Mendelian randomization study reveals a causal relationship between adiponectin and LDL cholesterol in Africans

Abstract

Adiponectin has been associated with cardiometabolic traits in observational studies across populations, yet it is unclear if these associations are causal. We performed Mendelian randomization (MR) analysis to assess the relationship between adiponectin and cardiometabolic traits in sub-Saharan Africans. We constructed a polygenic risk score (PRS) for adiponectin levels across 3354 unrelated sub-Saharan Africans. The PRS was used as the instrumental variable in two-stage least-squares MR analysis to assess its association with insulin resistance, HDL, LDL, total cholesterol, triglycerides, blood pressure, Type 2 Diabetes (T2D), and hypertension. The adiponectin PRS was causally related with LDL (β = 0.55, 95%CI 0.07-1.04, P-value = 0.024) but not the other traits. This association was observed in both overweight/obese and normal weight individuals, but only reached statistical significance among overweight/obese individuals (β = 0.55, 95%CI 0.01-1.08, P-value = 0.045). In normal weight individuals, the adiponectin PRS was associated with T2D (OR = 0.13, 95%CI 0.02-0.73, P-value = 0.021), and in men with HDL (β = 1.03, 95%CI 0.14-1.92, P-value = 0.023). The findings of this first MR study in sub-Saharan Africans support a causal relationship of adiponectin with LDL, with T2D in normal weight individuals only, and with HDL in men only. These observations add to the small but growing literature on adiponectin MR studies.

Figure 1

Conceptual model.

Figure 2

Bar plot of model fit of the adiponectin PRS at multiple P-value thresholds.

Figure 3

Causal associations between adiponectin and cardiometabolic outcomes using the two-stage least-squares method. Values are presented are adjusted for age, sex, population stratification, BMI, alcohol consumption, and smoking. SBP and DBP are additionally adjusted for blood pressure medication use. HOMA-IR and Triglycerides were log transformed. All continuous variables were Z-standardized. T2D cases were excluded for all HOMA-IR analyses.

Figure 4

Causal associations between adiponectin and cardiometabolic outcomes in normal weight and overweight/obese individuals using the two-stage least-squares method. Values are presented are adjusted for age, sex, population stratification, BMI, alcohol consumption, and smoking. SBP and DBP are additionally adjusted for blood pressure medication use. HOMA-IR and Triglycerides were log transformed. All continuous variables were Z-standardized. T2D cases were excluded for all HOMA-IR analyses.

Figure 5

Causal associations between adiponectin and cardiometabolic outcomes in men and women using the two-stage least-squares method. Values are presented are adjusted for age, population stratification, BMI, alcohol consumption, and smoking. SBP and DBP are additionally adjusted for blood pressure medication use. HOMA-IR and Triglycerides were log transformed. All continuous variables were Z-standardized. T2D cases were excluded for all HOMA-IR analyses.

Figure 6

Concordance between P-values of two-stage least-squares method and generalized method of moments (GMM) for all models.