Abstract
OBJECTIVE: This study examined the associations of variability and patterns in BMI, body weight (BW), and waist circumference (WC) with cognitive decline. METHODS: A total of 4304 participants (aged ≥65 years) from the National Health and Aging Trends Study between 2011 around 2021 were analyzed. Adjusted mixed-effect models assessed BW and body composition variability metrics linked to cognitive function z scores over 11 years, including standard deviation (SD), coefficient of variation (CV), root mean square error, last-to-first assessment change groups, and overall pattern over the follow-up. RESULTS: Participants in the highest SD variability quartile had the fastest cognitive decline (β = -0.036 [95% CI: -0.044 to -0.028] z scores per year) compared with the lowest variability BMI quartile (β = -0.019 [95% CI: -0.027 to -0.010] z scores per year, p values for interaction, < 0.001). Similar trends were observed for BMI CV and root mean square error, BW SD and CV, and WC CV. Compared with the stable or gain group, participants with ≥5% loss in BMI and BW had the fastest cognitive decline (both p values for interaction, < 0.0002). The cognitive decline rates among the stable, loss, gain and cycling patterns in BMI, BW, or WC were not significantly different. CONCLUSIONS: Greater variability and loss in BW and body composition were linked to accelerated cognitive decline in older adults.