Determinants of Peak Bone Mass: Clinical and Genetic Analyses in a Young Female Canadian Cohort


  • Dr. Laurence A. Rubin,

    Corresponding author
    1. Division of Rheumatology and the Multidisciplinary Osteoporosis Program, Women's College Campus, Sunnybrook and Women's, College Health Sciences Centre; Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    2. Department of Medicine, The Toronto Hospital, Toronto, Ontario, Canada
    • Women's College Campus Sunnybrook and Women's Health Sciences Centre 60 Grosvenor St., Suite 416 Toronto, Ontario M5S 1B6, Canada
    Search for more papers by this author
    • Drs. Rubin and Hawker are co-first authors of this manuscript.

  • Gillian A. Hawker,

    1. Division of Rheumatology and the Multidisciplinary Osteoporosis Program, Women's College Campus, Sunnybrook and Women's, College Health Sciences Centre; Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    2. Department of Medicine, The Toronto Hospital, Toronto, Ontario, Canada
    Search for more papers by this author
    • Drs. Rubin and Hawker are co-first authors of this manuscript.

  • Vanya D. Peltekova,

    1. Division of Rheumatology and the Multidisciplinary Osteoporosis Program, Women's College Campus, Sunnybrook and Women's, College Health Sciences Centre; Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Lynda J. Fielding,

    1. Division of Rheumatology and the Multidisciplinary Osteoporosis Program, Women's College Campus, Sunnybrook and Women's, College Health Sciences Centre; Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Rowena Ridout,

    1. Division of Endocrinology, Women's College Campus, Sunnybrook and Women's College Health Sciences Centre; Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • David E. C. Cole

    1. Department of Medicine, The Toronto Hospital, Toronto, Ontario, Canada
    2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada
    3. Department of Paediatrics (Genetics), University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author


Peak bone mass has been shown to be a significant predictor of risk for osteoporosis. Previous studies have demonstrated that skeletal mass accumulation is under strong genetic control, and efforts have been made to identify candidate loci. Determinants of peak bone mass also include diet, physical activity, hormonal status, and other clinical factors. The overall contribution of these factors, genetic and nongenetic, and their interaction in determining peak bone density status have not been delineated. Six hundred and seventy-seven healthy unrelated Caucasian women ages 18–35 years were assessed. A detailed, standardized interview was conducted to evaluate lifestyle factors, menstrual and reproductive history, medical conditions, medication use, and family history of osteoporosis. Bone mineral density (BMD) was measured at the lumbar spine (L2–L4) and the femoral neck (hip) using dual-energy X-ray absorptiometry. Genotyping of the vitamin D receptor (VDR) locus at three polymorphic sites (BsmI, ApaI, and TaqI) was performed. In bivariate analyses, BMD at the lumbar spine and hip was positively correlated with weight, height, body mass index (BMI), and level of physical activity, both now and during adolescence, but negatively correlated with a family history of osteoporosis. Hip, but not spine BMD, correlated positively with dietary intake of calcium, and negatively with amenorrhea of more than 3 months, with caffeine intake, and with age. Spine, but not hip BMD, correlated positively with age and with number of pregnancies. VDR haplotype demonstrated significant associations with BMD at the hip, level of physical activity currently, and BMI. In multivariate analysis, independent predictors of greater BMD (at the hip or spine) were: age (younger for the hip, older for the spine), greater body weight, greater height (hip only), higher level of physical activity now and during adolescence, no family history of osteoporosis, and VDR genotype (hip only). Weight, age, level of physical activity, and family history are independent predictors of peak BMD. Of these factors, weight accounts for over half the explained variability in BMD. VDR alleles are significant independent predictors of peak femoral neck, but not lumbar spine BMD, even after adjusting for family history of osteoporosis, weight, age, and exercise. However, the overall contribution of this genetic determinant is modest. Taken together, these factors explained ∼17% and 21% of the variability in peak spine and hip BMD, respectively, in our cohort. Future research should be aimed at further evaluation of genetic determinants of BMD. Most importantly, understanding the critical interactive nature between genes and the environment will facilitate development of targeted strategies directed at modifying lifestyle factors as well as earlier intervention in the most susceptible individuals.


Epidemiologic and twin studies suggest that heritable factors account for up to 80% of the variability in bone mineral density (BMD),(1-5) a major determinant of bone strength and resistance to fracture.(6) In twin studies, monozygotic pairs show considerably less variation in BMD than do dizygotic pairs.(7,8) BMD of mothers in early menopause and their premenopausal daughters are distributed similarly around age-adjusted means, and BMD in daughters of mothers who have suffered fractures is lower than in daughters of women who have not fractured.(9,10) While twin data and family studies are useful first steps in delineating the contribution of genetic factors to this quantitative phenotype, such data cannot be extrapolated to populations. Furthermore, the impact of any specific genetic factor may vary considerably between ethnic and racial groups as a consequence of exposure to, or interaction with, population-specific clinical and environmental determinants.(11)

The genetic influence on BMD may lessen, or at least be less appreciable, with age, particularly after menopause. Therefore, their greatest impact is likely to be as predictors of peak BMD.(12) Several candidate genes have been examined, but only the gene encoding the vitamin D receptor (VDR) has been subjected to extensive scrutiny worldwide.(13) Reports examining the relationship between polymorphic alleles of the VDR gene and BMD have yielded conflicting results.(14,15) Even in studies where a significant correlation is demonstrable, the genotype–phenotype relationship is not uniform, with evidence for population-, age-, and site-specific variability. Furthermore, the majority of these studies have focused on postmenopausal populations, in whom interactions may be confounded by age and estrogen-related covariates. However, based on these studies, it appears that VDR genotype accounts for a small but statistically significant fraction of BMD variability.(16,17)

To date, there have been few large-scale studies of peak BMD determinants in unselected, unrelated young women. To investigate the relationships and relative contributions of recognized clinical and genetic determinants to peak BMD attainment, we recruited a large group of Canadian Caucasian women, age 18–35 years, for study. The associations between environmental, hormonal, and nutritional factors, polymorphic alleles of the VDR, and BMD were assessed. Multivariate regression analyses were performed to determine which of these factors were significant independent predictors of peak BMD.



Through advertisements in local newspapers and posted flyers, 993 young women were recruited for screening by telephone interview. Of these, 202 were excluded: 36 were ineligible because of age or comorbid conditions known to be associated with secondary bone loss (Crohn's disease with a history of prior bowel surgery, symptomatic hyperthyroidism, rheumatoid arthritis, bilateral oophrectomy, or use of systemic corticosteroid therapy for more than 3 months duration at any time in the past); 55 refused to participate; and 111 did not attend the scheduled interview and test session. Of the remaining 791, 94 women with non-Caucasian background were not included in this initial analysis. Only unrelated individuals were included, and data on a total of 677 women, with equivalent numbers of subjects in each of the three age groups (18–25, 26–30, and 31–35 years) were studied. The study protocol was approved by the Ethics Review Board of Women's College Hospital (Toronto, Ontario, Canada).

After obtaining written consent, each subject completed a standardized, pretested questionnaire about lifestyle factors (daily calcium intake based on diet and use of calcium supplements, level of physical activity, smoking, and alcohol intake), menstrual and reproductive history, past history of specific medical conditions (eating disorders, endometriosis, malignancy, asthma, inflammatory bowel disease, thyroid disease, liver disease, and psoriasis), current and prior medication use (including short courses of corticosteroids and use of anticonvulsants), history of fractures, and family history of osteoporosis. Daily dietary calcium intake was based on self-reported intake of both dairy and nondairy sources of calcium using a food frequency questionnaire and analyzed both as a continuous variable and by quartiles. To evaluate the level of physical activity, each woman was asked to report their overall level of activity now and as an adolescent (physically active, not physically active), their level of activity at work (work: on a 4-point categorical scale from “sedentary” to “predominantly manual/active all day”), distance walked on average each week (walk; on a 4-point categorical scale from “less than 1/4 km per week” to “more than 6 km per week”), whether they engaged in any regular activity long enough to “work up a sweat” (sweat: yes/no), and the time engaged in recreational exercise per week (recreational exercise: a 4-point categorical scale from “none” to “2+ hours per week of heavy exercise, such as aerobics”). A composite physical activity score was then calculated for each woman as the sum of response scores for physical activity items (work score + walk score + recreational exercise score + sweat score) in tertiles (0 = inactive; 1 = moderately active; 2 = very active).

All self-reported fractures were documented by site, age, and mechanism by which they occurred and then classified as either traumatic or low trauma using World Health Organization criteria.(18) Family history of osteoporosis was defined as a self-reported family history of any of the following: physician diagnosis of osteoporosis, presence of a Dowager's hump, height loss of more than 2 in, or history of atraumatic fracture in one or more first-degree (mother/father/sister/brother) or second-degree (maternal or paternal grandparent, uncle or aunt) family members. Counts of first- and second-degree relatives were summed to create a categorical variable termed “number of relatives with osteoporosis” (none, 1 relative, or 2 or more relatives). Body mass index (BMI) was calculated as (weight in kilograms)/(height in meters)2 using current measurements.


BMD at the spine (L2–L4) and hip (femoral neck, Ward's triangle, trochanter) were assessed by dual-energy X-ray absorptiometry using a DPX-L absorptiometer (Lunar Corp., Madison, WI, U.S.A.). BMD was reported as grams per square centimeter as well as the Z score (number of SD above or below the mean BMD for age/weight-matched controls) and T score (number SD above or below the mean BMD of young white female controls aged 20–45 years). Only composite lumbar (L2–L4) and femoral neck values are reported here.

VDR genotyping and haplotyping

Genomic DNA was extracted from peripheral blood lymphocytes and polymerase chain reaction amplified using specific VDR oligonucleotide primers as previously described.(19) Amplicons were subjected to restriction enzyme digestion with BsmI, ApaI, or TaqI sequentially, and the final products separated by gradient gel electropheresis followed by visualization using a silver staining technique.(20) The presence or absence of the restriction sites are denoted as b/B, a/A and t/T, respectively. Of the eight potential haplogroups, we have identified seven.(20) The prevalent alleles in the Caucasian population are haplogroups H1 (baT or the T allele) and H2 (BAt or the t allele). Tests for Hardy–Weinberg equilibrium(21) did not show significant deviation from the expected frequencies (p > 0.5 for the alleles at any of the three restriction sites; data not shown).

Statistical methods

All clinical data were entered into a Paradox database (Version 5.0; Borland International Inc., Scotts Valley, CA, U.S.A.); double data entry was performed to ensure data accuracy. Genetic information was entered into a separate Microsoft Excel database and similarly validated. These data sets were merged and analyses performed with SAS software Version 6.12 (SAS Institute Inc., Cary, NC, U.S.A.). Correlations between bone density (L2–L4, femoral neck) and VDR genotype, lifestyle factors, menstrual and reproductive factors, comorbid conditions, and fractures were assessed using Spearman's rank correlation coefficients. Interactions between explanatory variables were also assessed using Spearman or Pearson rank correlations (continuous variables) or chi-square and Fisher exact tests for categorical variables. Because we postulated that VDR genotype may exert its effect on BMD through modification of the effect of lifestyle and other factors on bone turnover, we additionally assessed for associations between VDR genotype and other potential predictors of BMD. All potential explanatory variables were then entered into separate stepwise multivariable linear regression analyses to determine independent predictors of lumbar spine and femoral neck BMD. Statistical significance was at a two-tailed level of 0.05.



The mean age was 27.5 years; more than half were single (64.9%), and most (79.4%) had received post–secondary school education. The mean weight (kg) was 63.1 (min–max = 38.6–119.1), mean height (m) 1.64 (1.47–1.88), and mean BMI (kg/m2) (± SD) was 21.69 (±6.26). None had previously been assessed for osteoporosis. Most (70.9%) had never smoked, and the mean alcohol consumption was ≤1 alcoholic drink/day (89.4%).

The mean age at menarche for the cohort was 12.7 years and excluding the first year after, over half (55.3%) reported their menstrual periods had always been regular, excluding periods of time when using oral contraceptives. Among those reporting at least one episode of amenorrhea (>3 months), 38% had experienced more than 6 months of amenorrhea. Over one-third of the women (268, 39.6%) were currently using oral contraceptives, while 43.9% and 16.4% were past- or never-users of oral contraceptives, respectively (2 women could not recall). One woman had undergone hysterectomy with unilateral oophrectomy, and six had received exogenous estrogen (n = 3) and/or progestin therapy (n = 5) as treatment for endometriosis. None were receiving or had received standard postmenopausal hormone replacement therapy. One-quarter of the cohort (n = 159) had experienced at least one pregnancy (min–max = 1–5), and of these women, the majority (90.5%) had breastfed. The mean total daily calcium intake, based on diet and use of calcium supplements, was 562 mg (range 0–2630 mg). The women consumed on average 1.5 cups per day of coffee, tea, or cola.

Most women described themselves as physically active currently (79.8%) and as an adolescent (79.8%). Almost all of the women (91.2%) exercised at least 30 minutes/week, and for most (77.4%) the exercise was sufficiently vigorous to work up a sweat.

Nearly half (338/677, 49.9%) reported at least one previous medical problem, although none were clinically active requiring medication (e.g., corticosteroids) at the time of the study. Self-reported medical histories included ulcerative colitis (n = 11) or Crohn's disease (n = 1), thyroid disease (n = 30), liver disease (n = 3), psoriasis (n = 23), asthma (n = 130), endometriosis (n = 22), infertility requiring drug therapy (n = 19), or a history of an eating disorder (n = 64). Four women had taken stable doses of phenytoin for more than 1 years' duration. Nearly one-third of the women had experienced at least one fracture, but only 28 fractures in 24 women could be considered to have been low trauma by World Health Organization criteria.(18)

Bone mineral density

Mean lumbar spine (L2–L4) and femoral neck BMD were 1.19 g/cm2 (SD = 0.13; min–max = 0.70–1.59) and 1.01 g/cm2 (SD = 0.12; min–max = 0.68–1.36), respectively. The mean Z score at the lumbar spine was +0.08 (SD 1.03; min–max = −3.52 to +3.44), and at the femoral neck +0.24 (SD = 0.95; min–max = −2.7 to +3.13). The mean T score was −0.06 (SD = 1.09; min–max = −3.44 to +3.22) at the spine and +0.23 (SD = 1.02; min–max = −2.48 to +3.15) at the femoral neck. Both the Z and T scores were normally distributed about the means.

Genotyping and haplotyping of the VDR Locus

Six haplotypes were detected and frequencies of the four most common are listed in Table 1. Composite haplotype frequencies are also listed in Table 1, with results comparable to those reported by Uitterlinden et al. in an elderly Dutch cohort.(19) The overall concordance between b and T (or inversely B and t) alleles in this sample set revealed no significant departures from the expected linkage disequilibrium for the alleles defined by these restriction sites. Furthermore, only the H1(baT) and H2(BAt) haplotypes were significantly correlated with BMD in either the bi- or multivariate analyses. In particular, we found no correlation for H3 (bAT), as reported by Uitterlinden et al.,(19) nor for any of the other uncommon recombinant haplotypes (data not shown). We therefore analyzed results by genotype (T/t) only.

Table Table 1..  Genotype/Haplotype Frequencies*
original image

Results of bivariate analyses

For lumbar spine BMD, significant positive correlations were found with age, weight, height, BMI, and level of physical activity as an adolescent (Table 2). Negative correlations were noted with family history of osteoporosis (number of affected relatives and history of a first-degree male relative with osteoporosis, but not a first-degree female relative). At the hip, strong positive correlations were noted between BMD and weight, height, BMI, and calcium intake (Table 3). In contrast to vertebral BMD, both level of physical activity as an adolescent and currently (all variables) were positively correlated with hip BMD. Significant negative correlations with BMD at this site were found for age, family history of osteoporosis, and caffeine consumption. VDR genotype was significantly correlated with BMD at the hip only. There were small differences between correlations for each of the polymorphic sites. The correlation was greatest for TaqI (p = 0.001; Fig. 1), but significant correlations were also demonstrated for BsmI and ApaI (data not shown).

Figure FIG. 1..

Differences in mean ± SEM Z score for lumbar spine (LS) and femoral neck (LF) for each TaqI genotype (TT, Tt, tt).

Table Table 2..  Bivariate Analysis: Lumbar Spine
original image
Table Table 3..  Bivariate Analysis: Femoral Neck
original image

A significant association was observed between BMD and calcium intake (quartiles) at the femoral neck (Table 3), so we assessed the combined effect of calcium intake and genotype on BMD. As shown in Fig. 2, there is a significant interaction between the two explanatory variables (p < 0.01). A significant interaction was also observed between VDR genotype (TaqI) and current level of physical activity affecting hip and BMD (p = 0.01).

Figure FIG. 2..

Interaction of femoral neck BMD (g/cm2), TaqI genotype and calcium intake. Mean daily calcium intakes (mg/day) listed by quartiles: A > 684, B = 461–683, C = 293–460, D < 292 mg/day.

Results of multivariate analyses

After controlling for all other variables, greater lumbar spine BMD was found to be significantly associated with greater body weight, older age, higher level of physical activity as an adolescent and currently (sweat variable), and no family history of osteoporosis (Table 4). This final model explained 16.6% of the observed variance in lumbar spine BMD. The VDR genotype was not found to be predictive of peak lumbar spine BMD whether the analyses were performed with or without family history of osteoporosis as a covariate.

Table Table 4..  Multivariate Regression Analysis of BMD
original image

For femoral neck BMD, significant independent predictors of higher BMD were greater weight and height, younger age, higher level of physical activity as an adolescent and currently, no family history of osteoporosis, and VDR genotype. After adjusting for these variables, there was a trend suggesting a negative correlation between a history of amenorrhea of greater than 3 months' duration and BMD (p = 0.054). The final model explained 21.3% of the observed variance in femoral neck BMD (Table 4).

Regression analysis was used to examine the effect of each (VDR polymorphism) on BMD, without concomitant adjustment for self-reported family history. The results of these analyses were not different from those reported above. Similarly, we performed regression analyses excluding those subjects with any past history of medical conditions that might affect bone and mineral metabolism (i.e., those who used phenytoin or inhaled steroids, or those with any history of thyroid disease, endometriosis, inflammatory bowel disease, asthma, or eating disorders). Again, no substantial differences in the resulting regression models were observed (results not shown).


The risk of developing osteoporosis and the resultant risk of fracture is in large part determined by the peak bone mass achieved in adolescence and early adulthood.(22,23) In this large cohort of young women, a comprehensive assessment of clinical and lifestyle determinants was designed to estimate their relative contribution to BMD. Our results reveal that the most significant independent determinants of BMD were weight, age, physical activity (currently as well as during adolescence), and family history of osteoporosis. While a positive correlation was found between age and BMD in the spine, a negative correlation was found at the hip. Although peak bone mass in women has been documented to occur in late adolescence (age 18 years),(24,25) these data suggest that there may be ongoing accumulation at the lumbar spine beyond this age.

While our final model explained only 21.3% of the total variance in peak BMD at the hip and 16.6% at the spine, body weight accounted for over half of the explainable variability at both sites (Table 4). Increased physical activity, both currently and as an adolescent, explained a further proportion of the variability in BMD, consistent with previously reported studies of younger women,(26-28) as well as older, postmenopausal populations.(29,30) We also detected a significant interaction between VDR genotype, level of current physical activity, and BMD. Geusens et al.(31) have reported that VDR genotype correlated with muscle strength in their cohort of elderly (>70 years) nonobese women. Because the association of VDR and BMD was not present in this cohort after adjusting for muscle strength, they proposed that the VDR effect on BMD was mediated through its action on muscle. In this regard, it is known that VDR is expressed in striated muscle,(32) and it is possible that allelic variation may in some way influence the physiological response of muscle and consequently BMD to environmental factors such as exercise level.

Though calcium intake is reported to be of greatest importance in determining bone mass acquisition during the period of rapid skeletal growth in late childhood and early adolescence,(33) we did not find calcium intake to be a significant independent determinant of peak BMD in our subjects. Welten et al.(34) also could not demonstrate an independent effect of calcium intake on BMD in Dutch subjects followed longitudinally from age 13–28 years. They concluded that weight-bearing activity in males and body weight in females were the most significant positive contributors to peak lumbar spine BMD. In young Finns, femoral neck BMD increased with increasing calcium intake, but calcium intake was not an independent predictor of BMD at either site after adjusting for age and weight.(35) While we did not detect an independent association of BMD with daily calcium intake, we did find evidence of a significant interaction between VDR genotype and calcium intake effects on BMD at the femoral neck, suggesting that VDR may play a role in modulating calcium's effect on BMD.

Previous studies have shown that in older subjects with low calcium intake, significant differences in fractional intestinal calcium absorption and changes in BMD were VDR genotype dependent.(36-38) In premenopausal women, VDR genotype has also been reported to correlate with BMD only in individuals with low calcium intake(39) or with changes in calcium absorption.(40) In a recent prospective analysis of prepubertal girls, Ferrari et al.(41) reported that the BMD response to calcium supplementation varied according to VDR genotype. They found significant differences in BMD accrual at the proximal hip and radius, but not the spine, in subjects with the Bb as compared with bb VDR genotype. Taken together, these findings support a significant interaction between the VDR and calcium metabolism with respect to BMD accumulation, possibly related to a threshold level of calcium intake, below which the VDR effect becomes significant.(36)

Among the variables that failed to show any significant correlation with peak BMD was age at menarche. Even when examined by age groups (18–24, 25–29 and 30–35 years of age), no relationship was found with BMD. While a number of studies have previously identified this as an important determinant of peak bone mass,(42) recent reports point to cumulative ovarian and gonadotropin hormone exposure as the more relevant factor.(43,44) Several of the “menstrual function” variables that were evaluated as potential determinants of peak bone mass in this study may be considered proxies for overall estrogen exposure. These include a history of amenorrhea lasting more than 3 months, history of irregular menstrual periods, number of pregnancies, as well as age at menarche. While no statistically significant associations were found between BMD at either site and any of these variables, there was a trend suggesting a negative correlation between a history of amenorrhea of greater than 3 months' duration and BMD (p = 0.054 at the femoral neck). Future analyses specifically detailing hormonal exposure may clarify any potential relationship between ovulatory function and peak BMD in this cohort.

Since Morrison and colleagues(45) reported an association between VDR alleles and BMD, there have been conflicting follow-up data from around the world. A recent metanalysis(16) and a more extensive literature review(17) support the initial hypothesis that allelic polymorphism in the 3′ region of the VDR gene contributes to BMD, and that this genetic effect is more evident at the hip than spine. Our findings confirm a significant association between VDR genotype and BMD at the proximal femur, but not lumbar spine. In our group, individuals homozygous for the tt (or BB) genotype have the highest femoral neck BMD. These results are consistent with those of Uitterlinden et al.(19) in a large elderly Dutch population sample, and Houston et al.(46) in an older Scottish cohort. Although the first association of VDR genotype and BMD in a group of young (age 20–29 years) women was inverse to what we have reported, multivariate regression analysis demonstrated that VDR genotype was an independent determinant of BMD only at the femoral neck.(47)

The existence of an extended stretch of polymorphic variation in the 3′ region of the VDR gene remains a problematic aspect of any VDR association study. The three polymorphisms show strong, but not complete, linkage disequilibrium. There are significant population-specific differences and the degree to which any one of the 3′ RFLP's can “represent” one or more haplotypes is difficult to accurately determine. Uitterlinden et al.(19) observed that the recombinant haplotype, bAT (H3) was more predictive of BMD in his Dutch cohort than either of the more common baT (H1) or BAt (H2) haplotypes. We did not find any such effect in our younger Canadian cohort, allowing us to analyze the “Tt” genotype as the representative VDR variable.

Our results, along with those of Uitterlinden et al.(19) and Houston et al.,(46) bring to three the number of studies showing an association of the so-called “large B (BAt)” allele with increased BMD. This constitutes further support for the argument that the functional locus is elsewhere and only linked to these RFLP sites. In vitro studies favor the view that none of the 3′ RFLP's—Bsm I, Apa I, or Taq I—leads directly to differential VDR gene expression.(48,49) However, the precise role of flanking VDR polymorphisms—for example, the polyA microsatellite, ∼5 kb pairs downstream of the BbAaTt region in the 3′ untranslated region—remains to be determined.(50)

Our confidence that the “big B” effect is valid in our cohort has been strengthened by confirmatory results in two other regional studies. In 72 patients with primary biliary cirrhosis independently ascertained from the same geographic area, we found a significant positive correlation between BMD and the BAt allele.(51) In a cohort of 205 kidney stone formers with idiopathic hypercalciuria, we have also found that VDR genotype was significantly associated with increased calcium excretion.(52) While it remains possible that all three studies identified significant VDR associations by chance, it is much more likely that we have found valid association with a functional site, either elsewhere in the gene itself or in close genetic linkage to VDR on chromosome 12.

To determine whether we underestimated the independent contribution of the specific candidate genetic factor (VDR), we performed our analyses two ways: by excluding and including the “family history of osteoporosis” covariates. Although exclusion of all family history variables did not substantially change our findings, the component of BMD variability explained by VDR genotype almost doubled, consistent with the previously reported metanalysis.(16)

There are potential limitations to our study. Since it was designed to assess genetic risk for osteoporosis, we expect that women concerned about their risk, and possibly with a known family history of osteoporosis, might have been more likely to volunteer for the study. If so, this should have resulted in a higher that expected proportion of individuals with low BMD. However, this was not the case; BMD values for the group overall were normally distributed, and not statistically significantly different from Lunar dual-energy X-ray absorptiometry age-matched North American control values. While we anticipate that the prevalence of a family history of osteoporosis is higher than would be found in a population-based cohort, this would not be expected to alter our observed findings. Since interviews to determine lifestyle and clinical determinants of peak BMD were conducted prior to determination of BMD and VDR genotype status, we consider it highly unlikely that our results are due to systemic over-reporting of family history, or any other risk factor for osteoporosis, in women with low BMD or in the subsequently determined high-risk genotype group.

In this study, we did not exclude subjects who had been pregnant or who had breastfed, nor those taking oral contraceptives, because we were interested in examining the effects of these factors on peak BMD status. We also allowed participation by individuals who had previously been diagnosed with a variety of medical conditions that potentially might impair bone mass accumulation or secondarily result in bone loss. All subjects were clinically well and not on any medication that might cause a reduction of BMD at the time of the study. Most importantly, a separate analysis of the subgroup that excluded all individuals with any previous medical problems did not significantly alter our results.

In summary, the contribution of both genetic and environmental factors as determinants of peak bone mass is well recognized. Our study (which evaluates the largest cohort of young Caucasian women assessed to date) reveals that about 20% of the variability in BMD can be explained by clinical and environmental factors and a single candidate gene (the VDR). While unmeasured environmental variables may account for a proportion of the variance in peak bone mass, the bulk of the variance is likely due to as yet undetected genetic factors. The polygenic basis for peak bone mass acquisition and subsequent regulation has recently been more extensively characterized in several murine models.(53) This approach may uncover novel candidate genes for further study in both animal and human subjects. Ultimately, more comprehensive characterization of the genetic contribution to human bone mass attainment and regulation will require genome-wide scans. By utilizing genome-wide linkage disequilibrium mapping,(54,55) candidate regions can be identified in large populations and their genes mapped in detail. All such studies will need strong epidemiologic support and corroboration to integrate and clarify the ubiquitous effect of environmental and lifestyle factors on genetic predisposition. As such, our study serves as the foundation for such an effort.


We thank Edra Spevack, Yvonne Garcia, and Elizabeth McDonald for their participation in study implementation (recruiting, interviewing, and data entry), Cheryl Chase who performed all BMD assessments, Dr. David Hamilton and Mr. Wade Blanchard (Dalhousie University, Halifax, Nova Scotia) who provided statistical consultation, Dr. Millan Patel for his review and many helpful suggestions, Ms. A. Pavlova and Dr. J. Evroski for technical support, and Ms. L. Sauvé for assistance in preparation of this manuscript. G.A.H. is a Scholar of the Medical Research Council of Canada. This work has been supported by grants from the Physicians' Services Incorporated Foundation (Grant #95–22) and Women's College Hospital Research Foundation.