Abstract 4146374: Geographic polygenic heterogeneity of obesity across the United States

Introduction: The rising prevalence of obesity represents a public health crisis. Genetic analyses have established the strong heritable basis of body-mass index (BMI), which can be estimated by polygenic risk scores (PRS). However, the geographic distribution of the genetic risk of obesity is unknown. Hypothesis: BMI and BMI PRS vary across US regions and social indices interplay with BMI PRS. Methods: We use the nationwide All of Us Research Program to quantify BMI and BMI PRS heterogeneity in the European ancestry across US regions. Social indices include median income, vacant housing, deprivation index, having health insurance and poverty. PRS was computed using LDpred2 with 681275 individuals from GIANT consortium and UK Biobank. We compared heterogeneity within/between ZIP codes and states using linear mixed models. BMI and BMI PRS were adjusted with the first 100 principal components (PC) of ancestry. Our analyses were adjusted for age, sex and the first 10 PCs. Results: Across 93825 individuals, the mean (standard deviation,SD) age was 58.5(16.9) years, 56369(60%) were female, and the mean (SD) BMI was 29.2(7.2) kg/m 2 . BMI PRS explained 11.8%(95%CI 11.4-12.1%,P<2x10 -16 ) of BMI variance. Pennsylvania had the highest significant variance explained (R 2 ) by PRS with 13.2%(N=14288,P<2x10 -16 ). BMI and BMI PRS were both spatially clustered (Moran’s I spatial autocorrelation=0.83 and 0.77, P<10 -16 ,respectively). Urban areas had 0.12 SD-lower BMI compared to rural areas(P=4x10 -9 ). Within each state, BMI and BMI PRS varied significantly, e.g. east Pennsylvania has a significantly lower BMI than the west. BMI variance explained by ZIP code was higher than by state (R 2 =2%,P=2x10 -37 vs R 2 =0.6%,P=7x10 -14 ,respectively). High vacant housing and low median income were the strongest social indices associated with high BMI(P<2x10 -16 ). Per 1-SD higher PRS,1-SD higher vacant housing and lower median income additionally increased BMI by 0.01 and 0.02 SD(interaction P<2x10 -16 ),respectively. High poverty,no health insurance and low deprivation index are independently associated with low BMI. Conclusion: Understanding BMI differential distributions by genetic effects and social indices may uncover drivers of obesity across the US.