We aimed to determine the effect of manipulation under anesthesia in frozen shoulder patients. A blinded randomized trial with a 1-year follow-up was performed at 3 referral hospitals in Southern Finland. We randomly assigned 125 patients with clinically verified frozen shoulder to the manipulation group (n = 65) or control group (n = 60). Both the intervention group and the control group were instructed in specific therapeutic exercises by physiotherapists. Clinical data were gathered at baseline and at 6 weeks and 3, 6, and 12 months after randomization. The 2 groups did not differ at any time of the follow-up in terms of shoulder pain or working ability. Small differences in the range of movement were detected in favor of the manipulation group. Perceived shoulder pain decreased during follow-up equally in the 2 groups, and at 1 year after randomization, only slight pain remained. Manipulation under anesthesia does not add effectiveness to an exercise program carried out by the patient after instruction.