Fig. 3: Influence of PSA-related index event bias on prostate cancer GWAS.
From: Genetically adjusted PSA levels for prostate cancer screening
a, Conceptual diagram depicts how selection on PSA levels induces an association between genetic variant G and U, a composite confounder that captures polygenic and non-genetic factors. This selection induces an association with prostate cancer (PrCa) via path G–U→PrCa, in addition to the direct G→PrCa effects. Bi-directional dotted lines show that PSA is not only a disease biomarker but also influences screening behavior and the likelihood of prostate cancer detection. b,c, The impact of bias correction is shown for 209 prostate cancer risk variants. Independent risk variants were selected from the PRACTICAL GWAS meta-analysis (85,554 cases and 91,972 controls of European ancestry) by Conti et al.32 using LD clumping (LD r2 < 0.01, P < 5 × 10−8). For each variant, associations with PSA (βPSA) are based on an inverse-variance-weighted fixed-effects meta-analysis in men of European ancestry (n = 85,824). b, GWAS effect sizes for prostate cancer (βPrCa) are aligned to the risk-increasing allele. Bias-adjusted effect sizes (βadj) are denoted by triangles. c, Two-sided GWAS P values for prostate cancer (PPrCa) were derived from an inverse-variance-weighted fixed-effects meta-analysis. Two-sided bias-adjusted P values (Padj), denoted by triangles, were calculated from a chi-squared test statistic based on βadj and corresponding standard errors. Genome-wide significance threshold (P < 5 × 10−8) is indicated by the horizontal dotted line.
