Feskanich D, Singh V, Willett WC, Colditz GA. Vitamin A Intake and Hip Fractures Among Postmenopausal Women. JAMA. 2002;287(1):47–54. doi:10.1001/jama.287.1.47
Author Affiliations: Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School (Drs Feskanich, Willett, and Colditz), and Departments of Nutrition (Dr Willett) and Epidemiology (Drs Willett and Colditz), Harvard School of Public Health, Boston, Mass; and Roche Vitamins Inc, Human Nutrition Research, Parsipanny, NJ (Dr Singh).
Context Ingestion of toxic amounts of vitamin A affects bone remodeling and
can have adverse skeletal effects in animals. The possibility has been raised
that long-term high vitamin A intake could contribute to fracture risk in
Objective To assess the relationship between high vitamin A intake from foods
and supplements and risk of hip fracture among postmenopausal women.
Design Prospective analysis begun in 1980 with 18 years of follow-up within
the Nurses' Health Study.
Setting General community of registered nurses within 11 US states.
Participants A total of 72 337 postmenopausal women aged 34 to 77 years.
Main Outcome Measures Incident hip fractures resulting from low or moderate trauma, analyzed
by quintiles of vitamin A intake and by use of multivitamins and vitamin A
supplements, assessed at baseline and updated during follow-up.
Results From 1980 to 1998, 603 incident hip fractures resulting from low or
moderate trauma were identified. After controlling for confounding factors,
women in the highest quintile of total vitamin A intake (≥3000 µg/d
of retinol equivalents [RE]) had a significantly elevated relative risk (RR)
of hip fracture (RR, 1.48; 95% confidence interval [CI], 1.05-2.07; P for trend = .003) compared with women in the lowest quintile
of intake (<1250 µg/d of RE). This increased risk was attributable
primarily to retinol (RR, 1.89; 95% CI, 1.33-2.68; P
for trend <.001 comparing ≥2000 µg/d vs <500 µg/d). The
association of high retinol intake with hip fracture was attenuated among
women using postmenopausal estrogens. Beta carotene did not contribute significantly
to fracture risk (RR, 1.22; 95% CI, 0.90-1.66; P
for trend = .10 comparing ≥6300 µg/d vs <2550 µg/d). Women
currently taking a specific vitamin A supplement had a nonsignificant 40%
increased risk of hip fracture (RR, 1.40; 95% CI, 0.99-1.99) compared with
those not taking that supplement, and, among women not taking supplemental
vitamin A, retinol from food was significantly associated with fracture risk
(RR, 1.69; 95% CI, 1.05-2.74; P for trend = .05 comparing ≥1000
µg/d vs <400 µg/d).
Conclusions Long-term intake of a diet high in retinol may promote the development
of osteoporotic hip fractures in women. The amounts of retinol in fortified
foods and vitamin supplements may need to be reassessed.
Ingestion of toxic amounts of vitamin A has long been known to have
adverse skeletal effects.1,2 In
animals, retinoic acid suppresses osteoblast activity,3
stimulates osteoclast formation,4 and antagonizes
the ability of vitamin D to maintain normal serum calcium levels,5 all of which may contribute to the accelerated bone
resorption and fractures observed with hypervitaminosis A.6
In humans, long-term ingestion of high amounts of vitamin A can lead to hypercalcemia
and bone abnormalities.7 In addition, lower
bone mass8 and decreased biochemical markers
for bone turnover9 have been observed in patients
treated with retinoids for skin diseases.
These documented effects on bone from acute toxic exposure or chronic
high-dose intake of vitamin A have led to speculation that long-term consumption
of diets high in vitamin A may contribute to osteoporosis and hip fractures.
In Sweden, where both vitamin A intake and hip fracture incidence are high,
Melhus et al10 reported a 10% reduction in
femoral bone mineral density and a doubling of the risk of hip fracture among
women consuming more than 1500 µg/d of retinol compared with women who
consumed less than 500 µg/d. However, in contrast with these findings,
there was no evidence that vitamin A intake was associated with radial bone
loss among postmenopausal women participating in a 4-year calcium supplementation
trial,11 and neither serum retinol level nor
vitamin A supplement use was related to either radial bone mass or fracture
history among a sample of postmenopausal women in Iowa.12
To further investigate this issue, we conducted an 18-year prospective
analysis of vitamin A intake and hip fractures among postmenopausal women
in the Nurses' Health Study. We anticipated an increased risk of hip fracture
among the women with the highest retinol consumption from foods or supplements,
but no significant increase with higher carotenoid intakes since long-term
supplementation with beta carotene does not increase serum retinol concentrations.13
The Nurses' Health Study (NHS) began in 1976 when 121 700 female
registered nurses aged 30 to 55 years and living in 1 of 11 US states responded
to the initial mailed questionnaire. A medical history and information on
behaviors and lifestyle were collected at that time. Subsequent follow-up
questionnaires have been mailed every 2 years to update data, collect information
on new risk factors of interest, and identify incident diseases. Dietary data
were first collected in 1980. At least 90% of the cohort has responded in
each 2-year follow-up cycle. Deaths are confirmed through the National Death
Index. Approximately 98% of the cohort is white.
Only postmenopausal women (via natural or surgical menopause) who responded
to the dietary questionnaire in 1980 were included in this analysis. Women
entered analysis in 1980 or in the follow-up cycle in which they first reported
being postmenopausal. We excluded women at baseline with a previous hip fracture
or a diagnosis of cancer, heart disease, stroke, or osteoporosis because these
conditions could have caused a change in usual dietary habits. A total of
72 337 women ranging in age from 34 to 77 years contributed to this analysis.
On the 1982 questionnaire, participants were asked to report all previous
hip fractures along with the circumstances and the date of fracture. Incident
fractures were similarly reported on subsequent questionnaires. Only fractures
due to low or moderate trauma (eg, tripping, slipping, falling from the height
of a chair) were considered cases for this study. About 15% of the reported
hip fractures occurred with high-trauma events (eg, motor vehicle accidents,
skiing, horseback riding) and were therefore excluded. We anticipated a high
degree of accuracy for self-reported hip fractures in a cohort of registered
nurses. Indeed, in a small validation study of 30 reported hip fractures,
medical records confirmed all reports.14
Dietary intake was assessed in 1980, 1984, 1986, 1990, and 1994 with
a semiquantitative food frequency questionnaire (FFQ), which consisted of
a list of foods with a selection of 9 responses ranging from "never" to "6
or more times per day" for reporting the frequency of consumption of the specified
standard portion size. The FFQ food list began with 61 items in 1980 and increased
to 116 items in 1984 and to more than 130 items on the later FFQs. Nutrient
contents of foods were derived primarily from US Department of Agriculture
sources and supplemented with data from food manufacturers and published research.
Carotenoid contents were obtained from a US Department of Agriculture and
National Cancer Institute database.15,16
Use of brand-specific multivitamins and single vitamin or mineral supplements
also was reported and added to total nutrient intakes (assessment of beta
carotene supplement use did not begin until 1984, at which time it was used
by less than 1% of the cohort). The FFQ also requested specification of the
type of fat or oil and brand of margarine used in cooking and baking, and
this information was used to calculate nutrients in fried and baked foods.
Exposure data were derived for total vitamin A, retinol, and provitamin
A carotenoids, including beta carotene, alpha carotene, and beta cryptoxanthin.
All measures were adjusted for total energy intake using regression analysis.17 In these analyses, the vitamin A, retinol, and carotenoid
intakes were cumulatively updated during follow-up. That is, at the beginning
of each 2-year follow-up cycle, the nutrient intake was calculated as the
mean of all reported intakes up to that time. For example, at the beginning
of analysis in 1980, vitamin A intake was simply the value from the 1980 FFQ;
in the 2-year cycle beginning in 1994, vitamin A intake was calculated as
the mean of the values from the 1980, 1984, 1986, 1990, and 1994 FFQs.
Statistical models also included the following dietary variables that
are potential confounders of the association between vitamin A and hip fractures:
calcium, vitamin D, vitamin K, protein, alcohol, and caffeine. As was done
for the vitamin A and carotenoid measures, the covariate nutrient intakes
that are correlated with total energy intake (all except alcohol and caffeine)
were adjusted for total energy, and all were cumulatively updated during follow-up.
The food and nutrient assessments from the FFQ have been validated in
several substudies within the NHS cohort. In a comparison of the 1986 FFQ
with two 1-week diet records among 191 women, correlation coefficients were
0.79 for total vitamin A from food only and 0.59 for retinol from food only.18 In another substudy,19
the correlation between beta carotene intake from the 1986 FFQ and plasma
beta carotene levels among 186 women was 0.31 (P<.05).
Smoking status, number of cigarettes smoked per day, use of postmenopausal
hormones, and body weight were requested on all biennial questionnaires. Body
mass index was calculated at each 2-year cycle from the current weight and
the height reported on the 1976 questionnaire. The number of hours of physical
activity per week spent in leisure-time activities was reported in 1980, 1982,
1986, 1988, 1992, 1994, and 1996, and values were cumulatively updated in
analysis. Use of thiazide diuretics was assessed in 1980, 1982, 1988, 1994,
Person-time was accrued for each participant from the return date of
the 1980 questionnaire or the questionnaire on which she first reported being
postmenopausal until the occurrence of a hip fracture, death, or the end of
follow-up on June 1, 1998. The 72 337 women in this analysis contributed
a total of 860 355 person-years. For analyses of vitamin A, retinol,
and beta carotene from food sources only, women were excluded when they reported
taking a multivitamin or a specific vitamin A or beta carotene supplement.
These analyses included 34 386 women and 313 138 person-years. Current
exposure and covariate information was used to allocate person-time to the
appropriate category for each variable at the beginning of each 2-year follow-up
cycle. In the main analyses, women were categorized into quintiles of vitamin
A, retinol, and carotenoid intakes, and age-adjusted incidence rates were
calculated for each quintile. Relative risks (RRs) were then calculated as
the ratio of risk in each quintile compared with the risk in the first, or
referent, quintile. We used proportional hazards models to calculate multivariate
RRs, adjusting simultaneously for the potential confounding variables. To
examine linear trend, the vitamin A, retinol, and carotenoid exposure variables
were entered into the models as continuous values. All analyses were carried
out using SAS, version 6.12 (SAS Institute Inc, Cary, NC).
Table 1 outlines the changes
in vitamin A, retinol, and beta carotene intakes and in consumption patterns
of supplements and foods that were major contributors to vitamin A intake
in the NHS cohort during follow-up. Between 1980 and 1994, mean total retinol
intake decreased from 1378 to 1114 µg/d and mean total beta carotene
intake increased from 4278 to 5908 µg/d. Multivitamins were the primary
contributors to total retinol (35%-43% of intake), and carrots contributed
the most to total beta carotene intake (30%-41% of intake). Liver was the
primary food source of retinol, although its contribution declined between
1980 and 1994 while the percentages from milk and breakfast cereals increased.
In 1994, mean retinol intake limited to food sources was 546 µg/d,
similar to the 527 µg/d reported for women aged 51 to 70 years in the
1988-1994 Third National Health and Nutrition Examination Survey (NHANES III).20 Beta carotene intake from food in the NHS cohort
remained fairly stable during follow-up at about 4500 µg/d, much higher
than the mean intake of 2665 µg/d reported from NHANES III.
Multivitamins were used by 34% of the cohort in 1980; this increased
to 53% by 1996. Vitamin A supplement use remained between 3% and 5% during
the follow-up period. No more than 3% of the cohort used beta carotene supplements
between 1984 and 1992. This increased to 10% in 1994.
The recommended dietary allowance for vitamin A has recently been set
at 700 µg/d for women, with a tolerable upper limit of 3000 µg/d.20 During follow-up, about 14% of NHS women had intakes
below the recommended dietary allowance while 21% consumed more than the tolerable
Table 2 shows the age-standardized
characteristics of the study population by quintiles of vitamin A intake.
Women with higher vitamin A intakes, whether from food only or from food plus
supplements, were less likely to smoke and were more physically active. Those
who consumed more vitamin A from food had diets higher in calcium, protein,
vitamin D, and vitamin K and somewhat lower in alcohol. They were also more
likely to take a multivitamin, vitamin A, or beta carotene supplement. In
the highest quintile of vitamin A intake from food plus supplements, 67% of
the women were taking a multivitamin, compared with only 17% in the lowest
Among the postmenopausal women in the study population from 1980 to
1998, we identified 603 cases of nontraumatic hip fractures. The mean age
at fracture was 64 years.
Associations between hip fractures and cumulatively updated intakes
of vitamin A, retinol, and beta carotene are shown in Table 3. Because of a lower prevalence of other risk factors among
women with higher vitamin A intake, associations between vitamin A and risk
of hip fracture were substantially stronger after controlling for these in
multivariate analyses. For women in the highest quintile of vitamin A intake
from food plus supplements compared with those in the lowest quintile, the
multivariate RR of hip fracture was 1.48 (95% confidence interval [CI], 1.05-2.07)
and the linear trend was significant (P = .003).
For retinol from food plus supplements, the increase in risk was even greater
(RR, 1.89; 95% CI, 1.33-2.68; P for trend <.001).
For beta carotene, we observed only a weak and nonsignificant increase in
risk of hip fracture (RR, 1.22; 95% CI, 0.90-1.66; P
for trend = .10). To compare our results with those from Melhus et al,10 we reanalyzed our data comparing women with retinol
intakes of more than 1500 µg/d with women consuming less than 500 µg/d.
We found an RR of 1.64 (95% CI, 1.14-2.35) in the high-intake category, similar
to the odds ratio of 1.54 (95% CI, 1.06-2.24) reported by Melhus et al from
their model, which, like ours, controlled for calcium intake.
To assess risk of fracture from food sources only, we excluded women
from further follow-up when they reported taking a multivitamin or a specific
vitamin A or beta carotene supplement. These intakes from food do, however,
include vitamin A fortification of milk, margarine, and breakfast cereals.
Relative risks were elevated in the highest quintiles of vitamin A intake
from food (RR, 1.82; 95% CI, 0.97-3.40; P for trend
= .24) and retinol from food (RR, 1.69; 95% CI, 1.05-2.74; P for trend = .05), similar to the results we observed for intakes
from food plus supplements, despite the fact that the cut points for the highest
quintiles in the food analyses were lower than those used in the analyses
that included supplements. Beta carotene from food showed a weak and nonsignificant
positive association with fracture risk. Although beta carotene is the primary
dietary carotenoid with provitamin A activity, we examined intakes of alpha
carotene and beta cryptoxanthin and found no appreciable increase in risk
of hip fracture with higher intakes of either of these carotenoids (data available
on request from the authors).
Consumption of liver and carrots, the major food sources of retinol
and beta carotene, respectively, in this population, confirmed the results
we observed in the nutrient analyses. Women who consumed liver 1 or more times
per week had an elevated although nonsignificant RR of 1.69 (95% CI, 0.95-3.04)
for hip fracture compared with women who never consumed liver. For women who
consumed carrots 1 or more times per day compared with less than once per
week, the RR was 1.05 (95% CI, 0.61-1.82).
Using intake as a continuous variable, for every additional 500-µg/d
increase in retinol intake, hip fracture risk increased significantly: 15%
(95% CI, 8%-22%) for retinol from food plus supplements and 33% (95% CI, 9%-64%)
for retinol from food only. For a 2000-µg daily increase in beta carotene,
the increase in fracture risk was not statistically significant: 7% (95% CI, −1%
to 15%) for total beta carotene and 2% (95% CI, −10% to 17%) for beta
carotene from food only.
Results from analyses using only the 1980 baseline diet to characterize
nutrient intakes over the 18 years of follow-up were weaker than the results
from the Table 3 analyses in which
nutrient intakes were cumulatively updated over time. In these baseline analyses,
the RRs for hip fracture in the highest quintiles of intake were 1.10 (95%
CI, 0.82-1.47) for baseline vitamin A from food plus supplements and 1.17
(95% CI, 0.87-1.58) for baseline retinol from food plus supplements.
To examine the relative contributions of total retinol and beta carotene
to the risk of hip fractures, we put both nutrient intakes in the same multivariate
model. Results were similar to those reported from separate models (data available
on request from the authors). We also examined the association between beta
carotene and hip fractures within strata of retinol intake. Since absorption
of beta carotene and its conversion to retinol is greater when retinol intake
is low, any elevated risk of hip fracture with higher beta carotene intake
might be seen most clearly among women with the lowest retinol intakes. However,
among women in the lowest tertile of retinol intake from food plus supplements,
those in the highest compared with the lowest quintile of beta carotene intake
from food plus supplements had an RR of hip fracture of 0.94 (95% CI, 0.53-1.68).
We performed other stratified analyses by intakes of calcium (<800 vs ≥800
mg/d) and vitamin D (<5 vs ≥5 µg/d) and by smoking status (never,
past, or current) and observed no significant differences in associations
between hip fractures and vitamin A, retinol, or beta carotene intakes within
the strata (data available on request from the authors).
We also examined intakes of vitamin A, retinol, and beta carotene stratified
by postmenopausal hormone use. For retinol from food plus supplements, the
risk of hip fracture was significantly elevated only among the women not using
postmenopausal hormones (Table 4).
In comparison to the reference group of current hormone users in the lowest
quintile of retinol intake, the RRs in the highest retinol quintiles were
2.52 (95% CI, 1.48-4.31) among women not using postmenopausal hormones and
only 1.26 (95% CI, 0.68-2.33) among current hormone users. Unlike retinol,
higher beta carotene intake did not confer a greater risk of hip fracture
among the women not using postmenopausal hormones (data available on request
from the authors).
Associations between hip fractures and use of multivitamins, vitamin
A supplements, and beta carotene supplements are shown in Table 5. Risk of hip fracture was elevated with current use of either
vitamin A supplements (RR, 1.40; 95% CI, 0.99-1.99) or multivitamins (RR,
1.32; 95% CI, 1.04-1.67), although we did not observe a linear increase in
risk with longer duration of use for either one. Risk also was elevated among
past users, though time since last use was low: 41% of past multivitamin users
and 31% of past vitamin A supplement users had quit within the 2-year period
since the previous questionnaire. Multivitamins are a source of vitamin D
and calcium as well as retinol; therefore, these analyses were controlled
for total vitamin D and calcium intakes. Risk of hip fracture was even greater
for current vitamin A supplements users (RR, 1.75; 95% CI, 1.09-2.80) and
multivitamin users (RR, 1.59; 95% CI, 1.12-2.26) among women with low intakes
of retinol from food sources (<600 µg/d). Use of beta carotene supplements
did not confer any increase in risk of hip fractures.
In this prospective cohort study of postmenopausal women, the risk of
hip fracture was almost doubled among women with retinol intakes of about
2000 µg/d or more compared with those with intakes of less than about
500 µg/d. Current use of vitamin A supplements alone was associated
with a 40% (nonsignificant) increase in risk. In contrast with retinol, higher
intakes of beta carotene did not significantly increase the risk of hip fracture.
Our results are consistent with those reported by Melhus et al in Sweden.10 In that nested case-control study with 247 hip fracture
cases among women aged 40 to 76 years, the risk of fracture was doubled for
a retinol intake of greater than 1500 µg/d compared with an intake of
less than 500 µg/d, and risk was found to increase linearly (P = .006). With calcium in the model, the odds ratio was reduced to
1.54 but remained statistically significant. Using the same comparison groups,
we found a similar RR of 1.64.
Among the postmenopausal women in the NHS cohort, we observed a strong
positive association between retinol intake and hip fractures only among those
not currently using postmenopausal hormones. We had no prior hypothesis for
this finding; therefore, it needs to be examined in other studies.
Due to the high rate of absorption and the large storage capacity for
retinol in the human body,21 vitamin A toxicity
can result from acute ingestion of a very high dose, generally more than 100 000
IU,22 or from repeated exposure for several
weeks or months to lesser dosages (eg, between 25 000 and 50 000
IU/d)6 (1 IU = 0.3 µg of retinol). Vitamin
A toxicity has not been observed with excessive intakes of beta carotene,23 presumably because of limitations on its absorption
and conversion to retinol.24
Excess vitamin A is known to have teratogenic effects on bone growth
in mice25,26 and is likely to
affect human fetal bone development as well.27
In growing animals, hypervitaminosis A can produce bone fractures28- 30 and other skeletal
anomalies.31,32 Isolated cases
of hypervitaminosis A in humans were associated with bone pain, hypercalcemia,
and increased bone resorption.6 Also, patients
undergoing long courses of treatment with retinoids, such as isotretinoin
and etretinate for skin conditions, have experienced progressive calcification
of ligaments, modeling abnormalities of long bones, and osteoporosis.33,34 Direct effects of vitamin A on bone
osteoclasts and osteoblasts have been demonstrated in vitro that increase
bone resorption and decrease formation.3,4,29,35
If this mechanism occurs in vivo, long-term exposure to high vitamin A amounts
could plausibly lead to low bone density and fractures. This may be exacerbated
in older adults with diminished capacity to clear high levels of ingested
retinol.36 A deleterious effect of vitamin
A on bone may also operate through its antagonistic relationship with vitamin
D. In a recent study in humans, a vitamin A dose corresponding to about 1
serving of liver was shown to severely diminish the ability of vitamin D to
increase intestinal calcium absorption.37
Melhus et al10 noted that hip fracture
incidence in Europe is highest in Sweden and Norway, countries that also have
the highest intakes of vitamin A, primarily from fish oils and dairy foods.
In the United States, high consumption is also common and easily attainable.
Due to the loss of vitamin A with the removal of fat, skim and low-fat milk
are fortified to at least 350 µg/c (1500 µg/L). One tbsp (15 mL)
of fortified margarine also contains about 150 µg of vitamin A and breakfast
cereals are often fortified with up to 375 µg per serving. In addition,
multivitamins may contain 1500 µg of retinol per tablet, although in
more recent formulations part of the vitamin A content may be contributed
by beta carotene. It is also possible to purchase vitamin A supplements in
doses of 3000 µg.
Supplements contribute significantly to vitamin A intake in the United
States. In a sample of adult women in NHANES III, the mean intake of vitamin
A from supplements was 1338 µg/d,20 and
in the NHS cohort, more than 20% of the vitamin A in the diet was provided
by multivitamins and about 5% came from vitamin A supplements. However, multivitamins
contain other essential nutrients and have been associated with lower risk
of coronary disease,38 colon cancer,39 and breast cancer40
in this same cohort. Also, the findings from this study do not imply that
supplementation with vitamin A should be discontinued when used for conditions
such as retinitis pigmentosa, for which benefits have been documented,41 although monitoring of bone health may be appropriate.
This study was conducted in a mostly white population of women and results
may not be generalizable to other racial/ethnic groups. Results are also limited
to self-reported hip fractures with no further identification of exact fracture
site. The strengths of this study are the 90% follow-up rate over 18 years
and the repeated assessment of diet and supplement use. Results were greatly
attenuated when only the initial reported diet was used for analyses because
of real changes in diet over time and random error from a single measure,
which is reduced when dietary intakes are cumulatively averaged with each
Our findings provide further evidence that chronic intake of excessive
vitamin A, particularly from retinol, may contribute to the development of
osteoporotic hip fractures in women. The amounts of retinol in fortified foods
and vitamin supplements may need to be reassessed since these add significantly
to total retinol consumption in the United States.