Evaluation of causal association and risk stratification of body mass index polygenic risk scores with cardiovascular risk factors in African populations

Abstract

Introduction Cardiovascular diseases (CVDs) are a major cause of morbidity and mortality globally and in sub-Saharan Africa. Body mass index (BMI) is a crucial risk factor for non-communicable diseases, such as CVD. However, the genetic mechanisms by which it contributes to non- communicable diseases are poorly understood. Multivariant recall by genotype (RbG) studies, using polygenic risk scores (PRS) offer an opportunity to unravel mechanistic insights into how BMI causes CVD. However, a major limitation is the poor performance of Europe-designed PRS in the African population. The aim was to compare the causal association and risk stratification of BMI PRSs using the RbG design in Africa’s East, South, and West regions. Methods This cross-sectional secondary data analysis study utilised the Africa Wits-INDEPTH Partnership for Genomic Research (AWI-Gen) data set to compare seven validated BMI PRS methodologies using the RbG approach to evaluate their utility for causal inferences in carotid intima-medial thickness (CIMT) and blood pressure (BP). The AWI-Gen data set comprised 12,000 participants aged 29 and above from East, West and South regions of Africa. Linear regression was used for predictivity and causal inference analysis. Confounding analyses were performed using the T-test and the Chi-square test. BMI was an outcome variable, and PRSs were exposure variables. Age, sex, education and socioeconomic status were included as confounders. The analyses were performed for the entire study population and per region. Results In 10,457 continental Africans, we utilised the RbG approach to evaluate the causal association of BMI with average CIMT and BP. In linear models, BMI was positively associated with CIMT (in AWI-Gen: Beta 0.0022 (p<0.0001)) and blood pressure (In AWI-Gen: Beta 0.5065 (p<0.0001) for diastolic BP; Beta 0.72667 (p<0.0001) for systolic BP) and African iv regions. The predictivity of BMI PRSs was low, ranging from 0% for LDPred2_MA to 1.67% for PRSice_MAMA. In the RbG, the PRSs failed to control for confounding of all the covariates. Conclusion The BMI PRSs are not useful for stratification and causal inference through RbG due to their low predictivity in continental Africans. There is a need for genome-wide association studies (GWAS) in continental Africa, which could compute more predictive PRS that can be used in RbG approaches for causal inference.