Findings and interobserver agreement in radiography and ultrasonography of the vertebral column of a large population of normally performing horses

Summary

Diagnostic imaging is a mainstay in the investigation of equine neck and back pain, but interpretation of radiographic and ultrasonographic findings in the spinal column of horses is not straightforward for a variety of reasons including individual anatomical variations, progressive degenerative nature of most pathologies and superimposition as well as technical limits. These issues are even more relevant in the context of examinations of apparently sound horses with absent, mild or unclear complaints, like it is often the case in pre-purchase examinations or cases of failure to meet expected performance. The first aim was to report on the spectrum, degree and location of first-line imaging findings in the spine of a large population of normally performing horses. Limited data is available about agreement of interpretation of equine vertebral column imaging by radiologists. The second aim of this prospective study was to determine interobserver agreement on radiographic and ultrasonographic diagnostic imaging findings in the vertebral column of the same population between multiple observers with longstanding experience in equine diagnostic imaging at two different institutions. Seventy-one horses randomly selected from a larger population of 250 normally performing horses participating in a swiss project on equine back health were examined at one referral center. Radiographic and ultrasonographic examinations were performed in a standardized fashion and images graded separately by two experienced radiologists at two different institutions. Focus was placed on osteoarthrosis of the synovial intervertebral articulations (SIVAs), impinging and overriding of the spinous processes (SPs) and spondylosis. Cohen’s weighted kappa was calculated for each pathology, location and segment in each modality. Interobserver agreement was calculated for findings at specific locations, single pathologies and single grades. Most horses showed no changes and were allocated grades 0 at most locations. Few abnormalities were found throughout the vertebral column, with clusters of abnormalities of the SIVAs in the caudal cervical segment and cranial lumbar segment as well as a cumulation of changes at the dorsal spinous processes in the caudal thoracic segment. These coincide with previously reported predilection sites of imaging findings in symptomatic populations. Overall mean value of agreement for imaging findings was moderate (k= 0.7). Agreement was moderate for imaging findings regarding the synovial intervertebral articulations of the thoracolumbar spine in radiography (k=0.66) but weak in ultrasonography (k=0.58). There was moderate agreement in the imaging findings of the cervical spine in ultrasonography (k=0.61) as well as radiography (k=0.62). Strong agreement was found in the radiographic assessment of changes of the thoracolumbar spinous processes (k=0.80). Almost perfect agreement was found in the radiographic assessment of thoracolumbar spondylosis (k=0.95). Agreement between radiologists in detection and grading of pathologies of the equine vertebral column is weak to almost perfect depending on pathology and location. Our results confirm findings about distribution of specific pathologies found in other studies and support the importance of interpreting imaging findings along with clinical findings for definitive case management and decision making. Additional studies are needed for determination of the correlation of imaging findings among different modalities and correlation of diagnostic imaging with clinical findings.