One hundred and thirty-six patients with secondary frozen shoulder were randomized to receive manipulation under anesthesia (n=68) or celecoxib (n=68), both in addition to an exercise program. The primary outcome of interest was shoulder function as measured by the Constant-Murley Score. Secondary outcomes of interest included the McGill Pain Questionnaire containing Pain Rating Index and Pressure Pain Intensity, and range of motion. Outcomes were measured up to 12 weeks follow-up. Constant-Murley scores were significantly in favour of the manipulation group at 1 day, 2 weeks, 4 weeks, and 12 weeks follow-up. Range of motion and Pressure Pain Intensity were also significantly in favour of the manipulation group at all time points. Pain Rating Index was in favour of the manipulation group from 2-12 weeks follow-up.