Serum 25-hydroxyvitamin D levels and bone mineral density in 16–20 years-old girls: lack of association


MD PhD Gunnar Sigurdsson , ProfessorDepartment of Internal Medicine, Reykjavik Hospital, University of Iceland, 108 Reykjavik, Iceland (fax: + 354–525-1552,


Kristinsson JÖ, Valdimarsson Ö, Sigurdsson G, Franzson L, Olafsson I & Steingrimsdottir L. (Reykjavik Hospital, University of Iceland, Reykjavik, Iceland). Serum 25-Hydroxyvitamin D Levels and bone mineral density in 16–20 years old girls: lack of association. J of Intern Med 1998; 243: 381–88

In conclusion, our results indicate that the vitamin D requirements for normal bone mineral accrual in adolescent and young adult girls may be lower than presumed and suggest that calcium and phosphate concentrations in plasma may be regulated in this age group to some extent by other factors which needs to be further clarified. However, the important question of the effect of repeated, cyclical vitamin-D nadir on bone mass in young people needs to be explored in a longitudinal survey.

Objective. Hypovitaminosis D has been shown to be associated with low bone mineral density in middle-aged and elderly women. The aim of this study was to evaluate whether such an association might exist in adolescent and young adult girls, approaching peak bone mass.

Design. Cross-sectional study carried out in late winter.

Setting. Reykjavik area at latitude 64°N.

Subjects. Two-hundred and fifty-nine Icelandic Caucasian girls, aged 16, 18 and 20 years, randomly selected from the registry of Reykjavik.

Main outcome measures. Bone mineral density in lumbar spine, hip, distal forearm and total skeleton was measured with dual-energy X-ray absorptiometry (DEXA) and compared with 25-hydroxyvitamin D levels [25 (OH)D] in serum, measured by radioimmunoassay. Calcium and vitamin-D intake were also assessed by a questionnaire.

Results. 18.5% of the girls were below 25 nmol L−1 in serum 25 (OH)D which has been recognized as the lower normal limit for adults. No significant association was found between 25 (OH)D levels and bone mineral density.

Conclusions. Normal calcium and phosphate concentrations in plasma and normal bone mineral density are maintained in adolescent and young adult girls at lower 25 (OH)D levels than published ‘normal’ levels for middle-aged and elderly.


It is well known that severe vitamin D deficiency leads to rickets in children and osteomalacia in adults. In recent years it has been shown that subclinical vitamin D deficiency in the elderly leads to secondary hyperparathyroidism and low bone mineral density (BMD) [ [1][2][3]1–4]. Plasma 25-hydroxyvitamin D [25 (OH)D] is the most sensitive clinical index of vitamin D status and has been found to be positively related to BMD in middle-aged and elderly women [ 2, 5, 6]. Vitamin D is derived from dietary intake and cutaneous synthesis initiated by solar irradiation of the skin. At higher latitudes there is reduced cutaneous production of vitamin D for 4–6 months of the year [ 7], and seasonal variation of vitamin D metabolites, with high summer/fall and low winter values, is well established and is generally more pronounced in Northern latitudes [ [8][9][10][11][12][13]8–14]. Some [ [15][16][17]15–18], but not all [ 19], studies have indicated a seasonal difference in BMD in adults. The relation between BMD and 25 (OH)D in adolescence has not been studied thoroughly. The 25 (OH)D concentration in the winter is likely to be low in the inhabitants of Reykjavik, Iceland, with latitude 64°N. Whether this has any relationship with high prevalence of fractures in the Nordic countries has not been excluded. It was therefore of interest to study whether there is an association between 25 (OH)D and BMD in a random sample of adolescent and young adult girls in this area. The relationship between BMD and calcium intake was also evaluated as was the association between BsmI Vitamin D receptor (VDR) gene polymorphism and BMD which has been implicated with different bone mass in adults and elderly [ 20].

Materials and methods


Three hundred and sixty-six girls aged 16, 18 and 20 (born 1980, 1978 and 1976) were selected randomly from the registry of Reykjavik and invited to participate in the study. Fifty-one declined to participate for personal reasons, 49 could not be reached and seven were pregnant. Two hundred and fifty-nine participated (71%), and informed written consent was obtained from the girls or their parents. Five were excluded because of insufficient data, one because of a medical disease and four had taken medication known to affect bone metabolism, (i.e. steroids or anticonvulsants). The final study group included 249 healthy, Icelandic, Caucasian girls, 71 in the 16th year, 60 in the 18th year and 118 in the 20th year. The assessments were performed in February, March and April 1996.

The study was approved by the Ethical Committee of the Reykjavik Hospital.

Assessment of bone mineral density

Dual-energy X-ray absorptiometry (DEXA) (Hologic QDR-2000 plus, Hologic Inc. Waltham, MA, software version 8.00 A) was used to measure bone mineral content (BMC, g) and bone area (BA, cm2) at the lumbar vertebrae (L2-L4), the proximal end of the left femur, dominant forearm and total skeleton. From these two measurements, areal bone mineral density (BMD, g/cm2) was calculated. The reproducibility for replicate scans of the same individual with repositioning was 1.0% at the spine and total skeleton, 0.9% for forearm and 1.6% at the proximal femur (CV, coefficient of variation). Phantom calibration was made daily. The total radiation dose received for a complete set of scans was about 11.4 μSv.

Measurements of the spine were made at L2-L4, of the proximal femur at three sites (femoral neck, trochanteric and intertrochanteric region, the sum called total hip) and of the dominant forearm at three sites (1/3 distal, mid-distal and ultra-distal regions). The sum of the three forearm sites was called total forearm.

Assessment of anthropometric findings

Measurements of height (cm) and weight (kg) were made on the same day as bone mineral measurements. Height was measured to the nearest 0.5 cm with patients barefoot using a SECA stadiometer. Weight was measured to the nearest 0.5 kg with subject barefoot and in light clothing using a SECA electronic scale.


Data relating to lifestyle factors were collected by a questionnaire. These included age at menarche, regularity of menstrual cycle, use of oral contraceptives, medical history, smoking habits and consumption of alcohol. The 20-year-old girls were asked if they had been exposed to whole body artificial ultraviolet light during the last three months, with the answers recorded as yes or no. Data are not available for the 16 and 18-year-old girls. The menarcheal age was defined as the time since menarche.

Assessment of dietary intake

Energy, calcium, vitamin D and protein intake were assessed using a validated, semiquantitative food frequency questionnaire with internal checks, developed by Overvad et al [ [21]21, 22]. and adapted by the Icelandic Nutrition Council. The adapted questionnaire includes 150 different food items and has been validated against a weighted 7-day food record from 40 individuals (unpublished). Portion sizes were estimated by allowing patients to select a picture most closely resembling their usual portion from a series of photographs depicting three different portion sizes of common foods. Fluids, including milk and cod liver oil, were assessed using household measures. Energy, protein, calcium and vitamin-D were calculated using the national nutrient database.

Biochemical measurements

Serum calcium and phosphate concentrations were measured by a Vitros 750 analyser (Clinical Diagnostics Division of Johnson & Johnson, Rochester, NY, USA). The interassay coefficient of variation (CV%) for calcium was 2.73% and 2.05% at 2.20 mmol L−1 and 2.92 mmol L−1, respectively, and for phosphate 1.77% and 4.55% at 1.13 mmol L−1 and 2.20 mmol L−1, respectively.

Serum 25-hydroxyvitamin D was determined by radioimmunoassay (INCSTAR Corporation, Stillwater, MN, USA). The reference values are 25–100 nmol L−1. The lower detection limit is 7.5 (3.8) nmol L−1. This assay measures both D2 and D3 metabolites. At levels of 34.4 nmol L−1 and 133.4 nmol L−1 the interassay coefficients of variation were 11.9% and 12.5%, respectively. The girls, aged 16 and 18, were measured in February, and the 20-year-old in March and April.

Vitamin D receptor gene polymorphism

Genomic DNA was isolated from 10 mL of peripheral blood as described [ 23]. The polymerase chain reaction primers for the amplification of the 3′-untranslated region of the VDR gene comprising the polymorphic BsmI site have been described elsewhere [ 20]. Taq polymerase and BsmI were obtained from Pharmacia Biotech (Allerød, Denmark) and AMERSHAM Inc., respectively. The amplification products were digested with BsmI and restriction fragments separated by electrophoresis on 2% agarose gels containing ethidium bromide. VDR genotypes were determined according to the presence (b) or absence (B) of the polymorphic BsmI restriction site.

Statistical analysis

Testing whether Pearson's correlation coefficients (r) were significantly different from zero was performed by assuming t = r n-2/1-r2, to follow the Student's t-distribution with n-2 degrees of freedom, if there were no true correlations (n, number of observations). The significance of differences between correlation coefficients was evaluated by assuming z = 1/2 ln [ (1 + r)/(1-r)] to follow normal distribution with variance 1/(n-3). A backward, stepwise multiple linear regression was performed in a hierarchical model using the SPIDA program package [ 24]. P-values less than 0.05 were considered significant.

The dependent variables used were BMC and BMD. The independent variables were body weight, height, energy, protein and calcium intake, 25 (OH)D levels in serum, menarcheal age and months of smoking. Comparison between means of groups were performed using a Student's t-test.


General characteristics

Table 1 summarises patients' characteristics, including weight, height, total energy, calcium and vitamin D intake and biochemical values, serum 25 (OH)D, calcium and phosphate concentration. The intake of calcium was similar amongst the girls at all ages. The mean consumption of the 16-year-old was highest at 1613 mg day−1, but was not significantly different (t-test, P > 0.1) from the other groups. The vitamin D intake was, on the other hand, highest in the girls aged 20, but the difference was not significant (t-test, P > 0.1). The total range of vitamin D intake for the three groups was 0.9–41.2 μg/day. The mean serum 25 (OH)D concentration was 43.9 nmol L−1 for all ages, and there was no significant difference (t-test, P > 0.1) between the groups, the range being 1.7–132.0 nmol L−1. Serum calcium and phosphate were both significantly higher in the youngest age group compared to the other two (P < 0.01–0.001). The mean protein intake was 90 g day−1. All the girls except one had entered menarche. The mean age at menarche was 13 years, 1.6 months.

Table 1.  Characteristics of the girls studied (mean ± SD) Thumbnail image of

25-hydroxyvitamin D

Figure 1 shows an insignificant correlation between 25 (OH)D and total BMD for the whole group. The only significant correlation with 25 (OH)D was found for total forearm BMC and BMD in the 16-year-old girls (r= 0.30 and 0.27 P < 0.05, respectively, in univariate analysis and in multivariate analysis P < 0.01).

Figure 1.

The correlation between serum 25 (OH)D in nmol L−1 and total skeleton BMD (g/cm2) for the whole study group (r= 0.03, n.s.).

In a univariate analysis a significant correlation was found between vitamin D intake and serum 25 (OH)D concentration (r= 0.25; P < 0.001: total group, similar in all age groups). This relation is shown in Fig. 2, which displays the distribution of the vitamin D intake and serum 25 (OH)D concentration in the total group. Forty-six (18.5%) of the girls had a serum 25 (OH)D concentration below 25 nmol L−1, a level, which commonly is defined as the lower normal limit for adults [ 2]. Forty of those 46 girls (87%) consumed less than 10 μg/day of vitamin D, which is the recommended daily intake in many countries for teenagers [ 25]. The bone mineral measurements of this subgroup with low serum 25 (OH)D levels were not significantly different from those of the girls with normal 25 (OH)D concentration (> 25 nmol L−1) or compared with those with a concentration greater than 60 nmol L−1 (n<> = 51). The serum calcium and phosphate concentrations and the Ca × P product were not different in the patients who had a 25 (OH)D concentration below 25 nmol L−1. The girls in the oldest age group who had been exposed to ultraviolet light from sun lamps for the last three months (n<> = 75) had a 55% higher concentration of 25 (OH)D, compared with the others who were not exposed (52.7 nmol vs−1. 33.9 nmol L−1; P < 0.01).

Figure 2.

Correlation between vitamin-D intake and 25 (OH)D concentration in the total group (n= 237; r= 0.25; P < 0.001). The bold horizontal line defines the lower normal limit of 25 (OH)D (25 nmol L−1), and the bold vertical line the recommended dietary allowance for teenagers (10 (μg) [ 25].

Calcium intake and BMD

Table 2 presents bone mineral content (BMC) and bone mineral density (BMD) measured at different sites along with the correlation coefficients between these measurements and calcium intake. The univariate analysis showed no significant correlation between calcium intake and BMC or BMD in any age group ( Table 2). Only in the subgroup of 21 girls aged 18 who consumed less than 1000 mg day−1 of calcium, a significant correlation between bone mineral density of total hip and calcium intake was found (r= 0.50; P= 0.02). The r-value for total skeletal BMD and BMC in this group was 0.42 (P= 0.051). This subgroup was, however, too small for multivariate analysis.

Table 2.  Bone mineral content (BMC, g) and bone mineral density (BMD, g/cm2) in the study population (mean ± SD) and the coefficients of correlation between calcium intake and BMC or BMD in a bracket above Thumbnail image of

In a multivariate analysis the only significant correlations were in the youngest age group, between spine BMD and calcium intake (P= 0.02) and in the total group between total forearm BMD and calcium intake (P= 0.015).

When the total group was divided into tertiles with respect to calcium intake, the mean total skeletal BMC or BMD between the tertiles was not significantly different (P= 0.37), the lowest tertile consumed less than 800 mg (n<> = 30) with the mean total skeletal BMD of 1.019 g cm−2, the middle tertile 800–1200 mg (n<> = 65) with the mean BMD of 1.025 g cm−2 and the highest tertile more than 1200 mg of calcium daily (n<> = 142) with mean total skeletal BMD of 1.032 g cm−2.

In the subgroup of girls (n<> = 9) with 25 (OH)D levels < 25 nmol L−1 and calcium intake less than 1000 mg day−1, the mean total BMD was 1.019 g cm−2 or not significantly different from the rest of the girls (P > 0.2). No significant correlation was found in a univariate or multivariate analysis between protein intake or calcium/protein ratio and BMD.

VDR polymorphism

The frequencies of the VDR genotypes amongst the 249 girls were BB 22.1%, Bb 44.6% and bb 33.3%. There was no significant difference in mean BMC or BMD at any of the skeletal sites between the three genotypes in any age group (see Table 3). There was also no difference in 25 (OH)D levels between the VDR genotypes.

Table 3.  Frequencies of VDR genotypes and bone mineral density (BMD, g/cm2), at spine, femur and forearm in relation to genotypes (mean ± SD). The total group (n 249) Thumbnail image of


There was no significant correlation between alcohol consumption or use of oral contraceptives and BMD or BMC. One-hundred and twelve girls (45%) had at sometime used the contraceptive pill, 83 (70%) of the oldest age group. Eighty-four (34%) girls had smoked (mean 0.5 package-years). Smoking was most frequent in the group aged 18, 29 (45%) had smoked (mean 0.7 package-years). In that group there was a significant negative correlation in univariate analysis between smoking and BMD and BMC at total hip and total body (r= 0.32–0.38; P < 0.01). There was also a significant negative correlation between weight and smoking (r=−0.28; P < 0.05) and in a multivariate analysis including body weight, smoking was still significant in this group of 18-year-old (P < 0.005 for BMD and P < 0.05 for BMC).


This cross-sectional study of girls, aged 16, 18 and 20, found no correlation between 25 (OH)D levels and bone mineral density The total skeletal BMD increased by 5% and total skeletal BMC by 12% from age 16 years to 20 years. The study group was a random sample from an area (Reykjavik) of high latitude 64°N. The study was performed late in the winter when the 25 (OH)D levels are supposed to be at the lowest level. Our hypothesis was that the BMD means would differ in girls with low and high 25 (OH)D levels. Some studies in adults have found a positive relationship between 25 (OH)D levels and BMD [ 2, 5, 6], with an increase in serum 25 (OH)D of 40 nmol L−1 associated with a 5–10% increase in bone density in 45–65-year-old women [ 5]. Other studies have reported a seasonal variation in BMD in adults [ [15][16]15–17]. Krall et al. have reported an inverse relation between parathyroid hormone levels and 25 (OH)D levels in postmenopausal women [ 26] and Dawson-Hughes et al [ 18]. significantly reduced late wintertime bone loss in healthy postmenopausal women with vitamin-D supplementation.

Hypovitaminosis D has been defined as a serum 25 (OH)D level below 25 nmol L−1 [ 2]. In this study a surprisingly large group of girls (18.5%) had a 25 (OH)D level below the lower normal limit (25 nmol L−1). However, only seven girls in our study group were below 10 nmol L−1. Although the 25 (OH)D levels were low in part of the group in this study, their mean BMD was not significantly lower. Other Scandinavian studies of young adults describe hypovitaminosis D in about 4% to 9% during winter and in up to 5% during summer [ [11]11, 14].

The Ca × P product was the same in the girls with low and high 25 (OH)D levels, suggesting that factors other than vitamin D are effective in maintaining these concentrations in the blood which are essential for normal bone mineralization [ 27]. One of these factors might be growth hormone which has been shown to increase calcium and phosphate absorption from the gut [ 28], and the levels of which are high in adolescence [ 29]. It also increases the conversion of 25 (OH)D into 1,25 (OH)2D [ 30], and the latter has been shown to be higher in growing adolescents [ 31]. Another explanation may be the effect of estrogen which is known to augment 1, alfa-hydroxylase activity [ 32]. In our study, serum calcium and phosphate levels were significantly higher in the 16-year-old girls who presumably had higher growth hormone levels. Higher serum calcium levels has also been described in pubertal girls [ 10]. Nutritional osteomalacia has rarely been described in adolescents, mainly in Asian immigrants in the UK [ 33].

Our data suggest that 10 μg/day of vitamin-D in the diet is required to achieve 25 (OH)D levels above the lower normal limit (25 nmol L−1), but it is notable that even at this northern latitude (64°) only a small part of the circulating 25 (OH)D is obtained from dietary intake with the correlation coefficient between vitamin-D intake and 25 (OH)D concentration of 0.25 (R2 = 0.06) (see Fig. 2). However, measurement errors might probably lead to a falsely low R2.

Most earlier studies on 25 (OH)D concentrations have been performed in adults. Our levels of 25 (OH)D were lower than those observed in earlier studies during the winter on adults in Denmark [ [11]11, 13], Sweden [ 10] or USA [ 26] and on adolescents in Norway [ 31]. On the other hand, our 25 (OH)D levels were higher than those found during the winter in a study on adolescents in Finland [ 8] and in middle-aged women in Britain [ 5]. The low winter levels of 25 (OH)D in young Icelandic girls seem to last from December to April (unpublished results).

In our study group, the calcium intake was, on average high with the mean level well above the recommended dietary allowance (RDA) for all age groups [ 25] although 60 girls (24%) were under the RDA for their age group. This high calcium intake might partly overcome the low 25 (OH)D levels. However, even in the subgroup with low 25 (OH)D levels and calcium intake below 1000 mg (n<> = 9), the mean BMD seemed not to be any lower, suggesting the overwhelming importance of hormonal factors in maintaining BMD in adolescents.

There may be a threshold effect in calcium balance [ 34], so that a calcium intake above a certain level does not increase bone mineral density. In the present study there was generally no correlation between calcium intake and BMD except in the subgroup of girls aged 18 with a calcium intake of less than 1000 mg day−1, where the r-value between BMD of total hip was 0.5 (P= 0.02). This might suggest a threshold of 1000 mg for the calcium requirement in this age group, as our previous results [ 35] as well as other studies [ 34] have suggested. In multivariate analysis a significant correlation was, however, found in the 16-year-old girls between calcium intake and spinal BMD and total forearm BMD. The insignificant trend shown in our total study group of increasing BMD with increased calcium intake may be real, but a study of this size would not have the power to detect small differences such as these.

In this group of girls approaching peak bone mass, the vitamin D receptor gene polymorphisms seemed not to be associated with different BMD levels. Most studies have been on older patients, and results have been conflicting, some suggesting a relationship between VDR and BMD [ [20]20, 36] others not [ [37]37, 38]. We have found a significant effect of the bb genotype being associated with increased BMD in adults [ 39]. We did not find this relation in this young age group, which may reflect an effect of long–term interaction of VDR genotypes and nutrition on the maintenance of BMD in adults, rather than the attainment of peak bone mass.



We thank D. Oskarsdottir for performing all DEXA measurements, D.N. Magnusdottir for VDR-genotyping, S. Stefansson and S. Valdimarsson for their assistance, H. Sigvaldason for statistical analysis and the lab technicians at the Department of Clinical Chemistry for their excellent assistance. This study was supported by a grant from the Icelandic Research Council and Reykjavik Hospital Science Fund.