Methods
We conducted a comprehensive literature search of the English language literature using Medline and the Cochrane Database of Systematic Reviews. The search covered the period from January 1980 to November 2010. We used combinations of the following search terms "labor, induced/or induction of labor; prostaglandin or prostaglandins, misoprostol; Cytotec; 16,16,-dimethylprostaglandin E2 or E2; dinoprostone; Prepidil; Cervidil: Dinoprost; Carboprost or hemabate; prostin, oxytocin, misoprostol, prostaglandins, membrane sweeping or membrane stripping, amniotomy, balloon catheter or Foley catheter, hygroscopic dilators, laminaria, dilapan, saline injection, nipple stimulation, intercourse, acupuncture, castor oil, herbs". Titles and abstracts were reviewed for possible exclusion by two reviewers (KK or EM and JC). If both reviewers excluded a citation, we eliminated that publication from further review. If at least one reviewer felt the citation might be included or if there was insufficient information to make a determination from the title and abstract, we obtained the full article for review. We identified additional articles for consideration of inclusion through cross checks of relevant bibliographies. Reference lists were created and full-text articles were retrieved for further consideration for inclusion.
In accordance with published guidelines for a "best evidence" review,[
3‐
5] this study included high-quality systematic reviews and randomised controlled trials in a hierarchical fashion. If a high-quality systematic review was available, only randomised controlled trials (RCT) published after the search date for the systematic review were included, except in the instance in which we found a RCT that had not been identified by the systematic review's search or a RCT that had been identified by the systematic review's search but which was awaiting classification. In addition, we included studies with at least one other comparison group (control, placebo or another method) for women undergoing induction of labour at term with a live fetus. We excluded systematic reviews dealing exclusively with subgroups of participants, such as nulliparas or women with prelabour rupture of membranes or with only a particular dose or formulation of the method under study (i.e. low dose or sustained-release preparations). We excluded dose-ranging studies, comparisons of two different formulations of the same method and studies in which subjects in one or more treatment arm received several different methods of labour induction. We did not exclude studies in which subjects received oxytocin augmentation after cervical ripening.
If five or more randomised controlled trials involving a method of induction were published subsequent to the search date of the most recent included systematic review or were "awaiting classification" in the systematic review, we conducted meta-analyses of the primary outcomes reported in these studies. Two authors, VR and UP extracted data independently. Differences were resolved by a third reviewer (EM) after careful review of each manuscript. The new data were added to the data on the comparison available in the Cochrane review. We computed risk ratios and 95% confidence intervals for the main outcome measures reported in these subsequent studies using Comprehensive Meta-Analysis, Version 2, Englewood, NJ. We used the fixed effects method for these analyses in order to match the measures of effect reported by the included Cochrane reviews.
We arranged the methods of labour induction according to types including pharmacologic methods, non-pharmacologic methods, complementary and alternative medicine methods, and investigational methods. However, for comparisons of methods with each other, we followed the pre-specified hierarchy used for the series of induction of labour Cochrane Reviews and arranged labour induction methods in that specific order [
2]. In each subsection of this paper, we compare each method with those methods prior to it on this list. (see Table
1)
Table 1
Induction of labour methods; hierarchy of comparisons[
2]
(1) | placebo/no treatment |
(2) | vaginal prostaglandin E2 |
(3) | intracervical prostaglandin E2 |
(4) | intravenous oxytocin |
(5) | amniotomy |
(6) | intravenous oxytocin with amniotomy |
(7) | vaginal misoprostol |
(8) | oral misoprostol |
(9) | mechanical methods including extra-amniotic Foley catheter |
(10) | membrane sweeping |
(11) | extra-amniotic prostaglandins |
(12) | intravenous prostaglandins |
(13) | oral prostaglandins, excluding misoprostol |
(14) | mifepristone |
(15) | oestrogens with or without amniotomy |
(16) | corticosteroids |
(17) | relaxin |
(18) | hyaluronidase |
(19) | castor oil, bath, and/or enema |
(20) | acupuncture |
(21) | breast stimulation |
(22) | sexual intercourse |
(23) | homoeopathic methods |
(24) | isosorbide mononitrate |
(25) | buccal or sublingual misoprostol |
(26) | hypnotic relaxation |
All full text articles were independently reviewed by two authors (EM and KK) for possible inclusion. In order to be included in this review, trials had to report on one or more of the outcomes of interest specified by the Cochrane Collaboration induction of labour generic protocol [
2]. The Cochrane generic protocol identified the most clinically important benefits and harms of labor induction as the outcomes of interest. These included the following five primary outcomes which were felt to be of most clinical importance: vaginal delivery not achieved within 24 hours (or period specified by authors), uterine hyperstimulation with fetal heart rate (FHR) changes, caesarean section, serious neonatal morbidity or perinatal death (e.g. seizures, birth asphyxia defined by trialists, neonatal encephalopathy, disability in childhood), serious maternal morbidity or death (e.g. uterine rupture, admission to intensive care unit, septicaemia) [
2].
Secondary outcomes included unfavourable or unchanged cervix after 12 or 24 hours, need for oxytocin augmentation, uterine hyperstimulation without FHR changes, uterine rupture, epidural analgesia, instrumental vaginal delivery, meconium stained amniotic fluid, Apgar score less than seven at five minutes, neonatal intensive care admission, neonatal encephalopathy, perinatal death, disability in childhood, maternal side effects including nausea, vomiting and diarrhea. Other secondary outcomes included postpartum hemorrhage, serious maternal complications, maternal infections including chorioamnionitis and endometritis, and neonatal infections including meningitis, pneumonia, and sepsis. Maternal satisfaction data were included when available. For each of the methods of induction, we reported the significant measures of effect (odds ratios or risk ratios) on our outcomes of interest from the included systematic reviews and RCTs.
Due to the large number of methods, comparisons and outcomes, we did not include discussion of subgroup analyses. However, because of the importance of cervical status as a determinant of failure of induction of labor to achieve vaginal birth, we reported on the effect of induction methods on caesarean deliveries for the subgroup with unfavorable cervices, where available in the Cochrane reviews.
Two authors (EM and JC) assigned quality scores to each included full-text article based on the Scottish Intercollegiate Guidelines Network (SIGN) quality assessment instruments. These quality assessment instruments are designed to assess the internal validity of each study, and the degree to which the studies' performance minimized bias [
6]. The Scottish Intercollegiate Guidelines Network publishes methodology checklists for critical appraisal of both randomised controlled trials and for systematic reviews [
6].
We systematically reviewed benefits and harms of each induction method and calculated number needed to treat (NNT) and number needed to harm (NNH) for each significant comparison among methods. For comparisons including only one trial, we used the "treat as one trial" method of calculating the NNT [
7]. When more than one trial was included in the comparison, we calculated NNT from pooled odds ratios and risk ratios reported in the included meta-analyses using the
Visual Rx, version 2; this method is less prone to bias than the "treat as one trial" method of NNT calculation [
8,
9]. For the purpose of NNT calculations from pooled estimates, we used risk ratios or odds ratios where reported for adverse outcomes and odds ratios to calculated NNT from positive outcomes [
9]. When odds ratios were not available in the source studies, we calculated them from available data using
Comprehensive Meta-Analysis, Version 2, Englewood, NJ. NNT estimates were rounded up to the next whole number whereas NNH estimates were rounded down to the nearest whole number [
7,
10].
For each method of induction, two authors (EM and KK) assigned a level of evidence based on the "GRADE" (Grading of Recommendations Assessment, Development and Evaluation) system [
11]. In this system, the overall strength of evidence is assigned not only based on study design and conduct, but also on factors such as the consistency and precision of the results and the likelihood of publication bias. Overall strength of evidence is classified in the GRADE system as high, intermediate, low or very low. The levels of evidence were assigned in the following manner. If the preponderance of evidence supporting a particular method of labor induction for the outcomes of interest is strong enough that further research would be unlikely to change the reviewers' confidence in the estimate of effect, the evidence quality was assessed as high [
11]. If further research would be likely to have an important impact on confidence in the estimate, the evidence quality was assessed as moderate [
11]. If further research would be very likely to have an important impact in the estimate of effect, the quality of evidence was assessed as low, and if the estimate of effect is very uncertain, the evidence was assessed as very low [
11].
These same authors (EM and KK) also assigned a balance of benefits and harms and a grade of recommendation according to GRADE system guidelines [
11,
12]. For each clinical intervention under study, the balance of benefits and harms is assessed, and a grade of recommendation is classified as strong or weak. This systematic review does not have a "stand alone" study protocol. In reporting outcomes from included study, we followed PRISMA guidelines [
13].
This is a systematic review of previously-published data and as such does not require ethics approval.
Discussion
Our best-evidence review of the literature suggests that many commonly-recommended methods for induction of labour have important trade-offs between benefits and harms. Compared with placebo, use of vaginal and cervical prostaglandin E2 was consistently associated with reduced likelihood of failure to deliver vaginally within 24 hours but increased risk for hyperstimulation with and without FHR changes. Vaginal misoprostol reduced failure to achieve vaginal delivery within 24 hours compared with vaginal and cervical PGE2, but increased uterine contractile abnormalities. Likewise, vaginal misoprostol reduced caesarean deliveries compared with IV oxytocin, but increased uterine hyperstimulation. Mechanical methods for induction of labour were associated with reduced rates of uterine hyperstimulation compared with vaginal PGE2 and vaginal misoprostol, but were also associated with increased risk for maternal and neonatal infectious complications in the one included systematic review that compared mechanical methods with all other methods pooled. Intravenous oxytocin with and without amniotomy did not appear to have significant benefits compared with vaginal PGE2.
Of the non-pharmacologic methods, membrane sweeping appeared to have the strongest evidence-base. It was successful in reducing post-term gestations without increasing clinically-important harms. There is not enough evidence of benefit/safety to recommend the other non-pharmacologic methods of breast stimulation and sexual intercourse.
Our review included evaluation of several investigational methods of induction of labour, of which hyaluronidase appears the most promising. In one small trial, hyaluronidase reduced the need for oxytocin augmentation and for caesarean delivery. These findings need to be confirmed in large, appropriately-powered randomised controlled trials.
Our review may have been limited by restricting our search to the English-language literature and by publication bias. Because we used the Cochrane hierarchy, we compared each method of labour induction only with methods above them on the Cochrane hierarchy list. This may have limited the total number of comparisons made. Likewise, the included studies contained heterogeneous populations of women with both intact and ruptured membranes and cervices favourable and unfavourable for induction. The large number of methods of induction considered in our review precluded subgroup analyses according to membrane status. Likewise, we were not able to consider variation in pharmacologic preparation and dosing of the different compounds under study. In our review of methods of induction of labor in the setting of unfavorable cervix, we did not identify a clear best choice for induction of labor in this setting.
Despite the large amount of evidence that we were able to bring to bear on several common methods of labour induction, we also found considerable imprecision surrounding benefits and harms of many of the included methods. Numbers of included women in most induction randomized trials were too small to exclude differences in rare adverse outcomes such as uterine rupture, amniotic fluid embolism, or perinatal asphyxia. Further research is necessary to identify potential risks and benefits of both commonly-used and investigational methods of induction of labour.
Competing interests
Drs. Mozurkewich, Romero, Berman, and Perni, are co-investigators on an ongoing multi-centre, industry-sponsored randomized controlled trial comparing the misoprostol vaginal insert with the dinoprostone vaginal insert.
Authors' contributions
KK and JC performed the literature searches required for this review and reviewed all abstracts. EM and JC performed an updated search of the literature and abstract review when it became necessary during manuscript preparation. EM and KK reviewed all full text articles. EM and JC performed all assessments of study quality. EM, KK, DB, VR, UP, and VK wrote and edited the manuscript. KK and VK participated in the formulation of the methods of this review and EM and KK assigned the evidence grades. All authors read and approved the final manuscript.