Search

Study design We will conduct an umbrella systematic review to identify eligible randomized controlled trials (RCTs) that compare the safety and efficacy of pharmacological, mechanical, or combined methods for inpatient cervical ripening and labor induction. We assume that participants who fulfilled the inclusion criteria will be equally eligible to be assigned randomly to any of the interventions of interest (to meet the assumption of transitivity). This study is exempt from Institutional Review Board approval. The pre-specified primary outcome will be the interval to vaginal delivery within 24 hours. Secondary outcomes to be assessed will include cesarean delivery, uterine hyperstimulation, neonatal and maternal morbidity and mortality, operative vaginal delivery, NICU admissions, and Apgar score < 7 at 5 minutes. Electronic literature search Initially we will perform and umbrella systematic review search using PubMed, EMBASE, and the Cochrane Central Register of Controlled Trials, ClinicalTrials.gov for narrative or systematic reviews and meta-analyses assessing the effectiveness and safety of pharmacological, mechanical, or combined methods for cervical ripening and labor induction. The computerized search will include systematic reviews from inception dates of each database and until May 2021. To ensure that recent trials will be included, we will also search individual RCTs testing interventions for labor induction in the same databases up to May 2021. In addition, individual RCTs will be identified from these reviews and data from each will be manually checked and subsequently employed to perform the network meta-analysis. The search strategy will include related text words and medical subject headings regarding cervical ripening, labor induction (see supplementary material). We will review references of included studies for additional related articles but will not search conference proceedings or the grey literature. A list of the unique PubMed identification numbers of all relevant articles will be compiled, and a search for related articles will be performed. This technique has been shown to be highly effective in the identification of relevant studies. Criteria for inclusion and exclusion of studies We plan to include randomized controlled trials that compare FDA-approved agents and methods currently employed for cervical ripening and labor induction. Such pharmacological and mechanical methods include intravenous oxytocin, misoprostol (oral or vaginal routes of administration), dinoprostone (vaginal and intracervical), and mechanical methods (single, double-balloon catheters and hygroscopic dilators) used alone or concomitantly with pharmacological agents. The doses for misoprostol will be stratified (less or greater than 50 mcg). We included studies with at least one other comparison group (placebo or no treatment) or another method or agent currently used for labor induction. Studies were limited to women undergoing labor induction at term with a live fetus. We will exclude studies describing methods that are no longer or infrequently employed in current clinical practice. Studies published in languages other than English will be included. Screening, data extraction, and risk of bias assessment Titles and abstracts will be screened independently by 2 investigators and those that do not meet eligibility criteria will be excluded. After omitting the duplicated and unrelated studies, we will review the full texts of the remaining studies and ascertain whether they meet inclusion criteria. Any discrepancies will be solved by a third reviewer. If multiple publications are derived from the same dataset, the study with the most complete data and the longest follow-up will be included. Two investigators, using a systematic review reference tool (Covidence), will independently extract the information from the original studies using a standardized data abstraction list that will include study characteristics, patient characteristics, and intervention details for each treatment group and outcome measures. Authors will be contacted for clarification as needed, and data will be recalculated into a form that is appropriate for analysis when needed. Any disagreements regarding data extraction will be resolved by discussion with a third author. Two investigators will appraise the risk of bias for individual studies according to the Cochrane Handbook. The criteria for assessment will involve randomization, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, incomplete outcome data, selective reporting, and other biases. Each of the domains will be determined as “low risk,” “unclear risk,” or “high risk.” Studies with a high risk of bias in ≥1 key items will be regarded to be at a high risk of bias. For example, for the domain of random sequence generation, low risk of bias will include studies in which randomization was performed by random number generation. A high risk of bias for this domain will include subjects randomly assigned based on day of the week (even or odd). If insufficient detail is provided in the article/manuscript, the domain will be labelled “unclear risk.” Studies with a low risk of bias in all key items will be regarded to be at a low risk of bias. Otherwise, they will be regarded to be at an unclear risk of bias. Disagreements will be resolved via a discussion with a third author. Methods for evidence synthesis We will begin with a narrative overview of the clinical and methodologic characteristics of the included trials, thereby helping to explore the assumptions of homogeneity and consistency for direct and indirect comparisons. We will generate descriptive statistics for all relevant trials and study population characteristics to provide a transparent representation of the patients in this analysis. Whenever possible, statistical analyses will be based on an intent-to-treat method and include all randomly allocated women. For each outcome, we will review the network geometry of all comparisons to confirm that the network is connected. Standard pairwise and network meta-analyses will be conducted to calculate odds ratios analyzed on the logarithmic scale for dichotomous outcomes and weighted mean differences for continuous outcomes. Direct estimates for traditional pairwise meta-analyses will be calculated with the random effects model using restricted maximum-likelihood estimation (REML). Direct pairwise comparison results will be reported as odds ratios (ORs), corresponding 95% confidence intervals (CIs), and 95% prediction intervals, which describe the range of true effects in future trials. Heterogeneity will be assessed with the I2 and Q chi-squared test. An I2 value of ≥50% or a p-value of .05 indicates that the direct and indirect comparisons are in agreement within the network. The comparative efficacy of the cervical ripening and labor induction agents that will be included in this analysis will be assessed with placebo or no treatment as the reference group. The probability that each agent employed is the best among those analyzed will be determined by evaluation of the rank probabilities and surface under the cumulative ranking curve (SUCRA) for the efficacy results of the network meta-analysis. A higher SUCRA value indicates better performance for the respective intervention. Results from the network meta-analyses will be presented as a summary of relative effect sizes for each possible pair of treatments and reported as log ORs with 95% CIs. A league table will be constructed to tabulate all comparisons estimated from the network. Sensitivity and subgroup analyses performed After each pairwise direct comparison, depending on the number of studies assessing a particular outcome, we will investigate the influence of each individual study on the overall summary estimate by re-estimating the meta-analysis after sequentially omitting each study. An individual study will be suspected of having excessive influence if the point estimate of its “omitted” analysis lies outside the confidence interval of the “combined” analysis. If possible, subgroup analyses and meta-regression will be conducted to explore the impact of potentially important effect modifiers on findings from network meta-analysis. These separate analyses include covariates in meta-regression models that consider the main confounders such as cervical score, gestational age, parity and others. In addition, subgroup analyses of the treatment networks will be conducted to compare the individual agents used for cervical ripening. Software considerations We will use the user-written commands from the package “metan” for Stata SE software (version 16.0; StataCorp, College Station, TX) to perform the pairwise direct comparison meta-analyses. Similarly, for network meta-analysis, we will use several network “meta“ user-written commands to perform multivariate random-effects meta-analysis and multivariate random-effects meta-regression. In case that Stata commands do not provide feaures for some purposes in our analyses, we will also consider using the R package “netmeta” (version 1.4-0). No funding was provided for this study.