The American Journal of Clinical Nutrition · Nov 2026
Data-driven thresholds for annualized changes in body mass index and waist circumference and all-cause mortality in United States adults aged 65 years or older: a secondary analysis of a prospective cohort
By Xu F, Woolf K, Earp JE et al.
Background Although longitudinal changes in body mass index [BMI (in kg/m²)] provide prognostic value beyond single measurements, and waist circumference (WC) adds complementary information about abdominal size, standardized annualized percentage change (APC) thresholds for mortality risk remain unestablished. Objectives To identify data-driven BMI and WC APC thresholds associated with all-cause mortality in United States adults aged ≥65 y. Methods This secondary analysis utilized National Health and Aging Trends Study Rounds 1-14 data (2011‒2024). A common analytic framework was applied within BMI (n = 5321; 2010 deaths) and WC (n = 5375; 2082 deaths), respectively, to identify candidate thresholds using training samples (80%) via the Contal and O’Quigley methods. Sensitivity analyses evaluated sex-specific, alternative follow-up horizons, unintentional weight loss, and final follow-up exclusions. Thresholds were further evaluated in a validation sample (20%) using Kaplan–Meier survival analysis and Cox proportional hazards models. Joint BMI/WC associations were evaluated using Cox models within the full eligible sample (n = 5130; 1947 deaths). Results Data-driven BMI (‒1.6%/+ 0.9%) and WC (‒1.2%/+ 1.2%) thresholds predicted higher mortality risk. Results remained consistent across sensitivity analyses, with slight variations by sex and final follow-up exclusions. Decreases in BMI and WC were associated with increased risk in the training [BMI: hazard ratio (HR): 4.03; 95% confidence interval (CI): 3.46, 4.69; WC: HR: 3.89; 95% CI: 3.35, 4.52] and validation samples (BMI: HR: 3.87; 95% CI: 2.77, 5.41; WC: HR: 3.54; 95% CI: 2.55, 4.92). A similar risk pattern occurred for increases in BMI and WC. Joint 9-category BMI/WC analyses showed a significant multiplicative interaction (Pinteraction < 0.001); participants with discordant BMI and WC change directions exhibited the highest adjusted HRs. Conclusions In older United States adults, data-driven BMI and WC APC thresholds stratify mortality risk. Joint BMI/WC change categories suggest heterogeneous mortality risk patterns, particularly when BMI and WC change in opposite directions. These thresholds require external validation before clinical application.