Overview of the evidence
Eighteen studies were identified for inclusion.132–149 The studies investigated the clinical effectiveness of the following interventions to prevent preterm birth in women with twin and triplet pregnancies:
Where evidence from systematic reviews of RCTs or individual RCTs was identified in relation to a particular intervention and associated outcomes prioritised for consideration by the GDG, evidence from study designs lower in the hierarchy of evidence (such as observational studies, including cohort studies and case–control studies) was excluded.
Women in two studies were advised to abstain from sexual intercourse. In one study that investigated hospital bed rest versus home bed rest, women in the intervention group only were advised to abstain from intercourse.134 In another study that investigated the effectiveness of vaginal progesterone in twin pregnancies, both the intervention and control groups were advised to abstain from intercourse.139
Bed rest
A Cochrane review reported meta-analysis of evidence relating to the effectiveness of routine hospital bed rest compared to no bed rest for preventing preterm birth in women with twin and triplet pregnancies.132 The review included six RCTs and one quasi-randomised controlled trial; five of the studies involved women with twin pregnancies and two involved women with triplet pregnancies. One of the RCTs was conducted in Finland, two in Australia and four in Zimbabwe.
One retrospective observational study, conducted in Denmark and involving women with twin pregnancies, compared hospital bed rest with bed rest at home or with no bed rest at all.133 Another retrospective study, conducted in the USA, involved women with triplet pregnancies and also compared hospital bed rest with home bed rest but, in addition, all women in the study were advised to discontinue vaginal intercourse at 20 weeks of gestation.134
One RCT examined the effectiveness of hospital bed rest and prophylactic oral salbutamol (as a combined intervention) compared to hospitalisation for bed rest alone.135 The study involved women with twin and triplet pregnancies and was conducted in Finland.
Progesterone
Seven RCTs evaluated the clinical effectiveness of progesterone compared to placebo in the prevention of preterm birth in women with twin or triplet pregnancies.136–142
Five of the studies reported data separately for spontaneous preterm birth.136;137;139;140;142 One study did not specify whether or not the reported preterm birth data included iatrogenic preterm births,138 and in another study, data for preterm birth and intrauterine death were reported together.141 This last study reported clinical effectiveness data from the Study of Progesterone for the Prevention of Preterm Birth in Twins (STOPPIT; a double blind, randomised, placebo-controlled study; see https://www.charttrials.abdn.ac.uk/stoppit/). A published economic evaluation relating to STOPPIT was identified separately (see later).150
Two of the six studies reported on daily vaginal progesterone gel and they were conducted in the UK.139;141 The other studies reported on weekly intramuscular progesterone:136–138;140;142 one of these studies was conducted in Finland136 and the other four in the USA.137;138;140;142
Two of the RCTs involved women with triplet pregnancies140;142 and the other five involved women with twin pregnancies.136–139;141
Cervical cerclage
One RCT,143 one prospective observational study144 and four retrospective observational studies145–148 evaluated the effectiveness of cervical cerclage in the prevention of preterm birth in women with twin or triplet pregnancies.
The RCT was conducted in Israel and involved women with twin pregnancies conceived after ovulation induction.143 Women in the intervention group underwent elective cervical suture (McDonald) at 13 weeks of gestation and they were compared with a control group who received no cervical suture.
The prospective observational study involved women with a short cervix, all of whom rested at home or in the hospital.144 The study was conducted in the USA. The other four studies used a retrospective cross-sectional review of medical records of women with triplet pregnancies who had undergone cervical cerclage compared to women who had not.145–148 Three of the studies were conducted in the USA145–147 and one in Israel.148
Tocolytic therapy
A Cochrane review involving women with twin pregnancies assessed the clinical effectiveness of prophylactic tocolytic therapy.149 The review included five RCTs, each examining a different betamimetic agent (salbutamol, feneterol, isoxurpine, ritodrine or terbutaline). The trials were conducted in the UK, Ireland, Sweden, South Africa and Zimbabwe.
No studies examining the role of other tocolytic agents were identified for inclusion.
Sexual abstinence
No studies examining the effectiveness of sexual abstinence alone were identified for inclusion.
Published health economic evidence
One published health economic evaluation was identified in relation to this question,150 which was prioritised for health economic analysis. The published economic evaluation related to STOPPIT, for which a separate publication reporting clinical effectiveness data only141 was included in the review of clinical evidence (see above). The main outcomes of the STOPPIT trial were birth or fetal death before 34 weeks of gestation. The use of progesterone in this population of women did not reduce the incidence of preterm birth. There was a tendency towards increased neonatal stay in special care units in the progesterone group. The mean hospital costs for the progesterone group were about £28,000 compared to £25,000 in the placebo group.
Cost effectiveness acceptability curves reported in the economic evaluation showed the probability of prophylactic vaginal progesterone being cost effective as a function of the decision makers’ willingness to pay to prevent a case of spontaneous preterm birth against the alternative of not providing prophylactic progesterone. In the health economic analysis, progesterone was 20% cost effective at a willingness to pay value of £30,000 per preterm birth prevented. The net benefit statistic confirmed the finding that progesterone was unlikely to be cost effective at £30,000 per preterm birth prevented, as the net benefit was negative (−£3,637, 95% CI−£3,853 to −£3,420), meaning that there would be a financial loss to the health service. The authors calculated the expected value of perfect information, which showed that using placebo consistently produced higher net health benefits.
The authors of the economic evaluation concluded that the probability of prophylactic vaginal progesterone being cost effective was low in women with twin pregnancies, and sensitivity analysis showed the findings to be robust. This was a well conducted and presented health economic analysis. The study did not consider quality adjusted life years (QALYs) as an outcome, although the QALY is NICE’s preferred measure of outcome. However, the GDG believes the QALY approach would be unlikely to change the conclusions of the analysis, since preterm birth and fetal death are good proxies for the quality and quantity of life that would be needed to calculate QALYs.
Health economics profile
One well-conducted health economic analysis was identified for inclusion and this concluded that the probability of prophylactic vaginal progesterone being cost effective in women with twin pregnancies is low. The findings were shown to be robust in sensitivity analysis.
This question was prioritised for further health economic evaluation. The question is linked to the question considering tests to predict preterm birth (see Section 8.1), in that this question addresses the cost effectiveness of interventions to prevent preterm birth once it has been predicted. None of the interventions considered for preventing preterm birth, including bed rest, cervical cerclage, progesterone and tocolytic drugs, was found to be clinically effective and so a formal health economic analysis was not required. Data from NHS reference costs show that a cervical cerclage procedure will cost the NHS about £320, while the British National Formulary (BNF) 59 shows that tocolytic drugs will cost £56 on average and progesterone £90 per pregnant woman. Given that the interventions are not clinically effective, these resources could be used for other more clinically effective and, therefore, cost-effective interventions.
Evidence to recommendations
Relative value placed on the outcomes considered
Primary outcomes:
neonatal:
maternal:
length of stay
maternal side effects (infection, haemorrhage, drug effects, tachycardia, caesarean section).
Secondary outcomes:
The GDG considered all outcomes to be important but believed perinatal mortality to be the most critical.
Trade-off between clinical benefits and harms
Preventing preterm birth can lead to better short- and long-term outcomes for the baby. This will also result in less use of healthcare resources. The clinical harms associated with preventing preterm birth include keeping a woman in an environment she does not wish to be in (for example, hospitalisation for bed rest), which may not be beneficial in the long term. Interventions may have unexpected adverse side effects for women and babies, and may result in higher preterm labour rates.
Trade-off between net health benefits and resource use
This review question was prioritised for health economic analysis but there was no evidence of clinical effectiveness for any of the interventions considered by the GDG (bed rest at home or in hospital, intramuscular or vaginal progesterone, cervical cerclage, oral tocolytics or sexual abstinence) and so no formal health economic analysis was conducted because the GDG was not going to recommend use of any of the interventions. NHS reference costs show that cervical cerclage will cost the NHS about £320 while BNF 59 shows that tocolytic drugs will cost on average £56 and progesterone £90 per pregnant woman. These cost data illustrate that these are expensive interventions that should not be used. Given that the interventions are not clinically effective, these resources could be freed for more clinically effective, and hence cost-effective, interventions.
Quality of evidence
Evidence for bed rest ranged from very low to moderate quality (mainly low); for progesterone from very low to high quality (mainly low); for cervical cerclage it was very low or low quality; and for tocolytics (oral betamimetics) it was low or very low quality.
Other considerations
Although the GDG recommended that bed rest (at home or in hospital), intramuscular or vaginal progesterone, cervical cerclage and oral tocolytics should not be used routinely to prevent spontaneous preterm birth in twin or triplet pregnancies, this does not preclude their use when clinically indicated (that is, targeted use appropriate to individual circumstances).
It was not possible to determine whether chorionicity affected the effectiveness of the methods used to prevent preterm birth.
Since the GDG had identified some evidence to suggest that the risk of spontaneous preterm birth could be predicted accurately (see Section 8.1), the group included a recommendation for further research to evaluate interventions for preventing spontaneous preterm birth in women with twin and triplet pregnancies, including those at high risk.