Body composition measurements for nutrition research.

INTRODUCTION: THE BASIS OF MEASUREMENTS ‘A proper evaluation of the individual living body and its functional characteristics is not possible from the simple classification system of the butcher.. . ’. With these remarks Keys & Brozek (1 953) dismissed the contributions of anatomists to body composition measurements and introduced the concepts that still sustain modern investigators in this field. These are that meaningful nutritional and physiological investigations in energy metabolism are those focussed on a distinction between the relatively inert components of the body and those that are metabolically more active. The former consist of extracellular fluid, bone mineral and depot fat and the rest, muscle and muscle-free lean tissue. Similarly, Siri (1956) suggested that a distinction between fat, minerals, protein and water provided a physiologically useful compartmentalization. These views were in fact simply more elaborate statements of an earlier idea (Behnke et al. 1942) that the body could be thought of as consisting of two components, fat and lean tissue. Either way these propositions indicate first, what it may be useful to measure and second, that, with the addition of weight measurements, the size of a single unmeasured compartment can be calculated if estimates are available for the other components of the system. Unfortunately, the simplicity of these arguments conceals a considerable difficulty. That is that since the size of an individual compartment is rarely measured directly, rather some constituent or property of it, the precision and accuracy of the basic measurement cannot easily be translated into equivalent values for the variable of interest. Account has to be taken of the consistency of the relationship between the measured constituent or property and the compartment to which it belongs and, when the size of an undetermined compartment is measured by difference between body-weight and the weight of other compartments, relative compartmental size determines the magnitude of the proportional errors. Examples of these problems will emerge in subsequent paragraphs.