Osteochondritis Dissecans of the Knee: an Interrater Reliability Study of Magnetic Resonance Imaging Characteristics.

Background: Imaging characteristics of osteochondritis dissecans (OCD) lesions quantified by magnetic resonance imaging (MRI) are often used to inform treatment and prognosis. However, the interrater reliability of clinician-driven MRI-based assessment of OCD lesions is not well documented. Purpose: To determine the interrater reliability of several historical and novel MRI-derived characteristics of OCD of the knee in children. Study Design: Cohort study (diagnosis); Level of evidence, 3 Methods: A total of 42 OCD lesions were evaluated by 10 fellowship-trained orthopaedic surgeons using 31 different MRI characteristics, characterizing lesion size and location, condylar size, cartilage status, the interface between parent and progeny bone, and features of both the parent and the progeny bone. Interrater reliability was determined via intraclass correlation coefficients (ICCs) with 2-way random modeling, Fleiss kappa, or Krippendorff alpha as appropriate for each variable. Results: Raters were reliable when the lesion was measured in the coronal plane (ICC, 0.77). Almost perfect agreement was achieved for condylar size (ICC, 0.93), substantial agreement for physeal patency (ICC, 0.79), and moderate agreement for joint effusion (ICC, 0.56) and cartilage status (ICC, 0.50). Overall, raters showed significant variability regarding interface characteristics (ICC, 0.25), progeny (ICC range, 0.03 to 0.62), and parent bone measurements and qualities (ICC range, –0.02 to 0.65), with reliability being moderate at best for these measurements. Conclusion: This multicenter study determined the interrater reliability of MRI characteristics of OCD lesions in children. Although several measurements provided acceptable reliability, many MRI features of OCD that inform treatment decisions were unreliable. Further work will be needed to refine the unreliable characteristics and to assess the ability of those reliable characteristics to predict clinical lesion instability and prognosis.