Linear and Nonlinear Associations Between Lipoprotein(a) and the Risks of Atherosclerotic Cardiovascular Disease

Lipoprotein(a) [Lp(a)] is a recognized risk factor for atherosclerotic cardiovascular disease (ASCVD), but the shape and potential nonlinearity of its association remain uncertain. We assessed the linear and nonlinear associations between Lp(a) levels and ASCVD risk using observational and Mendelian randomization (MR) approaches.

We analyzed 351,858 UK Biobank participants (2006-2023), stratified into Lp(a) percentiles: <70th, 70th-<80th, 80th-<90th, and ≥90th. Outcomes included ASCVD events from hospital, primary care, self-report, and death registry data. Cox models estimated the hazard ratios (HRs). MR analyses used a polygenic risk score from 10 Lp(a)-associated single-nucleotide polymorphisms, with nonlinearity tested by doubly ranked MR. Higher Lp(a) levels were associated with increased ASCVD risk. Compared with the <70th percentile, adjusted HRs (95% confidence interval) were 1.11 (1.07-1.16), 1.18 (1.14-1.22), and 1.25 (1.21-1.30) for the 70th-<80th, 80th-<90th, and ≥90th groups. Kaplan-Meier curves diverged early by group. Spline models suggested nonlinearity with an inflection near 130 nmol/L (P=0.007). MR showed a 2% higher ASCVD risk per 10 nmol/L genetically predicted Lp(a) (P<2×10^-16). Nonlinear MR suggested steeper gradients at higher levels, though not statistically significant (P=0.087).

Elevated Lp(a) concentrations were causally associated with ASCVD risk, showing a predominantly graded relationship with possible nonlinearity at very high levels, supporting routine Lp(a) measurement and the development of Lp(a)-lowering therapies.