Background: We evaluated the need for arthroscopic capsular release (ACR) in refractory primary frozen shoulder (FS) by comparing clinical outcomes of patients treated with ACR and manipulation under anesthesia (MUA). Methods: We assessed patients with refractory primary FS, 54 patients (group A) who were treated with MUA and 22 patients (group B) who were treated with ACR. In group A, manipulation including a backside arm-curl maneuver was performed under interscalene brachial block. In group B, manipulation was performed only to release the inferior capsule before arthroscopic circumferential capsular release, which was carried out for the unreleased capsule after manipulation. Pain, range of shoulder motion, and American Shoulder and Elbow Society score were recorded at 1 week, 3 months, 6 months, and 1 year after surgery. We compared outcome variables between treatment groups and between diabetics and non-diabetics and also evaluated the numbers of patients receiving additional intra-articular steroid injection.Results: Outcome variables at 3 months after surgery and improvements in outcome variables did not differ between groups. Group A showed significantly better results than group B in the evaluation of pain and range of motion at 1 week. Diabetics showed comparable outcomes to non-diabetics for most variables. Eleven patients required additional steroid injections between 8 to 16 weeks after surgery: 12.2% in group A, 18.2% in group B. Additional injections were given three times more often in diabetics compared to non-diabetics.Conclusions: MUA alone can yield similar clinical outcomes to ACR in refractory FS.