Abstract
Background Given the increasing prevalence of diabetes mellitus (DM) in the United States, estimating the effects of population-level increases in obesity on incident DM has substantial implications for public health policy. Therefore, we determined the population attributable fraction, which accounts for the prevalence and excess risk of DM associated with obesity. Methods and Results We included non-Hispanic White, non-Hispanic Black, and Mexican American participants without DM at baseline from MESA (Multi-Ethnic Study of Atherosclerosis) with available data on body mass index and key covariates from 2000 to 2017 to calculate unadjusted and adjusted (age, study site, physical activity, diet, income, and education level) hazard ratios (HR) for obesity-attributable DM. We calculated national age-adjusted prevalence estimates for obesity using data from NHANES (National Health and Nutrition Examination Survey) in 4 pooled cycles (2001-2016) among adults with similar characteristics to MESA participants. Last, we calculated unadjusted and adjusted population attributable fractions from the race/ethnic and sex-specific HR and prevalence estimates. Of 4200 MESA participants, the median age was 61 years, 46.8% were men, 53.9% were non-Hispanic White, 32.9% were non-Hispanic Black, and 13.3% were Mexican. Among MESA participants, incident DM occurred in 11.6% over a median follow-up of 9.2 years. The adjusted HR for obesity-related DM was 2.7 (95% CI, 2.2-3.3). Adjusted population attributable fractions were 0.35 (95% CI, 0.29-0.40) in 2001 to 2004 and 0.41 (95% CI, 0.36-0.46) in 2013 to 2016, and greatest among non-Hispanic White women. Conclusions The contribution of obesity towards DM in the population remains substantial and varies significantly by race/ethnicity and sex, highlighting the need for tailored public health interventions to reduce obesity. Registration URL: https://www.clinicaltrials.gov; Unique identifiers: NC00005487, NCT00005154.