Direct Comparison of the Precision of the New Hologic Horizon Model With the Old Discovery Model

Previous publications suggested that the precision of the new Hologic Horizon densitometer might be better than that of the previous Discovery model, but these observations were confounded by not using the same participants and technologists on both densitometers. We sought to study this issue methodically by measuring in vivo precision in both densitometers using the same patients and technologists. Precision studies for the Horizon and Discovery models were done by acquiring spine, hip, and forearm bone mineral density twice on 30 participants. The set of 4 scans on each participant (2 on the Discovery, 2 on the Horizon) was acquired by the same technologist using the same scanning mode. The pairs of data were used to calculate the least significant change according to the International Society for Clinical Densitometry guidelines. The significance of the difference between least significant changes was assessed using a Wilcoxon signed-rank test of the difference between the mean square error of the absolute value of the differences between paired measurements on the Discovery (Δ-Discovery) and the mean square error of the absolute value of the differences between paired measurements on the Horizon (Δ-Horizon). At virtually all anatomic sites, there was a nonsignificant trend for the precision to be better for the Horizon than for the Discovery. As more vertebrae were excluded from analysis, the precision deteriorated on both densitometers. The precision between densitometers was almost identical when reporting only 1 vertebral body. (1) There was a nonsignificant trend for greater precision on the new Hologic Horizon compared with the older Discovery model. (2) The difference in precision of the spine bone mineral density between the Horizon and the Discovery models decreases as fewer vertebrae are included. (3) These findings are substantially similar to previously published results which had not controlled as well for confounding from using different subjects and technologists.