Inpatient vaginal dinoprostone vs outpatient balloon catheters for cervical ripening in induction of labor: An individual participant data meta‐analysis of randomized controlled trials

Abstract

Introduction

Outpatient cervical ripening and induction of labor might offer potential benefits. There are a few randomized controlled trials (RCTs) comparing outpatient balloon catheters with inpatient vaginal dinoprostone, but the reported outcomes among these trials were inconsistent, justifying the need for a meta‐analysis. We aimed to evaluate the effectiveness and safety of inpatient vaginal dinoprostone compared to outpatient balloon catheters for cervical ripening in labor induction.

Material and Methods

Eligible RCTs were identified using MEDLINE, Emcare, Embase, Scopus, CINAHL Plus, Cochrane Pregnancy and Childbirth Group's Trials Register, WHO International Clinical Trials Registry Platform, and clinicaltrials.gov from inception to July 2024. Women with live singleton pregnancies at 34 or more weeks of gestation were eligible. The authors of eligible trials were invited to share their de‐identified data. The main outcomes were vaginal birth and a composite adverse perinatal and maternal outcome. All analyses were adjusted for age and parity. Two‐stage random effects meta‐analysis was the main analysis strategy with the intention‐to‐treat principle. This meta‐analysis was registered with PROSPERO (CRD42022313183) on 27‐04‐2022.

Results

We identified three eligible RCTs, and all three shared data (N = 1636); inpatient vaginal dinoprostone (n = 832), outpatient balloon catheter (n = 804). The odds of vaginal birth were higher after inpatient vaginal dinoprostone than outpatient balloon catheter (67.8% vs 61.7%, adjusted odds ratio [aOR] 1.30, 95% CI 1.05–1.62, I ² = 0%). There was no significant difference in the composite adverse perinatal outcome (13.7% vs 13.1%, aOR 1.09, 95% CI 0.75–1.58, I ² = 28.7%) or the composite adverse maternal outcome (16.6% vs 19.8%, aOR 0.81, 95% CI 0.61–1.07, I ² = 11.5%). The difference in effect on vaginal birth rate varied according to body mass index. Overweight and obese women had a lower vaginal birth rate after outpatient induction, whereas for those with underweight/normal weight, the rates of vaginal birth were similar.

Conclusions

Balloon catheter used in an outpatient labor induction setting probably leads to fewer vaginal births compared to vaginal dinoprostone in an inpatient setting. In pre‐planned subgroup analysis, for pregnant women with underweight/normal weight, both inpatient vaginal dinoprostone and outpatient balloon catheter methods are viable options, but balloon catheter has a lower vaginal birth rate in women with overweight and obesity during pregnancy. The perinatal and maternal safety profiles are comparable.

Cervical ripening in the induction of labor with outpatient balloon catheters is probably less effective than inpatient vaginal dinoprostone, specifically in maternal overweight and obesity during pregnancy. The perinatal and maternal safety profiles are comparable.

Abbreviations

aOR

adjusted odds ratios

BMI

body mass index

CI

confidence intervals

IOL

induction of labor

IPD

individual participant data

MD

mean difference

RCT

randomized controlled trial

Key message.

Cervical ripening in the induction of labor with outpatient balloon catheters is probably less effective than inpatient vaginal dinoprostone, specifically in maternal overweight and obesity during pregnancy. The perinatal and maternal safety profiles are comparable.

1. INTRODUCTION

Induction of labor (IOL) is the process of triggering birth by stimulating the changes in the cervix and the onset of labor. It is one of the most common obstetric interventions, occurring in about 30% of births in high‐resource countries and one in ten births globally. ¹ , ² The rate of IOL has seen a significant increase in several countries. For instance, in the United States, it rose from 9% in 1989 to 30% in 2020. ³ Similarly, Australia witnessed an increase in the IOL rate from 26% in 2011 to 34% in 2021. ² In New Zealand, the rate increased from 19.4% in 2009 to 26.4% in 2018. ⁴

IOL is performed when the risks of continuing the pregnancy appear to outweigh the benefits of doing so, as perceived by the woman and her caregivers. In women with an unfavorable cervix, cervical ripening is essential for successful IOL. ⁵ Pharmacological preparations, including exogenous use of prostaglandin preparations such as misoprostol or dinoprostone, or mechanical methods such as balloon catheters and cervical osmotic dilators, are employed. ⁶ , ⁷ Balloon catheters exert pressure and traction on the cervix, releasing local prostaglandins and inflammation, whereas prostaglandins promote cervical remodeling and contractility of the myometrium, leading to cervical ripening in IOL. ⁶ , ⁷ Over the last decade, low‐dose oral misoprostol has been increasingly recommended due to the higher vaginal birth rate, cost‐effectiveness, and lower risk of uterine hyperstimulation. ¹ , ⁸ , ⁹ , ¹⁰ , ¹¹ IOL is not without risks. Approximately 20% of IOL results in progression to cesarean section in women with an unfavorable cervix. ¹² Therefore, identifying evidence‐based and optimal labor induction methods is crucial to promote positive outcomes for all mothers and newborns. Recently, there has been increased interest in outpatient induction. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ There is limited evidence from qualitative research that outpatient IOL yields positive birth experiences in terms of support from their partner and increased freedom of movement and self‐expression as long as shared decision‐making takes place. ¹⁶ , ¹⁷

The latest Cochrane review of mechanical methods for IOL suggests balloon catheters may have comparable effectiveness to vaginal dinoprostone with regard to vaginal delivery not achieved within 24 h (risk ratio 1.01, 95% confidence interval 0.82–1.26; 7 studies with 1685 women; low‐quality evidence). ¹⁸ Balloon catheters were also shown to lower the risk of uterine hyperstimulation with fetal heart rate changes and neonatal severe morbidity and mortality. ¹⁸ , ¹⁹ However, no firm conclusion can be drawn for the maternal safety profile. ¹⁸ , ¹⁹ None of the existing aggregate data meta‐analyses on outpatient versus inpatient settings performed an exclusive comparison investigating clinical outcomes derived from outpatient care using balloon catheters and vaginal dinoprostone in the inpatient setting, the two most widely used labor induction methods in outpatient and inpatient settings, respectively. ¹⁴

To date, a few randomized controlled trials (RCTs) have compared outpatient balloon catheters with inpatient vaginal dinoprostone for IOL, ²⁰ , ²¹ , ²² but the reported outcomes among these trials were inconsistent, justifying the need to pool the data in a meta‐analysis. Importantly, the rare incidence of adverse outcomes can make it challenging for aggregate data meta‐analyses to generate meaningful conclusions on safety outcomes. An individual participant data (IPD) meta‐analysis can overcome these issues by ensuring standardized definitions, assessing unreported outcomes and data trustworthiness, improving statistical power, and allowing evaluations on subgroup analysis. ¹⁹ , ²³ , ²⁴ , ²⁵ , ²⁶ We, therefore, performed an IPD meta‐analysis that compares the effectiveness (rate of vaginal delivery) and safety of inpatient vaginal dinoprostone and outpatient balloon catheters for IOL. We hypothesized that cervical ripening in IOL with inpatient vaginal dinoprostone is equally effective and safe as cervical ripening in IOL with an outpatient balloon catheter.

2. MATERIAL AND METHODS

This IPD meta‐analysis was prospectively registered in PROSPERO (CRD42022313183) and received ethical approval from the Monash Health Human Research Ethics Committee (Reference No. RES‐22‐0000‐119Q). All the included RCTs had ethical approval from appropriate ethical committees. We report findings in this paper following the PRISMA‐IPD statement. ²⁵

2.1. Data sources

We searched databases including Ovid MEDLINE, Ovid Embase, Ovid Emcare, CINAHL Plus, Scopus, Cochrane Pregnancy and Childbirth Group's Trials Register, WHO International Clinical Trials Registry Platform, and ClinicalTrials.gov, for published and unpublished RCTs from inception to July 2024. Search strategy keywords (“Foley” OR “Cook” OR “Atad” OR “misoprostol” OR “dinoprostone” OR “prostaglandin E1” OR “prostaglandin E2” OR “cervical ripening” OR “Labor, induction”) AND (“outpatient” OR “ambulatory care” OR “home” OR “out of hospital”) were combined with MeSH terms to identify suitable trial reports. Reference checking was also conducted to find potentially eligible studies.

2.2. Outcomes and effect modifiers

The primary outcomes were vaginal delivery, a composite measure of adverse perinatal and maternal outcomes. The perinatal composite outcome included one or more neonatal deaths, stillbirth, acidosis (pH <7.1), neonatal Apgar score <7 at 5 min, neonatal seizures, neonatal intensive care unit admission, hypoxic–ischemic encephalopathy of any stage, meconium aspiration syndrome, cord prolapse, endotracheal intubation, external cardiac compressions, and neonatal infection either clinically suspected (as defined by neonatal antibiotic administration) or proven with culture. The maternal composite outcome comprised one or more of maternal mortality, maternal admission to the intensive care unit, post‐partum hemorrhage ≥1000 mL, uterine rupture, or maternal infection (defined as a temperature ≥38°C at any time during labor or delivery, antibiotics use, or clinically diagnosed infection, such as endometritis).

Secondary outcomes included unassisted vaginal birth rate, cesarean section rate, instrumental delivery rate, indications for cesarean and instrumental delivery, requirement for oxytocin infusion, uterine hyperstimulation (defined as either tachysystole or hypertonus with a non‐reassuring fetal heart‐rate tracing on cardiotocography), need for more methods for cervical ripening and IOL, modified Bishop score, induction‐to‐delivery time, admission‐to‐delivery interval, total maternal and neonatal stay, and use of epidural and opioid analgesia during labor. Additionally, we reported individual components of the composite measure of adverse perinatal outcomes and the composite measure of adverse maternal outcomes.

2.3. Eligibility criteria

We selected RCTs comparing outpatient balloon catheters and inpatient vaginal dinoprostone for labor induction in women with an unfavorable cervix. Cluster RCTs and cross‐over trials were excluded. Only data from women with a live singleton fetus with more than 34 weeks of gestation were analyzed. No restrictions were placed based on membrane status, preterm labor, previous cesarean, or scar pregnancy. RCTs that used other forms of prostaglandins or alternative routes for the administration of dinoprostone were excluded. Two investigators (FC and MP) independently screened the titles, abstracts, and full texts of the potentially relevant RCTs using the online platform COVIDENCE. ²⁷ Any discrepancies were reviewed by a third researcher (WL), and a consensus was reached.

2.4. Data extraction

We liaised with each eligible RCT's primary or corresponding author with the request to share de‐identified raw data. For cases where there was no response after two rounds of fortnightly email invitations, we attempted to contact the co‐authors and the last authors involved in the trials. If that was also unresponsive, we approached the institutions associated with these authors and any researcher with whom the trial investigators may have recently collaborated Discussion and agreement on the protocol and methodology were undertaken with the authors of the included studies.

We requested the principal authors to complete an Excel datasheet or send their participant‐level data along with the data code dictionary. Data for all randomized participants in these RCTs were deidentified and password‐protected upon transfer. The received data were checked for excluded or missing data, internal consistency, errors, the pattern of data presentation and randomization, and consistency with the publication. Any identified issues were clarified with the trial investigators before being accepted for the analysis. Contributing trial investigators were invited to comment on the scope of the project, draft protocol, result interpretation, and manuscript editing. They also participated in teleconference meetings as the project progressed.

2.5. Assessment of risk of bias

Two investigators (FC and MP) independently conducted the risk of bias (RoB) assessment using the Cochrane Risk of Bias 2 tool, with a third reviewer (WL) to resolve disagreements. ²⁸ Trialists were consulted if the information was inadequate.

2.6. Data analysis

All three datasets were checked for integrity and reproducibility with the respective publications. ²⁰ , ²¹ , ²² They were cleaned according to a single data code dictionary. All analyses followed the intention‐to‐treat principle. Data were synthesized using a two‐stage random‐effects model to accommodate the clinical and methodological heterogeneity between studies. All analyses adjusted for maternal age and parity. Stage one of the analysis involved evaluating the primary and binary secondary outcomes using adjusted odds ratios (aOR) with 95% confidence intervals (CI) for each trial. The mean difference (MD) was computed for the hospital stay. Stage two of the analysis involved pooling the summary estimates derived from the first stage through a random‐effects model with the DerSimonian Laird method. For induction to vaginal birth time, subdistribution hazard ratios and 95% CIs using the subdistribution hazard model were estimated, which considered cesarean section as a competing risk. Statistical heterogeneity was quantified using I ² statistics, with heterogeneity considered substantial if I ² exceeded 50%. For adverse outcomes, an aOR >1.0 indicates that the vaginal dinoprostone group may confer a greater risk, whereas, for positive outcomes (e.g, vaginal delivery), a subdistribution hazard ratio or OR >1.0 indicates that the vaginal dinoprostone group may be more favorable.

We examined treatment–covariate interactions for the primary outcome, vaginal delivery rate between treatment and baseline covariates, including gestational age at induction, parity, maternal age, body mass index (BMI), and initial modified Bishop score. Adjusted odds ratios (aOR) for the interaction terms were combined with a random‐effects model identical to the primary analysis. Statistical analysis was performed with STATA 17.0 (Stata Corp, College Station, TX, USA) and SPSS (SPSS Statistics version 29.0, IBM corporation). All findings were reported as inpatient vaginal dinoprostone compared with outpatient balloon catheters (reference group). The combination of the setting and methods is being evaluated in this IPD meta‐analysis.

3. RESULTS

3.1. General characteristics of the studies

Of the 1734 studies identified from the literature search, 36 articles underwent full‐text screening. Three RCTs were eligible for inclusion, and all authors agreed to contribute to this IPD meta‐analysis (Figure 1). The final analysis comprised data on 1636 women who were randomized, with 832 (50.9%) allocated to the inpatient vaginal dinoprostone group and 804 (49.1%) to the outpatient balloon catheter group.

FIGURE 1.

Trial identification (PRISMA‐IPD flow diagram). RCTs, randomized controlled trials.

The participating RCTs were undertaken across Australia and New Zealand. The baseline characteristics of women randomized to the inpatient vaginal dinoprostone and outpatient balloon catheter groups were comparable with respect to age, gestation, and pre‐induction modified Bishop score. The mean BMI values across contributing trials were 23.5, 28.3, and 27.5, as detailed in Table 1. Overall, more women were nulliparous (75.8%), of Caucasian ethnicity (68.0%), and undergoing IOL due to prolonged pregnancy (45.5%) (Table 1). Detailed information about the study characteristics of each included RCT with its eligibility criteria and types of interventions, including the median doses of dinoprostone received, balloon type, volume inflated, and duration, has been summarized in Table 2.

TABLE 1.

Baseline characteristics of each included randomized controlled trial.

Characteristics	Inpatient group	Outpatient group
Henry 2013 (Australia)	N = 51	N = 50
Maternal age (years), mean (SD)	32.9 (5.1)	32.7 (4.4)
BMI kg/m2, mean (SD)	23.0 (4.2)	24.1 (4.2)
Gestational age (weeks), median (IQR)	41.2 (1.1)	41.2 (0.8)
Ethnicity
Caucasian	41/51 (80.4)	41/50 (82.0)
Hispanic	‐	‐
Asian	9/51 (17.6)	5/50 (10.0)
Black	‐	‐
Other	1/51 (2.0)	4/50 (8.0)
Initial Bishop score, median (IQR)	2.9 (1.6)	2.7 (1.7)
Parity, n/N (%)
0	46/51 (90.2)	45/50 (90.0)
≥1	5/51 (9.8)	5/50 (10.0)
Indication for IOL, n/N (%)
Other/unknown	4/51 (7.8)	2/50 (4.0)
Hypertensive disorders	2/51 (3.9)	3/50 (6.0)
Post‐dated pregnancy	35/51 (68.6)	38/50 (76.0)
Diabetes/gestational diabetes	6/51 (11.8)	3/50 (6.0)
Fetal growth restriction	3/51 (5.9)	1/50 (2.0)
Obstetric cholestasis	1/51 (1.9)	3/50 (6.0)
No missing data
Beckmann 2019 (Australia)	N = 233	N = 215
Maternal age (years), mean (SD)	30.7 (5.1)	30.2 (5.3)
BMI, mean (SD)	28.5 (6.6)	28.1 (6.3)
Gestational age (weeks), median (IQR)	41.0 (0.9)	41.0 (1.1)
Ethnicity
Caucasian	194/233 (83.3)	172/215 (80.0)
Hispanic	‐	‐
Asian	23/233 (9.9)	21/215 (9.8)
Black	1/233 (0.4)	2/215 (0.9)
Other	14/233 (6.0)	20/215 (9.3)
Initial Bishop score, median (IQR)	3.0 (2.0)	4.0 (2.0)
Parity, n/N (%)
0	155/233 (66.5)	157/215 (73.0)
≥1	78/233 (33.5)	58/215 (27.0)
Indication for IOL, n/N (%)
Other/unknown	18/233 (7.7)	21/215 (9.8)
Hypertensive disorders	0	0
Post‐dated pregnancy	172/233 (73.8)	152/215 (70.7)
Diabetes/gestational diabetes	10/233 (4.3)	8/215 (3.7)
Fetal growth restriction	0	0
Advanced maternal age	9/233 (3.9)	10/215 (4.6)
Elective/social reasons	24/233 (10.3)	24/215 (11.2)
Wise 2023 (New Zealand)	N = 548	N = 539
Maternal age (years), mean (SD)	32.3 (5.0)	32.3 (5.5)
BMIa, mean (SD)	27.6 (6.7)	27.4 (6.6)
Gestational age (weeks), median (IQR)	40.0 (2.1)	40.0 (2.2)
Ethnicity
Caucasian	350/548 (63.9)	315/539 (58.4)
Hispanic	‐	‐
Asian	92/548 (16.8)	109/539 (20.2)
Black	‐	‐
Other	106/548 (19.3)	115/539 (21.3)
Initial Bishop scorea, median (IQR)	3.0 (2.0)	3.0 (2.0)
Parity, n/N (%)
0	417/548 (76.1)	420/539 (77.9)
≥1	131/548 (23.9)	119/539 (22.1)
Indication for IOL, n/N (%)
Other/unknown	68/548 (12.4)	64/539 (11.9)
Hypertensive disorders	36/548 (6.6)	41/539 (7.6)
Post‐dated pregnancy	177/548 (32.3)	171/539 (31.7)
Diabetes/gestational diabetes	136/548 (24.8)	146/539 (27.0)
Fetal growth restriction	54/548 (9.8)	51/539 (9.5)
Oligohydramnios	5/548 (0.9)	3/539 (0.6)
Advanced maternal age	28/548 (5.1)	34/539 (6.3)
Obstetric cholestasis	11/548 (2.0)	12/539 (2.2)
Elective/social reasons	33/548 (6.0)	17/539 (3.1)

Abbreviations: BMI, body mass index; IQR, interquartile range; SD, standard deviation.

TABLE 2.

Study characteristics of each included randomized controlled trial.

Study characteristics	Henry 2013 (N = 101)	Beckmann 2019 (N = 448)	Wise 2023 (N = 1087)
Eligibility criteria of the included trials
Inclusion criteria	Primip + multip low‐risk women with a singleton, vertex presentation, ≥18 years old, ≥37 weeks of gestation, Bishop Score <7 and cervical dilation <2 cm, normal CTG	Primip + multip low‐risk women with a singleton, vertex presentation, ≥37 weeks of gestation, undergoing IOL for low‐risk indications including post‐term, “social” or “elective” reasons and advanced maternal age (≥40 years) and Bishop Score <7	Primip + multip low‐risk women with a singleton, vertex presentation, ≥37 weeks of gestation, intact membranes, normal non‐stress test, Bishop score <7, and able to remain within 1 h of the hospital with someone who could speak sufficient English
Exclusion criteria	Unsuitable for outpatient management, Unsuitable for randomization to either dinoprostone (e.g., previous CS) or catheter use (e.g., latex allergy), or prior attempted IOL in this pregnancy, Bishop score ≥7 or cervical dilatation ≥2 cm, ruptured membranes, or evidence of regular uterine contractions at time of booked induction, multiple pregnancy or non‐vertex presentation, Unable to give informed consent	Major congenital abnormality, SGA, previous CS, Bishops score ≥7, high head (≥4/5), residing >60 min from hospital, or any contraindication to vaginal birth	Previous CS, major fetal anomaly, suspected severe FGR, and a maternal or fetal condition for which the clinician felt that outpatient care was contraindicated
Types of interventions in the included trials
Inpatient group	Inpatient initial 2 mg dinoprostone gel for nulliparous and 1 mg dinoprostone gel for parous women into posterior vaginal fornix and repeated in 6 h, with a median (range: minimum–maximum) dose of 2 (1–4). Only five patients received three doses and one patient received four doses	Inpatient dinoprostone either as 2 mg dinoprostone vaginal gel (Prostin®) or 10 mg dinoprostone controlled‐release vaginal tape (Cervidil®) and dinoprostone gel was repeated six hourly, with a median (range: minimum–maximum) dose of 1 (1–3). Eleven patients received three doses	Inpatient dinoprostone either as dinoprostone gel (Prostin E2) 1 or 2 mg, depending on clinician preference or a controlled‐release dinoprostone 10 mg pessary (Cervidil®). After 6 h (for gel) or 12–24 h (for the pessary), reassessed and ARM performed if possible or repeated dinoprostone administration until ARM was possible (maximum 6 gels or 2 pessaries) or spontaneous rupture of membranes or labor occurred or if the patient or clinician wanted to change to second method (inpatient balloon), with a median (range: minimum–maximum) dose of 1 (1–6). Of note, 24 patients received four doses of dinoprostone gel. Eight patients received five doses of dinoprostone gel and one patient received five doses of dinoprostone 10 mg pessary. Two patients received six doses of dinoprostone gel
Outpatient group	Outpatient 16 F standard latex Foley Catheter inflated with 30 mL of sterile water; discharged after reassuring CTG	Outpatient double‐balloon catheter (Cook®) inflated with 80 mL of sterile water for a maximum of 12 h; post‐balloon CTG was not routinely done, only 30 min of observation	Outpatient single 50 mL Foley balloon catheter (Bard, 2‐way, 20F) for 18–24 h. On return to hospital, ARM was performed if possible or switched to a second method (inpatient dinoprostone). Did not need a routine CTG after balloon catheter placement, unless there is clinical concern
Additional/second methods after the first method failed
Inpatient group	Further dinoprostone was given or inpatient Foley catheter inserted	A dose of dinoprostone gel was administered followed by an attempt at an amniotomy 6 h later. If this was again unsuccessful, another dose of dinoprostone gel was administered. If at the subsequent review, the amniotomy was still not possible and there remained a clear indication for delivery, CS was offered	Inpatient Foley balloon catheter
Outpatient group	Inpatient dinoprostone	A dose of dinoprostone gel was administered followed by an attempt at an amniotomy 6 h later. If this was again unsuccessful, another dose of dinoprostone gel was administered. If at the subsequent review, the amniotomy was still not possible and there remained a clear indication for delivery, CS was offered	Inpatient dinoprostone either as dinoprostone gel (Prostin E2) or a controlled‐release dinoprostone 10 mg pessary (Cervidil®)

Abbreviations: ARM, artificial rupture of membranes; CS, cesarean section; CTG, cardiotocogram; FGR, fetal growth restriction; IOL, induction of labor.

3.2. Risk of bias of included studies

On screening for risk of bias, two RCTs were labeled as having “some concerns” owing to the inability to blind participants to the outpatient procedure. ²¹ , ²² One RCT was categorized as overall ‘high risk’ given that there were protocol deviations and that a per‐protocol analysis was used ²⁰ (Figure 2). However, all three RCTs were eligible for inclusion after data checking.

FIGURE 2.

Risk of bias assessment of included randomized controlled trials.

3.3. Synthesis of the results

3.3.1. Primary outcomes

The odds of vaginal birth were significantly higher in the inpatient vaginal dinoprostone group than in the outpatient balloon catheter groups (three RCTs, 1626 women, 67.8% vs 61.7%, aOR 1.30, 95% CI 1.05–1.62, p = 0.02, I ² = 0%) (Table 3; Figure 3A). The risk of composite adverse perinatal outcome was comparable between the inpatient vaginal dinoprostone and outpatient balloon catheter groups (three RCTs, 1636 women, 13.7% vs 13.1%, aOR 1.09, 95% CI 0.75–1.58, p = 0.64, I ² = 28.7%) (Table 3; Figure 3B). The composite adverse maternal outcome rates were comparable between groups (three RCTs, 1626 women, 16.6% vs 19.8%, aOR 0.81, 95% CI 0.61–1.07, p = 0.14, I ² = 11.5%) (Table 3; Figure 3C).

TABLE 3.

Outcomes for inpatient vaginal dinoprostone compared to the outpatient balloon catheter for cervical ripening in the induction of labor.

Outcome	No. of trials	No. of women	Crude incidence (%)—inpatient vaginal dinoprostone vs outpatient balloon catheter	aOR (95% CI)	Heterogeneity, I 2 (%)	p‐value
Primary outcomes
Odds of vaginal birth	3	1636	67.8 vs 61.7	1.30 (1.05–1.62)	0	0.02
Composite adverse perinatal outcome	3	1636	13.7 vs 13.1	1.09 (0.75–1.58)	29	0.64
Composite adverse maternal outcome	3	1636	16.6 vs 19.8	0.81 (0.61–1.07)	11	0.14
Secondary outcomes
Delivery outcomes
Unassisted vaginal birth	3	1636	51.1 vs 41.7	1.53 (1.15–2.04)	25	0.003
Instrumental vaginal birth	3	1636	16.7 vs 20.0	0.82 (0.60–1.11)	18	0.19
Instrumental vaginal birth for fetal compromise	3	1636	10.2 vs 11.9	0.86 (0.63–1.17)	0	0.34
Instrumental vaginal birth for failure to progress	3	1490	5.1 vs 5.7	0.89 (0.50–1.56)	22	0.67
Cesarean delivery	3	1636	32.2 vs 38.3	0.77 (0.62–0.95)	0	0.02
Cesarean delivery for fetal compromise	3	1626	13.5 vs 17.0	0.77 (0.58–1.02)	0	0.07
Cesarean delivery for failure to progress	3	1626	15.5 vs 17.4	0.90 (0.68–1.17)	0	0.42
Time to vaginal birth	3	1563	‐	sHR 1.23 (1.09–1.39)	0	0.95
Labor progression outcomes
Uterine hyperstimulation	3	1534	8.5 vs 8.0	2.11 (0.30–14.98)	72	0.46
Uterine tachysystole	‐	‐	‐	Insufficient data	‐	‐
Uterine hypertonus	‐	‐	‐	Insufficient data	‐	‐
Oxytocin use	3	1634	62.6 vs 77.6	0.38 (0.21–0.68)	73	<0.001
Need for more methods for cervical ripening a	3	1626	7.6 vs 19.2	0.56 (0.10–3.14)	92	0.51
Use of epidural analgesia during labor	3	1636	58.2 vs 68.2	0.65 (0.52–0.80)	0	<0.001
Use of opioid analgesia during labor	‐	‐	‐	Insufficient data	‐	‐
Change in modified Bishop score	‐	‐	‐	Insufficient data	‐	‐
Total hospital stay for mothers	2	548	‐	MD −1.02 (−10.35 to 8.30)	37	0.83
Total hospital stay for neonates	2	546	‐	MD 3.23 (−2.68 to 9.13)	0	0.28
Perinatal safety outcomes
Stillbirth	3	1636	No events	Insufficient data	‐	‐
Neonatal death	3	1636	No events	Insufficient data	‐	‐
Apgar score of <7 at 5 min	3	1532	1.7 vs 3.0	0.58 (0.30–1.14)	0	0.11
Arterial umbilical cord pH of <7.10	3	1087	7.0 vs 3.3	2.28 (1.27–4.06)	0	0.005
NICU admission	3	1636	6.6 vs 6.8	1.13 (0.59–2.19)	41	0.71
Neonatal infection	3	1285	5.8 vs 4.0	1.30 (0.51–3.32)	58	0.58
Maternal safety outcomes
Maternal infection	3	1535	5.5 vs 7.0	0.83 (0.55–1.26)	0	0.38
Severe postpartum hemorrhage	3	1624	10.5 vs 11.0	0.88 (0.53–1.44)	38	0.60
Maternal intensive care unit admission	2	1535	0.3 vs 0.1	Insufficient data	‐	‐
Maternal death	3	1636	No events	Insufficient data	‐	‐
Uterine rupture	1	1085	0 vs 0.2	Insufficient data	‐	‐

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; MD, mean difference; NICU, neonatal intensive care unit; sHR, subdistribution hazard ratio.

^a This applied to women who had a second method/s for cervical ripening when the initially allocated method was not successful.

FIGURE 3.

Forest plots for the primary outcomes and time to vaginal birth. (A) Odds of vaginal birth. (B) Composite adverse perinatal outcome. (C) Composite adverse maternal outcome. (D) Time to vaginal birth. 95% CI, 95% confidence interval; DL, DerSimonian‐Laird method; OR, adjusted odds ratio; sHR, subdistribution hazard ratio.

3.3.2. Secondary outcomes

Table 3 shows the results for secondary outcomes. The odds of unassisted vaginal birth were significantly higher in the inpatient vaginal dinoprostone group (three RCTs, 1636 women, aOR 1.53, 95% CI 1.15–2.04, p = 0.003, I ² = 24.9%). There was no statistically significant difference in the odds of instrumental vaginal birth for fetal compromise (three RCTs, 1636 women, aOR 0.86, 95% CI 0.63–1.17, p = 0.34, I ² = 0%) or failure to progress (three RCTs, 1490 women, aOR 0.89, 95% CI 0.50–1.56, p = 0.67, I ² = 21.9%) between the two methods. A time‐to‐event analysis for the subdistribution hazard ratio of vaginal birth showed a higher cumulative rate for the inpatient vaginal dinoprostone group (three RCTs, 1563 women, subdistribution hazard ratio 1.23, 95% CI 1.09–1.39, p = 0.001, I ² = 0%) (Table 3; Figure 3D).

Women in the inpatient vaginal dinoprostone group were less likely to have epidural analgesia (three RCTs, 1636 women, aOR 0.65, 95% CI 0.52–0.80, p < 0.001, I ² = 0%) and labor augmentation with oxytocin (three RCTs, 1634 women, aOR 0.38, 95% CI 0.21–0.68, p < 0.001, I ² = 73.1%) than their outpatient balloon catheter counterparts. There were no statistically significant differences between the two methods for uterine hyperstimulation (three RCTs, 1534 women, aOR 2.11, 95% CI 0.30–14.98, p = 0.46, I ² = 72.1%), the need for more methods for cervical ripening (three RCTs, 1626 women, aOR 0.56, 95% CI 0.10–3.14, p = 0.51, I ² = 91.7%), and total hospital stay in hours for mothers (two RCTs, 548 women, MD −1.02, 95% CI −10.35 to 8.30, p = 0.83, I ² = 37.2%) or neonates (two RCTs, 546 women, MD 3.23, 95% CI −2.68 to 9.13, p = 0.28, I ² = 0%).

Table 3 also shows the analysis of the components of the composite perinatal and maternal outcomes. The use of inpatient dinoprostone led to a higher risk of neonatal arterial umbilical cord pH of <7.10 than outpatient balloon catheter (three RCTs, 1087 women, aOR 2.28, 95% CI 1.27–4.06, p = 0.005, I ² = 0%). There was no statistically significant difference between the two interventions in the rate of neonatal intensive care unit admissions (three RCTs, 1636 women, aOR 1.13, 95% CI 0.59–2.19, p = 0.71, I ² = 41.2%), neonatal infection (three RCTs, 1285 women, aOR 1.30, 95% CI 0.51–3.32, p = 0.58, I ² = 57.6%), or Apgar score <7 at 5 min (three RCTs, 1532 women, aOR 0.58, 95% CI 0.30–1.14, p = 0.11, I ² = 0%). There was no statistically significant result indicating that either intervention was superior regarding maternal infection (three RCTs, 1535 women, aOR 0.83, 95% CI 0.55–1.26, p = 0.38, I ² = 0%) or severe post‐partum hemorrhage (three RCTs, 1624 women, aOR 0.88, 95% CI 0.53–1.44, p = 0.60, I ² = 37.7%).

3.3.3. Intervention–covariate interactions

Pregnant women with higher BMI were more likely to benefit from inpatient vaginal dinoprostone in terms of a higher vaginal birth rate (three RCTs, 1624 women, interaction p = 0.02). In a stratified analysis according to the BMI categories (Underweight/Normal weight: BMI <25, Overweight: BMI ≥25–29.9 and Obese: ≥30 in kg/m²), those who were overweight (aOR 1.84, 95% CI 1.18–2.88) or obese (aOR 1.59, 95% CI 1.06–2.40) had a higher vaginal birth rate when using inpatient vaginal dinoprostone than outpatient balloon catheter, but not for those who were underweight or had normal weight (aOR 0.93, 95% CI 0.67–1.29) (Figure 4). However, there was no intervention–BMI interaction on the perinatal or maternal composite safety outcomes (Table 4).

FIGURE 4.

Odds ratio of vaginal birth with cervical ripening after inpatient vaginal dinoprostone vs outpatient balloon catheter across body mass index categories.

TABLE 4.

Treatment–covariate interactions for primary outcomes.

Effect modifiers	No. of trials	No. of women	aOR (95% CI) for interaction	Heterogeneity, I 2 (%)	p‐value for interaction
Vaginal birth
Gestational age at induction	3	1626	0.87 (0.72–1.05)	0	0.14
Parity (multiparous vs nulliparous)	3	1535	1.78 (0.59–5.36)	50	0.31
Maternal age	3	1626	1.00 (0.96–1.05)	0	0.91
Initial modified Bishop score	3	1611	1.08 (0.94–1.25)	0	0.28
BMI	3	1624	1.04 (1.01–1.08)	0	0.02
Composite adverse perinatal outcome
Gestational age at induction	3	1626	1.20 (0.93–1.55)	0	0.17
Parity (multiparous vs nulliparous)	3	1535	0.50 (0.22–1.16)	0	0.10
Maternal age	3	1626	0.99 (0.94–1.05)	0	0.74
Initial modified Bishop score	3	1611	0.97 (0.80–1.17)	0	0.73
BMI	3	1624	0.99 (0.95–1.03)	0	0.66
Composite adverse maternal outcome
Gestational age at induction	3	1626	1.20 (0.96–1.51)	0	0.11
Parity (multiparous vs nulliparous)	3	1535	0.87 (0.40–1.86)	2	0.71
Maternal age	3	1626	1.00 (0.95–1.07)	17	0.87
Initial modified Bishop score	3	1611	0.94 (0.79–1.12)	0	0.52
BMI	3	1624	1.00 (0.94–1.06)	31	0.97

Abbreviations: aOR, adjusted odds ratio; BMI, body mass index; CI, confidence interval.

There was no differential effect on vaginal birth rate identified for the other baseline variables, including gestational age (three RCTs, 1626 women, interaction p = 0.15), parity (nulliparous vs multiparous, three RCTs, 1535 women, interaction p = 0.31), maternal age (three RCTs, 1626 women, interaction p = 0.91), and initial modified Bishop score (three RCTs, 1611 women, interaction p = 0.28) (Table 4).

4. DISCUSSION

This IPD meta‐analysis found that outpatient balloon catheter was probably a less effective strategy compared to inpatient vaginal dinoprostone for cervical ripening in IOL. In the pre‐planned subgroup analysis, it appears that this is particularly the case for women with overweight or obese during pregnancy. For women with underweight or normal weight, both inpatient vaginal dinoprostone and outpatient balloon catheter methods are viable options. Adverse perinatal and maternal outcomes were comparable between the two methods.

A key strength of our study is that we obtained data from all eligible RCTs identified in the literature search, ²⁰ , ²¹ , ²² which contained a large sample size of 1636 women evenly distributed across the two groups. Utilizing IPD meta‐analysis for data synthesis enabled us to standardize outcomes, increase the sample size, and ensure the data were thoroughly harmonized. This approach also allowed us to explore potential interactions between treatment and covariates.

Our study also has limitations. First, not all relevant data were collected in every RCT. For example, two eligible RCTs did not report data for meconium aspiration syndrome. ²¹ , ²² Second, there is clinical heterogeneity among included RCTs. While I ² statistics for most investigated outcomes were 0%, reflecting an absence of statistical heterogeneity, clinical heterogeneity may exist. ²⁰ , ²¹ , ²² As such, differences in indications of IOL and slight differences in the balloon catheter volumes and vaginal dinoprostone dosages were noted. Third, not all secondary outcomes had sufficient statistical power, and their results should be interpreted cautiously. Last, the existing data on this topic all come from Australia and New Zealand, with a significant proportion of the participants being Caucasian. Therefore, caution is needed when generalizing the results to other populations.

Aggregate data meta‐analyses on the method of IOL have used vaginal birth within 24 h as a measure of effectiveness. ¹⁸ , ²⁹ , ³⁰ They could not identify a significant difference between vaginal dinoprostone and balloon catheters. However, these analyses did not consider the setting where IOL took place. In our IPD meta‐analysis, we considered both the method and setting of labor induction and found that an outpatient balloon catheter for IOL was probably a less effective strategy compared to inpatient vaginal dinoprostone. These differences might be explained by several factors. In an inpatient setting, patients receiving dinoprostone for IOL are typically monitored more closely by healthcare providers. If the labor induction process is not progressing well, healthcare providers could intervene with additional measures or proceed to alternative methods. In contrast, in an outpatient setting, delays in seeking medical attention or reaching the hospital may happen and impact the rate of vaginal birth. Also, patients in an inpatient setting may have better compliance and adherence to instructions and protocols, as they are under the direct supervision of healthcare professionals. In an outpatient setting, there may be variations in patient compliance, which could affect the effectiveness of the induction method.

In terms of safety, a recent IPD meta‐analysis of RCTs conducted primarily in inpatient settings demonstrated that balloon catheters lead to fewer adverse perinatal outcomes than vaginal prostaglandins, including arterial umbilical cord pH levels of less than 7.1. ¹⁹ Similarly, the present IPD meta‐analysis found consistent evidence that inpatient dinoprostone leads to a higher risk of neonatal acidosis than outpatient balloon catheters. However, the composite perinatal outcome analysis did not reach statistical significance. The observed perinatal safety advantages associated with balloon catheters in inpatient settings might be offset by the lack of close monitoring when using balloon catheters outside of the hospital. This could be a potential area for further research. Our study is consistent with previous evidence for maternal safety. ¹⁸ , ¹⁹ This shows that there is no significant safety difference between dinoprostone and balloon catheters, even when the latter are used outpatient.

A recent trend of outpatient IOL is gaining popularity, as it might minimize hospital crowding and reduce infection risks. ³¹ This approach allows pregnant women to initiate labor at home, aligning with public health measures and patient preference for a less clinical birthing experience. However, our study did not find that outpatient balloon catheter IOL is as effective as inpatient dinoprostone in achieving vaginal birth. We also did not find clear perinatal safety benefits of using balloon catheters in an outpatient setting comparable to those in an inpatient setting. Therefore, outpatient IOL is unlikely to be a universal solution for all patients. The secondary analysis of PROBAAT trials has shown a non‐significant high cesarean section rate among women with obesity who had a Foley balloon for IOL. ³² This information can be useful for updating clinical guidelines and shared decision‐making.

Summarizing all available evidence, while outpatient IOL with balloon catheters might offer convenience, resource management benefits, and a positive birth experience, ¹⁶ , ¹⁷ it may not be optimal for those who are overweight or obese. Clinicians should carefully select patients for outpatient IOL, considering the potential trade‐offs between convenience and clinical outcomes. Informed choices, shared decision‐making, enhanced monitoring protocols, and patient education are essential to ensure the effectiveness and safety of outpatient balloon IOL, particularly for women who are overweight or obese during pregnancy. Meanwhile, the lack of robust data on consumer experience and satisfaction should be considered another limitation in the current evidence that warrants future research.

5. CONCLUSION

Cervical ripening in IOL with outpatient balloon catheters is probably a less effective method than inpatient vaginal dinoprostone; in a pre‐planned subgroup analysis, this appears to relate to women with overweight and obesity during pregnancy. Therefore, we do not recommend home induction using balloon catheters for these patients. For pregnant women with underweight or normal weight, both inpatient vaginal dinoprostone and outpatient balloon catheter methods are viable options.

AUTHOR CONTRIBUTIONS

Fei Chan and Malitha Patabendige (co‐first authors) principally managed the project and the collaborative process and conducted the design, data collection, and IPD checking. Fei Chan, Malitha Patabendige, and Wentao Li performed data synthesis, writing, and editing. Malitha Patabendige performed the final editing and submission of the manuscript and addressed all the revisions. Ben W. Mol and Wentao Li conceived the research idea and were responsible for overseeing all aspects of conduct. Malitha Patabendige and Wentao Li contributed to eligibility screening and risk of bias assessment as the second and third reviewers. BWM provided clinical oversight. Fei Chan and Malitha Patabendige designed and conducted the literature searches. All authors were involved in the decision to submit the manuscript. All contributing trial investigators had opportunities to comment on the initial scope and draft protocol and participated in teleconference meetings as the project progressed. Trial investigators also prepared and supplied data and answered questions about their RCTs.

FUNDING INFORMATION

Two NHMRC Investigator Grants supported this study (GNT1176437 for BWM and GNT2016729 for WL). MP is supported by a Research Training Stipend provided by the Australian Government.

CONFLICT OF INTEREST STATEMENT

Malitha Patabendige is supported by a Research Training Stipend provided by the Australian Government. Ben W. Mol is supported by an NHMRC Investigator grant (GNT1176437), reports consultancy, travel support, and research funding from Merck, and consultancy for Organon and Norgine, and holds stock from ObsEva. Wentao Li declared a grant from NHMRC that supported this work. The remaining authors report no conflict of interest.

Chan F, Patabendige M, Wise MR, et al. Inpatient vaginal dinoprostone vs outpatient balloon catheters for cervical ripening in induction of labor: An individual participant data meta‐analysis of randomized controlled trials. Acta Obstet Gynecol Scand. 2025;104:1041‐1055. doi: 10.1111/aogs.15092

DATA AVAILABILITY STATEMENT

The protocol, statistical analysis plan, and codebook are available upon request from the corresponding author (MP). The trial investigators who shared IPD for the purposes of the meta‐analysis retain ownership of their trial data, and any requests for access to IPD should be made directly to them.