Should we predict maximum oxygen uptake using fat-free mass or body mass in a non-athletic population

Background: Our purpose was to assess whether fat-free mass (FFM) or body mass (M) is the more appropriate body-size variable to predict maximum oxygen uptake (VO2max) (L⋅min−1) in a non-athletic population.Methods: We adopted the well-known allometric/power function-model VO2max=a·Xb⋅e to predict VO2max (L⋅min−1) using either FFM or M as the predictor variable. These allometric models can be linearised with a log-transformation, and linear regression (or ANCOVA) can then be used to estimate the unknown parameters.Results: Initially, ignoring the effect of age and sex, FFM was the strongest predictor of VO2max. When we predicted Ln(VO2max) using only Ln(FFM) adjusting for the effects of age and sex, the explained variance was R2=0.718 (AIC=-1882.5), with the FFM exponent b=0.658. However, when we predicted Ln(VO2max) allowing body mass AND bodyfat% as separate covariates also adjusting for age and sex, the explained variance increased further to R2=0.733 (AIC=-2077.4), with the M exponent b=0.636. The difference in the explained variances, but more importantly, the difference in the AIC’s was 194.88, confirming the superiority of predicting VO2max using separate M and bodyfat% covariates, i.e., adopting FFM alone was less successful compared to using M and bodyfat% separately. Our analyses identified a greater negatively weighted bodyfat% term that improved the prediction of VO2max, found by our subsequent/final prediction models, to be due to excess central adiposity (waist circumference). Note that subsequent/final M and FFM exponents were both estimated to be b=0.67, confirming that VO2max should be normalized using VO2max (mL·M−2/3·min−1) or VO2max (mL·FFM−2/3·min−1) rather than VO2max (mL·FFM−1·min−1) as recommended by previous research.Conclusion: Incorporating FFM into equations to predict VO2max fails to adequately explain the negative effect of excess central adiposity. However, by incorporating body mass (M) and BF% separately into the equations, a greater negatively weighted BF% coefficient is capable of compensating for this apparent omission.