Calcified Tissue International

, Volume 73, Issue 4, pp 332–338

Vitamin D Receptor Gene Polymorphism Predicts Height and Bone Size, Rather Than Bone Density in Children and Young Adults


    • Department of Pediatrics, div. EndocrinologySophia Children’s Hospital, 3000 CB Rotterdam
    • Department of RadiologyDijkzigt University Hospital, Rotterdam
    • Department of Internal MedicineDijkzigt University Hospital, Rotterdam
  • S. M. P. F. de Muinck Keizer-Schrama
    • Department of Pediatrics, div. EndocrinologySophia Children’s Hospital, 3000 CB Rotterdam
  • E. P. Krenning
    • Department of Nuclear MedicineDijkzigt University Hospital, Rotterdam
  • H. A. P. Pols
    • Department of Internal MedicineDijkzigt University Hospital, Rotterdam
  • A. G. Uitterlinden
    • Department of Internal MedicineDijkzigt University Hospital, Rotterdam
Clinical Investigation

DOI: 10.1007/s00223-002-2130-2

Cite this article as:
van der Sluis, I., de Muinck Keizer-Schrama, S., Krenning, E. et al. Calcif Tissue Int (2003) 73: 332. doi:10.1007/s00223-002-2130-2


Peak bone mass is considered to be under strong genetic control. We studied the association among anthropometry, bone density and vitamin D receptor (VDR) genotype in an ethnically homogeneous group of 148 Caucasian children and young adults. Bone density was measured by dual energy X-ray absorptiometry (DXA) and VDR genotype was determined by a direct haplotyping procedure of the BsmI, ApaI, and TaqI restriction fragment length polymorphisms. A second DXA measurement was made after approximately 4 years. Results are expressed as age- and sex-adjusted standard deviation scores (SDS). Previously, the collagen IA1 Sp1 polymorphism was studied in this population. We found VDR genotype to be associated with a 0.4 SDS increased height per allele copy of haplotype ‘3’ (P = 0.04) and a 0.4 SDS increased width of the lumbar vertebral body in the haplotype ‘3’ allele carriers (P = 0.05). We observed a trend towards a 0.3 SDS decreased bone mineral apparent density of lumbar spine (BMAD) per copy of haplotype ‘3’ allele (P = 0.10). In contrast, no association with areal bone mineral density (BMD) was observed. In the follow-up analyses, no differences in height or bone gain among the VDR genotypes were demonstrated. By combining the risk alleles of VDR and collagen IA1 Sp1 genotype, an additive genotype effect on height (P = 0.006) and vertebral body width (P = 0.001) was found. In this exploratory study we found VDR genotype to be associated with frame size and BMAD. The VDR genotype effects on stature and bone size seem to neutralize the effect on areal BMD.


Vitamin D receptorPolymorphismHeightBone sizeBone mineral densityChildrenGenetic factors

Peak bone mass is considered to be under strong genetic control [1, 2, 3]. Bone mass acquisition in childhood is also associated with weight, height, hormonal status, and lifestyle factors such as physical activity and calcium intake [4, 5, 6, 7]. Some of these determinants might be under genetic control as well. Identification of the genes mediating effects on bone mass may lead to better understanding of the pathogenesis of osteoporosis, and might help us to identify subjects at risk and/or predict response to treatment.

In the extensive search for candidate genes, which are associated with osteoporosis, the vitamin D receptor (VDR) polymorphisms have been most frequently studied [5, 8, 9, 10]. The hormone 1,25-dihydroxyvitamin D3 is required for the mineralization of bone, intestinal calcium absorption, control of calcium and phosphate homeostasis, and the regulation of parathyroid hormone secretion. The effects of vitamin D are mediated by the VDR. These characteristics make the VDR gene a logical candidate gene to analyze for effects on variations in bone mass. The VDR gene maps to chromosome 12q13-14, and several sites of sequence variation in the VDR gene have been described to date. For example, a cluster of linked sites exists near exon 9 and the 3- UTR (untranslated region) and are detected by BsmI, ApaI, and TaqI, as restriction fragment length polymorphisms (RFLPs) [10]. The majority of the association studies have been performed in pre- and postmenopausal women because the skeleton undergoes many changes in this period of life and most fractures occur in the elderly. However, variation in the attainment of peak bone mass plays an important role in the development of osteoporosis in later life [11]. In this respect, polymorphisms of the VDR gene may also play a role in the attainment of peak bone mass.

Few pediatric populations have been studied, but so far with conflicting results, similar to what was found for adult populations. Sainz et al. [12] found an association between bone density measured with quantitative computed tomography (QCT) and the BsmI and ApaI polymorphism in prepubertal girls. Others could not confirm this finding in a group of healthy children and young adults [13, 14]. Some studies reported an association between the BsmI polymorphism and frame size, i.e., height at birth and final height [15, 16]. Therefore, possible effects of VDR polymorphism on bone might be mediated by other frame size factors. We therefore investigated the association among VDR polymorphism, anthropometry and bone density characteristics in a group of healthy Dutch children and young adults. In view of the known influence of vitamin D on collagen synthesis [17, 18] and previous observations on collagen Iα1 (COLIA1) Sp1 genotype effects on frame size in this cohort [19], we also studied possible interaction between both genotype effects.
Figure 1

Interaction of effects of VDR genotype and COLIA1 genotype on height and vertebral body width. The combined genotypes were analyzed as a risk score: ‘0’ those not carrying a risk genotype of VDR nor COLIA1, ‘1’ those carrying a risk genotype of either VDR or COLIA1, and ‘2’ those carrying a risk genotype of VDR as well as COLIA1. P-values of regression analyses are given.

Materials and Methods


In 1994–1995, 500 healthy Dutch children (403 Caucasian) from the Rotterdam region participated in a study to obtain reference values for dual energy X-ray absorptiometry (DXA) measurements. Children with diseases or using drugs known to affect bone metabolism were excluded. The cross-sectional results of this first study have been presented previously [4, 20]. All participants were approached to volunteer in a follow-up study. One hundred forty-eight Caucasian children and young adults (57 boys and 91 girls) agreed; 6 children participated only at follow-up at which time blood samples were taken. The mean follow-up period was 4.4 years (range 3.2–6.7 years). The study was limited to Caucasian children. This study was approved by the medical ethics committee of the University Hospital Rotterdam. Written informed consent was obtained from the parents and all children older than 12 years of age.

Anthropometric and Bone Density Measurements

Height was measured without shoes on a fixed stadiometer and weight was measured without shoes on a standard clinical balance. Body mass index (BMI) was calculated as weight/height2. Height and BMI were expressed as age- and sex-matched standard deviation score (SDS) [21, 22]. As validated previously [23], pubertal development was evaluated by self-assessment of breast and pubic hair stage in girls and genitalia and pubic hair stage in boys, according to the method of Tanner [24]. We classified Tanner stage 1 as prepubertal, Tanner stages 2, 3, 4 as pubertal, and Tanner stage 5 as postpubertal. A questionnaire was administered to determine calcium intake, physical activity, medical history, and menarche.

Bone mineral density (BMD, g/cm2) of lumbar spine (LS) and total body (TB) were determined by DXA (Lunar DPX-L, Madison, WI). For children with weight below 30 kg pediatric software was used. To correct for bone size we calculated apparent BMD (BMAD, g/cm3) of lumbar spine with the model BMADLS = BMDLS × [4/(π × width)]. ‘Width’ represents the mean width of the second to fourth lumbar vertebral body. This model was validated by in vivo volumetric data obtained from magnetic resonance imaging of lumbar vertebrae [25]. Lean body mass was assessed by total body DXA. Bone density and body composition were expressed as age- and sex-adjusted standard deviation scores (SDS) using our own reference data [26]. We used the same data to obtain reference values for vertebral width, allowing us to calculate SD scores.

Genotyping Procedure

DNA was isolated from lymphocytes according to standard procedures, as described previously [8]. Three clustered anonymous RFLPs in the 3′ end of the VDR gene were determined by PCR and enzymatic digestion of the products with BsmI, ApaI, and TaqI using a direct haplotyping PCR procedure. In this procedure, the clustered restriction site polymorphisms at the VDR gene locus were monitored simultaneously and individually. This method has been described previously [8] and allows direct determination of the haplotypes based on 3 RFLPs without analysis of parental DNA. The alleles were named as described earlier [8, 10] for alleles, defined by individual RFLPs; in haplotypes, for example “BAt”, capital letters denote absence and lowercase letters denote the presence of the site for the restriction enzymes BsmI (B/b), ApaI (A/a), and TaqI (T/t) on each of the alleles. In the direct haplotyping, number 1 stands for baT, 2 = BAt, 3 = bAT, 4 = BAT, 5 = bAt. We determined the G to T substitution in the polymorphic Sp1 binding site in the first intron of the COLIA1 gene, as described previously [19].

Statistical Analysis

The analysis was limited to the ethnically homogeneous group of Caucasian children. To take possible age and sex differences between the genotype groups into account we calculated age- and sex-matched standard deviation scores. We grouped subjects by allele copy number (0, 1, 2) for the most common haplotype alleles 1, 2, and 3 and individual RFLPs BsmI and ApaI. Because there is a 99% concordance between b and T (and B and t), we do not present the results for TaqI. Hardy Weinberg Equilibrium was calculated according to standard procedures using Chi square analysis. We tried to avoid multiple testing artifacts by only testing consistent effects (i.e., only those observed at baseline and follow-up) and by accepting only three possible genetic models to explain differences among groups, i.e., an allele dose effect, a dominant effect, or a recessive effect. Allele dose was defined as the number of copies of a certain allele in the genotype. In case of a consistent trend reflected as an allele dose effect, which showed no significant deviation of linearity, a linear regression analysis was performed to quantify the association. This results in a regression coefficient ‘β’, i.e., the increment or decrement in SDS per allele copy. In the case of a dominant or recessive effect of the test allele, independent sample T-tests were performed to test for differences between two genotype groups. For dominant alleles we compared test allele carriers versus non-carriers. For recessive effects, homozygous subjects for the test allele were compared to heterozygous carriers combined with non-carriers. We searched for possible interaction between risk alleles of the VDR gene and the polymorphic Sp1 site in the COLIA1 gene. Non-haplotype ‘3’ alleles and the T-allele were assigned as risk alleles for decreased height and vertebral body width. As a consequence, the non-risk genotypes of VDR were [1, 3], [2, 3] or [3, 3] and of COLIA1 this was [GG], whereas the risk genotypes of VDR were [1, 1], [1, 2], [1, 4] or [2, 2] and of COLIA1 they were [GT] or [TT]. We compared three risk groups, i.e., those not carrying any risk genotype (=0), those carrying one risk genotype, either from VDR or from COLIA1 (=1), and those carrying risk a genotype from both VDR as well as COLIA1 (=2). In case of an additive effect, we performed a linear regression analysis. P-values ≤0.05 were considered to be significant.



Baseline characteristics are presented in Table 1. SD scores of height, vertebral body width, and bone density parameters showed a normal distribution and the means did not differ from the expected zero. No significant gender differences in baseline characteristics were found. The mean age at menarche was 13.1 years (range 11–16 yrs), which is the mean age of menarche in Dutch girls [21]. At follow-up, 31 children were prepubertal, 48 pubertal and 69 postpubertal. The VDR haplotype and genotype frequencies are given in Table 2. No differences in VDR haplotype and genotype distribution were present between boys and girls. There was no evidence of Hardy Weinberg disequilibrium.

Table 1

Baseline characteristics









Age (years)*

10.6 (4.3–18.5)

11.8 (4.3–19.9)


Pubertal stage



31 (59%)

37 (42%)



15 (28%)

34 (38%)



7 (13%)

18 (20%)


Calcium intake(mg/day)*

1208 (460–1897)

1187 (302–4356)


Physicalactivity (hours/week)*

9.3 (1.5–23)

7.7 (1.8–22)


*Mean (range)

Table 2

VDR haplotype and genotype distribution



Total (%)






141 (49.6)




103 (36.3)




38 (13.4)




2 (0.7)













1, 2

53 (37.2)



1, 1

31 (21.8)



1, 3

21 (14.8)



2, 2

18 (12.7)



2, 3

14 (9.8)



3, 3

3 (2.1)



1, 4

2 (1.4)




142 (100)


HWE p-value




*Ranking according to frequency in the Caucasian population

HWE = Hardy Weinberg Equilibrium

Anthropometry and Bone Characteristics

We first analyzed the subjects grouped by their VDR genotype based on copy number for the three most frequent haplotype alleles, haplotype 1, 2, and 3. The genotype effects by VDR haplotype ‘3’ allele are reported in Table 3. No clear genotype effects of haplotype ‘1’ and ‘2’ were found, except for a weak association between haplotype ‘2’ and BMAD. At baseline, height SDS was significantly increased in VDR haplotype ‘3’ allele carriers with evidence of an allele dose effect (β = 0.35, standard error (SE) = 0.17, P = 0.04), while a similar trend was found at follow-up (β = 0.28, SE = 0.17, P = 0.10).

Table 3

Antropometric characteristics and bone density by VDR haplotype 3 allele at baseline


Number of VDR allele copies


Haplotype allele
















0.06 (0.94)

0.33 (1.15)

1.17 (0.75)




Vertebral width

0.08 (0.89)

0.51 (1.27)

0.41 (0.40)



Bone density



0.00 (1.05)

−0.23 (0.86)

−0.79 (0.61)





0.06 (0.99)

0.04 (0.98)

−0.52 (1.03)



−0.12 (0.94)

−0.03 (1.14)

−0.47 (0.76)



0.05 (0.84)

0.25 (1.29)

0.36 (1.06)


*Mean standard deviation score (SD); β is the regression coefficient: the increment or decrement in SDS per allele copy**Independent sample T-test in case of a dominant test allele effect (0 versus 1&2 allele copies) Vertebral width, mean width of the lumbar vertebral body L2-4; BMAD, bone mineral apparent density; BMD, bone mineral density; BMC, bone mineral content; LS, lumbar spine; TB, total body

Interestingly, we observed a similar trend when we analyzed vertebral body width as another parameter of frame size. At baseline, vertebral body width was 0.4 SDS increased in haplotype ‘3’ carriers vs non-carriers (P = 0.05), although this did not reach significance at follow-up (P = 0.47). No significant associations between the VDR genotypes and BMI and LBM were found.

When we analyzed parameters of bone density, no genotype effects on BMD and BMC were observed at baseline and follow-up (Table 3). However, we observed a consistent association between haplotype allele ‘3’ carriers and decreased BMAD, at baseline as well as at follow-up. Although obvious trends were found, the differences did not reach significance. At baseline, BMAD decreased 0.3 SDS (SE 0.17) per copy of VDR haplotype allele ‘3’ (P = 0.10) and at follow-up (P = 0.11). An opposite trend in BMAD was observed for carriers of haplotype 2 (P = 0.09, data not shown).

No association between the VDR genotypes and physical activity was observed.

Longitudinal Data

We compared baseline with follow-up measurements. The change in height and vertebral body width SDS between baseline and follow-up was not associated with VDR genotype. No correlation was found between VDR genotype and bone gain, expressed as ΔBMC, ΔBMD, and ΔBMAD SDS (data not shown).

Individual RFLPs

We also analyzed the association between bone density and VDR genotype, as defined by the individual BsmI and ApaI restriction site polymorphisms. No consistent effects were found for height, vertebral body width, BMD, or BMC. In concordance with the results of the direct haplotyping for the BsmI RFLP, a B-allele dose effect on BMAD was found at baseline and follow-up (β = 0.25, SE = 0.12, P = 0.05). No differences in BMAD between the ApaI genotype groups were observed.

Interaction of VDR and COLIA1 Genotype

The above-described results identified non-haplotype 3 alleles as risk alleles for decreased height and vertebral body width. In this cohort, we previously found the T allele of the G to T substitution in the polymorphic Sp1 binding site of the collagen IA1 gene to as be associated with decreased height as well as vertebral body width SDS [19]. We therefore analyzed height and vertebral body width SDS according to VDR genotype combined with COLIA1 genotype [Figure 1]. This risk genotype score analysis revealed a dose-dependent decrease in height (β = -0.36, SDS; SE = 0.13; P = 0.006), as well as in vertebral body width (β = -0.45, SDS; SE = 0.13; P = 0.001). We performed multiple linear regression analysis with VDR haplotype ‘3’, COLIA1, and an interaction term (VDR3 × COLIA1) as independent variables, and height and vertebral body width as dependent variable. The interaction term was not significant in the model for height (P = 0.63) nor for vertebral body width SDS (P = 0.11).


In this study we investigated the relationship among VDR genotype, anthropometry, and bone density in children and young adults. We found a significantly increased height and lumbar vertebral body width SDS in carriers of a particular haplotype of three adjacent polymorphic sites (haplotype ‘3’ = bAT). In addition, we found this haplotype to be associated with decreased BMAD, although this trend did not reach significance. This suggests that VDR polymorphism affects several frame size characteristics, in addition to an effect on apparent BMD, that may influence areal bone density.

Our observations of the effect of VDR genotype on aspects of frame size are in line with the previously reported associations of VDR genotype with birth weight, height, growth to final height, and bone area [15, 16, 27]. Taken together these observations suggest a role of the vitamin D endocrine system in regulating growth of the skeleton which is reflected in genotype-dependent differences in frame size. Vitamin D is likely to affect growth as chondrocytes in the growth plate have receptors for 1,25-dihydroxyvitamin D3 [28, 29] and vitamin D is one of the regulators of chondrocyte proliferation in the growth plate [30, 31]. Indeed, growth retardation is a well-known clinical feature of rickets [32, 33]. Vitamin D-dependent rickets (VDDR) type I or pseudo vitamin D-deficiency rickets is caused by a mutation in the gene coding for P450c1αa (which catalyzes 1α hydroxylation of 25OHD), while VDDR type II is an autosomal recessive disease caused by loss of function mutations of the VDR. Furthermore, studies with animal models of VDDR type I [34, 35] and type II [36, 37] also show growth retardation. So, clinical and experimental findings strongly suggest that variations in vitamin D levels or VDR influence growth.

With respect to interpretation of our results we must note that the polymorphisms used are anonymous. They are likely to act as markers through linkage disequilibrium for truly functional sequence variation elsewhere in the gene. In view of our analysis of linkage disequilibrium across the VDR gene it is unlikely that the linkage disequilibrium with the Bsm-Apa-Taq haplotypes extends far outside the VDR gene (Fang et al. unpublished data). It is thereby unlikely that another (nearby) gene is explaining the associations found. One approach to partly overcome this drawback is to increase genetic resolution by combining informative alleles in multi-allelic haplotype markers. Uitterlinden et al. [8] developed a direct molecular haplotyping PCR test to monitor three clustered RFLPs at the VDR gene locus simultaneously. This method was used in a study of 1782 elderly Caucasian men and women [8] and in a cohort of 814 young Canadian women [5]. We found VDR haplotype frequencies similar to those reported in these Caucasian populations. Uitterlinden et al. [8] reported an association between haplotype ‘3’ and a mildly decreased BMD; in women, haplotype ‘2’ showed a weak trend towards higher BMD. Rubin et al. [5] showed that haplotype ‘1’ and ‘2’ were significantly correlated with BMD, but found no association for haplotype ‘3’. Together with our findings these results suggest that haplotype ‘3’ may be assigned as a risk allele, while haplotype ‘2’ may be assigned as a protective allele with respect to bone characteristics.

Most studies in children have a relatively limited sample size. Sainz et al. [12] reported that VDR gene alleles predicted bone density in 100 prepubertal girls of Mexican descent. Similar results were found in a group of 75 young adult Finns [38]. A study of healthy girls suggested that the effect of VDR genotype on BMD was associated with environmental factors like calcium intake [39]. These studies indicate the presence of heterogeneity for the associations of VDR with bone density. Identification of the functional allele of the VDR gene in linkage disequilibrium with these anonymous marker alleles is necessary to solve the issue. These results are different, however, from Gunnes et al. [13] and Lorentzon et al. [15] both of whom found no association between VDR polymorphism and BMD. In these cases, areal BMD was assessed by single photon absorptiometry and DXA, while no correction for bone size was made. Although the trend is weak and the power is low, we found that VDR polymorphism was particularly associated with apparent (‘volumetric’) BMD and not with areal BMD. Indeed, Sainz et al. [12] used volumetric density assessed by quantitative computed tomography when they found an association with VDR polymorphisms. Furthermore, SPA and DXA will overestimate BMD in tall stature. Therefore, the fact that haplotype ‘3’ allele carriers are taller, while their BMAD is lower, might neutralize the effect on areal BMD. These findings suggest that the association might be more evident with volumetric density in growing individuals. Thus, areal BMD may be a less suitable end-point for studying effects of VDR genotype.

Generally, differences between studies examining BMD in relation to VDR genotype might be explained in various ways. Firstly, VDR genotype is not involved in determining BMD or BMAD and the findings are due to chance. Secondly, the association is true but there may be ‘allelic heterogeneity’ between populations. Thirdly, environmental and genetic influences might interact, and such influences might be different in various developmental stages, and may also be site-dependent according to physical loading and bone composition.

The change in BMAD or BMD SDS between baseline and follow-up did not associate with the VDR genotypes in the present study. Gunnes et al. [13] also studied bone gain and did not find an association between the individual BsmI polymorphism and bone gain in the forearm. This suggests that children follow their own ‘BMD accrual’ curve, similar to their growth curve.

By combining risk alleles of two candidate genes, we found additive effects of the VDR haplotype ‘3’ and COLIA1 Sp1 polymorphisms, which affect both height and bone size. Carrying either the VDR risk allele or the COLIA1 risk allele had an effect on frame size while carrying risk alleles at both loci further enhanced the effect. The interaction term, however, was not significant, indicating an additive effect rather than a ‘true’ interactive effect. While these are epidemiological observations, this notion is supported by molecular biological experimental evidence. The VDR is a transcription factor that, amongst other factors regulates the expression of the COLIA1 gene [17, 18]. It can thus be hypothesized that VDR-regulated expression of the COLIA1 gene differs across VDR and COLIA1 alleles. Recently, a similar interaction was observed between the VDR gene and COLIA1 gene in the susceptibility for fracture in an elderly population [40]. Although the exact mechanism remains to be elucidated, together these results suggest this intergenic interaction to have effects at different stages in life.

With regard to our study population, some characteristics should be noted. The group of children we studied was a random sample of the previous study population. We only analyzed Caucasian children, so it was an ethnically homogenous group. We found VDR genotypes to be in Hardy Weinberg Equilibrium, suggesting that no severe selection bias has occurred. The effects we observed are considerable in terms of effect size but most were of borderline significance. This reflects the limited statistical power we could develop given the limited sample size. For example, only three children were homozygous for haplotype ‘3’, so results should be interpreted carefully, and larger study populations are needed. For the same reason, no conclusion could be drawn from our study on whether the strength of the VDR genotype effect depends on pubertal stage.

In conclusion, VDR genotype, as defined by haplotypes constructed of the BsmI, ApaI, and TaqI polymorphisms, is associated with height and vertebral body width in Caucasian children and young adults in this study. We observed additive interaction between genotype effects of VDR polymorphism and the COLIA1 Sp1 polymorphism on frame size. VDR is weakly associated with bone mineral apparent density and not with areal bone density. The VDR genotype effect on areal BMD is neutralized by increased height and bone size, therefore only effects on BMAD were found. However, the results should be interpreted carefully because of the limited sample size; larger study populations are needed.


We gratefully thank Pascal Arp for his assistance in DNA analysis.

Copyright information

© Springer-Verlag 2003