BACKGROUNDIntra-articular glenohumeral joint injections are essential procedures for treating various shoulder disorders. Fluoroscopy-guided injections have been extensively used; however, they pose a risk of radiation exposure and are expensive and time-consuming. Recently, it has been suggested that ultrasound-guided injections are accurate and cost effective procedures.AIMTo evaluate the accuracy of ultrasound-guided glenohumeral injections using a posterior approach that is confirmed using magnetic resonance arthrography (MRA).METHODSThe study included 179 shoulders of patients with recurrent anterior instability (150 patients; 103 and 76 right and left shoulders, respectively; 160 males and 19 females; average age = 20.5 years; age range: 14-63 years) who underwent MRA for preoperative diagnosis. They were injected with 12 mL lidocaine (1%) using the ultrasound-guided posterior approach and then underwent magnetic resonance imaging. Two shoulder surgeons, except for the injector, evaluated the transverse relaxation (T2)-weighted images of axial planes and classified the intra-articular condition of injected contrast into three groups based on one of the three following scenarios: no leakage, injection into the glenohumeral joint without leakage; minor leakage, practical intra-articular injection with some leakage outside the posterior rotator cuffs; and major leakage, inaccurate injection with mass leakage without any contrast into the joint. The inter-rater reliability between two assessors was also evaluated by calculating Cohen’s kappa coefficient. The learning curve was assessed regarding the inaccurate injection rate by analyzing Spearman’s rank correlation coefficient.RESULTSOf the 179 injections, 163 shoulders (91.0%) had no leakage, 10 shoulders (5.6%) had minor leakage, and six shoulders (3.4%) had major leakage. In total, 173 shoulders (96.6%) were intra-articularly injected; thus, we could detect anterior labrum and capsular pathologies. Regarding the inter-rater reliability, the kappa coefficient was 0.925, indicating consistency in the evaluations by both examiners. Regression analysis of the inaccurate injection rate for assessingtechnical learning showed a logarithmic curve with a downward trend (R2 = 0.887, P < 0.001). Three (50%) of the six inaccurate injections classified into “major leakage” were observed in the first 30 injections, indicating that the accurate injection showed a leaning effect.CONCLUSIONUltrasound-guided intra-articular glenohumeral injections using a posterior approach had high accuracy; however, injection accuracy depends on clinical experience.