Abbott RD, White LR, Ross GW, Masaki KH, Curb JD, Petrovitch H. Walking and Dementia in Physically Capable Elderly Men. JAMA. 2004;292(12):1447-1453. doi:10.1001/jama.292.12.1447
Author Affiliations: Division of Biostatistics and Epidemiology, University of Virginia School of Medicine, Charlottesville (Dr Abbott); Pacific Health Research Institute (Drs Abbott, White, Ross, Masaki, Curb, and Petrovitch), Department of Veterans Affairs (Drs White, Ross, and Petrovitch), Honolulu-Asia Aging Study, Kuakini Medical Center (Drs Abbott, White, Ross, Masaki, Curb, and Petrovitch), and Departments of Geriatric Medicine and Medicine, John A. Burns School of Medicine, University of Hawaii (Drs White, Ross, Masaki, Curb, and Petrovitch), Honolulu.
Context Evidence suggests that physical activity may be related to the clinical
expression of dementia. Whether the association includes low-intensity activity
such as walking is not known.
Objective To examine the association between walking and future risk of dementia
in older men.
Design Prospective cohort study.
Setting and Participants Distance walked per day was assessed from 1991 to 1993 in 2257 physically
capable men aged 71 to 93 years in the Honolulu-Asia Aging Study. Follow-up
for incident dementia was based on neurological assessment at 2 repeat examinations
(1994-1996 and 1997-1999).
Main Outcome Measures Overall dementia, Alzheimer disease, and vascular dementia.
Results During the course of follow-up, 158 cases of dementia were identified
(15.6/1000 person-years). After adjusting for age, men who walked the least
(<0.25 mile/d) experienced a 1.8-fold excess risk of dementia compared
with those who walked more than 2 mile/d (17.8 vs 10.3/1000 person-years;
relative hazard [RH], 1.77; 95% confidence interval [CI], 1.04-3.01). Compared
with men who walked the most (>2 mile/d), an excess risk of dementia was also
observed in those who walked 0.25 to 1 mile/d (17.6 vs 10.3/1000 person-years;
RH, 1.71; 95% CI, 1.02-2.86). These associations persisted after accounting
for other factors, including the possibility that limited amounts of walking
could be the result of a decline in physical function due to preclinical dementia.
Conclusions Findings suggest that walking is associated with a reduced risk of dementia.
Promoting active lifestyles in physically capable men could help late-life
Physical and environmental factors associated with the risk of dementia
remain largely undefined. Although equivocal, evidence suggests that physical
activity may have a relationship with the clinical expression of dementia.1- 7 Whether
the association includes low-intensity activity such as regular walking is
not known. One study showed that a composite measure of physical activity,
partially based on walking histories, is associated with a reduced risk of
dementia.1 In a large cohort of women, those
who walked more had significantly smaller declines in a modified Mini-Mental
State Examination score over a 6- to 8-year period of follow-up.2 Others
describe relations that are weak while also providing contrasting evidence
for an important and protective effect of cognitive activities on risk of
dementia.8- 10 While
cognitive activity appears to be a predictor of future dementia, overlap with
physical activity may make it difficult to detect a relation between walking
The purpose of this report is to examine the association between walking
and future risk of dementia. Findings are based on a sample of physically
capable elderly men enrolled in the Honolulu-Asia Aging Study. To account
for the possibility that limited amounts of walking in late life could be
the result of a decline in physical function due to preclinical dementia,
adjustments were made for genetic susceptibility to dementia, cognitive performance,
and declines in physical activity since mid adulthood.
From 1965 to 1968, the Honolulu Heart Program began following up 8006
men of Japanese ancestry living on the island of Oahu, Hawaii, for development
of cardiovascular disease.11 Beginning with
examinations given from 1991 to 1993, the Honolulu-Asia Aging Study was launched
as an expansion of the Honolulu Heart Program for the study of neurodegenerative
diseases and cognitive function in elderly persons.12 Participants
included 3734 men aged 71 to 93 years (approximately 80% of the survivors
in the original Honolulu Heart Program cohort). Findings for this report are
based on follow-up for incident dementia based on neurological assessment
at 2 repeat examinations (1994-1996 and 1997-1999). Procedures were in accordance
with institutional guidelines and approved by an institutional review committee.
Written informed consent was obtained from the study participants.
Men who died (n = 377) before the scheduling of the second cycle of
cognitive assessments (1994-1996) were excluded from follow-up, as were an
additional 145 cases of prevalent dementia. Men with poor cognitive function
(n = 75) whose dementia status could not be confirmed were also excluded.
To reduce the confounding effects of Parkinson disease and stroke on the capacity
to walk, 39 men with prevalent Parkinson disease and 116 with prevalent stroke
were excluded. Among the remaining sample, 194 men had missing data on physical
To help isolate the association of walking from that of work-related
activities, 143 men were excluded because of continued employment. To reduce
confounding due to disability on the relation between walking and risk of
dementia, only men who were physically capable were considered for follow-up.
Men were considered to be physically capable if they presented for a baseline
clinical examination at the Kuakini Medical Center and reported undertaking
slight or moderate activities in a typical 24-hour period. Here, based on
the Physical Activity Index,13,14 "walking
on level ground" and "gardening or carpentry" were used as references to help
define slight and moderate activities, respectively. There were 124 men who
failed to present for a clinic visit (122 received home visits) and 76 whose
daily activities failed to meet the criteria for being slight or moderate.
Because cigarette smoking reduces the health benefits of being physically
active,14,15 an additional 161
smokers were removed. After these exclusions, there remained 1 man who used
a walker and 26 who used a cane. These men were also excluded from follow-up.
The final sample for this report includes 2257 men.
Cases of dementia were identified through a system of screening for
cognitive function following a rigid study protocol.12 Initial
screening considered a participant's age and cognitive performance on the
Cognitive Abilities Screening Instrument (CASI). The latter is a comprehensive
measure of intellectual function that has been developed and validated for
use in cross-cultural studies.16 Performance
scores range from 0 to 100, with high scores indicating better cognitive function
than low scores. Scores lower than 74 were selected a priori as an indicator
of possible dementia during dementia screening.12 The
value of 74 corresponds closely to a score of 22 on the Mini-Mental State
The CASI was administered twice at the baseline examination (1991-1993)
as part of 3 phases of screening.12 All men
with an initial CASI score of lower than 74 were invited to return for a second
phase of evaluation. At phase 2, if a repeat CASI score was also lower than
74 or if a score on the Informant Questionnaire on Cognitive Decline in the
Elderly18 was 3.6 or higher, a return visit
was requested for a complete dementia assessment. Among the remaining men,
recruitment for subsequent phases was based on a sampling scheme that increased
the likelihood for selection in older men and in men with intermediate vs
high CASI scores. Through this process of screening, 145 cases of prevalent
dementia were observed. As noted, these men were excluded from this study.
For these men with prevalent dementia, 96% had a CASI score lower than 74.
Among men without dementia, 11% had a CASI score lower than 74. As noted,
75 men with poor cognitive function (defined as an initial CASI score <74)
whose dementia status could not be confirmed were excluded from follow-up.
For the follow-up examinations used to identify incident cases of dementia
reported in this study, the CASI was administered once.19 For
the first follow-up examination (1994-1996), participants were recruited for
complete dementia assessment if 1 of the following occurred: the repeat CASI
score declined at least 9 points from the initial baseline CASI score; the
repeat CASI score was 77 or lower and the participant had less than 12 years
of education; or the repeat CASI score was 79 or lower and the participant
had 12 or more years of education. At the second follow-up examination (1997-1999),
complete assessment was requested when a CASI score was lower than 70. In
all instances, trained technicians administered the CASI without regard to
physical function and activity.
For diagnosis of dementia, information from a variety of sources was
considered, including a history given by a family member, a standardized neuropsychological
evaluation, and a thorough neurological examination. Laboratory findings and
computed tomography were also used for the classification of dementia. Final
diagnoses were assigned by a consensus panel consisting of a neurologist and
additional physicians with expertise in dementia in the absence of information
on physical activity. Participants with dementia met criteria based on the Diagnostic and Statistical Manual of Mental Disorders, Revised
Third Edition.20 Research criteria established
by the National Institute of Neurological and Communicative Disorders and
Stroke and the Alzheimer's Disease and Related Disorders Association were
used in the diagnosis of Alzheimer disease.21 Alzheimer
disease was defined to include cases of dementia in which Alzheimer disease
was judged to be the sole or primary cause. Diagnoses of vascular dementia
adhered to the criteria of the California Alzheimer's Disease Diagnostic and
Treatment Centers.22 Vascular dementia included
cases in which a vascular cause was considered the sole or primary cause.
At the beginning of follow-up (1991-1993), study participants were asked
about the average amount of distance walked per day. To account for the possibility
that a limited amount of walking in late life could be the result of a decline
in physical function due to preclinical dementia, adjustments were made for
genetic susceptibility to dementia (presence of 1 or 2 apolipoprotein ∊4
baseline measure of cognitive function (CASI assessed in 1991-1993), and declines
in physical activity since mid adulthood.
Presence of apolipoprotein ∊4 alleles was identified through genotyping
performed at Duke University, Durham, NC, following conventional methods.24 Physical activity was measured by the Physical Activity
Index, a common measure of daily metabolic output that is inversely related
to the risk of cardiovascular disease.13,14 Because
high levels of the Physical Activity Index are associated with more active
lifestyles than low levels, a decline in physical activity since mid adulthood
(as a continuous measure) was defined as the Physical Activity Index at the
time of study initiation (1965-1968) minus the index when follow-up began
for the current report (1991-1993).
Among the other characteristics, physical function was assessed based
on performance on a battery of tests, including ability to walk 10 ft (3 m)
with a normal gait, ability to walk on toes and heels, measures of balance,
ability to stand from a sitting position, and other factors related to physical
function. Based on a weighted average of these items, a "physical performance
score" was created.25 High scores indicate
better physical function than low scores. Additional characteristics included
age, years of education, body mass index (calculated as weight in kilograms
divided by the square of height in meters), childhood years spent living in
Japan, status as a skilled professional, hypertension, diabetes, prevalent
coronary heart disease, and total and high-density lipoprotein cholesterol
levels. For this analysis, a diagnosis of hypertension was made when systolic
or diastolic blood pressure was at least 160 or 95 mm Hg, respectively, or
when a study participant was receiving medication for treatment of hypertension.
Diabetes was defined on the basis of medical history or use of insulin or
oral hypoglycemic therapy. Prevalent coronary heart disease included myocardial
infarction, angina pectoris, and coronary insufficiency.26 A
timed walk was also performed at the baseline examination (1991-1993).
To describe the association between walking and dementia, estimates
of the age-adjusted incidence of dementia are provided across ranges of distance
walked based on standard analysis of covariance methods and logistic regression
models.27,28 Similar procedures
were used to describe the association between walking and the markers of preclinical
dementia and the other characteristics.
To estimate the relative hazard (RH) of dementia (and 95% confidence
interval [CI]) between ranges of distance walked, proportional hazards regression
models were used.29 Time to dementia was defined
as the time to diagnosis of dementia cases observed in the course of follow-up.
Men who died without a diagnosis of dementia prior to the end of follow-up
were censored at the time of death and those who remained alive were censored
at the close of the second repeat examination (1999). Adjustments were made
for age, the markers of preclinical dementia, and the other characteristics
as separate independent variables in a single regression model. Presence of
apolipoprotein ∊4 alleles (yes vs no), status as a skilled professional
(yes vs no), hypertension, diabetes, and prevalent coronary heart disease
were modeled as dichotomous variables and the other characteristics were modeled
as continuous variables. All reported P values were
based on 2-sided tests of significance and P values
<.05 were considered statistically significant. Statistical analyses were
carried out using SAS software, version 8.02 (SAS Institute Inc, Cary, NC).
Table 1 describes the association
between walking and the markers of preclinical dementia and the other characteristics.
On average, men who reported walking longer distances were younger than those
who reported walking less (P<.001). Among the
markers of preclinical dementia, walking was unrelated to presence of apolipoprotein
∊4 alleles. Men who walked the most had the highest CASI scores and the
lowest declines in physical activity since mid adulthood. The physical performance
score (P<.001) and years of education (P = .03) tended to be higher in men who walked the most, although associations
were modest. Possibly in response to the diagnosis of coronary heart disease,
prevalent coronary heart disease tended to be more common in men who walked
the longest distances. Prevalence of diabetes became progressively less frequent
as walking distances increased. For both coronary heart disease and diabetes,
however, associations with walking were not statistically significant. There
were no clear relations between walking and the remaining characteristics.
During the course of follow-up, 158 cases of dementia were identified
(15.6/1000 person-years). The mean time from baseline examination (1991-1993)
to diagnosis was 4.7 years (range, 2.4-7.4 years) with nearly 7 years of follow-up,
on average, for the study participants. The mean age at diagnosis was 84 years
(range, 75-98 years). Among the cases of dementia, 101 (10.0/1000 person-years)
were attributed to Alzheimer disease as the sole or primary cause and 30 (3.0/1000
person-years) were attributed to vascular dementia as the sole or primary
cause. For the remaining 27 cases (2.7/1000 person-years), Alzheimer disease
and vascular dementia were mixed as contributing causes with Parkinson disease
or atypical parkinsonism in 7 cases, Lewy bodies in 6 cases, and a variety
of other contributing factors in 12 cases. For 2 cases, a cause could not
The proportion of men who presented for the first repeat examination
(1994-1996) was 85% (1920/2257) and the proportion of survivors who presented
for the second examination (1997-1999) was 74% (1495/2025). Although efforts
are ongoing to determine dementia status, 199 men who were alive at the end
of follow-up had yet to receive a repeat examination. Compared with those
who were examined, those without a repeat examination were 1 year older and
had 1 less year of education on average (P<.001).
Those without a repeat examination were also more likely to have a lower CASI
score at baseline (85.0 vs 88.1 in those with an examination; P<.001). Based on age, education, CASI performance, and the other
study characteristics, we projected that 16 cases of dementia may have been
missed in this sample. Among 666 men who received a repeat examination and
met the CASI criteria for complete neurological evaluation, 76 (11%) had failed
to return for follow-up assessment. Among this group, we projected that 8
cases of dementia may have been missed.
Table 2 describes the incidence
of dementia according to ranges of distance walked. After adjusting for age,
men who walked the least (<0.25 mile/d [<0.40 km/d]) experienced a 1.8-fold
excess of total dementia compared with those who walked more than 2 mile/d
(>3.2 km/d) (17.8 vs 10.3/1000 person-years; RH, 1.77; 95% CI, 1.04-3.01).
Compared with men who walked the most (>2 mile/d), an excess of dementia was
also observed in those who walked 0.25 to 1 mile/d (17.6 vs 10.3/1000 person-years;
RH, 1.71; 95% CI, 1.02-2.86). The association between walking and dementia
also persisted in both those who did and those who did not have apolipoprotein
∊4 alleles. Although the number of men with apolipoprotein ∊4 alleles
was small (n = 365), the age-adjusted incidence of dementia in those who walked
less than 0.25 mile/d was 26.0 per 1000 person-years vs 16.1 per 1000 person-years
in men who walked more than 2 mile/d (RH, 1.63; 95% CI, 0.57-4.67). Among
those with no apolipoprotein ∊4 allele, incidence of dementia was 16.9
per 1000 person-years in those who walked the least vs 8.0 per 1000 person-years
in those who walked the most (RH, 2.16; 95% CI, 1.17-4.01).
There was a 1.8-fold excess of Alzheimer disease in men who walked 2
mile/d or less vs those who walked more than 2 mile/d (10.9 vs 6.0/1000 person-years;
RH, 1.81; 95% CI, 0.97-3.40). Although an association with other dementia
subtypes was not statistically significant, incidence of mixed and other dementia
declined from 3.3 to 1.1 per 1000 person-years as walking increased from less
than 0.25 to more than 2 mile/d. An association between walking and vascular
dementia was less apparent (Table 2).
To help determine whether the excess in dementia in those who walked
the least could be attributed to confounding by other factors, the relation
between walking and dementia was further adjusted for the characteristics
in Table 1. After adjustment (Table 3), a 1.9-fold excess risk of total
dementia occurred in men who walked less than 0.25 mile/d compared with men
who walked more than 2 mile/d (RH, 1.93; 95% CI, 1.11-3.34). Compared with
the most active men, those who walked 0.25 to 1 mile/d experienced a 1.7-fold
excess in dementia risk (RH, 1.75; 95% CI, 1.03-2.99). Risk of Alzheimer disease
was 2.2-fold higher in men who walked less than 0.25 mile/d vs those who walked
the most (RH, 2.21; 95% CI, 1.06-4.57).
The focus of this report is on day-to-day activity; however, a faster
timed walk at the baseline examination (1991-1993) was also associated with
a decreased age-adjusted incidence of dementia. For men who walked 10 ft (3
m) in more than 6 seconds, incidence was 20.2/1000 person-years compared with
13.1 per 1000 person-years for those with walking times of 3 seconds or less
(RH, 1.57; 95% CI, 0.77-3.21). Among men who were able to walk 10 ft in 4
seconds or less (n = 1682), incidence was 15.3 per 1000 person-years in those
who walked 1 mile/d or less vs 8.0 per 1000 person-years in those who walked
more than 2 mile/d (RH, 1.93; 95% CI, 1.04-3.57). In those who were slower
walkers (>4 seconds to walk 10 ft), an association was also present; risk
also appeared to be increased, although nonsignificantly. Based on a small
sample of 568 men, incidence of dementia declined from 26.1 to 19.8 per 1000
person-years as distance walked increased from less than 0.25 mile/d to more
than 2 mile/d (RH for those who walked <0.25 vs >2 mile/d, 1.31; 95% CI,
Our findings suggest that physically capable elderly men who walk more
regularly are less likely to develop dementia. The capacity to walk quickly
during a timed walk also appears to be associated with a reduced risk of dementia,
although these results should be confirmed. There is also a need to better
characterize walking behaviors and patterns that have relations with late-life
cognition, including metabolic equivalents that are easily adopted by elderly
individuals. Although this study did not enroll women, observations in women
of an association between walking and changes in cognitive function over time
suggest that the relationship between walking and dementia may apply to them
as well.2 Promoting active lifestyles may have
important effects on late-life cognitive function.
There are no clear explanations for the relation between walking and
dementia. Although associations were independent of other study characteristics
that were determined at the time when walking was assessed, it may be that
men who walk frequently are more resistant to risk factor changes or transitions
into adverse risk factor states. Although changes in risk factor status in
the course of follow-up were not considered in the current study (nor were
such data always available), it would be important to determine if men who
walk regularly are less prone to development of intervening conditions that
have a closer link with dementia.
It also is likely that the relationship of walking and dementia is modulated
by many factors, including environmental and lifestyle exposures that could
influence cognitive capacity. Walking and dementia may be related through
general effects on overall vitality and biological aging. Other mechanisms
may involve links between cardiovascular health and dementia,30 tracking
with preclinical dementia, and, possibly, direct influences on brain plasticity
and structural and functional brain reserves.31,32 The
role of diet and its relation with physical activity also needs to be examined,
as does constitutional frailty.
While our study was observational, we took several measures to avoid
residual confounding. We included only physically capable men, which diminishes
the possibility that relations between walking and dementia could have been
through associations with overt disability and physical impairment. Even when
focusing on men who were capable of walking 0.25 to 1 mile/d, there was an
excess of dementia vs men who walked more than 2 mile/d. Among the 76 men
who were neither slightly nor moderately active, 6 eventually developed dementia
and all reported walking less than 0.25 mile/d.
Based on the observation that cigarette smoking can reduce the benefits
of being physically active for outcomes such as mortality and stroke,14,15 cigarette smokers were also excluded
from follow-up. The smokers who were excluded in the current study (n = 161)
had no association of dementia with walking. Here, the age-adjusted incidence
of dementia was 9.8, 34.7, 35.5, and 12.6 per 1000 person-years as the range
of distance walked increased from less than 0.25 mile/d to more than 2 mile/d.
An earlier report from the Honolulu-Asia Aging Study also suggests that smoking
increases the risk of dementia.33
Studying a sample of elderly men in the mild climate of Hawaii has the
advantage that walking on a continuous basis may be more easily sustained
throughout the year. Self-reported activity may be more consistent with actual
behavior since recall is less likely to be interrupted by months of inclement
weather. With a focus on retired men, the walking that occurred was also likely
to be related to domestic needs or a modifiable decision to walk for leisure.
We did not obtain further data on participants' purposes of walking, however.
The stable environment in Hawaii that permits year-round walking may
also explain why a relation between walking and dementia has not been observed
elsewhere. Identifying a relation between walking and cognition may require
samples where levels of walking are continuous rather than sporadic. Intensity
may also be important. Since all men in the Honolulu sample were retired and
older than 70 years, however, variation in intensity may be low. In addition,
nearly 90% of the men were born in Hawaii, and outmigration has been rare
(about 1 per 1000 per year), further suggesting that associations could reflect
lifelong patterns of walking behaviors.
In combination with a low rate of outmigration, the continuous efforts
to identify cases of dementia in the Honolulu-Asia Aging Study are likely
to have resulted in case finding that is more complete than in general population–based
settings, where cognitive impairment is often unrecognized.34 Within
the current study, the CASI criterion for neurological assessment at the second
follow-up examination (score <70) was lowered from levels that were used
at the first follow-up examination (1994-1996). Although more cases were identified
at the second follow-up examination (82/158), we may have identified fewer
cases because we reduced the sensitivity of the screening test. Had the CASI
criterion for the second follow-up examination been used at the first follow-up
examination, 20 cases would have been missed. Here, dementia incidence would
become 14.8, 16.0, 12.0, and 8.6 per 1000 person-years as distance walked
increases from less than 0.25 to more than 2 mile/d. The consequence of this
loss is minimal, however, since the excess of dementia in men who walked 1
mile/d or less remains significantly higher than in men who walked more than
2 miles/day (RH after age and risk factor adjustment, 1.80; 95% CI, 1.05-3.07).
In addition, the 20 dementia cases that would have been missed tended to walk
less than the overall cohort (8/20 walked <0.25 mile/d while 3/20 walked >2
mile/d), suggesting that the observed association between walking and dementia
could have been even stronger had dementia case finding been more complete
at the second follow-up examination.
Consistent with previous reports,8- 10 cognitive
performance in the Honolulu sample (as measured by the CASI) was a markedly
stronger predictor of incident dementia than was walking. Nevertheless, walking
continued to be associated with reduced risk of dementia independent of cognitive
function, suggesting that the risk of dementia could include important factors
other than cognition. Other studies have suggested an association between
physical activity and cognition.1- 7 Even
in the Bronx Aging Study, in which the associations between walking and dementia
were weak, participants who rarely walked had a 1.5-fold excess risk of dementia
compared with those who walked frequently.8 Although
the latter was not significant, findings were based on a small sample of 469
men and women, limiting the power to detect a difference. Walking in the cohort
of elderly men in Hawaii has also been associated with a lower risk of coronary
heart disease, total mortality, and death due to cancer.26,35 Although
complex, this study and past evidence suggest that walking and active lifestyles
in general are associated with a reduced risk of dementia.