Abstract
We read with interest the study by Du et al. comparing CT-based pedicle Hounsfield units (HU) with MRI-based pedicle bone quality (PBQ) scores for predicting pedicle screw loosening (PSL) following posterior lumbar interbody fusion. 1 The authors should be commended for addressing an important question, as reliable preoperative prediction of PSL remains a clinical challenge. While the findings suggest HU values may be more predictive than PBQ, we believe some additional considerations warrant attention.
First, the follow-up period of 12 months may underestimate the true incidence and clinical significance of screw loosening. PSL can continue to appear beyond the first postoperative year, particularly in osteoporotic patients, and later-occurring loosening is often more clinically meaningful. 2 A longer follow-up would help clarify the durability of the predictive value of HU and PBQ measurements.
Second, the proposed HU cutoff values (106-110) raise concerns regarding generalizability. Hounsfield units are influenced by multiple technical factors, including CT scanner model, acquisition parameters, reconstruction algorithm, and slice thickness. Without cross-calibration, applying a single cutoff across centres and equipment may be unreliable. 3 Multicenter validation with standardised imaging protocols would strengthen external applicability.
Third, the statistical modelling approach deserves further scrutiny. By analysing CT and MRI predictors in separate regression models, the study cannot fully establish their relative contribution to PSL risk. A combined multivariate analysis including HU, PBQ, and clinical covariates would allow direct comparison within the same framework and might reveal additive or complementary predictive value. 4
Taken together, these points suggest that while Du et al provide valuable data supporting the use of CT-based HU, caution is required before adopting specific thresholds or concluding superiority over MRI-based methods. Longer-term follow-up, multicenter reproducibility studies, and integrated multivariate modelling will be essential to confirm and extend these promising results.
