Introduction

Spontaneous preterm birth (PTB) is due to delivery at less than 37 completed weeks’ gestation, and it remains the leading cause of neonatal mortality and morbidity in western countries [1].

Most scientific Societies recommend tocolysis to treat spontaneous preterm labour [2]. Acute tocolysis over the first 48 hours is an established practice to delay delivery until antenatal corticosteroids are administered and/or pregnant women are transferred to a tertiary care center. The efficacy of tocolysis beyond the 48 hours (also known as maintenance or secondary tocolysis) is still debated [3]. Indeed, many of the agents tested (magnesium, beta-agonists, etc..) provided no efficacy [4].

Progestogens were found to be useful to prevent PTB in 2 at risk categories: a) women with a prior history of PTB and/or late pregnancy loss [5], and b) women with short cervix at mid gestation [6]. However, most of the PTBs occur among women presenting with signs and symptoms of preterm labour. Although 75% of them remain undelivered after hospital admission, their risk of PTB remains high, as up to 30% delivers prior to 37 weeks’ gestation [7]. Several plausible mechanisms of progesterone supplementation to prevent PTB have been proposed, including reduced gap-junction formation, oxytocin antagonism (leading to uterine relaxation), maintenance of cervical integrity and anti-inflammatory effects [8].

Progestogens for maintenance tocolysis were shown to effectively delay delivery according to small low quality randomized controlled trials, but not to larger high-quality RCTs [9]. Furthermore, heterogeneous conclusions were drawn by previous meta-analyses, as different drugs were studied while routes of administration were not adequately accounted for [10, 11]. Thus, the objective of this systematic review and pairwise meta-analysis is to evaluate the efficacy of maintenance tocolysis with 17-alpha-hydroxyprogesterone caproate (17-HP), vaginal progesterone (P) or oral P when compared to placebo or no treatment. We decided to investigate the individual role played by each progestogen as they have different molecular structures and their biological effects are mediated by different receptors [12].

Materials and methods

We performed this systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [13]. The protocol was registered with PROSPERO (registration number: CRD42020219966; http://www.crd.york.ac.uk/PROSPERO/).

Results

After duplicates were removed, a total of 4390 records were found. We acquired 21 articles in full text, as potentially relevant. Four studies were excluded: 2 because they were not RCTs [16, 17], 1 because participants and interventions did not match the inclusion criteria [18], and 1 because we were unable to retrieve the full text [19] (Fig 1 [13]).

Download:

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

Fig 1. Flowchart of included studies.

https://doi.org/10.1371/journal.pone.0277563.g001

We finally included 17 RCTs [9, 20–35] with 2152 participants: 12 studies compared vaginal P to placebo or no treatment [20–24, 26–28, 32–35], 5 tested 17-HP versus placebo or no treatment [25, 29, 31, 32] while 1 compared oral P to placebo30.

The mean participants’ age was 27.3 ± 4.4 years. Mean gestational age at randomization was 30.6 weeks’ for vaginal P studies and 28.9 weeks’ for 17-HP studies. Among studies on vaginal P, 7 employed a dose of 200 mg/day [9, 20, 21, 23, 26, 27, 32], only one used 90 mg/day [35], while 4 used 400 mg/day [22, 28, 33, 34]. Among RCTs on intramuscular 17-HP, 2 used a 250 mg weekly dose [24, 29], 1 employed 500 mg twice weekly [25], while 2 used 341 mg, either every four days [31] or weekly [32]. Finally, the study investigating oral P utilized a daily dose of 200 mg [30].

Four studies were conducted in Iran [26, 28, 33, 34], 2 in Italy [31, 32], 2 in India [22, 30], 1 enrolled women from Switzerland and Argentina [21], while Canada [9], USA [29], Nepal [24], France [25], Turkey [27], Sweden [35], Portugal [20] and Spain [23] contributed with 1 study each. Four studies were multicenter [23, 28, 29, 30].

Six studies were funded by public institutions [26, 28, 29, 31, 33, 34]; the remaining did not report their funding source. Table 1 summarizes the characteristics and results of each trial.

Download:

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

Table 1. Characteristics of included studies.

https://doi.org/10.1371/journal.pone.0277563.t001

Risk of bias of included studies

Seven studies were considered at low risk of selection bias because random sequence generation and allocation concealment were appropriate [9, 24, 26, 28, 29, 31, 33]. Furthermore, random sequence generation was appropriate in 7 [21, 22, 25, 27, 30, 32, 34] studies but information on allocation concealment was not provided. The work of Areia et al was judged at high risk of bias for allocation concealment [20], while Briery et al did not describe how random sequence was generated despite adequate allocation concealment [29]. Eight studies were considered at high risk of performance bias because they were open label [20, 22, 25, 26, 30, 32, 33], while the remaining were judged at low risk as they were placebo controlled, in a double blind design. All the included RCTs were judged at low risk of detection bias because the study outcomes were objective and unlikely to be biased by lack of blinding. Two RCTs were rated high for risk of attrition bias [27, 28]. The study protocol was available only for five studies [9, 26, 29, 31, 33], where the outcomes reported in the final publication coincided with the ones listed in the original protocol; for taball the remaining studies the protocol was not available, therefore they were considered at unclear risk of selective outcome reporting bias (Fig 2).

Download:

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

Fig 2. Risk of bias of included studies.

https://doi.org/10.1371/journal.pone.0277563.g002

Primary outcomes (S2 Table)

PTB < 34 weeks.

Preterm birth <34 weeks’ gestation did not differ among women receiving vaginal P as opposed to placebo/no treatment (RR 1.21, 95%CI 0.91 to 1.61, 7 studies, 1077 participants, moderate certainty of evidence), or receiving oral P as opposed to placebo (RR 0.89, 95%CI 0.38 to 2.10, 1 study, 90 participants, low certainty of evidence). Instead, 17-HP significantly reduced the outcome when compared to placebo/no treatment (RR 0.72, 95%CI 0.54 to 0.95, 4 studies, 450 participants, moderate certainty of evidence) (Fig 3).

Download:

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

Fig 3. Forest Plot PTB < 34 weeks.

A. Forest plot comparison (Vaginal P compared with no treatment or placebo) for outcome: preterm birth < 34 weeks’ gestation. B. Forest plot comparison (17-HP compared with no treatment or placebo) for outcome: preterm birth < 34 weeks’ gestation. C. Forest plot comparison (Oral P compared with no treatment or placebo) for outcome: preterm birth < 34 weeks’ gestation. CI, confidence interval; Vaginal P, vaginal progesterone; Oral P, oral progesterone; 17-HP, 17-alfa hydroxyprogesterone.

https://doi.org/10.1371/journal.pone.0277563.g003

PTB < 37 weeks.

Preterm birth < 37 weeks’ gestation did not differ among women receiving vaginal P as opposed to placebo/no treatment (RR 0.95, 95%CI 0.72 to 1.26, 8 studies, 1231 participants, moderate certainty of evidence), while oral P significantly reduced the outcome when compared to placebo (RR 0.58, 95%CI 0.36 to 0.93, 1 study, 90 participants, low certainty of evidence). Finally, frequency of PTB < 37 weeks did not differ between treatment with 17-HP and placebo/no treatment (RR 0.86, 95%CI 0.60 to 1.21, 4 studies, 450 participants, low certainty of evidence) (Fig 4).

Download:

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

Fig 4. Forest Plot PTB < 37 weeks’.

A. Forest plot comparison (Vaginal P compared with no treatment or placebo) for outcome: preterm birth < 37 weeks’ gestation. B. Forest plot comparison (17-HP compared with no treatment or placebo) for outcome: preterm birth < 37 weeks’ gestation. C. Forest plot comparison (Oral P compared with no treatment or placebo) for outcome: preterm birth < 37 weeks’ gestation. CI, confidence interval; Vaginal P, vaginal progesterone; Oral P, oral progesterone; 17-HP, 17-alfa hydroxyprogesterone.

https://doi.org/10.1371/journal.pone.0277563.g004

Discussion

This systematic review is the largest up-do-date, including 17 RCTs and 2152 singleton gestations, 410 (23%) women more than previous meta-analysis [9].

The quantitative meta-analysis showed for the first time, with moderate certainty of evidence, that 17-HP given to women remaining undelivered after an episode of threatened preterm labour significantly reduced their risk of PTB <34 weeks’ gestation. It should be underlined that the relatively small number of events impacts on the precision of the estimate. We considered PTB <34 weeks’ gestation as the primary outcome because clinical management of threatened preterm labour substantially differs after this time point. Acute tocolysis is not recommended in the late preterm period [36] (i.e. delivery between 34 and 37 weeks’) [37] while antenatal corticosteroids treatment is still debated [38].

Although our primary outcome had been addressed in previous meta-analyses [9, 39], our conclusions differ due to a larger study population, that now includes data recovered from 2RCTs [31, 32] originally reporting PTB <35 weeks’ as primary outcome. This allows almost doubling the number of participants in our actual study.

Despite such positive effect on PTB <34 weeks’ gestation, we confirmed that 17-HP did not affect PTB <37 weeks’. The apparent inconsistency may be due to a limitation of primary studies, which do not systematically differentiate between spontaneous and iatrogenic PTB (defined as planned delivery for maternal and/or adverse fetal conditions), the latter being the most common delivery cause in the late preterm period [37].

Based on a large study population (>2000 women) deriving from different countries, our meta-analysis also corroborated the previous suspect that vaginal P is not helpful for PTB prevention after an episode of preterm labour [9, 39, 40]. Nationality represents a pivotal issue, as a secondary analysis of the “4P trial” demonstrated that vaginal P increased the risk of PTB in women from Switzerland while it was ineffective among Argentinians [41], hinting towards potential harmful effects, as already hypothesized by other Authors [9].

Distinct molecular structures, biological effects and sites of actions (genomics and non-genomics) may account for differences in the effectiveness of P and 17-HP on PTB rates [42, 43]. Accordingly, we decided to analyse RCTs using the 2 progestogens separately, and not to combine data on vaginal and oral P, as the 2 routes of administration lead to different metabolites and clearance [44]. Thus, we cannot drew definitive conclusions on oral P since only one, low-quality small trial was included in our review.

Concerning secondary outcomes we confirmed that both 17-HP [11] and vaginal P [10] are associated with prolongation of pregnancy when compared to placebo/no treatment, leading to higher birth-weights, despite unchanged NICU admission rates. The fact the vaginal P treatment is not associated with reduction of PTB < 34 weeks but with significantly increased latency could be related to the gestational age at randomization. Indeed, vaginal P studies randomized women almost two weeks later compared to 17-HP studies. Latency is therefore an outcome conditioned by gestational age at randomization, so it must be interpreted in relation to it. Moreover, the three trials that reported a significant latency did not investigate PTB <34 weeks as an outcome [22, 33, 34]. Finally, it also seems that perinatal mortality and RDS were reduced among women treated with vaginal P, while data on 17-HP showed no difference [11]. The apparent conflicting results between effects of the 2 progestogens on primary outcomes and lack of related effects on morbidity, could be ascribed to the small number of events included in the MA for low birth weight and RDS. The small number of events, together with the small sample size and the relatively small number of included studies strongly affect the precision of the estimate, returning in wide confidence intervals crossing the line of no effect.

The strengths of this systematic review include its comprehensive search strategy and rigorous statistical analysis. Our findings do not solely rely on the higher number of included studies, but also on the systematic search of unpublished data for the two primary outcomes. We performed a subgroup analysis on the impact of additional risk factors for PTB, such as previous PTB/late pregnancy loss, showing for the first time that vaginal P did not affect PTB rates, independently of patients’ obstetric history. Furthermore, we found a trend towards higher PTB rates < 34 weeks’ gestation when vaginal P was given to populations with previous PTB.

Our work is not without limitations: different potential sources of risk of bias, as well as lack of availability of the majority of study protocols, increased the risk of selective outcomes reporting. This is the reason why we drew definitive conclusions only on the primary outcomes. We could not perform the subgroup analysis on the history of PTB/late pregnancy loss on women treated with 17-HP due to the paucity of studies. Similarly, subgroup analysis for mean cervical length at baseline was not feasible because 5 studies did not report the measure, and all but 2 of the remaining studies reported a mean cervical length < 25 mm. These limitations remind us of the importance of international networks to encourage database exchanges. For instance, the recent Individual Patient Data meta-analysis on the effects of progestogens in women at risk for positive obstetric history or asymptomatic short cervix had such a large population that its conclusions could impact clinical practice [45].

In conclusion, with a moderate certainty of evidence we found that 17-HP prevents PTB < 34 weeks’ gestation among women that remained undelivered after an episode of threatened preterm labour. Vaginal P was ineffective, although it seems to prolong pregnancy. However, there are insufficient data to generate recommendations for the use of progestagens as a maintenance tocolytic in clinical practice. Prior to any change in patients’ management, a large multicentre international RCT is urged to detect possible differences in PTB prevention among women treated with vaginal P, intramuscular 17-HP or placebo.

Supporting information