Levels and changes in body mass index decomposed into fat and fat-free mass index: relation to long-term all-cause mortality in the general population.

Abstract

In the general population, body mass index (BMI = weight (kg)/(height (m))2) shows a U-shaped relation to mortality, which is attributable to a combination of an inverse association with fat-free mass index (FFMI) and a direct association with fat mass index (FMI). However, preceding changes in body composition related to diseases, health behaviors, or social conditions that are also influencing later mortality may confound these associations. To examine associations of FFMI and FMI, adjusted for preceding changes in FFMI and FMI over a 6 years period, with all-cause mortality in a healthy general population. The study population was a random subset of adult Danes, participating in the Danish MONICA project; 989 men and 962 women, born 1922, 1932, 1942, and 1952, and examined in 1987-88 and 1993-94. They had no known major co-morbidities until start of follow-up in 1993-94, and were followed up for 18 years. Measures included height, weight, and bio-impedance, from which BMI, FFMI, and FMI were calculated, and information on educational level, smoking, alcohol drinking, leisure-time physical activity, which were obtained by questionnaires. We analyzed the relation between body composition and all-cause mortality by Cox proportional hazards model with splines, stratified by birth cohorts, and with adjustment for preceding changes in body composition and for the covariates including gender. We estimated hazard ratios (HR) with 95% confidence intervals (CI) relative to HR = 1.00 at the median values of BMI, FMI, and FFMI. During 18 years of follow-up, 286 men and 200 women died. BMI showed the well-known U-shaped association with mortality, and FMI was directly and FFMI inversely associated with mortality. Associations were not significantly modified by gender. Preceding changes in BMI, FMI, and FFMI were only weakly and not significantly associated with mortality. Associations for FMI and FFMI were monotonic, but curve-linear with a higher mortality above and below the respective median values of FMI and FFMI: at the 5th percentiles of FMI and FFMI, HRs were 0.80 (CI 0.57-1.13) and 2.01 (1.24-3.27), and at the 95th percentiles, HRs were 2.16 (1.38-3.38) and 0.81 (0.52-1.27), respectively. In an apparently healthy general population, a large fat mass and a small fat-free mass are associated with greater risk of early mortality, also after adjusting for preceding changes in body composition, health behaviors, and educational level.