Association between late-life body mass index and dementia: The Kame Project.
ABSTRACT To examine the association between body mass index (BMI), waist circumference (WC), and waist-to-hip ratio (WHR) and risk of dementia and its subtypes in late life.
Participants were members of the Kame Project, a population-based prospective cohort study of 1,836 Japanese Americans living in King County, WA, who had a mean age of 71.8 years and were dementia-free at baseline (1992-1994), and were followed for incident dementia through 2001. Cox proportional hazards models were used to estimate the risk of dementia, Alzheimer disease (AD), and vascular dementia (VaD) controlling for demographic and lifestyle characteristics and vascular comorbidities as a function of baseline BMI, WC, and WHR and change in BMI over time.
Higher baseline BMI was significantly associated with a reduced risk of AD (hazard ratio [HR] = 0.56, 95% confidence interval [CI] = 0.33-0.97) in the fully adjusted model. Slower rate of decline in BMI was associated with a reduced risk of dementia (HR = 0.37, 95% CI = 0.14-0.98), with the association stronger for those who were overweight or obese (HR = 0.18, 95% CI = 0.05-0.58) compared to normal or underweight (HR = 1.00, 95% CI = 0.18-5.66) at baseline.
Higher baseline body mass index (BMI) and slower declining BMI in late life are associated with a reduced risk of dementia, suggesting that low BMI or a faster decline in BMI in late life may be preclinical indicators of an underlying dementing illness, especially for those who were initially overweight or obese.
- [Show abstract] [Hide abstract]
ABSTRACT: The occurrence of obesity, commonly estimated using body mass index (BMI), and the most common late-onset dementia, Alzheimer's disease (AD), are increasing globally. The year 2013 marked a decade of epidemiologic observational reports on the association between BMI and late-onset dementias. In this review, we highlight epidemiological studies that measured both mid- and late-life BMI in association with dementia. Studies investigating the association between midlife BMI and risk for dementia demonstrated generally an increased risk among overweight and obese adults. When measured in late-life, elevated BMI has been associated with lower risk. In addition, being underweight and/or having a decrease in BMI in late-life are associated with higher dementia risk compared to BMI in the normal range or stable BMI. In this review, a decade (2003-2013) of epidemiologic observational studies on associations between BMI and AD is highlighted. These observations provide a strong base for addressing biological mechanisms underlying this complex association.Journal of Alzheimer's disease: JAD 08/2014; · 3.61 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Obesity impairs cognition and health-related quality of life (HRQOL) in older adults; however, the appropriate treatment of obese older adults remains controversial. The objective was to determine the independent and combined effects of weight loss and exercise on cognition, mood, and HRQOL in obese older adults. One hundred seven frail, obese older adults were randomly assigned to a control, weight-management (diet), exercise group or to a weight-management-plus-exercise (diet-exercise) group for 1 y. In this secondary analysis, main outcomes were Modified Mini-Mental State Examination (3MS) and total Impact of Weight on Quality of Life-Lite (IWQOL) scores. Other outcomes included Word Fluency Test, Trail Making Test Parts A and B, and Geriatric Depression Scale (GDS) scores. Scores on the 3MS improved more in the diet (mean ± SE: 1.7 ± 0.4), exercise (2.8 ± 0.4), and diet-exercise (2.9 ± 0.4) groups than in the control group (0.1 ± 0.4) (between-group P = 0.0001-0.04); scores in the diet-exercise group improved more than in the diet group but not more than in the exercise group. Scores on the Word Fluency Test improved more in the exercise (4.1 ± 0.8) and diet-exercise (4.2 ± 0.7) groups than in the control group (-0.8 ± 0.8; both P = 0.001). For the Trail Making Test Part A, scores in the diet-exercise group (-11.8 ± 1.9) improved more than in the control group (-0.8 ± 1.9) (P = 0.001); a similar finding was observed for the Trail Making Test Part B. Scores on the IWQOL improved more in the diet (7.6 ± 1.6), exercise (10.1 ± 1.6), and diet-exercise (14.0 ± 1.4) groups than in the control group (0.3 ± 1.6) (P = 0.0001-0.03); scores in the diet-exercise group improved more than in the diet group but not more than in the exercise group. In the diet-exercise group, peak oxygen consumption and strength changes were independent predictors of 3MS changes; weight and strength changes were independent predictors of IWQOL changes. GDS scores did not change. Weight loss and exercise each improve cognition and HRQOL, but their combination may provide benefits similar to exercise alone. These findings could inform practice guidelines with regard to optimal treatment strategies for obese older adults. This trial was registered at clinicaltrials.gov as NCT00146107.American Journal of Clinical Nutrition 04/2014; · 6.50 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Rationale Although dementia and nutritional status have been shown to be strongly associated, differences in body composition (BC) among elderly with dementia have not yet to be firmly established. Objective To assess the BC through conventional and vector bioimpedance analysis (BIA and BIVA, respectively) in a sample of institutionalized elderly men with and without dementia, in order to detect dementia-related BC changes. Methods Forty-one institutionalized men aged 65 years or older (23 without dementia- CG- and 18 with dementia-DG-) were measured with BIA and interpreted with BIVA and predictive equations. Results Age (74.4 and 75.7 y) and BMI (22.5 and 23.6 kg/m2) were similar for DG and CG, respectively. Resistance and resistance/height ratio did not differ significantly between groups Reactance and reactance/height ratio were 21.2 and 20.4% lower in DG than in CG. Phase angle was significantly lower in DG (mean: 4.0; 95% CI: 3.6-4.3 degrees) than in CG (mean: 4.7; 95% CI: 4.3-5.1 degrees). Mean fat mass index (6.0 and 7.0 kg/m2), and mean fat-free mass index (16.4 and 16.6 kg/m2) were similar in DG and CG. BIVA showed a significant downward migration of the ellipse in DG with respect to CG (T2=15.1, p<0.01). Conclusion Conventional BIA showed no significant differences in BC between DG and CG, even though reactance and reactance/height were about 21% lower in DG. Nevertheless, a body cell mass depletion and an increase in the extracellular/intracellular water-ratio were identified in DG using BIVA. BIVA reflects dementia-related changes in BC better than BIA.Nutrition 07/2014; · 3.05 Impact Factor
Association between late-life body mass
index and dementia
The Kame Project
A.R. Borenstein, PhD,
E. Schofield, MPH
Y. Wu, PhD
E.B. Larson, MD, MPH
Objective: To examine the association between body mass index (BMI), waist circumference (WC),
and waist-to-hip ratio (WHR) and risk of dementia and its subtypes in late life.
Methods: Participants were members of the Kame Project, a population-based prospective cohort
study of 1,836 Japanese Americans living in King County, WA, who had a mean age of 71.8 years
and were dementia-free at baseline (1992–1994), and were followed for incident dementia
through 2001. Cox proportional hazards models were used to estimate the risk of dementia,
Alzheimer disease (AD), and vascular dementia (VaD) controlling for demographic and lifestyle
characteristics and vascular comorbidities as a function of baseline BMI, WC, and WHR and
change in BMI over time.
Results: Higher baseline BMI was significantly associated with a reduced risk of AD (hazard ratio
[HR] ? 0.56, 95% confidence interval [CI] ? 0.33–0.97) in the fully adjusted model. Slower rate of
decline in BMI was associated with a reduced risk of dementia (HR ? 0.37, 95% CI ? 0.14–0.98),
with the association stronger for those who were overweight or obese (HR ? 0.18, 95% CI ? 0.05–
0.58) compared to normal or underweight (HR ? 1.00, 95% CI ? 0.18–5.66) at baseline.
Conclusion: Higher baseline body mass index (BMI) and slower declining BMI in late life are asso-
ciated with a reduced risk of dementia, suggesting that low BMI or a faster decline in BMI in late
life may be preclinical indicators of an underlying dementing illness, especially for those who were
initially overweight or obese. Neurology®2009;72:1741–1746
AD ? Alzheimer disease; BMI ? body mass index; CI ? confidence interval; DSM-IV ? Diagnostic and Statistical Manual of
Mental Disorders, 4th edition; HR ? hazard ratio; VaD ? vascular dementia; WC ? waist circumference; WHR ? waist-to-hip
Evidence suggests that weight loss precedes the diagnosis of dementia1-4and may be the result
of preclinical pathophysiologic changes.5Increased adiposity also is associated with an in-
creased risk of dementia.6-8These paradoxical findings are likely related to the long preclinical
phase of dementia and the problem that associations between various risk or protective factors
and dementia depend upon when they are measured in relation to the clinical onset of disease.
In this regard, overweight or obesity in midlife may be more appropriately considered a risk
factor, while declining weight in late life may be considered a preclinical indicator of the
The purpose of the current study is to examine the relation between late-life adiposity, measured
dementia, Alzheimer disease (AD), and vascular dementia (VaD). We used data from the Kame
of dementia, AD, and VaD as a function of baseline adiposity, and to further determine whether
change in BMI is associated with risk. We hypothesized that higher adiposity at baseline and slower
rate of decline in BMI would be associated with decreased risk of dementia and its subtypes.
Address correspondence and
reprint requests to Dr. Tiffany F.
Hughes, Department of
Psychiatry, University of
Pittsburgh School of Medicine,
3811 O’Hara St., Pittsburgh, PA
From the Department of Psychiatry (T.F.H.), University of Pittsburgh, PA; the Department of Epidemiology and Biostatistics (A.R.B., E.S., Y.W.),
University of South Florida, Tampa; and the Group Health Center for Health Studies (E.B.L.), University of Washington, Seattle.
Disclosure: The authors report no disclosures.
Copyright © 2009 by AAN Enterprises, Inc.
METHODS Study population. Participants were members
of the Kame Project, a population-based prospective study of
community- and institution-dwelling Japanese Americans 65
years and older living in King County, WA. The study was car-
ried out between May 1992 and December 2001 and consisted
of five time points (baseline and follow-ups at approximately 2,
4, 6, and 8 years). The study was approved by the University of
Washington Human Subjects Committee and supported by a
Japanese American Community Advisory Board, and written in-
formed consent was obtained from all participants. A more de-
tailed description of the study has been presented elsewhere.9
From the 3,045 participants enumerated in a study census of
Japanese Americans in King County, WA, in November 1991,
1,985 individuals participated in the baseline examination be-
tween May 1992 and September 1994 (65.2%). Of these, 149
were identified as prevalent cases of dementia and 1,836 were
dementia-free at baseline and eligible for follow-up to detect in-
cident dementia. Of these 1,836 participants, 1,615 (88.0%)
had anthropometric data, of whom 137 were missing follow-up
data due to death, loss to follow-up, or refusal to participate,
leaving 1,478 (80.5%) participants for the current analysis. Dur-
ing the entire study period (mean ? 7.8 years; SD ? 0.3), 129
incident dementia cases, 71 incident AD cases, and 22 incident
VaD cases were documented in the sample with complete data
for this analysis.
Dementia diagnosis. The diagnosis of dementia was based on a
two-stage case ascertainment process consisting of cognitive screen-
ing followed by a clinical diagnostic evaluation. Trained interview-
scored 86 or less of a possible 100 points were referred for full stan-
dard clinical and neuropsychological evaluation. The clinical evalu-
by study physicians9and informant interviews including the Clini-
cal Dementia Rating Scale.11Trained psychometrists administered
test battery and other tests.9A consensus committee determined the
presence of dementia and its subtypes based on the DSM-IV13crite-
ria for dementia, the National Institute of Neurological and Com-
municative Disorders and Related Disorders Association14criteria
for AD, and a number of criteria for VaD.15,16A more detailed de-
scription of the diagnostic procedure can be found elsewhere.9
Anthropometric measures. At the baseline examination, an-
thropometric measurements including standing height, weight,
WC, and hip circumference were taken by trained interviewers
and recorded for each participant. Only weight was measured at
each of the follow-up examinations. BMI (weight [kg] over
height squared [m2]) was considered the primary index of body
weight since it is scaled according to height, and was categorized
as obese (BMI ?25.0), overweight (BMI ? 23.0–24.9), normal
(BMI ? 18.5–22.9), and underweight (BMI ?18.5) according
to cutoffs proposed by the International Obesity Taskforce for
Asian populations17for descriptive purposes. WC (inches) and
WHR (WC [in] over hip circumference [in]) were considered
secondary measures of adiposity. Fixed slope parameters for each
participant were calculated using random effects modeling and
served as the measure of rate of change in BMI across the study
period (mean ? ?0.07, SD ? 0.16).
Covariates. The demographic characteristics of gender (men/
women) and education (dichotomized from the original contin-
uous variable as less than high school/high school or greater)
were included as covariates. In addition, baseline values were
elicited for current or past smoking status (yes/no), current or
past alcohol consumption (yes/no), and regular exercise (yes/no).
Self-reported history of cardiovascular conditions including hy-
pertension (yes/no), hypercholesterolemia (yes/no), angina pec-
toris (yes/no), diabetes (yes/no), heart attack (yes/no), TIA (yes/
no), and stroke (yes/no) were collected at baseline. Information
on ApoE genotype status was available for 1,056 (57.2% [of
1,836]) participants from the first biennial assessment who also
had adiposity measures.
Statistical analyses. The association between the adiposity
measures and risk of incident dementia, AD, and VaD was ex-
amined using Cox proportional hazard regression models to esti-
mate hazard ratios (HRs) and 95% confidence intervals (CIs),
with age at onset as the time scale and age at entry as the trunca-
tion variable.18Three separate models were calculated for contin-
uous baseline BMI, WC, and WHR and continuous change in
BMI: 1) adjusting for age, 2) additionally adjusting for gender
and education, and 3) additionally adjusting for smoking status,
alcohol consumption, regular exercise, hypertension, hypercho-
lesterolemia, angina pectoris, diabetes, heart attack, TIA, and
stroke. We also added a quadratic term for BMI when investigat-
ing the association between baseline BMI and dementia to ac-
count for the nonlinear relation observed graphically, which
produced a better fitting model. A multiplicative interaction
term between baseline BMI and change in BMI was estimated in
fully adjusted proportional hazard regression models for demen-
tia to determine whether the association between change in BMI
and dementia depended upon baseline BMI. All analyses were
conducted using SAS version 919with p values less than 0.05
(two-tailed) interpreted as being significant.
RESULTS The average age of the participants at
baseline (n ? 1,478) was 71.8 years, 55.3% were
women, and 75.8% had at least high school educa-
tion or greater. The average BMI was 24.3 (range
15.4–47.3), with 39.6% obese, 24.6% overweight,
32.7% normal weight, and 3.1% underweight. In
terms of the other covariates at baseline, 48.9% were
current or past smokers; 37.1% were current or past
drinkers of alcohol; 65.2% reported regular physical
activity; 46.8% reported hypertension; 5.3% re-
ported having a coronary artery attack; 3.3% re-
ported having a TIA; 2.6% reported a stroke; 16.6%
reported diabetes; 12.3% reported hypercholesterol-
emia; 5.5% reported angina; and 20.6% were ApoE
?4 allele positive. The current sample did not differ
from the dementia-free cohort (n ? 1,836) with re-
spect to any of the covariates except age and educa-
tion, where the current sample was 0.9 years younger
(t ? ?2.78; p ? 0.004) and 46.0% had greater than
or equal to high school education compared to the
dementia-free cohort (54.0%; ?2? 7.13, p ?
The characteristics of the sample by BMI category
at baseline are shown in table 1. Those who were
underweight were more likely to be women, less
likely to be a current or past smoker or alcohol
drinker, and less likely to have high cholesterol or
Neurology 72May 19, 2009
diabetes. Those who were of normal weight were the
least likely to have hypertension.
Table 2 shows the associations between baseline
and change in BMI and risk of dementia, AD, and
VaD. BMI at baseline was inversely associated with
dementia, AD, and VaD, but was only significant for
AD in the fully adjusted model. We did not find that
the risk of dementia, AD, and VaD was associated
Table 1 Baseline sample characteristics by BMI category among 1,478 participants 65 years and older in
the Kame Project (1992–2001)
<18.5 (n ? 46)18.5–22.9 (n ? 484)23.0–24.9 (n ? 363)
>25.0 (n ? 585)p Value*
Dementia, n ? 129
Alzheimer disease, n ? 71
Vascular disease, n ? 22
Age, mean (SD), n ? 1,478
72.99 (6.47)72.10 (5.80)71.79 (5.06) 71.41 (4.43)0.06
Sex, % female, n ? 1,478
82.61 71.0752.34 41.88
Education, % > high school,
n ? 1,478
71.74 76.8674.10 76.410.71
Alcohol, % yes, n ? 1,478
Smoking, % yes, n ? 1,478
34.78 41.1248.48 56.58
Hypertension, % yes, n ? 1,478
39.13 36.7848.76 54.53
Hypercholesterolemia, % yes,
n ? 1,465
Diabetes mellitus, % yes, n ? 1,467
Angina pectoris, % yes, n ? 1,460
Stroke, % yes, n ? 1,471
TIA, % yes, n ? 1,469
Physical activity, % regular,
n ? 1,431
50.00 64.6768.26 64.840.12
ApoE, % ?4 positive,
n ? 1,056
*Analysis of variance for continuous variables or ?2test for categorical variables.
BMI ? body mass index.
Table 2Hazard ratios for incident dementia by baseline BMI and change in BMI over study period among
participants 65 years and older in the Kame Project (1992–2001)
DementiaAlzheimer diseaseVascular dementia
unaffectedHR (95% CI)
unaffectedHR (95% CI)
unaffectedHR (95% CI)
129/1,3490.93 (0.51?1.69)69/1,3980.63 (0.35?1.12)22/1,418 0.82 (0.20?3.32)
129/1,349 0.89 (0.49?1.61) 69/1,3990.60 (0.34?1.06)22/1,4180.73 (0.19?2.88)
108/1,2940.78 (0.42?1.44)59/1,333 0.56 (0.33?0.97)19/1,3510.66 (0.13?3.36)
74/9430.80 (0.38?1.68)43/9710.68 (0.31?1.51) 12/9910.40 (0.06?2.51)
129/1,3490.58 (0.22?1.49)69/1,3980.37 (0.17?1.15)22/1,4180.73 (0.07?7.35)
129/1,3490.59 (0.23?1.54)69/1,3980.37 (0.11?1.22)22/1,418 0.80 (0.08?7.97)
108/1,2940.37 (0.14?0.98)59/1,3330.32 (0.09?1.08)19/1,3510.41 (0.03?5.34)
74/9430.31 (0.09?1.02)43/9710.21 (0.06?0.80) 12/9910.43 (0.02?10.60)
*Model adjusted for age.
†Model adjusted for sex and education.
‡Model additionally adjusted for alcohol, smoking, hypertension, hypercholesterolemia, diabetes, angina pectoris, stroke,
TIA, and physical activity.
§Model additionally adjusted for ApoE ?4 status.
HR ? hazard ratios; BMI ? body mass index; CI ? confidence interval.
Neurology 72May 19, 2009
with WC and WHR at baseline (data not shown).
The risk of dementia and AD was reduced with a
slower rate of BMI decline during the study period in
the fully adjusted model where an average BMI de-
cline of 1.06 units less per year (1.15 for AD) was
associated with a 63% reduced risk (68% for AD),
although the point estimate for AD was marginally
significant. Adjusting for the presence of the ApoE
?4 allele did not substantially change the results, but
did strengthen the association between rate of change
in BMI and the risk of dementia and AD. A signifi-
cant interaction between baseline BMI and change in
BMI was found for dementia (HR ? 0.73, 95%
CI ? 0.53–0.99, p for interaction ? 0.048). The
reduction in risk of dementia with slower BMI de-
cline was greater with increasing BMI at baseline
(figure) where there was a significant reduction in
risk for those whose baseline BMI was overweight or
obese (HR ? 0.18, 95% CI ? 0.05–0.58) com-
pared to those who were normal or underweight
(HR ? 1.00, 95% CI ? 0.18–5.66).
DISCUSSION In this analysis from the Kame
Project, we report that the risk for AD was reduced
with higher late-life BMI at baseline and that the
risks of dementia and AD were reduced with a slower
rate of BMI decline over a follow-up period of ap-
proximately 8 years. More importantly, the extent to
which change in BMI was associated with dementia
depended upon baseline BMI such that overweight
or obese participants at baseline had a more pro-
nounced reduction in risk with slower decline in
BMI compared to normal or underweight partici-
pants. These findings suggest that late-life adiposity
is associated with the risk of dementia, where high
and slowly declining BMI reduce the risk; or con-
versely, that low or fast declining BMI may be pre-
clinical indicators for dementia.
The risk for dementia is believed to develop across
the lifespan as the pathologic hallmarks have been
detected decades before its clinical presentation. Be-
cause of its long prodromal period, assessing the
characteristics of a risk factor is time-dependent and
the potential for reverse causality exists. Our finding
of a reduced risk of AD with higher baseline BMI is
suggestive of a protective effect of higher BMI in late
life, similar to findings from the Kungsholmen
Project6and the Chicago Health and Aging
Project.20This is different from what has been shown
in midlife, where overweight or obesity increased the
risk of dementia and AD.8Our findings are in accord
with others who have shown weight loss in later life
to be a risk factor for dementia1,2and for weight loss
to precede the diagnosis of dementia.3,4Hence, it
may be that a nonlinear association exists where
higher adiposity in midlife increases the risk of de-
mentia and its subtypes, and that pathophysiologic
changes associated with dementia then lead to de-
clines in adiposity in late life.5
Measures of central fat distribution, including
WC and WHR, are known to increase the risk of
coronary heart disease21and mortality22more than
total body weight. Evidence suggests that central
obesity in midlife23and late life24increases the risk of
dementia and AD. Our findings do not support a
link between WC or WHR in late life and dementia
risk. These findings are partially in accord with those
of the Northern Manhattan study where continuous
WC was not associated with dementia, AD, or de-
mentia associated with stroke. The same study did,
however, find that the risk of dementia associated
with stroke was increased for the largest WC quartile
compared to the smallest WC quartile.7Additional
studies at both midlife and late life are needed to
further elucidate whether central adiposity is associ-
ated with the risk for dementia and its subtypes.
The finding that overweight or obese participants
had a reduced risk of dementia with slower decline in
BMI compared to those who were normal or under-
weight may reflect a floor effect. Those who are nor-
mal or underweight at baseline have less weight to
lose compared to those who are overweight or obese,
which would lessen their rate of change in BMI over
the course of the study. It may also be that those who
were normal or underweight at the beginning of the
study and were losing weight at a faster rate were lost
to follow-up since this weight loss could affect overall
health. Taken together, the results suggest that hav-
ing a slow rate of a decline in weight if previously
overweight or obese may reduce risk more than being
overweight or obese alone in late life.
The x-axis shows baseline body mass index (BMI), and the
y-axis shows the hazard ratio (HR) for change in BMI over
the study period and dementia controlling for baseline BMI,
age, and education.
Neurology 72May 19, 2009
The use of BMI categories for Asian populations
in our study resulted in a high number (39.6%) of
obese (BMI ?25) participants at baseline compared
to the United States prevalence of 12% in the early
1990s.25Using the Caucasian BMI cutpoints re-
duced this number to 5.0% for obese (BMI ?30)
with the remaining 3.1% underweight (BMI
?18.5), 57.3% normal (BMI ? 18.5–25), and
34.6% overweight (BMI ? 25–30). Since 96% of
the participants in the Kame Project were 100% Jap-
anese,9and twin studies have shown that genetic in-
fluences on BMI are substantial,26we considered the
Asian, rather than Caucasian, categories to be more
appropriate for our sample. Furthermore, studies
have shown that Asians in general have higher body
fat, greater centralized distribution of body fat, and
higher WHR than Caucasians with lower or similar
BMIs, which highlights the importance of redefining
the categories to assess health risks in our sample.17
Despite this descriptive difference, we believe that
our main findings are generalizable beyond Japanese
American or Asian populations since we used contin-
uous measures of baseline BMI and change in BMI
that were independent of BMI categories.
Several biologic processes may explain the associ-
ation between high and slower decline in BMI and
reduced risk of dementia. Higher weight in late life
may offer protection by increasing insulin-growth
factor I levels,27increasing leptin hormone levels
known to be involved in regulation of synaptic plas-
ticity in the hippocampus,28and increasing the pro-
duction of estrogen,29all of which have been shown
to be associated with better cognitive perfor-
mance.30,31Slower decline in BMI over the course of
the study may indicate that preclinical changes asso-
ciated with dementia are not occurring. Brain areas
that control weight (i.e., mesial temporal cortex)32
are affected during the preclinical dementia phase
that may lead to weight loss. Weight loss may also
result from predementia apathy,33reduced olfactory
function,34difficulty in eating,35or inadequate nutri-
tion36due to cognitive impairment.
There are both strengths and limitations of this
study that should be acknowledged. The first
strength is that the association between various mea-
sures of adiposity and incident dementia and two of
its subtypes were examined. Also, the study included
both community and institutionalized individuals,
minimizing any selection bias that may result from
including only those healthy enough to remain inde-
pendent in the community. Finally, we adjusted for
multiple variables that would likely confound any as-
sociations between adiposity and dementia.
Limitations of this study include the relatively
short follow-up period, which increases the potential
for reverse causality to explain the findings. Also, our
sample was limited to Americans who were of Japa-
nese ancestry and thus may limit the generalizability
to other populations. We also assumed that height
was constant throughout the study period even
though studies have shown that women and men lose
0.2 to 0.3 cm per year between the ages of 70 and 90
years37; however, any error introduced would likely
nondifferentially overestimate the rate of decline in
BMI and not bias the findings. It is also possible that
using BMI as our measure of weight may have under-
estimated adiposity in the elderly who generally have
less lean body mass,38which would have attenuated
our findings. Finally, we had insufficient power to
detect a relationship between adiposity and VaD, but
did find similar point estimates as dementia and AD.
Others studies with more power have shown a stron-
ger effect for VaD than AD,7suggesting that adipos-
ity may exert its influence on dementia with a
1.Buchman AS, Wilson RS, Bienias JL, Shah RC, Evans DA,
Bennett DA. Change in body mass index and risk of inci-
dent Alzheimer disease. Neurology 2005;65:892–897.
2.Stewart R, Masaki K, Xue Q-L, et al. A 32-year prospective
study of change in body weight and incident dementia.
Arch Neurol 2005;62:55–60.
3.Johnson DK, Wilkins CH, Morris JC. Accelerated weight
loss may precede diagnosis in Alzheimer disease. Arch
4.Knopman DS, Edland SD, Cha RH, Petersen RC, Rocca
WA. Incident dementia in women in preceded by weight
loss by at least a decade. Neurology 2007;69:739–746.
5.Buchman AS, Schneider JA, Wilson RS, Bienias JL, Ben-
nett DA. Body mass index in older persons is associated
with Alzheimer disease pathology. Neurology 2006;67:
6.Atti AR, Palmer K, Volpato S, Winblad B, De Ronchi D,
Fratiglioni L. Late-life body mass index and dementia inci-
dence: nine-year follow-up data from the Kungsholmen
Project. J Am Geriatr Soc 2008;56:111–116.
7.Luchsinger JA, Patel B, Tang M-X, Schupf N, Mayeux R.
Measures of adiposity and dementia risk in the elderly.
Arch Neurol 2007;64:392–398.
8.Whitmer RA, Gunderson EP, Quesenberry CP, Zhou J,
Yaffe K. Body mass index in midlife and risk of Alzheimer
disease and vascular dementia. Curr Alzheimer Res 2007;
9.Graves AB, Larson EB, Edland SD, et al. Prevalence of
dementia and its subtypes in the Japanese American popu-
lation of King County, Washington state, The Kame
Project. Am J Epidemiol 1996;144:760–771.
10.Teng EL, Hasegawa K, Homma A, et al. The Cognitive
Abilities Screening Instrument (CASI): a practical test for
cross-cultural epidemiologic studies of dementia. Int Psy-
Neurology 72May 19, 2009
11.Hughes CP, Berg L, Danziger WL, Coben LA, Martin RL.
A new clinical scale for the staging of dementia. Br J Psy-
Morris JC, Heyman A, Mohs RC, et al. The Consortium
to Establish a Registry for Alzheimer’s Disease (CERAD):
part I: clinical and neuropsychological assessment of Alz-
heimer’s disease. Neurology 1989;39:1159–1165.
American Psychiatric Association. Diagnostic and Statisti-
cal Manual of Mental Disorders: DSM-IV. Washington,
DC: American Psychiatric Association; 1994.
McKhann G, Drachman D, Folstein M, Katzman R, Price
D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: re-
port of the NINCDS-ADRDA Work Group under the aus-
pices of the Department of Health and Human Services Task
Force on Alzheimer’s Disease. Neurology 1984;34:939–944.
Chui HC, Victoroff JI, Margolin D, Jagust W, Shankle R,
Katzman R. Criteria for the diagnosis of ischemic vascular
dementia proposed by the State of California Alzheimer’s
Disease Diagnostic and Treatment Centers. Neurology
Roman GC, Tatemichi TK, Erkinjuntti T, et al. Vascular
dementia: diagnostic criteria for research studies: report of
the NINDS-AIREN International Workshop. Neurology
Kanazawa M, Yoshiike N, Osaka T, Numba Y, Zimmet P,
Inoue S. Criteria and classification of obesity in Japan and
Asia-Oceania. Asia Pacific J Clin Nutr 2002;11:S732–S737.
Korn EL, Graubard BI, Midthune D. Time-to-event anal-
ysis of longitudinal follow-up of a survey: choice of the
time-scale. Am J Epidemiol 1997;145:72–80.
SAS Institute. SAS System for Microsoft Windows (ver-
sion 9). Cary, NC: SAS Institute Inc; 2003.
Evans DA. Body mass index and cognitive decline in a biracial
Yarnell JW, Patterson CC, Thomas HF, Sweetnam PM.
Central obesity: predictive value of skinfold measurements
for subsequent ischemic heart disease at 14 years follow-up
in the Caerphilly Study. Int J Obes Relat Metab Disord
Mason C, Craig CL, Katzmarzyk PT. Influence of central
and extremity circumferences on all-cause mortality in
men and women. Obesity 2008;16:2690–2695.
Whitmer RA, Gustafson DR, Barrett-Connor E, Haan
MN, Gunderson EP, Yaffe K. Central obesity and in-
creased risk of dementia more than three decades later.
Razay G, Vreugdenhil A, Wilcock G. Obesity, abdominal
obesity and Alzheimer disease. Dement Geriatr Cogn Dis-
25.Mokdad AH, Serdula MK, Dietz WH, Bowman BA,
Marks JS, Koplan JP. The spread of the obesity epidemic
in the United States, 1991–1998. JAMA 1999;282:1519–
Stunkard AJ, Harris JR, Pedersen NL, McClearn GE. The
body-mass index of twins who have been reared apart. N
Engl J Med 1990;322:1483–1487.
Yamamoto H, Kato Y. Relationship between plasma
insulin-growth like factor I (IGF-I) levels and body mass
index (BMI) in adults. Endocr J 1993;4:41–45.
Harvey J, Solovyova N, Irving A. Leptin and its role in
hippocampal synaptic plasticity. Prog Lipid Res 2006;45:
Singh M, Dykens JA, Simpkins JW. Novel mechanisms
for estrogen-induced neuroprotection. Exp Biol Med
Okereke O, Kang JH, Ma J, Hankinson SE, Pollak MN,
Grodstein F. Plasma IGF-I levels and cognitive perfor-
mance in older women. Neurobiol Aging 2007;28:135–
Oomura Y, Hori N, Shiraishi T, Fukunaga K, Takeda H,
Tsuji M. Leptin facilitates learning and memory perfor-
mance and enhances hippocampal CA1 long-term potenti-
ation and CaMK II phosphorylation in rats. Peptides
Grundman M, Corey-Bloom J, Jerigan T, Archibald S,
Thal LJ. Low body weight in Alzheimer’s disease is associ-
ated with mesial temporal cortex atrophy. Neurology
Friedland RP, Fritsch T, Smyth KA, Koss E, Lerner AJ,
Chen CH. Patients with Alzheimer’s disease have re-
duced activities in midlife compared with healthy
control-group members. Proc Natl Acad Sci USA 2001;
Djordjevic J, Jones-Gotman M, De Sousa K, Chertkow H.
Olfaction in patients with mild cognitive impairment and
Alzheimer’s disease. Neurobiol Aging 2008;29:693–706.
Chang CC, Roberts BL. Feeding difficulty in older adults
with dementia. J Clin Nurs 2008;17:2266–2274.
Shatenstein B, Kergoat MJ, Reid I. Poor nutrient intakes
during 1-year follow-up with community-dwelling older
adults with early-state Alzheimer dementia compared to
cognitively intact matched controls. J Am Diet Assoc
Sorkin JD, Muller DC, Andres R. Longitudinal change in
the heights of men and women: consequential effects on
body mass index. Epidemiol Rev 1999;21:247–260.
Elmadfa I, Meyer AL. Body composition, changing physi-
ological functions, and nutrient requirements of the el-
derly. Ann Nutr Metab 2008; 52 suppl 1: 2–5.
Neurology 72May 19, 2009