Scaling cardiac structural data by body dimensions: a review of theory, practice, and problems.

Robust estimates of the "true" bivariate relationship between body size (X) and heart size (Y) have seldom been determined empirically. The removal of the covariate influence of body size from cardiac dimension variables facilitates both correct inter- or intra-group comparisons, and the construction of reference standards for normality. In the literature to date this "scaling" or normalisation of cardiac dimensions has been performed typically via a per-ratio standards method, (Y/X), with body surface area chosen as the size denominator. This review demonstrates that the per-ratio standards approach may be theoretically, mathematically, and empirically flawed. The most appropriate scaling procedure appears to be a curvilinear, allometric model of the general form Y = aXb. The cardiac dimension variable (Y) may be regressed upon the body size variable (X) to derive a power function ratio (Y/Xb) that is allegedly size-independent. The current consensus is that an estimate of fat-free mass (FFM) provides the most appropriate body size variable. In the scaling literature allometric modelling procedures have generally yielded FFM exponents (b) consistent with the theory of geometric similarity. We suggest that cardiac dimension data should be scaled by appropriate powers of FFM, derived from allometric modelling. However, despite the potential superiority of FFM as a scaling denominator, reference standards for normality based on FFM have not been developed or proposed. Future research should examine the robustness of the FFM-cardiac dimension relationship in large samples.