Progesterone to prevent miscarriage in women with early pregnancy bleeding: the PRISM RCT

Progesterone in pregnancy

Progesterone is an endogenous hormone that is essential to achieve and maintain a healthy pregnancy. Progesterone prepares the lining of the uterus (endometrium) to allow the implantation of the early embryo and stimulates glands in the endometrium to secrete nutrients for the embryo. During the first 8 weeks of pregnancy, progesterone is produced by the corpus luteum; however, between 8 and 12 weeks, the placenta takes over the progesterone-producing role and maintains the pregnancy thereafter.

The physiological importance of progesterone has prompted researchers, physicians and patients to consider progesterone supplementation during early pregnancy to prevent miscarriages. Progesterone supplementation in early pregnancy has been attempted in two contexts: the first is to prevent miscarriages in asymptomatic women with a past history of recurrent miscarriages and the second is to rescue a pregnancy in women who have started to bleed in early pregnancy. 1 Our Progesterone in Recurrent Miscarriage (PROMISE) study, published in the New England Journal of Medicine, addressed the first context. 2 In 2012, the National Institute for Health and Care Excellence (NICE) Clinical Guideline 1543 called for a large randomised placebo-controlled clinical trial to test whether or not progesterone therapy in the first trimester could reduce the risk of miscarriage in women with a history of threatened miscarriage. In response, the current study was designed to address this question and focuses on the rescue context in women with vaginal bleeding in early pregnancy.

Burden of disease

Miscarriage is the most common complication of early pregnancy; one in five clinically recognised pregnancies end in a miscarriage. 4 This has a substantial impact on physical and psychological well-being: research shows that the level of distress associated with miscarriage can be equivalent to that of a stillbirth of a term baby and can induce post-traumatic stress disorder. 5 An estimated 140,000 women per year miscarry in the UK. 3

Costs to the NHS

It is estimated that miscarriage costs the NHS > £350M each year. 3 This value includes the costs of diagnosis (blood tests and ultrasonography), management of miscarriages (expectant, medical or surgical), investigations of causes of miscarriages (e.g. antiphospholipid syndrome, parental karyotype and uterine cavity tests) and hospital inpatient costs. There are also the associated costs of complications following treatment of miscarriages (e.g. uterine perforation, infection, bleeding or visceral damage) and any long-term health consequences of miscarriages or miscarriage management (including complications of intrauterine infections and adhesions). Furthermore, the societal costs (including days lost from work and out-of-pocket expenses for patients and partners) can be expected to be far greater.

Progesterone in clinical use for threatened miscarriage

The Progesterone in Spontaneous Miscarriage (PRISM) study was conceived to address the possibility that progesterone therapy in the first trimester of pregnancy may reduce the risk of miscarriage in women presenting with early pregnancy bleeding. We conducted a UK clinician survey (n = 222) in October 2012. In the UK, the majority of clinicians (212 out of 222; 95.5%) do not use progesterone to prevent miscarriage in women with early pregnancy bleeding. The key reason for non-use is the lack of robust evidence. Therefore, it is not surprising that the majority of clinicians (201 out of 222; 91%) called for a definitive trial. We also conducted a survey of international practitioners at the International Federation of Gynaecology and Obstetrics (FIGO) 2012 conference in Rome. Surprisingly, this survey found that the majority of clinicians (61 out of 68; 90%) already use progesterone in women with early pregnancy bleeding, although the vast majority (56 out of 66; 85%) were willing to recruit into a randomised trial, presumably indicating a lack of confidence in the available evidence.

Effectiveness of progesterone in threatened miscarriage

The first trial of progesterone therapy in women with early pregnancy bleeding was published in 1967, and since then six trials have studied this question, which have previously been summarised in a Cochrane systematic review. 1 In 2014, prior to conducting the PRISM trial, we performed a systematic review of trials on the use of progestogens in women with early pregnancy bleeding, and identified seven studies. 6–12 These studies are listed in Table 1. The seven studies included a total of 744 women. These studies were small and of poor quality, with none reporting the method of allocation concealment. Only three out of seven studies were placebo controlled and five out of seven studies were not blinded. The modified Jadad quality score varied from 1 out of 6 to 3 out of 6. Outcome data were available for miscarriage rates. Individual studies were too small to show an effect, but a meta-analysis of these seven studies (Figure 1) showed a statistically significant reduction in miscarriage rate with progestogen use [relative rate (RR) 0.53, 95% confidence interval (CI) 0.39 to 0.73]. There was no heterogeneity across the studies (I² = 0%), suggesting that there was consistency across the studies.

TABLE 1 - Randomised trials of progestogens vs. placebo or no treatment

Study	Intervention	Duration of treatment	Comparison	Risk of bias
Ehrenskjöld et al., 19676 (n = 153)	20 mg of oral dydrogesterone	20 mg then tapering (20 mg after 12 hours/20 mg every 8 hours until symptoms ceased/10 mg twice daily for 5 days/5 mg twice daily for at least 7 days)	No treatment	Method of randomisation unclear, allocation concealment adequate, blinding of patients and study personnel adequate
El-Zibdeh and Yousef 20097 (n = 146)	10 mg of oral dydrogesterone twice daily	From enrolment until 1 week after bleeding stopped	No treatment	Quasi-randomised (allocated according to the day of the week), no allocation concealment, no blinding for participants or study personnel
Gerhard et al., 19878 (n = 34)	25-mg progesterone vaginal suppositories twice daily	Until miscarriage or for 14 days after bleeding stopped	Placebo	Method of randomisation unclear, allocation concealment unclear, no blinding for participants or study personnel
Mistò,19679 (n = 16)	20 to 40 mg of oral dydrogesterone	Once daily for 6–15 days, sometimes for longer periods and for several cycles	Placebo	Method of randomisation unclear, allocation concealment adequate, blinding of patients and study personnel adequate
Omar et al., 200510 (n = 154)	Dydrogesterone	40 mg of dydrogesterone followed by 10 mg twice daily until bleeding stopped	No treatment	Method of randomisation unclear, no allocation concealment, no blinding of patients and study personnel
Palagiano et al., 200411 (n = 50)	90 mg of progesterone (Crinone® 8% Central Pharma Ltd, Bedford, UK) vaginal suppositories	Once daily for 5 days	Placebo	Method of randomisation unclear, allocation concealment adequate, no blinding for participants or study personnel
Pandian, 200912 (n = 191)	Oral dydrogesterone	40 mg of oral dydrogesterone followed by 10 mg of dydrogesterone twice daily, until 16 weeks of gestation	No treatment	Method of randomisation and allocation concealment adequate, no blinding of participants or study personnel

FIGURE 1.

Meta-analysis of studies of progesterone in women with early pregnancy bleeding (literature review conducted in 2014). M–H, Mantel–Haenszel.

More recently, a Cochrane review on this question summarised evidence from seven studies (see Table 1). The review found that the studies were small with methodological weaknesses (the largest study had a sample size of 191) but the pooled analysis found a significantly lower risk of miscarriages among women who received progesterone than among those who received placebo or no treatment (risk ratio 0.64, 95% CI 0.47 to 0.87). 1

Primary objective

• The primary objective of the PRISM trial was to test the hypothesis that in women presenting with vaginal bleeding in the first trimester, receiving progesterone (400 mg vaginal capsules, twice daily) as soon as possible after identification of a visible intrauterine gestation sac with a scan until 16 completed weeks of gestation increases pregnancies with live births at ≥ 34 completed weeks by at least 5% compared with placebo.

Secondary objectives

• To test the hypothesis that progesterone improves other pregnancy and neonatal outcomes, including gestational age at birth and survival at 28 days of neonatal life.

• To test the hypothesis that progesterone, compared with placebo, is not associated with serious adverse effects for the mother or the neonate, including chromosomal anomalies in the newborn.

• To explore differential or subgroup effects of progesterone in prognostic subgroups, including age, fetal heart activity, gestation at presentation, amount of bleeding, body mass index and the number of previous miscarriages.

• To perform a cost-effectiveness analysis, with cost per additional birth over 34 weeks of gestation from an NHS and NHS/Personal Social Services (PSS) perspective. We will also model longer-term outcomes to the extent that the data permit.

Sequence generation and minimisation

Computer-generated random numbers were used, and participants were randomised online via a secure internet facility. This third-party independent ITMS was designed, developed and delivered by MedSciNet^® (MedSciNet UK Ltd, St Thomas’ Hospital, London, UK) in accordance with the standards of the International Organisation for Standardisation 2700015 and the requirements of the US Food and Drug Administration (FDA) CFR21:11. 16,17

Participants were randomised to receive progesterone or placebo in a 1 : 1 ratio. A ‘minimisation’ procedure via computer-based algorithm based on the method described by Pocock and Simon18 was used to avoid chance imbalances in important stratification variables. A random element was incorporated to make the treatment group less predictable. 19 The stratification variables (equally weighted) used for minimisation are listed below:

• age (< 35 or ≥ 35 years)

• body mass index (BMI) (< 30 or ≥ 30 kg/m²)

• fetal heart activity (present or absent)

• estimated gestational age at presentation (< 42 or ≥ 42 days)

• amount of bleeding [pictorial bleeding assessment chart (PBAC)]20 score of ≤ 2 or ≥ 3).

Allocation

When all of the eligibility criteria and baseline data items were entered online, the ITMS generated a trial number that took into account the minimisation variables recorded for the individual and that was linked to a specific trial intervention pack. The pack number was revealed via e-mail to the local principal investigator (PI), the relevant trial pharmacist (see Blinding) and the research nurse or midwife performing the randomisation. The trial intervention pack was dispensed to the patient by the clinical trial pharmacist at the randomising hospital. Each trial intervention pack contained either progesterone or an identical-looking placebo pessary.

Interventions

Each participant in the PRISM trial received either progesterone or placebo pessaries, to be administered vaginally. Both products were supplied by Besins Healthcare International (Besins Healthcare, Montrouge, France), a global pharmaceutical company with a manufacturer’s licence for tablets and capsules, in compliance with good manufacturing practice standards,21 good clinical practice requirements22 and Medicines for Human Use (Clinical Trials) Regulations 2004. 23 Besins Healthcare also provided qualified person release of the trial drug under the requirements of the Medicines for Human Use (Clinical Trials) Regulations 2004. 23

Progesterone pessaries

The IMP was a 400 mg dose of progesterone [i.e. two 200 mg pessaries of Utrogestan^® (micronised vaginal progesterone, Utrogestan^®, Besins Healthcare, Montrouge, France)] taken vaginally twice daily (every morning and every evening) for the duration of treatment. The product had all of the properties of endogenous progesterone with induction of a full secretory endometrium and, in particular, gestagenic, antiestrogenic, slightly antiandrogenic and antialdosterone effects.

Placebo pessaries

Placebo pessaries were vaginal pessaries, composed of sunflower oil, soybean lecithin, gelatin, glycerol, titanium dioxide and purified water, encapsulated in the same form as the IMP, and identical in colour, shape and weight, for use in the placebo arm of the PRISM trial. The dose, route and timing of administration were also identical to those in the active progesterone arm of the study.

Dose

The biologically effective dosage of progesterone pessaries ranged from 200 mg once daily to 400 mg twice daily according to the summary of product characteristics24 and the British National Formulary (BNF). 25 Our choice of 400 mg twice daily was made after a careful review of the existing literature and an extensive survey of clinicians in the UK (see Chapter 1, Progesterone in clinical use for threatened miscarriage). We also reviewed other related evidence. For example, progesterone vaginal capsules are commonly used for luteal support in assisted conception at a treatment dose of 400 mg twice daily, with no specific concerns for safety raised on this dose. 26,27 In addition, the findings from the PROMISE trial, which used the same dose, showed no safety concerns. 13 Therefore, after evaluating the evidence, we considered the dosage of 400 mg vaginal progesterone twice daily to be an acceptable regimen to ensure a clinically effective dose and to minimise the risk of a negative trial result from therapy with a suboptimal dose.

Timing of dose

Treatment commenced as soon as possible after confirmation of an intrauterine pregnancy sac and within 4 days of vaginal bleeding and continued until the gestational age of 16 weeks. Our rationale to discontinue the treatment at 16 weeks was that production of progesterone by the corpus luteum becomes less important when compared with the placental production of progesterone after 16 weeks of gestation. Furthermore, the largest (n = 191) and most recent of the seven previously published trials12 continued treatment until 16 weeks and found a large and statistically significant reduction in miscarriage risk (risk ratio 0.44, 95% CI 0.24 to 0.82). There was also overwhelming agreement among the clinicians and researchers involved in the preparation of this application that we should continue the progesterone until 16 weeks’ gestation.

Route

An immunomodulatory effect of progesterone at the trophoblastic–decidual interface is the key presumed mechanism for preventing miscarriage. 28–31 Our choice to use the vaginal route was, therefore, rational to deliver a greater proportion of the drug to the relevant site (the uterus) using the ‘first uterine pass’ effect. 32,33 Furthermore, studies that have used vaginal progesterone in the prevention of preterm birth have shown its effectiveness when given via this route. 34–36 For example, 14 out of 36 studies of second and/or third trimester progesterone to prevent preterm birth (identified by a recent systematic review) used vaginal progesterone, with significant improvements being observed for various clinical outcomes, confirming the biological effects of vaginal progesterone. 37

The acceptability and availability of interventional drugs were also important considerations supporting the vaginal route of drug delivery. Our discussions with consumer representatives confirmed that a vaginal formulation would be more acceptable to women than an intramuscular injection. These findings were further supported by a study in which 12% of participants were unable to tolerate the intramuscular progesterone preparation and declined participation or withdrew from that trial. 38 Of those who did continue, 34% complained of localised soreness around the injection site. Moreover, the Miscarriage Association conducted a survey to identify women’s opinions regarding acceptability of administering vaginal or rectal medications. The findings showed that the vaginal route of administration of medicines was acceptable to 100 out of 111 (90%) women, and the rectal route was acceptable to 91 out of 111 (82%) women. The pessary formulation of the PRISM trial is widely available in the UK and worldwide.

Instructions to participants

Each participant commenced the trial intervention on the day it was received and continued administration until it was finished, at 16 completed weeks of gestation, unless the pregnancy had ended before this time. Each participant was given instructions on how to administer the pessaries. In addition, each participant was asked for consent to notify her general practitioner (GP) by letter that she was participating in the trial. Moreover, each participant was given a card with contact details of local PRISM investigators and the central trial co-ordinating centre (TCC), the Birmingham Clinical Trials Unit, to inform any directing clinicians in case of potential drug interactions.

Concomitant non-trial treatments

Concomitant therapy was provided at the discretion of the care-providing clinicians, and all concomitant treatment and medications were documented via the ITMS. Other than identified contraindicated drugs (see Participants) and other progestogen preparations, the initiation of treatment for another indication did not necessitate withdrawal from the PRISM trial.

Compliance monitoring

Our previous experience of research and clinical care for women with miscarriage demonstrated that they would be highly motivated and compliant with therapy advice. However, compliance with the PRISM trial was evaluated by ‘pill counting’ in the first instance. Participants were asked to return completed, partially used and unused treatment packs to the trial centres. The research nurses and midwives at each study centre documented the pessaries returned by each participant, and the trial pharmacists kept their own accountability logs.

In an effort to improve compliance, women who failed to return their empty or unused blister packs were provided with an envelope to return them to the research team. Finally, if neither of these two approaches was successful, where possible, the patients were contacted directly by the research team via telephone and asked to give an honest assessment of their drug compliance in terms of what percentage of treatment they felt that they took.

Good compliance with the intervention was defined as taking > 80% of trial medicines from the date of allocation up to 16 weeks of gestation.

Participant withdrawal from treatment

Following discussion with the trial management group, participants in the PRISM trial could be withdrawn from the trial treatment if it became medically necessary in the opinion of the investigator(s) or clinician(s) providing patient care. In the event of such premature treatment cessation, study nurses and midwives made every effort to obtain and record information about the reasons for discontinuation and to follow-up all safety and efficacy outcomes as appropriate. Providing that the patient gave their continued consent, the follow-up information for these patients was still collected. Participants in the PRISM trial could also voluntarily decide to cease taking the study treatment at any time. If a woman stopped taking the trial treatment but permitted further data collection, she was followed up and outcome assessments were undertaken for the remainder of the study.

Primary outcome

Live births at or beyond 34 completed weeks of gestation (≥ 34 weeks), as a proportion of all women randomised.

Secondary outcomes

Secondary outcomes were as follows (as a proportion of those randomised unless stated):

• Time from conception to pregnancy end (any reason). Conception date was estimated using the date of last menstrual period or, failing that, the date from the ultrasound scan at 9–14 weeks.

• Ongoing pregnancy at 12 weeks of gestation.

• Miscarriage rate (defined as delivery before 24 weeks of gestation).

• Other pregnancy end outcomes – live birth at < 34 weeks’ gestation, ectopic pregnancy, termination, stillbirth, molar pregnancy, resolved pregnancy of unknown location (PUL), failed PUL, twin live births, gestational age at miscarriage.

• When there is live birth at ≥ 24 weeks’ gestation – time from conception to delivery (gestational age), gestational age < 28/< 32/< 37 weeks’ gestation, mode of delivery [unassisted vaginal, instrumental vaginal, elective Caesarean section (C-section), emergency C-section, vaginal breech delivery, other], birthweight adjusted for gestational age and sex, small for gestational age and sex (< 10th centile), arterial and venous cord pH, Apgar scores.

• Antenatal complications – pregnancy-induced hypertension, pre-eclampsia, obstetric cholestasis, cervical cerclarge, preterm (< 37 weeks’ gestation) pre-labour rupture of membranes, gestational diabetes mellitus (other complications will be tabulated but not formally analysed).

• Intrapartum complications – chorioamnionitis, intrauterine growth restriction, macrosomia (other complications will be tabulated but not formally analysed).

• Postpartum complications – haemorrhage (other complications will be tabulated but not formally analysed).

• Maternal complications – admission to a high-dependency unit (HDU), admission to an intensive therapy unit (ITU) (other complications will be tabulated but not formally analysed).

• Neonatal complications – discharge to hospital, early infection, retinopathy of prematurity, necrotising enterocolitis, intraventricular haemorrhage, congenital and chromosomal abnormalities, respiratory distress syndrome, ventilation or oxygen support (other complications will be tabulated but not formally analysed).

• Survival at 28 days of neonatal life.

• Maternal unexpected AEs (tabulated but not formally analysed).

• SAEs.

Resource use outcomes

These are detailed in Chapter 4.

Future outcomes

Each participant in the PRISM study was asked to consent for the future evaluation of themselves, the child who was born and the health records of both. Although long-term follow-up will remain outside the scope of this trial, we plan to conduct further studies, as discussed in Chapter 6, Recommendations for research.

Outcome generation

Details of how outcome measures were generated are given in Table 2. The ITMS was utilised to capture baseline and outcome data, and to maintain an audit trail. Relevant trial data were transcribed directly into the ITMS. Source data comprised the research clinic notes, hospital notes, hand-held pregnancy notes, laboratory results and self-reports.

Second outcome assessment (end of pregnancy)

The second outcome assessment was conducted at, or after, birth (Figure 3). The research nurse or midwife at each study site used the patient’s hospital notes to obtain pregnancy outcome data, such as the mode of delivery, gestation, weight and Apgar score at birth. If for any reason the research nurse or midwife was unable to access the hospital records, a telephone call was made to the patient to obtain as much follow-up information as possible.

FIGURE 3.

Participant care pathway and outcome assessment. OA, outcome assessment points; OA1, ongoing pregnancy beyond 12 weeks (range 11–14 weeks); OA2, live birth at > 34 weeks; OA3, survival at 28 days of neonatal life; R, randomisation.

Notes on adverse events and serious adverse events

All of the trial participants were asked to report any hospitalisations, consultations with other medical practitioners, disability, incapacity or any other AEs to their local research team. If the local study nurse or midwife was unavailable for any reason, they were able to report the events to the trial manager or trial co-ordinator via telephone at any time. Moreover, at the time of each outcome assessment, investigators, research nurses and midwives at each study centre proactively asked each participant about any AEs in the preceding weeks. AEs were assessed by clinical investigators, further reported as appropriate and recorded on the ITMS.

Serious adverse events and serious adverse reactions (SARs) were recorded on a purpose-designed SAE form and notified by local investigators to the TCC within 24 hours of the local investigators becoming aware of these events. In addition, local investigators were responsible for reporting SAEs to their host institutions in accordance with local regulations and instituting supplementary investigations as appropriate based on clinical judgement of the causative factors. Any SAE or SAR that was outstanding at the end of the trial treatment period was followed up at least until the final outcome was determined, even if this provision necessitated follow-up beyond 28 days post partum. The TCC reported all SAEs to the Data Monitoring and Ethics Committee (DMEC) approximately every 6 months. The DMEC viewed data blinded to treatment but was able to review unblinded data if requested.

Suspected unexpected serious adverse reactions (SUSARs) were unblinded, as appropriate, reviewed by the trial manager within 24 hours of reporting and further reported to the Medicines and Healthcare products Regulatory Agency and the regional ethics committee by the TCC as soon as possible for any event, within 15 days (or 7 days in the case of fatal or life-threatening SUSARs).

Secondary maternal outcome: pregnancy outcomes

There was evidence to suggest that progesterone increased the rate of ongoing pregnancy at 12 weeks: 1672 out of 2025 (83%) women in the progesterone group and 1602 out of 2013 (80%) in the placebo group remained pregnant at 12 weeks (adjusted relative risk 1.04, 95% CI 1.01 to 1.07; p = 0.01). However, there was no convincing evidence of a reduction in the number of miscarriages, with 410 out of 2025 (20%) women in the progesterone group and 451 out of 2013 (22%) in the placebo group experiencing a miscarriage (adjusted relative risk 0.91, 95% CI 0.81 to 1.01; p = 0.09). The median gestational age at the time of miscarriage was 8 weeks (IQR 7–10 weeks) for both groups. Time from conception to the end of pregnancy for any reason is graphically displayed in Figure 7.

FIGURE 7.

Subgroup forest plot. a, PBAC, Pictorial Blood Assessment Chart. 10

Other secondary maternal outcomes

Twenty-nine women in the progesterone group and 22 women in the placebo group gave birth to twins (see Table 7). There was evidence to suggest that women in the progesterone group were less likely to deliver via emergency C-section (15% in the progesterone group vs. 19% in the placebo group; adjusted relative risk 0.80, 95% CI 0.69 to 0.94; p = 0.006). The results of other secondary maternal outcomes appeared similar in both groups, with no significant differences.

Secondary neonatal outcomes

Overall, the distribution of gestational age at delivery in those women with a live birth was very similar in both groups. Live births were delivered at 38 ^+ 4 weeks on average in both groups. There were 498 (16%) preterm births (< 37 weeks) observed, but the numbers were very similar in both groups (17% in the progesterone group vs. 15% in the placebo group; adjusted relative risk 1.07, 95% CI 0.91 to 1.25; p = 0.42). Birthweights appeared similar across both groups (mean difference –21 g, 95% CI –67 to 25 g; p = 0.37), with no evidence of any differences in the numbers of infants being large or small for their gestational age (plus other covariates listed in Table 7). No differences were noted in other outcomes. Eight neonatal deaths were observed by 28 days in the progesterone group, compared with two in the placebo group (adjusted relative risk 3.84, 95% CI 0.80 to 18.40; p = 0.09).

Pregnancy-related complications

Complication rates of antenatal, intrapartum, postpartum and neonatal complications appeared similar for both groups (Table 8). The denominators throughout Table 8 differ across each outcome as they are based on the number of completed responses for that relevant outcome.

Safety data

The number of SAEs was similar in both groups: 105 out of 2025 (5%) compared with 98 out of 2013 (5%), in the progesterone and placebo groups, respectively (Table 9). The SAEs were categorised in body systems as detailed in Table 10.

Sensitivity analyses

A sensitivity analysis assuming that all missing responses were treatment failures for live birth at ≥ 34 weeks was statistically significant (adjusted relative risk 1.04; 95% CI 1.00 to 1.08; p = 0.04), but not when missing responses were simulated using multiple imputation. An analysis carried out where the dating scan was prioritised over the randomisation scan when calculating gestational age had no effect on the output (Table 11).

Subgroup analyses

The output from the subgroup analyses for the primary outcome of live birth at ≥ 34 weeks can be seen in Table 12 and Figures 7 and 8. All tests for subgroup by treatment group interaction were non-significant apart from number of previous miscarriages. Here, there was evidence that the number of live births was higher in the progesterone group than in the placebo group (72% in the progesterone group vs. 57% in the placebo group; relative risk 1.28, 95% CI 1.08 to 1.51; p = 0.007) for those who had three or more previous miscarriages. Further post hoc subgroup analysis on the number of previous miscarriages, where the subgroup was split into none compared with ≥ 1 previous miscarriage, suggested that progesterone was effective in those who had ≥ 1 previous miscarriage (75% in the progesterone group vs. 70% in the placebo group; relative risk 1.09; 95% CI 1.03 to 1.15; p = 0.01).

FIGURE 8.

Post hoc subgroup analyses. The results are reported as point estimates and 95% CIs. The widths of the CIs have not been adjusted for multiplicity, so the intervals should not be used to infer definitive treatment effects.

Subgroup analyses for the outcome of miscarriage (at < 24 weeks) can be seen in Table 13. The results were similar to the primary outcome subgroup analyses, with the only evidence of a differential effect observed in the subgroup of number of previous miscarriages. Here, in those with three or more previous miscarriages, there was some evidence that progesterone was effective (23% in the progesterone group vs. 36% in the placebo group; relative risk 0.58, 95% CI 0.40 to 0.83; p = 0.003).

Outcomes

The primary outcome of the CEA was live births at ≥ 34 completed weeks of gestation based on the principal outcome of the clinical trial. Gestational age was estimated based on participants’ ultrasonography result and 11–14 weeks, when available, or otherwise based on the ultrasonography at randomisation. An additional outcome of the PRISM trial was neonatal survival at 28 days post partum, and this was explored as a secondary outcome in the CEA.

Resource use and costs

Data on all resources consumed in the hospital setting by each woman from randomisation to hospital discharge were collected prospectively using researcher-recorded trial collection forms. The use of other services provided by the community during the same period was captured retrospectively via health services self-completed questionnaires. Unit costs for each resource item (Table 14) were identified from established national sources. 47,50 All costs were expressed in 2017–18 Great British pounds. Cost estimates from earlier years were inflated to 2017–18 prices using the Hospital and Community Health Services (HCHS) pay and prices index. 50 Hospital-related unit costs values were obtained from the NHS Reference Costs 2016/1747 where available. Otherwise, these costs were obtained from reference costs for earlier years or from other sources, such as the Personal Social Services Research Unit (PSSRU) costs. In cases where there are different categories associated with a resource use, weighted averages were used (see Table 14).

Resource use data within the hospital setting (inpatient and outpatient) focused on the:

• quantity of progesterone administered

• antenatal period

• intrapartum period

• postnatal period (maternal and neonatal).

Data were collected for primary care resources such as:

• contacts with GPs

• contacts with community midwives.

Progesterone

The quantity of progesterone vaginal pessaries was calculated based on the number of days they were used from randomisation until the end of 16 weeks of gestation (or earlier if miscarriage occurred before 16 weeks). The cost of progesterone was identified from the BNF46 as £21 for a 21-pessary pack. In the trial, each woman used four pessaries daily, which is equivalent to a cost of £4 per day (see Table 14).

Antenatal period

For the antenatal period, resource use data were collected on the number of hospital, day assessment unit (DAU) and emergency visits as well as the number of inpatient day admissions (for a stay of < 24 hours) and nights of inpatient admissions. Based on the descriptions in the trial literature, the costs of antenatal hospital (routine observation) and DAU (antenatal specialised non-routine ultrasound scan) visits were obtained from the NHS Reference Costs 2016/17. 47

For inpatient day admission (< 24 hours), this was described in the trial as a day-case management of an antenatal disorder. 47 Emergency visit (antenatal diagnostic procedures) costs were provided from the NHS Reference Costs 2013 to 2014,48 whereas inpatient night admissions costs were obtained from an earlier PSSRU cost. 49 Because all participants in the trial underwent routine ultrasonography at specified times in the study, the cost of an ultrasound was not included in the analysis.

Intrapartum period

The resource use collected at the end of pregnancy varied depending on whether or not the baby was born alive. Where live births occurred, the onset of labour was recorded as spontaneous, augmented, pre-labour C-section or induced. Information on the mode of delivery included unassisted vaginal deliveries, instrumental vaginal deliveries, elective C-sections, emergency C-sections or vaginal breech deliveries with or without intrapartum complications. Deliveries such as water births and home deliveries were captured as ‘others’.

The Healthcare Resource Group (HRG) unit costs47 for delivery mode are categorised as normal vaginal delivery, assisted vaginal delivery, planned C-section and emergency C-section, corresponding to unassisted vaginal delivery, instrumental vaginal delivery, elective C-section and emergency C-section on the case report form (CRF), respectively. Each category is grouped further as with or without complications. Weighted averages of the unit costs for the different levels of complications were calculated. As there were no HRG unit costs available for breech delivery or ‘other’, in consultation with the clinical team, we assumed that the costs were the same as those for a normal vaginal delivery. For labour onset, we assumed that these costs were already accounted for by the delivery mode and, hence, these were not costed separately.

Where babies were not born alive, the outcome was recorded as miscarriage, ectopic pregnancy, termination or stillbirth. The management of such outcomes was recorded as spontaneous resolution, surgical management or medical management. The costs of management were provided in the NHS Reference Costs as threatened or spontaneous miscarriage with intervention (medical or surgical management) and without intervention (spontaneous resolution). 47

Postnatal period

During the postnatal period, data were collected for both maternal and neonatal resource use, from pregnancy end to 28 days post discharge.

Maternal resource use

The unit costs for hospital visits, DAU visits, emergency visits and others that applied to the antenatal period were also relevant to the postnatal period resource use; relevant unit costs are presented in Table 14. For postnatal DAU visits and inpatient admissions, the corresponding costs for the antenatal indices were used. The cost of a postnatal emergency visit was obtained using a weighted average of emergency medicine for patients requiring category 0–2 treatment and category 0–1 investigation. 47 Postnatal hospital visit costs were obtained from the PSSRU. 49

Data collected for immediate maternal postnatal care resource use included admissions to a HDU (level 2 care) or an ITU (level 3 care). Using the UK definitions for level 2 care (patient receiving a single organ support) and level 3 care (patient receiving at least two organ supports),53 the unit costs for these admissions were obtained from the NHS Reference Costs 2016/17. 47 For level 2 care (HDU admissions), ‘adult critical care – one organ supported’ was used; for level 3 care (ITU admissions), a weighted average was taken across five HRGs [adult critical care (two organs supported) to adult critical care (six or more organs supported)].

Neonatal resource use

The immediate neonatal care resource use included the number of nights receiving intensive care, high-dependency care and special care. Costs for these resources were obtained from the costs schedule. 47

Serious adverse events

Information on SAEs was collected using SAE forms. In this study, a SAE was defined as an untoward event resulting in maternal death, stillbirth, hospitalisation, persistent or significant disability, congenital anomaly or birth defect. Only clinically specified SAEs deemed to have arisen from the trial intervention were considered to be relevant to the economic analysis. Because there were no SAEs that were clearly related to the use of progesterone, we did not include such costs in the analysis.

Primary care services

Service use questionnaires, completed by a subsample of women, captured data on primary care service resource use in the trial period. These included the number of visits to the GP, practice/community midwife, practice nurse, psychologist (or counsellor), health visitor, social worker and other community services. However, these services were recorded as the number of visits without a specified duration.

The costs for primary care services were obtained from the Unit Costs of Health and Social Care 2017. 50 To cost each primary care resource use, we used the recommended average duration of 9.22 minutes for a GP face-to-face visit, 30 minutes for a midwife visit, 15.5 minutes for a GP nurse visit and 20 minutes for the remaining variables. Because no telephone contacts with the GP or practice nurse were recorded, we did not include these costs.

Primary economic analysis

A within-trial incremental CEA was conducted to estimate the relative costs and benefits of progesterone compared with placebo. The cost-effectiveness of progesterone was expressed in terms of the cost per additional live birth after ≥ 34 complete weeks of gestation. The base-case primary analysis focused on the hospital-related (inpatient and outpatient) costs for the participants incurred in the trial period.

Using study-specific resource use and costs, the total cost over the trial period was calculated by multiplying the resource items used by the corresponding unit cost and adding up all items. The mean total costs and mean total resource use for participants across the trial arms were calculated. Given the skewness inherent in most cost data and the concern of economic analyses with mean costs, we calculated 95% CIs around mean differences through the analyses of 1000 resamples using the bias-corrected and accelerated (BCa) bootstrap method. 54

To explore heterogeneity in the trial population, multivariate cost analyses were performed using seemingly unrelated regressions. 55,56 Seemingly unrelated regression has been shown to be robust to skewed data and allows for a correlation in the error terms between costs and outcomes. 57 Model covariates included baseline data on age, BMI, the quantity of bleeding and the number of previous miscarriages. The selection of covariates was informed by the prognostic variables used by the clinical team. All results were presented as mean values with SD and, where applicable, as mean differences in costs and effects with 95% CIs.

Incremental cost-effectiveness ratios (ICERs) were calculated by dividing the difference in mean total cost between the trial arms by the difference in the number of live births at ≥ 34 weeks. The ICER is a measure that depicts the additional cost ascribed to an additional effect. To calculate ICERs, the formula below was used, with C representing cost and E representing effects:

To quantify the uncertainty that typically occurs as a result of variations in sampling, we used non-parametric bootstrapping to resample the joint distribution in the mean cost and outcome difference. 58 This generated 5000 paired estimates of incremental costs and outcomes, which were plotted in a cost-effectiveness plane as a scatterplot. 59 A cost-effectiveness plane is a four-quadrant plane depicting bootstrap estimates. Based on the location of the scatterplot dots on the quadrant, an intervention may be deemed more effective and less costly (south-east), more effective and more costly (north-east), less effective and more costly (north-west) or less effective and less costly (south-west) than the alternative intervention.

A cost-effectiveness acceptability curve (CEAC) was constructed to show the probability that progesterone is a cost-effective intervention compared with placebo across a range of values, representing the decision-maker’s willingness to pay (WTP) for an additional benefit. 44 Currently, there is no standard valuation for an additional live birth. 2,60 In the UK, NICE typically uses a WTP threshold of £20,000–30,000 per quality-adjusted life-year (QALY) in approving a health-care intervention. 44

Secondary economic analyses

Secondary analyses involving ICER calculations on the primary outcome are based on the following costs.

• Hospital-related costs for resource use for women only: the hospital-related costs for women included antenatal, intrapartum and postnatal care costs. We excluded the neonatal care costs accrued by infants from the total cost for this analysis to examine which aspect of cost had the largest effect on the result – the women’s cost or the neonatal care cost.

• Hospital and primary care costs: here we added primary care costs such as GP and practice nurse visits to the hospital costs. This analysis was conducted to explore the perspective of the NHS/PSS.

• Adjusting for missing data: the main base-case analysis was carried out on only individuals for whom there were outcome data. For this analysis, we included participants who were lost to follow-up. We assumed that all women lost to follow-up had a miscarriage and we imputed the costs for this subgroup. Missing costs were imputed using multiple imputations61 by applying chained equations with predictive mean matching across 60 imputations. 62

• Hospital-related costs for women with three or more previous miscarriages: preliminary results suggested that there was clinical effectiveness for women with three or more previous miscarriages. Therefore, we conducted a CEA for this subgroup.

We also carried out a secondary incremental CEA based on the final end point using hospital-related costs for participants with complete data. We reported the analysis in terms of cost per additional baby who survived beyond 28 days after birth for each woman.

Sensitivity analysis

Deterministic sensitivity analyses were conducted to explore the inherent uncertainty in key assumptions and variations in the analytical methods used, and to consider the broader issue of the generalisability of the results. This involved varying some of the parameters while leaving others at their baseline value. A number of sensitivity analyses were conducted, as detailed below.

• Fixed cost of treatment until 16 weeks of gestation. In the PRISM trial, women in early pregnancy started treatment from randomisation until 16 weeks’ gestation. If it is assumed that all women started treatment at approximately 7.4 weeks (52 days) with no miscarriage and continued treatment until 16 weeks (112 days) of gestation, this would translate to 60 days of treatment. Based on the progesterone vaginal pessary cost of £4 per day (as described above for four pessaries) and assuming that on day 1 participants used two pessaries, the expected cost of progesterone is £238 [2 + (59 × 4)]. In a real-life scenario, the intervention would be provided for the expected treatment period (from 6 to 8 weeks) until 16 weeks, and hence we explored the impact of a fixed cost of progesterone until 16 weeks of gestation.

• Unit costs. We explored the impact of alternative cost estimates. For inpatient night of admission for both the antenatal and postnatal periods, we replaced the cost used in the primary analysis with the cost of excess bed-days (£311). 47 The cost of management of miscarriage used in the main analyses was obtained from the NHS Reference Costs. 47 However, the NICE guideline on miscarriage management3,63 provides an estimate of £1522 for medical management and £1827 for surgical management. These values have been used by other studies. 13 For the sensitivity analysis, we replaced the costs of management with these costs.

• Primary care costs. We imputed the missing costs for primary care costs using multiple imputations61 by applying chained equations with predictive mean matching across 60 imputations. 62 All statistical analyses were performed using Stata version 14 (StataCorp LP, College Station, TX, USA).

Model-based analysis

Preliminary results showed that there was no clinically detectable effect on neonatal outcomes as a result of the PRISM trial that was a result of progesterone. This finding is in keeping with an earlier study in which women were given progesterone. 2,64 In view of this result, modelling costs and outcomes beyond the trial period was not deemed necessary.

Outcomes

The details of the major outcomes of the trial are presented in Table 15. At the end of pregnancy, 1513 (74.72%) and 1459 (72.48%) women in the progesterone and placebo arms, respectively, had live births after 34 completed weeks of pregnancy. This translates to an effect difference of approximately 2.2% (0.022, 95% CI –0.004 to 0.050). Among women who had live births during the trial period, babies born to 1538 out of 2025 (75.95%) women in the progesterone arm and 1487 out of 2013 (73.87%) women in the placebo arm survived beyond 28 days of birth.

Resource use and costs

A breakdown of the resource use data by trial arm is presented in Table 16. Mean health-care costs per participant by trial arm are presented in Table 17. During the trial, 2023 women in the intervention arm received progesterone and 2009 women in the non-intervention arm received placebo. Based on the mean number of days that participants utilised progesterone pessaries, the average cost of the intervention was calculated to be £204 (95% CI £200 to £207) per woman (see Table 17). The most substantial costs accrued during the trial by participants were from antenatal hospital visits, with a mean cost of £2339 (SD £2672) per woman for the progesterone arm and £2334 (SD £2665) per woman for the placebo arm.

During the antenatal period, women allocated to the progesterone arm had, on average, a higher frequency of antenatal and DAU visits but a smaller number of emergency room visits and hospital admissions than women in the placebo arm. During the postnatal period, women in the progesterone arm utilised similar services more than those in the placebo arm except for emergency hospital visits, which were the same for both arms. However, women in the placebo arm had more admissions to the HDU [mean 0.06 (SD 0.46) for placebo vs. mean 0.05 (SD 0.30) for the progesterone arm]. Similarly, babies born to women in the placebo arm had, on average, a greater number of admissions to the HDU [mean 0.52 (SD 3.90) for placebo vs. mean 0.42 (SD 4.02) for the progesterone arm] and neonatal special care [mean 1.16 (SD 4.95) for placebo vs. mean 1.02 (SD 5.07) for the progesterone arm].

For delivery mode, women in the placebo arm had, on average, more emergency C-sections than women in the intervention arm. On average, women in the placebo arm utilised more neonatal care services than those in the intervention arm.

In keeping with the mean resource use, antenatal care costs for DAU and hospital visits were higher in the intervention arm, whereas emergency visits and hospital admissions costs were higher in the placebo arm.

In terms of cost differences, the greatest mean value for the participants was for emergency C-section with complications, which was greater in the placebo group than in the treatment group (–£137, 95% CI –£246 to –£28). The highest cost difference as a result of the intervention was for elective C-section without complication with a mean difference of £48 (95% CI –£11 to £108) per participant. Neonatal care variables were consistently lower in the progesterone group [ITU, –£10 (95% CI –£442 to £421); HDU, –£93 (95% CI –£344 to £159) and special care unit, –£76 (95% CI –£260 to £109)].

Mean total costs

The mean total costs by trial group for different variables are presented in Table 18. The average hospital-related service costs per woman for the trial period was £7452 in the progesterone group and £7572 in the placebo group, generating a mean cost difference of –£127 (BCa mean –£127, 95% CI –£759 to £505). The inclusion of the intervention cost (£204) generated a mean difference of £83 (adjusted mean £76, 95% CI –£559 to £711) per woman, which increased slightly to £78 (95% CI –£563). The differences in cost were not statistically significant.

Primary analysis

The primary (base-case) cost-effectiveness outcome of the PRISM trial was the cost for an additional live birth after ≥ 34 completed weeks of pregnancy. The progesterone intervention appeared to be slightly more effective than placebo, resulting in an additional two live births per 100 women (an effect difference of 0.022, 95% CI –0.004 to 0.050) at ≥ 34 weeks’ gestation. The ICER, which combines the differences in costs in both groups, is presented in Table 19. The administration of progesterone resulted in an estimated additional cost of £83 per woman (adjusted mean £76, 95% CI –£559 to £711).

Given the differences in costs and effects, the point ICER estimate for progesterone compared with placebo was calculated at £3305 per additional live birth.

Figure 9 shows the results of 5000 bootstrap replications plotted on the cost-effectiveness plane for the primary analysis. Each point on the plane depicts a pair of incremental cost and incremental effectiveness estimates for the comparison between progesterone and placebo. This suggests that progesterone is likely to be more effective, given that the majority of the scatterplots are in the south-east and north-east quadrants. However, it is uncertain whether the progesterone intervention is likely to be more costly (north-east) or less costly (south-east) than no intervention.

FIGURE 9.

Cost-effectiveness plane for the primary analysis using hospital-related costs for the participants. Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

The CEAC for the primary analysis (Figure 10) shows the probability of progesterone being cost-effective at various values of decision-makers’ WTP per additional live birth. Figure 10 indicates that, for thresholds of WTP per additional live birth of > £15,000, there is > 80% probability that progesterone is cost-effective. For WTP thresholds of > £30,000, the probability of cost-effectiveness exceeds 90%.

FIGURE 10.

The CEAC for the primary analysis using hospital-related costs for the participants. Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Secondary analysis I (hospital-related costs for participants minus neonatal care costs)

For the first secondary analysis, we removed neonatal care costs from the hospital costs. This resulted in an adjusted mean cost difference of £170 (95% CI –£113 to £453) (Table 20).

TABLE 20 - Point estimate ICER for secondary analysis I (maternal hospital costs)

Treatment	Mean cost (£)	Mean effect	Cost difference (£) (95% CI)	Effect difference	ICER (£)
Progesterone	6467	0.745	170, –113 to 453	0.022	7370
Placebo	6298	0.726	–	–	–

The ICER was calculated as £7370 per additional live birth beyond 34 weeks of gestation. The increased ICER value for this analysis is probably because, on average, women in the placebo arm utilised more neonatal hospital resources than women in the progesterone arm. Hence, the removal of neonatal care costs resulted in a higher cost difference.

The cost-effectiveness plane (Figure 11) shows that progesterone is the more effective and more costly intervention, with the majority of the bootstrap replications in the south-east quadrant. The CEAC (Figure 12) suggests that, for WTP thresholds of > £12,000 per additional live birth, there is > 95% probability that progesterone is a cost-effective intervention.

FIGURE 11.

Cost-effectiveness plane for secondary analysis I (maternal hospital costs).

FIGURE 12.

The CEAC for secondary analysis I (maternal hospital costs).

Secondary analysis II (hospital and primary care costs for participants)

In another secondary analysis, we included hospital-related costs and primary care costs for the participants. First, we explored the total primary care cost for women with complete primary care service use data; complete data were available for 272 participants. In this subgroup, the mean total cost was £120 for women in the progesterone arm and £98 for women in the placebo arm. For each woman in the progesterone arm, we added £120 to the total hospital-related cost; for each woman in the placebo arm, we added £98 to the total hospital-related cost (Table 21). We calculated an additional cost of £106 (adjusted mean £98, 95% CI –£537 to £733) and an ICER of £4264 per additional live birth.

TABLE 21 - Point estimate ICER for secondary analysis II (hospital and primary care costs)

Treatment	Mean cost (£)	Mean effect	Cost difference (£) (95% CI)	Effect difference	ICER (£)
Progesterone	7776	0.747	98, –537 to 733	0.022	4264
Placebo	7670	0.725	–	–	–

Again, the cost-effectiveness plane (Figure 13) clearly suggests that progesterone is more effective. Figure 14 depicts the CEAC for this analysis. For WTP thresholds of > £15,000, per additional live birth beyond 34 weeks’ gestation, the probability of progesterone being more effective than placebo is over 80%. The probability of cost-effectiveness exceeds 90% for WTP thresholds of > £30,000.

FIGURE 13.

Cost-effectiveness plane for secondary analysis II (hospital and primary care costs).

FIGURE 14.

The CEAC for secondary analysis II (hospital and primary care costs).

Secondary analysis III (hospital costs for participants including those lost to follow-up)

For this analysis, we included both the women used for the primary base-case analysis and those who were lost to follow-up. We explored a worst-case scenario and assumed that all women lost to follow-up had a miscarriage. We used multiple imputations to impute missing costs and re-ran the primary analysis. This included 2069 participants in the progesterone arm and 2054 participants in the placebo arm.

The results (Table 22) showed that progesterone was slightly more costly (cost difference £29, 95% CI –£593 to £651) and more effective than no progesterone for this subgroup, with an ICER of £1378 per additional live birth beyond 34 weeks of gestation.

TABLE 22 - Point estimate ICER for secondary analysis III (hospital costs for participants including those lost to follow-up)

Treatment	Mean cost (£)	Mean effect	Cost difference (£) (95% CI)	Effect difference (95% CI)	ICER (£)
Progesterone	7788	0.731	29, –593 to 651	0.021, –0.007 to 0.049	1378
Placebo	7750	0.710	–	–	–

From the cost-effectiveness plane (Figure 15), it is evident that progesterone is more effective; however, it is uncertain which intervention is more costly. The CEAC (Figure 16) shows that, for WTP thresholds of > £15,000 per additional live birth beyond 34 weeks of gestation, the probability of progesterone being cost-effective is approximately 85%, and the probability exceeds 90% for WTP thresholds of > £30,000.

FIGURE 15.

Cost-effectiveness plane for secondary analysis III (hospital costs for participants including those lost to follow-up).

FIGURE 16.

The CEAC for secondary analysis III (hospital costs for participants including those lost to follow-up).

Secondary analysis IV (hospital costs for participants with three or more previous miscarriages)

We conducted a subgroup analysis of women with three or more previous miscarriages. This included 137 women in the intervention arm and 148 women in the placebo arm. The intervention was more effective, with an additional gain of 15 live births per 100 women, beyond ≥ 34 weeks' gestation (Table 23). An ICER of £11,606 per additional live birth beyond 34 weeks’ gestation was calculated for this subgroup.

TABLE 23 - Point estimate ICER for participants with three or more previous miscarriages

Treatment	Mean cost (£)	Mean effect	Cost difference (£) (95% CI)	Effect difference (95% CI)	ICER (£)
Progesterone	9304	0.715	1754, –1041 to 4550	0.151, 0.042 to 0.260	11,606
Placebo	7803	0.574	–	–	–

Similarly, the cost-effective plane (Figure 17) depicts that progesterone is more effective and more costly with the majority of the bootstrap replications in the south-east quadrant. The CEAC (Figure 18) shows over > 90% probability of the intervention being cost-effective for WTP thresholds of > £15,000.

FIGURE 17.

Cost-effectiveness plane for participants with three or more previous miscarriages. Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

FIGURE 18.

The CEAC for participants with three or more previous miscarriages. Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Reproduced from Okeke Ogwulu et al. 52 © 2020 The Authors. BJOG: An International Journal of Obstetrics and Gynaecology published by John Wiley & Sons Ltd on behalf of Royal College of Obstetricians and Gynaecologists. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Analyses using secondary outcomes

Incremental CEAs were conducted for the final end point of the PRISM trial (Table 24). The analysis was based on the incremental cost of the intervention for an additional baby survival for each woman at 28 days post partum. The effect difference was 0.021 (95% CI –0.005 to 0.048). The ICER for this analysis was £3037 per additional baby who survived beyond 28 days post birth.

Sensitivity analyses

We conducted sensitivity analyses in which we explored different scenarios (Table 25). Using alternative cost estimates for nights of hospital admissions and miscarriage management appeared to have a limited effect on the resulting ICER (see Table 25).

However, using a fixed cost of progesterone until 16 weeks increased the ICER to £4977 per additional live birth beyond 34 weeks’ gestation (see Table 25). The cost-effectiveness plane (Figure 19) showed dominance in the north-east and south-east quadrants, which indicated that progesterone was more effective and either less costly or more costly. The CEAC (Figure 20) shows over 90% probability of the intervention being cost-effective for WTP thresholds of > £25,000. Likewise, the removal of the cost of delivery increased the ICER to £3743 per additional live birth beyond 34 weeks (see Table 25).

FIGURE 19.

Cost-effectiveness plane for a fixed cost of progesterone until 16 weeks’ gestation.

FIGURE 20.

The CEAC for a fixed cost of progesterone until 16 weeks’ gestation.

In secondary analysis II, we added mean totals of the primary care costs to the total costs for each woman depending on the trial arm. From this, we calculated an ICER of £4264 per additional live birth beyond 34 weeks of gestation (see Table 21). The sensitivity analysis in which we imputed costs via multiple imputations did not have much effect on the ICER, with a slight increase to £4321 per additional live birth beyond 34 weeks (see Table 25).

Principal findings

We evaluated the cost-effectiveness of progesterone in preventing miscarriage and leading to a live birth at ≥ 34 weeks of pregnancy in women who presented with bleeding in early pregnancy. Our results suggest that progesterone is more effective and slightly more costly than placebo. More specifically, progesterone resulted in an additional two live births per 100 women (0.022, 95% CI –0.004 to 0.050) at ≥ 34 weeks of gestation relative to placebo, with an additional cost of £83 (adjusted mean difference £76, 95% CI –£559 to £711) per woman. The additional cost was mainly attributable to the cost of progesterone administration (mean cost £204). The ICER was estimated at £3305 per live birth at ≥ 34 weeks. A conclusion on the cost-effectiveness of the PRISM trial would depend on the amount that society is willing to pay to increase the chances of an additional live birth at ≥ 34 weeks of pregnancy. For potentially acceptable WTP threshold values for an additional live birth,44 the probability of progesterone being cost-effective for this population group exceeds 90%.

The National Institute for Health and Care Excellence attached a value to an averted stillbirth of 25 QALYs. 65 This assumes that life lost has a typical life expectancy in good health, but when discounting is applied to the expected years in full health, it yields 25 discounted QALYs. To interpret the primary CEA results in relation to QALYs, we used the NICE value (25 QALYs) as a proxy. If we assume that babies born alive at ≥ 34 weeks live in full health and divide the ICER (£3305 per additional live birth) by 25, then the cost per QALY is likely to be £132. If a baby did live in full health for the anticipated life expectancy, then on the basis of this ICER (£132) the intervention is cost-effective. 44 Furthermore, evidence from the NHS Reference Costs schedule47 indicates that the upper cost quartile of the most expensive delivery is about £15,000, which could go much higher if we allow for the cost of excess bed-days. This further suggests that progesterone intervention is cost-effective.

The ICER for the final end point (secondary outcome) of the trial was £3037 per additional baby surviving beyond 28 days after birth. The intervention was more effective, with a gain of three neonates per 100 women surviving beyond 28 days post partum. A subgroup analysis of women with three or more previous miscarriages led to an increase of 15 live births per 100 women in the intervention group.

Strengths and limitations of the economic analyses

The strength of the CEA is that it was based on a large, robust, multicentre randomised controlled trial involving over 4000 participants, making this the largest study to explore whether or not progesterone provides value for the public health-care resources. The outcome and resource use data were prospectively collected at different points in the trial using CRFs. Unit costs were obtained from established national sources. In cases where HRGs did not clearly depict our variables, we liaised with the clinical team to decide on the most appropriate HRG. The CEA also benefited from the robustness of the main analyses and the sensitivity analyses. However, data on primary care services were available for < 10% of the participants: we accounted for this by imputing missing costs in our analyses. A limitation of this analysis was the failure to explore the wider societal costs to the participants. However, this was beyond the scope of the study and beyond the requested resource.

Comparison with the literature

To our knowledge, this is the first UK study to investigate the cost-effectiveness of progesterone in preventing miscarriage and achieving a live birth beyond 34 weeks of gestation. A similar study investigated the cost-effectiveness of progesterone in preventing miscarriages in women with a history of recurrent miscarriages and leading to a live birth beyond 24 weeks of gestation. 2 The authors reported that the total mean cost of the intervention was £332.17 higher in the progesterone arm than in the placebo arm and an ICER of £18,053 per additional live birth beyond 24 weeks for the base-case analysis, with a cost-effectiveness probability of 50% at this value.

Implications for policy

The results of the CEA suggest that progesterone is likely to be considered a cost-effective intervention by decision-makers for women presenting with early pregnancy bleeding (threatened miscarriage) within 12 weeks of gestation.

Summary of health economic findings

• For the primary analysis, the mean total cost was higher in the progesterone group (£7655) than in the placebo group (£7572), with an additional cost of £83 (1% higher cost than usual care).

• The additional difference in the mean probability of a live birth beyond ≥ 34 completed weeks of gestation was 0.022 (95% CI –0.004 to 0.050), indicating that the progesterone intervention resulted in an additional two live births per 100 women at ≥ 34 weeks.

• For the primary analysis, the ICER per additional live birth beyond 34 weeks of gestation was calculated as £3305.

• There is > 80% confidence that progesterone is cost-effective if decision-makers are prepared to pay £15,000 per additional live birth and > 90% if the WTP threshold is £30,000.

• Currently, in the UK, progesterone is not routinely given to women who are at high risk of miscarriage. The results of the CEA suggest that progesterone is likely to be considered cost-effective, particularly for women (with one or more miscarriages) who present with bleeding in early pregnancy.