CKD is a major contributor to morbidity and mortality globally, affecting 10% of the population worldwide,1 with differences between men and women.2 Specifically, kidney function declines faster in men than in women in diabetic and nondiabetic CKD,3 and in nondiabetic kidney disease of different etiologies (mixed etiology, IgA nephropathy, idiopathic membranous nephropathy, and autosomal dominant polycystic kidney disease).4 The all-cause mortality rate is higher in men than women at all levels of eGFR.2 The sex difference in CKD progression might be due to differences in lifestyle (such as smoking and dietary intake), kidney structure, or sex hormones.5 The percentage of the sex difference explained by lifestyle is unclear; however, sex hormones rather than renal structure and function could be relevant to the sex disparity because “manipulation of the hormonal environment by gonadectomy or administration by gonadectomy or administration of sex hormones reproduces these effects on renal disease progression.”5 Androgens may play a key role in this sex disparity, by directly and indirectly affecting various cellular processes, through modulating the synthesis of various cytokines, growth factors, and vasoactive agents.3 Genetically predicted testosterone is related to a higher risk of CKD and worse kidney function in men, but not women.6 Sex hormone binding globulin (SHBG), which binds to sex hormones, regulates sex hormone levels and activity,7 specifically lowering sex hormone bioactivity,7 and thereby could play a role in CKD. Observationally, higher SHBG is associated with better kidney function, measured by higher eGFR, in men but not in women,8 but the associations are difficult to verify in randomized controlled trials. Given the close association of SHBG with sex hormones, it is also difficult to disentangle its role from the role of sex hormones in observational studies. Moreover, no study has specifically examined the sex-specific role of SHBG in CKD and kidney function.
In this situation where experimental evidence is unavailable, a Mendelian randomization (MR) study design, which uses naturally occurring SHBG-related genetic variants resulting in life-long differences in endogenous exposures, provides a way forward. As the genetic variants, specifically single nucleotide polymorphisms (SNPs), are determined at conception, MR can minimize confounding by socioeconomic position or other confounders, and obtain estimates without any potentially harmful interventions.9 MR has been used to assess the role of SHBG in diabetes,10–12 asthma,13 and breast cancer.14 Here, we used a sex-specific MR study to examine the role of genetically predicted SHBG in CKD and kidney function (indicated by albuminuria and eGFR), by applying published genome-wide significant SNPs predicting SHBG in men and women15 to CKD and kidney function in the United Kingdom (UK) Biobank, by far the largest cohort in Europeans.16 We also assessed its role in BP, a key risk factor in CKD. To understand the role of SHBG independent of testosterone in men, we used multivariable MR controlling for testosterone in men.
METHODS
Study Design
To assess the sex-specific associations of SHBG with CKD and kidney function, we used published genome-wide significant SNPs predicting SHBG in men and women from the latest and largest genome-wide association study (GWAS) of SHBG15 applied to CKD and kidney function in the UK Biobank. We also replicated the analysis using genetic variants for SHBG from the SHBG gene identified in a previous GWAS.17 We assessed the sex-specific association of genetically predicted SHBG with systolic BP (SBP) and diastolic BP (DBP), given that BP is a key risk factor for CKD. To assess the role of SHBG independent of testosterone, the main sex hormone in men, we used multivariable MR controlling for testosterone, using genetic instruments for testosterone from the largest published GWAS of testosterone applied to individual level measures of kidney function in the UK Biobank. Estradiol is the main sex hormone in women; however, published genetic instruments for estradiol in women are not available in the largest sex-specific GWAS in the UK Biobank,15 so the role independent of estradiol in women was not assessed.
Genetic Predictors for SHBG Used in Univariable MR
MR estimates were derived from Wald estimates, namely, genetic association with outcome (here CKD and kidney function) divided by genetic association with the exposure (here SHBG). The genetic variants for SHBG were obtained from the largest sex-specific GWAS conducted in the UK Biobank (178,782 white British men and 230,454 white British women). The GWAS associations were also replicated in three independent studies (Cohorts for Heart and Aging Research in Genomic Epidemiology Consortium, Twins UK, and the European Prospective Investigation into Cancer and Nutrition-Norfolk).15 There were 357 genome-wide significant (P<5×10−8), uncorrelated (r2<0.05) SNPs for SHBG in men and 359 uncorrelated SNPs for SHBG in women, with minor allele frequency >0.1%, and an imputation quality score >0.5.15 The genetic variants and their associations with SHBG are shown in Supplemental Tables 1 and 2. To check for potential pleiotropy, that is, the genetic variants affecting CKD or kidney function via another factor that is not on the causal pathway from SHBG to CKD, we also checked for their association with factors affecting CKD or possibly affecting CKD, but which are not a downstream factor of SHBG, such as potential confounders of SHBG on CKD. As the causes of CKD are unclear, we selected factors known to play a role in major chronic diseases, including Townsend index, smoking, alcohol drinking, and physical activity,18–20 and checked for genetic associations using the UK Biobank summary statistics at genome-wide significance (P<5×10−8). Body mass index (BMI) might be a driver21 or a downstream factor of SHBG,15 which means it is a potential confounder (the former) or mediator (the latter), so we also checked for its associations with BMI. We included all SNPs in the main analysis, and repeated the analysis without the SNPs related to potential confounders in a sensitivity analysis.
In the sensitivity analysis, we used another set of published genetic variants from another study, a GWAS meta-analysis of 21,791 participants (12,401 men, 9390 women) from ten studies, and validated in 7046 participants (2537 men, 4509 women) in another six studies.17 As previously,13 we used three SNPs in the SHBG gene. As the GWAS was not conducted in men and women separately, for ease of comparison, we obtained the sex-specific association of each SNP with SHBG in the UK Biobank, using summary statistics provided by the Neale Lab (http://www.nealelab.is/uk-biobank/), adjusted for age, age2, and 20 principal components.
Genetic Predictors for Multivariable MR
Genetic predictors for SHBG in men were closely related to total testosterone,15 so we used multivariable MR controlling for testosterone. Specifically, we included genetic predictors for SHBG (357 SNPs as in univariable MR) and total testosterone. The genetic predictors for total testosterone were from the GWAS in the UK Biobank, specifically, 231 SNPs in men,15 as shown in Supplemental Table 3. After combining the genetic predictors for SHBG and testosterone, we dropped duplicate SNPs and checked for the correlation of the remaining SNPs in MR-Base and LDlink (https://ldlink.nci.nih.gov/). We dropped correlated SNPs (r2>0.05) and SNPs whose correlations with other SNPs were not available. The remaining SNPs were used for multivariable MR.
Genetic Associations with CKD and Kidney Function
Genetic associations with CKD and kidney function were obtained from individual-level data in the UK Biobank (under application 42468). The UK Biobank is a large, ongoing, prospective cohort study, with median follow up time of 11.1 years.16 The UK Biobank recruited 502,713 people (aged 40–69 years, mean age 56.5 years, 45.6% men) from 2006 to 2010 from England, Scotland, and Wales, 94% of self-reported European ancestry. As previously,6 to control for population stratification, we restricted our analysis to participants of white British ancestry, and for quality control we further excluded participants with inconsistent self-reported and genotyped sex; excess relatedness (more than ten putative third-degree relatives); abnormal sex chromosomes (such as XXY); or poor-quality genotyping (heterozygosity or missing rate >1.5%).
CKD events included fatal and nonfatal CKD, that is, mortality and hospitalization due to chronic renal failure and self-reported chronic renal failure (International Classification of Diseases, Tenth revision [ICD-10] code: N18; ICD-9 code: 5859; and self-report code: 1192), indications of RRT (ICD-10 code Z99.2 [dependence on renal dialysis], Z94.0 [kidney transplant status], and Z49 [encounter for care involving renal dialysis]). Information was obtained from a nurse-led interview at recruitment, ongoing follow-ups via record linkage to inpatient admission and death registration,16 and eGFR, calculated at baseline on the basis of the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula using serum creatinine,22 <60 ml/min per 1.73 m2. We identified 6016 CKD patients in 179,916 white British men, and 5958 CKD patients in 212,079 white British women updated to December 2019.
Secondary outcomes included albuminuria, eGFR, and BP. Albuminuria was identified on the basis of urine albumin-to-creatinine ratio >30 mg/g23; participants with microalbumin lower than the detection limit were considered to be free from albuminuria. We identified 8845 men and 8275 women with albuminuria. eGFR was calculated on the basis of the CKD-EPI formula using serum creatinine (eGFR_cr)22 and using both creatinine and cystatin C (eGFR_crcys).24
Logistic regression was used to obtain the association of each SNP with CKD and albuminuria, linear regression was used for eGFR. We controlled for age, 20 principal components, assay array (the UK BiLEVE array and UK Biobank Axiom array), smoking, BMI, SBP, and DBP. Smoking, BMI, and BP were controlled for because they are risk factors in common for various chronic diseases, such as CKD and cardiovascular disease (CVD),25,26 controlling for common risk factors may partly address selection bias from competing risk from other diseases before recruitment.26,27 Sex-specific genetic associations with SBP and DBP were obtained from the Neale Lab (http://www.nealelab.is/uk-biobank/), adjusted for age, age2, and 20 principal components.
Statistical Analysis for MR Estimates Using Univariable MR
After aligning the SNPs on the basis of allele letter related to higher SHBG, we meta-analyzed the Wald estimates28,29 (i.e., the genetic association with CKD or kidney function divided by the genetic association with SHBG), to obtain the overall association of genetically predicted higher SHBG with CKD and kidney function. Similarly, we assessed the associations of genetically predicted higher SHBG with BP. In the sensitivity analysis, we also aligned the SNPs on the basis of allele letter related to lower SHBG and examined the association of genetically predicted lower SHBG with CKD and kidney function. When there are no outliers, we used inverse variance weighting (IVW) with multiplicative random effects, which is applicable to both individual-level and summary data, and provides equivalent estimates to two-stage least squares.30,31 Specifically, we obtained genetic associations with SHBG from a published GWAS in the UK Biobank,15 and obtained genetic associations with CKD and kidney function from individual-level information from the UK Biobank. The strength of these genetic instruments was assessed from the F statistic, calculated using the square of SNP-exposure associations divided by the square of its SEM.32 To examine whether the MR estimates were different in men and women, we calculated the P value for the differences in sex-specific estimates (log ORs for CKD and albuminuria, and beta coefficients for eGFR), where we used a well-established formula to calculate the z statistic (Supplemental Text), and then obtained the two-tailed P value.33
MR rests on three assumptions, namely, the genetic variants are strongly related to the exposure, are not related to confounders of the exposure-outcome relation, and only influence the outcome via the exposure.34 The IVW estimate is valid when all of the genetic variants meet these assumptions.31 As we used hundreds of genetic variants predicting SHBG, and possibly not all of the genetic variants meet these assumptions, in the sensitivity analysis, we used other statistical methods robust to pleiotropy, specifically, MR pleiotropy residual sum and outlier (MR-PRESSO), the contamination mixture method, and a weighted median. MR-PRESSO is able to identify outliers with potential horizontal pleiotropy when using multiple genetic variants as an instrument, and provides a corrected estimate after removing these outliers.35 The contamination mixture method is a newly developed method for dealing with hundreds of genetic variants in MR,31 as in this study. It assumes the true causal effect is the value taken by the largest number of genetic variants, so it is robust to outliers,31 but provides a more conservative estimates than MR-PRESSO.36 As such, if outliers were detected, we used the estimates from the contamination mixture method in the main analysis. The weighted median estimate is robust to invalid instruments, and able to provide consistent estimation even when up to 50% of the weight is from invalid SNPs.37
For the power calculation, we used the approximation that the sample size required for MR is the sample size needed in the conventional observational study divided by R2 (the variance in SHBG explained by the SNPs).38 The required sample size for CKD events and microalbuminuria were calculated on the basis of the log OR, the ratio of patients to nonpatients, and R2,39 the required sample size for eGFR was calculated on the basis of the effect size and R2.38
Statistical Analysis for MR Estimates Using Multivariable MR
Multivariable MR extends conventional univariable MR. Multivariable IVW has been developed to take two or more exposures into account, by taking advantage of pleiotropic SNPs predicting two or more exposures, and can be used to estimate the effect of one exposure after controlling for the other exposure(s),40 here the role of SHBG adjusted for testosterone. We used a weighted regression-based method, which gives equivalent estimates to two-stage least squares,41 by compiling all genetic variants predicting SHBG and testosterone, removing the duplicate and correlated (r2>0.05) genetic variants, and obtaining their associations with SHBG, testosterone, and the outcome (CKD or kidney function). The analysis was conducted using the “mr_mvivw” command of the “MendelianRandomization” package. Similar assumptions of univariable MR also apply to multivariable MR, specifically, the genetic variants are related to SHBG and/or testosterone, not related to the confounders in the association of SHBG and testosterone with CKD or kidney function, and not associated with CKD or kidney function other than via SHBG and/or testosterone.40 To assess instrument strength, we calculated the conditional F-statistic, namely, the Sanderson-Windmeijer F-statistic.42 To assess potential pleiotropy, we estimate modified Cochran’s Q statistic.42 Given multivariable IVW estimates might be biased when their assumptions are violated, we also used multivariable MR-Egger, which can detect whether the genetic predictors were acting other than via SHBG or testosterone (directional pleiotropy) indicated by a nonzero intercept and provides a corrected estimate when there is an indication of pleiotropy.43 Given SHBG is the exposure of interest, we orientated the multivariable MR-Egger on SHBG.
All statistical analyses were conducted using R version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria), the “clumping” function of MR-Base and the R packages “MendelianRandomization” and “MVMR.”
Ethics Approval
This research has been conducted using the UK Biobank resource under application number 42,468. No original data were collected for the MR study. Ethics approval for each of the studies included in the investigation can be found in the original publications (including informed consent from each participant). The UK Biobank has already received ethical approval from the Research Ethics Committee and participants provided written informed consent. The analysis of other publicly available summary statistics does not require additional ethics approval.
Results
Genetic Predictors
In univariable MR, we used the previously published 357 genome-wide significant SNPs for SHBG in men, and 359 SNPs for SHBG in women in the main analysis. All SNPs had F statistic >10, with average F statistic 183.4 in men, and 124.1 in women. The genetic variants explain 3%–3.5% variance in SHBG, results of power calculations are shown in Supplemental Table 4. One SNP (rs1260326) was related to alcohol drinking in men and women; six SNPs were related to BMI in men, and four SNPs were related to BMI in women. To address any potential pleiotropy, we conducted analysis including and excluding these SNPs. We repeated the analysis using three SNPs (rs12150660, rs1625895, and rs1641537) in SHBG.
The Role of SHBG in CKD and Kidney Function
Using univariable MR, genetically predicted higher SHBG was associated with a lower risk of CKD and higher eGFR in men (Table 1). The genetic associations of each genetic variant (effect allele refers to the allele increasing SHBG) with CKD and kidney function are shown in Supplemental Tables 5 and 6. Genetically predicted lower SHBG was associated with a higher risk of CKD and lower eGFR in men (Supplemental Table 7). Genetically predicted SHBG was not related to CKD or kidney function in women (Table 1). The difference by sex was significant for CKD (P value for sex difference 0.01) and for eGFR (P values for sex difference 0.01 for eGFR_cr and 0.0003 for eGFR_crcys). Using the three SNPs in the SHBG gene showed similar patterns of association (Table 2). The associations remained after excluding SNPs related to alcohol drinking (Supplemental Table 8) and BMI (Supplemental Table 9), and were robust to using different methods (Supplemental Table 10). Genetically predicted higher SHBG was associated with lower SBP in men and women, and with lower DBP in women using MR-PRESSO, but was not robust to other analytic methods (Table 3).
Table 1. -
Associations of genetically predicted SHBG with CKD and kidney function in the UK Biobank using univariable MR in men and women
| Outcomes |
#SNPs |
OR (Per SD Increase in Genetically Predicted SHBG) |
95% CI |
P Value |
| CKD |
|
|
|
| Men |
357 |
0.78 |
0.65 to 0.93 |
0.01 |
| Women |
359 |
1.11 |
0.92 to 1.33 |
0.24 |
| Albuminuria |
|
|
|
| Men |
357 |
0.97 |
0.78 to 1.19 |
0.75 |
| Women |
359 |
0.73 |
0.61 to 1.12 |
0.32 |
|
|
Beta-Coefficient (per SD Increase in Genetically Predicted SHBG) |
|
|
| eGFR_cr (ml/min per 1.73 m2) |
| Men |
357 |
0.58 |
0.15 to 1.02 |
0.01 |
| Women |
359 |
−0.18 |
−0.56 to 0.20 |
0.28 |
| eGFR_crcys (ml/min per 1.73 m2) |
| Men |
357 |
1.17 |
0.50 to 1.84 |
0.001 |
| Women |
359 |
−0.11 |
−0.30 to 0.09 |
0.07 |
The contamination mixture method was used in the analysis. OR, odds ratio; 95% CI, 95% confidence interval.
Table 2. -
Associations of genetically predicted SHBG with CKD and kidney function in the UK Biobank using another set of genetic variants in
SHBG gene in men and women using different analytic methods
| Outcomes |
#SNPs |
OR (Per SD Increase in Genetically Predicted SHBG) |
95% CI |
P Value |
| CKD |
|
|
|
|
| Men |
|
|
|
|
| IVW |
3 |
0.90 |
0.79 to 1.02 |
0.09 |
| WM |
|
0.87 |
0.77 to 0.99 |
0.04 |
| Women |
|
|
|
|
| IVW |
3 |
1.04 |
0.88 to 1.24 |
0.62 |
| WM |
|
1.03 |
0.86 to 1.24 |
0.72 |
| Albuminuria |
|
|
|
|
| Men |
|
|
|
|
| IVW |
3 |
0.97 |
0.88 to 1.08 |
0.62 |
| WM |
|
0.98 |
0.88 to 1.09 |
0.70 |
| Women |
|
|
|
|
| IVW |
3 |
1.06 |
0.91 to 1.24 |
0.43 |
| WM |
|
1.06 |
0.90 to 1.25 |
0.49 |
|
|
Beta-Coefficient (per SD Increase in Genetically Predicted SHBG) |
|
|
| eGFR_cr (ml/min per 1.73 m2) |
|
|
|
|
| Men |
|
|
|
|
| IVW |
3 |
0.09 |
−0.16 to 0.34 |
0.47 |
| WM |
|
0.16 |
−0.10 to 0.41 |
0.24 |
| Women |
|
|
|
|
| IVW |
3 |
0.004 |
−0.32 to 0.32 |
0.98 |
| WM |
|
0.11 |
−0.23 to 0.45 |
0.53 |
| eGFR_crcys (ml/min per 1.73 m2) |
|
|
|
|
| Men |
|
|
|
|
| IVW |
3 |
0.30 |
0.04 to 0.56 |
0.03 |
| WM |
|
0.38 |
0.11 to 0.66 |
0.01 |
| Women |
|
|
|
|
| IVW |
3 |
0.16 |
−0.17 to 0.48 |
0.36 |
| WM |
|
0.29 |
−0.06 to 0.64 |
0.10 |
OR, odds ratio; 95% CI, 95% confidence interval; WM, weighted median.
Table 3. -
Associations of genetically predicted SHBG with BP in the UK Biobank in men and women using different analytic methods
| Outcomes |
#SNPs
a
|
Beta (Per SD Increase in Genetically Predicted SHBG) |
95% CI |
P Value |
| Systolic BP |
| Men |
356 |
−0.03 |
−0.06 to 0.02 |
0.14 |
| Conmix model |
|
−0.003 |
−0.06 to 0.06 |
0.93 |
| Weighted median |
|
−0.09 |
−0.13 to −0.05 |
0.00003 |
| MR-PRESSO |
356 |
−0.03 |
−0.06 to 0.02 |
0.14 |
| Women |
|
|
|
|
| Conmix model |
355 |
−0.19 |
−0.24 to −0.14 |
0.0005 |
| Weighted median |
|
−0.007 |
−0.06 to 0.04 |
0.78 |
| MR-PRESSO |
|
−0.19 |
−0.23 to −0.14 |
8.7×10−15
|
| Diastolic BP |
| Men |
|
|
|
|
| Conmix model |
356 |
0.02 |
−0.01 to 0.05 |
0.29 |
| Weighted median |
|
0.01 |
−0.05 to 0.07 |
0.82 |
| MR-PRESSO |
|
−0.04 |
−0.08 to 0.002 |
0.06 |
| Women |
|
|
|
|
| Conmix model |
357 |
0.01 |
−0.03 to 0.04 |
0.59 |
| Weighted median |
|
0.03 |
−0.02 to 0.08 |
0.28 |
| MR-PRESSO |
|
−0.12 |
−0.17 to −0.07 |
2.1×10−6
|
Conmix model, contamination mixture model. 95% CI, 95% confidence interval.
aGenetic variants that might be directly associated with BP (P<5×10−8) were dropped, specifically one SNP (rs11419346) related to systolic and diastolic BP was dropped in men, four SNPs (rs1047891, rs198358, rs548235873, and rs72681869) related to systolic BP in women, and two SNPs (rs198358 and rs12787706) related to diastolic BP in women.
In multivariable MR in men, in addition to the 357 SNPs for SHBG in men, we included 231 SNPs for total testosterone. After excluding 39 duplicate SNPs, 87 correlated SNPs, and two SNPs with unclear correlation information, 460 SNPs remained and were used. The protective associations of SHBG with CKD and eGFR remained after controlling for testosterone in multivariable MR (Table 4). Confidence intervals for the association of genetically predicted SHBG with eGFR_crcys included the null value, possibly due to larger variation in the genetic association with eGFR_crcys than with eGFR_cr, but the direction of association was consistent with that for eGFR_cr (Table 4). Genetically predicted higher SHBG was also related to lower risk of albuminuria in multivariable MR (Table 4). The conditional F statistics for SHBG and testosterone were 14.6 and 17.1, respectively, indicating no weak instrument. The testing of Cochran’s Q statistic rejected the null hypothesis of no heterogeneity. The estimates were robust to using multivariable MR-Egger (Supplemental Table 11).
Table 4. -
Associations of genetically predicted SHBG with CKD and kidney function in the UK Biobank using multivariable MR in men
| Outcomes |
#SNPs |
OR (Per SD Increase in Genetically Predicted SHBG) |
95% CI |
P Value |
| CKD |
460 |
0.61 |
0.45 to 0.82 |
0.001 |
| Albuminuria |
460 |
0.69 |
0.51 to 0.93 |
0.01 |
|
|
Beta-Coefficient (per SD Increase in Genetically Predicted SHBG) |
|
|
| eGFR_cr (ml/min per 1.73 m2) |
460 |
2.48 |
1.34 to 3.62 |
2.0×10−5
|
| eGFR_crcys (ml/min per 1.73 m2) |
460 |
1.04 |
−0.20 to 2.27 |
0.10 |
OR, odds ratio; 95% CI, 95% confidence interval.
Discussion
Using MR to minimize confounding, our study suggests SHBG affects CKD and kidney function sex-specifically. SHBG might be an underlying protective factor against CKD in men but not in women. The effect in men remained after controlling for testosterone.
Our findings extend the limited evidence concerning the role of SHBG in CKD and kidney function, by showing a beneficial association of SHBG with CKD and kidney function in men. SHBG is related to lower bioavailable testosterone in men,15 genetically predicted lower bioavailable testosterone is related to lower risk of CKD and better kidney function in men.6 Despite these consistent beneficial associations, these findings need to be interpreted cautiously. MR assesses lifetime effects of an endogenous exposure rather than the short-term effects of an exogenous exposure. As such, the effect on CKD and kidney function might not be comparable to the acute renal effect of factors regulating SHBG. However, a low-protein diet, which increases SHBG,44 is recommended for patients with CKD. Moreover, given the limited studies on the role of SHBG, the underlying pathways are unclear. BP is a key risk factor for CKD, although a recent MR study suggests BP is not a causal factor for CKD.45 This finding is consistent with the unclear association of SHBG with BP found here. Although we cannot exclude the possibility that genetically predicted higher SHBG lowers BP, our findings suggest that other pathways beyond BP might exist. Inflammation may underlie the pathology of CKD.46 SHBG suppresses inflammation in in vitro experiments, and the effect is not altered by cosupplementation with testosterone or estradiol,47 supporting a pathway via inflammation. In addition, insulin resistance has recently been identified to affect CKD in a sex-specific way.48 Previous MR studies suggest SHBG lowers the risk of type 2 diabetes and possibly lowers insulin resistance,10–12 so insulin resistance may also mediate the renal effect of SHBG. Assessing these pathways and exploring other potential pathways, especially those relevant to the sex-specific response to SHBG and kidney function, would be valuable. Moreover, the sex difference in CKD progression and the role of sex hormones in the sex disparity may differ by etiology of CKD. For example, the sex difference in diabetic kidney disease and nondiabetic autosomal dominant polycystic kidney disease is less consistent across different studies than other types of CKD.3,4 As such, it would be valuable to understand the etiology of CKD, and examine whether the role of sex hormones differs in CKD with different etiologies in further studies.
Despite the novelty, our study has several limitations. First, MR rests on three assumptions, namely, the genetic variants are strongly related to the exposure, are not related to the confounders of the exposure-outcome relation, and only influence the outcome via the exposure.34 To satisfy these assumptions, we used genetic variants sex-specifically predicting SHBG from the largest published GWAS conducted in the UK Biobank and replicated in three independent studies.15 We also replicated the analysis using another set of genetic variants in the SHBG gene. To address the second assumption, we checked for their associations with potential confounders, such as socioeconomic position and lifestyle. To address the third assumption, we used several analysis methods robust to pleiotropy, which gave consistent estimates. Similar assumptions are also required for multivariable MR, except that the assumptions are extended to multiple exposures (here SHBG and testosterone). To satisfy these assumptions, we used genetic variants strongly related to SHBG and testosterone, identified from by far the largest GWAS.15 We also used multivariable MR-Egger, which is more robust to pleiotropy.43 Second, SHBG has a high variability,49 so measurement error might arise from the single time-point assessment of SHBG and wide age range in UK Biobank participants. Measurement error is expected to be nondifferential, thus bias is toward the null, which may explain the wide confidence intervals for some estimates. However, this will only lower the precision of MR estimates, rather than affect the directions of association. Replication in a larger sample would be worthwhile. Third, if the genetic associations with exposure and outcome are from the same sample, MR estimates with weak instruments might be biased.50 The bias is proportional to 1/F statistic.50 In this study, the sex-specific genetic variants for SHBG and the genetic associations with kidney function were both from the UK Biobank; however, weak instrument bias is not expected given the SNPs from the large sex-specific GWAS are strong instruments with high F statistics (average F statistic 183.4 in men and 124.1 in women). Moreover, replication using SNPs from a nonoverlapping sample showed a consistent pattern of associations. Fourth, our study could be affected by survivor bias,27 and by competing risk, namely, by an event (such as CVD) whose occurrence precludes the occurrence of CKD.26 As such, the associations with CKD will be underestimated due to competing events. To control for the inevitable bias, we adjusted for common causes of CKD and CVD,26 specifically smoking, BMI, and BP. Fifth, in MR studies, a common limitation is that the genetic variants only explain a small proportion of variance in the exposure,39 here the genetic variants explain 3%–3.5% of the variance in SHBG. As such, an MR study provides a unconfounded but less-precise estimate, and usually requires large sample sizes.38 To maximize power, we used by far the largest cohort study in Europeans. Sixth, the associations in Europeans may not apply to other populations, such as Asians. However, causal effects should be consistent across settings, for example, the effect of SHBG on diabetes has been reported in several studies conducted in different settings.10–12 Nevertheless, the effect size might vary by population according to levels of sex hormones,51,52 as such, replication in other populations, such as Asians, would be needed. Seventh, compensatory processes or feedback mechanisms may dilute the genetic effects, thus biasing the MR estimates toward the null. However, this cannot explain the observed beneficial associations of SHBG with CKD and kidney function. Lastly, because of the limited follow-up data on kidney function, we used baseline rather than decline in kidney function. It would be worthwhile to assess the role of SHBG in CKD progression, when sufficient follow-up data in the UK Biobank are available.
Our findings suggest SHBG is more than a transporter of sex hormones. SHBG and drivers of SHBG may serve as potential targets for CKD. Whether SHBG has a sex-specific role in other chronic conditions might also be considered. Medications, dietary factors or lifestyle, such as calorie restriction,53 which modulate SHBG might be effective for the prevention and treatment of CKD, especially in men. Exploring such factors and understanding the pathways would be relevant to drug repositioning, new drug development, and dietary recommendations.
Genetically predicted higher SHBG was associated with lower risk of CKD and better kidney function in men, but not in women. The role of SHBG in men might be independent of testosterone. Identifying drivers of SHBG and the underlying pathways could provide new insights into CKD prevention and treatment.
Disclosures
All authors have nothing to disclose.
Funding
None.
Ethics Approval and Consent to Participate
This research has been conducted using the UK Biobank Resource under Application number 42468. No original data were collected for the MR study. Ethics approval for each of the studies included in the investigation can be found in the original publications (including informed consent from each participant). The UK Biobank has already received ethics approval from the Research Ethics Committee and participants provided written informed consent. The analysis of other publicly available summary statistics does not require additional ethics approval.
Data Sharing Statement
The access of data from the UK Biobank can be obtained by application to the UK Biobank (http://biobank.ctsu.ox.ac.uk/crystal/). The summary statistics can be downloaded from the Neale Lab website (http://www.nealelab.is/uk-biobank/).
The main outcomes are from the UK Biobank under application 42468. The authors would like to thank the UK Biobank for approving our application. The authors would also like to thank the Neale Lab (http://www.nealelab.is/uk-biobank/) for conducting the GWAS of biomarkers and providing the sex-specific genetic associations in the UK Biobank.
Supplemental Material
This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020050659/-/DCSupplemental.
Supplemental Table 1. Genetic predictors for SHBG in men.
Supplemental Table 2. Genetic predictors for SHBG in women.
Supplemental Table 3. Genetic predictors for total testosterone in men.
Supplemental Table 4. Power calculation in men and women.
Supplemental Table 5. Association of each SHBG-related genetic variant with CKD and kidney function in men.
Supplemental Table 6. Association of each SHBG-related genetic variant with CKD and kidney function in women.
Supplemental Table 7. Associations of genetically predicted lower SHBG with CKD and kidney function in the UK Biobank using univariable MR.
Supplemental Table 8. Associations of SHBG with CKD and kidney function in the UK Biobank using excluding genetic variants related to alcohol drinking.
Supplemental Table 9. Associations of SHBG with CKD and kidney function in the UK Biobank using excluding genetic variants related to body mass index.
Supplemental Table 10. Associations of genetically predicted SHBG with CKD and kidney function in the UK Biobank using different analysis methods.
Supplemental Table 11. Associations of genetically predicted SHBG with CKD and kidney function in the UK Biobank using multivariable MR-Egger in men.
Supplemental Text. Formula of testing sex difference.
References
1. Jha V, Garcia-Garcia G, Iseki K, Li Z, Naicker S, Plattner B, et al.: Chronic kidney disease: Global dimension and perspectives. Lancet 382: 260–272, 201323727169
2. Carrero JJ, Hecking M, Chesnaye NC, Jager KJ: Sex and gender disparities in the epidemiology and outcomes of chronic kidney disease. Nat Rev Nephrol 14: 151–164, 201829355169
3. Neugarten J, Golestaneh L: Gender and the prevalence and progression of renal disease. Adv Chronic Kidney Dis 20: 390–395, 201323978543
4. Neugarten J, Acharya A, Silbiger SR: Effect of gender on the progression of nondiabetic renal disease: A meta-analysis. J Am Soc Nephrol 11: 319–329, 200010665939
5. Neugarten J, Golestaneh L: Influence of sex on the progression of chronic kidney disease. Mayo Clin Proc 94: 1339–1356, 201931272577
6. Zhao JV, Schooling CM: The role of testosterone in chronic kidney disease and kidney function in men and women: A bi-directional Mendelian randomization study in the UK Biobank. BMC Med 18: 122, 202032493397
7. Laurent MR, Hammond GL, Blokland M, Jardí F, Antonio L, Dubois V, et al.: Sex hormone-binding globulin regulation of androgen bioactivity
in vivo: Validation of the free hormone hypothesis. Sci Rep 6: 35539, 201627748448
8. Kim C, Ricardo AC, Boyko EJ, Christophi CA, Temprosa M, Watson KE, et al.; Diabetes Prevention Program Research Group: Sex hormones and measures of kidney function in the diabetes prevention program outcomes study. J Clin Endocrinol Metab 104: 1171–1180, 201930398516
9. Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G: Mendelian randomization: Using genes as instruments for making causal inferences in epidemiology. Stat Med 27: 1133–1163, 200817886233
10. Perry JR, Weedon MN, Langenberg C, Jackson AU, Lyssenko V, Sparsø T, et al.; MAGIC: Genetic evidence that raised
sex hormone binding globulin (SHBG) levels reduce the risk of type 2 diabetes. Hum Mol Genet 19: 535–544, 201019933169
11. Ding EL, Song Y, Manson JE, Hunter DJ, Lee CC, Rifai N, et al.: Sex hormone-binding globulin and risk of type 2 diabetes in women and men. N Engl J Med 361: 1152–1163, 200919657112
12. Wang Q, Kangas AJ, Soininen P, Tiainen M, Tynkkynen T, Puukka K, et al.: Sex hormone-binding globulin associations with circulating lipids and metabolites and the risk for type 2 diabetes: Observational and causal effect estimates. Int J Epidemiol 44: 623–637, 201526050255
13. Arathimos R, Granell R, Haycock P, Richmond RC, Yarmolinsky J, Relton CL, et al.: Genetic and observational evidence supports a causal role of sex hormones on the development of asthma. Thorax 74: 633–642, 201930936389
14. Dimou NL, Papadimitriou N, Gill D, Christakoudi S, Murphy N, Gunter MJ, et al.:
Sex hormone binding globulin and risk of breast cancer: A Mendelian randomization study. Int J Epidemiol 48: 807–816, 201931143958
15. Ruth KS, Day FR, Tyrrell J, Thompson DJ, Wood AR, Mahajan A, et al.; Endometrial Cancer Association Consortium: Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med 26: 252–258, 202032042192
16. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al.: UK biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 12: e1001779, 201525826379
17. Coviello AD, Haring R, Wellons M, Vaidya D, Lehtimäki T, Keildson S, et al.: A genome-wide association meta-analysis of circulating sex hormone-binding globulin reveals multiple Loci implicated in sex steroid hormone regulation. PLoS Genet 8: e1002805, 201222829776
18. McDaid AF, Joshi PK, Porcu E, Komljenovic A, Li H, Sorrentino V, et al.: Bayesian association scan reveals loci associated with human lifespan and linked biomarkers. Nat Commun 8: 15842, 201728748955
19. Larsson SC, Burgess S, Mason AM, Michaëlsson K: Alcohol consumption and cardiovascular disease: A mendelian randomization study. Circ Genom Precis Med 13: e002814, 202032367730
20. Zhuang Z, Gao M, Yang R, Li N, Liu Z, Cao W, et al.: Association of physical activity, sedentary behaviours and sleep duration with cardiovascular diseases and lipid profiles: A Mendelian randomization analysis. Lipids Health Dis 19: 86, 202032384904
21. Maggio M, Lauretani F, Basaria S, Ceda GP, Bandinelli S, Metter EJ, et al.:
Sex hormone binding globulin levels across the adult lifespan in women--the role of body mass index and fasting insulin. J Endocrinol Invest 31: 597–601, 200818787375
22. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, et al.; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration): A new equation to estimate glomerular filtration rate [published correction appears in
Ann Intern Med 155: 408, 2011]. Ann Intern Med 150: 604–612, 200919414839
23. National Institute of Diabetes and Digestive and Kidney Diseases: Quick reference on UACR & GFR. Available at:
https://www.niddk.nih.gov/health-information/professionals/advanced-search/quick-reference-uacr-gfr. Accessed November 10, 2020
24. Inker LA, Schmid CH, Tighiouart H, Eckfeldt JH, Feldman HI, Greene T, et al.; CKD-EPI Investigators: Estimating glomerular filtration rate from serum creatinine and cystatin C [published correction appears in
N Engl J Med 367: 681, 2012]. N Engl J Med 367: 20–29, 201222762315
25. Wang T, Ma X, Tang T, Jin L, Peng D, Zhang R, et al.: Overall and central obesity with insulin sensitivity and secretion in a han Chinese population: A Mendelian randomization analysis. Int J Obes 40: 1736–1741, 201627569682
26. Schooling CM, Lopez P, Yang Z, Zhao JV, Au Yeung SL, Huang JV: Use of multivariable Mendelian randomization to address biases due to competing risk before recruitment. Available at:
https://www.biorxiv.org/content/10.1101/716621v4. Accessed November 10, 2020
27. Schooling CM: Selection bias in population-representative studies? A commentary on Deaton and Cartwright. Soc Sci Med 210: 70, 201829754782
28. Palmer TM, Sterne JA, Harbord RM, Lawlor DA, Sheehan NA, Meng S, et al.: Instrumental variable estimation of causal risk ratios and causal odds ratios in Mendelian randomization analyses. Am J Epidemiol 173: 1392–1403, 201121555716
29. Didelez V, Sheehan N: Mendelian randomization as an instrumental variable approach to causal inference. Stat Methods Med Res 16: 309–330, 200717715159
30. Burgess S, Dudbridge F, Thompson SG: Combining information on multiple instrumental variables in Mendelian randomization: Comparison of allele score and summarized data methods. Stat Med 35: 1880–1906, 201626661904
31. Burgess S, Foley CN, Allara E, Staley JR, Howson JMM: A robust and efficient method for Mendelian randomization with hundreds of genetic variants. Nat Commun 11: 376, 202031953392
32. Bowden J, Del Greco M F, Minelli C, Davey Smith G, Sheehan NA, Thompson JR: Assessing the suitability of summary data for two-sample Mendelian randomization analyses using MR-Egger regression: The role of the I2 statistic. Int J Epidemiol 45: 1961–1974, 201627616674
33. Paternoster R, Brame R, Mazerolle P, Piquero A: Using the correct statistical test for the equality of regression coefficients. Criminology 36: 859–866, 1998
34. Davies NM, Holmes MV, Davey Smith G: Reading Mendelian randomisation studies: A guide, glossary, and checklist for clinicians. BMJ 362: k601, 201830002074
35. Verbanck M, Chen CY, Neale B, Do R: Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50: 693–698, 201829686387
36. Slob EAW, Burgess S: A comparison of robust Mendelian randomization methods using summary data. Genet Epidemiol 44: 313–329, 202032249995
37. Bowden J, Davey Smith G, Haycock PC, Burgess S: Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 40: 304–314, 201627061298
38. Freeman G, Cowling BJ, Schooling CM: Power and sample size calculations for Mendelian randomization studies using one genetic instrument. Int J Epidemiol 42: 1157–1163, 201323934314
39. Burgess S: Sample size and power calculations in Mendelian randomization with a single instrumental variable and a binary outcome. Int J Epidemiol 43: 922–929, 201424608958
40. Burgess S, Thompson SG: Multivariable Mendelian randomization: The use of pleiotropic genetic variants to estimate causal effects. Am J Epidemiol 181: 251–260, 201525632051
41. Burgess S, Dudbridge F, Thompson SG: Re: “Multivariable Mendelian randomization: The use of pleiotropic genetic variants to estimate causal effects”. Am J Epidemiol 181: 290–291, 201525660081
42. Sanderson E, Davey Smith G, Windmeijer F, Bowden J: An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. Int J Epidemiol 48: 713–727, 201930535378
43. Rees JMB, Wood AM, Burgess S: Extending the MR-Egger method for multivariable Mendelian randomization to correct for both measured and unmeasured pleiotropy. Stat Med 36: 4705–4718, 201728960498
44. Longcope C, Yosha S, Young RA, Baker S, Braverman LE: The effects of low-protein diet and testosterone on sex hormone-binding globulin capacity in male rabbits. Metabolism 36: 703–707, 19873600283
45. Yu Z, Coresh J, Qi G, Grams M, Boerwinkle E, Snieder H, et al.: A bidirectional Mendelian randomization study supports causal effects of kidney function on blood pressure. Kidney Int 98: 708–716, 202032454124
46. Li L, Tang W, Yi F: Role of inflammasome in chronic kidney disease. Adv Exp Med Biol 1165: 407–421, 201931399976
47. Yamazaki H, Kushiyama A, Sakoda H, Fujishiro M, Yamamotoya T, Nakatsu Y, et al.: Protective effect of sex hormone-binding globulin against metabolic syndrome:
In vitro evidence showing anti-inflammatory and lipolytic effects on adipocytes and macrophages. Mediators Inflamm, 2018 10.1155/2018/3062319 30046278
48. Zhao JV, Schooling CM: Sex-specific associations of insulin resistance with chronic kidney disease and kidney function: A bi-directional Mendelian randomisation study. Diabetologia 63: 1554–1563, 202032409868
49. Dahan MH, Goldstein J: Serum sex hormone-binding globulin levels show too much variability to be used effectively as a screening marker for insulin resistance in women with polycystic ovary syndrome. Fertil Steril 86: 934–941, 200617027360
50. Burgess S, Davies NM, Thompson SG: Bias due to participant overlap in two-sample Mendelian randomization. Genet Epidemiol 40: 597–608, 201627625185
51. Krakowsky Y, Conners W, Morgentaler A: Serum concentrations of sex hormone-binding globulin vary widely in younger and older men: Clinical data from a men’s health practice. Eur Urol Focus 5: 273–279, 201928753796
52. Yu S, Qiu L, Liu M, Li S, Tao Z, Zhang Q, et al.: Establishing reference intervals for sex hormones and SHBG in apparently healthy Chinese adult men on the basis of a multicenter study. Clin Chem Lab Med 56: 1152–1160, 201829729137
53. Cangemi R, Friedmann AJ, Holloszy JO, Fontana L: Long-term effects of calorie restriction on serum sex-hormone concentrations in men. Aging Cell 9: 236–242, 201020096034
