Hip bone loss is attenuated with 1000 IU but not 400 IU daily vitamin D3: A 1-year double-blind RCT in postmenopausal women



Few year-long vitamin D supplementation trials exist that match seasonal changes. The aim of this study was to determine whether daily oral vitamin D3 at 400 IU or 1000 IU compared with placebo affects annual bone mineral density (BMD) change in postmenopausal women in a 1-year double-blind placebo controlled trial in Scotland. White women aged 60 to 70 years (n = 305) were randomized to one of two doses of vitamin D or placebo. All participants started simultaneously in January/February 2009, attending visits at bimonthly intervals with 265 (87%) women attending the final visit and an additional visit 1 month after treatment cessation. BMD (Lunar iDXA) and 1,25-dihydroxyvitamin D[1,25(OH)2D], N-terminal propeptide of type 1 collagen [P1NP], C-terminal telopeptide of type I collagen [CTX], and fibroblast growth factor-23 [FGF23] were measured by immunoassay at the start and end of treatment. Circulating PTH, serum Ca, and total 25-hydroxyvitamin D [25(OH)D] (latter by tandem mass spectrometry) were measured at each visit. Mean BMD loss at the hip was significantly less for the 1000 IU vitamin D group (0.05% ± 1.46%) compared with the 400 IU vitamin D or placebo groups (0.57% ± 1.33% and 0.60% ± 1.67%, respectively) (p < 0.05). Mean (± SD) baseline 25(OH)D was 33.8 ± 14.6 nmol/L; comparative 25(OH)D change for the placebo, 400 IU, and 1000 IU vitamin D groups was −4.1 ± 11.5 nmol/L, +31.6 ± 19.8 nmol/L, and +42.6 ± 18.9 nmol/L, respectively. Treatment did not change markers of bone metabolism, except for a small reduction in PTH and an increase in serum calcium (latter with 1000 IU dose only). The discordance between the incremental increase in 25(OH)D between the 400 IU and 1000 IU vitamin D and effect on BMD suggests that 25(OH)D may not accurately reflect clinical outcome, nor how much vitamin D is being stored. © 2013 American Society for Bone and Mineral Research.


Vitamin D has been synonymous with bone health since its discovery in the early 20th century.[1] It has a clear role in preventing rickets and osteomalacia and in helping absorb sufficient dietary calcium necessary for mineralizing bones, but we are still unclear about the extent of its involvement in protecting against osteoporosis. A number of trials have investigated the effect of vitamin D treatment on fracture risk with mixed results, which may be explained by type of vitamin D treatment (vitamin D2 versus vitamin D3); dose (daily or less frequent); duration of treatment; and administration route (oral/injection). Subsequent meta-analyses, of which there are many, have concluded that either calcium is also required[2-4] or the daily vitamin D dose needs to be higher than 800 IU.[5] The overall benefit appears to be weighted by one trial in which institutionalized elderly were treated with vitamin D and calcium;[6] the large, pragmatic trials involving healthier elderly, which did not show any benefit, have been criticized for lack of compliance.[7, 8] Observational studies have suggested a relationship between 25-hydroxyvitamin D [25(OH)D], the major circulating metabolite of vitamin D, and bone mineral density (BMD)[9-11] or reduced fracture,[12, 13] but even then, the data are equivocal. There are some discrepancies about how much oral vitamin D is required to increase 25-hydroxyvitamin D [25(OH)D] by a fixed amount.[14, 15] The 25(OH)D increase may be greater if the starting point is lower, and an initial high starting point of 25(OH)D (mean 71 nmol/L) was suggested as a reason for the nonresponse of BMD to vitamin D supplementation in a recent trial.[16, 17] A recent dose-response study found that increasing amounts of vitamin D resulted in successively smaller incremental increases in 25(OH)D.[18]

Although the starting 25(OH)D concentration may be critical in determining the study outcome, for many studies that have investigated the role of vitamin D on bone health, there is confounding from the underlying effects of season, as 25(OH)D is higher in summer and lower in winter. The relationship between change in vitamin D status and bone status in healthier older adults remains unclear. Few studies have included estimates of dietary intakes and sunlight exposure in their adjustments.

The aim of this study was to test whether vitamin D3 supplementation at a daily dose of 400 IU (10 µg) or 1000 IU (25 µg) for 1 year starting in January/February affects bone mineral density loss in women aged 60 to 70 years living in northeast Scotland, a population with low circulating 25(OH)D. Secondary outcomes included markers of vitamin D and bone metabolism, including total 25-hydroxyvitamin D [25(OH)D], parathyroid hormone (PTH), 1,25 dihydroxyvitamin D [1,25(OH)2D], serum calcium, fibroblast growth hormone factor 23 (FGF23), N-terminal propeptide of type 1 collagen (P1NP), and C-terminal telopeptide of type I collagen (CTX).

Materials and Methods


The women were healthy postmenopausal, nonsmoking women living in northeast Scotland aged 60 to 70 years who were taking part in a vitamin D intervention study (VItamin D and CardiOvascular Risk, VICtORy) to investigate risk of cardiovascular disease (CVD).[19] At the time of recruitment, they were not suffering from any condition (diabetes, asthma, malabsorption, blood pressure >160 mm Hg systolic or >99 mm Hg diastolic) or taking medication (hypotensive, hypolipemic, anti-inflammatory, oral corticosteroid) likely to affect vitamin D metabolism or CVD risk. Women on thyroxine treatment were included if stable, as assessed by free T4 and thyroid stimulating hormone concentrations, and their dose had not changed in the 3 months before study entry. Exclusion criteria were planned frequent trips or long periods abroad that would result in an increased exposure to UVB light, or an abnormal biochemical profile on screening. All women provided written informed consent. Ethical permission was obtained from Grampian Research Ethics Committee (08/S0802/73). There were no differences in characteristics between the women who were analyzed for the VICtORy study[19] and the women who had dual-energy X-ray absorptiometry (DXA) scans and could be included in determining the effect of the treatment on bone outcomes (data not shown). Details of the numbers recruited are included in Fig. 1.

Figure 1.

Participant flow through the study. Fewer measurements are available at the hip because of total hip replacements.

Randomization and intervention

Capsules containing vitamin D (400 IU and 1000 IU) and identical placebo capsules were manufactured by Pure Encapsulations (Sudbury, MA, USA) and sent directly to Bilcare (Crickhowell Powys, UK), where they were packed and coded. The certificate of analysis from the manufacturer showed that each capsule contained 400 IU and 1040 IU of vitamin D. Independent analysis (by Eurofins Laboratories Ltd, Wolverhampton, UK) at the end of the study gave a vitamin D content of 346 IU and 832 IU, respectively, which they accepted as being within their quality-control specifications. Participants were randomized in January to March 2009 by the Health Services Research Unit, University of Aberdeen (telephone service) to one of three groups: placebo, 400 IU (10 µg) vitamin D3, or 1000 IU (25 µg) vitamin D3, using minimization criteria for body mass index (BMI; <18.5, 18.5 to 24.99, 25 to 29.99, 30 to 30.99, or >40). At each visit, the participants were given a bottle of capsules (n = 65), sufficient to last 2 months, and instructed to take one capsule at the same time each day with food. Compliance was estimated by counting the unused capsules at each subsequent visit (and supported by later 25(OH)D measurements). The investigators remained blinded throughout the study until after the statistical analysis of the outcomes had been performed. This was achieved by ensuring that the 25(OH)D data was analyzed by a separate researcher (HMM) using a different set of codes from the researcher carrying out the analysis of the main outcome measures (ADW).

Anthropometric measurements

Women were weighed (Seca, Hamburg, Germany) and their height measured using a stadiometer (Holtain Ltd, Crymych, UK) at baseline, 6 months, and 12 months. They had DXA scans to estimate BMD at the hip (total) and spine (L1 to L4), total bone mass, total lean mass, and total fat mass (Lunar iDXA, GE Medical Systems Inc., Madison, WI, USA) before the intervention started and at the end of the study. Daily phantom measurements were performed. In vivo precision was obtained using repeat scans from 60 volunteers and was 0.54% for spine (L2 to L4) BMD and 0.56% for mean total hip BMD (left hip only 0.68%; right hip only 0.75%).

Markers of vitamin D status and bone health

Overnight fasted serum and plasma samples collected at each visit were stored at −80°C, with each participant's complete set batched together before analysis. Measurements for 25(OH)D, PTH, and FGF23 were undertaken under the direction of WDF in the Department of Clinical Chemistry, University of Liverpool, Liverpool, UK, which takes part in the quality-control group for vitamin D, DEQAS. Serum was analyzed for 25(OH)D3 and 25(OH)D2 using dual tandem mass spectrometry using the NIST (National Institute of Standards and Technology, US) standard that is recommended.[20] Interassay coefficients of variation for the assay were <10% for both 25(OH)D2 and 25(OH)D3. The sum of the two are reported as total 25(OH)D. The half-life of circulating 25(OH)D was calculated using the following formula: half life = time since treatment ceased × ln(2) / (ln(25(OH)D) at end of treatment / 25(OH)D after the period of treatment cessation). Parathyroid hormone (PTH) was measured in plasma samples using an electrochemiluminescent immunoassay (ECLIA) on a Modular Analytics E170 analyzer (Roche Diagnostics, Burgess Hill, UK). Inter/intra-assay coefficient of variation was <4% from 1 to 30 pmol/L. The assay sensitivity (replicates of the zero standard) was 0.8 pmol/L. Serum 1,25(OH)2D was measured by radioimmunoassay with an 125I-labelled 1,25(OH)2D derivative tracer and Sac-cell separation after immunoextraction of 1,25(OH)2D using a mini column containing a solid-phase monoclonal antibody (Immunodiagnostic Systems, Boldon, Tyne and Wear, UK). Inter/intra-assay coefficient of variation was <10% over the concentrations analyzed.[21] FGF23 was measured by ELISA with an anti-human FGF-23 mouse monoclonal antibody (Kainos Laboratories, Tokyo, Japan). It has quantification range of 3 to 800 pg/mL.

Serum calcium adjusted for albumin was measured by Clinical Biochemistry, Aberdeen Royal Infirmary, UK, using standard automated systems (ADIVA 2400 Chemistry System, Siemens, Surrey, UK). Inter/intra-assay coefficient of variations over the concentrations analyzed were <2.2% for calcium and <1.8% for albumin. Serum N-terminal propeptide of type 1 collagen (P1NP) and plasma C-terminal telopeptide of type I collagen (CTX) were measured at the start and end visits only by enzyme chemiluminescent immunoassay (ECLIA) (Roche Products Ltd, Penzberg, Germany) using a Diagnostics Elecsys 2010 Immunoassay System (Roche Diagnostics, Mannheim, Germany or Burgess Hill, UK) (coefficient of variation was <4% for P1NP and <5% for CTX).

Assessment of dietary vitamin D intake, sunlight exposure and physical activity

Dietary vitamin D was assessed by the Scottish Collaborative Group food frequency questionnaire (FFQ) (http://www.foodfrequency.org) that was completed after each visit; sunlight exposure was determined by polysulphone badges.[22] These were worn at the lapel on outside clothing for the week after the study visit and returned by post. The difference in absorption at 330 nm before and after the badge had been worn, ΔA330, was used to obtain the weekly standard erythema dose (SED) as follows: SED = 10.7 [ΔA330] + 14.3 [ΔA330]2 − 26.4 [ΔA330]3 + 89.1 [ΔA330]4 (Perkin Elmer UV/VIS Lamda 2 Spectrophotometer). At each visit, the women were questioned about time spent outside and how much of the body was uncovered (5% face only to 60% for face, hands, and arms or legs, plus trunk). The women were questioned about any holidays taken in the previous 2 months. Physical activity was estimated using a bone-specific questionnaire that has been validated in this age group.[23] Energy expended (MET hours per week) and the magnitude of the mechanical load on the skeleton were estimated using the responses to 29 questions about work- and home-related activities and leisure-time pursuits.

Statistical analysis

Using data from a previous 2-year intervention trial of women aged 55 to 65 years,[24] where the annual percentage bone loss was −1.02 ± 1.61 at the lumbar spine (LS) and −0.82 ± 1.12 at the hip, we estimated that with 75 subjects in each group (with 100 allowing a 25% dropout as originally powered on markers of cardiovascular risk), a mean LS BMD loss of 0.75% for the placebo group would enable us to detect a significant benefit (90% power, p = 0.05) in the treatment group, if BMD increased by 0.11%. At 80% power, the detectable difference would be significant if the treatment group lost 0.01% BMD or less. Similarly, for total hip BMD, assuming a loss of 0.65% in the placebo group, we would have sufficient power to detect a significant treatment effect (p = 0.05) if the treatment group lost 0.05% or 0.13% BMD (90% and 80% power, respectively).

Statistical analysis was carried out using PASW Statistics 18 (release 18.0.2) (IBM SPSS Statistics, http://www.spss.com/software/statistics/academic/). One-way ANOVA was used to analyze the BMD change between the treatment groups with Dunnetts post hoc ANOVA test to test the difference between groups. A General Linear Model was used to allow adjustment for major confounders (dietary calcium intake, physical activity). Repeated measures analysis of variance (ANOVA) with treatment × time interactions tested using Pillai's Trace for 25(OH)D and markers of bone turnover. Analysis was carried out on an “intention to treat” basis. The per protocol analysis excluded 4 women who withdrew but continued with the study visits and >80% compliance (total n = 255).


There were no differences in mean subject characteristics between the treatment groups (Table 1). Overall dietary vitamin D intake was low. Dietary calcium intakes were adequate. There were no differences in mean baseline BMD between the women who took part in the study and those who did not meet the study criteria, nor compared with the women who did not have a final bone scan (data not shown). There were no serious adverse events related to the treatment. One woman (who withdrew from the study) had elevated calcium thought to be a result of unmasking mild primary hyperparathyroidism, which was later monitored by her general practitioner. She had been on the 1000 IU daily vitamin D dose. There were 7 fractures during the study (wrist/lower arm [n = 3], foot/ankle [n = 3], and clavicle [n = 1]), with one woman having two fractures on two separate occasions. These are reported as adverse events, as the study was not powered to examine fractures. Three of the women were on placebo and three on the 400 IU vitamin D dose.

Table 1. Subject Characteristics at Baseline
Subject characteristicsPlacebo400 IU1000 IUp1
Mean ± SDn = 90n = 84n = 90ANOVA
Age (years)64.6 ± 2.364.2 ± 1.964.9 ± 2.20.121
Weight (kg)69.9 ± 12.268.1 ± 11.469.4 ± 11.60.570
Height (cm)161.4 ± 6.2160.3 ± 6.8160.9 ± 5.20.463
Waist (cm)86.5 ± 11.485.6 ± 10.187.0 ± 11.50.693
BMI (kg/m2)25.9 ± 3.825.3 ± 3.925.2 ± 3.40.327
Fat mass appendicular2 (kg)12.8 ± 3.812.4 ± 3.412.6 ± 3.30.738
Fat mass trunk (kg)14.5 ± 4.914.2 ± 4.514.5 ± 5.10.944
Lean mass appendicular2 (kg)16.6 ± 2.316.0 ± 2.416.5 ± 2.40.213
Lean mass trunk (kg)18.9 ± 2.018.5 ± 2.218.9 ± 2.10.260
 n = 90n = 83n = 88 
Dietary calcium intake (mg/d)1281 ± 4901253 ± 4841305 ± 5660.763
Dietary calcium intake with supplements (mg/d)1291 ± 4921261 ± 4881306± 5680.846
Dietary vitamin D (µg/d)5.6 ± 3.04.6 ± 2.55.3 ± 2.90.081
Energy intake (MJ/d)9.5± 3.08.9 ± 2.69.4 ± 3.00.412
Energy intake to basal metabolic rate ratio1.71 ± 0.581.63 ± 0.491.68 ± 0.550.565
 n = 90n = 81n = 86 
Physical activity (MET h/week)71.5 ± 31.374.7 ± 29.777.1 ± 37.30.628
Mechanical component of physical activity (peak score)4.6 ± 1.94.6 ± 1.85.0 ± 1.70.353
 n = 88n = 81n = 87Kruskal-Wallis 1-way ANOVA
Median sunlight exposure (SED) baseline0.
Median sunlight exposure (SED) v3 summer4.
  • SED = standard erythemal dose.
  • 1Variable log transformed if required (dietary Ca vitamin D).
  • 2Appendicular refers to the sum of the measurements for legs and arms.
National deprivation category (n) (1, 2, 3, 4, 5, 6)28, 41, 4, 10, 2, 127, 38, 3, 8, 2, 420, 53, 4, 5, 5, 00.210
BMI category (%) (BMI kg/m2 <20, 20–25, 25–30, 30–35, >35)3, 28, 49, 17, 36, 32, 46, 12, 43, 33, 47, 14, 20.962
Underactive thyroid (%)12480.108
Thyroxine (stable dose) (%)7440.620

The capsule count indicated mean overall compliance of 92% (range 72% to 98%), which did not differ significantly between the three groups. There were no adverse events reported during the course of the study that were attributable to the treatment. Total 25(OH)D increased and remained high for the vitamin D-treated groups. For the placebo group, there was an increase in 25(OH)D during summer, which reflected sun exposure in the spring/summer and then a decrease during winter (Fig. 2). Sunlight exposure was positively skewed. Higher erythemal doses (SED) were received in summer compared with winter, but no differences were seen between treatment groups. Physical activity also changed throughout the study, increasing in summer and then decreasing, whereas dietary intakes of vitamin D, calcium, and dietary energy intake remained the same throughout the year (data not shown).

Figure 2.

25-hydroxyvitamin D (mean, 95% CI) according to visit. The 25(OH)D data were normally distributed except for visit 0. Mean (SEM) pmol/L for placebo, 400 IU, and 1000 IU, respectively, were:

  • Visit 0 (January–March): 35.8 (1.7); 33.4 (1.4); 33.2 (1.5)
  • Visit 1(March–May): 34.5 (1.6); 57.3 (1.5); 63.5 (1.7)
  • Visit 2 (May–July): 43.7 (2.0); 64.8 (1.7); 70.7 (1.8)
  • Visit 3 (July–September): 55.0 (2.0); 68.8 (1.7); 76.9 (2.0)
  • Visit 4 (September–November): 48.1 (2.2); 69.5 (1.9); 77.7 (2.1)
  • Visit 5 (November–January): 38.1 (1.9); 66.0 (1.8); 75.8 (2.0)
  • Visit 6 (January–March): 32.0 (1.6); 65.0 (2.1); 76.0 (2.0).

There was no change in CTX, P1NP, FGF23 or 1,25(OH)2 D as a result of treatment (Table 2). There was a significant difference in PTH, which was observed before and after treatment (Table 2), and throughout the year as shown by repeated measures analysis (Fig. 3). At the end of treatment, PTH had decreased from baseline for both vitamin D treatment groups compared with placebo (p < 0.001 and p = 0.031 for the comparison with 1000 IU vitamin D and 400 IU vitamin D, respectively). There were small changes in adjusted serum calcium (−0.016 mmol/L, −0.009 mmol/L, and 0.005 mmol/L for placebo, 400 IU vitamin D, and 1000 IU vitamin D, respectively), with the difference significant only between placebo and 1000 IU vitamin D (p = 0.027).

Table 2. Changes in Markers of Vitamin D Status and Bone Health
Mean ± SDBaseline (visit 0) Final (visit 6) Difference 
Placebo400 IU1000 IUpPlacebo400 IU1000 IUpPlacebo400 IU1000 IUp
n = 90n = 84n = 90 n = 89n = 84n = 90 n = 89n = 84n = 90 
  1. 25(OH)D = total 25-hydroxyvitamin D; 1,25(OH)2D = total 1,25-dihydroxyvitamin D; PTH = parathyroid hormone; CTX = plasma beta C-terminal telopeptide; P1NP = serum N-terminal propeptide of type 1 collagen; BMD = bone mineral density.
  2. ANOVA *p < 0.05; p < 0.01; p < 0.001 (variable log transformed if required).
  3. Dunnett post hoc ANOVA tests as follows: 25(OH)D difference p < 0.001 for all comparison groups; PTH difference between placebo and 1000 IU vitamin D, p < 0.001; between 400 IU vitamin D and 1000 IU vitamin D, p = 0.031; and between placebo and 400 IU vitamin D, p = 0.071; adjusted serum calcium difference between placebo and 1000 IU, p < 0.027; percentage change in hip BMD difference between placebo and 1000 IU vitamin D, p = 0.027; and between 400 IU vitamin D and 1000 IU vitamin D, p = 0.043. For all other comparisons, p > 0.10.
  4. CTX and P1NP had one extreme outlier that was removed for that analysis only. Similarly, FGF23 had a different extreme outlier that was removed. Including the outlier did not change the outcome.
25(OH)D (nmol/L)35.8 ± 16.433.4 ± 13.233.2 ± 13.8 32.0 ± 14.965.0 ± 19.775.9 ± 18.9−4.1 ± 11.5+31.6 ± 19.8+42.6 ± 18.9
1,25(OH)2D (pmol/L)143 ± 43138 ± 40138 ± 46 138 ± 43139 ± 43139 ± 45 −6 ± 43+1 ± 36+1 ± 41 
PTH (pmol/L)5.3 ± 1.34.8 ± 1.35.1 ± 1.3*5.1 ± 1.24.4 ± 1.14.4 ± 1.0−0.2 ± 0.8−0.5 ± 0.8−0.8 ± 0.9
Serum Ca (mmol/L)2.35 ± 0.072.33 ± 0.072.32 ± 0.07 2.33 ± 0.072.33 ± 0.072.33 ± 0.07 −0.016 ± 0.053−0.009 ± 0.049+0.005 ± 0.065*
         Change (%) 
 n = 84n = 77n = 82 n = 80n = 83n = 90 n = 77n = 75n = 82 
P1NP (mg/L)45.1 ± 23.045.7 ± 20.243.8 ± 18.6 44.9 ± 22.643.9 ± 18.742.3 ± 15.6 4.9 ± 41.82.9 ± 27.10.3 ± 21.2 
CTX (mg/L)0.38 ± 0.170.41 ± 0.150.37 ± 0.17 0.39 ± 0.170.40 ± 0.150.37 ± 0.16 2.7 (36.4)5.1 (33.3)4.6 (25.9) 
 n = 59n = 63n = 70 n = 59n = 63n = 70 n = 59n = 63n = 70 
FGF23 (ng/L)71.9 ± 40.083.9 ± 73.574.8 ± 67.6 96.5 ± 165.992.3 ± 90.087.4 ± 116.2 21.2 ± 62.411.9 ± 34.313.9 ± 31.9 
 n = 88n = 83n = 88 n = 88n = 83n = 88 n = 88n = 83n = 88 
Mean total hip BMD (g/cm2)0.920 ± 0.1180.917 ± 0.1020.923 ± 0.132 0.914 ± 0.1180.912 ± 0.1030.923 ± 0.135 −0.60 ± 1.66−0.57 ± 1.33−0.05 ± 1.46*
 n = 90n = 84n = 90 n = 88n = 83n = 88 n = 88n = 83n = 88 
Lumbar spine (L1 to L4) BMD (g/cm2)1.081 ± 0.1531.075 ± 0.1411.068 ± 0.161 1.076 ± 0.1531.071 ± 0.1351.070 ± 0.164 −0.46 ± 2.79−0.23 ± 2.69+0.23 ± 2.88 
Figure 3.

Parathyroid hormone (mean, 95% CI) according to visit. Mean (SEM) pmol/L for placebo, 400 IU, and 1000 IU, respectively, were:

  • Visit 0 (January–March): 5.32 (0.14); 4.84 (0.14); 5.13 (0.13)
  • Visit 1(March–May): 5.24 (0.13); 4.58 (0.14); 4.73 (0.13)
  • Visit 2 (May–July): 5.06 (0.14); 4.33 (0.12); 4.39 (0.11)
  • Visit 3 (July–September): 4.84 (0.11); 4.51 (0.14); 4.40 (0.10)
  • Visit 4 (September–November): 4.88 (0.12); 4.31 (0.12); 4.43 (0.11)
  • Visit 5 (November–January): 5.16 (0.15); 4.53 (0.13); 4.46 (0.12)
  • Visit 6 (January–March): 5.13 (0.13); 4.38 (0.12); 4.35 (0.11).

Bone loss at the hip was significantly greater for the placebo and 400 IU vitamin D groups (losing BMD at a mean rate of 0.6% and 0.6%, respectively) compared with the 1000 IU vitamin D group, which essentially showed no change (−0.05%) (ANOVA post hoc comparison with 1000 IU vitamin D treatment: p = 0.027 for placebo; p = 0.043 for 400 IU vitamin D). The BMD change at the lumbar spine (L1 to L4) was not significantly different between the treatment groups (−0.5%, −0.2%, and 0.2% for placebo, 400 IU vitamin D, and 1000 IU vitamin D, respectively). Adjustment for confounders (including dietary calcium, sunlight exposure, or physical activity) did not change the outcome, nor did repeating the analysis excluding women who stopped treatment but continued study visits and those with <80% compliance. Body weight, change in CTX, change in PTH, and change in 1,25(OH)2D, but not changes in serum calcium, P1NP, and FGF23, were found to be additional independent predictors of percentage hip BMD change by ANOVA (Table 3).

Table 3. Predictors of Percentage Hip BMD Change by Analysis of Covariance
Variables included in the modelDependent variable: mean hip BMD change (%)
  1. Height, physical activity, mean dietary calcium, mean dietary vitamin D, sunlight exposure (total standard erythemal dose), FGF23 difference, and adjusted serum calcium difference were also tested but were not found to be significant predictors of mean hip BMD change.
  2.  1Placebo [reference].
Weight (kg)0.0180.0080.029
1,25(OH)2D difference−0.0060.0020.013
PTH difference0.2430.1160.037
CTX difference−2.3440.9920.019
Treatment group  0.015
400 IU vitamin D10.2500.2430.305
1000 IU vitamin D10.7010.2430.004

There were few women with 25(OH)D >75 nmol/L at baseline (n = 6 in total: 3 from the placebo and 3 from the 1000 IU vitamin D group) and an additional 29 in total (placebo n = 14; 400 IU vitamin D n = 9; 1000 IU vitamin D n = 6) with 25(OH)D >50 nmol/L. The differences between treatment groups for mean percentage hip BMD change was still significant if these women were excluded.

The relationship between bone markers (Table 4) showed that at baseline (i) 25(OH)D was inversely associated with PTH; (ii) serum Ca was inversely associated with the bone markers CTX and P1NP; (iii) there were inverse associations between the bone turnover markers and BMD; and (iv) 1,25(OH)2D was inversely associated with FGF23 and BMD. At the end of the treatment period, the associations were similar except in addition (i) 25(OH)D was now also positively associated with 1,25(OH)2D; (ii) serum Ca was inversely associated with CTX and no longer with P1NP; and (iii) 1,25(OH)2D was only negatively associated with LS BMD, not hip BMD or FGF23.

Table 4. Associations Between Markers of Bone Health at Baseline (Spearman Correlations)
Baseline (visit 0) correlations25(OH)D (nmol/L)1,25(OH)2D (pmol/L)PTH (pmol/L)Serum Ca (mmol/L)P1NP (µg/L)CTX (µg/L)FGF23 (ng/L)Mean total hip BMD (g/cm2)Lumbar spine BMD (g/cm2)
25(OH)D (nmol/L)1.000        
 n = 263        
1,25(OH)2D (pmol/L)0.0851.000       
 n = 263n = 264       
PTH (pmol/L)−0.3180.0821.000      
 n = 259n = 259n = 259      
Serum Ca (mmol/L)0.053−0.020−0.0631.000     
 n = 263n = 263n = 259n = 263     
P1NP (µg/L)0.0260.073−0.0190.160*1.000    
 n = 243n = 243n = 241n = 243n = 243    
CTX (µg/L)0.0120.0810.0370.1680.7781.000   
 n = 241n = 241n = 239n = 241n = 241n = 241   
FGF23 (ng/L)0.021−0.143*0.0920.061−0.016−0.0631.000  
 n = 192n = 192n = 192n = 192n = 182n = 181n = 192  
Lumbar spine BMD (g/cm2)0.089−0.138*−0.1470.021−0.167−0.2430.0961.000 
 n = 263n = 264n = 259n = 263n = 243n = 241n = 192n = 264 
Mean total hip BMD (g/cm2)−0.003−0.157*−0.061−0.084−0.223−0.3190.1080.6241.000
 n = 258n = 259n = 254n = 258n = 239n = 237n = 190n = 259n = 259
Final (visit 6) correlations         
  • 25(OH)D = total 25-hydroxyvitamin D; 1,25(OH)2D = total 1,25-dihydroxyvitamin D; PTH = parathyroid hormone; CTX = plasma beta C-terminal telopeptide; P1NP = serum N-terminal propeptide of type 1 collagen; BMD = bone mineral density.
  • *p < 0.05;
  • p < 0.01;
  • p < 0.001 (variable log transformed if required).
25(OH)D (nmol/L)1.000        
 n = 263        
1,25(OH)2D (pmol/L)0.126*1.000       
 n = 263n = 263       
PTH (pmol/L)−0.3240.1081.000      
 n = 258n = 258n = 259      
Serum Ca (mmol/L)0.058−0.055−0.0011.000     
 n = 261n = 261n = 258n = 262     
P1NP (mg/L)−0.0570.0220.0480.0441.000    
 n = 252n = 252n = 249n = 253n = 253    
CTX (mg/L)−0.0640.0520.0190.140*0.7981.000   
 n = 249n = 249n = 246n = 250n = 249n = 250   
FGF23 (ng/L)0.045−0.1390.0250.1240.0630.0091.000  
 n = 191n = 191n = 192n = 191n = 187n = 186n = 192  
Lumbar spine BMD (g/cm2)−0.083−0.191−0.0410.016−0.173−0.2550.1161.000 
 n = 263n = 263n = 259n = 262n = 253n = 250n = 192n = 264 
Mean total hip BMD (g/cm2)−0.083−0.098−0.019−0.036−0.234−0.2860.1220.6291.000
 n = 258n = 258n = 254n = 257n = 248n = 245n = 190n = 259n = 259

One month after treatment cessation, mean 25(OH)D had decreased for the two vitamin D groups but not for the placebo group. There was a large variation in the response to treatment cessation and the calculated half-life in relation to 25(OH)D decay (ie, 25(OH)D decrease occurring 1 month after treatment cessation). The number below the key cutoffs for 25(OH)D (<25 nmol/L; <50 nmol/L, and <75 nmol/L) showed significant differences between the treatment groups (Table 5).

Table 5. Changes in 25-Hydroxyvitamin D Status After Treatment Cessation
Median (IQR)Final treatment visit One month after final visit 
Placebo400 IU1000 IU Placebo400 IU1000 IU 
n = 89n = 85n = 90 n = 86n = 81n = 86 
  1. 25(OH)D = total 25-hydroxyvitamin D; IQR = interquartile range.
  2.  p < 0.001 for Kruskal-Wallis test or chi-square test for categorical variables.
25(OH)D3 (nmol/L)28.5 (18.6)62.9 (24.8)75.4 (25.8)25.8 (13.8)48.3 (21.4)63.5 (15.1)
Total 25(OH)D (nmol/L)30.0 (20.0)63.7 (25.0)75.4 (25.8)28.0 (14.3)50.3 (21.1)65.0 (14.9)
25(OH)D <25 nmol/L (% subjects)
25(OH)D <50 nmol/L (% subjects)87.618.86.794.348.116.3
25(OH)D <75 nmol/L (% subjects)98.974.147.898.992.682.6
Half-life 25(OH)D (months)    1.7 (7.4)2.2 (3.3)2.6 (4.4) 
Excluding subjects where 25(OH)D difference< > −1 or 1    n = 77n = 75n = 80 
Half-life 25(OH)D (months)    1.7 (5.8)2.1 (2.8)2.1 (4.1) 


This study showed that the group supplemented with 1000 IU vitamin D a day showed negligible mean hip BMD losses of 0.05% after 1 year of treatment compared with losses of 0.6% for the placebo and the 400 IU vitamin D groups, the latter bone loss being similar to the assumption for our a priori power calculation (0.65%). We found no differences in LS BMD loss between the three groups, although the overall bone loss at this site was much lower than that we predicted (mean bone loss for the placebo group was 0.5% and we expected this to be 0.75%). In this age group, LS BMD loss of some women may be masked by osteophyte growth[25] or possibly fractures that may result in artefactually increased BMD in the region of the collapsed vertebra.[26]

Our findings are in agreement with the conclusions that more than 400 IU a day may be required to improve bone health[27] and that the mean 25(OH)D concentration at which benefit occurs is >74 nmol/L (mean 25(OH)D reached in our study was 76 nmol/L for the 1000 IU treatment).[5] When we excluded women with baseline 25(OH)D >50 nmol/L, which is the threshold above which the Food and Nutrition Board of the Institute of Medicine (IOM) concluded that 97.5% of the population would be vitamin D replete,[28] our findings did not change. Their recommendations to meet this are 600 IU vitamin D a day for adults. A recent 1-year intervention study in women aged 50 to 80 years compared two doses of vitamin D (daily 800 IU and 6500 IU, the latter as 20,000 IU twice weekly in addition to 800 IU daily) and found no differences in BMD change (no change from baseline for LS and a small increase in BMD for the hip) between the groups.[17] However, they did not have a placebo comparison. Both groups started with much higher circulating 25(OH)D than our study (71 nmol/L).

The increase in 25(OH)D per unit of ingested vitamin D in our study was greater than others have reported,[29] but it is known that the 25(OH)D response to vitamin D supplementation is greater if the starting 25(OH)D is lower. However, it is unclear why a small additional increase in 25(OH)D would result in reduced hip BMD loss. There was a marked increase in 25(OH)D observed for the 400 IU vitamin D group compared with the placebo with only a small additional increment in 25(OH)D for the 1000 IU vitamin D group (in spite of the latter being two and a half times the dose). Even for the summer visit, when serum 25(OH) D was at its peak for the placebo and there was a plateau for the treatment groups, the change from baseline was twice that for the 400 IU compared with the placebo, but the 1000 IU dose produced only a 20% additional increment in 25(OH)D above that for the 400 IU dose. The diminishing returns of increased vitamin D supplementation on 25(OH)D response has been noted by others.[18] One could speculate that the additional vitamin D in the higher-dose group was not completely converted to 25(OH)D and the excess may have been deposited into tissue stores or perhaps degraded by the body and excreted. It is also possible that less vitamin D was absorbed from the gastrointestinal tract for the 1000 IU dose compared with the 400 IU dose. Discrepancies could arise because of differences in the vitamin D content of the capsules (perhaps because of different degradation rates). However, independent analysis after the study finished showed that although the high-dose capsule contained 16.8% less than the target amount and the low dose 13.5% less, the vitamin D content was still within the analytical specifications and the ratio of the two doses at the two time points was almost identical (the final ratio being 2.4 instead of the expected 2.5). We observed a small increase in serum calcium for the 1000 IU vitamin D treatment group, and it is possible that transient increases in serum calcium after ingestion of the higher-dose vitamin D contribute to a bone anabolic effect (although analysis of serum P1NP, a bone formation marker, showed no change). Alternatively, our data may reflect increased mineralization of undermineralized bone.[30] The magnitude of change for serum calcium was very small, with the final mean concentration being the same for all three groups, but the differences from baseline between treatment groups were statistically significant. Alternatively, the degradation of excess vitamin D could stimulate metabolic pathways that might be beneficial to bone, although this is largely speculative. Other metabolites of vitamin D, produced as a result of 24-hydroxylation, for example, may be involved when higher vitamin D doses are ingested.[31] Although it is possible that additional vitamin D improved muscle strength and increased physical activity resulted in improved BMD, our study was not designed to test this.

We did not observe any between-group differences in circulating bone turnover markers (final or change from baseline) that could help explain the mechanism behind our findings (although CTX, 1,25(OH)2D, and PTH were independent predictors of change in BMD). The only marker for which there was a difference between groups was PTH. Although there appeared to be a dose response with the reduction in PTH for the 1000 IU vitamin D treatment being greater than that for the 400 IU, the final circulating PTH for the two vitamin D–treated groups was identical. The inclusion of FGF23 in the analysis did not influence the outcomes. This growth factor is important in regulating circulating phosphate. It is known that FGF23 increases with severity of chronic kidney disease, but manipulating dietary phosphate in healthy individuals does not appear to affect FGF23, as shown when six healthy males were given oral phosphate binders in combination with low dietary phosphate intake for 2 days followed by 3 days of repletion with inorganic phosphate.[32] In contrast, a study with 10 healthy subjects was able to induce changes in FGF23 (and 1,25[OH]2D) by varying dietary phosphate and calcium.[33] Our population was relatively healthy and this marker may be more meaningful in those with declining kidney function.

Vitamin D binding protein (VDBP) may modulate the activity of vitamin D metabolites, which could explain the discrepancies seen between different studies examining the relationship between 25(OH)D and bone health.[34] It has also been speculated that the VDBP binding of 25(OH)D may protect 25(OH)D against degradation, which provides a mechanism by which the decrease in 25(OH)D in winter at high latitudes could be minimized. Unfortunately, we did not have VDBP measurements to test this hypothesis. Measurements of half-life of 25(OH)D range from a few weeks[35] to 2 months,[36] but the “vitamin D winter,” the period when vitamin D cannot be synthesized, can last 6 months or more at high latitude. We found a wide variation in the calculated half-life for 25(OH)D, with a median of around 2 months for both treatment groups, which is at the upper end of published half-lives for 25(OH)D.[37] It was notable that the percentage of women with 25(OH)D below 25 nmol/L 1 month after treatment stopped was essentially zero for the treatment groups compared with 40% who had no vitamin D treatment. For the 50 nmol/L cutoff, the figure given by the IOM as covering most of the population's vitamin D needs[28] showed that 16% and 50% of women were below this for 1000 IU and 400 IU treatment groups, respectively, compared with 94% for the placebo. If a higher cutoff of 75 nmol/L is considered optimal, even for the 1000 IU vitamin D dose, the number of women below this cutoff increased rapidly from 48% to 83% within 1 month. Although this indicates that the treatment should be continued to sustain 25(OH)D at high circulating concentrations, we do not know if treatment has longer-term benefits in keeping people above the lower threshold for risk of deficiency. Further, the assumption is that 25(OH)D is reflecting vitamin D stores in the body.

Our findings and those of others raise the issue of whether there is some resistance to the body achieving higher 25(OH)D with oral vitamin D. This area is one of controversy, with calls for higher circulating 25(OH)D because these concentrations are observed in outdoor workers such as lifeguards.[38, 39] There are contrasting data to show that less than half of those with an outdoor lifestyle, living in an area where UVB radiation is sufficient all year, could reach 75 nmol/L 25(OH)D,[40] although some have argued that this is because use of sunscreen would have blocked vitamin D synthesis. It is possible that the breakdown of excess vitamin D, which normally occurs to regulate cutaneous vitamin D synthesis, is compromised in UVR-overexposed skin and this could explain high circulating concentrations of 25(OH)D in some individuals. If one accepts that sunlight can increase 25(OH)D to circulating concentrations in excess of 100 nmol/L in everyone, the dose-response rate for oral vitamin D is puzzling because it indicates that many-fold increases in vitamin D intake are required to raise mean 25(OH)D by relatively small additional increments. There are studies underway using daily doses of 4000 IU and 6000 IU that are below the no observed adverse effect level of 10,000 IU but above safe upper limit of 4000 IU suggested by IOM, but it would appear that perhaps not all of this is converted to circulating 25(OH)D. If the storage capacity of vitamin D is limited as has been suggested, resolving what happens to the excess vitamin D may be important in determining the optimal dose for benefit. Although risk of hypercalcemia is extremely small, there may be other adverse outcomes that are affected by higher-dose vitamin D. There are already studies that suggest that high loading doses of vitamin D may be inadvisable.[41]

The limitations of our study include the recruitment being restricted to postmenopausal women, who were healthy but with low circulating 25(OH)D, so that the findings may not be applicable to other populations; the duration of the study being 1 year only; and that the end points were surrogate markers of bone health (BMD and markers of bone metabolism), not fractures. The strengths were the study design with all subjects starting the intervention at the beginning of the year; careful monitoring of the participants, being seen at regular intervals of 2 months; the precision of 25(OH)D measurements, standardized to NIST; and good subject retention throughout.

In conclusion, we found a small effect of vitamin D on BMD change in healthy postmenopausal women >5 years past menopause, in which 1000 IU but not 400 IU vitamin D appeared to attenuate bone loss over 1 year. There was a dose-response effect on PTH and serum calcium, albeit small, but not on any of the other measured markers including 1,25(OH)2D and bone turnover markers.


All authors state that they have no conflicts of interest.


Trial registration: Vitamin D effects on cardiovascular disease risk (VICtORy) study at controlled-trials.com as ISRCTN20328039 (http://controlled-trials.com/ISRCTN20328039/). This work was funded partly by the UK Food Standards Agency, the Department of Health, and the National Osteoporosis Society.

We thank Prof Roger Francis (University of Newcastle, UK) for his role as Trial Steering Committee Chair; Prof Juliet Compston (University of Cambridge, UK) and Prof Bernard Keavney and Dr Mark Pearce (both of University of Newcastle, UK) as members of the Data Monitoring Committee; and the following staff at the University of Aberdeen, UK: Mrs Lismy Cheripelli, Registered General Nurse who assisted with the volunteer visits; Mrs Lana Gibson and Mrs Jennifer Scott for performing dual-energy X-ray absorptiometry scans; Ms Kelly Dean and Mrs Denise Tosh for bone marker measurements; Mrs Gladys McPherson (Health Services Research Unit), who provided the randomization service; and Ms Katrina Galbreith for help with data entry.

Authors' roles: HMM (principal investigator) had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. She was involved in the study design, study management, and the writing of the manuscript. ADW was responsible for the day-to-day running of the study and involved in the interpretation of the study data. LSA offered statistical advice and interpretation. AJB had responsibility for the volunteers' welfare and gave assistance with the study design. WDF offered analysis and interpretation of measurements for 25-hydroxyvitamin D, PTH, and FGF23. AM contributed to the study design and interpretation of physical activity data. DMR (director of the Aberdeen Prospective Osteoporosis Screening Study from where the volunteers were recruited) contributed to the study design and interpretation. KS (the research nurse in charge of the study volunteers) contributed to the study design. WGS was responsible for routine biochemistry tests, including serum calcium and interpretation. FT was involved in study design and interpretation. HMM, ADW, AM, FT, and DMR were involved in obtaining study funding. All authors critically appraised the manuscript.

Role of the sponsor: The sponsor, Research and Innovation, University of Aberdeen, UK, was responsible for confirming proper arrangements to initiate, manage, monitor, and finance this RCT as designated by the Scottish Executive Health Department Research Governance Framework for Health and Community Care and the Department of Health Research Governance Framework for Health and Social Care 2nd Edition (2006)–Confirmation of the Role of Sponsor.