Brought to you by

Reply to the Letter to the Editor: “Proposed diagnostic volumetric bone mineral density thresholds for osteoporosis and osteopenia at the cervicothoracic spine in correlation to the lumbar spine”

by Sebastian Rühling, Jan S. Kirschke (jan.kirschke@tum.de)

Proposed diagnostic volumetric bone mineral density thresholds for osteoporosis and osteopenia at the cervicothoracic spine in correlation to the lumbar spine

Dear Editor,

With great interest we have read the Letter to the Editor by Xiao Chen and colleagues regarding our publication “Proposed diagnostic volumetric bone mineral density thresholds for osteoporosis and osteopenia at the cervicothoracic spine in correlation to the lumbar spine” [1]. Here, we would like to address their valuable feedback.

One of the main goals of our study was to emphasise that volumetric bone mineral density (vBMD) at the cervicothoracic spine is significantly different compared to the lumbar spine. Moreover, differences are not only found between cervicothoracic and lumbar regions, but also for every level within the thoracic and cervical spine. In line with current recommendations, we agree with Chen and colleagues that vBMD values should be averaged over at least two vertebrae. However, the vBMD gradient at the cervicothoracic spine necessitates averaging with level-specific thresholds, in contrast to the lumbar spine where vBMD values are very similar for each vertebral level. It may not matter whether cervicothoracic vBMD values are converted to their “lumbar equivalent” values based on level specific conversion factors and averaged afterwards, or whether the level-specific cut-off values are averaged for the proposed levels. As stated by Chen and colleagues, the diagnostic performance (i.e., prediction of incident vertebral fractures) of such level-specific thoracolumbar vBMD thresholds goes beyond the purpose of the presented study and was in parallel investigated by our group in May 2022 [2] (see Fig. 2 and 3). As expected, thresholds for the thoracic spine were slightly decreased compared to the lumbar spine, yet highly significant.

In line with a previous commentary by Engelke and Keaveny, we fully agree with Chen and colleagues that different CT scanners substantially influence extracted HU (and possibly vBMD) measurements [3]. To obtain accurate results, we have indeed used a dedicated phantom (Anthropomorphic Abdomen Phantom, QRM Quality Assurance in Radiology and Medicine) for asynchronous calibration of all CT scanners in this study [4]. This allows for scanner-specific HU-to-BMD conversion equations [1]. Our group has recently shown in another study that this method has a small, acceptable residual error compared to dedicated QCT accuracy with a phantom within the same scan [5] and demonstrated that the established thresholds may be used with similar diagnostic accuracy to distinguish patients with and without osteoporotic fractures. In line with Chen and colleagues, we would like to emphasize that further prospective studies with larger sample sizes are required to validate the thresholds to a generalisable ground truth and to consider other valuable parameters, such as prevalent fractures.

As correctly pointed out by Chen and colleagues, the caption of Figure 4 is not consistent to the figure itself and the manuscript text. The colour descriptors (brown) and (green) were unfortunately mixed up in the figure caption (but not in the figure legend).

Finally, we would like to thank Chen and colleagues for their time and effort to validate our results. Their study is designed similar to the study by Sollmann et al. and similarly no patient crossed the threshold from healthy to osteoporotic and vice versa [5]. As vBMD is a continuous variable that is categorized in normal, osteopenic and osteoporotic by two hard cut-offs, already small errors around those cut-offs can naturally lead to a category shift. In such a setting, a kappa coefficient of >0.61 may be considered substantial. It is assumed that Chen and colleagues excluded vertebrae with degenerative changes; in this regard, it would be interesting to compare kappa values reported by Chen and colleagues with kappa values resulting from a comparison of different levels within the lumbar region itself. Additionally, it would be very informative to report, for example, the root-mean-square errors (RMSE) of thoracic vs. lumbar vBMD measurements or AUCs of different levels to differentiate patients with osteoporotic fractures from patients without.

Overall, we fully agree with Chen et al. that further prospective studies are needed to validate our proposed thresholds. However, we think that our study clearly pointed out that (1) different vBMD thresholds are needed for the cervicothoracic spine compared to the lumbar spine and that (2) high correlations between vBMD of different vertebral levels were found, confirming that a scan only visualizing the cervicothoracic spine may also be used for opportunistic diagnosis of osteoporosis.