Haan MN, Shemanski L, Jagust WJ, Manolio TA, Kuller L. The Role of APOE ∊4 in Modulating Effects of Other Risk Factors for Cognitive Decline in Elderly Persons. JAMA. 1999;282(1):40–46. doi:10.1001/jama.282.1.40
Author Affiliations: Departments of Epidemiology and Preventive Medicine (Dr Haan) and Neurology (Dr Jagust), University of California, School of Medicine, Davis; Department of Biostatistics, School of Public Health, University of Washington, Seattle (Dr Shemanski); Departments of Epidemiology and Biometry, National Heart, Lung, and Blood Institute, Bethesda, Md (Dr Manolio); and Department of Epidemiology, School of Public Health, University of Pittsburgh, Pittsburgh, Pa (Dr Kuller).
Context Cognitive decline in elderly persons is often an early predictor of
dementia. Subclinical cardiovascular disease (CVD) and diabetes mellitus may
contribute to substantial decline in cognitive function in the elderly. These
risks may be modified by gene-environment interactions between apolipoprotein
E (APOE) genotype and CVD risk factors or subclinical
Objectives To examine the association between subclinical CVD and decline in cognitive
functioning in the elderly and to examine effect modification by the APOE genotype of the association between subclinical disease
and cognitive decline.
Design The Cardiovascular Health Study, a population-based, prospective cohort
Setting and Population A total of 5888 randomly selected Medicare-eligible participants from
Sacramento County, California; Forsyth County, North Carolina; Washington
County, Maryland; and Pittsburgh, Pa, aged 65 years or older, who were recruited
in 1989-1990 (n = 5201) and in 1992-1993 (n = 687) and who were followed up
for 7 and 5 years, respectively.
Main Outcome Measures Change over time in scores on the Modified Mini-Mental State Examination
and the Digit Symbol Substitution Test as a function of APOE genotype, subclinical CVD, and diabetes mellitus.
Results Seventy percent of participants had no significant decline on the Modified
Mini-Mental State Examination. Systolic blood pressure, the ankle-arm brachial
index, atherosclerosis of the internal carotid artery, diabetes mellitus,
and several diagnoses of prevalent CVD were significantly associated with
declines in scores on the Modified Mini-Mental State Examination and the Digit
Symbol Substitution Test. The rate of cognitive decline associated with peripheral
vascular disease, atherosclerosis of the common and internal carotid arteries,
or diabetes mellitus was increased by the presence of any APOE ∊4 allele.
Conclusions Most healthy elderly people did not experience cognitive decline. Measures
of subclinical CVD were modest predictors of cognitive decline. Those with
any APOE ∊4 allele in combination with atherosclerosis,
peripheral vascular disease, or diabetes mellitus were at substantially higher
risk of cognitive decline than those without the APOE ∊4
allele or subclinical CVD. High levels of atherosclerosis increased cognitive
decline independently of APOE genotype.
The influence of cardiovascular diseases (CVDs) and risk factors on
cognitive decline in the elderly is of importance in the search for predictors
and pathways for prevention of dementia. Previous work has linked cognitive
decline and dementia to CVD,1 stroke,2 diabetes mellitus (DM) and glucose metabolism,2- 11
hypertension and high and low blood pressure,2,12- 17
peripheral vascular disease,18 atherosclerosis,19 genetic factors,20,21
and cerebral blood flow.22,23
Apolipoprotein E ∊4 (APOE ∊4) genotype
is associated with an increased risk of dementia and cognitive decline.24 Apolipoprotein E ∊4 alleles may account for
13% to 20% of dementia cases.25,26
Although APOE genotype is not a specific disease
locus, it may create host susceptibility that affects the rate of disease
progression.27 Gene-environment interactions
are important in the progression of cognitive impairment and the development
of dementia.28 Recent work by the Rotterdam
Study1 has suggested that DM, peripheral vascular
disease, and atherosclerosis are associated with an increased risk of vascular
dementia and Alzheimer disease (AD). Similar findings for a synergistic role
played by the APOE ∊4 genotype in the presence
of atherosclerotic diseases have been reported by Kuusisto et al.29
A previous Cardiovascular Health Study (CHS) article30
reported an association between cognitive decline measured using the Modified
Mini-Mental State Examination (3MSE) of at least 5 points from year 5 (when
magnetic resonance imaging was performed) through year 7 and several magnetic
resonance imaging measures (infarct on magnetic resonance imaging, high ventricular
volume, increased sulci width, and high white matter grade). Heckbert et al31 also found in CHS that a lower 3MSE score was associated
with white matter changes.
The purpose of this study was to examine the influence of subclinical
CVD and DM on 7-year changes in the 3MSE and the Digit Symbol Substitution
test (DSS) results and to examine effect modification by APOE ∊4 of the associations among cognitive decline and atherosclerosis,
peripheral vascular disease, or DM.
The CHS is a cohort study of 5888 randomly selected persons aged 65
years or older at baseline. Descriptions of the CHS study design and recruitment
of the CHS cohort have been published elsewhere.32- 35
The original sample of the CHS was composed of 5201 elderly adults who were
recruited from a defined sample of the Medicare files in 4 communities in
the United States: Forsyth County, North Carolina; Sacramento County, California;
Washington County, Maryland; and Pittsburgh, Pa. Eligible participants were
not institutionalized and were able to give informed consent. Recruitment
occurred between June 1989 and May 1990 (year 2). The sample at that time
was 57% women and 95% white. In 1992-1993 (year 5), 687 blacks were recruited
from Forsyth County, Sacramento, and Pittsburgh by similar methods as the
original recruitment, so 15% of the total sample were blacks. Out of the 5888
total study participants, only those who attended the first annual visit for
the original cohort (year 3) and attended baseline (year 5) for the new cohort
are included in these analyses (Table 1).
Each person took the 3MSE and the DSS at least twice (2 years) during
the follow-up period.
The 3MSE is scored from 0 to 100.36 It
was administered in all annual examinations, starting with the second annual
The DSS, a subtest of the Wechsler Adult Intelligence Scales, measures
the ability to link symbols with numbers in a timed test (90 seconds).37 Scores include the total test items completed, the
number correctly coded, and the number incorrectly coded. The score ranges
from 0 to 90 and is the total number done correctly. A higher score indicates
better cognitive status. It was administered in all annual examinations.
Methods of measurement for APOE alleles have
been published elsewhere.30APOE alleles were assessed for 4607 whites (93%), 850 blacks (93%),
and 37 others (82%). At the baseline for both cohorts, prevalence and extent
of clinical CVD were assessed by confirmed history of myocardial infarction,
angina, congestive heart failure (CHF), atrial fibrillation, coronary artery
bypass graft, the use of a pacemaker, stroke or transient ischemic attack,
carotid endarterectomy, intermittent claudication, or peripheral vascular
surgery. Subclinical CVD measures included an ankle-arm blood pressure index
of less than 0.9 mm Hg, internal or common carotid arterial wall thickness
(measured by ultrasonography), major abnormalities on electrocardiogram (ECG),38 and atrial fibrillation (ECG). The details of incident
events surveillance have been published elsewhere.39
Incident stroke and DM were both defined by self-report during biannual follow-up
visit, which was confirmed by review of medical records or a physician questionnaire.
Fasting plasma glucose and insulin levels were measured at each annual clinic
visit and analyzed at the CHS central laboratory. Two-hour glucose tolerance
was measured at the first annual follow-up visit for the original cohort and
the baseline visit for the new cohort.
The associations between cardiovascular risk factors measured at baseline
and cognitive scores were examined using longitudinal statistical methods.
Generalized estimation equations (GEEs), as previously described,40 were used to examine associations of cognitive scores
and risk factors, taking into account the lack of independence in a participant's
observations across time. In these analyses, the correlation structure between
a participant's cognitive scores at different visits was assumed to have compound
symmetry, meaning the correlation was assumed constant between any 2 cognitive
scores for a participant. This correlation structure was most representative
of the data compared with other investigated correlation structures, such
as autoregression. The basic GEE model included the covariates of age, sex,
education, and race, all at baseline, and incident stroke during follow-up.
Stroke was included as a covariate because it is a very strong known predictor
of cognitive decline during follow-up.2 To
minimize the bias resulting from participants with the greatest decline being
lost to follow-up, an indicator variable reflecting missing cognitive score
data at the last analyzed visit was also included in each model. Baseline
stroke, systolic and diastolic blood pressure, and DM were also added to the
model. The effect modification analyses between APOE ∊4
alleles and atherosclerosis, peripheral vascular disease, and DM were included
because previous reports of such analyses1,2,6,9
suggested that APOE ∊4 might modify the association
between these measures and cognitive status or dementia.
Tests of the associations between cognitive scores and risk factors
were performed using 2-way interactions of the risk factor with follow-up
years and the dependent variable was the cognitive score at each year. Modification
of the association between a prespecified risk factor and cognitive change
was examined by including a 3-way interaction term to the model, which was
equal to the product of follow-up year and the 2 risk factors of interest.
All analyses were performed using SAS statistical software,41
with P<.001 considered significant. The new cohort
of blacks was not analyzed separately because race differences were not the
focus of the study, nor was there adequate statistical power. Their data and
scores were added to years 3 through 7 in the analysis.
Table 2 shows the mean 3MSE
and DSS scores by age and sex groups. separately by race. Table 3 shows the mean scores by years of education and for presence
of baseline stroke. Women had higher 3MSE and DSS scores than men, and whites
had higher scores than blacks. Scores were lower at older ages, and both scores
were higher for those with more education. Those with prevalent stroke had
lower scores on both tests.
Table 4 shows descriptive statistics by the 3MSE and the DSS scores.
The mean 3MSE score declined by 4 points, and the mean DSS score declined
by more than 6 points in 7 years. The annual median change in the 3MSE was
0, and −0.67 points for the DSS. The 3MSE score did not change at all
for 14% of the people. Assuming a change of at least 5 in either direction
due to random variation or learning effects and considered clinically unimportant,
70% of the people exhibited little decline or improvement. Cumulatively over
7 years, 11.9% had a decline of 10 points or more and 4.8% had an improvement
of 10 points or more.
Table 5 shows the results from a GEE linear regression model for
the association between cognitive change and DM and glycemic status, APOE genotypes, and several circulatory disease measures.
Clotting factors (fibrinogen, factors VII and VIII), low-density lipoprotein,
high-density lipoprotein, and triglyceride levels (all P values >.05) were not significantly associated with the 3MSE or the
DSS (data not shown). For continuous variables, subjects at or above 1 SD
higher than the mean were compared with those who were below 1 SD above the
mean. Mean (SD) values for the fasting plasma glucose level were 6.17 (2.1)
mmol/L (111  mg/dL); 2-hour oral glucose tolerance test, 8.2 (3.2) mmol/L
(147  mg/dL); systolic blood pressure, 136.58 (21.84) mm Hg; ankle-arm
blood pressure, 1.06 (0.17) mm Hg; common carotid arterial thickness, 1.06
(0.22) mm; and internal carotid arterial thickness, 1.44 (0.57) mm.
Self-reported history of DM, a definite (medically confirmed) diagnosis
of DM, fasting plasma glucose level, and 2-hour glucose tolerance were all
negatively associated with change in the DSS scores. An abnormal 2-hour oral
glucose tolerance result was associated with decline in both the DSS and the
3MSE scores. Persons who were 1 SD above the mean on the oral glucose tolerance
test declined 0.54 points more on the 3MSE than those below 1 SD above the
oral glucose tolerance test mean over the 7 years. Persons with confirmed
DM declined by 1.8 points more on the DSS in 7 years compared with those without
The frequency distribution of the APOE genotype
was as follows: 2/2 (n=32), 2/3 (n=642), 2/4 (n=124), 3/3 (n=2894), 3/4 (n=940), and 4/4 (n=79). Approximately 25% of the sample had at least 1 APOE ∊4 allele. Table 5
compares the change in the 3MSE score for those with any APOE ∊4 allele to those without any APOE ∊4
alleles. Those with at least one APOE ∊4 allele
declined nearly 3 points more in 7 years on the 3MSE compared with those without APOE ∊4 alleles. Those with APOE ∊4 alleles declined 2 points on the DSS vs 0.29 points in those
without APOE ∊4 alleles over 7 years.
Those with higher systolic blood pressure had greater declines in cognitive
functioning for both the 3MSE and the DSS. An increase of 1 SD (21.84 mm Hg)
above the mean was associated with a decrease of 0.96 points in the 3MSE over
7 years. Presence of low ankle-arm blood pressure (both the continuous ratio
and dichotomized ratio) was associated with greater declines in both measures
of cognition. A person with an ankle-arm blood pressure of less than 0.90
mm Hg experienced a 4.62-point decline on his or her 3MSE score over the 7
years and a 2.73-point decline in the DSS score vs −0.66 and −0.39,
respectively, for those with an ankle-arm blood pressure of at least 0.90
mm Hg. Atrial fibrillation was associated with a 4.4-point decline in the
3MSE scores over 7 years vs 0.63 for those without atrial fibrillation. Higher
maximum wall thickness in both the internal and common carotid arteries was
associated with greater declines in DSS and 3MSE scores. Among prevalent CVDs,
the greatest declines in the 3MSE scores occurred in those with prevalent
stroke, followed by CHF and major ECG abnormal results.
Figure 1 shows the ratios
of the mean wall thickness of the internal carotid and common carotid arteries
for each allele combination compared with those with 2/2 alleles (P<.001). Those with 4/4 alleles had maximum wall thickness that was 40% greater in the
internal carotid artery and 80% greater in the common carotid artery. Maximum
wall thickness for those with 3/4 alleles was 30%
greater in the internal and 60% greater in the common carotid artery, and
they were 20% greater in both arteries for those with 2/4 alleles.
We examined whether the presence of an APOE ∊4
allele increased the rate of cognitive decline associated with several measures
of subclinical atherosclerotic disease, peripheral vascular disease, or a
confirmed diagnosis of DM (Table 6).
Annual rates of change in the 3MSE are presented for combinations of (1) no
CVD and no APOE ∊4 allele, (2) either CVD or
any APOE ∊4 allele, and (3) CVD and any APOE ∊4 allele. Only measures for which P<.001 are reported.
Among those having any APOE ∊4 allele
and low ankle-arm blood pressure, the annual rate of decline in the 3MSE score
was 8.3 times greater than subjects with neither. The annual rate of decline
in the 3MSE scores in those with only APOE ∊4
allele or only a low ankle-arm blood pressure was 2.94 points and 3.66 points
greater than those with neither, respectively. We found that the average annual
decline in the 3MSE scores associated with a 1 SD increase above the mean
wall thickness of the internal carotid artery was nearly 3.9 times greater
in the presence of any APOE ∊4 when compared
with those with neither APOE ∊4 alleles nor
those at the mean.
Among those with DM, subjects with the APOE ∊4
allele had a 1.67-fold cognitive decline in their 3MSE scores compared with
those with neither DM nor the APOE ∊4 allele.
The presence of APOE ∊4 alleles in persons who
did not have DM actually contributed to a somewhat greater decline in their
3MSE scores than among subjects with DM (3.01-fold increase among those who
did not have DM). Those with neither DM nor the APOE ∊4
allele had the least cognitive decline in 3MSE scores.
Figure 2 illustrates the slope
of change for each of 7 study years in the 3MSE scores by presence of an APOE ∊4 allele for each quartile of wall thickness
of the internal carotid artery. Those with an APOE ∊4
allele who were also in the fourth quartile of internal carotid artery wall
thickness had the greatest decline in 3MSE scores. The cognitive change in
those in the top 3 quartiles of atherosclerosis was modest despite the presence
of an APOE ∊4 allele. The highest risk groups
for decline were both groups (those with an APOE ∊4
allele and those without) in the top quartile of atherosclerosis. The final
mean 3MSE score for those in the lowest risk group (first quartile of atherosclerosis
and without the APOE ∊4 allele) vs the mean
score for those in the highest risk group (fourth quartile of atherosclerosis
and with the APOE ∊4 allele) were, respectively,
76.4 and 68.4, an 8-point difference.
At least 2 other studies have reported effect modification of the association
between atherosclerosis and cognitive decline or dementia by the APOE ∊4 allele. A cross-sectional study9
has suggested that both peripheral vascular disease and atherosclerosis are
associated with a substantially increased risk of vascular dementia and AD.
Similar results have been reported by Kuusisto et al.29
Skoog et al42 found that only those with both
the APOE ∊4 allele and white matter disease
had an increased risk of AD or vascular dementia.
other research on the association between cognitive impairment and DM have
been conflicting.43,44 Neuropsychological
case-control studies have reported cognitive impairment in participants with
DM.45- 47 Although
there have been reports of significant relationships between cognitive impairment
and poor glucose management as measured by glycosylated hemoglobin,47- 50 other
studies have reported a lack of such relationships.51,52
Kuusisto et al29 and Skoog et al42
have both reported an increased risk of AD in those with type 1 or type 2
DM. Finch and Cohen53 have argued for a complex
but important etiologic role for DM in AD and other research has reported
associations between various measures of glycemic status, including hyperglycemia
and insulinemia, and cognitive impairment.54,55
Since we adjusted for the incidence of stroke in these analyses, part of the
mechanism by which DM may influence cognitive function was controlled for.
A recent study by Curb et al56 based on a single
measure of the 2-hour oral glucose tolerance tests taken 25 years before dementia
assessment and a brief history of DM diagnosis 15 years previous to testing
did find an association with vascular dementia but not AD.
research on the association between the APOE ∊4
allele and cognitive change57,58
has found that those without the APOE ∊4 allele
perform better than those with the APOE ∊4 allele
on verbal learning tests,59 visual attention,
psychomotor rapidity, and the 3MSE.60 However,
research by Small et al61 has suggested that
the APOE ∊4 allele may not influence cognitive
performance in adults without dementia, and loss of recent verbal memory may
be related to impending dementia, not to normal age-related changes in cognition.
This may be a reflection of differences in progression to disease or dementia
in those with the APOE ∊4 allele.
Although atherosclerosis of the common and internal carotid arteries was associated
with decline in both cognitive measures, lipids and clotting factors were
not. The absence of this association and the modest but consistent association
of most forms of circulatory disease, systolic blood pressure, and peripheral
vascular disease on cognitive decline suggest that mechanisms other than lipids
or coagulation may be important and should be further examined. Previous work
done in CHS supports the notion that such changes underlie the link between
circulatory disease and long-term cognitive decline.30,31
This is a randomly selected, population-based cohort study. During the
7-year follow-up period, 15.1% died, and 5.2% did not participate in the study
for other reasons. Attrition between baseline and the first annual examination
was negligible (<5%). Other studies of the elderly have shown relatively
little cognitive decline in the first few years of follow-up in cohort studies.
This may be due to the self-selection of healthier volunteers as study participants.
Thus, there is little impact on our estimates of using the 3MSE and DSS scores
measured at the first annual visit. Since our analytic approach allows the
inclusion of all subjects who contributed a cognitive test score for each
year they participated in the study, we have reduced, but not eliminated,
bias that might have arisen from a differential dropout of more cognitively
impaired subjects. We reduced this potential for bias further by including
an indicator variable reflecting missing data at the final follow-up visit.
Since most of the homebound CHS cohort subjects are assessed by our staff
in their homes and relatively few are admitted to nursing homes, we have assessed
most study participants who are cognitively impaired. Any selection would
probably tend to bias any association toward the null hypothesis. A learning
effect may be of concern for repeated administrations of the 3MSE and the
DSS. This would improve test scores from repeated exposure to the same test.
The CHS neither altered the 3MSE nor specifically changed words for recall.
Nevertheless, our results have shown a consistent decline in these measures
over time. If there is a learning effect, it is counterbalanced by real decline
in cognition and our results would be conservative from this standpoint.
The results of our study may have important implications for the origin
of vascular dementia and of AD. This adds to the evidence of a link between
atherosclerosis and dementia. The modifying role of the APOE ∊4 allele identifies a high-risk group for cognitive decline.
Since the rate of cognitive decline is also much higher among those who have
high levels of atherosclerosis but who do not have the APOE ∊4 allele, circulatory disease appears to have a role independent
of the APOE ∊4 allele. Based on these results,
prevention of atherosclerosis could reduce the risk of dementia.