Concentration & genetic regulation of sex hormone binding globulin and fracture risk in older women

Abstract

Objective:

To examine the association between concentrations of sex hormone binding globulin (SHBG) and fracture risk in community-dwelling older women and explore whether this was explained by the genetic regulation of SHBG.

Methods:

This prospective cohort study examined 4871 women aged ≥70 years, not taking medications influencing SHBG concentrations. A genome-wide association study was undertaken to identify single nucleotide polymorphisms (SNPs) associated with SHBG concentrations. Incident fracture was confirmed by medical imaging and adjudicated by expert review committee.

Results:

Median age of participants was 74.0 years. Over 3.9 (SD 1.4) years follow-up, 484 participants had an incident fracture. There was a linear trend for a positive association between SHBG concentrations and fracture risk (p=0.001), with the highest SHBG quartile associated with a significantly greater fracture risk compared with the lowest quartile (HR=1.54, 95% CI 1.16-2.04, p=0.003), adjusting for age, BMI, alcohol consumption, smoking, diabetes, impaired renal function, treatment allocation, medications affecting bone, and high-density lipoprotein-cholesterol. Two independent SNPs were associated with SHBG concentrations, rs10822163 and rs727428, neither were associated with fracture risk.

Conclusion:

SHBG concentrations were positively associated with a greater fracture risk in community-dwelling women aged ≥70 years, which was not explained by genetic variants associated with SHBG regulation.

Introduction

The societal burden of fragility fracture is escalating due to the exponential increases in both osteoporosis and falls in the aging population[1, 2]. Women aged ≥65 years account for 74% of all fractures that share 89% of total related costs[3]. Most fragility fractures occur in women with normal bone mineral density (BMD) or osteopenia, not osteoporosis[4, 5]. Therefore, identification of community-dwelling older women at greatest fracture risk, other than by BMD[6], has the potential to reduce fracture morbidity and related healthcare costs.

Sex hormone binding globulin (SHBG), a glycoprotein that binds and transports sex steroids in the circulation, has been positively associated with fracture risk in women aged >50 years[7]. SHBG is also positively associated with high-density lipoprotein (HDL)-cholesterol[8, 9] which in turn is inversely associated with BMD[10] and positively associated with fracture risk in older women[11]. Genome-wide association studies (GWAS) have reported genetic correlations and causal associations between SHBG and heel estimated, and forearm, BMD[12, 13], for adult men together with women of all ages[12], or of specific age groups[13]. An association between higher genetically predicted SHBG concentrations and a greater fracture risk in women aged 40-69 years has been reported, without consideration of the effects of menopause, age, exogenous steroid therapy or thyroid hormone replacement on SHBG[14]. For women aged ≥70 years the data are conflicting, with two cohort studies supporting a positive association between SHBG and fracture risk[15, 16], and two population-based cohort studies showed no association[17, 18].

The aim of the present study was to investigate the association between SHBG concentrations and fracture risk in a large sample of community-dwelling older women not using any medications known to influence SHBG concentrations, and to examine the role of genetic regulation of SHBG in this association.

Methods

Study design and participants

The Sex Hormones in Older Women (SHOW) Study[19] was a sub-study comprising 6392 (69.6%) of the 9179 female Australian participants aged ≥70 years in the ASPirin in Reducing Events in the Elderly (ASPREE) trial who provided biobank samples at enrolment[20]. Inclusion eligibility included no prior major cardiovascular event, dementia or a score of <78 on Modified Mini-Mental State Examination[21], physical disability (major difficulty with independently performing any one of six basic activities of daily living)[22], clinically significant anaemia, uncontrolled hypertension, or a life expectancy <5 years. The current study excluded participants taking medications that might influence SHBG concentrations, including sex hormones, anti-estrogens, anti-androgens, systemic glucocorticoids, thyroxine, and metformin (Supp Fig 1). The ASPREE trial was registered with the International Standard Randomized Controlled Trial Number Register (ISRCTN83772183) and clinicaltrials.gov (NCT01038583), and approved by the ethics committee at each participating centre. The SHOW Study was approved by the Alfred Hospital Ethics Committee (616/15) and the Monash University Human Research Ethics Committee (CF16/10 - 2016000001). All participants provided written informed consent.

Blood concentration of sex hormone binding globulin

Non-fasting blood samples were collected within 12 months of recruitment. SHBG was measured by automated immunoassay (Roche Diagnostics; North Ryde, NSW, Australia) in a single plasma sample at the ANZAC Research Institute (Sydney, NSW, Australia), with intra- and inter-assay coefficient of variation of 1.0-2.0% and 3.0-5.0%, respectively[19].

Genome-wide association analysis

The DNA samples provided by the ASPREE participants were genotyped using the Axiom 2.0 Precision Medicine Diversity Array (Thermo Fisher Scientific, CA), which includes >850,000 markers selected for high genomic coverage from the 1000 Genomes Project Phase III. Variant calling was performed using a custom pipeline aligned to the human reference genome GRCh38, and imputation was conducted using the TOPMed Imputation Server with the TOPMed-r2 reference panel[23, 24]. Quality control removed variants with low imputation quality scores (R²<0.3). Consistent with our previous study[25], we performed a GWAS for SHBG concentrations in 4951 unrelated women of European ancestry from the SHOW study who were not taking sex hormones, anti-estrogens, anti-androgens, or systemic glucocorticoids. For the GWAS, we excluded variants with missing genotype rate >0.05, a Hardy-Weinberg equilibrium P<10⁻⁶ (chi-square test), and minor allele frequency (MAF) <0.01. Consequently, 8,234,463 common variants (MAF >0.01) remained for analyses.

To correct for departures from normality in the original SHBG concentration values, a rank-based inverse normal transformation was applied. The transformed SHBG values were then regressed using Models 1 to 3 described below, with the standardised residuals entered as the outcome variable in the GWAS. Population stratification was controlled using the Principal Components (PCs) derived from the genotype data[26] as covariates in Model 1. The top 10 PCs used in the previous genetic study were applied[25]. Age was then added to Model 1 (Model 2), followed by body mass index (BMI) (Model 3). GWAS analyses were performed using an additive genetic linear regression (for effect allele, non-carriers=0, heterozygous carriers=1, homozygous carriers=2) for SHBG concentration using the PLINK2.0 software[27]. A genome-wide significance p-value threshold of 5×10⁻⁸ was applied. To determine if there were multiple independent signals at each identified locus, single-single nucleotide polymorphisms (SNP) association analyses, conditional on the identified lead SNPs, using the GCTA-COJO package were performed[28]. An additional association was declared when the conditional p-value was below the genome-wide significance threshold (5×10⁻⁸).

Assessment of fractures

Fracture after randomisation, both traumatic and pathological, was captured in the ASPREE-Fracture sub-study[29]. Participant-reported fractures were identified during the annual visit and 6-month telephone follow-up. Fracture confirmation was by hospital admission notes and discharge summaries, medical imaging reports (radiographic imaging, magnetic resonance imaging, computed tomography, and bone scans), emergency department notes, and death certificates. Only cases confirmed by medical imaging were considered as fracture, reviewed by an end point adjudication committee blinded to treatment allocation[29].

Covariates

Baseline BMI was calculated from weight (kg)/height (m)². Date of birth, smoking status, alcohol consumption, medical comorbidities, and prescription medications were documented[30]. Fasting glucose and lipid levels were measured by a local pathologist. Diabetes was defined as a fasting plasma glucose concentration of ≥7 mmol/L, current treatment of diabetes, or self-report. Impaired renal function was defined as an estimated glomerular filtration rate of <60 mL/min per 1.73 m².

Statistical analysis

SHBG concentrations, with skewed distributions, are expressed by median (range) and examined by quartiles. Cox proportional hazards regression was used to estimate the hazard ratios (HR) and 95% confidence intervals (CI) for the first recorded fractures during the ASPREE trial in relation to each quartile of the SHBG concentrations. The first recorded fractures of the skull, face, neck, hand, and foot were excluded as the outcome event (i.e. participants were censored at the date of these fractures) as these fractures are more likely to be associated with trauma, and not fragility fractures. Follow-up for fractures (i.e. calculation of person-time) began at the date of ASPREE randomisation, and ended at the date of first fracture or the date of censoring. Participants were censored at either the date of death, or end of follow-up in the trial (i.e. the earliest occurring of complete withdrawal, loss to follow-up, or the cessation of the intervention phase of the study), whichever came first. The linear trend for the association between SHBG concentrations (in quartiles) and fracture risk was examined using the likelihood ratio test with p-value for trend calculated. HR curve (spline curve, using the median SHBG concentration as the reference) was used to investigate whether there was evidence for a nonlinear association. Analyses were adjusted for age, BMI, alcohol consumption, smoking status, diabetes, impaired renal function, treatment allocation (aspirin/placebo), and use of available medications that would affect bone structure or mineralization (bisphosphonates, strontium ranelate, denosumab, raloxifene, teriparatide). Given the associations between SHBG and HDL-cholesterol[8, 9] and between HDL-cholesterol and fracture risk[11], additional adjustment for HDL-cholesterol was performed. The explained variation in fracture risk by SHBG concentrations was measured by the sum of squares approach[31].

The association between identified SNPs and fracture risk was examined using Cox proportional hazards regression, with Model 1 adjusted for the top 10 genetic PCs, Model 2 adjusted for the top 10 genetic PCs, age, and BMI, and Model 3 with additional adjustment for alcohol consumption, smoking status, diabetes, impaired renal function, treatment allocation, medication affecting bone structure or mineralization, and HDL-cholesterol.

The interactions between age group (<75 vs ≥75 years based on the median age) or obesity status (BMI <30 kg/m² vs BMI ≥30 kg/m²) and SHBG concentrations for their associations with fracture risk were tested using the likelihood ratio test. Sensitivity analysis was performed excluding participants who reported treatment with medications that affect bone structure or mineralization at study entry. We defined minimal trauma fracture as those caused by a fall from less than or equal to standing height, and examined the association with SHBG concentrations and identified SNPS in a sensitivity analysis. Tests based on Schoenfeld residuals and graphical methods using Kaplan-Meier curves showed no evidence that proportional hazard assumptions were violated. All statistical tests were two-sided and a p-value of <0.05 was considered statistically significant. All statistical analyses were performed using Stata MP version 17.0 (StataCorp LP, College Station, TX).

Results

Of the 6392 Australian women providing biobank samples for the SHOW Study, 6345 (99.3%) had SHBG measured. After excluding 821 participants taking therapies that might influence SHBG concentrations, 4781 participants were included in the analyses (Supp Fig 1). The participant characteristics at study entry are presented in Table 1. Participants were aged 70.0 to 94.8 years, with a median age of 74.0 years (IQR 71.7-77.6). Their mean BMI was 27.8 kg/m², with approximately 70% of the participants classified as overweight or obese. Over a mean 3.9 (SD 1.4) years of follow-up, 484 participants had an incident fracture.

Table 1.

Characteristics of study participants at baseline

	N=4781
Age, years	74.0 (IQR 71.7-77.6)
Body mass index, kg/m2	27.8 (5.0)
Body mass index categories, no (%)
<18.5 kg/m2	39 (0.8)
18.5-24.9 kg/m2	1424 (30.0)
25.0-29.9 kg/m2	1887 (39.7)
≥30.0 kg/m2	1404 (29.5)
Current/former smoker, no (%)	1624 (34.0)
Current/former alcohol user, no (%)	3780 (79.1)
Aspirin group, no (%)	2372 (49.6)
Diabetes, no (%)	205 (4.3)
Impaired renal function, no (%)	824 (17.7)
Use of medications that affect bone structure or mineralization, no (%)	582 (12.2)
High-density lipoprotein-cholesterol, mmol/L	1.7 (range 0.1-4.4)
Sex hormone binding globulin, nmol/L	41.8 (range 2.4-167.6)
Quartile 1, n=1205	25.3 (range 2.4-31.4)
Quartile 2, n=1194	36.6 (range 31.5-41.8)
Quartile 3, n=1191	47.6 (range 41.9-54.8)
Quartile 4, n=1191	65.6 (range 54.9-167.6)

Data are presented as median [interquartile range (IQR) / range], mean (SD), or no (%)

Having an SHBG concentration in the highest quartile was associated with a significantly greater fracture risk compared with the lowest quartile (HR 1.63, 95% CI 1.23-2.16, p=0.001), after adjusting for age, BMI, alcohol consumption, smoking, diabetes, impaired renal function, treatment allocation, and medication affecting bone structure or mineralization (Model 1) (Table 2). Furthermore, there was a linear trend for a positive association between SHBG concentrations and fracture risk (p<0.001). With additional adjustment for HDL-cholesterol (Model 2), the linear trend for the positive association between SHBG concentrations and fracture risk remained (p=0.001), as did a significantly greater fracture risk associated with the highest quartile SHBG concentrations compared with the lowest quartile (HR 1.54, 95% CI 1.16-2.04, p=0.003). A linear association between SHBG concentration and fracture risk was also evidenced (Supp Fig 2). In Model 2, SHBG concentrations explained 1% of the variation in fracture risk, as R² increased from 0.1389 to 0.1506 with adding SHBG to the Cox model.

Table 2.

Plasma concentrations of sex hormone binding globulin and risk of fracture

	No of participants with fracture / person-years	Univariable		Multivariable Model 1		Multivariable Model 2
		Hazard ratio (95% CI)	P-value	Hazard ratio (95% CI)	P-value	Hazard ratio (95% CI)	P-value
Fracture - total population (n=4781)
SHBG quartile 1	92/4785	1.00 (ref)		1.00 (ref)		1.00 (ref)
SHBG quartile 2	107/4715	1.17 (0.89, 1.55)	0.26	1.13 (0.85, 1.51)	0.39	1.10 (0.82, 1.46)	0.52
SHBG quartile 3	121/4681	1.33 (1.02, 1.75)	0.04	1.25 (0.94, 1.66)	0.12	1.18 (0.89, 1.57)	0.26
SHBG quartile 4	164/4565	1.86 (1.44, 2.40)	<0.001	1.63 (1.23, 2.16)	0.001	1.54 (1.16, 2.04)	0.003
P for trend			<0.001		<0.001		0.001
Fracture - excluding women on medication affecting bone structure or mineralization (n=4199)
SHBG quartile 1	77/4402	1.00 (ref)		1.00 (ref)		1.00 (ref)
SHBG quartile 2	90/4222	1.21 (0.90, 1.65)	0.21	1.16 (0.85, 1.58)	0.36	1.12 (0.82, 1.53)	0.47
SHBG quartile 3	107/4104	1.48 (1.10, 1.98)	0.01	1.38 (1.01, 1.87)	0.04	1.30 (0.95, 1.78)	0.10
SHBG quartile 4	128/3752	1.95 (1.47, 2.58)	<0.001	1.72 (1.26, 2.33)	0.001	1.63 (1.19, 2.23)	0.002
P for trend			<0.001		<0.001		0.001
Minimal trauma fracture (n=4781)
SHBG quartile 1	76/4785	1.00 (ref)		1.00 (ref)		1.00 (ref)
SHBG quartile 2	80/4715	1.06 (0.78, 1.45)	0.71	1.02 (0.74, 1.40)	0.91	0.99 (0.72, 1.36)	0.94
SHBG quartile 3	104/4681	1.39 (1.03, 1.86)	0.03	1.23 (0.90, 1.67)	0.34	1.14 (0.83, 1.57)	0.41
SHBG quartile 4	132/4565	1.81 (1.37, 2.40)	<0.001	1.44 (1.05, 1.96)	0.02	1.37 (0.998, 1.87)	0.052
P for trend			<0.001		0.01		0.04

CI: confidence interval; SHBG: sex hormone binding globulin

Model 1: Adjusted for age, body mass index, alcohol consumption, smoking status, diabetes, impaired renal function, treatment allocation (aspirin/placebo), and medication affecting bone structure or mineralization

Model 2: Model 1 + high-density lipoprotein cholesterol

There was no interaction between either age group or obesity status and SHBG concentrations for their associations with fracture risk. A sensitivity analysis was performed excluding participants who reported use of medications that affect bone structure or mineralization at study entry (Table 2). Among the 4199 participants analysed, 402 participants had a fracture during the trial. The linear trend for a positive association between SHBG concentrations and fracture risk remained significant in the fully adjusted model (p=0.001), and a significantly greater fracture risk with the highest quartile SHBG concentrations compared with the lowest quartile (HR 1.63, 95% CI 1.19-2.23, p=0.002). Limiting the analysis to the 392 participants with incident minimal trauma fracture (Table 2), the linear trend for a positive association between SHBG concentrations and fracture risk in the fully adjusted model was significant (p=0.04).

The GWAS identified two independent loci associated with SHBG concentrations, below the genome-wide significance threshold (p<5×10⁻⁸) after adjusting for the top 10 genetic PCs, age and BMI (Supp Fig 3). These two lead SNPs mapped to known genes: rs10822163 (JMJD1C gene on 10q21.3, a candidate histone demethylase and potential coactivator for key transcription factors) and rs727428 (SHBG gene on 17p13.1, encoding a steroid-binding protein involved in the regulation of steroid responses) (Supp Table 1). No secondary signals were found at any loci (all p>5×10⁻⁸ when conditioned on the identified lead SNPs by a fully adjusted additive genetic linear regression), indicating that these lead SNPs represented the associations in their respective loci.

Of the 4951 unrelated women of European ancestry in the GWAS, 4285 participants were involved in the current incident fracture study. Among them, 429 had an incident fracture, and 372 had an incident minimal trauma fracture. Participants with the effect allele (G) in rs10822163 had significantly higher SHBG concentrations (p<0.0001), and participants with the effect allele (T) in rs727428 had significantly lower SHBG concentrations (p<0.0001), compared with those without the respective effect allele (Supp Table 2). However, neither SNPs were significantly associated with the risk of fracture, or minimal trauma fracture (Table 3).

Table 3.

Association between two lead genome-wide significant SNPs and risk of fracture (n=4285)

SNP ID	Genotype	No of participants with fracture / person-years	Multivariable Model 1		Multivariable Model 2		Multivariable Model 3
			Hazard ratio (95% CI)	P-value	Hazard ratio (95% CI)	P-value	Hazard ratio (95% CI)	P-value
Fracture
rs10822163	CC	122/4614	1.00 (ref)		1.00 (ref)		1.00 (ref)
	GaC	213/8646	0.92 (0.74, 1.15)	0.47	0.88 (0.70, 1.10)	0.25	0.86 (0.69, 1.08)	0.20
	GaGa	94/3636	0.96 (0.73, 1.25)	0.74	0.93 (0.71, 1.22)	0.61	0.96 (0.73, 1.26)	0.77
rs727428	CC	142/5118	1.00 (ref)		1.00 (ref)		1.00 (ref)
	TaC	205/8243	0.89 (0.72, 1.11)	0.30	0.93 (0.75, 1.15)	0.49	0.96 (0.77, 1.19)	0.69
	TaTa	82/3534	0.82 (0.63, 1.08)	0.17	0.84 (0.64, 1.10)	0.21	0.85 (0.65, 1.13)	0.27
Minimal trauma fracture
rs10822163	CC	117/4623	1.00 (ref)		1.00 (ref)		1.00 (ref)
	GaC	170/8655	0.77 (0.61, 0.98)	0.03	0.73 (0.57, 0.93)	0.01	0.73 (0.57, 0.93)	0.01
	GaGa	85/3639	0.91 (0.69, 1.20)	0.51	0.88 (0.67, 1.17)	0.38	0.92 (0.69, 1.22)	0.56
rs727428	CC	112/5127	1.00 (ref)		1.00 (ref)		1.00 (ref)
	TaC	186/8252	1.02 (0.81, 1.29)	0.85	1.06 (0.84, 1.34)	0.63	1.09 (0.85, 1.38)	0.50
	TaTa	74/3539	0.94 (0.70, 1.26)	0.68	0.95 (0.71, 1.27)	0.73	0.96 (0.71, 1.29)	0.77

SNP: single nucleotide polymorphism; CI: confidence interval; ^aDenotes the effect allele

Model 1: Adjusted for the top 10 genetic PCs; Model 2: Model 1 + age and body mass index; Model 3: Model 2 + alcohol consumption, smoking, diabetes, impaired renal function, treatment allocation (aspirin/placebo), medication affecting bone structure or mineralization, and high-density lipoprotein cholesterol

Discussion

This study reaffirms a statistically significant positive, but modest, independent association between SHBG concentrations and fracture risk in community-dwelling older women. The association was not explained by the common genetic variants associated with SHBG concentrations. Importantly, the proportion of the variation in fracture risk explained by SHBG was small, in keeping with SHBG not being a clinically useful indicator of fracture risk in older women.

We observed a linear trend for a positive association between SHBG concentrations and fracture risk, with consistent results for incident fractures excluding fractures of the skull, face, neck, hand, and foot, as well as incident minimal trauma fracture. These associations were independent of variables known to modulate fracture risk. Our finding aligns with a systematic review and meta-analysis of women aged >50 years[7]. However, evidence from cohort studies is limited and conflicting for older women. Among the 9 studies examining women included in this systematic review[7], four studies examined older women[15, 16, 32, 33] with only one prospective cohort study[16]. In the latter prospective study of 348 women aged ≥70 years, the highest SHBG quartile was associated with a greater risk of nonvertebral fractures and was a predictor strongly associated with osteoporotic fractures compared with all other participants[16]. Another two population-based prospective cohort studies showed no association of SHBG with atraumatic incident clinical fractures in 1499 women aged >70 years[17] or fall-related limb fractures in 187 women aged 75 years[18]. The discrepancies in study findings might be due to the different definition of fractures and varied follow-up duration for incident fractures.

The independent SNPs associated with SHBG concentrations identified in our study have been reported in GWAS studies of SHBG in different cohorts: rs10822163[34, 35] and rs727428[35, 36]. Neither SNPs were associated with incident fracture in our study, whereas a genetic association between SHBG and BMD, and fracture risk has been reported with different SNPs identified as associated with SHBG concentrations[12-14]. These findings are difficult to interpret, as these prior studies did not take into account menopause status, or the effects of common medications that alter SHBG concentrations, including sex steroid therapy (notably oral estrogen in contraceptive pills and menopausal hormone therapy), thyroxine and glucocorticoids[12-14], and two studies combined men and women in their analysis[12, 13].

The link between SHBG and fracture risk remains unclear. SHBG is an important transporter for estradiol, estrone and testosterone, and it has been hypothesized that high SHBG concentrations might bind these sex steroids and reduce their bioavailability for target tissues, such as bone[37], leading to decreased BMD[12, 13] and increased fracture risk. However, evidence that the unbound fractions of estradiol and testosterone are the most bioactive is lacking[38]. Furthermore, after menopause estrogen and testosterone are synthesised in bone from dehydroepiandrosterone (DHEA) and androstenedione[39]. SHBG only weakly binds DHEA and androstenedione, and does not bind DHEA sulfate, the circulating reservoir for DHEA[40]. Therefore, high concentrations of SHBG do not limit the supply of the estrogen and testosterone precursors to bone cells. Consequently, the SHBG blood concentration may not be a meaningful determinant of bone tissue estradiol or testosterone biosynthesis in postmenopausal women.

Striking however is the phenotype of women with high SHBG concentrations. Such women are more likely to have high HDL-cholesterol and low to normal BMI, both of which are independently associated with a greater likelihood of fragility fracture in older women[11, 41]. SHBG and HDL-cholesterol are inversely associated with leptin[42]. Leptin inhibits bone resorption and is positively associated with BMD[43]. Women with constitutional low body weight and hypothalamic amenorrhoea also have lower leptin concentrations[44]. Hence, the associations between high SHBG and high HDL-cholesterol, low BMI, and high fracture risk may be explained by the shared relationships between these variables and leptin.

Our study has several strengths. It included around 5000 healthy community-dwelling women aged 70 to 95 years, free of chronic disability at enrolment, and not taking a comprehensive list of medications that might influence blood SHBG concentrations. Our analyses adjusted for known potential confounders including age, BMI, alcohol consumption, smoking, comorbidities, medications affecting bone structure or mineralization, and HDL-cholesterol. The sensitivity analyses excluding women using medications affecting bone at baseline and analysis of incident minimal trauma fracture strengthen the results. There are limitations to our study. Being a prospective cohort study, causality could not be determined, and we cannot exclude possible residual confounding. BMD and bone turnover markers were not measured. Hence, we could not examine the association between SHBG and these variables, or their influence on our results. Fractures were confirmed by imaging and expert adjudication. However, fractures were participant reported and some incident fractures, such as vertebral fractures, may have not been imaged, and hence not captured in our study. Minimal trauma fractures may have been misclassified. However, these misclassifications of fractures would be more likely to be non-differential misclassification that would cause an underestimation of the magnitude of the associations observed. Our study included relatively healthy older volunteers in a long-term clinical trial who may have had a greater interest in their health, limiting the generalisability of our findings. As the participants were primarily of European ancestry, our findings may not be generalisable to other populations. GWAS, as a technique, does not capture all possible types of genetic regulation that may contribute to SHBG regulation (for example, rare genetic variants or structural variants).

In conclusion, the positive association between SHBG concentrations and fracture risk in community-dwelling women aged ≥70 years was not explained by two genetic variants associated with SHBG regulation. Furthermore, the small contribution of SHBG to fracture risk indicates SHBG offers no likely clinical utility as a biomarker for identifying older women most at risk of fracture.

Data availability statement

The data generated from this study will not be deposited in a public repository due to privacy and consent restrictions. Deidentified, coded data can be made available from the corresponding author upon reasonable request, subject to a data sharing agreement.