Bayesian posthoc trial analysis of cerclage management in preterm prelabor rupture of the membranes.

BACKGROUND

There is inadequate evidence on whether to retain or remove a cervical cerclage after preterm prelabor rupture of the membranes (PPROM). The PROMCerclage trial was the only study to randomize patients to cerclage retention or immediate cerclage removal after PPROM. The trial was terminated early, and results were not statistically significant but the cerclage retention group had shorter latency and more chorioamnionitis. Bayesian post-hoc trial analysis maximizes available clinical trial data by formally integrating information from preexisting studies or prior beliefs into statistical modeling to estimate the probability of a treatment effect.

OBJECTIVE

To re-analyze published results of the PROMCerclage trial using a Bayesian framework to estimate the probability of benefit and harm for cerclage retention compared to removal on latency and chorioamnionitis in the event of PPROM.

STUDY

DESIGN

We conducted a post-hoc Bayesian analysis of data from the PROMCerclage trial, which included 56 patients with a cerclage in place and subsequent PPROM who had been randomized to cerclage retention vs cerclage removal. Bayesian analysis uses prior information (priors) combined with study data to generate a posterior, or "final" effect size. The primary analysis assumed no association between cerclage management and delivery <1 week after randomization and chorioamnionitis (prior relative risk [RR]=1). Other priors were specified using results of published observational studies. For each outcome, using each prior, we estimated the probability of benefit (RR<1, lower risk of delivery <1 week from randomization / lower risk of chorioamnionitis), RR, and credible interval (CrI).

RESULTS

In the primary analysis comparing cerclage retention to removal, the posterior probability that delivery <1 week from randomization was less likely (ie, RR <1.0) was 21.9%. The posterior probability that chorioamnionitis was less likely with cerclage retention versus removal was 9.5%. Using priors incorporating results of previously published observational studies greatly influenced the results.

CONCLUSION

A Bayesian re-analysis of the only randomized trial of cerclage retention versus removal after PPROM found that shorter latency and more chorioamnionitis are more likely with cerclage retention than with immediate removal.