Body Composition by Computed Tomography vs Dual-Energy X-ray Absorptiometry: Long-Term Prediction of All-Cause Mortality in the Health ABC Cohort.

Abstract

Body composition assessment by computed tomography (CT) predicts health outcomes in diverse populations. However, its performance in predicting mortality has not been directly compared to dual-energy X-ray absorptiometry (DXA). Additionally, the association between different body compartments and mortality, acknowledging the compositional nature of the human body, is not well studied. Compositional data analysis, which is applied to multivariate proportion-type data set, may help to account for the interrelationships of body compartments by constructing log ratios of components. Here, we determined the associations of baseline CT-based measures of mid-thigh cross-sectional areas versus DXA measures of body composition with all-cause mortality in the Health ABC cohort, using both traditional (individual body compartments) and compositional data analysis (using ratios of body compartments) approaches. The Health ABC study assessed body composition in 2911 older adults in 1996-1997. We investigated the individual and ratios of (by compositional analysis) body compartments assessed by DXA (lean, fat, and bone masses) and CT (muscle, subcutaneous fat area, intermuscular fat, and bone) on mortality, using Cox proportional hazard models. Lower baseline muscle area by CT (hazard ratio [HR]men = 0.56; 95% confidence interval [95% CI]: 0.48-0.67, HRwomen = 0.60; 95% CI: 0.48-0.74) and fat mass by DXA (HRmen = 0.48; 95% CI: 0.24-0.95) were predictors of mortality in traditional Cox regression analysis. Consistently, compositional data analysis revealed that lower muscle area versus IMF, muscle area versus bone area, and lower fat mass versus lean mass were associated with higher mortality in both sexes. Both CT measure of muscle area and DXA fat mass (either individually or relative to other body compartments) were strong predictors of mortality in both sexes in a community research setting.