Comparative analysis of two modified TLICS systems in guiding surgical decision-making for thoracolumbar fractures

To compare the Novel Modified Thoracolumbar Injury Classification System (NmTLICS) and the China-modified Thoracolumbar Injury Classification System (CN-mTLICS) in guiding surgical decision-making for thoracolumbar fractures.

We retrospectively analyzed the complete imaging data of 101 patients with single-level thoracolumbar fractures admitted between January 2021 and December 2023. Two independent observers scored the patients using the traditional TLICS, NmTLICS, and CN-mTLICS systems. Using the actual clinical treatment decision as the "gold standard," inter-observer agreement was assessed with Cohen's Kappa coefficient. Discrimination was evaluated by the area under the ROC curve (AUC) with pairwise DeLong comparison; classification accuracy among the three related systems was compared with a global Cochran's Q-test followed by post hoc McNemar tests with Bonferroni correction; and model calibration and clinical utility were assessed by calibration analysis and decision curve analysis. Discrepant cases between the two modified systems were further analyzed.

All three scoring systems demonstrated excellent inter-observer agreement, with NmTLICS (κ = 0.896) and CN-mTLICS (κ = 0.866) outperforming traditional TLICS (κ = 0.849). Regarding diagnostic efficacy, both NmTLICS and CN-mTLICS showed significantly higher sensitivity and NPV than traditional TLICS (P < 0.01). They successfully identified 17 and 16 patients, respectively, who were recommended for conservative treatment by TLICS but actually required surgery. There was no statistically significant difference in specificity compared with TLICS (P > 0.05). Discrepancy analysis revealed significant clinical complementarity between the two modified systems: NmTLICS detected 6 cases of severe vertebral collapse (>50%) missed by CN-mTLICS, while CN-mTLICS identified 5 cases of severe intervertebral disc injury missed by NmTLICS.

Both NmTLICS and CN-mTLICS reduced the under-triage of severe burst fractures by TLICS, mainly through higher sensitivity and negative predictive value without a significant loss of specificity, and they address complementary dimensions (fracture morphology and intervertebral disc injury, respectively). As these are exploratory single-center findings based on treatment decisions rather than long-term outcomes, prospective multicenter validation is needed before routine clinical application.