Predicting VO Using Lung Function and Three-Dimensional (3D) Allometry Provides New Insights into the Allometric Cascade (M).

BACKGROUND

Using directly measured cardiorespiratory fitness (i.e. VO) in epidemiological/population studies is rare due to practicality issues. As such, predicting VO is an attractive alternative. Most equations that predict VO adopt additive rather than multiplicative models despite evidence that the latter provides superior fits and more biologically interpretable models. Furthermore, incorporating some but not all confounding variables may lead to inflated mass exponents (∝ M) as in the allometric cascade.

OBJECTIVE

Hence, the purpose of the current study was to develop multiplicative, allometric models to predict VO incorporating most well-known, but some less well-known confounding variables (FVC, forced vital capacity; FEV, forced expiratory volume in 1 s) that might provide a more dimensionally valid model (∝ M) originally proposed by Astrand and Rodahl.

METHODS

We adopted the following three-dimensional multiplicative allometric model for VO (l⋅min) = M·HT·WC·exp(a + b·age + c·age + d·%fat)·ε, (M, body mass; HT, height; WC, waist circumference; %fat, percentage body fat). Model comparisons (goodness-of-fit) between the allometric and equivalent additive models was assessed using the Akaike information criterion plus residual diagnostics. Note that the intercept term 'a' was allowed to vary for categorical fixed factors such as sex and physical inactivity.

RESULTS

Analyses revealed that significant predictors of VO were physical inactivity, M, WC, age, %fat, plus FVC, FEV. The body-mass exponent was k = 0.695 (M), approximately∝M. However, the calculated effect-sizes identified age and physical inactivity, not mass, as the strongest predictors of VO. The quality-of-fit of the allometric models were superior to equivalent additive models.

CONCLUSIONS

Results provide compelling evidence that multiplicative allometric models incorporating FVC and FEV are dimensionally and theoretically superior at predicting VO(l⋅min) compared with additive models. If FVC and FEV are unavailable, a satisfactory model was obtained simply by using HT as a surrogate.