[Site-specific Hounsfield units predict lumbar cage subsidence using predict]

To investigate and validate the utility of site-specific Hounsfield unit(HU) measurements at the posterior endplate (Index-Post) for predicting cage subsidence after lumbar fusion, and to evaluate its performance when combined with the Vertebral Bone Quality (VBQ) score.

A retrospective matched case-control study was conducted. Ninety patients who underwent single-level transforaminal lumbar interbody fusion (TLIF) were allocated into a subsidence group (n=30) and a non-subsidence group (n=60), matched for age, sex, and surgical level. Preoperative CT was used to measure site-specific HU values (Index-Post, Index-Mid, Index-Ant) and traditional sites (L₁ center, L₁-L₄ average). VBQ scores were calculated from preoperative MRI. Receiver operating characteristic(ROC) analysis, logistic regression, and decision curve analysis (DCA) were performed to assess predictive performance.

The incidence of cage subsidence was 33.3% (30/90). The Index-Post HU in the subsidence group (96.5±25.8) HU was significantly lower than that in the non-subsidence group (123.4±29.1) HU (P<0.001, Cohen's d=0.98). ROC analysis showed that the Index-Post HU had an area under the curve (AUC) of 0.78 [95% CI(0.69, 0.87)], which was superior to the traditional L1 vertebral center HU (AUC=0.65;DeLong test, P=0.012). The optimal cutoff value was 112 HU, yielding a sensitivity of 70.0% and specificity of 68.3%. The composite model (HU+VBQ) achieved an AUC of 0.83. Multivariate logistic regression identified Index-Post HU (OR=0.965, P=0.006) and age (OR=1.045, P=0.015) as independent predictors. Decision curve analysis demonstrated a net benefit within the threshold probability range of 20% to 50%.

Site-specific HU measurement at the posterior endplate is superior to traditional methods for predicting cage subsidence following TLIF. Combining this metric with the VBQ score further enhances predictive efficacy. However, standardized measurement protocols and multicenter validation are still required to confirm these findings and establish their clinical utility.