Interobserver variability of vertebral heart size measurements in dogs with normal and enlarged hearts.

The vertebral heart size (VHS) method by Buchanan is based on anatomic landmarks. A potential source of variation among observers is differences in the selection of measurement points. The objective was to test variability between observers with different levels of training in thoracic radiology and small animal clinical practice. Fifty sets of thoracic radiographs of cavalier King Charles spaniels, were divided into five groups; (Normal) normal cardiopulmonary structures, (I) slight cardiomegaly, (II) moderate cardiomegaly, (II +) moderate cardiomegaly with congestive heart failure, and (III +) severe cardiomegaly with congestive heart failure. Cardiomegaly was confirmed by echocardiography to be caused by mitral regurgitation because of myxomatous mitral valve disease. Sixteen observers representing four levels of experience (four observers/level) evaluated the radiographs; (1) European Diplomates in Veterinary Diagnostic Imaging, (2) Experienced small animal clinicians, (3) Trainees in small animal clinical practice (4) Veterinary students. Almost identical mean VHS values were found among the four experience levels for each of the five groups of radiographs with a low coefficient of variation, range 1.5-3.2%. Mean difference among the 16 observers was 1.05 +/- 0.32 vertebrae (v). Mean difference among individuals in each observer group was approximately 0.5 v for all but the groups of trainees were the difference was 0.6-0.9 v. The conclusion is that VHS method for heart size is independent of observer experience but dependent of individual observers selection of reference points and transformation of long and short axis dimensions into VHS units.