Health risk factors such as body mass index (BMI), serum cholesterol, and blood pressure are associated with many human common diseases1,2, e.g., being overweight is associated with increased risk to cardiovascular diseases (CVD)3 and type-2 diabetes (T2D)4. These associations are usually derived from observational studies that cannot distinguish whether the risk factors are “upstream” causal factors, “downstream” consequences of the diseases, or confounding factors associated with both the exposures and outcomes. The randomized controlled trial (RCT) is considered to be the gold standard approach to test for causality. For instance, LDL-cholesterol (LDL-c) was initially found to be associated with coronary artery disease (CAD) in an observational study5, and the association was subsequently confirmed to be causal by RCTs6,7. However, RCTs are time-consuming, expensive, and sometimes impractical or even unethical8. It is not feasible to design RCTs that can test many different interventions simultaneously. Genetic methods are useful to infer causality because genetic variants are present from birth and therefore unlikely to be confounded with environmental factors. Mendelian randomization (MR) is an analysis that uses genetic variants, which are expected to
