Differences between dual X-ray absorptiometry using pencil beam and fan beam modes and their determinants in vivo and in vitro

The aim of this study was to determine the influence of individual factors on differences in bone mineral density (BMD) using dual X-ray absorptiometry pencil beam (PB) and fan beam (FB) modes in vivo and in vitro. PB.BMD and FB.BMD of 63 normal Caucasian females ages 21-80 yr were measured at the lumbar spine and hip. Residuals of the FB/PB regression were used to assess the impact of height, weight, adiposity index (AI) (=weight/height(3/2)). back tissue thickness, and PB.BMD, respectively, on FB/PB difference. The Hologic Anthropomorphic Spine Phantom (ASP) was measured using the PB and FB modes at two different levels to assess the impact of scanning mode and focus distance. The European Spine Phantom (ESP) prototype, a geometrically well-defined phantom with known vertebral densities, was measured using PB and FB modes and analyzed manually to determine the impact of bone density on FB/PB difference and automatically to determine the impact of edge detection on FB/PB difference.. Population BMD results were perfectly correlated, but significantly overestimated by 1.5% at the lumbar spine and underestimated by 0.7% at the neck, 1.8% at the trochanter, and 2.0% at the total hip, respectively, when using the FB compared with PB mode. At the lumbar spine, the FB/PB residual correlated negatively with height (r = 0.34, p < 0.01) and PB.BMD (r = 0.48, p < 0.0001) and positively with Al (r = 0.26, p < 0.05). At the hip, residual of trochanter correlated positively with weight (r = 0.36, p < 0.01) and Al (r = 0.36, p < 0.01). The FB mode significantly increased ASP BMD by 0.7% compared with PB. Using the FB mode, increasing focus distance significantly (p < 0.001) decreased area and bone mineral content, but not BMD. By contrast, increasing focus distance significantly decreased PB,BMD by 0.7%. With the ESP, the PB mode supplied accurate projected area of the bone (AREA) results but significant underestimation of specified BMD in the manual analysis. The FB mode significantly underestimated PB.AREA by 2.9% but fitted specified BMD quite well. FB/PB overestimation was larger for the low-density (+8.7%) than for the high-density vertebra (+4.9%). The automated analysis resulted in more than 14% underestimation of PB.AREA (low-density vertebra) and an almost 13% overestimation of PB.BMD (high-density vertebra) using FB, In conclusion, FB and PB measurements are highly correlated at the lumbar spine and hip with small but significant BMD differences related to height, adiposity, and BMD, In clinical practice, it can be erroneous to switch fi om one method to another, especially in women with low bone density.