Abstract
Background
The Endothelial Activation and Stress Index (EASIX) serves as a biomarker of endothelial dysfunction. Although EASIX is prognostic in critical illnesses such as sepsis, its specific association with mortality in critically ill patients with diabetes mellitus (DM) is undetermined. This study was designed to evaluate this relationship.
Methods
This retrospective study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. EASIX was calculated as lactate dehydrogenase (U/L) × creatinine (mg/dL) / platelet count (10⁹/L) and analyzed as log2-transformed values and by tertiles. The primary outcome was 30-day mortality, with 365-day mortality as a secondary outcome. Cox proportional hazards regression, Kaplan-Meier analysis, restricted cubic splines, subgroup analyses, and mediation analysis with lactate as a mediator were employed to evaluate the prognostic association. Sensitivity analyses, including propensity score matching, were performed to assess the robustness of the findings.
Results
This study included 4,175 critically ill patients with DM. After full adjustment, each log2‑unit increase in the EASIX score was associated with a significantly increased risk of both 30‑day mortality (HR 1.20, 95% CI 1.16–1.24) and 365‑day mortality (HR 1.14, 95% CI 1.11–1.17). Mortality risk demonstrated a graded increase across ascending EASIX tertiles. A linear dose‑response relationship was confirmed, and this association remained consistent across most predefined subgroups. Propensity score matching analysis further supported this association. Mediation analysis indicated that lactate levels partially mediated the observed relationship between EASIX and mortality.
Conclusions
EASIX is independently associated with increased 30‑day and 365‑day mortality in critically ill patients with DM, an effect partially mediated by lactate.
Clinical trial number
Not applicable.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12902-026-02298-8.
Keywords: Endothelial Activation and Stress Index (EASIX), Diabetes Mellitus (DM), Mortality, Biomarker, Intensive Care Unit, MIMIC-IV
Introduction
Diabetes mellitus (DM) represents a major global health challenge, affecting over half a billion people worldwide and contributing substantially to global morbidity and mortality [1, 2]. The underlying pathophysiology, characterized by chronic inflammation, immune dysregulation, and microvascular impairment, not only increases susceptibility to severe infections but also accelerates the progression of cardiovascular, renal, and cerebrovascular diseases in hospitalized patients [3–8]. In critically ill patients, diabetes has been linked to a higher incidence of acute organ failure and subsequent mortality [9]. Some studies have reported higher mortality rates in critically ill patients with diabetes compared to those without diabetes [10]. These considerations underscore the clinical importance of identifying practical biomarkers for prognosis and risk stratification in this patient population.
The Endothelial Activation and Stress Index (EASIX), calculated from lactate dehydrogenase (LDH), creatinine, and platelet count, represents a practical biomarker derived from routine laboratory parameters [11]. Initially established in the setting of allogeneic stem cell transplantation, EASIX demonstrated strong predictive value for outcomes in patients with steroid-refractory acute graft-versus-host disease, a condition marked by endothelial injury and thrombotic microangiopathy [12]. EASIX has been reported to correlate with established biomarkers of endothelial injury after transplantation, including soluble urokinase-type plasminogen activator receptor, growth differentiation factor-15, and soluble C5b-9 [13]. Subsequent studies have demonstrated its prognostic value across multiple critical conditions, including sepsis, acute respiratory distress syndrome, acute kidney injury, and cardiovascular diseases, where it consistently correlates with mortality risk and adverse clinical outcomes [14–18]. However, its prognostic significance specifically in critically ill patients with DM remains to be determined.
This study aimed to evaluate the association between EASIX and mortality in critically ill patients with DM and to assess its ability to identify high-risk individuals, using data from the MIMIC-IV database.
Materials and methods
Data source
This retrospective cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, v3.1), which contains de-identified clinical records of ICU patients at Beth Israel Deaconess Medical Center (Boston, MA, USA) from 2008 to 2019. Ethical approval with a waiver of informed consent was granted by the Institutional Review Boards of both MIT and Beth Israel Deaconess Medical Center. Author Rong Ding completed the required NIH human research protection course (Certification No. 64760223) for database access. This study adhered to the STROBE guidelines and the Declaration of Helsinki [19].
Study population
This study initially identified patients with DM from the database using relevant International Classification of Diseases (ICD)-9/10 codes. We included only adult patients (age ≥ 18 years) and considered each patient’s first ICU admission. Those with an ICU stay of less than 24 h or missing data for serum creatinine, lactate dehydrogenase, or platelet count from the first ICU day were subsequently excluded. After applying these criteria, the final cohort comprised 4,175 patients eligible for analysis (Fig. 1). This sample, derived from including all eligible patients in the database, provides sufficient statistical power for multivariable adjustments, subgroup analyses, and propensity score matching.
Fig. 1.
Flowchart of the Study Cohort. ICU, Intensive Care Unit; MIMIC-IV, Medical Information Mart for Intensive Care IV
Demographic and laboratory variables
Demographic and laboratory variables were retrieved from the MIMIC-IV database via Structured Query Language. The extracted parameters encompassed: (1) general characteristics, including age, sex, and weight; (2) vital signs, such as heart rate, respiratory rate, and mean arterial pressure; and (3) laboratory measurements, comprising white blood cell count, hemoglobin, glucose, anion gap, bicarbonate, blood urea nitrogen, calcium, chloride, sodium, potassium, international normalized ratio, creatinine, lactate dehydrogenase, platelets, among others. All data were obtained within the first 24 h following ICU admission. We further gathered comorbidities (e.g., hypertension, atrial fibrillation), illness severity scores (OASIS, SAPS II, SOFA), and first-day treatment information, including the use of vasoactive agents, mechanical ventilation, continuous renal replacement therapy (CRRT), and extracorporeal membrane oxygenation (ECMO). All variables had missing data proportions below 5%, with detailed distributions presented in Supplementary Table 1. Missing values were handled through single imputation using Bayesian Ridge regression.
Research variable and outcomes
The primary exposure was EASIX, calculated according to the formula: EASIX = [LDH (U/L) × creatinine (mg/dL) / platelet count (109/L)]. Due to its right-skewed distribution, EASIX values were log2-transformed for analysis. Patients were categorized into three groups according to tertiles of log2(EASIX): T1 (< 0.53, n = 1,392), T2 (0.53–2.05, n = 1,391), and T3 (≥ 2.05, n = 1,392).
The primary outcome was 30-day mortality, and the secondary outcome was 365-day mortality.
Statistical analysis
Continuous variables are presented as mean ± SD or median (IQR), and categorical variables as number (percentage). Group comparisons were performed using ANOVA or Kruskal-Wallis tests for continuous variables, based on distributional normality, and chi-square or Fisher’s exact tests for categorical variables, as appropriate.
Multivariable Cox regression was used to assess the relationship between log2-transformed EASIX and mortality. The final model (Model 3) adjusted for age, sex, weight, heart rate, mean arterial pressure, respiratory rate, glucose, white blood cell count, hemoglobin, calcium, sodium, potassium, hypertension, atrial fibrillation, congestive heart failure, cerebrovascular disease, renal disease, malignant cancer, sepsis, vasoactive agent use, and mechanical ventilation. Log2(EASIX) was analyzed both as a continuous variable and by tertiles, with P for trend calculated across tertiles. We assessed multicollinearity in the model using generalized variance inflation factors (GVIF), and all values were below 2, indicating no substantial multicollinearity (Supplementary Table 2).
Survival outcomes across tertiles of log2(EASIX) were visualized using Kaplan-Meier curves and compared with the log-rank test. The potential nonlinear association between log2(EASIX) as a continuous variable and the risks of 30- and 365-day mortality was examined using restricted cubic splines with three knots placed at the 10th, 50th, and 90th percentiles within Cox models. To evaluate the robustness of the primary findings, subgroup analyses were performed across predefined clinical strata, and interactions were tested using likelihood ratio tests. The results of these subgroup analyses are summarized in forest plots.
We conducted a mediation analysis using a bootstrap approach to decompose the total effect of log2(EASIX) on mortality into the direct effect and the indirect effect operating through lactate. This analysis also quantified the proportion of the total effect mediated by lactate and was performed separately for 30-day and 365-day mortality.
To address potential confounding, we performed 1:1:1 propensity score matching (PSM) with a caliper of 0.2 in the diabetic cohort, with matching variables being all covariates from the multivariable Cox model (model 3). Covariate balance after matching was evaluated using standardized mean differences (SMD). Additionally, to further enhance the robustness of effect estimates, we applied doubly robust estimation by combining propensity score matching with multivariable Cox regression adjustment in the matched cohort [20].
All analyses was performed using R 4.2.2 (http://www.Rproject.org; The R Foundation, Vienna, Austria) and the Free Statistics software (version 2.2; Beijing FreeClinical Medical Technology Co., Ltd, Beijing, China). Statistical significance was indicated by P < 0.05.
Results
Baseline characteristics of study subjects
The baseline characteristics of the study subjects stratified by log2(EASIX) tertiles are presented in Table 1. Higher tertiles showed increased respiratory rate but decreased mean arterial pressure. Laboratory trends included elevated glucose, anion gap, blood urea nitrogen, potassium, international normalized ratio, creatinine, and lactate dehydrogenase, alongside reduced hemoglobin, bicarbonate, calcium, and platelets. Comorbidities including heart failure, renal disease, cancer, sepsis, ischemic cardiomyopathy, and maintenance dialysis were more prevalent in higher tertiles. Disease severity scores (OASIS, SAPS II, SOFA) were progressively higher with increasing EASIX tertiles. Similarly, the utilization of vasopressors, mechanical ventilation, CRRT, and ECMO also increased across these groups. Accordingly, 30-day and 365-day mortality demonstrated a graded increase across tertiles (Supplementary Table 3).
Table 1.
The clinical characteristics of critically ill patients with diabetes mellitus according to log2 (EASIX) levels
| Variables | Log2 (EASIX) | T1 (< 0.53) | T2 (0.53–2.05) | T3 (≥ 2.05) | P value |
|---|---|---|---|---|---|
| Total (n = 4175) | (n = 1392) | (n = 1391) | (n = 1392) | ||
| General characteristics | |||||
| Sex (male), n (%) | 2471 (59.2) | 714 (51.3) | 884 (63.6) | 873 (62.7) | < 0.001 |
| Age, years | 68.4 ± 13.6 | 67.3 ± 14.2 | 70.1 ± 13.0 | 67.7 ± 13.3 | < 0.001 |
| Weight, kg | 88.5 ± 26.0 | 86.8 ± 27.2 | 89.4 ± 25.1 | 89.4 ± 25.7 | 0.009 |
| Vital signs | |||||
| Heart rate, beats/min | 86.1 ± 16.9 | 86.4 ± 16.8 | 84.5 ± 16.1 | 87.4 ± 17.6 | < 0.001 |
| Respiratory rate, beats/min | 20.2 ± 4.1 | 19.8 ± 3.9 | 20.1 ± 3.9 | 20.7 ± 4.3 | < 0.001 |
| MAP, mmHg | 78.3 ± 11.3 | 80.3 ± 11.5 | 77.9 ± 10.7 | 76.8 ± 11.3 | < 0.001 |
| Laboratory parameters | |||||
| WBC, 109/L | 12.8 (9.1, 18.4) | 12.6 (9.3, 17.5) | 12.7 (9.2, 17.9) | 13.2 (8.7, 19.7) | 0.340 |
| Hemoglobin, g/dL | 9.8 ± 2.3 | 10.2 ± 2.3 | 9.9 ± 2.3 | 9.1 ± 2.2 | < 0.001 |
| Glucose, mg/dL | 168.8 (133.2, 213.7) | 162.7 (133.2, 204.6) | 168.8 (134.1, 214.1) | 175.2 (132.6, 222.8) | 0.001 |
| AG, mmol/L | 18.1 ± 5.8 | 16.2 ± 4.7 | 17.4 ± 5.2 | 20.8 ± 6.4 | < 0.001 |
| Bicarbonate, mmol/L | 20.1 ± 5.5 | 21.9 ± 5.2 | 20.4 ± 5.2 | 18.0 ± 5.5 | < 0.001 |
| BUN, mmol/L | 31.0 (20.0, 53.0) | 20.0 (14.0, 28.0) | 32.0 (21.0, 47.0) | 53.0 (35.0, 77.0) | < 0.001 |
| Calcium, mg/dL | 8.1 ± 0.9 | 8.2 ± 0.9 | 8.1 ± 0.9 | 7.9 ± 1.0 | < 0.001 |
| Chloride, mmol/L | 100.2 ± 7.2 | 100.5 ± 6.5 | 100.8 ± 7.3 | 99.2 ± 7.6 | < 0.001 |
| Sodium, mmol/L | 135.8 ± 6.0 | 136.0 ± 5.6 | 136.1 ± 6.3 | 135.3 ± 6.2 | < 0.001 |
| Potassium, mmol/L | 4.8 ± 1.0 | 4.5 ± 0.8 | 4.8 ± 0.9 | 5.1 ± 1.1 | < 0.001 |
| INR | 1.4 (1.2, 1.7) | 1.2 (1.1, 1.5) | 1.4 (1.2, 1.7) | 1.5 (1.2, 2.1) | < 0.001 |
| Creatinine, mg/dL | 1.3 (0.9, 2.3) | 0.9 (0.7, 1.1) | 1.4 (1.0, 1.9) | 2.6 (1.6, 4.4) | < 0.001 |
| LDH, IU/L | 279.0 (206.0, 420.5) | 210.0 (172.0, 267.0) | 285.0 (223.0, 391.0) | 426.5 (284.0, 769.2) | < 0.001 |
| Platelet, 109/L | 204.6 ± 111.6 | 270.1 ± 121.2 | 197.2 ± 82.7 | 146.5 ± 90.1 | < 0.001 |
| Log2 (EASIX) | 1.2 (0.2, 2.6) | -0.2 (-0.8, 0.2) | 1.2 (0.9, 1.6) | 3.2 (2.6, 4.2) | < 0.001 |
| Comorbidities, n (%) | |||||
| Hypertension | 3347 (80.2) | 1073 (77.1) | 1138 (81.8) | 1136 (81.6) | 0.002 |
| Atrial fibrillation | 1424 (34.1) | 427 (30.7) | 512 (36.8) | 485 (34.8) | 0.002 |
| Congestive heart failure | 1745 (41.8) | 418 (30) | 655 (47.1) | 672 (48.3) | < 0.001 |
| Cerebrovascular disease | 702 (16.8) | 275 (19.8) | 242 (17.4) | 185 (13.3) | < 0.001 |
| Renal disease | 1549 (37.1) | 222 (15.9) | 554 (39.8) | 773 (55.5) | < 0.001 |
| Malignant cancer | 609 (14.6) | 187 (13.4) | 192 (13.8) | 230 (16.5) | 0.042 |
| COPD | 381 ( 9.1) | 124 (8.9) | 151 (10.9) | 106 (7.6) | 0.012 |
| Hyperlipidemia | 2021 (48.4) | 657 (47.2) | 742 (53.3) | 622 (44.7) | < 0.001 |
| Ischemic heart disease | 1289 (30.9) | 293 (21.0) | 437 (31.4) | 559 (40.2) | < 0.001 |
| Maintenance Dialysis | 131 ( 3.1) | 3 (0.2) | 13 (0.9) | 115 (8.3) | < 0.001 |
| Sepsis | 2714 (65.0) | 740 (53.2) | 937 (67.4) | 1037 (74.5) | < 0.001 |
| Disease severity scores, scores | |||||
| OASIS | 34.9 ± 9.8 | 32.5 ± 8.9 | 34.7 ± 9.5 | 37.6 ± 10.3 | < 0.001 |
| SAPS II | 41.6 ± 15.2 | 34.9 ± 12.8 | 41.0 ± 13.5 | 48.8 ± 15.7 | < 0.001 |
| SOFA | 5.9 ± 3.7 | 3.6 ± 2.7 | 5.6 ± 3.1 | 8.4 ± 3.7 | < 0.001 |
| Treatments on the first day of ICU admission, n (%) | |||||
| Vasoactive agent | 1344 (32.2) | 331 (23.8) | 450 (32.4) | 563 (40.4) | < 0.001 |
| Mechanical ventilation | 1557 (37.3) | 451 (32.4) | 550 (39.5) | 556 (39.9) | < 0.001 |
| CRRT | 270 (6.5) | 8 (0.6) | 43 (3.1) | 219 (15.7) | < 0.001 |
| ECMO | 9 ( 0.2) | 0 (0.0) | 2 (0.1) | 7 (0.5) | 0.009 |
EASIX, Endothelial Activation and Stress Index; T, tertile; MAP, Mean Arterial Pressure; WBC, White Blood Cell Count; AG, Anion Gap; BUN, Blood Urea Nitrogen; INR, International Normalized Ratio; LDH, lactate dehydrogenase; COPD, Chronic Obstructive Pulmonary Disease; OASIS, Oxford Acute Severity of Illness Score; SAPS II, Simplified Acute Physiology Score II; SOFA, Sequential Organ Failure Assessment; CRRT, Continuous Renal Replacement Therapy; ECMO, Extracorporeal Membrane Oxygenation
Multivariable cox regression
Multivariable Cox proportional hazards regression was used to assess the association between EASIX and mortality in critically ill patients with DM. As a continuous variable, log2(EASIX) was significantly associated with 30-day mortality in both unadjusted (HR 1.23, 95% CI 1.19–1.26; P < 0.001) and fully adjusted models (HR 1.20, 95% CI 1.16–1.24; P < 0.001). When analyzed categorically, the highest log2(EASIX) tertile was associated with a 169% increased risk of 30-day mortality compared to the lowest tertile in unadjusted analyses (HR 2.69, 95% CI 2.29–3.17; P < 0.001); this association persisted in the fully adjusted model (HR 2.23, 95% CI 1.86–2.68; P < 0.001). A similar trend was observed for 365-day mortality (Table 2).
Table 2.
Multivariate cox regression analyses for 30-day and 365-day mortality
| Variable | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| HR (95%CI) | P-value | HR (95%CI) | P-value | HR (95%CI) | P-value | |
| 30-day mortality | ||||||
| Log2 (EASIX) Continuous | 1.23 (1.19 ~ 1.26) | < 0.001 | 1.25 (1.22 ~ 1.29) | < 0.001 | 1.20 (1.16 ~ 1.24) | < 0.001 |
| Log2 (EASIX) Tertiles | ||||||
| T1 | 1(Ref) | 1(Ref) | 1(Ref) | |||
| T2 | 1.53 (1.28 ~ 1.83) | < 0.001 | 1.44 (1.21 ~ 1.73) | < 0.001 | 1.33 (1.10 ~ 1.59) | 0.003 |
| T3 | 2.69 (2.29 ~ 3.17) | < 0.001 | 2.72 (2.30 ~ 3.20) | < 0.001 | 2.23 (1.86 ~ 2.68) | < 0.001 |
| Trend test | < 0.001 | < 0.001 | < 0.001 | |||
| 365-day mortality | ||||||
| Log2 (EASIX) Continuous | 1.18 (1.15 ~ 1.21) | < 0.001 | 1.20 (1.17 ~ 1.23) | < 0.001 | 1.14 (1.11 ~ 1.17) | < 0.001 |
| Log2 (EASIX) Tertiles | ||||||
| T1 | 1(Ref) | 1(Ref) | 1(Ref) | |||
| T2 | 1.41 (1.24 ~ 1.60) | < 0.001 | 1.34 (1.17 ~ 1.52) | < 0.001 | 1.19 (1.04 ~ 1.36) | 0.010 |
| T3 | 2.20 (1.95 ~ 2.48) | < 0.001 | 2.22 (1.97 ~ 2.51) | < 0.001 | 1.76 (1.54 ~ 2.01) | < 0.001 |
| Trend test | < 0.001 | < 0.001 | < 0.001 | |||
Log2 (EASIX): T1 (< 0.53), T2 (0.53–2.05), T3 (≥ 2.05). HR, Hazard Ratio; CI, Confidential Interval
Model 1: No adjusted
Model 2: Adjusted for age, sex, weight
Model 3: Adjusted for age, sex, weight, heart rate, mean arterial pressure, respiratory rate, glucose, white blood cell, hemoglobin, calcium, sodium, potassium, hypertension, atrial fibrillation, congestive heart failure, cerebrovascular disease, renal disease, malignant cancer, sepsis, vasoactive agent, and mechanical ventilation
Kaplan-Meier survival analysis
Kaplan-Meier survival analysis revealed significant differences in both 30-day and 365-day mortality among critically ill patients with DM stratified by log2(EASIX) tertiles (Fig. 2). The log-rank test demonstrated a statistically significant gradient across tertiles (P < 0.0001).
Fig. 2.
Kaplan-Meier survival analysis for 30-day (A) and 365-day (B) mortality stratified by EASIX tertiles: T1 (< 0.53), T2 (0.53–2.05), T3 (≥ 2.05). Each panel includes a main plot and a magnified inset of the survival curves for enhanced visualization
Restricted Cubic Spline Analysis
As shown in Fig. 3, log2(EASIX) demonstrated a significant linear positive association with both 30-day and 365-day mortality (P for nonlinearity = 0.053 and 0.370, respectively) in critically ill patients with DM.
Fig. 3.
Restricted cubic spline curves showing the relationship of log2(EASIX) with 30-day (A) and 365-day (B) mortality. Adjusted for age, sex, weight, heart rate, mean arterial pressure, respiratory rate, glucose, white blood cell, hemoglobin, calcium, sodium, potassium, hypertension, atrial fibrillation, congestive heart failure, cerebrovascular disease, renal disease, malignant cancer, sepsis, vasoactive agent, and mechanical ventilation
Subgroup analysis
Subgroup analyses were conducted across sex, age, atrial fibrillation, congestive heart failure, malignant cancer, and sepsis status (Fig. 4). Elevated log2(EASIX) levels were consistently associated with increased risks of both 30-day and 365-day mortality across all subgroups. Although significant multiplicative interactions were observed for sepsis status (P for interaction = 0.003) in 30-day mortality, and for malignant cancer (P for interaction < 0.001) and sepsis (P for interaction < 0.001) in 365-day mortality, the direction of the association remained consistent throughout all subgroups.
Fig. 4.
Subgroup analyses of the association between log2 (EASIX) and mortality at 30-day and 365-day. Adjusted for age, sex, weight, heart rate, mean arterial pressure, respiratory rate, glucose, white blood cell, hemoglobin, calcium, sodium, potassium, hypertension, atrial fibrillation, congestive heart failure, cerebrovascular disease, renal disease, malignant cancer, sepsis, vasoactive agent, and mechanical ventilation
Mediation analysis
The mediation analysis revealed that lactate played a significant mediating role in the relationship between log2(EASIX) and mortality. The mediating effect was more pronounced for 30-day mortality than for 365-day mortality, with lactate accounting for 14.82% and 11.03% of the total effect, respectively (Fig. 5).
Fig. 5.
Mediation analysis of the association between EASIX and mortality by lactate: (A) 30-day mortality; (B) 365-day mortality
Sensitivity analysis
We performed sensitivity analyses to assess the robustness of the primary outcome. The final propensity score-matched cohort comprised 1419 patients. Baseline characteristics before and after matching are presented in Supplementary Table 4. After matching, all covariates demonstrated standardized mean differences (SMDs) < 0.1 (Fig. 6), indicating good balance across baseline characteristics.
Fig. 6.
Standardized Mean Differences (SMD) of covariates between original and matched cohorts
Consistent with the primary multivariable regression analysis, propensity score matching (PSM) demonstrated significant associations between log2(EASIX) and 30-day mortality. In the matched cohort, patients in the highest log2(EASIX) tertile (T3) showed a 76% increased mortality risk compared to those in the lowest tertile (T1) (HR 1.76, 95% CI 1.37–2.27). Doubly robust estimation produced comparable risk estimates, with T3 patients exhibiting 90% higher mortality versus T1 (HR 1.90, 95% CI 1.46–2.45). Qualitatively consistent hazard ratios were observed across multivariable, PSM, and doubly robust models (all P for trend < 0.001) (Table 3).
Table 3.
Primary outcome analysis with different models
| Method | Variable | 30-Day Mortality | |
|---|---|---|---|
| HR (95%CI) | P value | ||
| Multivariate | T1 | 1(Ref) | |
| T2 | 1.33 (1.10 ~ 1.59) | 0.003 | |
| T3 | 2.23 (1.86 ~ 2.68) | < 0.001 | |
| P for trend | < 0.001 | ||
| Matched | T1 | 1(Ref) | |
| T2 | 0.84 (0.63 ~ 1.13) | 0.258 | |
| T3 | 1.76 (1.37 ~ 2.27) | < 0.001 | |
| P for trend | < 0.001 | ||
| Doubly robust with all covariates | T1 | 1(Ref) | |
| T2 | 0.90 (0.67 ~ 1.20) | 0.464 | |
| T3 | 1.90 (1.46 ~ 2.45) | < 0.001 | |
| P for trend | < 0.001 | ||
Log2(EASIX): T1 (< 0.53), T2 (0.53–2.05), T3 (≥ 2.05). HR, Hazard Ratio; CI, Confidential Interval
Discussion
This study demonstrates that elevated EASIX levels at ICU admission are independently associated with increased mortality in critically ill patients with DM. This association remained significant after comprehensive adjustment for confounders and was further supported by propensity score matching analysis. A significant linear relationship was observed between EASIX and mortality risk. The association was consistent across most predefined subgroups, although a notable multiplicative interaction was observed in patients with sepsis. Lactate was identified as a significant mediator of this relationship.
DM imposes a substantial global health burden and is strongly associated with the development of multi-system complications. It induces widespread endothelial damage through interconnected pathways involving chronic hyperglycemia, insulin resistance, and oxidative stress, which sustain a state of chronic inflammation [21–23]. This results in a dysfunctional endothelium characterized by impaired vasodilation, increased permeability, and a pro-thrombotic phenotype [24, 25]. In critical illness, this pre-existing vulnerability is amplified, compromising microvascular regulation and accelerating capillary leak, tissue hypoperfusion, and multi-organ failure [26]. Experimental evidence substantiates that the diabetic state exacerbates endothelial activation and barrier dysfunction in response to acute insults, providing a direct mechanistic link to the poor clinical outcomes observed in this population [27, 28].
There is growing interest in biomarkers capable of quantifying endothelial dysfunction. EASIX has emerged as a promising candidate, valued for its computational simplicity and capacity to reflect integrative endothelial status [29]. The study by Pedraza et al. in sequential cohorts of HSCT recipients demonstrated that EASIX increased in parallel with established biomarkers of endothelial injury, including VCAM-1, TNFR1, and VWF: Ag, and effectively predicted the development of acute graft-versus-host disease, confirming its role as a reliable surrogate of endothelial damage [30]. Within the context of COVID-19, EASIX has been established as a strong predictor of mortality, further supporting its prognostic relevance beyond hematological conditions [31]. Subsequent studies have expanded the utility of EASIX to broader critical care populations. Elevated EASIX levels are significantly associated with an increased risk of 28-day and 90-day all-cause mortality in patients with sepsis [32].
EASIX reflects endothelial injury through the specific pathologies captured by its individual components. LDH elevation signals widespread cellular ischaemia and necrosis, marking the extent of tissue hypoxia that compromises vascular integrity [33, 34]. Thrombocytopenia tracks ongoing platelet consumption within the microvasculature: as endothelial surfaces become pro-thrombotic, circulating platelets are trapped, fragmented and depleted, so the falling count flags the intensity of systemic micro-clot formation [35, 36]. Serum creatinine rises not only with renal impairment but also with worsening systemic perfusion and inflammatory stress, paralleling drivers of endothelial dysfunction [37, 38]. Collectively, EASIX integrates these complementary signals of cellular hypoxia, thrombotic consumption, and systemic stress into a single composite index of endothelial dysregulation.
The observed association between EASIX and mortality invites further mechanistic interpretation, informed by our mediation and subgroup findings. The identification of lactate as a significant mediator raises the possibility that tissue hypoperfusion and impaired oxygen utilization could represent one plausible pathway through which the endothelial and microcirculatory disturbances reflected by EASIX contribute to fatal outcomes. However, it should be noted that some studies have reported EASIX and lactate as independent predictors of mortality, suggesting that the relationship between these variables is complex and may involve both independent and mediated effects. Additionally, the significant interaction observed in the sepsis subgroup indicates that the relationship between EASIX and mortality is not constant but may vary depending on the clinical setting. The attenuated association observed in patients with sepsis may be attributed to the overwhelming systemic inflammation and coagulopathy inherent to this condition, which could serve as dominant competing risks, thereby partially obscuring the specific contribution of endothelial injury to mortality. In non-septic patients, endothelial dysfunction quantified by EASIX may play a more influential role in determining outcome, with tissue hypoxia serving as a key mediating pathway.
This study identifies EASIX as a potential tool for stratifying prognosis in critically ill diabetic patients. The identification of lactate as a significant mediator suggests that tissue hypoperfusion may be a potential treatable pathway linking endothelial dysfunction to mortality, though further studies are needed to validate this mediation mechanism and to disentangle the independent versus mediated effects of these biomarkers. This insight reinforces the importance of targeting microcirculatory optimization and systemic oxygen delivery in the management of this high-risk population.
Several limitations must be acknowledged. First, the retrospective observational design precludes causal inference, and unmeasured confounding may persist despite multivariate adjustment; in addition, the study protocol was not pre-registered, which may introduce potential bias. Second, the use of data from a single critical care database may affect the generalizability of our findings. Third, only the baseline EASIX at ICU admission was analyzed; its dynamic changes during the ICU stay, which may carry additional prognostic information, were not evaluated. Fourth, while EASIX is posited to reflect endothelial injury, its components are not specific. Elevations in LDH and creatinine, as well as reductions in platelet count, can be influenced by numerous non-endothelial factors such as hemolysis, hepatic or renal dysfunction, bleeding, or transfusion. This lack of specificity may confound its interpretation as a direct measure of endothelial damage in individual patients. Fifth, a considerable number of patients were excluded due to missing data. Although we assumed these data to be missing at random, this assumption cannot be verified, and selection bias may still exist. Sixth, due to database limitations, we were unable to assess long-term glycemic control (e.g., HbA1c) or outpatient diabetes treatment regimens, which may influence disease severity and introduce additional confounding.
Conclusions
Elevated EASIX at ICU admission is independently associated with increased mortality in critically ill patients with diabetes, an association partially mediated by lactate. These findings suggest that EASIX may serve as a useful adjunctive marker for early risk stratification in this vulnerable population.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
The authors thank the Laboratory for Computational Physiology at MIT (LCP-MIT) for providing access to the MIMIC-IV database. We acknowledge Jie Liu (People’s Liberation Army General Hospital), Qilin Yang (The Second Affiliated Hospital of Guangzhou Medical University), and Haibo Li (Fujian Maternity and Child Health Hospital) for their contributions to study design and statistical analysis.
Abbreviations
- EASIX
Endothelial Activation and Stress Index
- DM
Diabetes Mellitus
- MIMIC-IV
Medical Information Mart for Intensive Care IV
- ICU
Intensive Care Unit
- LDH
Lactate Dehydrogenase
- STROBE
Strengthening the Reporting of Observational Studies in Epidemiology
- ICD
International Classification of Diseases
- OASIS
Oxford Acute Severity of Illness Score
- SAPS II
Simplified Acute Physiology Score II
- SOFA
Sequential Organ Failure Assessment
- HR
Hazard Ratio
- CI
Confidence Interval
- SD
Standard Deviation
- IQR
Interquartile Range
- GVIF
Generalized Variance Inflation Factor
- PSM
Propensity Score Matching
- SMD
Standardized Mean Differences
Author contributions
The study was designed by RD and HZ. RD performed data collection, analysis, and wrote the initial manuscript. RS, YD, XN, JL, JW, and LW contributed to data analysis and figure preparation. HZ participated in writing and editing the manuscript. All authors reviewed and approved the final version of the manuscript.
Funding
This work was supported by The Hospital Fund Project of the Second Hospital of Shanxi Medical University (NO. 202404-10).
Data availability
The data underlying this article were accessed from PhysioNet (https://physionet.org/content/mimic-iv/, version 3.1). The derived data generated in this research will be shared on reasonable request to the corresponding author with permission from PhysioNet.
Declarations
Ethics approval and consent to participate
Approval for using the MIMIC-IV database was obtained from the Institutional Review Board of the Massachusetts Institute of Technology and the Institutional Review Board of Beth Israel Deaconess Medical Center. The MIMIC database’s existing ethical approval applies to the data in this study, eliminating the requirement for additional ethical approval or informed consent. This retrospective study was conducted in accordance with 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.
References
- 1.Ong KL, Stafford LK, McLaughlin SA, Boyko EJ, Vollset SE, Smith AE, et al. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the global burden of disease study 2021. Lancet. 2023;402:203–34. 10.1016/S0140-6736(23)01301-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB, et al. IDF diabetes atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119. 10.1016/j.diabres.2021.109119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Lin X, Xu Y, Pan X, Xu J, Ding Y, Sun X, et al. Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025. Sci Rep. 2020;10:14790. 10.1038/s41598-020-71908-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Hossain MJ, Al-Mamun M, Islam MR. Diabetes mellitus, the fastest growing global public health concern: early detection should be focused. Health Sci Rep. 2024;7:e2004. 10.1002/hsr2.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Krinsley JS. Glycemic variability and mortality in critically 111 patients: the impact of diabetes. J Diabetes Sci Technol. 2009;3:1292–301. 10.1177/193229680900300609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Liu L, Zhao Y-B, Cheng Z-T, Li Y-H, Liu Y. Development and validation of a prognostic model for critically ill type 2 diabetes patients in ICU based on composite inflammatory indicators. Sci Rep. 2025;15:3627. 10.1038/s41598-025-87731-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Jandeleit-Dahm K, Cooper M. The role of AGEs in cardiovascular disease. Curr Pharm Des. 2008;14:979–86. 10.2174/138161208784139684. [DOI] [PubMed] [Google Scholar]
- 8.Carey IM, Critchley JA, DeWilde S, Harris T, Hosking FJ, Cook DG. Risk of infection in type 1 and type 2 diabetes compared with the general population: a matched cohort study. Diabetes Care. 2018;41:513–21. 10.2337/dc17-2131. [DOI] [PubMed] [Google Scholar]
- 9.Slynkova K, Mannino DM, Martin GS, Morehead RS, Doherty DE. The role of body mass index and diabetes in the development of acute organ failure and subsequent mortality in an observational cohort. Crit Care. 2006;10:R137. 10.1186/cc5051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Christiansen CF, Johansen MB, Christensen S, O’Brien JM, Tønnesen E, Sørensen HT. Type 2 diabetes and 1-year mortality in intensive care unit patients. Eur J Clin Invest. 2013;43:238–47. 10.1111/eci.12036. [DOI] [PubMed] [Google Scholar]
- 11.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:e007635. 10.1136/jitc-2023-007635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Luft T, Benner A, Jodele S, Dandoy CE, Storb R, Gooley T, et al. EASIX in patients with acute graft-versus-host disease: a retrospective cohort analysis. Lancet Haematol. 2017;4:e414–23. 10.1016/S2352-3026(17)30108-4. [DOI] [PubMed] [Google Scholar]
- 13.Gavriilaki E, Demosthenous C, Evangelidis P, Bousiou Z, Batsis I, Vardi A, et al. Soluble urokinase-type plasminogen activator receptor (suPAR), growth differentiation factor-15 (GDF-15), and soluble C5b-9 (sC5b-9) levels are significantly associated with endothelial injury indices in CAR-T cell recipients. Int J Mol Sci. 2024;25:11028. 10.3390/ijms252011028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Li N, Yin J, Zeng Y, Li S, Huang C, Chen G, et al. Association between endothelial activation and stress index and all-cause mortality risk in sepsis patients: a retrospective cohort analysis utilizing the MIMIC-IV database. BMC Infect Dis. 2025;25:966. 10.1186/s12879-025-11363-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zhang Z, Wang C, Wang J, Luo Z, Zou K, Xu Q. Endothelial activation and stress index elevation associated with adverse prognosis in sepsis-associated acute kidney injury patients: validation based on two critical care cohorts. Shock [Internet]. 2025. 10.1097/SHK.0000000000002715. [cited 2025 Oct 7]. [DOI] [PubMed] [Google Scholar]
- 16.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. 10.3389/fphys.2025.1570988. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zhang Z, Li Y, Wei Y, Yang Y, Luo Z, Xie J, et al. The association between endothelial activation and stress index and the development and prognosis of acute kidney injury in elderly patients with critical illness. Ren Fail. 2025;47:2577174. 10.1080/0886022X.2025.2577174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Yin F, Wang K. Independent prognostic importance of endothelial activation and stress index (EASIX) in critically ill patients with heart failure: modulating role of inflammation. Front Med. 2025;12:1560947. 10.3389/fmed.2025.1560947. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Cuschieri S. The STROBE guidelines. Saudi J Anaesth. 2019;13:31. 10.4103/sja.SJA_543_18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.McCaffrey DF, Griffin BA, Almirall D, Slaughter ME, Ramchand R, Burgette LF. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med. 2013;32:3388–414. 10.1002/sim.5753. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Potenza M, Gagliardi S, Nacci C, Carratu M, Montagnani M. Endothelial dysfunction in diabetes: from mechanisms to therapeutic targets. Curr Med Chem. 2009;16:94–112. 10.2174/092986709787002853. [DOI] [PubMed] [Google Scholar]
- 22.Giacco F, Brownlee M. Oxidative stress and diabetic complications. Schmidt AM, editor. Circ Res. 2010;107:1058–70. 10.1161/CIRCRESAHA.110.223545. [DOI] [PMC free article] [PubMed]
- 23.Kim J, Montagnani M, Koh KK, Quon MJ. Reciprocal relationships between insulin resistance and endothelial dysfunction: molecular and pathophysiological mechanisms. Circulation. 2006;113:1888–904. 10.1161/CIRCULATIONAHA.105.563213. [DOI] [PubMed] [Google Scholar]
- 24.Vanhoutte PM, Shimokawa H, Feletou M, Tang EHC. Endothelial dysfunction and vascular disease - a 30th anniversary update. Acta Physiol. 2017;219:22–96. 10.1111/apha.12646. [DOI] [PubMed] [Google Scholar]
- 25.Li Y, Liu Y, Liu S, Gao M, Wang W, Chen K, et al. Diabetic vascular diseases: molecular mechanisms and therapeutic strategies. Signal Transduct Target Ther. 2023;8:152. 10.1038/s41392-023-01400-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Marik PE, Bellomo R. Stress hyperglycemia: an essential survival response! Crit Care. 2013;17:305. 10.1186/cc12514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Collado A, Humoud R, Kontidou E, Eldh M, Swaich J, Zhao A, et al. Erythrocyte-derived extracellular vesicles induce endothelial dysfunction through arginase-1 and oxidative stress in type 2 diabetes. J Clin Invest. 2025;135:e180900. 10.1172/JCI180900. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Yang D-R, Wang M-Y, Zhang C-L, Wang Y. Endothelial dysfunction in vascular complications of diabetes: a comprehensive review of mechanisms and implications. Front Endocrinol. 2024;15:1359255. 10.3389/fendo.2024.1359255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wu L, Dong G, Wu T, Gu X, Zhang Y, Wang S. The endothelial activation and stress index as a risk factor for all-cause mortality in patients with anemia: a nationwide retrospective cross-sectional study. Hematology. 2025;30:2542634. 10.1080/16078454.2025.2542634. [DOI] [PubMed] [Google Scholar]
- 30.Pedraza A, Salas MQ, Rodríguez-Lobato LG, Escribano-Serrat S, Suárez-Lledo M, Martínez-Cebrian N, et al. Easix score correlates with endothelial dysfunction biomarkers and predicts risk of acute graft-versus-host disease after allogeneic transplantation. Transpl Cell Ther. 2024;30:187e1. 10.1016/j.jtct.2023.11.016. [DOI] [PubMed] [Google Scholar]
- 31.Jeong JH, Heo M, Park S, Lee SH, Park O, Kim T, et al. Association between age-adjusted endothelial activation and stress index and intensive care unit mortality in patients with severe COVID-19. Tuberc Respir Dis. 2024;87:524–31. 10.4046/trd.2024.0081. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Xu H-B, Ye Y, Xue F, Wu J, Suo Z, Zhang H. Association between endothelial activation and stress index and 28-day mortality in septic ICU patients: a retrospective cohort study. Int J Med Sci. 2023;20:1165–73. 10.7150/ijms.85870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Cheng X, Liu L, Tian Y, Lin Y. Serum lactate dehydrogenase as a prognostic marker for 90-day mortality in connective tissue disease patients receiving glucocorticoids and hospitalized with pneumonia: a cohort study. Sci Rep. 2025;15:16806. 10.1038/s41598-025-01721-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Yan H, Liang X, Du J, He Z, Wang Y, Lyu M, et al. Proteomic and metabolomic investigation of serum lactate dehydrogenase elevation in COVID-19 patients. Proteomics. 2021;21:2100002. 10.1002/pmic.202100002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Engelmann B, Massberg S. Thrombosis as an intravascular effector of innate immunity. Nat Rev Immunol. 2013;13:34–45. 10.1038/nri3345. [DOI] [PubMed] [Google Scholar]
- 36.Levi M, Van Der Poll T. Coagulation and sepsis. Thromb Res. 2017;149:38–44. 10.1016/j.thromres.2016.11.007. [DOI] [PubMed] [Google Scholar]
- 37.Kellum JA, Lameire N, for the KDIGO AKI Guideline Work Group. Diagnosis, evaluation, and management of acute kidney injury: a KDIGO summary (Part 1). Crit Care. 2013;17:204. 10.1186/cc11454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Vassalotti JA, Francis A, Soares Dos Santos AC, Correa-Rotter R, Abdellatif D, Hsiao L-L, et al. Are your kidneys ok? Detect early to protect kidney health. Kidney Int. 2025;107:370–7. 10.1016/j.kint.2024.12.006. [DOI] [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 data underlying this article were accessed from PhysioNet (https://physionet.org/content/mimic-iv/, version 3.1). The derived data generated in this research will be shared on reasonable request to the corresponding author with permission from PhysioNet.