Abstract
Background
Chronic obstructive pulmonary disease (COPD) remains a major global health burden and is currently the third leading cause of death worldwide. Acute exacerbations accelerate disease progression and contribute substantially to mortality, underscoring the urgent need for reliable prognostic biomarkers. The endothelial activation and stress index (EASIX), a composite indicator of endothelial dysfunction, has demonstrated prognostic utility across diverse critical illnesses. However, its association with clinical outcomes in critically ill patients with COPD has not been clearly established.
Methods
In this retrospective cohort study, data of critically ill patients with COPD were extracted from the Medical Information Mart for Intensive Care (MIMIC) database. Participants were stratified into tertiles based on EASIX values, and intergroup differences in clinical characteristics were analyzed. The relationship between EASIX and 28-day all-cause mortality was examined using Kaplan–Meier survival analysis, Cox proportional hazards regression, and restricted cubic spline modeling. The Boruta algorithm was applied to assess the relative importance of candidate predictors, and prognostic models were subsequently developed using six machine learning algorithms.
Results
A total of 4,590 patients met the inclusion criteria. The incidence of 28-day ICU mortality increased progressively across higher EASIX tertiles (p < 0.001). EASIX was independently associated with 28-day ICU all-cause mortality, with both unadjusted and fully adjusted Cox models confirming this relationship (unadjusted HR = 1.21, p < 0.001; adjusted HR = 1.082, p < 0.001). Subgroup analyses demonstrated that the association between elevated EASIX and mortality risk remained consistent across demographic and comorbidity categories (p for interaction > 0.05 for all). The Boruta algorithm identified EASIX as one of the most important predictors of 28-day mortality. Among the six machine learning models evaluated, the XGBoost algorithm yielded the highest discriminative (AUC = 0.823), calibration and clinical application.
Conclusions
EASIX serves as an independent prognostic marker for 28-day all-cause mortality in critically ill COPD patients. Furthermore, the EASIX-based machine learning model demonstrated strong predictive accuracy, supporting its potential as a valuable clinical tool or early risk stratification and decision-making in intensive care settings.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12890-026-04299-8.
Keywords: Chronic obstructive pulmonary disease, Endothelial activation and stress index, Mortality, Risk factor, Machine learning
Background
Chronic obstructive pulmonary disease (COPD) is a prevalent and progressive respiratory disorder arising from a complex interplay of genetic predisposition and environmental exposures. With a global prevalence of approximately 10.3%, COPD ranks as the third leading cause of mortality worldwide [1–3]. Acute exacerbations—episodes characterized by an acute worsening of respiratory symptoms beyond normal daily variation—accelerate disease progression and contribute substantially to morbidity and mortality [4, 5]. These events are associated with rapid declines in lung function, increased requirements for mechanical ventilation, prolonged hospital stays, and higher short-term mortality rates [4, 6]. Such adverse outcomes emphasize the urgent need for reliable prognostic biomarkers that can facilitate early identification of high-risk patients and guide tailored management strategies.
Prognostic assessment of critically ill patients with COPD currently relies primarily on clinical indices and pulmonary function parameters. Although scoring systems such as the Acute Physiology and Chronic Health Evaluation II (APACHE II) and the Pneumonia Severity Index (PSI) are used in routine practice, there remains a critical need for more accessible indicators rather than cumbersome scoring systems to evaluate prognosis in this highly heterogeneous population [7–9]. Recent studies have explored hematologic and inflammatory biomarkers, including the platelet-to-lymphocyte ratio and the systemic immune-inflammation index, for their potential prognostic utility in critically ill COPD patients [10, 11]. However, these parameters demonstrate limited specificity and discriminative ability, underscoring the continued need for simple, objective, and accurate tools for early risk stratification in intensive care settings.
Emerging evidence has highlighted the pivotal role of endothelial dysfunction in the pathogenesis and progression of COPD [12, 13]. Endothelial impairment appears early in the disease course, intensifies with advancing severity, and has been closely linked to the development of emphysema severity and the frequency of acute exacerbations [14]. The endothelial activation and stress index (EASIX) is a novel quantitative marker derived from routine laboratory parameters—lactate dehydrogenase (LDH), serum creatinine (Scr), and platelet count (PLT)—and reflects the degree of microvascular injury and endothelial stress [15]. Initially established to predict complications following hematopoietic stem cell transplantation, EASIX has since demonstrated prognostic relevance in various critical illnesses, including acute pancreatitis, acute kidney injury, and coronary artery disease [16–19]. Its predictive value has also been validated in several respiratory disorders such as asthma, small cell lung cancer, and acute respiratory distress syndrome, suggesting broad applicability as an indicator of endothelial dysfunction and systemic stress [20–22].
Despite the growing recognition of EASIX in diverse clinical contexts, its relationship with outcomes in critically ill patients with COPD has not been systematically investigated. Given the central contribution of endothelial dysfunction to COPD pathophysiology, elucidating this association holds important clinical implications. Therefore, this study aimed to comprehensively examine the relationship between EASIX and 28-day all-cause mortality in critically ill patients with COPD using multiple analytical approaches, including regression and machine learning methods. By evaluating the prognostic significance of EASIX, this study seeks to provide a robust theoretical and practical basis for early risk stratification and personalized management in this vulnerable patient population.
Methods
Study design
This retrospective study used data extracted from the Medical Information Mart for Intensive Care (MIMIC-III and MIMIC-IV) databases, developed and maintained by the Laboratory for Computational Physiology at the Massachusetts Institute of Technology. These publicly available databases contain de-identified, high-quality clinical data from intensive care unit (ICU) admissions at Beth Israel Deaconess Medical Center, including detailed information on demographics, vital signs, laboratory test results, and therapeutic interventions. Access to the databases and data extraction were approved under certificate number 52,663,507.
Ethics approval and consent to participate
The studies involving humans were approved by the Institutional Review Boards of Beth Israel Deaconess Medical Center and the Massachusetts Institute of Technology. The studies were conducted in accordance with local legislation and institutional requirements. Since the database contains no protected health information, the requirement for informed consent was waived. Accordingly, no additional ethical approval was required for this retrospective analysis. All procedures were conducted in accordance with relevant guidelines and regulations, including the Declaration of Helsinki.
Inclusion and exclusion criteria
Eligible participants were adult patients (aged ≥ 18 years) with an ICU stay exceeding 24 h and a confirmed diagnosis of COPD, identified using International Classification of Diseases (ICD) codes (ICD-10: J44, J440, J441, J449; ICD-9: 49120, 49121, 49122, 496) [23]. Inclusion required availability of serum lactate dehydrogenase (LDH), serum creatinine (Scr), and platelet count (PLT) data within the first 24 h of ICU admission. Critically ill COPD patients included in the study did not receive invasive mechanical ventilation before ICU admission. Patients with missing laboratory data within 48 h or those with multiple ICU admissions were excluded, with only the first admission retained for analysis. The final cohort comprised 4,590 patients, who were subsequently stratified into tertiles based on EASIX levels (Fig. 1).
Fig. 1.
The flow chart of this study
Data extraction
Clinical data were extracted using PostgreSQL (version 16.10) via Structured Query Language (SQL). The extracted variables included baseline demographic characteristics; comorbidities (defined by ICD codes); vital signs; laboratory indices at admission (hematologic, biochemical, coagulation, and arterial blood gas parameters); and illness severity scores—Acute Physiology Score III (APS III), Sequential Organ Failure Assessment (SOFA), Simplified Acute Physiology Score II (SAPS II), Oxford Acute Severity of Illness Score (OASIS), Glasgow Coma Scale (GCS), Acute Physiology and Chronic Health Evaluation II (APACHE II), and Systemic Inflammatory Response Syndrome (SIRS) score—recorded within the first 24 h of ICU admission. Treatment-related variables, including mechanical ventilation, endotracheal intubation, and follow-up duration from admission to death or discharge, were also collected. Additional extraction details are provided in the Supplementary Materials.
EASIX was calculated according to the established formula [15]:
 |
To minimize bias from missing data, variables with > 30% missingness were excluded. For variables with ≤ 30% missing data, multiple imputation via the “MICE” package was performed. For details on missing data per parameter, see the Supplementary Materials (TableS2).
Clinical outcomes
The primary endpoint was defined as all-cause mortality within the ICU at 28 days after ICU admission. Secondary endpoints included all-cause in-hospital mortality at 28 days and the requirement for mechanical ventilation after hospital admission.
Statistical analysis
As a retrospective analysis, no a priori sample size calculation was conducted. Multicollinearity among covariates was assessed using the variance inflation factor (VIF), with variables exhibiting a VIF > 5 excluded. Continuous variables were expressed as mean ± standard deviation (SD) or median (interquartile range, IQR) and compared using ANOVA or Kruskal–Wallis tests, as appropriate. Categorical variables were summarized as counts (percentages) and compared using the chi-square or Fisher’s exact test. Survival distributions across EASIX tertiles were estimated using Kaplan–Meier curves and compared with the log-rank test.
The relationship between EASIX and clinical outcomes (ICU and in-hospital mortality, mechanical ventilation) was examined using Cox proportional hazards regression. EASIX was analyzed both as a continuous variable and as a categorical variable (by tertiles, with the lowest tertile serving as reference). Variables with a univariate association at P < 0.05 and clinically relevant factors were included in multivariable models. Three models were constructed: Model 1 (unadjusted), Model 2 (adjusted for age, sex, and body mass index), and Model 3 (fully adjusted for demographic, clinical, laboratory, and severity score covariates). Results are reported as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). A two-tailed P < 0.05 was considered statistically significant. All analyses were performed using R software (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria).
Restricted cubic splines
Restricted cubic spline (RCS) regression analysis was performed to evaluate potential nonlinear associations between baseline EASIX values and clinical outcomes, including 28-day all-cause ICU mortality, 28-day all-cause in-hospital mortality, and the need for mechanical ventilation.
Subgroup analysis
To assess the consistency of the prognostic value of EASIX across clinically relevant populations, prespecified subgroup analyses were conducted. Patients were stratified by sex, age (< 60 and ≥ 60 years), and body mass index (BMI; <30 and ≥ 30 kg/m2), as well as by the presence or absence of common comorbidities, including hypertension (HTN), liver cirrhosis (LC), hepatitis (HEP), type 2 diabetes mellitus (T2DM), type 1 diabetes mellitus (T1DM), heart failure (HF), myocardial infarction (MI), cancer (CA), chronic kidney disease (CKD), acute kidney injury (AKI), pneumonia (PNA), hyperlipidemia (HLD). The primary objective of these analyses was to determine whether the association between EASIX and the primary and secondary outcomes remained stable across these groups.
Establishment and validation of the prediction models
The Boruta algorithm was applied exclusively to the training set to identify the most important features in the dataset [24]. Its core mechanism involves comparing the Z-score of each original feature against the maximum Z-score of a set of randomly shuffled “shadow features.” In this process, all original features are duplicated, shuffled to create shadow features, and then evaluated using a random forest model to obtain their respective Z-scores. Based on this comparison over multiple independent trials, a feature is classified as “important” (also termed a confirmed variable) if its Z-score is significantly higher than the maximum Z-score of the shadow features. Features that do not meet this threshold are categorized as either “tentative” or “unimportant” (together termed rejected variables). The confirmed predictors were subsequently used to construct predictive models for 28-day ICU and in-hospital mortality, as well as for mechanical ventilation risk, among critically ill COPD patients. Six supervised machine learning algorithms—AdaBoost, Gradient Boosting, Random Forest, Support Vector Machine, XGBoost, and Decision Tree—were implemented using a 7:3 training-to-validation split. The training and validation sets were utilized for model establishment and evaluation, respectively. Grid search optimization was utilized to identify the most suitable hyperparameters for each algorithm. The discrimination was quantitatively evaluated by the AUC of the ROC curve, sensitivity, recall, accuracy, and F1 score. Decision curve analysis (DCA) was employed for assessing clinical effectiveness, while calibration curves were used to evaluate the accuracy of the model in predicting absolute risk. For the best-performing diagnostic models, we revalidated their generalization ability and robustness using 10-fold cross-validation to prevent overfitting. Because SOFA and SAPS II scores were used as common tools for predicting the illness severity and prognosis in critically ill patients, we also compared the predictive abilities of ML-based predictive models with those of the conventional scoring systems by using Net Reclassification Improvement (NRI) or Integrated Discrimination Improvement (IDI). Additional details regarding model development and validation are provided in the Supplementary Materials.
Results
Baseline characteristics
Table 1 summarizes the baseline characteristics of critically ill COPD patients stratified by EASIX tertiles (T1: 0.0738–0.9778, T2: 0.9784–2.1446, T3: 2.1452–7.9973). Patients in the highest EASIX tertile (T3) exhibited higher BMI, more severe illness scores, and a greater comorbidity burden—including HTN, AKI, CKD, T2DM, LC, HLD, HF, and MI—compared with those in the lower tertiles. Laboratory findings in the high-EASIX group showed elevated red blood cell distribution width (RDW), red blood cells, albumin, calcium total, chloride, potassium, anion gap, blood glucose, sodium, PCO2, PH and partial pressure of oxygen (PO2). Patients in T3 also experienced poorer clinical outcomes, including longer ICU and hospital stays, higher ICU and in-hospital mortality rates, and a greater need for mechanical ventilation and intubation (all p < 0.05). Additionally, a comparison of baseline characteristics between ICU survivors and non-survivors revealed significantly higher EASIX values among non-survivors (1.987 vs. 1.33, p < 0.001), as detailed in the Supplementary Materials (Table S1).
Table 1.
Characteristics and outcomes of participants categorized by EASIX
| Categories | Overall (n = 4590) |
T1 (n = 1530) | T2 (n = 1530) | T3 (n = 1530) | p-value |
|---|---|---|---|---|---|
| Demographic | |||||
| Age, years | 72 (22–100) | 70 (39–100) | 73.5 (28–99) | 72 (22–97) | < 0.001 |
| Gender | < 0.001 | ||||
| F | 2139 (46.60) | 830 (54.25) | 705 (46.08) | 604 (39.48) | |
| M | 2451 (53.40) | 700 (45.75) | 825 (53.92) | 926 (60.52) | |
| BMI, kg/m2 | 27.427 (0.354–95.135) | 26.122 (0.354–87.64) | 27.856 (7.824–95.135) | 28.542 (9.816–72.266) | < 0.001 |
| Language | < 0.001 | ||||
| Chinese | 24 (0.52) | 7 (0.46) | 5 (0.33) | 12 (0.78) | |
| English | 3656 (79.65) | 1180 (77.12) | 1237 (80.85) | 1239 (80.98) | |
| Portuguese | 28 (0.61) | 3 (0.20) | 16 (1.05) | 9 (0.59) | |
| Russian | 36 (0.78) | 5 (0.33) | 13 (0.85) | 18 (1.18) | |
| Spanish | 73 (1.59) | 20 (1.31) | 21 (1.37) | 32 (2.09) | |
| Other | 773 (16.84) | 315 (20.59) | 238 (15.56) | 220 (14.38) | |
| Insurance | 0.395 | ||||
| Government | 23 (0.50) | 10 (0.65) | 7 (0.46) | 6 (0.39) | |
| Medicaid | 514 (11.20) | 184 (12.03) | 161 (10.52) | 169 (11.05) | |
| Medicare | 3349 (72.96) | 1096 (71.63) | 1130 (73.86) | 1123 (73.40) | |
| Private | 653 (14.23) | 229 (14.97) | 216 (14.12) | 208 (13.59) | |
| Self pay | 6 (0.13) | 1 (0.07) | 3 (0.20) | 2 (0.13) | |
| Other | 45 (0.98) | 10 (0.65) | 13 (0.85) | 22 (1.44) | |
| Marital status | 0.199 | ||||
| Divorced | 454 (9.89) | 147 (9.61) | 155 (10.13) | 152 (9.93) | |
| Married | 1833 (39.93) | 574 (37.52) | 614 (40.13) | 645 (42.16) | |
| Single | 1038 (22.61) | 382 (24.97) | 335 (21.90) | 321 (20.98) | |
| Widowed | 783 (17.06) | 265 (17.32) | 268 (17.52) | 250 (16.34) | |
| Other | 482 (10.50) | 162 (10.59) | 158 (10.33) | 162 (10.59) | |
| Vital signs | |||||
| HR | 88 (31–191) | 89 (31–191) | 88 (31–189) | 88 (32–175) | 0.132 |
| Nbps | 120 (38–230) | 122 (38–211) | 119 (55–218) | 118 (51–230) | < 0.001 |
| Nbpd | 65 (15–6868) | 68 (15–190) | 65 (23–170) | 64 (15–6868) | 0.254 |
| Nbpm | 79.5 (9–6350) | 81 (13–6350) | 79 (9–174) | 78 (32–6116) | 0.353 |
| RR | 19 (0–61) | 19 (0–46) | 19 (0–61) | 20 (0–45) | 0.024 |
| SpO2 | 97 (58–100) | 97 (74–100) | 97 (58–100) | 97 (58–100) | < 0.001 |
| Temperature | 98.1 (0–104.5.5) | 98.1 (34.6–103.1.6.1) | 98.1 (35.3–103.4.3.4) | 98.1 (0–104.5.5) | 0.044 |
| Commorbidities | |||||
| Hypertension | 1828 (39.83) | 744 (48.63) | 651 (42.55) | 433 (28.30) | < 0.001 |
| Acute Kidney Injury | 1978 (43.09) | 307 (20.07) | 667 (43.59) | 1004 (65.62) | < 0.001 |
| Liver Cirrhosis | 328 (7.15) | 62 (4.05) | 95 (6.21) | 171 (11.18) | < 0.001 |
| Hepatitis | 149 (3.25) | 25 (1.63) | 45 (2.94) | 79 (5.16) | < 0.001 |
| Pulmonary Tuberculosis | 176 (3.83) | 62 (4.05) | 52 (3.40) | 62 (4.05) | 0.554 |
| Pneumonia | 1933 (42.11) | 659 (43.07) | 624 (40.78) | 650 (42.48) | 0.412 |
| Cerebrovascular Accident | 442 (9.63) | 136 (8.89) | 153 (10.00) | 153 (10.00) | 0.485 |
| Chronic Kidney Disease | 995 (21.68) | 101 (6.60) | 316 (20.65) | 578 (37.78) | < 0.001 |
| Cancer | 861 (18.76) | 294 (19.22) | 283 (18.50) | 284 (18.56) | 0.853 |
| Type 2 Diabetes | 1482 (32.29) | 419 (27.39) | 492 (32.16) | 571 (37.32) | < 0.001 |
| Type 1 Diabetes | 34 (0.74) | 10 (0.65) | 12 (0.78) | 12 (0.78) | 0.888 |
| Hyperlipidemia | 1581 (34.44) | 470 (30.72) | 522 (34.12) | 589 (38.50) | < 0.001 |
| Heart Failure | 2227 (48.52) | 548 (35.82) | 801 (52.35) | 878 (57.39) | < 0.001 |
| Myocardial Infarction | 402 (8.76) | 62 (4.05) | 136 (8.89) | 204 (13.33) | < 0.001 |
| Scales | |||||
| SOFA | 5 (0–19) | 3 (0–17) | 5 (0–16) | 7 (0–19) | < 0.001 |
| APSIII | 46 (6–166) | 41 (7–166) | 44.5 (6–141) | 53 (13–149) | < 0.001 |
| SAPSII | 39 (7–99) | 35 (7–89) | 39 (8–99) | 43 (13–98) | < 0.001 |
| OASIS | 34 (7–67) | 33 (7–65) | 33 (10–62) | 34 (11–67) | < 0.001 |
| GCS | 15 (3–15) | 15 (3–15) | 15 (3–15) | 15 (3–15) | 0.008 |
| APACHEII | 25 (5–52) | 23 (5–48) | 24 (7–52) | 26 (5–52) | < 0.001 |
| SIRS | 0.735 | ||||
| 0 | 21 (0.46) | 8 (0.52) | 8 (0.52) | 5 (0.33) | |
| 1 | 405 (8.82) | 129 (8.43) | 147 (9.61) | 129 (8.43) | |
| 2 | 1308 (28.50) | 450 (29.41) | 438 (28.63) | 420 (27.45) | |
| 3 | 1903 (41.46) | 632 (41.31) | 614 (40.13) | 657 (42.94) | |
| 4 | 953 (20.76) | 311 (20.33) | 323 (21.11) | 319 (20.85) | |
| Laboratory tests | |||||
| Rdw | 15.2 (11.1–33.1) | 14.8 (11.6–31.1) | 15.1 (11.1–31.6) | 15.6 (12–33.1.1) | < 0.001 |
| Red blood cells | 3.54 (1.08–6.84) | 3.61 (1.54–6.26) | 3.57 (1.45–6.84) | 3.425 (1.08–6.36) | < 0.001 |
| White blood cells | 11.1 (0.1–216.4.1.4) | 11.4 (0.3–63.2) | 10.9 (0.2–168.6.2.6) | 10.9 (0.1–216.4.1.4) | 0.239 |
| Albumin | 3.1 (1–5.5.5) | 3.1 (1–5) | 3.1 (1–5.3.3) | 3 (1.3–5.5) | < 0.001 |
| Anion gap | 14 (0–42) | 13 (0–31) | 14 (5–42) | 15 (3–38) | < 0.001 |
| Calcium total | 8.4 (0–15.4.4) | 8.5 (0–12) | 8.4 (2.4–13.9) | 8.3 (4.4–15.4) | < 0.001 |
| Chloride | 103 (67–137) | 102 (67–137) | 103 (71–132) | 103 (79–134) | < 0.001 |
| Glucose | 129 (35–2048) | 124 (35–2048) | 131 (38–993) | 133 (38–905) | < 0.001 |
| Potassium | 4.2 (1.6–10) | 4.1 (1.6–10) | 4.15 (2.2–9.7) | 4.4 (1.7–8.7) | < 0.001 |
| Sodium | 139 (106–167) | 139 (106–167) | 139 (109–161) | 138.5 (111–167) | 0.001 |
| Calculated total CO2 | 27 (0–61) | 28 (10–61) | 26.5 (4–57) | 25 (0–61) | < 0.001 |
| PCO2 | 45 (13–151) | 46 (20–151) | 45 (13–143) | 44.5 (16–125) | < 0.001 |
| PH | 7.36 (6.56–7.68) | 7.37 (6.69–7.66) | 7.37 (6.56–7.68) | 7.34 (6.84–7.68) | < 0.001 |
| PO2 | 87 (12–649) | 89 (19–567) | 89.5 (12–649) | 83.5 (17–560) | 0.011 |
| Alt | 22 (0–4286) | 19 (1–2976) | 22.5 (2–2390) | 26 (0–4286) | < 0.001 |
| Ast | 29 (4–6414) | 23 (4–4746) | 30 (4–4746) | 41.5 (4–6414) | < 0.001 |
| Bilirubin total | 0.5 (0.1–53.6) | 0.4 (0.1–24) | 0.6 (0.1–24.5) | 0.6 (0.1–53.6) | < 0.001 |
| Urea nitrogen | 23 (0–171) | 17 (0–95) | 23 (3–144) | 35 (2–171) | < 0.001 |
| Lactate | 1.5 (0.4–15.3) | 1.3 (0.4–14.4) | 1.5 (0.4–12.2) | 1.6 (0.4–15.3) | < 0.001 |
| Platelet count | 212 (21–1647) | 277 (63–1647) | 203 (48–1013) | 159 (21–821) | < 0.001 |
| Creatinine | 1 (0.2–13.2) | 0.7 (0.2–3.2) | 1.1 (0.3–4.9) | 1.6 (0.3–13.2) | < 0.001 |
| LDH | 250 (31–4055) | 203 (31–774) | 256 (56–1244) | 322.5 (87–4055) | < 0.001 |
| EASIX | 1.438 (0.074–7.997) | 0.608 (0.074–0.978) | 1.438 (0.978–2.145) | 3.577 (2.145–7.997) | < 0.001 |
| Events | |||||
| Mechanical Ventilation, n(%) | 3577 (77.93) | 1150 (75.16) | 1188 (77.65) | 1239 (80.98) | 0.001 |
| Mechanical Ventilation Hour, h | 50 (0.117–1957.9) | 43.76 (0.283–1957.9) | 48.79 (0.117–1153.55.117.55) | 57 (0.15–1211.58.15.58) | 0.082 |
| Intubation, n(%) | 637 (13.88) | 186 (12.16) | 199 (13.01) | 252 (16.47) | 0.001 |
| LOS hospital, days | 10.81 (−0.03-181.69) | 10.13 (−0.03-181.69) | 10.24 (0.54–102.8) | 12.055 (0.13–170.31.13.31) | < 0.001 |
| Death within hospital 28days, (%) | 813 (17.71) | 180 (11.76) | 258 (16.86) | 375 (24.51) | < 0.001 |
| LOS ICU, days | 3.53 (1–136.03.03) | 3.16 (1–136.03.03) | 3.42 (1–56) | 3.89 (1–59.04.04) | 0.025 |
| Death within ICU 28days, (%) | 832 (18.13) | 184 (12.03) | 263 (17.19) | 385 (25.16) | < 0.001 |
Abbreviations: BMI Body Mass Index, HR Heart Rate, Nbps Noninvasive Systolic Blood Pressure, Nbpd Noninvasive Diastolic Blood Pressure, Nbpm Noninvasive Mean Blood Pressure, RR Respiratory Rate, SpO2 Peripheral Oxygen Saturation, PCO2 Partial Pressure of Carbon Dioxide, PO2 Partial Pressure of Oxygen, Alt Alanine Aminotransferase, Ast Aspartate Aminotransferase, LDH Lactate Dehydrogenase, EASIX Endothelial Activation and Stress Index, LOS length of stay, ICU Intensive care unit
EASIX index: T1: 0.0738–0.9778, T2: 0.9784–2.1446, T3: 2.1452–7.9973
Association between EASIX and clinical outcomes
Kaplan–Meier survival analysis demonstrated that patients in the highest EASIX tertile had significantly increased risks of 28-day ICU mortality, in-hospital mortality, and mechanical ventilation compared with those in lower tertiles (p < 0.001; Fig. 2), supporting a robust association between elevated EASIX levels and adverse outcomes in critically ill COPD patients.
Fig. 2.
Kaplan–Meier curves showing cumulative probability of all-cause mortality in ICU (a), all-cause mortality in hospital (b) and mechanical ventilation (c) according to EASIX index at 28 days
Cox proportional hazards regression analyses confirmed the independent association between EASIX and 28-day all-cause ICU mortality. When analyzed as a continuous variable, higher EASIX values were consistently associated with increased mortality risk across unadjusted (HR: 1.21, 95% CI: 1.171–1.251, p < 0.001), partially adjusted (HR: 1.222, 95% CI: 1.182–1.263, p < 0.001), and fully adjusted models (HR: 1.082, 95% CI: 1.037–1.129, p < 0.001). When categorized by tertiles, patients in T3 had a significantly higher mortality risk than those in T1 across all models (fully adjusted HR: 1.405, 95% CI: 1.138–1.735, p = 0.002). EASIX also remained a significant independent predictor of 28-day in-hospital mortality and the requirement for mechanical ventilation in multivariable Cox analyses (Table 2). Restricted cubic spline analysis based on the fully adjusted model further demonstrated a linear dose-response relationship between EASIX and all three clinical endpoints (P for nonlinearity = 0.114, 0.116 and 0.52, respectively; Fig. 3).
Table 2.
Cox proportional hazard ratios (HR) for all-cause mortality and mechanical ventilation
| Categories | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | p-value | HR (95% CI) | p-value | HR (95% CI) | p-value | |
| ICU mortality | ||||||
| Continuous variable per unit | 1.21 (1.171–1.251) | 0.000 | 1.222 (1.182–1.263) | 0.000 | 1.082 (1.037–1.129) | 0.000 |
| Tertile | ||||||
| T1(n = 1491) | Ref | Ref | Ref | |||
| T2(n = 1490) | 1.482 (1.228–1.789) | 0.000 | 1.441 (1.191–1.743) | 0.000 | 1.244 (1.021–1.514) | 0.03 |
| T3(n = 1491) | 2.277 (1.91–2.714) | 0.000 | 2.333 (1.95–2.792) | 0.000 | 1.405 (1.138–1.735) | 0.002 |
| Hospital mortality | ||||||
| Continuous variable per unit | 1.207 (1.167–1.247) | 0.000 | 1.218(1.178–1.26) | 0.000 | 1.077 (1.032–1.124) | 0.001 |
| Tertile | ||||||
| T1(n = 1491) | Ref | Ref | Ref | |||
| T2(n = 1490) | 1.488 (1.23–1.8) | 0.002 | 1.444 (1.192–1.751) | 0.000 | 1.245 (1.02–1.519) | 0.031 |
| T3(n = 1491) | 2.257 (1.889–2.696) | 0.000 | 2.312 (1.929–2.772) | 0.000 | 1.384 (1.118–1.713) | 0.003 |
| Mechanical Ventilation | ||||||
| Continuous variable per unit | 1.037 (1.018–1.056) | 0.000 | 1.033 (1.014–1.053) | 0.001 | 1.03 (1.006–1.054) | 0.013 |
| Tertile | ||||||
| T1(n = 1491) | Ref | Ref | Ref | |||
| T2(n = 1490) | 1.114 (1.027–1.208) | 0.009 | 1.102 (1.015–1.197) | 0.021 | 1.068 (0.98–1.163) | 0.132 |
| T3(n = 1491) | 1.182 (1.091–1.281) | 0.000 | 1.162 (1.071–1.262) | 0.000 | 1.11 (1.007–1.223) | 0.037 |
EASIX index: T1: 0.0738–0.9778, T2: 0.9784–2.1446, T3: 2.1452–7.9973
Model 1: unadjusted
Model 2: adjusted for age, sex and BMI
Model 3: adjusted for age, sex, BMI, hypertension, acute kidney injury, pneumonia, chronic kidney disease, heart failure, cancer, OASIS scale, SOFA scale, APACHEII scale, rdw, red blood cells, white blood cells, albumin, anion gap, calcium total, glucose, potassium., calculated total CO2, ph, PO2, RR and SpO2
Fig. 3.
Restricted cubic spline curve for the EASIX index hazard ratio. Heavy central lines represent the estimated adjusted hazard ratios, with shaded ribbons denoting 95% confdence intervals. a Restricted cubic spline for icu mortality. b Restricted cubic spline for hospital mortality. c Restricted cubic spline for mechanical ventilation. HR, hazard ratio; CI, confdence interval; ICU, intensive care unit
Subgroup analysis
Subgroup analyses were conducted to examine whether the association between EASIX and prognosis varied according to demographic factors (sex, age, BMI) or comorbidities, including HTN, LC, HEP, T2DM, T1DM, HF, MI, CA, CKD, AKI, PNA, and HLD (Fig. 4). The associations between elevated EASIX and poor outcomes remained consistent across all subgroups, with no significant interactions observed.
Fig. 4.
Forest plots of hazard ratios for the ICU mortality (a), hospital mortality (b) and mechanical ventilation rate (c) in different subgroups. HR, hazard ratio; CI, confidence interval; BMI, body mass index
Feature importance based on the Boruta algorithm
Feature selection using the Boruta algorithm identified EASIX as one of the most important predictors of adverse outcomes. The algorithm classified variables as “confirmed,” “tentative,” or “rejected” based on their relative predictive importance. Across all endpoints, EASIX was consistently categorized as a confirmed feature (Fig. 5).
Fig. 5.
Importance of potential risk factors of 28-day ICU mortality (a), hospital mortality (b) and mechanical ventilation rate (c) ranked by Boruta algorithm. The horizontal axis is the name of each variable, and the vertical axis is the Z value of each variable. The box plot shows the importance value of each variable during model calculation. The red boxes represent confirmed variables, the yellow boxes represent rejected variables, and the pink boxes represent tentative variables
Establishment and validation of the machine learning prediction model
According to the predefined protocol, the cohort was randomly divided into a development set (n = 3,213) and a validation set (n = 1,377). Six machine learning algorithms—AdaBoost, Gradient Boosting, Random Forest (RF), Support Vector Machine (SVM), XGBoost, and Decision Tree—were trained to predict 28-day ICU mortality, in-hospital mortality, and mechanical ventilation. Model performance was evaluated using ROC curves and AUC values (Fig. 6). Among these models, the XGBoost algorithm demonstrated the highest predictive accuracy for all outcomes, including 28-day ICU mortality (AUC = 0.823), in-hospital mortality (AUC = 0.791), and mechanical ventilation (AUC = 0.776). Table S3 presents a set of detailed performance metrics for six prediction models of the primary clinical outcomes. The 10-fold cross-validation results of the optimal machine learning algorithm model in this study are provided in the Supplementary Materials (FigureS1). The clinical net benefit evaluation of the EASIX-based machine learning prediction model for 28-day ICU mortality is presented in Fig. 7, while the secondary study outcomes can be found in the Supplementary Materials (FigureS2). According to the DCA curves (Fig. 7), the XGBoost model exhibited greater net benefit along with the threshold probability compared with other models, indicating that the XGBoost model was the optimal model with favorable clinical utility. For comparison with SOFA in predicting 28-day ICU mortality, our model yielded an NRI of 0.6675 and an IDI of 0.111; compared with SAPS II, the corresponding values were an NRI of 0.5736 and an IDI of 0.0849; compared with APACHE II, the corresponding values were an NRI of 0.7706 and an IDI of 0.162. Detailed results are provided in the Supplementary Materials (FigureS3).
Fig. 6.
Validate the efficacy of machine learning prediction models for 28-day ICU mortality (a), hospital mortality (b) and mechanical ventilation rate (c) through ROC curve. AdaBoostTEST: Adaptive Boosting, GBDTTEST: Gradient Boosting Decision Tree, RFTEST: Random Forest Machine, SVMTEST: Support Vector Machine, XGBTEST: eXtreme Gradient Boosting Machine
Fig. 7.
Evaluate the clinical net benefit of an EASIX-based machine learning prediction model for 28-day ICU mortality. a Decision Curve Analysis (DCA) of the EASIX-based prediction model. b Calibration plot of the EASIX-based prediction model
Discussion
This study evaluated the association between the EASIX and short-term clinical outcomes in critically ill patients with COPD using data from the MIMIC-III and MIMIC-IV databases (n = 4,590). Elevated EASIX values were significantly associated with increased 28-day ICU mortality in both unadjusted (HR: 1.21, 95% CI: 1.171–1.251, p < 0.001) and fully adjusted Cox regression models (HR: 1.082, 95% CI: 1.037–1.129, p < 0.001). Kaplan–Meier survival analysis further demonstrated that patients in the highest EASIX tertile exhibited higher risks of 28-day ICU and in-hospital mortality, as well as an increased requirement for mechanical ventilation. These associations remained consistent across subgroups and were corroborated by Boruta-based feature selection, which identified EASIX as a key predictor of adverse outcomes. Among the six machine learning algorithms evaluated, the XGBoost model achieved the best predictive performance (AUC = 0.823), underscoring its potential utility in clinical prognostication.
EASIX is a composite marker derived from routine laboratory values (LDH, Scr, PLT) that reflects endothelial dysfunction and microvascular injury [15, 25]. Originally developed to predict complications such as graft‑versus‑host disease after stem cell transplantation, it has since shown prognostic value in various critical illnesses [16–19, 26–28]. In respiratory diseases, elevated EASIX has been linked to higher mortality and increased risk of mechanical ventilation in COVID‑19, acute respiratory distress syndrome, and asthma [20–22, 29]. Consistent with these findings, the present study confirms that elevated EASIX is an independent risk factor for ICU and in-hospital mortality, as well as mechanical ventilation requirement, in critically ill patients with COPD.
Endothelial dysfunction plays a central role in the pathogenesis of COPD, as supported by structural, functional, and imaging evidence [30, 31]. Studies show that pulmonary capillary density is reduced in COPD patients, and endothelial-dependent vasodilation—mediated by nitric oxide—is impaired early in the disease, correlating with the severity of airflow limitation [32, 33]. Ultrasonography and imaging further link endothelial dysfunction to decreased exercise capacity, lower FEV₁, heightened emphysema severity, and loss of the peripheral microvascular network [34–37]. These findings collectively highlight endothelial dysfunction as an early and progressive contributor to impaired gas exchange, alveolar destruction, and clinical decline in COPD [38]. The current findings, demonstrating EASIX as a surrogate marker of endothelial stress and a robust predictor of short-term mortality, provide further support for the critical role of endothelial dysfunction in the clinical trajectory of COPD.
In the early stages of COPD, smoking serves as a key initiating factor in pulmonary vascular injury [38, 39]. Components of cigarette smoke directly impair endothelial cell structure and function, resulting in microvascular rarefaction, intimal thickening, and smooth muscle hyperplasia in small pulmonary arteries [30, 40]. These alterations contribute to diminished endothelium-dependent vasodilation and infiltration of inflammatory cells. A hallmark of this process is reduced nitric oxide (NO) bioavailability, primarily due to suppressed endothelial nitric oxide synthase (eNOS) activity, oxidative stress, and systemic inflammation [13, 25, 41]. In advanced stages, chronic hypoxia drives endothelial dysfunction through HIF signaling, promoting vascular remodeling [42–44]. Concurrent nuclear factor kappa-B (NF-κB) activation shifts endothelium toward a pro-inflammatory state, sustaining inflammation during exacerbations [45]. Endothelial apoptosis, driven by oxidative stress and associated with impaired cystic fibrosis transmembrane conductance regulator (CFTR) signaling and α1-antitrypsin deficiency, further contributes to emphysematous change. Vascular endothelial growth factor (VEGF), a critical regulator of angiogenesis, exhibits stage-dependent alterations throughout disease progression: early upregulation promotes vascular remodeling, whereas decreased expression of specific VEGF isoforms in later stages fosters endothelial apoptosis [46–48]. Combined with CFTR dysfunction and reduced α1-antitrypsin activity, these changes exacerbate microvascular loss and alveolar destruction [49, 50]. Additionally, endothelial senescence diminishes reparative capacity and disrupts barrier integrity, intensifying pulmonary inflammation and vascular instability [51]. Collectively, these mechanisms illustrate how endothelial dysfunction—mediated by complex molecular interactions—drives COPD progression from early endothelial injury to advanced structural remodeling and chronic inflammation.
Given that adverse clinical outcomes are influenced by multiple interrelated factors that cannot be fully captured through conventional statistical approaches, this study applied the Boruta algorithm to evaluate variable importance comprehensively. EASIX emerged as a significant predictor, achieving a high Z-score and being classified as a “confirmed” variable in the Boruta plot. This finding reinforces its independent association with mortality risk in critically ill COPD patients and provides a strong rationale for its inclusion in prognostic modeling. Subsequently, prediction models were developed using six widely applied machine learning algorithms—Decision Tree, RF, SVM, AdaBoost, GBD, and XGB. All models demonstrated satisfactory predictive performance, with XGBoost achieving the highest discriminative ability (AUC = 0.823). Besides, according to the DCA, the XGBoost model demonstrated greater net benefit across threshold probabilities compared to other models. Although this XGBoost model is still in its early research phase, this finding enhances confidence in its potential for further clinical application. Furthermore, several variables identified as “tentative” predictors in the Boruta analysis may also contribute to adverse outcomes. Future studies should investigate their potential interactions with EASIX to achieve a more comprehensive understanding of prognostic determinants in this patient population. In this study, traditional severity scoring systems such as SOFA, SAPS II, and APACHE II performed less effectively than the machine learning models, indicating that they may not adequately predict short-term mortality in critically ill COPD patients. Although these traditional scores can be used to assess the risk of adverse outcomes in critically ill patients, their assessments rely heavily on clinician experience. Furthermore, because they exclude analysis of a substantial number of valuable variables, the predictive performance of these scoring systems is generally inferior to that of multivariable models.
This study has several methodological strengths. It employed a comprehensive analytical framework integrating Kaplan–Meier survival analysis, Cox proportional hazards regression, restricted cubic spline modeling, subgroup analyses, the Boruta feature selection algorithm, multiple machine learning approaches, and ROC curve evaluation to systematically assess the prognostic value of EASIX in critically ill COPD patients. The combined use of these complementary methods enhances the robustness of the findings and offers multidimensional insights into the clinical utility of EASIX for mortality prediction. Nevertheless, several limitations inherent to the retrospective design should be acknowledged. Despite rigorous multivariable adjustment, the potential for selection bias and residual confounding cannot be fully excluded. Besides, the conclusions of this study are based solely on validation within an internal cohort, which limits the generalizability of the results. Therefore, the performance of this model still requires future validation in multicenter clinical studies. Additionally, the MIMIC database does not contain detailed data on COPD severity stratification (such as GOLD stages or pulmonary function results), disease phenotypes, or acute exacerbation history. This lack of information may confound the correlation between EASIX and clinical outcomes and limits the potential for deeper subgroup analyses. It should be noted that the findings are derived from a cohort primarily consisting of European and American individuals in the MIMIC database, which may not be fully representative of other ethnic groups, particularly Asian populations. Variations in genetics, environment, and healthcare systems could also limit the applicability of the results. Thus, further validation through large-scale, prospective studies is needed to confirm their generalizability.
Conclusions
This study identifies EASIX as an independent prognostic indicator of short-term mortality in critically ill COPD patients. The EASIX-based machine learning model demonstrated strong predictive performance, supporting its potential application in early risk stratification and individualized clinical decision-making. However, its clinical utility should be further confirmed through multicenter, prospective validation studies.
Supplementary Information
Acknowledgements
We thank Phoebe Chi, MD, from Liwen Bianji (Edanz) (www.liwenbianji.cn), for editing a draft of this manuscript.
Authors’ contributions
Jianyi Niu, Qiaoyun Huang, and Yanqi Dong participated in the idea and design of the study. They also helped to draft the manuscript that was submitted and critically review it for significant intellectual content. Jianyi Niu and Qiaoyun Huang handled the data management and statistical analysis. All authors were involved in the revisions to the paper, overall responsibility for the work, and approval of the manuscript’s final draft.
Funding
This study is supported by the National Natural Science Foundation of China (grant no. 82200046), Basic Research Project of Guangzhou Municipal Science and Technology Bureau (grant no. 2024A03J1202), Guangdong Basic and Applied Basic Research Foundation (grant no. 2022A1515110233), Tertiary Education Scientific research project of Guangzhou Municipal Education Bureau (grant no. 202235392), Guangzhou Health Science, Technology Project of Guangzhou Municipal Health Commission (grant no. 20231A011079), Open Project of State Key Laboratory of Respiratory Disease (grant no. SKLRD-OP-202312) and Basic and Applied Basic Research Project of Guangzhou Municipal Science and Technology Bureau (grant no. 2023A04J0571).
Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
The studies involving humans were approved by the Institutional Review Boards of Beth Israel Deaconess Medical Center and the Massachusetts Institute of Technology. The studies were conducted in accordance with local legislation and institutional requirements. Since the database contains no protected health information, the requirement for informed consent was waived. Accordingly, no additional ethical approval was required for this retrospective analysis. All procedures were conducted in accordance with relevant guidelines and regulations, including the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Jianyi Niu, Qiaoyun Huang and Yanqi Dong contributed equally to this work.
Contributor Information
Rongchang Chen, Email: chenrcstatekeylab@gmail.com.
Lili Guan, Email: dr_nickguan@163.com.
References
- 1.Celli BR, Drazen JM, Wedzicha JA. Update on Clinical Aspects of Chronic Obstructive Pulmonary Disease. N Engl J Med. 2019;381(13):1257–66. [DOI] [PubMed] [Google Scholar]
- 2.Ritchie AI, Wedzicha JA. Definition, Causes, Pathogenesis, and Consequences of Chronic Obstructive Pulmonary Disease Exacerbations. Clin Chest Med. 2020;41(3):421–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Suissa S, Dell’Aniello S, Ernst P. Long-term natural history of chronic obstructive pulmonary disease: severe exacerbations and mortality. Thorax. 2012;67(11):957–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Prediletto I, Giancotti G, Nava S. COPD Exacerbation: why it is important to avoid ICU admission. J Clin Med. 2023;12(10):3369. [DOI] [PMC free article] [PubMed]
- 5.Ko FW, Chan KP, Hui DS, Goddard JR, Shaw JG, Reid DW, Yang IA. Acute exacerbation of COPD. Respirology. 2016;21(7):1152–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Whittaker Brown S-A, Braman S. Recent Advances in the Management of Acute Exacerbations of Chronic Obstructive Pulmonary Disease. Med Clin North Am. 2020;104(4):615–30. [DOI] [PubMed] [Google Scholar]
- 7.Afessa B, Morales IJ, Scanlon PD, Peters SG. Prognostic factors, clinical course, and hospital outcome of patients with chronic obstructive pulmonary disease admitted to an intensive care unit for acute respiratory failure. Crit Care Med. 2002;30(7):1610–5. [DOI] [PubMed] [Google Scholar]
- 8.Gayaf M, Karadeniz G, Güldaval F, Polat G, Türk M. Which one is superior in predicting 30 and 90 days mortality after COPD exacerbation: DECAF, CURB-65, PSI, BAP-65, PLR, NLR. Expert Rev Respir Med. 2021;15(6):845–51. [DOI] [PubMed] [Google Scholar]
- 9.Anupam Goel RG, Pinckney, Littenberg B. APACHE II predicts long-term survival in COPD patients admitted to a general medical ward. J Gen Intern Med. 2003;18(10):824–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cai C, Zeng W, Wang H, Ren S. Neutrophil-to-Lymphocyte Ratio (NLR), Platelet-to-Lymphocyte Ratio (PLR) and Monocyte-to-Lymphocyte Ratio (MLR) as Biomarkers in Diagnosis Evaluation of Acute Exacerbation of Chronic Obstructive Pulmonary Disease: A Retrospective, Observational Study. Int J Chronic Obstr Pulm Dis. 2024;19:933–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Zinellu A, Zinellu E, Mangoni AA, Pau MC, Carru C, Pirina P, et al. Clinical significance of the neutrophil-to-lymphocyte ratio and platelet-to-lymphocyte ratio in acute exacerbations of COPD: present and future. Eur Respiratory Rev. 2022;31(166):220095. [DOI] [PMC free article] [PubMed]
- 12.Screm G, Mondini L, Salton F, Confalonieri P, Trotta L, Barbieri M, et al. Vascular endothelial damage in COPD: where are we now. Where will we go? Diagnostics. 2024;14(9):950. [DOI] [PMC free article] [PubMed]
- 13.Green CE, Turner AM. The role of the endothelium in asthma and chronic obstructive pulmonary disease (COPD). Respir Res. 2017;18(1):20. [DOI] [PMC free article] [PubMed]
- 14.Szucs B, Szucs C, Petrekanits M, Varga JT. Molecular characteristics and treatment of endothelial dysfunction in patients with COPD: a review article. Int J Mol Sci. 2019;20(18):4329. [DOI] [PMC free article] [PubMed]
- 15.Luft T, Benner A, Jodele S, Dandoy CE, Storb R, Gooley T, Sandmaier BM, Becker N, Radujkovic A, Dreger P, et al. EASIX in patients with acute graft-versus-host disease: a retrospective cohort analysis. Lancet Haematol. 2017;4(9):e414–23. [DOI] [PubMed] [Google Scholar]
- 16.Penack O, Luft T, Peczynski C, Benner A, Sica S, Arat M, et al. Endothelial Activation and Stress Index (EASIX) to predict mortality after allogeneic stem cell transplantation: a prospective study. J Immunother Cancer. 2024;12(1):e007635. [DOI] [PMC free article] [PubMed]
- 17.Wang J, Chen X, Qin C, Zeng X, Du X, Wang D. The endothelial activation and stress index is a potential prognostic indicator for patients with acute pancreatitis managed in the intensive care unit: a retrospective study. Front Med. 2024;11:1498148. [DOI] [PMC free article] [PubMed]
- 18.Deng F, Lin J, Lin H. Association between EASIX and acute kidney injury in critically ill cancer patients. BMC Nephrol. 2024;25(1):453. [DOI] [PMC free article] [PubMed]
- 19.Finke D, Hund H, Frey N, Luft T, Lehmann LH. EASIX (endothelial activation and stress index) predicts mortality in patients with coronary artery disease. Clin Res Cardiol. 2024;114(8):1008–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.He Y, Li Y, Xiaojin Y, Wu D, Jiang W, Xie X. Endothelial activation and stress index for prediction of mortality in asthma. Front Med. 2025;12:1622944. [DOI] [PMC free article] [PubMed]
- 21.Chen J, Lv J, Liu M, Dai X, Deng W. Between endothelial activation and stress index (EASIX) and mortality in acute respiratory distress syndrome (ARDS) patients: a multicenter retrospective study. Front Physiol. 2025;16:1570988. [DOI] [PMC free article] [PubMed]
- 22.Go S-I, Park S, Kang MH, Kim H-G, Kang JH, Kim JH, Lee G-W. Endothelial activation and stress index (EASIX) as a predictive biomarker in small cell lung cancer. Cancer Biomarkers. 2022;35(2):217–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Chen Y, Liu A, Zhang L, Wan X, Jiao J, Sun Z. Association between stress hyperglycemia ratio and all-cause mortality in patients with chronic obstructive pulmonary disease. Sci Rep. 2025;15(1):32892. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Yue S, Li S, Huang X, Liu J, Hou X, Zhao Y, et al. Machine learning for the prediction of acute kidney injury in patients with sepsis. J Translational Med. 2022;20(1):215. [DOI] [PMC free article] [PubMed]
- 25.Flammer AJ, Anderson T, Celermajer DS, Creager MA, Deanfield J, Ganz P, Hamburg NM, Lüscher TF, Shechter M, Taddei S, et al. The Assessment of Endothelial Function. Circulation. 2012;126(6):753–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kalicińska E, Biernat M, Rybka J, Zińczuk A, Janocha-Litwin J, Rosiek-Biegus M, et al. Endothelial Activation and Stress Index (EASIX) as an early predictor for mortality and overall survival in hematological and non-hematological patients with COVID-19: multicenter cohort study. J Clin Med. 2021;10(19):4373. [DOI] [PMC free article] [PubMed]
- 27.Escribano-Serrat S, Rodríguez-Lobato LG, Charry P, Martínez-Cibrian N, Suárez-Lledó M, Rivero A, Moreno-Castaño AB, Solano MT, Arcarons J, Nomdedeu M, et al. Endothelial Activation and Stress Index in adults undergoing allogeneic hematopoietic cell transplantation with post-transplant cyclophosphamide-based prophylaxis. Cytotherapy. 2024;26(1):73–80. [DOI] [PubMed] [Google Scholar]
- 28.Estler B, Fröhlich H, Täger T, Hund H, Frey N, Luft T, et al. Endothelial dysfunction assessed by the endothelial activation and stress index (EASIX) predicts risk of mortality in chronic heart failure patients. Int J Cardiol. 2025;438:133566. [DOI] [PubMed]
- 29.Jeong JH, Heo M, Park S, Lee SH, Park O, Kim T, Yeo HJ, Jang JH, Cho WH, Yoo J-W. Association between Age-Adjusted Endothelial Activation and Stress Index and Intensive Care Unit Mortality in Patients with Severe COVID-19. Tuberc Respir Dis. 2024;87(4):524–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Peinado VI, Pizarro S, Barberà JA. Pulmonary Vascular Involvement in COPD. Chest. 2008;134(4):808–14. [DOI] [PubMed] [Google Scholar]
- 31.Voelkel NF, Gomez-Arroyo J, Mizuno S. COPD/Emphysema: The Vascular Story. Pulmonary Circulation. 2011;1(3):320–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Wiebe BM, Laursen H. Lung morphometry by unbiased methods in emphysema: bronchial and blood vessel volume, alveolar surface area and capillary length. APMIS. 1998;106(6):651–6. [DOI] [PubMed] [Google Scholar]
- 33.Estépar RSJ, Kinney GL, Black-Shinn JL, Bowler RP, Kindlmann GL, Ross JC, Kikinis R, Han MK, Come CE, Diaz AA, et al. Computed Tomographic Measures of Pulmonary Vascular Morphology in Smokers and Their Clinical Implications. Am J Respir Crit Care Med. 2013;188(2):231–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Minet C, Vivodtzev I, Tamisier R, Arbib F, Wuyam B, Timsit J-F, Monneret D, Borel J-C, Baguet J-P, Lévy P, et al. Reduced six-minute walking distance, high fat-free-mass index and hypercapnia are associated with endothelial dysfunction in COPD. Respir Physiol Neurobiol. 2012;183(2):128–34. [DOI] [PubMed] [Google Scholar]
- 35.Vukic Dugac A, Ruzic A, Samarzija M, Badovinac S, Kehler T, Jakopovic M. Persistent endothelial dysfunction turns the frequent exacerbator COPD from respiratory disorder into a progressive pulmonary and systemic vascular disease. Med Hypotheses. 2015;84(2):155–8. [DOI] [PubMed] [Google Scholar]
- 36.Iyer KS, Newell JD, Jin D, Fuld MK, Saha PK, Hansdottir S, Hoffman EA. Quantitative Dual-Energy Computed Tomography Supports a Vascular Etiology of Smoking-induced Inflammatory Lung Disease. Am J Respir Crit Care Med. 2016;193(6):652–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Washko GR. The Role and Potential of Imaging in COPD. Med Clin North Am. 2012;96(4):729–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Peinado VI, Barbera JA, Ramirez J, Gomez FP, Roca J, Gimferrer LJJM, Rodriguez-Roisin R. Endothelial dysfunction in pulmonary arteries of patients with mild COPD. Am J Physiol. 1998;274(6):L908–913. [DOI] [PubMed] [Google Scholar]
- 39.Vignola AM, Paganin F, Capieu L, Scichilone N, Bellia M, Maakel L, Bellia V, Godard P, Bousquet J, Chanez P. Airway remodelling assessed by sputum and high-resolution computed tomography in asthma and COPD. Eur Respir J. 2004;24(6):910–7. [DOI] [PubMed] [Google Scholar]
- 40.Borgas D, Chambers E, Newton J, Ko J, Rivera S, Rounds S, Lu Q. Cigarette Smoke Disrupted Lung Endothelial Barrier Integrity and Increased Susceptibility to Acute Lung Injury via Histone Deacetylase 6. Am J Respir Cell Mol Biol. 2016;54(5):683–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Endemann DH, Schiffrin EL. Endothelial Dysfunction. J Am Soc Nephrol. 2004;15(8):1983–92. [DOI] [PubMed] [Google Scholar]
- 42.Semenza GL. Hypoxia-Inducible Factor 1 and Cardiovascular Disease. Annu Rev Physiol. 2014;76(1):39–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Sheikh AQ, Saddouk FZ, Ntokou A, Mazurek R, Greif DM. Cell Autonomous and Non-cell Autonomous Regulation of SMC Progenitors in Pulmonary Hypertension. Cell Rep. 2018;23(4):1152–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Tan Q, Kerestes H, Percy MJ, Pietrofesa R, Chen L, Khurana TS, Christofidou-Solomidou M, Lappin TRJ, Lee FS. Erythrocytosis and Pulmonary Hypertension in a Mouse Model of Human HIF2A Gain of Function Mutation. J Biol Chem. 2013;288(24):17134–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Tran TTV, Jeong Y, Kim S, Yeom JE, Lee J, Lee W, et al. PRMT1 ablation in endothelial cells causes endothelial dysfunction and aggravates COPD attributable to dysregulated NF-κB signaling. Advanced Sci. 2025;12(19):e2411514. [DOI] [PMC free article] [PubMed]
- 46.Santos S, Peinado VI, Ramírez J, Morales-Blanhir J, Bastos R, Roca J, Rodriguez-Roisin R, Barberà JA. Enhanced Expression of Vascular Endothelial Growth Factor in Pulmonary Arteries of Smokers and Patients with Moderate Chronic Obstructive Pulmonary Disease. Am J Respir Crit Care Med. 2003;167(9):1250–6. [DOI] [PubMed] [Google Scholar]
- 47.Kanazawa H, Yoshikawa J. Elevated Oxidative Stress and Reciprocal Reduction of Vascular Endothelial Growth Factor Levels With Severity of COPD. Chest. 2005;128(5):3191–7. [DOI] [PubMed] [Google Scholar]
- 48.Kasahara Y, Tuder RM, Cool CD, Lynch DA, Flores SC, Voelkel NF. Endothelial cell death and decreased expression of vascular endothelial growth factor and vascular endothelial growth factor receptor 2 in emphysema. Am J Respir Crit Care Med. 2001;163:737–44. [DOI] [PubMed] [Google Scholar]
- 49.Dransfield M, Rowe S, Vogelmeier CF, Wedzicha J, Criner GJ, Han MK, Martinez FJ, Calverley P. Cystic Fibrosis Transmembrane Conductance Regulator: Roles in Chronic Obstructive Pulmonary Disease. Am J Respir Crit Care Med. 2022;205(6):631–40. [DOI] [PubMed] [Google Scholar]
- 50.Ahmad S, Lockett AD, Brown MB, Santos-Falcon N, Rush NI, Oueini H, et al. Active trafficking of Alpha 1 antitrypsin across the lung endothelium. PLoS ONE. 2014;9(4):e93979. [DOI] [PMC free article] [PubMed]
- 51.Paschalaki K, Rossios C, Pericleous C, MacLeod M, Rothery S, Donaldson GC, Wedzicha JA, Gorgoulis V, Randi AM, Barnes PJ. Inhaled corticosteroids reduce senescence in endothelial progenitor cells from patients with COPD. Thorax. 2022;77(6):616–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.