Investigating the impact of body composition on the estimation of resting metabolic rate: new equations for adults aged ≥65 years developed using cross-sectional data.

Background: Due to changes in body composition during aging, the inclusion of body composition measures as a variable within equations to predict resting metabolic rate (RMR) may improve their predictive accuracy.

Objectives: This analysis of cross-sectional data aimed to develop and validate new RMR equations for older adults (≥65 y) incorporating variables for body composition, to predict performance and accuracy, and to explore the relative contribution of body composition variables acting directly or potentially via fat-free mass (FFM) to RMR.

Methods: Analyses were conducted utilizing a unique international dataset of gold standard measures developed for this purpose. RMR was predicted from potential predictive variables using stepwise multiple regression. Predictive performance of the final model was assessed using double cross-validation. The new prediction equation was compared with published prediction equations for similar populations and with previously published RMR prediction equations that did not include FFM. Direct associations between the determined predictor variables and RMR with indirect effects mediated via FFM were examined using mediation final (or pathway) analysis.

Results: The dataset contained 1238 participants. The predictive equations {utilizing either FFM (Equation 1) or lean body weight [LBW](Equation 2)} follow. Equation 1: RMR = 8.645 × height + 23.684 × weight - 29.717 × age + 38.213 × FFM + 209.637 × sex + 2693.223; Equation 2: RMR = -30.570 × age + 80.736 × LBW - 186.825 × sex + 3956.822 where RMR (kJ/d); height (cm); weight (kg); age (y); FFM (kg); LBW (kg); sex (M = 1, F = 0). The equation performed similarly to some anthropometric-based prediction equations. Predictors using FFM performed marginally better than those using LBW. All variables had significant (P < 0.001) direct effects upon RMR and significant (P < 0.001) indirect effects for sex, weight, and height.

Conclusions: New prediction equations predict RMR at the population level with minimal bias; however, the difference in performance with anthropometry-based equations is minimal. This may be explained by the contribution of FFM to weight, whereby equations that include weight are already accounting for FFM.