The Effect of Blood Lipids on the Left Ventricle: A Mendelian Randomization Study

Background Cholesterol and triglycerides are among the most well-known risk factors for cardiovascular disease. Objectives This study investigated whether higher low-density lipoprotein (LDL) cholesterol and triglyceride levels and lower high-density lipoprotein cholesterol level are causal risk factors for changes in prognostically important left ventricular (LV) parameters. Methods One-sample Mendelian randomization (MR) of 17,311 European individuals from the UK Biobank with paired lipid and cardiovascular magnetic resonance data was performed. Two-sample MR was performed by using summary-level data from the Global Lipid Genetics Consortium (n = 188,577) and UK Biobank Cardiovascular Magnetic Resonance substudy (n = 16,923) for sensitivity analyses. Results In 1-sample MR analysis, higher LDL cholesterol was causally associated with higher LV end-diastolic volume (β = 1.85 ml; 95% confidence interval [CI]: 0.59 to 3.14 ml; p = 0.004) and higher LV mass (β = 0.81 g; 95% CI: 0.11 to 1.51 g; p = 0.023) and triglycerides with higher LV mass (β = 1.37 g; 95% CI: 0.45 to 2.3 g; p = 0.004). High-density lipoprotein cholesterol had no significant association with any LV parameter. Similar results were obtained by using 2-sample MR. Observational analyses were frequently discordant with those derived from MR. Conclusions MR analysis demonstrates that LDL cholesterol and triglycerides are associated with adverse changes in cardiac structure and function, in particular in relation to LV mass. These findings suggest that LDL cholesterol and triglycerides may have a causal effect in influencing cardiac morphology in addition to their established role in atherosclerosis. (J Am Coll Cardiol 2020;76:2477-88).

B oth the incidence and prevalence of ischemic heart disease, as well as its long-term sequelae such as heart failure, are on the rise (1,2). Cardiac imaging is an important and widely used tool in guiding the diagnosis and treatment of these patients (3). Left ventricular (LV) parameters derived from cardiac imaging modalities such as end-diastolic volume, ejection fraction, and mass are known to be prognostically important with respect to subsequent major adverse cardiovascular events and cardiovascular death (4,5). Low-density lipoprotein (LDL) cholesterol is one of the best publicized and most unequivocally implicated risk factors in the development of ischemic heart disease; its causal involvement in atherosclerotic plaque formation in the arterial system is well elucidated (6). For triglycerides, a causal relationship with cardiovascular disease has also been demonstrated (7), and for high-density lipoprotein (HDL) cholesterol, low levels are associated with increased risk of cardiovascular disease, but causality has not been established (8). However, to our knowledge, no study has established the causative impact of lipids on the structure and function of the LV. The overall study protocol has been described in detail previously (11), as has the CMR protocol and reference ranges (12,13 I n s t r u m e n t a l v a r i a b l e a n a l y s i s . MR was per-

RESULTS
Demographic and clinical characteristics of study participants, as well as median CMR parameter values, are outlined in Table 1     Analyses controlling for HDL and triglycerides genetic instruments did not change the significant results (Supplemental Table 4), indicating that the causal relationships between LDL cholesterol and LVEDV and LV mass were robust to confounding from other lipid fractions. One-sample MR was additionally performed following adjustment for all covariates and yielded similar results (which are detailed, along with results for HDL and triglycerides, in Supplemental Table 5). Models examining the association between LV parameters and genetically determined lipid levels where phenotypic lipid levels were included as a covariate showed no significant attenuation of these associations, indicating an independent effect (Supplemental Table 6). Sensitivity analysis performed using a restricted list of variants in the GRS following exclusion of potentially pleiotropic variants (69, 81, and 50 variants remained for LDL, HDL, and triglycerides, respectively) yielded concordant results to the primary analysis (Supplemental Table 7).
Based on Egger intercept p-values, no directional horizontal pleiotropy was detected. Sensitivity analyses by MR-Egger, weighted median, and weighted mode methods produced associations with concordant effect directions, although the confidence intervals were much wider, as expected (Supplemental Table 9, Supplemental Figure 3). The MR pleiotropy residual sum and outlier method found evidence of horizontal pleiotropy for the association between LDL cholesterol and LVEDV, but upon removal of outlier variants, the effect estimates were not significantly changed as per the distortion tests (Supplemental Table 10). The MR-Steiger test suggested that the assumption of causal directionality for the relationships between the LDL GRS and the LV parameters was correct (Supplemental Table 11). The variance of phenotypic LV parameters explained by observed lipid measurements and lipid genetic risk scores is presented in Supplemental Table 12. HDL CHOLESTEROL. In multivariate analysis, higher phenotypic HDL cholesterol levels were associated with higher LVEDV (b ¼ 8.27 ml; 95% CI: 7.00 to 9.53 ml; p < 0.0001) and higher LV mass (b ¼ 1.   Table 1. Lipids and the LV median, and weighted mode methods (Supplemental Table 9, Supplemental Figure 4).  Table 4). Sensitivity analysis performed by using a restricted list of variants in the GRS following the exclusion of potentially pleiotropic variants yielded concordant results to the primary analysis (Supplemental Table 7).   Table 8). There was evidence of horizontal pleiotropy for the association between triglycerides and LVEDV, but removal of the outlier variants did not significantly change the effect estimates as indicated by the distortion tests (Supplemental Table 10). Sensitivity analysis examining the association of triglycerides and LV mass using MR-Egger, weighted median, and weighted mode methods did not reach significance because of wider CIs, although they showed concordant effect directions (Supplemental Table 9, Supplemental Figure 5). The assessment of causal directionality using the MR Steiger test supported what this study's hypothesis has proposed (Supplemental Table 11).  Table 13).

DISCUSSION
To our knowledge, this study is the first to conduct    Tables 1 and 2.

Non-User
Statin User Non-User