Glenoid notch MRI findings do not predict normal variants of the anterior and superior labrum

AIM: To determine (1) the relationship of a glenoid notch to the presence of a normal labral variant in the anterioresuperior glenoid labrum; (2) the inter- and intra-observer reliability of recognising a glenoid notch; and (3) whether magnetic resonance arthrography (MRA) is more reliable than non-contrast magnetic resonance imaging (MRI) in visualising a glenoid notch. MATERIALS AND METHODS: From 1995 through 2010,104 patients underwent MRI or MRA before diagnostic shoulder arthroscopy by the senior author. Five blinded musculoskeletal radiologists independently read the images twice to evaluate for the presence or absence of a glenoid notch. Fifty-nine (57%) patients had normal anterioresuperior labral variants. The authors calculated the relationship of the readings to the arthroscopically determined presence or absence of a normal labral variant and the reading's diagnostic performance and rater reliability. RESULTS: On average, 38% (range 9-65%) of the glenoid scans were read as notched. The sensitivity, specificity, positive predictive value, and negative predictive value of the notch relative to the presence of a normal variant were 43.1%, 71.2%, 70.2%, and 48% versus 44.3%, 77.5%, 79.4%, and 56.1% for MRI and MRA, respectively. The overall average intra-observer κ-values were 0.438 (range 0.203-0.555) and 0.346 (range e0.102 to 0.570) for MRI and MRA, respectively. The average interobserver intra-class correlation coefficient reliability values were 0.730 (range 0.693-0.760) and 0.614 (range 0.566-0.662) for MRI and MRA, respectively. CONCLUSIONS: A notched glenoid on MRI lacks sufficient diagnostic performance and rater reliability for the clinical detection and prediction of normal anterioresuperior labral variants.