American Journal of Epidemiology
Copyright © 1998 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 148, No. 1
Printed in U.S.A.
Use of Antioxidant Supplements and Its Association with Cognitive Function
in a Rural Elderly Cohort
The MoVIES Project
Aaron B. Mendelsohn,1 Steven H. Belle,1 Gary P. Stoehr,2 and Mary Ganguli 13
There has been much interest recently in the therapeutic benefits of antioxidants, including a possible
protective role in preventing or delaying cognitive decline. This study describes the use of antioxidant
supplements among 1,059 rural, noninstitutionalized elderly residents of southwestern Pennsylvania who are
participants in the Monongahela Valley Independent Elders Survey. The data were collected during the
survey's second wave of follow-up (1989-1991). The mean age of participants was 74.5 years (standard
deviation 5.5), and 57.3% were women. Current use of nutritional supplements containing vitamin A, C, or E,
/3-carotene, zinc, or selenium was measured through self-report. Subjects were administered a battery of 15
neuropsychological tests measuring performance in several cognitive domains. Of the 1,059 persons, 342
(32.3%) were taking antioxidant supplements. Women and persons with higher levels of education were more
often antioxidant users. Antioxidant use did not vary significantly by age, race, or income. In univariate
analyses, antioxidant use was significantly and positively associated with performance on several cognitive
tests. However, after adjustment for age, education, and sex, there were no significant differences in cognitive
test performance between antioxidant users and nonusers. This study is one of only a few that have analyzed
the relation between antioxidants and cognition in a community-based sample. After potentially confounding
factors are accounted for, the results do not support the hypothesis that antioxidant supplement use is
associated with cognitive function. Am J Epidemiol 1998; 148:38-44.
aged; antioxidants; cognition; minerals; nutrition; rural population; vitamins
expensive, widely available, and generally nontoxic
(6). The elderly are frequent users of such supplements
(7).
Although there has been some scientific attention to
the relation between antioxidants and cognition, the
majority of studies in this area have been conducted on
nonrepresentative samples of the population (1), such
as demented persons and nondemented controls ( 8 10), healthy volunteers (11), or institutionalized persons (12). Few studies have analyzed random community samples. Furthermore, few studies have examined
more than general mental status.
There were two specific objectives for the present
report. The first aim was to describe the use of antioxidant supplements in a random sample of rural,
community-based elders, who have less access to
health care and are poorer and less educated than the
nonrural elderly (13). The second aim was to examine
whether use of antioxidant supplements is associated
with cognitive function, using data from a battery of
neuropsychological tests assessing performance in a
range of cognitive domains.
The prevalence of cognitive impairment is expected
to rise with the pending increase in the elderly population (1, 2). Antioxidants might play a role in preventing or delaying age-related cognitive decline (1)
by functioning as scavengers of free radicals (3),
which have been shown to be involved in aging processes (4) and in cerebrovascular disease pathology
(5). The possibly beneficial effects of antioxidant use
with regard to cognitive functioning are still unproven.
Nevertheless, antioxidant supplements are attractive to
the elderly population because they are relatively inReceived for publication April 11,1997, and in final form November 13, 1997.
Abbreviations: CERAD, Consortium to Establish a Registry for
Alzheimer's Disease; MoVIES, Monongahela Valley Independent
Elders Survey.
1
Department of Epidemiology, Graduate School of Public
Health, University of Pittsburgh, Pittsburgh, PA.
2
School of Pharmacy, University of Pittsburgh, Pittsburgh, PA.
3
Division of Geriatrics and Neuropsychiatry, Department of Psychiatry, School of Medicine, University of Pittsburgh, Pittsburgh, PA.
Reprint requests to Aaron B. Mendelsohn, Epidemiology Data
Center, Graduate School of Public Health, University of Pittsburgh,
127 Parran Hall, 130 DeSoto Street, Pittsburgh, PA 15261.
38
Antioxidants and Cognitive Function in the Elderly
MATERIALS AND METHODS
Study population
Participants in this study were enrolled in the
Monongahela Valley Independent Elders Survey
(MoVIES). The sampling and recruiting methods and
research design have been described elsewhere (14,
15). Briefly, the MoVES Project is an ongoing, prospective community study of cognitive impairment and dementia in a sample of elderly individuals living in a
rural setting. Participants in this study all reside in the
mid-Monongahela Valley, located in southwestern
Pennsylvania. In 1987, an age-stratified 1:13 random
sample of the population was drawn from voter registration and senior citizen mailing lists supplied by
the Monongahela Valley Community Health Center
(14). Given that the voter registration lists contained
all individuals who had ever voted in this area of low
levels of in- and outmigration, we are reasonably assured of their completeness and appropriateness. Requirements for participation were: a minimum age of
65 years, a minimum educational level of sixth grade,
and fluency in English; in addition, the participant
could not be in a long-term care institution at the time
of sampling (14, 15).
Repeat screening of the study cohort occurred, on
average, every 2 years after baseline. The data presented here are from the second wave of follow-up
(1989-1991), during which information on use of
nutritional supplements was first collected in detail. Of
the 1,366 randomly selected study participants at baseline, 1,059 seen in the second wave were included in
this analysis. The difference was due to 137 deaths,
108 subjects who deferred participation to a future
wave, 46 dropouts, 12 relocations, and 2 subjects who
were untestable. Two other persons had incomplete or
missing data on the variables of interest and were also
excluded from these analyses.
Data collection
During each wave of follow-up, trained research
assistants interviewed participants and reliable informants, whenever the latter were available. These semistructured interviews were conducted primarily in the
participants' homes and included cognitive screening
and collection of self-reported data on demographic,
social, and functional factors and medication use (14).
The cognitive test battery consisted of 15 tests measuring performance in various domains known to be
affected in dementia (16). The cognitive tests included
the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) neuropsychological battery
(17), consisting of: the Mini-Mental State Examination (18); the 15-item CERAD version of the Boston
Am J Epidemiol
Vol. 148, No. 1, 1998
39
Naming Test (19); Immediate Learning, Delayed Recall, and Delayed Recognition (originals and foils) of
a 10-item CERAD word list (17); the CERAD Constructional Praxis (circle, diamond, cube, cross) Test
(20); and Verbal Fluency for Categories (names of
fruits and animals) (21). Additionally, the Mo VIES
battery included: Verbal Fluency for Letters (words
beginning with the letters "P" and "S") (21); the Clock
Drawing Test (22); Immediate Retell and Delayed
Recall of an 18-item story (23); the Trailmaking Test,
parts A and B (24); and the Temporal Orientation Test
(25). Descriptions of these tests have been provided
previously (16).
Data on medication use were obtained by asking
participants, "Do you take any over-the-counter medicines (for which you don't need a prescription), such
as for sleep, headaches, arthritis, constipation, heartburn, colds, itching/rashes, vitamins, tonics, etc.?"
Whenever possible, research assistants verified medication use by reading the labels of drugs and supplements that respondents reported taking (15). For
purposes of this analysis, any over-the-counter medication or nutritional supplement containing vitamin A,
C, or E or /3-carotene, zinc, or selenium was considered an antioxidant. Although zinc and selenium are
not antioxidants, these minerals were defined as antioxidants for consistency with other analyses (10, 26)
and because these nutrients participate in antioxidant
enzyme activities (27). Data on dosages of antioxidant
supplements taken were not collected, because of the
complexity and time involved in standardizing the
amounts of active ingredients found in these supplements (28). Furthermore, self-reported information on
duration of use was found to be unreliable during the
baseline interviews and was not collected at follow-up.
Statistical analysis
Characteristics of antioxidant users. Chi-square
tests were utilized to assess trends in antioxidant use
by age (65-74, 75-84, and 2:85 years), education (less
than high school, high school graduation, and more
than high school), and annual income (<$ 10,000,
$10,000-$15,999, and >$16,000).
A test for differences in proportions was used to
determine whether antioxidant use differed by sex
or race.
Cognitive function. On the basis of moving averages, we determined whether to analyze scores on
each cognitive test in continuous or dichotomous
form. The moving average is a smoothing technique
that can be used as a graphic aid to assess relations
such as those between antioxidant supplement use
and cognitive test scores (29). When this technique
showed a constant proportion of antioxidant supple-
40
Mendelsohn et al.
ment use across a range of cognitive test scores and a
discontinuity or trend toward more (or less) antioxidant use with changing test scores, a "natural cutpoint"
was considered to be present.
When the test scores were examined with respect to
the proportions of antioxidant users, moving averages
revealed natural cutpoints for the 10 tests listed in
table 2. Natural cutpoints found for these tests were all
near scores that represented the lowest 10th percentile
of test scores in this sample. The 10th percentile has
been used in other MoVIES analyses to represent the
threshold below which individuals are identified as
likely to be cognitively impaired and are asked to
undergo further clinical evaluation (14). Hence, to be
consistent with previous MoVIES analyses, we used
the 10th percentile as the cutpoint for test scores
analyzed dichotomously.
Pearson chi-square statistics were used to assess the
significance of the association between antioxidant use
and scores on the 10 cognitive tests analyzed as
dichotomous variables. Multiple logistic regression
analyses adjusting for age (continuous variable), sex
(reference group: women), and education (more than
high school (reference group), high school, or less than
high school) were used to examine the independent
relations between antioxidant use and these cognitive
test scores.
For five tests (Story Immediate Retell, Trailmaking
part B, Word List Delayed Recall, Word List Immediate Learning, and Verbal Fluency for Categories),
the moving averages technique did not reveal any
natural cutpoints, and the test scores were analyzed on
a continuous scale (table 3).
The distributions of test scores for antioxidant users
and nonusers were compared by Student's t test for the
five cognitive tests with scores analyzed on a continuous scale. All assumptions of this parametric test
were met. Multiple linear regression models adjusting
for age, sex, and education (as categorized above)
were used to examine the independent relations between antioxidant use and scores on these five cognitive tests. The Trailmaking part B scores were normalized for regression modeling by taking the square root
of the number of connections made per second.
For linear regression analyses, plots of the predicted
values from the cognitive tests by the residual values
were examined to assess whether the residuals were
randomly distributed. Additionally, the square of the
correlation coefficient was computed to measure the
percentage of the variance in the cognitive test scores
explained by the independent variables.
For logistic regression models, Hosmer-Lemeshow
goodness-of-fit statistics were calculated (30). When
the Hosmer-Lemeshow goodness-of-fit statistic was
significant, independent variables were recategorized
to achieve adequate model fit.
Determination of significance. For all analyses, ap
value less than 0.05 was chosen a priori to indicate
statistical significance.
RESULTS
Characteristics of the MoVIES sample
The 1,059 subjects described in this report had a
mean age of 74.5 years (standard deviation 5.5), with
a range of 66-97 years; 57.3 percent were women, and
96.9 percent were white. The median educational level
was high school graduation, and the median annual
income was in the $10,000-$15,999 range.
Characteristics of antioxidant users
Of the 1,059 subjects, 342 (32.3 percent) reported
taking at least one nutritional supplement containing
antioxidants. Information was obtained from a proxy
for 18 (5.3 percent) of the antioxidant supplement
users and 38 (5.3 percent) of the nonusers. The majority of antioxidant users (70.5 percent) were taking
multivitamin preparations. Vitamin C (28.4 percent)
and vitamin E (27.5 percent) supplements were also
common sources of antioxidants.
A higher percentage of women (35.8 percent) than
of men (27.7 percent) reported taking antioxidants
TABLE 1. Use of nutritional supplements containing antioxidants, by selected demographic characteristics: The
MoVIES* Project, 1989-1991
No.
% using
antioxidants
1,059
32.3
Sex
Women
Men
607
452
35.8
27.7
0.005
Educational level
<High school
High school
>High school
449
359
251
27.4
35.7
36.3
0.01
Age (years)
<75
75-84
285
598
400
61
33.3
30.5
34.4
0.61
33
1,026
24.2
32.6
0.32
287
324
309
31.4
32.1
31.4
0.98
Characteristic
Total
Race
Black
White
Annual income (dollars)
<10,000
10,000-15,999
216,000
1
P
value
MoVIES, Monongahela Valley Independent Elders Survey.
Am J Epidemiol
Vol. 148, No. 1, 1998
Antioxidants and Cognitive Function in the Elderly
41
TABLE 2. Results from univariate and multivariate regression analyses of the associations between antioxidant supplement use
and dichotomously defined (above vs. below 10th percentile) cognitive test scores: The MOVIES* Project, 1989-1991
Univariate analyses
Cognitive
test
Antioxidant users
10th
percentile t
Boston Naming
Clock Drawing
Mini-Mental State
Story Delayed Recall
Trailmaking part A
(seconds)
Temporal Orientation
Construction Praxis
Verbal Fluency (Letters)
Word List Delayed
Recognition of Originals
Word List Delayed
Recognition of Foils
No.
No. ot
participants
Multiple regression
analyses!
Nonusers
P
No.
below
10th
percentile
%
No. of
participant!:
below
10th
%
value§
OR*,H
95% Cl*
percentile
<13
<6
<24
<1.5
332
332
338
330
34
26
27
26
10.2
7.8
8.0
7.9
700
697
712
702
82
56
67
86
11.7
8.0
9.4
12.3
0.48
0.91
0.45
0.04
1.04
0.95
0.99
1.46
0.66-1.63
0.57-1.58
0.59-1.64
0.89-2.38
>83
>1
<8
<13
324
336
332
332
25
45
37
32
7.7
13.4
11.1
9.6
692
703
698
700
76
70
80
69
11.0
10.0
11.5
9.9
0.11
0.10
0.88
0.91
1.31
0.68
0.91
0.87
0.79-2.19
0.45-1.02
0.59-1.41
0.55-1.38
<9
328
32
9.8
699
104
14.9
0.02
1.42
0.92-2.19
<10
328
33
10.1
699
53
7.6
0.18
0.71#
0.45-1.13
* MoVIES, Monongahela Valley Independent Elders Survey; OR, odds ratio; Cl, confidence interval.
t Logistic regression models with cognitive test score as the dependent variable and antioxidant supplement use (yes/no), age (continuous
variable), education (>high school (reference group), high school, <high school), and sex (reference group: women) included.
f Lowest 10th percentile of scores for the entire sample (n = 1,059).
§ x 2 test.
H Odds ratio for the likelihood of scoring above the lowest 10th percentile for those taking antioxidant supplements compared with those
not taking antioxidant supplements.
# For this model, the education categories of high school and <high school were combined.
(p = 0.005) (table 1). Those in the higher educational
categories were also more likely to be taking antioxidants (p = 0.01). There were no significant differences (p ^ 0.05) in antioxidant consumption according to age, race, or income.
Cognitive performance and antioxidant status
The proportion of antioxidant users scoring above
the 10th percentile was higher than that of nonusers on
several of the tests analyzed as dichotomous variables;
however, the difference was statistically significant
(p < 0.05) only for Story Delayed Recall and Word
List Delayed Recognition of Originals (table 2).
There were no significant associations between antioxidant use and scores on any of the cognitive tests
analyzed in logistic regression models, after adjustment for age, sex, and education (table 2). Furthermore, the odds ratios for several of the tests were
nearly 1, suggesting that antioxidant use had little
association with scoring above or below the 10th percentile on these cognitive tests. The null findings are
probably not due to lack of power, since this sample
provided approximately 79 percent power to detect an
odds ratio of 2 for antioxidant use at an alpha level of
0.05 after adjustment for age, sex, and education.
The Hosmer-Lemeshow statistic was significant
(p < 0.05) for the model in which Word List Delayed
Am J Epidemiol
Vol. 148, No. 1, 1998
Recognition of Foils was the outcome variable. The
education categories of high school and less than high
school were combined, and a new model for Word List
Delayed Recognition of Foils was generated. The new
model had adequate fit.
Table 3 shows the mean values (and standard errors)
in the two antioxidant groups for the cognitive tests
that were analyzed as continuous variables. Persons
taking antioxidants scored better than those not taking
antioxidants on all tests; differences were statistically
significant (p < 0.05) for all tests except Trailmaking
part B. After controlling for age, sex, and education,
there were no significant associations between antioxidant use and the cognitive tests analyzed in linear
regression models. However, antioxidant users performed better than nonusers on all of these tests. Table
3 shows the parameter estimates and p values for the
associations between antioxidant use and cognitive
test performance for these tests. Also included are the
R2 values for these models, which ranged from 0.19 to
0.29. The plots of the predicted values versus the
residuals revealed that there was no systematic bias in
the models.
To see whether lack of adequate statistical power
might explain why associations were not found between the cognitive test scores and antioxidant supplement use, we performed post hoc power calcula-
42
Mendelsohn et al.
TABLE 3. Results from univariate and multivariate regression analyses of the associations between antioxidant supplement use
and continuously defined cognitive test scores: The Mo VIES* Project, 1989-1991
Univariate analyses
Cognitive
test
Story Immediate Retell
TrailmaMng part B#
Word List Delayed Recall
Word List Immediate Learning
Verbal Fluency (Categories)
MuR^le regression
analyses!
Nonusers
Antioxidant users
No. of
participants
Mean
score
SE*
No. of
participants
Mean
score
SE
P
value*
Parameter
estimate
(P)
SE
331
319
329
329
332
6.3
0.44
6.5
19.4
26.5
0.16
0.006
0.11
0.23
0.36
703
660
698
698
699
5.8
0.43
6.1
18.6
25.3
0.11
0.004
0.08
0.16
0.25
0.005
0.08
0.01
0.004
0.008
0.26
0.005
0.19
0.41
0.71
0.18
0.007
0.13
0.25
0.40
P
valueH
0.19
0.29
0.19
0.22
0.19
0.13
0.47
0.13
0.10
0.08
* MoVIES, Monongahela Valley Independent Elders Survey; SE, standard error.
t Linear regression models with cognitive test score as the dependent variable and antioxidant supplement use (yes/no), age (continuous variable), education
(>hlgh school (reference group), high school, <high school), and sex (reference group: women) included.
t Student's nest.
§ Square of the correlation coefficient.
1) Test of whether the parameter estimate = 0.
# Square root of the number of connections made per second.
tions. This sample provided 55-91 percent power to
detect as significant (at an alpha level of 0.05) increases in the R2 values (in linear regression models
containing age, sex, and education) as small as 0.003
that were due to the addition of a term for antioxidant
use.
DISCUSSION
Nearly one third of MoVIES participants reported
using nutritional supplements containing antioxidants.
This figure is consistent with the percentages of elderly nutritional supplement users reported using National Center for Health Statistics survey data (31, 32).
Additionally, the pattern of nutritional supplement use
in the MoVIES population is similar to the patterns
reported by other investigators (7, 31), with women
and persons of higher educational levels being the
most common users of antioxidants. Whites were also
more likely to use antioxidants than were blacks, but
this difference was not statistically significant—perhaps because the MoVIES sample contained only 33
blacks (3 percent of the total sample). Antioxidant use
was not related to age in this elderly cohort, even after
adjustment for education. If antioxidants are truly associated with decreased mortality, as has been suggested by the outcomes of several studies (26, 33-35),
then it would be expected that the "oldest old" would
include the highest proportion of antioxidant users.
There was also no significant association between
income and antioxidant use in the MoVIES population, perhaps because there is little variation in income
reported by MoVIES participants.
Antioxidant users scored better than nonusers on
several cognitive tests, but after data were controlled
for age, sex, and education, there were no statistically
significant relations between antioxidant use and cognitive test performance. One possible explanation for
this finding is that the relation between antioxidant
supplement use and cognitive functioning is confounded by factors related to use of these products.
Potentially, individuals in this sample who use antioxidant supplements are those who are less likely to be
cognitively impaired—i.e., women and those with
higher educational levels. Perhaps these individuals
are more able to keep abreast of popular literature on
the benefits of antioxidants because they are cognitively intact, or perhaps they can afford to purchase
these supplements because of their higher socioeconomic status (as reflected by educational level). Lack
of statistical power does not seem to explain these
results, since power analyses showed that this sample
provided adequate power to detect small differences in
cognitive test scores between antioxidant users and
nonusers.
To our knowledge, only one other study of antioxidants and cognitive impairment utilized a populationbased sample. Jama et al. (1) studied 5,182 elderly
persons in Rotterdam, The Netherlands, and found
that dietary and nutritional supplement intake of /3carotene was inversely associated with cognitive impairment, even after adjustment for age, sex, education, smoking, total caloric intake, and consumption of
other antioxidants. The discrepancy between the results of this study and those of the Rotterdam Study
could be due to differences between the study populations, such as a difference in age distribution or
socioeconomic status or the exclusion of demented
persons in the Rotterdam Study. Methodological discrepancies, including the fact that Jama et al. analyzed
dietary data in addition to data on nutritional supplements and analyzed the different antioxidants separately (1), could also explain the conflicting results.
Although Jama et al. found an association between
/3-carotene and cognition, they did not find similar
results with vitamin E or C.
Am J Epidemiol
Vol. 148, No. 1, 1998
Antioxidants and Cognitive Function in the Elderly
Previous studies of antioxidants and cognitive impairment used only a few neuropsychological measures (1, 2, 8, 11). This study analyzed data from 15
tests that were selected to assess performance in a
range of cognitive domains. This is an advantage of
this analysis, since the other studies might not have
been able to detect relations between antioxidants and
specific areas of cognition.
There are several potential limitations of this study.
Antioxidant consumption was self-reported, and data
on dosage and duration of antioxidant supplement use
were not collected. Although most products containing
antioxidants are marketed for daily use, there was no
verification of the frequency of use, and data were not
collected on dietary intake or on plasma levels of
antioxidants. If there is a dose-response effect for
antioxidants, or if the protective effect of antioxidants
is cumulative, it is likely that this cross-sectional study
would not have detected beneficial effects in those
persons with a limited duration of antioxidant use or
minimal antioxidant intake.
Bias would arise if cognitively impaired individuals
were less likely to accurately report their consumption
of supplements (11). To minimize this problem, we
compared medication labels with the self-reported data
and discussed medication use with informants whenever they were available. In addition, it may be possible that a beneficial effect of antioxidants on cognitive functioning is stronger among persons with
moderate to severe cognitive impairment than among
persons with mild cognitive impairment, the latter
group being more likely to be found in community
studies such as this one. However, in a recent clinical
trial carried out among patients with Alzheimer's disease of moderate severity, Sano et al. (36) did not find
that a-tocopherol (vitamin E) improved cognition.
In summary, even with the inflated rate of type I
error which would arise due to our examining individually the relation between antioxidant use and each of
15 cognitive tests, this cross-sectional analysis does
not support the hypothesis that antioxidant supplement
use is associated with cognitive function after controlling for the effects of age, sex, and education. Additional research on the benefit of antioxidant supplements in elderly populations is necessary to answer the
question of whether these products can legitimately be
used for primary or secondary prevention of cognitive
impairment.
ACKNOWLEDGMENTS
This study was supported in part by grants AG07562 and
AG00312 from the National Institute on Aging, US Department of Health and Human Services.
Am J Epidemiol
Vol. 148, No. 1, 1998
43
The authors are grateful to Drs. Graham Ratcliff and
Lewis Kuller for critical review of the manuscript. The
authors further acknowledge the support of the Mo VIES
Project staff and the cooperation of the Monongahela Valley
Community Health Center.
REFERENCES
1. Jama JW, Launer LJ, Witteman JC, et al. Dietary antioxidants
and cognitive function in a population-based sample of older
persons: The Rotterdam Study. Am J Epidemiol 1996;144:
275-80.
2. Kalmijn S, Feskens EJ, Launer LJ, et al. Polyunsaturated fatty
acids, antioxidants, and cognitive function in very old men.
Am J Epidemiol 1997; 145:33-41.
3. Gutteridge JM. Antioxidants, nutritional supplements and lifethreatening diseases. Br J Biomed Sci 1994;51:288-95.
4. Harman D. The aging process. Proc Natl Acad Sci U S A
1981;78:7124-8.
5. Evans PH. Free radicals in brain metabolism and pathology.
Br Med Bull 1993;49:577-87.
6. Meyers DG, Maloley PA, Weeks D. Safety of antioxidant
vitamins. Arch Intern Med 1996;156:925-35.
7. Schneider CL, Nordlund DJ. Prevalence of vitamin and mineral supplement use in the elderly. J Fam Pract 1983;17:
243-7.
8. Burns A, Marsh A, Bender DA. Dietary intake and clinical,
anthropometric and biochemical indices of malnutrition in
elderly demented patients and non-demented subjects. Psychol
Med 1989;19:383-91.
9. Zaman Z, Roche S, Fielden P, et al. Plasma concentrations of
vitamins A and E and carotenoids in Alzheimer's disease. Age
Ageing 1992;21:91-4.
10. Jeandel C, Nicolas MB, Dubois F, et al. Lipid peroxidation
and free radical scavengers in Alzheimer's disease. Gerontology 1989;35:275-82.
11. Goodwin JS, Goodwin JM, Garry PJ. Association between
nutritional status and cognitive functioning in a healthy elderly
population. JAMA 1983;249:2917-21.
12. Sram RJ, Binkova B, Topinka J, et al. Effect of antioxidant
supplementation in an elderly population. Basic Life Sci 1993;
61:459-77.
13. van Nostrand JF. Common beliefs about the rural elderly:
What do the national data tell us? Hyattsville, MD: National
Center for Health Statistics, 1993. (Vital and health statistics,
series 3: analytic and epidemiological studies, no. 28) (DHHS
publication no. (PHS) 93-1412).
14. Ganguli M, Belle S, Ratcliff G, et al. Sensitivity and specificity for dementia of population based criteria for cognitive
impairment: The MoVIES Project. J Gerontol 1993;48:M15261.
15. Ganguli M, Seaberg E, Belle S, et al. Cognitive impairment
and the use of health services in an elderly rural population:
The MoVIES Project. Monongahela Valley Independent Elders Survey. J Am Geriatr Soc 1993;41:1065-70.
16. Ganguli M, Ratcliff G, Huff FJ, et al. Effects of age, gender,
and education on cognitive tests in a rural elderly community
sample: norms from the Monongahela Valley Independent
Elders Survey. Neuroepidemiology 1991;10:42—52.
17. 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 Alzheimer's
disease. Neurology 1989;39:1159-65.
18. Folstein MF, Folstein SE, McHugh PR. "Mini-mental state": a
practical method of grading the cognitive state of patients for
the clinician. J Psychiatr Res 1975;12:189-98.
19. Kaplan E, Goodglass H, Weintraub S. The Boston Naming
Test. Philadelphia, PA: Lea and Febiger, 1983.
44
Mendelsohn et al.
20. Rosen WG, Mohs RC, Davis KL. A new rating scale for
Alzheimer's disease. Am J Psychiatr 1984;141:1356-64.
21. Benton AL, Hamsher KD. Multilingual Aphasia Examination.
Iowa City, LA: University of Iowa Press, 1976.
22. Freedman M, Leach L, Kaplan, et al. Clock Drawing: a
neuropsychological analysis. New York, NY: Oxford University Press, 1994.
23. Becker JT, Boiler F, Saxton J, et al. Normal rates of forgetting
of verbal and non-verbal material in Alzheimer's disease.
Cortex 1987,23:59-72.
24. Reitan RM. The relation of the Trailmaking Test to organic
brain damage. J Consult Psychol 1955;19:393-4.
25. Benton AL, Hamsher KD, Varney NR, et al. Contributions to
neuropsychological assessment: a clinical manual. New York,
NY: Oxford University Press, 1983.
26. Chen J, Geissler C, Parpia B, et al. Antioxidant status and
cancer mortality in China. Int J Epidemiol 1992;21:625-35.
27. Bettger WJ. Zinc and selenium: site-specific versus general
antioxidation. Can J Physiol Pharmacol 1993;71:721-4.
28. Stoehr GP, Ganguli M, Seaberg EC, et al. Over-the-counter
medication use in an older rural community: The MoVLES
Project. J Am Geriatr Soc 1997;45:158-65.
29. Kendall M. Time series. 2nd ed. London, England: Charles
Griffin and Company Ltd, 1976.
30. Hosmer DW, Lemeshow S. Applied logistic regression. New
York, NY: John Wiley and Sons, Inc, 1989.
31. Subar AF, Block G. Use of vitamin and mineral supplements:
demographics and amounts of nutrients consumed. The 1987
Health Interview Survey. Am J Epidemiol 1990;132:1091101.
32. Block G, Cox C, Madans J, et al. Vitamin supplement use, by
demographic characteristics. Am J Epidemiol 1988;127:297309.
33. Gaziano JM, Manson JE, Branch LG, et al. A prospective
study of consumption of carotenoids in fruits and vegetables
and decreased cardiovascular mortality in the elderly. Ann
Epidemiol 1995;5:255-60.
34. Kushi LH, Folsom AR, Prineas RJ, et al. Dietary antioxidant
vitamins and death from coronary heart disease in postmenopausal women. N Engl J Med 1996;334:1156-62.
35. Eichholzer M, Stahelin HB, Gey KF, et al. Prediction of male
cancer mortality by plasma levels of interacting vitamins:
17-year follow-up of the prospective Basel study. Int J Cancer
1996;66:145-50.
36. Sano M, Ernesto C, Thomas RG, et al. A controlled trial of
selegiline, alpha-tocopherol, or both as treatment for Alzheimer's disease. The Alzheimer's Disease Cooperative Study.
N Engl J Med 1997;336:1216-22.
Am J Epidemiol
Vol. 148, No. 1, 1998