Induction of labor occurs in about 20% of term pregnancies but is associated with a lower rate of spontaneous vaginal delivery in comparison with spontaneously occurring labor.1
The physiologic basis of membrane sweeping leading to labor is well established and has been shown to release endogenous prostaglandins,2–4 phospholipase A3 and oxytocin (Ferguson’s reflex).5 Uterine contraction frequency is also increased by membrane sweeping.3 The release of prostaglandins can last at least 6 hours.4
Membrane sweeping alone is an effective but less efficient method of labor induction compared with more established methods such as vaginal prostaglandin or amniotomy with oxytocin infusion.6–8 A recent Cochrane meta-analysis of membrane sweeping trials concluded that although it reduces the number of women progressing to postterm gestation and the need for formal labor induction, sweeping is not associated with a reduction in cesarean or instrumental delivery, and it can cause discomfort and vaginal bleeding.8
We did a literature search on PubMed (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi) using search terms “membrane sweeping” or “membrane stripping” and “labor induction” on November 24, 2005, for studies in all languages published between January 1966 and November 2005 inclusive, looking for trials of membrane sweeping or membrane stripping and labor induction. Although many studies of membrane sweeping have been reported, these studies focus on attempts to reduce incidence of postdates and formal labor induction as identified in the Cochrane review.6 We found only 1 study that dealt specifically with membrane sweeping in conjunction with formal labor induction.9 That study demonstrated positive benefits of membrane sweeping only among nulliparas requiring prostaglandin for labor induction.
Because membrane sweeping is uncomfortable,6 we were interested in determining whether 1 systematic membrane sweep done at initiation of formal labor induction had any beneficial effects.
MATERIALS AND METHODS
A randomized trial was done to compare the effect of membrane sweeping to no sweeping at initiation of labor induction at term. The study was approved by the local institutional review board, and written consent was obtained in all cases. This study was done at a tertiary referral university hospital with more than 5,000 deliveries per annum in Kuala Lumpur, Malaysia.
Based on an earlier study,9 which reported a nonsignificant increase in the spontaneous vaginal delivery rate from 75% to 85% with sweeping, sample size calculation using an α of 0.05 and β of 0.8 indicated that 270 women were needed in each group for an appropriately powered randomized study on the effect of membrane sweeping in conjunction with formal labor induction. We analyzed our results after having recruited 274 women in total.
Women at term (37 to 42 weeks), with a singleton fetus, intact membranes, and cephalic presentation were recruited when they presented to the induction bay of the delivery suite for labor induction. Recruitment was carried out by the investigators or the staff of the induction bay. Women with a previous cesarean delivery, intrauterine fetal death, or known gross fetal anomalies were excluded. Recruitment took place from March 2004 to February 2005. All women who were approached agreed to participate (Fig. 1). Membrane sweeping for prior cervical ripening was not routinely offered in our institution.
Women who agreed to participate were first stratified into nulliparas and multiparas groups. Randomization was effected by the use of sealed opaque envelopes with a piece of paper inside bearing the legend “Sweep” or “No Sweep”. Envelopes were prepared in blocks of 20 (10 sweep and 10 no sweep) for each stratified group. Envelopes were then shuffled and placed in boxes marked “Nulliparas” and “Multiparas.” Boxes were refilled as required with blocks of 20 envelopes. For random assignment to treatment groups, an envelope was withdrawn from the appropriate box and allocated to the woman. Once allocated, an envelope was discarded if a woman chose to withdraw or recruitment was in error. The allocated envelope was opened by the clinician performing the initial vaginal examination just before that examination. A sticker bearing the identification of the randomized woman was affixed to the paper bearing the legend “Sweep” or “No Sweep,” and the paper was placed in a sealed drop box until unblinding at the end of the study.
The treatment allocation was not revealed to the women, and only the clinician performing the initial vaginal examination at initiation of labor induction was aware of the allocation. The investigators were also blinded to the treatment allocation. The treatment allocation was not recorded in the women’s charts. There was a potential weakness in the masking process because the clinician performing the sweep might have been involved in the continuing management of the woman. However, we had institutional guidelines on labor induction and labor, and there was no indication of systematic breach of the guidelines. Data were collected in a prospective manner.
The method of induction of labor was decided at the initial vaginal examination on the induction bay. Women assigned to “sweep” had their cervix swept by inserting the examining finger as high as possible past the internal cervical os, and the membranes were swept off the lower pole of the uterus by a complete circular sweep of the finger, once clockwise and once counterclockwise. In the event of a closed internal or external cervical os, the cervical canal or external cervix was swept with 2 circular motions. Dinoprostone (3 mg) pessary placement in the posterior fornix or amniotomy immediately followed. Bishop score was then recorded. Women assigned to “no sweep” had a straightforward cervical assessment for Bishop score followed immediately by either dinoprostone or amniotomy. We applied universal electronic fetal monitoring to our induced women.
Membrane sweeping was done only at the initiation of labor induction for sweep-assigned women. No subsequent sweeping was performed on the first day of labor induction for any women, even if the induction process was continuing. Women assigned to “no sweep” did not have any sweeping in conjunction with labor induction on the first day of induction. However, we did not prohibit sweeping in labor or sweeping if the woman had not gone into labor after the first 24 hours. All randomly assigned women received their allocated treatment.
A visual analog score (VAS) based on a scale of 0 to 10, with 0 representing no pain to 10 representing unbearably severe pain, was obtained from the women immediately after commencement of labor induction by a blinded investigator.
In our institution, dinoprostone pessary was used for labor induction if the Bishop score was unfavorable (≤ 4), and amniotomy usually was performed when the cervix was 3 cm or more dilated and the presenting part was low. In unfavorable women, after dinoprostone pessary insertion, the woman was assessed 6 hours later, at which time depending on the cervical dilatation and presence of contractions a further dinosprostone pessary might be inserted or amniotomy performed. We allowed a maximum of 2 doses of dinoprostone per day. We reassessed routinely again 6 hours later, and if the cervix remained unfavorable, women with nonurgent indications were usually rested overnight and the process repeated the following morning. Following amniotomy for labor induction, oxytocin was usually started within 2 hours if contractions were inadequate. The start of labor induction was taken as time of insertion of the first dinoprostone pessary or of amniotomy, depending on the method of labor induction used.
Once in established labor (regular contractions and cervical dilatation ≥ 3 cm), vaginal assessment was usually done every 4 hours initially unless otherwise indicated. Oxytocin was started for labor augmentation when labor progress fell below the action line in the partogram. Our intravenous oxytocin infusion regime started at 2 mU/min. A doubling in dosage was made every 30 minutes until contractions of 4 to 5 every 10 minutes were achieved or there was cardiotocograph abnormality. The maximum oxytocin dose was 32 mU/min. Once started, oxytocin infusion was continued to delivery unless otherwise indicated.
As soon as possible after delivery, another VAS was obtained by the investigator from the women to gauge their perception of the birth process before their hospital discharges. The scale ranged from 0 to 10, with 0 representing very satisfied to 10 representing very dissatisfied.
Outcome measures collected included mode of delivery, induction to delivery interval, any use of oxytocin in labor, duration of oxytocin infusion in labor, dose of dinoprostone used, and VAS results. Secondary outcomes recorded included meconium staining of liquor, Apgar score at 5 minutes, umbilical cord blood pH value at delivery, admission for neonatal care, indications for neonatal care admissions, indications for operative delivery, induction to delivery interval 12 hours or less, and induction delivery interval of 24 hours or less. Secondary analysis based on parity was also planned.
Data were entered into a statistical software package SPSS 11 (SPSS Inc., Chicago, IL), and GraphPad Instat and QuickCalcs software (GraphPad Software Inc., San Diego CA) were also used for data analysis. The SISA software (Uitenbroek, DG; http://home.clara.net/sisa/fiveby2.htm, accessed November 12, 2005) was used to perform Fisher exact tests with data sets larger than 2 × 2. The Kolmogorov-Smirnov test was used to check for normal distribution. The t test was used to analyze means and distributions with the Mann-Whitney U test used to check for consistency in the event that a t test was applied on nonnormal data and ordinal data, the Fisher exact test was used for categorical data sets (up to 2 × 5), χ2 for larger categorical data sets, and relative risk (RR) and its 95% confidence interval (CI) calculated using GraphPad Instat program. Numbers needed to treat and its 95% CI were generated with GraphPad QuickSeal. P < .05 in any test was considered statistically significant, and all tests used 2-tailed results.
We recruited a total of 274 women (138 nulliparas and 136 multiparas). Ten multiparas were excluded from the analysis (4 sweep and 6 no sweep) because of exclusion criteria infringement (6 with previous cesarean section scars, 3 prelabor rupture of membranes, and 1 with known gross fetal anomaly). This left 264 women for analysis; 138 nulliparas (70 sweep and 68 no sweep) and 126 multiparas (66 sweep and 60 no sweep). Two hundred twenty-eight (86.4%) women underwent dinoprostone induction and, only 36 (13.6%) had primary amniotomy for labor induction. There was no significant difference between the groups in any characteristic listed (Table 1 and Table 2).
Distributions of duration of oxytocin use, induction to delivery interval, dose of dinoprostone use, and VAS were found to be nonnormal. Because our sample sizes were large, we used the t test for primary analysis and presented our finding as mean ± standard deviation. The Mann-Whitney U test was also used for analysis on these outcomes. No significant inconsistency was found between t test and Mann-Whitney U test P values.
Swept women had higher spontaneous vaginal delivery rate (69% compared with 56%, P = .041), shorter induction to delivery interval (mean 14 hours compared with 19 hours, P = .003), fewer that required oxytocin use (46% compared with 59%, P = .037), shorter duration of oxytocin infusion (mean 2.6 hours compared with 4.3 hours, P = .001) and improved VAS for birth process satisfaction (mean 4.0 compared with 4.7, P = .015). The reduction in dinoprostone dose used (mean 1.2 compared with 1.3, P = .082) was not significant. Postsweeping VAS for pain (mean 4.7 compared with 3.5, P < .001) was significantly increased (Table 3).
The number needed to treat to achieve an additional spontaneous vaginal delivery was 8 (95% CI 4–78). The RR for operative delivery was 0.71 (95% CI 0.51–0.97). The number needed to treat to avoid the use of oxytocin in labor was also 8 (95% CI 4–90). The RR for any oxytocin use was 0.78 (95% CI 0.62–0.98). Even if we analyzed women (n = 139; 63 swept 76 no sweep) who only required oxytocin in labor, the mean ± standard deviation duration of use was still significantly shorter among swept women (5.6 ± 3.5 hours compared with 7.2 ± 3.5 hours, P = .007).
Secondary outcomes of induction delivery interval 12 hours or less (57% sweep compared with 48% no sweep, P = .175) and induction delivery interval 24 hours or less (86% sweep compared with 77% no sweep, P = .057) did not show any significant differences.
Planned secondary analysis based on subgroups of nulliparas and multiparas (Table 3) showed significant beneficial results among swept nulliparas in any oxytocin use, duration of oxytocin use, induction to delivery interval, and VAS satisfaction score. Swept multiparas had significant beneficial results in induction to delivery interval and mean dose of dinosprostone use. However, if analysis of dinoprostone use in multiparas was based on multiparas who had undergone dinoprostone induction excluding those who had amniotomy induction, the result was no longer significant (P = .078). Swept nulliparas and multiparas had significantly higher VAS pain score. All the other main outcomes analyzed favored swept multiparas, even if statistical significance was not achieved. This study was not sufficiently powered to examine the effect of sweeping on subgroups.
Relative risks for operative delivery among a number of subgroups are shown on Figure 2. Statistically significant reductions in RR among swept women were obtained in the whole and Bishop score 5 or more groups only. All other groups also had RR less than 1, but 95% CI crossed 1.
Among swept women who had amniotomy induction, despite the small numbers (20 sweep, 16 no sweep), significant benefits, expressed as mean ± standard deviation (unless otherwise stated), were demonstrated in oxytocin infusion duration (3.5 ± 3.1 hours compared with 6.9 ± 3.7 hours, P = .005), induction to delivery interval (5.3 ± 2.4 hours compared with 7.7 ± 3.5 hours, P = .019), and birth process VAS (mean 2.4 ± 0.9 compared with 4.0 ± 2.2, P = .005). This result must be interpreted with caution because the numbers were small.
Labor complications and adverse neonatal outcomes were similar between sweep and no sweep groups (Table 2). However, our study was not powered to investigate the uncommon adverse neonatal events. Indications for operative delivery were also similar in both sweep and no sweep groups (Table 2).
Initial Bishop score was similar for both groups, 5.1 ± 1.8 compared with 4.7 ± 2.0 (P = .098). Further exploration of this possible trend indicated 1 component of Bishop score (cervical dilatation) showed a statistically significant difference (Fisher exact test, P = .024), with the swept group having fewer women with closed cervical os (Fig. 3). The other 4 components of Bishop score (cervical length, P = .63; cervical consistency, P = .15; station of the head, P = .99; and cervical position P = .36) did not show any difference between the sweep or no sweep groups.
Induction of labor involves starting a process which is probably self-sustaining once labor is established, and it is believed to involve a self-perpetuating cascade of endogenous release of prostaglandins.10
The results of this study indicated that membrane sweeping at initiation of formal labor induction in conjunction with established methods of labor induction had beneficial effects. The women recruited had their labor induced for a broad range of indications and represented the majority of women undergoing labor induction today.1,11,12 Membrane sweeping is simple and quick, requires no equipment, and possibly needs only to be performed at the initiation of formal labor induction. Swept women expressed higher satisfaction with the birth process, even though sweeping was initially more painful. The finding that sweeping is painful has been shown.13,14 Neonatal outcomes (Table 2) were similar, and this is in keeping with the findings from a recent meta-analysis.6
The additive beneficial effect of membrane sweeping with vaginal prostaglandin gel has been demonstrated by Doane and McCarty,15 who show that compared with no intervention, sweeping alone, or prostaglandin gel alone, the combined approach reduces postterm pregnancies and antenatal visits. The combined treatment group has a median intervention-to-delivery interval of 1 day, indicating that it is an effective mode of labor induction. In our study, the median induction-to-delivery interval was 11 hours in the sweep group. Our study is consistent with the earlier finding that membrane sweeping with vaginal prostaglandin may have synergistic effects.
Foong et al9 reported the only other randomized study of membrane sweeping in conjunction with formal labor induction and found reductions in cesarean delivery, induction labor interval, and total oxytocin dose used, but only in the subgroup of nulliparas who had required dinoprostone induction. In contrast, beneficial effects of sweeping were found across subgroups (Fig. 2) in our study. In our study, the number of women undergoing amniotomy induction was low, and further powered study is required. There are important differences between the study of Foong et al9 and our study. Our study involved 1 sweep at initiation of labor induction compared with repeated sweeping until established labor. The earlier study had more patients amenable to amniotomy induction (33.5% compared with 13.6%) and lower overall operative delivery rate (19.4% compared with 37.1%) compared with this study. Because it was the standard protocol in both our institutions to perform amniotomy when the cervix was 3 cm or more dilated, higher cervical dilatation in their subjects may be the reason for less operative delivery, because initial cervical dilatation is the best predictor for operative delivery.11,12 The RR for operative delivery in our study was 0.71 (95% CI 0.51–0.97) compared with 0.60 (95% CI 0.35–1.02) for the earlier study.
The above indicated that membrane sweeping at initiation of formal induction was clearly beneficial but because our study did not prohibit subsequent sweeping after 24 hours nor formally assess subsequent sweeping, it was not possible to determine the precise effect of repeat sweeping. Further study is needed to clarify this issue. However, because 83% of our women delivered within 24 hours from commencement of labor induction and because we did not perform any additional sweeps in conjunction with labor induction within the first 24 hours in any study women, the positive effect of repeat sweeping was likely to be low. Because prostaglandin release after sweeping lasts for at least 6 hours,6 repeat sweeping may not be necessary in formal labor induction where additional interventions are planned within 6 hours. Serial membrane sweeping during established nulliparous labor is not useful16
Women who were randomly assigned to sweeping in this study were found to have a decreased incidence of having a closed cervix at the initial Bishop score assessment (Fig. 3). This was unlikely to be due to a chance randomization event, because all other Bishop score components and all other subject characteristics were similar (Tables 1 and 2). Our study was pragmatically designed with only 1 vaginal examination at initiation of labor induction. Because Bishop score was routinely recorded after sweeping, some clinicians may thus have recorded the postsweep cervical dilatation for the Bishop score, assuming that sweeping caused the cervix to dilate. This may help explain sweeping’s beneficial effect in our study, because the mechanically dilated and disrupted cervix might offer less resistance to further dilation. If membrane sweeping did cause cervical dilatation, more swept women might be expected to have undergone amniotomy induction. Amniotomy induction was 14.5% compared with 12.7% in sweep compared with no sweep women, respectively, in our study, which was not significant. This was expected, because our study women generally had closed or minimally dilated cervix (Fig. 3), and the limited upgrading probably did not permit significantly more amniotomy induction to be performed. Interestingly in the study of Foong et al,9 where amniotomy induction was commoner, the ratio for amniotomy induction was 39.5% compared with 27.4% (P = .059) among sweep and no sweep women, respectively, despite prior matching for method of induction during randomization. Further study is required to see whether membrane sweeping has immediate effects on cervical dilation and resistance. Based on the results of our study, membrane sweeping had beneficial results, and sweeping at initiation of formal labor induction at term should be considered.
2. Mitchell MD, Flint AP, Bibby J, Brunt J, Arnold JM, Anderson AB, et al. Rapid increases in plasma prostaglandin concentrations after vaginal examination and amniotomy. Br Med J 1977;2:1183–5.
3. McColgin SW, Bennett WA, Roach H, Cowan BD, Martin JN Jr., Morrison JC. Parturitional factors associated with membrane stripping. Am J Obstet Gynecol 1993;169:71–7.
4. Manabe Y, Yoshimura S, Mori T, Aso T. Plasma levels of 1314-dihydro-15-keto prostaglandin F2alpha, estrogens, and progesterone during stretch-induced labor at term. Prostaglandins 1985;30:141–52.
5. Chard T, Gibbens GL. Spurt release of oxytocin during surgical induction of labor in women. Am J Obstet Gynecol 1983;147:678–80.
6. Boulvain M, Stan C, Irion O. Membrane sweeping for induction of labor. Cochrane Database Syst Rev 2005; 25.
7. Kelly AJ, Kavanagh J, Thomas J. Vaginal prostaglandin (PGE2 and PGF2a) for induction of labor at term. Cochrane Database Syst Rev 2003;4:CD003101.
8. Howarth GR, Botha DJ. Amniotomy plus intravenous oxytocin for induction of labor. Cochrane Database Syst Rev 2001;3:CD003250.
9. Foong LC, Vanaja K, Tan G, Chua S. Membrane sweeping in conjunction with labor induction. Obstet Gynecol 2000;96:539–42.
10. Chez RA. Cervical ripening. Clin Obstet Gynecol 1998;41:606–10.
11. Wing DA, Tran S, Paul RH. Factors affecting the likelihood of successful induction after intravaginal misoprostol application for cervical ripening and labor induction. Am J Obstet Gynecol 2002;186:1237–40.
12. Vrouenraets FP, Roumen FJ, Dehing CJ, van den Akker ES, Aarts MJ, Scheve EJ. Bishop score and risk of cesarean delivery after induction of labor in nulliparous women. Obstet Gynecol 2005;105:690–7.
13. Boulvain M, Fraser WD, Marcoux S, Fontaine JY, Bazin S, Pinault JJ, et al. Does sweeping of the membranes reduce the need for formal induction of labor? A randomised control trial. Br J Obstet Gynaecol 1998;105:34–40.
14. Wong SF, Hui SK, Choi H, Ho LC. Does sweeping of membranes beyond 40 weeks reduce the need for formal induction of labor? BJOG 2002;109:632–6.
15. Doany W, McCarty J. Outpatient management of uncomplicated postdate pregnancy with intravaginal prostaglandin E2 gel and membrane stripping. J Matern Fetal Med 1997;6:71–8.
16. Chanrachakul B, Suthutvoravut S, Sangthawan M, Herabutya Y. Effects of lower uterine segment sweeping on progress of labor in nullipara. J Med Assoc Thai 2001;84:1582–6.