Abstract

Among the explanations proposed for the weak and inconsistent association between BMI and mortality in the elderly are the lack of adjustment for waist circumference (WC) and that the association varies with health status. This work examines the independent association of BMI and WC with mortality in older adults, and the influence of health status on this association. A cohort of 3,536 persons representative of the Spanish population aged >= 60 years was selected in 2000 and 2001, and followed prospectively until 2007. The analyses were performed with Cox models and adjusted for the main confounders. During follow-up, 659 persons died (18.6% of the cohort). Before adjusting for WC, mortality in the upper quartile of BMI was 15% lower than in the lower quartile (hazard ratio (HR): 0.85; 95% confidence interval (CI): 0.66-1.08; P for linear trend = 0.076). After adjusting for WC, the association was even stronger, so that mortality in the upper quartile of BMI was 37% lower than in the lower quartile (HR: 0.63; 95% CI: 0.45-0.88; P for linear trend 0.003). Before adjusting for BMI, no association was observed between WC and mortality. After adjusting for BMI, WC was positively associated with mortality (HR for upper vs. lower quartile of WC: 1.48; 95% CI: 1.07-2.05; P for linear trend = 0.008). These associations were mainly observed in those with limitations in mobility and agility. BMI has an inverse, and WC has a direct, independent association with mortality in older adults, particularly in those with worse health status.