecting estradiol’s attenuation of eosinophil peroxidase release and leukotriene synthesis.37 However, the precise mechanisms underlying sex hormone-eosinophil interactions in COPD remain elusive, particularly regarding androgen receptor modulation of eosinophil survival, menopause-related hormonal shifts, and tissue-specific inflammatory regulation.38,39 Future investigations integrating hormone profiling with single-cell transcriptomic analysis of airway eosinophils are warranted to delineate these pathways and their therapeutic implications.

Our study identifies an ECOPD phenotype distinguished by reduced eosinophil counts (Cluster 3), which correlates with inferior hospital outcomes. The data indicate that low eosinophil levels are associated with more severe exacerbation and elevated hospitalization rates, thereby serving as prognostic indicators for clinical deterioration.40 It has confirmed that non-eosinophilic patients exhibit higher mortality rates compared to their eosinophilic counterparts, thereby reinforcing our findings from Cluster 3.41 Furthermore, patients in Cluster 3 display a substantial morbidity burden, signifying a complex profile marked by systemic inflammatory responses and multiple comorbidities. The effective management of these patients requires a comprehensive approach that tackles both pulmonary and systemic issues.

Systemic inflammation, which is directly correlated with adverse clinical outcomes, serves as an indicator of the severity of COPD.42 Elevated CRP levels are particularly noteworthy in patients with severe COPD, signifying not only the occurrence of exacerbations but also persistent chronic inflammation.43,44 The elevated CRP levels in Cluster 3 patients are associated with poorer in-hospital adverse outcomes compared to other clusters, likely attributable to their heightened systemic inflammatory responses exacerbated by multiple comorbidities. Mediation analysis further elucidates that elevated CRP levels significantly mediate the relationship between ECOPD phenotypes and in-hospital mortality. Consequently, CRP emerges as a crucial prognostic biomarker in ECOPD.45–47 Further research is essential to comprehensively understand these relationships and assess the potential impact of interventions designed to reduce CRP, particularly in high-risk groups such as Cluster 3. Moreover, there is a pressing need for personalized management strategies that concurrently address systemic inflammation and associated comorbidities. Besides, albumin is a crucial biomarker for assessing health status in ECOPD patients. In Cluster 3, more than 90% of patients exhibit levels below 40 g/L.48 This reduced albumin level indicates inadequate nutrition, thereby exacerbating clinical outcomes.49 Hypoalbuminemia in Cluster 3 may arise due to inadequate dietary intake and persistent inflammation, exacerbated by respiratory stress and recurrent exacerbations.50 This cycle of declining health extends beyond malnutrition, impacting nutritional status, immune response, and tissue repair, thereby heightening the risk of severe outcomes.51

This study’s strengths include a large, rigorously curated cohort of 20,890 hospitalized ECOPD patients, enabling robust phenotype identification through multidimensional data integration. We systematically addressed high-dimensional complexity by applying exploratory factor analysis and multiple correspondence analysis, distilling 130 parameters to 12 clinically interpretable biomarkers, thereby balancing methodological transparency with replicability. However, our study has several limitations. Firstly, we used blood eosinophil counts as a surrogate for sputum eosinophils to predict the eosinophilic phenotype, although this method has been validated.52 Secondly, despite employing a broad range of variables and objective statistical methods, some subjectivity in variable selection remains.53 Additionally, we did not identify the causes of exacerbation, including viral and bacterial profiles.54 Lastly, the generalizability of our findings is limited since patient phenotypes were identified from a single center without external validation, despite the robust sample size.

Conclusion

This study defines three ECOPD phenotypes stratified by sex, multimorbidity, and inflammatory endotypes, advancing eosinophil-guided exacerbation thresholds and prognostic risk stratification. We establish a framework integrating acute-phase eosinophil dynamics with systemic inflammation to personalize management, considering neutrophilic inflammation and multimorbidity as therapeutic targets. Our findings could optimize acute-care precision and phenotype-driven resource allocation.

Abbreviations

ECOPD, Exacerbations of chronic obstructive pulmonary disease; ALB, Albumin; BMI, Body mass index; COT, conventional oxygen therapy; COTE, index -the COPD, specific co-morbidity test; CVD, cardiovascular disease; CRP, C-reactive protein; CI, Confidence interval; HFNC, high-Flow Nasal Cannula therapy; IMV, invasive mechanical ventilation; ICU, Intensive care unit; ICS, inhaled corticosteroids; GOLD, the Global Initiative for Obstructive Lung Diseases; LABA, long-acting beta-agonists; LAMA, long-acting muscarinic antagonists; LOS, Length of stay; MAP, mean arterial pressure; NIV, non-invasive ventilation; NLR, Neutrophil-to-lymphocyte ratio; PCA, Principal component analysis; PLR, Platelet-to-lymphocyte ratio; PIE, the pure indirect effect; PaO2, Partial pressure of oxygen; OR, Odds ratio; ORadj, Odds ratio after adjusting for confounders; SABA, short-acting beta-agonists; SAMA, short-acting beta-agonists; SE, Severe exacerbation in the preceding year.

Data Sharing Statement

Supplementary data to this article can be found online. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethical Approval

The study was approved by the Institutional Review Board (IRB) of West China Hospital, Sichuan University (Chengdu, China) (No. 2023–1882). All patients provided written informed consent.

Acknowledgments

We thank all participating patients and all departments, healthcare providers and professionals involved in this study.

Author Contributions

All authors made a significant contribution to the work reported, whether in conception, study design, execution, data acquisition, analysis, and interpretation, or across all these areas; participated in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; agreed on the journal to which the article was submitted; and agreed to be accountable for all aspects of the work.

Funding

This study was supported by 1.3.5 project for disciplines of excellence-Clinical Research Fund, West China Hospital, Sichuan University (2023HXFH045), 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (ZYGD23002), Science and Technology Foundation of Sichuan Province (2022NSFSC1278), National Natural Science Foundation of China (82100032, 82300036, 82200037), Post-Doctor Research Project, West China Hospital, Sichuan University (2023HXBH093), China Postdoctoral Science Foundation (2022M722294), and National Clinical Research Center for Geriatrics, West China Hospital, Sichuan University (Z20192005).

Disclosure

The authors declare that they have no conflicts of interest to disclose.

References

1. Jacobs DM, Noyes K, Zhao J, et al. Early hospital readmissions after an acute exacerbation of chronic obstructive pulmonary disease in the nationwide readmissions database. Ann Am Thorac Soc. 2018;15(7):837–845. doi:10.1513/AnnalsATS.201712-913OC

2. Myers LC, Faridi MK, Hasegawa K, Hanania NA, Camargo CA. The hospital readmissions reduction program and readmissions for chronic obstructive pulmonary disease, 2006-2015. Ann Am Thorac Soc. 2020;17(4):450–456. doi:10.1513/AnnalsATS.201909-672OC

3. Shah T, Churpek MM, Coca Perraillon M, Konetzka RT. Understanding why patients with COPD get readmitted: a large national study to delineate the medicare population for the readmissions penalty expansion. Chest. 2015;147(5):1219–1226. doi:10.1378/chest.14-2181

4. Myers LC, Faridi MK, Hasegawa K, Camargo CA. Pulmonary rehabilitation and readmission rates for medicare beneficiaries with acute exacerbation of chronic obstructive pulmonary disease. Chronic Obstr Pulm Dis. 2021;8(4):427–440. doi:10.15326/jcopdf.2020.0193

5. Burgel PR, Paillasseur JL, Caillaud D, et al. Clinical COPD phenotypes: a novel approach using principal component and cluster analyses. Eur Respir J. 2010;36(3):531–539. doi:10.1183/09031936.00175109

6. Agusti A, Calverley PMA, Celli B, et al. Characterisation of COPD heterogeneity in the ECLIPSE cohort. Respir Res. 2010;11(1):122. doi:10.1186/1465-9921-11-122

7. Ruan H, Zhang H, Wang J, Zhao H, Han W, Li J. Readmission rate for acute exacerbation of chronic obstructive pulmonary disease: a systematic review and meta-analysis. Respir Med. 2023;206:107090. doi:10.1016/j.rmed.2022.107090

8. Suissa S, Dell’Aniello S, Ernst P. Long-term natural history of chronic obstructive pulmonary disease: severe exacerbations and mortality. Thorax. 2012;67(11):957–963. doi:10.1136/thoraxjnl-2011-201518

9. Pepin JL, Lemeille P, Denis H, et al. Health trajectories before initiation of non-invasive ventilation for chronic obstructive pulmonary disease: a French nationwide database analysis. Lancet Reg Health Eur. 2023;34:100717. doi:10.1016/j.lanepe.2023.100717

10. Chen X, Li J, Liu G, et al. Identification of distinct clinical phenotypes of heterogeneous mechanically ventilated ICU patients using cluster analysis. JCM. 2023;12(4):1499. doi:10.3390/jcm12041499

11. Zhang X, Zhang L, Wang G, et al. Clinical phenotypes of patients hospitalized for an asthma exacerbation: prognostic implications. J Allergy Clin Immunol Pract. 2021;9(2):830–841.e14. doi:10.1016/j.jaip.2020.09.031

12. Castaldi PJ, Dy J, Ross J, et al. Cluster analysis in the COPDGene study identifies subtypes of smokers with distinct patterns of airway disease and emphysema. Thorax. 2014;69(5):415–422. doi:10.1136/thoraxjnl-2013-203601

13. Gregory A, Xu Z, Pratte K, et al. Clustering-based COPD subtypes have distinct longitudinal outcomes and multi-omics biomarkers. BMJ Open Respir Res. 2022;9(1):e001182. doi:10.1136/bmjresp-2021-001182

14. Venkatesan P. GOLD COPD report: 2024 update. Lancet Respir Med. 2024;12(1):15–16. doi:10.1016/S2213-2600(23)00461-7

15. Mirza S, Clay RD, Koslow MA, Scanlon PD. COPD guidelines: a review of the 2018 GOLD report. Mayo Clin Proc. 2018;93(10):1488–1502. doi:10.1016/j.mayocp.2018.05.026

16. Simpson JL. Airway inflammation in COPD: is it worth measuring and is it clinically meaningful? Respirol. 2020;25(1):47–48. doi:10.1111/resp.13656

17. Divo M, Cote C, De Torres JP, et al. Comorbidities and risk of mortality in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2012;186(2):155–161. doi:10.1164/rccm.201201-0034OC

18. Shiwaku K, Anuurad E, Enkhmaa B, Kitajima K, Yamane Y. Appropriate BMI for asian populations. Lancet. 2004;363(9414):1077. doi:10.1016/S0140-6736(04)15856-X

19. Norris M, Lecavalier L. Evaluating the use of exploratory factor analysis in developmental disability psychological research. J Autism Dev Disord. 2010;40(1):8–20. doi:10.1007/s10803-009-0816-2

20. Zhang X, Deng K, Yuan Y, et al. Body composition-specific asthma phenotypes: clinical implications. Nutrients. 2022;14(12):2525. doi:10.3390/nu14122525

21. Zakrzewska M, Gruszka R, Stawiski K, et al. Expression-based decision tree model reveals distinct microRNA expression pattern in pediatric neuronal and mixed neuronal-glial tumors. BMC Cancer. 2019;19(1):544. doi:10.1186/s12885-019-5739-5

22. Imai K, Keele L, Tingley D. A general approach to causal mediation analysis. Psychol Methods. 2010;15(4):309–334. doi:10.1037/a0020761

23. VanderWeele TJ. Mediation analysis: a practitioner’s guide. Annu Rev Public Health. 2016;37(1):17–32. doi:10.1146/annurev-publhealth-032315-021402

24. Chi WE, Huang S, Jeon M, Park ES, Melguizo T, Kezar A. A practical guide to causal mediation analysis: illustration with a comprehensive college transition program and nonprogram peer and faculty interactions. Front Educ. 2022;7:886722. doi:10.3389/feduc.2022.886722

25. Richiardi L, Bellocco R, Zugna D. Mediation analysis in epidemiology: methods, interpretation and bias. Int J Epidemiol. 2013;42(5):1511–1519. doi:10.1093/ije/dyt127

26. Fuchs L, Chronaki CE, Park S, et al. ICU admission characteristics and mortality rates among elderly and very elderly patients. Intensive Care Med. 2012;38(10):1654–1661. doi:10.1007/s00134-012-2629-6

27. Waeijen-Smit K, Crutsen M, Keene S, et al. Global mortality and readmission rates following COPD exacerbation-related hospitalisation: a meta-analysis of 65 945 individual patients. ERJ Open Res. 2024;10(1):00838–02023. doi:10.1183/23120541.00838-2023

28. MacDonald MI, Osadnik CR, Bulfin L, et al. MULTI-PHACET: multidimensional clinical phenotyping of hospitalised acute COPD exacerbations. ERJ Open Res. 2021;7(3):00198–02021. doi:10.1183/23120541.00198-2021

29. Fei G, Dai MY, Qiao JP, Xu YH. Respiratory infectious phenotypes in acute exacerbation of COPD: an aid to length of stay and COPD assessment test. COPD. 2015;2257. doi:10.2147/COPD.S92160

30. Reddy AT, Lakshmi SP, Banno A, Reddy RC. Glucocorticoid receptor α mediates roflumilast’s ability to restore dexamethasone sensitivity in COPD. Int J Chronic Obstr Pulm Dis. 2020;15:125–134. doi:10.2147/COPD.S230188

31. Bafadhel M, McKenna S, Terry S, et al. Blood eosinophils to direct corticosteroid treatment of exacerbations of chronic obstructive pulmonary disease: a randomized placebo-controlled trial. Am J Respir Crit Care Med. 2012;186(1):48–55. doi:10.1164/rccm.201108-1553OC

32. Gedebjerg A, Szépligeti SK, Wackerhausen LMH, et al. Prediction of mortality in patients with chronic obstructive pulmonary disease with the new global initiative for chronic obstructive lung disease 2017 classification: a cohort study. Lancet Respir Med. 2018;6(3):204–212. doi:10.1016/S2213-2600(18)30002-X

33. Bhatt SP, Rabe KF, Hanania NA, et al. Dupilumab for COPD with blood eosinophil evidence of type 2 inflammation. N Engl J Med. 2024;390(24):NEJMoa2401304. doi:10.1056/NEJMoa2401304

34. Artham S, Juras PK, Goyal A, et al. Estrogen signaling suppresses tumor-associated tissue eosinophilia to promote breast tumor growth. Sci Adv. 2024;10(39):eadp2442. doi:10.1126/sciadv.adp2442

35. Caspard H, Ambrose CS, Tran TN, Chipps BE, Zeiger RS. Associations between individual characteristics and blood eosinophil counts in adults with asthma or COPD. J Allergy Clin Immunol Pract. 2020;8(5):1606–1613.e1. doi:10.1016/j.jaip.2019.12.019

36. DiSantostefano RL, Hinds D, Le HV, Barnes NC. Relationship between blood eosinophils and clinical characteristics in a cross-sectional study of a US population-based COPD cohort. Respir Med. 2016;112:88–96. doi:10.1016/j.rmed.2016.01.013

37. Oliveira SHP, Taub DD, Nagel J, et al. Stem cell factor induces eosinophil activation and degranulation: mediator release and gene array analysis. Blood. 2002;100(13):4291–4297. doi:10.1182/blood.V100.13.4291

38. Brightling C, Greening N. Airway inflammation in COPD: progress to precision medicine. Eur Respir J. 2019;54(2):1900651. doi:10.1183/13993003.00651-2019

39. Gonsalves NP, Aceves SS. Diagnosis and treatment of eosinophilic esophagitis. J Allergy Clin Immunol. 2020;145(1):1–7. doi:10.1016/j.jaci.2019.11.011

40. Han MK, Quibrera PM, Carretta EE, et al. Frequency of exacerbations in patients with chronic obstructive pulmonary disease: an analysis of the SPIROMICS cohort. Lancet Respir Med. 2017;5(8):619–626. doi:10.1016/S2213-2600(17)30207-2

41. Ramakrishnan S, Jeffers H, Langford-Wiley B, et al. Blood eosinophil-guided oral prednisolone for COPD exacerbations in primary care in the UK (STARR2): a non-inferiority, multicentre, double-blind, placebo-controlled, randomised controlled trial. Lancet Respir Med. 2024;12(1):67–77. doi:10.1016/S2213-2600(23)00298-9

42. Barnes PJ. Inflammatory mechanisms in patients with chronic obstructive pulmonary disease. J Allergy Clin Immunol. 2016;138(1):16–27. doi:10.1016/j.jaci.2016.05.011

43. Butler CC, Gillespie D, White P, et al. C-reactive protein testing to guide antibiotic prescribing for COPD exacerbations. N Engl J Med. 2019;381(2):111–120. doi:10.1056/NEJMoa1803185

44. Sacks HS. In primary care, CRP testing reduced antibiotic use in COPD exacerbations without worsening health. Ann Intern Med. 2019;171(10):JC51. doi:10.7326/ACPJ201911190-051

45. Joppa P, Petrasova D, Stancak B, Tkacova R. Systemic inflammation in patients with COPD and pulmonary hypertension. Chest. 2006;130(2):326–333. doi:10.1378/chest.130.2.326

46. Prins HJ, Van Der Werf TS, Boersma WG. Is CRP-guided antibiotic treatment a safe way to reduce antibiotic use in severe hospitalised patients with exacerbations of COPD?. Eur Respir J. 2019;54(4):1901597. doi:10.1183/13993003.01597-2019

47. Prins HJ, Duijkers R, van der Valk P, et al. CRP-guided antibiotic treatment in acute exacerbations of COPD in hospital admissions. Eur Respir J. 2019;53(5):1802014. doi:10.1183/13993003.02014-2018

48. Wang Y, Zhang J, Zhang J, et al. Association between per- and polyfluoroalkyl substances exposure and prevalence of chronic obstructive pulmonary disease: the mediating role of serum albumin. Sci Total Environ. 2024;925:171742. doi:10.1016/j.scitotenv.2024.171742

49. Mador MJ, Sethi S. Systemic inflammation in predicting COPD exacerbations. JAMA. 2013;309(22):2390–2391. doi:10.1001/jama.2013.5733

50. Brightling CE. Biomarkers that predict and guide therapy for exacerbations of chronic obstructive pulmonary disease. Ann Am Thorac Soc. 2013;10(Supplement):S214–S219. doi:10.1513/AnnalsATS.201302-023AW

51. Brett AS, Al-Hasan MN. COPD exacerbations - a target for antibiotic stewardship. N Engl J Med. 2019;381(2):174–175. doi:10.1056/NEJMe1905520

52. Zhang XY, Simpson JL, Powell H, et al. Full blood count parameters for the detection of asthma inflammatory phenotypes. Clin Exp Allergy. 2014;44(9):1137–1145. doi:10.1111/cea.12345

53. Hwang J, Shin H, Jung MC. Joint motion pattern classification by cluster analysis of kinematic, demographic, and subjective variables. Appl Ergon. 2013;44(4):636–642. doi:10.1016/j.apergo.2012.12.005

54. Marques A, Souto-Miranda S, Machado A, et al. COPD profiles and treatable traits using minimal resources: identification, decision tree and stability over time. Respir Res. 2022;23(1):30. doi:10.1186/s12931-022-01954-6