Linear and nonlinear analyses of the association between low-density lipoprotein cholesterol and diabetes: The spurious U-curve in observational study

Abstract

Objective: Hyperlipidemia is traditionally considered a risk factor for diabetes. The effect of low-density lipoprotein cholesterol (LDL-C) is counterintuitive to diabetes. We sought to investigate the relationship between LDL-C and diabetes for better lipid management.

Methods: We tested the shape of association between LDL-C and diabetes and created polygenic risk scores of LDL-C and generated linear Mendelian randomization (MR) estimates for the effect of LDL-C and diabetes. We evaluated for nonlinearity in the observational and genetic relationship between LDL-C and diabetes.

Results: Traditional observational analysis suggested a complex non-linear association between LDL-C and diabetes while nonlinear MR analyses found no evidence for a non-linear association. Under the assumption of linear association, we found a consistently protective effect of LDL-C against diabetes among the females without lipid-lowering drugs use. The ORs were 0.84 (95% CI, 0.72-0.97, P=0.0168) in an observational analysis which was more prominent in MR analysis and suggested increasing the overall distribution of LDL-C in females led to an overall decrease in the risk of diabetes (P=0.0258).

Conclusions: We verified the liner protective effect of LDL-C against diabetes among the females without lipid-lowering drug use. Non-linear associations between LDL-C against diabetes in observational analysis are not causal.

Keywords: cardiometabolic disease; diabetes; fasting blood glucose (FBG); linear causal association; low-density lipoprotein cholesterol (LDL-C).