J-Shaped Association Between Serum Glucose and Functional Outcome in Acute Ischemic Stroke

Hyperglycemia is commonly encountered in patients with acute ischemic stroke, with estimates varying and depending on the frequency of glucose measurements and the criteria used to define hyperglycemia.¹ The incidence of poststroke hyperglycemia is estimated at 45% in studies with frequent glucose measurements and a threshold value of 7 mmol/L to define hyperglycemia.² Because the prevalence of previously diagnosed diabetes mellitus in stroke patients is estimated between 10% and 20%, diabetes mellitus is obviously not the only underlying pathophysiologic mechanism of poststroke hyperglycemia.³ Previously undiagnosed diabetes mellitus and impaired glucose tolerance account for a further 5% to 28%.³ In addition, 10% to 20% of stroke patients present with hyperglycemia with normal glycosylated hemoglobin.⁴ This is considered as a neurohumoral stress response, although studies on the association between serum cortisol and poststroke hyperglycemia yielded conflicting results.⁴

Hyperglycemia after stroke is independently associated with infarct volume in magnetic resonance and spectroscopy studies,⁵ and poor functional outcome.⁶ However, the UK Glucose Insulin in Stroke Trial (GIST-UK) showed no clinical benefit of treating hyperglycemia rapidly with glucose–potassium–insulin infusion.⁷ Moreover, mortality was higher in those patients with greatest glucose reductions (>2 mmol/L), implying a possibly deleterious effect when glucose levels decrease to less than a critical threshold.⁷ An animal model study demonstrated that the association between serum glucose and cerebral infarct volume is described by a U-shaped curve with a nadir of approximately 7 mmol/L.⁸ However, a similar curve in human studies was never reported.

The primary aim of the present study is to investigate the association between serum glucose and functional outcome, and the possible presence of a similar curve. Our secondary objective is to define a limit or range for serum glucose levels in the acute phase of stroke that may be associated with better functional outcome.

Subjects and Methods

We analyzed all consecutive patients who were admitted to the stroke unit or intensive care unit of the Centre Hospitalier Universitaire Vaudois (Lausanne, Switzerland) between January 2003 and December 2008 who were registered in the Acute Stroke Registry and Analysis in Lausanne (ASTRAL) and who had serum glucose measured within 24 hours after stroke onset. Briefly, the ASTRAL includes all patients with a main diagnosis of acute ischemic stroke admitted within 24 hours after stroke onset, whereas patients with transient ischemic attack (TIA), intracerebral hemorrhage, subarachnoid hemorrhage, cerebral sinus venous thrombosis, stroke mimics (such as hypoglycemia, based on acute and subacute glucose measurements) and late admission (>24 hours after stroke onset) are excluded.

Demographic data (age, gender, ethnicity, insurance), metabolic parameters (serum glucose, creatinine, cholesterol), full blood count, and vital signs (skin temperature, heart rate, systolic and diastolic blood pressure) are routinely collected in the first 24 hours after stroke onset. National Institutes of Health Stroke Scale (NIHSS) scores (at admission, at 4 to 6 hours, and 24 hours after admission) were prospectively recorded for each patient by NIHSS-certified medical personnel. We also recorded vascular risk factors like arterial hypertension, atrial fibrillation, diabetes mellitus, valvulopathy, coronary artery disease, and smoking that are either already known or newly diagnosed. We also assessed prestroke Rankin score, previous cerebrovascular events, and medications at the time of stroke. Finally, we recorded silent lesions, leukoaraiosis, and early ischemic changes in brain CT, as well as significant arterial stenosis/occlusion, because all patients underwent brain parenchymal (mostly CT) and arterial (mostly CT angiography) imaging. Rankin score was assessed by Rankin-certified personnel at 3 months in the outpatient clinic. At 12 months, and for patients not able to attend the outpatient clinic, Rankin score was assessed during a telephone interview. Favorable outcome was defined as Rankin score ≤2. All patients received proper management according to current international guidelines.⁹ Stroke patients with hypoglycemia are being treated in the emergency department and in the stroke unit with 5% glucose perfusions. Patients with hyperglycemia >10 mmol/L receive subcutaneous long-acting insulin; starting in 2008, patients with hyperglycemia on admission are administered an intravenous insulin protocol aimed at achieving normoglycemia.

A patient was considered as having diabetes if any of the following was present either before the stroke or at least 1 week after the stroke in the absence of significant complications of stroke: fasting plasma glucose >126 mg/dL (7.0 mmol/L); a 2-hour value in the oral glucose tolerance test >200 mg/dL (11.1 mmol/L); a random plasma glucose concentration >200 mg/dL (11.1 mmol/L) in the presence of symptoms; or when already using antidiabetic medications or insulin.

Statistical Analysis

Continuous variables are reported as median±interquartile range. All parameters presented in Table 1 were analyzed in univariate and multivariate models to identify significant associations with favorable outcome at 3 and 12 months. Level of significance was set at 95% (P=0.05). The model was further analyzed in a robust regression model based on fractional polynomial (−2–2) functions.

Results

Between January 2003 and December 2008, we encountered 3165 patients with acute cerebrovascular events in our stroke unit or intensive care unit. We excluded 1423 patients because of intracerebral hemorrhage, TIA, late admission, and ocular ischemia; 296 (9.3%) patients were excluded because of no documentation of acute blood glucose within the first 24 hours or >50% of the other parameters. The main characteristics of 1446 patients who were finally included in our analysis are summarized in Table 1.

Initial NIHSS score, age, prestroke Rankin score, serum glucose, early ischemic changes in brain imaging (mostly noncontrast CT), and leukoaraiosis were independently associated with Rankin score at 3 months. The correlations remained significant at 12 months, with the exception of neuroimaging (Table 2); the estimated area under the receiver-operator characteristic curve was 0.869. The association between serum glucose and 12-month Rankin score is described by a J-shaped curve with a nadir of 5 mmol/L (Figure A). A similar curve was generated for the association between serum glucose values at admission and 24-hour NIHSS score (Figure B). Both curves consist of 2 parts: a slow downward drift of Rankin score (or NIHSS score in Figure B) when serum glucose is being reduced to a nadir of ≈5 mmol/L and a steep upward drift when glucose values are further reduced below this critical threshold. As is evident in Figure A and B, the association between serum glucose and outcome is disproportional between the 2 parts of each curve: for glucose values >5 mmol/L, large differences in serum glucose correspond to only mild differences in stroke outcome. On the contrary, for glucose values <5 mmol/L, small differences in serum glucose correspond to disproportionally large differences in stroke prognosis. The range of serum glucose values that correspond to Rankin score ≤2 is 3.7 to 7.3 mmol/L (Figure A). Similarly, the range for 24-hour NIHSS score <7 is 4.0 to 7.2 mmol/L (Figure B).

• Download figure
• Download PowerPoint

Discussion

The present study is the first to our knowledge to demonstrate a J-shaped association between serum glucose levels at admission and functional outcome in patients with acute ischemic stroke with a nadir of ≈5 mmol/L. This result is corroborated by a similar relationship between serum glucose at admission and NIHSS score at 24 hours. Also, this study defines a range of serum glucose values at admission (3.7–7.3 mmol/L) that is associated with long-term (12 months after stroke) favorable outcome.

Currently, European Stroke Organization guidelines recommend treatment of serum glucose with insulin titration at levels >10 mmol/L, but not routine use of insulin infusion in patients with moderate hyperglycemia.⁹ However, the American Stroke Association guidelines recommend insulin treatment when glucose values exceed 140 to 185 mg/dL (7.7–10.3 mmol/L).¹⁰ However, these recommendations are based mainly on expert consensus. The present study shows that values >7.3 mmol/L are associated with poor outcome.

Our findings provide a possible explanation for the negative results of GIST-UK, which failed to detect a net clinical benefit of active treatment of hyperglycemia with glucose–potassium–insulin infusion compared to infusion of normal saline.⁷ First, the mean glucose reduction between the 2 arms of GIST-UK was only 0.57 mmol/L, which would correspond to only a small improvement in outcome when projected to the curve of Figure A. More substantial reductions can be achieved with sliding-scale insulin; in a small pilot study, glucose decreased from 14.7±4.9 mmol/L to 7.3±1.1 mmol/L in patients with hyperglycemia after stroke.¹¹ Second, too aggressive glucose control in some patients in GIST-UK probably contributed to poor outcome, as indicated by a post hoc analysis showing that patients with large glucose reduction (≥2 mmol/L) had significantly higher mortality (34% vs 22%; P=0.009) than those patients with smaller reductions (<2 mmol/L).⁷

Recently, another group showed better glycemic control with intravenous insulin than with subcutaneous insulin administration (9.7 vs 6.0 mmol/L; P<0.001).¹² However, aggressive reduction of hyperglycemia below the threshold of 5 mmol/L seems risky, as shown by our data (Figure B). In a similar fashion, the NICE-SUGAR study reported recently that intensive glucose control increased mortality among critically ill patients. In particular, an intermediate blood glucose target of ≤10 mmol/L resulted in lower mortality than did a target of 4.6 to 6 mmol/L.¹³ In myocardial infarction, a U-shaped relationship between 30-day mortality and initial blood glucose has also been shown.¹⁴ Finally, cerebral microdialysis in patients with various types of acute brain damage indicates that low-normal cerebral glucose values are associated with higher markers of stress than mildly elevated levels.¹⁵

Conclusions

The present study has several limitations. First, this is an observational study; therefore, the implicit suggestion that maintenance of glycemic control within the range of 3.7 to 7.3 mmol/L is associated with better outcome needs to be confirmed in intervention trials. Second, this is a single-center study and our results also should be confirmed in other populations. Another limitation of our study is that we did not have quality data for glucosylated hemoglobin A1c, which might have allowed identification of further associations. It should be also kept in mind that we measured serum glucose, and not whole blood (capillary) glucose, because the values reported by other glucose studies may use different samples for glucose measurement. Also, we do not have data about the efficacy of hyperglycemia (or hypoglycemia) treatment in patients with hyperglycemia (or those with hypoglycemia), which may have influenced our results. Finally, the number of observations (of patients) at both edges of the curves is relatively low compared to the number of observations at the base of the curve.

In conclusion, the present study shows that the association between serum glucose levels and functional outcome is described by a J-shaped curve. The shape of the curve suggests that considerable reductions of glucose are necessary to significantly improve outcome. However, there is little tolerance toward the lower side of reducing serum glucose values. Serum glucose values within the range of 3.7 to 7.3 mmol/L are associated with favorable outcome.

Footnotes