Agreement and precision of periprosthetic bone density measurements in micro-CT, single and dual energy CT.

The objective of this study was to test the precision and agreement between bone mineral density measurements performed in micro CT, single and dual energy computed tomography, to determine how the keV level influences density measurements and to assess the usefulness of quantitative dual energy computed tomography as a research tool for longitudinal studies aiming to measure bone loss adjacent to total hip replacements. Samples from 10 fresh-frozen porcine femoral heads were placed in a Perspex phantom and computed tomography was performed with two acquisition modes. Bone mineral density was calculated and compared with measurements derived from micro CT. Repeated scans and dual measurements were performed in order to measure between- and within-scan precision. Mean density difference between micro CT and single energy computed tomography was 72 mg HA/cm3 . For dual energy CT, the mean difference at 100 keV was 128 mg HA/cm3 while the mean difference at 110-140 keV ranged from -84 to -67 mg HA/cm3 compared with micro CT. Rescanning the samples resulted in a non-significant overall between-scan difference of 13 mg HA/cm3 . Bland-Altman limits of agreement were wide and intraclass correlation coefficients ranged from 0.29 to 0.72, while 95% confidence intervals covered almost the full possible range. Repeating the density measurements for within-scan precision resulted in ICCs >0.99 and narrow limits of agreement. Single and dual energy quantitative CT showed excellent within-scan precision, but poor between-scan precision. No significant density differences were found in dual energy quantitative CT at keV-levels above 110 keV.