Efficacy and safety of oral and sublingual versus vaginal misoprostol for induction of labour: a systematic review and meta-analysis

Objective Misoprostol is a synthetic PGE1 analogue that is used for induction of labour. Current guidelines support the use of doses that do not exceed 25 mcg in order to limit maternal and neonatal adverse outcomes. The present meta-analysis investigates the efficacy and safety of oral compared to vaginally inserted misoprostol in terms of induction of labor and adverse peripartum outcomes. Methods We searched Medline, Scopus, the Cochrane Central Register of Controlled Trials CENTRAL, Google Scholar, and Clinicaltrials.gov databases from inception till April 2022. Randomized controlled trials that assessed the efficacy of oral misoprostol (per os or sublingual) compared to vaginally inserted misoprostol. Effect sizes were calculated in R. Sensitivity analysis was performed to evaluate the possibility of small study effects, p-hacking. Meta-regression and subgroup analysis according to the dose of misoprostol was also investigated. The methodological quality of the included studies was assessed by two independent reviewers using the risk of bias 2 tool. Quality of evidence for primary outcomes was evaluated under the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, ranging from very low to high. Results Overall, 57 studies were included that involved 10,975 parturient. Their risk of bias ranged between low-moderate. There were no differences among the routes of intake in terms of successful vaginal delivery within 24 h (RR 0.90, 95% CI 0.80) and cesarean section rates (RR 0.92, 95% CI 0.82, 1.04). Sublingual misoprostol was superior compared to vaginal misoprostol in reducing the interval from induction to delivery (MD – 1.11 h, 95% CI – 2.06, – 0.17). On the other hand, per os misoprostol was inferior compared to vaginal misoprostol in terms of this outcome (MD 3.45 h, 95% CI 1.85, 5.06). Maternal and neonatal morbidity was not affected by the route or dose of misoprostol. Conclusion The findings of our study suggest that oral misoprostol intake is equally safe to vaginal misoprostol in terms of inducing labor at term. Sublingual intake seems to outperform the per os and vaginal routes without increasing the accompanying morbidity. Increasing the dose of misoprostol does not seem to increase its efficacy. Clinical trial registration Open Science Framework (https://doi.org/10.17605/OSF.IO/V9JHF).


Introduction
Induction of labour is used for decades in obstetrics; however, during the last years a steep rise in the percentage of deliveries that are medically induced is observed which from anecdotal reports seems to reach approximately 9.6% of all deliveries worldwide [1]. In the United States these rates are considerably higher reaching approximately 27%, while in Europe they seem to range between 6.8 and 33% [2,3]. Whereas several methods have been used to induce cervical ripening as well as onset of labor, including oxytocin, prostaglandins, laminaria tents and foley balloon catheter, prostaglandins and oxytocin are the most widely accepted.
The American College of Obstetricians and Gynecologists suggests the use of prostaglandins for cervical ripening and labor induction at intervals that should be at least 3-6 h apart to avoid the risk of uterine tachysystole, although the minimum interval that is considered safe has not been standardized yet [4]. Regarding misoprostol, the World Health Organization recommends either oral (25 μg, 2-hourly) or vaginal route (25 μg, 6-hourly) for induction of labour [1]. In a previous meta-analysis that was conducted by Alfirevic et al., misoprostol combined with oxytocin and amniotomy has been proven to be the best method for achieving vaginal delivery within 24 h [5].
Misoprostol was originally licensed for oral use; however, vaginal and sublingual routes of administration are becoming more and more popular supported by pharmacokinetic studies focusing on the systemic bioavailability parameters achieved [6]. While the sublingual route seems to have the greater bioavailability, safety concerns have been raised from studies that were conducted in first trimester terminations [7].
Several articles have addressed the efficacy of the various routes of misoprostol in inducing a successful vaginal delivery and a previous meta-analysis that was published in 2008 suggested that both sublingual and vaginal routes seem to be comparable in terms of achieving vaginal delivery [8]. More recently, Alfirevic et al. published a systematic review comparing oral misoprostol to other methods of induction of labor (including oxytocin) and observed that it was associated with fewer cesarean sections [9]. Since then, several randomized trials have been published and an update is indicated to review current knowledge as, to date, there is no consensus on the optimal route of misoprostol intake for induction of labour.
The purpose of this meta-analysis is to investigate whether oral (either per os or sublingual) misoprostol intake is superior to vaginal administration in terms of inducing labor and leading to a vaginal delivery. Taking in mind the safety concerns that have been raised in the field of first trimester pregnancy complications, we also investigated its impact on maternal and neonatal morbidity outcomes.

Protocol and registration
The present meta-analysis was designed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [10]. The study was based in aggregated data that have been already published in the international literature. Patient consent and institutional review board approval were not retrieved as they are not required in this type of studies. The study's protocol was published in open science framework, prior to the conduct of this review (Registration https:// doi. org/ 10. 17605/ OSF. IO/ V9JHF).

Eligibility criteria
Eligibility criteria for the inclusion of studies were predetermined. Randomized trials that compared the efficacy of oral (both per os or sublingual) to that of vaginal misoprostol in terms of induction of labour were considered eligible for inclusion. Quasi-randomized trials as well as observational studies (prospective and retrospective studies) were omitted from the systematic review. Only studies investigating induction of labor outcomes in singleton vertex presentations were included. Studies were included irrespective of the actual reason for induction of labour (antenatal pathology or prolonged pregnancy).  April 30, 2022. No date restrictions were applied. Articles were limited to English language. The search strategy included the text words "induction; labour; misoprostol; oral; vaginal; sublingual; per os; oral" and is presented in brief in "Appendix".

Information sources and search strategy
Studies were selected in three consecutive stages. Following deduplication, the titles and abstracts of all electronic articles were independently screened by three authors (V.P., M.P and T.K.) to assess their eligibility. The decision for inclusion of studies in the present meta-analysis was taken after retrieving and reviewing the full version of articles that were considered potentially eligible. Discrepancies that arose in this latter stage were resolved by consensus from all authors.

Study selection and data extraction
Outcome measures were predefined during the design of the present systematic review. Data extraction was performed using a modified data form that was based in Cochrane`s data collection form for intervention reviews for RCTs and non-RCTs [11].
The main outcomes of our study were the rates of successful vaginal delivery within 24 h, as well as cesarean section rates. Secondary outcomes included successful vaginal delivery beyond the limit of 24 h, interval to delivery, risk of uterine tachysystole, risk of postpartum hemorrhage and its accompanying side effects (risk of disseminated intravascular coagulation and need for obstetric hysterectomy) and neonatal side effects (including Apgar score values at 5` and need for admission to the NICU).

Assessment of risk of bias and quality of evidence
The methodological quality of the included studies was assessed by two independent reviewers using the risk of bias 2 (RoB 2) tool. RoB2 incorporates five domains that include assessment of (i) risk of bias that arises from the randomization process, (ii) risk of bias due to deviations from the intended interventions, (iii) missing outcome data, (iv) risk of bias in the measurement of the outcome, and (v) risk of bias in the selection of the reported result.
Quality of evidence was evaluated under the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, ranging from very low to high. More specifically, credibility of evidence will be assessed by taking into account the following domains: study limitations, directness, consistency, precision and publication bias. In particular, study limitations were evaluated based on risk of bias assessments (RoB 2 score), while directness was judged using the PICOS (population, intervention, comparison, outcome, study type) approach.

Synthesis of results
Statistical meta-analysis was performed with RStudio using the meta and metafor functions (RStudio Team (2015). RStudio: Integrated Development for R. RStudio, Inc., Boston, MA URL http:// www. rstud io. com/). Statistical heterogeneity was not considered during the evaluation of the appropriate model of statistical analysis as the anticipated methodological heterogeneity of included studies did not leave space for assumption of comparable effect sizes among studies included in the meta-analysis [12]. Confidence intervals were set at 95%. We calculated pooled mean differences (MD) and 95% confidence intervals (CI) with the Hartung-Knapp-Sidik-Jonkman instead of the traditional Dersimonian-Laird random effects model analysis (REM).
The decision to proceed with this type of analysis was taken after taking into consideration recent reports that support its superiority compared to the Dersimonian-Laird model when comparing studies of varying sample sizes and betweenstudy heterogeneity [13]. When variables where expressed as median (range), median (interquartile range) or interquartile range and sample size transformation where performed to acquire the mean and standard deviation to include the studies in the meta-analysis [14].
Publication bias was assessed by examining the possibility of small-study effects through the visual inspection of funnel plots. The asymmetry of funnel plots was statistically evaluated using the Egger's regression and Begg-Mazumdar's rank correlation tests. The Duval and Tweedie`s trim and fill method was applied to impute missing effects irrespective of the asymmetry of the funnel plot.
Publication bias was evaluated by examining the potential presence of small-study effects through the visual inspection of funnel plots. Rücker's Limit Meta-Analysis was applied to account for bias that arises due to small-study effects for primary outcomes. Outlier analysis was also undertaken to evaluate the effect of outlier studies on the overall effect size.

Prediction intervals
Prediction intervals (PI) were also calculated, using the meta function in RStudio, to evaluate the estimated effect that is expected to be seen by future studies in the field. The estimation of prediction intervals takes into account the inter-study variation of the results and express the existing heterogeneity at the same scale as the examined outcome.

Subgroup analysis
Subgroup analysis was designed following the retrieval of studies as several articles evaluated differences among sublingual and vaginal misoprostol as well as between per os and vaginal misoprostol; hence, oral misoprostol was subgrouped (as per os and sublingual) to evaluate for differences in summary effect estimates among the two methods.
Given the fact that in their previous meta-analysis Alfirevic et al. commented the lack of a substantial number of studies that could permit subgroup analysis based on the dose of misoprostol [9] in the present meta-analysis we performed meta-regression analysis of primary and secondary outcomes based on differences in misoprostol dosage (per dose administered and not total dose) among the two groups, to evaluate the potential effect of differences in the effect estimate. Subgroup analyses were also performed after arbitrary grouping of studies according to the maximum dose administered (Low dose = both groups received a dose of < 50 mcg of misoprostol, Intermediate = at least one of the two groups received = 50 < 100 mcg of misoprostol, High dose = at least one of the two groups received ≥ 100 mcg of misoprostol).

Study selection and study characteristics
Following completion of the electronic search strategy we were able to identify 1436 potentially relevant articles. After reading the abstracts and, when needed, full texts we managed to limit them to an overall number of 69 articles of which we finally selected 57 randomized trials that involved 10,975 parturient . The methodological characteristics of included studies as well as patient characteristics are depicted in the "Appendix" and reveal comparable groups of oral vs vaginal misoprostol intake in terms of maternal age and body mass index, gestational age at delivery parity, Bishop score prior to the start of induction and neonatal birthweight.

Risk of bias of included studies
The evaluation of the methodological quality of included studies with the RoB2 tool revealed low risk of bias for the majority of studies, whereas some concerns were raised in 12 studies and high risk of bias was revealed in 9 studies ( Fig. 1 and "Appendix").

Synthesis of results
Overall, effect sizes of outcomes regarding per os and sublingual misoprostol compared to vaginal are presented in Table 1, while summary effect sizes along with subgroup analysis according to maximum misoprostol dose are presented in Table 2.
The various methods that were used were comparable in terms of successful vaginal delivery within 24 h (Fig. 2). Significant heterogeneity was observed among included studies. Small studies seem to influence the results and Rucker's meta-analysis revealed significant differences among the two groups in favor of the oral route (Table 3). Outlier analysis also indicated that several studies affected the overall effect with a p value closer to statistical significance for the adjusted estimate (p = 0.0857).
Cesarean section rates were comparable among the various routes of administration of misoprostol (Fig. 3). The dose of misoprostol did not influence these findings (RR 0.88, 95% CI 0.49, 1.60 for low doses, RR 0.94, 95% CI 0.81, 1.08 for intermediate doses and RR 0.89, 95% CI 0.75, 1.07 for high doses). Outliers, small study effects and meta-regression analysis did not influence the overall effect estimate.
Summary effect sizes of secondary outcomes are presented in Table 2. Briefly, the route of misoprostol intake did not influence delivery rates within 48 h from induction of labor or interval to delivery. Similarly, uterine tachysystole rates, risk of postpartum hemorrhage and risk of disseminated intravascular coagulation (DIC) were not affected by the route of administration. Concerning neonatal outcomes, risk of admission to the neonatal intensive care unit (NICU) was comparable among cases that received oral misoprostol to those that had intravaginal placement.

Publication bias
The significance of the results of secondary outcomes were not altered by the trim-fill and small study effects (Rucker's) analysis. Contour enhanced funnel plots did not reveal significant bias with the exception of the uterine tachysystole outcome ("Appendix"). Trim and fill analysis did not reveal significant alterations in the level of statistical significance of results ( Table 3).

Quality of evidence
The overall quality of the evidence for the evaluation of misoprostol's efficacy in successfully inducing labor and resulting in normal delivery was evaluated as high due to the low study limitations, very low possibility of indirectness and inconsistency of results, precision of findings and low risk of publication bias.

Principal findings
The findings of our meta-analysis suggest that vaginal misoprostol increases the odds of successful vaginal delivery within 24 h compared to per os intake. However, sublingual intake might be superior to vaginal misoprostol. Of note, in terms of absolute time, sublingual misoprostol seems to be superior as it reduces the actual time from induction to delivery by approximately 1.11 h (compared to vaginal intake), whereas per os misoprostol seems to be the less effective  mode as it increases this interval by approximately 3.45 h (compared to vaginal intake). However, direct comparison of these two modes of administration was not available in a substantial number of studies; hence, a network metaanalysis was not undertaken to evaluate the optimal route of intake.

Comparison with existing literature
A previous meta-analysis that investigated differences in labour outcomes among women receiving sublingual and vaginal misoprostol revealed increased rates of uterine tachysystole in the sublingual group; however, its findings were based in 5 studies [8], whereas our analysis included outcomes from 39 studies (10 of them regarding sublingual intake) and revealed that both oral (either per os or sublingual) and vaginal intake are equally safe in inducing labour. In 2001 Crane et al. in a retrospective cohort of 519 women observed that the incidence of tachysystole and hyperstimulation varied among the various routes of misoprostol intake [71]. The results of the present metaanalysis; however, seem to contradict this and leave plenty of space for decision making in clinical practice.
In 2014 Alfirevic et al. conducted a meta-analysis and concluded that the appropriate dose of misoprostol intake should be considered the one that ranges between 20 and 25 mcg and that it is preferable to use oral intake compared to vaginal, especially in the setting of developing countries in order to help reduce the risk of maternal and neonatal infection. Our meta-analysis did not reveal substantial differences among the three groups in terms of safety, however, it should be noted that between subgroups (Low, intermediate, high doses), substantial heterogeneity was noted that revealed a trend towards increased intervals form induction to delivery in higher and intermediate doses, whereas lower doses had a comparable effect estimate compared to vaginal misoprostol. To date, there is limited evidence on the actual impact of the vaginal route on maternal and neonatal infection rates (including chorioamnionitis, post-partum endometritis, neonatal infection and neonatal sepsis) and given the indirectness of reporting meta-analysis was not performed for these outcomes.

Strengths and limitations
The present meta-analysis is based in a large number of randomized controlled trials of low-moderate risk of bias. The secondary analyses that were performed permit safe interpretation concerning the accuracy of estimated effect sizes and estimation of risk of bias. Publication bias was thoroughly evaluated and indicated that our analysis is based mainly on studies of low-moderate risk of bias. Furthermore, for the first time it was found that the dose of misoprostol does not seem to affect rates of normal delivery (within 24and 48-h).
On the other hand, several other confounders may partially limit the findings of this study, including co-administration of oxytocin for induction and augmentation of labor, parity, Bishop score at start of induction, presence of antenatal pathology (that may result in antepartum fetal distress, therefore, resulting in cesarean section) and gestational week at induction. However, meta-analyses that are based on aggregate data are always prone to these factors. In our analysis the Hartung-Knapp-Sidik-Jonkman random effects model was used which outperforms the traditional DerSimonian-Laird model as it consistently results in more adequate error rates when heterogeneous studies are considered for analysis; hence, confounders can be partially overlooked [13].

Conclusions and implications
Considering the results of the present meta-analysis it seems reasonable to conclude that sublingual misoprostol seems to offer better results in terms of a shorter interval to delivery and successful vaginal delivery within 24 h compared to vaginal misoprostol which seems to perform better than the per os intake (both as a pill or oral solution). Maternal and neonatal morbidity is not affected by either mode of intake. Intermediate doses can be acceptable according to the outcomes of the present study and seem to be the most efficacious in terms of shortening the interval to delivery. Table 3 Trim and fill analysis and p-curve analysis Meta-regression analysis was based on the difference of the dosage between the two groups (Misoprostol dosage in the per os group minus the dosage that was delivered in the vaginal group). Trim and fill, p-curve analysis and Rucker's analysis are based on the sum of the studies that were included. P-curve analysis was omitted in outcomes that were not based in at least 2 studies that indicated a potentially significant (p < 0.05) of the route of misoprostol intake. (CI NA = Confidence intervals were not available following analysis). All secondary analyses (trimfill, meta-regression, p-curve and Rucker's analysis for small study effects) are based on the sum of studies Clinicians should be aware of these findings and could consider using misoprostol orally or sublingually instead of vaginally. Furthermore, there is need for future research, especially randomized controlled trials comparing oral and sublingual routes of administration as there is limited data on this direct comparison.

Search plot
(See Tables 4, 5) Fig. 3 Forest plots of risk of delivering with cesarean section among parturient induced with oral misoprostol (subgrouped to sublingual and per os intake) and those induced with vaginally inserted misoprostol. (Vertical line = "no difference" point between the two groups. Blue squares = risk ratios; Diamonds = pooled risk ratios and 95 confidence intervals for all studies; Horizontal black lines = 95% CI; Horizontal red line = pooled 95% prediction intervals) ◂    Delivery with cesarean section 1 3

Secondary outcomes (contour funnel plots, p-curves if available analysis exists)
Interval to delivery (experimental = oral, control = vaginal) Risk of meconium-stained amniotic fluid Risk of 5-min Apgar score < 7 1 3

Outlier analyses
Delivery within 24 h