Relation of Hyperglycemia Early in Ischemic Brain Infarction to Cerebral Anatomy, Metabolism, and Clinical Outcome Michael Kushner, MD,"? Patrizia Nencini, MD,' Martin Reivich, MD,"? Mario Rango, MD,' Dara Jamieson, MD,"? Franz Fazekas, MD," Robert Zimmerman, MD,§ John Chawluk, MD,"+ Abass Alavi, MD,? and Wayne Alves, PhD'Z
We studied the relation of serum glucose level measured in the first 12 hours of symptoms to the clinical findings, results of computed tomography (CT), and patterns of cerebral metabolism in 39 patients who had acute ischemic cerebral infarction. Structural damage was assessed by CT. Metabolic disruption was assessed using "F-fluorodeoxyglucose and positron emission tomography (PET). Median initial serum glucose concentration was 155 mddl(6.7 d) Clinical . recovery was significantly poorer in patients with initial serum glucose levels higher than the median ( p < 0.05, chi square). PET tended to show normal results or minor abnormalities in patients with initial glucose levels less than the median, as opposed to lobar or multilobe abnormalities in patients with levels that were higher than the median ( p < 0.05, Kendall's Tau b). The severity of hypometabolism in the ischemic region, expressed as the percent asymmetry of local cerebral glucose metabolism between homologous brain regions, was greater in patients with initial glycemia concentrations higher than the median ( p < 0.001, t test). Relationships of serum glucose level with metabolic derangement and structural damage, but not outcome, held true in patients without a history of diabetes mellitus. Kushner M, Nencini P, Reivich M, Rango M, Jamieson D, Fazekas F, Zimmerman R, Chawluk J, Alavi A, Alves W. Relation of hyperglycemia early in ischemic brain infarction to cerebral anatomy, metabolism, and clinical outcome. Ann Neurol 1990;28:129-135
Observations in animals and humans have suggested that an elevated serum glucose concentration at or about the time of an ischemic insult is associated with enhanced postischemic damage of cerebral tissue f 17). A variety of putative mechanisms now have been proposed whereby glucose, or its metabolites, might serve to augment postischemic damage r8- 111. Pulsinelli and colleagues {GI and Levy and associates [73 first suggested that moderate ictal hyperglycemia, in excess of 120 mg/dl (6.7 mM), was associated with poorer clinical outcome in hospitalized stroke patients. Some studies suggested that there is a relationship between ictal hyperglycemia and the extent of the ischemic lesion observed on imaging studies {12, 131. A potential implication of these observations is that early measures to control serum glucose concentrations after
infarction might serve to limit stroke severity in the clinical setting. Despite these previous reports, other experimental studies of hyperglycemia in stroke yielded contrary, or mixed, results C14-161. Nedergaard and Diemer [161 found that ictal hyperglycemia does not impact on stroke severity, while Ginsberg and colleagues r141 reported a protective influence of hyperglycemia following ischemia. In this report we describe the results of OUT observations on the relationships between serum glucose level at the time of an ischemic ictus with the clinical findings, computed tomography (CT) results, and patterns of cerebral metabolism measured using "F-fluorodeoxyglucose (FDG) and positron emission tomography (PET). This study was undertaken to determine
From the "Cerebrovascular Research Center and the Departments of *Neurology, $Neurosurgery, and $Radiology, University of Pennsylvania, Philadelphia, PA.
Address correspondence to Dr Kushner, Cerebrovascular Research Center, Room 429, Johnson Pavilion, University of Pennsylvania, Philadelphia, PA 19104.
Received Jun 30, 1989, and in revised form Dec 14, 1989, and Jan 18, 1990. Accepted for publication Jan 25, 1990.
Copyright 0 1990 by the American Neurological Association
129
whether or not initial serum glucose level was related to stroke severity, the extent of tissue damage, or metabolic disruption.
Methods Patients hospitalized for acute cerebral infarction at the Hospital of the University of Pennsylvania or the Graduate Hospital were eligible for inclusion in this study. Informed consent was obtained for each patient under the guidelines of the Committee on Studies Involving Human Subjects of the University of Pennsylvania [ 17).
R.Z.). The CT-detected lesions were mapped on neuroanatomical templates derived from a standardized tomographicanatomical atlas [2 1). The size of the CT lesions was graded according to a noncontinuous incremental scale described previously [17}: I = no lesion, I1 = small lesion less than 2.0 cm in diameter, 111 = lesion larger than 2.0 cm but contained within one cerebral lobe, and IV = multilobe abnormality. We used the last CT scan performed in order to reduce error arising from a CT with no findings performed soon after the onset of clinical manifestations.
Metabolic Methods The criteria for patient selection were clinical and CT evidence of acute nonhemorrhagic cerebral infarction with persistent focal neurological deficits at least until the time of PET study. Exclusionary criteria included: history of cerebral infarction; age less than 21 years; presence of stupor or coma; pregnancy or child-bearing potential; CT evidence of intracerebral hemorrhage; presence of primary subarachnoid hemorrhage; diabetic ketoacidosis or hyperosmolar state; and severe systemic disorders such as myocardial infarction or renal, pulmonary, or hepatic failure. Clinical data were acquired prospectively using standardized medical history, neurological history, and physical exarnination forms derived from standardized data base registries { 18-20). Patient follow-up was also performed according to protocol. Clinical data were scored so as to rank neurological disability, degree of impairment of activities of daily living, and extent of neurological recovery. The degree of recovery was classified into four groups based on the relative degree of restoration of function compared to the ictal nadir according to a scale previously described [17). Patients were classified as having made either a complete, good, moderate, or poor recovery. Records from the emergency medical systems, hospital emergency department, and inpatient charts were reviewed in order to obtain data concerning the first serum blood glucose level determined within 12 hours of the ischemic ictus. If a blood glucose level value within 12 hours of the ictus was not available, or if no blood glucose value was obtained prior to the initiation of dextrose-containing intravenous infusions or insulin therapy, the patient was excluded from study. Serum glucose level was determined by the enzymatic method. Hospital records were reviewed, and patients and next of kin were interviewed in order to determine any history of diabetes mellitus or previous treatment with hypoglycemic agents, and the results of laboratory tests were used to determine the presence or absence of diabetes mellitus. Patients who had previously established diabetes mellitus or treatment with hypoglycemic agents were considered nondiabetic.
All patients underwent imaging to determine local cerebral glucose metabolism (LCMRgl) using FDG and PET. The first PET study was performed as close to the time of the ischemic ictus as possible, with a mean of 7 days after infarction. Fifty-nine percent of patients had repeat PET studies from 2 weeks to 6 months after infarction. The technique employed has been described previously C22, 231. The measured value of the lumped constant, 0.542, was used [24]. Images of LCMRgl were obtained using a modified version of the positron emission transverse tomograph (PE'IT V) camera C251. Local glucose consumption waq calculated using the method of Sokoloff and associates, in which a correction for the small loss of label due to phosphatase activity was used 123, 261. The PET data and images were analyzed both quantitatively and qualitatively. Quantitative analysis of the PET data was performed as follows. The whole brain metabolic rate for glucose (WBMRgI) was calculated for each patient. LCMRgl was calculated in the region of the cerebral ischemia as follows. All 14 intercalated PET slices were reviewed independent of the CT or clinical data (by M. K. and F. F.). The slice containing the region with the most severely disturbed metabolism was then selected. Using a visual display unit and irregular regions of interest overlay programs, a region of interest (ROI) was inscribed so as to be totally contained within thc area of the maximum metabolic abnormality. In order to avoid partial volume effects we attempted to make the ROI span at least one image resolution element (17 mm), but when needed, smaller ROIs of lesser span were employed in order to inscribe totally the ROI within the area of maximum metabolic abnormality. LCMRgl was also determined from a ROI of similar size and shape at the homologous site in the opposite hemisphere; brainstem lesions were excluded from quantitative analysis. When possible, the correctness of the PET localization of the lesion was assessed post hoc by reference to the latest CT scan. For this study the severity of metabolic disturbance was expressed as the percentage asymmetry between LCMRgl for the lesion and homologous normal brain by the expression:
RadiologicaL Methods
(contralateral LCMRgl - lesion LCMRgl) (contralateral LCMFQl)
All patients underwent conventional x-ray CT of the head using either a General Electric (Milwaukee, WI) 9800 or 8800 machine. Sixty-two percent of patients underwent follow-up CT, because of a normal or nondiagnostic initial study, from 5 days to 6 months following the ictus. For each case the last CT scan performed was reviewed blindly (by
In cases where no clear-cut metabolic abnormality could be identified, lesion ROIs were not defined and the percent asymmetry of LCMRgl was considered equal to zero. Qualitative analysis of the PET image was performed as follows. All metabolic images were reviewed blindly by three
Clinical Methods
130 Annals of Neurology Vol 28 No 2 August 1990
independent observers (M. K., A. A., and J. C.), who related the observed metabolic abnormalities to neuroanatomical templates derived from a standardized tomographic-anatomical atlas {21]. Ail PET readings were performed independent of the CT findings. Readers were asked to outline on the templates the area of most intense hypometabolism, which was presumed to represent the core of the ischemic focus, and the extent of surrounding hypometabolism. The overall spatial extent of the metabolic abnormality (e.g., ischemic core and remote hypometabolic regions) was defined when at least two readers agreed. PET images were scored qualitatively and assigned to categories that have been described previously [17]. In brief, the PET scans were rated as follows: I = normal scan, I1 = small abnormality on the order of an image resolution element (17 mm), I11 = abnormal metabolism confined to one cerebral lobe, and IV = multilobe metabolic abnormality.
Data Analysis Statistical analyses were designed to compare the relationship between serum glucose level at admission and the clinical, CT, and metabolic findings. Ictal serum glucose level was also compared with demographic data, such as age and gender, and the presence or absence of other risk factors for vascular disease, such as hypertension and diabetes mellitus. These comparisons were performed nonparametrically using the median test or Fisher's exact probability, depending on cell size. Within each analysis, patients were stratdied according to the median serum glucose value. A separate correlational analysis was also performed for patients with and those without a history of diabetes mellitus. The qualitative CT and PET gradings were coded according to the ordinal rank, and correlation coefficients were computed using Kendall's Tau b, a nonparametric counterpart of Pearson's r. Correlations between continuous data scales were performed using Spearman's rank correlation coefficient (rs).The quantitative PET data including WBMFigl, lesion LCMRgl, and asymmetry values were analyzed using Student's t test.
Results Clinical Results Thirty-nine patients (21 men, 18 women) fulfilled the conditions for entry into the study. The median age
was 65 years (range, 5 1 to 82 years). Thirty-six patients had a hemispheric lesion; 44% of these were in the right hemisphere and 56% in the left. In 13 patients the ischemic area was subcortical and in 3 patients the lesion was localized to the brainstem. The median serum glucose value at admission was 155 mg/dl for the total study group (range, 78 to 286 mgldl). The follow-up period ranged from 14 to 180 days. Average duration of clinical follow-up was 104 days for the entire group. The analyses of initial serum glucose values versus patient clinical characteristics are summarized in Table 1. Recovery was poorer when initid serum glucose levels were higher ( p < 0.05, chi square). History of diabetes mellitus was prominent in the hyperglycemic group ( p < 0.01, chi square). No relationship was noted between initial serum glucose level and age, history of hypertension, previous transient ischemic attacks, or location of the lesion (e.g., cortical versus subcortical). A separate analysis showed that history of diabetes mellims also was associated with worse recovery ( p < 0.05, chi square). Radiological Findings The mean interval between the ischemic ictus and the latest CT scan was 87 days. Grades of the CT-detected lesions did not differ across the range of glucose values (Table 2). No significant relationship was noted between the size of the CT lesion and a history of diabetes mellitus (Kendall's Tau b = 0.147, p = 0.16, not significant). Metabolic Findings Overall 2 hemispheric subcortical lesions and 2 subtentorial lacunes presented no clear-cut abnormalities on PET. The PET scans showed that the extent of the metabolic abnormality tended to be greater than expected, on the basis of the CT data, in the patients whose initial glucose values were 155 mg/dl or greater (see Table 2). No disparity between CT and PET data
Table I . Clinical Characteristics versus Initial Serum Glucose Levela Level < 155 mg/dl (n = 19) Age (yr) 5 65 > 65 Recovery Complete or good Moderate or poor Diabetes mellitus Yes No
Level 2 155 mg/dl (n = 20)
9
11
10
9
13
6
6
14
1 17
17
4
Statistical Values Chisquare
p
=
=
0.023
=
4.32
NS
Chi square
p < 0.05 Chi square = 19.2
p < 0.001
"Values are numbers of patients.
NS
=
not statistically sigriificant
Kushner et
al:
Hyperglycemia in Ischemic Brain Infarction
131
Table 2. C T and PET ResultS oers'slls Initial Serum Glucose Level Level < 155 mgid Lesion Grade
(n = 19)
Table 3. Comparison of WBMRgl and LCMRgl Values versuJ Initial Serum Glucose Level Level 2 155 mgidl (n = 20)
Level < 155 mddl
WBMR~I~
CT"
I
Lesion LCMRgl'
I1 I11
Asymmetry
IV PETb
I I1 I11 IV
3 6 0 11
5.9 +- 1.3 (n = 17)
Level 2 155 mddl
* 1.0
< 0.02
(n = 19) 1.3 t 1.1 (n = 17)
< 0.05
3.8
2.3 ? 1.0 (n = 11) 18.6 7.1525 44.9 -t 17.2% (n = 17) (n = 19)
*
p Value"
< 0.001
"Probability values are derived from Student's t test. bMean WBMRgl in age-matched elderly control subjects is 5.3 f 0.8 mg/lOO g d m i n (n = 14). CLCMRglvalues are mean i 1 standard deviation (units are mgl100 gdminute).
"Kendall's Tau b = 0.147; p value is not significant for CT lesion grade versus glucose level. bKendall's Tau b = 0.267, p < 0.05 for PET lesion grade versus glucose level.
WBMRgl = whole brain metabolic rate for glucose; LCMRgl = local cerebral glucose metabolism.
was noted for when initial glucose values were less than 155 mg/dl. Qualitative analysis of the PET scans showed that lobar or multilobe hypometabolism was associated with initial glucose levels of 155 mg/dl or more (Kendall's Tau b = 0.267, p < 0.05). No correlation was found between the extent of the PET-detected lesion and history of diabetes mellitus (Kendall's Tau b = 0.18, p value was not significant). Quantitative PET data were available for 36 patients (92%); three studies suffered from technical difficulties such as the lack of arterial blood samples or other technical deficiencies. The quantitative PET analyses are summarized in Table 3. Overall WBMRgl was significantly lower in the group with glucose concentrations that were above the median ( p < 0.02, t test). In the 26 patients with a clear-cut supratentord lesion on the PET scan, significantly lower glucose consumption in the core of the ischemic area was found for the group with initial hyperglycemia ( p < 0.05, t test). A more severe metabolic disturbance within the ischemic core, expressed as the percentage of asymmetry of LCMRgl in homologous brain regions, was present in the group with initial hyperglycemia (44.9 k 17.2% versus 18.6 t 7.196; p < 0.001, t test).
Twenty-two patients were identified as being nondiabetic. The median ictal serum glucose value was 123 mgldl. No significant relationship was found between the extent of recovery and initial serum glucose value (Table 4). Ranking of the relative sizes of the ultimate ischemic lesion found on last CT scans did show a significant correlation between lesion size and ictal glucose level (Spearman's ~t(= 0.433,df = 20, p < 0.05, data not shown). Both quantitative and qualitative metabolic analyses showed significant relationships between ictal serum glucose level and the PET data (see Table 4). The severity of lesion hypometabolism was significantly greater in patients with initial hyperglycemia. Also a significantly larger lesion as detected by PET was associated with initial hyperglycemia.
Nondiabetic Patients As mentioned, diabetes mellitus also was associated with poorer clinical outcome. Not surprisingly diabetes mellitus was associated with higher ictal serum gluose levels as well. Thus, interaction between diabetes mellitus and hyperglycemia might confound analysis over the entire group. In order to control for the influences of diabetes mellitus, a separate analysis of the data was undertaken in the patients without a history of diabetes mellitus. 132 Annals of Neurology Vol 28 No 2 August 1990
Discussion Before we discuss the implications of these findings, some potential sources of error should be mentioned. PET measurements of LCMRgl under ischemic conditions require some caution in their interpretation {27). Ischemia may perturb the deoxyglucose model by altering the rate constants or the lumped constant r28, 29}. Moreover, the lumped constant may be expected to increase appreciably at plasma glucose concentrations lower than 70 mgldl(3.9 mM), and be marginally depressed when serum glucose levels are greater than 300 mgldl(16.7 mM) 130, 311. Theoretically the effect of using a lumped constant that is artifactually low, or high, is to increase or decrease respectively calculated LCMRgl values. Protective from errors arising from rate constant alterations is the demonstration by Sokoloff 132) that LCMRgl is little perturbed if measurements are integrated over 40 minutes, as is customary in our laboratory. Practically speaking, Lu-
Table 4. Rerovely and CT and PET Lesion Grade versus Initial Serum Glucose Level in Nondiabetic Patient5 Serum Glucose Level
> 123 mgidl (n = 11)
Statistical Values
6 5
p
3 3 3
1 2
-
5
4
p
0
4
6 2 1
1 4 1
2 10.5
5
< 123 mgidl (n Recovery Complete or good Moderate or poor CT lesion grade I I1 111 IV PET lesion grade I I1 111
IV Asymmetry of lesion LCMRgl
=
11)
8
?
3.7%
34.7
= 0.24 Fisher's exact probability
= 0.54 Kendall's Tau b
-
p < 0.05 ?
4.2%
Kendall's Tau b p < 0.002, t test
LCMRgl = local cerebral glucose metabolism.
cignani and associates [331 found in animals that the 14C-deoxyglucose method for determining local cerebral glucose utilization is relatively insensitive to elevated plasma glucose concentrations. Nedergaard and coworkers {34} did note that LCMRgl could be depressed under conditions of extreme hyperglycemia. All in all these conditions of extreme hyperglycemia studied in animal models are not generally encountered in the clinical setting 1351. While ictal glucose values varied widely in our series, at the time of the PET (median, 7 days) all patient glucose values were within the range of 77 to 279 mg%. Also the determination of LCMRgl in small or irregular brain regions might produce falsely elevated values due to partial volume effects {361. This would not be a concern with larger lesions (relative to PET camera resolution). In our study, small or inapparent CT-detected lesions occurred with approximately equal frequency across the range of glucose levels. Larger CT-detected lesions were slightly more common when glucose levels were higher, but this difference was not significant. This actual distribution is protective against such ROI size errors. The decision to set the metabolic asymmetry to zero in normal scans was an arbitrary one designed to increase the number of cases for statistical reasons. Despite these caveats, several points concerning initial hyperglycemia in ischemic brain infarction emerge from our observations. Our results lend support to the notion that initial hyperglycemia is associated with poorer clinical outcome [G, 121. The blood glucose level at admission is significantly correlated with the extent of metabolic brain abnormahties seen on the acute PET, but not the final CT. Mainly our study differs from previous efforts because of the narrower
range of neurological disability (due to the exclusion of comatose or moribund patients) and the metabolic studies. Most previous clinical studies of this problem have focused on extreme clinical end points such as death and overall impairment of activities of dady living; no prior PET study has dealt with this problem [b, 7, 371. Certainly, stroke-related death and state of consciousness are important considerations, but in the majority of stroke patients long-term prognosis is more of a concern. A seemingly small, or even inapparent, structural lesion seen on CT or magnetic resonance imaging (MRI) may be associated with profound neurological deficits. Generally speaking, functional imaging with PET has shown that this situation often reflects widespread disturbance in cerebral metabolism that is not confined to the suspected ischemic zone 138, 391. Thus, the PET image of cerebral metabolism may be expected to be sensitive to changes in local ischemic tissue and remote functional disturbances in undamaged tissue. The term diaschisis has found increasing favor as an umbrella for these remote circulatorymetabolic effects {38). The exact nature of the relationships between the clinical findings, local ischemic tissue changes, and remote metabolic effects remains uncertain, although we found that widespread abnormalities detected on PET scans predict severe neurological deficits and more complex symptoms [40}. The data generated by emission tomography is multidimensional by nature, and conveys information about metabolism, anatomy, function, and cerebral state, among other variables. In this series the qualitative aspects of our PET results suggest that functional brain imaging may provide a more sensitive indicator of cerebral function than either bedside clinical testing
Kushner et
al:
Hyperglycernia in Ischemic Brain Infarction
133
or anatomical imaging. In a study group the size of ours, conclusions regarding initial hyperglycemia would not be possible based on the clinical and CT data alone. Differing animal models of cerebral infarction have yielded mixed results about the effects of hyperglycemia 11-5, 14-16]. Prado and colleagues 141) suggested that the site and vascular anatomy of the ischemic focus may be a determinant of the relative risk of ictal hyperglycemia. Those results suggested that ischemic regions with residual collateral circulation might be more vulnerable to ictal hyperglycemia than endarterial distributions such as lacunar infarction 142). Our series is too small to examine this point. The patient with acute stroke may experience hyperglycemia as a result of a stress reaction E433. Thus it is conceivable that such elevations in glucose levels may be an epiphenomenon of stroke itself. Melamed [44) found that severely affected patients with hemorrhagic infarction or coma exhibited the greatest elevation in fasting blood glucose levels after stroke. Candelise and coworkers 112) found a correlation between infarct size as determined by CT and the fasting glucose value obtained up to 48 to 72 hours after stroke. Our serum glucose data are not directly comparable to these previous findings because we examined the first serum glucose measurement before treatment, whether fasting or not, within 12 hours of the ictus. Fasting glucose values were obtained the day of the PET study in preparation for the FDG injection and these ranged no higher than the initial levels at admission.
This study was supported in part by Program Project grant NS14867-10 from the U.S. Public Health Service. Dr Kushner is the recipient of Clinical Investigator Development Award 1 KO8 NS00999-5.
References 1. Pulsinelli WA, Waldman S, Rawlinson D, Plum F. Moderate hyperglycemia augments ischemic brain damage: a neuropathologic study in the rat. Neurology 1982;32:1239-1246 2. Siemkowicz E, Hansen AJ. Clinical restitution following cerebral ischemia in hypo-, normo- and hyperglycemic rats. Acta Neurol Scand 1978;58:1-8 3. Siemkowicz E. Hyperglycemia in the reperfusion period hampers recovery from cerebral ischemia. Acta Neurol Scand 1981; 64207-216 4. Myers RE, Yamaguchi S. Nervous system effects of cardiac arrest in monkeys. Arch Neurol 1977;34:65-74 5. Brint S, Kraig R, Kiessling M, Pulsinelli W. Hyperglycemia augments infarct size in focal experimental brain ischemia. Ann Neurol 1985;18:12 7 6. Pulsinelli WA, Levy DE, Sigsbee B, et al. Increased damage after ischemic stroke in patients with hyperglycemia with or without established diabetes mellitus. Am J Med 1983;74:540544
134 Annals of Neurology Vol 28 No 2 August 1990
7. Levy DE, Pulsinelli WA, Scherer PB, Plum F. Effect of admission glucose level on recovery from acute stroke. Ann Neurol 1985;18: 122 8. Kraig RP, Petito CK, Plum F, Pulsinelli WA. Hydrogen ions kill brain at concentrations reached in ischemia. J Cereb Blood Flow Metab 1987;7:379-386 9. Chopp M, Frinak S, Walton DR, et al. Intracellular acidosis during and after cerebral ischemia: in vivo nuclear magnetic resonance study of hyperglycemia in cats. Stroke 1987;18:919923 10. Welsh FA, Ginsberg MD, Rieder W, Budd WW. Deleterious effect of glucose pretreatment on recovery from diffuse cerebral ischemia in the cat. 11. Regional metabolite levels. Stroke 1980;11:355-363 11. Plum F. What causes infarction in ischemic brain?: the Robert Wartenberg lecture. Neurology 1983;33:22 2-23 3 12. Candelise L, Landi G , Orazio EN, Boccardi E. Prognostic significance of hyperglycemia in acute stroke. Arch Neurol 1985; 42:661-663 13. Berger L, Hakim AM. The association of hyperglycemia with edema in stroke. Stroke 1986;17:865-871 14. Ginsberg MD, Prado R, Dietrich WD,et al. Hyperglycemia reduces the extent of cerebral infarction in rats. Stroke 1987; 18:570-574 15. Ibayashi S, Fujishima M, Sadoshima S, et al. Cerebral blood flow and tissue metabolism in experimental cerebral ischemia of spontaneously hypertensive rats with hyper-, normo-, and hypoglycemia. Stroke 1986:17:261-266 16. Nedegaard M, Diemer NH. Influence of hyperglycemia on ischemic brain damage after middle cerebral artery (MCA) occlusion in rat. Acta Neuropathol (Berl) 1985;73:131-137 17. Kushner M, Reivich M, Fieschi C, et al. Metabolic and clinical correlates of acute ischemic infarction. Neurology 1987;37: 1103-11 10 18. Mohr JP, Caplan LR, Melski JW, et al. The Harvard cooperative stroke registry. A prospective registry. Neurology 1978;28: 754-762 19. Shinar D, Gross CR, Mohr JP, et al. Inter-observer variability in the assessment of neurologic history and examination in the stroke data base. Arch Neurol 1985;42:557-565 20. Wade DT, Skilbeck CE, Hewer RL. Predicting Barthel A.D.L. at 6 months after an acute stroke. Arch Phys Med Rehabil 1983;64:24-28 21. Matsui R, Hirano A. An atlas of the human brain for computerized tomography. Tokyo: Igaku-Shoin, 1978 22. Reivich M, Kuhl DE, Wolf A, et al. The 18F-fluorodeoxyglucose method for the measurement of local cerebral glucose utihzation in man. Circ Res 1979;44:127-137 23. Sokoloff L, Reivich M, Kennedy C, et al. The ('*C) deoxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure, and normal values in the conscious and anesthetized albino rat. J Neurochem 1977;28:897-916 24. Reivich M, Alavi A, Wolf A, et al. Glucose metabolic rate kmetic model parameter determination in man: the lumped constant and rate constants for '8F-fluorodeoxyglucose and "Cdeoxyglucose. J Cereb Blood Flow Metab 1985;5:179-192 25. Ter Pogossiari MM, Mullani NA, Hood JT. Design considerations for a positron emission transverse tomograph (PETT V) for the imaging of the brain. J Comput Assist Tomogr 1978; 2:539-544 26. Sokoloff L. The radioactive deoxyglucose method; theory, procedure, and applications for the measurement of local glucose utilization in the central nervous system. In: Agranoff BW, Aprison MH, eds. Advances in neurochemistry, vol4. New York: Plenum, 1982:l-82 27. Gjedde A, Wienhard K, Heiss W-D, et al. Comparative regional analysis of 2-fluorodeoxyglucose and methylglucose up-
28.
29.
30.
3J .
32. 33.
34.
take in brain of four stroke patients: with special reference to the regional estimation of the lumped constant. J Cereb Blood Flow Metab 1985;5:163-178 Hawkins RA, Phelps ME, Huang SC, Kuhl DE. Effect of ischemia on quantification of local cerebral glucose metabolic rate in man. J Cereb Blood Flow Metab 1981;1:37-51 Ginsberg MD, Reivich M. Use of the 2-deoxyglucose method of local cerebral glucose utilization in the abnormal brain: evaluation of the lumped constant during ischemia. Acta Neurol Scand 1979;60(suppl 72k226-227 Schuier F, Orzi F, Suda S, et al. The lumped constant for the ('4C)deoxyglucose method in hyperglycemic rats. J Cereb Blood Flow Metab 1981;l(suppl 1):S63 Pettigrew KD, Sokoloff L, Patlak CS. A theoretical derivation of the lumped constant for the 2-deoxyglucose method for measuring local glucose metabolism in the brain. J Cereb Blood Flow Metab 1981;l(suppl 1):S89-90 Sokoloff L.The ('4C)deoxyglucose method. Acta Neurol Scand 1979;60(suppl 1721640-649 Lucignani G, Dow-Edward D, Namba H, et al. Local cerebral glucose utilization in hyperglycemia.J Cereb Blood Flow Metab 1987;7(suppl 1):S499 Nedergaard M, Jakobsen J, Diemer N H . Autoradiographc determination of cerebral glucose content, blood flow, and glucose utilization in focal ischemia of the rat brain: influence of the plasma glucose concentration. J Cereb Blood Flow Metab 1988;8:100-108
35. Duverger D, MacKenzie ET. The quantification of cerebral infarction following focal ischemia in the rat: influence of strain,
36. 37.
38.
39. 40.
41.
42. 43. 44.
arterial pressure, blood glucose concentration, and age. J Cereb Blood Flow Metab 1988;8:649-461 Mazziotta JC, Phelps ME, Plummer D, Kuhl DE. Quantitation in positron emission computed tomography. V. Physical-anatomic effect. J Comput Assist Tomogr 1981;5:734-743 Woo E, Chan YW, Yu YL, Huang CY. Admission glucose level in relation to mortality and morbidity outcome in 252 stroke patients. Stroke 1988;19:185-191 Feeney DM, Baron JC. Diaschisis. Stroke 1986;17:817-830 Kushner M, Tobin M, Alavi A, et al. Cerebellar glucose consumption in normal and pathological states. J Nucl Med 1987; 2811667-1 670 Kushner M, Reivich M. Functional imaging of brain ischemia. In: Weinberger J, ed. Frontiers in clinical neuroscience, vol 5 . New York: Alan R. Liss, 1989163-174 Prado R, Ginsberg M, Dietrich WD, et al. Hyperglycemia increases infarct size in collaterally perfused but not end-arterial vascular territories. J Cereb Blood Flow Metab 1988;8:186192 Fisher CM. Lacunar strokes and infarcts: a review. Neurology 1982;32:871-876 Feibel JH, Hardy PM, Campbell RG, et al. Prognostic value of the stress following stroke. JAMA 1977;238:1374-1376 Melamed E. Reactive hyperglycemia in patients with acute stroke. J Neurol Sci 1076;29:267-275
Kushner er al: Hyperglycemia in Ischemic Brain Infarction 135
We studied the relation of serum glucose level measured in the first 12 hours of symptoms to the clinical findings, results of computed tomography (CT), and patterns of cerebral metabolism in 39 patients who had acute ischemic cerebral infarction. Structural damage was assessed by CT. Metabolic disruption was assessed using "F-fluorodeoxyglucose and positron emission tomography (PET). Median initial serum glucose concentration was 155 mddl(6.7 d) Clinical . recovery was significantly poorer in patients with initial serum glucose levels higher than the median ( p < 0.05, chi square). PET tended to show normal results or minor abnormalities in patients with initial glucose levels less than the median, as opposed to lobar or multilobe abnormalities in patients with levels that were higher than the median ( p < 0.05, Kendall's Tau b). The severity of hypometabolism in the ischemic region, expressed as the percent asymmetry of local cerebral glucose metabolism between homologous brain regions, was greater in patients with initial glycemia concentrations higher than the median ( p < 0.001, t test). Relationships of serum glucose level with metabolic derangement and structural damage, but not outcome, held true in patients without a history of diabetes mellitus. Kushner M, Nencini P, Reivich M, Rango M, Jamieson D, Fazekas F, Zimmerman R, Chawluk J, Alavi A, Alves W. Relation of hyperglycemia early in ischemic brain infarction to cerebral anatomy, metabolism, and clinical outcome. Ann Neurol 1990;28:129-135
Observations in animals and humans have suggested that an elevated serum glucose concentration at or about the time of an ischemic insult is associated with enhanced postischemic damage of cerebral tissue f 17). A variety of putative mechanisms now have been proposed whereby glucose, or its metabolites, might serve to augment postischemic damage r8- 111. Pulsinelli and colleagues {GI and Levy and associates [73 first suggested that moderate ictal hyperglycemia, in excess of 120 mg/dl (6.7 mM), was associated with poorer clinical outcome in hospitalized stroke patients. Some studies suggested that there is a relationship between ictal hyperglycemia and the extent of the ischemic lesion observed on imaging studies {12, 131. A potential implication of these observations is that early measures to control serum glucose concentrations after
infarction might serve to limit stroke severity in the clinical setting. Despite these previous reports, other experimental studies of hyperglycemia in stroke yielded contrary, or mixed, results C14-161. Nedergaard and Diemer [161 found that ictal hyperglycemia does not impact on stroke severity, while Ginsberg and colleagues r141 reported a protective influence of hyperglycemia following ischemia. In this report we describe the results of OUT observations on the relationships between serum glucose level at the time of an ischemic ictus with the clinical findings, computed tomography (CT) results, and patterns of cerebral metabolism measured using "F-fluorodeoxyglucose (FDG) and positron emission tomography (PET). This study was undertaken to determine
From the "Cerebrovascular Research Center and the Departments of *Neurology, $Neurosurgery, and $Radiology, University of Pennsylvania, Philadelphia, PA.
Address correspondence to Dr Kushner, Cerebrovascular Research Center, Room 429, Johnson Pavilion, University of Pennsylvania, Philadelphia, PA 19104.
Received Jun 30, 1989, and in revised form Dec 14, 1989, and Jan 18, 1990. Accepted for publication Jan 25, 1990.
Copyright 0 1990 by the American Neurological Association
129
whether or not initial serum glucose level was related to stroke severity, the extent of tissue damage, or metabolic disruption.
Methods Patients hospitalized for acute cerebral infarction at the Hospital of the University of Pennsylvania or the Graduate Hospital were eligible for inclusion in this study. Informed consent was obtained for each patient under the guidelines of the Committee on Studies Involving Human Subjects of the University of Pennsylvania [ 17).
R.Z.). The CT-detected lesions were mapped on neuroanatomical templates derived from a standardized tomographicanatomical atlas [2 1). The size of the CT lesions was graded according to a noncontinuous incremental scale described previously [17}: I = no lesion, I1 = small lesion less than 2.0 cm in diameter, 111 = lesion larger than 2.0 cm but contained within one cerebral lobe, and IV = multilobe abnormality. We used the last CT scan performed in order to reduce error arising from a CT with no findings performed soon after the onset of clinical manifestations.
Metabolic Methods The criteria for patient selection were clinical and CT evidence of acute nonhemorrhagic cerebral infarction with persistent focal neurological deficits at least until the time of PET study. Exclusionary criteria included: history of cerebral infarction; age less than 21 years; presence of stupor or coma; pregnancy or child-bearing potential; CT evidence of intracerebral hemorrhage; presence of primary subarachnoid hemorrhage; diabetic ketoacidosis or hyperosmolar state; and severe systemic disorders such as myocardial infarction or renal, pulmonary, or hepatic failure. Clinical data were acquired prospectively using standardized medical history, neurological history, and physical exarnination forms derived from standardized data base registries { 18-20). Patient follow-up was also performed according to protocol. Clinical data were scored so as to rank neurological disability, degree of impairment of activities of daily living, and extent of neurological recovery. The degree of recovery was classified into four groups based on the relative degree of restoration of function compared to the ictal nadir according to a scale previously described [17). Patients were classified as having made either a complete, good, moderate, or poor recovery. Records from the emergency medical systems, hospital emergency department, and inpatient charts were reviewed in order to obtain data concerning the first serum blood glucose level determined within 12 hours of the ischemic ictus. If a blood glucose level value within 12 hours of the ictus was not available, or if no blood glucose value was obtained prior to the initiation of dextrose-containing intravenous infusions or insulin therapy, the patient was excluded from study. Serum glucose level was determined by the enzymatic method. Hospital records were reviewed, and patients and next of kin were interviewed in order to determine any history of diabetes mellitus or previous treatment with hypoglycemic agents, and the results of laboratory tests were used to determine the presence or absence of diabetes mellitus. Patients who had previously established diabetes mellitus or treatment with hypoglycemic agents were considered nondiabetic.
All patients underwent imaging to determine local cerebral glucose metabolism (LCMRgl) using FDG and PET. The first PET study was performed as close to the time of the ischemic ictus as possible, with a mean of 7 days after infarction. Fifty-nine percent of patients had repeat PET studies from 2 weeks to 6 months after infarction. The technique employed has been described previously C22, 231. The measured value of the lumped constant, 0.542, was used [24]. Images of LCMRgl were obtained using a modified version of the positron emission transverse tomograph (PE'IT V) camera C251. Local glucose consumption waq calculated using the method of Sokoloff and associates, in which a correction for the small loss of label due to phosphatase activity was used 123, 261. The PET data and images were analyzed both quantitatively and qualitatively. Quantitative analysis of the PET data was performed as follows. The whole brain metabolic rate for glucose (WBMRgI) was calculated for each patient. LCMRgl was calculated in the region of the cerebral ischemia as follows. All 14 intercalated PET slices were reviewed independent of the CT or clinical data (by M. K. and F. F.). The slice containing the region with the most severely disturbed metabolism was then selected. Using a visual display unit and irregular regions of interest overlay programs, a region of interest (ROI) was inscribed so as to be totally contained within thc area of the maximum metabolic abnormality. In order to avoid partial volume effects we attempted to make the ROI span at least one image resolution element (17 mm), but when needed, smaller ROIs of lesser span were employed in order to inscribe totally the ROI within the area of maximum metabolic abnormality. LCMRgl was also determined from a ROI of similar size and shape at the homologous site in the opposite hemisphere; brainstem lesions were excluded from quantitative analysis. When possible, the correctness of the PET localization of the lesion was assessed post hoc by reference to the latest CT scan. For this study the severity of metabolic disturbance was expressed as the percentage asymmetry between LCMRgl for the lesion and homologous normal brain by the expression:
RadiologicaL Methods
(contralateral LCMRgl - lesion LCMRgl) (contralateral LCMFQl)
All patients underwent conventional x-ray CT of the head using either a General Electric (Milwaukee, WI) 9800 or 8800 machine. Sixty-two percent of patients underwent follow-up CT, because of a normal or nondiagnostic initial study, from 5 days to 6 months following the ictus. For each case the last CT scan performed was reviewed blindly (by
In cases where no clear-cut metabolic abnormality could be identified, lesion ROIs were not defined and the percent asymmetry of LCMRgl was considered equal to zero. Qualitative analysis of the PET image was performed as follows. All metabolic images were reviewed blindly by three
Clinical Methods
130 Annals of Neurology Vol 28 No 2 August 1990
independent observers (M. K., A. A., and J. C.), who related the observed metabolic abnormalities to neuroanatomical templates derived from a standardized tomographic-anatomical atlas {21]. Ail PET readings were performed independent of the CT findings. Readers were asked to outline on the templates the area of most intense hypometabolism, which was presumed to represent the core of the ischemic focus, and the extent of surrounding hypometabolism. The overall spatial extent of the metabolic abnormality (e.g., ischemic core and remote hypometabolic regions) was defined when at least two readers agreed. PET images were scored qualitatively and assigned to categories that have been described previously [17]. In brief, the PET scans were rated as follows: I = normal scan, I1 = small abnormality on the order of an image resolution element (17 mm), I11 = abnormal metabolism confined to one cerebral lobe, and IV = multilobe metabolic abnormality.
Data Analysis Statistical analyses were designed to compare the relationship between serum glucose level at admission and the clinical, CT, and metabolic findings. Ictal serum glucose level was also compared with demographic data, such as age and gender, and the presence or absence of other risk factors for vascular disease, such as hypertension and diabetes mellitus. These comparisons were performed nonparametrically using the median test or Fisher's exact probability, depending on cell size. Within each analysis, patients were stratdied according to the median serum glucose value. A separate correlational analysis was also performed for patients with and those without a history of diabetes mellitus. The qualitative CT and PET gradings were coded according to the ordinal rank, and correlation coefficients were computed using Kendall's Tau b, a nonparametric counterpart of Pearson's r. Correlations between continuous data scales were performed using Spearman's rank correlation coefficient (rs).The quantitative PET data including WBMFigl, lesion LCMRgl, and asymmetry values were analyzed using Student's t test.
Results Clinical Results Thirty-nine patients (21 men, 18 women) fulfilled the conditions for entry into the study. The median age
was 65 years (range, 5 1 to 82 years). Thirty-six patients had a hemispheric lesion; 44% of these were in the right hemisphere and 56% in the left. In 13 patients the ischemic area was subcortical and in 3 patients the lesion was localized to the brainstem. The median serum glucose value at admission was 155 mg/dl for the total study group (range, 78 to 286 mgldl). The follow-up period ranged from 14 to 180 days. Average duration of clinical follow-up was 104 days for the entire group. The analyses of initial serum glucose values versus patient clinical characteristics are summarized in Table 1. Recovery was poorer when initid serum glucose levels were higher ( p < 0.05, chi square). History of diabetes mellitus was prominent in the hyperglycemic group ( p < 0.01, chi square). No relationship was noted between initial serum glucose level and age, history of hypertension, previous transient ischemic attacks, or location of the lesion (e.g., cortical versus subcortical). A separate analysis showed that history of diabetes mellims also was associated with worse recovery ( p < 0.05, chi square). Radiological Findings The mean interval between the ischemic ictus and the latest CT scan was 87 days. Grades of the CT-detected lesions did not differ across the range of glucose values (Table 2). No significant relationship was noted between the size of the CT lesion and a history of diabetes mellitus (Kendall's Tau b = 0.147, p = 0.16, not significant). Metabolic Findings Overall 2 hemispheric subcortical lesions and 2 subtentorial lacunes presented no clear-cut abnormalities on PET. The PET scans showed that the extent of the metabolic abnormality tended to be greater than expected, on the basis of the CT data, in the patients whose initial glucose values were 155 mg/dl or greater (see Table 2). No disparity between CT and PET data
Table I . Clinical Characteristics versus Initial Serum Glucose Levela Level < 155 mg/dl (n = 19) Age (yr) 5 65 > 65 Recovery Complete or good Moderate or poor Diabetes mellitus Yes No
Level 2 155 mg/dl (n = 20)
9
11
10
9
13
6
6
14
1 17
17
4
Statistical Values Chisquare
p
=
=
0.023
=
4.32
NS
Chi square
p < 0.05 Chi square = 19.2
p < 0.001
"Values are numbers of patients.
NS
=
not statistically sigriificant
Kushner et
al:
Hyperglycemia in Ischemic Brain Infarction
131
Table 2. C T and PET ResultS oers'slls Initial Serum Glucose Level Level < 155 mgid Lesion Grade
(n = 19)
Table 3. Comparison of WBMRgl and LCMRgl Values versuJ Initial Serum Glucose Level Level 2 155 mgidl (n = 20)
Level < 155 mddl
WBMR~I~
CT"
I
Lesion LCMRgl'
I1 I11
Asymmetry
IV PETb
I I1 I11 IV
3 6 0 11
5.9 +- 1.3 (n = 17)
Level 2 155 mddl
* 1.0
< 0.02
(n = 19) 1.3 t 1.1 (n = 17)
< 0.05
3.8
2.3 ? 1.0 (n = 11) 18.6 7.1525 44.9 -t 17.2% (n = 17) (n = 19)
*
p Value"
< 0.001
"Probability values are derived from Student's t test. bMean WBMRgl in age-matched elderly control subjects is 5.3 f 0.8 mg/lOO g d m i n (n = 14). CLCMRglvalues are mean i 1 standard deviation (units are mgl100 gdminute).
"Kendall's Tau b = 0.147; p value is not significant for CT lesion grade versus glucose level. bKendall's Tau b = 0.267, p < 0.05 for PET lesion grade versus glucose level.
WBMRgl = whole brain metabolic rate for glucose; LCMRgl = local cerebral glucose metabolism.
was noted for when initial glucose values were less than 155 mg/dl. Qualitative analysis of the PET scans showed that lobar or multilobe hypometabolism was associated with initial glucose levels of 155 mg/dl or more (Kendall's Tau b = 0.267, p < 0.05). No correlation was found between the extent of the PET-detected lesion and history of diabetes mellitus (Kendall's Tau b = 0.18, p value was not significant). Quantitative PET data were available for 36 patients (92%); three studies suffered from technical difficulties such as the lack of arterial blood samples or other technical deficiencies. The quantitative PET analyses are summarized in Table 3. Overall WBMRgl was significantly lower in the group with glucose concentrations that were above the median ( p < 0.02, t test). In the 26 patients with a clear-cut supratentord lesion on the PET scan, significantly lower glucose consumption in the core of the ischemic area was found for the group with initial hyperglycemia ( p < 0.05, t test). A more severe metabolic disturbance within the ischemic core, expressed as the percentage of asymmetry of LCMRgl in homologous brain regions, was present in the group with initial hyperglycemia (44.9 k 17.2% versus 18.6 t 7.196; p < 0.001, t test).
Twenty-two patients were identified as being nondiabetic. The median ictal serum glucose value was 123 mgldl. No significant relationship was found between the extent of recovery and initial serum glucose value (Table 4). Ranking of the relative sizes of the ultimate ischemic lesion found on last CT scans did show a significant correlation between lesion size and ictal glucose level (Spearman's ~t(= 0.433,df = 20, p < 0.05, data not shown). Both quantitative and qualitative metabolic analyses showed significant relationships between ictal serum glucose level and the PET data (see Table 4). The severity of lesion hypometabolism was significantly greater in patients with initial hyperglycemia. Also a significantly larger lesion as detected by PET was associated with initial hyperglycemia.
Nondiabetic Patients As mentioned, diabetes mellitus also was associated with poorer clinical outcome. Not surprisingly diabetes mellitus was associated with higher ictal serum gluose levels as well. Thus, interaction between diabetes mellitus and hyperglycemia might confound analysis over the entire group. In order to control for the influences of diabetes mellitus, a separate analysis of the data was undertaken in the patients without a history of diabetes mellitus. 132 Annals of Neurology Vol 28 No 2 August 1990
Discussion Before we discuss the implications of these findings, some potential sources of error should be mentioned. PET measurements of LCMRgl under ischemic conditions require some caution in their interpretation {27). Ischemia may perturb the deoxyglucose model by altering the rate constants or the lumped constant r28, 29}. Moreover, the lumped constant may be expected to increase appreciably at plasma glucose concentrations lower than 70 mgldl(3.9 mM), and be marginally depressed when serum glucose levels are greater than 300 mgldl(16.7 mM) 130, 311. Theoretically the effect of using a lumped constant that is artifactually low, or high, is to increase or decrease respectively calculated LCMRgl values. Protective from errors arising from rate constant alterations is the demonstration by Sokoloff 132) that LCMRgl is little perturbed if measurements are integrated over 40 minutes, as is customary in our laboratory. Practically speaking, Lu-
Table 4. Rerovely and CT and PET Lesion Grade versus Initial Serum Glucose Level in Nondiabetic Patient5 Serum Glucose Level
> 123 mgidl (n = 11)
Statistical Values
6 5
p
3 3 3
1 2
-
5
4
p
0
4
6 2 1
1 4 1
2 10.5
5
< 123 mgidl (n Recovery Complete or good Moderate or poor CT lesion grade I I1 111 IV PET lesion grade I I1 111
IV Asymmetry of lesion LCMRgl
=
11)
8
?
3.7%
34.7
= 0.24 Fisher's exact probability
= 0.54 Kendall's Tau b
-
p < 0.05 ?
4.2%
Kendall's Tau b p < 0.002, t test
LCMRgl = local cerebral glucose metabolism.
cignani and associates [331 found in animals that the 14C-deoxyglucose method for determining local cerebral glucose utilization is relatively insensitive to elevated plasma glucose concentrations. Nedergaard and coworkers {34} did note that LCMRgl could be depressed under conditions of extreme hyperglycemia. All in all these conditions of extreme hyperglycemia studied in animal models are not generally encountered in the clinical setting 1351. While ictal glucose values varied widely in our series, at the time of the PET (median, 7 days) all patient glucose values were within the range of 77 to 279 mg%. Also the determination of LCMRgl in small or irregular brain regions might produce falsely elevated values due to partial volume effects {361. This would not be a concern with larger lesions (relative to PET camera resolution). In our study, small or inapparent CT-detected lesions occurred with approximately equal frequency across the range of glucose levels. Larger CT-detected lesions were slightly more common when glucose levels were higher, but this difference was not significant. This actual distribution is protective against such ROI size errors. The decision to set the metabolic asymmetry to zero in normal scans was an arbitrary one designed to increase the number of cases for statistical reasons. Despite these caveats, several points concerning initial hyperglycemia in ischemic brain infarction emerge from our observations. Our results lend support to the notion that initial hyperglycemia is associated with poorer clinical outcome [G, 121. The blood glucose level at admission is significantly correlated with the extent of metabolic brain abnormahties seen on the acute PET, but not the final CT. Mainly our study differs from previous efforts because of the narrower
range of neurological disability (due to the exclusion of comatose or moribund patients) and the metabolic studies. Most previous clinical studies of this problem have focused on extreme clinical end points such as death and overall impairment of activities of dady living; no prior PET study has dealt with this problem [b, 7, 371. Certainly, stroke-related death and state of consciousness are important considerations, but in the majority of stroke patients long-term prognosis is more of a concern. A seemingly small, or even inapparent, structural lesion seen on CT or magnetic resonance imaging (MRI) may be associated with profound neurological deficits. Generally speaking, functional imaging with PET has shown that this situation often reflects widespread disturbance in cerebral metabolism that is not confined to the suspected ischemic zone 138, 391. Thus, the PET image of cerebral metabolism may be expected to be sensitive to changes in local ischemic tissue and remote functional disturbances in undamaged tissue. The term diaschisis has found increasing favor as an umbrella for these remote circulatorymetabolic effects {38). The exact nature of the relationships between the clinical findings, local ischemic tissue changes, and remote metabolic effects remains uncertain, although we found that widespread abnormalities detected on PET scans predict severe neurological deficits and more complex symptoms [40}. The data generated by emission tomography is multidimensional by nature, and conveys information about metabolism, anatomy, function, and cerebral state, among other variables. In this series the qualitative aspects of our PET results suggest that functional brain imaging may provide a more sensitive indicator of cerebral function than either bedside clinical testing
Kushner et
al:
Hyperglycernia in Ischemic Brain Infarction
133
or anatomical imaging. In a study group the size of ours, conclusions regarding initial hyperglycemia would not be possible based on the clinical and CT data alone. Differing animal models of cerebral infarction have yielded mixed results about the effects of hyperglycemia 11-5, 14-16]. Prado and colleagues 141) suggested that the site and vascular anatomy of the ischemic focus may be a determinant of the relative risk of ictal hyperglycemia. Those results suggested that ischemic regions with residual collateral circulation might be more vulnerable to ictal hyperglycemia than endarterial distributions such as lacunar infarction 142). Our series is too small to examine this point. The patient with acute stroke may experience hyperglycemia as a result of a stress reaction E433. Thus it is conceivable that such elevations in glucose levels may be an epiphenomenon of stroke itself. Melamed [44) found that severely affected patients with hemorrhagic infarction or coma exhibited the greatest elevation in fasting blood glucose levels after stroke. Candelise and coworkers 112) found a correlation between infarct size as determined by CT and the fasting glucose value obtained up to 48 to 72 hours after stroke. Our serum glucose data are not directly comparable to these previous findings because we examined the first serum glucose measurement before treatment, whether fasting or not, within 12 hours of the ictus. Fasting glucose values were obtained the day of the PET study in preparation for the FDG injection and these ranged no higher than the initial levels at admission.
This study was supported in part by Program Project grant NS14867-10 from the U.S. Public Health Service. Dr Kushner is the recipient of Clinical Investigator Development Award 1 KO8 NS00999-5.
References 1. Pulsinelli WA, Waldman S, Rawlinson D, Plum F. Moderate hyperglycemia augments ischemic brain damage: a neuropathologic study in the rat. Neurology 1982;32:1239-1246 2. Siemkowicz E, Hansen AJ. Clinical restitution following cerebral ischemia in hypo-, normo- and hyperglycemic rats. Acta Neurol Scand 1978;58:1-8 3. Siemkowicz E. Hyperglycemia in the reperfusion period hampers recovery from cerebral ischemia. Acta Neurol Scand 1981; 64207-216 4. Myers RE, Yamaguchi S. Nervous system effects of cardiac arrest in monkeys. Arch Neurol 1977;34:65-74 5. Brint S, Kraig R, Kiessling M, Pulsinelli W. Hyperglycemia augments infarct size in focal experimental brain ischemia. Ann Neurol 1985;18:12 7 6. Pulsinelli WA, Levy DE, Sigsbee B, et al. Increased damage after ischemic stroke in patients with hyperglycemia with or without established diabetes mellitus. Am J Med 1983;74:540544
134 Annals of Neurology Vol 28 No 2 August 1990
7. Levy DE, Pulsinelli WA, Scherer PB, Plum F. Effect of admission glucose level on recovery from acute stroke. Ann Neurol 1985;18: 122 8. Kraig RP, Petito CK, Plum F, Pulsinelli WA. Hydrogen ions kill brain at concentrations reached in ischemia. J Cereb Blood Flow Metab 1987;7:379-386 9. Chopp M, Frinak S, Walton DR, et al. Intracellular acidosis during and after cerebral ischemia: in vivo nuclear magnetic resonance study of hyperglycemia in cats. Stroke 1987;18:919923 10. Welsh FA, Ginsberg MD, Rieder W, Budd WW. Deleterious effect of glucose pretreatment on recovery from diffuse cerebral ischemia in the cat. 11. Regional metabolite levels. Stroke 1980;11:355-363 11. Plum F. What causes infarction in ischemic brain?: the Robert Wartenberg lecture. Neurology 1983;33:22 2-23 3 12. Candelise L, Landi G , Orazio EN, Boccardi E. Prognostic significance of hyperglycemia in acute stroke. Arch Neurol 1985; 42:661-663 13. Berger L, Hakim AM. The association of hyperglycemia with edema in stroke. Stroke 1986;17:865-871 14. Ginsberg MD, Prado R, Dietrich WD,et al. Hyperglycemia reduces the extent of cerebral infarction in rats. Stroke 1987; 18:570-574 15. Ibayashi S, Fujishima M, Sadoshima S, et al. Cerebral blood flow and tissue metabolism in experimental cerebral ischemia of spontaneously hypertensive rats with hyper-, normo-, and hypoglycemia. Stroke 1986:17:261-266 16. Nedegaard M, Diemer NH. Influence of hyperglycemia on ischemic brain damage after middle cerebral artery (MCA) occlusion in rat. Acta Neuropathol (Berl) 1985;73:131-137 17. Kushner M, Reivich M, Fieschi C, et al. Metabolic and clinical correlates of acute ischemic infarction. Neurology 1987;37: 1103-11 10 18. Mohr JP, Caplan LR, Melski JW, et al. The Harvard cooperative stroke registry. A prospective registry. Neurology 1978;28: 754-762 19. Shinar D, Gross CR, Mohr JP, et al. Inter-observer variability in the assessment of neurologic history and examination in the stroke data base. Arch Neurol 1985;42:557-565 20. Wade DT, Skilbeck CE, Hewer RL. Predicting Barthel A.D.L. at 6 months after an acute stroke. Arch Phys Med Rehabil 1983;64:24-28 21. Matsui R, Hirano A. An atlas of the human brain for computerized tomography. Tokyo: Igaku-Shoin, 1978 22. Reivich M, Kuhl DE, Wolf A, et al. The 18F-fluorodeoxyglucose method for the measurement of local cerebral glucose utihzation in man. Circ Res 1979;44:127-137 23. Sokoloff L, Reivich M, Kennedy C, et al. The ('*C) deoxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure, and normal values in the conscious and anesthetized albino rat. J Neurochem 1977;28:897-916 24. Reivich M, Alavi A, Wolf A, et al. Glucose metabolic rate kmetic model parameter determination in man: the lumped constant and rate constants for '8F-fluorodeoxyglucose and "Cdeoxyglucose. J Cereb Blood Flow Metab 1985;5:179-192 25. Ter Pogossiari MM, Mullani NA, Hood JT. Design considerations for a positron emission transverse tomograph (PETT V) for the imaging of the brain. J Comput Assist Tomogr 1978; 2:539-544 26. Sokoloff L. The radioactive deoxyglucose method; theory, procedure, and applications for the measurement of local glucose utilization in the central nervous system. In: Agranoff BW, Aprison MH, eds. Advances in neurochemistry, vol4. New York: Plenum, 1982:l-82 27. Gjedde A, Wienhard K, Heiss W-D, et al. Comparative regional analysis of 2-fluorodeoxyglucose and methylglucose up-
28.
29.
30.
3J .
32. 33.
34.
take in brain of four stroke patients: with special reference to the regional estimation of the lumped constant. J Cereb Blood Flow Metab 1985;5:163-178 Hawkins RA, Phelps ME, Huang SC, Kuhl DE. Effect of ischemia on quantification of local cerebral glucose metabolic rate in man. J Cereb Blood Flow Metab 1981;1:37-51 Ginsberg MD, Reivich M. Use of the 2-deoxyglucose method of local cerebral glucose utilization in the abnormal brain: evaluation of the lumped constant during ischemia. Acta Neurol Scand 1979;60(suppl 72k226-227 Schuier F, Orzi F, Suda S, et al. The lumped constant for the ('4C)deoxyglucose method in hyperglycemic rats. J Cereb Blood Flow Metab 1981;l(suppl 1):S63 Pettigrew KD, Sokoloff L, Patlak CS. A theoretical derivation of the lumped constant for the 2-deoxyglucose method for measuring local glucose metabolism in the brain. J Cereb Blood Flow Metab 1981;l(suppl 1):S89-90 Sokoloff L.The ('4C)deoxyglucose method. Acta Neurol Scand 1979;60(suppl 1721640-649 Lucignani G, Dow-Edward D, Namba H, et al. Local cerebral glucose utilization in hyperglycemia.J Cereb Blood Flow Metab 1987;7(suppl 1):S499 Nedergaard M, Jakobsen J, Diemer N H . Autoradiographc determination of cerebral glucose content, blood flow, and glucose utilization in focal ischemia of the rat brain: influence of the plasma glucose concentration. J Cereb Blood Flow Metab 1988;8:100-108
35. Duverger D, MacKenzie ET. The quantification of cerebral infarction following focal ischemia in the rat: influence of strain,
36. 37.
38.
39. 40.
41.
42. 43. 44.
arterial pressure, blood glucose concentration, and age. J Cereb Blood Flow Metab 1988;8:649-461 Mazziotta JC, Phelps ME, Plummer D, Kuhl DE. Quantitation in positron emission computed tomography. V. Physical-anatomic effect. J Comput Assist Tomogr 1981;5:734-743 Woo E, Chan YW, Yu YL, Huang CY. Admission glucose level in relation to mortality and morbidity outcome in 252 stroke patients. Stroke 1988;19:185-191 Feeney DM, Baron JC. Diaschisis. Stroke 1986;17:817-830 Kushner M, Tobin M, Alavi A, et al. Cerebellar glucose consumption in normal and pathological states. J Nucl Med 1987; 2811667-1 670 Kushner M, Reivich M. Functional imaging of brain ischemia. In: Weinberger J, ed. Frontiers in clinical neuroscience, vol 5 . New York: Alan R. Liss, 1989163-174 Prado R, Ginsberg M, Dietrich WD, et al. Hyperglycemia increases infarct size in collaterally perfused but not end-arterial vascular territories. J Cereb Blood Flow Metab 1988;8:186192 Fisher CM. Lacunar strokes and infarcts: a review. Neurology 1982;32:871-876 Feibel JH, Hardy PM, Campbell RG, et al. Prognostic value of the stress following stroke. JAMA 1977;238:1374-1376 Melamed E. Reactive hyperglycemia in patients with acute stroke. J Neurol Sci 1076;29:267-275
Kushner er al: Hyperglycemia in Ischemic Brain Infarction 135
