American Journal of Epidemiology
Copyright O 2000 by The Johns Hopkins University School of Hygiene and Public Hearth
All rights reserved
Vol. 151, Mo. 8
Printed in U.SA.
Effect of Alcohol Intake on Bone Mineral Density in Elderly Women
The EPIDOS Study
Olivier Ganry,' Claude Baudoin,2 and Patrice Fardellone3 for the EPIDOS Group
To study potential associations between alcohol consumption and bone mineral density in women aged 75
years or older, the authors analyzed 7,598 ambulatory women (mean age, 79.9 years; standard deviation, 3.8
years) recruited at five centers in France between 1992 and 1994. The current alcohol intake was assessed
using a self-questionnaire. Bone mineral density was measured by dual-photon X-ray absorptiometry of the
proximal femur and total body and adjusted for age, weight, and height (Z score). Compared with nonusers,
women who drank 11-29 g of alcohol per day (g/day) had higher bone mineral density values at the trochanteric
site (p = 0.0017). Neither 1-10 g/day nor 530 g/day users had increased bone mineral density levels. These
results were unrelated to estrogen replacement therapy use, dietary calcium intake, current smoking status,
usual physical activity, educational attainment, household monthly income, and general health status. Alcohol
intake was not associated with bone mineral density at the femoral neck. Total body bone mineral density was
lower in subjects with alcohol intakes 530 g/day (p = 0.047). Our data suggest that moderate drinking (e.g., 1-3
glasses of wine per day) is associated with an increase in trochanteric bone mineral density in elderly
ambulatory women. However, higher intakes may have detrimental effects on bone mass. Am J Epidemiol
2000; 151:773-80.
aged; alcohol drinking; bone density; osteoporosis; women
Heavy drinking is well known to be associated with
osteoporosis and osteoporotic fractures in chronic alcoholics (1, 2). Epidemiologic studies (3-5) have shown
that long-term heavy drinkers have multiple risk factors
for bone loss, including low dietary calcium and other
nutritional deficiencies, low body weight, smoking, and
a high caffeine intake (6, 7). Although the detrimental
effects of alcohol on bone metabolism have been confirmed by animal studies (8, 9), most of these involved
administration of large amounts of alcohol. Several
recently published epidemiologic studies have suggested that moderate drinking in social settings may be
associated with higher bone mineral density values
(10-14). The results were somewhat conflicting, however, and definitions of alcohol intake categories lacked
accuracy and varied across studies. Also variable were
the accuracy of the alcohol intake questionnaires used
and the length of the study period. In most studies, the
sample size was small (6, 14-17), a fact that required
use of a limited number of alcohol intake categories.
Few studies specified the type of alcoholic beverage
used (wine, beer, or liquor), a potentially important
parameter since an estrogen-like compound has
recently been identified in some bourbon whiskeys
(18). Few studies included a large sample of women
(19, 20). As a rule, the statistical analysis did not
involve adjustment for the effects of potentially important confounding variables, such as smoking, dietary
calcium intake, estrogen replacement therapy, or physical activity. The rationale underlying the choice of
covariates for adjusting the crude data was unclear
(21). Finally, the anatomic sites of bone mineral density
measurement varied across studies.
Because postmenopausal women are the population
at highest risk for osteoporosis, we evaluated potential
associations between alcohol intake and bone mineral
density measured at the proximal femur and total body
in 7,598 women aged 75 years or older. These women
were participants in a prospective French study of
risk factors for hip fractures (Epiddmiologie de
TOsteoporose (EPIDOS) Study). Our analysis took
into account the main potential confounding factors
reported in the literature, namely, estrogen replacement therapy, smoking history, calcium intake, physi-
Received for publication October 5, 1998, and accepted for publication July 7, 1999.
Abbreviations: FF, French francs; EPIDOS, Epidemiologie de
I'0st6oporose.
1
Service conformation Medicate et d*Epidemiologie, CHU Hopital
Nord, Amiens, France.
2
INSERM Unite 349, Paris, France.
3
Service de Rhumatologie, CHU Hopital Nord, Amiens, France.
Reprint requests to Dr. Claude Baudoin, INSERM Unite 349, 2,
rue Ambroise Pare, 75475 Paris Cedex 10, France (e-mail:
brigitte.gouin © insenm.lrb.ap-hop-pa ris.fr).
773
774
Ganry et al.
cal activity, social status (educational attainment and
family income), and general health status (6, 14, 15).
MATERIALS AND METHODS
Recruitment
The recruitment procedure has been described previously (22, 23). In brief, between January 1992 and
January 1994, women aged 75 years or older from five
cities throughout France (Amiens, Paris, Lyon,
Toulouse, Montpellier) were randomly selected from
registries such as voting or healthcare lists. Each
woman was sent a letter inviting her to participate in a
study conducted at a university hospital in her city.
Based on findings from a history and physical examination, women with a history of hip fracture or bilateral
hip replacement, Paget's disease, renal failure, hyperthyroidism, or treatment for hypothyroidism were
excluded. Further letters were sent until 1,500 women
had been included at each center. The final study population was composed of 7,598 ambulatory women,
evenly distributed over the five centers. The baseline
evaluation was done at the center clinics by trained
nurses and included administration of a structured
questionnaire, a physical examination, and a functional
assessment. The study subjects completed the questionnaire at the hospital, during the recruitment visit, with
the help of a trained interviewer if necessary. The study
was approved by an ethics committee, and all women
gave their written informed consent.
Alcohol intake
Each woman completed a food consumption selfquestionnaire including 58 items of which the last five
were on alcohol consumption. One item (translated to
English) was "Do you drink wine, beer, or liquor?" and
was to be answered by "yes" or "no"; subjects who
answered "yes" completed an item asking how many
glasses of each alcoholic beverage they drank on average per day, during or outside meals. To obtain the
total daily alcohol intake, these answers were converted to grams as follows (24, 25): one glass of wine
(4 oz or 120 ml) = 10 g of alcohol; one glass of beer
(4 oz or 120 ml) = 4 g of alcohol; and one standard
shot of liquor = 10 g of alcohol.
Bone mineral density measurement
Bone mineral content was measured by dual-photon
X-ray absorptiometry using a Lunar DPX unit (Lunar
Corporation, Madison, Wisconsin) and Lunar version
3.61 software. The measurement sites were the total
body and the trochanter, femoral neck, and Ward's tri-
angle of the right femur. Lumbar spine bone mineral
density was not measured. The results were expressed
as bone mineral density in grams per square centimeter
obtained by dividing the bone mineral content by the
projected area of the region scanned. Bone mineral densities in the EPIDOS Study women had a normal distribution, consistent with data from other populations (26).
Covariates
Estrogen replacement therapy. Questionnaire items
on estrogen replacement therapy asked whether estrogens had ever been used to relieve menopausal symptoms, the brand(s) used, the age at which use was initiated, and the total number of years of use. The use of
hormones other than estrogens to alleviate menopausal
symptoms was not evaluated. Categories for duration
of estrogen replacement therapy use were never used,
<5 years, and £5 years.
Smoking status. Based on current smoking status,
patients were categorized as nonsmokers or smokers,
with the latter being defined as subjects who smoked
at least one cigarette per day. Because only 2 percent
of the study subjects smoked more than 20 cigarettes
per day, this category was not studied separately. No
information was collected on cigars and pipes, which
are rarely used by women in France.
Calcium intake.
The current dietary calcium
intake was assessed using a food frequency questionnaire. Dietary calcium intake was calculated from the
calcium content of each food, typical portion sizes,
and the frequency of consumption per week. In a previous study (27), we found that the calcium intake
estimates obtained using this questionnaire were
closely correlated with those derived from food frequency methods. Calcium was analyzed as a quantitative variable.
Physical activity. Usual physical activity was estimated by items on three types of activity, namely, shopping or walking outdoors to satisfy daily needs, exercise
(walking, working out, cycling, gardening, swimming),
and heavy housework (e.g., vacuum cleaning or mopping floors). These items were answered by "yes" or
"no." The women were categorized into four groups
based on whether they engaged in none, one, two, or
three of these activities.
Education. Educational attainment was categorized in four groups as follows: 5 years of formal education, more than 5 but less than 12 years, high school
graduate, and college education.
Family income.
Monthly family incomes were
evaluated in three groups (<6,000 French francs
(FF)/month, 6,000-8,500 FF/month, and >8,500
FF/month; equivalent approximately from 920 to
1,300 US dollars, with 1 dollar = 6.50 FF).
Am J Epidemiol
Vol. 151, No. 8, 2000
Alcohol Intake and Bone Density in Elderly Women
Health status. Self-rated health was classified in
three categories based on whether the response to the
question "Compared with women of your age, how do
you rate your health status?" was "better," "similar," or
"worse."
Estrogen replacement therapy use, current smoking
status, and usual physical activity were determined
during a face-to-face interview. Dietary calcium
intake, educational attainment, monthly household
income, and general health status were assessed from
the self-administered questionnaire. Height and weight
were measured while the women were dressed in light
clothes with shoes removed.
Statistical analysis
Bone mineral density measurements were adjusted for
the age, height, and weight of non-alcohol users according to a three-stage procedure previously used by several
authors (28). First, within the non-alcohol user subgroup,
we estimated the specific parameters of the multiple linear regression of bone mineral density according to age,
height, and weight. Second, for each woman, irrespective
of alcohol use, we calculated the difference between measured bone mineral density values and bone mineral density values predicted by multiple linearregression.Third,
we derived a standardized value, the Z score, by dividing
all these differences by the standard deviation of the differences in the non-alcohol user subgroup. Consequently,
the Z score in the non-alcohol user subgroup was on a
scale with zero mean and unit standard deviation. The
Z score measures the deviation of a woman's bone mineral density above or below the expected value for a
non-alcohol user of the same age, height, and weight.
The univariate analysis of the effect of alcohol use
alone on the Z score was evaluated using one-way
analysis of variance and multivariate analysis, using
multiple linear regression. In this last analysis, the statistical assessment of the alcohol effect was based on
the lack-of-fit test. If this test was significant, we partitioned the alcohol effect, first using cubic polynomial
contrast to check for a nonlinear distribution of the Z
score over alcohol intake and then using Dunnett's post
hoc test to compare each group of alcohol users with
the non-alcohol user group. Contingency tables were
analyzed using the chi-square test. Data were analyzed
using S-Plus software (29). The level of significance
was set at 0.05. All probability values were two-tailed.
RESULTS
Alcohol intake
Table 1 shows the distribution of alcohol intakes.
The range was 0-80 g/day, 60 percent of the women
Am J Epidemiol Vol. 151, No. 8, 2000
775
TABLE 1. Type and quantity of alcohol consumed by
EPIDOS* Study women.t France, 1992-1994
Type of alcohol (%)
Alcohol
intake
Light (1-10 g/day)
Moderate (11-29 g/day)
High (£30 g/day)
Aperitif
Wire*
Beer
and/or
hard liquor§
97.2
97.4
98.9
0.5
5.0
7.3
0.3
3.7
27.5
Distribution of alcohol Intakes!]
p/day
Light (1-10 g/day)
Moderate (11-29 g/day)
High £30 g/day)
No. of
women
Total no. of
women by
category
8
10
1,832
8
1,840
12
16
18
20
26
28
2
17
6
991
27
1
1,044
30
36
38
40
46
50
60
80
109
10
1
44
2
7
4
1
178
* EPIDOS, Epidemiologie de I'Osteoporose.
t In response to questions about alcohol consumption, 4,536
women did not drink alcohol, and only 23 women failed to answer.
$ The consumption of wine in glasses per day was 1 = 1,867
women, 2 = 996 women, 3 = 82 women, 4 = 41 women, 5 = 5
women, and 6 = 1 woman.
§ In France, aperitif drinks are commonly taken before meals,
supposedly to stimulate the appetite, and hard liquor after meals,
supposedly to aid digestion.
Tl The distributions display all values as they were collected,
without gathering.
reported no alcohol intake, and 3 percent drank more
than 30 g/day. Among the 3,062 alcohol users, 2,932
(96 percent) gave answers clustered on three values
(10 g/day, 20 g/day, and >30 g/day). The remaining
130 patients reported the following alcohol intakes:
1-9 g/day (n = 8), 11-19 g/day (n = 25), 21-29 g/day
(n = 28), and >30 g/day (n = 69). We consequently
created four categories: 1) no alcohol use (0 g/day); 2)
light drinking (1-10 g/day); 3) moderate drinking
(11-29 g/day); 4) heavy drinking (30 g/day). Table 1
also displays links between the amount and type of
alcohol consumed. Wine was the most frequently consumed alcoholic beverage: consumption of wine alone
was reported by 96.8 percent, 91.1 percent, and 65.7
percent of light, moderate, and heavy alcohol users,
respectively. Increases in alcohol intake from one category to the next were due primarily to consumption of
776
Ganry et al.
women at the higher end of the socioeconomic spectrum (based on educational attainment and family
income) increased with alcohol intake.
other alcoholic beverages in addition to wine (0.4 percent, 6.6 percent, and 33.1 percent of light, moderate,
and heavy users, respectively). Heavy users drank
wine and strong liquor (25.8 percent) or beer (5.6 percent). Among alcohol users who did not drink wine,
light users drank only alcoholic beverages traditionally
taken before or after meals (2.3 percent), whereas
moderate users drank only beer (1.7 percent).
Study patient characteristics are shown in table 2.
No significant relation was found between alcohol
intake and age, age at menopause, estrogen replacement therapy use, or physical activity. The percentages
of smoking and low calcium intake increased with the
alcohol intake. The heaviest drinkers were taller and
heavier than the other study subjects. The proportion
of women who felt their health status was better than
that of women of the same age and the proportion of
TABLE 2.
Bone mineral density
The unadjusted analysis found no significant effect
of alcohol consumption on bone mineral density measured at the proximal femur (table 3). After stratification for age and weight conducted by dividing patients
into two subgroups based on whether these two
parameters were above or below the median value, the
alcohol effect was statistically significant at the
trochanteric site (p = 0.032). A more powerful adjustment on age, weight, and height was then performed
using the Z score. This showed a highly significant
effect of alcohol intake at the trochanteric site (p =
Clinical characteristics of EPIDOS* Study women by alcohol Intake, France, 1992-1994
Mean
age
(years)
Mean
height
(cm)
ERT
Mean
*
U8era
meZmuse
(kg)
<722
(%)t
actiVity
menopause
b8tter
education)
FF'/
month)
(years)
Alcohol Intake
None (0 g/day; n = 4,536)
Light (1-10 g/day; n =
1,840)
Moderate (11-29 g/day;
n-1,044)
High (£30 g/day; n = 178)
Unadjusted p values
79.9 (3.8)§
153(6.0)
60.1(10.0)
48.7(5.9)
6.0
22
49.5
59.2
14.3
24.7
30.1
79.9(3.8)
154(6.0)
59.8(9.9)
48.8(5.0)
5.6
3.4
47.9
60.6
15.4
30.6
34.7
79.6(3.9)
79.9 (3.8)
153(6.0)
155 (6.1)
59.0(9.8)
60.3 (11.0)
48.7(5.9)
49.2 (5.7)
4.8
9.0
6.4
13.5
54.5
56.2
59.3
64.0
17.1
28.1
33.8
42.7
36.7
38.7
0.67
0.01
0.06
0.66
0.13
0.0001
0.002
0.45
<0.0001
<0.0001 <0.0001
• EPIDOS, EpWemtologle de I'Osteoporose; ERT, estrogen replacement therapy; FF, French francs,
t Durations of ERT use were as follows: <5 years, 68.5%; 5-9 years, 9.3%; £10 years, 22.4%.
X Median for all the women included in the study.
§ Numbers in parentheses, standard deviation.
TABLE 3.
Bone mineral content, according to alcohol Intake, In EPIDOSf Study women, France, 1992-1994
Bone mineral density (g/cm1)
Trochanter
Alcohol consumption
None (0 g/day)
Ught (1-10 g/day)
Moderate (11-29 g/day)
High (£30 g/day)
Nonadjusted p values
Adjusted p values
On age and weighty
On covariabtes#
0.638 (0.120)§
0.640(0.115)
0.642 (0.121)
0.636(0.112)
Femoral
neck
Ward's
triangle
0.716(0.114) 0.584 (0.129)
0.716(0.108) 0.583(0.123)
0.711 (0.116) 0.579(0.131)
0.714(0.110) 0.579(0.120)
Z sco ret
Total
body
0.971 (0.095)
0.968 (0.095)
0.962 (0.094)'
0.952 (0.088)'
0.67
0.55
0.62
0.007
0.032
0.94
0.97
0.046
Trochanter
Femoral
neck
0.0(1.0)
0.0(1.0)
0.047 (0.976) 0.008(0.958)
0.131 (1.059)' 0.000(1.049)
0.004 (0.982) -0.033 (0.979)
Ward's
triangle
Total
body
0.0(1.0)
0.0(1.0)
0.004(0.973) -0.023(1.021)
0.000(1.047) -0.031 (1.063)
-0.051(0.937) -0.213(0.958)'
0.0017
0.96
0.92
0.047
0.0004
0.86
0.70
0.039
* p< 0.05 tor the comparison of light moderate, and heavy drinkers with nondrlnkers, using a multiple comparison test (Durmetfs test), after adjustment
t EPIDOS, EptdemtotogJe de I'Osteoporose.
t Bone mineral density values were adjusted on age, height, and weight and then standardized to obtain a 0 mean wfth a standard deviation equal to 1 in the
nondrinker group (see Statistical analysis).
§ Numbers in parentheses, standard deviation.
U Adjusted on age and weight, with dfchotomization for both parameters based on whether the value was above or below the corresponding median (age <
80 years or 2 80 yeare; weight < 60 kg or £ 60 kg).
# Adjusted on estrogen replacement therapy, current smoking status, usual physical activity, dietary calcium Intake, educational attainment family monthly
income, and general health status.
Am J Epidemiol
Vol. 151, No. 8, 2000
Alcohol Intake and Bone Density in Elderly Women
0.0017). When the covariates were taken into account,
the effect of alcohol intake on the Z score at the
trochanteric site remained highly significant (p =
0.0004).
To determine whether the relation between the Z
score and the alcohol intake was linear, we fit a polynomial regression of degree 3. We found a significant
quadratic term (linear p = 0.54, quadratic p = 0.032,
cubic p = 0.054), suggesting a nonlinear relation. The
multiple comparison test of each of the three alcohol
user groups with the non-alcohol user group showed
that this effect existed only for the moderate drinkers
(11-29 g/day), who had the highest Z scores (table 3;
figure 1).
Neither femoral neck and Ward's triangle bone mineral densities nor their Z scores were significantly
linked to alcohol intake.
For total body bone mineral density, the difference
among alcohol intake groups was highly significant
before adjustment (p = 0.007), but it became less so
after adjustment on age and weight using the rough
adjustment procedure described above (p = 0.046) or
the Z score (p — 0.047). The multiple comparisons
tests showed that this difference was entirely ascribable to lower Z-score values in the heavy users (^30
g/day). After adjustment on the covariates, the difference remained close to the level of significance.
alcohol
ERT
not known
no
neck
ward
trochanter
total body
0
10
20
30
40
alcohol intake (grams / day)
FIGURE 1. Proximal femur and total body Z scores by alcohol
intake, Ranee, 1992-1994. To rate the relative magnitude of change
according to alcohol Intake, we juxtaposed the mean value in women
who did and did not use estrogen replacement therapy (ERT). BMD,
bone mineral density.
Am J Epidemiol
Vol. 151, No. 8, 2000
777
DISCUSSION
We found that moderate alcohol use (11-29 g/day)
was associated with a significant increase in trochanteric bone mineral density in elderly women. This
increase corresponded to one tenth of the standard deviation, which was similar to the increase seen in estrogen
replacement therapy used under the same study conditions (figure 1). Compared with nonusers, light and
heavy users showed no increase in bone mineral density.
The results suggest that the beneficial effect on bone
mineral density occurred only with the moderate alcohol intake provided by three glasses of wine, three
drinks of hard liquor, or almost eight beers per day.
Table 4 shows previously published results on the
link between moderate alcohol consumption and bone
mineral density among postmenopausal women. All
these studies measured bone mineral density at the
proximal femur, although one (11) did not measure the
trochanteric site. In three of these studies (11, 12, 14),
there was no significant relation between alcohol
intake and trochanteric bone mineral density.
However, the numbers of patients were small, so that
only two alcohol intake categories could be studied
(nonusers and users); furthermore, the statistical tests
used in these studies were not adjusted on covariates.
Finally, our results are very similar to those of the
Framingham Study (10), which included a large sample of women with approximately the same age distribution as in our sample. Thus, both the Framingham
and EPIDOS studies support a beneficial effect on
bone mass of moderate alcohol use.
As in the Framingham Study (10), we found that
age, weight, and height were significant covariates in
the analysis of the link between alcohol intake and
bone mineral density. Because this type of adjustment
may produce spurious positive associations, we
strengthened our conclusions by using two different
adjustment methods: the adjusted Z score, which is a
powerful tool but requires stringent statistical conditions of application, and a bone mineral density analysis with two age and two weight categories, which is a
less powerful but more robust test. The two methods
yielded similar results.
An important issue was to determine the threshold
or range of alcohol consumption likely to be beneficial
or detrimental for bone mass. In the Framingham
Study, bone mineral density was higher at the upper
end of the range of alcohol intake studied (207-414
ml/week). The broader alcohol intake range in our
study allowed us to show that the beneficial effect of
alcohol intake on bone mass was no longer present
above three glasses of wine per day.
There are a number of limitations to our study. The
EPIDOS Study population is composed of women who
778
Ganry et al.
TABLE 4. Epldemlologlc studies of associations between alcohol Intake and bone mineral density in women
SetBngt
of
study
19884:
Sydney, Australia
Rancho Bernardo, CA
1988-1991
Framlngham Study, MA 1988-1989
Copenhagen, Denmark
1989
Helsinki, Finland
1991$
EPIDOS44 Study, France
1998
Mean
age
(years)
59(1)§
60(8)§§
76(5)
63(2)
59(6)
80(4)
Bone mineral density (significance)
Alcohol use (%)
N on users
35H
Statistical
analysis
Users
36
71
196UU
384
325ttt
71
50§§§
37###
22
4,536
3,062
Comments#,tt
Comment**
Commenl44-l
Comment*
Commentsf.tt
Trochanter
NS#
NS
NS
**•
Femoral
neck
Ward's
triangle
NS
NS
NS
NS
NS
NS
NS
NS
NS
•
NS
Spine,
L2-L4
**
*
NS
*•
—
Radius
shaft
-n
NS
•
Nsum
—
—
Distal
radius
*
NS
*
—
—
—
* p < 0.05; •• p < 0.01; **• p < 0.001.
t Sydney, Australia (Felson et al., Am J Epidemiol 1995;142:485-92); Rancho Bernardo, CA (Hoibrook et al.. BMJ 1993,306:1506-9); Framlngham Study, MA
(Angus et al., Bone Miner 1988;4:265-77); Copenhagen, Denmark (Hansen et al., Osteoporos Int 1991 ;1:95-102); Helsinki, Finland (Laitinen et al., Catdf Tissue
Int 1991;48:224-31).
t Year of pubflcatlon.
§ Numbers in parentheses, standard deviation (SD).
D Nonusers and users, not defined. Mean consumption: 12 (SD, 6) g/day; range: 1-43 g/day.
#The analysis was not adjusted.
t t The analysis was based on the correlation coefficient (Pearson's or Spearman's).
$t NS, not significant; —, not performed; EPIDOS, Epidemlologie de rOsteoporose.
§§ Recruited over 45 years.
UD Nonusere, <0.1 g/week; users (three classes), according to the tertfles (48.7 g/week and 120.5 g/week).
## Adjusted for age, estrogen replacement therapy, body mass index, smoking, and exercise.
t t t Nonusers, <1 oz (29.57 ml/week); users (three classes), £1-3 oz (88.71 ml/week); >3-7 oz (206.99 ml/week); >7-14 oz (414 ml/week). There were no
women who consumed over 14 oz.
4=t=t Adjusted for age, age at menopause, number ol years of estrogen replacement therapy use, weight height, and mean number of cigarettes smoked.
§§§ Regular intake registered as weekly consumption of one or more units of alcohol for at least 1 year. None of the participants abused alcohol.
UUH A statistically significant difference (p < 0.01) In bone mineral content (In percentage) was found between the first measurement performed In 1979 and
the second measurement in 1989. This difference was highest In nonusers.
### Nonusers and users, not defined. Mean consumption: 36 g/week (95% confidence Interval: 11, 77); 26% consumed 220 g/week.
volunteered for participation in the study and may,
therefore, not be representative of the population at
large. Because we recruited only ambulatory elderly
women, the ranges of alcohol intakes and bone mineral
density values in our study sample may not reflect
those in the general population. Our study patients
were probably higher up in the socioeconomic spectrum and healthier than most women of the same age.
Recruitment bias may have occurred because the goal
of our study explained in the letter sent to potential
participants was perhaps unlikely to hold appeal for
alcoholic women. Nevertheless, the alcohol intakes in
our population were similar to those in earlier studies
of the elderly (10, 30).
The link between alcohol intake and bone mass
identified in our study may be questionable, however.
Subjects who lead an active life and have a good health
status have a higher bone mass, which in turn has been
shown to be associated with a longer life span (31).
The authors of the Framingham Study pointed out that
the bone mineral density increase in heavier drinkers
in their study may have been due to either alcohol
intake or the longer survival of women who were both
alcohol users and had high bone densities. As in earlier
studies (10, 32, 33), moderate or heavy users were
more likely than nonusers to rate their health as good
or excellent, to report exercising regularly, to have
higher incomes, and to have a better socioeconomic
status. In our study, however, the women with the
highest socioeconomic status and/or educational
attainment and those who reported the best general
health status had alcohol intakes above the range associated with benefits on bone mass.
Although our sample size was sufficient to allow
comparisons of nonusers, light users, and moderate
users, the statistical power of our study was limited
for the heavy user group. To study our alcohol users,
we used only three alcohol intake categories, corresponding approximately to one, two, and three
glasses of wine per day. This categorization corresponded exactly to the answers given by 96 percent
of alcohol users. We do not know whether the very
small proportion of answers indicating intake of
intermediate quantities was due to the form of our
questions (we asked how many glasses the respondents took per day ) or reflected drinking habits, particularly the custom of drinking wine by the glass. If
the first hypothesis is true, our questionnaire probably resulted in a number of random misclassifications
between two adjacent intake categories, which would
not affect our conclusions. Reported alcohol intakes
may have been lower than actual alcohol intakes, particularly in the heaviest drinkers. This would produce
a systematic misclassification, either toward the lowest alcohol intake categories or toward a virtual dose
considered "socially acceptable." If underreporting
of the alcohol intake increased as the intake itself
increased, then the difference between one intake category and the next would have been blunted, decreasing the likelihood of demonstrating a significant
association between intake and bone mineral density.
Another source of attenuation of the effect of intake
Am J Epidemiol
Vol. 151, No. 8, 2000
Alcohol Intake and Bone Density in Elderly Women
on bone mineral density may be the absence, in our
study, of data on past alcohol consumption, which did
not allow us to differentiate exusers from never users
in our nonuser category. Several studies have shown
that moderate users, particularly women, keep their
alcohol intake constant, whereas heavy users older
than 65 years of age tend to decrease their intake over
time (33). It has been shown that even a single
administration of questionnaires similar to ours provides useful estimates of alcohol intake over an
extended period of time (25).
The potential causal link between moderate alcohol
consumption and increased bone mineral density
deserves discussion. As seen in table 4, bone measurement sites varied across studies. The skeleton is
composed of trabecular and cortical bone. Trabecular
bone predominates at some sites (spine, 66 percent;
ultradistal radius, 66 percent; trochanteric region, 50
percent) and cortical bone at others (femoral neck, 75
percent; total body, 80 percent; distal radius, 80 percent; midradius, 99 percent) (34). In most studies,
moderate alcohol intakes were associated with
increased bone mass at the trochanter and spine,
which are predominantly trabecular. No studies found
any significant effect at the femoral neck. Effects on
midradius and total body bone mineral density were
less consistent. Taken in aggregate, these findings
suggest a protective effect of moderate drinking
against loss of trabecular bone. Because trabecular
bone has a higher rate of bone turnover and a faster
metabolic rate than cortical bone, the magnitude of
any effect on bone mass tends to be greatest at heavily trabecular sites. The bone mass-protecting effect of
moderate drinking may involve estrogen-like effects
on bone stimulation. Estrogen has well-known effects
on bone formation and remodeling. In experimental
studies (35) and clinical studies in postmenopausal
women (36, 37), moderate alcohol consumption
increased estradiol levels. In addition, alcohol may
affect osteoblasts directly (38) and/or indirectly via
other factors such as hormones (39).
In conclusion, our data suggesting that social drinking in elderly ambulatory women may have a positive
effect on bone mass at heavily trabecular sites such as
the trochanter are reassuring. The detrimental effect of
alcohol consumption on bone may occur only above
three glasses of wine per day. Multiple factors are
probably involved in the beneficial effects of moderate
alcohol consumption on bone metabolism. Alcohol in
modest amounts may have estrogen-like effects on
bone. However, daily alcohol intake is influenced by
many factors, including nutritional and physical activity habits, which may have their own effects on bone
mass.
Am J Epidemiol Vol. 151, No. 8, 2000
779
ACKNOWLEDGMENTS
This work was part of the EPIDOS Study supported by an
INSERM/MSD-Chibret contract.
The following investigators were members of the
EPIDOS Study Group: Coordinators: G. Breart and P.
Dargent-Molina (epidemiology), P. J. Meunier and A. M.
Schott (clinical work), D. Hans (bone densitometry and
ultrasound quality control), and P. D. Delmas (biochemistry); Principal Investigators (center): C. Baudoin and J. L.
Sebert (Amiens), M. C. Chapuy and A. M. Schott (Lyon), F.
Favier and C. Marcelli (Montpellier), C. J. Menkes, C.
Cormier, and E. Hausherr (Paris), and H. Grandjean and C.
Ribot (Toulouse).
REFERENCES
1. Diamond T, Stiel D, Lunzer M, et al. Ethanol reduces bone formation and may cause osteoporosis. Am J Med 1989;86:
282-8.
2. Schapira D. Alcohol abuse and osteoporosis. Semin Arthritis
Rheum 199O;19:371-6.
3. Saville PD. Changes in bone mass with age and alcoholism. J
Bone Joint Surg 1965;47:492-9.
4. Crilly RG, Anderson C, Hogan D. Bone histomorphometry,
bone mass and related parameters in alcoholic males. Calcif
Tissue Int 1988;43:269-76.
5. Bikle DD, Genant HK, Cann C, et al. Bone disease in alcohol
abuse. Ann Intern Med 1985; 103:42-8.
6. Laitinen K, Valimaki M. Alcohol and bone. Calcif Tissue Int
1991;49:S70-3.
7. Cummings SR, Kelsey JL, Nevitt MC, et al. Epidemiology of
osteoporosis and osteoporotic fractures. Epidemiol Rev 1985;
7:178-208.
8. Peng TC, Garner SC, Frye GD, et al. Evidence of a toxic effect
of ethanol on bone in rats. Alcoholism 1982;6:96-9.
9. Turner RT, Greene VS, Bell NH. Demonstration that ethanol
inhibits bone matrix synthesis and mineralization in the rat. J
Bone Miner Res 1987;2:61-6.
10. Felson D, Zhang Y, Hannan MT, et al. Alcohol intake and bone
mineral density in elderly men and women. Am J Epidemiol
1995; 142:485-92.
11. Holbrook TL, Barrett-Connor E. A prospective study of alcohol consumption and bone mineral density. BMJ 1993;306:
1506-9.
12. Angus RM, Sambrook PN, Pocock NA. Dietary intake and
bone mineral density. Bone Miner 1988;4:265-77.
13. Hansen MA, Overgraad K, Riis BJ, et al. Potential risk factors
for development of postmenopausal osteoporosis examined
over a 12-year period. Osteoporos Int 1991; 1:95-102.
14. Laitinen K, Valimaki M, Keto P. Bone mineral density measured by dual-energy X-ray absorptiometry in healthy Finnish
women. Calcif Tissue Int 1991;48:224-31.
15. Sowers MR, Clark MK, Hollis B, et al. Radial bone mineral
density in pre and perimenopausal women: a prospective study
of rates and risk factors for bone loss. J Bone Miner Res 1992;
7:647-57.
16. Kroger H, Tuppurainen M, Honkanen R, et al. Bone mineral
density and risk factors for osteoporosis. Calcif Tissue Int
1994;55:l-7.
17. Stevenson JC, Lees B, Devenport M, et al. Determinants of
bone density in normal women: risk factors for future osteoporosis. BMJ 1989;289:924-8.
18. Van Thiel DH, Galvao-Teles G, Monterio E, et al. The phytoestrogens present in de-ethanolized bourbon are biologically
780
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
Ganry et al.
active: a preliminary study in postmenopausal women.
Alcohol Clin Exp Res 1991;15:822-3.
Orwoll ES, Bauer DC, Vogt TM, et al. Axial bone mass in
older women. Ann Intern Med 1996;124:187-96.
Bauer D, Browner W, Cauley J, et al. Factors associated with
appendicular bone mass in older women. Ann Intern Med
1993;118:657-65.
Newcombe RG, Clements D, Evans W. Alcohol and bone density. (Letter). BMJ 1993;307:323.
Dargent-Molina P, Favier F, Grandjean H, et al. Fall-related
factors and risk of hip fracture: the EPIDOS prospective study.
Lancet 1996;348:145-9.
Hans D, Dargent-Molina P, Schott AM, et al. Ultrasonographic
heel measurements to predict hip fracture in elderly women:
the EPIDOS prospective study. Lancet 1996;348:511-14.
Fuchs CS, Stampfer MJ, Colditz GA, et al. Alcohol consumption and mortality among women. N Engl J Med 1995;332:
1245-50.
Giovannucci E, Colditz G, Stampfer MJ, et al. The assessment
of alcohol consumption by a simple self-administered questionnaire. Am J Epidemiol 1991;133:810-17.
Loocker AC, Wanner HW, Dunn WL, et al. Proximal femur
bone mineral levels of US adults. Oseoporos Int 1995;5:
389-409.
Fardellone P, Sebert JL, Bouraya M, et al. Evaluation de la
teneur en calcium du rdgime alimentaire par autoquestionnaire
fr6quentiel. (In French). Rev Rhum 1991;58:99-103.
Seeman E, Hooper JL, Bach LA, et al. Reduced bone mass in
daughters of women with osteoporosis. N Engl J Med 1989;
320:554-8.
29. S-PLUS 4 guide to statistics. Seattle, WA: Data Analysis
Products Division, MathSoft, 1997.
30. Goodwin JS, Sanchez CJ, Thomas P, et al. Alcohol intake in a
healthy elderly population. Am J Public Health 1987;77:
173-7.
31. Browner WS, Seeley DG, Vogt TM, et al. Non-trauma mortality in elderly women with low bone mineral density. Lancet
1991;338:355-8.
32. Andrews FK, Layne N. Drinking patterns in Canada: variations in drinking frequencies and demographic characteristics
of current drinkers. Can J Public Health 1985;76:38-42.
33. Adams WL, Cox NS. Epidemiology of problem drinking
among elderly people. Int J Addict 1995;30:1693—716.
34. Bonnick SL. Bone densitometry in clinical practice.
Application and interpretation. Current clinical practice.
Totowa, NJ: Humana Press, 1998.
35. Yamamoto A, Sekino A, Tajina M, et al. Effect of long-term
alcohol administration on bone metabolism in rats. J Nutr Sci
Vitaminol (Tokyo) 1997;43:369-75.
36. Gavaler JS, Deal SR, Van Thiel DH, et al. Alcohol and estrogen levels in postmenopausal women: the spectrum of effect
Alcohol Clin Exp Res 1993; 17:786-90.
37. Gavaler JS. Alcohol effects on hormone levels in normal postmenopausal women and postmenopausal women with alcohol
induced cirrhosis. Recent Dev Alcohol 1995; 12:199-205.
38. Purohit A, Flanagan AM, Reed MJ. Estrogen synthesis by
osteoblast cell lines. Endocrinology 1992; 131:2027-9.
39. Schweikert H, Wolf L, Romalo G. Oestrogen formation from
androstenedione in human bone. Clin Endocrinol 1995;43:
37^2.
Am J Epidemiol Vol. 151, No. 8, 2000