Hypodensity of the Cerebral n t e Matter in Patients with Transient Ischemic Attack or Minor Stroke: Influence on the Rate of Subsequent Stroke J. C. van Swieten, MD," L. J. Kappelle, MD," A. Algra, MD," J. C. van Latum, MD,f P. J. Koudstaal, MI),? and J. van Gijn, MD,* for the Dutch TIA Trial Study Group
In a prospective study of 3,017 patients with transient ischemic attack or minor ischemic stroke from the Dutch Transient Ischemic Attack Trial, the presence or absence of diffuse hypodensity of the white matter on a baseline computed tomography (CT) scan of the brain was related to the occurrence of subsequent stroke. On entry, 337 patients were judged to have diffuse hypodensity of the white matter on CT; they were older (71.4 +- 7.4 years versus 64.4 +- 9.9 years), more often had hypertension (50% versus 41%), and more often had lacunar infarcts on CT scan (40% versus 26%) than did patients with normal white matter. Strokes, fatal or nonfatal, occurred in 51 (15%) of the patients with diffuse hypodensity of the cerebral white matter, compared to 217 (8%) in the group with normal white matter (crude hazard ratio, 2.0; 95% confidence interval, 1.4-2.7). After adjustment for age and other relevant entry variables, the hazard ratio was 1.6 (95% confidence interval, 1.2-2.2). In patients younger than 7 0 years the crude hazard ratio was 2.7 (95% confidence interval, 1.7-4.2). The distribution between the main subtypes of stroke was similar for patients with and those without diffuse hypodensity of the cerebral white matter: Intracerebral hemorrhage occurred in 6 and 996, cortical infarction in 47 and 4595, and lacunar infarction in 34 and 2996, respectively. We conclude that hypodensity of the cerebral white matter in patients with transient ischemic attack or minor stroke is associated with an extra risk of future stroke, from large as well as from small vessels, and particularly in patients under 7 0 years old; this increase of risk is independent of other risk factors for stroke. van Swieten JC, Kappelle LJ, Algra A, van Latum JC, Koudstaal PJ, van Gijn J. Hypodensity of the cerebral white matter in patients with transient ischemic attack or minor stroke: influence on the rate of subsequent stroke. Ann Neurol 1992;32:177-183
Diffuse hypodensity of the cerebral white matter (HWM) on computed tomography (CT) frequently is seen in elderly patients [1-3). Subcortical dementia with this type of white matter abnormality is known as Binswanger's disease. These patients usually have hypertension and often a history of one or more strokes [4-61. Moreover, macroscopic study of the brains of these demented patients at autopsy or by CT usually shows lacunar infarcts in the deep white matter or basal gangha {5-91. This type of white matter disease has been attributed to small-vessel disease, as arteriolosclerosis is always present [lo]. A similar but less marked HWM on CT has also been found in about 10% of patients with cerebral ischemia without dementia [ll, 12). The pathological findings in these patients resemble those of Binswanger's disease 12, 9, 13, 141.
It is a reasonable but unproved assumption that those patients with a transient ischemic attack (TIA) or a nondisabling ischemic stroke who also show HWM on CT have more generalized vascular disease and are more at extra risk for stroke recurrence. Until now, only advanced age, hypertension, the severity of the first stroke, and a history of myocardial infarction have been identified as risk factors for the recurrence of stroke in patients with cerebral ischemia [15--181, whereas antihypertensive treatment, antiplatelet treatment, and carotid surgery in appropriate patients have been found to reduce the risk of stroke [19-21) The aim of the present study was to investigate the relation between the presence of HWM on the baseline CT and the incidence of subsequent stroke in a large and prospective series of patients with TIA or nondisabling stroke, who participated in a clinical trial
From the *Department of Neurology, University Hospital Utrecht, Utrecht, and the tDepartment of Neurology, University Hospital Rotterdam, Rotterdam, The Netherlands.
Received Oct 3, 1991, and in revised form Jan 13, 1992. Accepted for publication Feb 23, 1992. Address correspondence to Dr van Swieten, Department of Neurology, University Hospital Utrecht, Heidelberglaan 100, 3584 CX Utrecht. The Netherlands.
Copyright 0 1992 by the American Neurological Association
177
of antiplatelet agents. Furthermore, we addressed the question of whether the type of stroke on follow-up differs between patients with and those without HWM.
Patients and Methods Between March 1986 and March 1989, 3,150 patients with a TIA or nondisabling ischemic stroke (Rankin grade 3 or less) were randomized in the Dutch TIA Trial [22, 231. In this trial the preventive effects of 30 mg of acetylsalicylic acid were compared with those of 283 mg, in a double-blind design, and 50 mg of atenolol versus placebo was similarly tested. Exclusion criteria were identifiable causes of ischemia other than arterial thrombosis or thromboembolism, an interval of more than 3 months since the last event, and an indication or contraindication for the trial drugs. At randomization, the history of each patient was recorded on a checklist worded in ordinary language, a method reasonably consistent between observers 1241. A TIA was defined as a focal neurological deficit with abrupt or relatively rapid onset, persisting less than 24 hours; and a minor stroke, as a similarly focal neurological deficit persisting more than 24 hours. A lacunar stroke, transient or permanent, was defined as the presence of unilateral motor or sensory symptoms without cortical deficits {25J Information about risk factors for stroke (hypertension, diabetes mellitus, smoking), previous cerebral ischemic events, atherosclerotic heart disease, and peripheral vascular disease was recorded. Baseline investigations included electrocardiography (EKG), chest radiograph, hematocrit, and serum glucose and creatinine levels. From the standard 12-lead EKG, we measured P-terminal force for left atrial enlargement, Casale index for left ventricle hypertrophy, Q wave as a measure for an old myocardial infarct, and ST segment depression as a measure for cardiac ischemia [26, 27). At randomization, C T scanning was mandatory as a baseline measure and also in order to exclude causes other than ischemia, except in case of transient monocular blindness. The present study comprised the 3,017 patients in whom a CT scan of the brain was performed. All C T scans were made within 1 month after the qualifying event, and were reviewed soon after randomization, by at least two neurologists, independently from the clinical features, and before the occurrence of any outcome events [28}. Cerebral infarcts were defined as intraparenchymal hypodense lesions, and were subdivided into lacunar infarcts (small deep infarcts), cortical infarcts (superficial, involving the cortex), and border zone infarcts (between the territory of two major cerebral arteries, or between the deep and superficial branches of the middle cerebral artery). According to the clincial signs and symptoms, the relevance of the cerebral infarcts on the C T scan was assessed. HWM on C T was distinguished from infarction by its illdefined boundaries (Fig 1). The severity of HWM was expressed separately for the frontal and parietooccipital areas of the white matter separately in one of two degrees: In grade 1 the hypodensity was restricted to the region adjoining the ventricles; in grade 2 the hypodensity extended as far as the cortex. The reliability of this simple scale was confirmed in a separate interobserver study [291. HWM was rated as mild for grade 1 found in either the frontal or the parietooccipital
178 Annals of Neurology
Vol 32 No 2 August 1992
Fig I . CT scan of a 69-year-old man presenting with a right hemiparesis shows dzfluse hypodensity of the frontal and occz$ital white matter, and two (ldt and ri&, arrowheads) lacmar infarcts in the basal ganglia. area of the white matter, and as moderate to severe if there were abnormalities in both regions. After randomization, patients were seen every 4 months until their last outpatient visit, between March 1, 1990, and July 1, 1990. Patients unable or unwilling to continue visiting the hospital were followed up through their general practitioners. Outcome events were distinguished for the purpose of the present study into fatal or nonfatal stroke, and fatal or nonfatal cardiac event (death from documented heart disease, sudden death, and nonfatal myocardial infarction). The diagnosis of stroke included relevant clinical features persisting more than 24 hours and were correlated with a fresh infarct or hemorrhage on a repeat C T scan. If sudden and focal neurological deficits without C T changes (or if no C T scan was available) caused an increase in disability of at least one grade on the modified Rankin scale C30, 3 11, the event was also classified as a stroke. Strokes and cardiac events were independently classified by at least three members of the End Point Committee C223. O n the basis of the clinical features and the C T scan, two investigators (J. C. v. S., L. J. K.) classified the strokes into supratentorial lacunar syndromes [25), supratentorial cortical syndromes in the territories of the anterior and middle cerebral arteries (asphasia, hemianopia, neglect, motor deficits in only one limb), and vertebrobasilar syndromes (at least two of the following features: diplopia, ataxia, dysarthria, or difficulty in swallowing, or hemianopia without other cortical sgns).
Data Analysis The major aim of the data analysis was to study the relationship between hypodensity and recurrent stroke by comparing
the incidence of stroke between patients with and those without HWM. A Kaplan-Meier curve was used for graphic comparison [32]. The occurrence of strokes was compared in terms of hazard ratio (HR), which may be interpreted as a relative risk; that is, the risk of stroke in the patients with HWM divided by the risk of stroke in those with normal white matter. Two issues may disturb this relationship. Firstly, there may be an uneven distribution of other risk factors for stroke between patients with and those without HWM (“confounding”). Hence, the ratios, obtained by use of the Cox proportional hazards model, were adjusted for baseline incomparabilities in all relevant variables [33].The precision of the hazard ratio estimates was described by means of 95% confidence intervals (CIS) obtained from the Cox model. Secondly, the strength of the relationship of HWM and stroke may differ between subgroups (“effect modification”). Statistical difference between hazard ratios of these subgroups was determined by use of the coefficient and its standard deviation of the interaction term of HWM and the subgroup factor in the Cox model. In addition, chisquare tests were used when appropriate.
Resuits CT showed diffuse HWM in 337 (11%) of 3,017 patients with a TIA or nondisabling stroke. The HWM
was mild in 133 patients ( 4 % ) and moderate to severe in 204 patients (7%). The HWM was the only abnormality in 180 patients, while in the remaining 157 patients there was also evidence of infarction. The patients with HWM differed from the patients with normal white matter in other aspects (Table 1); they had higher age (71.4 f 7.4 versus 64.4 f 9.9 years) and a higher prevalence of hypertension (50 versus 41%). The qualifying event included motor symptoms in 81% of patients with HWM, aphasia in 2196, hemianopia in 596, and dysarthria in 26%; the occurrence of these neurological deficits was similarly distributed for patients with normal white matter. Patients with HWM also had a higher proportion of minor stroke and lacunar syndrome as a qualifying event (81 versus 68% and 64 versus 57%) and a higher proportion of lacunar infarcts (40 versus 27%) and multiple infarcts (12 versus 7%, see Table 1). Finally, an increased Pterminal force (> 0.03 msec mV) on standard 12-lead EKG was significantly more common in the HWM group (54% versus 43%). Other risk factors for stroke and the occurrence of heart disease or peripheral vascular disease were evenly distributed between the two groups (see Table 1). Between patients with mild
Table 1. Baseline Characteristics in Patients with and Those without Hypodensity of the White Matter
Age (yr) Male gender Cardiovascular risk factors Hypertension Diabetes mellitus Smoking History of Myocardial infarction Angina pectoris Intermittent claudication Qualifying event TIA Minor stroke Lacunar stroke Infarct on CT > I infarct Type of infarct Iacunar Cortical Border zone Cerebellar Hematocrit >0.45 Standard 12-lead EKG Left ventricle hypertrophy ST segment depression Q wave P-terminal force >0.03
White Matter Hypodensity (n = 337)
Normal White Matter (n = 2,680)
71.4 5 7.4 189 (56%)
64.4 f 9.9 1,781 (67%)
In a prospective study of 3,017 patients with transient ischemic attack or minor ischemic stroke from the Dutch Transient Ischemic Attack Trial, the presence or absence of diffuse hypodensity of the white matter on a baseline computed tomography (CT) scan of the brain was related to the occurrence of subsequent stroke. On entry, 337 patients were judged to have diffuse hypodensity of the white matter on CT; they were older (71.4 +- 7.4 years versus 64.4 +- 9.9 years), more often had hypertension (50% versus 41%), and more often had lacunar infarcts on CT scan (40% versus 26%) than did patients with normal white matter. Strokes, fatal or nonfatal, occurred in 51 (15%) of the patients with diffuse hypodensity of the cerebral white matter, compared to 217 (8%) in the group with normal white matter (crude hazard ratio, 2.0; 95% confidence interval, 1.4-2.7). After adjustment for age and other relevant entry variables, the hazard ratio was 1.6 (95% confidence interval, 1.2-2.2). In patients younger than 7 0 years the crude hazard ratio was 2.7 (95% confidence interval, 1.7-4.2). The distribution between the main subtypes of stroke was similar for patients with and those without diffuse hypodensity of the cerebral white matter: Intracerebral hemorrhage occurred in 6 and 996, cortical infarction in 47 and 4595, and lacunar infarction in 34 and 2996, respectively. We conclude that hypodensity of the cerebral white matter in patients with transient ischemic attack or minor stroke is associated with an extra risk of future stroke, from large as well as from small vessels, and particularly in patients under 7 0 years old; this increase of risk is independent of other risk factors for stroke. van Swieten JC, Kappelle LJ, Algra A, van Latum JC, Koudstaal PJ, van Gijn J. Hypodensity of the cerebral white matter in patients with transient ischemic attack or minor stroke: influence on the rate of subsequent stroke. Ann Neurol 1992;32:177-183
Diffuse hypodensity of the cerebral white matter (HWM) on computed tomography (CT) frequently is seen in elderly patients [1-3). Subcortical dementia with this type of white matter abnormality is known as Binswanger's disease. These patients usually have hypertension and often a history of one or more strokes [4-61. Moreover, macroscopic study of the brains of these demented patients at autopsy or by CT usually shows lacunar infarcts in the deep white matter or basal gangha {5-91. This type of white matter disease has been attributed to small-vessel disease, as arteriolosclerosis is always present [lo]. A similar but less marked HWM on CT has also been found in about 10% of patients with cerebral ischemia without dementia [ll, 12). The pathological findings in these patients resemble those of Binswanger's disease 12, 9, 13, 141.
It is a reasonable but unproved assumption that those patients with a transient ischemic attack (TIA) or a nondisabling ischemic stroke who also show HWM on CT have more generalized vascular disease and are more at extra risk for stroke recurrence. Until now, only advanced age, hypertension, the severity of the first stroke, and a history of myocardial infarction have been identified as risk factors for the recurrence of stroke in patients with cerebral ischemia [15--181, whereas antihypertensive treatment, antiplatelet treatment, and carotid surgery in appropriate patients have been found to reduce the risk of stroke [19-21) The aim of the present study was to investigate the relation between the presence of HWM on the baseline CT and the incidence of subsequent stroke in a large and prospective series of patients with TIA or nondisabling stroke, who participated in a clinical trial
From the *Department of Neurology, University Hospital Utrecht, Utrecht, and the tDepartment of Neurology, University Hospital Rotterdam, Rotterdam, The Netherlands.
Received Oct 3, 1991, and in revised form Jan 13, 1992. Accepted for publication Feb 23, 1992. Address correspondence to Dr van Swieten, Department of Neurology, University Hospital Utrecht, Heidelberglaan 100, 3584 CX Utrecht. The Netherlands.
Copyright 0 1992 by the American Neurological Association
177
of antiplatelet agents. Furthermore, we addressed the question of whether the type of stroke on follow-up differs between patients with and those without HWM.
Patients and Methods Between March 1986 and March 1989, 3,150 patients with a TIA or nondisabling ischemic stroke (Rankin grade 3 or less) were randomized in the Dutch TIA Trial [22, 231. In this trial the preventive effects of 30 mg of acetylsalicylic acid were compared with those of 283 mg, in a double-blind design, and 50 mg of atenolol versus placebo was similarly tested. Exclusion criteria were identifiable causes of ischemia other than arterial thrombosis or thromboembolism, an interval of more than 3 months since the last event, and an indication or contraindication for the trial drugs. At randomization, the history of each patient was recorded on a checklist worded in ordinary language, a method reasonably consistent between observers 1241. A TIA was defined as a focal neurological deficit with abrupt or relatively rapid onset, persisting less than 24 hours; and a minor stroke, as a similarly focal neurological deficit persisting more than 24 hours. A lacunar stroke, transient or permanent, was defined as the presence of unilateral motor or sensory symptoms without cortical deficits {25J Information about risk factors for stroke (hypertension, diabetes mellitus, smoking), previous cerebral ischemic events, atherosclerotic heart disease, and peripheral vascular disease was recorded. Baseline investigations included electrocardiography (EKG), chest radiograph, hematocrit, and serum glucose and creatinine levels. From the standard 12-lead EKG, we measured P-terminal force for left atrial enlargement, Casale index for left ventricle hypertrophy, Q wave as a measure for an old myocardial infarct, and ST segment depression as a measure for cardiac ischemia [26, 27). At randomization, C T scanning was mandatory as a baseline measure and also in order to exclude causes other than ischemia, except in case of transient monocular blindness. The present study comprised the 3,017 patients in whom a CT scan of the brain was performed. All C T scans were made within 1 month after the qualifying event, and were reviewed soon after randomization, by at least two neurologists, independently from the clinical features, and before the occurrence of any outcome events [28}. Cerebral infarcts were defined as intraparenchymal hypodense lesions, and were subdivided into lacunar infarcts (small deep infarcts), cortical infarcts (superficial, involving the cortex), and border zone infarcts (between the territory of two major cerebral arteries, or between the deep and superficial branches of the middle cerebral artery). According to the clincial signs and symptoms, the relevance of the cerebral infarcts on the C T scan was assessed. HWM on C T was distinguished from infarction by its illdefined boundaries (Fig 1). The severity of HWM was expressed separately for the frontal and parietooccipital areas of the white matter separately in one of two degrees: In grade 1 the hypodensity was restricted to the region adjoining the ventricles; in grade 2 the hypodensity extended as far as the cortex. The reliability of this simple scale was confirmed in a separate interobserver study [291. HWM was rated as mild for grade 1 found in either the frontal or the parietooccipital
178 Annals of Neurology
Vol 32 No 2 August 1992
Fig I . CT scan of a 69-year-old man presenting with a right hemiparesis shows dzfluse hypodensity of the frontal and occz$ital white matter, and two (ldt and ri&, arrowheads) lacmar infarcts in the basal ganglia. area of the white matter, and as moderate to severe if there were abnormalities in both regions. After randomization, patients were seen every 4 months until their last outpatient visit, between March 1, 1990, and July 1, 1990. Patients unable or unwilling to continue visiting the hospital were followed up through their general practitioners. Outcome events were distinguished for the purpose of the present study into fatal or nonfatal stroke, and fatal or nonfatal cardiac event (death from documented heart disease, sudden death, and nonfatal myocardial infarction). The diagnosis of stroke included relevant clinical features persisting more than 24 hours and were correlated with a fresh infarct or hemorrhage on a repeat C T scan. If sudden and focal neurological deficits without C T changes (or if no C T scan was available) caused an increase in disability of at least one grade on the modified Rankin scale C30, 3 11, the event was also classified as a stroke. Strokes and cardiac events were independently classified by at least three members of the End Point Committee C223. O n the basis of the clinical features and the C T scan, two investigators (J. C. v. S., L. J. K.) classified the strokes into supratentorial lacunar syndromes [25), supratentorial cortical syndromes in the territories of the anterior and middle cerebral arteries (asphasia, hemianopia, neglect, motor deficits in only one limb), and vertebrobasilar syndromes (at least two of the following features: diplopia, ataxia, dysarthria, or difficulty in swallowing, or hemianopia without other cortical sgns).
Data Analysis The major aim of the data analysis was to study the relationship between hypodensity and recurrent stroke by comparing
the incidence of stroke between patients with and those without HWM. A Kaplan-Meier curve was used for graphic comparison [32]. The occurrence of strokes was compared in terms of hazard ratio (HR), which may be interpreted as a relative risk; that is, the risk of stroke in the patients with HWM divided by the risk of stroke in those with normal white matter. Two issues may disturb this relationship. Firstly, there may be an uneven distribution of other risk factors for stroke between patients with and those without HWM (“confounding”). Hence, the ratios, obtained by use of the Cox proportional hazards model, were adjusted for baseline incomparabilities in all relevant variables [33].The precision of the hazard ratio estimates was described by means of 95% confidence intervals (CIS) obtained from the Cox model. Secondly, the strength of the relationship of HWM and stroke may differ between subgroups (“effect modification”). Statistical difference between hazard ratios of these subgroups was determined by use of the coefficient and its standard deviation of the interaction term of HWM and the subgroup factor in the Cox model. In addition, chisquare tests were used when appropriate.
Resuits CT showed diffuse HWM in 337 (11%) of 3,017 patients with a TIA or nondisabling stroke. The HWM
was mild in 133 patients ( 4 % ) and moderate to severe in 204 patients (7%). The HWM was the only abnormality in 180 patients, while in the remaining 157 patients there was also evidence of infarction. The patients with HWM differed from the patients with normal white matter in other aspects (Table 1); they had higher age (71.4 f 7.4 versus 64.4 f 9.9 years) and a higher prevalence of hypertension (50 versus 41%). The qualifying event included motor symptoms in 81% of patients with HWM, aphasia in 2196, hemianopia in 596, and dysarthria in 26%; the occurrence of these neurological deficits was similarly distributed for patients with normal white matter. Patients with HWM also had a higher proportion of minor stroke and lacunar syndrome as a qualifying event (81 versus 68% and 64 versus 57%) and a higher proportion of lacunar infarcts (40 versus 27%) and multiple infarcts (12 versus 7%, see Table 1). Finally, an increased Pterminal force (> 0.03 msec mV) on standard 12-lead EKG was significantly more common in the HWM group (54% versus 43%). Other risk factors for stroke and the occurrence of heart disease or peripheral vascular disease were evenly distributed between the two groups (see Table 1). Between patients with mild
Table 1. Baseline Characteristics in Patients with and Those without Hypodensity of the White Matter
Age (yr) Male gender Cardiovascular risk factors Hypertension Diabetes mellitus Smoking History of Myocardial infarction Angina pectoris Intermittent claudication Qualifying event TIA Minor stroke Lacunar stroke Infarct on CT > I infarct Type of infarct Iacunar Cortical Border zone Cerebellar Hematocrit >0.45 Standard 12-lead EKG Left ventricle hypertrophy ST segment depression Q wave P-terminal force >0.03
White Matter Hypodensity (n = 337)
Normal White Matter (n = 2,680)
71.4 5 7.4 189 (56%)
64.4 f 9.9 1,781 (67%)
