Is there a good agreement between MRI readers for Thaunat's classification in arthroscopically-proven meniscal ramp lesions?

Abstract

To describe and evaluate the reproducibility by MRI of an arthroscopically-based classification for meniscal ramp lesions. We hypothesize that MRI would present good interobserver and intraobserver reliability to evaluate meniscal ramp lesions. Twenty MRI of the knee with arthroscopically-proven meniscal ramp lesions were independently assessed by two skilled musculoskeletal radiologists and a third-year radiology resident. Reading was performed in a randomized and anonymous manner, in two steps, with a minimum of 1-month interval between each. Cohen's kappa coefficient statistic was used to analyze intra and interobserver reading agreement. Associate findings were also categorized. From 20 subjects, 17 were male, with mean age of 35 years. MRI reading showed type IV ramp lesion as most prevalent with eight cases (37%), followed by type V - four (21%), type I - four (20%), type III - three, (16%) and type II - one (6%). Regarding ramp lesion types, intraobserver agreement was substantial for both skilled readers (Kappa = 0.72), and moderate for the less experienced reader (Kappa = 0.51); interobserver agreement was moderate. Results between most experienced readers were also analyzed in two categories: stable (types I and II) and unstable (types III, IV and V), also resulting in moderate agreement (Kappa = 0.54). Intraobserver agreement was substantial for both readers (Kappa = 0.68). The most common associate findings were joint effusion (85%), posteromedial capsular structures injury (60%), and medial meniscus extrusion (60%). The arthroscopy classification for meniscal ramp lesions stability adapted for MRI has good reproducibility when applied by trained musculoskeletal radiologists.