SEARCH

SEARCH BY CITATION

Keywords:

  • bone mass;
  • bone densitometry;
  • genetics;
  • family screening;
  • family studies

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

Low bone mass in adults is a major risk factor for low-impact fractures and is considered of complex origin because of interaction of environmental and genetic factors, each with modest effect. The objective was to assess the relative impact of genetics and environment and quantify the risk in relatives of osteopenic individuals. We studied 440 Icelandic nuclear families with 869 first-degree relatives of both sexes. Index cases (male or female) had BMD in the lumbar spine or hip >1.5 SD less than sex-matched controls. Heritability of BMD was estimated by maximum likelihood method, and variance component analysis was used to partition the genetic and environmental effects. Relative risk of low BMD (< −1 SD) in first-degree relatives was estimated, and heritable decrement in BMD was calculated compared with controls. Heritability was estimated as 0.61–0.66. Relative risk among first-degree relatives was 2.28, and the yield of screening was as high as 36%. The genetic influence was consistent with one or a few genes with considerable effect in addition to multiple genes each with a small effect. The genetic deficit in BMD was already present before 35 yr of age and equaled bone loss during 8–30 yr after menopause. We confirmed that genetics are more important than environment to low bone mass in adults. Our results are consistent with a few underlying genes with considerable effect. The prevalence among first-degree relatives of both sexes is common, suggesting that screening them should be cost effective and informative to elucidate the underlying genetics.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

Osteoporosis is an important public health problem worldwide, and early identification of women and men at risk for fracture is a major focus of research projects with the goal to invent preventive measures. Osteoporosis is characterized by low bone mass and alterations in bone microarchitecture leading to increased brittleness and susceptibility to fractures.[1] BMD, a surrogate for bone mass and bone strength, is a primary predictor of osteoporotic (low trauma) fractures.[2] Osteoporosis as defined by BMD[3] is one of the so-called common complex diseases that are thought to have both environmental and genetic factors in their risk. For decades, the studies of BMD have focused on nongenetic factors and showed that BMD is affected by an array of environmental and medical factors.[4] In recent years, data have been gathered that unequivocally show that variation in BMD is also under genetic control.[4-8]

Bone mass in adulthood is a culmination of the peak bone mass acquired during childhood and puberty and the rate of bone loss in adulthood. The familial resemblance in BMD shown in many studies is acquired quite early because the BMD of prepubertal offspring is correlated to that of their mother.[9, 10] Heritability estimates of BMD are higher in premenopausal than in postmenopausal daughters and seem to peak between the ages of 13 and 26 years and decrease with age, suggesting that the major genetic effect is on the attainment of peak bone mass rather than bone loss later in life.[10] Contrary to general conception, it seems that the effects of the environment shared by those living in the same household contribute little to similarity in BMD among family members.[6, 11, 12]

BMD in the general population is a complex trait, with a heritability of 0.6–0.8, assumed to reflect to large extent cumulative effects of many genes with small effect (polygenic effect).[8] Evidence for a major gene effect on BMD has been shown in some subgroups[13-16] but has been unconfirmed by others.[12]

Our cohort consisted of 440 nuclear families including 869 first-degree relatives, which is a formidable cohort in comparison with most other similar studies. Our objective was 4-fold: (1) to assess the relative impact of genetics and environment on low bone mass; (2) to estimate the relative risk of low bone mass to first-degree relatives; (3) to translate the genetic vulnerability into accumulated years of bone loss compared with individuals of same age; and (4) to test the hypothesis that only one or a few gene(s) play a significant role in low bone mass as determined by BMD in the hip and spine, the most important sites of low-trauma fractures.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

Study group

Four hundred forty nuclear families (parents and offspring), each with a member with low BMD, were used in this study. These families were created from 17,244 white individuals (2,189 men and 15,055 women) who were >18 yr of age and who had undergone bone densitometry at the University Hospital in Reykjavik in 1998–2006. By linking this database of BMD measurements to the extensive database on genealogy of all Icelanders, 31,075 nuclear families were found. Only those families that had two or more family members with BMD measurements were used for statistical analysis. Men and women below −1.5 SD (age, sex, and weight corrected) in lumbar spine or hip BMD were chosen as low BMD probands (altogether n = 440). This cut-off corresponds to the lowest eighth percentile of the age-matched population. The recruited relatives of these probands had participated in linkage studies by deCODE (n = 809).[17] In addition, 60 family members not recruited by us but who had undergone bone densitometry and therefore were listed in the database were included. For subsequent statistical analysis, we fractionated these three groups, probands, recruited relatives, and nonrecruited relatives, because it was essential for correcting for the recruitment bias created by our genetic studies.[18, 19]

Individuals who were <40 or >130 kg in weight were not included because of uncertainty of their BMD measurements, nor were those with missing age, height, or weight information. No further exclusions were made among the relatives; those with secondary osteoporosis or with diseases or taking medications that influence BMD were included, making this group fully representative of all of those who have had BMD measurements in Iceland.

There were 174 families that included two family members with BMD measurements, 131 families with three family members, 76 families with four members, 38 families with five members, and 21 families with six or more family members with BMD measurements, resulting in a total number of 440 nuclear families available for statistical analysis.

Informed consent was obtained from all the participants, and the study was approved by the National Bioethics Committee (NBC) in Iceland. All names listed in the bone densitometry registry and the genealogic database were encrypted through a process approved by the NBC and the data protection authority before being analyzed.

BMD measurements and standardization

BMD (g/cm2) at the lumbar spine (L1–L4) and hip (combined values at the femoral neck, trochanter, and intertrochanter region) was measured using a Hologic QDR4500A DXA scanner (Hologic, Waltham, MA, USA). The machine was calibrated daily. The CVs of the DXA measurement of the spine and hip were 1.0% and 1.8%, respectively. Weight (kg) and height (m) were measured using standard methods.

A random population-based cohort of both sexes, age 30–85 yr,[20, 21] was used as a reference population for BMD standardization purposes. Healthy individuals from this group (i.e., with no diseases or medications that might affect BMD such as hormone replacement therapy) served as a reference population (530 women and 493 men). The standardization was done based on the age and weight of the individual. Each sex was normalized separately. The normalized BMD measurements we used throughout the study have therefore been corrected for age, weight, and sex.

Statistical methods

The standardized BMD values for the population are assumed to obey a normal distribution. A variance component analysis was performed with three components: polygenic/environmental, allelic, and residual. The sizes of the variance components were estimated using a maximum likelihood approach considering only the likelihood of recruited individuals. The BMD measurements of these individuals can be considered to obey a normal distribution conditioned on the BMD measurements of the relatives (see Appendix for further details).

The relative risk for offsprings of individuals with low BMD (BMD < −1.5) was defined as the risk of having BMD < −1.5 in the offspring divided by the prevalence of having BMD < −1.5 in the general population of Iceland.[21] More precisely, if E denotes the event that a parent has BMD = −1.5 and R denotes the event that the offspring has BMD < −1, relative risk for that relation is defined as the ratio P(R|E)/P(R).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

BMD corrections

An age-, sex-, and weight-corrected Z-score was computed by performing regression of the effects of age and weight in the randomly recruited reference group. BMD was found to increase linearly with weight in both sexes, which explained 21.1% and 22.5% of hip BMD variance but 11.9% and 7.0% of lumbar spine BMD variance in females and males, respectively.

For women, a model where BMD decreases with age after the age of 45 was found to fit the observed data. For men, we found only a smaller decrease in hip BMD with age and only an insignificant trend for decrease in spine BMD with age.

Characteristics of study group

Table 1 shows the characteristics of the low BMD probands and their first-degree relatives. A few individuals (n = 36) are probands for two nuclear families. The age of the probands ranged from 20 to 85 yr (only 13 that were <30 yr of age) and the relatives from 19 to 86 yr (20 that were <30 yr of age). The probands were ∼1.6–2.2 SD, and the relatives were ∼0.5 SD below their age-matched mean in BMD. All BMD values were corrected for the confounding age and body weight as described.

Table Table 1.. Characteristics of the Study Group
Thumbnail image of

Figure 1 clearly shows the shift toward lower BMD in the relatives of the probands compared with the reference population.

thumbnail image

Figure FIG. 1.. The distribution of hip BMD among the relatives in the families compared with the normal reference population (age, sex, and weight matched). −o−, normal reference population; -------, relatives of probands.

Download figure to PowerPoint

Heritability, risk ratio estimates, and quantification of bone deficit caused by the genetic factor

To study the familial clustering of low BMD, we first estimated the heritability (h2) of low BMD for the hip and spine in these nuclear families. The heritability defined as the proportion of phenotypic variation that is attributed to genetic components for the whole study group was 0.61 for lumbar spine BMD and 0.66 for total hip BMD. The correlation between hip and spine BMD was found to be 0.79.

We estimated the relative risk of low BMD in the hip in these nuclear families compared with the population-based control group,[21] shown in Table 2. The data indicate that screening among first-degree relatives will yield as high as 36% of them having hip BMD below −1 SD and 17% < −1.5 SD, after being age, sex, and weight corrected depending on the probands' BMD. The results were almost identical if proband's lumbar spine BMD was low, with a relative risk as high as 2.16 and absolute risk as high as 35.1% for a first-degree relative having lumbar spine BMD < −1 SD if the proband's BMD was ≤ −2.0 SD. Somewhat lower risk was observed for relatives of probands with BMD in the range −1.5 to −2 SD, but relative risk was still ∼1.8.

Table Table 2.. Relative (RR) and Absolute (%) Risk of Low BMD in the Hip Among First-Degree Relatives According to the Hip BMD of the Proband
Thumbnail image of

We compared the BMD in the group of relatives to the reference group at all ages (cross-sectional in both groups) to assess if there was any difference between the two groups according to age. We observed the same deviation from the reference group in the relatives at all age groups (i.e., this difference in BMD in families versus controls was also present in younger age groups <35 yr of age as in the older group, and we found no age-related difference in the decrement between the two groups).

In Table 3, we calculated how the observed genetic deficit in peak bone mass can be translated into number of years of bone loss in women after the age of 45 yr. If a parents' hip BMD is < −2 SD of the mean, their daughters can expect to experience the equivalent of 15 yr of bone loss, on average, above what other women have at the same age. Therefore, at the age of 50 yr, these daughters will have a BMD equivalent to that of 65-yr-old women. Furthermore, if both their parents have a hip BMD < −2 SD, they will have a BMD equivalent to that of 80-yr-old women when they are only 50 yr of age themselves.

Table Table 3.. Comparison of the Observed Genetic Deficit in Peak Bone Mass in the Families With the Expected Equivalent of Age-Related Bone Loss After Age 45 in Years
Thumbnail image of

In the more relaxed criteria of a parent's hip BMD < −1.5 SD, the years of accumulated bone loss in their daughters was equivalent to 11.6 yr and 23.2 yr if both parents had hip BMD < −1.5 SD. Similar results were obtained for spine BMD.

Genetic factor partitioning

We attempted to partition the genetic factors underlying this low BMD more specifically by testing two genetic models: a polygenic model allowing for many genes (or common environmental factors shared by family members) versus a monogenic model, with only one gene with a larger effect (allelic component). Table 4 shows the partition of the variation in BMD by the variance component analyses on pedigree data by these two different statistical models: the polygenic (H0) versus the monogenic one (H1). We found that the allelic model is significantly more likely to fit the observed data than the nonallelic model (p < 0.0001 for the hip and 0.004 for the spine). The allelic additive component with a contribution from a single gene (few genes) explained ∼47% in the total hip BMD and 27% in lumbar spine BMD variability.

Table Table 4.. Percentage of Variance of BMD Estimated by a Mixed Model Including an Allelic Component (H1) Compared With a Model Assuming No Allelic Component (H0)
Thumbnail image of

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

We evaluated the contribution of genetic and environmental factors to low BMD in the hip and lumbar spine by using 440 nuclear families with 321 male and 548 female relatives >18 yr of age. Heritability estimates were high in our study sample (0.61–0.66) after adjusting for age and weight, as others have found previously.[6] Weight invariably affected BMD, more in the hip than in spine in both sexes. The high heritability unequivocally showed that the major determining factors of BMD variation are genetic in nature, accounting for >60% of the variation in BMD, both in the general population and in the BMD range characteristic of osteoporosis. We did not exclude individuals with diseases or taking medications (such as glucocorticoids) known to affect BMD from our study group, which could possibly explain somewhat lower h2 than seen in some other studies using various approaches.[22-24]

The high heritability of BMD would certainly provide a reason to look for low BMD in close relatives for early risk assessment of osteoporosis and subsequent prevention. Little attention has been paid to the possible yield of such screening among relatives, and family history of low bone mass is often low on the list of likely risk factors, although the importance of positive family history of fracture, especially maternal history of hip fracture, has been recognized.[5] Our results clearly showed the usefulness of such screening within families. This has also been highlighted by Duncan et al.[25] From our results, we estimated that if the proband has a BMD that is 1–2 SD below the age-matched mean, the absolute yield of finding low bone mass at the same skeletal site in first-degree relatives will be in the range of 28–36%. Our results also show that there is a clinically significant correlation of BMD between any first-degree relative pairs, not only mother-daughter pairs, stressing the importance that they too should be screened for osteoporosis. It is worth remembering in this respect that for every 1 SD decrease below the age-adjusted mean for hip BMD, the risk of hip fracture increases by a factor that is greater than two,[2] but the absolute risk increment is of course very dependent on the age of the individual concerned and other factors.

Importantly, to put the relative risk assessment into perspective, our results showed that daughters of parents who have low BMD may have a bone mass deficit equivalent to a cumulative bone loss of 8–30 yr, implying that women will be that many years older in terms of their BMD than the average woman of their age. Furthermore, in terms of the WHO criteria of osteoporosis,[3] daughters of < −2 SD parents can expect to enter the osteoporosis range of BMD (< −2.5 T-score) in the hip by 15 yr and in the lumbar spine by 10.6 yr before most other women. These results therefore indicate the necessity of BMD measurements in individuals of both sexes in these families at a younger age than is generally recommended today (women > 65 yr),[26] especially in women around menopause before they start losing bone. Further support for the early risk assessment is our observation that the deficit in BMD caused by the genetic effect is already present before 35 yr of age. A similar decrement in BMD with age was observed in our families as in our reference group, indicating that the genetic effect is most likely affecting only acquisition of peak bone mass, although we realize that the cross-sectional design of our study does not allow us to directly assess the magnitude of heritability of bone loss. Brown et al.[11] have also provided data that genetic factors influencing variation in BMD are common to both pre- and postmenopausal women, and heritability seems to peak before age 30.[10] Yang et al.[27] have also shown that 67% of postmenopausal BMD variation is attributable to the premenopausal BMD (peak BMD), whereas only 29% is attributable to bone loss rate after menopause. There is very little evidence from other studies showing that bone loss is controlled by genetic factors,[11, 27] with some exceptions.[28] Under certain conditions (diseases or medications), bone loss can be exaggerated, but this would be most deleterious when the peak bone mass is also low.

Our study group included all age groups >18 yr of age, and these genetic effects are therefore most likely common to both sexes of all adult ages. Some studies have, however, indicated a sex- and age-specific effect,[25, 29, 30] whereas others have not.[11, 22, 31-33]

For further understanding of the genetic effect on BMD variation, we performed a variance component analysis to partition variation in BMD into genetic and environmental effects. We attempted to separate the genetic component into an allelic component assuming that variation can be explained by a single locus dependent on number of alleles shared by two individuals or caused by a common familial environment. Several studies have, however, shown that the impact of shared environmental effects is negligible to cause covariation between family members.[6, 11, 12] This is, however, not saying that environmental factors under other stronger conditions are not of influence but indicates that common environmental factors experienced by family members living in the same household may be only a small subset of all environmental factors influencing BMD.[6] In accordance with that, the environmental factor, with no correlation between any pair of individuals within the family, accounts for 34–41% of the BMD variation. For the genetic factor, our analysis is more consistent with a single gene with a large effect rather than the assumption of many genes each with a small overall effect. The genetic component of the BMD variation (66% of total hip BMD variation) can be explained by a single gene responsible for two thirds of the effect in the hip (47% of total variation) and one third caused by many genes with a small effect (18.1% of the total variation). Similarly, for the spine, 26.9% were caused by one or few genes versus 34.7% caused by multiple genes. Our model is, however, likely to be a simplification, and a large allelic component may also be observed when the true model is oligogenic; we therefore conclude that the observed effect is either caused by a single gene or a few genes each with considerable effect.

Liu et al.[22] found a significant major gene effect only in the hip but not the spine. Our results showing higher h2 and allelic effect in hip than in lumbar spine might reflect age-related artefacts in measurements of lumbar spine BMD (e.g., osteoarthritis, aortic calcification), which makes the phenotypic variation more difficult to assess at that site, especially in old age. Support for such a major gene effect has also been suggested by some studies, especially in young individuals with idiopathic osteoporosis,[13] especially in males,[16] but unconfirmed in others.[12] Livshits et al.[14] have advocated and provided evidence for a pleiotropic genetic effect in the hip and spine, and Deng et al.[15] showed evidence for a major gene responsible for 16% of variation in spine, hip, and wrist BMD jointly. These studies have mostly been based on index cases with extremely low BMD, whereas our index group is made up of individuals with the lowest 8% of age- and sex-matched BMD, which we do not consider extreme values but reflecting the group where the public health problem of osteoporosis is derived from.

It should be recognized, however, that low-trauma fractures are not only related to low bone mass, which has been the topic of our study, but are also heavily dependent on external factors such as falls with completely different etiology and possibly genetics.

Bone densitometry with DXA as in our study is used in general practice for the diagnosis of osteoporosis and assessment of fracture risk. We confirmed that genetics are by far the most important factor predisposing to low bone mass. A screening of first-degree relatives of both sexes of individuals with low bone mass may therefore often be of help in the search for underlying etiological factors but is also of importance for the relatives themselves because we have shown that the yield of such screening is considerable. We provided support for the notion that a considerable part of the genetic effect may be caused by only one or a few genes with considerable effect and that these genes may differ between families and populations. Clearly the evidence for major genes in our study is indirect and, only the identification of such genes will prove their existence. Our results should, however, encourage the search for such genes and support the notion of the importance of major genes in the etiology of complex diseases.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

The authors thank all the participants in the study, Diana Oskarsdottir, radiographer, for supervising all the DXA measurements, and Maria Henley for preparation of the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

APPENDIX: STATISTICAL METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX: STATISTICAL METHODS

The standardized BMD values for a randomly recruited individual are assumed to obey a standard normal distribution, and the likelihood that this individual has a standardized BMD value x can be computed as

  • equation image

If there is a genetic effect to BMD values, the measurements of two related individuals will necessarily be correlated. The likelihood of observing BMD measurements x1, x2,…, xn for related individuals p1, p2,…, pn randomly recruited from the population is:

  • equation image

where Σ is the covariance matrix for the BMD values of p1, p2,…, pn.

We assume an additive model for inheritance and use a variance components method. There are three different components for the variance:

  • 1.
    An allelic component, Σa. Here we assume that the variation in BMD can be explained by a single locus. The correlation between two individuals in this factor is dependent on the number of alleles shared between two individuals (i.e., if two individuals share zero alleles, their correlation will be zero; if they share one allele their correlation will be ½; and if they share two alleles, their correlation will be one). This component is computed over all possible ways that the members of the pedigree can share alleles. Parent offspring pairs will necessarily share one allele, but sibling pairs may share zero or one or two alleles.
  • 2.
    A polygenic/environmental component, Σg. The correlation in this factor between two individuals is the expected sharing between two individuals: ½ for sibpairs and parent offspring pairs.
  • 3.
    A residual component, Σr, where the correlation is zero between any pair of individuals.

Given the sharing of the alleles between the individuals in the pedigree, the correlation matrix can be constructed as Σ = saΣa + SgΣg + srΣr. Because the population variance is known to be 1, we can constrain sa + sg + se = 1, and the heritability (h2) of the trait can be estimated as the sa + sg. The likelihood of the observed values can be computed as before.

Because the BMD measurements in the population are assumed to obey a normal distribution, the BMD measurements of the relatives can be assumed to follow a normal distribution with mean and SD that are computed conditionally of the values of the index cases. Because our database of BMD measurements is quite extensive, we also at times have by chance the BMD measurements of a number of relatives of low BMD individuals (n = 60). We treat the BMD measurements of these individuals as known values, and we only treat the BMD measurements of the recruited individuals as random variables.

If we order the individuals in a pedigree such that the recruited individuals precede the nonrecruited individuals, the covariance matrix of the BMD measurements in a pedigree can be written as

  • equation image

The covariance matrix of the recruited individuals given the BMD values of the nonrecruited can be computed as equation image = Σ11 – Σ12Σmath imageΣ21. If we let x1 be the BMD values of the nonindex cases and x2 be the BMD of the index cases, the distribution of x1 conditional on x2 = a is multivariate normal with mean equation image = Σ12Σmath imagea.

The relative sizes of the variance components were estimated using a maximum likelihood approach.