Stress hyperglycemia is associated with in‐hospital mortality in patients with diabetes and acute ischemic stroke

Abstract Background and Objective Stress hyperglycemia may occur in diabetic patients with acute severe cerebrovascular disease, but the results regarding its association with stroke outcomes are conflicting. This study aimed to examine the association between stress‐induced hyperglycemia and the occurrence of in‐hospital death in patients with diabetes and acute ischemic stroke. Research Design and Methods All data were from the Chinese Stroke Center Alliance (CSCA) database and were collected between 2016 and 2018 from >300 centers across China. Patients’ demographics, clinical presentation, and laboratory data were extracted from the database. The primary endpoint was in‐hospital death. The ratio of fasting blood glucose (FBG) to HbA1c was calculated, that is, the stress‐induced hyperglycemia ratio (SHR), to determine stress hyperglycemia following acute ischemic stroke. Results A total of 168,381 patients were included. The mean age was 66.2 ± 10.7, and 77,688 (43.0%) patients were female. The patients were divided into two groups: survivors (n = 167,499) and non‐survivors (n = 882), as well as into four groups according to their SHR quartiles (n = 42,090–42,099/quartile). There were 109 (0.26%), 142 (0.34%), 196 (0.47%), and 435 (1.03%) patients who died in the Q1, Q2, Q3, and Q4 quartiles, respectively. Compared with Q1 patients, the death risk was higher in Q4 patients (odds ratio (OR) = 4.02) (adjusted OR = 1.80, 95% confidence interval [CI] = 1.10–2.92, p = 0.018 after adjustment for traditional cardiovascular risk factors). The ROC analyses showed that SHR (AUC = 0.667, 95% CI: 0.647–0.686) had a better predictive value for mortality than that of fasting blood glucose (AUC = 0.633, 95% CI: 0.613–0.652) and HbA1c (AUC = 0.523, 95% CI: 0.504–0.543). Conclusions The SHR may serve as an accessory parameter for the prognosis of patients with diabetes after acute ischemic stroke. Hyperglycemia in stroke patients with diabetes mellitus is associated with a higher risk of in‐hospital death.


| BACKG ROU N D
Acute stress hyperglycemia is a common manifestation found in patients presenting to the emergency room with acute cerebrovascular disease. Acute stress hyperglycemia is not only associated with the severity of stroke 1,2 but also with poor outcomes of stroke, especially in patients without diabetes mellitus. 3 On the other hand, the association between acute hyperglycemia and the outcomes of patients with diabetes mellitus is controversial, not only for stroke but also for other critical illnesses. 1,4,5 Indeed, a meta-analysis of stroke and hyperglycemia demonstrated that stress hyperglycemia in nondiabetic patients was associated with an increased risk of mortality after stroke (pooled relative risk (RR) = 3.07, 95% confidence index (CI): 2.50-3.79), but this was not observed in patients with a history of diabetes. 6 This phenomenon is supported by other cohort studies 4,7 and was also observed in other critical illnesses, that is, that acute hyperglycemia with pre-existing diabetes mellitus led to lower mortality and shorter length of ICU stay than in patients without diabetes. 6,8,9 There is no unified definition of stress hyperglycemia, 10 and the patients are generally classified as known diabetes, newly diagnosed diabetes, and hospital-related hyperglycemia. 3,11,12 Most of the previous studies simply used fasting glucose or initial blood glucose at admission to determine the presence of stress hyperglycemia, without considering the usual glucose levels before stroke onset. This might explain why stress hyperglycemia cannot predict the outcome of stroke in patients with a history of diabetes because of high background glucose levels. Therefore, background blood glucose levels should be considered when assessing the relationship between stress hyperglycemia and the outcomes of critical illness, especially for patients with pre-existing diabetes. The stress hyperglycemia ratio (SHR) is a new method for determining blood glucose stress. It also considers glycated hemoglobin (HbA1c) (which represents the blood glucose levels over the last 2-3 months) and random blood glucose after the stress events. 13 A study on acute ischemic stroke in patients with diabetes showed that the use of the glycemic gap and SHR as indicators of stress hyperglycemia could be better predictors for the severity and poor outcome of stroke, 14 but this study was a single-center, small sample study, limiting the generalizability of its results. In addition, most previous studies used admission glucose, which can be influenced by the diabetic status and the food consumed over the previous hours. 10,13,[15][16][17] Therefore, fasting blood glucose (FBG) instead of random or admission glucose could be a more reliable marker, as previously suggested. 18 China is a large country with a high burden of cerebrovascular diseases. 19,20 With the support of the Chinese government, an important database was built at Tiantan Hospital. The aim of the present study was to examine the association between stress-induced hyperglycemia and the occurrence of in-hospital death in patients with diabetes and acute ischemic stroke. The study population included patients with acute cerebral infarction within 72 h of onset and with a previous history of diabetes or previous use of hypoglycemic drugs for > 6 months. Diabetes diagnosis was based on a self-reported history of diabetes confirmed by the medical records. Potential patients were excluded from the study if they had incomplete information on in-hospital mortality or had missing HbA1c and FBG data.

| Data collection and outcome assessment
In the CSCA, data are directly entered by each center using a webbased patient data collection interface by trained registrars. 21 Baseline information was extracted from the database, including demography, vascular risk factors, infarction location, laboratory data, and clinical data. Vascular risk factors included a medical history of hypertension, atrial fibrillation, coronary artery disease, and current smoking. Clinical data included systolic and diastolic blood pressure, baseline National Institutes of Health Stroke Score (NIHSS), leukocytes, alcohol intake, body mass index (BMI), and brain natriuretic peptide (BNP) at admission. Current smoking was defined as smoked at least one cigarette per day for the previous year or more. Laboratory data included baseline HbA1c, fasting blood sugar, and routine blood biochemical variables that were obtained following an 8-12 h fast within the first 24 h after admission. Laboratory data were from the certified central hospital laboratories.

HbA1c, and SHR
To determine the presence of stress hyperglycemia following an acute ischemic stroke, the ratio of AG (random glucose at admission) to HbA1c was used, which is named the SHR. 13 In order to eliminate the confounding effect of food, AG in the formula was replaced by FBG (fasting blood glucose) in the present study, as supported by a previous study.

| Statistical analysis
The data were tested for normal distribution using the Kolmogorov-Smirnov test. The continuous variables were expressed as means ± SD or as medians with interquartile ranges according to their distribution (normal or skewed). Statistical comparisons of continuous variables were performed using Student's t test or Wilcoxon rank-sum test (comparisons of two groups), or ANOVA or the Kruskal-Wallis U test (comparison of more than two groups). Categorical variables were expressed as numbers and percentages and analyzed using the chi-square test or Fisher's exact test. We evaluated the association between SHR and in-hospital death using multivariable logistic regression analysis adjusted for potential confounders, including age, sex, BMI, NIHSS on admission, hypertension, atrial fibrillation, previous ischemic stroke, previous myocardial infarction, SAH, antiplatelets, anticoagulation, lipid-lowering drug, smoking, alcohol, LDL-C, FBG, HbA1c, eGFR, HCY, systolic blood pressure, and diastolic blood pressure.
Receiver operating characteristics (ROC) curves were generated to examine the predictive value of SHR, FBG, and HbA1c, based on the area under the ROC curve (AUC). All statistical analyses were performed using SAS 9.4 (SAS Institute). A SAS macro named %ggBaseline was used to analyze and report the baseline characteristics automatically. 22 P-values < 0.05 (two-sided) were deemed statistically significant.

| Characteristics of the patients
A total of 181,111 diabetic patients with acute ischemic stroke were available in the CSCA database; 341 patients were excluded for incomplete data regarding in-hospital mortality, and 12,389 patients were excluded for missing glucose data. Therefore, 168,381 patients were included in this study ( Figure 1). The mean age was 66.2 ± 10.7, and 77,688 (43.0%) patients were female.

| Characteristics of the patients according to the vital status
The patients were divided into two groups according to the outcome during hospitalization: survivors (n = 167,499) and nonsurvivors (n = 882). The characteristics of the two groups are shown in Table 1. The patients in the non-survivor group were more likely to display traditional cardiovascular risk factors such as being older (p < 0.0001), hypertension (p = 0.0004), history of F I G U R E 1 Patient flow diagram. CSCA, China Stroke Center Alliance myocardial infarction (p < 0.0001), higher initial fasting blood glucose (p < 0.0001), higher HbA1c (p = 0.018), lower eGFR (p < 0.0001), higher homocysteine (p < 0.0001), and higher systolic blood pressure (p < 0.0001), as well as factors associated with poor stroke prognosis, such as higher NIHSS (p < 0.0001), history of transient ischemic attack (p = 0.024), history of stroke (p < 0.0001), cerebral hemorrhage (p = 0.033), subarachnoid hemorrhage (p = 0.004).
Regarding treatments, the non-survivors were more likely to be on antiplatelet (p = 0.001), anticoagulants (p < 0.0001), and lipidlowering drugs (p = 0.003) and to have received reperfusion therapy (p < 0.001), suggesting that they had the indications for such treatments. Otherwise, the non-survivors showed characteristics usually associated with a better prognosis, such as not smoking (p < 0.0001) and not drinking (p = 0.007). Regarding the specific objective of the present study, the non-survivors had a higher SHR compared with the survivor group (p < 0.0001).

| SHR and in-hospital death
The patients were divided into four groups according to their SHR quartiles. The characteristics are presented in Table 2. We identified a potential association between the occurrence of in-hospital death and SHR in diabetic patients with acute ischemic stroke in the unadjusted (P for trend <0.0001) and adjusted (P for trend = 0.0141) models (Table 3). Multicollinearity was investigated, and no multicollinearity was found. There were 109 (0.26%), 142 (0.34%), 196 (0.47%), and 435 (1.03%) patients who died in the Q1, Q2, Q3, and

| DISCUSS ION
Stress hyperglycemia can occur in diabetic patients with acute severe cerebrovascular disease, 1,3,11,14 but the results regarding its association with stroke outcomes are conflicting. Therefore, this study aimed to examine the association between stress-induced hyperglycemia and the occurrence of in-hospital death in patients with diabetes and acute ischemic stroke. The study used the data from the CSCA, which is a national initiative to improve the burden of stroke in China. 21 The results suggest that the SHR may serve as an accessory parameter for the prognosis of patients with diabetes after acute ischemic stroke.
Previous studies showed that stress hyperglycemia is associated with poor outcomes in patients with severe illnesses such as acute myocardiac infarction, trauma, and acute ischemic stroke. 6,8,9 Zarean et al. 23 showed that the glycemic gap was associated with mortality in patients with diabetes and hemorrhagic stroke. Nevertheless, the association between stress hyperglycemia and acute ischemic stroke in patients with diabetes was rarely reported. Several studies showed that admission hyperglycemia was significantly associated with poor short-term outcomes of ischemic stroke after mechanic thrombectomy. [15][16][17] Nevertheless, these studies did not further analyze the association of admission hyperglycemia with the outcome of acute ischemic stroke in patients with diabetes. It is believed that acute hyperglycemia has a distinct association with the increased risk of poor outcome or mortality in non-diabetic patients. 3,24,25 For patients with pre-diabetes history, this relationship is controversial.
Because hyperglycemia is a common trait among patients with DM, absolute hyperglycemia cannot reflect the changes in glucose under critical conditions without consideration of the basic glucose level, especially when blood glucose is poorly controlled. This might be a reasonable explanation for the paradoxical relationship between absolute hyperglycemia and mortality in patients with diabetes. 3,7 Therefore, the relative stress hyperglycemia indicators of glycemic gap and SHR, which take background blood glucose into consideration, should have a better prediction of the poor outcome of critical illness. By using the glycemic gap, Zarean et al. 23 could exclude the impact of poor glycemic control in the outcome of diabetic patients with hemorrhagic stroke. Lee et al. 26 showed that SHR could predict the in-hospital mortality in critically ill patients across the glycemic spectrum, while absolute glycemia could not. A recent study showed that the SHR was a predictor of poor outcomes in non-diabetic patients, but that study included only 18.1% (29/160) of diabetic patients. 27 Merlino et al. 28 showed in 414 patients (irrespective of the diabetes status) that stress hyperglycemia was associated with poor outcomes in patients with acute ischemic stroke after intravenous thrombolysis and that the quartiles of SHR had significant trends for poor outcomes and mortality. In patients with minor stroke or TIA, patients with newly diagnosed diabetes had a risk of recurrence similar to patients with known diabetes, while patients with stress hyperglycemia had a markedly higher risk of recurrence. 29 Yuan et al. 10 reported that stress hyperglycemia was associated with the risk of hemorrhagic transformation, which could explain, at least in part, the higher mortality observed with high SHR. In the present study, the glucose-to-HbA1c ratio was used, which also accounts for glycemic control. Indeed, blood glucose levels fluctuate widely during the TA B L E 3 Adjusted hazard ratios of outcomes at 12 months according to glucose-to-HbA1c ratio quartiles course of the day and with disease conditions, while HbA1c levels represent the general glycemic control over the past 3 months. 13 In addition, the AUC of SHR for mortality was higher than for FBG and HbA1c.
Although the underlying mechanisms are too complex to be fully understood, it was proposed that the hyperactivated oxidative stress response, insulin resistance, inflammation, cytokine production, and hormonal derangements may account for the association between stress hyperglycemia and poor outcomes. 11,30 The excess mortality has been hypothesized to be due to the combined effects of glucagon, growth factors, catecholamines, and glucocorticoids, leading to gluconeogenesis, inflammation, and insulin resistance, 31 contributing to neuroinflammation and oxidative stress that exacerbates the brain injury. 32 In addition, high blood glucose is associated with brain-blood barrier breakdown, brain edema, and increased apoptosis. [33][34][35] Furthermore, elevated blood glucose could aggravate inflammation and oxidative stress response, potentially creating a vicious cycle that leads to further hyperglycemia. 36,37 Meanwhile, hyperglycemia promotes the release of excessive circulating free fatty acids, which also aggravates hyperglycemia. The overlapping interaction of glucotoxicity, lipotoxicity, and inflammation might contribute to a detrimental physiopathological vicious cycle. In addition, hyperglycemia accelerates neuronal damage in hypoxic brain tissue. 38 Hyperglycemia also increases the production of thrombinantithrombin complexes and the tissue factor pathway to stimulate coagulation. 39 The activation of protein kinase C and NADPH oxidase increases reactive oxygen species (ROS) levels and reduces nitric oxide synthase, thereby leading to decreased reperfusion and possibly neuron damage. 39 Evidence from MRI studies showed that admission hyperglycemia was associated with expanded infarction core and reduced penumbra salvage. 40,41 Therefore, the multiple intricate molecular mechanisms and pathological changes might lead to poor outcomes after stroke with stress hyperglycemia. These mechanisms probably affect nerve repair after stroke, impair collateral circulation, increase vascular permeability, increase platelet cohesion, and increase the occurrence of complications in patients.
The exact mechanisms will have to be examined.
Still, poor glycemic control has been shown to be associated with poor functional outcomes after stroke, 42 indicating that long-term glycemic stress and damage are involved in the functional prognosis of stroke, while acute hyperglycemia after stroke might be a predictor of death. Unfortunately, the present study had no data about the functional outcomes after stroke. This will have to be examined in future studies.
Even though stress hyperglycemia is associated with poor outcomes after stroke, intensive glycemic control is not actually recommended during the acute phase. So far, no evidence showed that tight blood glucose control was associated with better outcomes after acute ischemic stroke. A meta-analysis including 1296 patients with acute ischemic stroke from seven trials demonstrated that tight blood glucose control (i.e., maintaining glucose levels between 4.0 and 7.5 mmol/L) increased the risk of hypoglycemia events compared with the control group (OR = 25.9, 9.2-72.7). 43 The current stroke guidelines recommend that blood glucose be controlled F I G U R E 2 Receiver operating characteristic (ROC) analysis of stressinduced hyperglycemia ratio (SHR), fasting blood glucose, and HbA1c for predicting in-hospital death between 140 and 180 mg/dl, 44 but the guidelines do not make any distinction between patients with or without diabetes history.

The recently completed phase 3 trial Stroke Hyperglycemia Insulin
Network Effort (SHINE) demonstrated that intensive blood glucose control (80-130 mg/dl) did not improve the functional outcomes and even increased the risk of severe hypoglycemia. 45 To be worth mentioning, about 80% of the included patients had a history of diabetes. In view of the risks for diabetic patients, SHR is more meaningful than absolute hyperglycemia. Such a view is supported by a study in critically ill patients. 26 As suggested by a recent review, acute management of stroke could include aggressive glucose management in patients with ischemic stroke and hyperglycemia. 46 The present study has several limitations. First, we did not assess the blood glucose control during hospitalization because such data are not included in the database. Second, the design of this study was observational, and a cause-effect relationship cannot be determined. Although it is a national multi-center registration study with a huge amount of data, most of the enrolled patients have mild stroke severity, which makes the study inevitable choice bias. Third, no distinction is made between types 1 and 2 diabetes in the database. Thus, a prospective randomized large-scale study is expected to clarify this relationship.

| CON CLUS ION
This study showed that the SHR is significantly associated with an increased risk of in-hospital mortality in diabetic patients after acute ischemic stroke. This finding suggests that careful glycemic management is important for diabetic patients after stroke onset.

ACK N OWLED G M ENTS
We thank all the participating centers in the CSCA program for their hard work in data collection.

CO N FLI C T O F I NTE R E S T
All authors declare that they have no competing interests.

AUTH O R CO NTR I B UTI O N S
DHM and ZXL wrote the manuscript and interpreted the data. HQG and YYJ contributed to review the manuscript and statistical problems. XQZ and YLW reviewed the manuscript. YJW interpreted data, reviewed and edited the manuscript, and contributed to discussions.
All authors read and approved the final manuscript.