Bioelectrical impedance validation studies: alternative approaches to their interpretation

BACKGROUND/OBJECTIVES: Cross-validation of methods of body composition assessment necessitates statistical evaluation of the degree to which the two methods are in agreement. Typically, impedance-based methods for predicting body composition are assessed against other methods using limits of agreement and correlation analysis. Alternative approaches are presented with reference to example body composition data obtained using bioimpedance spectroscopy (BIS) and dual-energy X-ray absorptiometry (DXA). SUBJECTS/METHODS: A randomly selected data set, drawn from a body composition database, was analysed by limits of agreement analysis and error grid analysis. RESULTS: The precision of BIS-derived predictions of percentage body fat relative to that of DXA can be determined from limits of agreement analysis. The importance of knowing the precision of the reference method in such analyses was highlighted. Error grid analysis has the potential to aid interpretation of method comparison data in an intuitively understandable way. CONCLUSIONS: Alternative ways of comparing analytical methods that are in use in other branches of biomedical research may prove useful when evaluating the utility of impedance-based methods and other methods for the assessment of body composition in cross-validation studies.