Minimising twins in in vitro fertilisation: a ... - Wiley Online Library

Fertility and assisted reproduction

DOI: 10.1111/j.1471-0528.2011.02966.x www.bjog.org

Minimising twins in in vitro fertilisation: a modelling study assessing the costs, consequences and cost–utility of elective single versus double embryo transfer over a 20-year time horizon GS Scotland,a D McLernon,b JJ Kurinczuk,c P McNamee,a K Harrild,b H Lyall,d M Rajkhowa,e M Hamilton,f S Bhattacharyag a

Health Economics Research Unit, University of Aberdeen, Aberdeen, UK b Section of Population Health, University of Aberdeen, Aberdeen, UK c National Perinatal Epidemiology Unit, University of Oxford, Oxford, UK d Assisted Conception Services Unit, NHS Glasgow, Glasgow, UK e Birmingham Women’s Fertility Centre, Birmingham Women’s Hospital, Birmingham, UK f Department of Obstetrics and Gynaecology, Aberdeen Maternity Hospital, Aberdeen, UK g Department of Obstetrics and Gynaecology, University of Aberdeen, Aberdeen, UK Correspondence: Dr GS Scotland, Health Economics Research Unit, Polwarth Building, University of Aberdeen, Aberdeen AB25 2ZD, UK. Email [email protected] Accepted 25 February 2011. Published Online 8 April 2011.

Objectives To assess the cumulative costs and consequences of

double embryo transfer (DET) or elective single embryo transfer (eSET) in women commencing in vitro fertilisation (IVF) treatment aged 32, 36 and 39 years. Design Microsimulation model. Setting Three assisted reproduction centres in Scotland. Sample A total of 6153 women undergoing treatment at one

of three Scottish IVF clinics, between January 1997 and June 2007. Methods A microsimulation model, populated using data inputs

derived from a large clinical data set and published literature, was developed to compare the costs and consequences of using eSET or DET over multiple treatment cycles. Main outcome measures Disability-free live births; twin

pregnancy rate; women’s quality-adjusted life-years (QALYs); health service costs.

Results Not only did DET produce a higher cumulative live birth rate compared with eSET for women of all three ages, but also a higher twin pregnancy rate. Compared with eSET, DET ranged from costing an additional £27 356 per extra live birth in women commencing treatment aged 32 years, to costing £15 539 per extra live birth in 39-year-old women. DET cost £28 300 and £20 300 per additional QALY in women commencing treatment aged 32 and 39 years, respectively. Conclusions Considering the high twin pregnancy rate associated

with DET, coupled with uncertainty surrounding QALY gains, eSET is likely to be the preferred option for most women aged £36 years. The cost-effectiveness of DET improves with age, and may be considered cost-effective in some groups of older women. The decision may best be considered on a case-by-case basis for women aged 37–39 years. Keywords Cost-effectiveness, in vitro fertilisation, single embryo

transfer.

Please cite this paper as: Scotland G, McLernon D, Kurinczuk J, McNamee P, Harrild K, Lyall H, Rajkhowa M, Hamilton M, Bhattacharya S. Minimising twins in in vitro fertilisation: a modelling study assessing the costs, consequences and cost–utility of elective single versus double embryo transfer over a 20-year time horizon. BJOG 2011;118:1073–1083.

Introduction Multiple pregnancy is the commonest major complication associated with in vitro fertilisation (IVF) treatment1 and is caused by the replacement of more than one embryo in an attempt to enhance pregnancy rates. The increased clinical

risks associated with multiple pregnancies has prompted changes in practice, leading to a reduction in the number of embryos transferred from three to two in many countries, but in 2005/06 twins still accounted for 19.9% of all IVF live births in Europe2 compared with a natural incidence of 1–1.5% in spontaneously conceived births.3

ª 2011 The Authors BJOG An International Journal of Obstetrics and Gynaecology ª 2011 RCOG

1073

Scotland et al.

This is of significant concern not just to fertility specialists but also to obstetricians, neonatologists and primary-care providers as twin pregnancies are associated with increased maternal and neonatal morbidity and mortality,4 and increased costs to the health service.5 Considerations of safety have been a driver in many European countries for implementation of elective single embryo transfer (eSET) as an effective means of minimising the twin pregnancy rate associated with IVF. This decision has been based largely on data from randomised controlled trials, which have shown that eSET, followed by replacement of a single frozen/thawed embryo in those who fail to become pregnant, can virtually eliminate twins and yet maintain live birth rates comparable with those achieved through a double embryo transfer (DET) cycle in women aged under 36 years.6 The Human Embryology and Fertilisation Authority (HFEA), which regulates IVF in the UK, now advocates the use of eSET as an effective way of reducing twins associated with IVF7 and recommends a national target of a twin birth rate below 10% (www.hfea.gov.uk/530.html). However, the existing randomised and observational studies have not considered the effectiveness and cost-effectiveness of using eSET and DET over multiple complete treatment cycles, where all women have the opportunity to move to a frozen cycle (if spare embryos are available) or subsequent fresh treatment cycle, if the previous treatment fails. There is also limited evidence regarding the cost-effectiveness of using eSET in women who did not meet the inclusion criteria of existing randomised controlled trials. A major source of concern among clinicians and women is current evidence that the use of eSET will reduce overall live birth rates.8 Women in particular are deterred from taking up eSET, fearing that it will reduce their chances of live birth or result in a need for additional treatment cycles.9,10 Healthcare commissioners lack information on how alternative embryo transfer strategies will affect health service costs, pregnancy rates and twin pregnancy rates over time in different subgroups of women.7 To address these concerns, this study models the effectiveness and cost-effectiveness of using eSET versus DET in up to three full treatment cycles (i.e. three fresh cycles with associated frozen/thawed embryo transfers when needed) for women of different ages.

Methods A microsimulation economic model was developed to assess the cumulative costs and consequences of adopting either a DET policy or an eSET policy over multiple treatment cycles. Under the DET policy, modelled patients were allowed up to three fresh treatment cycles11 with DET followed by replacement of frozen/thawed embryos, two at

1074

a time, for those failing to achieve a live birth. Under the eSET policy, all women with at least two good-quality embryos available (i.e. eligible for eSET) received eSET followed by replacement of frozen/thawed embryos, one at a time, for those failing to achieve a live birth (Figure 1). Women with fewer than two good embryos available were modelled to receive DET or non-elective SET, depending upon the total number of embryos available. Treatment was discontinued when a simulated patient completed three full treatment cycles, delivered a surviving live birth, stopped treatment for other reasons, or reached the age of 45 years. We make the assumption that women would have the opportunity to complete a three-cycle programme after commencing treatment. An expert advisory group consisting of IVF specialists, an embryologist and a neonatologist was convened to inform the model structure and assumptions. The analysis was conducted from the health and social care perspective.

Clinical inputs All clinical inputs used in the model are presented in the Supporting Information (see Table S1). The total number of embryos available and the number of good-quality embryos available, for each simulated patient at each fresh cycle, were drawn from age- and cycle-specific probability distributions derived from a large clinical data set. The data set included 6153 women undergoing treatment at one of three Scottish IVF clinics, between January 1997 and June 2007 (10 511 fresh cycles; 3106 associated frozen/thawed cycles). Logistic regression models derived from a subsample of 4643 women undergoing their first fresh treatment cycle with DET were used to estimate the probability of having a positive pregnancy test, and the probability of singleton and twin clinical pregnancy for each simulated patient undergoing their first fresh DET cycle. Cycles involving donated gametes were excluded from the analysis. The mean age (SD) of women in the data set was 33.7 (4.1) years and mean (SD) BMI was 24.7 (4.0) kg/m2. Median (interquartile range) duration of infertility was 36 (24–53) months. Of the couples, 29.7% had tubal infertility, 6.1% had ovulatory problems, 13.3% had endometriosis, 45.2% had male factor problems, 22.8% had unexplained infertility and 1.2% had other causes of infertility recorded. The median (interquartile range) number of embryos available for transfer was five (three to eight), and 82.0% of couples had at least two good-quality embryos available for transfer in their first fresh cycle. Women’s age, number of embryos available and the quality of embryos transferred were included as the predictor variables in each logistic regression model. The clinical data set was also used to estimate age band-specific (£35, 36–39, 40–43, ‡43 years) probabilities for biochemical pregnancy, clinical pregnancy and twin


Cost-effectiveness of eSET versus DET

Treatment (eSET x3 or DET x3)

Pregnancy outcomes

Neonatal outcomes

Long-term outcomes

No pregnancy

First fresh treatment cycle

Subsequent fresh/frozen cycles

Miscarriage/still birth Neonatal death

Singleton live birth

Twin live birth Discontinued

NICU admission

Preterm birth

Term delivery

Healthy singleton infant Disabled singleton infant

Twins (both healthy) Twins (1 healthy/1 disabled) Twins (both disabled)

Figure 1. Schematic representation of the model structure.

pregnancy in frozen/thawed embryo transfer cycles following DET, as well as miscarriage rates for singleton and twin pregnancies. These age bands were chosen because they represent cutoffs at which marked changes in IVF success rates occur; this was based on the advice of the expert advisory group for the study. The probability of biochemical pregnancy following fresh eSET was estimated by applying a relative risk estimate, derived from a large randomised controlled trial,12 to the estimated probability of pregnancy following DET for each simulated patient. A relative risk estimate was also applied to the probability of pregnancy following frozen DET, to estimate the expected pregnancy rate with frozen SET.13 A twinning rate of 0.9% was applied to clinical pregnancies following SET.14 For pregnancies carried to 24 weeks, published data were used to estimate plurality-dependent probabilities for stillbirth, preterm birth, neonatal mortality,15 neonatal morbidity,16,17 and longer-term childhood complications including cerebral palsy, cognitive impairment and visual impairment.18 Probabilities for miscarriage, stillbirth, preterm birth, neonatal morbidity, neonatal mortality and longterm complications were adjusted upwards for monozygotic twins following eSET, using relative risk estimates derived from the literature.19–23 In addition, we adjusted the disability incidence upwards for twin survivors where the co-twin was stillborn or died during the neonatal period,

by applying an odds ratio of 2.424 to the incidence rate observed in surviving twin pairs.18 As neonatal event rates for twins are generally reported on a per twin infant basis, some assumptions were required regarding the concordance of adverse neonatal outcomes among liveborn twin pairs. For admissions to neonatal intensive care units, a 90% concordance rate was assumed (based on expert opinion); i.e. nine pairs of twins in every 20 twin infants admitted. For neonatal deaths and disability outcomes, concordance rates of 20.5% and 6.7% were applied respectively.25 In extrapolating long-term costs, we assumed that the modelled disability outcomes would have an adverse effect on life expectancy. For this reason, mortality rates reported for preterm infants with cerebral palsy were used to model survival in disabled infants,26,27 while standard UK life tables were used to model survival in infants without disability.

Linking cycles We assumed that all women failing to conceive after fresh DET, but with cryopreserved embryos, would attempt a frozen cycle. We further assumed that 100% of women failing to conceive following eSET would have at least one frozen embryo available and would therefore attempt a frozen/thawed cycle and that 80% of embryos would survive the freezing process. The impact of reducing embryo


1075

Scotland et al.

survival to 70% was assessed through sensitivity analysis. Finally, we applied observed age-specific probabilities of going on to subsequent frozen cycles following failed frozen cycles, and assumed that these probabilities would be the same following eSET and DET.28,29 Age-specific discontinuation rates following completed treatment cycles (i.e. fresh cycles with associated frozen/thawed cycles) were also estimated from the clinical data set and the same rates were applied for eSET and DET. Age-dependent probabilities of failing to reach embryo transfer were applied (Scottish clinical data set) in subsequent fresh cycles, with a proportion of these failures assumed to occur before egg recovery.30 Women experiencing cancellation before egg recovery could proceed with another fresh cycle, without the cancelled cycle counting as one of their three fresh attempts. Cancellations following egg recovery were counted as one of the three fresh attempts and these women were only eligible to proceed with associated frozen cycles or a final fresh cycle. For women with at least two embryos available in subsequent fresh cycles, the same approach as that in the first cycle was used to estimate singleton and twin clinical pregnancy rates following DET and eSET. However, pregnancy rates were adjusted downward (compared with the first cycle) to reflect the diminishing probability of success with increasing number of fresh cycles undertaken.31 Those women with only one embryo available in subsequent fresh cycles were modelled to receive non-elective SET and were assigned age-specific probabilities of pregnancy as estimated from the clinical data set.

Cost inputs All cost parameters used in the model are presented in the Supporting Information (see Table S2). Costs associated with IVF treatment, early follow up, ovarian hyperstimulation and antenatal and obstetric care were estimated using a detailed resource use data set for women undergoing their first fresh treatment cycle in Aberdeen.31 Costs associated with miscarriage, ectopic pregnancy and stillbirth were obtained from the National Tariff for the Department of Health, Health Care Resource Groups (version 3.5) (www.dh. gov.uk/en/Publicationsandstatistics/Publications/Publications PolicyAndGuidance/DH_082571). Costs associated with frozen/thawed cycles were taken from Dixon et al.32 For costs associated with neonatal admissions, we used the average length of stay reported for singletons and twins admitted to neonatal units in Scotland16 and applied an average per diem cost for admissions to neonatal units (including neonatal intensive care units and special care baby units) in Scotland (www.isdscotland.org/isd/3601. html). Additional autopsy and counselling costs associated with neonatal deaths were taken from a study by Petrou et al.33

1076

For health service costs in the first 2 years of life, we applied annual costs for singletons and twins reported by Henderson et al.34 but removed 80% of year one admission costs to avoid double counting; i.e. 80% of year one admissions would be expected to occur during the neonatal period,35 which had already been captured as described above. Following this, we applied annual health and social care costs appropriate to the long-term modelled health outcome of each infant, using data reported by Mangham et al.36 All costs were adjusted to a common base year (2007/08) using the Hospital and community health services inflation index (www.pssru.ac.uk/uc/uc2009contents. htm).

Utility weights Time spent by women in different states of the model was adjusted using quality (utility) weights reflecting the relative desirability of those states on a scale where zero represents death and one represents full health (see Table S3). This approach enabled the estimation of women’s qualityadjusted life-years (QALYs). A utility decrement associated with being ‘infertile with the desire for a child’ was derived from a US study37 in which this state was mapped to the preference weighted Health State Utilities Index Mark 2 (HUI2) classification system.38 By subtracting the resultant utility weight from the HUI2 US population norm,39 we obtained a utility decrement of 0.07 and subtracted it from Euroqol EQ-5D (EQ-5D) UK population norms in our model.40 We assumed that a successful healthy birth outcome would return women to the age-specific UK population norm, whereas giving birth to a child with a disability would be associated with an EQ-5D utility decrement of 0.07.41 For women suffering a stillbirth or neonatal/infant death followed by no subsequent pregnancy, we applied a utility weight of 0.543 elicited directly from infertile women in Grampian,42 because no general population preference weights were available for this outcome. A secondary analysis was conducted where women’s lifeyears were further adjusted using utility weights for IVF outcomes previously obtained from a group of women waiting to receive IVF treatment in Aberdeen.42 In this previous study, standard gamble utility weights were elicited for: ‘premature delivery/neonatal admission’; ‘giving birth to a child with disability’; ‘experiencing a neonatal death followed by no subsequent pregnancy’; and ‘treatment failure’ followed by life-long childlessness.

Analysis The analysis was conducted separately for three cohorts of women commencing treatment at different ages: 32, 36 and 39 years. The first age group (32 years) is representative of the mean age of women included in existing randomised



controlled trials of eSET, whereas 36 years is the age at which an observed marked decline in the IVF success rate begins.43 At 39 years, women face a more rapidly decreasing success rate and a limited time period in which to achieve a live-birth (www.hfea.gov.uk/fertility-treatmenttrends.html). It is also the oldest age at which women in the UK currently have access to NHS-funded treatment. Women’s QALYs were used to assess the cost–utility of the alternative strategies over a 20-year time horizon from the health and social care perspective. Future costs and QALYs were discounted at a rate of 3.5% per year (www. hm-treasury.gov.uk/green_book_guidance_discounting.htm). Monte Carlo simulation was used to simulate the movement of women through the model and, at the same time, characterise uncertainty in projected costs and consequences arising from input parameter uncertainty. The uncertainty surrounding each model input was characterised by assigning an appropriate probability distribution (Tables S1–S3). The model was then analysed by simulating the passage of 10 000 women through the model 1000 times, with a value for each parameter being drawn at random from its assigned distribution for each of the 1000 runs. This probabilistic approach gives an estimated distribution for each modelled output. These output distributions were used to generate 95% credible intervals (CI) for the differences in costs and outcomes between eSET and DET. Cost-effectiveness acceptability curves were used to present the probability of DET being cost-effective in comparison with eSET given different values of society’s willingness to pay for a QALY.

Deterministic sensitivity analysis Given the uncertainty surrounding the probability of women moving to cycles involving transfer of frozen/ thawed embryos, and the relative chances of success in these cycles following failed eSET and DET, deterministic sensitivity analysis was conducted to ascertain the influence of three key parameters: (i) the background probabilities of moving to a first frozen/thawed transfer following failed DET and eSET; (ii) the relative probability of moving to subsequent frozen/thawed cycles following failed frozen SET; and (iii) the relative probability of pregnancy following frozen SET compared with frozen DET. Further, we assessed the impact of changes to the twin pregnancy rate following eSET, to represent the possibility of less stringent implementation. The impact on findings of varying the size of the utility decrement associated with infertility was assessed, as was the impact of applying a temporary decrement for neonatal admissions that resulted in a good long-term outcome. Finally, we conducted an analysis to assess how cost-effectiveness would change if women were to be allowed access to publicly funded treatment aged 40 years and above.

Results The modelled clinical outcomes associated with the alternative strategies are presented in Table 1. The cumulative live birth rate is higher with DET than with eSET across all three cohorts. Elective SET is associated with a higher singleton live birth rate than DET in the 32- and 36-year-old cohorts, whereas the two strategies are almost equivalent with respect to this outcome in the 39-year-old cohort. There is a significantly lower twinning rate with eSET across all ages, and consequently fewer stillbirths, neonatal deaths and infants affected by long-term disability. The twinning rate associated with DET also appears to decrease with age. The health service cost associated with both eSET and DET declines with age because of the decreasing probability of pregnancy and live birth. Elective SET is less costly than DET in all three cohorts, although the additional cost associated with DET also tends to decrease as women’s age increases. This is because DET is associated with a lower twin pregnancy rate in older women, which results in a lower relative increase in the cost per additional live birth with DET. From the data presented in Table 1, the incremental cost per live birth for DET versus eSET comes to £27 356, £18 580 and £15 539 in the 32-, 36- and 39-year-old cohorts, respectively. Considering the QALYs accruing to women (Table 2) DET is associated with a greater number of QALYs in all three cohorts and the incremental cost per additional QALY associated with DET also decreases with women’s age. Figure 2 indicates the probability of DET being considered cost-effective for women of different ages, given different values of society’s willingness to pay for a QALY.

Deterministic sensitivity analysis The findings were most sensitive to the changes in the probabilities of moving to subsequent frozen/thawed treatment cycles following failed eSET and DET, and the size of the utility decrement associated with infertility. Doubling the probability of moving on to subsequent frozen/thawed cycles in the eSET arm, relative to the observed probability of moving to subsequent frozen/ thawed cycles following DET, slightly reduced the costeffectiveness of DET in 32-year-old women [incremental cost-effectiveness ratio (ICER): £29 500 per additional QALY], but had limited impact on the overall findings in older women. Changing the relative probability of pregnancy with frozen/thawed SET compared with frozen/thawed DET (from 0.615 to 0.70), slightly reduced the cost-effectiveness of DET in women of all ages (ICERs £32 300, £25 100, and


1077

1078

2.9 (0.8 to 5.0) 8.0

4.0 (1.5 to 6.7)

2.9

5.8

3.4 (1.1 to 5.8) 7.7 4.3 4.6 (1.7 to 7.7)

4.0

10.5 6.0

Stillbirths and neonatal deaths per 1000 women treated

5.0

10.6

5.6 (2.6 to 8.8)

6.5 (3.0 to 10.3) 7.5 No. of infants with long-term disability per 1000 women treated

14.0

17.2 (14.5 to 20.0) 19.1 1.9 21.1 (18.7 to 23.3) 23.4 2.3 25.1 (23.5 to 26.8) 2.5 Twin live birth rate (as % age of all live births)

27.6

0.0 ()3.3 to 3.3) 28.7 28.7 )5.0 ()8.5 to )1.7) 39.3 48.9 Singleton live birth rate (%)

40.2

)8.7 ()12.7 to )4.9)

34.3

4.3 (1.2 to 7.3) 30.9 26.5 2.2 ()1.2 to 5.4) 38.6 36.4 1.4 ()2.3 to 4.8) 45.4 Term live birth rate (%)

46.8

7.6 (4.1 to 11.0) 37.1 29.4 6.9 (3.2 to 10.4) 47.4 40.5 8.1 (4.2 to 11.7) 58.5 50.4 Live birth rate (%)

£1181 (802 to 1599) £10 390 £9209 £1282 (878 to 1679) £11 732 £10 451 £2215 (1760 to 2624) £13 405 £11 190 Health service costs (mean)

Diff. (95% CI) DET Diff. (95% CI) DET eSET

DET

Diff. (95% CI)

eSET

36-year-old women 32-year-old women

Table 1. Cumulative costs and outcomes following up to three fresh treatment cycles with eSET or DET (with associated frozen cycles)

eSET

39-year-old women

Scotland et al.

£21 600 per QALY in the 32-, 36- and 39-year-old cohorts respectively). Applying the assumption that 70% (as opposed to 80%) of frozen embryos survive the freezing process, slightly improved the cost-effectiveness of DET in all age groups. Varying the twin pregnancy rate following eSET, to reflect potential variation in implementation, had a substantial impact on cost-effectiveness of the alternative strategies. For example, increasing the twin pregnancy rate to 5% following planned eSET reduced the ICER for DET versus eSET to £23 900 per QALY in 32-year-old women, £18 500 per QALY in 36-year-old women, and £17 300 per QALY in 39-year-old women. Raising the twin pregnancy rate following planned eSET to 10% increased costeffectiveness of DET even further (e.g. to £20 100 per extra QALY in 32-year-old women). However, caution is required when interpreting these results as we could not ascertain how imperfect implementation of eSET would affect the live birth rate. Reducing the size of the utility decrement associated with infertility by 25% substantially reduced the costeffectiveness of DET (ICERs £41 300, £35 100 and £32 900 per QALY in the 32-, 36- and 39-year-old cohorts, respectively). Applying a large but temporary utility decrement of 0.5 over the duration of the neonatal period when neonates were admitted, to reflect the potential stress associated with this event, had little impact on the results. We also assessed the impact of assuming that women’s quality of life would return to baseline levels over the 2 years following a stillbirth or neonatal death. Given the low absolute probability of this event, this change also had little impact on findings. When women’s life-years were quality adjusted using standard gamble utility weights previously elicited from women waiting to undergo IVF treatment in Aberdeen,42 the apparent costeffectiveness of DET improved substantially in women of all ages; the ICERs for DET versus eSET being £9400, £7300 and £5700 per additional QALY in the 32-, 36and 39-year-old cohorts respectively. Finally, raising the starting age of women to 40 resulted in the health service costs being marginally lower with DET, whereas the live birth rate remained marginally higher; the mean difference (95% CI) in cost and live birth rate coming to –£126 ()620 to 487) and +2.4% ()0.8 to 5.6), respectively. These differences were not significant in the 40-year-old cohort because of the diminishing pregnancy rates and higher discontinuation rates limiting the potential for DET to exert its per cycle superiority. In terms of QALYs, there was very little to choose between the strategies; the mean difference (95% CI) for DET versus eSET equating to )0.01 QALYs ()0.057 to 0.037). The twin pregnancy rate associated with DET in 40-year-old women was 12.1%.



Table 2. Incremental cost–utility of DET versus eSET by age of women when they start treatment Strategy (age group)

32 years eSET DET 36 years eSET DET 39 years eSET DET

Mean health service costs

Difference in health service costs

Women’s QALYs

Difference in women’s QALYs

Additional cost per additional QALY (Health service perspective)

£11 190 £13 405

* £2215 (1760–2624)

13.226 13.304

* 0.078 (0.027–0.133)

* £28 263

£10 451 £11 732

* £1282 (878–1679)

13.077 13.136

* 0.059 (0.012–0.110)

* £21 722

£9209 £10 390

* £1181 (802–1599)

12.933 12.991

* 0.058 (0.012–0.108)

* £20 278

*Comparison group.

Strengths and weaknesses

1

Probability cost-effective

0.9 0.8 0.7 0.6 0.5

39 years

0.4

36 years

0.3

32 years

0.2 0.1 0 £0

£10,000 £20,000 £30,000 £40,000 Willingness to pay per QALY

£50,000

Figure 2. Cost-effectiveness acceptability curves for DET versus eSET by age at which women commence treatment.

Discussion A strategy of elective DET was associated with not only a higher cumulative live birth rate compared with eSET across women of all ages, but also a higher rate of twin pregnancy and associated complications (stillbirths, neonatal deaths and adverse child health outcomes), particularly in younger women. From the health service perspective, the DET strategy varied from costing an additional £27 356 per extra live birth in 32-year-old women to an extra £15 539 per extra live birth in 39-year-old women. Compared with eSET, DET cost £28 300 per additional QALY for women commencing treatment aged 32 years, £21 700 per extra QALY for women commencing treatment aged 36 years, and £20 300 per extra QALY in women starting treatment aged 39 years.

The greatest challenge in assisted reproduction is to formulate effective and cost-effective strategies of embryo transfer such that the risk of multiples is minimised without sacrificing success rates. This is the first study to assess the cumulative costs and consequences of using alternative embryo transfer strategies over multiple treatment cycles for women of different ages. The model structure was developed through consultation with a panel of experts, and we used a large observational data set of contemporaneous data from three of four teaching-hospital-based IVF centres in Scotland, combined with data from randomised controlled trials, to estimate singleton and twin pregnancy rates following eSET and DET. To our knowledge this is also the first study to estimate QALYs associated with alternative approaches to assisted reproduction. It has been argued that QALYs are unsuitable for comparison of assisted reproductive technologies, on the grounds that they are not geared towards capturing the benefit of creating life, but rather the benefit of improving health.44 However, if we view infertility as ‘a disruption to normal function that limits an individual’s opportunity to live a good life’,45 and assisted reproductive technologies (ART) as interventions that couples access to overcome their infertility and improve their quality of life, then the benefits of ART to prospective patients can in theory be measured using QALYs. We chose to concentrate on women’s QALYs rather than the expected QALYs for both partners, as women are the main recipient of IVF treatment. This approach can be seen as a way of putting ART on an equal footing with chronic diseases; economic evaluations of other medical interventions for chronic, noninfectious diseases generally assess the health benefits accruing to the one main recipient of treatment, rather than the


1079

Scotland et al.

health improvements accruing to all individuals affected by the decision to treat. As relatively few eSET procedures have been performed in the UK until recently,43 we were unable to estimate the probability of pregnancy following eSET directly. To overcome this problem we estimated pregnancy rates following eSET by applying a constant relative reduction to the pregnancy rate that would be expected with DET. To ensure the applicability of this relative reduction to our modelled cohorts, we applied the same selection criteria for eSET as was used in the randomised controlled trials from which the multiplier was derived.46 This approach yielded live birth rates following eSET that were congruent with those observed in other randomised trials.6 A further weakness was that we were unable to assess relative cost-effectiveness of transferring blastocyst-stage embryos as opposed to cleavage-stage embryos. This is a question that we will be able to address in the future, as this practice gains wider clinical acceptance. Finally, the process of modelling multiple fresh and frozen cycles over a prolonged time horizon, introduced an additional degree of uncertainty into the analysis. This was the result of assumptions being required regarding the probability of women moving to frozen/thawed cycles, and subsequent fresh treatment cycles, following failed eSET and DET.

Comparison with other studies A recent study47 simulated cumulative outcomes from a single complete cycle with eSET (with frozen embryos thawed one at a time) and a single complete cycle with DET (with frozen embryos thawed two at time), using statistical models developed from a large observational data set. Roberts et al.47 found that DET produced a higher expected live birth rate than eSET (30.8% versus 27.8%) when success rates for frozen embryos matched those observed in their clinical data set, but their analysis suggested that eSET might outperform DET if outcomes following frozen embryo transfer were to improve. Although differences in the structure and assumptions of our model prevent a direct comparison of predicted success rates, we found that DET retained a higher probability of live birth compared with eSET even when we increased the relative probability of success in frozen single embryo transfers. This apparent difference in findings is probably because we applied observed cryopreservation rates and discontinuation rates following failed frozen cycles, rather than the assumption that women freeze and use all goodquality embryos associated with a single egg recovery before proceeding to a new cycle. Our findings are broadly consistent with those of previous economic evaluations that found a single fresh DET to be more costly and marginally more effective than a

1080

single cycle with eSET followed by replacement of a single frozen/thawed embryo in those who failed to becme pregnant.48 In the only other economic evaluation of eSET that took women’s age into consideration, Dixon et al.32 compared the costs and consequences of one fresh DET and one fresh eSET (followed by one frozen/thawed SET) and also found that the cost-effectiveness of DET increases with age. Previous attempts to model multiple treatment cycles over time29,49 came to the conclusion that DET would generate a higher live birth rate at higher cost to the health service and society. However, our study suggests that the additional health service costs associated with DET decrease as women get older. Further comparisons between this and previous studies are hampered by the fact that previous studies did not assess cost-effectiveness by women’s age or only considered costs incurred up to 6 weeks after delivery.29,49

Interpretation This study suggests that a DET strategy would produce not only a higher live birth rate than eSET for women of all ages but also a higher twin pregnancy rate. The difference in the twin pregancy rates between the strategies decreases as women’s age increases, and as a consequence the additional health service costs associated with DET also decrease as women age. It is the higher twinning rate that drives the higher cost increases associated with DET in younger women. The preferred strategy for women commencing treatment at different ages depends upon how much society is willing to pay for an additional live birth, coupled with concerns about the acceptability of the observed twin pregnancy and complication rates. No guidance exists on how much society should be willing to pay for a live birth so we estimated women’s QALYs associated with the alternative strategies. The additional cost per extra QALY was found to decrease with women’s age because of the lower probability of twin pregnancy and associated complications with DET in older women. In the UK, interventions that cost below £20 000 to £30 000 per additional QALY are often considered cost-effective.50 Applying this threshold range, and applying the primary utility assumptions, DET would be more likely to be considered cost-effective for women commencing treatment aged 36 years or over, and less likely to be considered costeffective in younger women (Figure 2). However, limited availability of population preference weights for the different IVF treatment/outcome scenarios makes it difficult to draw firm conclusions from the QALY analysis. The incremental cost per QALY was found to be highly sensitive to the size of the quality of life decrement associated with infertility. Reducing this decrement from



0.07 to 0.05 resulted in DET costing over £30 000 per additional QALY in women of all ages. Cost–utility was also sensitive to the source of the preference values. Application of preference weights previously elicited from women waiting to undergo IVF treatment in Aberdeen,42 resulted in a more favourable incremental cost per QALY ratio for DET across women of all ages. The driver behind this change was the low utility weight that infertile women assigned to the outcome of being infertile following treatment failure (childlessness for life). This is consistent with the findings of previous studies suggesting that many women have a clear preference for DET despite being aware of the risks associated with twin pregnancy.9,10

members of the Advisory Group on Multiple Births to the HFEA.

Contribution to authorship SB, GS and KH were responsible for the original idea and developed the study protocol and funding application with JJK, MR and HL. DM advised on statistical aspects of the study design and undertook the statistical analyses underpinning the cost-effectiveness model. PM and MH advised on the design and analysis of the cost-effectiveness models. All the authors were involved in the discussion and interpretation of results. GS conducted the economic modelling and drafted the paper. All authors contributed to the writing of further drafts. GS is the guarantor for the study.

Policy implications The HFEA in the UK has recently introduced a twin pregnancy target of no more than 10% of all IVF births, with which clinics are obliged to comply. Given the very high twin pregancy rates expected with DET in 32- and 36-yearold women, coupled with higher health service costs and uncertainty relating to the expected QALY gains, eSET is likely to be the favoured strategy in these women. However, with the twinning rate associated with DET decreasing and the difference in health service costs between the strategies becoming smaller as women age, DET may be considered cost-effective in some groups of older women. For women aged 37–39 years, the decision over embryo transfer may best be considered on a case-by-case basis, though our model suggests that DET will have a higher likelihood of being considered cost-effective for older women within in this age range. If women were allowed to commence publicly funded treatment aged 40 years, our model suggests that while eSET and DET may be more or less equivalent in terms of costs to the health service, DET may produce a marginally higher live birth rate (though not significant at the 5% level). This suggests DET may offer the most efficient approach to IVF in women aged 40 years and over, should NHS-funded treatment be made available to these women.

Details of ethics approval

Conclusions

The following supplementary materials are available for this article: Table S1. Model parameters, (clinical inputs). Table S2. Model parameters, (costs). Table S3. Model parameters, (utility values and decrements). Additional Supporting Information may be found in the online version of this article. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author. j

Given the high twin pregnancy rates associated with DET in women aged 36 years and under, coupled with increased costs to the health service for uncertain QALY gains, a selective eSET policy is likely to be the preferred option for women in this age group. As women get older, the likelihood of DET being considered cost-effective from the health service perspective increases.

Disclosure of interests One of the authors (MH) is Chair of the British Fertility Association. Two of the authors (MH and SB) are

The North of Scotland Research Ethics Committee confirmed that a formal ethics application was not required for this study based on anonymised routine data and published literature.

Funding This study was funded by the Chief Scientist Office of the Scottish Governments Health Directorates (SGHD). The views expressed here are those of the authors and not necessarily those of the SGHD.

Acknowledgements The authors wish to thank the members of the advisory group for the project: Dr Anthony Harrold (Consultant Gynaecologist, NHS Tayside), Dr Maybeth Jamieson (Consultant Embryologist, Assisted Conception Unit Glasgow), Dr Mark Hamilton (Consultant and Subspecialist in Reproductive Medicine, Aberdeen Fertility Centre), Dr Clem Tay (Consultant and Subspecialist in Reproductive Medicine, Simpson Fertility Centre) and Dr Sumesh Thomas (Consultant Neonatologist, Aberdeen Maternity Hospital).

Supporting information


1081

Scotland et al.

References 1 ESHRE Campus Course Report. Prevention of twin pregnancy after IVF/ICSI by single embryo transfer. Hum Reprod 2001;16: 790–800. 2 De Mouzon J, Goossens V, Bhattacharya S, Castilla JA, Ferraretti AP, Jorsak V, et al. Assisted reproductive technology and intrauterine inseminations in Europe, 2006: results generated from European registers by ESHRE. Hum Reprod 2010;25:1851–62. 3 Bortolus R, Parazzini F, Chatenoud L, Benzi G, Bianchi MM, Marini A. The epidemiology of multiple births. Hum Reprod Update 1999;5: 179–87. 4 Blondel B, Kaminski M. Trends in the occurrence, determinants, and consequences of multiple births. Semin Perinatol 2002;26:239–49. 5 Ledger WL, Anumba D, Marlow N, Thomas CM, Wilson EC. Cost of Multiple Births Study Group (COMBS Group). The costs to the NHS of multiple births after IVF treatment in the UK. BJOG 2006;113: 21–5. 6 Pandian Z, Templeton A, Serour G, Bhattacharya S. Number of embryos for transfer after IVF and ICSI: a Cochrane review. Hum Reprod 2005;20:2681–7. 7 El-Toukhy T, Khalaf Y, Braude P. IVF results: optimize not maximize. Am J Obstet Gynecol 2006;194:322–31. 8 Roberts SA, Fitzgerald CT, Brison DR. Modelling the impact of single embryo transfer in a national health service IVF programme. Hum Reprod 2009;24:122–31. 9 Newton CR, McBride J, Feyles V, Tekpetey F, Power S. Factors affecting patients’ attitudes toward single and multiple embryo transfer. Fertil Steril 2007;87:269–78. 10 Blennborn M, Nilsson S, Hillervik C, Hellberg D. The couple’s decision making in IVF: one or two embryos at transfer? Hum Reprod 2005;20:1292–7. 11 National Collaborating Centre for Women’s and Children’s Health on behalf of the National Institute for Clinical Excellence. Fertility: assessment and treatment for people with fertility problems. Clinical Guideline. Royal College of Obstetricians and Gynaecologists, London: RCOG Press, 2004. 12 Thurin A, Hausken J, Hillensjo¨ T, Jablonowska B, Pinborg A, Strandell A, et al. Elective single embryo transfer versus double embryo transfer in in vitro fertilization. N Engl J Med 2004;351:2392–402. 13 Le Lannou D, Griveau J-F, Laurent M-C, Gueho A, Veron E, Morcel K. Contribution of cryopreservation to elective single embryo transfer in IVF-ICSI. Reprod BioMed Online 2006;13:368–75. 14 Vitthala S, Gelbaya TA, Brison DR, Fitzgerald CT, Nardo LG. The risk of monozygotic twins after assisted reproductive technology: a systematic review and meta-analysis. Hum Rep Update 2009;15:45–55. 15 Pinborg A, Lidegaard Ø, La Cour Freiesleben N, Anderson AN. Consequences of vanishing twins in IVF/ICSI pregnancies. Hum Reprod 2005;20:2821–9. 16 Penny G. Scottish Perinatal and Infant Mortality and Morbidity Report 2001. Edinburgh: Information & Statistics Division, NHS Scotland, 2002. 17 Helmerhorst FM, Perquin DAM, Donker D, Keirse MJNC. Perinatal outcome of singletons and twins after assisted conception: a systematic review of controlled studies. BMJ 2004;328:261. 18 Stro¨mberg B, Dahlquist G, Ericson A, Finnstro¨m O, Ko¨ster M, Stjernqvist K. Neurological sequelae in children born after in-vitro fertilisation: a population-based study. Lancet 2002;359:461–5. 19 Sperling L, Kiil C, Larsen LU, Qvist I, Schwartz M, Jørgensen C, et al. Naturally conceived twins with monochorionic placentation have the highest risk of fetal loss. Ultrasound Obst Gyn 2006;28:644–52. 20 Pharaoh POD. Neurological outcome in twins. Semin Perinatol 2002;7:223–30.

1082

21 Hack KEA, Derks JB, Elias SG, Franx A, Roos EJ, Voerman SK, et al. Increased perinatal mortality and morbidity in monochorionic versus dichorionic twin pregnancies: clinical implications of a large Dutch cohort study. BJOG 2008;115:58–67. 22 Petterson B, Stanley F, Henderson D. Cerebral palsy in multiple births in Western Australia: genetic aspects. Am J Med Genet 1990;37: 346–51. 23 Glinianaia SV, Pharoah POD, Wright C, Rankin JM. Fetal and infant death in twin pregnancy: neurodevelopmental consequences for the survivor. Arch Dis Child Fetal Neonatal Ed 2002;86:F9–15. 24 Scher AI, Petterson B, Blair E, Ellenberg JA, Grether JK, Haan E, et al. The risk of mortality or cerebral palsy in twins: a collaborative population based study. Pediatr Res 2002;52:671–81. 25 Pinborg A, Loft A, Schmidt L, Greisen G, Rasmussen S, Anderson AN. Neurological sequelae in twins born after assisted conception: controlled national cohort study. BMJ 2004;329:311. 26 Hutton JL, Phaoah POD, Rosenbloom L. Effects of cognitive, motor and sensory disabilities on survival in cerebral palsy. Arch Dis Child 2002;86:84–9. 27 Strauss D, Brooks J, Rosenbloom L, Shavelle R. Life expectancy in cerebral palsy: an update. Dev Med Child Neurol 2008;50:487–93. 28 van Montfoort P, Fiddelers AA, Janssen JM, Derhaag JG, Dirksen CD, Dunselman GAJ, et al. In unselected patients, elective single embryo transfer prevents all multiples, but results in significantly lower pregnancy rates compared with double embryo transfer: a randomized controlled trial. Hum Reprod 2006;21:338–43. 29 Fiddelers AAA, Dirksen CD, Dumoulin JCM, van Montfoort APA, Land JA, Janssen JM, et al. Cost-effectiveness of seven IVF strategies: results of a Markov decision-analytic model. Hum Reprod 2009;24:1648–55. 30 Maheshwari A, Scotland G, McTavish A, Bell J, Hamilton M, Bhattacharya S. Direct health services costs of providing assisted reproduction services in older women-Retrospective cross-sectional analysis. Fertil Steril 2010;93:527–36. 31 Templeton A, Morris JK, Parslow W. Factors that affect outcome of in-vitro fertilisation treatment. Lancet 1996;348:1402–6. 32 Dixon S, Nasiri FF, Ledger WL, Lenton EA, Duenas A, Sutcliff P, et al. Cost-effectiveness analysis of different embryo transfer strategies in England. BJOG 2008;115:758–66. 33 Petrou S, Bischof M, Bennet C, Elbourne D, Field D. Cost-effectiveness of neonatal extracorporeal membrane oxygenation based on 7year results from the United Kingdom Collaborative EMCO trial. Pediatrics 2006;117:1640–9. 34 Henderson J, Hockley C, Petrou S, Goldacre M, Davidson L. Economic implications of multiple births: inpatient hospital costs in the first 5 years of life. Arch Dis Child Fetal Neonatal Ed 2004;89:F542–5. 35 Pinborg A, Loft A, Rasmussn S, Anderson AN. Hospital care utilisation of IVF/ICSI twins until 2–7 years of age: a controlled Danish national cohort study. Hum Reprod 2004;19:2529–36. 36 Mangham LJ, Petrou S, Doyle LW, Draper ES, Marlow N. The cost of preterm birth throughout England and Wales. Pediatrics 2009;123: e312–27. 37 Institute of Medicine. Neisseria Gonorrhea. In: Stratton KR, Durch JS, Lawrence RS, editors. Vaccines for the 21st Century: A Tool for Decision Making. Washington, DC: National Academy Press; 2000. pp. 257–65. 38 Torrance GW, Feeny DH, Furlong WJ, Barr RD, Zhang Y, Wang Q. Multi-attribute preference functions for a comprehensive health status classification system: Health Utilities Index Mark 2. Med Care 1996;34:702–22. 39 Luo N, Johnson JA, Shaw JW, Feeny D, Coons SJ. Self-reported health status of the general adult US population as assessed by EQ-5D and Health Utilities Index. Med Care 2005;43:1078–86.



40 Kind P, Hardman G, Macran S. UK Population Norms for EQ-5D. Discussion Paper 172. York: Centre for Health Economics, University of York; 1999. 41 Kuhlthau K, Kahn R, Hill KS, Gnanasekaran S, Ettner SL. The wellbeing of parental caregivers of children with activity limitations. Matern Child Health J 2010;14:155–63. 42 Scotland GS, McNamee P, Peddie VL, Bhattacharya S. Safety versus success in elective single embryo transfer: women’s preferences for outcomes of in vitro fertilisation. BJOG 2007;114:977–83. 43 Human Fertilisation and Embryology Authority.A long term analysis of the HFEA Register data (1991–2006). Human Fertilisation and Embryology Authority 2008; [www.hfea.gov.uk]. Last accessed 17 March 2011. 44 Devlin N, Parkin D. Funding fertility: issues in the allocation and distribution of resources to assisted reproduction technologies. Hum Fertil 2003;6:S2–6. 45 McMillan J. Allocating fertility services by medical need. Hum Fertil 2001;4:11–3.

46 Kjellberg AT, Carlsson P, Bergh C. Randomised single versus double embryo transfer: obstetric and paediatric outcome and a cost-effectiveness analysis. Hum Reprod 2006;21:210–6. 47 Roberts SA, McGowan L, Hirst WM, Brison DR, Vail A, Lieberman BA. Towards single embryo transfer? Modelling clinical outcomes of potential treatment choices using multiple data sources: predictive models and patient perspectives. Health Technol Assess 2010;14: 1–237. 48 Fiddelers AAA, Severens JL, Dirksen CD, Dumoulin JCM, Land JA, Evers JLH. Economic evaluations of double- versus single- embryo transfer in IVF. Hum Reprod Update 2007;13:5–13. 49 De Sutter P, Gerris J, Dhont M. A health-economic decision-analytic model comparing double with single embryo transfer in IVF/ICSI. Hum Reprod 2002;17:2891–6. 50 Appleby J, Devlin N, Parkin D. NICE’s cost effectiveness threshold. How high should it be?. BMJ 2007;335:358–9.


1083