Interobserver and Intraobserver Reliability of an MRI-Based Classification System for Injuries to the Ulnar Collateral Ligament.
Ramkumar PN., Frangiamore SJ., Navarro SM., Lynch TS., Forney MC., Kaar SG., Akhavan S., Moutzouros V., Westermann RW., Farrow LD., Schickendantz MS.
BACKGROUND: Despite improvements in understanding biomechanics and surgical options for ulnar collateral ligament (UCL) tears, there remains a need for a reliable classification of UCL tears that has the potential to guide clinical decision making. PURPOSE: To assess the intra- and interobserver reliability of the newly proposed magnetic resonance imaging (MRI)-based classification for UCL tears. Secondary objectives included assessing the effect of additional views, discrimination between distal and nondistal tears, and correlation of imaging reads with intraoperative findings of the UCL. STUDY DESIGN: Cohort study (diagnosis); Level of evidence, 2. METHODS: Nine fellowship-trained specialists from 7 institutions independently completed 4 surveys consisting of 60 elbow MRI scans with UCL tears using a newly proposed 6-stage classification system. The first and third surveys contained 60 coronal images, while the second and fourth contained the same images with coronal and axial views presented in a random order to assess intraobserver variability via the weighted kappa value and the effect of additional imaging views. Weighted kappa values were also calculated for each of the 4 surveys to acquire interobserver reliability. Reliability analysis was repeated through a 2-group classification analysis for distal and nondistal tears. Observer readings were compared with intraoperative UCL findings. RESULTS: For the newly proposed 6-stage MRI-based classification, intra- and interobserver reliability demonstrated near perfect and substantial agreement, respectively. These values increased only when substratified into the 2-group distal and nondistal tear classification ( P < .05). The additional axial view did not statistically improve the agreement within and among readers. When compared with intraoperative findings from 30 elbows, observer readings were accurate for tear grade (partial and complete), proximal location, and distal location but not midsubstance tears. CONCLUSION: The newly proposed 6-stage MRI-based classification utilizing grade and location of the injury had substantial to near perfect agreement among and within fellowship-trained observers.