Systematic review and meta-analysis of magnetic resonance imaging features for diagnosis of adhesive capsulitis of the shoulder

To perform a systematic review and meta-analysis to identify magnetic resonance imaging (MRI) features that will aid in the diagnosis of adhesive capsulitis of the shoulder (ACS) and provide a summary of the diagnostic accuracy of the identified features The MEDLINE and EMBASE databases were searched for studies assessing the diagnostic accuracy of MRI features of ACS. Overlapping descriptors used to denote the same imaging finding in different studies were subsumed under a single feature. The pooled accuracy including the diagnostic odd ratios (DORs) with 95% confidence intervals (CIs) of the identified features was calculated using a bivariate random-effects model. In total, 15 studies were included, and 74 overlapping descriptors were subsumed under six features. All six features were found to be informative for ACS diagnosis [coracohumeral ligament thickening: DOR, 13; 95% CI, 6-29; fat obliteration of the rotator interval (RI): DOR, 8; 95% CI, 3-24; RI enhancement: DOR, 44; 95% CI, 14-141; axillary joint capsule enhancement: DOR, 52; 95% CI, 27-98; inferior glenohumeral ligament (IGHL) hyperintensity: DOR, 31; 95% CI, 8-115; IGHL thickening: DOR, 28; 95% CI, 11-70]. The sensitivity and specificity of enhancement of the RI and axillary joint capsule and IGHL hyperintensity were > 80%. Six informative MRI features for ACS diagnosis were identified in this study with RI and axillary joint capsule enhancement and IGHL hyperintensity showing the highest diagnostic accuracy. Informative features observed on non-arthrogram MRI can be as helpful as features observed on direct magnetic resonance arthrography for ACS diagnosis. • Six informative MRI features for ACS diagnosis were identified (diagnostic odds ratio > 1). • RI and axillary joint capsule enhancement and IGHL hyperintensity showed high sensitivities/specificities (> 80%). • The use of non-arthrogram MRI is recommended for ACS diagnosis.