The Deleterious Effect of Admission Hyperglycemia on Survival and Functional Outcome in Patients With Intracerebral Hemorrhage
Background and Purpose—We aimed to evaluate the association between blood glucose (BG) levels at admission and both functional outcome at discharge and 1-month mortality after intracerebral hemorrhage (ICH).
Methods—All cases of first-ever ICH were identified from the population-based Stroke Registry of Dijon, France from 1985 to 2009. Clinical and radiological information was recorded. BG was measured at admission. Multivariate analyses were performed using logistic and Cox regression models. Multiple imputation was used as a sensitivity analysis.
Results—We recorded 465 first-ever ICH. BG at admission was obtained in 416 patients (89.5%) with a median value of 6.92 mmol/L. In multivariate analyses, BG in the highest tertile (≥8.6 mmol/L) was an independent predictor of functional handicap (odds ratio, 2.51; 95% CI, 1.43–4.40; P=0.01) and 1-month mortality (hazard ratio, 2.51; 95% CI, 1.23–2.43; P=0.002). The results were consistent with those obtained from multiple imputation analyses.
Conclusions—Admission hyperglycemia is associated with poor functional recovery at discharge and 1-month mortality after ICH. These results suggest a need for trials that evaluate strategies to lower BG in acute ICH.
Admission hyperglycemia is associated with poor prognosis in patients with ischemic stroke,1 but its impact on early outcome has been less extensively assessed in patients with spontaneous intracerebral hemorrhage (ICH). Previous studies showed conflicting results and were limited by small population size, hospital-based setting, or lack of control for potentially confounding factors.1–7
We aimed to evaluate the effects of blood glucose (BG) levels at admission on both functional outcome at discharge and 1-month mortality in patients with ICH.
From 1985 to 2009, all cases of first-ever ICH that occurred within the city of Dijon, France (150 000 inhabitants) were identified from the Dijon Stroke Registry. The case collection procedure was based on multiple overlapping sources of information to identify fatal and nonfatal, hospitalized and nonhospitalized, ICH patients.8
Functional impairment at discharge was measured with a self-constructed handicap scale (1=walking alone; 2=walking with support; 3=walking stick; 4=wheelchair; 5=bedridden; 6=dead). Various cut-offs were analyzed for severe, moderate-to-severe, and mild-to-severe handicap (respectively ≥4, ≥3, and ≥2). Any-cause mortality was recorded at 1 month.
Vascular risk factors, prestroke treatments, and clinical severity at onset (level of consciousness, aphasia, and motor deficit) were recorded. The ICH location (lobar, deep, cerebellum, brain stem, intraventricular extension) was reported according to the findings of initial neuroimaging. BG was measured immediately at admission.
Admission BG was not analyzed as a continuous variable because of its nonlinear relationship with the log odds of severe handicap (regression splines) and violation of the log–linear assumption (martingale residual plots). It was transformed into categorical variables (second tertile of admission BG chosen as the cut-off point). Associations between functional outcome or 1-month mortality and admission BG were studied using multivariate backward logistic and Cox regression models, respectively. We performed sensitivity analyses using multiple imputation based on the predictive mean matching method to handle missing BG data for 49 patients (10.5%). Stratum-specific analyses were performed by including variables with significant statistical interactions along with admission BG in multivariate models. Significance was set at P<0.05. SAS 9.2 (SAS Institute, Inc) was used for the statistical analyses.
We recorded 465 first-ever ICH. BG at admission was obtained in 416 patients (89.5%) with a median value of 6.92 mmol/L (interquartile range, 5.7–10.2 mmol/L; Supplemental Table I, http://stroke.ahajournals.org).
In multivariate, complete-case analyses, admission BG in the upper tertile (Table 1) and quartile (Supplemental Table II) was associated with increased risk of severe handicap with heterogeneity across age strata (probability value for interaction=0.046; Table 1). Similar results were obtained with moderate-to-severe and mild-to-severe handicap (Supplemental Tables III and IV).
One-month survival was lowest for patients in the upper tertile of admission BG (Figure) and positively related with admission BG tertiles (Supplemental Figure I) and quartiles (Supplemental Figure II). In multivariate, complete-case analysis, admission BG in the upper tertile (Table 2) and quartile (Supplemental Table V) were predictors of 1-month mortality with heterogeneity among patients with and without peripheral artery disease (probability value for interaction=0.021; Table 2). All results were consistent with those obtained with multiple imputation analyses.
This population-based study demonstrated that admission hyperglycemia was independently associated with early mortality in ICH patients. An increase in cerebral complications, including more widespread bleeding, new bleeding, and brain edema, could account for this finding.2–7 Admission hyperglycemia was also associated with poor functional outcome at discharge. This result is consistent with that of a systematic overview of studies assessing the association between admission hyperglycemia and poor functional recovery in nondiabetic stroke patients.1
Several pathophysiological mechanisms could explain our findings. Animal experiments on ICH have suggested that hyperglycemia may promote vasogenic brain edema and perihematomal neuronal death.9 In contrast, some authors suggested that admission hyperglycemia could only reflect severity of the ICH as a stress reaction to a serious brain injury.3 Because our results were not altered after adjustment for items of clinical severity, we could argue that hyperglycemia has pathophysiological significance and is not simply an epiphenomenon of the stress response.
Several limitations must be acknowledged. BG was measured at admission, and the delay between ICH onset and the blood sample may have varied from one patient to another. Values of BG were missing in 10.5% of patients. However, multiple imputation analyses found similar results to those from complete-case analyses. Finally, hematoma volume and the National Institutes of Health Stroke Scale were not measured and were replaced with features that reflect severity, but which may be imperfect. These limitations may not have had a major impact on the validity of our findings, although some residual confounding bias is possible.
To conclude, these results indicate that additional studies are necessary to assess strategies to lower BG in ICH. The disappointing results from glucose-lowering therapy in ischemic stroke10 probably would not apply in hemorrhagic stroke because the mechanisms are different.
Sources of Funding
The Dijon Stroke Registry is supported by the French Institute for Public Health Surveillance (InVS) and Inserm.
We thank Mr Philip Bastable for reviewing the English.
The online-only Data Supplement is available at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.632950/-/DC1.
- Received July 17, 2011.
- Accepted July 29, 2011.
- © 2012 American Heart Association, Inc.
- Capes SE,
- Hunt D,
- Malmberg K,
- Pathak P,
- Gerstein HC
- Passero S,
- Ciacci G,
- Ulivelli M
- Fogelholm R,
- Murros K,
- Rissanen A,
- Avikainen S
- Béjot Y,
- Aboa-Eboulé C,
- Durier J,
- Rouaud O,
- Jacquin A,
- Ponavoy E,
- et al
- Song EC,
- Chu K,
- Jeong SW,
- Jung KH,
- Kim SH,
- Kim M,
- et al