Fourteen studies were included: 4 RCTs (164 patients), one non-randomised controlled trial (120 patients), one multiple baseline design (9 patients), 5 case series (85 patients) and 3 case reports (4 patients).
Overall, the trials were of poor methodological quality with methodological scores ranging from 2 to 25 out of a possible 48 points. Eight of the 14 studies scored less than 10 points. Nine of the 14 included studies did not use a control group. The information in many studies was inadequate in terms of the following: the study population; the duration of the shoulder pain; the exact intervention; blinding of the patient, therapist and assessor; drop-outs; follow-up; cointerventions; and statistical methods.
The success rates for treatment ranged from 0 to 100% (10 studies). It was not possible to calculate the success rates in 4 studies because either the results were only reported at group level or the number of patients who were treated successfully was not reported. The success rates were 46, 52 and 67% in the 3 studies that reported success rates on an intention to treat basis.
Results from the 3 methodologically flawed RCTs and the 1 non-randomised controlled trial are reported below.
One crossover RCT (20 patients, 3 to 16 weeks after stroke) found no significant difference in pain between one week of EMG biofeedback and one week of relaxation exercises after 2 weeks. Both treatments were applied without a washout period, and the follow-up period was different for both treatment groups. One multicentre RCT (85 patients) found no significant difference in pain at 12 weeks for cryotherapy followed by exercise, compared with the Bobath approach, but the frequency of the occurrence of pain was less for the Bobath group. Twenty patients withdrew before completion of the treatment.
One non-randomised controlled trial (120 patients) found that more patients who received FES, compared with no additional FES, were without pain at 3 and 24 months' post-treatment: with FES, 70% of the patients were without pain at 3 months versus 36% for the control, and 81% were without pain at 24 months versus 55% for the control. One RCT (26 patients after a recent stroke) found no difference in indirect assessment of pain (using range of motion) for FES compared with no FES. There were insufficient data to allow a direct assessment of the change in pain. One RCT (33 patients) found no difference for exercise versus control.