Eight trials (n=461, range 14 to 125) were included. Four achieved a PEDro score of 6 or above and were considered good quality; the other four scored 4 or less and were considered poor quality.
Three trials used a non-exercise comparator: one compared to placebo; one used a waiting list control; and one used a functional brace worn during the day and at night if possible. The placebo-controlled trial found statistically significant differences between groups for pain and Neer score at six months, but not after two and a half years (patients were no longer in their randomised groups at this point). The waiting list controlled trial found a statistically significant improvement for the intervention group for pain, function, range of abduction and extension and quality of life (short-form 36). The trial with the brace control found no statistically significant differences between groups.
Five trials included exercise as part of the comparator. One found that the intervention group had significantly less pain over a 24-hour period and on the subacromial compression test after treatment. One trial found a significantly greater pain reduction at 12 months for the intervention group. The other three trials did not report any statistically significant differences between groups.