Abstract
BACKGROUND: Mendelian randomization studies suggest a causal effect of lipoprotein(a) (Lp(a)) on atherosclerotic cardiovascular disease. Noncardiovascular effects (eg, diabetes risk) are inadequately investigated. OBJECTIVES: In this noninterventional phenome-wide association study designed to better understand the potential causal role of Lp(a), direct causal phenotypic effects of exposure to Lp(a) were estimated. Also, the association between LPA null allele rs41272114 with type 2 diabetes was assessed, and ancestry-specific Lp(a) thresholds were determined. METHODS: In the UK Biobank (n = 425,677 adults, 55% female), we studied 1,456 phenotypes spanning 18 classes using 4 ancestry-specific polygenic risk scores and false discovery rate multiple testing correction. Network deconvolution Mendelian randomization was leveraged to separate direct from indirect (ie, associations via mediating variables) causal phenotypic effects and account for confounding, reverse causation, and bidirectionality. RESULTS: Lp(a) was significantly associated with 80 phenotypes across 7 classes. Higher Lp(a) exposure had significant direct causal effects, independent of low-density lipoprotein cholesterol, on coronary artery disease (OR: 1.36; 95% CI: 1.21-1.54) and glycated hemoglobin (HbA1c; β = 0.099; 95% CI: 0.051-0.15) only. Very low Lp(a) exposure was not associated with type 2 diabetes (OR: 0.92; 95% CI: 0.64-1.31) or HbA1c (β = -0.016; 95% CI: -0.062 to 0.030). Among European and African ancestries, 86 (77th percentile) and 93 (59th percentile) nmol/L optimally discriminated myocardial infarction risk, respectively. CONCLUSIONS: Increasing Lp(a) exposure had direct, independent causal effects on coronary artery disease and HbA1c only; very low Lp(a) exposure is suggested to not be causally associated with type 2 diabetes. The optimal European and African ancestry threshold to stratify cardiovascular risk is comparable, and below 125/105 nmol/L in current U.S./European medical professional society guidelines.