Objective: We investigated the long-term in vivo reproducibility of quantitative CT (QCT) examinations that were conducted in conformity with a standard and well established methodology.
Materials and methods: The long-term reproducibility in vivo of QCT vertebral densitometry was studied in 12 normal postmenopausal women (mean age 51 years), who underwent four to five examinations over the same 2 year period.
Results: One group of six patients demonstrated good reproducibility with a coefficient of variation (CV) of bone mineral density (BMD) of < 2.4% and were considered "good cases." The other group of six patients showed poor reproducibility with a CV of BMD of > 3.2% and were considered "problem cases." A statistical study of the technical parameters of the QCT examination was performed to determine and correct for the factors that are correlated with reproducibility errors. Analyzing uncalibrated variations in Hounsfield numbers, we found that the surrounding soft tissues such as muscle, aorta, liver, and fat were correlated in both the good and the problem cases. However, only in the good cases did we find that the Hounsfield numbers of the vertebral bodies correlated strongly with the soft tissues and with parameters characteristic of the calibration regression line.
Conclusion: Essentially, the calibration procedure appeared to fail in some examinations of the problem cases, causing poor long-term reproducibility. This calibration failure may be related to positioning of the subject in the gantry and to variations in electrical parameters of the X-ray tube. A CV > 1% between the four slopes of calibration for the individual vertebrae measured at the same visit may indicate calibration problems and may suggest the need for repeat scanning.