Factors Associated with Obesity Adjusting for Self-Reported Skin Color-Ethnicity Versus Genomic Ancestry in Admixed Brazilian Population

Abstract Objectives To evaluate differences in factors associated with obesity, considering self-reported color-ethnicity information in comparison with genomic ancestry in admixed Brazilian population. Methods Data from 563 free-living unrelated individuals aged ≥12 years (mean = 47; min. = 12; max = 93) of cross-sectional population-based Health Survey of São Paulo with Focus on Nutrition (2015 ISA-Nutrition) were evaluated, including self-reported skin color-ethnicity, blood samples and other information collected in 2015. Genotyping was accessed from blood using Axiom™ PMR Array, resulting in 555,064 high-performing markers, after quality control and merging with 1000 Genomes Project data. Population structure was evaluated using principal component analysis (PCA). Factors associated with obesity were assessed using two logistic regression models: 1) using self-reported skin color-ethnicity (white, black, mixed, or other), and 2) using the first four components of PCA. Other variables included in models were sex, age, age squared, income, education, marital status, smoking status, physical activity, diet quality, high blood pressure, diabetes, and dyslipidemia. Results Prevalence of obesity in the sample was 24.9%. In model 1, factors positively associated with obesity were: female (OR = 2.1; 95%CI = 1.3–3.3, ref. = male), educational level adequate for age (OR = 1.7; 95%CI = 1.0–2.9, ref. = lower education level for age), being divorced (OR = 2.8; 95%CI = 1.4–5.6, ref. = married), and having high blood pressure (OR = 2.4; 95%CI = 1.4–3.9, ref. = normal pressure), whilst income was negatively associated (OR = 0.99; 95%CI = 0.997–0.999). Variables associated with obesity in model 1 remained significant in model 2, with similar OR and 95%CI. Other variables were also significant in model 2: age (OR = 1.07; 95%CI = 1.0–1.14), age squared (OR = 0.99; 95%CI = 0.998–0.999), being widow/er (OR = 0.47; 95%CI = 0.22–0.99, ref. = married), having diabetes (OR = 1.8; 95%CI = 1.02–3.0, ref. = without diabetes). Post-estimation statistics showed that both models performed similarly, with model 2 presenting slightly better results. Conclusions The use of genomic ancestry data added important information in the investigation of obesity in this admixed population. Funding Sources São Paulo Municipal Health Department, FAPESP, and CNPq.