analysis to avoid possible issues (Bowden et al., 2015; Hartwig et al., 2016). The variants included in each genetic instrument used in the present analysis are listed in online Supplemental Table S3. For each exposure, two instrumental variables were built considering GWS loci (P < 5× 10−8) and suggestive loci (P < 5× 10−5). We verified these MR estimates using the MR-RAPS approach, which is a method designed to identify and estimate confounded causal effects using weak genetic instrumental variables (Zhang et al., 2018). To ensure the reliability of the significant findings, we performed heterogeneity tests based on three different methods: inverse-variance weighted, MR-Egger regression, and maximum likelihood (online Supplemental Table S4). To further confirm the absence of possible distortions due to heterogeneity and pleiotropy, we tested the presence of horizontal pleiotropy among the variants included in the genetic instrument using MR-PRESSO (Verbanck et al., 2018). Finally, the funnel plot and leave-one-out analysis were conducted to identify potential outliers among the variants included in the genetic instruments tested. The MR analyses were conducted using the TwoSampleMR R package (Hemani et al., 2018b).
