Stress hyperglycemia ratio and machine learning model for prediction of all-cause mortality in critically ill patients with acute kidney injury: A cohort study from MIMIC-IV

Abstract

Objectives

Acute kidney injury (AKI) is marked by a rapid decline in renal function, often identified by elevated serum creatinine or reduced urine output. Although stress hyperglycemia ratio (SHR) has been linked to adverse outcomes in various conditions, its association with clinical prognosis in AKI patients remains unclear.

Methods

This cohort study analyzed data from critically ill patients with AKI extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) database. The primary outcomes were 28-day and 365-day all-cause mortality, while the secondary outcomes included ICU mortality and in-hospital mortality. The association between SHR and all-cause mortality was explored by Cox proportional hazards regression. The discriminative performance of SHR was evaluated through the Boruta feature selection model, followed by the development of a prognostic prediction model utilizing advanced machine learning.

Results

The analysis encompassed 3640 patients with AKI. Multivariable Cox regression analysis demonstrated that elevated SHR significantly predicted increased 28-day mortality [adjusted hazard ratio (HR) 1.19, 95% confidence interval (CI): 1.11–1.29, P < .001, Model 3] and 365-day mortality (HR: 1.17, 95% CI: 1.08∼1.27, P < .001, Model 3). Upon categorizing SHR into quartiles, individuals in the highest quartile (Q4) faced a substantially elevated risk, with a 101% greater likelihood of 28-day death and a 34% elevated hazard of 365-day death relative to those in the lowest quartile (Q1). Boruta feature selection analysis identified SHR as a significant predictor. Among various predictive models evaluated, the CatBoost classifier exhibited the most robust discriminative performance for 28-day and 365-day mortality, achieving an area under the receiver operating characteristic curve of 0.83 and 0.82, respectively, showing comparable discriminative performance to the other models.

Conclusion

The SHR demonstrated a nonlinear association all-cause mortality among critically ill patients AKI, suggesting its potential utility as a reliable prognostic indicator for predicting unfavorable results in AKI patients.

Graphical abstract

This is a visual representation of the abstract.

Keywords

Stress hyperglycemia glucose mortality acute kidney injury MIMIC-IV

Introduction

Acute kidney injury (AKI), defined by a sudden deterioration in renal function, is clinically diagnosed through increased serum creatinine (SCr) concentrations or diminished urine output.¹ This prevalent condition occurs in approximately 20%-50% of individuals requiring intensive care unit (ICU) treatment.^2,3 AKI is significantly linked to a range of unfavorable clinical outcomes, such as prolonged hospitalization durations, increased requirement for kidney replacement therapy (KRT), substantial healthcare expenditures, and elevated mortality rates.^4–7

The stress hyperglycemia ratio (SHR), combining hemoglobin A1c (HbA1c) and blood glucose, effectively reflect acute stress-induced hyperglycemia levels.⁸ The SHR index has been widely recognized as a reliable indicator of stress-induced hyperglycemia and serves as a valuable prognostic marker for assessing disease severity in critically ill patients.⁹ Previous studies have demonstrated that elevated SHR was associated with an increased risk of mortality and adverse outcomes in patients with cardiogenic shock, atrial fibrillation, chronic heart failure, sepsis, and sepsis-associated AKI.^10–14 Li et al.¹⁵ reported a U-shaped relationship between SHR and AKI in patients with chronic heart failure. An et al.¹⁶ also found a shaped re” relationship between SHR and one-year mortality in patients with chronic kidney disease (CKD). Among diabetes mellitus patients, SHR is a strong indicator of both immediate and long-term prognosis.¹⁷ However, the relationship between SHR and short-term and long-term mortality in AKI patients was unknown.

Given the emerging evidence linking the SHR to adverse outcomes in various clinical settings, this study sought to examine its association on critically ill patients with AKI. We hypothesized that elevated SHR levels are associated with an increased risk of mortality in critically ill patients with AKI.

Methods

Source of database

The data for this study were obtained from Medical Information Mart for Intensive Care IV (MIMIC-IV version 3.1) database (https://mimic.mit.edu/).¹⁸ The database includes a variety of details on 94,458 admissions of critically ill patients admitted to Beth Israel Deaconess Medical Center in Boston during 2008 to 2022.¹⁹ The database encompasses a comprehensive array of data elements, including survival outcomes, physiological parameters, diagnostic information, lab test results, and treatment methods. Due to the MIMIC database's comprehensive and high-quality dataset as well as other advantages, it is increasingly being utilized by researchers for academic studies.^20–22 One of the authors, Yingxiu Huang was granted access to the database after successfully completing an online training course and examination (Certificate ID: 56513391). As the dataset ensures patient privacy, consent for participation was not required. The work described was conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) and reported following the REporting of studies Conducted using Observational Routinely-collected Data (RECORD) guideline.

AKI diagnosis and AKI stage

The diagnosis of AKI following the Kidney Disease Improving Global Outcomes (KDIGO) criteria¹: (a) an elevation of serum creatinine (sCr) levels by at least 50% within a seven-day period, (b) an escalation in sCr serum creatinine (sCr) of ≥0.3 mg/dL within 48 h, or (c) a persistent reduction in urine output to below 0.5 milliliters per kilogram per hour sustained over a period exceeding six hours. AKI Stage 1 was characterized by a maximum-to-baseline change of ≥0.3 mg/dL or a ratio between 1.5 and 1.9 or urine output <0.5 mL/kg/h for 6–12 h. Stage 2 was characterized by a ratio of 2.0–2.9 or urine output <0.5 mL/kg/h for ≥12 h. Stage 3 included a ratio ≥3.0, serum creatinine (sCr) ≥ 4.0 mg/dL, or urine output <0.3 mL/kg/h for ≥24 h or anuria for ≥12 h or initiation of KRT.²³

Inclusion and exclusion criteria

Adult patients (clusion criteria serum creatinine (sCr)ine output <0.5kilogram per hour sustained over a period exceeding six houPatients were excluded if they met the following states: without AKI diagnosis, missing baseline data of glucose and HbA1c. Patients with repeated ICU admissions will be filtered to retain only the first admission record.

Exposure

The primary exposure of interest is the SHR. The SHR is determined as the index derived from admission glucose and HbA1c within the first 24 h of ICU admission, according to the formula below^24,25:

\begin{aligned} SHR & = [(admission blood glucose (mg / dl)) / \\ \times (28 .7 \times HbA 1 c (%) - 46 .7)] \end{aligned} .

The SHR was evaluated in two ways: as a continuous measure and as a categorical indicator. When it was a categorical variable, SHR was divided into four groups according to the quartile division of SHR: Q1 group <0.881, Q2 group 0.881–1.062, Q3 group 1.063–1.318, Q4 group >1.318.

Outcome

The primary outcomes were 28-day and 365-day all-cause mortality, defined as death from any cause within 28 days and 365 days after ICU admission, respectively. The secondary outcomes were in-hospital and ICU all-cause mortality, defined as death occurring during the index hospitalization and during the ICU stay, respectively. The primary outcomes were analyzed from ICU admission as the time zero. To mitigate potential immortal time bias due to AKI occurring after ICU admission, a supplementary 7-day landmark analysis was performed, in which only patients who developed AKI within the first 7 ICU days and survived to the landmark time were included.

Data extraction

Data was extracted from PostgreSQL using structured query language (SQL).²⁶ Demographic data such as age, sex, and race were gathered. Details on pre-existing conditions, including type 2 diabetes, chronic kidney disease, severe liver disease, heart failure, malignancy, congestive heart failure (CHF), cerebrovascular disease, and chronic pulmonary disease, were identified using International Classification of Diseases (ICD) codes (Supplementary Table S1), and sepsis was defined according to the Sepsis-3 criteria by the presence of documented or suspected infection and an acute increase in the SOFA score of ≥2 points. Vital signs, including heart rate, mean blood pressure (MBP), respiratory rate, and peripheral oxygen saturation (SpO₂), along with laboratory results such as serum glucose, HbA1c, hemoglobin, white blood cell count (WBC), platelets, electrolytes (sodium, potassium, chloride), anion gap, bicarbonate, pH, international normalized ratio (INR), prothrombin time (PT), blood urea nitrogen (BUN), and serum creatinine, were collected. The first 24-h urine output was also collected. Additionally, the severity of illness was evaluated using several scoring systems, including the Sequential Organ Failure Assessment (SOFA), Acute Physiology Score III (APS III), Charlson Comorbidity Index (CCI), Oxford Acute Severity of Illness Score (OASIS), and Simplified Acute Physiology Score II (SAPS II). Intervention included KRT, vasoactive use, and ventilation. Baseline data collection window is shown in Supplementary Figure S1.

Statistical analysis

Patients were divided into four groups according to their SHR quartiles. Descriptive statistics were used in the summary of the attributes. For data with a normal distribution, continuous variables were presented as mean and standard deviation (SD); for data with a skewed distribution, they were presented as median and interquartile range (IQR). Proportions (percentages) were used to display categorical variables. To analyze baseline characteristics, we compared data using the chi-square test for categorical data and the Kruskal‒Wallis test for continuous factors. Patients with missing baseline glucose or HbA1c data were excluded from the study. For covariates with missing data rates below 20%, missing values were imputed using the K-Nearest Neighbors (KNN) method.²⁷

We employed multivariable Cox proportional hazard regression models to evaluate the independent relationship between SHR and outcome, yielding HRs and 95% CIs for risk assessment. The proportional hazards assumption was tested using residual-based methods. Accordingly, hazard ratios from the Cox models were interpreted as average relative effects over the follow-up period. A refined Cox regression model was applied, which were adjusted for multiple variables. Covariates were chosen based on their established clinical relevance and prior evidence. The reference group was defined as the lowest quartile of SHR. Model 1: Unadjusted. In Model 2, the analysis was adjusted for demographic factors including age, race, and sex. In Model 3, the analysis accounted for age, sex, race, diabetes, CHF, severe liver disease, sepsis, chronic kidney disease (CKD), chronic pulmonary disease, cancer, KRT, ventilation, vasoactive drug use, SOFA score, WBC, MBP, platelets, and INR. Trend tests were carried out, wherein the median values of each quartile of SHR were incorporated as continuous variables. Additionally, survival curves were produced by Kaplan‒Meier analysis and compared using log-rank analysis. Restricted cubic splines (RCS) were employed to examine the potential non-linear relationship between the SHR and clinical outcomes. Subgroup analysis assessed SHR on mortality varied by subgroup. The secondary outcome was adjusted for confounders in Model 3. Sensitivity analyses were conducted by excluding patients with ICU stays shorter than 24 h, and by using AKI onset within the first 2 ICU days to further assess the robustness of the findings. The predictive performance of the SHR, SOFA, and SAPSII was compared using Receiver Operating Characteristic (ROC) curve analysis.

The feature selection process was conducted using the Boruta algorithm, a machine learning-based approach that ranks variables according to their relative importance in predicting all-cause mortality. To avoid data leakage, Boruta was applied exclusively to the training set. Statistically significant predictors identified through this process were subsequently incorporated into predictive modeling. The dataset was randomly split into 70% training and 30% testing sets. Five distinct machine learning algorithms - logistic regression (LR), random forest (RF), gradient boosting classifier (GBC), LightGBM, and CatBoost - were systematically implemented to evaluate 28-day and 365-day all-cause mortality risk in AKI patients. Default hyperparameter settings were used for all models, without additional tuning. Model descriptive performance was quantitatively assessed through receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) calculations. All analyses were performed with a fixed random seed of 123 to ensure reproducibility.

Free Statistics software version 2.0²⁸ and R statistical software version 4.2.2 (http://www.R-project.org, The R Foundation) were used for all studies. A statistically significant test was defined as a two-tailed test with a significance level of p < .05.

Result

Baseline characteristics of participants

The cohort study population comprised 3640 critical AKI patients selected from the MIMIC-IV v3.1 dataset (Figure 1). The cohort included 58% male participants with a mean age of 68 years. During the 28-day follow-up period, 609 deaths occurred, with an overall mortality rate reaching 16.7%. Within 365 days, the mortality rate reached 27.6%, with 1005 patients died. The detailed baseline characteristics of the study population are summarized in Table 1.

Figure 1.

Flow chart of the study.

Table 1.

Baseline characteristics of patients with AKI.

Characteristics		SHR				P value
Characteristics	Total(n = 3640)	Q1 (<0.881)(n = 910)	Q2 (0.881‒1.062)(n = 910)	Q3 (1.063‒1.318)(n = 910)	Q4 (>1.318)(n = 910)	P value
Sex, male, n (%)	2119 (58.2)	535 (58.8)	547 (60.1)	513 (56.4)	524 (57.6)	.41
Race, white, n (%)	2007 (55.1)	494 (54.3)	505 (55.5)	535 (58.8)	473 (52)	.031
Age (years)	68.1 ± 15.0	68.2 ± 15.1	67.6 ± 15.2	68.9 ± 14.8	67.9 ± 14.7	.271
Heart rate (beats/min)	101.2 ± 20.4	98.3 ± 20.0	99.0 ± 19.4	101.0 ± 19.3	106.4 ± 21.7	<.001
MBP (mmHg)	85.0 ± 12.2	86.1 ± 12.4	87.1 ± 12.3	85.0 ± 12.3	82.0 ± 11.1	<.001
Respiratory rate (beats/min)	27.8 ± 6.1	27.1 ± 5.7	27.6 ± 6.1	27.9 ± 6.1	28.7 ± 6.3	<.001
Temperature (°C)	37.4 ± 0.7	37.3 ± 0.7	37.4 ± 0.6	37.4 ± 0.6	37.4 ± 0.8	.079
Spo₂ (%)	91.8 ± 5.5	91.8 ± 5.1	92.0 ± 5.1	92.0 ± 5.1	91.6 ± 6.4	.241
HbA1c (%)	6.5 ± 1.9	7.2 ± 2.3	6.2 ± 1.5	6.3 ± 1.7	6.5 ± 1.9	<.001
Blood glucose (mg/dL)	164.0 ± 99.3	114.9 ± 43.8	127.0 ± 42.1	156.7 ± 57.6	257.2 ± 141.2	<.001
SHR	1.2 ± 0.5	0.7 ± 0.1	1.0 ± 0.1	1.2 ± 0.1	1.8 ± 0.7	<.001
Hemoglobin (g/dL)	11.4 ± 2.4	11.6 ± 2.3	11.8 ± 2.3	11.5 ± 2.3	10.9 ± 2.5	<.001
Platelets(10⁹/L)	203.4 ± 81.6	209.3 ± 85.4	207.2 ± 77.9	201.5 ± 76.1	195.5 ± 86.0	.001
WBC (10⁹/L)	13.3 ± 7.9	11.5 ± 9.0	12.2 ± 6.2	13.5 ± 8.0	16.0 ± 7.2	<.001
Anion gap(mmol/L)	16.4 ± 4.9	15.7 ± 4.5	15.4 ± 3.9	15.9 ± 4.2	18.6 ± 6.2	<.001
Bicarbonate (mmol/L)	24.1 ± 3.8	24.5 ± 3.8	24.2 ± 3.3	24.1 ± 3.7	23.5 ± 4.0	<.001
BUN (mg/dL)	20.0 (15.0, 30.0)	19.0 (14.0, 28.0)	18.0 (13.2, 25.8)	19.0 (15.0, 29.0)	25.0 (17.0, 38.0)	<.001
Calcium (mmol/L)	8.9 ± 0.7	8.9 ± 0.8	8.9 ± 0.6	8.9 ± 0.7	8.8 ± 0.8	.024
Chloride (mmol/L)	105.2 ± 6.0	105.4 ± 5.8	104.9 ± 5.3	105.0 ± 5.4	105.6 ± 7.4	.04
Creatinine (mg/dL)	1.1 (0.8, 1.5)	1.0 (0.8, 1.4)	1.0 (0.8, 1.3)	1.0 (0.8, 1.4)	1.2 (0.9, 2.0)	<.001
Sodium (mmol/L)	140.3 ± 4.9	140.4 ± 4.5	140.0 ± 4.2	140.0 ± 4.4	140.7 ± 6.3	.006
Potassium(mmol/L)	4.6 ± 0.9	4.5 ± 0.9	4.4 ± 0.8	4.5 ± 0.9	4.8 ± 0.9	<.001
INR	1.4 ± 0.8	1.3 ± 0.6	1.3 ± 0.5	1.4 ± 0.8	1.6 ± 1.1	<.001
PT	15.3 ± 8.6	14.7 ± 6.1	14.5 ± 5.5	15.1 ± 7.3	17.0 ± 13.1	<.001
PH	7.3 ± 0.1	7.3 ± 0.1	7.3 ± 0.1	7.3 ± 0.1	7.3 ± 0.1	<.001
24-h urine output (ml)	1425.0 (900.0, 2125.0)	1425.0 (900.0, 2187.5)	1475.0 (963.8, 2150.0)	1400.0 (925.0, 1991.0)	1396.0 (793.8, 2196.2)	.082
Myocardial infarct, n (%)	1048 (28.8)	230 (25.3)	240 (26.4)	259 (28.5)	319 (35.1)	< .001
Congestive heart failure, n (%)	1155 (31.7)	284 (31.2)	230 (25.3)	276 (30.3)	365 (40.1)	< .001
Cerebrovascular disease, n (%)	1875 (51.5)	503 (55.3)	531 (58.4)	499 (54.8)	342 (37.6)	< .001
Chronic pulmonary disease, n (%)	680 (18.7)	172 (18.9)	155 (17)	155 (17)	198 (21.8)	.03
CKD, n (%)	733 (20.1)	192 (21.1)	140 (15.4)	168 (18.5)	233 (25.6)	< .001
Malignant cancer, n (%)	218 (6.0)	55 (6)	32 (3.5)	68 (7.5)	63 (6.9)	.002
Diabetes, n (%)	1385 (38.0)	407 (44.7)	246 (27)	306 (33.6)	426 (46.8)	< .001
Severe liver disease, n (%)	76 (2.1)	5 (0.5)	13 (1.4)	18 (2)	40 (4.4)	< .001
Sepsis, n (%)	1611 (44.3)	356 (39.1)	337 (37)	384 (42.2)	534 (58.7)	< .001
CCI	6.4 ± 2.7	6.5 ± 2.9	6.1 ± 2.6	6.5 ± 2.6	6.6 ± 2.7	<.001
SOFA	3.0 (2.0, 5.0)	2.0 (1.0, 4.0)	2.0 (1.0, 4.0)	3.0 (2.0, 5.0)	4.0 (2.0, 7.0)	<.001
APSIII	46.3 ± 22.7	41.7 ± 20.4	41.1 ± 18.9	44.7 ± 20.8	57.5 ± 26.0	<.001
SAPSII	34.4 ± 12.7	32.1 ± 11.7	32.1 ± 11.8	34.0 ± 11.6	39.5 ± 14.1	<.001
OASIS	33.2 ± 8.9	31.4 ± 8.5	31.9 ± 8.3	33.1 ± 8.7	36.2 ± 9.3	<.001
AKI stage, n (%)					0.003
1	1013 (27.8)	258 (28.4)	276 (30.3)	244 (26.8)	235 (25.8)
2	1881 (51.7)	491 (54)	458 (50.3)	484 (53.2)	448 (49.2)
3	746 (20.5)	161 (17.7)	176 (19.3)	182 (20)	227 (24.9)

KRT, n (%)	65 (2.6)	11 (1.7)	8 (1.2)	15 (2.4)	31 (4.8)	<.001
Vasoactive use, n (%)	540 (21.2)	111 (17.4)	113 (17.6)	127 (20.3)	189 (29.3)	<.001
Ventilation, n (%)	836 (32.8)	164 (25.7)	182 (28.4)	197 (31.5)	293 (45.4)	< .001
Outcome
28-day mortality, n (%)	609 (16.7)	91 (10)	118 (13)	149 (16.4)	251 (27.6)	< .001
ICU-mortality, n (%)	339 (9.3)	47 (5.2)	56 (6.2)	70 (7.7)	166 (18.2)	<.001
In-hospital mortality, n (%)	471 (12.9)	66 (7.3)	90 (9.9)	101 (11.1)	214 (23.5)	< .001
365-day mortality, n (%)	1005 (27.6)	213 (23.4)	201 (22.1)	247 (27.1)	344 (37.8)	< .001

Note: SHR: stress hyperglycemia ratio; WBC: white blood cell; BUN: blood urea nitrogen; PH: potential of Hydrogen; INR: international normalized ratio; PT: prothrombin time; HbA1c: glycosylated hemoglobin; CKD: chronic kidney disease; SOFA: sequential organ failure assessment; CCI: charlson comorbidity index; APS III: acute physiology score III; SAPS II: simplified acute physiology score II; OASIS: oxford acute severity of illness score; KRT: kidney replacement therapy.

Kaplan‒Meier survival curve analysis

The Kaplan‒Meier (KM) curve demonstrated that the 28-day and 365-day cumulative survival rates were significantly lower in the Q4 group compared to the Q1 groups, indicating a clear disparity in survival outcomes among the quartiles (P < .0001) (Figure 2A and 2B).

Figure 2.

Kaplan‒Meier survival analysis curves for all-cause mortality: (A) 28-day mortality; (B) 365-day mortality.

Landmark analyses

To mitigate immortal time bias, we conducted a landmark analyses at 7 days post-ICU admission. Survival curves before and after the landmark were plotted according to SHR quartiles. As illustrated in Figure 3, the highest SHR quartile (Q4) was associated with the most rapid decline in survival by day 28 (Figure 3A) and, consistently, with the lowest cumulative survival over 365 days (Figure 3B).

Figure 3.

Landmark survival analysis from day 7 for (A) 28-day and (B) 365-day mortality, according to SHR quartiles.

The association between SHR and 28-day and 365-day mortality in patients with AKI

Three models were employed to assess the relationship between SHR and primary outcomes. In the unadjusted model (Model 1), each unit increase in SHR was associated with a 33% higher risk of 28-day mortality (HR: 1.33, 95% CI: 1.26–1.4; P.4; 001) and a 17% higher risk of 365-day mortality (HR = 1.17, 95% CI: 1.08–1.1.27; P.1.2001) (Table 2). After controlling for confounding variables in Model 2, SHR showed a 34% increase in 28-day mortality (HR = 1.34, 95% CI: 1.27–1.41; P.34,001) and a 33% rise in 365-day mortality (HR = 1.33, 95% CI: 1.26–1.40; P.33,001). After accounting for all potential confounders in Model 3, SHR remained significantly associated with 28-day mortality (HR = 1.19, 95% CI: 1.11–1.29, P.19,001) and 365-day mortality (HR = 1.17, 95% CI: 1.08–1.27, P.17,001). When SHR was categorized into quartiles, the Q4 group exhibited a markedly elevated risk of 28-day (HR = 2.01, 95% CI: 1.51–2.68, P.01,001) and 365-day mortality (Model 3: HR = 1.34, 95% CI: 1.09–1.65, P = .006) compared to the Q1 group. The P-values for trend test were both less than .05 in all three models.

Table 2.

Cox proportional hazard models for 28-day and 365-day all-cause mortality.

Variable	Model 1		Model 2		Model 3
Variable	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
28-day all-cause mortality
SHR	1.33 (1.26∼1.4)	<.001	1.34 (1.27∼1.41)	<.001	1.19 (1.11∼1.29)	<.001
SHR quartile
Q1	1(Ref)		1(Ref)		1(Ref)
Q2	1.32 (1∼1.73)	.049	1.36 (1.03∼1.79)	.028	1.1 (0.79∼1.53)	.563
Q3	1.7 (1.31∼2.21)	<.001	1.71 (1.32∼2.22)	<.001	1.4 (1.03∼1.9)	.032
Q4	3.11 (2.44∼3.95)	<.001	3.11 (2.45∼3.96)	<.001	2.01 (1.51∼2.68)	<.001
P for trend		<.001		<.001		<.001
365-day all-cause mortality
SHR	1.31 (1.24∼1.38)	<.001	1.33 (1.26∼1.4)	<.001	1.17 (1.08∼1.27)	<.001
SHR quartile
Q1	1(Ref)		1(Ref)		1(Ref)
Q2	0.95 (0.79∼1.15)	.617	0.98 (0.81∼1.19)	.859	0.89 (0.71∼1.13)	.339
Q3	1.22 (1.01∼1.46)	.037	1.21 (1.01∼1.45)	.042	1.09 (0.87∼1.35)	.464
Q4	1.88 (1.59∼2.23)	<.001	1.9 (1.6∼2.26)	<.001	1.34 (1.09∼1.65)	.006
P for trend		<.001		<.001		.001

Note: Model 1: Unadjusted. Model 2: Adjusted for sex, age, race. Model 3: Adjusted further for diabetes, CHF, sepsis, CKD, chronic pulmonary disease, severe liver disease, malignant cancer, SOFA score, KKRT, ventilation, vasoactive drug use, WBC, MBP, platelets, and INR.

The non-linear association between SHR and 28-day and 365-day mortality

The restricted cubic spline (RCS) analysis, adjusted for covariates including age, sex, race, diabetes, CHF, sepsis, CKD, chronic pulmonary disease, cancer, severe liver disease, KKRT, SOFA score, ventilation, vasoactive drug use, WBC, MBP, platelets, and INR, revealed a non-linear relationship linking SHR with primary outcomes. Specifically, P-values indicating non-linearity were 0.014 for 28-day mortality (Figure 4A) and 0.015 for 365-day mortality (Figure 4B). The RCS curve inflection points were identified at approximately SHR ≈ 0.92 for 28-day mortality and SHR ≈ 0.75 for 365-day mortality, indicating a threshold effect in the SHR-mortality relationship (Table 3).

Table 3.

The threshold effect of SHR on 28-day and 365-day all-cause mortality.

Inflection point (SHR)	Breakpoint.HR (95% CI)	P value
28-day all-cause mortality
<0.92	0.39 (0.072, 2.124)	.276
≥0.92	1.51 (1.259, 1.811)	< .001
Non-linear test		.029
365-day all-cause mortality
<0.75	0.29 (0.027, 2.09)	.196
≥0.75	1.41 (1.215, 1.626)	<.001
Non-linear test		.012

Note: Adjusted for age, race, sex, diabetes, CHF, sepsis, CKD, chronic pulmonary disease, severe liver disease, malignant cancer, SOFA score, KRTKRT, ventilation, vasoactive drug use, WBC, MBP, platelets, and INR. 99.8% data was shown.

Subgroup analyses

A comprehensive subgroup analysis was conducted, stratifying by age (<65 years and ≥65 years), sex, race, diabetes, chronic pulmonary disease, CHF, and myocardial infarct, to evaluate the robustness of the association between SHR and all-cause mortality at 28 and 365 days. The results presented in Figure 5, demonstrated a consistent correlation between SHR and mortality within different subgroups. Interaction tests revealed no significant interactions between subgroups except age (P for interaction > .05).

Figure 4.

RCS analysis of SHR with all-cause mortality: (A) 28-day mortality; (B) 365-day mortality. Adjusted for age, race, sex, diabetes, CHF, sepsis, CKD, chronic pulmonary disease, severe liver disease, malignant cancer, SOFA score, KRT, ventilation, vasoactive drug use, WBC, MBP, platelets, and INR. 99.8% data was displayed. 99.8% of data was shown.

Sensitivity analysis

To assess the robustness of our primary findings, sensitivity analyses were conducted by excluding patients with ICU stays shorter than 24 h, and by using AKI onset within the first 2 ICU days. The consistency between the sensitivity analyses and the primary analysis supports the robustness of our findings (Table 4).

Table 4.

Sensitivity analysis of SHR on 28-day and 365-day mortality excluding patients with ICU stay less than 24 h and AKI onset within the first 2 ICU days.

Variable	Crude model		Adjusted model
Variable	HR (95%CI)	P value	HR (95%CI)	P value
Excluding patients with an ICU stay of less than 24 h
28-day mortality
SHR continuous	1.31 (1.25∼1.39)	<.001	1.19 (1.1∼1.28)	<.001
SHR quartile
Q1	1(Ref)		1(Ref)
Q2	1.3 (0.98∼1.71)	.068	1.12 (0.8∼1.57)	.498
Q3	1.65 (1.27∼2.16)	<.001	1.42 (1.04∼1.94)	.029
Q4	2.96 (2.31∼3.78)	<.001	1.99 (1.49∼2.67)	<.001
P for trend		<.001		<.001
365-day mortality
SHR continuous	1.29 (1.22∼1.37)	<.001	1.17 (1.07∼1.27)	<.001
SHR quartile
Q1	1(Ref)		1(Ref)
Q2	0.94 (0.77∼1.14)	.54	0.9 (0.71∼1.14)	.372
Q3	1.17 (0.97∼1.41)	.108	1.08 (0.86∼1.35)	.498
Q4	1.81 (1.52∼2.15)	<.001	1.33 (1.07∼1.65)	.009
P for trend		<.001		.002
AKI onset 2days with ICU admission
28-day mortality
SHR continuous	1.33 (1.26∼1.39)	<.001	1.2 (1.11∼1.29)	<.001
SHR quartile	SHR quartiles
Q1	1(Ref)		1(Ref)
Q2	1.26 (0.93∼1.7)	.136	1.1 (0.76∼1.6)	.613
Q3	1.92 (1.46∼2.53)	<.001	1.75 (1.25∼2.44)	.001
Q4	3.28 (2.53∼4.23)	<.001	2.33 (1.7∼3.18)	<.001
P for trend		<.001		<.001
365-day mortality
SHR continuous	1.31 (1.25∼1.38)	<.001	1.18 (1.1∼1.28)	<.001
SHR quartile	SHR quartiles
Q1	1(Ref)		1(Ref)
Q2	0.94 (0.76∼1.16)	.542	0.89 (0.68∼1.15)	.368
Q3	1.31 (1.07∼1.59)	.007	1.21 (0.96∼1.54)	.111
Q4	1.99 (1.66∼2.39)	<.001	1.48 (1.18∼1.86)	.001
P for trend		<.001		<.001

Note: Adjusted for age, race, sex, diabetes, CHF, sepsis, CKD, chronic pulmonary disease, severe liver disease, malignant cancer, SOFA score, KRT, ventilation, vasoactive drug use, WBC, MBP, platelets, and INR.

Secondary outcome

KM curve analysis showed that in-hospital and ICU all-cause mortality were significantly greater in Q4 group compared to Q1 group (both P < .001). The association between SHR and both in-hospital and ICU all-cause mortality were assessed using Cox proportional hazard analysis. The findings revealed that elevated SHR levels correlated with higher in-hospital mortality (adjusted HR = 1.17, 95% CI: 1.07–1.27, P.17,001) and ICU-mortality (adjusted HR = 1.18, 95% CI: 1.08–1.29, P.18,001) (Table 5). RCS analysis further revealed a non-linear relationship linking SHR to mortality during hospitalization and in the ICU.

Table 5.

Secondary outcome.

Outcome	Crude model		Adjusted model
Outcome	HR (95% CI)	P value	HR (95% CI)	P value
In-hospital mortality
SHR continuous	1.34 (1.27∼1.41)	<.001	1.17 (1.07∼1.27)	<.001
SHR quartile
Q1	1(Ref)		1(Ref)
Q2	1.38 (1.01∼1.9)	.046	1.3 (0.89∼1.9)	.168
Q3	1.59 (1.16∼2.16)	.004	1.27 (0.88∼1.85)	.205
Q4	3.62 (2.75∼4.77)	<.001	2.32 (1.66∼3.23)	<.001
P for trend		<.001		<.001
ICU mortality
SHR continuous	1.36 (1.28∼1.43)	<.001	1.18 (1.08∼1.29)	<.001
SHR quartile
Q1	1(Ref)		1(Ref)
Q2	1.2 (0.82∼1.77)	.348	1.08 (0.67∼1.72)	.754
Q3	1.54 (1.06∼2.22)	.023	1.22 (0.78∼1.9)	.39
Q4	3.89 (2.81∼5.37)	<.001	2.33 (1.58∼3.44)	<.001
P for trend		<.001		<.001

Receiver operating characteristic analysis

To evaluate the discriminative performance of SHR for both short- and long-term mortality, we conducted ROC analyses for 28-day and 365-day outcomes (Figure 6). For 28-day mortality (Figure 6A ), SHR demonstrated moderate discriminative ability with an AUC of 0.633 (95% CI: 0.607–0.658), whereas SOFA (95% CI: 0.669–0.714) and SAPS II (95% CI: 0.74–0.78) showed higher discriminative performance with AUCs of 0.692 and 0.760, respectively. For 365-day mortality (Figure 6B), SHR exhibited a lower AUC of 0.579 (95% CI: 0.557–0.60), while SOFA and SAPS II maintained higher predictive capacities with AUCs of 0.666 (95% CI: 0.647–0.685) and 0.755 (95% CI: 0.739–0.772), respectively.

Figure 5.

Subgroup analysis forest plot for all-cause mortality: (A) 28-day mortality; (B) 365-day mortality.

Figure 6.

ROC curve analysis of SHR, SOFA, and SAPSII. (A) 28-day mortality. (B) 365-day mortality. SHR: stress hyperglycemia ratio; SOFA: sequential organ failure assessment; SAPSII: simplified acute physiology score II; AUC: area under the curve.

Figure 7.

Importance of potential risk factors of 28-day mortality ranked by Boruta algorithm. The horizontal axis is the name of each variable, and the vertical axis is the Z value of each variable. The box plot shows the Z value of each variable during model calculation. The green boxes represent important variables, the red boxes represent unimportant variables, and the yellow boxes represent potentially important variables. CHF: congestive heart failure; MI: myocardial infarction; KRT: kidney replacement therapy; BUN: blood urea nitrogen; SHR: stress hyperglycemia ratio; WBC: white blood cell; INR: international normalized ratio; SOFA: sequential organ failure assessment; CCI: charlson comorbidity index; APS III: acute physiology score III; SAPS II: simplified acute physiology score II; OASIS: Oxford acute severity of illness score; KRT: kidney replacement therapy.

Boruta algorithm

The Boruta algorithm-based feature selection results are shown in Figure 7. In the Boruta method, variables in the red area are deemed irrelevant, while those in the green area are identified as critical features.

Machine learning

The discriminative performance of five machine learning models—Random Forest (RF), Gradient Boosting Classifier (GBC), Logistic Regression (LR), Light Gradient Boosting Machine (LightGBM), and Categorical Boosting (CatBoost)—was evaluated for predicting 28-day and 365-day mortality. Among them, the CatBoost classifier achieved the highest AUC values of 0.83 and 0.82 for 28-day and 365-day death, respectively, showing comparable discriminative performance to the other models (Figure 8A and 8B). Comparable performance was seen across additional metrics, including accuracy (0.83 and 0.79), recall (0.45 and 0.48), precision (0.48 and 0.66), F1-score (0.47 and 0.56), Brier score (0.17 and 0.21), and Matthews correlation coefficient (MCC; 0.36 and 0.43) (Supplementary Table S1).

Figure 8.

The machine learning algorithm predicts in-hospital and 360-day all-cause mortality. AUC: Area under the curve; LR: Logistic regression; RF: Random forest; GBC: Gradient boosting classifier; LightGBM: Light gradient boosting machine; and CatBoost: Categorical boosting.

Discussion

Our retrospective cohort analysis revealed a non-linear relationship between SHR and clinical outcomes, including 28-day, in-hospital, ICU, and 365-day all-cause mortality, among critically ill patients with AKI. To our knowledge, this is the first study to examine the relationship between SHR levels and both short- and long-term clinical outcomes in patients with AKI.

Stress hyperglycemia, reflecting inflammatory and neuro-hormonal disturbances during acute conditions or critical illness, is a more reliable biomarker than admission glucose levels or fasting blood glucose.^15,16 The SHR was introduced by Roberts et al. to quantify the stress hyperglycemia response.²⁹ Stress-induced hyperglycemia is common in critically ill patients and is associated with increased risk of all-cause mortality.⁹ Previous studies have shown that SHR was independently associated with adverse outcomes in various conditions, including cardiogenic shock, atrial fibrillation, chronic heart failure, and sepsis.^10–13 For instance, Xia et al.³⁰ reported that SHR was independently associated with ICU mortality in AKI patients from MIMIC III database. Similarly, Zhou et al.¹⁴ found that higher SHR correlated strongly with 30-day and 1-year mortality in 1161 patients with sepsis-associated AKI. Cao et al.³¹ also demonstrated a J-shaped relationship between SHR and all-cause mortality, and a U-shaped link with cardiovascular mortality in CKD patients. Collectively, these studies suggest that increased SHR index correlate with greater risks of unfavorable clinical outcomes. Our study adds further evidence supporting this association in critical patients with AKI.

Our ROC analyses demonstrated that SHR has moderate discriminative ability for both short- and long-term mortality, with AUCs of 0.633 and 0.579 for 28-day and 365-day mortality, respectively, compared with higher AUCs for SOFA (0.692 and 0.666) and SAPS II (0.760 and 0.755). Although SHR alone is less predictive than these established scoring systems, it offers several practical advantages. Calculation of SHR requires only two routinely available laboratory values—admission glucose and HbA1c—making it feasible for rapid bedside assessment. This simplicity is particularly valuable in resource-limited settings or when immediate severity scoring is not available. Therefore, SHR could serve as a complementary marker alongside conventional scores like SOFA, providing additional prognostic information to improve early risk stratification and guide clinical decision-making in critically ill patients.

The influence of SHR on clinical outcomes may be mediated through various biological pathways. Elevated SHR levels could drive adverse effects via mechanisms including insulin resistance, oxidative stress, and inflammatory reactions.³² Under stress conditions, the neuroendocrine system activates, increasing sympathetic activity and levels of catecholamines, glucocorticoids, inflammatory cytokines, and glucagon.³³ These alterations may trigger insulin resistance by accelerated breakdown of hepatic glycogen and enhanced gluconeogenesis. Furthermore, increased SHR may amplify the systemic inflammatory response by stimulating the renin–angiotensin–aldosterone system (RAAS) and inducing the release of pro-inflammatory cytokines, such as interleukin-1, interleukin-6, and tumor necrosis factor-α. These cytokines can further injure the kidneys.³⁴ Additionally hyperglycemia stimulates the production of plasminogen activator inhibitor type 1 (PAI-1). Elevated PAI-1 levels not only reduce local fibrinolysis but also increase the risk of thrombus formation, thereby exacerbating the deterioration of renal function.^35,36 While these mechanisms are plausible and supported by prior studies, our findings only demonstrate an association between SHR and clinical outcomes in AKI; causal pathways remain to be investigated in future studies.

To address the potential risk of immortal time bias arising from the time-dependent nature of AKI and stress hyperglycemia assessment, we performed landmark analyses as a robustness evaluation. The results of the landmark analyses were consistent with those of the primary analysis, with similar direction and magnitude of the association between SHR and all-cause mortality in critically ill patients with AKI. Therefore, the landmark analyses did not materially alter the interpretation of the study outcomes but rather reinforced the robustness of our main findings.

This study has several advantages. First, it leverages a large-scale sample from the high-quality MIMIC-IV dataset, a real-world dataset. Second, we employed diverse statistical methods, including RCS analysis, KM analysis, subgroup analysis, sensitivity analysis, Cox regression analysis, the Boruta algorithm, and machine learning, to comprehensively investigate and validate the association between SHR and clinical prognosis.

Limitations

However, the research also has limitations. First, its retrospective nature may introduce selection bias. Second, patients with missing baseline glucose or HbA1c data were excluded, which may have introduced additional selection bias and could limit the representativeness of our study population. Third, we only used SHR data on the first day of ICU admission, limiting our ability to assess SHR variations and potentially affecting the precision of our result. Fourth, as an observational study, we could not confirm the mechanism linking higher SHR levels to AKI prognosis. Fifth, another limitation of our study is that it was conducted solely using the MIMIC-IV database, which may limit generalizability. External validation of SHR in independent or prospective cohorts, ideally spanning multiple centers and diverse populations, is warranted to confirm its prognostic value across varied clinical settings. Finally, landmark analyses were conducted to reduce immortal time bias and evaluate the robustness of the primary findings. However, these analyses may introduce depletion of susceptible, as patients who experience early events are excluded, potentially attenuating the estimated effect size. Therefore, while landmark analyses strengthen the confidence in our results, their inherent selection process should be considered when interpreting the findings.

Conclusion

In this study, SHR emerged as a non-invasive prognostic marker for both short-term and long-term mortality in patients with AKI. These findings suggest that SHR may be useful for risk stratification and for identifying patients at higher risk of adverse outcomes. Further prospective studies with larger sample sizes are needed to confirm these findings and explore the clinical utility of SHR in improving AKI outcomes.

Supplemental Material

sj-docx-1-dhj-10.1177_20552076261421140 - Supplemental material for Stress hyperglycemia ratio and machine learning model for prediction of all-cause mortality in critically ill patients with acute kidney injury: A cohort study from MIMIC-IV

Supplemental material, sj-docx-1-dhj-10.1177_20552076261421140 for Stress hyperglycemia ratio and machine learning model for prediction of all-cause mortality in critically ill patients with acute kidney injury: A cohort study from MIMIC-IV by Yingxiu Huang, Ting Ao, Ming Hu and Peng Zhen in DIGITAL HEALTH

Footnotes

Abbreviations

Acknowledgments

The authors thank Dr Qilin Yang of Department of Critical Care, The Second Affiliated Hospital of Guangzhou Medical University and the Physician Scientist Team for guidance on data extraction and analysis.

ORCID iDs

Yingxiu Huang

Ting Ao

Ming Hu

Peng Zhen

Ethics approval and consent to participate

MIMIC-IV database used in the present study was approved by the Institutional Review Boards (IRB) of Institutional Review Boards of Beth Israel Deaconess Medical Center (2001-P-001699/14) and the Massachusetts Institute of Technology (No. 0403000206). The individual information of the participants included in this database was anonymous, and informed consent was waived.

Author contribution statement

Yingxiu Huang designed the protocol and wrote the manuscript. Ting Ao extracted, collected, and analyzed data. Ming Hu prepared tables and figures. Peng Zhen designed the protocol, wrote the manuscript, reviewed, and edited the manuscript. All authors reviewed the manuscript and all authors have approved to the final version of the manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

The entirety of the raw data employed in this study originates from the MIMIC IV (version 3.1) database, a publicly accessible repository. The specific dataset accessed for this investigation can be retrieved from the designated URL: .

Supplemental material

Supplemental material for this article is available online.

References

Khwaja

. KDIGO Clinical practice guidelines for acute kidney injury. Nephron Clin Pract 2012; 120: c179–c184.

Havaldar

Sushmitha

EAC

Shrouf

, et al. Epidemiological study of hospital acquired acute kidney injury in critically ill and its effect on the survival. Sci Rep 2024; 14: 28129.

Alba Schmidt

De Rosa

Müller

, et al. Acute kidney injury predicts mortality in very elderly critically-ill patients. Eur J Intern Med 2024; 127: 119–125.

Hoste

EAJ

Bagshaw

Bellomo

, et al. Epidemiology of acute kidney injury in critically ill patients: the multinational AKI-EPI study. Intensive Care Med 2015; 41: 1411–1423.

Chen

J-J

Lee

T-H

Chan

M-J

, et al. Electronic alert systems for patients with acute kidney injury: a systematic review and meta-analysis. JAMA Netw Open 2024; 7: e2430401.

Gameiro

Marques

Lopes

. Long-term consequences of acute kidney injury: a narrative review. Clin Kidney J 2021; 14: 789–804.

Lee

C-C

Kuo

Chan

M-J

, et al. Characteristics of and outcomes after dialysis-treated acute kidney injury, 2009-2018: a Taiwanese multicenter study. Am J Kidney Dis 2023; 81: 665–674.e1.

Yang

Kim

T-H

Yoon

K-H

, et al. The stress hyperglycemia ratio, an index of relative hyperglycemia, as a predictor of clinical outcomes after percutaneous coronary intervention. Int J Cardiol 2017; 241: 57–63.

Zhang

Shen

H-C

Liang

W-R

, et al. Relationship between stress hyperglycemia ratio and allcause mortality in critically ill patients: results from the MIMIC-IV database. Front Endocrinol 2023; 14: 1111026.

10.

Yan

Chen

Quan

, et al. Association between the stress hyperglycemia ratio and 28-day all-cause mortality in critically ill patients with sepsis: a retrospective cohort study and predictive model establishment based on machine learning. Cardiovasc Diabetol 2024; 23: 63.

11.

H-M

Zheng

S-W

Xie

Y-Y

, et al. Simultaneous assessment of stress hyperglycemia ratio and glycemic variability to predict mortality in patients with coronary artery disease: a retrospective cohort study from the MIMIC-IV database. Cardiovasc Diabetol 2024; 23: 61.

12.

Xie

Chen

, et al. The prognostic impact of stress hyperglycemia ratio on mortality in cardiogenic shock: a MIMIC-IV database analysis. Diabetol Metab Syndr 2024; 16: 12.

13.

Cheng

Shen

Han

, et al. Association between stress hyperglycemia ratio index and all-cause mortality in critically ill patients with atrial fibrillation: a retrospective study using the MIMIC-IV database. Cardiovasc Diabetol 2024; 23: 63.

14.

Zhou

Zhong

, et al. The association between stress hyperglycemia ratio and clinical outcomes in patients with sepsis-associated acute kidney injury: a secondary analysis of the MIMIC-IV database. BMC Infect Dis 2024; 24: 1263.

15.

Ding

Zheng

, et al. Relationship between stress hyperglycemia ratio and acute kidney injury in patients with congestive heart failure. Cardiovasc Diabetol 2024; 23: 29.

16.

Che

, et al. Predictive value of stress hyperglycemia ratio on one-year mortality in chronic kidney disease patients admitted to intensive care unit. BMC Nephrol 2024; 25: 58.

17.

Lai

Meng

Zhou

, et al. Association of stress hyperglycemia ratio with presence and severity of chronic kidney disease among US adults with diabetes mellitus. Front Endocrinol 2024; 15: 1446390.

18.

Johnson

AEW

Bulgarelli

Shen

, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data 2023; 10: 1. DOI: https://doi.org/10.1038/s41597-022-01899-x.

19.

Johnson

Bulgarelli

Pollard

, et al. Mimic-iv. DOI: 10.13026/HXP0-HG59.

20.

Zhao

Zhang

, et al. The association between triglyceride glucose-body mass index and all-cause mortality in critically ill patients with atrial fibrillation: a retrospective study from MIMIC-IV database. Cardiovasc Diabetol 2024; 23: 64. DOI: https://doi.org/10.1186/s12933-024-02153-x.

21.

Wang

Gao

Dantona

, et al. DRG-LLaMA : tuning LLaMA model to predict diagnosis-related group for hospitalized patients. NPJ Digit Med 2024; 7: 16.

22.

Liu

X-Y

Xiao

, et al. Association between the ROX index and mortality in patients with acute hypoxemic respiratory failure: a retrospective cohort study. Respir Res 2024; 25: 43.

23.

Khwaja

. KDIGO Clinical practice guidelines for acute kidney injury. Nephron Clin Pract 2012; 120: c179–c184.

24.

Ding

Zhang

Dai

, et al. The prognostic value of the stress hyperglycemia ratio for all-cause and cardiovascular mortality in patients with diabetes or prediabetes: insights from NHANES 2005–2018. Cardiovasc Diabetol 2024; 23: 84.

25.

Qiao

Ruan

, et al. Stress hyperglycemia ratio as an independent predictor of acute kidney injury in critically ill patients with acute myocardial infarction: a retrospective U.S. Cohort study. Ren Fail 2025; 47: 2471018.

26.

Zhang

. Obesity and 1-year all-cause survival of adult intensive care patients with heart failure: data from the MIMIC-IV. Diabetol Metab Syndr 2024; 16: 90.

27.

El-Menyar

Naduvilekandy

Asim

, et al. Machine learning models predict triage levels, massive transfusion protocol activation, and mortality in trauma utilizing patients hemodynamics on admission. Comput Biol Med 2024; 179: 108880.

28.

Huang

Zhen

, et al. Association between serum bicarbonate and 28-day mortality in critically ill patients with infective endocarditis: a cohort study from MIMIC-IV. Sci Rep 2025; 15: 12.

29.

Roberts

Quinn

Valentine

, et al. Relative hyperglycemia, a marker of critical illness: introducing the stress hyperglycemia ratio. J Clin Endocrinol Metab 2015; 100: 4490–4497.

30.

Xia

Kuang

, et al. Predictive value of glycemic gap and stress glycemia ratio among critically ill patients with acute kidney injury: a retrospective analysis of the MIMIC-III database. BMC Nephrol 2023; 24: 27.

31.

Cao

Guo

D-T

, et al. The association between stress-induced hyperglycemia ratio and cardiovascular events as well as all-cause mortality in patients with chronic kidney disease and diabetic nephropathy. Cardiovasc Diabetol 2025; 24: 55.

32.

Pan

Xiong

Huang

, et al. Association between stress hyperglycemia ratio with short-term and long-term mortality in critically ill patients with ischemic stroke. Acta Diabetol 2024; 61: 859–868.

33.

Scheen

Giraud

Bendjelid

. Stress hyperglycemia, cardiac glucotoxicity, and critically ill patient outcomes current clinical and pathophysiological evidence. Physiol Rep 2021; 9: e14713.

34.

Stentz

Umpierrez

Cuervo

, et al. Proinflammatory cytokines, markers of cardiovascular risks, oxidative stress, and lipid peroxidation in patients with hyperglycemic crises. Diabetes 2004; 53: 2079–2086.

35.

Rikitake

Liao

. Rho-kinase mediates hyperglycemia-induced plasminogen activator inhibitor-1 expression in vascular endothelial cells. Circulation 2005; 111: 3261–3268.

36.

Iwasaki

Okamoto

Kato

, et al. High glucose induces plasminogen activator inhibitor-1 expression through rho/rho-kinase-mediated NF-kappaB activation in bovine aortic endothelial cells. Atherosclerosis 2008; 196: 22–28.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.05 MB