Meta-analysis of COL1A1 Sp1 polymorphism in relation to bone mineral density and osteoporotic fracture

Bone 32 (2003) 711–717

www.elsevier.com/locate/bone

Meta-analysis of COL1A1 Sp1 polymorphism in relation to bone mineral density and osteoporotic fracture V. Mann1 and S.H. Ralston* Bone Research Group, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, UK Received 10 July 2002; revised 21 January 2003; accepted 11 February 2003

Abstract Genetic factors play an important role in the pathogenesis of osteoporosis and several candidate gene polymorphisms have been implicated in the regulation of this process. One of the most widely studied is the Sp1 binding site polymorphism in the COL1A1 gene. This polymorphism has been associated with BMD and osteoporotic fracture in several studies, but the data from different studies have been conflicting. Here we have attempted to clarify the association between COL1A1 Sp1 alleles, BMD, and osteoporotic fracture by conducting a meta-analysis of 26 published studies including 7849 participants. Under a fixed effects model, BMD values at the lumbar spine (6800 subjects) were significantly lower in the “Ss” genotype group when compared with “SS” homozygotes (standardized mean difference ⫽ 0.131 [95% CI, 0.06,0.16], P ⫽ 0.00005) but the difference was not significant for the “ss” comparison (0.09 [⫺0.03,0.21], P ⫽ 0.13). At the femoral neck (6750 subjects) BMD values were lower in the “Ss” genotype (0.14 [0.08,0.19], P ⬍ 0.00001) and lower still in the “ss” genotype group (0.19 [0.07,0.31], P ⫽ 0.001). Similar results were found when the data were analyzed under a random effects model. Analysis of fracture data (6961 subjects) showed an increased odds ratio for any fracture in “Ss” subjects (1.26 [95% CI 1.09,1.46], P ⫽ 0.002) and an even greater increase in “ss” subjects (1.78 [1.30,2.43], P ⫽ 0.0003). Subgroup analysis showed that increased risk was largely attributable to vertebral fracture where the odds ratio was 1.37 [1.15,1.64] for “Ss” (P ⫽ 0.0004) and 2.48 [1.69,3.65] for “ss” (P ⬍ 0.00001). The risk of nonvertebral fracture was not increased in relation to the COL1A1 genotype, although power to detect an effect was limited by the fact that fewer studies had analyzed nonvertebral fracture. We conclude that the COL1A1 Sp1 alleles are associated with a modest reduction in BMD and a significant increase in risk of osteoporotic fracture, particularly vertebral fracture. © 2003 Elsevier Science (USA). All rights reserved.

Introduction Osteoporosis is a common disease which is characterized by low bone mass and an increased risk of fragility fractures. Twin and family studies indicate that genetic factors play a major role in regulating BMD and other determinants of fracture risk, but the molecular genetic basis of osteoporosis remains poorly understood [1]. While linkage studies in man have identified many of the genes responsible for monogenic bone diseases, less is known about the genes which predispose to osteoporosis in the general population. Although it is generally agreed that several genes contrib-

* Corresponding author. E-mail address: [email protected] (S.H. Ralston). 1 Current address: Musculoskeletal Research Unit, University of Edinburgh Medical School, Edinburgh EH8 9AG, United Kingdom.

ute, each has a relatively small effect on BMD and other determinants of fracture risk [1]. Candidate genes which have been studied in relation to BMD and osteoporotic fractures include the vitamin D receptor [2], the estrogen receptor [3], the COL1A1 gene [4], transforming growth factor beta-1 [5], and many others [1]. For most of these candidate genes there has been a lack of consistency between studies, making interpretation of the overall effect difficult. These discrepancies may partly be explained by the fact that many studies have been small and insufficiently powered to detect subtle differences in BMD and fracture risk which result from the effect of an individual candidate gene [6,7]. The COL1A1 Sp1 binding site polymorphism initially described in 1996 [4] is one of the most widely studied candidate genes for osteoporosis [1]. Functional studies have shown that the polymorphism alters binding of Sp1 to its recognition site in DNA and is associated with

8756-3282/03/$ – see front matter © 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S8756-3282(03)00087-5

712

V. Mann, S.H. Ralston / Bone 32 (2003) 711–717

disturbances in COL1A1 transcription, collagen protein production, and the biomechanical properties of bone [8]. Differences in ultrasound properties of bone, femoral neck geometry [9], and response to bisphosphonate therapy have also been reported in subjects of different genotype [10]. While-several investigators have found positive associations between COL1A1 alleles, BMD, and fracture, no significant association has been found in other studies; therefore in order to clarify and quantify the strength of association between COL1A1 alleles and predisposition to osteoporosis we have used the technique of meta-analysis to explore the relationship between the Sp1 binding site polymorphism of the COL1A1 gene, BMD, and osteoporotic fracture.

Methods Identification of relevant studies Clinical studies in which the COL1A1 Sp1 polymorphism had been related to BMD and/or osteoporotic fracture were identified by electronic searches of MEDLINE between October 1996 and December 2001, using several combinations of search terms including “collagen,” “COL1A1,” “polymorphism,” “genetics,” “fracture,” and “BMD.” We also screened references of retrieved articles and review articles to identify potentially eligible studies that might have been missed in the electronic search. Only studies that had been published in peer-reviewed journals within this period were included; we did not include data published in abstract form. We excluded studies that simply recorded the prevalence of COL1A1 alleles in different populations [11] and studies of populations in which the polymorphism was absent [12,13]. Data recording For each study we recorded the number of individuals in each genotype group with fractures, the genotype specific mean and standard deviation of BMD values at the lumbar spine and femoral neck, and the mean and standard deviation of weight height and body mass index. Data that were not available in the source publications were obtained from the corresponding author whenever possible. Statistical methods Data were analyzed using the Revman 4.1 software package available from the Cochrane Collaboration (www. cochrane.dk). Two comparisons were made: (a) “SS” homozygotes with “Ss” heterozygotes; and (b) “SS” homozygotes with “ss” homozygotes. For analysis of categorical variables, we calculated the odds ratio under fixed effect and random effect models. For continuous variables we calculated standardized mean difference under fixed effects and random effects models. The random effect

model incorporates between-study variation, but if there is no significant heterogeneity as determined by a ␹2 test (i.e., P ⬎ 0.10), the results from both models are comparable. Funnel plots were performed to look for evidence of publication bias [14].

Results Twenty-six eligible studies which included a total of 7849 individuals were identified for inclusion in the metaanalysis. Details of these studies are summarized in Table 1. Lumbar spine bone mineral density Twenty studies were identified with a total of 7849 participants for which spine BMD had been measured. Values for spine BMD were significantly lower in the “Ss” (n ⫽ 2090) versus “SS” (n ⫽ 4380) genotype groups. Under a fixed effects model, the difference was 0.11 [0.06,0.16] Z-score units (P ⫽ 0.00005) (Fig. 1a) compared with 0.11 [0.02,0.19] for a random effects model (P ⫽ 0.01). For the “SS” (n ⫽ 4380) versus “ss” (n ⫽ 330) comparison the BMD difference was 0.09 [⫺0.03,0.21] (P ⫽ 0.13) under a fixed effects model (Fig. 1b) and 0.10 [⫺0.09,0.28] under a random effects model (P ⫽ 0.3). Since vertebral compression fractures can cause spurious elevations in BMD, a subgroup analysis was performed in which spine BMD values from cohorts of individuals with vertebral fracture were excluded. These comprised the studies of Aerssens et al. [15], Langdahl et al. [16], and McGuigan et al.. [17]. This subgroup analysis also showed significantly lower BMD values in carriers of the “s” allele with evidence of a gene-dose effect. For the “Ss” (n ⫽ 1802) versus “SS” (n ⫽ 3738) comparison, the between genotype difference was 0.09 [0.04,0.15] Z-score unit under a fixed effects model (P ⫽ 0.001) and 0.08 [⫺0.01,0.16] for a random effects model (P ⫽ 0.08). For the “SS” (n ⫽ 3738) versus “ss” (n ⫽ 285) comparison, the BMD difference was 0.16 [0.04,0.29] P ⫽ 0.009 under both fixed and random effects models since the ␹2 test for heterogeneity was not significant (P ⫽ 0.56). Femoral neck bone mineral density Twenty-one studies were identified including a total of 6750 participants in which femoral neck BMD had been measured and this also showed lower BMD values in carriers of the “s” allele when compared with “SS” homozygotes. The BMD difference between genotypes for the “SS” (n ⫽ 4357) and “Ss” (n ⫽ 2072) comparison under a fixed effects model was 0.14 [0.08,0.19] Z-score units (P ⬍ 0.00001) (Fig. 2a) and 0.24 [0.07,0.40] for a random effects model (P ⫽ 0.005). For the “SS” (n ⫽ 4357) versus “ss” (n ⫽ 321) comparison the BMD difference was 0.19 [0.07,0.31] (P ⫽ 0.001) under both fixed and random effects models since the ␹2 test for heterogeneity was nonsignificant (P ⫽ 0.26) (Fig. 2b).

V. Mann, S.H. Ralston / Bone 32 (2003) 711–717

713

Table 1 Studies included in the meta-analysis Author

Country

Study design

Number of participants

Age

Sex

Aerssens et al. [15] Ashford et al. [30] Alvarez et al. [31] Berg et al. [32]

Belgium UK Spain Norway

Case–control Population-based cohort Case–control Population-based cohort

374 (cases 135, controls 239) 314 44 (cases 20, controls 24) 269

69–87 75 25–52 12–13

Braga et al. [33] Efstathiadou et al. [34] Garnero et al. [35] Grant et al. [4] Hampson et al. [36] Harris et al. [22]

Italy Greece France UK UK USA

Clinic referrals Clinic referrals Population-based cohort Case–control Clinic referrals (Intervention study)

715 154 220 299 72 243

63–64 47–61 34–46 50–70 37–48 65

Heegaard et al. [24] Hustmyer et al. [37] Keen et al. [25] Langdahl et al. [16]

Denmark USA UK Denmark

Case–control Twin study Case–control Case–control

133 (cases 16, controls 117) 78 185 (cases 55, controls 130) 249 (cases 105, controls 144)

45–54 21–49 45–64 28–70

Liden et al. [38] McGuigan et al. [17] McGuigan et al. [26]

Sweden UK UK

Case–control Case–control Clinic referral

136 (cases 64, controls 72) 181 (cases 93, controls 88) 341

58–77 69–75 56–77

Female Female Female Male ⫽ 126 Female ⫽ 143 Female Female Female Female Female Male ⫽ 108 Female ⫽ 135 Female Female Female Male ⫽ 95 Female ⫽ 154 Female

Peris et al. [39] Roux et al. [40] Sainz et al. [41] Sowers et al. [42] Tao et al. [43]

Spain France USA USA Australia

Case–control Case–control Population-based cohort Population-based cohort Population-based cohort

95 (cases 35, controls 60) 217 (cases 110, controls 107) 109 259 258

31–71 45–90 6–12 28–48 7.0–8.9

Uitterlinden et al. [21] Valimaki et al. [44] Van Pottelbergh et al. [45] Weichetova et al. [46]

The Netherlands Finnish Belgium Czech Republic

Population-based cohort Population-based cohort Population-based cohort Case–control

1778 601 352 252 (cases 126, control 126)

44–54 85–98 71–86 55–68

Fracture Seventeen studies were identified including a total of 6961 individuals in which fracture data were available. These included 1326 patients with fractures and 5635 controls. The odds ratio for fracture was 1.26 [1.09,1.46] for the “Ss” versus “SS” comparison under a fixed effects model (P ⫽ 0.002) (Fig. 3a) and 1.27 [1.05,1.54] under a random effects model (P ⫽ 0.01). For the “ss” versus “SS” comparison, the odds ratio for fracture was 1.78 [1.30,2.43] under a fixed effects model (P ⫽ 0.0003) (Fig. 3b) and 1.84 [1.13,3.00] under a random effects model (P ⫽ 0.01). A subgroup analysis was performed analysing vertebral and nonvertebral fracture separately. For vertebral fracture (899 patients with fractures and 3757 controls), the odds ratio was 1.37 [1.15,1.64] for the “Ss” versus “SS” comparison under a fixed effects model (P ⫽ 0.0004) and 1.41 [1.12,1.78] under a random effects model (P ⫽ 0.003). For the “ss” versus “SS” comparison, the odds ratio for vertebral fracture was 2.48 [1.69,3.65] under a fixed effects model (P ⬍ 0.00001) and 2.52 [1.39,4.57] under a random effects model (P ⫽ 0.002) (Table 2a). For nonvertebral fracture (399 patients with fractures and 1812 controls), the odds ratio was 1.11 [0.84,1.46] for the “Ss” versus “SS” comparison under a fixed effects model (P ⫽ 0.5) and 1.10

Male ⫽ 156 Female ⫽ 185 Male Female Female Female Male ⫽ 120 Female ⫽ 138 Female Female Male Female

[0.76,1.60] under a random effects model (P ⫽ 0.6). For the “ss” versus “SS” comparison, the odds ratio for vertebral fracture was 1.46 [0.80,2.67] under a fixed effects model (P ⫽ 0.2) and 1.63 [0.56,4.75] under a random effects model (P ⫽ 0.4) (Table 2b). Body mass index, height, weight, and age Data on body mass index were available in 17 studies, data on weight in 14 studies, and data on height in 12 studies. Body mass index was significantly lower in carriers of the “s” allele (Table 3). For the “SS” versus “Ss” comparison, the difference was 0.08 [0.02,0.14] under a fixed effects model (P ⫽ 0.008) and 0.08 [⫺0.02,0.18] under a random effects model (P ⫽ 0.12). Values for the “SS” versus “ss” comparison were 0.11 [⫺0.02,0.24] under both fixed and random effects models (P ⫽ 0.09) since the ␹2 test for heterogeneity was not significant (P ⫽ 0.76). Body weight was also lower in carriers of the “s” allele (Table 3). For the “SS” versus “Ss” comparison the difference was 0.10 [0.03, 0.16] P ⫽ 0.002 under fixed effects model and 0.08 [⫺0.03,0.19] P ⫽ 0.16 under a random effects model. For the “SS” versus “ss” comparison the difference was 0.06 [⫺0.08,0.20] under both fixed and random effects models (P ⫽ 0.4), since the ␹2 test for heterogeneity was

714

V. Mann, S.H. Ralston / Bone 32 (2003) 711–717

Fig. 1. Meta-analysis for COL1A1 polymorphism and association with Lumbar Spine-BMD (LS-BMD). (a) “SS” homozygote versus “Ss” heterozygote. (b) “SS” homozygote versus “ss” homozygote. Each study is shown as the point estimate of the standardized mean difference with 95% confidence intervals as analyzed using a fixed effect model. Data for BMD in males (M)/females (F) and for BMD in control (a)/fracture groups (b) were treated individually in the meta-analysis. The diamond shows the overall effect with an increased association denoted as that above 0. Where the diamond lies toward the right of the vertical line this indicates a reduced LS-BMD in (a) the “Ss” compared with the “SS” and (b) the “ss” compared with the “SS”.

Fig. 2. Meta-analysis for COL1A1 polymorphism and association with Femoral Neck-BMD (FN-BMD). (a) “SS” homozygote versus “Ss” heterozygote. (b) “SS” homozygote versus “ss” homozygote. Again as for LS-BMD each study is shown as the point estimate of the standardized mean difference with 95% confidence intervals as analyzed using a fixed effects model. Data for BMD in males (M)/females (F) and in control (a)/fracture groups (b) were treated individually in the meta-analysis. The diamond shows the overall effect with an increased association denoted as that above 0. Where the diamond lies toward the right of the vertical line this indicates a reduced FN-BMD in (a) the “Ss” compared with the “SS” and (b) the “ss” compared with the “SS”.

V. Mann, S.H. Ralston / Bone 32 (2003) 711–717

715

Fig. 3. Meta-analysis for COL1A1 polymorphism and association with fracture. Odds ratio (OR) for fracture is reported with 95% confidence intervals as analyzed using a fixed effects model for (a) “Ss” genotype and (b) “ss” genotype. Where data was available for males (M) and females (F) they were reported individually within the meta-analysis as was data for vertebral (1) and non-vertebral (2) fracture from the study of Uitterlinden et al. [21]. The diamond shows the overall risk and where it lies toward the right of the vertical line, above 0, this indicates an increased risk of fracture associated with genotype.

nonsignificant (P ⫽ 0.79). “SS” individuals were taller than those who carried the “s” allele (Table 2); when the “SS” versus “Ss” comparison for height was analyzed the difference was ⫺0.13 [⫺0.20,⫺0.07] P ⫽ 0.00009 under a fixed effects model and ⫺0.31 [⫺0.60,⫺0.01] P ⫽ 0.04 under a random effects model. For the “SS” versus “ss” comparison the difference was ⫺0.92 [⫺1.61, ⫺0.23] P ⫽ 0.009 using a fixed effects model and ⫺0.71[⫺2.50,1.07] P ⫽ 0.4 using a random effects model. Individuals who had sustained vertebral fracture were removed from the analysis. There was no overall difference in age of the subjects studied between the different genotype groups (data not shown).

Population attributable risk of fracture due to COL1A1 Sp1 alleles We calculated the attributable fraction of fractures caused by the COL1A1 polymorphism using the method described by Efstathiadou et al. [18] from the prevalence of the “Ss” and “ss” genotypes in the fracture population and the relative risk of fracture associated with each genotype. This gave estimates of 13.6% for all fractures and 15.7% for vertebral fracture.

Discussion

We analyzed funnel plots for all outcome measures to look for evidence of publication bias, but these were symmetrical, providing evidence against selective publication of positive studies. An example of a funnel plot for lumbar spine BMD is shown in Fig. 4, demonstrating symmetry for “SS” versus “Ss” genotype comparison.

This meta-analysis confirms and extends the results of two previously reported meta-analyses in which the COL1A1 Sp1 binding site polymorphism was found to be significantly associated with prevalent osteoporotic fracture [18] and BMD [8] in Caucasian subjects from various countries in Europe and from the USA. The genotypespecific differences in LS BMD and FN BMD reported here were slightly smaller than in a previous meta-analysis which included 16 studies [8], but the differences were not signif-

Table 2a Vertebral fractures (899 patients and 3757 controls)

Table 2b Nonvertebral fractures (399 patients and 1812 controls)

Funnel plot analysis

Comparison

Effect size odds ratio [95% CI]

P value

SS vs Ss

1.37 [1.15,1.64]Fixed Effect 1.41 [1.12,1.78]Random Effect 2.48 [1.69,3.65]Fixed Effect 2.52 [1.39,4.57]Random Effect

P P P P

SS vs ss

⫽ ⫽ ⬍ ⫽

0.0004 0.003 0.00001 0.002

Comparison

Effect size odds ratio [95% CI]

P value

SS vs Ss

1.11 [0.84,1.46]Fixed Effect 1.10 [0.76,1.60]Random Effect 1.46 [0.80,2.67]Fixed Effect 1.63 [0.56,4.75]Random Effect

P ⫽ 0.5 P ⫽ 0.6 P ⫽ 0.2 P ⫽ 0.4

SS vs ss

716

V. Mann, S.H. Ralston / Bone 32 (2003) 711–717

icant since the confidence intervals of the estimates overlapped. The estimates for fracture risk reported in this study are also largely similar to those reported previously (8;18), in showing an odds ratio of 1.26 –1.27 for the “Ss” genotype and 1.78 –1.84 for the “ss” genotype, with a greater risk when the analysis was limited to vertebral fracture. This is probably due to the fact that nonskeletal risk factors such as falls and reduced visual acuity contribute more to the pathogenesis of nonvertebral fracture [19,20] and that fewer investigators have analyzed the COL1A1 genotype in relation to nonvertebral fracture. Several previous studies have shown that the COL1A1 genotype predicts osteoporotic fracture by mechanisms which are partly independent of BMD, indicating that the polymorphism may act as a marker for bone quality as well as bone density [16,21,25,26]. In keeping with these observations, we have previously presented evidence to suggest that the COL1A1 Sp1 polymorphism is a functional variant which modulates Sp1 binding and COL1A1 gene regulation, resulting in production of an increased amount of collagen ␣1 (I) chain relative to ␣2 (I) and a decrease in bone strength as measured by biomechanical testing of bone cores from patients of different genotype [8]. Associations have also been reported with femoral neck geometry [9] and ultrasound properties of bone in some studies, although the data for quantitative ultrasound values are conflicting. The data presented here are consistent with a deleterious effect of COL1A1 alleles on bone quality, since the observed increase in vertebral fracture risk for “Ss” heterozygotes was ⫹41% which is substantially greater than the risk which would have been predicted [27,28] by the modest genotype-specific reduction in spine BMD (⫹11%) and BMI (⫹4%). The difference was even greater for “ss” homozygotes where the vertebral fracture risk was ⫹148%, compared with an expected fracture risk of ⫹9% for the difference in BMD and ⫹5% for the difference in BMI. The value of meta-analysis in this situation is that it can assist in estimating population-wide effects of genetic risk factors in human disease [29]. Our data not only support the hypothesis that the COL1A1 Sp1 polymorphism is a clinically relevant predictor of osteoporotic fracture in the general population, but also show that COL1A1 genotypes predict fractures by mechanisms which seem largely independent of the effect on BMD.

Table 3 Covariables and genotype Comparison BMI Weight Height

SS SS SS SS SS SS

vs vs vs vs vs vs

Ss ss Ss ss Ss ss

Effect size SMD fixed effect [95% CI]

P value

0.08 [0.02,0.14] 0.11 [⫺0.02,0.24] 0.10 [0.03,0.16] 0.06 [⫺0.08,0.20] ⫺0.13 [⫺0.20,⫺0.07] ⫺0.92 [⫺1.61,⫺0.23]

P ⫽ 0.008 P ⫽ 0.09 P ⫽ 0.002 P ⫽ 0.4 P ⫽ 0.00009 P ⫽ 0.009

Fig. 4. Funnel plot of LS BMD comparison of “SS” vs “Ss” genotype to determine publication bias. The genotype values for LS BMD from individual studies are plotted on the X axis while the measure of each study’s sample size is plotted on the Y axis with effect size increasing as the sample size of component studies increases.

References [1] Stewart TL, Ralston SH. Role of genetic factors in the pathogenesis of osteoporosis. J Endocrinol 2000;166(2):235– 45. [2] Morrison NA, Qi JC, Tokita A, et al. Prediction of bone density from vitamin D receptor alleles (Erratum). Nature 1997;387:106. [3] Kobayashi S, Inoue S, Hosoi T, Ouchi Y, Shiraki M, Orimo H. Association of bone mineral density with polymorphisms of the estrogen receptor gene in post-menopausal women. J Bone Miner Res 1996;11:306 –11. [4] Grant SFA, Reid DM, Blake G, Herd R, Fogelman I, Ralston SH. Reduced bone density and osteoporosis associated with a polymorphic Sp1 site in the collagen type I alpha 1 gene. Nat Genet 1996; 14:203–5. [5] Langdahl BL, Knudsen JY, Jensen HK, Gregersen N, Eriksen EF. A sequence variation: 713-8delC in the transforming growth factor-beta 1 gene has higher prevalence in osteoporotic women than in normal women and is associated with very low bone mass in osteoporotic women and increased bone turnover in both osteoporotic and normal women. Bone 1997;20(3):289 –94. [6] Cooper GS, Umbach DM. Are vitamin D receptor polymorphisms associated with bone mineral density? A meta-analysis. J Bone Miner Res 1996;11:1841–9. [7] Rubin LA, Hawker GA, Peltekova VD, Fielding LJ, Ridout R, Cole DE. Determinants of peak bone mass: clinical and genetic analyses in a young female Canadian cohort. J Bone Miner Res 1999;14(4):633– 43. [8] Mann V, Hobson EE, Li B, et al. A COL1A1 Sp1 binding site polymorphism predisposes to osteoporotic fracture by affecting bone density and quality. J Clin Invest 2001;107(7):899 –907. [9] Qureshi AM, McGuigan FE, Seymour DG, Hutchison JD, Reid DM, Ralston SH. Association between COLIA1 Sp1 alleles and femoral neck geometry. Calcif Tissue Int 2001;69(2):67–72. [10] Qureshi AM, Herd RJ, Blake GM, Fogelman I, Ralston SH. COLIA1 Sp1 polymorphism predicts response of femoral neck bone density to cyclical etidronate therapy. Calcif Tissue Int 2002;70(3):158 – 63. [11] Beavan S, Prentice A, Dibba B, Yan L, Cooper C, Ralston SH. Polymorphism of the collagen type I alpha 1 gene and ethnic differences in hip-fracture rates. N Engl J Med 1998;339:351–2. [12] Nakajima T, Ota N, Shirai Y, et al. Ethnic difference in contribution of Sp1 site variation of COL1A1 gene in genetic predisposition to osteoporosis. Calcif Tissue Int 1999;65(5):352–3. [13] Lim S-K, Li SZ, Won YJ, Shin W-Y, Lee HC, Huh KB. Lack of association between a polymorphic Sp1 binding site in collagen type 1 alpha 1 gene and osteoporosis in Korean women. J Bone Miner Res 1997;12(Suppl):S491 [Abstract].

V. Mann, S.H. Ralston / Bone 32 (2003) 711–717 [14] Egger M, Davey SG, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. Br Med J 1997;315(7109): 629 –34 [see comments]. [15] Aerssens J, Dequeker J, Peeters J, Breemans S, Broos P, Boonen S. Polymorphisms of the VDR, ER and COL1A1 genes and osteoporotic hip fracture in elderly postmenopausal women. Osteoporos Int 2000; 11(7):583–91. [16] Langdahl BL, Ralston SH, Grant SFA, Eriksen EF. An Sp1 binding site polymorphism in the COL1A1 gene predicts osteoporotic fractures in men and women. J Bone Miner Res 1998;13:1384 –9. [17] McGuigan FEA, Reid DM, Ralston SH. Susceptibility to osteoporotic fracture is determined by allelic variation at the Sp1 site, rather than other polymorphic sites, at the COL1A1 locus. Osteoporosis Int 2000;11(4):338 – 43. [18] Efstathiadou Z, Tsatsoulis A, Ioannidis JP. Association of collagen I alpha 1 Sp1 polymorphism with the risk of prevalent fractures: a meta-analysis. J Bone Miner Res 2001;16(9):1586 –92. [19] Cummings SR, Nevitt MC, Browner WS, et al. Risk factors for hip fracture in white women. Study of Osteoporotic Fractures Research Group. N Engl J Med 1995;332(12):767–73. [20] Dargent-Molina P, Favier F, Grandjean H, et al. Fall-related factors and risk of hip fracture: the EPIDOS prospective study. Epidemiologie de l’osteoporose. Lancet 1996;348(9021):145–9. [21] Uitterlinden AG, Burger H, Huang Q, et al. Relation of alleles of the collagen type I ␣ 1 gene to bone density and risk of osteoporotic fractures in postmenopausal women. N Engl J Med 1998;338:1016–22. [22] Harris SS, Patel MS, Cole DE, Dawson-Hughes B. Associations of the collagen type I alpha1 Sp1 polymorphism with five-year rates of bone loss in older adults. Calcif Tissue Int 2000;66(4):268 –71. [23] MacDonald HM, McGuigan FA, New SA, et al. COLlA1 Sp1 polymorphism predicts perimenopausal and early postmenopausal spinal bone loss. J Bone Miner Res 2001;16(9):1634 – 41. [24] Heegaard A, Jorgensen HL, Vestergaard AW, Hassager C, Ralston SH. Lack of influence of collagen type I alpha1 Sp1 binding site polymorphism on the rate of bone loss in a cohort of postmenopausal Danish women followed for 18 years. Calcif Tissue Int 2000;66(6): 409 –13. [25] Keen RW, Woodford-Richens KL, Grant SFA, Ralston SH, Lanchbury JS, Spector TD. Polymorphism at the type I collagen (COL1A1) locus is associated with reduced bone mineral density, increased fracture risk and increased collagen turnover. Arthritis Rheum 1999; 42:285–90. [26] McGuigan FE, Armbrecht G, Smith R, Felsenberg D, Reid DM, Ralston SH. Prediction of osteoporotic fractures by bone densitometry and COL1A1 genotyping: a prospective, population-based study in men and women. Osteoporos Int 2001;12(2):91– 6. [27] Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. Br Med J 1996;312(7041):1254 –9. [28] Lunt M, Felsenberg D, Reeve J, et al. Bone density variation and its effects on risk of vertebral deformity in men and women studied in thirteen European centers: the EVOS study. J Bone Miner Res 1997; 12(11):1883–94. [29] Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of genetic association studies. Nat Genet 2001; 29(3):306 –9. [30] Ashford RU, Luchetti M, McCloskey EV, et al. Studies of bone density, quantitative ultrasound, and vertebral fractures in relation to collagen type I alpha I alleles in elderly women. Calcif Tissue Int 2001;68(6):348 –51.

717

[31] Alvarez L, Oriola J, Jo J, et al. Collagen type I alpha1 gene Sp1 polymorphism in premenopausal women with primary osteoporosis: improved detection of Sp1 binding site polymorphism in the collagen type 1 gene. Clin Chem 1999;45(6 Pt 1):904 – 6. [32] Berg JP, Lehmann EH, Stakkestad JA, Haug E, Halse J. The Sp1 binding site polymorphism in the collagen type I alpha 1 (COL1A1) gene is not associated with bone mineral density in healthy children, adolescents, and young adults. Eur J Endocrinol 2000;143(2):261–5. [33] Braga V, Mottes M, Mirandola S, et al. Association of CTR and COL1A1 alleles with BMD values in peri- and postmenopausal women. Calcif Tissue Int 2000;67(5):361– 6. [34] Efstathiadou Z, Kranas V, Ioannidis JP, Georgiou I, Tsatsoulis A. The Sp1 COL1A1 gene polymorphism, and not vitamin D receptor or estrogen receptor gene polymorphisms, determines bone mineral density in postmenopausal Greek women. Osteoporos Int 2001;12(4): 326 –31. [35] Garnero P, Borel O, Grant SFA, Ralston SH, Delmas PD. Collagen I ␣ 1 polymorphism, bone mass and bone turnover in healthy French pre-menopausal women: the OFELY study. J Bone Miner Res 1998; 13(5):813– 8. [36] Hampson G, Evans C, Petitt RJ, et al. Bone mineral density, collagen type 1 alpha 1 genotypes and bone turnover in premenopausal women with diabetes mellitus. Diabetologia 1998;41(11):1314 –20. [37] Hustmyer FG, Lui G, Johnston CC, Christian J, Peacock M. Polymorphism at an Sp1 binding site of COL1A1 and bone mineral density in pre-menopausal female twins and elderly fracture patients. Osteoporosis Int 1999;9:346 –50. [38] Liden M, Wilen B, Ljunghall S, Melhus H. Polymorphism at the Sp1 binding site in the collagen type I alpha 1 gene does not predict bone mineral density in postmenopausal women in sweden. Calcif Tiss Int 1998;63(4):293–5. [39] Peris P, Alvarez L, Oriola J, et al. Collagen type I alpha1 gene polymorphism in idiopathic osteoporosis in men. Rheumatology (Oxford) 2000;39(11):1222–5. [40] Roux C, Dougados M, Abel L, Mercier G, Lucotte G. Association of a polymorphism in the collagen I ␣ 1 gene with osteoporosis in French women. Arthritis Rheum 1998;41:187– 8. [41] Sainz J, Van Tornout JM, Sayre J, Kaufman F, Gilsanz V. Association of collagen type 1 alpha 1 gene polymorphism with bone density in early childhood. J Clin Endocrinol Metab 1999;84(3):853–5. [42] Sowers M, Willing M, Burns T, et al. Genetic markers, bone mineral density and serum osteocalcin levels. J Bone Miner Res 1999;14(8): 1411–9. [43] Tao C, Garnett S, Petrauskas V, Cowell CT. No association was found between collagen alpha type 1 gene and bone density in prepubertal children. J Clin Endocrinol Metab 1999;84(11):4293– 4 [letter]. [44] Valimaki S, Tahtela R, Kainulainen K, et al. Relation of collagen type I alpha 1 (COL1A 1) and vitamin D receptor genotypes to bone mass, turnover, and fractures in early postmenopausal women and to hip fractures in elderly people. Eur J Intern Med 2001;12(1):48 –56. [45] Van Pottelbergh I, Goemaere S, Nuytinck L, De Paepe A, Kaufman JM. Association of the type I collagen alpha 1 Sp1 polymorphism, bone density and upper limb muscle strength in community-dwelling elderly men. Osteoporos Int 2001;12(10):895–901. [46] Weichetova M, Stepan JJ, Michalska D, Haas T, Pols HA, Uitterlinden AG. OL1A1 polymorphism contributes to bone mineral density to assess prevalent wrist fractures. Bone 2000;26(3):287–90.