Published health economic evidence

No published health economic evidence was identified, although this question was prioritised for health economic analysis.

Evidence profiles

Evidence profiles for this question are presented in Tables 8.1 to 8.8. One study¹²² reported some results for the predictive value of cervical length measurements in the form of diagnostic test accuracy measures and other results in the form of relative risks (RRs). The evidence from this study is, therefore, presented in a separate evidence profile (Table 8.2). The RRs were considered by the GDG, but they did not influence its final recommendations and so Table 8.2 below does not include RRs (see the full evidence profile in Appendix J for these results).

Table 8.1

GRADE summary of findings for cervical length measurement in twin pregnancies (diagnostic accuracy studies reporting diagnostic accuracy measurements only).

Table 8.2

GRADE summary of findings for cervical length measurement in twin pregnancies (diagnostic accuracy studies reporting relative risks and diagnostic accuracy measurements).

Table 8.3

GRADE summary of findings for cervical length measurement in triplet pregnancies.

Table 8.4

GRADE summary of findings for fetal fibronectin test in twin pregnancies.

Table 8.5

GRADE summary of findings for combined cervical length measurement and fetal fibronectin test in twin pregnancies.

Table 8.6

GRADE summary of findings for home uterine activity monitoring (with or without nursing contact) versus no monitoring in twin pregnancies.

Table 8.7

GRADE summary of findings for home uterine activity monitoring and daily contact with a nurse versus daily contact alone versus weekly contact in twin pregnancies.

Table 8.8

GRADE summary of findings for obstetric history (preterm singleton birth in the previous pregnancy) in twin pregnancies.

Evidence statement

Evidence was identified for cervical length measurement, fibronectin testing, additional antenatal care and previous obstetric history in predicting preterm birth in twin and triplet pregnancies. The quality of the evidence ranged from very low to high.

There was evidence that a short cervical length, especially less than 25 mm, at 18–24 weeks of gestation in twin pregnancies is a good predictor of preterm birth at up to 35 weeks of gestation (high or moderate quality). A short cervix was, however, not predictive of birth before 37 weeks (high quality evidence).

There was evidence that a cervical length less than 25 mm measured at 14–20 weeks in triplet pregnancies was associated with spontaneous preterm birth before 32 weeks (low quality).

There was no association between a positive fetal fibronectin test result and the risk of spontaneous preterm birth in twin pregnancies (moderate to high quality). However, if used in conjunction with cervical length, the ability of the test to identify women who were at a significantly higher risk of preterm birth was improved (very low quality).

There was evidence that home uterine activity monitoring in twin pregnancies was not effective in predicting spontaneous preterm birth (very low to moderate quality).

There was evidence in twin pregnancies that the occurrence of a singleton preterm birth in the previous pregnancy significantly increased the risk of preterm birth in twin pregnancies (very low quality).

Relative value placed on the outcomes considered

Sensitivity is the proportion of pregnancies that resulted in preterm birth that were predicted to be preterm (true positive). One hundred minus sensitivity (100 − sensitivity) shows how many of these pregnancies were predicted to be term at scan (false negative).

Specificity is the proportion of pregnancies that resulted in a term birth that were predicted to be term (true negative). One hundred minus specificity (100 − specificity) shows how many of these pregnancies were predicted to be preterm at scan (false positive).

PPV is the proportion of pregnancies that were predicted to be preterm and that resulted in a preterm birth. One hundred minus PPV (100 − PPV) shows how many of these pregnancies resulted in a term birth.

NPV is the proportion of pregnancies that were predicted to be term that resulted in a term birth. One hundred minus NPV (100 − NPV) shows how many of these pregnancies resulted in a preterm birth.

The positive likelihood ratio (LR⁺) shows how much the odds of a birth being preterm increase when preterm birth is predicted. The negative likelihood ratio (LR⁻) shows how much the odds of a pregnancy being preterm decrease when a scan predicts term birth.

The GDG prioritised likelihood ratios and sensitivity when considering the evidence for different methods of predicting preterm birth.

Trade-off between clinical benefits and harms

Correctly identifying women who are at risk of preterm birth (true positives) potentially allows more careful monitoring. It allows decisions to be made in consultation with women who are fully aware of the risks involved, including planning for an earlier birth date. If women are predicted to be at risk of preterm birth but then deliver at term (false positives), this may result in unnecessary extra monitoring, which could cause maternal anxiety and unnecessary interventions from healthcare professionals. Correctly identifying women who are not at risk of preterm birth (true negatives) saves resources and prevents anxiety by avoiding unnecessary extra monitoring. It allows the woman and her healthcare team to plan for an accurate birth date. Failing to identify women who go on to miscarry or deliver extremely prematurely (false negatives) could result in the delivery of preterm fetuses in a setting where neonatal facilities are suboptimal: these women would also miss the opportunity for antenatal administration of corticosteroids.

Trade-off between net health benefits and resource use

A more accurate test may be worth an extra cost if there is a way to reduce preterm birth in women who are true positives. The treatments that have been investigated with the intention of preventing preterm birth were not clinically effective (see Section 8.2) and so the GDG did not proceed with the planned health economic analysis. Ultrasound scans cost about £71 and, assuming two scans are conducted for each woman, could cost the NHS around £140 per pregnant woman (no additional scanning appointments would be needed but the duration of these appointments would be increased by approximately 15 minutes per scan, compared with routine antenatal care). The GDG concluded that it could not recommend routine ultrasound scanning to predict preterm birth.

Quality of evidence

All the evidence in triplet pregnancies was low quality. For predicting preterm birth in twin pregnancies the quality of evidence varied, as follows:

• cervical length alone: low to high
• fetal fibronectin: moderate to high (mainly moderate)
• combined cervical length and fetal fibronectin tests: very low
• home uterine activity monitoring: moderate to very low
• obstetric history: very low.

Other considerations

Some evidence was identified regarding preterm birth in triplet pregnancies. However, there was not enough evidence to compare the effectiveness of all methods of preventing preterm birth in twin and triplet pregnancies. The ‘Antenatal care’ (NICE clinical guideline 62)¹⁴ recommends against screening for preterm birth. The GDG believes that screening for preterm birth is an important consideration in twin and triplet pregnancies as there is a greater risk of preterm birth in such pregnancies. Cervical length measurement of less than 25 mm at 18–24 weeks of gestation in twin pregnancies and 14–20 weeks of gestation in triplet pregnancies predicts the risk of spontaneous preterm birth and the test may be improved further with the addition of fetal fibronectin. However, the GDG does not recommend that this test is undertaken routinely in twin or triplet pregnancies because no effective interventions to reduce this risk have been identified (see Section 8.2) and testing can create anxiety or offer false reassurance.

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 pregnancies^140;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 studies^145–148 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 USA^145–147 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.

Evidence statement

Evidence was identified for all outcomes prioritised in relation to the interventions used to prevent preterm birth in twin and triplet pregnancies, although the studies varied in the number of outcomes reported. The quality of the evidence was mostly low or very low, with some being of moderate or high quality.

Relative value placed on the outcomes considered

Primary outcomes:

• neonatal:
  • spontaneous preterm birth
  • gestational age at delivery
  • perinatal mortality and morbidity
• maternal:
  • length of stay
  • maternal side effects (infection, haemorrhage, drug effects, tachycardia, caesarean section).

Secondary outcomes:

• neonatal unit admission
• low birthweight and very low birthweight
• respiratory distress syndrome
• intraventricular haemorrhage
• necrotising enterocolitis
• neonatal length of stay
• maternal quality of life
• maternal satisfaction.

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.

Published health economic evidence

No published health economic evidence was identified and this question was not prioritised for health economic analysis.

Evidence statement

Limited evidence was identified for the effectiveness of routine (elective) corticosteroids for reducing perinatal morbidity in twin and triplet pregnancies. The evidence compared different aspects of treatment and was mostly very low in quality.

The evidence that was reported addressed neonatal mortality, neurodevelopmental outcomes at 1 year, respiratory distress syndrome, intraventricular haemorrhage, necrotising enterocolitis, neonatal length of stay, birthweight and composites of these outcomes.

No studies were identified that examined development of gestational diabetes, development of gestational hypertension, maternal satisfaction or neurodevelopmental outcomes after 1 year in twin or triplet pregnancies treated with corticosteroids.

The corticosteroids for which data were reported were betamethasone and dexamethasone. No data were reported that allowed a direct comparison between the two corticosteroids, between different routes of administration for the same corticosteroid or between different doses of the same corticosteroid. There were limited data comparing the number of courses of dexamethasone or betamethasone, with a single course showing a lower mortality rate than multiple courses (very low quality evidence). However, the significance of the difference was not reported.

Relative value placed on the outcomes considered

All outcomes specified in the review protocol, including neonatal mortality and morbidity (respiratory distress syndrome, necrotising enterocolitis, intraventricular haemorrhage) and long-term neurodevelopmental outcomes, were considered by the GDG to be critical to the formulation of recommendations for clinical practice. Birthweight was considered a particularly important outcome given the potential harm of multiple courses of corticosteroids.

Trade-off between clinical benefits and harms

No clear evidence of benefit in giving routine single or multiple courses of antenatal corticosteroids in twin or triplet pregnancies was identified and there is limited evidence of harm from multiple courses compared to no treatment or targeted (rescue) treatment. However, the effect of corticosteroids on long-term neurodevelopmental outcomes is unknown.

Trade-off between net health benefits and resource use

The cost impact and opportunity costs of using corticosteroids prophylactically can be significant. Intramuscular corticosteroids cost up to £4.70 per dose and oral corticosteroids cost up to £1.40 per dose. If multiple doses were to be used (in one of the included studies up to eight doses were used) a course of treatment could cost almost £40 per pregnant woman. A recommendation not to use routine antenatal corticosteroid prophylaxis in twin and triplet pregnancies will save the NHS money because routine antenatal corticosteroid prophylaxis is sometimes used in current practice.

Quality of evidence

Few studies were identified for inclusion, with only one being an RCT. The resulting body of evidence considered by the GDG was generally of very low quality. The evidence for perinatal and neonatal mortality was very low in quality, as was the evidence for long-term neurodevelopmental outcomes, respiratory distress syndrome, intraventricular haemorrhage and necrotising enterocolitis. The evidence for a composite of mortality and morbidity was low in quality. Evidence for birthweight and neonatal length of stay was low quality.

Other considerations

The majority of the studies identified did not report chorionicity and no studies reported ethnicity. Some evidence specific to triplet pregnancies was identified, although most studies focused on twin pregnancies. It is not possible to extrapolate twin data to triplets because triplets have a higher preterm birth rate. There was very little evidence in relation to choice of corticosteroids (for example betamethasone or dexamethasone), dosages or route of administration, or gestational age at administration or delivery. The studies identified were of poor quality and significant differences in birthweight between corticosteroid and no corticosteroid groups may be due to corticosteroid exposure.

It is unclear whether antenatal corticosteroids should be given routinely or targeted. There is no strong evidence of the benefit or harm of a single course of corticosteroids compared to multiple courses in twin and triplet pregnancies. A recommendation not to use antenatal corticosteroids routinely (as prophylaxis) in twin and triplet pregnancies does not preclude targeted (or rescue) administration when indicated (for example when preterm labour or birth is imminent).

The GDG considered reporting of gestational age at birth to be very important as differences in gestational age may account for some observed differences in outcomes: in one study, logistic regression showed that gestational age was the best predictor of survival.¹⁵³ The incidence of respiratory distress syndrome was very high in the control and experimental groups in another study,¹⁵⁴ and no explanation was provided by the study authors.