Using serial trichotomization with multiple radiographic parameters to diagnose atlantoaxial instability in predisposed toy-breed dogs

Objective radiographic measurements for determining atlantoaxial instability (AAI) in dogs have been reported. However, choosing a single optimal parameter among different studies is challenging. This study aimed to identify optimal parameters for AAI in small-breed dogs by simultaneously evaluating the flexed and nonflexed lateral radiographs. Moreover, the potential of using serial trichotomization with multiple parameters was investigated. Radiographs of 10 dogs diagnosed with AAI on magnetic resonance imaging (MRI) were compared with 26 control dogs to select parameters with the area under the curve ≥0.9 or ≤0.1. Parameters with outstanding discrimination included cranial border distance and C1–C2 angle on nonflexed radiographs and dens length on ventrodorsal radiographs. Using cutoff values reflecting 100% sensitivity and 100% specificity, the subject was determined into AAI diagnosed, ruled-out or indeterminate. The accuracy rate was 100% for each of the parameters. However, 31–47% of dogs were classified as indeterminate. Serial trichotomization in the order of dens length, cranial border distance, and C1–C2 angle significantly reduced the proportion of dogs in the uncertainty zone while maintaining accuracy. The established method was tested on 19 new subjects. A similar trend of high accuracy (100%) and small uncertainty zone (5%) was observed in the validation cohort. For certain breeds, serial trichotomization with dens length on the ventrodorsal radiograph and cranial border distance and C1–C2 angle on the nonflexed lateral radiograph showed good potential for determining AAI and identifying dogs in which advanced imaging is indicated for diagnosis.