Abstract

To estimate the potential for a nonmechanical association between fat and bone mineral density (BMD), the authors evaluated a sample of 868 13-year-old girls attending schools in Porto, Portugal (2003-2004). Adiposity was estimated using fat mass from bioelectrical impedance analysis and fat area from a skinfold equation. Forearm BMD was measured by dual-energy x-ray absorptiometry. Associations were quantified with coefficients obtained using standard multivariate regression and residuals regression. Path analysis was used to assess the plausibility of the causal assumptions. Crude adiposity was positively associated with BMD; standardized menarcheal-age-adjusted coefficients (b(st)) for fat mass and fat area were 20.7 (95% confidence interval (CI): 17.6, 23.9) and 17.8 (95% CI: 14.6, 21.0), respectively. After standard weight adjustment, the association with adiposity decreased (b(st) = -2.30 (95% CI: -12.0, 7.37) and b(st) = 1.96 (95% CI: -2.84, 6.76), respectively), and the effect of residuals was small. In path analysis, direct effects were negligible (b(st) = -1.57 (95% CI: -11.9, 8.75) and b(st) = 1.89 (95% CI: -3.13, 6.91), respectively), and adiposity did not contribute to the overall model fit. The association between adiposity and BMD was strongly mediated by weight, and estimates were consistent between methods. No specific benefit in bone quality should be expected from relative increases in adiposity.