Sequences of Regressions Distinguish Nonmechanical from Mechanical Associations between Metabolic Factors, Body Composition, and Bone in Healthy Postmenopausal Women. Issue 4 (9th March 2016)

• Record Type: Journal Article
• Authors: Solis-Trapala, Ivonne; Schoenmakers, Inez; Goldberg, Gail R; Prentice, Ann; Ward, Kate A
• Is Part Of: Journal of nutrition. Volume 146:Issue 4(2016)
• Page Start: 846
• Page End: 854
• Publication Date: 2016-03-09
• Subjects: peripheral quantitative computed tomography -- bone -- adiponectin -- leptin -- osteocalcin -- graphical Markov model -- postmenopausal women -- fat-free mass -- fat mass -- BMI; Nutrition -- Periodicals; Diet -- Periodicals; 613.205
• DOI: 10.3945/jn.115.224485
• Languages: English
• ISSNs: 0022-3166

Abstract

Abstract: Background: There is increasing recognition of complex interrelations between the endocrine functions of bone and fat tissues or organs. Objective: The objective was to describe nonmechanical and mechanical links between metabolic factors, body composition, and bone with the use of graphical Markov models. Methods: Seventy postmenopausal women with a mean ± SD age of 62.3 ± 3.7 y and body mass index (in kg/m 2 ) of 24.9 ± 3.8 were recruited. Bone outcomes were peripheral quantitative computed tomography measures of the distal and diaphyseal tibia, cross-sectional area (CSA), volumetric bone mineral density (vBMD), and cortical CSA. Biomarkers of osteoblast and adipocyte function were plasma concentrations of leptin, adiponectin, osteocalcin, undercarboxylated osteocalcin (UCOC), and phylloquinone. Body composition measurements were lean and percent fat mass, which were derived with the use of a 4-compartment model. Sequences of Regressions, a subclass of graphical Markov models, were used to describe the direct (nonmechanical) and indirect (mechanical) interrelations between metabolic factors and bone by simultaneously modeling multiple bone outcomes and their relation with biomarker outcomes with lean mass, percent fat mass, and height as intermediate explanatory variables. Results: The graphical Markov models showed both direct and indirect associations linking plasma leptin and adiponectin concentrations with CSA and vBMD. At the distal tibia, lean mass, height, and adiponectin-UCOC interaction were directly explanatory of CSA ( R 2 = 0.45); at the diaphysis, lean mass, percent fat mass, leptin, osteocalcin, and age-adiponectin interaction were directly explanatory of CSA ( R 2 = 0.49). The regression models exploring direct associations for vBMD were much weaker, with R 2 = 0.15 and 0.18 at the distal and diaphyseal sites, respectively. Lean mass and UCOC were associated, and the global Markov property of the graph indicated that this association was explained by osteocalcin. Conclusions: This study, to our knowledge, offers a novel approach to the description of the complex physiological interrelations between adiponectin, leptin, and osteocalcin and the musculoskeletal system. There may be benefits to jointly targeting both systems to improve bone health.