Osteoporosis International

, Volume 18, Issue 9, pp 1157–1175

Searching for genes underlying susceptibility to osteoporotic fracture: current progress and future prospect

Authors

  • S.-F. Lei
    • Laboratory of Molecular and Statistical Genetics and the Key Laboratory of Protein Chemistry and Developmental Biology of Ministry of Education, College of Life SciencesHunan Normal University
  • H. Jiang
    • Laboratory of Molecular and Statistical Genetics and the Key Laboratory of Protein Chemistry and Developmental Biology of Ministry of Education, College of Life SciencesHunan Normal University
  • F.-Y. Deng
    • Laboratory of Molecular and Statistical Genetics and the Key Laboratory of Protein Chemistry and Developmental Biology of Ministry of Education, College of Life SciencesHunan Normal University
    • Laboratory of Molecular and Statistical Genetics and the Key Laboratory of Protein Chemistry and Developmental Biology of Ministry of Education, College of Life SciencesHunan Normal University
    • Department of Orthopedic Surgery, School of MedicineUniversity of Missouri-Kansas City
Review

DOI: 10.1007/s00198-007-0402-4

Cite this article as:
Lei, S., Jiang, H., Deng, F. et al. Osteoporos Int (2007) 18: 1157. doi:10.1007/s00198-007-0402-4

Abstract

Introduction

Osteoporotic fracture (OF) is a public health problem. It is a common practice in the genetics of osteoporosis that bone mineral density (BMD) was studied as a major surrogate phenotype in gene search for risk of OF (ROF) because of their high phenotypic correlation. However, some studies indicate that the genetic correlation between BMD and ROF is very low. This implies that most genes found important for BMD may not be relevant to ROF. Ideally, employing OF per se as a direct study phenotype can directly find the relevant genes underlying ROF.

Evidence

Here, we summarized some evidence supporting ROF under moderate genetic control, and the current progress of molecular genetic studies employing OF as the direct study phenotype, then give our consideration on the future prospects in the genetics of ROF.

Keywords

Association studyCandidate genesGenetic determinationOsteoporotic fractureRisk of osteoporotic fracture

Introduction

Osteoporosis is a systemic skeletal disease characterized by low bone mineral density (BMD) and micro-architectural deterioration of bone tissue, with a consequent increase in susceptibility to fracture. Osteoporosis has become a public health problem. Between 30% and 50% of women and 13% and 30% of men will suffer from a fracture related to osteoporosis in their lifetime worldwide [1, 2]. The number of hip fractures estimated in 1990 was 1.66 million, and it is predicted to rise to 4.5–6.26 million by the year 2050 [3, 4].

Osteoporotic fracture (OF) is the most serious clinic consequence of osteoporosis. The risk of osteoporotic fracture (ROF) is a complex disease regulated by both environmental and genetic factors. Low BMD is an important risk factor for fracture [5]. To date, extensive molecular genetic studies have used BMD as a major surrogate phenotype to search for genes underlying susceptibility to OF. However, the genetic correlation between BMD and ROF is very low [6, 7]. Deng et al. [6] reported that the genetic correlation between BMD and ROF was non-significant and less than 1% of additive genetic variance was shared by them. Andrew et al. [7] indicated that risk of wrist fracture in women was heritable and was influenced by genes independent of those influencing BMD. Many studies focusing on the relationship of candidate genes and ROF also support the above findings. They found that candidate genes were associated with ROF independent of BMD [810]. These observations imply that despite of the importance of genetic factors for the etiology of ROF and BMD, on the whole, most genes found important for BMD may not be relevant to ROF. These also suggest that the documented association between ROF and BMD may be more likely to be explained by “environmental” factors, particularly those unique to the individual, such as the combined effects of a random low fall and the reduced mechanical strength associated with low bone mass [7].

Currently, extensive molecular studies using BMD as a major proxy risk factor for OF may expect that BMD and ROF share the same genetic risk factors. However, the low genetic correlation between them suggests that investigators should pursue different research strategies for finding genes that contribute to BMD and ROF. Ideally, employing OF per se as a direct phenotype can directly find the relevant genes underlying ROF. Here, we summarize some evidence supporting ROF under strong genetic control, and the current progress of molecular genetic studies employing OF as the direct study phenotype, then give our consideration on the future prospects in the genetics of ROF.

Evidence for genetic contribution to ROF

Searching for genes underlying ROF should begin with reliable evidence for a significant heritable component. Indirect evidence supports the susceptibility to OF being regulated by genetic factors. Genetic epidemiological studies show that a family history of fracture is a significant risk factor for OF [1115]. Cummings et al. [5] found a maternal history of hip fracture doubled the risk of hip fracture (RO = 2.0), and the increase in risk of fracture remained significant after adjustment for BMD. Keen et al. [11] reported that a history of wrist fracture in a female first-degree relative was associated with an increased risk of prevalent fracture at both appendicular and vertebral sites in a cross-sectional study design. The risk of hip fracture increases among daughters whose mothers have a prior history of fragility fractures after the age of 50 years [1618]. A meta-analysis, which included 34,928 men and women from seven prospectively studied cohorts followed for 134,374 person-years, found that a family history of hip fracture in parents was associated with a significant risk both of all OFs (RR = 1.54) and of hip fracture (RR = 2.27) and concluded that a parental history of fracture (particularly hip fracture) conferred an increased risk of fracture independent of BMD [19].

Recently, several studies provided formal and direct evidence supporting a significant and moderate genetic determination of ROF [6, 7, 2022]. Deng et al. [20] estimated that the narrow-sense heritability of Colles’ fracture was approximately 0.25 in a cohort of American Caucasian women, thus accounting for approximately one-quarter of the variation in total Colles’ fracture risk observed. A study in 6,570 white healthy female volunteer twins between 18 and 80 years of age identified and validated 220 nontraumatic wrist fracture cases, and found the additive polygenic heritability of ROF was ∼54% [21]. Another recent study reported after age adjustment less than 20% of the overall fracture variance could be explained by genetic variation [22]. In 50 extended Caucasian pedigrees, we estimated that risk of hip fracture had a heritability of ∼0.53 [6]. The estimations of heritability for ROF heavily depend upon the used models (additive or dominant) and studied samples. Generally, twin studies, as the optimum method of determining heritability, show higher heritabilities than the family studies for the same type OF, as the family studies may be under-estimating effects due to confounding by common environment effects. Assuming a polygenic liability, one finds the involvement of oligogenic genes for ROF with moderate heritability may be less than for complex traits with high heritabilities. This may increase our difficulty in finding genes underlying susceptibility to OF. This is because, compared to complex diseases with high heritability, those with moderate heritability of liability are less likely to include genes with large effect and therefore are more likely to undetected. On the other hand, moderate heritability for ROF means that the remaining larger percentage of phenotypic variation for ROF is explained by environmental factors. Due to the confounding effects of environmental factors, they should be carefully controlled when finding genes for ROF. In a word, accumulative evidence unambiguously demonstrated that the ROF is regulated by moderate genetic factors.

Current progress on molecular genetics for ROF

To date, to identify genes underlying ROF, many studies have focused their efforts using OF as direct studied phenotype. Candidate gene association study is the commonly used method to identify genes underlying susceptibility to OF.
  1. a.

    Vitamin D receptor

    Vitamin D receptor (VDR) when bound to vitamin D plays an important role in calcium homeostasis by regulating bone cell growth and differentiation, intestinal calcium absorption and parathyroid hormone secretion. The polymorphisms of VDR gene in association studies are focused on BsmI, ApaI, TaqI and FokI.

    Many studies find that some polymorphisms are associated with the risk of hip, vertebral, or other fractures (Table 1) [2326]. Some negative associations between the VDR polymorphisms and ROFs are also found (Table 1). The associations between the haplotypes from the polymorphisms in the VDR gene and ROF are reported, but the results are inconsistent (Table 1) [2729]. For example, Haplotype “baT” is associated with the risk of vertebral fracture, but not with the risk of nonvertebral fracture [27].

    Recently, a meta-analysis in 6,067 fracture patients and 20,175 controls detected that the FokI, BsmI, ApaI and TaqI polymorphisms were not significantly associated with the risk of both vertebral fracture and nonvertebral fracture, but the Cdx2 A-allele carriers had 9% (p = 0.039) risk reduction [30]. Another meta-analysis also supports the negative relationship between the BsmI and TaqI polymorphisms and ROF [31].

     
  2. b.

    Estrogen receptor

    Estrogens, through binding to two different estrogen receptors (ERs), ER-α and ER-β, are known to play an important role in regulating bone homeostasis and preventing postmenopausal bone loss. Numerous studies show that ER-α polymorphisms may influence the incidence of OFs. We briefly summarized some of the published associations of ER gene in Table 2.

    Some positive associations are found between the polymorphisms in the ER gene and ROF [3234]. Negative associations between the ER polymorphisms and ROFs are also found (Table 2).The mean number of TA repeats is lower in patients with vertebral fracture than in normal controls (p < 0.01) and a significantly increased OR of OFs is also found in individuals with 11–18 repeats [32]. Two polymorphisms PvuII and XbaI in the intron 1 are widely studied (Table 2). Allcroft et al. [34] found the distribution of the alleles of the polymorphism PvuII was similar in the patients with symptomatic vertebral crush fractures and male control subjects. A recent mete-analysis indicates that XX homozygotes have decreased risk of fractures when compared with carriers of the x allele (OR = 0.66), whereas the PvuII polymorphism is not associated with fracture risk [35]. In another study on vertebral fracture cases during a mean follow-up period of 7 years, researchers find the haplotype px of ER-α is dose-dependent, associated with increased vertebral fracture risk [28].

     
  3. c.

    Type I collagen

    Type I collagen α1 (COLIA1) is the most abundant protein in bone, and mutations in the genes encoding collagen type I α1 and collagen type I α2 are estimated to be responsible for up to 90% of cases of the Mendelian disease osteogenesis imperfecta [36]. A G to T polymorphism binding Sp1 site in COLIA1 gene receives most of attention of researchers on detecting the relationship of COLIA1 and ROFs. Many studies on the relationship of COL1A1 gene polymorphisms and ROFs are summarized in Table 3.

    Extensive molecular studies support that Sp1 polymorphism is associated with ROF [29, 3740]. A study in postmenopausal women on the same polymorphism showed that women with the TT genotype had a 5.9-fold increased risk when compared with the other genotypes (GG + GT) [38]. Other polymorphisms in the COL1A1 gene such as MspI, RsaI, MnlI are also studied with vertebral fracture, but only the Sp1 polymorphism acted as an independent predictor of fracture (OR = 2.26) [41]. Importantly, a meta-analysis including 13 eligible studies with 3,641 subjects found that the COLIA1 Sp1 polymorphism showed a dose-response relationship with the prevalence of fractures and the risk of non-vertebral fracture was not increased in relation to the COL1A1 genotype [42], which was similar with the results from another meta-analysis [43].

     
  4. d.

    Other candidate genes

    There are many other studies focusing on the relationship of ROFs and other candidate genes, such as transforming growth factor- β (TGF-β) [44, 45], methylenetetrahydrofolate reductase (MTHFR) [46], LRP5 [47], runt-related gene 2 (RUNX2)/core binding factor A1 (CBFA1) [48], interleukin-1 receptor antagonist (IL-1ra) [49], apolipoprotein E [50, 51], osteoprotegerin (OPG) [52], and so on (Table 4). Compared with the VDR, ER and COL1A1 genes, only a few studies focus on the above genes, so we may not make any conclusion on each association.

     
Table 1

Association studies between VDR polymorphisms and ROFs

Marker

Subject

Sex

Phenotype

 

Case/Control (N)

Effect originally reported

Allele [Frequency#]

Allele p-value (Odd Ratio)

Genotype [Frequency#]

Genotype p-value

Ref.

BsmI

  

Hip fracture

54

108

BB > bb (twofold increased risk of hip fracture).

NA

NA

NA

 

[26]

   

Forearm fracture

163

108

Little association of forearm fracture with VDR gene.

NA

NA

NA

  

TaqI

Caucasian (≥ 65 yrs)

F

Hip fracture

181

692

NS

T [60.5;56.6]

0.317 (1.186)

NA

 

[71]

   

Vertebral fracture

127

482

NS

T [59.1;56.7]

0.624 (1.104)

NA

  
   

Other fractures

223

368

NS

T [57.6;56.3]

0.785 (1.048)

NA

  

ApaI

  

Hip fracture

181

692

NS

A [51.9;55.6]

0.373 (0.862)

NA

  
   

Vertebral fracture

127

482

NS

A [53.1;57.0]

0.385 (0.841)

NA

  
   

Other fractures

223

368

NS

A [54.0;58.6]

0.245 (0.820)

NA

  

FokI

Caucasian (45–72 yrs)

F

Vertebral fracture

68

332

ff in vertebral fracture women > ff in controls (p = 0.003, equivalent to a relative risk of 2.58 )

F [52.9;65.1]

0.059 (0.604)

Ff [25.0;11.4] Ff [44.1;47.0] FF [30.9;41.6]

0.010

[72]

BsmI

Caucasian

F & M

Vertebral fracture

110

153

BB, Bb are more frequent in cases (p = 0. 06)

B [52.8;46.4]

0.312 (1.288)

bb [22.7;33.3] Bb [49.1;40.5] BB [28.2;26.1]

0.161

[73]

FokI

     

NS

F [61.5;63.1]

0.794 (0.935)

ff [13.8;15.7] Ff [49.5;42.5] FF [36.7;41.8]

0.528

 

ApaI

     

NS

A [56.5;54.5]

0.734 (1.089)

aa [15.0;20.4] Aa [57.0;50.3] AA [28.0;29.3]

0.493

 

TaqI

     

NS

T [57.1;61.5]

0.497 (0.841)

tt [15.0;15.5] Tt [56.1;45.9] TT [29.0;38.5]

0.214

 

BsmI

Caucasian (≥ 60 yrs)

F

Hip fracture

135

239

NS

B [57.1;61.5]

0.246 (0.777)

bb [36.3;25.9] Bb [44.4;52.3] BB [19.3;21.8]

0.108

[74]

BsmI ApaI TaqI

Caucasian (≥ 55 yrs)

F

Osteoporotic fracture

97

907

The “baT” VDR haplotype overrepresented among cases (p = 0.009 )

baT [59.8;48.0]

0.027 (1.614)

(baT)0 [15.5;28.4] (baT)1 [49.5;47.2] (baT)2 [35.1;24.4]

0.009

[27]

BsmI

Caucasian (> 85 yrs)

F

Hip fracture

64

108

NS

B [41.4;34.3]

0.402 (1.313)

Bb [31.3;40.7] Bb [54.7;50.0] BB [14.1;9.3]

0.371

[75]

BsmI ApaI TaqI

Caucasian (≥ 55 yrs)

F

Vertebral fracture

85

549

Haplotype “baT” is overrepresented in cases

BaT [54.7;47.6]

0.259 (1.302)

(baT)0 [22.4;27.9] (baT)1 [45.9;49.0] (baT)2 [31.8;23.1]

0.199

[28]

   

Nonvertebral fracture

131

931

NS

baT [52.7;48.9] BAt [36.6;40.8] bAT [10.7;10.3]

0.479 (1.180) 0.449 (0.833) 0.954 (1.022)

bbaaTT [26.7;25.3] BbAaTt [42.0;37.2] bbAaTT [9.9;9.9] BBAAtt [12.2;17.9] BbAATt [6.9;8.6] bbAATT [2.3;1.1]

0.455

 

Cdx2

Different ethnicities(55–80 yrs)

F & M

Vertebral fracture

217

1698

NS

A [15.2;19.3]

0.145 (0.749)

GG [71.9;65.6] GA [25.8;30.1] AA [2.3;4.3]

0.121

[24]

   

Nonvertebral fracture

248

2600

Negative correlation between the A-allele and hip fracture incidence rates (p = 0.006 for men and p = 0.02 for women)

A [16.1;19.0]

0.268 (0.820)

GG [69.8;66.2] GA [28.2;29.5] AA [2.0;4.3]

0.178

 

BsmI

Caucasian (31–89 yrs)

F

Vertebral fracturea

34

469

Allele B is significantly and dose dependently overrepresented in women with fracture (p = 0.04)

B [38.2;37.7]

0.954 (1.021)

bb [38.2;38.4] Bb [47.1;47.8] BB [14.7;13.9]

0.990

[8]

   

Nonvertebral fracture

86

469

Allele B is significantly and dose dependently overrepresented in women with fracture (p = 0.01)

B [50.0;37.7]

0.033 (1.650)

bb [23.3;38.4] Bb [53.5;47.8] BB [23.3;13.9]

0.010

 

TaqI

Caucasian

F

Hip fracture

69

608

VDR CC genotype had an increased risk of hip fracture (OR = 2.6)

C [45.7;39.3]

0.256 (1.335)

TT [34.8;35.9] TC [39.1;49.7] CC [26.1;14.4]

0.034

[9]

FokI, BsmI ApaI, TaqI

 

F & M

Vertebral fracture

2088

20175

NS NS NS NS

NA

NA

NA

NA

[30]*

Cdx2

     

9% (p = 0.039) risk reduction for the Cdx2 A-allele

NA

NA

NA

NA

 

BsmI TaqI

Caucasian

F & M

Fracture

1632

5203

NS

NA

NA

NA

NA

[31]*

*Meta-analysis studies; NS: non-significant; OR: odds ratio; [Frequency#] frequency in case and control (%); NA: not available; Effect originally reported: the reported effects in original paper; Allele p-value: the p value of Chi-square test for the distribution of allele in case and control; Genotype p-value: the p value of Chi-square test for the distribution of genotype in case and control.

Table 2

Association studies between ER-α gene polymorphisms and ROFs

Marker

Subject

Sex

Phenotype

Case/Control (N)

 

Effect originally reported

Allele [Frequency#]

Allele p-value (Odd Ratio)

Genotype [Frequency#]

Genotype p-value

Ref.

(TA)n

Caucasian (50–73 yrs)

F

Vertebral fracture

73

537

Increased fracture risk in women with a low number of repeats (TA < 15), (OR = 2.9)

NA

NA

NA

NA

[33]

(TA)n

Caucasisian (33–79 yrs)

F & M

Vertebral fracture

190

197

The mean number of TA repeats is lower in patients with osteoporotic fractures, (p < 0.01)

TA# [17.3;18.6]

NA

NA

NA

[32]

BstUI

   

110

153

NS

B [91.4;87.9]

0.319 (1.530)

Bb [0.0;0.7] Bb [17.3;22.9] BB [82.7;76.5]

0.380

 

PvuII

     

NS

P [39.1;47.4]

0.186 (0.714)

Pp [36.4;30.7] Pp [49.1;43.8] PP [14.5;25.5]

0.104

 

XbaI

     

NS

X [28.6;33.0]

0.359 (0.793)

Xx [50.0;47.7] Xx [42.7;38.6] XX [7.3;13.7]

0.236

 

PvuII,

Caucasian (≥ 60 yrs)

F

Hip fracture

135

239

NS

P [44.4;45.8]

0.828 (0.954)

Pp [31.1;28.0] Pp [48.9;52.3] PP [20.0;19.7]

0.786

[74]

XbaI

      

X [34.1;36.6]

0.652 (0.903)

xx [42.2;40.2] Xx [47.4;46.4] XX [10.4;13.4]

0.689

 

PvuII

Caucasian (27–77 yrs)

M

Vertebral fracture

53

29

NS

P [48.1;39.7]

0.614 (1.265)

pp [28.3;34.5] Pp [47.2;51.7] PP [24.5;13.8]

0.507

[34]

PvuII XbaI

Japanese (45–75 yrs)

F

Vertebral fracture

48

149

Useful genetic markers for predicting vertebral fracture in relatively young postmenopausal women.

NA

NA

PPXX [4.1;2.0] PPXx [6.1;10.1] PPxx [8.2;4.1] PpXx [20.4;16.9] Ppxx [28.6;30.4] ppXx [0.0;0.7] ppxx [32.6;35.8]

0.781

[76]

(TA)n PvuII XbaI

Caucasian (≥ 55 yrs)

F & M

Vertebral fracture

152

1032

An increased vertebral fracture risk for an allele dose effect, (OR = 2.2 for haplotype ‘px’, OR = 2.0 for a low number of (TA)n repeats) (n < 18)

PX [65.8;53.1]

0.003 (1.698)

(PX)0 [14.5;22.1] (PX)1 [39.5;49.6] (PX)2 [36.1;28.3]

0.013

[10]

PvuII XbaI

Caucasian (≥ 55 yrs)

F

Vertebral fracture

85

549

Haplotype px of ER-α is dose-dependently associated with increased vertebral fracture risk, (p < 0.001, corresponding to an OR of 1.9)

px [68.2;51.5] PX [22.4;37.2] Px [9.4;11.3]

0.004 (2.019) 0.008 (0.487) 0.607 (0.816)

ppxx [47.1;26.8] PpXx [30.6;38.3] PPxx [11.8;11.1] PPXX [3.5;13.8] PPXx [7.1;8.6] PPxx [0.0;1.5]

0.002

[28]

   

Nonvertebral fracture

131

931

NS

px [52.7;53.1] PX [36.3;36.1] Px [11.1;10.8]

0.933 (0.984) 0.902 (1.024) 0.836 (1.063)

ppxx [24.4;28.5] PpXx [46.6;37.4] PPxx [9.9;11.9] PPXX [7.6;13.7] PPXx [10.7;7.3] PPxx [0.8;1.2]

0.119

 

XbaI

Caucasian

F & M

Osteoporotic fracture

484

1107

A protective effect for XX, (OR = 0.66)

X [34.0;40.1]

0.023 (0.772)

xx [42.8;37.0] Xx [46.5;45.7] XX [10.7;17.3]

0.002

[35]*

PvuII

   

655

1574

NS

P [42.2;45.7]

0.125 (0.866)

pp [34.5;29.7] Pp [46.6;49.4] PP [18.9;20.8]

0.081

 

*Meta-analysis studies; NS: non-significant; OR: odds ratio; [Frequency#] frequency in case and control (%); NA: not available; Effect originally reported: the reported effects in original paper; Allele p-value: the p value of Chi-square test for the distribution of allele in case and control; Genotype p-value: the p value of Chi-square test for the distribution of genotype in case and control.

Table 3

Association studies between COL1A1 gene polymorphisms and ROFs

Marker

Subject

Sex

Phenotype

Case/Control (N)

 

Effect originally reported

Allele [Frequency#]

Allele p-value (Odd Ratio)

Genotype [Frequency#]

Genotype p-value

Ref.

Sp1

Caucasian (≥ 55 yrs)

F

Nonvertebral fracture

111

1667

Overrepresented of the Ss and ss genotype in nonvertebral fracture women (per copy of the s allele, OR = 1.5)

S [75.2;82.4]

0.042 (0.632)

ss [7.2;3.0] Ss [35.1;29.2] SS [57.7;67.8]

0.014

[77]

   

Vertebral fracture

74

1339

NS

S [78.4;82.2]

0.401 (0.784)

ss [2.7;3.1] Ss [37.8;29.3] SS [59.5;67.6]

0.292

 

Sp1

Caucasian (22–80 yrs)

F & M

Vertebral fracture

105

144

Over-representation of the ss genotype in the fracture patients (OR = 11.83, with the ss genotype)

S [69.0;81.9]

0.014 (0.481)

ss [14.3;1.4] Ss [33.3;33.3] SS [52.4;65.3]

0.000

[39]

Sp1

Caucasian

F

Vertebral fracture

36

67

NS

S [81.9;80.6]

0.996 (0.997)

ss [0.0;4.5] Ss [36.1;29.9] SS [63.9;65.7]

0.385

[78]

Sp1

Caucasian (45–64 yrs)

F

Vertebral & appendicular fracture

55

130

“s” allele is associated with an increased risk of total fracture (p = 0.04)

S [75.5;80.8]

0.498 (0.769)

ss [0.0;3.8] Ss [49.1;30.8] SS [50.9;65.4]

0.030

[79]

Sp1

Caucasian (21–49 yrs)

F & M

Vertebral fracture

56

78

NS

S [79.5;84.6]

0.519 (0.744)

ss [3.6;5.2] Ss [33.9;20.5] SS [62.5;74.3]

0.214

[80]

Sp1

Caucasian (≥ 60 yrs)

F

Hip fracture

135

239

NS

S [81.9;78.5]

0.409 (1.255)

ss [5.2;6.3] Ss [25.9;30.5] SS [68.9;63.2]

0.537

[75]

Sp1

Caucasian (45–70 yrs)

F

Wrist fracture

126

126

The overall gene-dose effect is an OR of 2.1 per copy of the “s” allele

S [78.6;86.5]

0.097 (0.572)

Ss [5.6;1.6] Ss [31.7;23.8] SS [62.7;74.6]

0.064

[81]

Sp1

Caucasian

F & M

Vertebral fracture

93

88

Significant allele distribution between patients and controls ( p = 0.005)

S [75.8;89.2]

0.020 (0.390)

ss [6.5;1.1] Ss [35.5;19.3] SS [58.0;79.5]

0.005

[41]

MspI

     

NS

M [33.9;28.4]

0.385 (1.322)

mm [47.3;55.7] Mm [37.6;31.8] MM [15.1;12.5]

0.530

 

RsaI

     

NS

R [28.5;18.8]

0.128 (1.709)

rr [46.2;64.8] Rr [50.5;33.0] RR [3.2;2.3]

0.043

 

MnlI

     

NS

M [29.0;37.5]

0.226 (0.682)

mm [50.5;45.5] Mm [40.9;34.1] MM [8.6;20.5]

0.074

 

Sp1

Caucasian (≥ 75 yrs)

F

Spine fracture

69

198

NS

S

NA

Ss/Ss [18.8; NA] SS [81.2;NA]

NA

[82]

   

No-spine fracture

64

198

NS

S

NA

ss/Ss [26.6;NA] SS [73.4;NA]

NA

 

Sp1

Caucasian (≥ 55 yrs)

F

Osteoporotic fracture

97

907

Significant different genotype distribution between cases and controls (p = 0.004)

S [73.7;83.1]

0.015 (0.554)

ss [7.2;2.8] Ss [38.1;28.2] SS [54.6;69.0]

0.004

[27]

Sp1

Caucasian (> 85 yrs)

F

Hip fracture

64

108

NS

S [86.7;90.7]

0.331 (0.624)

ss [6.3;0.0] Ss [14.1;18.5] SS [79.7;81.5]

0.027

[75]

Sp1

Spanish

F

Osteoporotic fracture

82

139

The women with the ss genotype had a 5.9-fold increased risk when compared with the other genotypes (SS + Ss)

S [62.2;71.2]

0.165 (0.665)

ss [25.6;6.5] Ss [24.4;44.6] SS [50.0;48.9]

0.000

[38]

    

82

98

 

S [62.2;71.4]

0.189 (0.658)

ss [25.6;6.1] Ss [24.4;44.9] SS [50.0;49.0]

0.000

 

Sp1

Caucasian

F

Vertebral fracture

43

101

Significant overrepresentation of the ’s’n allele in fractured women (p = 0.029)

S [60.5;74.8]

0.074 (0.503)

ss [9.3;4.0] Ss [60.5;42.6] SS [30.2;53.4]

0.029

[37]

Sp1

Caucasian (≥ 75 yrs)

F

Wrist fracture

181

964

An increased presence of the “s” allele in wrist fracture patients (OR = 2.73)

S [79.8;83.7]

0.171 (0.757)

ss [5.0;2.7] Ss [30.4;27.3] SS [64.6;70.0]

0.154

[40]

   

Hip fracture

48

964

NS

S [79.2;83.7]

0.407 (0.739)

ss [4.2;2.7] Ss [33.3;27.3] SS [62.5;70.0]

0.513

 

Sp1

Caucasian

F

Hip fracture

69

608

COLIA1 TT genotype had an increased risk of hip fracture (OR = 3.8)

S [72.5;80.3]

0.129 (0.647)

ss [10.1;4.4] Ss [34.8;30.6] SS [55.1;65.0]

0.070

[9]

Sp1

  

Fracture

1325

5635

ss vs Ss, OR = 1.37, p = 0.0004; ss vs SS, OR = 2.48, p < 0.00001 for vertebral fracture

NA

NA

NA

NA

[43]*

   

Vertebral fracture

899

3757

      
   

Nonvertebral fracture

399

1812

      

Sp1

Caucasian

F & M

Fracture

950

2701

Ss vs. SS, OR = 1.25; ss vs. SS, OR = 1.68; ss vs. Ss, OR = 1.35.

S [78.2;82.5]

0.004 (0.762)

ss [5.2;3.3] Ss [33.3;28.5] SS [61.6;68.2]

0.000

[42]*

*Meta-analysis studies; NS: non-significant; OR: odds ratio; [Frequency#] frequency in case and control (%); NA: not available; Effect originally reported: the reported effects in original paper; Allele p-value: the p value of Chi-square test for the distribution of allele in case and control; Genotype p-value: the p value of Chi-square test for the distribution of genotype in case and control.

Table 4

Association studies between other gene polymorphisms and ROFs

Marker

Subject

Sex

Phenotype

Case/ Control (N)

 

Effect originally reported

Allele [Frequency#]

Allele p-value (Odd Ratio)

Genotype [Frequency#]

Genotype p-value

Ref.

TGF-β (713-8delC C788T)

Caucasian (21–80 yrs)

F

Spinal fracture

161

131

The prevalence of 713-8delC is significantly higher in the osteoporotic group (p < 0.05)

NA

NA

NA

NA

[83]

TGF-β (T869C)

Caucasian (≥ 70 yrs)

F

Fracture

1337

 

The TGF-β C allele is associated with an increase prevalent fracture (OR = 1.37)

C [44.9;51.1]

NA

NA

NA

[45]

MTHFR (C677T)

Caucasian

F & M

Vertebral Fracture

388

336

TT genotype is significantly more common in women with vertebral fractures, (p < 0.05)

T [33.4;19.1]

0.000 (2.141)

CC [45.1;51.0] CT [43.1;39.8] TT [11.9;9.2]

0.239

[46]

MTHFR (C677T)

Caucasian

F

Forearm Fracture

74

207

Individuals homozygotic for the C-allele have high risk comparing to those homozygotic for the T-allele (OR  = 3.93) for forearm fracture

T [22.3;32.4]

0.130 (0.623)

CC [56.8;44.4] CT [41.9;46.4] TT [1.4;9.2]

0.033

[84]

   

Hip fracture

41

51

Individuals homozygotic for the C-allele have high risk comparing to those homozygotic for the T-allele (OR = 6.99) for hip fracture

T [20.7;38.2]

0.063 (0.408)

CC [61.0;35.3] CT [36.6;52.9] TT [2.4;11.8]

0.028

 

LRP5 (3357 A > G)

Caucasian (70–85 yrs)

F

Fracture

5–year follow-up period, 227 subjects

 

The incident fracture rate is significantly increased in subjects homozygous for the GG polymorphism (RR = 1.61, p = 0.027).

NA

NA

GG [24.8;16.8]

NA

[47]a

LRP5 (Ala1330Val)

Caucasian (≥ 55 yrs)

F & M

Prevalent vertebral fractures

354

2835

NS

Val

NA

AlaAla [71.8;73.1] (AlaVal/ValVal) [28.2;26.9]

0.594

[87]

   

Incident osteoporotic fractures

950

5423

NS

Val

NA

AlaAla [72.1;73.3] (AlaVal/ValVal) [27.9;26.7]

0.444

 
   

Incident fragility fractures

409

5964

Male carriers of the LRP5 1330-valine variant had a 60% increased risk for fragility fractures.

Val

NA

AlaAla [69.2;73.4] (AlaVal/ValVal) [30.8;26.6]

0.063

 

LRP6 (I1062V)

Caucasian (≥ 55 yrs)

F & M

Prevalent vertebral fractures

357

2820

NS

Val

NA

IleIle [65.5;66.0] (IleVal/ValVal) [34.5;34.0]

0.867

 
   

Incident osteoporotic fractures

953

5462

NS

Val

NA

IleIle [63.6;64.9] (IleVal/ValVal) [36.4;35.1]

0.433

 
   

Incident fragility fractures

411

6004

Male carriers of the LRP6 1062-valine allele variant had a 60% increased risk for fragility fractures.

Val

NA

IleIle [62.3;64.9] (IleVal/ValVal) [27.7;35.1]

0.093

 

RUNX2/CBFA1 (A/G)

Caucasian (≥ 35 yrs)

F

Colles’ fracture

103

224

Allele counts in the Colles’ fracture group are significantly different from the control (p = 0.005).

G [88.3;85.3] A [4.4;10.9] 11Ala [7.3;3.8]

0.452 (1.310) 0.059 (0.425) 0.156 (2.012)

GG [76.7;72.8] GA [8.7;18.3] AA [0.0;1.3] G11Ala [14.7;6.7] A11Ala [0.0;0.9] Unique [0.0;0.0]

0.022

[48]

   

Hip fracture

67

224

NS

G [85.9;85.3] A [10.2;10.9] 11Ala [3.9;3.8]

0.791 (1.113) 0.951 (0.972) 0.549 (1.12)

GG [68.7;72.8] GA [19.4;18.3] AA [0.0;1.3] G11Ala [7.5;6.7] A11Ala [0.0;0.9] Unique [4.5;0.0]

0.039

 
   

Spine fracture

108

224

NS

G [85.5;85.3] A [9.3;10.9] 11Ala [5.1;3.8]

0.984 (0.993) 0.682 (0.850) 0.527 (1.405)

GG [70.4;72.8] GA [18.5;18.3]AA [0.0;1.3] G11Ala [10.2;6.7] A11Ala [0.0;0.9] Unique [0.9;0.0]

0.339

 

IL-1ra (MspA1I)

Caucasian (21–79 yrs)

F & M

Osteoporotic fracture

8

153

NS

M [68.8;72.9]

0.438 (0.808)

Mm [8.3;8.5] Mm [45.9;37.3] MM [45.9;54.2]

0.678

[49]

IL-1ra (86 bp VNTR)

Caucasian (21–79 yrs)

F & M

Osteoporotic fracture

102

146

The A1A1/A3 genotypes of the IL-1ra VNTR polymorphism are significantly more frequent in osteoporotic patients compared with age-matched normal controls (p = 0.043).

A1 A2 A3**

NA

A1A1/A3 [56.2;43.3]

NA

 

IL-1β (AvaI)

Caucasian (21-79 yrs)

F & M

Osteoporotic fracture

109

153

NS

A [33.5;33]

0.830 (1.059)

Aa [45.0;45.1] Aa [43.1;42.5] AA [11.9;12.4]

0.991

[49]

(BsoFI)

     

NS

B [65.6;63.7]

0.658 (1.123)

bb [11.9;12.4] Bb [45.0;47.7] BB [43.1;39.9]

0.869

 

(TaqI)

     

NS

T [26.1;24.8]

0.876 (1.046)

tt [56.0;57.5] Tt [35.8;35.3] TT [8.3;7.2]

0.939

 

IL-1α (rs17561)

Caucasian

F & M

Vertebral fracture

291

283

NS

G [60.7;68.1]

0.065 (0.724)

TT [9.0;11.5] GT [42.6;40.3] GG [48.4;48.2]

0.541

[85]

(rs207137)

     

NS

C [33.2;28.1]

0.189 (1.269)

TT [43.3;50.9] CT [47.1;42.0] CC [9.6;7.1]

0.158

 

(rs2856838)

     

The C allele of the rs2856838 polymorphism tended to be more common among patients with vertebral fractures (P = 0.06).

T [35.0;40.5]

0.168 (0.788)

CC [43.1;33.6] CT [43.8;51.9] TT [13.1;14.5]

0.061

 

(rs1800794)

     

NS

T [31.4;32.0]

0.820 (0.960)

CC [46.4;47.3] CT [44.3;41.3] TT [9.3;11.3]

0.637

 

APOE polymorphism

Caucasian

F

Forearm fracture

73

211

NS

E2 [13.7;19.4] E3 [95.9;95.3] E4 [31.5;28.4]

0.271 (0.658) 0.560 (1.161) 0.619 (1.158)

NA

NA

[50]

   

Hip fracture

43

53

NS

E2 [27.9;30.2] E3 [95.3;92.5] E4 [25.6;20.8]

0.807 (0.895 0.443 (1.673) 0.576 (1.313)

NA

NA

 

APOE polymorphism

Caucasian

F

Hip fracture

about 7-year follow-up period, 1,750 women

 

Women with at least one APOE*4 allele had an increased risk of hip fracture, relative hazard (RH)= 1.90, compared with women without APOE*4

APOE*2 [9.2;8.3] APOE*3 [75.8;83.4] APOE*4 [15.0;8.4]

NA

NA

NA

[51]a

   

Wrist fracture

about 7-year follow-up period, 1,750 women

 

Women with at least one APOE*4 allele had an increased risk of wrist fracture, RH = 1.67 compared with women without APOE*4

APOE*2 [7.7;8.4] APOE*3 [78.8;83.3] APOE*4 [13.5;8.3]

NA

NA

NA

 
   

Vertebral fracture

about 7-year follow-up period, 1,750 women

  

APOE*2 [9.7;8.8] APOE*3 [79.8;82.7] APOE*4 [10.5;8.5]

NA

NA

NA

 
   

Other fracture

about 7-year follow-up period, 1,750 women

  

APOE*2 [8.6;9.3] APOE*3 [84.1;82.9] APOE*4 [7.4;8.8]

NA

NA

NA

 

OPG (A163-G)

Caucasian

F & M

Vertebral fracture

268

289

Allele G of A163-G polymorphism is more common among fracture patients (OR = 1.44, p < 0.05)

G [19.7;14.6]

0.100 (1.450)

AA [65.8;73.5] AG [28.9;23.7] GG [5.3;2.8]

0.093

[52]

(T245-G)

     

Variant allele G of T245-G polymorphism is more common in osteoporotic patients (OR = 2.0, p < 0.02)

G [6.5;3.3]

0.113 (1.890)

TT [87.6;93.4] TG [11.6;6.6] GG [0.7;0.0]

0.038

 

(G1181-C)

     

The CC genotype of G1181-C polymorphism is less common among fracture patients (p < 0.01).

C [54.1;58.2]

0.339 (0.849)

GG [18.0;21.1] GC [55.6;42.9] CC [26.3;36.7]

0.006

 

(T950-C)

     

NS

C [49.6;51.6]

0.649 (0.926)

TT [23.7;24.9] TC [53.4;47.1] CC [22.9;28.0]

0.262

 

A6890-C)

     

NS

C [16.6;14.6]

0.538 (1.155)

AA [68.4;73.7] AC [30.1;23.5] CC [1.5;2.8]

0.138

 

ESR2 (T-1213C)

Chinese (50-84 yrs)

F

Hip fracture Forearm fracture Vertebral fracture

80

425

The C allele of T-1213C had a 2.74-fold increased risk of fracture in menopausal women (P = 0.009).

NA

NA

NA

NA

[86]

SEMA7A (+15667G > A,+15775C > G,+16285C > T,+19317C > T,+22331A > G)

Korean

F

Vertebral fracture

91

429

GCCCC showed an association with risk of vertebral fracture (OR = 1.87-1.93, P = 0.02-0.03).

A [49.5;48.9] G [17.6;19.7] T [29.1;25] T [17.5;13.6] G [52.4;46.8] GCCCA [39.0;37.9] AGCCG [8.8;12.5] ACTTG [11.0;12.1] GCCCC [18.1;10.4]

0.931 (1.020) 0.625 (0.863) 0.471 (1.204) 0.314 (1.365) 0.307 (1.266) 0.934 (1.020) 0.310 (0.669)0.762 (0.895) 0.056 (1.820)

NA

NA

[88]

ADRB3 (Trp64Arg)

Caucasian (≥ 60 yrs)

F & M

Hip fracture

51

489

NS

NA

NA

TrpTrp [88.2;85.1] (TrpArg/ArgArg) [11.8;14.9]

0.543

[89]

   

Vertebral fracture

80

489

NS

NA

NA

TrpTrp [87.5;85.1] (TrpArg/ArgArg) [12.5;14.9]

0.568

 
   

Wrist fracture

63

489

NS

NA

NA

TrpTrp [90.5;85.1] (TrpArg/ArgArg) [9.5;14.9]

0.249

 
   

Any fractures

222

489

NS

NA

NA

TrpTrp [86.9;85.1] (TrpArg/ArgArg) [13.1;14.9]

0.511

 

a longitudinal studies; **A1 = 4 repeats, A2 = 2 repeats, and A3 = 5 repeats; NS: non-significant; OR: odds ratio; [frequency#] frequency in case and control (%); NA: not available; effect originally reported: the reported effects in original paper; Allele p-value: the p value of Chi-square test for the distribution of allele in case and control; genotype p-value: the p value of chi-square test for the distribution of genotype in case and control.

Significant gene-gene interactions between candidate genes on ROFs are reported in several studies [27, 28]. Interactions between ER-α and VDR gene polymorphisms lead to increased risk of osteoporotic vertebral fractures in women [28]. Uitterlinden et al. [27] investigated both the interaction effects of the VDR and COLIA1 polymorphisms on ROF in 1004 postmenopausal women, and concluded that interlocus interaction was likely to be an important component of ROF.From statistical point, the inconsistency may be due to false positive studies, false negative studies or true variability in association [53, 54]. To qualify the replication studies per association for the same design OF studies, we calculated the number of positive and negative studies according to statistical threshold at p < 0.05. For each study, we re-evaluated the distribution of alleles or genotypes in cases and controls and calculated p values by chi-square analysis. Table 5 lists the number of positive and negative studies per association. For the VDR and ER genes, the number of total studies for each association is relatively small; moreover, most of these studies are negative. For the association between the Sp1 polymorphism in the COL1A1 gene and ROF, five of nine total studies support that Sp1 variant contributes to susceptibility to vertebral fracture. Although many studies with small sample size and negative association are not listed here, the high replications on this association could not be reasonably explained by publication bias. Therefore, this evidence, together with the positive association in meta-analysis, supports that Sp1 variant contributes to susceptibility to ROF with a high confidence.
Table 5

Number of association significant studies for vertebral and hip fractures

Associated gene, phenotypes

Variant

Number of studies total

Number of studies positive

Number of studies negative

VDR, vertebral fracture

BsmI

3

0

3

VDR, vertebral fracture

TaqI

3

0

3

VDR, vertebral fracture

ApaI

3

0

3

VDR, vertebral fracture

FokI

3

1

2

VDR, vertebral fracture

Cdx2

2

1

1

VDR, vertebral fracture

BsmI, ApaI & TaqI

1

0

1

VDR, hip fracture

BsmI

3

1

2

VDR, hip fracture

TaqI

2

1

1

VDR, hip fracture

ApaI

1

0

1

ERα, vertebral fracture

PvuII

2

0

2

ERα, vertebral fracture

XbaI

1

0

1

ERα, vertebral fracture

PvuII & XbaI

3

2

1

ERα, hip fracture

PvuII

1

0

1

ERα, hip fracture

XbaI

1

0

1

COL1A1, vertebral fracture

Sp1

9

5

4

COL1A1, hip fracture

Sp1

4

1

3

To qualify the replication studies per association for the same design OF studies, we calculated the number of positive and negative studies according to statistical threshold at p < 0.05. For each study, we re-evaluated the distribution of alleles or genotypes in cases and controls and calculated p values by chi-square analysis. We only calculate the number of positive and negative association studies between VDR, ER and COL1A1 genes and vertebral and hip fractures, because VDR, ER and COL1A1 genes are three extensively studied genes, and vertebral and hip fractures commonly occur in human.

To date only limited candidate genes have been tested their importance on the ROF (mostly in Caucasians). The results lack reproducibility according to skeletal sites and polymorphisms, so we may not confidently make a conclusion on any association. Here, we addressed the major potential reasons for inconsistent association results for ROF.

First, heterogeneities in study design, population characteristics, sampling scheme, age-related character of fracture, and skeletal sites, are the major potential source for the inconsistent association.

Second, the lack of sufficient power may easily generate unreliable and unreproducible results. Sample size is the primary factor related with statistical power. As suggested by Shen et al. [55] because most of the individual determinants of complex diseases have small effects [56], considering the modest genetic effects as such, sample sizes at the scale of thousands are generally needed to generate reliable and replicable associations. However, very few of the present association studies for ROF reach this level, and most of them have used sample sizes smaller than 1,000. Therefore, the current sample size is probably not suitable to reach a reliable conclusion.

Third, the population-based candidate gene association is widely used to identify genes underlying susceptibility to OF because it is easy to recruit samples and powerful to detect small genetic effects. But the population admixture easily yields to false positive or false negative population association results [57] and thus produces the inconsistent association results. TDT is a family-based association analysis to avoid the effects of population admixture [58]. Its sampling unit consists of two parents with more than one-affected children. OFs mainly appear at the older stage of one’s life, and it is difficult to collect families which have more than one affected children for TDT design.

Fourth, the complicated nature of etiology of ROF including low effects of genes, allelic heterogeneity, epistasis, gene-gene and gene-environmental interaction, incomplete penetrance and pleiotropy may easily result in inconsistent results. Current statistical method cannot simultaneously consider these complicated natures. Additionally, the development of OF is a multistage process, and each stage in the various cell types involved is probably regulated differently and by a different set of genes and factors.

Fifth, the definition of OF is difficult to standardize, thus easily resulting in phenotyping errors. OF is phenotypically associated with many factors such as BMD, bone size and structure, bone loss, and propensity to fall [59]. In clinical practice, self-report with subsequent verification by medical records is a typical way of ascertainment of OFs, but the accuracy of ascertainment of fractures at different sites is limited [60, 61] and thus easily results in phenotyping errors [62]. Phenotyping and genotyping errors can significantly diminish power and even cause false-positive results [63, 64].

Future prospects

Several well-written articles extensively address some problems in genetic dissection of complex diseases and provide useful guidelines for performing and interpreting such studies [55, 65]. Shen et al. [55] proposed some potential remedies and suggestions for current association studies. For instance, the future studies should not only calculate the statistical power to select sample size but also consider the suboptimal conditions, such as weak effects and rare alleles. For population admixture, potential stratification should be assessed and controlled using the method of genomic control [66] or structured association [67].

In my opinion, for future progress, a large homogenous sample is required to enhance the statistical power for OF association. The relatively small numbers would have tended to limit the power of these associations and to increase the possibility of false positive errors. Large samples may augment weak association signals and are less confused by random statistical fluctuations.

Multi-center international collaborations are the most expedient strategy in the identification of genes for ROF. These collaborations would be through the efforts of a consortium to merge and jointly analyze all extant data sets for association. This will rapidly generate large samples collected by using standardized protocols in study design, population characteristics, sampling scheme, age-related character of fracture, phenotypic measurement, and marker sets etc. The GENOMOS consortium is a good example in this respect in the field of genetics of osteoporosis [68]. Current meta-analyses will be a useful adjunct to allow a compilation and limited quantitative treatment of existing association data. However, meta-analyses have well-known limitations, and will unlikely resolve the inconsistent associations without standardized phenotyping and standardized genotyping [69].

Currently only limited candidate genes in searching for OF risk genes have been investigated because traditional candidate gene association analysis can only investigate limited genes at a time. However, the ROF may be determinated by multiple loci. Therefore, to systemically understand the complex etiology, it is necessary to systemically screen the loci underlying susceptibility to OF. Recent rapid progress in SNP genotyping technology and in HapMap project have now made it feasible and timely to pursue a powerful whole genome association scan studies to systemically identify genes for ROF at the DNA level.

Precisely designed repetitive studies on novel candidate genes are the first important step to finally identify genes underlying susceptibility to OF. When we find consistent association at the DNA level in identifying genes underlying ROF, how should we take the next step? A definite conclusion about finding genes underlying ROF needs substantial and complementary evidence from three levels of gene function (i.e., DNA, RNA and protein). Therefore, after the candidate genes are conclusively identified to be associated with ROF, the next step is to determine how the genes contribute to the ROF on the RNA and protein levels. On the RNA level, Northern blots and real-time reverse transcription PCR (RT-PCR) are extensively used methods. On the protein level, the most commonly used techniques are Western blotting and enzyme-linked immunosorbent assay (ELISA). If all these tasks are successfully completed, one may make a conclusion of gene discovery for ROF.

Currently, with the rapid development of proteomics and genomics technologies, a high-throughput strategy for ROF gene identification becomes available [70]. This strategy traces back from protein to RNA and further to DNA via utilizing high-throughput methods of the technologies of proteomics and genomics. The technologies of proteomics and microarrays are the latest breakthroughs in experimental molecular biology, which allow for systemically monitoring the protein and RNA expression, respectively. The first step for this strategy is to find differentially expressed proteins in tissues of subjects with vs. without OF by protein expression proteomics technologies. Then at the RNA level, high-throughput technologies, such as the microarray method, are used to validate the differently expressed genes. Finally, genetic epidemiology analyses are performed to confirm the association results at the DNA level.

Conclusion

The accumulated evidence supports that ROF is regulated by moderate genetic factors. Inconsistent associations between candidate genes and ROF at the DNA level occur now, and it is difficult to make a conclusion on one association. Future studies in large homogenous sample are required. Multi-center international collaborations are the most expedient strategy in the identification of genes for ROF. A high-throughput strategy for OF gene identification backward from protein to RNA and further to DNA becomes available.

Identifying genetic determination of ROF can lead to new diagnostics for the early detection of individuals with high risk of fracture and also help identify responder /non-responder groups for particular therapies, and thus can result in the development of better diagnosis, prevention and treatment strategies of OF.

Acknowledgement

The study was supported by grants from Natural Science Foundation of China (30600364, 30230210, and 30470534).

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2007