Cholesterol paradox in atrial fibrillation: is it solved in lipid-lowering drugs users?

Abstract Background In recent years, a growing number of prospective observational studies have consistently observed a paradoxical inverse association between atherogenic lipid fractions, particularly low-density lipoprotein cholesterol (LDL-C) and incident atrial fibrillation (AF), suggesting a protective role of hypercholesterolemia. However, elevated lipid levels at baseline study examinations increase likelihood of lipid-lowering therapy (LLT) initiation during follow-up, introducing substantial time-varying confounding. Consequently, patients under LLT may represent a more homogeneous clinical group for further exploration of the "cholesterol paradox" in AF. Purpose We aimed to investigate the association of baseline serum and genetically-predicted LDL-C levels with risk of AF development in patients receiving LLT. Methods We conducted a prospective cohort study of 90,114 UK Biobank participants aged 40–69 years, who did not have history of AF at baseline and who reported use of any kind of LLT during a nurse-led interview. Baseline LDL-C was measured in non-fasted serum samples using a direct assay. For genetically-predicted LDL-C, we used a previously published polygenic risk score (PRS). We used Cox proportional hazards regression to estimate AF risk associated with serum and genetically-predicted LDL-C levels. We considered two covariate adjustment sets: Model 1 - including age, sex and body mass index; and Model 2 further including ethnicity, deprivation index, smoking status, alcohol intake, systolic and diastolic blood pressure, glycated hemoglobin, anti-hypertensive and anti-diabetic medication use, prevalent coronary heart disease and valvular heart disease. Potential non-linearity was explored using restricted cubic splines. Results During a median of 13.1 years (1,199,436 person-years) of follow-up, 11,164 participants developed incident AF. The median baseline serum LDL-C was 2.9 mmol/L interquartile range: 2.4-3.4 mmol/L. In Model 1, a 1 mmol/L increase in LDL-C level was associated with lower risk of AF (hazard ratio HR: 0.75, 95% confidence interval CI: 0.73-0.77). Following further covariate adjustment (Model 2), the inverse association between LDL-C and risk of AF was attenuated (HR: 0.98, 95%CI: 0.96-1.01). There was no evidence for nonlinearity (Figure 1) of these associations (P-nonlinearity=0.74 and 0.63). PRS for LDL-C was neither associated with AF in Model 1 (HR per 1 standard deviation increase: 0.96, 95%CI: 0.92-1.00) nor Model 2 (HR: 0.97, 95%CI: 0.93-1.02). Conclusions In a cohort of patients receiving LLT, an inverse link between LDL-C and AF disappeared after adjustment for other clinical factors. Additionally, the absence of genetic evidence supporting a protective role of lipids further suggests that the "cholesterol paradox" in AF may be attributed to residual confounding rather than a causal relationship.