Fig. 4: Evaluation of the effect of body size on breast cancer at different ages.
From: Causal mediation analysis for time-varying heritable risk factors with Mendelian randomization
a Two-sided 95% confidence/credible intervals of childhood body size (from UK Biobank) and adult body mass index (BMI) on breast cancer risk estimated by MVMR-IVW, GRAPPLE, and FLOW-MR (without multiple testing correction). Error bar centers indicate point estimates (same for panels (b, d, e). Number of SNPs for each p-value threshold: 10−8 (56 SNPs), 10−6 (119 SNPs), 10−4 (305 SNPs), and 0.001 (570 SNPs). b Two-sided 95% confidence/credible intervals of childhood BMI and adult BMI on breast cancer risk estimated by these three methods (without multiple testing correction). Number of SNPs for each p-value threshold: 10−8 (56 SNPs), 10−6 (118 SNPs), 10−4 (304 SNPs), and 0.001 (571 SNPs). c Estimated causal DAG using FLOW-MR at p-value threshold 0.001. The red (blue) arrows indicate significant positive (negative) direct effects at significance level α = 0.05 (two-sided, without multiple testing correction). 95% Two-sided credible intervals estimated by FLOW-MR are also provided (without multiple testing correction) for (d) all the direct effects and e indirect effects of 1-year-old BMI. SNP counts for each threshold: 10−8 (55 SNPs), 10−6 (117 SNPs), 10−4 (303 SNPs), and 0.001 (570 SNPs). Source data are provided with this paper.
