PURPOSE: This study evaluates the relationship between body mass index (BMI), fat mass index (FMI) and fat-free mass index (FFMI) changes and mortality in persons ≥65 years.

METHODS: Adults ≥65 years with at least two body composition measurements (BCM) between 1990 and 2011 were included. We excluded persons who died within one month of the second BCM and who had two single BCM in a one-month timeframe. Mortality data was retrieved until December 2012. For each person, we calculated the regression slopes for BMI, FMI and FFMI changes. Significant positive slopes were categorized as "gain", negative slopes as "loss" and the others as "maintenance". The impact of body composition changes was evaluated by Cox regression models while adjusting for sex, age, co-morbidities and body composition at the last measurement.

RESULTS: We included 791 persons with 3049 BCM. After adjustment for sex, and age and co-morbidities, a loss of FFMI, but not of FMI or BMI, increased the risk of mortality (HR 2.02, 95%CI 1.28-3.19). The prediction of mortality with FFMI loss remained significant when further adjusting for FMI loss and the last available body composition (HR 1.68, 95%CI 1.04-2.70).

CONCLUSIONS: FFMI loss is related to increased mortality in older persons.