Professional Documents
Culture Documents
Copyright 2001 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Is the Risk of Developing Alzheimers Disease Greater for Women than for
Men?
Liesi E. Hebert,1 Paul A. Scherr,2 Judith J. McCann,1 Laurel A. Beckett,1 and Denis A. Evans1
Alzheimers disease; incidence; mortality; prevalence; prospective studies; risk factors; sex factors; women
Data were collected in East Boston, Massachusetts, a center of the Established Populations for Epidemiologic Study
Received for publication August 11, 1999, and accepted for publication February 15, 2000.
Abbreviations: AD, Alzheimers disease; CI, confidence interval;
EPESE, Established Populations for Epidemiologic Study of the
Elderly.
1
Rush Institute for Healthy Aging, Rush University and RushPresbyterian-St. Lukes Medical Center, Chicago, IL.
2
Health Care and Aging Studies Branch, National Center for
Chronic Disease Prevention and Health Promotion, Centers for
Disease Control, Atlanta, GA.
Reprint requests to Dr. Liesi E. Hebert, Rush Institute for Healthy
Aging, 1645 West Jackson Blvd., Suite 675, Chicago, IL 60612 (email: lhebert1@rush.edu).
Diagnosis of AD
People in both samples received the same structured clinical evaluation, including neuropsychological tests, a neurologic examination, a brief psychiatric evaluation, laboratory
132
A large proportion of people with Alzheimers disease (AD) are women; however, it is not clear whether this
is due to higher risk of disease or solely to the larger number of women alive at ages when AD is common.
Beginning in 1982, two stratified random samples of people aged 65 years in East Boston, Massachusetts
underwent detailed, structured clinical evaluation for prevalent (467 people) and incident (642 people from a
cohort previously ascertained to be disease-free) probable AD. The prevalence sample was followed for mortality
for up to 11 years (through December 1992). The age-specific incidence of AD did not differ significantly by sex
(for men vs. women, odds ratio = 0.92; 95% confidence interval (CI): 0.51, 1.67). Controlled for age, prevalence
also did not differ significantly by sex (for men vs. women, odds ratio = 1.29; 95% CI: 0.67, 2.48). The increase
in risk of mortality due to AD did not vary by sex. The odds ratio for women with AD compared with women
without AD was 2.07 (95% CI: 1.21, 3.56). For men, the odds ratio was 2.22 (95% CI: 1.02, 4.81). These findings
suggest that the excess number of women with AD is due to the longer life expectancy of women rather than
sex-specific risk factors for the disease. Am J Epidemiol 2001;153:1326.
Age, formal schooling (in years), income (in 10 categories, modeled continuously), occupational status (24),
smoking (indicators for current or past smoking and cumulative pack-years), and alcohol intake (current daily consumption) were measured at the initial interview.
Statistical methods
RESULTS
Incidence of AD among women and men
There was no significant sex difference in the agespecific prevalence of AD. Of the 467 persons clinically
evaluated for AD, 84 of the 262 women and 50 of the 205
men had AD. In analysis controlling for single year of age
and sample design, the odds ratio for AD among men compared with women was 1.29 (95 percent CI: 0.67, 2.48).
Controlling for education, income, occupational status,
smoking, and alcohol intake did not affect the results. Linear
terms for age and education were adequate. The effect of sex
on disease risk did not vary significantly with age.
Risk of death due to AD among women and men
Womens and mens risks of prevalent and incident disease were compared using logistic regression analysis, controlling for age and follow-up interval for incident disease
and adjusting for the complex sampling (25, 26). The form
of the curve relating age to disease risk was evaluated by
adding higher-order terms for age and interaction terms for
age and sex to the models and by examining residual plots.
Similar model-fitting procedures were used to test the form
of the association with education.
Sex-specific risk of death with AD was assessed with a
sample-weighted proportional hazards model (27), using
age as the time line. AD was modeled as a time-varying
covariate (28) to account for development of AD during
follow-up. Other predictors were sex and sex AD interaction. The form of the age and sex relations to mortality and
interactions with AD were tested in the same way as in the
logistic regression models. Bootstrap resampling (29) by
stratum was used to calculate the variance.
To assess the effects of differing mortality on lifetime risk of
AD, we created life tables summarizing the outcomes for hypothetical cohorts of 100,000 women and 100,000 men, starting
disease-free at age 65 years and progressing to age 90. At each
successive year of age, those still alive and free of disease could
either die before the next year, develop AD, or remain free of
AD. Those alive with AD at the start of a year could either die
or survive. The only difference between the tables was that one
used womens death rates and the other used mens. The US
rates for Whites were used to provide stable single-year rates.
Identical age-specific (single years) but not sex-specific AD
incidence rates from East Boston were used for both sexes.
Total deaths were divided between persons with AD and persons without AD for each sex, so that the odds of dying among
persons with AD relative to the odds of those without AD was
the East Boston odds of 2.14 for both sexes.
134
Hebert et al.
for statistical significance. Of 19 previous population studies, 11 reported a higher prevalence for women (111), one
reported a higher prevalence for men (30), and seven found
no difference (3137). Of the eight studies that tested for
significance, four (14) found a significantly higher prevalence in women and four found no difference (3134). Most
population-based incidence studies have found less difference between women and men.
It has been difficult to achieve adequate sample sizes for
studies of AD incidence, especially because fewer men than
women survive during follow-up at the oldest ages, at which
disease is most common. Of 12 incidence studies (1219,
3841), nine included fewer than 20 male incident cases
(1217, 19, 38, 39). Only one study reported higher incidence among women with a probability less than 0.05 (38);
another found that the risk was significantly greater for men
under age 80 and greater for women over age 80 (40). Two
recent reports in which data from several studies were combined by meta-analysis (42) or pooled analysis (43) found
significantly higher incidence among women. The pooled
EURODEM analysis resulted in an odds ratio of 0.65 for
men compared with women (43), which is just below the 95
percent confidence interval of our results, while our estimate
of 0.92 is above the 95 percent confidence limit of the
pooled analyses. Such pooled or meta-analyses have the
advantage of large numbers but the disadvantages of variation in methods across studies (44) and inability to account
for the effect of complex sampling in computations of effect
estimates or confidence intervals.
Precise control for age is also a critical methodological
component in assessment of sex differences in AD risk.
Nearly all previous studies have assessed differences by
comparing rates in age groups rather than modeling risk.
Controlling by age groups may not be adequate, because the
risk of disease is not uniform within each age group and
there are more women at the older, higher-risk end of each
group. The wider the age range in each age group, the
greater the distortion. Ten of 13 studies reporting a single
rate for everyone over age 65 found that women had 23
times the AD rate of men (3, 5, 7, 8, 10, 17, 18, 32, 39, 41).
The four studies that found statistically significant sex differences in prevalence controlled for age in 10-year or larger
age groups (14). The two studies (38, 40) and the one metaanalysis (42) that found sex differences in incidence controlled by 5-year age groups.
Comparable ascertainment of disease for each risk group
is also important for assessing the impact of risk factors. In
some incidence studies, AD status was determined by direct
examination of survivors and by questioning of informants
or review of medical records for people who died during
follow-up (10, 12, 1517, 19, 38, 41). Because alternative
sources of data collection tend not to find all cases (17, 41),
the higher mortality among men may result in apparent sex
differences. The pooled analysis that found a higher incidence for women (43) had limited ability to control for interstudy variation, which may have the effect of comparing
women and men from substantively different populations.
Prevalence studies face an additional challenge in
addressing this issue, because prevalence is determined by
ACKNOWLEDGMENTS
REFERENCES
1. Bachman DL, Wolf PA, Linn R, et al. Prevalence of dementia
and probable senile dementia of the Alzheimer type in the
Framingham Study. Neurology 1992;42:11519.
2. Canadian Study of Health and Aging Working Group.
Canadian Study of Health and Aging: study methods and
prevalence of dementia. Can Med Assoc J 1994;150:899913.
Am J Epidemiol Vol. 153, No. 2, 2001
both the incidence rate and the duration of survival after disease onset. Because risk of dying from all causes is lower
for women than for men, on average, women have AD for a
longer period of time than men do. We found no difference
by sex in the risk of all-cause mortality among persons with
AD relative to those without AD. Few previous studies have
examined AD mortality risk (4549). The only populationbased study of sex differences (49) found a larger increase
in mortality due to AD among men, after controlling for
many other factors. That result is not directly comparable to
ours, because we calculated the increased risk of death from
all causes among persons with AD.
Our study had several strengths. Prevalence of AD and
incidence of AD were measured separately in the same community, using uniform structured clinical evaluations. We
excluded people who died during the 11-year follow-up
interval from the population at risk for incident disease,
rather than attempting to evaluate their disease status. The
sample size and numbers of AD cases, especially among
men, were larger than those in most previous studies of AD
prevalence or incidence. We used multivariate logistic
regression models, an approach that provides a summary of
the effect of sex over all ages and permits precise adjustment
for year of age and simultaneous control of other confounding variables. However, this study also had limitations.
Participants were from one community; results may differ
for other populations. Despite the sample size, the possibility of a small to moderate increased risk for either sex cannot be excluded, and numbers of participants were insufficient to estimate any change in risk over the age of 90 years.
Disease incidence was measured once at the end of followup.
Our study and most previous population studies of AD
incidence suggest that women and men have the same risk
of AD onset at each year of age. Despite this, women bear
most of the burden of AD: Many more women than men
have or will develop AD in their lifetimes, and women will
survive longer with the disease.
136
Hebert et al.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
27. Cox DR. Regression models and life tables (with discussion).
J R Stat Soc B 1972;34:187220.
28. Prentice RL, Williams BJ, Peterson AV. On the regression
analysis of multivariate failure time data. Biometrika 1981;
68:3739.
29. Efron B, Tibshirani RJ. An introduction to the bootstrap. New
York, NY: Chapman and Hall Ltd, 1993.
30. Lobo A, Saz P, Guillermo M, et al. Prevalence of dementia and
depression in the elderly community in a Southern Europe population: The Zargoza Study. Arch Gen Psychiatry 1995;52:
497506.
31. Elby E, Parhad I, Hogan D, et al. Prevalence and types of
dementia in the very old: results from the Canadian Study of
Health and Aging. Neurology 1994;44:1593600.
32. Yamada M, Sasaki H, Mimori Y, et al. Prevalence and risks of
dementia in the Japanese population: RERFs adult health
study Hiroshima subjects. J Am Geriatr Soc 1999;47:18995.
33. Pfeffer RI, Afifi AA, Chance JM. Prevalence of Alzheimers
disease in a retirement community. Am J Epidemiol 1987;125:
42036.
34. Liu H, Lin K, Teng E, et al. Prevalence and subtypes of dementia in Taiwan: a community survey of 5279 individuals. J Am
Geriatr Soc 1995;43:1449.
35. Rorsman B, Hagnell O, Lanke J. Prevalence and incidence of
senile and multi-infarct dementia in the Lundby Study: a comparison between the time periods 19471957 and 19571972.
Neuropsychobiology 1986;15:1229.
36. Prencipe MA, Casini A, Ferretti C, et al. Prevalence of dementia in an elderly rural population: effects of age, sex and education. Neurology 1996;60:62833.
37. Skoog I, Nilsson L, Palmertz B, et al. A population-based study
of dementia in 85-year-olds. N Engl J Med 1993;328:1538.
38. Fratiglioni L, Viitanen M, von Strauss E, et al. Very old women
at highest risk of dementia and Alzheimers disease: incidence