Objective To evaluate the efficacy and safety of different site injection of compound betamethasone injection, ropivacaine and sodium hyaluronate for treatment of frozen shoulder at early stage. Methods A Total of 68 participants were included from May 2015 to May 2017 and randomly assigned to the glenohumeral joint and subacromial space group (IA+SA group, n=34) and glenohumeral joint group (IA group, n=34). In the IA+SA group, a solution of 1 mL corticosteroid, 6 mL ropivacaine, 2 mL sodium hyaluronate, and 8 mL normal saline were prepared and injected to glenohumeral joint, and a solution of 1 mL corticosteroid, 2 mL ropivacaine, 2 mL sodium hyaluronate, and 2 mL normal saline were injected to subacromial space. In the IA group, participants were given the same dose of drugs to the glenohumeral joint. The Visual Analogue Scale (VAS) score and Constant-Murley score were used to assess pain and function of shoulder respectively. The change of VAS score and Constant-Murley score after treatment were used to evaluate pain relief and shoulder function improvement. Results Of the 68 participants, two in each group were lost to follow up and one in the IA+SA group dropped out. There was significant effect on pain relief and shoulder function improvement on all measurement in both groups (P<0.001) during the 12 weeks after treatment. In the IA group, group- by-time interaction were significant for pain relief at 6 and 12 weeks comparing with that at 3 weeks (P<0.001), while no significant difference at 6 weeks comparing with 12 weeks. In the IA+SA group, group-by-time interaction were significant for pain relief at all endpoints (3 weeksvs. 6 weeks: P<0.001; 3 weeksvs. 12 weeks: P<0.001; 3 weeksvs. 6 weeks: P=0.034). In both groups, there was significant effect on shoulder function improvement when compared at each endpoint within group (P<0.001). Between-group comparison revealed no significant effect on pain relief (P=0.386) or shoulder function improvement (P=0.685). There was also no significant effect on pain relief (3 weeks: P=0.898; 6 weeks: P=0.448; 12 weeks: P=0.216) and shoulder function improvement (3 weeks: P=0.120; 6 weeks: P=0.152; 12 weeks: P=0.868) at each same endpoint. Conclusions Different site injection can effectively release pain and improve shoulder function for the patients with frozen shoulder at early stage and be well tolerated. However, it is not found that two site injection is inferior to single site injection.