Table 2.

Quality assessment: risk of bias

Study ID	Random sequence generated	Allocation concealed	Blinding		Attrition infrequent a	Free of selective reporting
			Patients & clinicians	Outcome assessors
Bian 2016 [35]	Probably yes	Probably yes	No	Probably no	Probably no	Probably yes
Hu 2018 [36]	Probably yes	Probably yes	No	Probably no	Probably no	Probably yes
Huang 2015 [37]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Li 2017a [38]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Li 2017b [39]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Tian 2018 [40]	Yes	Yes	No	Probably no	NR	No
Wang 2016 [41]	Yes	Yes	No	Probably no	NR	Probably yes
Wang 2018 [42]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Wei 2019 [43]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Yao 2017 [44]	Probably yes	Probably yes	No	Probably no	Probably yes	Probably yes
Zhang 2015 [45]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Zhang 2019 [46]	Probably yes	Probably yes	No	Probably no	NR	Probably yes
Zhu 2018 [47]	Yes	Yes	No	Probably no	NR	Probably yes

Notes: ^a Yes defined as less than 10% attrition to all outcome and those excluded not likely to have made a material difference in outcomes. All answers as: yes, probably yes, probably no, no. NR not reported