Results
Colocalization analysis within the GDF15 gene supported the model with a shared causal genetic variant (posterior probability > 0.99; Figure) for GDF15 levels and BMI, with rs16982345 identified as the most likely shared causal variant. In Mendelian randomization analysis using the rs16982345 variant, genetically proxied higher circulating GDF15 levels were associated with increased BMI (change in standard deviation [SD] units per one SD increase in GDF15 levels: 0.021 [95% confidence interval 0.014 to 0.028], p =\(4\times 10^{-9}\)). Exploring bi-directional effects, there was no evidence of genetically proxied BMI being associated with GDF15 levels (change in SD units per one SD increase in BMI: 0.005 [95% confidence interval -0.133 to 0.143], p = 0.95). There was no strong evidence for colocalization of GDF15 levels and type 2 diabetes liability (posterior probability = 0.16).