Letter to the editor: Single-slice Hounsfield unit measurement for predicting pedicle screw loosening: methodological considerations
Article information
Dear Editor,
I read with interest the article “Single-slice computed tomography-derived pedicle screw trajectory Hounsfield unit values predict screw loosening after lumbar fusion: a retrospective study in Japan,” by Yamamoto et al. [1], published in the Asian Spine Journal 2026;20(3):498-505; online 18 November 2025). The authors propose a simplified, single-slice computed tomography (CT) method for estimating bone quality along the planned pedicle screw trajectory using Hounsfield units (HU), a computed-tomography-derived measure that has been well validated against bone mineral density and strength [2]. This approach could meaningfully reduce the preoperative workload compared with multi-slice or volumetric HU assessment. Two methodological issues, in particular, bear directly on how the reported findings should be interpreted: the small number of loosening events relative to the complexity of the multivariable model, and the validity of representing trajectory-wide bone quality with a single CT slice.
First, the number of loosening events is small relative to the complexity of the multivariable model used. In this cohort of 47 patients, screw loosening was identified on follow-up CT in 11.8% of screws, corresponding to approximately five to six loosening events. Multivariate logistic regression conventionally requires an events-per-variable (EPV) ratio of at least ten to yield stable coefficient estimates [3]; with several covariates reportedly entered, including screw diameter and length, the achieved EPV in this analysis falls well short of that threshold, and the model carries a meaningful risk of overfitting. The reported cutoff of 123 HU should therefore be regarded as hypothesis-generating rather than a validated clinical threshold until it is confirmed in a larger, adequately powered cohort.
Second, the premise that a single CT slice can adequately represent bone quality along the entire screw trajectory requires further justification. Trabecular density varies considerably along the screw path, particularly between the pedicle isthmus and the vertebral body entry and tip, so a measurement taken at one level may not represent bone quality at the actual site of screw purchase. Studies using multi-level or volumetric HU sampling have reported cutoffs that are similar but not identical to the one proposed here (e.g., 117 HU in a multi-institutional minimally invasive transforaminal lumbar interbody fusion cohort) [4]. It remains unclear whether the single-slice measurement in the present study was obtained at a pre-specified, standardized anatomical level across all screws, and how reproducibly slice selection can be applied beyond the reported kappa statistic.
Several additional points are worth acknowledging but are not developed further here: the timing of follow-up CT used to define loosening, osteoporosis treatment, cage type and position, sagittal alignment correction, and the general need for validation in larger cohorts. These are limitations common to most retrospective observational studies of this kind [5], and are less specific to the present work than the two issues raised above.
These concerns do not diminish the clinical utility of a simplified screening method; rather, they suggest that external validation in a larger, multicenter cohort—with a pre-specified, anatomically standardized single-slice location and adjustment for osteoporosis treatment and construct-related variables—would considerably strengthen confidence in the proposed 123 HU threshold before its adoption into routine preoperative planning.
Notes
Conflict of Interest
No potential conflict of interest relevant to this article was reported.
Author Contributions
All authors conceived, drafted, critically revised, and approved the final version of this letter.
