a b s t r a c t

Acute hyperglycemia is a common condition among patients with diabetes who are admitted to the emergency department (ED) for acute ischemic stroke . Previous findings regarding the association between hypergly- cemia at admission and adverse outcomes among patients with diabetes and AIS have been inconsistent. When investigating this association, it is necessary to consider premorbid blood glucose control. The objective of the current study was to assess whether HbA1c-based adjusted glycemic variables were associated with unfavorable outcomes among patients admitted to the hospital for AIS. We retrospectively analyzed data from 309 patients who were hospitalized for AIS at a single medical center in Taiwan between January 1, 2013, and October 31, 2015. We found that 1) HbA1c-based adjusted glycemic variables, including the glycemic gap and stress hyper- glycemia ratio, were associated with both AIS severity and neurological status at discharge; additionally, 2) HbA1c-based adjusted glycemic variables showed superior discriminative power compared with acute hypergly- cemia regarding the development of severe AIS. We conclude that both the glycemic gap and stress hyperglyce- mia ratio might be useful in assessing the disease severity and prognosis of patients presenting with AIS. Further prospective long-term follow-up studies should be performed to validate these findings.

(C) 2017

Introduction

Acute hyperglycemia frequently occurs in patients admitted to the emergency department (ED) for acute ischemic stroke [1,2].A sig- nificant association between initial glucose level and mortality among non-diabetic ischemic stroke patients has been observed [3]. However, whether hyperglycemia itself leads to neuronal damage and results in a worse prognosis or whether this factor simply reflects greater neuro- logical severity through stress-induced hyperglycemia (SIH) remains unclear. SIH is common among patients with critical illness [4]. SIH is attributed to excess levels of counter-regulatory hormones, anti-

? The work was performed at Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan.

* Corresponding author at: Department of Emergency Medicine, Tri-Service General Hospital, National Defense Medical Center, No. 325, Sec. 2, Cheng-Kung Road, Taipei, Taiwan.

E-mail address: [email protected] (S.-H. Tsai).

inflammatory cytokines, increased gluconeogenesis and hepatic insulin resistance [5-7]. Acute and mean hyperglycemia during hospitalization are associated with adverse outcomes. Nevertheless, the findings re- garding the association between hyperglycemia and adverse outcomes are inconsistent among acutely ill patients with diabetes mellitus . Because hyperglycemia is the cardinal feature of DM, pre-morbid hyperglycemia must be considered in patients with diabetes when in- vestigating the association between SIH and adverse outcomes. Long- term average glucose levels can be estimated by converting HbA1c values using an equation derived from an international multicenter study on HbA1c-derived average glucose (ADAG) [8]. Recent studies have shown the utility of several HbA1c-based adjustments of glycemic variables. For instance, our previous studies showed that an elevated glycemic gap was associated with adverse outcomes among patients with diabetes receiving care for community-acquired pneumonia and Liver abscesses at an intensive care unit (ICU) [9-11]. Prior research has also shown that the hyperglycemia ratio, which controls for baseline

glycemia, is a biomarker of risk for critical illness [12].

http://dx.doi.org/10.1016/j.ajem.2017.03.049

0735-6757/(C) 2017

Although hyperglycemia has been well studied as an index of poor prognosis among non-Diabetic patients with AIS, controversy remains regarding the use of acute hyperglycemia as a prognostic marker among patients with diabetes [13-15]. A previous study showed that hyperglycemia is a poor prognostic marker in patients with good glu- cose control (regardless of diabetic status), whereas the results have been inconsistent regarding patients with diabetes with poor glucose control [14]. Most clinical trials that have examined the prognosis of aggressive hyperglycemia control during the acute phase of AIS have revealed negative findings [15,16]. Nevertheless, studies examining the role of HbA1c-based adjusted glycemic variables in the assessment of disease severity and patient outcomes remain limited with regard to comorbid diabetes and AIS. We hypothesized that HbA1c-based ad- justment of acute hyperglycemia variables would be associated with AIS severity. The present study further explored the correlations be- tween two HbA1c-based glycemic variables (i.e., glycemic gap and stress hyperglycemia ratio) and the severity of AIS and clinical outcomes of patients admitted to the ED for AIS.

Materials and methods

Patients

We conducted a retrospective observational study of consecutive patients with diabetes who were admitted for AIS between January 1, 2013, and October 31, 2015. The institutional review board for human investigations at a tertiary referral medical center within Tri-Service General Hospital approved this study and waived the need for informed consent. The study was implemented according to the approved guide- lines. Patients with AIS were identified by searching for cases with Inter- national Classification of Diseases (9th revision) codes 433, 434, and

436. A neurologist evaluated all patients upon hospital admission to es-

tablish initial disease severity using the National Institutes of Health Stroke Scale . All patients received a cranial computed tomogra- phy scan at the ED. Data from the individual identified patients were then reviewed to select patients with AIS who had available plasma glu- cose level data within 30 min of initial presentation and data on HbA1c levels within 1 month before or immediately after admission. Patients who were younger than 18 years old, provided incomplete data, used steroids, were hypoglycemic (blood glucose b 70 mg/dL), were diag- nosed with diabetic ketoacidosis or a hyperosmolar hyperglycemic state, or received blood transfusions were excluded from this study.

Methods

We retrospectively reviewed patients’ medical records to collect data regarding patient age, gender, and underlying comorbidities as well as to collect laboratory data including plasma glucose level at initial presentation, HbA1c levels (measured within 1 month before or after admission), and length of hospital stay. We further assessed the severity of the indexed AIS events using the NIHSS, the Modified Rankin scale (MRS) and the Barthel index [17]. The patients were divided into those with NIHSS scores b 8 (mild stroke) and NIHSS scores ? 8 (moder-

ate to severe stroke) [18].

Serum glucose leve”>Determination of serum glucose levels, the HbA1c glycemic gap and stress hyperglycemia ratio

Serum glucose level was defined as the level measured upon ED ad- mission. HbA1c assays were performed using a blood analyzer (Primus CLC 385; Primus Corporation, Kansas City, MO, USA) equipped with a high-performance liquid chromatography system. Long-term average glycemic levels were estimated by converting HbA1c values using the equation derived from an international multicenter study on ADAG. The formula 28.7 x HbA1c – 46.7 was used to convert HbA1c levels to es- timated ADAG over 3 months [8]. The glycemic gap, which represents

the changes in serum glucose levels during the index event, was calcu- lated as the glucose level at ED admission minus the ADAG; the stress hyperglycemia ratio was calculated as the serum glucose level at ED ad- mission divided by HbA1c.

Statistical analyses

Continuous data were expressed as the mean +- standard deviation and were analyzed using two-tailed Student’s t-tests. Categorical data were expressed as frequencies (%) and evaluated using chi-square test or Fisher’s exact test. One-way analyses of variance were used to assess the significance of various characteristics, laboratory data, and adverse outcomes. A receiver-operator characteristic curve (ROC) was generat- ed to analyze the discriminative power of the Prediction tools, and the area under the ROC (AUROC) and its 95% confidence interval (CI) were calculated. Univariate and multivariate logistic regression analyses were performed to identify the risk factors associated with moderate to severe AIS. Variables associated with p b 0.05 in the univariate analysis were entered into the multivariate logistic regression analysis. The data were analyzed using SPSS version 22.0 (SPSS, Inc., Chicago, IL, USA), and differences with p values b 0.05 were considered significant.

Results

Demographic data and clinical outcomes

We initially identified 369 patients with diabetes whose HbA1c was assessed within one month or immediately after admission to the ED for AIS. Patients were excluded from the study for steroid use (n = 12), hypoglycemia (n = 16), diabetic ketoacidosis or a hyperosmolar hyper- glycemic state (n = 7) and incomplete data (missing NIHSS at dis- charge, n = 25). We enrolled 309 patients after a subsequent chart review. The demographic data and AIS-related clinical features of the enrolled patients are shown in Table 1. Compared with patients with mild stroke, patients with moderate to severe stroke were significantly older, more likely to smoke, had a higher glycemic gap, a higher stress hyperglycemia ratio and stayed longer at the hospital.

Correlations between acute hyperglycemia, long-term blood glucose control, glycemic gap, stress hyperglycemia ratio, and AIS severity

As Table 2 shows, the multivariate logistic regression analysis revealed that the odds ratios of the glycemic gap (mg/dL) and stress hyperglycemia ratio for moderate to severe AIS were 1.007 (95% CI = 1.003-1.012, p = 0.001) and 1.061 (95% CI = 1.025-1.098, p =

0.001), respectively. As Fig. 1 shows, significant correlations were found between the glycemic gap and NIHSS at admission (r = 0.285,

Table 1

Demographic data of the enrolled patients.

	Mild stroke (n = 198)	Moderate to severe stroke (n = 111)	p Value
Age (yrs)	69.4 +- 15.4	73.7 +- 13.3	0.009?
Male	123 (62.1%)	73 (65.8%)	0.520
Hypertension	152 (76.8%)	92 (82.9%)	0.206
Diabetes mellitus	165 (81.3%)	94 (84.7%)	0.757
Dyslipidemia	71 (35.9%)	30 (27.0%)	0.269
Atrial fibrillation	25 (12.6%)	11 (9.9%)	0.475
Coronary artery disease	67 (33.8%)	50 (45.0%)	0.051
peripheral artery disease	4 (2.0%)	3 (2.7%)	0.699
Smoking	76 (38.4%)	30 (27.0%)	0.044?
Glucose at admission, mg/dL	200.2 +- 78.1	212.6 +- 90.1	0.211
HbA1c, %	8.4% +- 2.2	7.9 +- 2.2	0.075
Glycemic gap, mg/dL	6.5 +- 54.2	32.2 +- 70.3	b0.001	*
Stress hyperglycemia ratio	23.8 +- 6.5	27.0 +- 9.0	0.001?
Large vessel stroke	99 (50%)	81 (73.0%)	b0.001?
Hospital stay (days)	9.2 +- 10.5	15.9 +- 11.9	b0.001?

* p b 0.05.

Table 2

Univariate and multivariate logistic regression analyses regarding the development of moderate to severe AIS.

	Univariate	Multivariate
	HR (95% CI)	p Value		HR (95% CI)	p Value
Age	0.991 (0.976-1.007)	0.283	Age	0.991 (0.975-1.008)	0.317
Gender	1.353 (0.827-2.213)	0.229	Gender	1.152 (0.684-1.940)	0.594
Hypertension	1.465 (0.809-2.654)	0.207	Atrial fibrillation	2.279 (1.095-4.745)	0.028?
Diabetes mellitus	1.106 (0.585-2.092)	0.757	Smoking	0.589 (0.347-1.001)	0.050
Hyperlipidemia Atrial fibrillation	0.673 (0.406-1.116) 2.119 (1.091-4.432)	0.125	Glycemic gap	1.007 (1.003-1.012)	0.001?
Coronary artery disease	1.603 (0.996-2.579)	0.052	–	–	–
Peripheral artery disease	0.708 (0.135-3.172)	0.683	Age	0.992 (0.975-1.009)	0.329
Smoking	0.595 (0.358-0.988)	0.045?	Gender	1.159 (0.689-1.950)	0.579
BS at admission	1.002 (0.999-1.005)	0.211	Atrial fibrillation	2.281 (1.095-4.745)	0.028?
HbA1c	0.905 (0.811-1.001)	0.076	Smoking	0.592 (0.349-1.005)	0.052
Glycemic gap Stress hyperglycemia ratio	1.007 (1.003-1.012) 1.060 (1.025-1.096)	0.001?	Stress hyperglycemia ratio	1.061 (1.025-1.098)	0.001?

0.028?

BS: Blood sugar; HR: hazard ratio; CI, confidence interval.

0.001?

Multivariate analysis included all variable with p values b 0.05 in the univariate analysis, age and gender.

* p b 0.05.

p b 0.001), NIHSS at discharge (r = 0.285, p b 0.001), MRS at discharge (r = 0.250, p b 0.001) and the Barthel index at discharge (r = 0.286, p b 0.001). Significant correlations were also found between the stress hyperglycemia ratio and NIHSS at admission (r = 0.284, p b 0.001),

NIHSS at discharge (r = 0.297, p b 0.001), MRS at discharge (r = 0.254, p b 0.001) and the Barthel index at discharge (r = 0.257, p b 0.001) (Fig. 2). Compared with acute hyperglycemia (defined as a blood glucose level >= 250 mg/dL), long-term Glycemic control and the

Fig. 1. Correlations between the glycemic gap and neurological outcomes.

stress hyperglycemia ratio (0.588, 95% CI = 0.521-0.655), the glycemic gap showed the greatest AUROC (0.590, 95% CI = 0.524-0.657) regard- ing the development of moderate to severe stroke (Fig. 3). We deter- mined an optimal cutoff value of 45 mg/dL using the maximal Youden index, which showed a sensitivity, specificity, positive predictive and negative predictive value of 39.6%, 75.8%, 47.8% and 69.1% for the devel- opment of moderate to severe stroke, respectively. As Table 3 shows, no significant differences were found with regard to comorbidities be- tween patients with and without an elevated glycemic gap. An addition- al analysis revealed that patients with an elevated glycemic gap (N 45 mg/dL) showed significantly higher NIHSS at admission (p b 0.001), poorer MRS at discharge (p = 0.003), a poorer Barthel index at discharge (p = 0.022) and worse NIHSS (p = 0.006) at discharge compared with patients with a glycemic gap b 45 mg/dL. We then used the maximal Youden index to determine an optimal cutoff value for the stress hyperglycemia ratio of 27.59, which yielded a sensitivity, specificity, positive predictive and negative predictive value of 46.2%, 75.0%, 52.9% and 69.6% for the development of moderate to severe stroke, respectively. As Table 4 shows, an additional analysis revealed that patients with an elevated stress hyperglycemia ratio (N 27.59) had significantly higher NIHSS at admission (p b 0.001), poorer MRS at discharge (p = 0.008), a poorer Barthel index at discharge (p = 0.049) and poorer NIHSS at discharge (p = 0.001) compared with pa- tients with a stress hyperglycemia ratio b 27.59. There were no

statistically significant differences regarding the rates of large vessel dis- eases based on the above-mentioned cut-off values of glycemic gap and stress hyperglycemia ratio. As Table 5 shows, an additional analysis re- vealed that there were no significant differences in HbA1c-based adjust- ed glycemic variables between large vessel and small vessel stroke.

In addition, Table 6 shows that patients with adequate glycemic con- trol in terms of having HbA1c values b 7% had a significantly higher mortality rate (p = 0.048), higher NIHSS upon admission (p b 0.013) and higher NIHSS at discharge (p b 0.037) than patients with poor gly- cemic control.

Discussion

The major findings of the present study were as follows: 1) HbA1c- based adjusted glycemic variables including the glycemic gap and stress hyperglycemia ratio eliminated the confounding effects of premorbid glycemic control and were associated with AIS severity and neurological status at discharge; 2) deteriorating effects of SIH might occur during the early stages of AIS; and 3) compared with acute hyperglycemia, HbA1c-based adjusted glycemic variables showed superior discrimina- tive power regarding the development of moderate to severe AIS among patients with diabetes.

In a non-diabetic population, acute hyperglycemia in patients with AIS has been strongly associated with an increased risk of poorer

Fig. 2. Correlations between the stress hyperglycemia ratio and neurological outcomes.

Fig. 3. ROC of acute hyperglycemia, chronic blood glucose control, glycemic gap, stress hyperglycemia ratio and the development of moderate to severe stroke in patients admitted to the ED for AIS.

outcomes and mortality [19,20]. However, the association between ini- tial glucose levels and the risk of death was not significant among pa- tients with diabetes and AIS, especially those with poor glycemic control [15,21]. Similarly, glucose level at admission was independently associated with increased mortality in non-patients with diabetes, whereas several studies have reported a relatively weak relationship between hyperglycemia and acute illness-related mortality among

Table 3

Comparison of patients with and without an elevated glycemic gap.

Table 4

Comparison of patients with and without an elevated stress hyperglycemia ratio.

	Stress hyperglycemia ratio b 27.59 (n = 207)	Stress hyperglycemia ratio >= 27.59 (n = 102)	p Value
Age (yrs)	70.4 +- 14.2	72.3 +- 12.7	0.237
Male	132 (63.8%)	62 (60.8%)	0.610
Hypertension	160 (77.3%)	84 (82.4%)	0.305
Diabetes mellitus	178 (86.0%)	81 (79.4%)	0.140
Dyslipidemia	72 (34.8%)	31 (30.4%)	0.441
Coronary artery disease	78 (37.7%)	39 (38.2%)	0.925
Peripheral artery disease	5 (2.4%)	2 (2.0%)	0.801
Atrial fibrillation	26 (12.6%)	10 (9.8%)	0.478
Smoking	74 (35.7%)	32 (31.4%)	0.446
NIHSS at admission Modified Rankin’s scale	7.0 +- 6.6	10.7 +- 9.0	b0.001? 0.008?
at discharge
0-3	138 (66.7%)	52 (51.0%)
4-6 Barthel index at discharge	69 (33.3%)	54 (49.0%)	0.049?
5-100	201 (97.1%)	94 (92.2%)
0	6 (2.9%)	8 (7.8%)
neurological complications	20 (9.7%)	11 (10.8%)	0.757
Non-neurological	51 (24.6%)	25 (24.5%)	0.980
complications
Large vessel stroke	117 (56.5%)	63 (61.8%)	0.379
NIHSS at discharge	6.4 +- 7.1	10.1 +- 10.1	0.001?
Hospital stay (days)	10.6 +- 11.2	11.8 +- 11.2	0.377
Mortality	10 (4.8%)	7 (6.9%)	0.461

NIHSS: National Institute of Health Stroke Scale; neurological complications: included de- terioration in NIHSS N 2 after admission, herniation, and Hemorrhagic transformation; non-neurological complications: included pneumonia, urinary tract infection, upper gastrointestinal tract bleeding, acute coronary syndrome, acute pulmonary edema, and pressure sore.

* p b 0.05.

critically ill patients with diabetes [13,14,22-24]. Even the highest glu- cose values during the first 24 h after ICU admission did not predict hospital mortality [25]. One plausible explanation for the inconsistent results is the lack of consideration of patients’ premorbid glycemic control. Adjusting for ADAG may control for the possible influence of long-term glycemic control on severity or outcomes. This belief is con-

sistent with the “diabetes paradox”, in which diabetes and glycemic

	Glucose-ADAG b 45 mg/dL (n = 217)	Glucose-ADAG >= 45 mg/dL (n = 92)	p Value
Age (yrs)	70.3 +- 14.2	72.5 +- 12.6	0.191
Male	131 (60.4%)	63 (68.5%)	0.177
Hypertension	169 (77.9%)	75 (81.5%)	0.473
Diabetes mellitus	187 (86.2%)	72 (78.3%)	0.084
Dyslipidemia	75 (34.6%)	28 (30.4%)	0.482
Coronary artery disease	82 (37.8%)	35 (38.0%)	0.966
Peripheral artery disease	5 (2.3%)	2 (2.2%)	0.944
Atrial fibrillation	27 (12.4%)	9 (9.8%)	0.505
Smoking	76 (35.9%)	28 (30.4%)	0.351
NIHSS at admission Modified Rankin’s scale at	7.1 +- 6.6	11.0 +- 9.2	b0.001? 0.003?

control are not independently associated with the severity of illness of critically ill patients. In this context, the debate concerning the associa- tions between acute hyperglycemia, long-term glucose control, and certain adverse clinical outcomes can be reasonably explained [26]. The results of the present study were also consistent with those of our previous work regarding the association between an elevated glycemic gap and adverse outcomes among patients with diabetes [9-11]. Al- though the discriminative power of the glycemic gap and stress hyper- glycemia ratio regarding the development of moderate to severe AIS was limited, we propose that HbA1c-adjusted glycemic variables (and not acute hyperglycemia alone) might be useful for identifying disease

discharge

0-3 145 (66.8%) 45 (48.9%)

4-6 72 (33.2%) 47 (51.1%)

Barthel index at discharge 0.022?

severity in patients with diabetes.

SIH might initiate a vicious cycle by increasing insulin resistance, free fatty acid production, Vascular inflammation, vascular permeability,

5-100	211 (97.2%)	84 (91.3%)	nitric oxide inactivation and Reactive oxygen species production and
0	6 (2.8%)	8 (8.7%)	thereby create a prothrombotic state [27,28]. Patients with SIH might

be more susceptible to Ischemia-reperfusion injury due to increased oxidative stress, pro-inflammation status and activation of stress-

Neurological complications	20 (9.4%)	11 (11.9%)	0.565
Non-neurological	53 (24.4%)	23 (25.0%)	0.953
complications
Large vessel stroke	125 (57.6%)	55 (59.8%)	0.722
NIHSS at discharge	6.4 +- 7.3	9.7 +- 10.0	0.006?
Hospital stay (days)	10.7 +- 11.1	11.7 +- 11.3	0.542
Mortality	11 (5.1%)	6 (6.5%)	0.609

Table 5

Comparison of A1c-based adjusted glycemic variables in patients with large and small ves- sel diseases.

NIHSS: National Institute of Health Stroke Scale; neurological complications: included de-

terioration in NIHSS N 2 after admission, herniation, and hemorrhagic transformation; non-neurological complications: included pneumonia, urinary tract infection, upper gastrointestinal tract bleeding, acute coronary syndrome, acute pulmonary edema, and pressure sore.

Glycemic gap	18.3 +- 67.0	12.2 +- 53.3	0.395
Stress hyperglycemia ratio	25.2 +- 8.2	24.6 +- 6.8	0.539

* p b 0.05.

Large vessel stroke (n = 180)

Small vessel stroke (n = 129)

p Value

Table 6

Clinical outcomes versus chronic glycemic control among patients with AIS.

the outcomes. The present study did not specifically address the effects of glycemic control during hospitalization. Although the discriminative

HbA1c b 7%

(n = 103)

7% <= HbA1c b 9%

(n = 108)

HbA1c >= 9%

(n = 98)

p Value

power of the glycemic gap and stress hyperglycemia ratio was limited in this AIS setting compared with in our previous studies, [9-11] we

Mortality 9 (8.7%) 7 (6.5%) 1 (1.0%) 0.048?

NIHSS at admission 9.9 +- 9.1 7.9 +- 7.6 6.8 +- 5.8 0.013?

NIHSS at discharge 9.6 +- 10.5 7.9 +- 9.0 6.4 +- 6.5 0.037?

MRS at discharge 0.442

0-3 60 (58.3%) 64 (59.3%) 65 (66.3%)

4-6 43 (41.7%) 44 (40.7%) 33 (33.7%)

BI at discharge 0.306

5-100	95 (92.2%)	103 (95.4%)	95 (96.9%)
0	8 (7.8%)	5 (4.6%)	3 (3.1%)
Neurological 32(31.1%)		25 (23.1%)	19 (19.4%)	0.144
complications
Non-neurological	7 (18.4%)	12 (28.7%)	12 (26.5%)	0.220
complications Hospital stay (days)	11.8 +- 11.6	9.9 +- 8.3	12.2 +- 13.4	0.443

NIHSS: National Institute of Health Stroke Scale; MRS: Modified Rankin Scale; BI: Barthel index; neurological complications: included deterioration in NIHSS N 2 after admission, herniation, and hemorrhagic transformation; non-neurological complications: included pneumonia, urinary tract infection, upper gastrointestinal tract bleeding, acute coronary syndrome, acute pulmonary edema, and pressure sore.

* p b 0.05.

responsive kinases [29,30,31]. Previous experiments have suggested that hyperglycemia increases neuronal damage in hypoxic brain tissue [32]. Within the first 48 h post-AIS, hyperglycemia has shown deleteri- ous effects on brain injury by increasing Brain edema and causing hem- orrhagic transformation and brain herniation [6,33]. Mechanistically, glucose fluctuation can activate nuclear factor-kB and the protein kinase C pathway, which leads to an increased expression of adhesion mole- cules and the excess formation of advanced glycation end products compared with stable glucose levels in vitro [34,35].

The associations between HbA1c levels and clinical outcomes re- main controversial. Although some studies have shown that poorer pre-stroke glycemic control is associated with increased Stroke severity, unfavorable long-term functional outcomes and poor survival, others have demonstrated that HbA1c paradoxically could not be used to pre- dict the clinical outcomes of patients with AIS [15,21,36-38]. The current study again found that patients with adequate pre-morbid glycemic control showed poorer neurological outcomes than patients with higher HAb1c levels. Physiological and cellular readjustments occur in re- sponse to hyperglycemia over time in these patients [39]. We speculate that the readjustment of blood glucose to a higher set-point value might influence the subsequent biological and clinical effects of acute fluctua- tions of blood glucose. A recent study also suggested that attempts to at- tain tight glucose control did not improve Healthcare outcomes [40]. Currently, the American Heart Association and the American Stroke As- sociation recommend that the ideal glucose level after AIS is between 140 and 180 mg/dL. Tight glycemic control (glucose b 126 mg/dL or b 7 mmol/L) or hypoglycemia is harmful and should be avoided [41, 42]. The Stroke Hyperglycemia Insulin Network Effort (SHINE) trial will hopefully contribute to clarifying the role of intensive glycemic con- trol during the acute phase of stroke [43]. We believe that these studies should be re-evaluated using HbA1c-based adjusted glycemic variables to assess the true benefits of blood glucose control during the acute phase while considering “relative hypoglycemia” among patients who have readjusted to a higher set-point of blood glucose.

Limitations

Our study had several limitations. First, our design was retrospective and might have been subject to selection bias. We believe that a pro- spective study on the use of the glycemic gap in non-diabetes patients should be conducted to further elucidate the usefulness of the glycemic variables that might assist with prognosis prediction. Second, the ade- quacy of glycemic control during hospitalization might have influenced

were still able to demonstrate that glycemic variables were superior to acute hyperglycemia and have the potential to serve as risk prediction items in future studies.

Conclusion

HbA1c-based glycemic variables were associated with disease sever- ity and neurological outcomes in patients presenting with AIS. Both the glycemic gap and stress hyperglycemia ratio might be used to assess the severity and prognosis of patients presenting with AIS. Further prospec- tive long-term follow-up studies should be performed to validate these findings.

Author contributions

C. J. Y. had full access to all of the study data and takes responsibility for the integrity and accuracy of the analysis; W. I. L., C. W. H. and S. H. T. designed the study; J. C. W, C. H. L and S. H. T. reviewed the literature and wrote the manuscript; C. L. T, J. T. L. and G. S. P collected the data;

C. J. Y. performed the statistical analysis; all of the authors reviewed and approved the final manuscript.

Additional information

The authors declare that they have no competing financial interests.

Acknowledgments

Grants from Tri-Service General Hospital, the National Defense Medical Center, Taipei, Taiwan (TSGH-C106-048), and the Ministry of Science and Technology, Taiwan (MOST-104-2314-B-106-043-MY2), supported this study.

References

1. Weir CJ, Murray GD, Dyker AG, Lees KR. Is hyperglycaemia an independent predictor of poor outcome after acute stroke? Results of a long-term follow up study. BMJ 1997;314(7090):1303-6.
2. Fuentes B, Castillo J, San Jose B, Leira R, Serena J, Vivancos J, et al. The prognostic value of capillary glucose levels in acute stroke: the glycemia in acute stroke (GLIAS) study. Stroke 2009;40(2):562-8.
3. Topczewska K, Drozdowski W, Chorazy M, Kulakowska A, Malyszko J. Hyperglyce- mia and diabetes have different impacts on outcome of ischemic and hemorrhagic stroke. Arch Med Sci AMS 2017;13(1):100-8.
4. Marik PE, Bellomo R. Stress hyperglycemia: an essential survival response! Crit Care

   2013;17(2):305.

   McCowen KC, Malhotra A, Bistrian BR. Stress-induced hyperglycemia. Crit Care Clin 2001;17(1):107-24.

5. Kavanagh BP, McCowen KC. Clinical practice. Glycemic control in the ICU. N Engl J Med 2010;363(26):2540-6.
6. Kavanagh BP, McCowen KC. Glycemic control in the ICU. N Engl J Med 2010; 363(26):2540-6.
7. Nathan DM, Kuenen J, Borg R, Zheng H, Schoenfeld D, Heine RJ. Translating the A1C assay into estimated average glucose values. Diabetes Care 2008;31(8):1473-8.
8. Chen PC, Liao WI, Wang YC, Chang WC, Hsu CW, Chen YH, et al. An elevated glyce- mic gap is associated with adverse outcomes in diabetic patients with community- acquired pneumonia. Medicine 2015;94(34):e1456.
9. Liao WI, Sheu WH, Chang WC, Hsu CW, Chen YL, Tsai SH. An elevated gap between admission and A1C-derived average glucose levels is associated with adverse out- comes in diabetic patients with pyogenic liver abscess. PLoS One 2013;8(5):e64476.
10. Liao WI, Wang JC, Chang WC, Hsu CW, Chu CM, Tsai SH. Usefulness of glycemic gap to predict ICU mortality in critically ill patients with diabetes. Medicine 2015; 94(36):e1525.
11. Roberts GW, Quinn SJ, Valentine N, Alhawassi T, O’Dea H, Stranks SN, et al. Relative hyperglycemia, a marker of critical illness: introducing the stress hyperglycemia ratio. J Clin Endocrinol Metab 2015;100(12):4490-7.
12. Egi M, Bellomo R, Stachowski E, French CJ, Hart GK, Hegarty C, et al. Blood glucose concentration and outcome of critical illness: the impact of diabetes. Crit Care Med 2008;36(8):2249-55.
13. Egi M, Bellomo R, Stachowski E, French CJ, Hart GK, Taori G, et al. The interaction of chronic and acute glycemia with mortality in critically ill patients with diabetes. Crit Care Med 2011;39(1):105-11.
14. Roquer J, Giralt-Steinhauer E, Cerda G, Rodriguez-Campello A, Cuadrado-Godia E, Jimenez-Conde J, et al. Glycated hemoglobin value combined with initial glucose levels for evaluating mortality risk in patients with ischemic stroke. Cerebrovasc Dis (Basel, Switzerland) 2015;40(5-6):244-50.
15. Gray CS, Hildreth AJ, Sandercock PA, O’Connell JE, Johnston DE, Cartlidge NE, et al. Glucose-potassium-Insulin infusions in the management of post-stroke hyperglycaemia: the UK glucose insulin in stroke trial (GIST-UK). Lancet Neurol 2007;6(5):397-406.
16. Steger C, Pratter A, Martinek-Bregel M, Avanzini M, Valentin A, Slany J, et al. Stroke patients with atrial fibrillation have a worse prognosis than patients without: data from the Austrian Stroke registry. Eur Heart J 2004;25(19):1734-40.
17. Briggs DE, Felberg RA, Malkoff MD, Bratina P, Grotta JC. Should mild or moderate stroke patients be admitted to an intensive care unit? Stroke 2001;32(4):871-6.
18. Capes SE, Hunt D, Malmberg K, Pathak P, Gerstein HC. Stress hyperglycemia and prognosis of stroke in nondiabetic and diabetic patients: a systematic overview. Stroke 2001;32(10):2426-32.
19. Ahmed N, Davalos A, Eriksson N, Ford GA, Glahn J, Hennerici M, et al. Association of admission blood glucose and outcome in patients treated with intravenous throm- bolysis: results from the safe implementation of treatments in stroke international stroke thrombolysis register (SITS-ISTR). Arch Neurol 2010;67(9):1123-30.
20. Hu GC, Hsieh SF, Chen YM, Hsu HH, Hu YN, Chien KL. Relationship of initial glucose level and All-cause death in patients with ischaemic stroke: the roles of diabetes mellitus and glycated hemoglobin level. Eur J Neurol 2012;19(6):884-91.
21. Lheureux O, Preiser JC. Year in review 2013: critical care-metabolism. Crit Care 2014;18(5):571.
22. Falciglia M, Freyberg RW, Almenoff PL, D’Alessio DA, Render ML. Hyperglycemia-re- lated mortality in critically ill patients varies with admission diagnosis. Crit Care Med 2009;37(12):3001-9.
23. Kaukonen KM, Bailey M, Egi M, Orford N, Glassford NJ, Marik PE, et al. Stress hyperlactatemia modifies the relationship between stress hyperglycemia and out- come: a retrospective observational study. Crit Care Med 2014;42(6):1379-85.
24. Freire AX, Bridges L, Umpierrez GE, Kuhl D, Kitabchi AE. Admission hyperglycemia and other risk factors as predictors of hospital mortality in a medical ICU population. Chest 2005;128(5):3109-16.
25. Krinsley JS, Fisher M. The diabetes paradox: diabetes is not independently associated with mortality in critically ill patients. Hosp Pract 1995;2012(40(2)):31-5.
26. Angeli F, Reboldi G, Poltronieri C, Lazzari L, Sordi M, Garofoli M, et al. Hyperglycemia in acute coronary syndromes: from mechanisms to prognostic implications. Ther Adv Cardiovasc Dis 2015.
27. Jafar N, Edriss H, Nugent K. The effect of short-term hyperglycemia on the innate im- mune system. Am J Med Sci 2016;351(2):201-11.
28. Webster KA. Stress hyperglycemia and enhanced sensitivity to myocardial infarc- tion. Curr Hypertens Rep 2008;10(1):78-84.
29. Zhou J, Wu J, Zhang J, Xu T, Zhang H, Zhang Y, et al. Association of stroke clinical out- comes with coexistence of hyperglycemia and biomarkers of inflammation. J Stroke Cerebrovasc Dis 2015;24(6):1250-5.
30. Huber T, Kleine JF, Kaesmacher J, Bette S, Poppert H, Zimmer C, et al. Blood leuko- cytes as prognostic parameter in stroke thrombectomy. Cerebrovasc Dis (Basel, Switzerland) 2016;42(1-2):32-40.
31. Luitse MJ, Biessels GJ, Rutten GE, Kappelle LJ. Diabetes, hyperglycaemia, and acute ischaemic stroke. Lancet Neurol 2012;11(3):261-71.
32. Bruno A, Liebeskind D, Hao Q, Raychev R. Diabetes mellitus, acute hyperglycemia, and ischemic stroke. Curr Treat Options Neurol 2010;12(6):492-503.
33. Otsuka A, Azuma K, Iesaki T, Sato F, Hirose T, Shimizu T, et al. Temporary hyperglycaemia provokes monocyte adhesion to Endothelial cells in rat thoracic aorta. Diabetologia 2005;48(12):2667-74.
34. Azuma K, Kawamori R, Toyofuku Y, Kitahara Y, Sato F, Shimizu T, et al. Repetitive fluctuations in blood glucose enhance monocyte adhesion to the endothelium of rat thoracic aorta. Arterioscler Thromb Vasc Biol 2006;26(10):2275-80.
35. Hjalmarsson C, Manhem K, Bokemark L, Andersson B. The role of prestroke glycemic control on severity and outcome of acute ischemic stroke. Stroke Res Treat 2014; 2014:694569.
36. Lei C, Wu B, Liu M, Chen Y. Association between hemoglobin A(1)C levels and clin- ical outcome in ischemic stroke patients with or without diabetes. J Clin Neurosci 2015;22(3):498-503.
37. Shin SB, Kim TU, Hyun JK, Kim JY. The prediction of clinical outcome using HbA1c in acute ischemic stroke of the deep branch of Middle cerebral artery. Ann Rehabil Med 2015;39(6):1011-7.
38. Krinsley JS, Meyfroidt G, van den Berghe G, Egi M, Bellomo R. The impact of premorbid diabetic status on the relationship between the three domains of glyce- mic control and mortality in critically ill patients. Curr Opin Clin Nutr Metab Care 2012;15(2):151-60.
39. Hughes H, Zimmerman S. Does using tight glycemic control in patients with acute ischemic stroke decrease morbidity or mortality? Ann Emerg Med 2014;64(5): 535-6.
40. Rosso C, Corvol JC, Pires C, Crozier S, Attal Y, Jacqueminet S, et al. Intensive versus subcutaneous insulin in patients with hyperacute stroke: results from the random- ized INSULINFARCT trial. Stroke 2012;43(9):2343-9.
41. Wijdicks EF, Sheth KN, Carter BS, Greer DM, Kasner SE, Kimberly WT, et al. Recom- mendations for the management of cerebral and cerebellar infarction with swelling: a statement for healthcare professionals from the American Heart Association/ American Stroke Association. Stroke 2014;45(4):1222-38.
42. Connor JT, Broglio KR, Durkalski V, Meurer WJ, Johnston KC. The stroke hyperglyce- mia insulin network effort (SHINE) trial: an adaptive trial design case study. Trials 2015;16:72.