Evaluation of Radiographic Predictors of Left Heart Enlargement in Dogs

We sought to determine the diagnostic accuracy of vertebral left atrial size (VLAS), vertebral heart size (VHS), and radiologists’ assessment for predicting left atrial enlargement (radLAE) in a large population of dogs with and without echocardiographic left atrial enlargement (echoLAE).   Dogs that had an echocardiogram and thoracic radiographs performed within 24-hours were consecutively enrolled provided they did not have pericardial effusion or a cardiac mass. EchoLAE was defined by a left atrium to aortic ratio acquired from short-axis (LA/AoSx) and long-axis (LA/AoLx) >1.68 and >2.54, respectively. Radiographic assessments were blinded to echocardiographic assessments and vice versa. Cutoffs for VLAS and VHS were generated using receiver operating characteristic analyses. Logistic regression was utilized to identify associations with echoLAE.   One-hundred and eighty-three dogs were enrolled. Dogs were diagnosed with myxomatous mitral valve disease (56%; 35 stage B1, 35 stage B2, and 33 stage C/D), other cardiovascular diseases (25%) or no cardiovascular disease (19%). Prevalence of EchoLAE was 63%. Optimal cutoffs for VLAS ( >2.3 vertebrae; sensitivity [Sn] 88.9%; specificity [Sp] 73.3%) and VHS ( >11.1 vertebrae; Sn = 74.6%; Sp = 75.8%) were generated and compared with radLAE (yes/no). VLAS >2.3 (area under the curve [AUC] = 0.81) was more accurate than VHS >11.1 (AUC = 0.75) and radLAE (AUC = 0.76) for predicting echoLAE. VLAS >2.3 and radLAE (both P < 0.001) were independently associated with echoLAE but VHS >11.1 was not (P = 0.054).   VLAS is a useful radiographic measurement for predicting echoLAE, particularly if echocardiography or a radiologist’s assessment are unavailable.