Context Both bone mineral density (BMD) and fracture risk have a strong genetic
component. Estrogen receptor α (ESR1) is a
candidate gene for osteoporosis, but previous studies of ESR1 polymorphisms in this field were hampered by small sample size,
lack of standardization, and inconclusive results.
Objective To generate large-scale evidence on whether 3 common ESR1 polymorphisms (intron 1 polymorphisms XbaI
[dbSNP: rs9340799] and PvuII [dbSNP: rs2234693] and
promoter TA repeats microsatellite) and haplotypes thereof are associated
with BMD and fractures.
Design and Setting Meta-analysis of individual-level data involving standardized genotyping
of 18 917 individuals in 8 European centers.
Main Outcome Measures BMD of femoral neck and lumbar spine; all fractures and vertebral fractures
Results No between-center heterogeneity was observed for any outcome in any
genetic contrast. None of the 3 polymorphisms or haplotypes had any statistically
significant effect on BMD in adjusted or unadjusted analyses, and estimated
differences between genetic contrasts were 0.01 g/cm2 or less.
Conversely, we found significant reductions in fracture risk. In women homozygous
for the absence of an XbaI recognition site, the
adjusted odds of all fractures were reduced by 19% (odds ratio, 0.81 [95%
CI, 0.71-0.93]; P = .002) and vertebral
fractures by 35% (odds ratio, 0.65 [95% CI, 0.49-0.87]; P = .003). Effects on fractures were independent of BMD and
unaltered in adjusted analyses. No significant effects on fracture risk were
seen for PvuII and TA repeats.
Conclusions ESR1 is a susceptibility gene for fractures,
and XbaI determines fracture risk by mechanisms independent
of BMD. Our study demonstrates the value of adequately powered studies with
standardized genotyping and clinical outcomes in defining effects of common
genetic variants on complex diseases.
Osteoporosis is a common disease characterized by reduced bone mass
and an increased risk of fracture, which affects up to 30% of women and 12%
of men at some point during life. Bone mineral density (BMD) is an important
clinical predictor of fracture risk, and most of the variance in BMD is genetically
determined.1,2 Many other predictors
of fragility fracture are also under genetic control, however, including ultrasound
properties of bone, biochemical markers of bone turnover, and skeletal geometry.
A wide variety of candidate genes have been investigated in relation to osteoporosis
outcomes, but one of the most widely studied is the estrogen receptor α (ESR1) gene.3 In particular,
polymorphisms defined by the restriction enzymes XbaI
(dbSNP [database of single-nucleotide polymorphisms]: rs9340799) and PvuII (dbSNP: rs2234693) in the first intron of ESR1 have been evaluated to date in approximately 40 studies, with
inconclusive results. These 2 polymorphisms are 46 base pairs apart and in
strong linkage disequilibrium with a microsatellite TA-variable number of
tandem repeats (VNTR) polymorphism4 (dbSNP:
rs3138774) situated 2.1 kb upstream in the ESR1 promoter
region. The role of this VNTR in osteoporosis outcomes is controversial, and
interpretation is further limited by analytic inconsistencies across published
There is increasing recognition that, given the common lack of replication
of results of small studies,10-12 the
delineation and establishment of common genetic risk factors for complex multigenetic
disorders, such as osteoporosis, requires large-scale investigations to clarify
subtle, but clinically important, genetic effects.12,13 Standardization
is also essential to avoid misinterpreting as genuine genetic variability
whatever differences between study teams are caused by analytical inconsistencies.
We report the results of a collaborative study using standardized genotyping
methodology on 18 917 individuals, which tests the contribution of these
3 common ESR1 polymorphisms and haplotypes thereof
on BMD and fractures.
The GENOMOS (Genetic Markers for Osteoporosis) project involves the
study of several candidate gene polymorphisms in relation to osteoporosis-related
outcomes in approximately 20 000 individuals drawn from 8 European centers.14 Participating teams contributed information on sex,
age, height, weight, TA genotype (number of TA repeats in each allele), XbaI genotype, PvuII genotype,
BMD at lumbar spine (L2-4) and femoral neck (in g/cm2), fractures
at any site, vertebral fractures, and menopausal status.
The 2 largest cohorts in the meta-analysis (Rotterdam and Aberdeen)
genotyped their entire population, whereas other cohorts generally excluded
women with secondary causes of osteoporosis or those receiving drugs that
could affect bone metabolism. Study design aspects for each cohort in the
consortium are available from the author on request. All participating centers
have received institutional review board or ethics committee approval according
to their local regulations, and participant informed consent has been obtained
according to the requirements of each center.
Genotyping for the 3 polymorphisms was performed in different centers
by using polymerase chain reaction–restriction fragment-length polymorphism,
single-base extension sequencing and 5′ nuclease Taqman assays for the XbaI and PvuII polymorphisms and
capillary electrophoresis for TA-VNTR.5-7 For XbaI and PvuII, X and P denote the absence of the respective
restriction sites (G allele and Callele, respectively).
Each center checked its own genotyping by reanalyzing at least 5% of
the samples with random selection. To ensure standardization between centers,
50 randomly selected samples from 1 center (Rotterdam) were sent in blinded
fashion to all the other cohorts for independent analysis. Results were assembled
and compared at the coordinating center. For XbaI
and PvuII, only 1 sample gave discrepant results
for XbaI in 1 cohort. For the TA repeats, 2 cohorts
systematically estimated 1 fewer repeat, and 1 estimated systematically 2
fewer repeats. Thus, readings were adjusted in these cohorts by adding 1 or
2 repeats, respectively. Aside from these systematic differences, 21 of the
allele determinations across cohorts did not agree exactly with the predominant
determination, but with the exception of 6 alleles (error rate <1%), the
difference was less than 4 repeats. No data were obtained for TA repeats in
one study (Cambridge), whereas in another study (Florence) TA repeats had
been determined with a different method that showed extensive differences
in the pilot samples. Thus, these data were not considered in any analyses.
Hardy-Weinberg equilibrium was checked on all data.
Bone mineral density was assessed by dual-energy x-ray absorptiometry
with Hologic devices in the Barcelona, DOPS (Danish Osteoporosis Study), Aarhus,
and Florence studies, Norland in the Aberdeen study, Lunar DPX-L in Rotterdam,
and a variety of devices cross-calibrated with the European Spine Phantom
in the Oxagen and Cambridge cohorts.15 Syntheses
of BMD data across studies always include also a study effect that would account
both for genuine differences in populations and potential systematic differences
between these devices. The results of the meta-analysis for BMD should be
interpreted with emphasis on the BMD differences between the contrasted genotypes
and haplotypes and not on the absolute BMD values.
We analyzed genotypes for each of the 3 polymorphisms and long-range
haplotypes (LRHs) by combining all 3 polymorphisms. The microsatellite genotypes
were clustered in 2 groups of alleles according to the bimodal appearance
of the composite distribution of the number of repeats.4,5 The
low-repeat number group (L) was defined to extend
up to the trough of the distribution, and alleles with higher numbers of repeats
were grouped in the high-repeat number group (H).
The resulting genotypes are HH,HL, and LL. Long-range haplotypes (x-p-L[A], X-P-L[B], x-P-L [C],x-p-H[D], X-P-H[E], x-P-H[F], X-p-H[G], and X-p-L[H]) were imputed by using the PHASE program.16
The main outcomes included lumbar spine BMD; femoral neck BMD; any recorded
fractures based on clinical history or radiographic evaluation, as defined
in each study; and vertebral fractures based on clinical or radiographic evaluation,
according to the criteria of McCloskey et al.17 Prevalent
fractures (at BMD determination) were considered in all cohorts. Data on incident
fractures during prospective follow-up were also collected according to clinical
history for peripheral fractures and comparison of spine radiographs at follow-up
(average, 7.4 years) vs baseline on 3469 participants in the Rotterdam cohort,
clinical history in the Aberdeen cohort, and clinical history for peripheral
fractures and spine radiographs in the small Cambridge sample. Four cohorts
(Florence, Barcelona, Aarhus, and Cambridge) consistently excluded up-front
fractures caused by high-energy trauma. In 3 of the 4 remaining cohorts (Rotterdam,
Oxagen, and DOPS), we could also separate fractures without obvious trauma
(typically vertebral fractures observed on radiographs) and low-energy trauma
from those caused by high-energy trauma according to location (face, distal
foot, distal hand) or medical history (injury, fall from height, impact sports).
This separation was not possible in Aberdeen. Genotyping was performed blinded
to the clinical data and vice versa.
For all analyses, data in each cohort were first split according to
sex. In all studies participants were unrelated, with the exception of Oxagen
pedigrees. For analytic consistency, in the main data synthesis we used only
1 randomly selected individual per Oxagen pedigree. Sensitivity analysis using
all Oxagen participants yielded largely similar results (data not shown).
For each genotype of interest, we estimated the unadjusted mean BMD
and standard deviation in each study. We then synthesized BMD differences
between genotype contrasts across studies by using fixed- and random-effects
general variance models.18 Between-study heterogeneity
was assessed by the Q statistic (considered significant for P<.10). Random-effects models incorporated the between-study heterogeneity
and allowed for a different effect in each population.18 In
the absence of between-study heterogeneity, fixed and random effects are similar.
We also performed analyses adjusting for the potential independent effect
of each polymorphism, as well as age, weight, and height, plus menopausal
status and any hormone therapy for women. Separate adjusted analyses were
performed for the genotypes of each polymorphism and for the combinations
of LRHs stemming from the 2 most common LRHs (A and E), ie, comparing individuals
with 2, 1, or no copies of haplotype A and with 2, 1, or no copies of haplotype
E. We considered study as a random factor and allowed study × genotype
(or haplotype) interactions to account for potentially variable genetic effects
across studies. The overall significance of the genetic effects was evaluated
with an F test for between-participant effects. Marginal means were also obtained. Pvalues estimated for the comparison of estimated marginal
means tended to be smaller and should be interpreted with more caution.
For fractures, we estimated the number of individuals in each genotype
and haplotype group of interest, and pairwise genotype and haplotype comparisons
were performed by estimate of an odds ratio (OR) in each study. Genotype analyses
investigated recessive and dominant models for each polymorphism, and haplotype
contrasts were based on the 2 most common LRHs (A and E). In each analysis,
ORs were evaluated for between-study heterogeneity by using the Q statistic
(considered significant for P<.10) and then synthesized
with the Mantel-Haenszel (fixed-effects) and DerSimonian and Laird (random-effects)
methods.18 We also performed sensitivity analyses
limited to incident fractures, limited to no-trauma/low-energy-trauma fractures,
and limited to women who had not received hormone therapy. Adjusted logistic
regression analyses were also performed by considering age, height, weight,
and menopausal status for women, as well as BMD. Sensitivity analyses were
performed also for the adjusted estimates after further adjusting for indices
of physical activity and ability. Data on different such indices were available
for the 4 largest cohorts (Rotterdam, Aberdeen, Oxagen, and DOPS) and for
Analyses were conducted using SPSS version 11.0 (SPSS Inc, Chicago,
Ill) and Meta-Analyst (Joseph Lau, Boston, Mass). All reported Pvalues are 2-tailed and unadjusted for multiple comparisons.
Data on 18 917 individuals were assembled, of whom 14 622
were women (3555 with current or past use of hormone therapy) (Table 1). Data on lumbar spine BMD, femoral neck BMD, any fractures,
and vertebral fractures were available on 16 370, 15 926, 18 841,
and 14 039 participants, respectively. Across the database, age quartile
cutoffs were 47.4, 50.6, 58.0, and 69.0 years for women and 56.5, 62.1, 67.3,
and 73.8 years for men. The database included 4952 individuals with any fracture
and 1072 with vertebral fractures. There were 1779 individuals with incident
fractures, the vast majority derived from the Rotterdam (n = 1260)
and Aberdeen cohorts (n = 489). Only the Rotterdam cohort had a
meaningful number of analyzable incident radiographically screened vertebral
fractures (n = 176). There were 2536 participants with no-trauma/low-trauma
fractures across the 7 cohorts with relevant data (excluding Aberdeen). Standardized
data on XbaI, PvuII, and
TA repeat genotypes were obtained in 16 147, 16 135, and 10 902
individuals, respectively (Table 1).
All 3 polymorphisms were in strong linkage disequilibrium with each other.
The distribution of TA repeats was consistently bimodal in all studies, and
overall the trough of the distribution was clearly seen at 19 repeats (Figure 1). For all cohorts, the A haplotype (x-p-L) accounted consistently for about half of the alleles
(range, 47.1% to 53.4%) and the E haplotype (X-P-H)
for almost a third (range, 29.6% to 32.4%). (Frequencies of inferred LRHs
per cohort are available from the corresponding author on request.)
In unadjusted analyses, none of the 3 polymorphisms was statistically
significantly associated with BMD in the lumbar spine or in the femoral neck
for any of the tested genotype contrasts (Figure
2 and Figure 3), with the
exception of a slightly higher femoral neck BMD with XX as compared with xx (statistically significant
at P<.05 by fixed-effects only). There was no
statistically significant between-study heterogeneity for any of the comparisons
(heterogeneity P>.10 for all). The estimated differences
in BMD were 0.01 g/cm2 or less for all genetic contrasts (Figure 2 and Figure
3). The results were similar when limited to women only, with no
significant between-study heterogeneity and maximal estimated differences
in the same range. The more sparse data on men were consistent with this picture,
but estimates had more uncertainty (Figure 2 and Figure 3). Analyses adjusted for age, height,
weight, hormone therapy, and menopausal status also showed that none of the
3 polymorphisms had a statistically significant association with BMD (not
shown in detail). The typical trend for all these analyses involved a higher
BMD, with XX over Xx and xx and with PP over Pp and pp, but differences were small and
nonsignificant. Maximum differences for marginal means were less than 0.01
g/cm2 overall and for women.
Results were similar using haplotypes (not shown in detail). The adjusted
differences in BMD between all genetic contrasts in women were always 0.01
g/cm2 or less at either skeletal site (with maximal trends typically
showing a higher BMD in E haplotype homozygotes). No clear differences were
observed in men. Interactions of age or menopausal status with genotype were
not statistically significant for any of these analyses (data not shown).
Genotype Analyses. For recessive and dominant
models of genetic contrasts (Table 2),
there was no statistically significant between-study heterogeneity for any
of the comparisons either for all fractures or for vertebral fractures alone
(heterogeneity P>.10). Thus, fixed- and random-effects
results were similar, although in a few cases random effects were somewhat
more conservative in terms of the level of statistical significance (Table 2).
There was a highly significant protection conferred by the XX genotype against the overall fracture risk, with approximately 20%
reduction in the odds (Figure 4, P<.001 by both fixed and random effects), and the magnitude
of the effect was similar in women and men. The risk for individuals with
the Xx genotype did not differ from those with xx (fixed-effects OR, 1.02; 95% confidence interval [CI],
0.94-1.10), consistent with a recessive effect of XX on
fracture risk. A favorable trend with PP disappeared
when analyses excluded XX homozygotes (fixed-effects
OR, 1.03; 95% CI, 0.91-1.17). TA repeats showed no effect.
For vertebral fractures, there was an approximately 30% reduction in
the odds of fractures with XX (Figure 4, P<.001 by fixed effects and P = .02 by random effects) and no difference
in the fracture risk between Xx and xx (fixed-effects OR, 1.11; 95% CI, 0.96-1.28); dominant models also
showed significant protective effects in the absence of xx, pp or LL (Table 2). Results were largely consistent for women and men.
Haplotype Analyses. The results of haplotype
contrasts (Table 3) were consistent
with the results of genotype contrasts. There was no significant between-study
heterogeneity for any of the analyses (heterogeneity P>.10).
When all fractures were considered, there was 20% reduction in the odds of
fractures in women homozygous for the E haplotype (Figure 4). For vertebral fractures, 30% to 50% odds reductions were
observed with either homozygosity for E haplotype or lack of homozygosity
for the A haplotype (Figure 4), and
differences between these 2 genetic models were subtle.
Sensitivity and Adjusted Analyses
Analyses limited to incident fractures suggested a similar effect for XX in women: OR, 0.83; 95% CI, 0.68-1.01; P = .07 for any incident fracture (OR, 0.77; P = .04 for the Rotterdam cohort, in which a systematic effort
was made to record radiographically vertebral fractures) and 0.54 (95% CI,
0.26-1.13) for radiographically screened incident vertebral fractures (Rotterdam).
Data on men (from the Rotterdam cohort) showed no effect, but they were limited
and thus inconclusive (OR, 0.99; 95% CI, 0.65-1.51 and 0.89; 95% CI, 0.38-2.09,
respectively, for any fracture and vertebral fractures).
Analyses limited to no-trauma/low-energy-trauma fractures suggested
a similar effect for XX in women (OR, 0.74; 95% CI,
0.61-0.90; P = .002 by fixed effects and
OR, 0.79; 95% CI, 0.60-1.03; P = .08 by
random effects, with no significant between-study heterogeneity). Data on
men showed no effect, but they were limited and thus inconclusive (OR, 1.01
by fixed and random effects).
Analyses excluding women who had received any hormone therapy also showed
a strong protective effect for XX both for any fracture
(OR, 0.71; 95% CI, 0.61-0.84; P<.001, with no
between-study heterogeneity) and for vertebral fractures (OR, 0.60; 95% CI,
0.45-0.82; P = .001 by fixed effects and
OR, 0.65; 95% CI, 0.46-0.92; P = .002 by
random effects, with no significant between-study heterogeneity).
After adjustment for age, height, weight, and menopausal status, the
OR for any fractures in women and men with the XX genotype
vs Xx and xx was 0.81 (95%
CI, 0.71-0.93; P = .002) and 0.91 (95%
CI, 0.70-1.18; P = .48), respectively.
The respective adjusted ORs for vertebral fractures were 0.65 (95% CI, 0.49-0.87; P = .003) and 0.84 (95% CI, 0.51-1.37; P = .48). After further adjustment for BMD values,
the estimates remained largely unchanged. For example, for women the OR for
any fractures remained 0.81 (95% CI, 0.71-0.93; P = .003)
after adjustment for lumbar spine BMD and 0.83 (95% CI, 0.72-0.95; P = .006) after adjustment for femoral neck BMD, whereas
after adjustment for lumbar spine BMD, the OR for vertebral fractures became
0.61 (95% CI, 0.45-0.82; P = .001). With
further adjustment for physical activity and ability indices in the 5 cohorts
with available data, the results remained similar. For example, for women
the OR for any fractures remained 0.78 (95% CI, 0.67-0.91), and the OR for
vertebral fractures was 0.72 (95% CI, 0.46-1.14). Other adjusted estimates
were also similar to the unadjusted results, and there was no significant
interaction between genotype and age or menopausal status (not shown).
In this multicenter study including individual-level information from
almost 20 000 individuals, we found that the ESR1 gene
exerts differential genetic effects on BMD and fracture risk. Effects on BMD
were either absent or of small magnitude, whereas there was a statistically
significant, 20% reduction in the odds of fractures and a possibly even larger
protective effect against vertebral fractures in XX homozygous
Although we did not standardize BMD measurements across all centers,
the method of data analysis was based on genotype-related differences in BMD
within each center to circumvent between-center differences in the type or
model of densitometer used. In any case, the observed 20% to 40% reduction
in the risk of fractures we observed would correspond to BMD differences of
0.030 to 0.080 g/cm2 in epidemiologic cohorts,19 which
should have been easily detectable, given the sample size of almost 20 000
individuals studied here. Nonetheless, clinical trials of osteoporosis treatments
have suggested that fracture risk reduction may be disproportionately large
compared with the corresponding changes in BMD.20,21 For
example, in the Fracture Intervention Trial, changes in spine BMD explained
only 16% of the reduction in the risk of vertebral fracture with alendronate.22 Our findings are consistent with the hypothesis that
the XbaI polymorphism influences fracture risk independent
of BMD, even though BMD would have been a plausible biological mediator of
the clinical effect for polymorphisms involved in the estrogen pathway. Possibilities
include effects on bone quality, bone geometry, bone turnover, or other nonskeletal
risk factors for fracture, such as decreased cognition or muscle strength.
These candidate mediators need to be better studied, and there is a rapidly
increasing literature on pleiotropic actions of ESR1 on
various outcomes.23-26 Whatever
the mechanism, the observed association has potential clinical relevance because
it indicates that genotyping for the ESR1 XbaI polymorphism
provides information on fracture risk that cannot be obtained by BMD measurements
From a methodologic point of view, our study had the advantage of using
individual participant data to allow consistent standardization of definitions,
measurements, and genetic contrasts. Sampling and systematic errors are a
threat to molecular studies.27 We ensured the
consistency and reliability of genotype results across the participating cohorts.
Eventually, the results from all the diverse cohorts included in our consortium
were similar, and there was no significant between-study heterogeneity detected
in any of the analyses of interest. Between-study heterogeneity is observed
in about half the cases in which different teams publish data on the same
putative gene-disease association.10,13 Sometimes
this heterogeneity may be due to technical differences and lack of standardization
across different centers rather than to genuine genetic diversity. Furthermore,
although our consortium design does not accommodate all previously published
data, these are limited3 compared with the
evidence that we generated. A meta-analysis in which genotyping is performed
prospectively is immune to the problems of publication bias28 because
all prospective, standardized genotyping results are eventually included in
the analysis and inclusion is not determined by the direction or strength
of the findings. Publication bias against studies that find no significant
association may be a problem in genetic association studies13 and
may be another reason for the occurrence of variability among the results
of studies published in the literature.
The XbaI and PvuII
polymorphic sites are located in the first intron of the ESR1 gene, and so far their functional consequences are unknown. However,
introns may contain regulatory elements. For example, the PvuII polymorphism is located within a potential bMyb binding site
with regulatory effects on a reporter gene.29 In
the absence of definitive evidence for the functionality of these ESR1 variants, more research is needed on the potential biological
pathways that they may affect. Alternatively, other polymorphic sites in strong
linkage disequilibrium with those that we studied may be functional variants
affecting receptor structure or, more likely, messenger RNA and protein expression.
A comprehensive analysis of the ESR1 gene might require
the genotyping of a large number of gene variants. However, it is impractical
to perform meta-analyses of such large scale on an extended number of unselected
polymorphisms; targets for obtaining large-scale genetic evidence should be
selected carefully according to preliminary smaller studies, as in this case.
Osteoporosis risk may also be modulated by a large number of genetic markers
beyond ESR1, including polymorphisms of the vitamin
D receptor (VDR) gene,30 the
collagen I α1(COLIA1) gene,31 and several other candidate genes.2 Although
the clinical impact of each implicated gene polymorphism is modest, the cumulative
effect may be large. Moreover, clarification of the role of these genetic
variants with large-scale evidence may give us important biological insights,
such as the extent to which effects on fractures diverge from BMD effects.
The current meta-analysis emphasizes the need for large-scale studies
to clarify postulated genetic determinants of osteoporosis and other complex
multigenetic diseases.12,13 Meta-analyses
of individual-level data in other fields have also suggested that plausible
genetic associations may be refuted with larger-scale evidence32,33 or
may be partially replicated.34 Quantifying
genetic risks for fractures and other osteoporosis outcomes will require adequately
powered studies, standardization, and relevant disease end points.
Corresponding Author: John P. A. Ioannidis,
MD, Department of Hygiene and Epidemiology, University of Ioannina School
of Medicine, Ioannina 45110, Greece (email@example.com).
Author Contributions: Dr Ioannidis had full
access to all the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Ioannidis, Ralston,
Bennett, Grinberg, Karassa, Langdahl, Nogues, Uitterlinden.
Acquisition of data: Ralston, Bennett, Brandi,
Grinberg, Langdahl, van Meurs, Mosekilde, Scollen, Albagha, Carey, Dunning,
Enjuanes, Mavilia, Masi, McGuigan, Nogues, Pols, Reid, Schuit, Sherlock, Uitterlinden.
Analysis and interpretation of data: Ioannidis,
Ralston, Bennett, Brandi, Grinberg, Karassa, Langdahl, van Meurs, Bustamante,
Dunning, van Leeuwen, Nogues, Pols, Schuit, Uitterlinden.
Drafting of the manuscript: Ioannidis, Ralston,
van Meurs, Schuit, Uitterlinden.
Critical revision of the manuscript for important
intellectual content: Ioannidis, Ralston, Bennett, Brandi, Grinberg,
Karassa, Langdahl, van Meurs, Mosekilde, Scollen, Albagha, Bustamante, Carey,
Dunning, Enjuanes, van Leeuwen, Mavilia, Masi, McGuigan, Nogues, Pols, Reid,
Statistical analysis: Ioannidis, Karassa, van
Meurs, Schuit, Uitterlinden.
Obtained funding: Ioannidis, Ralston, Bennett,
Brandi, Reid, Uitterlinden.
Administrative, technical, or material support:
Ioannidis, Ralston, Bennett, Grinberg, Langdahl, van Meurs, Scollen, Albagha,
Bustamante, Carey, Enjuanes, van Leeuwen, Mavilia, Masi, McGuigan, Nogues,
Pols, Schuit, Sherlock, Uitterlinden.
Study supervision: Ioannidis, Ralston, Bennett,
Brandi, Grinberg, van Meurs, Mosekilde, Albagha, Carey, Dunning, Nogues, Uitterlinden.
Additional Investigators Participating in GENOMOS:University of Ioannina School of Medicine, Ioannina,
Greece: Despina G. Contopoulos-Ioannidis and Thomas A. Trikalinos; Department of Internal Medicine, Erasmus MC, Rotterdam, the Netherlands: Pascal P. Arp and Wendy Hugens; Institute of Medical
Sciences, University of Aberdeen Medical School, Aberdeen, Scotland:
Amelia Bassitti, Helen MacDonald, Alison Stewart; Oxagen
Limited, Abingdon, England: Bryan Dechairo and Ian Mackay; FAMOS: Juliet Compston (University of Cambridge, Cambridge, England),
Cyrus Cooper (University of Southampton, Southampton, England), Emma Duncan
(Nuffield Orthopaedic Centre, Oxford, England), Richard Keen, (University
College, London, England), Alastair McLellan (University of Glasgow, Glasgow,
Scotland) and John Wass (Nuffield Orthopaedic Centre); Department of Internal Medicine, University of Florence Medical School, Florence,
Italy: Annalisa Tanini; Department of Genetics, Faculty
of Biology, Barcelona, Spain: Susana Balcells; Hospital
del Mar, Barcelona, Spain: Leonardo Mellibovsky and Adolfo Diez-Perez; DOPS study: Kim Brixen (Department of Endocrinology, Odense
University Hospital, Odense, Denmark), Stig P. Nielsen (Department of Physiology,
Hillerod Sygehus, Hillerod, Denmark), and Ole H. Sorensen (Department of Endocrinology,
HS Hvidovre Hospital, Hvidovre, Denmark); EPOS Study Group: A. Lopes Vaz, C. J. Todd, J. Nijs, S. Grazio, L. I. Benevolenskaya,
T. W. O’Neill, J. Reeve, S. K. Kaptoge and M. Lunt; Department of Medical Genetics, University of Antwerp, Antwerp, Belgium:
Wim van Hul.
Funding/Support: The GENOMOS Project receives
funding from the European Commission (grant QLRT-2001-02629).
Role of the Sponsor: The European Commission
had no role in the design and conduct of the study; collection, management,
analysis, and interpretation of the data; or in the preparation, review, and
approval of the manuscript.
Financial Disclosures: Dr Carey and Ms Sherlock
have received funding from Oxagen Ltd.
Nguyen TV, Blangero J, Eisman JA. Genetic epidemiological approaches to the search for osteoporosis genes. J Bone Miner Res
. 2000;15:392-40110750553Google ScholarCrossref
Ralston SH. Genetic control of susceptibility to osteoporosis. J Clin Endocrinol Metab
. 2002;87:2460-246612050200Google ScholarCrossref
Ioannidis JP, Stavrou I, Trikalinos TA.
et al. Association of polymorphisms of the estrogen receptor alpha gene with
bone mineral density and fracture risk in women: a meta-analysis. J Bone Miner Res
. 2002;17:2048-206012412813Google ScholarCrossref
Becherini L, Gennari L, Masi L.
et al. Evidence of a linkage disequilibrium between polymorphisms in the human
estrogen receptor alpha gene and their relationship to bone mass variation
in postmenopausal Italian women. Hum Mol Genet
. 2000;9:2043-205010942433Google ScholarCrossref
van Meurs JB, Schuit SC, Weel AE.
et al. Association of 5′ estrogen receptor alpha gene polymorphisms
with bone mineral density, vertebral bone area and fracture risk. Hum Mol Genet
. 2003;12:1745-175412837697Google ScholarCrossref
Sano M, Inoue S, Hosoi T.
et al. Association of estrogen receptor dinucleotide repeat polymorphism with
osteoporosis. Biochem Biophys Res Commun
. 1995;217:378-3838526937Google ScholarCrossref
Langdahl BL, Lokke E, Carstens M, Stenkjaer LL, Eriksen EFA. TA repeat polymorphism in the estrogen receptor gene is associated
with osteoporotic fractures but polymorphisms in the first exon and intron
are not. J Bone Miner Res
. 2000;15:2222-223011092403Google ScholarCrossref
Albagha OM, McGuigan FE, Reid DM, Ralston SH. Estrogen receptor alpha gene polymorphisms and bone mineral density:
haplotype analysis in women from the United Kingdom. J Bone Miner Res
. 2001;16:128-13411149476Google ScholarCrossref
Chen HY, Chen WC, Tsai HD, Hsu CD, Tsai FJ, Tsai CH. Relation of the estrogen receptor alpha gene microsatellite polymorphism
to bone mineral density and the susceptibility to osteoporosis in postmenopausal
Chinese women in Taiwan. Maturitas
. 2001;40:143-15011716992Google ScholarCrossref
Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of genetic association studies. Nat Genet
. 2001;29:306-30911600885Google ScholarCrossref
Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN. Meta-analysis of genetic association studies supports a contribution
of common variants to susceptibility to common disease. Nat Genet
. 2003;33:177-18212524541Google ScholarCrossref
Ioannidis JP, Trikalinos TA, Ntzani EE, Contopoulos-Ioannidis DG. Genetic associations in large versus small studies: an empirical assessment. Lancet
. 2003;361:567-57112598142Google ScholarCrossref
Ioannidis JP, Rosenberg PS, Goedert JJ, O'Brien TR. Commentary: meta-analysis of individual participants' data in
genetic epidemiology. Am J Epidemiol
. 2002;156:204-21012142254Google ScholarCrossref
Kalender WA, Felsenberg D, Genant HK, Fischer M, Dequeker J, Reeve J. The European Spine Phantom: a tool for standardization and quality
control in spinal bone mineral measurements by DXA and QCT. Eur J Radiol
. 1995;20:83-927588873Google ScholarCrossref
Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population
data. Am J Hum Genet
. 2001;68:978-98911254454Google ScholarCrossref
McCloskey EV, Spector TD, Eyres KS.
et al. The assessment of vertebral deformity: a method for use in population
studies and clinical trials. Osteoporos Int
. 1993;3:138-1478481590Google ScholarCrossref
Pettiti D. Meta-Analysis, Decision Analysis and Cost-Effectiveness
Analysis. New York, NY: Oxford University Press; 1999
Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict
occurrence of osteoporotic fractures. BMJ
. 1996;312:1254-12598634613Google ScholarCrossref
Torgerson DJ, Campbell MK, Thomas RE, Reid DM. Prediction of perimenopausal fractures by bone mineral density and
other risk factors. J Bone Miner Res
. 1996;11:293-2978822354Google ScholarCrossref
Cranney A, Wells G, Willan A.
et al. Meta-analyses of therapies for postmenopausal osteoporosis, II: meta-analysis
of alendronate for the treatment of postmenopausal women. Endocr Rev
. 2002;23:508-51612202465Google ScholarCrossref
Cummings SR, Karpf DB, Harris F.
et al. Improvement in spine bone density and reduction in risk of vertebral
fractures during treatment with antiresorptive drugs. Am J Med
. 2002;112:281-28911893367Google ScholarCrossref
Weel AE, Uitterlinden AG, Westendorp IC.
et al. Estrogen receptor polymorphism predicts the onset of natural and surgical
menopause. J Clin Endocrinol Metab
. 1999;84:3146-315010487678Google ScholarCrossref
Schuit SC, van Meurs JB, Bergink AP.
et al. Height in pre- and postmenopausal women is influenced by estrogen receptor
alpha gene polymorphisms. J Clin Endocrinol Metab
. 2004;89:303-30914715865Google ScholarCrossref
Bergink AP, van Meurs JB, Loughlin J.
et al. Estrogen receptor alpha gene haplotype is associated with radiographic
osteoarthritis of the knee in elderly men and women. Arthritis Rheum
. 2003;48:1913-192212847685Google ScholarCrossref
Schuit SC, Oei HH, Witteman JC.
et al. Estrogen receptor alpha gene polymorphisms and risk of myocardial infarction. JAMA
. 2004;291:2969-297715213208Google ScholarCrossref
Bogardus ST Jr, Concato J, Feinstein AR. Clinical epidemiological quality in molecular genetic research: the
need for methodological standards. JAMA
. 1999;281:1919-192610349896Google ScholarCrossref
Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR. Publication bias in clinical research. Lancet
. 1991;337:867-8721672966Google ScholarCrossref
Herrington DM, Howard TD, Brosnihan KB.
et al. Common estrogen receptor polymorphism augments effects of hormone replacement
therapy on E-selectin but not C-reactive protein. Circulation
. 2002;105:1879-188211997270Google ScholarCrossref
Cooper GS, Umbach DM. Are vitamin D receptor polymorphisms associated with bone mineral density?
a meta-analysis. J Bone Miner Res
. 1996;11:1841-18498970884Google ScholarCrossref
Efstathiadou Z, Tsatsoulis A, Ioannidis JP. Association of collagen Ialpha 1 Sp1 polymorphism with the risk of
prevalent fractures: a meta-analysis. J Bone Miner Res
. 2001;16:1586-159211547828Google ScholarCrossref
Keavney B, Parish S, Palmer A.
et al. Large-scale evidence that the cardiotoxicity of smoking is not significantly
modified by the apolipoprotein E epsilon2/epsilon3/epsilon4 genotype. Lancet
. 2003;361:396-39812573381Google ScholarCrossref
Keavney B, McKenzie C, Parish S.
et al. Large-scale test of hypothesised associations between the angiotensin-converting-enzyme
insertion/deletion polymorphism and myocardial infarction in about 5000 cases
and 6000 controls. International Studies of Infarct Survival (ISIS) Collaborators. Lancet
. 2000;355:434-44210841123Google Scholar
Ioannidis JP, Rosenberg PS, Goedert JJ.
et al. Effects of CCR5-Delta32, CCR2-64I, and SDF-1 3′A alleles on HIV-1
disease progression: an international meta-analysis of individual-patient
data. Ann Intern Med
. 2001;135:782-79511694103Google ScholarCrossref