Bias and agreement for radiogrammetric estimates of cortical bone geometry.

RATIONALE AND OBJECTIVES Radiogrammetry of the second metacarpal assumes a circular or elliptical model, which does not conform to reality. This study assesses the degree of bias and limits of agreement (error) resulting from deviations from these assumed models for estimates of bone mass and strength. METHODS Forty-six left metacarpals were radiographed in orthogonal posteroanterior and mediolateral views for calculation of algebraic estimations of area and bending moments of area. The bones were then sectioned at midshaft and digitized for determination of actual values of these properties. Bias was determined as the mean difference between methods, and error as +/-2 standard deviation. RESULTS Radiogrammetric methods significantly (P < 0.05) overestimate actual values for measures of cortical area and bending strength. Elliptical models are less biased and prone to smaller ranges of error than are uniplanar circular models. CONCLUSIONS Radiogrammetric estimation of metacarpal bone mass is subject to considerable method error arising from use of an overly simplistic circular model for its midshaft geometry.