Influence of experimental atlantoaxial ligament transection on CT-based measurements of atlantoaxial instability in small-breed dog cadavers

Introduction Static CT measurements obtained in standardized head–neck positions (HNPs) have been proposed for the diagnosis of atlantoaxial instability (AAI), mainly on the basis of retrospective clinical studies. The objective of this cadaveric study was to evaluate how experimentally induced atlantoaxial ligament disruption affects CT-based measurements of AAI within the same small-breed dogs, using each cadaver as its own control before and after sequential ligament transection. Materials and methods Sagittal reconstructed CT images of eight small-breed dog cadavers were obtained in extended, neutral, and maximally flexed HNPs before and after sequential transection of the atlantoaxial ligaments via the foramen magnum. This approach created an artificial, assumed instability model while preserving other craniovertebral junction structures. Group comparisons were performed using exploratory repeated-measures ANOVA without adjustment for multiple testing. Receiver operating characteristic (ROC) analyses were performed to explore the ability of selected measurements to discriminate between intact and completely transected ligament states. Results Several CT measurements changed after ligament transection, including the ventral compression index (VCI), basion-dens interval, atlantoaxial distance, C1-C2 overlap, and cranial translation ratio. The VCI showed the most consistent within-model discrimination. For differentiation between intact and completely transected ligament states, a model-derived VCI cutoff of ≥0.42 in neutral HNP yielded sensitivity and specificity estimates of 100% (95% CI for each, 63.1–100%); in extended HNP, a cutoff of ≥0.36 yielded sensitivity and specificity estimates of 87.5% (95% CI for each, 47.3–99.7%). Discussion and conclusion In this experimental cadaveric model, VCI appeared to be the most promising single CT measurement for distinguishing intact from completely transected AA ligament states. Neutral HNP provided the highest point estimates of within-model discrimination, whereas extended HNP yielded slightly lower point estimates. The findings should be interpreted in light of the artificial nature of the instability model and the small number of cadavers.