1. Summary of Risk Of Bias In Non‐randomised Studies of Interventions (ROBINS‐I).

Author	Confounding	Selection bias	Classification of interventions	Deviation	Missing data	Measuring outcomes	Reporting bias
Aletti 2006a	Critical	Low	Low	Unclear	Moderate/high	Critical	Critical
Chang 2012a	Critical	Low	Low	Unclear	Moderate/high	Critical	Unclear
Luyckx 2012	Critical	Low	Low	Unclear	Moderate/high	Critical	Unclear

Risk of bias in included non‐randomised studies was assessed using the ROBINS‐I tool as outlined in Appendix 4(Sterne 2016).