Genetic Variants, Serum 25-Hydroxyvitamin D Levels, and Sarcopenia

This genetic association study investigates serum 25-hydroxyvitamin D levels and sarcopenia risk among participants in the UK Biobank.

This supplemental material has been provided by the authors to give readers additional information about their work.

The inclusion criteria and exclusion criteria in the present study
The inclusion criteria in the genetic and observational analyses for the present study were: 1) unrelated European ancestry; 2) individuals having complete data regarding serum 25(OH)D concentration and relevant covariates; and 3) having any data on sarcopenia status.The exclusion criteria were: 1) sex mismatches, excess heterozygosity; 2) closer than third-degree relatives; 3) non-European ancestry; 4) missingness of serum 25(OH)D concentration, 5) lack of relevant covariates for genetic and observational analyses; 6) having no data of diagnosis of sarcopenia.

Standard Mendelian randomization (MR) methods
The MR-Egger approach can detect and correct directional pleiotropy by the MR-Egger intercept, albeit it has relatively low statistical. 1MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO) can help identify and correct horizontal pleiotropic outliers. 2We also inspected Cochran's Q statistic to assess heterogeneity between individual genetic variants, an indication of the presence of invalid instruments. 3

Non-linear MR
Non-linear MR is a genetic risk score (GS)-based single-sample approach.In brief, the entire UK Biobank sample was stratified into 100 strata according to the residual variation of the serum 25(OH)D concentration after regressing on the GS.Then, we used the ratio-of-coefficients method to calculate the localized average causal effect (LACE) in each stratum.Further, meta-regression of LACE against the stratum-specific mean exposure was performed by fitting a range of fractional polynomial exposureoutcome models of Degrees 1 and 2. The best-fitting model was identified by the likelihood ratio test, and the polynomial test for non-linearity in which the best-fitting fractional polynomial model of Degree 1 was compared against the linear model was reported. 4

MR key assumptions
The validity of causal inferences drawn from MR relies on three key assumptions: (1) genetic variants are robustly associated with the exposure (instrument strength), (2) genetic variants are not associated with potential confounders (independence), and (3) genetic variants are associated with the outcome only via the exposure (no unbalanced horizontal pleiotropy).

Several steps to assess the potential violation of MR assumptions
To minimize the potential violation of MR assumptions, we took several steps to assess the validity of the genetic variants.First, we evaluated instrument strength using the F statistic, with an F-value greater than 10 indicating a strong instrument with a low potential for instrument bias. 5Second, for the independence assumption, we restricted analyses to self-reported European ancestry and adjusted for principal components to minimize the potential confounding by population stratification.Third, to evaluate the independence assumption, we excluded blood, metabolic, renal, and other traits associated variants in the original vitamin D-GS from the PhenoScanner search and re-generated GS for vitamin D independent of these traits.Forth, we examined associations between the Vitamin D-GS and potential confounders to explore the violation of the no horizontal pleiotropy assumptions.Finally, we used a broader set of SNVs that consist of 122 autosomal variants as instrumental variable to serum 25(OH)D concentration to test the robustness of the results of our non-linear MR analyses.

Sensitivity analyses to verify MR assumptions and evaluate bias
Our primary vitamin D-GS was robustly associated with serum 25(OH)D concentration in the UK Biobank, explaining 2.1% of the variation (F statistic = 6,139, P<.001).We found little evidence that vitamin D-GS was associated with potential confounders in the UK Biobank, including BMI, smoking, alcohol intake, physical activity, and Townsend deprivation index (P>.05for all, see eTable 5).For the sensitivity analyses, we re-generated vitamin D-GS by excluding blood, metabolic, renal, or other traitsassociated variants, separately, and repeated the non-linear MR analyses.The pleiotropic effects of these outliers identified by the PhenoScanner database are shown in eTable 6.The L-shaped associations of genetically-predicted 25(OH)D with sarcopenia and its indices were consistent and were not driven by a particular group of variants (eTable 7).In addition, vitamin D GS with 122 variants as an instrument provided similar results, confirming the L-shaped associations of genetically-predicted 25(OH)D with sarcopenia and its indices (eFigure 3).
© 2023 Sha T et al.JAMA Network Open.

1 . 2 . 5 . 2 . 3 .
Genome-wide significant vitamin D variants used for the genetic instruments for serum 25(OH)D concentrations eTable Standard linear Mendelian randomization estimates for the associations of genetically predicted serum 25(OH)D with sarcopenia and its indices using 35 SNVs to instrument eTable 3. Two-sample Mendelian randomization estimates for the association of genetically predicted serum 25(OH)D with sarcopenia and its indices using 35 SNVs to instrument eTable 4. Two-sample Mendelian randomization estimates for the association of genetically predicted serum 25(OH)D with sarcopenia and its indices using 35 SNVs to the instrument after excluding outliers eTable Association of vitamin D-GS with potential confounders in the UK Biobank eTable 6. Functional blocks-related traits used in the sensitivity analyses eTable 7. Sensitivity analysis for non-linear Mendelian randomization analysis of serum 25(OH)D on sarcopenia risk eFigure 1. Radial Mendelian randomization plots for serum 25(OH)D and sarcopenia and its indices using 35 SNVs to the instrument eFigure Age-stratified non-linear Mendelian randomization analyses of genetic associations of serum 25(OH)D with sarcopenia and its indices, (A) and (E) sarcopenia, (B) and (F) grip strength, (C) and (G) appendicular lean mass index, and (D) and (H) slow gait speed eFigure Genetic associations of serum 25(OH)D with sarcopenia and its indices using 122 SNVs to instrument, (A) sarcopenia, (B) grip strength, (C) appendicular lean mass index, and (D) slow gait speed eReferences

eTable 1 .
Genome-wide significant vitamin D variants used for the genetic instruments for serum 25(OH)

*
The summary statistics of serum 25(OH)D concentrations were extracted from 417,580 European-ancestry individuals in the UK Biobank study.† The summary statistics of serum 25(OH)D concentrations were extracted from 79,366 European-ancestry individuals in the SUNLIGHT Consortium study.BP: base-pair position, Chr: chromosome number; EAF: effect allele frequency; SE: standard error; SNV: single-nucleotide variant; vitamin D-GS: genetic instrument for serum 25(OH)D concentration; a SNV proxy in the SUNLIGHT consortium: rs1123571; b SNV proxy in the SUNLIGHT consortium: rs2186777.

eTable 2 .
Standard linear Mendelian randomization estimates for the associations of genetically predicted serum 25(OH)D with sarcopenia and its indices using 35 SNVs to instrument * 295,489 participants of unrelated European ancestry from the UK Biobank were included in the genetic analyses.† Residual serum 25(OH)D.‡ per 10 nmol/L increase in serum 25(OH)D, OR for sarcopenia and slow gait speed, while a beta for grip strength and appendicular lean mass index.ORs/betas were estimated by two-stage least squares regression method with an adjustment for age, age-square, sex, birth location, assessment center, top 20 genetic principal components, genotyping array in both stages, and nuisance factors, which could affect serum 25(OH)D measurements, including the month in which blood sample was taken, fasting time before the blood sample was taken, and sample aliquots for measurement.OR, odds ratio; CI, confidence interval.

eFigure 1 .
Radial Mendelian randomization plots for serum 25(OH)D and sarcopenia and its indices using 35 SNVs to the instrument.eFigure 2. Age-stratified non-linear Mendelian randomization analyses of genetic associations of serum 25(OH)D with sarcopenia and its indices, eFigure 3. Genetic associations of serum 25(OH)D with sarcopenia and its indices using 122 SNVs to instrument, (A) sarcopenia, (B) grip strength, (C) appendicular lean mass index, and (D) slow gait speed.The red dot represents the reference point of serum 25(OH)D of 50nmol/L.The gray lines represent the 95% confidence intervals.The adjustment includes age, age-square, sex, assessment center, birth location, top 20 genetic principal components, genotyping array in both stages, and nuisance factors, which could affect serum 25(OH)D measurements, including month in which blood sample was taken, fasting time before blood sample was taken, and sample aliquots for measurement.

eTable 3. Two-sample Mendelian randomization estimates for the association of genetically predicted serum 25(OH)D with sarcopenia and its indices using 35 SNVs to instrument *
The summary statistics of sarcopenia, grip strength, and appendicular lean mass index were extracted from 324,976 unrelated European-ancestry individuals in the UK Biobank study, and the summary statistics of slow gait speed were *

Two-sample Mendelian randomization estimates for the association of genetically predicted serum 25(OH)D with sarcopenia and its indices using 35 SNVs to the instrument after excluding outliers
3The summary statistics of sarcopenia, grip strength, and appendicular lean mass index were extracted from 324,976 unrelated European-ancestry individuals in the UK Biobank study, and the summary statistics of slow gait speed were extracted from 322,814 unrelated European-ancestry individuals in the UK Biobank study.†peroneunit increase in serum 25(OH)D, OR for sarcopenia and slow gait speed, while a beta for grip strength and appendicular lean mass index.‡PforCochran's Q statistic, which was used to assess the heterogeneity between SNVs.P for Cochran's Q statistic, which was used to assess the heterogeneity between SNVs.The detailed definition of Cochran's Q statistic was published elsewhere.3MR,Mendelianrandomization; SNV, single-nucleotide variant; OR, odds ratio; CI, confidence interval; MR-PRESSO, MR Pleiotropy RESidual Sum and Outlier; IVW, inverse-variance weighted.*Functionalblocks identified using trait associations identified through PhenoScanner V2.© 2023 Sha T et al.JAMA Network Open.eTable 7.

Sensitivity analysis for non-linear Mendelian randomization analysis of serum 25(OH)D on sarcopenia risk. Excluded functional block SNVs associated with the functional block Pnon-linear *
*Likelihood ratio test comparing the besting-fitting fractional polynomial model of degree 1 against the linear model © 2023 Sha T et al.JAMA Network Open.