Exploring the non‐linear association between BMI and mortality in adults with and without diabetes: the US National Health Interview Survey

Z. Wang, B. Dong, J. Hu, O. Adegbija, L. W. Arnold

To assess the non-linear relationship between BMI and mortality and to determine the BMI values with the lowest mortality risk in adults with and without diabetes. This observational study assessed the relationship between BMI and mortality with flexible parametric survival models using data from the US National Health Interview Survey. Participants included 25 458 adults with diabetes and 315 939 adults without diabetes, aged 18-84 years at baseline surveys, conducted from 1997 to 2009. Mortality status data were obtained from the linked mortality data up to 2011. We observed a U-shaped relationship between BMI and mortality in both adults with and without diabetes. With the BMI 25-29.9 kg/m(2) group as reference, hazard ratios (95% CI) of mortality for those with BMI <18.5, 18.5 to 24.9, 30-34.9, 35-39.9 and ≥40 kg/m(2) were 2.67 (2.12, 3.35), 1.26 (1.18, 1.35), 1.04 (0.98, 1.12), 1.12 (1.02, 1.22) and 1.37 (1.24, 1.51), respectively, for adults with diabetes, adjusting for age, sex, race and survey year. The corresponding hazard ratios for adults without diabetes were 2.97 (2.78, 3.17), 1.27 (1.23, 1.30), 1.07 (1.03, 1.12), 1.36 (1.27, 1.45), and 1.77 (1.62, 1.92), respectively. The BMI values associated with the lowest mortality were 29.1 kg/m(2) for adults with diabetes and 26.7 kg/m(2) for those without diabetes. Regardless of the presence of diabetes, there is a U-shaped relationship between BMI and mortality. The BMI values associated with the lowest mortality were above the current 'normal' range for adults with and without diabetes. This article is protected by copyright. All rights reserved.