Full text of DOI:10.1359/jbmr.2002.17.4.709

JOURNAL OF BONE AND MINERAL RESEARCH
Volume 17, Number 4, 2002
© 2002 American Society for Bone and Mineral Research
Can Vitamin D Supplementation Reduce the Risk of
Fracture in the Elderly? A Randomized Controlled Trial
HAAKON E. MEYER,1,2 GURO B. SMEDSHAUG,1,2 ELISABETH KVAAVIK,1,2 JAN A. FALCH,3
AAGE TVERDAL,1 and JAN I. PEDERSEN2
ABSTRACT
Randomized controlled trials have shown that a combination of vitamin D and calcium can prevent fragility
fractures in the elderly. Whether this effect is attributed to the combination of vitamin D and calcium or to
one of these nutrients alone is not known. We studied if an intervention with 10 ␮g of vitamin D3 per day could
prevent hip fracture and other osteoporotic fractures in a double-blinded randomized controlled trial.
Residents from 51 nursing homes were allocated randomly to receive 5 ml of ordinary cod liver oil (n ؍ 569)
or 5 ml of cod liver oil where vitamin D was removed (n ؍ 575). During the study period of 2 years, fractures
and deaths were registered, and the principal analysis was performed on the intention-to-treat basis.
Biochemical markers were measured at baseline and after 1 year in a subsample. Forty-seven persons in the
control group and 50 persons in the vitamin D group suffered a hip fracture. The corresponding figures for
all nonvertebral fractures were 76 persons (control group) and 69 persons (vitamin D group). There was no
difference in the incidence of hip fracture (p ؍ 0.66, log-rank test), or in the incidence of all nonvertebral
fractures (p ؍ 0.60, log-rank test) in the vitamin D group compared with the control group. Compared with
the control group, persons in the vitamin D group increased their serum 25-hydroxyvitamin D concentration
with 22 nmol/liter (p ؍ 0.001). In conclusion, we found that an intervention with 10 ␮g of vitamin D3 alone
produced no fracture-preventing effect in a nursing home population of frail elderly people. (J Bone Miner
Res 2002;17:709–715)
Key words: frail elderly, hip fracture, osteoporosis, randomized controlled trial, vitamin D
In observational studies, low vitamin D intake and status
has been associated with increased risk of hip fracture, and
intake of cod liver oil has been associated with reduced risk
of hip fracture.(4,5) It also has been shown that vitamin D
supplementation reduces bone loss from the femoral neck in
postmenopausal women.(6,7) However, the effectiveness of
vitamin D in the prevention of osteoporotic fractures is not
clear.(8) In both a French and an American randomized
controlled trial a combined intervention with vitamin D3
and calcium significantly reduced the number of hip frac-
tures(9) and the number of all nonvertebral fractures,(9,10)
INTRODUCTION
IP FRACTURE constitutes a large health problem in the
elderly, and nursing home residents are especially
H
prone to these fractures.(1,2) Vitamin D deficiency may be
one important contributing risk factor for osteoporotic frac-
ture in old people, partly because of reduced food intake,
reduced sun exposure, and a decline in the capacity of
producing vitamin D in the skin.(3)
The authors have no conflict of interest.
1National Health Screening Service, Oslo, Norway.
2Institute for Nutrition Research, University of Oslo, Oslo, Norway.
3Hormone Laboratory and Department of Internal Medicine, Aker University Hospital, Oslo, Norway.
709

710
MEYER ET AL.
TABLE 1. BASELINE CHARACTERISTICS OF THE
1144 STUDY PARTICIPANTS
expectancy of more than half a year, should not be perma-
nently bedridden, and should not have difficulties in taking
medicine. Persons already taking vitamin D supplements
(usually in the form of a multivitamin supplement) were
included as long as this supplementation did not exceed 10
␮g/day. We found it unethical to require discontinuation of
the supplementation in this population of frail elderly peo-
ple. However, at the start of the study the use of vitamin D
supplements was registered carefully.
The study was approved by the regional ethics commit-
tee, Norwegian Board of Health and the Data Inspectorate.
Oral consent was considered satisfactory for this study with
cod liver oil, which has been used as a vitamin supplement
in the Norwegian population for decades. As described
elsewhere,(13) permission also was given to include men-
tally impaired persons in the study, a group that constitutes
an important part of the nursing home population.
Control group Vitamin D group
Characteristic
(n ϭ 575)
(n ϭ 569)
Age (year)a
85.0 Ϯ 7.4
76.9
59.4 Ϯ 13.7
163 Ϯ 9.3
22.3 Ϯ 4.5
446 Ϯ 196
84.4 Ϯ 7.3
74.9
60.2 Ϯ 14.3
164 Ϯ 9.5
22.3 Ϯ 4.6
456 Ϯ 196
Female (%)
Weight (kg)a
Height (cm)a
BMI (kg/m2)a
Calcium intake from
milk and cheese (mg)a
Vitamin D
43.0
3.4
39.9
2.4
supplementation (%)b
Calcium supplementation
(%)
At least one fall during
the last 3 months (%)
Previous hip fracture (%)
No
31.2
35.0
Conduct of the trial and supplements
66.8
28.6
4.6
67.3
26.4
6.3
Yes
The study started up at the different institutions over a
3-year period from autumn 1995. After careful instruction,
the staff at the institutions included study participants.(13)
Lists with all residents who were eligible and willing to
participate were returned to the coordinating office. The
intended complete record on those not included was not
reported in a satisfactory way. Before the study started, the
days of the month (1–31 days) were divided randomly into
group A and group B, and based on the day of birth, a
participant was placed automatically in group A or group B
when registered in the study database. The nursing staff was
not aware of the details in the allocation procedures. At
baseline, a simple questionnaire was filled in for each par-
ticipant by the nursing staff, including questions concerning
calcium intake from cheese and milk, the use of vitamin D
and calcium supplements, height and weight, previous hip
fracture, fall during the preceding 3 months, and mobility
status (use of cane, wheel chair, etc.).
Cod liver oil is rich in vitamins D and A. Vitamin D given
in oily suspensions generally is considered to be absorbed
almost completely. Approximately 80% of this fat-soluble
vitamin is incorporated into the chylomicron fraction and
absorbed through the lymphatic system.(14) Cod liver oil
also is a natural source of very long chain ␻-3 fatty acids.
The intervention was ordinary cod liver oil, whereas the
placebo was ordinary cod liver oil in which vitamin D had
been removed (Peter Møller, avd. av Orkla ASA, Oslo,
Norway). Vitamin D was extracted from the cod liver oil
using the techniques of bleaching earth, deodorization, and
molecular distillation. Because vitamin A also was removed
by this process, it was later added in the form of retinyl
Do not know
Dependent on gait
support (%)
No
34.8
44.7
20.6
33.1
42.9
24.1
Yes, cane or crutch
Wheelchair-user
a Mean values Ϯ SD.
b At baseline, usually in the form of a multivitamin supplement.
whereas in a Dutch randomized controlled trial supplemen-
tation with vitamin D3 alone did not reveal a protective
effect.(11) Further investigation is required to study whether
vitamin D alone can prevent osteoporotic fractures.(8) The
objective of this randomized controlled trial was to study
the effectiveness(8,12) of vitamin D in the prevention of
osteoporotic fractures in a population of frail elderly people,
that is, would an intervention with 10 ␮g of vitamin D3 per
day lead to reduced rates of hip fracture and all nonvertebral
fractures in a large, unselected group of institutionalized
elderly people compared with a similar placebo group. We
also planned to do subgroup analyses to study the effect of
vitamin D supplementation stratified on dietary calcium
intake and vitamin D intake at baseline.
MATERIALS AND METHODS
Subjects (study population)
We aimed to include as many as possible of the nursing palmitate to obtain the same levels of vitamin A as in
home residents in two Norwegian cities. Fifty-one out of untreated cod liver oil. Retinyl palmitate is the naturally
106 institutions were able and willing to participate. The occurring form of vitamin A in cod liver oil. In the cod liver
participation rate at the individual institution was on aver- oil in which vitamin D was not removed, the levels of
age 22.4% and somewhat higher at the wards because, vitamins A and D were controlled by the producer. There
frequently, not all wards at the institutions participated. The were no differences in taste, smell, or color, and the levels
study comprised 1144 residents. Mean age at baseline was of vitamin A, ␻-3 fatty acids, and energy were similar in the
84.7 years (SD Ϯ 7.4 years) and 75% of these were women two types of cod liver oil. Control analyses of the vitamin
(Table 1). To be eligible, the residents should have a life content by high-performance liquid chromatography

CAN VITAMIN D SUPPLEMENTATION PREVENT FRACTURE
711
showed trace amounts of vitamin D3 in the treated cod liver
oil (0.1–0.2 ␮g/ml) compared with 2.2 ␮g/ml in the un-
treated, that is, a difference of 10 ␮g per 5-ml dose.
During the 2 year study period, the intervention group
received 5 ml ordinary cod liver oil daily, whereas the
control group received 5 ml cod liver oil daily in which
vitamin D had been removed. The individual participants
either received group A cod liver oil or group B cod liver
oil. The study was double-blinded because neither the par-
ticipants nor the nursing staff or the investigators were
aware of which group contained vitamin D. In general, the
daily dose of cod liver oil was delivered together with other
medicines by the staff. A person from the coordinating
office visited each ward several times during the study and
had regular contact by telephone.
Statistical analyses
Based on previous studies,(5,13,16) we conservatively es-
timated that 6% of the nursing home population of Oslo
sustained a new hip fracture annually. A sample size of
1113 participants in each group thus would give the study
80% power to detect a reduction in hip fracture incidence by
30% at the 5% significance level. If hip fractures constitute
ϳ50% of all nonvertebral fractures in such a population,(9)
the study would have similar power to detect a 20% reduc-
tion in all nonvertebral fractures.
Differences between the two groups in baseline charac-
teristics and in treatment status were tested by the ␹2 test for
categorical variables and by the independent samples t-test
for means. Concerning changes in biochemical markers
from baseline to year 1, differences between the two groups
were tested by the independent samples t-test. In those
instances in which the criteria for normal distribution were
not fulfilled, the variables were log-transformed, and the
significance test was repeated. Because they gave similar
results, they were not reported.
Two types of fracture endpoints were defined in the
protocol: hip fracture (defined as cervical or trochanteric
fracture) and all nonvertebral fractures (including hip frac-
ture). The principal analysis of fracture data was made on
the intention-to-treat basis(12) using Kaplan-Meier analyses.
The two groups were compared by the log-rank test.(12)
Unadjusted and multivariate-adjusted hazard rate ratios, in
the text called relative risks, were calculated by the Cox
proportional hazards regression. In the multivariate model
controlling for potential confounding variables, all variables
listed in Table 1 except weight and height were included.
Censoring of data occurred only for people who died and
people lost to follow-up. Because of matching to hospital
registers, we had a complete follow-up of hip fracture.
Active-treatment analysis also was performed wherein all
persons were followed as long as they took the cod liver oil.
The two groups were compared by the log-rank test. Finally,
an analysis including only the people treated and followed
for the whole study period was undertaken, comparing the
Follow-up
All participants were followed with respect to hip frac-
tures, other nonvertebral fractures, and death during the
2-year study period.(13) Only fractures verified via hospital
discharge letters or X-ray descriptions were included in the
analysis. Follow-up with respect to fractures also was done
for withdrawals. At the end of the study, vital status (and
date of death when applicable) for all participants was
checked in the Norwegian registry of vital statistics.
To validate the reporting of hip fractures at the institu-
tions and to ensure a complete follow-up, the computerized
discharge registers in all hospitals serving the included
institutions were checked against the study database to see
if participants were admitted for hip fractures. Unverified
hip fractures were confirmed in the participant’s medical
records at the hospitals. The person responsible for this
checking was not otherwise involved in the study and had
no knowledge about the study participants. 22 hip fractures
were identified in this way, which constitute 22.5% of all
identified hip fractures. We were not able to follow 26
participants for all fractures when they moved to other
institutions. However, they were followed for hip fracture
using hospital registers.
2
two groups by the ␹ test.
When withdrawal occurred, the active-treatment period
was from the date at entry to the date of withdrawal.
The effects of the intervention on fractures and biochem-
ical tests were evaluated before unblinding of the data.
In the protocol it also was planned to do subgroup anal-
yses stratifying on total calcium intake from cheese and
milk as registered in the questionnaire at the start of the
study (higher or lower than median intake of 400 mg/day)
and the use of vitamin D supplements at the start of the
study (yes/no).
People treated for the whole study period were labeled
“treated for 2 years.” This group also includes 34 persons
who were included so late that they had not participated for
2 years when the study ended in the summer of 1999.
Biochemical measurements
Because the Data Inspectorate required written consent
from all persons in whom a blood sample was drawn,
mentally impaired persons were excluded from this proce-
dure. In general, it was difficult to recruit persons for blood
tests, and we ended up with a smaller than intended group,
which was on average 2.2 years younger (p ϭ 0.02) and had
a lower mortality during the study period (p ϭ 0.000)
compared with the rest. The blood samples were analyzed
for 25-hydroxyvitamin D (calcidiol), osteocalcin, parathy-
roid hormone (PTH), and ionized calcium. All analyses
were performed by the Hormone Laboratory, Aker Hospital,
Oslo, Norway. The methods are described elsewhere.(15)
The blood samples were drawn at baseline and at 1 year
RESULTS
Five-hundred seventy-five residents were allocated to the
after the study started (Ϯ1 month), regardless of whether control group and 569 residents were allocated to the vita-
they still took the treatment or not. min D group. There were no statistically significant differ-

712
MEYER ET AL.
TABLE 2. TREATMENT STATUS (1144 PARTICIPANTS)
1.09 (95% CI, 0.73–1.63), and multivariate adjustment did
not change this estimate. Similar results were found for
active-treatment analysis (p ϭ 0.55, log-rank test) and in the
analysis restricted to participants treated for the whole study
Control group
n (%)
Vitamin D group
n (%)
2
period (p ϭ 0.97, ␹ test). There was no effect of the
Treated for 2 years
Cessation of treatment
Deatha
186 (32.3)
197 (34.6)
intervention with vitamin D on all nonvertebral fractures
(Fig. 1). Thus, the relative risk of any nonvertebral fracture
was 0.92 (95% CI, 0.66–1.27) in the vitamin D group
versus the control group, and multivariate adjustment had
negligible effect on the estimate. Similar results were found
for active-treatment analysis (p ϭ 0.67, log-rank test) and in
the analysis restricted to participants treated for the whole
163 (28.3)
226 (39.3)
575
169 (29.7)
203 (35.7)
569
Other reasonsb,c
Total
a Treated on average 10.2 months.
b Treated on average 9 months.
c Reasons for cessation of treatment: no longer wanted to receive
the treatment (n ϭ 184); the ward discontinued to participate in
the study (n ϭ 87); the participant stopped taking the treatment
because of illness; poor appetite; nausea, etc. (n ϭ 117); the
participant not taking vitamin D supplements at entry started to
2
study period (p ϭ 0.41, ␹ test).
Concerning hip fracture, subgroup analyses showed that
an effect of the intervention with vitamin D was evident
neither in those with a higher than median intake (p ϭ 0.60,
log-rank test) nor in those with a lower than median intake
of calcium at baseline (p ϭ 0.31, log-rank test). Likewise,
there was no effect in those not taking (p ϭ 0.18, log-rank
test) or in those taking (p ϭ 0.30, log-rank test) a vitamin D
supplement at the start of the study. Similar results were
take vitamin D supplements during the study (n
ϭ 7); the
participant moved to an other institution (n ϭ 26). For 8 persons
the exact reason for withdrawal was not stated.
ences between the two groups in baseline characteristics found for all nonvertebral fractures. For example, in those
(Table 1). not taking vitamin D supplements at baseline, 42 persons in
Of those included in the study, 383 persons received cod the control group and 40 persons in the vitamin D group
liver oil throughout the study period, 332 persons discon- were identified with any nonvertebral fracture during
tinued the treatment because of death (after an average follow-up, with a corresponding relative risk of 0.93 (95%
treatment period of 10.2 months), and 429 persons stopped CI, 0.61–1.44, vitamin D group vs. control group).
taking the treatment for a variety of other reasons (after an
average treatment period of 9 months; Table 2). Of these
429 persons, 159 died after they had stopped taking the
treatment and before end of follow-up. There was no dif-
DISCUSSION
ference in mortality between the control group and the
vitamin D group (p ϭ 0.87, log-rank test), and there was no
In this study, performed in a population of frail nursing
significant difference in the proportions who stopped taking home residents, an intervention with 10 ␮g (400 IU) of
the treatment for other reasons and in the proportions who vitamin D neither prevented hip fracture nor all nonverte-
were treated for 2 years. The study participants were fol- bral fractures. Subgroup analyses stratified on dietary cal-
lowed for 1728 person-years (intention-to-treat), and they cium intake and vitamin D intake at baseline gave similar
received cod liver oil for a period of 1365 person-years or in results. Our study is in concert with a large Dutch random-
79% of the observation time.
ized controlled trial intervening with 10 ␮g of vitamin
Results from serum biochemical tests at baseline and after D/day alone as tablets over 3–3.5 years in a somewhat
1 year are shown in Table 3. One-third had a serum con- healthier study population than ours (mean age, 80 years;
centration of 25-hydroxyvitamin D Ͻ 30 nmol/liter, and participants living independently, in apartments or homes
15% had Ͻ20 nmol/liter at baseline. During the first year of for the elderly).(11) In contrast to the lack of effect of
the study, persons in the intervention group increased their vitamin D alone reported in these studies, a fracture-
25-hydroxyvitamin D concentration by 22 nmol/liter (95% preventing effect of an intervention with vitamin D and
CI, 9.0–35.7) compared with the control group. There was calcium combined has been reported in a large French
no statistical significant difference between the two groups randomized controlled trial over 18 months(9) and a smaller
concerning change in serum PTH, serum ionized calcium, American randomized controlled trial over 3 years.(10) The
or serum osteocalcin.
participants in the French study were institutionalized and
During follow-up, 99 hip fractures were identified in 97 on average 84 years old, whereas the participants in the
persons and 169 nonvertebral fractures (hip fractures in- American study were on average 71 years old and were
cluded) were identified in 145 persons. Forty-seven persons living in the community. In both studies, a higher dose of
in the control group and 50 persons in the vitamin D group vitamin D was used (17.5–20 ␮g/day), and the calcium
suffered a hip fracture. The corresponding figures for all supplementation ranged from 500 to 1200 mg/day. Al-
nonvertebral fractures were 76 persons in the control group though it cannot be excluded that the higher vitamin D dose
and 69 persons in the vitamin D group (Table 4). There was is responsible for these differences, a study in elderly
no difference in the incidence of hip fracture between the women and men showed that a daily supplementation with
control group and the vitamin D group according to the 10 ␮g of vitamin D clearly increased 25-hydroxyvitamin D
intention-to-treat analysis (Fig. 1). The relative risk of hip in serum, whereas only a marginal additional effect was
fracture in the vitamin D group versus the control group was seen in those randomized to receive a dose of 20 ␮g/day.(17)

CAN VITAMIN D SUPPLEMENTATION PREVENT FRACTURE
713
TABLE 3. SERUM BIOCHEMICAL VALUES IN THE CONTROL GROUP AND VITAMIN D GROUP AT BASELINE AND AFTER 1 YEARA
Change
Baseline
1 Year
(1 year–baseline)
Serum 25-hydroxyvitamin D (nmol/liter)
Control group (n ϭ 31)
Vitamin D group (n ϭ 34)
Serum PTH (pmol/liter)
51 Ϯ 33
47 Ϯ 26
46 Ϯ 20
64 Ϯ 21
Ϫ5 Ϯ 28
ϩ17 Ϯ 26*
Control group (n ϭ 34)
5.6 Ϯ 3.3
6.5 Ϯ 3.1
7.2 Ϯ 4.7
7.5 Ϯ 4.2
ϩ1.6 Ϯ 3.2
ϩ1.0 Ϯ 2.5
Vitamin D group (n ϭ 36)
Serum ionized calcium (mmol/liter)
Control group (n ϭ 26)
Vitamin D group (n ϭ 31)
Serum osteocalcin (nmol/liter)
Control group (n ϭ 33)
1.24 Ϯ 0.04
1.23 Ϯ 0.05
1.24 Ϯ 0.06
1.23 Ϯ 0.06
Ϫ0.001 Ϯ 0.05
ϩ0.003 Ϯ 0.06
2.0 Ϯ 0.9
2.1 Ϯ 1.0
1.5 Ϯ 0.7
1.6 Ϯ 0.8
Ϫ0.5 Ϯ 0.8
Ϫ0.5 Ϯ 0.9
Vitamin D group (n ϭ 35)
a Mean values Ϯ SD.
* p ϭ 0.001 for the comparison between the study groups.
TABLE 4. NUMBER OF FIRST HIP FRACTURE AND NUMBER OF three nursing homes in Oslo, of which two participated in
a
FIRST NONVERTEBRAL FRACTURE IN THE
1144 STUDY PARTICIPANTS
our study, reported a median nutritional intake of vitamin D
of 4.5 ␮g/day in women and 5.3 ␮g/day in men, and that
additional supplementation with vitamin D was frequent.(19)
The serum 25-hydroxyvitamin D concentrations at baseline
in our subsample also suggest that serious vitamin D defi-
ciency was not very frequent in our study population. How-
ever, because we were not able to take blood tests on a
random sample of all participants, we cannot exclude the
possibility that the blood test subsample, which was some-
what younger and had lower mortality, had better vitamin D
status than the rest of the study population. It also should be
added that it might be problematic to compare 25-
hydroxyvitamin D levels in this study with other studies
because of interlaboratory variation.(20)
Serum 25-hydroxyvitamin D increased in the vitamin D
group compared with the control group. However, we did
not find a corresponding decrease in PTH, and the reason for
this can only be speculated on. We did not measure renal
function or 1,25-dihydroxyvitamin D, the metabolite acting
directly on the parathyroid gland.
It was of great importance to design a study that was
workable and demanded little extra effort from the nursing
staff because much of the practical work was carried out by
the employees at the large number of participating institu-
tions. Therefore, the practical procedures were as simple as
possible. Because milk and cheese constitute two-thirds of
the total calcium intake in an average Norwegian diet,(13)
calcium intake was estimated based on a few questions on
milk and cheese intake. Therefore, the categorization of the
participants into high/low intake might be subject to non-
differential misclassification, which could reduce the effect
of calcium in the analyses.
Control
group
Vitamin D
group
Hip fractureb
47
76
7
7
45
17
50
69
3
8
47
11
All nonvertebral fractures
Radius or ulna
Proximal humerus
Hipb
Other
a In addition, 16 persons had two fractures and 4 persons had
three fractures.
b Two persons in the control group and 3 persons in the vitamin
D group had a fracture other than hip fracture first.
Approximately 80% of the participant in our study lived
in Oslo, Norway, the city with the highest incidence rates of
hip fracture ever reported internationally.(16) It has been
estimated that the annual incidence of hip fracture in the
nursing home population of Oslo exceeds 7 per 100 per
year.(13) The incidence in our study population was 5.9 per
100 per year, which is twice as high as in the Dutch study
by Lips et al.,(11) and also higher than in the French study by
Chapuy et al.(9) In this frail population with a very high risk
of fracture, we did not find any effect of the intervention
with 10 ␮g of vitamin D. On the other hand, ϳ40% of the
participants took vitamin D supplements (usually in the
form of multivitamin supplements containing 10 ␮g of
vitamin D) at baseline. However, there was no indication of
a different effect of the intervention in those not taking
supplements compared with those taking supplements at
baseline.
Because cod liver oil also contains vitamin A, both the
control group and the vitamin D group received vitamin A
(0.26 mg/ml). For a long time, it has been suggested that
high intake of vitamin A is harmful for skeletal health, and,
Previous studies in Oslo, Norway have suggested that
poor vitamin D status in the elderly is prevalent and that low
vitamin D intake is associated with increased risk of hip recently, (after the start of this study) it has been questioned
fracture.(5,18) On the other hand, a recent study performed in whether an adverse effect also might be present at moder-

714
MEYER ET AL.
groups was doing better than the other and that such antic-
ipation would lead to different treatment and follow-up.
However, usually, the number of participants at each ward
was small, making comparison less likely. At the coordi-
nating office, data collection from the institutions and eval-
uation of fractures were done by different persons. In addi-
tion, the identification of fractures via the computerized
discharge registers at the hospitals enabled us to validate the
recording of hip fractures at the institutions independently.
Concerning compliance, the nursing staff was instructed
to register when and why a participant stopped taking the
treatment. This also was asked for when a person from the
coordinating office visited each ward several times during
the study. In a survey made at the end of the study at 63 of
the participating wards, 95% of the wards stated that the
participants had received the treatment regularly.(13) The
change in 25-hydroxyvitamin D also indicated an effect of
the intervention.
Originally, the study was planned to include nursing
homes only in Oslo, Norway. In spite of hard effort, the
recruitment was lower than anticipated, and the study was
expanded to include nursing home residents in Bergen, the
second largest city in Norway, and in Lier, a neighboring
municipality to Oslo. Even so, our primary goal of including
2250 participants was unrealistic. The closing date for treat-
ment was set to June 30, 1999. It could be argued that we
might have missed a clinical interesting difference because
of reduced power. On the other hand, the lack of any
suggestion of effect in this study and in the study from
Amsterdam strongly suggests that an intervention with 10
␮g of vitamin D alone does not have a substantial fracture-
preventive effect. It should be added that serious vitamin D
deficiency was uncommon in these two study populations,
and the results do not rule out the possibility of a fracture-
preventing effect of an intervention with vitamin D alone in
elderly individuals with poor vitamin D status.
FIG. 1. Cumulative percentage with (A) incident hip fracture and (B)
nonvertebral fracture in the control group and in the vitamin D group.
Intention-to-treat analyses.
In conclusion, in this randomized controlled trial we did
not find that an intervention with 10 ␮g of vitamin D over
ately increased intakes.(21,22) On the other hand, in a recent 2 years could prevent hip fracture and other osteoporotic
randomized controlled trial, one group received calcium and fractures in institutionalized elderly people.
vitamin D and the other received calcium and vitamin D as
a multivitamin formulation including 0.8 mg of retinol.(23)
No difference between the two groups, regarding bone
mineral density (BMD) of the spine or hip, was found. In
our study, the incidence of hip fracture was lower than the
estimated incidence in the general nursing home population,
which does not support a deleterious skeletal effect of the
supplemented vitamin A.
Our method of randomization by day of birth may be
defined as quasi-randomization.(8) However, because it was
double-blinded which group received the vitamin D cod
liver oil, it is difficult to see how our procedure of random-
ization could introduce selection bias. Concerning conceal-
ment of allocation, no judgment was made either by the
nursing staff or by anyone at the coordinating office on
whether a person should participate or not in light of which
group he or she belonged to. The only potentially serious
problem concerning our allocation procedure, as far as we
see it, was that because of the open grouping of participants,
nursing staff, etc. could get the impression that one of the
ACKNOWLEDGMENTS
We thank Thale Briseid and Anne-Marie Aas for their
contribution in the preparation of the study, Hilde and Roger
Antonsen for contribution in the data collection, Egil Haug
who was responsible for the blood analyses, Cathrine
Lofthus who checked the computerized discharge registers
for hip fracture, and the nursing staff and nursing home
residents who participated in the study. The trial was eco-
nomically supported by Peter Møller, avd. av Orkla ASA,
Norwegian Dairies, the City Council of Oslo, and Odd
Fellow. Cod liver oil was donated by Peter Møller, avd. av
Orkla ASA, Oslo, Norway.
REFERENCES
1. Johnell O, Sernbo I 1986 Health and social status in patients
with hip fractures and controls. Age Ageing 15:285–291.

CAN VITAMIN D SUPPLEMENTATION PREVENT FRACTURE
715
2. Cumming RG 1996 Nursing home residence and risk of hip 14. Holick MF 1998 Vitamin D: Intestinal absorption. In: Shils
fracture. Am J Epidemiol 143:1191–1194.
ME, Olson JA, Shike M (eds.) Modern Nutrition in Health and
Disease. Lea & Febiger, Philadelphia, PA, USA, pp. 335.
15. Falch JA, Steihaug S 2000 Vitamin D deficiency in Pakistani
premenopausal women living in Norway is not associated with
evidence of reduced skeletal strength. Scand J Clin Lab Invest
60:103–109.
16. Falch JA, Kaastad TS, Bohler G, Espeland J, Sundsvold OJ
1993 Secular increase and geographical differences in hip
fracture incidence in Norway. Bone 14:643–645.
17. Lips P, Wiersinga A, van Ginkel FC, Jongen MJ, Netelenbos
JC, Hackeng WH, Delmas PD, van der Vijgh WJ 1988 The
effect of vitamin D supplementation on vitamin D status and
parathyroid function in elderly subjects. J Clin Endocrinol
Metab 67:644–650.
18. Mowe M, Bohmer T, Haug E 1998 Vitamin D deficiency
among hospitalized and home-bound elderly. Tidsskr Nor Lae-
geforen 118:3929–3931.
19. Kvaavik E 1997 En forundrsøkelse til prosjektet “Forebyg-
gelse av lårhalsbrudd.” Inntak av kalsium og vitamin D på
sykehjem i Oslo. Institute for Nutrition Research, University
of Oslo, Oslo, Norway.
20. Lips P, Chapuy MC, Dawson-Hughes B, Pols HA, Holick MF
1999 An international comparison of serum 25-
hydroxyvitamin D measurements. Osteoporos Int 9:394–397.
21. Melhus H, Michaelsson K, Kindmark A, Bergstrom R, Holm-
berg L, Mallmin H, Wolk A, Ljunghall S 1998 Excessive
dietary intake of vitamin A is associated with reduced bone
mineral density and increased risk for hip fracture. Ann Intern
Med 129:770–778.
22. Binkley N, Krueger D 2000 Hypervitaminosis A and bone.
Nutr Rev 58:138–144.
23. Bæksgaard L, Andersen KP, Hyldstrup L 1998 Calcium and
vitamin D supplementation increases spinal BMD in healthy,
postmenopausal women. Osteoporos Int 8:255–260.
3. Feldman D, Malloy PJ, Gross C 1996 Vitamin D: Metabolism
and action. In: Marcus R, Feldman D, Kelsey J (eds.) Osteo-
porosis. Academic Press, San Diego, CA, USA, pp. 205–235.
4. Falch JA, Mowe M, Bøhmer T, Haug E 1992 Serum levels of
intact parathyroid hormone in elderly patients with hip fracture
living at home. Acta Endocrinol (Copenh) 126:10–12.
5. Meyer HE, Henriksen C, Falch JA, Pedersen JI, Tverdal A
1995 Risk factors for hip fracture in a high incidence area: A
case-control study from Oslo, Norway. Osteoporos Int 5:239–
246.
6. Ooms ME, Roos JC, Bezemer PD, van der Vijgh WJ, Bouter
LM, Lips P 1995 Prevention of bone loss by vitamin D
supplementation in elderly women: A randomized double-
blind trial. J Clin Endocrinol Metab 80:1052–1058.
7. Dawson-Hughes B, Harris SS, Krall EA, Dallal GE, Falconer
G, Green CL 1995 Rates of bone loss in postmenopausal
women randomly assigned to one of two dosages of vitamin D.
Am J Clin Nutr 61:1140–1145.
8. Gillespie WJ, Henry DA, O’Connell DL, Robertson J 1997
Vitamin D and vitamin D analogues in the prevention of
fractures in involutional and post-menopausal osteoporosis. In:
Gillespie WJ, Madhok R, Swiontkowski M, Robinson CM,
Murray GD (eds.) Musculoskeletal Injuries Module of The
Cochrane Database of Systematic Reviews. The Cochrane
Collaboration, Issue 2. Oxford University Press, London, UK.
9. Chapuy MC, Arlot ME, Duboeuf F, Brun J, Crouzet B, Arnaud
S, Delmas PD, Meunier PJ 1992 Vitamin D3 and calcium to
prevent hip fractures in the elderly women. N Engl J Med
327:1637–1642.
10. Dawson-Hughes B, Harris SS, Krall EA, Dallal GE 1997
Effect of calcium and vitamin D supplementation on bone
density in men and women 65 years of age or older. N Engl
J Med 337:670–677.
11. Lips P, Graafmans WC, Ooms ME, Bezemer PD, Bouter LM
1996 Vitamin D supplementation and fracture incidence in
elderly persons. A randomized, placebo-controlled clinical
trial. Ann Intern Med 124:400–406.
12. Pocock SJ 1991 Clinical Trials. A Practical Approach. John
Wiley & Sons, Chichester, UK.
Address reprint requests to:
Haakon Eduard Meyer, M.D., Ph.D.
Norwegian Institute of Public Health
P.O. Box 4404 Nydalen
0403 Oslo, Norway
13. Kvaavik E, Meyer HE, Smedshaug GB, Falch JA, Tverdal A,
Pedersen JI 2000 The intervention study “Prevention of Hip
Fractures.” Method and implementation. Nor J Epidemiol 10: Received in original form January 2, 2001; in revised form April 9,
78–85. 2001; accepted June 21, 2001.