Figure 2.

Risk of bias graph and summary. The low risk rate of random sequence generation was 40%, and low risk of blinding of patients was 60% (a). Incomplete outcome data and other bias had the low risk of 80% (a). The risk of allocation concealment and the blinding of outcome assessment were unclear (a). High risk appeared in incomplete outcome data and other bias with the rate of 20% (a). Esgalhado et al.'s study and Khosroshahi et al.'s study in 2018 was evaluated as high risks in one assessment. No study was estimated as low risk in all assessments (b).