Methodological quality and risk of bias
The quality of the cohort studies was characterized as good, as measured by the Newcastle-Ottawa scale, with most studies scoring ≥ 7 out of 9 points (Table 1). The RCTs were assessed using the Cochrane Collaboration risk of bias (RoB) tool (11). Although all included RCTs conducted randomization, five of them did not adequately describe their randomization methods. Only two studies used sequentially numbered sealed opaque envelopes for allocation, 1 used a third party for allocation, 1 used software-generated random number sequence for allocation and the others never mentioned an allocation scheme. Eleven studies had an unclear risk of bias due to performance and detection bias. The infants and personnel could not be blinded due to the nature of intervention and three studies were judged to have a high risk of bias due to performance bias; however, the outcome assessor was blinded, resulting in a low risk of bias due to detection bias (suppl. eFig. 1, eFig. 2).