Study identifies prevalence estimates of clinically relevant canine thoracic radiographic findings and insights into interobserver variability

Objective To estimate the prevalence of 17 clinically relevant thoracic radiographic findings in dogs and quantify interobserver variability among veterinary radiologists, with the hypothesis that findings would vary widely in prevalence while maintaining high interobserver agreement. Methods 4,000 lateral thoracic radiographic examinations created between May and December 2023 were randomly selected from IDEXX Telemedicine Consultants’ electronic medical records. After exclusion, 2,761 examinations were ultimately evaluated for 17 preselected conditions by 5 board-certified veterinary radiologists, subsequently reduced to 3 to expedite the study. Prevalence for each condition was estimated using a latent class expectation-maximization approach. Results 2,761 studies were ultimately evaluated for 17 conditions by a total of 12 radiologists (1,367 studies evaluated in triplicate, and 1,394 studies evaluated in quintuplicate); the most prevalent conditions were abnormal pulmonary pattern (60.9%) and redundant tracheal membrane (29.4%). The least prevalent were heart base mass (0.5%) and esophageal foreign body (0.3%). Interobserver agreement was generally high for radiographically distinct conditions, such as esophageal foreign body, heart base mass, and pneumothorax, as each achieved an area under the receiver-operating characteristic curve of 1.00, with a modest decrease for subjective or morphologically ambiguous conditions, such as suspected tracheal chondromalacia (0.76) and cranial mediastinal widening as a breed variant/fat deposition (0.78). Conclusions Thoracic radiographic findings in dogs vary substantially in prevalence, with pulmonary abnormalities being the most common. Interobserver agreement is condition dependent, with the greatest variability among subjectively assessed findings. Clinical Relevance Prevalence estimates and agreement profiles provide a framework for clinical benchmarking and highlight categories where standardized criteria may improve consistency.