Validity and Reliability of a Peripheral Quantitative Computed Tomography Scanner for Measuring Muscle Cross-sectional Area

Abstract

PURPOSE: To quantify the validity and test-retest reliability of muscle cross-section area (CSA) measures from a peripheral quantitative computed tomography (pQCT) scanner compared to magnetic resonance (MR) images. METHODS: Four women (age ± SD = 32 ± 9 yrs; stature = 171 ± 11 cm; mass = 66 ± 15 kg) and 7 men (32 ± 9 yrs; 177 ± 7 cm; 85 ± 20 kg) had axial MR images taken with a 1.5T short bore magnet (Signa, General Electric). Within 10 days of the MR scans, 2 separate axial pQCT images were taken at the same time of day separated by 2–3 d with a low-radiation peripheral CT scanner (XCT 3000, Orthometrix). Images were acquired by manually aligning the MR and pQCT scanners with marks placed at the midpoint between the greater trochanter and lateral epicondyle of the femur. Two separate investigators used DicomWorks v1.3.5 software to trace the quadriceps muscles (cm2). Coefficients of determination (R2), standard errors of estimate (SEE), constant errors (CE), total errors (TE), and limits of agreement (LOA) were used to examine validity, whereas test-retest reliability was assessed with intra-class correlation coefficients (ICC), standard errors of measurement (SEM), minimum differences (MD) to be considered real, and repeated measures ANOVAs. RESULTS: Muscle CSA values from the two investigators were (mean ± SD) 77.0 ± 24.2 and 77.5 ± 23.9 cm2 for the MR images and 78.9 ± 25.2 and 79.6 ± 24.2 cm2 from the first pQCT trial. For validity, R2 = 0.979 and 0.983, SEE = 3.677 and 3.297 cm2, CE= 1.993 and2.133cm2, TE = 4.058 and 3.735 cm2, and the bias ±95% confidence intervals (LOA) = −1.405 ± 6.815 and −2.391 ± 5.895 cm2. For test-retest reliability, ICC = 0.996 and 0.998, SEM = 1.660 and 1.101 cm2, MD = 4.601 and 3.051 cm2, and there were no differences between the two pQCT trials (P = 0.127 and 0.296). CONCLUSIONS: The R2 values were high, but the TE values were approximately 5% of the mean CSA. The LOA bias as well as the CE values indicated that the pQCT was consistently overestimating muscle CSA by approximately 2.5%, which may have been due to subject placement within the MR versus pQCT scanners. Test-retest reliability was high (>0.99), with low measurement errors (approximately 2%), however, the MD values indicated that a 4-5% change in muscle CSA may be required to be considered real. pQCT may provide adequately valid and reliable measures of muscle CSA.