This is the first study in which the reproducibility of BSI is tested on a phantom, showing that this new software has higher CoV and LSC% compared to BMD, being about two to three times higher. As a consequence, BSI reproducibility was lower than that of BMD, and it ranged from 96.1% (6 cm soft tissue thickness, HD-mode) to 98.3% (1 cm soft tissue thickness, HD-mode). BMD reproducibility confirmed to be very good, but was affected by the increase of soft tissue thickness.
This is not the first phantom study aimed to evaluate the reproducibility of DXA measurements. The important aspect of these studies relies on the fact that phantoms are usually less affected by external factors, and this represents a methodological advantage to better understand the working principle of new techniques, such as the BSI. Previous studies have been published with data regarding BMD and TBS reproducibility [16,17,18, 21]. Concerning TBS, phantom studies showed that its reproducibility is somewhat lower compared to that of BMD, being between 96.4–98.3%  and 97.7–98.3% . Usually, BMD reproducibility is known to be very good and typically represents the standard of reference for other DXA-based measurements. This was confirmed by our study, as BMD showed high values of reproducibility being around 99% in all the three scan modalities. On the other hand, the reproducibility of BSI was lower than that of BMD, and this finding has implication in clinical practice especially for patient’s monitoring. From a clinical point of view, the lower is the reproducibility the longer has to be the time of follow-up to observe a clinically significant variation.
From a practical point of view, the increase in soft tissue thickness is a condition that is frequently found in clinical DXA routine, in patients with high values of BMI and waist circumference. In our study, we tried to understand the behaviour of this new technology by simulating this frequent clinical scenario, showing that BSI, despite having lower reproducibility, was equally affected as BMD. The increase in soft tissue thickness was associated to a reduction of precision, but this was statistically significant for both BMD and BSI only for the HD-mode. Thus, we found a detrimental effect of increasing soft tissue on BSI reproducibility, similar to that of BMD.
When considering BMD and BSI mean values difference between 0 and 6 cm, we notice again a similar behaviour of these two parameters, as a statistically significant variation was found for both for BMD and BSI, despite the first increased its value and the latter showed a decrease. This divergence is expected, as higher BSI values are associated to lower BMD values and vice versa. Nevertheless, the amplitude of variation between 0 and 6 cm was slightly higher for BSI compared to BMD, a difference that was more evident for the FA-mode.
This study presents some limitations. The first one is that we evaluated BSI reproducibility on a phantom that was designed for BMD quality control. Thus, the intrinsic value of BSI measurement may not be directly transferred to clinical practice, despite the evaluation of reproducibility remains independent from BSI significance. Another limitation is that we performed this study on a single phantom with a single densitometer. Thus, these results may present variations when applied to other settings, and the precision of this software may vary when a different phantom or densitometer is used, and of course, the precision may vary among patients with different degrees of bone quality. Lastly, we simulated fat tissue using a material (pork rind) which is similar but not equal to human fat; therefore, we may not directly apply these results to real BMI variations.
In conclusion, this phantom study assessed the reproducibility of BSI compared to that of BMD in different scan modes and interposed fat thicknesses, showing that BSI reproducibility was overall lower than that of BMD and only slightly negatively influenced by interposed fat thickness increase respect to BMD.