(684) Bayesian post-hoc analysis of cerclage retention versus removal in preterm prelabor rupture of the membranes

To estimate the probability of benefit/harm for cerclage retention compared to removal on prolongation of pregnancy and chorioamnionitis in the presence of a cervical cerclage and subsequent preterm prelabor rupture of the membranes (PPROM) using a Bayesian framework.


Study Design:

A post-hoc Bayesian analysis of data from a randomized clinical trial (RCT, Removal Versus Retention of Cerclage in PPROM [PROMCerclage], Galyean et al. AJOG. 2014 Oct;211(4):399) that included 56 patients with a cerclage in place and subsequent PPROM who had been randomized to cerclage retention vs cerclage removal. The trial was designed to detect differences in delivery < 1 week and chorioamnionitis, but was terminated after randomization of 39% of the planned sample size based on an interim analysis showing futility. The PROMCerclage trial is the only published RCT on this topic and prior observational studies have the potential for significant confounding, so we chose a minimally informative prior (relative risk [RR] = 1) for the primary analysis. Other priors were specified to represent preexisting enthusiasm or skepticism for cerclage retention using results of published observational studies. For each prior, the RR, credible interval (CrI), and probability of benefit/harm for delivery < 1 week and chorioamnionitis were estimated.

 

Results:

In the primary analysis combining a minimally informative prior with the findings of the PROMCerclage trial, the posterior probability that cerclage retention is beneficial was 22.9% for delivery < 1 week (RR, 1.22 [95% CrI, 0.73-2.03]) and 11.5% for chorioamniontis (RR, 1.54 [0.76-3.10], Table). The posterior probability that cerclage retention is harmful was 77.1% for delivery < 1 week and 88.5% for chorioamnionitis. Using priors incorporating results of previously published observational studies greatly influenced the posterior distribution of RRs (Figure).

 

Conclusion:

A post-hoc Bayesian analysis of data from an RCT of patients with PPROM demonstrates that the probability cerclage retention is harmful is greater than the probability it is beneficial.