The Reliability to Determine “Healing” in Osteochondritis Dissecans From Radiographic Assessment

Abstract

Current literature supporting treatment strategies for osteochondritis dissecans (OCD) has limited prognostic utility. Presently, there is no gold standard for an OCD "healing" outcome. Accurate and reliable determination of the OCD healed status on the basis of radiographs would be valuable to provide a guide for evidence-based treatment of OCD. The purpose of this study was to determine interrater and intrarater reliability of radiographic assessment of OCD lesion healing in the knee. A hospital database identified 39 consecutive patients with OCD in 42 knees, who were treated conservatively for at least 6 months. Patients were an average age of 11.9 years (±2.4 y). A total of 102 presentation slides were prepared, each containing 2 similar radiographic views from the same patient taken 6 months (±1 mo) apart. These slides were then categorized by 4 reviewers (1 orthopaedic surgeon, 1 orthopaedic fellow, and 2 musculoskeletal radiologists) blinded to patient data, as healed, not healed, or unable to evaluate the OCD lesion. Reviewers repeated their assessment at a minimum of 3 weeks after their first read. Intrarater and interrater reliability were measured with the Cohen κ coefficient and Randolph's free-marginal multirater κ, respectively, and with the percent agreement. OCD lesion categorization demonstrated insufficient interrater reliability with a κ of 0.44 and 63% agreement. The notch view had the highest interrater reliability with a κ of 0.63% and 76% agreement, and the lateral view had the lowest interrater reliability with a κ of 0.29% and 53% agreement. The OCD lesion categorization demonstrated substantial intrarater reliability with a κ of 0.68% and 81% agreement. Reviewers did not consistently agree on the "healing" status of the OCD lesions on the basis of radiographic assessments. Standard criteria to assess healing are needed to consistently evaluate OCD knee lesions in children. Not applicable.