The HBA-3 score: a simple bedside tool to predict cage subsidence after TLIF

Cage subsidence is a common radiographic finding following transforaminal lumbar interbody fusion (TLIF), but its clinical relevance remains uncertain. Hounsfield Unit (HU) values from preoperative CT scans have emerged as surrogates for local bone quality, yet few studies have validated simple, clinically applicable prediction models or examined the relationship between subsidence and functional outcomes.

Retrospective analysis of 76 patients who underwent single-level TLIF (2019-2023) with preoperative CT imaging and 12-month follow-up. HU was measured at the fused vertebral bodies. Cage subsidence was defined as ≥ 2 mm vertical migration or disc height loss on CT at 6-12 months. Logistic regression, Random Forest, and XGBoost models were applied to identify key predictors. A three-point bedside score (HBA-3) was derived using ROC/Youden thresholds for HU (< 131), BMI (> 30), and age (≥ 60). Outcomes included Oswestry Disability Index (ODI), EQ-5D VAS, and patient satisfaction.

The overall subsidence rate was 31.6% (24/76). HU, BMI, and age consistently emerged as the strongest predictors across all models, while sex, smoking, alcohol, and cage type contributed minimal signal. The HBA-3 score stratified subsidence risk stepwise from 12.5% (0 points) to 50.0% (3 points), with good discrimination (cross-validated AUC = 0.73; logistic 0.70, RF 0.73, XGBoost 0.72). Inter-rater reliability of HU was excellent (ICC = 0.98). Despite these radiographic findings, subsidence was not associated with worse PROMs at 12 months (satisfaction p = 0.29; ODI p = 0.66; EQ-5D VAS p = 0.87). Higher HBA-3 scores were not associated with ODI at 12 months (p = 0.45) or ODI MCID attainment (p = 0.25). ODI improved significantly at 12 months (- 12.0, p < 0.0001), while EQ-5D showed no significant change.

The HBA-3 score (HU, BMI, age) provides a simple, reproducible tool for estimating subsidence risk after TLIF, with stepwise increases in risk across strata. However, radiographic subsidence did not predict functional recovery in this single-level degenerative cohort, underscoring the need to distinguish between radiographic and patient-centred outcomes.