Performance comparison of stress hyperglycemia ratio for predicting fatal outcomes in patients with thrombolyzed acute ischemic stroke

Background The stress hyperglycemia ratio (SHR), a newly developed metric, is used to assess adverse outcomes in patients with acute ischemic stroke (AIS). However, the relationship between SHR and fatal outcomes (in-hospital mortality [IHM], malignant cerebral edema [MCE], symptomatic intracerebral hemorrhage [sICH], 3-month mortality, and poor functional outcome) in AIS patients receiving recombinant tissue plasminogen activator (rt-PA) treatment is unclear, and determining the optimal threshold remains incomplete. Materials and methods We retrospectively enrolled a total of 345 AIS patients treated with rt-PA during 2015–2022 and collected data on various glucose metrics, including different types of SHR, glycemic gap (GG), random plasma glucose (RPG), fasting plasma glucose (FPG), and hemoglobin A1c (HbA1c). SHR and GG were calculated using these equations: SHR1, [FPG]/[HbA1c]; SHR2, [admission RPG]/[HbA1c]; SHR3, FPG/[(1.59 × HbA1c)−2.59]; SHR4, [admission RPG]/[(1.59 × HbA1c)−2.59]; GG, admission RPG − [(1.59 × HbA1c)−2.59]. We used multivariable logistic regression analysis (MVLR) to identify the association between different glucose metrics and outcomes while comparing their predictive values. Results SHR1 had the greatest predictive power and a more significant correlation with fatal outcomes than other continuous glucose metrics. The area under the curve of the SHR1 for IHM, MCE, and sICH, 3-month mortality, and poor functional outcome were 0.75, 0.77, 0.77, 0.76, and 0.73, respectively. SHR1 (per 1-point increases) was independently associated with IHM (Odds ratios [ORs] = 5.80; 95% CI [1.96, 17.17]; p = 0.001), MCE (ORs = 4.73; 95% CI [1.71, 13.04]; p = 0.003), sICH (ORs = 4.68, 95% CI [1.48–14.82]; p = 0.009), 3-month mortality (ORs = 10.87; 95% CI [3.56, 33.21]; p<0.001), and 3-month poor functional outcome (ORs = 8.05; 95% CI [2.77, 23.39]; p<0.001) after adjustment in MVLR. In subgroup analysis, elevated SHR1 was associated with fatal outcomes in patients with non-diabetes, SBP≥ 180 mmHg, and NIHSS <16. Conclusion SHR1 demonstrates an independent association with fatal outcomes in AIS patients treated with rt-PA, exhibiting superior predictive ability over other glucose metrics.


Introduction
Stroke is a prevalent neurological condition and a primary global cause of death, resulting in approximately 6 million annual fatalities [1].Stroke is the leading cause of death in Thailand, accounting for over 250,000 new cases and 50,000 annual fatalities [2].Recombinant tissue plasminogen activator (rt-PA) is recommended as a safe and effective treatment [3].Elevated blood sugar in 40-50% of acute stroke patients may exacerbate ischemic injury through heightened oxidative stress, endothelial dysfunction, and impaired fibrinolysis, resulting in larger infarctions, worse clinical outcomes, and increased mortality rates.[4,5].Stress hyperglycemia (SH) refers to transient hyperglycemia in the context of illness accompanied by diabetes mellitus (DM) or non-DM.Recently, Roberts et al. [6] introduced the stress hyperglycemia ratio (SHR) to evaluate SH.Hemoglobin A1c (HbA1c), a stable indicator, was used to assess glycemic management in DM patients over three months.SHR is calculated by dividing the admission glucose concentration by the estimated average glucose concentration derived from HbA1c [7].Different studies employed the glucose/HbA1c ratio to define SHR, aiming for its practical use in clinical settings [8][9][10].
Poor outcomes and symptomatic intracerebral hemorrhage (sICH) in acute ischemic stroke (AIS) patients treated with rt-PA were associated with hyperglycemia.According to the American Diabetes Association, patients were classified as DM, newly diagnosed DM, or experiencing transient hyperglycemia during hospitalization.The definition of SH remains unclear, but an abrupt increase in plasma glucose levels above the average blood glucose level serves as a reliable indicator [6,11].Two types of biological markers for SH, SHR and glycemic gap (GG), have been developed to represent SH [11].Recently, the SHR, a ratio of plasma glucose level to HbA1c, has emerged as a prognostic biomarker for poor outcomes in AIS patients receiving rt-PA treatment.
Although different SHR equations effectively predicted unfavorable outcomes or critical illness in AIS patients [12], the optimal threshold of SHR for assessing SH and predicting fatal outcomes (in-hospital mortality [IHM], malignant cerebral edema [MCE], sICH, 3-month mortality, and poor functional outcome) has not been definitively confirmed.Limited data currently exists regarding the comparative predictive value of various types of SHR, GG, absolute plasma glucose, and HbA1c in predicting fatal outcomes in AIS patients treated with rt-PA.Hence, this study aims to explore the predictive performance, optimal thresholds, and association between these variables in predicting fatal outcomes.

Study population
We conducted a retrospective observational cohort study by collecting data on 345 AIS patients who were treated with intravenous rt-PA at Saraburi Hospital, a stroke referral center of a provincial hospital in Thailand, between January 1, 2015 and July 31, 2022.Treatment involved administering intravenous rt-PA following the 2019 AIS management guideline [13].Inclusion criteria: (1) age � 18 years; (2) AIS within 4.5 hours of the last known normal; (3) acute anterior circulation ischemic stroke; and (4) rt-PA treatment only.Exclusion criteria: (1) minor stroke; (2) pregnancy; (3) ICH or infarction > 1/3 the middle cerebral artery (MCA) territory; (4) referred patients with unattainable follow-up; (5) missing data: National Institutes of Health Stroke Scale (NIHSS), non-contrast computed tomography (NCCT) imaging, and laboratory results.AIS patients receiving EVT were not included in this study.AIS patients suspected of large vessel occlusion (LVO) were not transferred for EVT during the study period due to limitations in Thailand's public health coverage, causing difficulties in accessing this treatment.All patients stayed in the hospital, regardless of LVO.Data comprising clinical and imaging information were retrieved from our electronic medical records, with diagnoses established using the International Classification of Diseases, 10th Revision codes (I63).The data were fully anonymized before we accessed them, and the ethics committee waived the requirement for informed consent.We don't collect patient-identifying information, including hospital numbers, admission numbers, identity card numbers, or birthdates.The study received ethical approval from the human research ethics committee of Saraburi Hospital on January 30, 2023 (Certificate No. EC004/2566).We accessed the data for research purposes on February 5, 2023.

Data collection
Demographic data, encompassing age, gender, initial clinical presentation, medical history, laboratory investigations, time from symptom onset to treatment, admission blood pressure, Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification, and neuroimaging results, were collected.The extent of early ischemic changes was evaluated using the Alberta Stroke Program Early Computed Tomography Score (ASPECTS).Certified neurologists utilized NIHSS to evaluate stroke severity upon admission.The plasma glucose level on admission was measured before thrombolytic treatment.The history of DM was established by reviewing the patient's medical diagnosis and their antidiabetic drug usage record.

Patient preparation.
We collected fasting plasma glucose (FPG) and HbA1c after an 8-14 hour fasting but not exceeding 16 hours, to avoid starvation, with a morning collection between 6:00 a.m. and 10:00 a.m.Random plasma glucose (RPG) was measured upon hospital arrival regardless of the time since the last meal.
2.3.2Methods for collecting and submitting specimens.Blood collection tubes with anticoagulants like sodium fluoride/potassium oxalate were used to collect 2 cc of blood for plasma glucose testing, and tubes with ethylene diamine tetra acetate were used for HbA1c testing.The samples were analyzed within 45 minutes of collection, with results reported within 1 hour.Blood samples for plasma glucose were analyzed using an automated analyzer (Beckman Coulter DxC 700 AU) with the Beckman Coulter glucose reagent.Blood samples for HbA1c were analyzed using an automated analyzer (Mindray BS-820M) with the HbA1c reagent.

Outcomes assessment
The primary outcome of the study was IHM defined as patients with thrombolyzed AIS who died in the hospital.The secondary outcomes were MCE, sICH, 3-month mortality, and poor functional outcome.The diagnostic criteria for malignant cerebral edema (MCE) were as follows: (i) acute complete MCA infarction with parenchymal hypodensity covering at least 50% of the MCA territory, along with sulcal effacement and lateral ventricle compression; (ii) excessive midline shift exceeding 5 mm and obliteration of basal cisterns; and (iii) neurological deterioration was characterized by an increase in NIHSS score (more than 2 points) and a decline in consciousness level (at least 1 point in item 1A of the NIHSS assessment) [18].Based on the National Institute of Neurological Disorders and Stroke criteria, sICH was defined as any deterioration in NIHSS score or mortality within 7 days of thrombolysis initiation, along with the presence of any type of intracerebral hemorrhage on posttreatment imaging.[19].The survival status was determined by utilizing mortality data derived from electronic medical records and death certificates, which were supplied by local municipal authorities for each study participant.3-month mortality referred to death within 90 days regardless of causes, and 3-month poor functional outcome was defined as mRS > 2 at 90 days after a stroke All included patients were followed up through telephone interviews conducted by stroke-trained nurses and/or physical therapy staff 90 days after the stroke.NCCT scans were done within 4.5 hours of symptom onset and repeated at 24 hours post-thrombolysis.An emergency NCCT would be performed for deteriorating neurological deficits.

Statistical analysis
We analyzed the data using Stata version 17 (StataCorp, Lakeway, Texas 77845, USA) and considered a two-tailed p-value <0.05 statistically significant.Continuous variables with a normal distribution were summarized using mean and standard deviation, while those non-normal distributed variables were described using median and interquartile range (IQR).Categorical variables were presented as frequencies and percentages.Statistical tests such as t-test, Mann-Whitney U-test, and chi-squared test were used to compare differences between the two groups, depending on variables.The predictive potential of different admission glucose measures (SHR1, SHR2, SHR3, SHR4, GG, FPG, admission RPG, and HbA1c) for fatal outcomes (IHM, MCE, sICH, 3-month mortality, and poor functional outcome) was evaluated.Sensitivity, specificity, positive and negative predictive values, likelihood ratios of positive and negative, and accuracy were analyzed.The multivariable logistic regression (MVLR) model, considering relevant factors, was employed to determine odds ratios (ORs) and 95% confidence intervals (CIs) for predicting fatal outcomes.We selected variables for adjustment in the MVLR model based on previous background knowledge.For IHM, the crude model represents univariable analysis.In model A, we adjusted for age and sex.In model B, we adjusted for variables in model A plus TOAST classification, NIHSS, and baseline ASPECTS �6.In model C, we additionally adjusted for comorbidities (DM, chronic kidney disease [CKD], myocardial infarction [MI], congestive heart failure [CHF], mRS, history of cancer), systolic blood pressure (SBP), and diastolic blood pressure (DBP) to assess the relationship between glucose metrics and fatal outcomes.
To maintain the integrity of variable relationships, we assessed the impact of the approach in the MVLR model.We examined confounding factors and their effect on SHR and conducted a subgroup analysis to explore the multiplicative interaction.The receiver operating characteristic (ROC) curves for glucose metrics were compared using a nonparametric method [20], and the optimal cut-off value of glucose metrics at admission was determined using the Youden's index method for predicting fatal outcomes.

Results
In this retrospective study, a total of 387 patients were diagnosed with thrombolysis-indicated AIS during January 1, 2015 to July 31, 2022.Five patients refused rt-PA treatment, nine patients were diagnosed with posterior circulation ischemic stroke, six patients were referred to another hospital, and 22 patients with missing brain CT data were excluded.The remaining 345 patients were included for analysis in this cohort.However, 11 patients (3.19%) had missing data on the 3-month mRS evaluation for assessing poor functional outcomes.(Fig 1)

Predictive value of glucose metrics and ROC curve for predicting fatal outcomes
The ROC curve analysis was employed to assess the predictive efficacy of SHR1

Relationship between glucose metrics and fatal outcomes
3.3.1 IHM.We evaluated the association between various types of glucose metrics and IHM, and found that all types of glucose metrics were determinants of IHM in all the evaluated models, except for HbA1c.SHR1 with each one-point increase was associated with IHM after adjustment for age, sex, TOAST classification, NIHSS, baseline ASPECTS �7, and comorbidities (DM, CKD, MI, CHF, preexisting disability, and history of malignancy) (OR = 5.80; 95% CI: 1.96, 17.17; p = 0.001), whereas HbA1c was not associated with the outcome by any model.Based on all evaluated models, SHR1 � 1.18, SHR2 �1.26, SHR3 � 1.01, SHR4 � 1.03, GG � 0.17 mmol/L, FPG � 6.91 mmol/L, and admission RPG � 6.47 mmol/L showed significant association with IHM after adjusting for the confounders in all models.Only HbA1c � 5.55 was not associated with IHM in any of the models.Among the continuous glucose metrics, SHR1 exhibited the strongest correlation with IHM when compared to others (Table 3).
SHR1 in continuous glucose metrics remained a striking predictor and had the greatest impact on fatal outcomes after thrombolysis.The association between FPG in glucose metric threshold had the strongest relationship with fatal outcomes, which was further strengthened after adjusting for fully adjusted Model C. The distribution of mRS score at time of hospital discharge and 3-month follow-up in group stratified according to cut-off value of SHR1.(S1 Fig) .IHM, MCE, 3-month mortality, and poor functional outcome rates were higher in patients with SHR1�1.18compared to those with SHR1<1.18(39.4% versus 8.9%; p<0.001, 32.1% versus 7.2%; p<0.001, 50.5% versus 11.9%; p<0.001, and 69.4% versus 27.9%; p<0.001, respectively), while the sICH rate was higher in patients with SHR1�1.12 compared to those with SHR1<1.12(26.0%versus 4.1%; p<0.001) (S2 Fig).

Results of subgroup analysis for the fatal outcomes
Surprisingly, we also found that elevated SHR1 (SHR1�1.18[for IHM, MCE, 3-month mortality, and poor functional outcome] and �1.12 [for sICH]) was independently associated with fatal outcomes in non-DM AIS patients.The ORs for IHM were 6.13 versus 3.36 (p < 0.001), for MCE were 4.29 versus 0.84 (p<0.001), for 3-month mortality were 14.9 versus 11.6 (p = 0.014), and for 3-month poor functional outcome were 6.48 versus 5.36 (p = 0.003) compared with those who had DM.Similarly, elevated SHR1 with SBP�180 mmHg and NIHSS<16 were associated with an increased 3-month mortality risk and poor functional outcome compared with those with SBP<180 mmHg and NIHSS�16, respectively (see S1 Table ).

Discussion
The relationship between the SHR and fatal outcomes in AIS patients treated with rt-PA was identified in our cohort study.The main findings of this study were as follows: (1) SHR1 had the greatest predictive power for fatal outcomes among other glucose metrics; (2) SHR1, as a continuous variable, had the strongest association with fatal outcomes after adjustment for potential confounders; (3) both SHR1 and SHR3 at admission, as continuous variables and at thresholds, were independently associated with impact on fatal outcomes in the MVLR model, indicating that these indicators might have an important role in glycemic control intervention in patients with SH; (4) in subgroup analysis, elevated SHR1 had a stronger association with   IHM, MCE, 3-month mortality, and poor functional outcome in patients aged � 70 years compared to those < 70 years., while elevated SHR1 had a stronger association with sICH in patients < 70 years compared to those � 70 years.Furthermore, elevated SHR1 was more strongly associated with fatal outcomes in AIS patients who were non-DM, with SBP� 180 mmHg, and NIHSS<16.To our knowledge, there were few studies discussing the association between SHR and fatal outcomes in thrombolyzed AIS patients.Our study findings aligned with previous research, indicating that SHR1 was a significant predictor of 3-month mortality and sICH in AIS patients who underwent both rt-PA treatment [21][22][23][24] and EVT [25].The mechanism of the relationship between SH and fatal outcomes is not yet clearly understood.Several mechanisms have been proposed.Firstly, SH may serve as a marker indicating the extent of ischemic damage following stroke.SH is characterized by rapid blood glucose elevation due to hypothalamic-pituitary-adrenal axis and sympathetic nervous system activation, which promotes excessive gluconeogenesis, glycogenolysis, and insulin resistance through the complex interaction of hormones, including cortisol, growth hormone, glucagon, and catecholamines.[26,27].The hyperglycemic state leads to a dramatic elevation of inflammatory cytokines and vasoconstrictive factors, contributing to fatal outcomes [28].Secondly, AIS patients' acute stressrelated inflammation leads to rapid accumulation of circulating free fatty acids and oxidative stress.These contribute to the decline in endothelial nitric oxide, a vasodilator, and an increase in plasminogen activator inhibitor, further worsening penumbra perfusion and ischemia [29].Thirdly, SH is related to the reperfusion injury after successful recanalization, which increases the risk of hemorrhagic transformation, one of the fatal outcomes in AIS patients [30].
We speculate that FPG represent the genuine blood glucose level affected from the stress without being confounded by meal-derived glucose as the RPG [31].Consequently, SHR1 and SHR3 which were calculated from FPG represented stronger association and higher predictive values to the fatal outcomes than SHR2 and SHR4.Increasing hyperglycemia in AIS patients escalates oxidative stress, neurohormonal derangement, and inflammatory cytokines, thereby perpetuating a vicious cycle that exacerbates hyperglycemia [23].Meta-analyses have shown that acute hyperglycemia is related to IHM and poor functional recovery in non-DM AIS patients [28].In a recent study, Merlino et al. demonstrated that diabetic status has a protective role on the decremental effects of SH in AIS patients receiving alteplase.Conversely, non-DM with severe SH had a higher incidence of poor outcomes at three months and sICH.[32] This is consistent with our study's subgroup analysis, demonstrating that elevated SHR1 was more strongly associated with fatal outcomes in non-DM patients than in DM AIS patients.The relationship between SH and poor outcomes in non-DM AIS patients has several explanations.First, the glycemic threshold for stress hyperglycemia on top chronic hyperglycemia in DM is higher than that in normoglycemic patients.Second, chronic hyperglycemia in DM patients affects their physiologic response, so they may develop tolerance mechanisms that diminish the deleterious metabolic effects [21].Third, acute hyperglycemia in non-DM individuals may indicate a more severe or prolonged stress response, which can contribute to poorer clinical outcomes.Lastly, DM patients tend to receive more intensive glycemic management due to healthcare providers' awareness.
Optimal glycemic control benefits survival in both DM and non-DM AIS patients by mitigating cerebral lactic acidosis and other detrimental metabolic processes that expedite ischemic brain damage.Nonetheless, more research is required to ascertain the clinical advantages of aggressive insulin treatment versus standard care.Animal models have suggested that lowering blood sugar with insulin can reduce ischemic brain damage [33].Previous studies, including the UK Glucose Insulin in Stroke Trial [34], have demonstrated that insulin infusions lead to a significant decrease in plasma glucose levels compared to the saline group.However, despite this reduction, no noticeable improvement in clinical outcomes was observed.The SHINE study [35] categorized AIS patients into two groups: those treated with subcutaneous insulin on a sliding scale and those treated with continuous intravenous insulin.The study found no significant difference in 90-day functional outcomes between the two treatment groups, but it is possible that the study was underpowered.
There are some cautions in determining SHR in AIS patients.Based on several previous studies [36][37][38], SH is defined as absolute hyperglycemia without deterioration of pre-illness glycemic control in patients with pre-existing DM, without considering background glucose levels, which makes it difficult to distinguish SH in DM patients.Relative hyperglycemia might be a better predictor of critical disease outcomes than absolute hyperglycemia [39].However, CKD, anemia, and hemoglobinopathies could affect HbA1c measurement accuracy [40].Therefore, it is important to consider the interpretation of SHR in patients with these conditions.
The research had some limitations.Firstly, it was a single-center, retrospective observational study, which emphasized the importance of conducting multi-site clinical trials.Secondly, we did not determine the glycemic variability in this study due to lack of precise information and retrospective nature.Thirdly, the study had a small number of DM patients, which may have resulted in low power to detect the same relative risk in DM patients as in non-DM patients.Lastly, the study included only Thai patients; therefore, these findings cannot be generalized to other ethnicities.

Conclusions
In summary, our study demonstrated that both SHR1 and SHR3 are independently associated with fatal outcomes.Notably, SHR1 emerged as the most valuable biomarker for predicting fatal outcomes following rt-PA treatment.We recommend close monitoring of patients with elevated SHR1 levels upon admission, especially in non-DM, SBP� 180 mmHg, and NIHSS<16.Future studies are required to determine whether enhancing clinical prediction models with the combination of SHR and traditional risk factors enhances the accuracy of predicting fatal ischemic stroke outcomes.

Table 3 .
(Continued) Variables adjusted for are as follows: model A is age and sex; model B is model A+ TOAST classification, NIHSS, and baseline ASPECTS �7; model C is model B + comorbidities (DM, CKD, MI, CHF, preexisting disability, and history of malignancy) ‡ Variables adjusted for are as follows: model A is age and sex; model B is model A+ TOAST classification, NIHSS, baseline ASPECTS �7, SBP, and DBP; model C is model B + comorbidities (Hypertension and DM) § Variables adjusted for are as follows: model A is age and sex; model B is model A+NIHSS, baseline ASPECTS �7, SBP, and DBP; model C is model B + comorbidities (Hypertension, DM, prior use antiplatelet, onset to treatment time, and antihypertensive before rt-PA) https://doi.org/10.1371/journal.pone.0297809.t003 †