Results from conditional analyses were largely consistent with those from unconditional analyses (Fig. 5; Supplementary Table 11), suggesting that most of the marginal effects are independent of the other risk factors analyzed in this study. Conditioning on the other risk factors, SBP, LDL-c and BMI were the three major risk factors for CAD, BMI was still a large risk factor for T2D and the protective effect of LDL-c on T2D remained largely unchanged (Supplementary Fig. 14). We show above that the GSMR analyses identified significant protective effects of HDL-c against CVD, CAD, T2D and hypertension (Supplementary Fig. 15). However, all the effects became non-significant conditioning on the covariates (i.e., BMI, LDL-c, TG, and SBP), suggesting that the marginal effects of HDL-c on the diseases are not independent of the covariates due to the bidirectional causative associations between HDL-c and the other risk factors as illustrated in Supplementary Fig. 13. It is difficult to distinguish whether the effects of HDL-c on the diseases are mediated or driven by the covariates (Supplementary Fig. 16) because of the complicated association network among risk
