Mendelian randomization highlights the causal role of normal thyroid function on blood lipid profiles
Abstract The association between thyroid function and dyslipidemia has been well documented in observational studies. However, observational studies are prone to confounding, making it difficult to conduct causal inference. We performed a two-sample bi-directional Mendelian randomization (MR) using summary statistics from large-scale genome-wide association studies (GWASs) of thyroid stimulating hormone (TSH), free thyroxine (FT4) and blood lipids. We chose inverse variance weighted (IVW) method as main analysis, and consolidated results through various sensitivity analyses involving six different MR methods under different model specifications. We further conducted genetic correlation analysis and colocalization analysis to deeply reflect the causality. The IVW method showed per one standard deviation (SD) increase in normal TSH was significantly associated with a 0.048 SD increase in total cholesterol (TC; P < 0.001) and a 0.032 SD increase in low-density lipoprotein cholesterol (LDL; P = 0.021). Per one SD increase in normal FT4 was significantly associated with a 0.056 SD decrease in TC (P = 0.014) and a 0.072 SD decrease in LDL (P = 0.009). Neither TSH nor FT4 showed causal associations with high-density lipoprotein cholesterol (HDL) and triglycerides (TG). No significant causal effect of blood lipids on normal TSH or FT4 can be detected. All results were largely consistent when using several alternative MR methods, and were re-confirmed by both genetic correlation analysis and colocalization analysis. Our study suggested that even within reference range, higher TSH or lower FT4 are causally associated with increased TC and LDL, while no reverse causal association can be found.