Introduction

The primary challenge in multiple pregnancies is the prevention of extremely or very preterm birth (PTB), defined as delivery before 28 and 32 weeks’ gestation, respectively, as approximately 10% results in delivery before 32 weeks’ gestation. Up to 40% will deliver spontaneously before 37 weeks’ gestation [1]. Early PTB in multiple pregnancies considerably contributes to increased perinatal mortality, neonatal morbidity, and long-term health issues in the offspring born from these pregnancies [2,3]. PTB associated adverse outcomes are more common in women with multiple pregnancies compared to women with singleton pregnancies [4,5]. For example, infant mortality rates are substantially higher, with 21 deaths per 1,000 live births in twins versus 5 per 1,000 in singletons [6,7]. The higher rate of spontaneous preterm birth (sPTB) in twin pregnancies is partly attributed to greater myometrial stretching, contributing to more frequent and intense myometrial contractility compared to singleton pregnancies [8,9]. Another important risk indicator for sPTB is a short cervical length (CL). In women with a singleton or a multiple pregnancy, progesterone has been demonstrated to effectively reduce PTB in cases of short cervix, with cut-offs suggested at 25 or 30 mm for singletons and multiples [10,11].

Three randomised clinical trials (RCTs) in women with multiple pregnancy with a short cervix (defined between <25 and <38mm) found pessary treatment to reduce sPTB rates of before 34 weeks’ gestation, therefore, a cervical pessary has been suggested as a low-cost, effective intervention for the prevention of PTB [12–14]. However, other RCTs could not confirm these outcomes [15–18]. In many RCTs, participants in both the pessary and non-pessary groups also received progesterone, thus hampering a direct comparison. In absence of direct comparisons, consensus on the preferred preventive treatment for multiples with a short cervix is lacking. Therefore, we compared the effectiveness of a cervical pessary and vaginal progesterone in the prevention of sPTB and adverse perinatal outcomes in women with a multiple pregnancy, no prior sPTB before 34 weeks’ gestation, and an asymptomatic mid-trimester shortened cervix.

Methods

Results

From 29th July 2014 to 8th September 2023, 276 participants provided written informed consent and were randomly assigned to receive either cervical pessary (n = 138) or vaginal progesterone (n = 138) (Fig 1). Based on complete data of 237 participants, the DSMC advised on 27th March 2023 to halt recruitment due to futility reasons, namely an expected apparent challenge in establishing superiority of pessary treatment. Although no formal futility stopping rules based on the statistical significance of the effect of the treatment were prespecified in the statistical analysis plan, the DSMC considered continued recruitment scientifically unjustified based on the observed data. Following an independent review by the METC of the Amsterdam UMC, it was concluded that the DSMC’s advice should be followed, leading to early termination of the trial. Follow-up ended in March 2024.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart inclusions Quadruple P Multiples study.

https://doi.org/10.1371/journal.pmed.1004586.g001

Five participants were inappropriately randomised for various reasons, including having a CL measured after 22 weeks GA (n = 2) or in retrospect not meeting the inclusion criteria at the time of randomisation (n = 3). In addition, of the seven participants a written informed consent form was not properly stored after randomisation and insufficient information was provided in their electronic patient files to confirm whether they had followed the correct informed consent procedure. Two participants withdrew consent for data collection. Ultimately, 262 participants were included in the intention-to-treat analysis (Fig 1).

Baseline characteristics were comparable (Table 1). Participants were randomised at a median GA of 19 weeks and 5 days in the pessary group and 19 weeks and 6 days in the progesterone group, with a mean CL of 29.0 mm (SD 7.3.mm) versus 30.2 mm (SD 6.2 mm), respectively. Pregnancies were monochorionic in 26% of the participants in the pessary group and in 31% of the participants in the progesterone group. A total of seven triplets were included (pessary n = 3, progesterone n = 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics.

https://doi.org/10.1371/journal.pmed.1004586.t001

The primary outcome occurred in 53 of 269 (19.7%) children in the pessary group compared to 36 of 262 (13.7%) children in the progesterone group (adjusted RR 1.42 95% CI [0.84, 2.41]; p = 0.19) (Table 2). The rates of (s)PTB < 28, 32, 34 and 37 weeks did not differ significantly between both groups (Table 3). The per-protocol analysis did not reveal any different insights, even in the sensitivity analyses on different cut-off values for compliance (Table 2, Tables A and B in S1 File). A sensitivity analysis on the variability across centres was comparable not statistically significant (p = 0.48), supporting the robustness and generalizability of the primary outcome.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Primary outcome: composite adverse neonatal outcome on child level.

https://doi.org/10.1371/journal.pmed.1004586.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Secondary outcomes on maternal level.

https://doi.org/10.1371/journal.pmed.1004586.t003

Mean time to delivery was 96 days (SD 35) after pessary treatment and 99 days (SD 33) after progesterone (Table 3), and did not significantly differ as is shown by the Kaplan–Meier curve (Fig 2). Other neonatal outcomes were comparable between groups, including perinatal death (26/269 (9.7%) versus 21/262 (8.0%), RR 1.21, 95% CI [0.58, 2.51]; p = 0.62), duration of NICU admittance (18 days, IQR 6–38 versus 12 days, IQR 3–28, p = 0.25) and mean birth weight (2,054 gram versus 2,107 gram, p = 0.35) (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Secondary outcomes on neonatal level.

https://doi.org/10.1371/journal.pmed.1004586.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Kaplan–Meier curve for time to delivery.

https://doi.org/10.1371/journal.pmed.1004586.g002

At maternal level, significantly more often ‘excessive discharge’ was reported by participants in the pessary group compared to the progesterone group (31/133 (24.2%) versus 14/129 (11.2%), RR 2.16, 95% CI [1.21, 3.87]; p = 0.009) (Table 3). Vaginal bleeding was reported equally frequent in both groups (13/133 (9.8%) versus 13/129 (10.2%), RR 0.97, 95% CI [0.47, 2.01]; p = 0.93). Other maternal outcomes were also comparable between groups, including the incidence of genital tract infections (6/133 (4.6%) versus 6/129 (4.8%), RR 1.20, 95% CI [0.59,2.46]; p = 0.95) and chorioamnionitis (4/133 (3.1%) versus 4/129 (3.1%), RR 0.96, 95% CI [0.32,2.90]; p = 0.98). Emergency cervical cerclages were placed in two participants (1.5%) in the pessary group compared to 3 (2.3%) in the progesterone group (RR 0.71, 95% CI [0.14,3.75]; p = 0.69).

Among all randomised participants, serious adverse events occurred in 5 (3.8%) allocated to pessary and 1 (0.8%) in the progesterone group (RR 4.85, 95% CI [0.57,40.9]; p = 0.15) (Table D in S1 File). The most serious adverse events involved two maternal ICU admissions in the pessary group, one due to renal failure from pre-eclampsia/HELLP and one admission due to sepsis from severe chorioamnionitis. The other four hospital admissions were for non-obstetric indications (tachycardia, headache, severe bladder lesion after caesarean and urinary incontinence in known cauda equina syndrome). None of these serious adverse events was considered to be associated with the allocated treatment.

In the predefined subgroup analyses, effect modification was seen for nulliparous participants compared to multiparous participants (with a history of a term birth or prior sPTB between 34 and 37 weeks’ gestation). In nulliparous participants, the composite outcome occurred more often in the children in the pessary group (45/150) compared to the progesterone group (25/157) (30.0% versus 15.9%, RR 1.88 95% CI [1.03,3.43], interaction p = 0.93) (Table 5). In the other subgroup analyses (monochorionic versus dichorionic, CL below 25 mm versus above 25 mm, twins versus triplets and born before 36 weeks’ gestation versus after 36 weeks), no effect modification was seen and thus no significant group differences could be identified (Tables 5 and 6). Table E in S1 File shows the exploratory analysis on PTB rates for CL 0–25 mm and above 25 mm, in which no notable differences were seen.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Subgroup analyses of parity, chorionicity, number of foetuses and gestational age at birth on the primary outcome composite adverse perinatal outcome.

https://doi.org/10.1371/journal.pmed.1004586.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Subgroup analyses on the primary outcome composite adverse perinatal outcome, sPTB < 34 and sPTB < 28 weeks.

https://doi.org/10.1371/journal.pmed.1004586.t006

Discussion

This RCT showed that in asymptomatic women with multiple pregnancies without a prior sPTB before 34 weeks’ gestation and a shortened CL, the use of a cervical pessary did not result in improved perinatal outcomes compared to vaginal progesterone (19.7% versus 13.7%, RR 1.43; 95% CI [0.85,2.4]; p = 0.18). Among nulliparous participants, the observed differences showed a trend in favour of progesterone (30.0% versus 15.9%, RR 1.88; 95% CI [1.03,3.43], interaction p = 0.93). Excessive discharge was more often reported in the cervical pessary group compared to the vaginal progesterone group (24.2% versus 11.2%, RR 2.16, 95% CI [1.21, 3.87], p = 0.009).

When we place the results in context, it is noticeable that the result of our trial showed the opposite effect of what we expected. We anticipated to find a reduction of the composite outcome to 12%, based on previous findings in the ProTwin trial [12]. In this study, in the subgroup with a CL below 38 mm, pessary treatment significantly reduced frequency of poor perinatal outcome and very preterm delivery compared to a control group. Considering our populations were assumed to be comparable and similar cut-off value for short CL were applied, we expected comparable outcomes. Instead, we found a higher incidence of composite adverse perinatal outcomes in the pessary group compared to the progesterone group (19.7% versus 13.7%). In addition, the composite adverse outcome in the pessary group in this trial was twice as high compared to the ProTwin pessary group (19.7% compared to 10.0%). Comparing study populations at baseline, 91% of the ProTwin population was white and 37% became pregnant after fertility treatment compared to 64% and 9.8%, respectively, in the QP trial. However, it is challenging to quantify the extent to which these baseline differences contributed to the differences in outcomes between the trials. This trial was subsequently halted due to concerns about futility, as it appeared highly unlikely that the superiority of the pessary could be demonstrated with the remaining number of participants. However, this decision leaves us uncertain whether the observed effect, specifically, better outcomes with progesterone treatment, might have reached statistical significance. If so, the findings could have been even more compelling.

This study is one of the few multicentre RCTs that directly compares the effectiveness of cervical pessary to vaginal progesterone in preventing PTB in women with a multiple pregnancy and an asymptomatic cervix of below 38 mm. Since these women had no history sPTB before 34 weeks’ gestation, they had not been offered preventive progesterone treatment during pregnancy. A cut-off of CL at 38 mm allowed for assessment of effectiveness of these treatments across a broader range of CLs, thereby assessing a larger at-risk population than would have been possible with a 25 mm cut-off. This approach makes the findings more applicable to a general pregnant population. Another important finding is that the percentages of sPTB was very high, particularly among participants with a CL of 25 mm or less. Independent of the intervention, 30% of this group delivered spontaneously before 28 weeks’ gestation, which has major implications for neonatal outcomes. This underscores the importance of measuring CL in the second trimester, also for this group of pregnant women.

The study also has limitations. First of all, the nature of the interventions made blinding impossible, potentially introducing bias. Additionally, self-reported medication compliance in the progesterone group was low, with fewer than 30% of participants returning their medication diaries. As a result, obstetric caregivers’ notes and verbal reports to research nurses were used to assess adherence, which may have led to an overestimation of actual compliance and an underestimation of progesterone’s preventive potential in the per-protocol analysis. Given that our findings already indicate that pessary treatment is not superior to progesterone, we can therefore assure the robustness of this trial’s conclusion. Furthermore, per-protocol analyses are inherently prone to confounding, as women who adhered may differ systematically from those who did not, and the findings should therefore be interpreted with caution. Another limitation concerns its design. While at the time of protocol development and in previous similar studies the trial was not considered as a cluster randomised design, we have corrected for clustering in the current analyses, which we consider sufficient; nevertheless, we acknowledge that in future studies such clustering should be addressed from the design phase, and that the lack of initial adjustment may have led to some degree of underpowering. Finally, the accuracy of the cut-off values for determining good or poor adherence remains challenging, given the uncertainty regarding the minimum usage requirement necessary to achieve a positive outcome with the interventions used [29–31]. We observed that the incidence of composite adverse perinatal outcomes did not differ between those with 100% adherence to those with 60% adherence in both intervention groups. This raises the question of whether 60% adherence is already sufficient for protection, or if the intervention itself may have limited effectiveness.

One other RCT directly compared a cervical pessary to vaginal progesterone in multiple pregnancies, with methods comparable to ours [14]. They found a significant reduction in composite adverse perinatal outcomes in the pessary group compared to the progesterone group (19% versus 27%; RR 0.70, 95% CI [0.43, 0.93]). However, this was a secondary outcome and did not include PVL. Our results show a similar outcome rate in the pessary group (19.7%), but a lower rate in the progesterone group (13.7%). Despite the benefit of the pessary on the composite outcome, Dang and colleagues found no significant benefit on PTB before 34 weeks (pessary 16% versus progesterone 22%; RR 0.73, 95% CI [0.46, 1.18]), similar to our findings. The main difference between these two trials is the progesterone dosage, 200 mg in our trial versus 400 mg in the trial of Dang and colleagues. The optimal progesterone dosage for twin pregnancies remains uncertain. Additionally, the lowest measured CL in Dang’s study was 18 mm, with funnelling occurring only in 5% of the pessary group. In our trial, 7.5% had a CL of 15 mm or below, with 18% funnelling in the pessary group, suggesting a higher baseline risk for sPTB in addition to a reduced pessary effectiveness in shorter cervixes. However, higher rates of adverse events were also seen in the progesterone group with a shorter cervix. Lastly, population differences, such as higher percentage of nulliparous participants (86.5%), higher education levels and conception via IVF/ICSI (94.0%) in the study from Dang and colleagues, may contribute to the variation in outcomes, though the extent remains unclear.

Since no other trials directly compare cervical pessary and vaginal progesterone in multiple pregnancies, we can only draw comparisons with singleton populations or meta-analyses comparing these interventions against expectant management in multiple pregnancies. In the Quadruple P singletons trial, similar findings were reported, with no superiority of a cervical pessary over vaginal progesterone in low-risk participants without a history of sPTB before 34 weeks and an asymptomatic short cervix [20]. However, progesterone showed an advantage over pessary in the subgroup of a short cervix of 25 mm or less. This effect was not seen in the multiples.

The most recent meta-analysis comparing a pessary to expectant management of Norman and colleagues including five randomised trials concluded that a cervical pessary did not significantly reduce PTB before 34 weeks’ gestation in patients with twin pregnancies and a short cervix regardless of their obstetric history and no differences were found in neonatal morbidity and mortality [17]. On the other hand, a recent meta-analysis by Conde-Agudelo and colleagues of six randomised trials comparing vaginal progesterone with expected management found that vaginal progesterone significantly reduced the risk of PTB before 28–34 weeks and neonatal morbidity and mortality in twin pregnancies with a CL of 30 mm or less and 25 mm or less compared to standard care [10]. Additionally, two large trials were recently halted due to significantly higher rates of perinatal mortality in the pessary groups compared to cervical cerclage, although these were in high-risk singleton populations [34,35]. Given these negative perinatal outcomes, the lack of evidence of superiority of the pessary to progesterone in our trial, along with the absence of a significant effect of the pessary in the meta-analyses, there is currently no evidence to support the use of a cervical pessary for preventing PTB and adverse perinatal outcomes in multiple pregnancies.

Our research findings provide an opportunity to update an individual patient data meta-analysis on the efficacy of both pessary and progesterone use in preventing PTB. Since we cannot demonstrate the superiority of the pessary, it is essential to explore which methods are more effective in managing these high-risk pregnancies. While progesterone may have a positive effect, although it may not be large. Currently, the attention in prevention of PTB in multiple pregnancies seems to be shifting to a third intervention: cerclage, both ultrasound-indicated and physical examination-indicated. Despite initial concerns about its safety in multiple pregnancies, recent studies have shown positive outcomes [36,37]. Consequently, the TWIN-Cerclage study [38] has been initiated to investigate the effect of ultrasound-guided and physical examination-indicated cerclage in twins with a short cervix of 25 mm or less on PTB and adverse perinatal outcomes.

To summarise, our study did not find superiority of pessary treatment over progesterone in the prevention of a composite adverse perinatal outcome in multiple pregnancies without a prior sPTB at less than 34 weeks’ gestation and with a midtrimester short cervix below 38 mm. In the subgroup analysis of nulliparous women, a pessary seemed less effective than progesterone in preventing a composite neonatal outcome.

Supporting information

Acknowledgments

We would like to acknowledge and thank all women who participated in this study, as well as the staff of the Dutch trial bureau of the Dutch Consortium, all participating hospitals and especially all the research nurses involved in this trial, especially J.J.H. Bakker and L. Videler-Sinke. We would also like to acknowledge and thank the other members of the Quadruple P Research Group, namely the local principal investigators from the hospitals that were initially involved, but withdrew their participation in the trial and stopped recruiting new patients: M.N. Bekker, University of Utrecht, Wilhelmina’s Children Hospital, UMC Utrecht, the Netherlands; J. Langeveld, Atrium Medisch Centrum Zuyderland, the Netherlands; D.N.N. Papatsonis, Amphia Ziekenhuis, the Netherlands; T.E.M. Verhagen, Slingeland Ziekenhuis, the Netherlands; E.C.J. Verheijen, Maasziekenhuis Pantein, the Netherlands; H. Visser, TerGooi Medisch Centrum, the Netherlands; Y.M. de Mooij, Zaans Medisch Centrum, the Netherlands en I.M. Evers, Meander Medisch Centrum, the Netherlands.