Secondary outcomes
Secondary outcomes included PTB less than 28, 32, 34, and 35 weeks.
Maternal secondary outcomes examined were rates of preterm premature
rupture of the membranes
(PPROM), chorioamnionitis (clinical or laboratory diagnosis), cervical
laceration and stenosis, caesarean section delivery (excluding planned
or elective), number of days between cerclage and delivery,
intraoperative membrane rupture at time of cerclage, cases requiring
repeat cerclage and difference in cervical length before and after
cerclage. Neonatal outcomes included birthweight, Apgar score
<7, and neonatal mortality, which included mortality up to one
month post-delivery or neonatal survival (this figure was then inverted
to mortality). Thorough definitions of these outcomes are referenced in
the protocol paper(9).
Data synthesis
A meta-analysis was performed by pooling studies together using
RevMan(24) and Covidence software(20). The heterogeneity of data was
examined using forest plots and quantified throughout calculation of the
I² value. An I² of greater than or equal to 50% was used to indicate
substantial heterogeneity and a random-effects model was used. For all
I2 less than 50%, a fixed effects model was used.
Outcomes with less than five studies were analysed using a fixed effects
model(25). For reporting consistency between outcomes, we made the
McDonald intervention the reference set for all analyses, standardising
the direction of effect across all primary and secondary outcomes.
Measures of treatment effect
Where applicable, trial data were combined and reported using
meta-analyses using the standard estimation of: 1) risk ratio (RR) and
95% confidence intervals (CI) for dichotomous outcome variables, and 2)
mean differences (MDs) or standardised mean differences (SMDs) and 95%
CIs for continuous outcome variables.
Sensitivity Analysis
Sensitivity analysis was conducted on the primary outcomes – PTB less
than 37 weeks gestation. This was performed by removing studies with an
overall high risk of bias to examine their impact on the effect
estimate.