Affiliation: | aBrain Research Laboratories, Department of Psychiatry, New York University School of Medicine, New York, NY 10016 bDivision of Endocrinology, Department of Medicine, University of Connecticut School of Medicine, Farmington CT 06030 cClaude Pepper Older Americans Independence Center, University of Connecticut Health Center, Farmington, CT dDepartment of Behavioral Sciences and Community Health, University of Connecticut Health Center, Farmington, CT eDivision of Biostatistics, Department of Community Medicine and Health Care, University of Connecticut School of Medicine, Farmington, CT 06030 |
Abstract: | Dual-energy X-ray absorptiometry (DXA) is widely used for bone mineral density (BMD) measurements. Prospective daily quality control procedures such as cumulative sum (CUSUM) plots and Shewhart charts are very important and commonly used for routine monitoring of DXA measurements. These procedures are less suitable for post hoc adjustment of DXA measurements for clinical research studies, if and when that is needed. Extending previous methods, we propose and illustrate a simple statistical method for retrospective quality evaluation that may be used to adjust BMD measurements before they are analyzed in clinical research studies. Using multivariate regression, this method allows for simultaneous adjustment of different types of temporal variation such as sudden jumps, simple linear trends, changes in the slope of these trends, quadratic terms, and seasonal fluctuations. Adjusting the measurements to account for different types of temporal trends decreases the variance of bone mass density measurements and is an inexpensive way to increase the power of a study without increasing the sample size. The method is illustrated with measurements of two different phantoms used on the same DXA, with 2 and 3 yr of data, respectively. Results were consistent for both phantoms, with a gradual trend and a sudden jump, as well as a seasonal fluctuation term. |