Fig. 2.

Summary of the quality assessment of the included studies. All of the selected studies were RCTs. Among the included studies, four studies (Liang B, Zhu M W, Ma C J, Nathalie B) generated the random sequence by computer and two studies(Sorensen L S, Wei Z) generated from sealed opaque envelopes containing randomized numbers, one study (Zhang C H) generated from random number method and one study (Liu Y M) carried out the randomisation by medical record number, seven studies (Liang B, Liu Y M, Ma C J, Nathalie B, Sorensen L S, Zhang C H, Zhu M W) described their blind-ing method. Four studies (Liang B, Liu Y M, Nathalie B, Wei Z) had no follow-up bias, five studies(Kiek S,Sorensen L, Zhang C H, Zhang Y, Zhu M W) might have follow-up bias, and one study (Ma C J) had follow-up bias. Six studies (Liang B, Liu Y M, Ma C J, Wei Z, Zhang C H, Zhu M W) had no reporting bias and Four studies had reporting bias(Kiek S, Nathalie B, Sorensen L S, Zhang Y). Nine studies with no other bias and one study(Wei Z) may have other biases.