Dementia in Stroke Survivors in the Stroke Data Bank Cohort

  Dementia in Stroke Survivors in the Stroke Data Bank Cohort.pdf (Size: 2.63 MB / Downloads: 0)
We determined the prevalence of dementia in 927 patients with acute ischemic stroke aged
years in the Stroke Data Bank cohort based on the examining neurologist's best judgment
Diagnostic agreement among examiners was 68% ( K = 0 3 4 ) . Of 726 testable patients, 116 (16%)
were demented. Prevalence of dementia was related to age but not to sex, race, handedness,
educational level, or employment status before the stroke. Previous stroke and previous
myocardial infarction were related to prevalence of dementia although hypertension, diabetes
mellitus, atrial fibrillation, and previous use of antithrombotic drugs were not Prevalence of
dementia was most frequent in patients with infarcts due to large-artery atherosclerosis and in
those with infarcts of unknown cause. Computed tomographic findings related to prevalence of
dementia included infarct number, infarct site, and cortical atrophy. Among 610 patients who
were not demented at stroke onset, we used methods of survival analysis to determine the
incidence of dementia occurring during the 2-year follow-up. Incidence of dementia was related
to age but not sex. Based on logistic regression analysis, the probability of new-onset dementia
at 1 year was 5.4% for a patient aged 60 years and 10.4% for a patient aged 90 years. With a
multivariate proportional hazards model, the most important predictors of incidence of
dementia were a previous stroke and the presence of cortical atrophy at stroke onset (Stroke
1990-^1:858-866)
Cerebrovascular disease is considered to be the
second most common cause of dementia;
20%-25% of cases of dementia are due to
stroke, and another 10%—15% are attributed to a
combination of vascular and Alzheimer's disease.1
From the Department of Neurology (T.K.T., J.P.M., J.R.H.),
Columbia-Presbyterian Medical Center, New York, New York;
the Biometry and Field Studies Branch (MAF.), National Institute
of Neurological Disorders and Stroke, Bethesda, Maryland;
the Department of Neurology, (D.B.H.), Michael Reese Hospital
and Medical Center, Chicago, Illinois; the Department of Neurology
(T.R.P.). University of Maryland School of Medicine, Baltimore,
Maryland; and the Department of Neurology (PA.W.),
Boston University School of Medicine, Boston, Massachusetts.
Presented in part at the 14th International Joint Conference on
Stroke and Cerebral Circulation, San Antonio, Texas, February
9-11, 1989.
Supported by the Stroke Data Bank under National Institute of
Neurological Disorders and Stroke contracts N01-NS-2-2302,
N01-NS-2-2384, N01-NS-2398, N01-NS-2-2399, and N01-
NS-6-2305 and in part by National Institutes of Health grant
R01-NS-2-6179-01.
Address for correspondence: Dr. Tatemichi, Stroke and Aging
Research Project, Neurological Institute, 710 West 168th Street,
New York, NY 10032-3784.
Received August 17, 1989; accepted February 23, 1990.
Although cerebral arteriosclerosis does not play a
causal role in Alzheimer's disease,2 there is still
controversy about the role of ischemic stroke in
producing global cognitive disorders classifiable as
dementia.3'4 The true risk of dementia caused by
ischemic stroke is incompletely understood because
most previous studies have focused on series of
selected or consecutively encountered patients with
dementia rather than assessing dementia in stroke
cohorts.
To address this question properly, patients with
stroke should be studied prospectively using neuropsychological
and functional examinations combined
with well-defined diagnostic criteria. In preparation
for such a study, we turned to the Stroke Data Bank5
as a source of preliminary data. Because the study of
dementia was not a major focus of the Stroke Data
Bank, the diagnosis of dementia was based only on
the best judgment of the examining neurologist. As a
result, the validity and reliability of this source of
information may be limited. Nonetheless, we believe
that an exploratory effort using this data source could
serve as a useful guide to future studies on this topic.
Downloaded from stroke.ahajournals.org by on May 18, 2010
Tatemichi et al Dementia in Stroke 859
Our aims were to determine the prevalence of
dementia after ischemic stroke in this cohort, to
identify potential risk factors (including stroke mechanism
and computed tomographic [CT] findings)
related to prevalence of dementia at stroke onset, to
determine the incidence of dementia in the remaining
dementia-free cohort followed longitudinally, and
to develop a predictor model for incidence of dementia
using information about potential risk factors.
Subjects and Methods
The Stroke Data Bank procedures have been
described.5 Neuropsychologjcal tests were not given,
and the diagnosis of dementia was based on the
clinical judgment of the examining neurologist. The
question was formulated as "Examiner believes
patient is demented," with the possible responses
being "yes," "no," and "untestable." Using this
approach, we probably identified only patients with
severe dementia. Patients with mild dementia require
more detailed testing for proper diagnosis. The most
likely cause of dementia, again based on the examiner's
judgment, was designated as Alzheimer's disease,
stroke, or other (including combined Alzheimer's
disease and stroke). The cause of dementia was
an opinion not subjected to further verification.
Examinations were planned at 0-7 days, 7-10 days,
and 3, 6, 12, and 24 months after stroke onset.
The design of the Stroke Data Bank included studies
of interexaminer reliability for elements of the neurologic
examination. Agreement on the dementia question
was 68% with a significant K of 0.34,6 reflecting
fair to good agreement.7 This level of interexaminer
variability was similar to that found for visual fields,
level of alertness, and degree of hemiparesis.
The study population consisted of 927 patients
^60 years of age with ischemic stroke. An elderly
population was selected to focus on patients at the
highest risk for dementia. Patients with parenchymatous
or subarachnoid hemorrhage were excluded since
they represent different pathophysiological groups.
The diagnosis of dementia was based on findings from
the 7-10 day examination where possible to permit the
hyperacute effects of stroke to subside, thus maximizing
the number of patients considered testable; when
patients were discharged before day 7, the 0-7 day
examination was used. For the incidence studies, we
used the examinations at each planned follow-up
whenever available up to 2 years.
Although the terms "prevalence" and "incidence"
properly refer to unselected populations, we defined
two cohorts using these terms to distinguish between
findings at stroke onset and those at follow-up. The
prevalence cohort was defined as patients whose
dementia status was testable during the first 10 days
after stroke onset. The incidence cohort was defined
as testable patients who were free of dementia at
admission and who were assessed neurologically during
follow-up.
We determined prevalence rates (as a percentage)
for each 5-year age group for the prevalence cohort.
Since many patients (21.7%) were untestable, we
examined the features that predicted testability using
a multiple logistic model8 to assess selection biases.
We next examined the relation between prevalence
of dementia and sociodemographic variables, vascular
risk factors, functional history, ischemic stroke
subtype, and CT findings. For functional history, we
assessed the patient's activities during the month
before the stroke including five domains from the
Barthel Index (eating, dressing, transferring, walking,
and continence). In this analysis we controlled for
previous stroke since this factor alone might account
for functional impairments. For ischemic stroke subtype,
we used a diagnostic algorithm9 that classified
patients as having infarction of unknown cause,
infarction with normal angiogram, infarction with
tandem arterial pathology, infarction due to cardiogenic
embolism, infarction from thrombosis due to
large-vessel atherosclerosis (atherothrombotic stroke),
or infarction due to lacunes. For CT findings, we
examined the first scan obtained in the first 10 days
after stroke onset showing a new, related lesion; if only
one scan was obtained, we used it for analysis whether
or not a lesion was present. CT was examined in 426 of
the 726 patients in the prevalence cohort. The methods
for evaluating and recording CT findings5 and
their interobserver reliabilities have been described.10
Although not a specified procedure in the Stroke Data
Bank, we believe that the CT findings were recorded
independently of the neurologic examination. Specific
CT findings of interest were the number of new lesions
related to the stroke, the number of old unrelated
lesions, the presence of cortical atrophy or hydreocephalus,
the volume of the infarct, and the presence
of periventricular lucencies. The primary cerebral site
of the infarction was defined by a combination of CT
findings and the clinical syndrome. For these crosssectional
comparisons (except site), we used univariate
tests of association, tests for trend with ordinal
data,11 and nonparametric tests as appropriate. Site
was not subjected to statistical analysis since the
cerebral sites examined represent a combination of
multiple sites.
We determined incidence rates using Kaplan-
Meier12 product-limit estimates of the time to first
evidence of dementia. We also used a logistic regression
model8 to estimate the probability of incidence
of dementia, using age as a prognostic indicator. In
this model we controlled for the effects of previous
stroke, cortical atrophy, and atherothrombotic
stroke, which were the variables significantly related
to prevalence of dementia. Cox's semiparametric
proportional hazards model13 provided an estimate
of the differences in time to incidence of dementia,
and a stepwise elimination procedure identified
potential predictors. Based on univariate analyses in
the prevalence cohort, we considered the factors age,
hypertension, previous myocardial infarction, atrial
fibrillation, diabetes mellitus, previous stroke, previous
use of antiplatelet or anticoagulant drugs,
atherothrombotic and cardioembolic stroke subtypes,
Downloaded from stroke.ahajournals.org by on May 18, 2010
860 Stroke Vol 21, No 6, June 1990
TABLE 1. TesUbllity and Prevalence of Dementia by Age at Stroke Onset In Stroke DaU Bank Cohort a 60 Years Old
With Iscbemic Stroke
Age
(yr)
60-64
65-69
70-74
75-79
80-84
>85
Total
N
158
199
191
155
132
92
927
No.
130
163
163
118
89
63
726
Testable
%
82.3
81.9
85.3
76.1
67.4
68.5
78.3
No.
117
145
139
93
67
49
610
Demented
Yes
No.
13
18
24
25
22
14
116
%
10.0
11.0
14.7
21.2
24.7
22.2
16
p<0.001 testability related to age (^=23.43, df=5).
p<0.001 demented related to age 0^=15.89, df=5).
and the CT findings of number of new related lesions,
number of old unrelated lesions, cortical atrophy,
hydrocephalus, and infarct volume.
Results
Among the study population of 927 patients with
ischemic stroke aged ^60 years, 726 (78.3%) were
testable and comprised the prevalence cohort. Testability
was related to age (Table 1) and to several
clinical features that impair communication, including
reduced alertness (occurring in 159 of 925
patients, 17.2%), aphasia (192 of 927 patients,
20.7%), and hemineglect (81 of 927 patients, 8.7%).
Among these variables, logistic regression analysis
showed that clinical features but not age were the
main determinants of testability. Although older
patients were underrepresented in the prevalence
cohort, patients were selected out by the presence of
these clinical features, which appeared to be more
common in the elderly. This selection bias is unavoid-
30
10
T
1
1 I
60-64 6S-M 70-74 75-79 80-84 80+
AGE INTERVAL (IN YEARS)
FIGURE 1. Bar graph. Mean±SEMprevalence of dementia
by age at stroke onset among 726 testable patients ^60 years
old with ischemic stroke in Stroke Data Bank.
able in any clinical study of dementia in stroke
patients.
Among the prevalence cohort of 726 testable
patients, 116 (16%) were demented. Prevalence of
dementia was significantly (/><0.001) related to age
(Figure 1). The largest increase in frequency
occurred between the age groups 70-74 and 75-79
years. No difference in prevalence of dementia
occurred between men and women. For 95 of the 116
demented patients, the examining neurologist estimated
the probable cause of dementia to be stroke in
37 (38.9%), Alzheimer's disease in 34 (35.8%), and
other (including mixed dementia) in 24 (25.3%).
The sociodemographic variables sex, race, handedness
(in 685 patients), educational level (in 665), and
employment status prior to the stroke (in 713) were
not significantly associated with prevalence of
dementia (Table 2). However, a trend was apparent
between educational level and dementia; patients
with <8 years of education were more frequently
demented than those with some college education.
Among the vascular risk factors, both a history of
stroke (p<0.001) and a history of myocardial infarction
(p =0.021) were significantly related to prevalence
of dementia (Table 3); hypertension, diabetes
mellitus, atrial fibrillation, and previous antiplatelet
or anticoagulant therapy showed no significant association
with prevalence of dementia.
Functional history was related to prevalence of
dementia, even after controlling for the effects of
previous stroke (Table 4). In each domain of function,
impairment was more frequent among demented
patients than among nondemented ones; moreover,
demented patients with a history of stroke showed the
most impairment. Previous stroke alone was not significantly
related to functional history.
Prevalence of dementia varied by ischemic stroke
subtype (Table 5). In the entire prevalence cohort
dementia was most frequent among patients with
atherothrombotic infarction and those with infarction
of unknown cause and was least frequent in
those with strokes due to tandem arterial disease or
lacunes. Since 18 of the 51 patients (35%) with
atherothrombotic infarction had suffered a previous