Data extraction
Two authors (Q.Y.Q. and W.K.E.) independently reviewed studies to extract potentially eligible studies. Any discrepancies were resolved by consensus with the corresponding author (X.Z.H). The main outcomes (reduction of FPG and HbA1c) for each study were extracted or calculated, and the adverse events (AEs) were collected. For consistency, we collected all per-protocol analysis data. In addition, the clinical characteristics of each study were also summarized.