China Kadoorie Biobank

The CKB is a prospective cohort of 512,713 adults aged 30 to 79 years. Individuals were recruited between the years of 2004–2008 from five urban and five rural regions across China, as previously described (94). Baseline information was collected via detailed questionnaire (including demographic/lifestyle factors and medical history) and physical measurements (which included anthropometry, blood pressure, and spirometry). A nonfasting blood sample was taken and separated into plasma and buffy coat fractions for long-term storage. Longterm follow-up is through electronic linkage of each participant’s unique national identification number to the Chinese national health insurance system and established regional registries for death and disease. Health insurance reports include detailed information [e.g., disease description, ICD-10 code, and procedure or examination codes] about each hospital admission. Vascular disease events have been reviewed and standardized by clinicians.

A total of 102,783 CKB participants were genotyped using two custom-designed Affymetrix Axiom arrays including up to 800,000 variants, optimized for genome-wide coverage in Chinese populations. Stringent QC included an SNP call rate of >0.98, plate effect P > 10⁻⁶, batch effect P > 10⁻⁶, HWE P > 10⁻⁶ (combined 10 df χ² test from 10 regions), biallelic, MAF difference from a 1000 Genomes Project (1KGP) EAS of <0.2, a sample call rate of >0.95, heterozygosity < mean + 3 SD, no chromosome XY aneuploidy, genetically determined sex concordant with database, and exclusion of recent immigrants to each study area as identified by region-specific PCA, resulting in genotypes for 532,415 variants present on both array versions in 94,592 individuals. Imputation into the 1000 Genomes phase 3 reference (EAS MAF, >0) using SHAPEIT3 and IMPUTE4 yielded genotypes for 10,276,634 variants with an MAF of >0.005 and an info of >0.3. rs188810925 was not imputed (1KGP EAS MAF, 0), and rs7209826 was imputed with an info of 0.986 and an MAF of 0.30.

Outcomes, phenotypes, and transformations are defined in table S14. Disease end points are for hospitalizations, from electronic linkage to the national health insurance system. Phenotypes were measured either at baseline or in a subset of individuals at the second resurvey of ~5% of the CKB cohort.

Ethical approval for CKB was obtained jointly from the University of Oxford, the Chinese Center for Disease Control and Prevention (CCDC), and the regional CCDC from the 10 study areas.

We attempted transethnic replication in CKB. In 21,547 individuals with eBMD measurements available, the rs7209826 variant showed no evidence of association (β = 0.0016 g/cm²; SE = 0.001; P = 0.14) and strong evidence of heterogeneity with the corresponding eBMD estimate in UKBB (P _{heterogeneity} = 1.35 × 10⁻²⁸; fig. S19). A regional analysis showed no evidence of a suitable genetic instrument (fig. S20). Transethnic replication in a large East Asian biobank (CKB) could not be reliably performed because of a lack of association of SOST variants with BMD; however, there was a lack of association with CHD and other outcomes in CKB (table S15). This provides an opportune means to illustrate that the multiple cardiometabolic risk factor and disease associations identified in the European datasets, wherein genetic variants of SOST act as valid instruments for inhibition of sclerostin, did not arise as a consequence of horizontal pleiotropy (that is, the genetic associations did not arise through a mechanism other than that which operates through inhibition of sclerostin). The absence of suitable instruments in CKB does not imply that pharmacological inhibition of sclerostin will not be an effective therapeutic strategy for raising BMD in non-European populations.