Abstract
Background
Hyperglycaemia during acute ischaemic stroke (AIS) has been associated with poorer outcomes yet attempts to normalise glucose have not demonstrated clinical benefit. One proposed explanation is that the participants are not routinely stratified by the pre-stroke glycaemic status (diabetic hyperglycaemia [DHG], non-diabetic hyperglycaemia [NDHG]), which may contribute to heterogeneity in metabolic responses and dilute treatment effects. We compared early endocrine and inflammatory responses (insulin, C-peptide, cortisol, and cytokines) between DHG and NDHG and examined exploratory associations with stroke severity and 90-day functional outcomes.
Methods
In this prospective multicentre cohort, 80 patients with AIS and hyperglycaemia (DHG or NDHG) were enrolled within 72 hours of admission. Biomarkers were measured on admission. Group comparisons were performed by diabetes status and stroke severity. Exploratory associations with 90-day functional outcome were assessed with LASSO regression.
Results
Compared with DHG, NDHG was associated with higher insulin (142 vs 75 pmol/L;p<0.05), C-peptide (1.15 vs 0.87 nmol/L;p<0.05), IL-8 (23 vs 13 pg/mL;p<0.01), and IL-6 (8 vs 3 pg/mL;p=0.094). Cortisol was numerically higher in DHG than NDHG (352 vs 271 nmol/L;p=0.32). Within NDHG, endocrine and inflammatory measures varied by stroke severity, while remaining generally higher than DHG.
Conclusion
Diabetes status was associated with differences in early endocrine and inflammatory profiles during AIS-associated hyperglycaemia. Given the observational design, single-time-point sampling, and modest sample size, these findings should be interpreted as hypothesis-generating. Future adequately powered studies with standardised sampling and detailed diabetes phenotyping are needed to determine whether diabetes-stratified approaches to hyperglycaemia management and prognostic modelling improve clinical utility.
Keywords: Acute Ischemic Stroke, Stress Hyperglycaemia, Metabolic Phenotyping, Precision Medicine, Inflammatory Biomarkers, Type 2 Diabetes Mellitus
INTRODUCTION
Acute ischaemic stroke (AIS) accounts for approximately 87% of all strokes worldwide (1) and 79.2% of cases in Malaysia (2). Hyperglycaemia is commonly observed during the acute phase of AIS, occurring in patients with established diabetes as well as in those without known diabetes (3).
Reactive hyperglycaemia after AIS is driven primarily by stress-mediated neuroendocrine activation and lesion-specific autonomic dysregulation. Activation of the hypothalamic-pituitary-adrenal (HPA) axis and the sympathetic nervous system increases cortisol and catecholamine levels, promoting hepatic gluconeogenesis, glycogenolysis, peripheral insulin resistance, and relative insulin deficiency (4-6). In parallel, ischemic injury to autonomic regulatory regions, particularly the right insular cortex, can directly induce sympathoadrenal overactivity and hyperglycaemia independent of infarct size or prior glycaemic status (7, 8). The resulting hyperglycaemia further worsens cerebral injury by enhancing oxidative stress, inflammation, endothelial dysfunction, impaired fibrinolysis, and anaerobic lactic acidosis, thereby amplifying metabolic and cellular damage in the ischemic penumbra (9-11).
C-peptide is independently associated with cardiovascular and all-cause mortality in patients with vascular disease, as well as markers of endothelial dysfunction and atherosclerosis severity (12). Cortisol levels are consistently elevated in the first week after stroke and represent a key mediator of the stress response (4, 5). Higher cortisol levels are associated with increased dependency, delirium, depression, and mortality after stroke (5). In multivariate analysis adjusting for stroke severity, cortisol remains an independent predictor of poor functional outcome and mortality (4,13). Plasma cortisol >13.6 μg/dL (≈ 375.36 nmol/L) has been identified as a risk factor for poor 90-day prognosis (6).
In total, acute hyperglycaemia in AIS has been associated with poorer clinical outcomes; however, clinical trials of intensive glucose normalisation in the acute setting have not consistently demonstrated benefit (3,14). One proposed explanation is that many studies have not stratified participants by their pre-stroke glycaemic status to those who have diabetes and nondiabetics (15), despite the possible differences in glucose regulation and insulin dynamics (16). They may present with similar degrees of hyperglycaemia, bur their outcomes and underlying metabolic context may differ (17, 18).
To reflect this heterogeneity, AIS-associated hyperglycaemia is sometimes conceptualised as two presentations: diabetic hyperglycaemia (DHG), referring to acute hyperglycaemia in patients with established diabetes (or meeting predefined criteria for chronic dysglycaemia), and non-diabetic hyperglycaemia (NDHG), referring to acute hyperglycaemia in patients without known diabetes and without evidence of chronic hyperglycaemia based on pre-stroke glycaemic assessment (15, 16). In T2DM, hyperglycaemia often reflects chronic insulin resistance with variable impairment of endogenous insulin secretion, whereas in patients without diabetes, acute hyperglycaemia may more strongly reflect a stress response with neuroendocrine activation and inflammatory signalling (16, 19).
On this basis, it has been hypothesised that metabolic and inflammatory responses, and potentially the most appropriate glucose targets, may differ between diabetic hyperglycaemia (DHG) and non-diabetic hyperglycaemia (NDHG) (15, 20). However, comparative early endocrine and inflammatory profiles between DHG and NDHG during AIS incompletely characterised. The objective of this study was to characterise early endocrine and inflammatory responses during AIS-associated hyperglycaemia, stratified by pre-stroke glycaemic status (DHG vs. NDHG), and to evaluate their associations with stroke severity and 90-day functional outcomes, irrespective of revascularisation therapy. Specifically, we aimed to: (i) compare early insulin, C-peptide, cortisol, and cytokine profiles between DHG and NDHG; (ii) examine associations between these biomarkers and stroke severity (NIHSS); and (iii) explore associations with 90-day functional outcome (mRS). We hypothesised that NDHG would be characterised by relatively preserved endogenous insulin secretion (higher insulin and C-peptide) and a distinct inflammatory profile compared with DHG; that greater stroke severity would be associated with higher inflammatory markers and lower endogenous insulin-related measures; and that early endocrine/inflammatory measures would be associated with 90-day functional outcome.
METHODS
Study design and setting
This prospective, multicentre observational cohort study was conducted at Hospital Universiti Sains Malaysia (HUSM), Kelantan, and Hospital Seberang Jaya (HSJ), Penang, from 15 Feb 2022 to 14 Feb 2023. Reporting followed the STROBE statement for observational studies. Ethical approval was obtained from the Medical Research & Ethics Committee, Ministry of Health Malaysia (NMRR-20-2730-57378 [IIR]) and the Human Research Ethics Committee of Universiti Sains Malaysia (USM/JEPeM/21110737). Written informed consent was obtained from all participants or their authorised representatives.
Participants
Inclusion Criteria
Adults (≥18 years) were eligible if they had imaging-confirmed acute ischaemic stroke (AIS; CT and/or MRI) and hyperglycaemia documented within 72 hours of admission, irrespective of receipt of revascularisation therapy (intravenous thrombolysis and/or mechanical thrombectomy).
For classification, we used the first recorded plasma glucose value within 72 hours of admission (typically the admission glucose). Hyperglycaemia classification was operationalised as follows:
NDHG: first recorded plasma glucose within 72 hours (typically admission) ≥8.3 mmol/L (3), with HbA1c <6.5% (48 mmol/mol) (21), no prior diabetes diagnosis, and no use of glucose-lowering medications.
DHG: first recorded plasma glucose within 72 hours (typically admission) ≥10 mmol/L (3,22,23), with HbA1c ≥6.5% (48 mmol/mol) (21) and/or documented medical diagnosis of diabetes and/or current oral hypoglycaemic therapy.
Exclusion Criteria
Participants were excluded if they had severe renal impairment (eGFR ≤15 mL/min/1.73 m²), pregnancy or breastfeeding, conditions that may invalidate HbA1c interpretation (e.g., haemolytic anaemia, acute blood loss, or recent transfusion) (24), prior type 1 diabetes mellitus or secondary diabetes, or (for DHG) receipt of insulin within 8 hours prior to any other short-acting (regular) insulin dosing (time may differ depending on the formulation).
Sample Size
A priori sample size calculation was performed using G*Power (version 3.1.9.7). An effect size f of 0.42, α= 0.05, and a power of 80%, indicated a minimum requirement of 68 participants for the Analysis of Variance (ANOVA) to detect between-group differences. This requirement afforded the sufficient power for both parametric and non-parametric comparisons, as well as the detection of moderate bivariate correlations (ρ= 0.33). Furthermore, this sample size satisfies the dimensionality requirements for the multivariable LASSO regression analysis, allowing for effective variable selection while minimizing the risk of overfitting.
Blood sampling and laboratory procedures
Venous blood (approximately 5 mL) was collected within 72 hours of admission at the first documented hyperglycaemic episode and prior to initiation of exogenous insulin therapy. Samples were allowed to clot at room temperature for 30-60 minutes and centrifuged at 3500 rpm; serum was aliquoted and stored at -20°C until analysis. Insulin, C-peptide, and cortisol were measured using electrochemiluminescence immunoassay (ECLIA; Cobas e601). Plasma glucose was measured using the glucose oxidase method (Architect c8000). IL-6 and IL-8 were quantified using ELISA (Elabscience®) according to manufacturer protocols.
Clinical variables and operational definitions
Baseline demographics, comorbidities, and medication history were extracted from medical records and standardised case report forms.
Body mass index (BMI) was classified according to the Malaysian Clinical Practice Guidelines: <18.5 kg/m2 as underweight, 18.5–22.9 kg/m2 as normal, 23.0–27.4 kg/m2 as overweight, and ≥27.5 kg/m2 as obese (25). It was reported by a certified hospital dietitian as part of routine nutritional assessment.
Stroke severity at presentation was assessed using the National Institutes of Health Stroke Scale (NIHSS), where scores of 0 indicate no symptoms, 1–4 minor stroke, 5–15 moderate stroke, 16–20 moderate-to-severe stroke, and 21–42 severe stroke (26). In this study, the minor stroke group included one asymptomatic patient, while the moderate group included six patients from the more severe groups (NIHSS:16–20 (n=2), and ≥21 (n=4)). Stroke aetiology and vascular territory were documented, including the Trial of Org 10172 in Acute Stroke Treatment (TOAST) and the Oxfordshire Community Stroke Project (OCSP) classifications.
Stroke functional outcomes at 90-days were categorized using the modified Rankin Scale (mRS) as favourable outcomes (physically independent) (mRS ≤2) or unfovourable outcomes (disable) (mRS>2), with mortality defined as mRS = 6 (27).The modified Rankin Scale (mRS) at admission and reassessed at approximately 90 days post-discharge via clinic follow-up.
The stress hyperglycaemia ratio (SHR) was calculated as admission glucose (mmol/L) divided by estimated chronic glucose using the formula: SHR = glucose on admission / (1.59 × HbA1c [%] − 2.59) (28).
Insulin resistance was estimated using the triglyceride-glucose (TyG) index, computed as TyG = ln[(fasting triglycerides [mg/dL] × fasting glucose [mg/dL]) / 2], where higher values indicate greater insulin resistance (29).
Bias Minimization
To reduce measurement variability, samples were collected prior to insulin initiation and processed using standardised laboratory procedures. NIHSS and mRS were assessed by study investigators using consistent scoring procedures.
Statistical analysis
All statistical analyses were conducted using R (version 4.5). Continuous baseline characteristics were summarized using means and standard deviation (SD) when distributional assumptions were met and as medians and interquartile ranges (IQR) otherwise; categorical baseline variables were presented as frequencies with percentages. For continuous variables, normality was assessed within each subgroup using the Shapiro-Wilk test, and homogeneity of variances was assessed using Levene’s test. For comparisons between two groups, Student’s t-test was used for normal variables with equal variances; Welch’s t-test was used when variances were unequal; and the Mann-Whitney U test was used when normality was not satisfied. For comparisons across more than two groups, one-way ANOVA was used for normal variables with homogeneous variances, Welch’s one-way test for unequal variances, and the Kruskal-Wallis test for non-normal distributions. Post-hoc pairwise comparisons between subgroups were performed, and p-values were adjusted using the Benjamini-Hochberg (BH) procedure to control the false discovery rate (FDR). Adjusted p-values < 0.05 were considered statistically significant. Pearson’s chi-square or Fisher’s exact test were used for categorical variables. Missing biomarker data were handled using multiple imputation by chained equations with predictive mean matching (five imputations). Insulin, C-peptide, and glucose had no missing values; cortisol had eight missing values (10%), IL-6, and IL-8 had seven missing values (8.75%). The primary 90-day outcome analysis were imputed with the Last Observation Carried Forward (LOCF) method from NIHSS and mRS scores on admission (30,31).
To explore associations with unfavourable 90-day outcome (mRS >2), a penalised logistic regression model with LASSO regularisation was fitted. Candidate variables included demographics, vascular risk factors, diabetes-related characteristics, inflammatory and metabolic biomarkers, and stroke classifications (OCSP and TOAST). The penalty parameter (λ) was selected by 5-fold cross-validation; predictors with non-zero coefficients at λmin were retained. Model robustness was examined using 1000 bootstrap replications to estimate variable selection stability. All tests were two-tailed with p < 0.05.
RESULTS
Demographic description
Patient recruitment and follow-up are summarised in the study flow chart (Fig. 1), the total number of patients who met the inclusion criteria (n=82). Two patients were excluded for having type-1 diabetes (baseline C-peptide < 0.2 nmol/L). Eighty patients with AIS-associated hyperglycaemia were recruited and included in the analysis, including 25 with NDHG and 55 with DHG. Fifty-eight patients completed the follow-up at approximately 90 days post-discharge (seventy-seven patients treated with mechanical thrombectomy, two with thrombolysis and one untreated patient with constituted stroke). As shown in Table 1, baseline characteristics were generally comparable between groups, with no statistically significant differences in sex, ethnicity, BMI, or body weight. Patients aged from 41 to 87 years with mean of 61 years and SD of 10 years. Patients with NDHG was older than those with DHG (mean 65 vs. 59 years; p = 0.02). A higher proportion of patients with NDHG were active smokers (44% vs. 22%), whereas the prevalence of hypertension (76% vs. 84%) and hyperlipidaemia (40% vs. 49%) was similar between groups. Stroke severity on admission differed between groups: patients with NDHG had higher NIHSS scores than those with DHG (median NIHSS 6 vs. 3; p = 0.046). Stroke territory did not differ significantly, with lacunar infarcts predominating in both groups (72% in NDHG vs. 80% in DHG) and posterior circulation involvement numerically higher in NDHG (28% vs. 20%). Stroke aetiology was also comparable, with large artery atherosclerosis being the most frequent subtype (36% in NDHG vs. 42% in DHG), followed by undetermined and cardioembolic causes, without significant group differences (p = 0.32).
Figure 1.
A Flowchart of the study patient recruitment and follow-up process.
Table 1.
Patient’s Characteristics Based on Diabetes Status (NDHG vs. DHG)
| Variable | Overall n = 80 | NDHG n = 25 | DHG n = 55 | p-value | |
|---|---|---|---|---|---|
| Demographics | Age (y) | 61 (10) | 65 (11) | 59 (10) | 0.02 |
| Gender | 0.52 | ||||
| Female | 31 (39%) | 11 (44%) | 20 (36%) | ||
| Male | 49 (61%) | 14 (56%) | 35 (64%) | ||
| Race | 0.55 | ||||
| Malay | 77 (96%) | 25 (100%) | 52 (95%) | ||
| Chinese | 3 (3.8%) | 0 (0%) | 3 (5.5%) | ||
| Weight (kg) | 65 (60-75) | 65 (58-72) | 65 (60-77) | 0.73 | |
| Height (cm) | 160 (153-166) | 159 (153-163) | 160 (153-168) | 0.42 | |
| BMI | |||||
| Lean | 18 (23%) | 5 (20%) | 13 (24%) | 0.89 | |
| Overweight | 36 (45%) | 12 (48%) | 24 (44%) | ||
| Obese | 26 (33%) | 8 (32%) | 18 (33%) | ||
| Medical History | Smoking Status | 0.14 | |||
| Healthy | 49 (60%) | 12 (48%) | 38 (65%) | ||
| Former | 9 (11%) | 2 (8.0%) | 7 (13%) | ||
| Active | 23 (29%) | 11 (44%) | 12 (22%) | ||
| Hypertension | 0.54 | ||||
| Yes | 65 (81%) | 19 (76%) | 46 (84%) | ||
| No | 10 (19%) | 6 (24%) | 4 (16%) | ||
| Hyperlipidemia | 0.45 | ||||
| Yes | 37 (46%) | 10 (40%) | 27 (49%) | ||
| No | 43 (54%) | 15 (60%) | 28 (51%) | ||
| DM Disease Progression | <0.001 | ||||
| No | 25 (31%) | 25 (100%) | 0 (0%) | ||
| Newly diagnosed | 11 (14%) | 0 (0%) | 11 (20%) | ||
| Former | 44 (55%) | 0 (0%) | 44 (80%) | ||
| OHA former user | <0.001 | ||||
| Yes | 31 (39%) | 0 (0%) | 31 (56%) | ||
| No | 49 (64%) | 25 (100%) | 24 (46%) | ||
| Stroke Characteristics | Stroke Event | 0.85 | |||
| Recurrent | 17 (21%) | 5 (20%) | 12 (22%) | ||
| First time | 63 (79%) | 20 (80%) | 43 (78%) | ||
| Stroke Severity | |||||
| NIHSS | 4.0 (3.0-7.0) | 6.0 (3.0-12.0) | 3.0 (2.0-7.0) | 0.046 | |
| MRS | 0.88 | ||||
| > 2 | 33 (41%) | 10 (40%) | 23 (42%) | ||
| ≤ 2 | 47 (59%) | 15 (60%) | 23 (58%) | ||
| Blockage Location | 0.56 | ||||
| LACI | 62 (78%) | 18 (72%) | 44 (80%) | ||
| POCI | 18 (22.5%) | 7 (28 %) | 11 (20%) | ||
| Stroke Subtype | 0.32 | ||||
| SVD | 12 (15%) | 2 (8.0%) | 10 (18%) | ||
| LAA | 32 (40%) | 9 (36%) | 23 (42%) | ||
| Undetermined | 22 (28%) | 7 (28%) | 15 (27%) | ||
| CE | 14 (18%) | 7 (28%) | 7 (13%) | ||
| SHR | 1.11 (0.92-1.25) | 1.24 (1.12-1.35) | 1.02 (0.83-1.16) | <0.001 | |
| Laboratory Parameters | IL-6 (pg/mL) | 4 (2-11) | 8 (4-11) | 3 (2-11) | 0.094 |
| IL-8 (pg/mL) | 15 (11-25) | 23 (16-35) | 13 (10-19) | 0.002 | |
| Cortisol (nmol/L) | 323 (182-503) | 271 (149-451) | 352 (197-530) | 0.25 | |
| eGFR (mL/min/1.73m2) | 74 (54-92) | 68 (57-85) | 76 (52-95) | 0.82 | |
| CrCl (mL/min) | 68 (26) | 63 (20) | 70 (28) | 0.32 | |
| Albumin (g/L) | 41.0 (40.0-44.0) | 43.0 (41.0-44.0) | 41.0 (39.0-44.0) | 0.37 | |
| WBC ×10⁹/L | 9.6 (8.0-11.6) | 9.1 (8.0-10.4) | 10.1 (8.0-12.0) | 0.14 | |
| Cholesterol (mmol/L) | 5.88 (1.32) | 5.84 (1.48) | 5.90 (1.25) | 0.85 | |
| Triglycerides (mmol/L) | 1.36 (1.11-1.66) | 1.27 (1.09-1.65) | 1.41 (1.15-1.66) | 0.45 | |
| HbA1C (%) | 8.59 (2.26) | 6.18 (0.54) | 9.69 (1.85) | <0.001 | |
| Glucose (mmol/L) | 11.3 (8.6-14.0) | 7.4 (6.3-8.8) | 12.9 (10.6-14.8) | <0.001 | |
| TyG index | 9.04 (0.48) | 8.73 (0.39) | 9.18 (0.45) | <0.001 | |
| C-peptide (nmol/L) | 1.05 (0.62-1.33) | 1.15 (0.75-2.01) | 0.87 (0.6-1.21) | 0.032 | |
| Endogenous Insulin (pmol/L) | 93 (49-166) | 142 (88-220) | 72 (46-166) | 0.032 |
Abbreviations: DHG: Diabetic Hyperglycaemia; NDHG: Non-Diabetic Hyperglycaemia; BMI: Body Mass Index; NIHSS: National Institutes of Health Stroke Scale; mRS: Modified Rankin Scale; LACI: lacunar infarct; POCI: posterior circulation infarct; SVD: small vessel disease; LAA: large artery atherosclerosis; CE: cardioembolism; SHR: Stress Hyperglycemia Ratio; IL-6: Interleukin-6; IL-8: Interleukin-8; eGFR: Estimated Glomerular Filtration Rate; CrCl: Creatinine Clearance; WBC: White Blood Cell count; HbA1c: Hemoglobin A1c; TyG: Triglyceride-Glucose Index; OHA: Oral Hypoglycemic Agents.
Data are presented as mean (SD) for normally distributed variables, median (IQR) for non-normally distributed variables, and number (percentage) for categorical variables. P-values were calculated using Student’s t-test or Welch’s t-test for parametric comparisons; the Mann-Whitney U Test for non-parametric comparisons; and Pearson’s chi-square test or Fisher’s exact test for categorical variables.
Metabolic responses to hyperglycaemia between patients with DHG or NDHG
As shown in Table 1, endocrine and inflammatory measures during AIS-associated hyperglycaemia differed between the DHG and NDHG groups. Compared with DHG, patients with NDHG had higher endogenous insulin concentrations (median 142 vs. 72 pmol/L; p = 0.032) and higher C-peptide levels (median 1.15 vs. 0.87 nmol/L; p = 0.032), consistent with relatively preserved endogenous insulin secretion. Patients with NDHG also had higher IL-8 levels (median 23 vs. 13 pg/mL; p = 0.002) and a non-significant trend toward higher IL-6 (median 8 vs. 3 pg/mL; p = 0.094). Cortisol levels did not differ significantly between groups.
Markers reflecting glycaemic context also differed; NDHG was associated with a higher stress hyperglycaemia ratio (median 1.24 vs. 1.02; p < 0.001), whereas DHG was associated with a higher TyG index (mean 9.18 vs. 8.73; p < 0.001), suggesting greater baseline insulin resistance. Stroke aetiology (TOAST) and vascular territory (OCSP) were broadly similar between groups, indicating that these biomarker differences were observed despite comparable distributions of stroke subtype and territory.
Associations Between Stroke Severity, Diabetes Status, and Acute Metabolic Responses
Stroke severity was categorised using NIHSS, with scores ≤4 indicating minor stroke and >4 indicating moderate stroke. Of 80 patients, 46 (57.5%) had minor strokes and 34 (42.5%) had moderate strokes. Table 2 summarises endocrine and inflammatory measures across four predefined subgroups (DHG-minor, DHG-moderate, NDHG-minor, and NDHG-moderate), with pairwise comparisons adjusted using the BH correction.
Table 2.
Comparison of Demographic, Metabolic, and Biomarker Profiles Across Diabetes and NIHSS Severity Groups during AIS
| Variable | Overall N = 80 |
DHG minor IS N = 34 |
DHG moderate IS N = 21 |
NDHG minor IS N = 12 |
NDHG moderate IS N = 13 |
p-value |
|---|---|---|---|---|---|---|
| Age (y) | 61 (10) | 58 (10) | 61 (9) | 63 (9) | 67 (13) | 0.07 |
| Weight (Kg) | 65 (60-75) | 65 (60-75) | 63 (60-77) | 70 (65-77) | 58 (55-69) | 0.13 |
| IL-6 (pg/mL) | 4 (2-11) | 3 (1-6) | 7 (2-17) | 6 (4-8) | 11 (3-17) | 0.046 |
| IL-8 (pg/mL) | 15 (11-25) | 13 (10-19) | 14 (10-20) | 18 (16-25) | 27 (18-50) a, b | 0.009 |
| Cortisol (nmol/L) | 323 (182-503) | 327 (195-501) | 389 (270-639) | 246 (168-405) | 274 (142-500) | 0.61 |
| TyG index | 9.04 (0.48) | 9.23 (0.4) | 9.1 (0.52) | 8.84 (0.47) a | 8.63 (0.29) a, b | <0.001 |
| WBC ×10⁹/L | 9.6 (8.0-11.6) | 10.4 (7.8-11.9) | 9.8 (8.0-12.7) | 9.1 (8.4-11.1) | 9.0 (7.6-10.4) | 0.48 |
| SHR | 1.11 (0.92-1.25) | 0.98 (0.83-1.14) | 1.06 (0.87-1.21) | 1.28 (1.13-1.43) a, b | 1.22 (1.11-1.33) a | 0.001 |
| C-peptide (nmol/L) | 1.05 (0.62-1.33) | 1.03 (0.63-1.28) | 0.71 (0.36-1.15) | 1.78 (1.11-2.18) a, b | 1.06 (0.52-1.15) c | 0.003 |
| Insulin (pmol/L) | 93 (49-166) | 81 (47-228) | 59 (38-108) | 252 (149-430) a, b | 92 (48-133) c | <0.001 |
Abbreviations: IS: Ischemic stroke; DHG: Diabetic Hyperglycaemia; NDHG: Non-Diabetic Hyperglycaemia; NIHSS = National Institutes of Health Stroke Scale; SHR = stress-hyperglycemia ratio; TyG index = triglyceride-glucose index; IL-6 = interleukin-6; IL-8 = interleukin-8.
Values are presented as mean (SD) for normally distributed variables or median (IQR) for non-normally distributed variables. Superscript letters indicate statistically significant differences derived from pairwise comparisons with Benjamini–Hochberg (BH) correction (BH-adjusted p < 0.05): a vs. DHG-minor; b vs. DHG-moderate; c vs. NDHG-minor. Global p-values were calculated using One-way ANOVA or the Kruskal-Wallis test as appropriate.
Endogenous insulin-related measures
Insulin and C-peptide differed significantly across subgroups. Patients with NDHG and minor stroke had the highest insulin and C-peptide concentrations (insulin 252 pmol/L [149-430]; C-peptide 1.78 nmol/L [1.11-2.18]). These values were higher than those observed in both DHG groups (DHG-minor (81 pmol/L [47-228]; 1.03 nmol/L [0.63-1.28]), DHG–moderate (59 pmol/L [38-108]; 0.71 nmol/L [0.36-1.15])), and NDHG–moderate (92 pmol/L [48–133]; 1.06 nmol/L [0.52-1.15]) (all BH-adjusted p<0.05). The NDHG-moderate subgroup had significantly lower values than the NDHG-minor group but did not differ significantly from the DHG subgroups.
Inflammatory markers
IL-6 did not differ significantly between subgroups, although median values were numerically higher in the moderate stroke groups, from DHG–minor (3 pg/mL [1-6]) and NDHG–minor (6 pg/mL [4-8]) to DHG–moderate (7 pg/mL [2-17]) and NDHG–moderate (11 pg/mL [3-17]). IL-8 demonstrated clearer separation; the NDHG-moderate subgroup had the highest IL-8 concentration (27 pg/mL [18-50]), which was significantly higher than both DHG-minor (13 pg/mL [10-19]) and DHG-moderate (14 pg/mL [10-20]) (BH-adjusted p < 0.05). IL-8 levels in the NDHG-minor group (18 pg/mL [16-25]) were intermediate and did not differ significantly from other groups.
Glycemic context and insulin resistance indices
SHR differed by diabetes status, with the highest values observed in the NDHG-minor group (1.28 [1.13-1.43]), which was significantly higher than both DHG-minor (0.98 [0.83-1.14]) and DHG-moderate (1.06 [0.87-1.21]) (BH-adjusted p < 0.05). The NDHG-moderate group (1.22 [1.11-1.33]) was significantly higher than DHG-minor but not DHG-moderate (BH-adjusted p = 0.08). In contrast, the TyG index was significantly elevated in the DHG groups. The NDHG-moderate group had the lowest TyG index (8.63 [0.29]), which was significantly lower than both DHG-minor and DHG-moderate (BH-adjusted p<0.01). The NDHG-minor group (8.84 [0.47]) was significantly lower than DHG-minor (p=0.01) but did not differ significantly from DHG-moderate.
Cortisol and white blood cell counts were broadly similar across subgroups. Age did not differ significantly across groups (overall p=0.07). The highest mean age was observed in the NDHG-moderate subgroup (67 [13]), though no BH-adjusted pairwise comparison reached statistical significance.
The relationship between direct and indirect stress responses during AIS hyperglycaemia
In the overall cohort, cortisol and C-peptide showed a modest negative correlation (Spearman’s ρ = –0.19), while C-peptide showed a weak positive correlation with the stress hyperglycaemia ratio (SHR) (ρ = 0.19). Admission NIHSS correlated positively with IL-6 (ρ = 0.29) and IL-8 (ρ = 0.25), and negatively with endogenous insulin (ρ = –0.32), C-peptide (ρ = –0.25), and the TyG index (ρ = –0.26). Collectively, these correlations were consistent with higher stroke severity being associated with higher inflammatory marker levels and lower endogenous insulin-related measures.
When stratified by diabetes status, several associations appeared larger in magnitude in the NDHG group than in the DHG group. In NDHG, the correlations between NIHSS and IL-6 (ρ = 0.28) and between NIHSS and IL-8 (ρ = 0.29) were numerically higher than those in DHG (ρ = 0.27 and ρ = 0.17, respectively). The inverse association between NIHSS and insulin was strong in NDHG (ρ = –0.85, p < 0.001) but weaker and not statistically significant in DHG (ρ = –0.18, p = 0.19). A similar pattern was observed for NIHSS and C-peptide (NDHG: ρ = –0.64, p < 0.001; DHG: ρ = –0.16, p = 0.25). Insulin and C-peptide were positively correlated in both groups, with a higher correlation coefficient in NDHG (ρ = 0.79, p < 0.001) than in DHG (ρ = 0.40, p = 0.02). These subgroup findings suggest that the coupling between stroke severity and endocrine/inflammatory measures may differ by diabetes status; however, formal comparisons of correlation coefficients between groups would be required to confirm such differences.
The association between stroke metabolic responses and the stroke functional outcomes 90 days post-discharge
Using LASSO-penalised logistic regression with 5-fold cross-validation, the minimum binomial deviance occurred at log (λmin) ≈ 2.8 (λ ≈ 0.059), corresponding to a model with five non-zero coefficients (Fig. 2; Supplementary Fig. 1). Cross-validated discrimination was moderate, with the highest AUC observed within the λmin range (Supplementary Fig. 2).
Figure 2.
Coefficient path plot displaying the shrinkage trajectories of all candidate predictors across a sequence of log λ penalty values. As the LASSO penalty increases (moving toward larger log λ), regression coefficients progressively shrink toward zero, with only the most stable predictors retaining non-zero coefficients. The vertical dashed line denotes the optimal λ selected via 5-fold cross-validation (λ ≈ 0.058, log λ ≈ –2.845), at which IL-6, C-peptide, and cortisol remained non-zero and were therefore retained in the final penalized model. This plot illustrates the variable selection behaviour of the LASSO procedure across the whole regularization path.
Supplementary Figure 1.
LASSO Cross-Validation Curve Using Binomial Deviance as the Performance Metric. The plot displays the 5-fold cross-validated binomial deviance across a sequence of penalty values (λ) in the LASSO model. Red points represent mean deviance at each λ, and grey bars indicate ±1 standard error. The left vertical dashed line marks λ₁-SE, the most parsimonious model within one standard error of the minimum deviance. The right dashed line marks λmin, the penalty value yielding the lowest cross-validated deviance. The numbers above the curve indicate the number of predictors retained in the model at each λ. Lower deviance reflects better model fit, with optimal performance observed at log(λ) ≈ -2.845.
Supplementary Figure 2.
LASSO Cross-Validation Curve Using AUC as the Performance Metric. The plot shows the 5-fold cross-validated area under the AUC across a sequence of penalty values (λ) in the LASSO model. Red points represent mean AUC for each λ, and grey bars indicate ±1 standard error. The vertical dashed line on the left marks λ₁-SE, the most parsimonious model within one standard error of the minimum AUC. The vertical dashed line on the right marks λmin, the penalty value yielding the highest cross-validated AUC. The numbers above the curve indicate the number of predictors retained in the model at each λ. The model achieved peak discrimination near log(λ) ≈ -2.845, with AUC values in the moderate range (0.58-0.72).
The variables retained at λmin were IL-6, C-peptide, cortisol, female sex, and TOAST-undetermined subtype. In this exploratory model, IL-6, cortisol, and TOAST-undetermined showed positive coefficient directions (consistent with higher odds of unfavourable outcome), whereas C-peptide and female sex showed negative coefficient directions. All other candidate variables, including diabetes status, TyG index, SHR, other TOAST categories, OCSP territories, and vascular risk factors, were shrunk to zero (Supplementary Table 1).
Bootstrap resampling (1000 iterations) indicated higher selection frequencies for C-peptide (97.5%) and IL-6 (86.4%), with lower but moderate selection frequencies for TOAST-undetermined, female sex, and cortisol (65–67%). Other variables had substantially lower and inconsistent selection frequencies (Fig. 3). These results should be interpreted as hypothesis-generating, given the modest sample size and the potential for model instability despite penalisation.
Figure 3.
Bootstrap stability plot summarizing the frequency with which each predictor was selected (i.e., assigned a non-zero coefficient) across 1000 bootstrap-repeated LASSO models. Predictors selected in ≥70–80% of runs are considered highly stable, whereas those selected less frequently show weaker, sample-dependent signals. In this analysis, IL-6 and C-peptide demonstrated the highest selection proportions, supporting their robustness as exploratory predictors despite the modest sample size. This plot provides an assessment of the reliability and reproducibility of variable selection under repeated resampling.
DISCUSSION
In this prospective multicentre cohort, AIS-associated hyperglycaemia was accompanied by different endocrine and inflammatory profiles in patients classified as NDHG versus DHG. Compared with DHG, NDHG was associated with higher insulin and C-peptide concentrations, higher SHR, and higher IL-8 (with a non-significant trend toward higher IL-6). These findings are compatible with prior conceptual models suggesting that acute hyperglycaemia without pre-existing diabetes may be characterised by stronger stress-related neuroendocrine activation with concurrent immune signalling, whereas hyperglycaemia in established diabetes occurs on a background of chronic dysglycaemia and variable impairment of endogenous insulin secretion (16, 32). Given the observational design and single acute-phase sampling, these results should be interpreted as descriptive and hypothesis-generating rather than mechanistic.
Stroke severity appeared to further modify these responses. In subgroup analyses, NDHG with minor stroke showed the highest insulin and C-peptide concentrations, whereas NDHG with moderate stroke demonstrated the highest IL-8 levels and numerically higher IL-6. DHG subgroups showed lower insulin-related measures and higher TyG indices. Together, these patterns suggest heterogeneity in immunometabolic responses influenced by both pre-stroke Glycemic status and neurological injury severity, consistent with reports from larger biomarker-based phenotyping studies in AIS (33). Our observation of higher SHR in NDHG aligns with multiple cohorts and meta-analyses linking elevated SHR to greater stroke severity and poorer functional outcomes, with some studies reporting stronger associations among patients without established diabetes (34-39). However, NDHG likely represents a spectrum that may include latent dysglycaemia, and residual misclassification cannot be entirely excluded despite HbA1c-based criteria.
In exploratory prognostic modelling, IL-6 and C-peptide were the most consistently selected variables in bootstrap stability analyses, with cortisol contributing more modestly. These findings are directionally consistent with prior literature associating IL-6 and cortisol with adverse post-stroke outcomes (40-42). Nonetheless, the retained biomarkers should be considered candidates for validation rather than definitive independent predictors. Interpretation of insulin concentrations in established diabetes is also complex because circulating insulin reflects both secretion and clearance. In contrast, C-peptide may better reflect endogenous insulin secretion due to co-secretion and lack of first-pass hepatic extraction (43). While C-peptide has been proposed to have vascular and anti-inflammatory effects, clinical relevance in AIS remains uncertain, and any protective interpretation should remain cautious (44).
Limitations
This study has several limitations. First, the sample size was modest (n = 80) with a limited number of outcome events, which restricts statistical power and increases the risk of model instability and overfitting, even with penalised regression. Although LASSO and bootstrap stability analyses were used to mitigate this risk, prognostic findings should be interpreted cautiously and considered hypothesis-generating. Second, group sizes were unequal, with fewer NDHG than DHG participants, which may reduce the precision of between-group estimates and limit the generalisability of subgroup comparisons. Third, biomarker sampling occurred within 72 hours of admission and the timing relative to stroke onset was not standardised. Given dynamic post-stroke trajectories and circadian variation (particularly for IL-6, IL-8, and cortisol), heterogeneity in sampling time may have introduced temporal variability and contributed to inter-individual differences. Fourth, although stroke aetiology (TOAST) and territory (OCSP) were captured, other established prognostic determinants, such as infarct volume, collateral status, detailed neuroimaging characteristics, rehabilitation intensity, and adherence to secondary prevention therapies, were not measured and may confound observed biomarker-outcome associations. Fifth, the clinical performance of the exploratory LASSO model was moderate (cross-validated AUC approximately 0.52–0.72), indicating that it is not suitable for standalone clinical prediction in its current form. Sixth, the number of patients who had thrombolysis therapy was limited to (n=2), which precluded meaningful treatment-stratified analyses (e.g., compared to patients with mechanical thrombectomy). Seventh, due to the observational study design, causal relationships cannot be inferred. Eighth, the absence of a healthy control group limits direct comparison of biomarker levels between patients with AIS and individuals without stroke. Ninth, there were few participants with moderate-to-severe or severe stroke (NIHSS ≥16; n = 6), limiting inference regarding biomarker behaviour at higher severity levels. Finally, the cohort was predominantly Malay, which may limit external validity to more ethnically diverse populations given reported differences in insulin sensitivity, β-cell function, and inflammatory responses. Overall, these limitations underscore the need for larger, well-powered multicentre studies with standardised sampling time points, more detailed diabetes phenotyping and neuroimaging characterisation, and external validation of prognostic models prior to clinical translation.
In conclusion, AIS-associated hyperglycemia presents as two distinct endocrine and inflammatory phenotypes based on prior diabetes status. Patients with NDHG showed higher acute stress responses (endogenous insulin, c-peptide, SHR, IL-8) compared to the DHG cohort. These hypothesis-generating findings, together with exploratory associations with stroke severity and 90-day functional outcomes, support larger diabetes-stratified studies with standardised sampling and external validation to determine whether tailored glycaemic targets and management strategies improve AIS outcomes.
Acknowledgments
We would like to thank the Health Director General of Malaysia for his permission to Publish this article.
Conflict of Interest
The authors declare that they have no conflict of interest.
Funding
This study was funded by the Ministry of Higher Education Malaysia for the Fundamental Research Grant Scheme with Project Code FRGS/1/2020/STG03/USM/02/2.
References
- 1.Tsao CW, Aday AW, Almarzooq ZI, Anderson CA M, Arora P, Avery CL, Baker-Smith C M, Beaton AZ, Boehme AK, Buxton A E, Commodore-Mensah Y, Elkind MSV, Evenson KR, Eze-Nliam C, Fugar S, Generoso G, Heard DG, Hiremath S, Ho JE, Kalani R, Kazi DS, Ko D, Levine DA, Liu J, Ma J, Magnani JW, Michos E D, Mussolino ME, Navaneethan SD, Parikh NI, Poudel R, Rezk-Hanna M, Roth GA, Shah NS, St-Onge M-P, Thacker E L, Virani SS, Voeks JH, Wang N-Y, Wong ND, Wong SS, Yaffe K, Martin SS, Epidemiology o b o t A H A C o, Committee P S, Subcommittee SS Heart Disease and Stroke Statistics - 2023 Update: A Report From the American Heart Association. Circulation. 2023;147:e93–e621. doi: 10.1161/CIR.0000000000001123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Aziz ZA, Lee YY, Ngah BA, Sidek NN, Looi I, Hanip MR, Basri HB. Acute Stroke Registry Malaysia, 2010-2014: Results from the National Neurology Registry. J Stroke Cerebrovasc Dis. 2015;24:2701–2709. doi: 10.1016/j.jstrokecerebrovasdis.2015.07.025. [DOI] [PubMed] [Google Scholar]
- 3.Johnston KC, Bruno A, Pauls Q, Hall CE, Barrett KM, Barsan W, Fansler A, Van de Bruinhorst K, Janis S, Durkalski-Mauldin V L, Network f t N E T T, Investigators t S T. Intensive vs Standard Treatment of Hyperglycemia and Functional Outcome in Patients With Acute Ischemic Stroke: The SHINE Randomized Clinical Trial. JAMA. 2019;322:326–335. doi: 10.1001/jama.2019.9346. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.O’Neill PA, Davies I, Fullerton K J, Bennett D. Stress hormone and blood glucose response following acute stroke in the elderly. Stroke. 1991;22:842–847. doi: 10.1161/01.str.22.7.842. [DOI] [PubMed] [Google Scholar]
- 5.Hodea FV, Voiculescu VM, Manole TG, Grosu-Bularda A, Hariga CS. Hormonal profile in severe burns. Acta endocrinol (Buchar) 2024;20(3):408–411. doi: 10.4183/aeb.2024.408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chen XG, Shi SY, Hu L, Chen Y, Sun HW, Zhou L, Lu ZB, Wang H, Wang XS, Yu J, Zhao YJ, Lu YM, Ye J. Longitudinal changes in the hypothalamic-pituitary-adrenal axis and sympathetic nervous system are related to the prognosis of stroke. Front Neurol. 2022;13:946593. doi: 10.3389/fneur.2022.946593. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Winder K, Seifert F, Ohnemus T, Sauer EM, Kloska S, Dörfler A, Hilz MJ, Schwab S, Köhrmann M. Neuroanatomic correlates of poststroke hyperglycemia. Ann Neurol. 2015;77:262–268. doi: 10.1002/ana.24322. [DOI] [PubMed] [Google Scholar]
- 8.Allport L, Butcher K, Baird T, MacGregor L, Desmond P, Tress B, Colman P, Davis S. Insular Cortical Ischemia Is Independently Associated With Acute Stress Hyperglycemia. Stroke: Journal of the American Heart Association. 2004;35:1886–1891. doi: 10.1161/01.STR.0000133687.33868.71. [DOI] [PubMed] [Google Scholar]
- 9.Luitse MJA, Biessels GJ, Rutten GEHM, Kappelle LJ. Diabetes, hyperglycaemia, and acute ischaemic stroke. The Lancet Neurology. 2012;11:261–271. doi: 10.1016/S1474-4422(12)70005-4. [DOI] [PubMed] [Google Scholar]
- 10.Kruyt ND, Biessels GJ, DeVries JH, Roos YB. Hyperglycemia in acute ischemic stroke: pathophysiology and clinical management. Nature Reviews Neurology. 2010;6:145–155. doi: 10.1038/nrneurol.2009.231. [DOI] [PubMed] [Google Scholar]
- 11.Li WA, Moore-Langston S, Chakraborty T, Rafols JA, Conti AC, Ding Y. Hyperglycemia in stroke and possible treatments. Neurol Res. 2013;35:479–491. doi: 10.1179/1743132813Y.0000000209. [DOI] [PubMed] [Google Scholar]
- 12.Hengky A, Pratama KG, Tandarto K. Mortality and cardiovascular risk reduction after reversion of prediabetes to normoglycemia: a systematic review. Acta Endocrinol (Buchar) 2024;20(1):74–79. doi: 10.4183/aeb.2024.74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Blum CA, Mueller C, Schuetz P, Fluri F, Trummler M, Mueller B, Katan M, Christ-Crain M. Prognostic value of dehydroepiandrosterone-sulfate and other parameters of adrenal function in acute ischemic stroke. PLoS One. 2013;8:e63224. doi: 10.1371/journal.pone.0063224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cerecedo-Lopez CD, Cantu-Aldana A, Patel NJ, Aziz-Sultan MA, Frerichs KU, Du R. Insulin in the Management of Acute Ischemic Stroke: A Systematic Review and Meta-Analysis. World Neurosurg. 2020;136:e514–e534. doi: 10.1016/j.wneu.2020.01.056. [DOI] [PubMed] [Google Scholar]
- 15.Long MT, Coursin DB, Krinsley JS. Treatment of Hyperglycemia in Patients With Acute Stroke. JAMA. 2019;322:2248. doi: 10.1001/jama.2019.16339. [DOI] [PubMed] [Google Scholar]
- 16.Reshi R, Streib C, Ezzeddine M, Biros M, Miller B, Lakshminarayan K, Anderson D, Ardelt A. Hyperglycemia in acute ischemic stroke: Is it time to re-evaluate our understanding? Med Hypotheses. 2017;107:78–80. doi: 10.1016/j.mehy.2017.08.017. [DOI] [PubMed] [Google Scholar]
- 17.Zhu B, Pan Y, Jing J, Meng X, Zhao X, Liu L, Wang Y, Wang Y, Wang Z. Stress Hyperglycemia and Outcome of Non-diabetic Patients After Acute Ischemic Stroke. Frontiers in Neurology. 2019;10:1003. doi: 10.3389/fneur.2019.01003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ye G, Gao Q, Qi P, Wang J, Hu S, Chen K, Tan T, Lu J, Wang D. The role of diabetes mellitus on the thrombus composition in patients with acute ischemic stroke. Interv Neuroradiol. 2020;26:329–336. doi: 10.1177/1591019919896940. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Tziomalos K, Dimitriou P, Bouziana SD, Spanou M, Kostaki S, Angelopoulou SM, Papadopoulou M, Giampatzis V, Savopoulos C, Hatzitolios AI. Stress hyperglycemia and acute ischemic stroke in-hospital outcome. Metabolism. 2017;67:99–105. doi: 10.1016/j.metabol.2016.11.011. [DOI] [PubMed] [Google Scholar]
- 20.Luethi N, Cioccari L, Biesenbach P, Lucchetta L, Kagaya H, Morgan R, Di Muzio F, Presello B, Gaafar D, Hay A, Crisman M, Toohey R, Russell H, Glassford NJ, Eastwood GM, Ekinci EI, Deane AM, Bellomo R, Mårtensson J. Liberal Glucose Control in ICU Patients With Diabetes: A Before-and-After Study*. Critical Care Medicine. 2018;46:935–942. doi: 10.1097/CCM.0000000000003087. [DOI] [PubMed] [Google Scholar]
- 21.Committee ADAPP 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes-2025. Diabetes Care. 2024;48:S27–S49. doi: 10.2337/dc25-S002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Fuentes B, Ntaios G, Putaala J, Thomas B, Turc G, Diez-Tejedor E, European Stroke O. European Stroke Organisation (ESO) guidelines on glycaemia management in acute stroke. Eur Stroke J. 2018;3:5–21. doi: 10.1177/2396987317742065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kleindorfer DO, Towfighi A, Chaturvedi S, Cockroft KM, Gutierrez J, Lombardi-Hill D, Kamel H, Kernan WN, Kittner SJ, Leira EC, Lennon O, Meschia JF, Nguyen TN, Pollak PM, Santangeli P, Sharrief AZ, Smith SC, Turan TN, Williams LS. 2021 Guideline for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack: A Guideline From the American Heart Association/American Stroke Association. Stroke. 2021;52:e364–e467. doi: 10.1161/STR.0000000000000375. [DOI] [PubMed] [Google Scholar]
- 24.Sacks DB, Arnold M, Bakris GL, Bruns DE, Horvath AR, Lernmark Å, Metzger BE, Nathan DM, Kirkman MS. Guidelines and Recommendations for Laboratory Analysis in the Diagnosis and Management of Diabetes Mellitus. Diabetes Care. 2023;46:e151–e199. doi: 10.2337/dci23-0036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Clinical Practice Guidelines: Management of Obesity. 2. Malaysia Health Technology Assessment Section (MaHTAS); 2023. [Google Scholar]
- 26.Hage V. The NIH stroke scale: a window into neurological status. Nursing Spectrum. 2011;24:44–49. [Google Scholar]
- 27.Sulter G, Steen C, De Keyser J. Use of the Barthel index and modified Rankin scale in acute stroke trials. Stroke. 1999;30:1538–1541. doi: 10.1161/01.str.30.8.1538. [DOI] [PubMed] [Google Scholar]
- 28.Roberts GW, Quinn SJ, Valentine N, Alhawassi T, O’Dea H, Stranks S N, Burt M G, Doogue M P. Relative Hyperglycemia, a Marker of Critical Illness: Introducing the Stress Hyperglycemia Ratio. J Clin Endocrinol Metab. 2015;100:4490–4497. doi: 10.1210/jc.2015-2660. [DOI] [PubMed] [Google Scholar]
- 29.Simental-Mendía L E, Rodríguez-Morán M, Guerrero-Romero F. The product of fasting glucose and triglycerides as surrogate for identifying insulin resistance in apparently healthy subjects. Metab Syndr Relat Disord. 2008;6:299–304. doi: 10.1089/met.2008.0034. [DOI] [PubMed] [Google Scholar]
- 30.Saver JL, Altman H. Relationship between neurologic deficit severity and final functional outcome shifts and strengthens during first hours after onset. Stroke. 2012;43:1537–1541. doi: 10.1161/STROKEAHA.111.636928. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Taleb S, Ji-Hyun Lee J, Asanad S, Starkman S, Hamilton S, Gornbein J, Conwit RA, Sanossian N, Saver JL. Modified Rankin Scale disability status at day 4 poststroke is an informative predictor of long-term day 90 outcome. J Stroke Cerebrovasc Dis. 2024;33:107902. doi: 10.1016/j.jstrokecerebrovasdis.2024.107902. [DOI] [PubMed] [Google Scholar]
- 32.White MG, Shaw JAM, Taylor R. Type 2 Diabetes: The Pathologic Basis of Reversible β-Cell Dysfunction. Diabetes Care. 2016;39:2080–2088. doi: 10.2337/dc16-0619. [DOI] [PubMed] [Google Scholar]
- 33.Ding L, Mane R, Wu Z, Jiang Y, Meng X, Jing J, Ou W, Wang X, Liu Y, Lin J, Zhao X, Li H, Wang Y, Li Z. Data-driven clustering approach to identify novel phenotypes using multiple biomarkers in acute ischaemic stroke: A retrospective, multicentre cohort study. eClinicalMedicine. 2022;53:101639. doi: 10.1016/j.eclinm.2022.101639. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Li W, Yan H, Gao Y, Gao C, Li W, Pan Y, Johnston SC, Amarenco P, Bath PM, Wang Y, Liao X. Stress Hyperglycemia Ratio and Adverse Outcomes in Acute Mild Ischemic Stroke or High-Risk Transient Ischemic Attack: A Secondary Analysis of the INSPIRES Trial. Stroke. 2026;57(2):349–361. doi: 10.1161/STROKEAHA.125.052987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Feng J, He C, San H, Zhan M, Zhang H, Gu J, Fu H, Yao J, Li L, Wang X. The influence of stress hyperglycemia on consciousness disturbance and short- and long-term outcomes in stroke patients without documented diabetes: Differences between ischemic and hemorrhagic stroke. PLOS One. 2025;20:e0331077. doi: 10.1371/journal.pone.0331077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Zhang Y, Yin X, Liu T, Ji W, Wang G. Association between the stress hyperglycemia ratio and mortality in patients with acute ischemic stroke. Sci Rep. 2024;14(1):20962. doi: 10.1038/s41598-024-71778-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Huang YW, Yin XS, Li ZP. Association of the stress hyperglycemia ratio and clinical outcomes in patients with stroke: A systematic review and meta-analysis. Front Neurol. 2022;13:999536. doi: 10.3389/fneur.2022.999536. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Yu Y, Rao Z, Duan T, Zhang Y, Zhu C. Association between stress hyperglycaemia ratio and admission stroke severity: a retrospective study in a Chinese single-centre cohort. BMJ Open. 2025;15:e105117. doi: 10.1136/bmjopen-2025-105117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Long Z, Liao S, Chen Y. Nonlinear association between stress hyperglycemia ratio and severe consciousness disorder in acute ischemic stroke: A MIMIC retrospective analysis. PLOS One. 2025;20:e0329678. doi: 10.1371/journal.pone.0329678. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.McCabe JJ, Harris K, Walsh C, Gorey S, Arnold M, De Marchis GM, Hervella P, Iglesias-Rey R, Jern C, Katan M, Li L, Miyamoto N, Montaner J, Purroy F, Rothwell PM, Stanne TM, Sudlow C, Ueno Y, Vicente-Pascual M, Whiteley W, Woodward M, Kelly PJ. Interleukin-6, C-Reactive Protein, and Vascular Recurrence After Stroke With and Without Atherosclerosis. Stroke. 2025;56:2588–2596. doi: 10.1161/STROKEAHA.125.052091. [DOI] [PubMed] [Google Scholar]
- 41.Dieplinger B, Bocksrucker C, Egger M, Eggers C, Haltmayer M, Mueller T. Prognostic Value of Inflammatory and Cardiovascular Biomarkers for Prediction of 90-Day All-Cause Mortality after Acute Ischemic Stroke-Results from the Linz Stroke Unit Study. Clinical Chemistry. 2017;63:1101–1109. doi: 10.1373/clinchem.2016.269969. [DOI] [PubMed] [Google Scholar]
- 42.Montellano FA, Ungethüm K, Ramiro L, Nacu A, Hellwig S, Fluri F, Whiteley WN, Bustamante A, Montaner J, Heuschmann P U. Role of Blood-Based Biomarkers in Ischemic Stroke Prognosis. Stroke. 2021;52:543–551. doi: 10.1161/STROKEAHA.120.029232. [DOI] [PubMed] [Google Scholar]
- 43.Maida CD, Daidone M, Pacinella G, Norrito RL, Pinto A, Tuttolomondo A. Diabetes and Ischemic Stroke: An Old and New Relationship an Overview of the Close Interaction between These Diseases. International Journal of Molecular Sciences. 2022;23:2397. doi: 10.3390/ijms23042397. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Pan Y, Chen W, Jing J, Zheng H, Jia Q, Li H, Zhao X, Liu L, Wang Y, He Y, Wang Y. Pancreatic β-Cell Function and Prognosis of Nondiabetic Patients With Ischemic Stroke. Stroke. 2017;48:2999–3005. doi: 10.1161/STROKEAHA.117.018203. [DOI] [PubMed] [Google Scholar]