Journal of the Neurological Sciences, 111 (1992) 59-64 © 1992 Elsevier Science Publishers B.V. All rights reserved 0022-510X/92/$05.00
59
JNS 03805
Blood glucose, glycosylated haemoglobin, and outcome of ischemic brain infarction K a r i M u r r o s a, R a i n e r F o g e l h o l m a, S a n n a K e t t u n e n b, A r j a - L i i s a V u o r e l a b a n d J u h a V a l v e c Departments of'a Neurology, b Radiology, and c Radiotherapy, Central Hospital of Central Finland, Jyviiskylii, Finland (Received 28 October, 1991) (Revised, received 24 February, 1992) (Accepted 28 February, 1992) Key words: Brain infarction; Blood glucose; Glycosylated haemoglobin; Computed tGmography; Outcome
Summary From August 1987 through December 1989 all consecutive conscious patients younger than 70 years with a recent ( < 48 h) brain infarction of the carotid territory were prospectively included in the study. Blood samples for fasting blood glucose and glycosylated haemoglobin ( H b A l c ) were taken after a median delay of 23 h of the onset of symptoms. The severity of hemiparesis was assessed on admission, at 1 week, 3 weeks, and 3 months. The functional outcome was assessed at 3 months. Computed cerebral tomography was performed on admission, and later on at 3 weeks or 3 months. The brain infarct volume was
measured from the CTs. The patients were diagnosed to have prestroke normoglycemia (n = 76) and prestroke hyperglycemia (n -- 23) on basis of the HbAlc level. The case fatality rate, severity of hemiparesis, functional outcome, and infarct size did not differ between these 2 groups. On the other hand, fasting blood glucose level of the non-diabetics correlated strongly with the severity of hemiparesis and predicted stroke outcome. A statistically significant correlation was observed between blood glucose values and the volumes of cortical infarcts in non-diabetics. Because prestroke blood glucose level, in contrast to post-stroke blood glucose level, did not have any predictive value concerning stroke outcome it is concluded that high fasting blood glucose values after stroke reflect a stress response to a more severe ischemic brain lesion.
Introduction Diabetes increases the risk of cardiovascular disease. The Framingham study revealed that the risk of ischemic brain infarction in diabetics, after adjustment for several other risk factors, is twice that of non-diabetics (Kannel and McGee 1979). Diabetes may also worsen both the short-term and long-term prognosis following stroke (Asplund et ai. 1980). The pathogenetic mechanisms underlying this increased risk and the worse outcome after stroke have remained unsolved. Among the several possibilities, e.g. micro- and macroangiopathy, hyperlipidemia and blood clotting disorders, the adverse effect of hyperglycemia on cerebral metabolism during ischemia have in the last decade been the focus of intensive research, both experimental Correspondence to: Karl Murros MD, Department of Neurology, Central Hospital of Central Finland, SF-40620 Jyviis~l[i, Finland. Tel.: (941)691811; Fax: (941)691098.
and clinical. Experimental studies have yielded diverging results. Hyperglycemia, especially predating cerebral ischemia may increase the size of an ischemic lesion of the brain and worsen the prognosis (Pulsinelli et al. 1982; Venables et al. 1985; De Courten-Myers et aL 1988; Prado et ai. 1988) but, on the other hand, smaller ischemic lesions and better outcome have been seen in both hyperglycemic (3ernigan et al. 1984; Ginsberg et al. 1987; Kraft et al. 1990) and normoglycemic (Siemkowicz and Hansen 1987) animals. In clinical strdies the emphasis has been on blood glucose values during the first few days after stroke The clinical relevance of the blood glucose level is difficult to assess as the level is modified by many factors such as diabetes, latent diabetic state, timing of blood sampling, intravenous infusions containing glucose, and stress. Most studies are in agreement that a high blood glucose level during the first days after stroke is associated with poor outcome (Melamed 1976; Pulsineili et al. 1983; Candelise et al. 1985; Berger and
60 Hakim 1986; Gray et al. 1987; Power et al. 1988; Woo et al. 1988a). However, Adams et al. (1988a) did not find a correlation between admission serum glucose concentration and neurologic outcome in patients with cerebral infarction. Some studies have included glycosylated haemoglobin (HbAlc) measurements in order to diagnose those patients with prestroke hyperglycemia. HbAlc is known to provide time-averaged values for blood glucose over the preceding 2-4 months (Health and Public Policy Committee 1984). In strokes, no correlation between neurologic outcome as case fatality and HbAlc concentration seems to exist (Cox and Lorains 1986; Gray et al. 1987; Power et al. 1988; Woo et al. 1988b). Thus, a crucial question concerning the association of post-stroke hyperglycemia and unfavourable prognosis is whether the high glucose values are the primary cause of more severe strokes or are they only a reflection, a stress response, of more severe strokes and brain lesions? To shed more light on these questions, our study was designed to focus both on preand post-stroke glucose levels in relationship with various outcome variables, including infarct size, in hemispheral brain infarction.
Patients and methods
From August 1987 through December 1989 all patients admitted to the Central Hospital of Central Finland, Jyv~iskylii, and fullfilling the following inclusion criteria were prospectively included in the study: (1) Adult patients younger than 70 years, with (2) the first ischemic brain infarction in the internal carotid artery territory, (3) admitted to the hospital within 48 h of onset of stroke, (4) being conscious during the first neurologic assessment, and (5) having no need for parenteral feeding. The clinical diagnosis of ischaemic stroke was confirmed in all cases by CT performed within 2 days of the onset of symptoms, and a second CT study was performed 3 weeks or 3 months later. A computerized planimetric method was employed for calculating the total infarct volume from the CT films. The infarct area was digitized for each CT slice, and multiplied by the corresponding slice distance to the next slice. The total infarct volume was calculated by summing these subvolumes. Because the size of an acute ischemic lesion varies markedly during the first week after stroke onset (Berger and Hakim 1986; Brott et al. 1989), measurement of infarct volume was based on the second CT examination either 3 weeks or 3 months after onset. In the analysis, infarctions were classified as cortical or deep by CT finding. Patients with infarcts localizing only in the region of internal capsule or adjacent basal ganglia (and corresponding to the clinical symptoms) were considered as deep infarcts.
Neurologic assessment was performed on days 1 (24-48 h after onse0 and 7, at 3 weeks, and at 3 months after admission. The severity of paresis of the upper and lower extremity was scored as follows: No paresis = 0, minimal weakness = 1, moderate weakness (able to lift the extremity against gravity) = 3, and total paralysis = 4. This scoring system gives a maximum of 8 points in case of total hemiparalysis. At the end of the 3 month follow-up, the functional status was assessed by the 5-point scale as proposed by Rankin (Rankin 1957). Blood samples for overnight fasting (7.00 a.m.) blood glucose and HbAlc estimation were taken within 3 days of the onset of symptoms. Blood glucose was measured by the hexokinase method, and the HbAlc concentration was measured, after removal of the labile components, by high-pressure liquid chromatography. According to our hospital standards, HbAlc values less than 7.0% are considered to indicate a normal glucose metabolism. The interassay coefficients of variation were 3.1% and 1.8% for the blood glucose and HbAlc analyses, respectively. Of the various cerebrovascular risk factors in the case history we were especially concerned about hypertension, coronary heart disease, and diabetes. As the patients in the present study participated in a multicenter study designed to evaluate the possible effects of peroral nimodipine on ischemic stroke, the effects of nimodipine were taken into consideration in the statistical analyses. To focus on glucose metabolism was not within the scope of the nimodipine trial. It is reported that nimodipine has no influence on plasma glucose level (Adams et al. 1988b). The significance of differences between the group means or medians was determined by the 2-tailed Student t-test or the Mann-Whitney U-test. The 95% confidence intervals for medians were calculated by the method proposed by Campbell and Gardner in 1989. The correlations between non-parametric data was carried out by calculating the Spearman rank correlation coefficients (rs). The differences between distributions of discrete variables were tested by the chi-square test.
Results
One hundred and five patients admitted on average 6 h (range 0.25-25 h) after onset of stroke met the inclusion criteria. Both I-IbAlc and fasting (7.00 a.m.) blood glucose samples were available in 99 (94%) cases. The blood samples were taken after a median delay of 23 h (range 2-57 h) of the onset of symptoms. Of the 99 patients, 76 were considered to have a normal prestroke glucose metabolism. The 23 patients with abnormally high HbAlc level ( > 7%) were 4 years older ( P > 0.05), and the prevalence of coronary heart
61 TABLE 1 CHARACTERISTICS OF STUDY PATIENTS BY HbAIc LEVEL
Number Sex (% male) Mean age (years, range) Prevalence of hypertension a Prevalence of CHD a Prevalence of diabetes a ,8-Glucose (mmol/I, mean, range) Severity of hemiparesis (mean) Score (95%C.I.) on day I Deaths _ 7%
76 53 (70%) 58.1 (32-69) 20 (26%) 21 (28%) 0 (0%) 5.9 (4.5-10.3)
23 15 (65%) 61.7 (40-69) 9 (39%) ! I (48%) 15 (65%) 10.7 (5.5-17.8)
4.3 (3.7-4.9) 11 (14%)
5.1 (4.1-6.1) 3 (13%)
By case history. disease was almost twice that ( P < 0.05) in patients with normal HbAlc values (Table 1). The difference in the prevalence of hypertension was not statistically significant. Of the 23 patients with abnormal HbAlc, 15 were diabetics by case history (8 on oral medication, 5 on insulin, and 2 on diet only). During the first week, only the 15 known diabetics received treatment for their diabetes. The mean fasting blood glucose values were 5.9 mmol/I (range 4.5-10.3 mmol/l) and 10.7 mmol/l (5.5-17.8 mmol/I) for the patients, with normal or high HbAlc values, respectively. During the first day of hospital stay, 8 non-diabetic patients received glucose containing i.v. infusions with a standard rate of 4 g / h of glucose. In each case, the influence of the infusion on blood glucose level was calculated to be marginal. Figure 1 shows the distribution of fasting blood glucose concentrations by HbAlc level. There was a higly significant correlation between H b A l c and fasting blood glucose values in the 23 patients with high H b A l c level (r = 0.83, P = 0.0001). No such cor-
Number of
petients 20J
m
B! []
HbA 1 c< 7~g HbAlc-" 7~g n=99
relation existed among the 76 patients with normal HbAlc (r = 0.06, P = 0.64). There were no statistically significant differences in blood glucose values based on sex, age ( < 60 years vs. > 60 years), presence of hypertension or coronary heart disease, or treatment with nimodipine. The blood glucose values of patients with normal I-lbAlc level correlated highly significantly with the scores of hemiparesis on days 1, 7, at 3 weeks and at 3 months (Table 2) but in patients with abnormal HbAlc level no such correlation was observed. Figure 2 provides a graphic representation of the blood glucose values by the severity of hemiparesis on day 1 in the patients with normal HbAlc. No correlation was observed between the HbAlc values and the severity of hemiparesis on days 1, 7, at 3 weeks and at 3 months (n = 9 9 rs--0.14, n ffi 95 r s = 0.13, n - 91 rs •0.15 and n = 85 rs = 0.16, respectively). At 3 months of follow-up, 14 (14%) of the patients had died; 5 of the primary, and 3 of a recurrent brain infarction, 4 of pulmonary embolism, 1 of pneumonia, and 1 of intracerebral haemorrhage. The case fatality rate was equal in men and women, in the 2 age groups ( < 60 vs. > 60 years), in the groups with different risk factors (hypertension, coronary heart disease, diabetes), in those treated and untreated with nimodipine, and in the groups with normal and high HbAlc concentration (Table 1). The median blood glucose values of the 14
TABLE 2 CORRELATION BETWEEN SEVERITY OF HEMIPARESIS (SCORES 0-8) AND FASTING BLOOD GLUCOSE VALUES, BY HbAlc LEVEL H b A l c < 7%
4
5
6
7
8
9
10
11
12
)!4
~-gluc (retool/i) Fig. l, Distribution of fasting blood glucose concentration by the HbAlc level,
Day I Day 7 3 weeks 3 months
HbAlc > 7%
n
rs
P-value
n
rs
P-value
76 72 69 65
0.52 0.48 0.42 0.33
0.0001 0.0O01 0.0006 0.008
23 23 22 20
0.26 0.15 0.21 0.01
0.23 0.47 0.33 0.78
62
Brei n i nfsrct volume ( m l ) 225. - -
mine1/1
10
.
*
,
|
200. 175
•
150 125,
:e •
•
:
T
•
e
l
÷÷,_
:
100,
• "
75
25, . ~ e •• 0 •O 0 "m
soverit9 of hemiparesis (points)
:
.00
HbAlc~ 7~g n=33 rs --0.46 p=0.01 ~
• °o
•, 0_
•
,0-g] uc ( ram0!/1)
Fig. 2. A scattergram of fasting blood glucose values by severity of hemiparesis on day 1 in patients with normal HbAlc. Median values (transverse lines) are presented for each group of hemiparesis.
Fig. 3. Correlation between fasting blood glucose values and the volumes of cortical infarcts with a normal HbAlc level.
patients who died, were higher than those of the 85 who were alive during the follow-up (7.3 and 5.8 ram•l/l, respectively, Mann-Whitney P < 0.05). The blood glucose values correlated with the scores of functional status at 3 months (n ffi 85 rs--0.32, P < 0.01) but the corresponding correlation between HbAlc values and functional status was non-significant (rs = 0.16). All 99 patients underwent a CT study within 2 days of the symptom onset. The second CT showed a hypodense area corresponding to clinical symptoms in 74 patients. The second study (3 weeks or 3 months) was not performed on 5 patients who were alive at 3 months. Of the 14 patients dying during the follow-up, 12 died before the scheduled second CT. In 8 cases, despite the clinically evident diagnosis of hemispheral infarction, both the initial and the second CT were normal. Table 3 shows the median infarct volumes of cortical and deep infarcts by the HbAlc level. The median volumes of cortical and deep infarcts were of the same magnitude irrespective of the HbAlc level. Even if the 8 patients with normal CT (0 ml) were taken into consideration, there was no statistically significant difference in volumes between the patients with normal and high HbAlc; the median (95% C.I.)
volumes were n = 63, 3.2 ml (1.4-11.9 ml) and n ffi 19, 11.4 ml (0.7-46.3 ml), respectively. Accordingly, there were no statistically significant differences in infarct volumes when analysed by sex, age ( < 60 vs. > 60 years), prevalence of hypertension, coronary heart disease, or treatment with nimodipine. A statistically significant correlation between blood glucose values and cortical infarct volumes was observed in the non-diabetics (Fig. 3). With respect to deep infarcts in non-diabetics, no such correlation was found (n = 24, rs = 0.10). Correspondingly, no statistically significant correlations were observed between blood glucose values and infarct volumes in the diabetics (HbAlc > 7%).
TABLE 3 MEDIAN (95% CONFIDENCE INTERVALS) VOLUMES (ml) OF CORTICAL AND DEEP INFARCTS BY HbAlc LEVEL HbAlc < 7%
HbAI > 7%
All
n
n
I!
Cortical 33 28.3 (11.9-69.2) 11 30.2 (11.4-75.5) 44 29.3 (17.0-63.6) Deep 24 1.0 (0.3-3.2) 6 0.8 (0.2-2.1) 30 0.9 (0.4-1.6)
Discussion
The results of our study do not support previous studies which stress the adverse effects of hyperglycemia on severity and prognosis of stroke (Pulsinelli et al. 1983; Oppenheimer et al. 1985). HbAlc concentration, reflecting the prestroke glucose balance, did not have any predictive value for outcome. When estimating the blood glucose level at the stroke ictus, it is reasonable to assume that, on the average, diabetics with high HbAlc have higher blood glucose levels than non-diabetics at the very moment of stroke. However, during the follow-up period of 3 months, the case fatality rates of the non-diabetics and the diabetics (including the patients with unrecognized prestroke hyperglycemia) were similar and secondly, increasing severity of hemiparesis and more severe defects in functional status were not correlated with HbAlc values. Cortical infarct volumes appeared to be slightly larger in diabetics than in non-diabetics but the differ-
63 ence was non-significant. It should, however, be noted that diabetics have several non-metabolic factors for poor outcome such as diabetic angiopathy, impaired autoregulation of the cerebral vessels, and rheologic changes in blood constituents (Helgason 1988). Our study considered cortical and deep infarcts separately because it is stated that glucose may protect the brain from severe, focal, ischemic neuronal damage. On the other hand, in areas having no collateral supply, the extent of tissue injury appears to be determined overwhelmingly by the topography on the blood flow reduction per se and is insensitive to modulations of blood glucose (Prado et al. 1988). Althoug the HbAlc level did not predict the stroke outcome, in the non-diabetics the higher mean fasting blood glucose values during the acute phase of cerebral infarction were associated with higher case fatality rate, increasing severity of hemiparesis, and more severe defects in the 3 months functional status. Not surprisingly, blood glucose values significantly correlated with cortical infarct volumes. Only a few previous studies have focused on infarct volumes. Candelise et al. (1985) found that fasting blood glucose values correlated with both the clinical outcome and diameter of CT lesion in non-diabetic patients with ischemic or haemorrhaglc stroke. In another study (Berger and Hakim 1986), no significant correlation was observed between blood glucose level and the infarct size in 39 patients with cerebral infarction. However, in this retrospective study the blood sampling and the timing of CT were not standardized, nor were the effects of diabetes taken into consideration when comparing infarct sizes between groups with different glucose levels. The present study was unable to show a statistically significant association between high blood glucose values and poor outcome in diabetics. Probably the small number of patients with diabetes exposes the result to type II error and secondly, diabetics presumably have more daily variations in their blood glucose levels than do non-diabetics. When focusing on post-stroke blood glucose values, the impact of stress is of special concern. Even if the patient is admitted within a few h after onset of symptoms, do the blood glucose values necessarily represent the situation at the very moment of stroke or are we, in fact, measuring the stress response? The possibility of postprandial hyperglycemia must also be kept in mind. Release of catecholamines and increase in serum cortisol levels due to stress probably induce hyperglycemia in acute stroke. In fact, a statistically significant correlation has been recently demonstrated between serum cortisol and blood glucose values in the acute phase of stroke (Murros 1991; O'Neill et al. 1991). It is reasonable to suppose that in many instances the H b A l c concentration gives a better overall estimation of the blood glucose level at the very moment of stroke than
the post-stroke blood glucose concentration. Against this backround, the present results are in concordance with studies which emphasize the importance of stress hyperglycemia as a predictor of stroke prognosis, and the severity of the cerebral lesion being the cause and not the result of hyperglycemia (Power et al. 1988; Woo et al. 1988b). References Adams, H.P., C.P. Olinger, J.R. Marler, J. Biller, T.G. Brott, W.G. Barsan and K. Banwart (1988a) Comparison of admission serum glucose concentration with neurologie outcome in acute cerebral infarction. Stroke, 19: 455-458. Adams, H.A., H. Miiller, B. v. Bormann, U. B6rner and U., G. Hempelmann (1988b) The influence of nimodipine on plasma catecholamines and the perioperative endocrine stress response. An~isth. Intensivther. Notfallmed., 23: 82-87. Asplund, K., E. H~igg, C. Helmers, F. Lithner, T. Strand and P.-O. Wester (1980) The natural history of stroke in diabetic patients. Acta Med. Scand., 207: 417-424. Berger, L. and L.A. Hakim (1986) The association of hyperglycemia with cerebral edema on stroke. Stroke, 17: 865-871. Brott, T., J.M. Marler, C.P. Olinger, H.P. Adams, T. Tomsick, W.G. Barsan, J. Biller, R. Eberle, V. Hertzberg and M. Walker (1989) Measurements of acute cerebral infarction: lesion size by computed tomography. Stroke, 20: 871-875. Campbell, M.J. and M.J. Gardner (1989) Calculating confidence intervals for some non-parametric analyses. In: MJ. Gardner and D.G. Altman (Eds), Statistics with Confidence, British Medical Journal, London, pp. 71-79. Candelise, L., G. Landi, E.N. Orazio and E. Boccardi (1985) Prognostic significance of hyperglycemia in acute stroke. Arch. New rol., 42: 661-663. Cox, N.H. and J.W. Lorains (1986) The prognostic value of blood glucose and glycosylated haemoglobin assays in the management and diagnosis of diabetes mellitus. Ann. Int. Med., 10h 710-713. De Courten-Myers, G.M., R.E. Myers and B.S. Schoolfield (1988) Hyperglycemia enlarges infarct size in cerebrovascular occlusion in cats. Stroke, 19: 623-630. Ginsberg, M.D., R. Prado, W.D. Dietrich, R. Busto and B.D. Watson (1987) Hyperglycemia reduces the extent of cerebral infarction in rats. Stroke, 18: 570-574. Gray, C.S., R. Taylor, J.M. French, K.G. Alberti, G.S. Venables, O.F. James, D.A. Shaw, N.E. Cartlidge and D. Bates (1987) The prognostic value of stress hyperglycemia and previously unrecognized diabetes in acute stroke. Diabetic Med., 4: 237-240. Health and Public Policy Committee. American College of Physicians (1984) Glycosylated hemoglobin assays in the management and diagnosis of diabetes mellitus. Ann. Intern. Med, 101: 710713. Helgason, C.M. (1988) Blood glucose and stroke. Stroke, 19: 10491053. Jernigan, J., O.B. Evans and H.S. Kirshner (1984) Hyperglycemia and diabetes improve outcome in a rat model of anoxia/ischemia Neurology, 34 (Suppl.): 262. Kannel, W.B. and D.L. McGee (1979) Diabetes and cardiovascular disease. The Framingham Study. J. Am. Med. Assoc., 241: 20352038. Kraft, S.A., C.P. Larson, L.M. Shuer, G.K. Steinberg, G.V. Benson and R.G. Petal (1990) Effect of hyperglycemia on neuronal changes in a rabbit model of focal cerebral ischemia. Stroke, 21: 447-450. Melamed, E. (1976) Reactive hyperglycemia in patients with acute stroke. J. Neurol. Sci., 29: 267-275.
64 Murros, K. (1991) Stress Reactions of Brain Infarction, Academic dissertation, Kuopio University, pp. 48-51 O'Neill, P.A., !. Davies, K.J. Fullerton and D. Bennett (1991) Stress hormone and blood glucose response following acute stroke in the elderly. Stroke, 22: 842-847. Oppenheimer, S.M., B.I. Hoffbrand, G.A. Oswald and J.S. Yudkin (1985) Diabetes mellitus and early mortality from stroke. Br. Med. J., 291: 1014-1015. Power, M.J., K.J. Fullerton and R.W. Stout (1988) Blood glucose and prognosis after stroke. Age Ageing, 17: 164-170. Prado, R., M.D. Ginsberg, W.D. Dietrich, B.D. Watsin and R. Busto, R. (1988) Hyperglycemia increases infarct size in collaterally peffused but not end-arterial vascular territories. J. Cereb. Flow Metab., 8: 186-192. Pulsinelli, W.A., S. Waldman, D. Rawlinson and F. Plum (1982) Moderate hyperglycemia augments ischemic brain damage: a neuropathologic study in the rat. Neurology, 32: 1239-1246. Pulsinelli, W.A., D.E. Levy, B. Sigsbee, P. Scherer and F. Plum
(1983) Increased damage after ischemic stroke in patients with hyperglycemia with or without established diabetes meilitus. Am. J. Med., 74: 540-544. Rankin, J. (1957) Cerebral vascular accidents in patients over the age of 60, !!. Prognosis. Scott. Med. J., 2; 200-215. Siemkowicz, E. and A.J. Hansen (1987) Clinical restitution following cerebral ischemia in hypo-, normo- and hyperglycemic rats. Acta Neurol. Scand., 58: 1-8. Venables, G.S., S.A. Miller, G. Gibson, J.A. Hardy and A.J. Strong (1985) The effects of hyperglycemia on changes during reperfusion following focal cerebral ischemia in the cat. J. NeuroL Neurosurg. Psychiatry, 48: 663-669. Woo, E., Y.W. Chan, Y.L Yu and C.Y. Huang (1988a) Admission glucose level in relation to mortality and morbidity outcome in 252 stroke patients. Stroke, 19: 185-191. Woo, E., J.T.C. Ma, J.D. Robinson and Y.L. Yu (1988b) Hyperglycemia is a stress response in acute stroke. Stroke, 19: 13591364.
59
JNS 03805
Blood glucose, glycosylated haemoglobin, and outcome of ischemic brain infarction K a r i M u r r o s a, R a i n e r F o g e l h o l m a, S a n n a K e t t u n e n b, A r j a - L i i s a V u o r e l a b a n d J u h a V a l v e c Departments of'a Neurology, b Radiology, and c Radiotherapy, Central Hospital of Central Finland, Jyviiskylii, Finland (Received 28 October, 1991) (Revised, received 24 February, 1992) (Accepted 28 February, 1992) Key words: Brain infarction; Blood glucose; Glycosylated haemoglobin; Computed tGmography; Outcome
Summary From August 1987 through December 1989 all consecutive conscious patients younger than 70 years with a recent ( < 48 h) brain infarction of the carotid territory were prospectively included in the study. Blood samples for fasting blood glucose and glycosylated haemoglobin ( H b A l c ) were taken after a median delay of 23 h of the onset of symptoms. The severity of hemiparesis was assessed on admission, at 1 week, 3 weeks, and 3 months. The functional outcome was assessed at 3 months. Computed cerebral tomography was performed on admission, and later on at 3 weeks or 3 months. The brain infarct volume was
measured from the CTs. The patients were diagnosed to have prestroke normoglycemia (n = 76) and prestroke hyperglycemia (n -- 23) on basis of the HbAlc level. The case fatality rate, severity of hemiparesis, functional outcome, and infarct size did not differ between these 2 groups. On the other hand, fasting blood glucose level of the non-diabetics correlated strongly with the severity of hemiparesis and predicted stroke outcome. A statistically significant correlation was observed between blood glucose values and the volumes of cortical infarcts in non-diabetics. Because prestroke blood glucose level, in contrast to post-stroke blood glucose level, did not have any predictive value concerning stroke outcome it is concluded that high fasting blood glucose values after stroke reflect a stress response to a more severe ischemic brain lesion.
Introduction Diabetes increases the risk of cardiovascular disease. The Framingham study revealed that the risk of ischemic brain infarction in diabetics, after adjustment for several other risk factors, is twice that of non-diabetics (Kannel and McGee 1979). Diabetes may also worsen both the short-term and long-term prognosis following stroke (Asplund et ai. 1980). The pathogenetic mechanisms underlying this increased risk and the worse outcome after stroke have remained unsolved. Among the several possibilities, e.g. micro- and macroangiopathy, hyperlipidemia and blood clotting disorders, the adverse effect of hyperglycemia on cerebral metabolism during ischemia have in the last decade been the focus of intensive research, both experimental Correspondence to: Karl Murros MD, Department of Neurology, Central Hospital of Central Finland, SF-40620 Jyviis~l[i, Finland. Tel.: (941)691811; Fax: (941)691098.
and clinical. Experimental studies have yielded diverging results. Hyperglycemia, especially predating cerebral ischemia may increase the size of an ischemic lesion of the brain and worsen the prognosis (Pulsinelli et al. 1982; Venables et al. 1985; De Courten-Myers et aL 1988; Prado et ai. 1988) but, on the other hand, smaller ischemic lesions and better outcome have been seen in both hyperglycemic (3ernigan et al. 1984; Ginsberg et al. 1987; Kraft et al. 1990) and normoglycemic (Siemkowicz and Hansen 1987) animals. In clinical strdies the emphasis has been on blood glucose values during the first few days after stroke The clinical relevance of the blood glucose level is difficult to assess as the level is modified by many factors such as diabetes, latent diabetic state, timing of blood sampling, intravenous infusions containing glucose, and stress. Most studies are in agreement that a high blood glucose level during the first days after stroke is associated with poor outcome (Melamed 1976; Pulsineili et al. 1983; Candelise et al. 1985; Berger and
60 Hakim 1986; Gray et al. 1987; Power et al. 1988; Woo et al. 1988a). However, Adams et al. (1988a) did not find a correlation between admission serum glucose concentration and neurologic outcome in patients with cerebral infarction. Some studies have included glycosylated haemoglobin (HbAlc) measurements in order to diagnose those patients with prestroke hyperglycemia. HbAlc is known to provide time-averaged values for blood glucose over the preceding 2-4 months (Health and Public Policy Committee 1984). In strokes, no correlation between neurologic outcome as case fatality and HbAlc concentration seems to exist (Cox and Lorains 1986; Gray et al. 1987; Power et al. 1988; Woo et al. 1988b). Thus, a crucial question concerning the association of post-stroke hyperglycemia and unfavourable prognosis is whether the high glucose values are the primary cause of more severe strokes or are they only a reflection, a stress response, of more severe strokes and brain lesions? To shed more light on these questions, our study was designed to focus both on preand post-stroke glucose levels in relationship with various outcome variables, including infarct size, in hemispheral brain infarction.
Patients and methods
From August 1987 through December 1989 all patients admitted to the Central Hospital of Central Finland, Jyv~iskylii, and fullfilling the following inclusion criteria were prospectively included in the study: (1) Adult patients younger than 70 years, with (2) the first ischemic brain infarction in the internal carotid artery territory, (3) admitted to the hospital within 48 h of onset of stroke, (4) being conscious during the first neurologic assessment, and (5) having no need for parenteral feeding. The clinical diagnosis of ischaemic stroke was confirmed in all cases by CT performed within 2 days of the onset of symptoms, and a second CT study was performed 3 weeks or 3 months later. A computerized planimetric method was employed for calculating the total infarct volume from the CT films. The infarct area was digitized for each CT slice, and multiplied by the corresponding slice distance to the next slice. The total infarct volume was calculated by summing these subvolumes. Because the size of an acute ischemic lesion varies markedly during the first week after stroke onset (Berger and Hakim 1986; Brott et al. 1989), measurement of infarct volume was based on the second CT examination either 3 weeks or 3 months after onset. In the analysis, infarctions were classified as cortical or deep by CT finding. Patients with infarcts localizing only in the region of internal capsule or adjacent basal ganglia (and corresponding to the clinical symptoms) were considered as deep infarcts.
Neurologic assessment was performed on days 1 (24-48 h after onse0 and 7, at 3 weeks, and at 3 months after admission. The severity of paresis of the upper and lower extremity was scored as follows: No paresis = 0, minimal weakness = 1, moderate weakness (able to lift the extremity against gravity) = 3, and total paralysis = 4. This scoring system gives a maximum of 8 points in case of total hemiparalysis. At the end of the 3 month follow-up, the functional status was assessed by the 5-point scale as proposed by Rankin (Rankin 1957). Blood samples for overnight fasting (7.00 a.m.) blood glucose and HbAlc estimation were taken within 3 days of the onset of symptoms. Blood glucose was measured by the hexokinase method, and the HbAlc concentration was measured, after removal of the labile components, by high-pressure liquid chromatography. According to our hospital standards, HbAlc values less than 7.0% are considered to indicate a normal glucose metabolism. The interassay coefficients of variation were 3.1% and 1.8% for the blood glucose and HbAlc analyses, respectively. Of the various cerebrovascular risk factors in the case history we were especially concerned about hypertension, coronary heart disease, and diabetes. As the patients in the present study participated in a multicenter study designed to evaluate the possible effects of peroral nimodipine on ischemic stroke, the effects of nimodipine were taken into consideration in the statistical analyses. To focus on glucose metabolism was not within the scope of the nimodipine trial. It is reported that nimodipine has no influence on plasma glucose level (Adams et al. 1988b). The significance of differences between the group means or medians was determined by the 2-tailed Student t-test or the Mann-Whitney U-test. The 95% confidence intervals for medians were calculated by the method proposed by Campbell and Gardner in 1989. The correlations between non-parametric data was carried out by calculating the Spearman rank correlation coefficients (rs). The differences between distributions of discrete variables were tested by the chi-square test.
Results
One hundred and five patients admitted on average 6 h (range 0.25-25 h) after onset of stroke met the inclusion criteria. Both I-IbAlc and fasting (7.00 a.m.) blood glucose samples were available in 99 (94%) cases. The blood samples were taken after a median delay of 23 h (range 2-57 h) of the onset of symptoms. Of the 99 patients, 76 were considered to have a normal prestroke glucose metabolism. The 23 patients with abnormally high HbAlc level ( > 7%) were 4 years older ( P > 0.05), and the prevalence of coronary heart
61 TABLE 1 CHARACTERISTICS OF STUDY PATIENTS BY HbAIc LEVEL
Number Sex (% male) Mean age (years, range) Prevalence of hypertension a Prevalence of CHD a Prevalence of diabetes a ,8-Glucose (mmol/I, mean, range) Severity of hemiparesis (mean) Score (95%C.I.) on day I Deaths _ 7%
76 53 (70%) 58.1 (32-69) 20 (26%) 21 (28%) 0 (0%) 5.9 (4.5-10.3)
23 15 (65%) 61.7 (40-69) 9 (39%) ! I (48%) 15 (65%) 10.7 (5.5-17.8)
4.3 (3.7-4.9) 11 (14%)
5.1 (4.1-6.1) 3 (13%)
By case history. disease was almost twice that ( P < 0.05) in patients with normal HbAlc values (Table 1). The difference in the prevalence of hypertension was not statistically significant. Of the 23 patients with abnormal HbAlc, 15 were diabetics by case history (8 on oral medication, 5 on insulin, and 2 on diet only). During the first week, only the 15 known diabetics received treatment for their diabetes. The mean fasting blood glucose values were 5.9 mmol/I (range 4.5-10.3 mmol/l) and 10.7 mmol/l (5.5-17.8 mmol/I) for the patients, with normal or high HbAlc values, respectively. During the first day of hospital stay, 8 non-diabetic patients received glucose containing i.v. infusions with a standard rate of 4 g / h of glucose. In each case, the influence of the infusion on blood glucose level was calculated to be marginal. Figure 1 shows the distribution of fasting blood glucose concentrations by HbAlc level. There was a higly significant correlation between H b A l c and fasting blood glucose values in the 23 patients with high H b A l c level (r = 0.83, P = 0.0001). No such cor-
Number of
petients 20J
m
B! []
HbA 1 c< 7~g HbAlc-" 7~g n=99
relation existed among the 76 patients with normal HbAlc (r = 0.06, P = 0.64). There were no statistically significant differences in blood glucose values based on sex, age ( < 60 years vs. > 60 years), presence of hypertension or coronary heart disease, or treatment with nimodipine. The blood glucose values of patients with normal I-lbAlc level correlated highly significantly with the scores of hemiparesis on days 1, 7, at 3 weeks and at 3 months (Table 2) but in patients with abnormal HbAlc level no such correlation was observed. Figure 2 provides a graphic representation of the blood glucose values by the severity of hemiparesis on day 1 in the patients with normal HbAlc. No correlation was observed between the HbAlc values and the severity of hemiparesis on days 1, 7, at 3 weeks and at 3 months (n = 9 9 rs--0.14, n ffi 95 r s = 0.13, n - 91 rs •0.15 and n = 85 rs = 0.16, respectively). At 3 months of follow-up, 14 (14%) of the patients had died; 5 of the primary, and 3 of a recurrent brain infarction, 4 of pulmonary embolism, 1 of pneumonia, and 1 of intracerebral haemorrhage. The case fatality rate was equal in men and women, in the 2 age groups ( < 60 vs. > 60 years), in the groups with different risk factors (hypertension, coronary heart disease, diabetes), in those treated and untreated with nimodipine, and in the groups with normal and high HbAlc concentration (Table 1). The median blood glucose values of the 14
TABLE 2 CORRELATION BETWEEN SEVERITY OF HEMIPARESIS (SCORES 0-8) AND FASTING BLOOD GLUCOSE VALUES, BY HbAlc LEVEL H b A l c < 7%
4
5
6
7
8
9
10
11
12
)!4
~-gluc (retool/i) Fig. l, Distribution of fasting blood glucose concentration by the HbAlc level,
Day I Day 7 3 weeks 3 months
HbAlc > 7%
n
rs
P-value
n
rs
P-value
76 72 69 65
0.52 0.48 0.42 0.33
0.0001 0.0O01 0.0006 0.008
23 23 22 20
0.26 0.15 0.21 0.01
0.23 0.47 0.33 0.78
62
Brei n i nfsrct volume ( m l ) 225. - -
mine1/1
10
.
*
,
|
200. 175
•
150 125,
:e •
•
:
T
•
e
l
÷÷,_
:
100,
• "
75
25, . ~ e •• 0 •O 0 "m
soverit9 of hemiparesis (points)
:
.00
HbAlc~ 7~g n=33 rs --0.46 p=0.01 ~
• °o
•, 0_
•
,0-g] uc ( ram0!/1)
Fig. 2. A scattergram of fasting blood glucose values by severity of hemiparesis on day 1 in patients with normal HbAlc. Median values (transverse lines) are presented for each group of hemiparesis.
Fig. 3. Correlation between fasting blood glucose values and the volumes of cortical infarcts with a normal HbAlc level.
patients who died, were higher than those of the 85 who were alive during the follow-up (7.3 and 5.8 ram•l/l, respectively, Mann-Whitney P < 0.05). The blood glucose values correlated with the scores of functional status at 3 months (n ffi 85 rs--0.32, P < 0.01) but the corresponding correlation between HbAlc values and functional status was non-significant (rs = 0.16). All 99 patients underwent a CT study within 2 days of the symptom onset. The second CT showed a hypodense area corresponding to clinical symptoms in 74 patients. The second study (3 weeks or 3 months) was not performed on 5 patients who were alive at 3 months. Of the 14 patients dying during the follow-up, 12 died before the scheduled second CT. In 8 cases, despite the clinically evident diagnosis of hemispheral infarction, both the initial and the second CT were normal. Table 3 shows the median infarct volumes of cortical and deep infarcts by the HbAlc level. The median volumes of cortical and deep infarcts were of the same magnitude irrespective of the HbAlc level. Even if the 8 patients with normal CT (0 ml) were taken into consideration, there was no statistically significant difference in volumes between the patients with normal and high HbAlc; the median (95% C.I.)
volumes were n = 63, 3.2 ml (1.4-11.9 ml) and n ffi 19, 11.4 ml (0.7-46.3 ml), respectively. Accordingly, there were no statistically significant differences in infarct volumes when analysed by sex, age ( < 60 vs. > 60 years), prevalence of hypertension, coronary heart disease, or treatment with nimodipine. A statistically significant correlation between blood glucose values and cortical infarct volumes was observed in the non-diabetics (Fig. 3). With respect to deep infarcts in non-diabetics, no such correlation was found (n = 24, rs = 0.10). Correspondingly, no statistically significant correlations were observed between blood glucose values and infarct volumes in the diabetics (HbAlc > 7%).
TABLE 3 MEDIAN (95% CONFIDENCE INTERVALS) VOLUMES (ml) OF CORTICAL AND DEEP INFARCTS BY HbAlc LEVEL HbAlc < 7%
HbAI > 7%
All
n
n
I!
Cortical 33 28.3 (11.9-69.2) 11 30.2 (11.4-75.5) 44 29.3 (17.0-63.6) Deep 24 1.0 (0.3-3.2) 6 0.8 (0.2-2.1) 30 0.9 (0.4-1.6)
Discussion
The results of our study do not support previous studies which stress the adverse effects of hyperglycemia on severity and prognosis of stroke (Pulsinelli et al. 1983; Oppenheimer et al. 1985). HbAlc concentration, reflecting the prestroke glucose balance, did not have any predictive value for outcome. When estimating the blood glucose level at the stroke ictus, it is reasonable to assume that, on the average, diabetics with high HbAlc have higher blood glucose levels than non-diabetics at the very moment of stroke. However, during the follow-up period of 3 months, the case fatality rates of the non-diabetics and the diabetics (including the patients with unrecognized prestroke hyperglycemia) were similar and secondly, increasing severity of hemiparesis and more severe defects in functional status were not correlated with HbAlc values. Cortical infarct volumes appeared to be slightly larger in diabetics than in non-diabetics but the differ-
63 ence was non-significant. It should, however, be noted that diabetics have several non-metabolic factors for poor outcome such as diabetic angiopathy, impaired autoregulation of the cerebral vessels, and rheologic changes in blood constituents (Helgason 1988). Our study considered cortical and deep infarcts separately because it is stated that glucose may protect the brain from severe, focal, ischemic neuronal damage. On the other hand, in areas having no collateral supply, the extent of tissue injury appears to be determined overwhelmingly by the topography on the blood flow reduction per se and is insensitive to modulations of blood glucose (Prado et al. 1988). Althoug the HbAlc level did not predict the stroke outcome, in the non-diabetics the higher mean fasting blood glucose values during the acute phase of cerebral infarction were associated with higher case fatality rate, increasing severity of hemiparesis, and more severe defects in the 3 months functional status. Not surprisingly, blood glucose values significantly correlated with cortical infarct volumes. Only a few previous studies have focused on infarct volumes. Candelise et al. (1985) found that fasting blood glucose values correlated with both the clinical outcome and diameter of CT lesion in non-diabetic patients with ischemic or haemorrhaglc stroke. In another study (Berger and Hakim 1986), no significant correlation was observed between blood glucose level and the infarct size in 39 patients with cerebral infarction. However, in this retrospective study the blood sampling and the timing of CT were not standardized, nor were the effects of diabetes taken into consideration when comparing infarct sizes between groups with different glucose levels. The present study was unable to show a statistically significant association between high blood glucose values and poor outcome in diabetics. Probably the small number of patients with diabetes exposes the result to type II error and secondly, diabetics presumably have more daily variations in their blood glucose levels than do non-diabetics. When focusing on post-stroke blood glucose values, the impact of stress is of special concern. Even if the patient is admitted within a few h after onset of symptoms, do the blood glucose values necessarily represent the situation at the very moment of stroke or are we, in fact, measuring the stress response? The possibility of postprandial hyperglycemia must also be kept in mind. Release of catecholamines and increase in serum cortisol levels due to stress probably induce hyperglycemia in acute stroke. In fact, a statistically significant correlation has been recently demonstrated between serum cortisol and blood glucose values in the acute phase of stroke (Murros 1991; O'Neill et al. 1991). It is reasonable to suppose that in many instances the H b A l c concentration gives a better overall estimation of the blood glucose level at the very moment of stroke than
the post-stroke blood glucose concentration. Against this backround, the present results are in concordance with studies which emphasize the importance of stress hyperglycemia as a predictor of stroke prognosis, and the severity of the cerebral lesion being the cause and not the result of hyperglycemia (Power et al. 1988; Woo et al. 1988b). References Adams, H.P., C.P. Olinger, J.R. Marler, J. Biller, T.G. Brott, W.G. Barsan and K. Banwart (1988a) Comparison of admission serum glucose concentration with neurologie outcome in acute cerebral infarction. Stroke, 19: 455-458. Adams, H.A., H. Miiller, B. v. Bormann, U. B6rner and U., G. Hempelmann (1988b) The influence of nimodipine on plasma catecholamines and the perioperative endocrine stress response. An~isth. Intensivther. Notfallmed., 23: 82-87. Asplund, K., E. H~igg, C. Helmers, F. Lithner, T. Strand and P.-O. Wester (1980) The natural history of stroke in diabetic patients. Acta Med. Scand., 207: 417-424. Berger, L. and L.A. Hakim (1986) The association of hyperglycemia with cerebral edema on stroke. Stroke, 17: 865-871. Brott, T., J.M. Marler, C.P. Olinger, H.P. Adams, T. Tomsick, W.G. Barsan, J. Biller, R. Eberle, V. Hertzberg and M. Walker (1989) Measurements of acute cerebral infarction: lesion size by computed tomography. Stroke, 20: 871-875. Campbell, M.J. and M.J. Gardner (1989) Calculating confidence intervals for some non-parametric analyses. In: MJ. Gardner and D.G. Altman (Eds), Statistics with Confidence, British Medical Journal, London, pp. 71-79. Candelise, L., G. Landi, E.N. Orazio and E. Boccardi (1985) Prognostic significance of hyperglycemia in acute stroke. Arch. New rol., 42: 661-663. Cox, N.H. and J.W. Lorains (1986) The prognostic value of blood glucose and glycosylated haemoglobin assays in the management and diagnosis of diabetes mellitus. Ann. Int. Med., 10h 710-713. De Courten-Myers, G.M., R.E. Myers and B.S. Schoolfield (1988) Hyperglycemia enlarges infarct size in cerebrovascular occlusion in cats. Stroke, 19: 623-630. Ginsberg, M.D., R. Prado, W.D. Dietrich, R. Busto and B.D. Watson (1987) Hyperglycemia reduces the extent of cerebral infarction in rats. Stroke, 18: 570-574. Gray, C.S., R. Taylor, J.M. French, K.G. Alberti, G.S. Venables, O.F. James, D.A. Shaw, N.E. Cartlidge and D. Bates (1987) The prognostic value of stress hyperglycemia and previously unrecognized diabetes in acute stroke. Diabetic Med., 4: 237-240. Health and Public Policy Committee. American College of Physicians (1984) Glycosylated hemoglobin assays in the management and diagnosis of diabetes mellitus. Ann. Intern. Med, 101: 710713. Helgason, C.M. (1988) Blood glucose and stroke. Stroke, 19: 10491053. Jernigan, J., O.B. Evans and H.S. Kirshner (1984) Hyperglycemia and diabetes improve outcome in a rat model of anoxia/ischemia Neurology, 34 (Suppl.): 262. Kannel, W.B. and D.L. McGee (1979) Diabetes and cardiovascular disease. The Framingham Study. J. Am. Med. Assoc., 241: 20352038. Kraft, S.A., C.P. Larson, L.M. Shuer, G.K. Steinberg, G.V. Benson and R.G. Petal (1990) Effect of hyperglycemia on neuronal changes in a rabbit model of focal cerebral ischemia. Stroke, 21: 447-450. Melamed, E. (1976) Reactive hyperglycemia in patients with acute stroke. J. Neurol. Sci., 29: 267-275.
64 Murros, K. (1991) Stress Reactions of Brain Infarction, Academic dissertation, Kuopio University, pp. 48-51 O'Neill, P.A., !. Davies, K.J. Fullerton and D. Bennett (1991) Stress hormone and blood glucose response following acute stroke in the elderly. Stroke, 22: 842-847. Oppenheimer, S.M., B.I. Hoffbrand, G.A. Oswald and J.S. Yudkin (1985) Diabetes mellitus and early mortality from stroke. Br. Med. J., 291: 1014-1015. Power, M.J., K.J. Fullerton and R.W. Stout (1988) Blood glucose and prognosis after stroke. Age Ageing, 17: 164-170. Prado, R., M.D. Ginsberg, W.D. Dietrich, B.D. Watsin and R. Busto, R. (1988) Hyperglycemia increases infarct size in collaterally peffused but not end-arterial vascular territories. J. Cereb. Flow Metab., 8: 186-192. Pulsinelli, W.A., S. Waldman, D. Rawlinson and F. Plum (1982) Moderate hyperglycemia augments ischemic brain damage: a neuropathologic study in the rat. Neurology, 32: 1239-1246. Pulsinelli, W.A., D.E. Levy, B. Sigsbee, P. Scherer and F. Plum
(1983) Increased damage after ischemic stroke in patients with hyperglycemia with or without established diabetes meilitus. Am. J. Med., 74: 540-544. Rankin, J. (1957) Cerebral vascular accidents in patients over the age of 60, !!. Prognosis. Scott. Med. J., 2; 200-215. Siemkowicz, E. and A.J. Hansen (1987) Clinical restitution following cerebral ischemia in hypo-, normo- and hyperglycemic rats. Acta Neurol. Scand., 58: 1-8. Venables, G.S., S.A. Miller, G. Gibson, J.A. Hardy and A.J. Strong (1985) The effects of hyperglycemia on changes during reperfusion following focal cerebral ischemia in the cat. J. NeuroL Neurosurg. Psychiatry, 48: 663-669. Woo, E., Y.W. Chan, Y.L Yu and C.Y. Huang (1988a) Admission glucose level in relation to mortality and morbidity outcome in 252 stroke patients. Stroke, 19: 185-191. Woo, E., J.T.C. Ma, J.D. Robinson and Y.L. Yu (1988b) Hyperglycemia is a stress response in acute stroke. Stroke, 19: 13591364.
