Pedicle screw loosening after degenerative lumbar fusion: a systematic review and meta-analysis of preoperative imaging-derived bone quality and paraspinal muscle metrics

Abstract

Radiographic pedicle screw loosening (PSL) remains a frequent postoperative finding and potential marker of screw-bone interface compromise after degenerative lumbar fusion, yet preoperative risk stratification remains inconsistent. This systematic review and meta-analysis aims to compare the association and discriminative performance of preoperative imaging-derived bone quality and paraspinal muscle metrics for predicting radiographic PSL after degenerative lumbar fusion. PubMed, Embase, and CENTRAL were searched until January 2026. Observational studies reporting radiographic PSL and evaluating preoperative imaging-based bone or muscle metrics were included. Bone metrics comprised CT-derived Hounsfield units (HU), CT-derived bone mineral density (CT-BMD), MRI-based vertebral bone quality (VBQ), and DEXA T-scores. Muscle measures included psoas and posterior paraspinal cross-sectional area (CSA) and composition. Radiographic PSL definitions and adjudication modalities (plain radiograph versus CT) varied across included studies and were considered explicitly during synthesis and interpretation. Continuous outcomes were synthesized using random-effects inverse-variance models. Diagnostic accuracy was evaluated using bivariate (Reitsma) generalized linear mixed models, with summary receiver operating characteristic (SROC)-derived area under the curve (AUC) used to quantify discrimination. Risk of bias was assessed using QUADAS-2. Twenty-nine studies (4711 patients; 25.4% PSL) were included. Patients with PSL demonstrated poorer preoperative bone quality across modalities, including lower HU (MD -24.7; P  < 0.01), lower CT-BMD (MD -28.05; P  < 0.05), lower DEXA T-scores (MD -0.18; P  < 0.01), higher VBQ (MD 0.46; P  < 0.01). Diagnostic discrimination was moderate and comparable across metrics (SROC-derived pooled AUC 0.71–0.76; meta-regression P  = 0.08), although pooled estimates reflect a composite radiographic endpoint with heterogeneous adjudication conventions rather than a single uniform reference standard. Among muscle measures, reduced psoas CSA showed consistent association with PSL, whereas posterior paraspinal metrics were heterogeneous and largely non-significant. Nomogram analyses demonstrated clinically meaningful risk reclassification but limited rule-in and rule-out capacity. Imaging-derived bone quality metrics identify substrate-level vulnerability to radiographic PSL and provide moderate preoperative discrimination. Because pooled discrimination reflects a composite radiographic endpoint rather than a uniform reference standard, these metrics are moderate risk stratifiers for radiographic PSL, not validated stand-alone tests for clinically meaningful construct failure. Limited and heterogeneous evidence suggests psoas size may represent a complementary frailty marker. Prospective validation and standardized thresholds are required.