Inter- and Intraobserver Reliability of CT Assessment of Canine Elbow Sclerosis in the Absence of Elbow Pathology

Abstract To establish a highly standardized elbow computed tomography (CT) examination protocol and evaluate inter- and intraobserver reliability of assessment of canine elbow sclerosis in the absence of elbow pathology. Dogs that presented for CT screening for elbow dysplasia prior to breeding were evaluated. Only dogs diagnosed with CT as being free of elbow dysplasia were included. The CT images were randomized and assessed by five blinded observers. Two observers re-reviewed 30 randomly selected studies on a second occasion. A standardized approach to CT image reconstruction to create consistent image planes was used. Sclerosis was subjectively as graded 0 (no sclerosis), 1 (mild sclerosis), and 2 (marked sclerosis). Objective assessment of sclerosis using Hounsfield units (HU) was measured within a standardized region of interest at the medial aspect of the humeral condyle (MAHC) and medial coronoid process (MCP). Complete agreement between observers in ordinal scoring of sclerosis was moderate for each region (MCP traditional 38.8%, MCP modified 28.6%, MAHC 26.5%). Intraobserver reliability of ordinal sclerosis scoring was poor to moderate. The inter- and intraobserver reliability of HU measurements was good to moderate, and good to excellent, respectively. Correlation between sclerosis scores and HU measurements was low to moderate. Subjective and objective assessment of sclerosis of the MCP and MAHC in the absence of elbow pathology is unreliable despite the use of a highly standardized protocol.