Towards Standardization of Dual X-ray Absorptiometry (DXA) at the Forearm

Abstract

Manufacturer-implemented regions of interest (ROIs) to determine the bone mineral density (BMD) at the forearm are currently not standardized across dual X-ray absorptiometry (DXA) devices. We hypothesized that their differences introduce considerable variation in measurement results for forearm BMD when taken on different devices, and that a ROIs common to all devices with standardized placement and size significantly improve device comparability. The common ROI was defined to have a fixed length of 2 cm and to extend proximally from the location where the ulna and radius bones superimpose on the DXA image. The effects of universal standardization of forearm BMD were combined with and compared to those of the common ROI. They were drawn on 91 female study participants (ages 20-80 years, 10 per decade) who were scanned on Hologic QDR-4500, Aloka DCS-600EX, GE Lunar PIXI and Norland pDEXA DXA scanners. For all device combinations, manufacturer-implemented ROI root mean-square errors were significantly higher than for the common ROI, suggesting that implementing an ROI with common design on all scanners is a good way to reduce interdevice variability. When manufacturer-implemented ROIs were universally standardized root mean-square error (RMSE) values were less different from that of the nonstandardized Common ROI, suggesting that universal standardization can further improve interdevice comparability even when a common ROI such as the one implemented here is used. ROI standardization dramatically improves interdevice comparability.