Accuracy of CT hounsfield units for predicting cage subsidence and pedicle screw loosening after lumbar interbody fusion: a systematic review and meta-analysis.

Abstract

Cage subsidence (CS) and pedicle screw loosening (PSL) are significant postoperative complications following lumbar interbody fusion (LIF) surgeries, often associated with poor bone quality. Hounsfield Unit (HU) measurements from preoperative CT scans have emerged as a potential non-invasive predictor of bone density. This study aims to evaluate the diagnostic accuracy of HU in predicting CS and PSL. A systematic review and meta-analysis were conducted according to PRISMA guidelines. Databases searched included PubMed, Europe PMC, ScienceDirect, and Google Scholar. Studies with irrelevant outcomes, review articles, case reports, case series, animal studies, cadaveric studies, and abstract only articles were excluded. Pooled estimates of standardized mean difference (SMD), sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were calculated using a random-effects model. Risk of bias was assessed using QUADAS-2, and the GRADE framework was used to evaluate evidence quality. Seventeen retrospective cohort studies involving 2,294 patients were included. HU values were significantly lower in patients with complications. For CS, the pooled SMD was - 37.43 [95% CI: -44.80 to -30.08], and for PSL, -24.55 [95% CI: -29.04 to -20.06]. The pooled sensitivity and specificity of HU for predicting CS were 0.83 [95% CI: 0.71-0.90] and 0.86 [95% CI: 0.75-0.93], with an AUC of 0.91. For PSL, the pooled sensitivity and specificity were 0.80 [95% CI: 0.72-0.86] and 0.68 [95% CI: 0.56-0.78], with an AUC of 0.81. GRADE assessment rated the evidence as moderate quality. HU values are reliable predictors of CS and PSL after LIF, with stronger accuracy for CS, that may serve as a practical tool in preoperative planning to identify high-risk patients and reduce complication rates. However, the retrospective nature and methodological variability among included studies may affect generalizability, that warrants for further large scale, prospective studies.