Abstract
Background
Acute myocardial infarction (AMI) in critically ill patients is associated with high mortality. The glucose-to-platelet ratio (GPR), derived from routine early glucose and platelet measurements, may provide additional prognostic information, but its value in ICU patients with AMI remains unclear.
Methods
This retrospective cohort study included 4,676 adult ICU patients with AMI from the MIMIC-IV version 3.1 database. GPR was calculated as the mean glucose concentration during the first 24 h after ICU admission divided by the first platelet count recorded within the same period. Outcomes were 30-day and 360-day all-cause mortality from ICU admission. Associations were assessed using multivariable Cox proportional hazards models, restricted cubic splines, and exploratory internally validated prediction models.
Results
Among 4,676 patients, the 30-day and 360-day mortality rates were 19.6% and 29.1%, respectively. In the fully adjusted model, each 1-unit increase in GPR was associated with higher 30-day mortality (HR, 1.10; 95% CI, 1.05–1.15) and 360-day mortality (HR, 1.12; 95% CI, 1.07–1.16; both P < 0.001). Compared with the lowest tertile, the highest GPR tertile was associated with increased 30-day mortality (HR, 1.60; 95% CI, 1.35–1.90) and 360-day mortality (HR, 1.29; 95% CI, 1.12–1.47; both P < 0.001). Spline analyses suggested nonlinearity only for 30-day mortality. In exploratory internal validation, ridge regression incorporating GPR achieved an area under the receiver operating characteristic curve of 0.816.
Conclusions
Higher GPR was independently associated with increased short- and long-term all-cause mortality in critically ill patients with AMI. GPR may complement conventional variables for early ICU risk stratification.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12872-026-06222-9.
Keywords: Glucose-to-platelet ratio, Acute myocardial infarction, Intensive care, All-cause mortality, Machine learning
Introduction
Acute myocardial infarction (AMI), characterized by myocardial necrosis caused by acute coronary ischemia, remains a major contributor to global cardiovascular mortality and disability [1]. Although advances in reperfusion therapy and critical care have improved the management of AMI, patients requiring intensive care unit (ICU) admission still represent a particularly high-risk subgroup [2]. In this population, mortality remains high because of the combined impact of multiorgan dysfunction, metabolic derangements, and coagulation abnormalities [3]. Identifying simple and reliable biomarkers that reflect systemic stress responses is therefore crucial for timely risk stratification and clinical assessment.
Beyond traditional risk scores, modern critical care increasingly emphasizes integrative pathophysiologic markers capable of capturing the interplay between metabolism, inflammation, and hemostasis [4, 5]. Recent studies in acute coronary syndrome (ACS) and ST-segment elevation myocardial infarction (STEMI) have shown that admission oxidative stress markers, such as serum malondialdehyde, and composite hematologic–nutritional indices, such as the hemoglobin–albumin–lymphocyte–platelet (HALP) score, are associated with in-hospital mortality [6, 7]. These findings suggest that early biomarkers reflecting systemic pathophysiologic stress may provide clinically relevant prognostic information. In critically ill AMI patients, where myocardial injury rapidly transitions into systemic dysfunction, single-organ indicators may be insufficient [8]. Therefore, composite indices that summarize multiple biologic processes may provide complementary information for early prognostication.
Hyperglycemia and thrombocytopenia are common complications in critically ill patients with AMI, each associated with negative outcomes [9, 10]. Stress-induced hyperglycemia reflects catecholamine- and cortisol-mediated metabolic activation, which promotes oxidative stress, endothelial injury, and immune dysregulation. Elevated glucose levels are associated with larger infarct sizes and higher short- and long-term mortality rates, regardless of preexisting diabetes [11]. Conversely, thrombocytopenia frequently reflects platelet activation and consumption, leading to microvascular obstruction, inflammatory responses, and disseminated intravascular coagulation. Decreased platelet counts have been identified as prognostic indicators of adverse outcomes in both critical illness and myocardial infarction [12]. However, these two parameters are typically evaluated separately, potentially overlooking their combined pathophysiologic significance.
The concurrence of hyperglycemia and platelet depletion may signal a broader “metabolic–hemostatic imbalance,” representing the body’s maladaptive stress response to ischemia and inflammation. From a mechanistic perspective, excessive glucose-driven oxidative injury impairs mitochondrial ATP synthesis [13], while platelet consumption depletes vascular repair mediators and increases microthrombotic burden [14]. The synergy between these processes can amplify myocardial injury and compromise organ perfusion. Recognizing this interdependence provides a rationale for developing integrative indices that quantify both metabolic activation and hemostatic exhaustion.
Recent studies have proposed composite indices such as the stress hyperglycemia ratio (SHR) and glucose-to-lymphocyte ratio (GLR) to describe the interplay between metabolic stress and immune or hemostatic dysfunction [15, 16]. However, evidence integrating glucose metabolism with platelet dynamics in AMI remains scarce. The glucose-to-platelet ratio (GPR), calculated as the ratio of blood glucose concentration to platelet count, has emerged as a ratio-based risk marker that reflects both metabolic activation and hemostatic exhaustion [17]. In previous work, Chen et al. identified an elevated GPR as a marker linked to systemic inflammation, endothelial dysfunction, and adverse outcomes among patients with stroke [17]. It is unclear whether this association applies to critically ill patients with AMI.
Examining GPR in critically ill patients with AMI may help determine whether the coexistence of early glycemic stress and platelet-related illness severity identifies a subgroup at increased mortality risk. This study investigated the relationship between the GPR and both short-term (30-day) and long-term (360-day) all-cause mortality in critically ill AMI patients, utilizing data from the MIMIC-IV database. In addition, exploratory prediction models were developed to evaluate the potential role of GPR in early ICU risk stratification as an integrative, ratio-based marker.
Methods
Research design
This cohort study, designed as a retrospective observational investigation, complied with the STROBE guidelines for the reporting of observational research [18]. The data used in this study were obtained from the MIMIC-IV database (version 3.1), which was developed by MIT’s Laboratory for Computational Physiology and Beth Israel Deaconess Medical Center. The database comprises de-identified clinical data from over 70,000 ICU admissions from 2008 to 2022, covering demographics, laboratory results, vital signs, treatments, and outcomes [19]. Jie Peng, having completed the Collaborative Institutional Training Initiative (CITI) program and obtained authorized database access, conducted the data extraction. SQL was executed using PostgreSQL 13.7.2 and Navicat Premium 16. The creation and use of MIMIC-IV were approved by the Institutional Review Boards of MIT and BIDMC, with waiver of informed consent. Because this secondary analysis used de-identified data available to credentialed users, additional institutional review board approval was not required.
This study enrolled adult patients aged ≥ 18 years who were admitted to the ICU with a diagnosis of AMI. AMI was identified using ICD-9 and ICD-10 diagnosis codes. Specifically, ICD-9-CM codes beginning with 410 and ICD-10-CM codes beginning with I21 or I22 were used to identify AMI. For patients with multiple ICU admissions, only the first ICU admission was included. Patients were excluded if their ICU length of stay was shorter than 24 h or if blood glucose or platelet measurements within the first 24 h after ICU admission were unavailable. Patients were stratified into tertiles according to GPR.
Data collection and variable definitions
Baseline characteristics encompassed demographics (age, sex, race, weight), vital signs (heart rate, SBP, DBP, respiratory rate, temperature, oxygen saturation), and severity indices (Logistic Organ Dysfunction Score [LODS], Sequential Organ Failure Assessment [SOFA], Charlson Comorbidity Index [CCI]). Comorbidities included congestive heart failure, atrial fibrillation, chronic pulmonary disease, liver disease, malignant cancer, diabetes, and hypertension. The laboratory parameters assessed were red blood cell count, white blood cell count, platelet count, sodium, potassium, calcium, blood glucose, blood urea nitrogen, creatinine, international normalized ratio, prothrombin time, partial thromboplastin time, and urine output. Treatment-related variables included epinephrine use, norepinephrine use, insulin use, statin use, beta-blocker use, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker (ACEI/ARB) use, mechanical ventilation, and continuous renal replacement therapy (CRRT). Unless otherwise specified, these treatment-related and organ-support variables were defined according to use within the first 24 h after ICU admission.
Physiologic and laboratory variables were extracted from the first 24 h after ICU admission according to prespecified summary rules. GPR was calculated as the mean glucose concentration during the first 24 h after ICU admission (mg/dL) divided by the first platelet count recorded within the same 24-hour window. Comorbidities such as congestive heart failure, chronic pulmonary disease, liver disease, malignant cancer, atrial fibrillation (AF), and diabetes were identified via the Charlson Comorbidity Index (CCI), while hypertension was classified using International Classification of Diseases (ICD) codes. The database outcomes comprised hospital and ICU lengths of stay (LOS), in-hospital and ICU mortality, as well as all-cause mortality at 30 and 360 days.
Variables with more than 20% missingness were excluded from the analysis. For variables with less than 20% missingness, multiple imputation was performed using a random forest–based method implemented in the mice package in R, generating five imputed datasets (m = 5). For association analyses, estimates from the imputed datasets were pooled according to Rubin’s rules. For prediction modeling, all preprocessing steps, including imputation, standardization, feature selection, and hyperparameter tuning, were performed within the training cohort and then applied to the validation cohort to minimize data leakage. For association analyses, exposure and outcome variables were included as predictors in the imputation model to preserve the association structure but were not themselves imputed. For prediction modeling, imputation models were fitted in the training cohort only and did not use outcome information from the validation cohort [20, 21]. The potential influence of missing-data handling was considered when interpreting the prediction analyses.
Clinical outcomes
The primary outcome was 30-day all-cause mortality after ICU admission, and the secondary outcome was 360-day all-cause mortality after ICU admission. Follow-up began at ICU admission and ended at death or was administratively censored at 30 or 360 days, respectively. Mortality status was ascertained from death information available in MIMIC-IV, including hospital records and linked state death records [19].
Statistical methods
Descriptive analyses were conducted for the overall cohort and across GPR tertiles. The distribution of continuous variables was assessed using the Shapiro–Wilk test and visual inspection. Continuous variables were summarized as mean ± standard deviation or median with interquartile range, as appropriate. Categorical variables were presented as counts and percentages. Across-group comparisons were performed using one-way analysis of variance or the Kruskal–Wallis test for continuous variables and the Chi-square test or Fisher exact test for categorical variables, as appropriate.
Analysis of the association between GPR and clinical outcomes
Survival across GPR tertiles was estimated using Kaplan–Meier curves and compared with the log-rank test. Cox proportional hazards models were used to evaluate the associations of GPR with 30-day and 360-day all-cause mortality, with results reported as hazard ratios (HRs) and 95% confidence intervals (CIs). GPR was analyzed both as a continuous variable and as tertiles, with the lowest tertile used as the reference group. P for trend was calculated by modeling GPR tertiles as an ordinal variable.
Candidate covariates for multivariable Cox models were selected based on univariable Cox regression, Boruta-based variable-importance analysis, and clinical relevance. Three Cox models with increasing adjustment were constructed. Treatment-related and organ-support variables were not forced into the primary fully adjusted model because they may represent downstream markers of disease severity and could introduce overadjustment or collider bias; these variables were assessed in exploratory subgroup analyses.
Multicollinearity among covariates in the fully adjusted model was assessed using variance inflation factors. The proportional hazards assumption for GPR was evaluated using Schoenfeld residuals. Restricted cubic spline (RCS) models with three knots placed at the 10th, 50th, and 90th percentiles of the GPR distribution were fitted within the fully adjusted Cox model to explore potential nonlinear associations between GPR and mortality. To further compare the discriminative performance of GPR with its individual components, exploratory AUROC analyses were performed for fixed 30-day and 360-day mortality status. The AUROCs of GPR, the mean glucose concentration used in the numerator, and the first platelet count used in the denominator were compared using the DeLong test. For platelet count, the risk direction was specified so that lower platelet values corresponded to higher mortality risk. Subgroup analyses were performed according to major treatment-related or organ-support variables, and interaction terms were used to assess effect modification.
Prediction model methodology
For exploratory prediction modeling of 30-day all-cause mortality, the dataset was randomly divided, stratified by 30-day mortality status, into a training cohort and a validation cohort at a 70:30 ratio. For prediction modeling, the training–validation split was performed before imputation and standardization. Imputation models, standardization parameters, feature selection, and hyperparameter tuning were fitted exclusively within the training cohort and then applied to the validation cohort to avoid data leakage. Candidate predictors were standardized within the modeling pipeline. LASSO regression with 10-fold cross-validation was used for feature selection in the training cohort.
Model performance was evaluated in the validation cohort using the area under the receiver operating characteristic curve (AUROC), area under the precision–recall curve (AUPRC), accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1-score, Brier score, calibration slope, and AB score. AUROC 95% CIs were estimated using bootstrap resampling. Threshold-dependent metrics, including accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and F1-score, were calculated using the optimal cutoff determined by the Youden index in the training cohort and then fixed in the validation cohort. The AB score was defined as 0.5 × AUROC + 0.5 × (1 − Brier score) and was used as a composite internal ranking metric integrating discrimination and overall probabilistic accuracy; higher AB scores indicated better overall performance [22]. Calibration curves and decision curve analysis were used to assess calibration and potential clinical relevance.
To explore the potential contribution of GPR to risk stratification, binary logistic regression models without and with GPR were compared for 30-day and 360-day mortality. The reference model included the same covariates as fully adjusted Model 3 but excluded GPR, whereas the extended model additionally included GPR. Continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI), together with their 95% CIs and P values, were calculated. NRI and IDI were calculated for fixed-time binary mortality status at 30 and 360 days. These prediction analyses were considered exploratory because only internal validation was performed.
Statistical analyses were conducted using R software and IBM SPSS Statistics. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.
Results
Baseline characteristics
The study flow diagram is shown in Fig. 1. A total of 4,676 critically ill patients with AMI were included in the final analysis and classified into three GPR tertiles: T1 (n = 1,543), T2 (n = 1,590), and T3 (n = 1,543). The missing data pattern and missingness proportions are summarized in Supplement S1.
Fig. 1.
Flowchart of patient selection. Flowchart showing patient selection and grouping according to GPR tertiles. Abbreviations: GPR, glucose-to-platelet ratio; ICU, intensive care unit
Table 1 summarizes the baseline characteristics and clinical outcomes according to GPR tertiles. The mean age was 70.1 ± 13.3 years, and 65.0% of patients were male. The median GPR was 0.8 (0.5–1.1) in the overall cohort, with tertile-specific values of 0.5 (0.4–0.5), 0.8 (0.7–0.8), and 1.3 (1.1–1.7) in T1, T2, and T3, respectively. Compared with patients in the lower tertiles, those in T3 had greater illness severity, including higher LODS, SOFA, and CCI values. GPR elevation was accompanied by higher glucose levels and lower platelet counts, as expected from the definition of the index. Patients in T3 also had higher BUN, creatinine, INR, PT, and PTT values, and a higher prevalence of liver disease, malignant cancer, diabetes, and hypertension. Treatment-related variables, including epinephrine use, norepinephrine use, insulin use, mechanical ventilation, and CRRT, were more frequent in T3. Mortality was highest in the T3 group: 30-day all-cause mortality was 15.2%, 17.2%, and 26.4% across T1, T2, and T3, respectively, while 360-day all-cause mortality was 25.5%, 25.4%, and 36.5%, respectively.
Table 1.
Baseline characteristics and clinical outcomes according to GPR tertiles
| Characteristic | Overall N = 4,676 |
T1 N = 1,543 |
T2 N = 1,590 |
T3 N = 1,543 |
P value |
|---|---|---|---|---|---|
| GPR | 0.8 (0.5, 1.1) | 0.5 (0.4, 0.5) | 0.8 (0.7, 0.8) | 1.3 (1.1, 1.7) | < 0.001 |
| Demographic characteristics | |||||
| Age (years) | 70.1 ± 13.3 | 68.8 ± 14.3 | 70.4 ± 12.8 | 71.0 ± 12.6 | < 0.001 |
| Sex, n (%) | < 0.001 | ||||
| Female | 1635 (35.0) | 642 (41.6) | 501 (31.5) | 492 (31.9) | |
| Male | 3041 (65.0) | 901 (58.4) | 1089 (68.5) | 1051 (68.1) | |
| Race, n (%) | 0.036 | ||||
| White | 2952 (63.1) | 1009 (65.4) | 1003 (63.1) | 940 (60.9) | |
| Other | 1724 (36.9) | 534 (34.6) | 587 (36.9) | 603 (39.1) | |
| Weight (kg) | 84.2 ± 22.4 | 82.7 ± 22.5 | 85.1 ± 22.1 | 84.7 ± 22.5 | 0.008 |
| Vital signs and disease severity | |||||
| Temperature (°C) | 36.8 ± 0.5 | 36.8 ± 0.4 | 36.8 ± 0.5 | 36.8 ± 0.7 | 0.923 |
| Heart rate (beats/minute) | 83.0 ± 15.0 | 82.4 ± 15.0 | 81.7 ± 14.2 | 85.0 ± 15.5 | < 0.001 |
| Respiratory rate (breaths/minute) | 19.6 ± 3.6 | 19.4 ± 3.5 | 19.3 ± 3.3 | 20.2 ± 4.0 | < 0.001 |
| SBP (mmHg) | 115.3 ± 15.1 | 115.7 ± 15.4 | 114.9 ± 14.6 | 115.2 ± 15.4 | 0.284 |
| DBP (mmHg) | 62.8 ± 10.8 | 64.3 ± 11.2 | 62.4 ± 10.4 | 61.7 ± 10.6 | < 0.001 |
| SpO₂ (%) | 96.8 ± 2.0 | 96.8 ± 1.8 | 96.9 ± 1.8 | 96.8 ± 2.4 | 0.102 |
| LODS | 5.0 ± 3.0 | 4.3 ± 2.7 | 4.7 ± 2.9 | 6.0 ± 3.2 | < 0.001 |
| SOFA | 1.0 (0.0, 3.0) | 1.0 (0.0, 2.0) | 1.0 (0.0, 3.0) | 2.0 (0.0, 4.0) | < 0.001 |
| CCI | 6.3 ± 2.8 | 5.9 ± 2.9 | 6.1 ± 2.8 | 6.9 ± 2.7 | < 0.001 |
| Comorbidities | |||||
| Congestive heart failure, n (%) | 2370 (50.7) | 758 (49.1) | 796 (50.1) | 816 (52.9) | 0.094 |
| Atrial fibrillation, n (%) | 1674 (35.8) | 476 (30.8) | 614 (38.6) | 584 (37.8) | < 0.001 |
| Chronic pulmonary disease, n (%) | 1080 (23.1) | 374 (24.2) | 352 (22.1) | 354 (22.9) | 0.372 |
| Liver disease, n (%) | 322 (6.9) | 67 (4.3) | 77 (4.8) | 178 (11.5) | < 0.001 |
| Malignant cancer, n (%) | 390 (8.3) | 123 (8.0) | 103 (6.5) | 164 (10.6) | < 0.001 |
| Diabetes, n (%) | 2009 (43.0) | 467 (30.3) | 645 (40.6) | 897 (58.1) | < 0.001 |
| Hypertension, n (%) | 3606 (77.1) | 1114 (72.2) | 1270 (79.9) | 1222 (79.2) | < 0.001 |
| Laboratory parameters | |||||
| RBC (10¹²/L) | 3.6 ± 0.7 | 3.7 ± 0.7 | 3.7 ± 0.7 | 3.5 ± 0.8 | < 0.001 |
| WBC (10⁹/L) | 12.0 (9.1, 15.7) | 12.2 (9.3, 15.9) | 11.9 (9.3, 15.2) | 11.7 (8.6, 15.8) | 0.050 |
| Platelet (10⁹/L) | 200.3 ± 89.9 | 272.7 ± 94.7 | 189.6 ± 55.3 | 138.9 ± 57.0 | < 0.001 |
| Sodium (mmol/L) | 137.9 ± 4.5 | 137.9 ± 4.2 | 137.9 ± 4.0 | 138.0 ± 5.1 | 0.628 |
| Potassium (mmol/L) | 4.3 ± 0.6 | 4.3 ± 0.6 | 4.3 ± 0.5 | 4.4 ± 0.6 | 0.049 |
| Calcium (mg/dL) | 8.4 ± 0.7 | 8.5 ± 0.7 | 8.4 ± 0.7 | 8.3 ± 0.7 | < 0.001 |
| Glucose (mg/dL) | 153.7 ± 64.8 | 120.8 ± 31.1 | 143.2 ± 42.8 | 197.5 ± 82.1 | < 0.001 |
| BUN (mg/dL) | 22.0 (15.0, 37.5) | 19.0 (13.5, 32.1) | 20.7 (14.8, 33.3) | 27.2 (17.5, 46.0) | < 0.001 |
| Creatinine (mg/dL) | 1.1 (0.8, 1.8) | 1.0 (0.8, 1.5) | 1.1 (0.8, 1.6) | 1.3 (0.9, 2.2) | < 0.001 |
| INR | 1.4 ± 0.6 | 1.4 ± 0.7 | 1.4 ± 0.6 | 1.5 ± 0.6 | < 0.001 |
| PT (s) | 15.4 ± 6.8 | 15.2 ± 7.3 | 15.0 ± 6.2 | 16.1 ± 6.7 | < 0.001 |
| PTT (s) | 45.6 ± 24.6 | 43.9 ± 23.0 | 45.4 ± 24.1 | 47.4 ± 26.5 | < 0.001 |
| Treatments and interventions | |||||
| Epinephrine use, n (%) | 424 (9.1) | 98 (6.4) | 129 (8.1) | 197 (12.8) | < 0.001 |
| Norepinephrine use, n (%) | 1260 (26.9) | 339 (22.0) | 359 (22.6) | 562 (36.4) | < 0.001 |
| Insulin use, n (%) | 3006 (64.3) | 747 (48.4) | 1039 (65.3) | 1220 (79.1) | < 0.001 |
| Statin use, n (%) | 3238 (69.2) | 1025 (66.4) | 1183 (74.4) | 1030 (66.8) | < 0.001 |
| Beta-blocker use, n (%) | 3562 (76.2) | 1184 (76.7) | 1266 (79.6) | 1112 (72.1) | < 0.001 |
| ACEI/ARB use, n (%) | 1616 (34.6) | 551 (35.7) | 589 (37.0) | 476 (30.8) | < 0.001 |
| Mechanical ventilation, n (%) | 3928 (84.0) | 1247 (80.8) | 1352 (85.0) | 1329 (86.1) | < 0.001 |
| CRRT, n (%) | 326 (7.0) | 70 (4.5) | 68 (4.3) | 188 (12.2) | < 0.001 |
| Clinical outcomes | |||||
| Urine output (mL) | 1528.0 (929.2, 2300.0) | 1650.0 (1011.0, 2455.0) | 1574.0 (975.0, 2345.0) | 1350.0 (809.5, 2145.0) | < 0.001 |
| Hospital length of stay (days) | 8.6 (5.1, 14.0) | 8.0 (4.6, 13.6) | 8.0 (5.1, 12.7) | 9.5 (5.8, 16.0) | < 0.001 |
| In-hospital mortality, n (%) | 631 (13.5) | 134 (8.7) | 172 (10.8) | 325 (21.1) | < 0.001 |
| ICU length of stay (days) | 2.6 (1.6, 4.8) | 2.3 (1.5, 4.2) | 2.3 (1.5, 4.2) | 3.1 (1.7, 6.0) | < 0.001 |
| ICU mortality, n (%) | 454 (9.7) | 87 (5.6) | 118 (7.4) | 249 (16.1) | < 0.001 |
| 30-day all-cause mortality, n (%) | 917 (19.6) | 235 (15.2) | 274 (17.2) | 408 (26.4) | < 0.001 |
| 360-day all-cause mortality, n (%) | 1360 (29.1) | 393 (25.5) | 404 (25.4) | 563 (36.5) | < 0.001 |
Continuous variables are presented as mean ± standard deviation or median (interquartile range). Categorical variables are presented as number and percentage. P values were calculated using one-way ANOVA, Kruskal–Wallis test, Chi-square test, or Fisher exact test, as appropriate
Abbreviations: ACEI Angiotensin-converting enzyme inhibitorm, ARB Angiotensin receptor blocker, BUN Blood urea nitrogen, CCI Charlson Comorbidity Index, CRRT Continuous renal replacement therapy, DBP Diastolic blood pressure, GPR Glucose-to-platelet ratio, ICU Intensive care unit, INR International normalized ratio, IQR Interquartile range, LODS Logistic Organ Dysfunction Score, PT Prothrombin time, PTT Partial thromboplastin time, RBC Red blood cell count, SBP Systolic blood pressure, SD Standard deviation, SOFA Sequential Organ Failure Assessment, SpO₂ Peripheral oxygen saturation, WBC White blood cell count
Association between GPR and 30-day and 360-day all-cause mortality
Kaplan–Meier curves showed lower survival probabilities in the highest GPR tertile during both follow-up periods, with significant between-group differences for 30-day and 360-day all-cause mortality (both P < 0.001; Fig. 2A and B).
Fig. 2.
Kaplan–Meier survival curves according to GPR tertiles. Kaplan–Meier curves for 30-day all-cause mortality (A) and 360-day all-cause mortality (B) according to GPR tertiles. Differences among groups were compared using the log-rank test. Abbreviations: GPR, glucose-to-platelet ratio; T1, lowest tertile; T2, middle tertile; T3, highest tertile
Univariable Cox regression and Boruta-based variable-importance analysis were used to support covariate screening (Supplement S2 and S3). Based on statistical evidence and clinical relevance, three Cox models were constructed: Model 1 was unadjusted; Model 2 was adjusted for age, sex, race, and weight; and Model 3 was fully adjusted for age, sex, race, weight, potassium, BUN, INR, SOFA, congestive heart failure, atrial fibrillation, chronic pulmonary disease, liver disease, malignant cancer, diabetes, and hypertension. Treatment-related and organ-support variables were not included in the primary fully adjusted model but were assessed in exploratory subgroup analyses.
Multicollinearity assessment showed no substantial collinearity among Model 3 covariates (Supplement S4). The proportional hazards assumption was satisfied for both the 30-day and 360-day mortality models, as indicated by non-significant Schoenfeld residual tests for GPR and the corresponding global tests (Supplement S5). Specifically, the tests for GPR showed no evidence of non-proportionality for 30-day mortality (χ² = 0.407, P = 0.523) or 360-day mortality (χ² = 0.579, P = 0.447). In the fully adjusted Model 3, each 1-unit increase in GPR was associated with higher 30-day mortality (HR, 1.10; 95% CI, 1.05–1.15; P < 0.001) and 360-day mortality (HR, 1.12; 95% CI, 1.07–1.16; P < 0.001). Compared with T1, T3 was associated with increased 30-day mortality (HR, 1.60; 95% CI, 1.35–1.90; P < 0.001) and 360-day mortality (HR, 1.29; 95% CI, 1.12–1.47; P < 0.001), whereas T2 was not statistically significant for either outcome. The trend across tertiles remained significant for both outcomes (both P for trend < 0.001; Table 2).
Table 2.
Association between GPR and 30-day and 360-day all-cause mortality in Cox proportional hazards models
| Variables | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | |
| 30-day | ||||||
| GPR | 1.21 (1.16–1.26) | < 0.001 | 1.21 (1.17–1.26) | < 0.001 | 1.10 (1.05–1.15) | < 0.001 |
| GPR Tertile | ||||||
| T1 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | |||
| T2 | 1.15 (0.96–1.36) | 0.125 | 1.12 (0.94–1.33) | 0.218 | 1.16 (0.98–1.39) | 0.090 |
| T3 | 1.88 (1.60–2.21) | < 0.001 | 1.81 (1.54–2.12) | < 0.001 | 1.60 (1.35–1.90) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | |||
| 360-day | ||||||
| GPR | 1.22 (1.18–1.26) | < 0.001 | 1.24 (1.20–1.28) | < 0.001 | 1.12 (1.07–1.16) | < 0.001 |
| GPR Tertile | ||||||
| T1 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | |||
| T2 | 1.01 (0.88–1.16) | 0.907 | 0.97 (0.85–1.12) | 0.703 | 1.00 (0.87–1.15) | 0.999 |
| T3 | 1.59 (1.40–1.81) | < 0.001 | 1.51 (1.33–1.72) | < 0.001 | 1.29 (1.12–1.47) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | |||
Hazard ratios were estimated using Cox proportional hazards models. Model 1 was unadjusted. Model 2 was adjusted for age, sex, race, and weight. Model 3 was further adjusted for potassium, BUN, INR, SOFA, congestive heart failure, atrial fibrillation, chronic pulmonary disease, liver disease, malignant cancer, diabetes, and hypertension
Abbreviations: AF Atrial fibrillation, BUN Blood urea nitrogen, CI Confidence interval, GPR Glucose-to-platelet ratio, HR Hazard ratio, INR International normalized ratio, SOFA Sequential Organ Failure Assessment
Restricted cubic spline analysis showed an overall association between GPR and both outcomes (both P for overall association < 0.001). The association was nonlinear for 30-day mortality (P for nonlinearity = 0.015) and approximately linear for 360-day mortality (P for nonlinearity = 0.335) (Fig. 3A and B). These findings describe the risk pattern but do not establish a clinically actionable threshold. Exploratory AUROC comparisons using the DeLong test showed that GPR did not demonstrate significantly better discrimination than its individual components, including the mean glucose concentration used in the numerator and the first platelet count used in the denominator, for either 30-day or 360-day mortality prediction.
Fig. 3.
Restricted cubic spline analyses of GPR and mortality. Restricted cubic spline curves showing the adjusted association between GPR and 30-day all-cause mortality (A) and 360-day all-cause mortality (B). Abbreviations: CI, confidence interval; GPR, glucose-to-platelet ratio; HR, hazard ratio
Interaction analysis of GPR and clinical variables on mortality risk
After adjustment for all covariates, exploratory subgroup analyses were performed to evaluate whether the association between GPR and mortality varied across major treatment-related and organ-support variables. No significant effect modification was observed for insulin use, beta-blocker use, norepinephrine use, ACEI/ARB use, mechanical ventilation, or CRRT for either the primary or secondary outcome, with all interaction P values > 0.05. Figure 4A and B present the subgroup analyses for 30-day and 360-day all-cause mortality, respectively.
Fig. 4.
Subgroup analyses of the association between GPR and mortality. Forest plots showing the association between GPR and 30-day all-cause mortality (A) and 360-day all-cause mortality (B) across prespecified exploratory treatment-related and organ-support subgroups. Hazard ratios were estimated per 1-unit increase in GPR. Subgroup-specific models were adjusted for the same covariates as Model 3, excluding the stratification variable when applicable. Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; CI, confidence interval; CRRT, continuous renal replacement therapy; GPR, glucose-to-platelet ratio; HR, hazard ratio
Exploratory prediction modeling and incremental value of GPR
LASSO-based feature selection for machine-learning model development is presented in Supplement S6. The minimum binomial deviance occurred at log(λ) = − 4.5, with 21 nonzero coefficients. The selected predictors included age, heart rate, respiratory rate, temperature, LODS, SOFA, CCI, GPR, BUN, INR, PT, PTT, urine output, norepinephrine use, statin use, beta-blocker use, CRRT, liver disease, malignant cancer, and diabetes. Supplement S6A presents the cross-validation curve for determining the optimal λ value, while Supplement S6B illustrates the coefficient trajectories for each variable. The validation-cohort performance of the six machine-learning algorithms is summarized in Supplement S7, and their ROC curves are shown in Supplement S8. Ridge regression and elastic net showed nearly identical validation performance, with both achieving an AUROC of 0.816 and an AB score of 0.848. Ridge regression was selected as a representative top-performing model because it showed a slightly higher AUPRC and a calibration slope closer to 1.0, although its performance was similar to that of elastic net. In the validation cohort, the ridge model incorporating GPR achieved an AUROC of 0.816 (95% CI, 0.787–0.844), AUPRC of 0.561, Brier score of 0.120, calibration slope of 1.031, and AB score of 0.848. Time-dependent AUROC at 30 days was calculated for the Cox model. The multivariable Cox model incorporating GPR had an AUROC of 0.757 (95% CI, 0.725–0.789) and a Brier score of 0.142. Decision curve analysis suggested that the ridge model had a higher net benefit than the Cox model across a range of threshold probabilities (Fig. 5). Because this comparison involved different modeling frameworks, these findings should be interpreted as exploratory and should not be used to infer that GPR is superior to individual laboratory parameters.
Fig. 5.
Exploratory comparison between the ridge regression model and the Cox model for 30-day mortality. Receiver operating characteristic curve (A), decision curve analysis (B), and calibration curve (C) comparing the ridge regression model with the multivariable Cox model for 30-day all-cause mortality. Abbreviations: Cox, Cox proportional hazards model; DCA, decision curve analysis; GPR, glucose-to-platelet ratio; ML, machine learning; ROC, receiver operating characteristic
The incremental value of adding GPR to conventional clinical covariates was further assessed using logistic regression models with and without GPR. Adding GPR was associated with modest changes in reclassification indices for 30-day mortality, with a continuous NRI of 0.1301 (95% CI, 0.0595–0.2007; P < 0.001) and an IDI of 0.0035 (95% CI, 0.0010–0.0060; P = 0.006). For 360-day mortality, reclassification analyses also showed modest differences after GPR was added to conventional covariates, with a continuous NRI of 0.0834 (95% CI, 0.0225–0.1444; P = 0.007) and an IDI of 0.0047 (95% CI, 0.0021–0.0073; P < 0.001) (Table 3).
Table 3.
Potential contribution to risk stratification of adding GPR to binary logistic regression models
| Outcome | Metric | Estimate | 95% CI | P value |
|---|---|---|---|---|
| 30-day mortality | Continuous NRI | 0.1301 | 0.0595–0.2007 | < 0.001 |
| IDI | 0.0035 | 0.0010–0.0060 | 0.006 | |
| 360-day mortality | Continuous NRI | 0.0834 | 0.0225–0.1444 | 0.007 |
| IDI | 0.0047 | 0.0021–0.0073 | < 0.001 |
For each outcome, potential contribution to risk stratification was assessed by comparing a reference binary logistic regression model without GPR with an extended model including GPR. Both models included the same covariates as the fully adjusted model 3
Abbreviations: CI Confidence interval, GPR Glucose-to-platelet ratio, IDI Integrated discrimination improvement, NRI Net reclassification improvement
Discussion
In this retrospective cohort of 4,676 critically ill patients with AMI, higher GPR during the first 24 h after ICU admission was independently associated with higher 30-day and 360-day all-cause mortality. After adjustment for demographics, baseline laboratory variables, illness severity, and comorbidities, each 1-unit increase in GPR was associated with a 10% higher hazard of 30-day mortality and a 12% higher hazard of 360-day mortality. Patients in the highest GPR tertile had higher mortality risk than those in the lowest tertile. Restricted cubic spline analyses suggested a nonlinear association for 30-day mortality and an approximately linear association for 360-day mortality. These findings support GPR as an associated prognostic marker, but not as evidence of a direct causal pathway.
Exploratory prediction analyses were consistent with the possibility that GPR provides incremental prognostic information beyond conventional covariates. In the validation cohort, the ridge model incorporating GPR showed higher discrimination and a lower Brier score than the Cox model (AUROC, 0.816 vs. 0.757; Brier score, 0.120 vs. 0.142). However, this comparison involved different modeling frameworks and should not be interpreted as direct evidence that GPR itself improved model performance. In conventional logistic models, adding GPR was associated with modest improvements in net reclassification improvement and integrated discrimination improvement for both outcomes. These findings support the potential incremental value of GPR, but they do not prove superiority over glucose or platelet count alone. Further external validation and prospective assessment are needed to clarify how GPR should be incorporated into risk stratification frameworks in critically ill patients with AMI.
The potential clinical value of GPR lies in combining two routine measurements rather than replacing clinical judgment. Hyperglycemia and thrombocytopenia are independently associated with adverse outcomes after AMI [23, 24]. This concept is consistent with recent ACS and STEMI studies showing that biomarkers reflecting oxidative stress, systemic inflammation, nutritional status, and hematologic reserve, including serum malondialdehyde and HALP score, are associated with in-hospital mortality [6, 7]. Hyperglycemia may promote oxidative stress, endothelial dysfunction, and prothrombotic signaling through pathways involving mitochondrial superoxide, advanced glycation end products, protein kinase C, glycocalyx injury, and PI3K/Akt signaling [25–28]. Low platelet count in ICU patients may reflect platelet activation and consumption, disseminated or localized coagulopathy, infection, liver dysfunction, or marrow suppression [29–31]. Thus, GPR should be interpreted as an integrated risk marker reflecting metabolic and hemostatic derangement, rather than as a pure marker of platelet depletion or a direct therapeutic target.
Compared with existing composite indices, GPR has a different role. SHR focuses on acute glucose relative to chronic glycemic exposure and requires HbA1c, whereas GLR and platelet-to-lymphocyte-type ratios emphasize immune-inflammatory status [15, 16, 32–34]. In the present database, HbA1c availability was insufficient for calculating SHR, limiting direct comparison. Therefore, the novelty of GPR is mainly conceptual and practical: it combines glycemic stress with platelet dynamics using measurements routinely available early in ICU care. However, exploratory comparative discrimination analyses did not demonstrate statistically significant AUROC differences between GPR and either mean glucose or platelet count alone. Therefore, the present findings do not support superior discriminative performance of GPR over its individual components. Rather, GPR should be interpreted as a simple ratio-based summary of concurrent glycemic stress and platelet-related illness severity, and as a complementary marker rather than a replacement for the separate interpretation of glucose and platelet count. The restricted cubic spline results describe risk patterns but should not be interpreted as defining a clinically actionable threshold. Additional prospective research is needed before any cutoff can guide treatment.
The observed short-term nonlinearity may be consistent with acute stress physiology, but the mechanistic interpretation remains speculative. At higher GPR levels, simultaneous hyperglycemia and platelet depletion could reflect more severe systemic inflammation, endothelial injury, microvascular dysfunction, and organ hypoperfusion [35–39]. Hyperglycemia may enhance reactive oxygen species generation, tissue factor expression, and endothelial injury [40–43], whereas excessive platelet activation and consumption may contribute to microvascular obstruction and impaired tissue perfusion [44, 45]. AMI-related ischemia-reperfusion injury, catecholamine and cortisol surges, and immunothrombotic pathways, including neutrophil extracellular traps and platelet-leukocyte aggregates, may further amplify these processes [46–50]. However, these mechanisms were not directly measured in this study and should be regarded as biological plausibility rather than conclusions drawn from the present retrospective analysis.
The choice to use mean glucose over the first 24 h was intended to reduce noise from single admission values, transient peaks, and treatment timing in the ICU. This approach reflects the real-world ICU glycemic environment, where fasting measurements are rarely available and insulin therapy may modify glucose values. However, it may also obscure clinically relevant early peaks, platelet nadirs, or later changes in stress hyperglycemia and thrombocytopenia. The exploratory subgroup analyses across insulin use, norepinephrine use, mechanical ventilation, and continuous renal replacement therapy showed no statistically significant effect modification. These subgroup findings should be interpreted as consistency checks within clinically relevant strata rather than as proof of effect homogeneity.
The present study has several limitations. First, the retrospective single-center database design may introduce selection bias, residual confounding, and limited generalizability. Excluding patients with ICU stays shorter than 24 h helped define the exposure window but may have excluded both patients who died very early and patients rapidly transferred out of ICU. Second, GPR may be a surrogate of disease severity, systemic inflammation, shock, sepsis, consumptive coagulopathy, or multiorgan dysfunction rather than an independent causal mediator. Although Model 3 adjusted for SOFA, comorbidities, and key baseline laboratory variables, unmeasured or incompletely measured factors remained, including nutritional status, steroid exposure, glycemic control strategies, inflammatory markers, antiplatelet therapy, heparin exposure, infection, liver dysfunction, and coagulation disorders. Missing-data handling may also have influenced association estimates or contributed to optimistic internally validated prediction performance. Time-dependent or landmark analyses were not feasible. Third, glucose and platelet values were summarized within the first 24 h only; subsequent changes could not be incorporated into the primary exposure definition, raising the possibility of exposure misclassification. Fasting glucose status was unavailable, and HbA1c missingness prevented calculation of SHR. In addition, exploratory AUROC comparisons did not demonstrate superior discrimination of GPR over mean glucose or platelet count alone; therefore, the incremental prognostic value of GPR beyond its individual components and established indices requires further validation. Finally, all-cause mortality was used because cause-specific mortality was unavailable in MIMIC-IV; non-cardiovascular deaths from sepsis or multiorgan failure may have diluted or biased the observed association. Further external validation is needed to clarify how GPR should be incorporated into risk stratification frameworks in critically ill patients with AMI.
In summary, elevated GPR in the first 24 h after ICU admission was independently associated with short- and long-term all-cause mortality in critically ill AMI patients. GPR may provide a simple summary of metabolic and hemostatic stress using routinely available laboratory data, but its role should remain complementary to established clinical assessment until externally validated evidence is available.
Conclusion
Among critically ill patients with AMI, elevated GPR was independently associated with increased short- and long-term mortality. GPR may serve as an accessible ratio-based risk marker that reflects the coexistence of metabolic stress and platelet-related illness severity. Its role should be considered complementary to, rather than a replacement for, established clinical assessment and the separate interpretation of glucose and platelet count. External validation is required before GPR can be incorporated into routine ICU risk stratification.
Supplementary Information
Acknowledgements
The authors sincerely thank the Massachusetts Institute of Technology (MIT), Beth Israel Deaconess Medical Center (BIDMC), and the Laboratory for Computational Physiology for maintaining and providing access to the MIMIC-IV database.
Abbreviations
- ACEI
Angiotensin-converting enzyme inhibitor
- ARB
Angiotensin receptor blocker
- BUN
Blood urea nitrogen
- CCI
Charlson Comorbidity Index
- CRRT
Continuous renal replacement therapy
- DBP
Diastolic blood pressure
- GPR
Glucose-to-platelet ratio
- ICU
Intensive care unit
- INR
International normalized ratio
- IQR
Interquartile range
- LODS
Logistic Organ Dysfunction Score
- PT
Prothrombin time
- PTT
Partial thromboplastin time
- RBC
Red blood cell count
- SBP
Systolic blood pressure
- SD
Standard deviation
- SOFA
Sequential Organ Failure Assessment
- SpO₂
Peripheral oxygen saturation
- WBC
White blood cell count
Authors’ contributions
Jie Peng designed the study and conducted data extraction. Data collection and analysis were conducted by Jie Peng and Xingzhan Zhang. Dakai Liang and Xia Luo prepared the tables and figures. Jie Peng and Ling Zhao reviewed the results, interpreted the data, and drafted the manuscript. All authors made substantial intellectual contributions to the study and approved the final version for submission.
Funding
None.
Data availability
The data supporting this study were obtained from the MIMIC-IV database, a publicly available resource accessible at [ https://mimic.physionet.org ](https://mimic.physionet.org).
Declarations
Ethics approval and consent to participate
The study adhered to the Declaration of Helsinki principles. The Institutional Review Boards of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC) approved the creation of the MIMIC-IV database. The study was retrospective, utilized de-identified data, and did not involve direct patient contact, and informed consent was waived because the data were de-identified.
Consent for publication
Not applicable. This study used de-identified data extracted from the publicly accessible MIMIC-IV database, which does not contain any personal or identifiable patient information. Therefore, consent for publication was not required.
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.Roth GA, Mensah GA, Johnson CO, Addolorato G, Ammirati E, Baddour LM, et al. Global Burden of Cardiovascular Diseases and Risk Factors, 1990–2019. J Am Coll Cardiol. 2020;76:2982–3021. 10.1016/j.jacc.2020.11.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Welt FGP, Batchelor W, Spears JR, Penna C, Pagliaro P, Ibanez B, et al. Reperfusion Injury in Patients With Acute Myocardial Infarction: JACC Scientific Statement. J Am Coll Cardiol. 2024;83:2196–213. 10.1016/j.jacc.2024.02.056. [DOI] [PubMed] [Google Scholar]
- 3.Heusch G, Gersh BJ. The pathophysiology of acute myocardial infarction and strategies of protection beyond reperfusion: a continual challenge. Eur Heart J. 2017;38:774–84. 10.1093/eurheartj/ehw224. [DOI] [PubMed] [Google Scholar]
- 4.Tiller C, Reindl M, Holzknecht M, Lechner I, Oberhollenzer F, von der Emde S, et al. Association of Intramyocardial Hemorrhage With Inflammatory Biomarkers in Patients With ST-Segment Elevation Myocardial Infarction. JACC Adv. 2025;4:101647. 10.1016/j.jacadv.2025.101647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Becirovic E, Becirovic M, Ljuca K, Becirovic A, Babic M, Ljuca N, et al. Integration of Hematologic and Metabolic Biomarkers for Outcome Prediction in Acute Coronary Syndromes Without ST Elevation. Cureus. 2025. 10.7759/cureus.86446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Karakayali M, Ogun M, Artac I, Ilis D, Arslan A, Omar T, et al. Serum malondialdehyde levels at admission as a predictor of inhospital mortality in patients with acute coronary syndrome. Coron Artery Dis. 2025;36:211–7. 10.1097/MCA.0000000000001469. [DOI] [PubMed] [Google Scholar]
- 7.Karakayali M, Omar T, Artac I, Ilis D, Arslan A, Altunova M, et al. The prognostic value of HALP score in predicting in-hospital mortality in patients with ST-elevation myocardial infarction undergoing primary percutaneous coronary intervention. Coron Artery Dis. 2023;34:483–8. 10.1097/MCA.0000000000001271. [DOI] [PubMed] [Google Scholar]
- 8.Schmitz T, Harmel E, Heier M, Peters A, Linseisen J, Meisinger C. Inflammatory plasma proteins predict short-term mortality in patients with an acute myocardial infarction. J Transl Med. 2022;20:457. 10.1186/s12967-022-03644-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Liu J, Zhou Y, Huang H, Liu R, Kang Y, Zhu T, et al. Impact of stress hyperglycemia ratio on mortality in patients with critical acute myocardial infarction: insight from american MIMIC-IV and the chinese CIN-II study. Cardiovasc Diabetol. 2023;22:281. 10.1186/s12933-023-02012-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Harrington RA, Sane DC, Califf RM, Sigmon KN, Abbottsmith CW, Candela RJ, et al. Clinical importance of thrombocytopenia occurring in the hospital phase after administration of thrombolytic therapy for acute myocardial infarction. J Am Coll Cardiol. 1994;23:891–8. 10.1016/0735-1097(94)90634-3. [DOI] [PubMed] [Google Scholar]
- 11.Nakamura T, Ako J, Kadowaki T, Funayama H, Sugawara Y, Kubo N, et al. Impact of acute hyperglycemia during primary stent implantation in patients with ST-elevation myocardial infarction. J Cardiol. 2009;53:272–7. 10.1016/j.jjcc.2008.11.011. [DOI] [PubMed] [Google Scholar]
- 12.Guenancia C, Hachet O, Stamboul K, Béjot Y, Leclercq T, Garnier F, et al. Incremental predictive value of mean platelet volume/platelet count ratio in in-hospital stroke after acute myocardial infarction. Platelets. 2017;28:54–9. 10.1080/09537104.2016.1203397. [DOI] [PubMed] [Google Scholar]
- 13.Pietrangelo D, Lopa C, Litterio M, Cotugno M, Rubattu S, Lombardi A. Metabolic Disturbances Involved in Cardiovascular Diseases: The Role of Mitochondrial Dysfunction, Altered Bioenergetics and Oxidative Stress. Int J Mol Sci. 2025;26:6791. 10.3390/ijms26146791. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Barale C, Russo I. Influence of Cardiometabolic Risk Factors on Platelet Function. Int J Mol Sci. 2020;21:623. 10.3390/ijms21020623. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Jin X, Deng W, Weng Q, Liu F, Min J. Correlation between stress hyperglycemia ratio and risk of acute respiratory distress syndrome in patients with sepsis: a MIMIC-IV cohort study. Sci Rep. 2025;15:34737. 10.1038/s41598-025-18349-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wang C, Cai Z, Xue S, Tu G. Relationship between GLR and 28-day all-cause mortality in patients with hypertensive heart disease complicated by heart failure: A retrospective analysis of the MIMIC-IV database. Med (Baltim). 2025;104:e44967. 10.1097/MD.0000000000044967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Chen L, Chen N, Lin Y, Ren H, Huang Q, Jiang X, et al. Glucose to Platelet Ratio: A Potential Predictor of Hemorrhagic Transformation in Patients with Acute Ischemic Stroke. Brain Sci. 2022;12:1170. 10.3390/brainsci12091170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–7. 10.1016/S0140-6736(07)61602-X. [DOI] [PubMed] [Google Scholar]
- 19.Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10:1. 10.1038/s41597-022-01899-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Austin PC, White IR, Lee DS, van Buuren S. Missing Data in Clinical Research: A Tutorial on Multiple Imputation. Can J Cardiol. 2021;37:1322–31. 10.1016/j.cjca.2020.11.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Blazek K, van Zwieten A, Saglimbene V, Teixeira-Pinto A. A practical guide to multiple imputation of missing data in nephrology. Kidney Int. 2021;99:68–74. 10.1016/j.kint.2020.07.035. [DOI] [PubMed] [Google Scholar]
- 22.Li W, Wang B, Li T, Ma Y, Jin H, Zhao J, et al. A Causal and interpretable machine learning framework for postcranioplasty risk prediction and surgical decision support. Npj Digit Med. 2026;9:184. 10.1038/s41746-026-02370-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Williams SB, Goldfine AB, Timimi FK, Ting HH, Roddy M-A, Simonson DC, et al. Acute Hyperglycemia Attenuates Endothelium-Dependent Vasodilation in Humans In Vivo. Circulation. 1998;97:1695–701. 10.1161/01.CIR.97.17.1695. [DOI] [PubMed] [Google Scholar]
- 24.Marfella R, Siniscalchi M, Esposito K, Sellitto A, de Fanis U, Romano C, et al. Effects of Stress Hyperglycemia on Acute Myocardial Infarction. Diabetes Care. 2003;26:3129–35. 10.2337/diacare.26.11.3129. [DOI] [PubMed] [Google Scholar]
- 25.Nishikawa T, Edelstein D, Du XL, Yamagishi S, Matsumura T, Kaneda Y, et al. Normalizing mitochondrial superoxide production blocks three pathways of hyperglycaemic damage. Nature. 2000;404:787–90. 10.1038/35008121. [DOI] [PubMed] [Google Scholar]
- 26.Patterson EK, Cepinskas G, Fraser DD. Endothelial Glycocalyx Degradation in Critical Illness and Injury. Front Med. 2022;9:898592. 10.3389/fmed.2022.898592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Nieuwdorp M, van Haeften TW, Gouverneur MCLG, Mooij HL, van Lieshout MHP, Levi M, et al. Loss of Endothelial Glycocalyx During Acute Hyperglycemia Coincides With Endothelial Dysfunction and Coagulation Activation In Vivo. Diabetes. 2006;55:480–6. 10.2337/diabetes.55.02.06.db05-1103. [DOI] [PubMed] [Google Scholar]
- 28.Varma S, Lal BK, Zheng R, Breslin JW, Saito S, Pappas PJ, et al. Hyperglycemia alters PI3k and Akt signaling and leads to endothelial cell proliferative dysfunction. Am J Physiol-Heart Circ Physiol. 2005;289:H1744–51. 10.1152/ajpheart.01088.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Gresele P, Guglielmini G, Del Pinto M, Calabrò P, Pignatelli P, Patti G, et al. Low platelet count at admission has an adverse impact on outcome in patients with acute coronary syndromes: from the START Antiplatelet registry. Sci Rep. 2024;14:14516. 10.1038/s41598-024-64113-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Li H, Tang C, Zhu X, Zhang W, Abudupataer M, Ding S, et al. Histamine deficiency facilitates coronary microthrombosis after myocardial infarction by increasing neutrophil-platelet interactions. J Cell Mol Med. 2020;24:3504–20. 10.1111/jcmm.15037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Zhang W, Jiang H, Wu G, Huang P, Wang H, An H, et al. The pathogenesis and potential therapeutic targets in sepsis. MedComm. 2023;4:e418. 10.1002/mco2.418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Pruc M, Peacock FW, Rafique Z, Swieczkowski D, Kurek K, Tomaszewska M, et al. The Prognostic Role of Platelet-to-Lymphocyte Ratio in Acute Coronary Syndromes: A Systematic Review and Meta-Analysis. J Clin Med. 2023;12:6903. 10.3390/jcm12216903. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Pu J, Zhang H, Wang F, Zhou Y, Liu D, Wang H, et al. The Glucose-to-Lymphocyte Ratio Predicts All-cause Mortality and Cardiovascular Mortality in ST-Elevation Myocardial Infarction Patients: A Retrospective Study. Rev Cardiovasc Med. 2025;26:26065. 10.31083/RCM26065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Liu J, Hu X. Association between glucose-to-lymphocyte ratio and in-hospital mortality in acute myocardial infarction patients. PLoS ONE. 2023;18:e0295602. 10.1371/journal.pone.0295602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Cai S, Wang Q, Ma C, Chen J, Wei Y, Zhang L, et al. Association between glucose-to-lymphocyte ratio and in-hospital mortality in intensive care patients with sepsis: A retrospective observational study based on Medical Information Mart for Intensive Care IV. Front Med. 2022;9. 10.3389/fmed.2022.922280. [DOI] [PMC free article] [PubMed]
- 36.Li L, Ding L, Zheng L, Wu L, Hu Z, Liu L, et al. U-shaped association between stress hyperglycemia ratio and risk of all-cause mortality in cardiac ICU. Diabetes Metab Syndr. 2024;18:102932. 10.1016/j.dsx.2023.102932. [DOI] [PubMed] [Google Scholar]
- 37.Tang Y, Chen Q, Liang B, Peng B, Wang M, Sun J, et al. A retrospective cohort study on the association between early coagulation disorder and short-term all-cause mortality of critically ill patients with congestive heart failure. Front Cardiovasc Med. 2022;9. 10.3389/fcvm.2022.999391. [DOI] [PMC free article] [PubMed]
- 38.Anand SS, Yi Q, Gerstein H, Lonn E, Jacobs R, Vuksan V, et al. Relationship of Metabolic Syndrome and Fibrinolytic Dysfunction to Cardiovascular Disease. Circulation. 2003;108:420–5. 10.1161/01.CIR.0000080884.27358.49. [DOI] [PubMed] [Google Scholar]
- 39.Bulger DA, Zhang Z, Hu R, Dave EK, Mehta PK, Griendling KK, et al. Redox Regulation of Microvascular Physiology and Pathophysiology: Insights into Therapeutic Strategies and Limitations. Antioxid Redox Signal. 2025;43:566–621. 10.1177/15230864251372607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kaur R, Kaur M, Singh J. Endothelial dysfunction and platelet hyperactivity in type 2 diabetes mellitus: molecular insights and therapeutic strategies. Cardiovasc Diabetol. 2018;17:121. 10.1186/s12933-018-0763-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Witkowski M, Witkowski M, Saffarzadeh M, Friebel J, Tabaraie T, Ta Bao L, et al. Vascular miR-181b controls tissue factor-dependent thrombogenicity and inflammation in type 2 diabetes. Cardiovasc Diabetol. 2020;19:20. 10.1186/s12933-020-0993-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Undas A, Wiek I, Stêpień E, Zmudka K, Tracz W. Hyperglycemia Is Associated With Enhanced Thrombin Formation, Platelet Activation, and Fibrin Clot Resistance to Lysis in Patients With Acute Coronary Syndrome. Diabetes Care. 2008;31:1590–5. 10.2337/dc08-0282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Safdar NZ, Kietsiriroje N, Ajjan RA. The Cellular and Protein Arms of Coagulation in Diabetes: Established and Potential Targets for the Reduction of Thrombotic Risk. Int J Mol Sci. 2023;24:15328. 10.3390/ijms242015328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Chaudhary PK, Kim S, Kim S. An Insight into Recent Advances on Platelet Function in Health and Disease. Int J Mol Sci. 2022;23:6022. 10.3390/ijms23116022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Badulescu O-V, Ciocoiu M, Vladeanu MC, Huzum B, Plesoianu CE, Iliescu-Halitchi D, et al. The Role of Platelet Dysfunctions in the Pathogenesis of the Hemostatic-Coagulant System Imbalances. Int J Mol Sci. 2025;26:2756. 10.3390/ijms26062756. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Algoet M, Janssens S, Himmelreich U, Gsell W, Pusovnik M, Van den Eynde J, et al. Myocardial ischemia-reperfusion injury and the influence of inflammation. Trends Cardiovasc Med. 2023;33:357–66. 10.1016/j.tcm.2022.02.005. [DOI] [PubMed] [Google Scholar]
- 47.Alkatiri AH, Qalby N, Mappangara I, Zainal ATF, Cramer MJ, Doevendans PA, et al. Stress hyperglycemia and poor outcomes in patients with ST-elevation myocardial infarction: a systematic review and meta-analysis. Front Cardiovasc Med. 2024;11. 10.3389/fcvm.2024.1303685. [DOI] [PMC free article] [PubMed]
- 48.Mutua V, Gershwin LJ. A Review of Neutrophil Extracellular Traps (NETs) in Disease: Potential Anti-NETs Therapeutics. Clin Rev Allergy Immunol. 2021;61:194–211. 10.1007/s12016-020-08804-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Li X, Weber NC, Cohn DM, Hollmann MW, DeVries JH, Hermanides J, et al. Effects of Hyperglycemia and Diabetes Mellitus on Coagulation and Hemostasis. J Clin Med. 2021;10:2419. 10.3390/jcm10112419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Gauer JS, Ajjan RA, Ariëns RAS. Platelet–Neutrophil Interaction and Thromboinflammation in Diabetes: Considerations for Novel Therapeutic Approaches. J Am Heart Assoc. 2022;11:e027071. 10.1161/JAHA.122.027071. [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 data supporting this study were obtained from the MIMIC-IV database, a publicly available resource accessible at [ https://mimic.physionet.org ](https://mimic.physionet.org).