Baseline BMI as determinant of frailty trajectories in elderly general population

Introduction Frailty is a state of vulnerability in elderly that increases the risk of disability and death. However, frailty is a dynamic process that progresses over time and some individuals may follow a more rapid frailty progression. Obesity is associated to a higher risk of frailty and could also be a determinant of frailty trajectories. The aim of this study was to examine whether individuals with different levels of baseline BMI might experience different trajectory of frailty progression. Methods We studied participants of the English Longitudinal Study of Ageing (ELSA) who attended wave 2 in 2004–2005. The determinant was baseline levels of BMI and the outcome was the longitudinal trajectory of a 36-item frailty index, assessed every 2 years (waves 2 to 7) over a 10-year period. In order to deal with missing data, we applied multiple imputation on the underlying variables and then calculated the frailty index. We fitted a mixed model with age as time scale, age/quadratic age as fixed-effects and frailty index trajectory as dependent variable allowing for intercept/slope random effects. We fitted two separate models for baseline BMI, one adjusting for sex and the other adjusted additionally for baseline values of: income, smoking status, physical activity, HbA1c, and haemoglobin. Quadratic terms and interactions with time/squared time were included in the models. We calculated yearly frailty index values at different BMI values (17, 22, 25, 30 and 35 kg/m 2 ). Results We analysed 5333 participants (mean age 71.2 years (SD 8.0) 44.4% men, 50% followed-up until 2014–2015). In the sex adjusted model, BMI was associated with frailty trajectories. Frailty trajectories varied with different values of baseline BMI, and the differences of frailty trajectories with different levels of BMI tended to be greater at 80 years and over. With the age-adjusted model, baseline BMI estimates were negative and significant [−0.017 (95% confidence interval = −0.022; −0.012)] with a positive significant interaction between BMI and time [4.177e-06 (95% confidence interval = 3.074e-06; 5.279e-06)]. With the fully adjusted model, all BMI estimates were still significant. Compared with participants with a BMI less than or equal to 25, participants with an initial BMI greater than 25 had frailty trajectories that increased over time. This trend was accentuated during follow-up. Conclusions Obese participants may experience a steeper deterioration of frailty with ageing. These effects were still significant when adjusting for potential confounders. Differently, participants with low values of BMI (17 kg/m 2 ) show slightly higher levels of frailty at 60 years, but with 70 years and over their frailty trajectories follow those of normal values of baseline BMI (22 and 25 kg/m 2 ).