Complications such as proximal humeral fracture [16, 43], glenoid fracture [33], shoulder dislocation [16, 50], and brachial plexus palsy [3, 36] have been reported. The proximal humeral fracture is reported to occur in 3–5 % of manipulation cases [16, 43]. Magnussen and Taylor used a short lever arm technique to apply force to the proximal humerus to avoid a humeral fracture [33]. They experienced a glenoid fracture instead in this case. They warned that surgeons should be aware of and alert patients to this potential complication even with use of a short lever arm technique. Numerous authors have reported no complications with this procedure [1, 11, 18, 40, 41, 46, 49, 53]. Thomas et al. performed MUA in 246 patients, in which there were no fractures or dislocations, no symptoms of acute rotator cuff tear, no neurological or other iatrogenic injuries [51]. Ibrahim et al. reported one case of vasovagal episode during the procedure among 42 cases [20]. Atoun et al. examined the integrity of the rotator cuff before and after MUA using ultrasonography in 33 patients [2]. They found no rotator cuff tear after MUA, and concluded that MUA had not been associated with rotator cuff tear.

In order to avoid serious complications, specific manipulation techniques to use short lever arm have been proposed [19, 40]. Some authors have recommended the use of translational manipulative force rather than conventional rotational manipulative force to reduce the risk of fractures [41, 42, 47]. Arthroscopic capsular release before manipulation may also decrease the risk of complications [39].

MUA creates a capsular tear, most commonly at the anterior and inferior portions of the capsule [29, 52]. This capsular tear makes it possible to gain an increase in the range of motion. Some additional damages such as SLAP lesions, partial rotator cuff tears, and anterior labral detachments were also reported [29]. Although there is no evidence that these iatrogenic articular lesions are of clinical importance, some prefer to use arthroscopic capsular release [29].

Wang et al. compared idiopathic frozen shoulder versus posttraumatic stiff shoulder versus postsurgical stiff shoulder [55]. They found less improvement in pain and motion in postsurgical stiff shoulders. They also compared MUA for those with and without noninsulin-dependent DM [56]. They found no significant difference between them. Even though DM is a predisposing factor to frozen shoulder, noninsulin-dependent DM alone does not influence both the short- and long-term outcomes of frozen shoulder. Jenkins et al. also compared those with DM (39 shoulders of 36 patients, DM group) and those without (274 shoulders of 256 patients, control group) [23]. They found significant improvement in range of motion and OSS with no difference between the groups at early or late follow-up. In another type of secondary stiff shoulders, Leonidou and Woods performed MUA for secondary stiff shoulder following breast cancer treatment [28]. There were 7 shoulders of 7 patients in the breast cancer group and 274 shoulders of 256 patients in the control group. None of the patients in this study were diabetic. The mean preoperative OSS was 31 for the cancer group and 27 for the control group, improving to 43 for both groups. At the long-term follow-up, 71 % in the breast cancer group were satisfied. Due to recurrence, 42 % in the cancer group and 15 % in the control group had second MUA. They concluded that MUA was effective in patients with secondary stiff shoulder after breast cancer treatment.

Ogilvie-Harris et al. in their prospective cohort study performed arthroscopy in 40 patients with persistent stiffness and pain for at least 1 year despite conservative treatment: (1) first 20 patients underwent arthroscopy, followed by manipulation, and then arthroscopy; (2) second 20 patients underwent arthroscopic capsular release [39]. At 2–5-year follow-up, no difference in the range of motion, but arthroscopic capsular release group did significantly better in pain and in function: 15/20 (75 %) treated with arthroscopic capsular release had an excellent result compared with 7/18 (39 %) treated with MUA. DM patients did worse initially but there was no difference at final follow-up. A recent systematic review has concluded that the quality of evidence available is low and the data available demonstrate little benefit for a capsular release instead of, or in addition to, an MUA [14]. A high-quality study is required.

Sharma et al. compared MUA and joint distension in 32 patients [49]. They recommended joint distension because it was easy to carry out and gave better results than MUA. Quraishi et al. compared 18 shoulders of 17 patients in the MUA group, 20 shoulders of 19 patients in the joint distension group in their prospective randomized clinical trial [44]. Pain and Constant score were significantly better in joint distension group, but no significant difference in the range of motion. At final follow-up of 6 months, 94 % satisfied in the joint distension group compared with 81 % in the MUA group. Jacobs et al. also performed a prospective randomized clinical trial with 53 patients with frozen shoulder treated by MUA versus steroid injection with joint distension [21]. At 2-year follow-up, there were no significant differences in the outcome (Constant score, visual analog scale, Short Form-36). Therefore, they recommended steroid injection with joint distension as an outpatient treatment, which has the same clinical outcome as MUA.

Kivimaki et al. performed a randomized clinical trial comparing those treated with MUA followed by home exercises and those treated by home exercises alone [27]. There were 65 patients in the MUA group and 60 patients in the home exercises group. They did not find any significant differences at any time point in terms of pain and working ability. Although small differences were observed in the range of motion in favor of MUA, MUA does not add effectiveness to an exercise program performed by the patient. However, the inclusion criteria of this study were those less than 140° of elevation and less than 30° of external rotation. With these criteria, those with very subtle limitation in the range of motion must have been included, who had little benefit from MUA. These wide inclusion criteria might have affected the outcome of this study.

Manipulation under anesthesia was estimated to cost £1,446 (≈ US$2,455) and capsular release £2,204 (≈ US$3,742), both of which included rehabilitation physiotherapy. Arthrographic distension was estimated to cost approximately £114.84 (≈ US$195), depending on the choice of steroid injection [35]. However, given the paucity of economic evidence, it is not possible to make any conclusions regarding the most cost-effectiveness intervention [35].

Manipulation under anesthesia is an effective treatment option for refractory frozen shoulder. It is as effective as but less costly than arthroscopic capsular release. Serious complications such as a fracture should be avoided with use of a short lever arm with slow and gentle motion. High-quality evidence needs to be established.