Systematic and random errors in compression testing of trabecular bone

We sought to quantify the systematic and random errors associated with end-artifacts in the platens compression test for trabecular bone. Our hypothesis was that while errors may depend on anatomic site, they do not depend on apparent density and therefore have substantial random components. Trabecular bone specimens were first tested nondestructively using newly developed accurate protocols and then were tested again using the platens compression test. Percentage differences in modulus between the techniques (bovine proximal tibia [n = 18] and humerus [n = 17] and human lumbar spine, [n = 9]) were in the range of 4-86%. These differences did not depend on anatomic site (p = 0.21) and were only weakly dependent on apparent density and specimen aspect ratio (r(2) < 0.10). The mean percentage difference in modulus was 32.6%, representing the systematic component of the end-artifact error. Neglecting the minor variations explained by density and specimen size (approximately 10%), an upper bound on the random error from end-artifacts in this experiment was taken as the SD of the modulus difference (+/-18.2%). Based on a synthesis of data taken from this study and from the literature, we concluded that the systematic underestimation error in the platens compression test can be only approximated and is in the range of 20-40%; the substantial random error (+/-12.5%) confounds correction, particularly when the sample size is small. These errors should be considered when interpreting results from the platens test, and more accurate testing techniques should be used when such errors are not acceptable.