Use of factor analysis to model relationships between bone mass and physical, dietary, and metabolic factors in frail and pre-frail older adults.

Bone mass and quality decline with age, and can culminate in osteoporosis and increased fracture risk. This investigation modeled associations between bone and physical, dietary, and metabolic factors in a group of 200 pre-frail/frail older adults using factor analysis and structural equation modeling (SEM). Exploratory (EFA) and confirmatory factor analysis (CFA) were conducted to compose factors and to assess their robustness. SEM was used to quantify associations between bone and the other factors. Factors arising from EFA and CFA were: bone (whole body, lumbar and femur bone mineral density, and trabecular bone score; good fit), body composition - lean (lean mass, body mass, vastus lateralis, and femoral cross-sectional area; good fit), body composition - fat (total fat mass, gynoid, android, and visceral fat; acceptable fit), strength (bench and leg press, handgrip, and knee extension peak torque; good fit), dietary intake (kilocalories, carbohydrate, protein, and fat; acceptable fit), and metabolic status (cortisol, insulin-like growth factor 1 (IGF-1), growth hormone (GH), and free testosterone; poor fit). SEM using isolated factors showed that body composition (lean) (β = 0.66, < 0.001), body composition (fat) (β = 0.36, < 0.001), and strength (β = 0.74, < 0.001) positively associated with bone. Dietary intake relative to body mass negatively associated with bone (β = -0.28, = 0.001), whereas in absolute terms, it showed no association (β = 0.01, = 0.911). In a multivariable model, only strength (β = 0.38, = 0.023) and body composition (lean) (β = 0.34, = 0.045) associated with bone. Resistance training programs that focus on improving lean mass and strength in older individuals may benefit bone in this population. We used factor analysis and structural equation modeling, which are rarely used in nutrition or exercise science, but constitute powerful tools that may overcome limitations of traditional analyses, combining individual related variables into factors or constructs of interest. Our investigation represents a starting point on this progressive pathway, providing useful insight and a working model for researchers and practitioners who wish to tackle complex problems such as the multifactorial causes of bone loss in older adults.