Arthroscopic capsular release is a most well-known technique with favorable outcomes for frozen shoulder. However, considering the surgical trauma and the improvement of multisite injection, we design a study to compare the pain relief and safety of multisite injection (MI) versus arthroscopic capsular release (ACR) for frozen shoulder. A total of 80 patients with unilater al frozen shoulder were enrolled in this study. Group RBT (n = 40) received multisite injection (0.5% lidocaine and triamcinolone acetonide, once a week, no more than 2 injections), while Group ACR received arthroscopic capsular release. The following parameters were employed to compare: visual analog scale (VAS), range of motion (ROM), the Disability of Arm, Hand, and Shoulder (DASH) score and Oxford shoulder score (OSS). Side effects were also recorded. The VAS, ROM, DASH and OSS all improved significantly (P < 0.001). Internal rotation and external rotation at 1 month after operation were better in ACR group (40.35 ± 4.79 Vs 36.58 ± 7.49, 40.55 ± 4.37 Vs 38.63 ± 4.01, P = 0.009,0.043). However, no significance in terms of functional results and ROM was found at 6 months after operation. The OSS, DASH and VAS in patients with diabetes were 44.25 ± 3.05, 2.29 ± 1.12 and 0.50 ± 0.72, compared with 43.89 ± 3.09, 2.34 ± 1.49 and 0.29 ± 0.56 in patients without diabetes (P = 0.636, 0.889, 0.157). Multisite injection and arthroscopic capsular release are both effective treatments in the treatment of frozen shoulder. However, multisite injection is a simple, cost effective and superior alternative.