Genetic determinants of BMI, diet, and fitness interact to partially explain anthropometric obesity traits but not the metabolic consequences of obesity in men and women

Authors: Arrington, Carmen; Tacad, Debra Kirsty; ALLAYEE, HOOMAN - University Of California (UCLA); SUTTON, KRISTEN - University Of Colorado; DOMBROWSKI, CATHERINE - University Of California, Davis; Keim, Nancy; Newman, John; Bennett, Brian

Submitted to: International Journal of Obesity
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/27/2026
Publication Date: 2/20/2026
Citation: Arrington, C.E., Tacad, D.M., Allayee, H., Sutton, K.J., Dombrowski, C., Keim, N.L., Newman, J.W., Bennett, B.J. 2026. Genetic determinants of BMI, diet, and fitness interact to partially explain anthropometric obesity traits but not the metabolic consequences of obesity in men and women. International Journal of Obesity. https://doi.org/10.1038/s41366-026-02027-0.
DOI: https://doi.org/10.1038/s41366-026-02027-0

Interpretive Summary: BMI has long been used to assess obesity, but it doesn’t fully reflect body fat or how it’s distributed in the body. This study reveals a major issue: while genetic risk scores for BMI can predict BMI itself, they don’t accurately predict other key obesity measures, such as total body fat or fat distribution (e.g., abdominal versus hip fat). The research identified important gaps where the genetic risk of BMI doesn't appropriately capture body adiposity accumulation or its distribution. What was found is that other factors—such as metabolism, physical fitness, and sex—contribute much more to these specific body fat measures. The study suggests that future research on obesity should focus on a more specific genetic risk variable for body adiposity, combined with these other key predictors, to better understand and predict obesity-related traits.

Technical Abstract: Objectives: To assess how combining genetic, diet, and lifestyle factors describe the variation in body composition and distribution phenotypes. Methods: A cohort of men and women 18-66y (n =230) balanced by age, sex, and BMI from the USDA Nutritional Phenotyping Study (NCT02367287) provided demographic data, body composition, fitness, resting metabolic rate, diet, and biological samples. Imputed genetic data was used to calculate a BMI polygenic risk score (PRS; PGS002313) via the pgs_calc pipeline. Principle components (PC) analysis was used to capture genetic variation. Linear regression and ANCOVA assessed variables impact on outcomes measured by dual-energy X-ray absorption (DXA) including body fat percentage (BF%), lean mass index (LMI), trunk fat percentage (TF%), and the android-to-gynoid ratio (AGR). Explained variance was evaluated using sum of squares and partial R², followed by model comparison with Akaike’s (AIC) and Bayesian (BIC) information criteria. Statistical analysis and visualizations were done using R. Results: The PRS explained 15.6% of the variance in BMI (R² =0.156) and predicted obesity (BMI > 30) with good discriminatory ability (AUROC =0.71). ANCOVA confirmed the PRS as a BMI predictor (p =1.1x10'7), accounting for 10.9% of variation after adjusting for covariates (age, sex, PC 1-5). Diet, fitness, and resting metabolic rate (RMR) showed mixed independent associations with body composition phenotypes. In models constructed with all variables, variance analysis revealed that: a) RMR explained 11.7% of BMI variation, 10.7% of TF%, and 8.0% of BF%, but less for other variables; b) Sex explained 37.0% of LMI variation and 27.1% of BF% variation; c) Diet and genetics had lower contributions, with BMI PRS accounting for 6.0% of BMI variation and 4.7% of LMI variation; d) Fitness explained 7-9% of variation in all phenotypes except LMI (2.2%). In final models, age, sex, PRS, and RMR were retained for all DXA outcomes by AIC and BIC. Healthy eating index (HEI) was excluded by BIC and fitness was excluded by both AIC and BIC in the model of LMI, while only fitness was excluded by BIC in the model of AGR. Conclusions: The BMI PRS significantly predicts BMI and obesity, explaining a portion of the variance in body composition outcomes. However, lifestyle factors such as sex, RMR, and fitness also play important roles, with their contributions varying across different measures of adiposity.