Evaluation of the relative cost-effectiveness of treatments for infertility in the UK

Z. Philips1,2,3, M. Barraza-Llorens2 and J. Posnett2

1 Trent Institute for Health Services Research, University of Nottingham, Nottingham, NG7 2UH, 2 York Health Economics Consortium, University of York, Heslington, York, YO10 5DD, UK


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
This paper aims to complement existing clinical guidelines by providing evidence of the relative cost-effectiveness of treatments for infertility in the UK. A series of decision–analytical models have been developed to reflect current diagnostic and treatment pathways for the five main causes of infertility. Data to populate the models are derived from a systematic review and routine National Health Service activity data, and are augmented with expert opinion. Costs are derived from an analysis of extra-contractual referral tariffs and private sector data. Sensitivity analysis has been carried out to take account of the uncertainty of model parameters and to allow results to be interpreted in the light of local circumstances. Results of the modelling exercise suggest in-vitro fertilization is the most cost-effective treatment option for severe tubal factors and endometriosis, with surgery the most cost-effective in the case of mild or moderate disease. Ovulatory factors should be treated medically with the addition of laparoscopic ovarian diathermy in the presence of polycystic ovarian syndrome. For other causes, stimulated intrauterine insemination (unexplained and moderate male factor) and stimulated donor intrauterine insemination (severe male) are cost-effective.

Key words: assisted conception/cost-effectiveness/decision analysis/infertility/modelling


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The range of available treatments for infertility has increased substantially in the past 20 years with the development of assisted conception techniques. Such techniques have received public attention and have fuelled considerable debate, not least because of the high cost associated with these treatments. Evidence of the clinical effectiveness of treatments was first reviewed in 1992 [Effective Health Care (EHC), 1992], and more recently the Royal College of Obstetricians and Gynaecologists (RCOG) has produced UK guidelines for clinical diagnosis and management (RCOG, 1998aGo,bGo). However, in the evolving culture of the new National Health Service (NHS), evidence of clinical effectiveness on its own is no longer sufficient to inform the commissioning process. Evidence is also required on the relative cost-effectiveness of alternative treatments.

There is already published evidence relating to the costs and effectiveness of infertility treatments, but most studies focus on in-vitro fertilization (IVF) or assisted reproductive techniques. A high proportion of studies are based on USA or Canadian data, and calculate costs according to insurance charges rather than the true cost of treatment. Insurance-based charges rely on a top-down approach to costing and are presented at an aggregate level. Furthermore, they cannot be applied to the UK NHS, not least because of the different incentives within insurance-based and publicly funded health care systems.

The few studies that examine other treatments for infertility (in addition to IVF) are based on cohort studies. Van Voorhis et al. conclude that intrauterine insemination [with clomiphene citrate, human menopausal gonadotrophin (HMG), or alone] should be offered before assisted reproductive techniques in the absence of tubal disease, and that IVF is a cost-effective treatment for women with blocked Fallopian tubes (Van Voorhis et al., 1997Go). Peterson et al. also suggest that intrauterine insemination with HMG should be offered as a first-line treatment because of its lower costs and equivalent success rates to IVF. However, the authors comment that such results may have occurred because of limited sample size, and they stress the need for further work to identify appropriate subsets for particular treatments (Peterson et al., 1994Go). Although it is recognized that random allocation to alternative treatments is unlikely in the case of infertility, it is important to remember the caveats to cohort analyses. These studies are likely to include heterogeneous couples receiving alternative treatments. Such case-mix differences, in particular those differences that occur because of differential causes of infertility or severity of disease, make it difficult to compare different treatments in order to derive cost-effectiveness estimates. There have been studies comparing IVF with tubal surgery, but the results are conflicting. One study (Holst et al., 1991Go) concluded that IVF was more cost-effective than surgery, yet another (Haan and Van Steen, 1992Go) could find no discernible difference between the two alternatives. Such inconclusive results warrant further investigation, in particular research that distinguishes between the severity of disease. We have been unable to identify any studies that describe the severity of disease as well as the cause of the infertility as independent predictors of both cost and outcome. This is disappointing in view of the evidence (Hull and Fleming, 1995Go; Hull, 1996Go; Collins et al., 1997Go).

A recent review of the cost-effectiveness of treatments for infertility concluded that further work is still required to evaluate the cost-effectiveness of treatments among different diagnostic groups (Van Voorhis et al., 1998Go). To our knowledge, there are no UK-based studies that address this question.

The aim of this work is to complement existing UK guidelines by providing evidence of the relative costeffectiveness of treatment alternatives for the main causes of infertility.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
A series of decision-analytical models has been developed to reflect current diagnostic and treatment pathways for couples presenting with symptoms of infertility (failure to conceive after 2 years during which there has been sexual intercourse and no use of contraception, (EHC, 1992) in the UK. The aim of these models is to evaluate the relative cost-effectiveness of treatment alternatives for each of the main causes of infertility (ovulatory, tubal and male factors, endometriosis and unexplained (EHC, 1992), by estimating expected costs and outcomes for a cohort of couples treated with the best available treatment alternatives. All analyses of the cost-effectiveness comparisons include the expected costs of a diagnostic work-up, although the cost-effectiveness of alternative diagnostic procedures has not been evaluated explicitly. Our approach has been to develop models that are of sufficient generality to allow them to cope easily with changes in baseline assumptions and to be adapted to include additional diagnostic or treatment options as they become available.

Diagnostic model
The diagnostic process is modelled as a decision tree (Figure 1Go). The aims of this part of the model are to estimate the expected costs of diagnosis for a cohort of couples presenting for the first time, and to estimate the expected breakdown of the cohort into groups defined by suspected principal cause. The diagnostic pathway is consistent with that recommended in recent RCOG guidance (RCOG, 1998aGo) and includes the most commonly used investigations which have been shown in prospective studies to have the ability to discriminate outcomes (British Fertility Society, 1995Go; Van den Eede, 1995Go; Hull, 1996Go; Snick et al., 1997Go).



View larger version (15K):
[in this window]
[in a new window]
 
Figure 1. Simplified diagnostic model. GP = general practitioner.

 
The Department of Health (DoH) recommends that every effort should be made to identify and immunize seronegative women (DoH, 1996Go). Women who present to the general practitioner (GP) with infertility problems offer a unique opportunity for rubella screening. Therefore, in accordance with guidelines (RCOG, 1998aGo) we have assumed the GP will check the rubella status of the woman and offer rubella vaccination if results are seronegative. The simple investigations undergone by the couple include a detailed history of both partners, their coital history, and examinations for previous abnormalities such as sexually transmitted diseases (STD), previous surgeries, pelvic inflammatory disease (PID), or significant systemic illness. The female partner will have her menstrual cycle assessed with serum progesterone measurements taken on day 21 of her cycle. Follicle stimulating hormone (FSH) and luteinizing hormone (LH) concentrations will also be assessed. Male partners will undergo two semen analyses. The model assumes that all couples who do not discontinue investigations at this early stage will undergo tubal patency tests with either a hysterosalpingography or laparoscopy or both and undergo a chlamydia serology. The results of the initial investigations, in addition to the tubal patency screening, will establish whether there are any leading features that might suggest a specific cause of infertility. Where a leading feature is identified, further tests may be offered. In the case of suspected ovulatory factor, these include assessment of prolactin concentrations, vaginal or pelvic ultrasound, and magnetic resonance imaging (MRI) scans. For suspected principal male causes, tests will include endocrinology, testicular biopsy, screening for cystic fibrosis, karotyping, and further semen microscopy. The costs and likelihood of all tests are used to derive an expected cost of diagnosis.

The analysis relates to couples in which the woman is under 40 years of age. It is assumed that all couples undergo a full range of recommended diagnostic procedures, and in this sense the diagnostic model applies only to couples presenting for the first time. Where this is not the case estimated costs of diagnosis may be overstated. It is not our purpose to evaluate the cost-effectiveness of alternative diagnostic techniques. An expected diagnostic work-up cost is derived, which is added to overall treatment costs.

Treatment models
The end-points in the diagnostic model represent the points at which the cause of infertility is determined and a decision is to be made about appropriate treatment. A separate treatment model has been developed for each of the five main causes, with treatment alternatives that are relevant to cause and severity. Multiple factors are classified by main cause and are not modelled separately. Treatment alternatives have been selected on the basis of known effectiveness (EHC, 1992Go; Barraza-Llorens, 1998Go) and are consistent with the recent recommendations of the Royal College of Obstetricians and Gynaecologists (RCOG, 1998bGo). In the treatment models, no assumptions are made about previous history. In particular, the cost-effectiveness analysis is not restricted to couples with primary infertility.

Treatment models are constructed in the form of a decision tree with embedded Markov processes (Figure 2Go). The Markov process assumes that couples will always be in one of a finite number of states at the end of each cycle of treatment. The length of a cycle and the number of cycles permitted vary according to the alternatives being compared. Couples move between states according to a set of transition probabilities.



View larger version (14K):
[in this window]
[in a new window]
 
Figure 2. Treatment model structure: example of tubal factor infertility. SIVF = stimulated in-vitro fertilization.

 
Transition probabilities
A probability matrix has been constructed for each of the five main causes and for each of the modelled treatment alternatives. The key probabilities in each model are the probability of clinical pregnancy (per cycle of treatment), the probability that treatment is stopped in mid-cycle, the probability of the couple's withdrawing before the end of the course of treatment, and the probability of complications [such as ovarian hyperstimulation syndrome (OHSS)].

The most appropriate way to determine the effectiveness of a particular treatment is from evidence derived from a prospective randomized controlled trial. However, for treatments for infertility, these types of studies are rare (Peterson et al., 1994Go). We were unable to identify sufficient trial data for all the parameters required by the model. For this reason, baseline probabilities and frequencies have been derived from a combination of expert judgement, routine NHS activity data, and evidence from the literature. The results of a systematic review of literature published in 1992 (EHC, 1992Go) have been supplemented by a systematic review of English-language literature published since 1992 on the effectiveness of treatments for infertility (Barraza-Llorens, 1998Go). Baseline probabilities, including outcome data, are detailed in the Appendix.

Costs
Estimates of the unit costs of diagnostic and treatment procedures have been derived from an analysis of 1998 extra-contractual referral (ECR) tariffs for NHS Trusts in England and from private sector providers. Mean tariff prices have been estimated for relevant health resource groups (HRG) after removing outliers (defined as Trusts with prices ± 2 SD from the mean). The costs of drugs are taken from the British National Formulary (BNF, 1998Go) or the Drug Tariff (Drug Tariff, 1998Go) and are average costs based on all drugs within each class, excluding discounts. Baseline costs are detailed in the Appendix.

Expected costs for each treatment pathway are derived from a cost model that combines procedures, probabilities, frequencies, and unit costs. Table IGo shows an example of a cost model used to derive the expected cost of a cycle of stimulated in-vitro fertilization (SIVF). The main sources of resource use are stimulatory and pituitary down-regulator drugs, monitoring, the IVF procedure itself, and resources associated with complications such as ovarian hyperstimulation syndrome (OHSS). In this example the probability that treatment is cancelled after the stimulatory phase is 10% and the probability of OHSS is 4%. Costs associated with OHSS are derived from a separate cost model.


View this table:
[in this window]
[in a new window]
 
Table I. Cost model for stimulated in-vitro fertilization (SIVF; tubal factor)
 
Outcome measures
Outcomes are measured in terms of a clinical pregnancy [fetal cardiac activity confirmed by ultrasound (Van Voorhis et al., 1995Go)]. It would be preferable to measure outcomes in terms of maternities because the use of stimulatory drugs is expected to lead to higher rates of multiple birth and to a higher incidence of low birthweight babies (Dawood, 1996Go; Corabian, 1997Go). An end-point of pregnancy also does not take into account increased risks of complications such as ectopic pregnancies and miscarriage (Schenker and Ezra, 1994Go). All of these have important cost consequences and are explored further in the Discussion. It has not proved possible to account for these factors with the information currently available in the literature, and the results reported here must be interpreted in this light.

Sensitivity analysis
It has not been possible to obtain estimates of all of the parameters in the treatment models from published studies. Where information is lacking we have used judgements from a clinical advisory group. This combination of evidence with expert judgement is unavoidable but unsatisfactory, and in order to test the assumptions embodied in the baseline models we have carried out extensive sensitivity analysis.

In addition, because costs and outcomes are expected to differ between centres it is essential that the results of the baseline models reported here are validated in a local context.

Tables of results show break-even values of all key parameters. The break-even value is the value of a parameter at which treatments are equivalent in terms of cost-effectiveness. By comparing the break-even value with evidence from local practice or elsewhere it is possible to assess whether the results of the analysis are sensitive to the assumed values of particular parameters. If the break-even value is judged to be feasible, the analysis is sensitive and further work is required.

Cost-effectiveness
The scope of the cost-effectiveness analysis is restricted to costs falling on the health care sector. The perspective is that of a health commissioner whose objective is to maximize outcomes (defined in this case in terms of pregnancies) subject to a budget constraint. The analysis does not include set-up costs. The focus is on costs that are variable in the choice of treatment or with the number of cycles of treatment offered.

In evaluating relative cost-effectiveness, treatment choices are reduced in each case to a comparison of mutually exclusive pairs. Cost-effectiveness is defined as follows: of two options to be compared, the option with the lowest expected cost is more cost-effective unless the incremental cost per additional pregnancy implicit in a decision to select the more expensive option is lower than the average cost per pregnancy implicit in the lower cost option.

In the case of treatments for infertility (but not in all cases) this is equivalent to a rule which says that the option with the lowest average cost per pregnancy is the more cost-effective option (assuming the options are perfectly divisible). If the available budget is fixed, selecting a more expensive option must imply that fewer couples can be treated. Assuming the objective is to maximize the number of pregnancies, it can never be cost-effective to select the more expensive option unless this option produces more pregnancies despite the reduction in the number of couples treated. This result will follow if the average cost per pregnancy is lower.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Costs of diagnosis
Table IIGo summarizes the results of the diagnostic model applied to a cohort of 1000 couples presenting for the first time. It shows the expected breakdown of the cohort by main cause, together with the expected cost per couple. Couples initially diagnosed as having multiple factors have been reclassified into one of the five main groups according to the dominant cause. The average expected cost of diagnosis is £344 per couple.


View this table:
[in this window]
[in a new window]
 
Table II. Results of the diagnostic model
 
Cost-effectiveness analysis
Ovulatory factor
Couples for whom ovulatory factor is the suspected main cause of infertility are divided into three groups (Hughes et al., 1997Go), each of which is modelled separately: (i) hyperprolactinaemia (microadenoma susceptible to medical treatment); (ii) hypothalamic pituitary dysfunction without evidence of ovarian failure or polycystic ovarian syndrome (non-PCOS); (iii) polycystic ovarian syndrome (PCOS)

Causes such as complex pituitary, hypothalamic, thyroid, and genetic causes, hypogonadal hypogonadism, macroadenomas, primary amenorrhoea, and premature ovarian failure are out of the scope of the model.

For the hyperprolactinaemia group (i), both treatment alternatives involve up to six cycles of treatment with dopamine-receptor stimulants (bromocriptine or cabergoline). Results are shown in Table IIIGo.


View this table:
[in this window]
[in a new window]
 
Table III. Results of the cost-effectiveness analysis for ovulatory factor infertility
 
Because outcomes with both treatment regimes are assumed to be the same in the baseline model, the relative cost-effectiveness of bromocriptine is determined by its lower cost. The expected cost per pregnancy (including the expected costs of diagnosis) is £71 (12.4%) higher with cabergoline. This result is sensitive to the cost of cabergoline and to the assumption that expected outcomes are the same. Equally significant is the fact that bromocriptine is sometimes associated with adverse side-effects (such as nausea and vomiting) which may affect a couple's preference between treatment options and the dropout rate.

The alternative treatments for the non-PCOS group (ii) are medical treatment alone or medical treatment followed by SIVF. Medical treatment involves up to six cycles of anti-oestrogens followed by up to four cycles of gonadotrophin therapy. In the SIVF, alternative couples who have not achieved a pregnancy after the full course of medical treatment receive 1–4 cycles of SIVF. We have not included an option in which SIVF is offered without prior medical treatment (Table IIIGo).

The addition of SIVF to the range of treatment options is not cost-effective for this group. This conclusion is based on a comparison of the expected average cost per pregnancy associated with medical treatment alone (£1937) and the incremental cost per additional pregnancy achieved with the addition of SIVF (£7589 after one cycle). The average cost per pregnancy with only one cycle of SIVF is £2408. Because the incremental cost of SIVF is greater than the average cost of medical treatment alone, the addition of SIVF is not cost-effective. This result holds for 1–4 cycles of SIVF.

Break-even values derived from the sensitivity analysis are also shown in Table IIIGo. These are the values required to make the two options equivalent in terms of cost-effectiveness. For example, for the medical treatment plus SIVF option to be cost-effective would require a success rate for SIVF of more than 100% per cycle. None of the break-even values appears plausible and on this basis the results of the analysis are robust to any realistic changes in the baseline parameters.

For the PCOS group (iii) the alternatives in this case are medical treatment (anti-oestrogens and gonadotrophins) and laparoscopic ovarian diathermy (LOD) or medical treatment with LOD followed by 1–4 cycles of SIVF. Details of medical treatments are as for the non-PCOS group. The model assumes that LOD is performed as a separate procedure and not as part of the diagnostic laparoscopy. This means that the costs associated with this procedure represent an upper band. In centres that choose to perform LOD at the time of a diagnostic laparoscopy, additional treatment costs for the LOD will be negligible.

The addition of SIVF to the range of available treatments is not cost-effective for this group, nor is this result sensitive to any feasible change in the parameters of the model (see Table IIIGo).

In order to investigate the cost-effectiveness of SIVF in a context in which LOD is not an available option, an alternative scenario has been modelled. In this scenario three options are compared: (i) medical treatment with LOD, (ii) medical treatment alone, and (iii) anti-oestrogen therapy followed by 1–4 cycles of SIVF.

The results of the baseline model and sensitivity analysis are confirmed by a comparison of medical treatment with LOD against the other two treatment options. From these results it can be argued that, when available, LOD combined with medical treatment is the most cost-effective treatment pathway for women presenting with PCOS. The positive outcomes derived from LOD compensate for the higher costs of gonadotrophin therapy.

Nevertheless, when LOD is not an available option, results from a comparison between medical treatment alone and anti-oestrogen therapy followed by 1–4 cycles of SIVF produce interesting results (Table IVGo). Providing one cycle of SIVF after anti-oestrogen therapy is more cost-effective. But for more than two cycles of SIVF, medical treatment alone represents the best option. Comparing these results with the results of the baseline model (Table IIIGo) it can be seen that LOD is a major driver of the positive outcomes obtained with medical treatment and LOD. If LOD is excluded, the cost per pregnancy for medical treatment alone is significantly higher.


View this table:
[in this window]
[in a new window]
 
Table IV. Alternative treatment options for PCOS couples
 
Tubal factor
Tubal factors are classified as mild, moderate or severe (Hull and Fleming, 1995Go). Each is modelled separately (Table VGo). The treatment options in all cases are surgery or up to four cycles of SIVF. For mild and moderate cases the assumed surgical procedure is laparoscopic ablation (diathermy or laser), performed as a separate procedure. Open surgery is assumed for severe cases. Here, as in the analysis of surgical treatments for endometriosis, it is assumed that appropriate skills and experience are available to carry out these procedures and to achieve the success rates reported in the literature. Where this is not the case locally, then either costs will be higher and/or outcomes will not be as good as those assumed. The impact on the relative cost-effectiveness of modelled alternatives can be judged from the break-even values shown in the tables of results.


View this table:
[in this window]
[in a new window]
 
Table V. Results of the cost-effectiveness analysis for tubal factor infertility
 
In the baseline model, surgery is the most cost-effective treatment for mild tubal factor (Table VGo). The average cost per pregnancy is £1986 (surgery) compared with £9400 (SIVF, one cycle). Surgery offers better outcomes at a lower total cost than up to two cycles of SIVF. SIVF is dominated in this case (Table VGo). At three or more cycles, outcomes are better with SIVF but the incremental cost per additional pregnancy ranges between £48 018 (four cycles) and £301 079 (three cycles).

The sensitivity analysis shows that this result is sensitive to the success rate and cost of surgery. In particular, SIVF is expected to be cost-effective if either the success rate of surgery is <10.8% per cycle (46% in the baseline), or the cost of surgery is >£3926 (£516 in the baseline).

Neither appears plausible except in extreme cases.

Although SIVF is not dominated in the case of moderate tubal factor, surgery remains the more cost-effective option. The average costs per pregnancy are £6162 (surgery) and £11 125 (SIVF, one cycle). The incremental cost per additional pregnancy implicit in a decision to offer SIVF to couples in this cohort is £24 027–£11 773 (1–4 cycles) (Table VGo).

The sensitivity analysis again highlights the importance of the assumed success rate and cost of surgery. In addition, the assumed success of SIVF becomes relevant. For SIVF to be the most cost-effective option for this group requires: (i) the expected success of surgery to be less than 7.2% (13% in the baseline), or (ii) the cost of surgery to be more than £1161 (£516 in the baseline), or (iii) the success of SIVF to be more than 36.1% (20% in the baseline).

One way to interpret these break-even values is to say that in particular centres, where the success rate of surgery is low or where the success rate of SIVF is particularly high, SIVF may be a viable option for couples in this group.

In the case of severe tubal factor, SIVF is the most cost-effective option in the baseline model. Expected outcomes are better and although total expenditure is higher with SIVF, the average cost per pregnancy is lower [£14 833 (SIVF, one cycle) compared with £16 221 (surgery)]. The incremental cost per additional pregnancy implicit in the choice of SIVF is £10 867–£12 474 (1–4 cycles), and this is lower than the average cost implicit in the surgery option.

As might be expected, this result is sensitive to the assumed success rates of both surgery and SIVF, and to the costs of the two treatments. Surgery would be the more cost-effective option for this patient group if (i) the success rate of surgery is more than 12.1% (10% in the baseline), or (ii) the success rate of SIVF is less than 12.5% per cycle (15% in the baseline), or (iii) the cost of surgery is less than £1066 (£1337 in the baseline), or (iv) the total cost of the SIVF treatment (including the stimulation phase) is greater than £2084 per cycle (£1717 in the baseline).

The extent to which any of these values is plausible will depend on local circumstances.

The superiority of SIVF is confirmed by the results of an alternative scenario in which only recombinant gonadotrophins (as opposed to both urinary and recombinant gonadotrophins) are used in the ovarian stimulation protocol of this treatment. In this case the total cost per cycle increases to £1916 (or £1960 if only the most expensive recombinant gonadotrophins are considered). Both figures are below the break-even value obtained in the sensitivity analysis at which surgery is as cost-effective as SIVF (£2084).

Endometriosis
Endometriosis is classified as mild, moderate, or severe (Acosta et al., 1993Go). Each is modelled separately. Treatment alternatives are surgery (laparoscopic ablation for mild, open surgery for moderate) or up to four cycles of SIVF for mild and moderate cases. It is assumed that laparoscopic ablation is performed as a separate procedure. In the case of severe disease, modelled alternatives are open surgery alone or open surgery followed by up to four cycles of SIVF. These treatment alternatives are consistent with recent RCOG guidance (RCOG, 1998bGo).

RCOG guidance suggests that surgical ablation of minimal and mild endometriosis improves fertility. The guidance also suggests that stimulated intrauterine insemination (SIUI) may be an effective option for this group. For moderate and severe endometriosis, the guidelines suggest that surgery may improve fertility and that IVF may be considered as an alternative to surgery or as a secondary treatment following unsuccessful surgery.

It has also been suggested that surgery should be regarded as a means of relieving pain, rather than as a treatment for infertility. If this is the case the relevant comparison would be between surgery alone and surgery followed by SIVF. This is the comparison that has been modelled in the case of severe disease. For mild and moderate cases we have not modelled this progression explicitly, but the evidence suggests that surgery alone is likely to be a dominant option for mild disease (as in the baseline model), while in moderate cases the relative cost-effectiveness of adding SIVF is very sensitive to the relative success rates of surgery and surgery plus SIVF. Results from the baseline model are similarly sensitive (Table VIGo).


View this table:
[in this window]
[in a new window]
 
Table VI. Results of the cost-effectiveness analysis for endometriosis
 
In the case of mild endometriosis, SIVF at less than two cycles is dominated by surgery. Expected outcomes are lower and the cost is higher with SIVF. At two cycles and more, expected outcomes are higher with SIVF but the additional cost is such that the average cost per pregnancy is always higher. The average cost per pregnancy is £2393 (surgery) and £9400–£8377 (SIVF, 1–4 cycles). The incremental costs per additional pregnancy implicit in a decision to offer SIVF are between £38 059 (two cycles) and £16 464 (four cycles). The break-even values generated by the sensitivity analysis do not appear to offer any plausible situation in which this result could be reversed.

Table VIIGo shows break-even values for the success rate per cycle of SIUI which would make SIUI equivalent to SIVF in terms of cost-effectiveness for mild disease. In the absence of evidence of the effectiveness of SIUI for this group, we are unable to judge the plausibility of these break-even values. However, since SIVF is dominated by surgery in any case, surgery is likely to remain the most cost-effective choice for mild disease.


View this table:
[in this window]
[in a new window]
 
Table VII. SIVF versus SIUI: break-even success rates for SIUI
 
The same result occurs in the case of moderate endometriosis. SIVF at less than two cycles is dominated by surgery for couples in this group. The average costs per pregnancy are higher than for mild disease: £8673 (surgery), £11750– £10 416 (SIVF, 1–4 cycles). Surgery is the more cost-effective option in the baseline model.

The relative cost-effectiveness of surgery is more sensitive in this case (Table VIGo). In particular, the baseline result is sensitive to the relative success rates and costs of the two treatments. SIVF will be cost-effective if: (i) the success rate of surgery is less than 16.0% (20% in the baseline), or (ii) the success rate of SIVF is more than 25.1% per cycle (20% in the baseline), or (iii) the cost of surgery exceeds £1952 (£1337 in the baseline), or (iv) the cost of the full SIVF treatment (including stimulation) is less than £1163 per cycle (£1717 baseline).

Because of the sensitivity of the result in this case, the relative cost-effectiveness of SIVF for this cohort may be different in different centres. In centres in which the success rate associated with surgery is low, or the success of SIVF is high, assisted conception may be the preferred option.

In the case of severe endometriosis, because outcomes are relatively poor, the expected cost per pregnancy is high for both of the treatment options: £34692 (Surgery), £19488 (SIVF, one cycle). SIVF is the most cost-effective option for this cohort.

This result is also sensitive to the relative success rates and costs of the two treatments, although none of the break-even values appears particularly plausible. Surgery will be the more cost-effective option if: (i) the success rate of surgery exceeds 11.4% (5% in the baseline), or (ii) the success rate of SIVF is less than 6.5% per cycle (15% in the baseline), or (iii) the cost of surgery is less than £364 (£1337 baseline), or (iv) the cost of the full SIVF treatment is more than £4859 (£2233 baseline).

Until a means is found of improving the success associated with surgery, the relative cost-effectiveness of SIVF appears to be robust.

As for severe tubal factor, an alternative scenario was modelled in order to test the superiority of SIVF when only recombinant gonadotrophins are included in the ovarian stimulation stage. In this case, the resulting total cost per cycle of SIVF (£1916 for an average price of recombinant gonadotrophins and £1960 for the most expensive) is well below the value required for surgery to become the most cost-effective option (£4859).

Male factor
Male factor infertility is classified as mild, moderate or severe, on the basis of observed sperm parameters. There is no recognized universal definition of severity for male factor infertility. For the purpose of this study, we have interpreted severe disease as obstructive azoospermia or severe oligoteratoasthenozoospermia; and mild disease where sperm parameters are just outside the normal range (ESHRE, 1994). Moderate cases are assumed to incorporate combinations of sperm abnormalities lying within these two extremes. Varicocele is out of the scope of this model. For mild male factor infertility the comparison is between no treatment and four cycles of SIUI. For moderate, the comparison is between four cycles of SIUI and 1–4 cycles of intracytoplasmic sperm injection (ICSI) with SIVF. For the severe cohort, the comparison is between six cycles of donor insemination (DI) and four cycles of stimulated donor intrauterine insemination (SDIUI); between six cycles of DI and 1–4 cycles of ICSI; and between SDIUI (four cycles) and 1–4 cycles of ICSI (Table VIIIGo).


View this table:
[in this window]
[in a new window]
 
Table VIII. Results of the cost-effectiveness analysis for male factor subfertility
 
On the assumptions in the baseline model, no treatment is the most cost-effective option for mild male factor (Table VIIIGo). Although SIUI offers better outcomes, the additional costs are such that the expected average cost per pregnancy is higher: £2292 (no treatment), £5104 (SIUI). The incremental cost per additional pregnancy implicit in a choice of SIUI is £8546.

This result is sensitive to the assumed success rates of the two alternatives. SIUI will be the more cost-effective option if: (i) the spontaneous pregnancy rate per month associated with no treatment is less than 2.0% (5% in the baseline), or (ii) the success rate of SIUI is more than 36.0% per cycle (15% in the baseline).

Information on the rate of spontaneous pregnancy for couples in the mild group is poor and it is difficult to judge the plausibility of the break-even value of 2.0% per month. An alternative is to compare one cycle of SIUI (rather than four cycles) with no treatment. On the assumption that 5% of the no treatment cohort will achieve a pregnancy within the period, SIUI is a more cost-effective option. However, these results are extremely sensitive to the spontaneous pregnancy rate. For example, at a spontaneous pregnancy rate of 5.74%, both options become equivalent.

For moderate male factor, in a comparison between SIUI (four cycles) and ICSI (up to four cycles), ICSI is dominated for the first cycle. As the number of cycles of ICSI is increased, outcomes improve but the increase in total cost is such that the average cost per pregnancy is always higher. The average costs per pregnancy are £5104 (SIUI) and £9229–£8273 (ICSI, 1–4 cycles). On this basis, SIUI is the most cost-effective option.

Average break-even values derived from the sensitivity analysis are shown in Table VIIIGo. The result does not appear to be sensitive to the assumptions in the baseline model for plausible values of the parameters. Unless either: (i) the success rate of SIUI is less than 8.8% per cycle (15% in the baseline), or (ii) the success rate of ICSI is more than 51.3% per cycle (31.0% in the baseline), the relative cost-effectiveness of SIUI is robust.

The robustness of these results is corroborated by an alternative scenario in which gonadotrophins, as opposed to anti-oestrogens plus gonadotrophins are used in the ovarian stimulation protocol of SIUI. In this case the total cost per cycle increases to £543. According to the results obtained from the sensitivity analysis, SIUI is still the most cost-effective treatment. A further alternative is to compare one cycle of each treatment. In this case, the baseline result is the same as above. However, the results are more sensitive to the costs and success rates of both treatments.

Three pair-wise comparisons are shown in the case of severe male factor. Donor insemination (DI, six cycles) compared with SDIUI (four cycles) (Table IXGo); DI (six cycles) compared with 1–4 cycles of ICSI (Table IXGo); and SDIUI (four cycles) compared with 1–4 cycles of ICSI (Table VIIIGo). The relationship between these options should be transitive.


View this table:
[in this window]
[in a new window]
 
Table IX. Supplementary results for severe male factor infertility
 
SDIUI is more cost-effective than DI (Table IXGo). The average costs per pregnancy are £5964 (DI) and £4362 (SDIUI). This result is sensitive to relatively small changes in the success rates and costs of the two procedures. For example, DI could be the more cost-effective option if: (i) the success rate of DI is more than 10.4% per cycle (7.5% in the baseline), or (ii) the success rate of SDIUI is less than 14.4% per cycle (20% in the baseline).

DI is more cost-effective than ICSI (Table IXGo). Although expected outcomes are better with ICSI, the additional cost is such that the average cost per pregnancy is always higher: £5964 (DI), £9229–£8273 (ICSI, 1–4 cycles). This result is less sensitive to changes in the baseline parameters. For example, ICSI could be the more cost-effective option if: (i) the success rate of DI is less than 4.8% per cycle (7.5% in the baseline), or (ii) the success rate of ICSI is more than 47.7% per cycle (31.0% in the baseline).

Since SDIUI has been shown to be more cost-effective than DI, and DI is shown to be more cost-effective than ICSI it should be the case than SDIUI is more cost-effective than ICSI. This is the case (Table VIIIGo).

SDIUI (four cycles) dominates ICSI at less than three cycles. At less than three cycles of ICSI, expected outcomes are better with SDIUI and costs are lower. In the third and fourth cycle of ICSI, outcomes are better than those expected with SDIUI, but the additional cost is such that the average cost per pregnancy is always higher. Average costs are £4362 (SDIUI) and £9229–£8273 (ICSI, 1–4 cycles).

Table VIIIGo also shows average break-even values derived from the sensitivity analysis. None of these values appear to be plausible and the cost-effectiveness of SDIUI relative to ICSI appears to be robust.

These results are subject to two important qualifications: (i) The superiority of SDIUI over DI is very sensitive to the assumed success rate of DI. The break-even value for the success of DI at 10.4% per cycle is likely to be exceeded in a number of centres, and in these centres DI will be the more cost-effective treatment option. More importantly, if it was possible to quantify the costs associated with the higher rates of multiple pregnancy and low birthweight expected with SDIUI, the relative cost-effectiveness of SDIUI and DI may easily be reversed. (ii) In addition, there may be good reasons for couples to prefer ICSI over DI or SDIUI because this technique does not require the use of donor spermatozoa. Where this is the case, the analysis indicates the implicit cost of this decision. A decision to purchase four cycles of ICSI in preference to the more cost-effective option involves additional expenditure of £3010 per couple. The expected number of pregnancies is higher with ICSI (9.3 compared with 7.2) but at an incremental cost per additional pregnancy of approximately £21 000 the choice of ICSI must be evaluated on the basis of other factors of importance to the couple, rather than on the basis of incremental cost alone. This analysis does not mean that ICSI should be denied to couples who may benefit.

Unexplained infertility
The unexplained infertility group represents almost a quarter of all treated couples, and this is the single most common classification of infertility even after recommended diagnostic tests have been performed. The treatment choices modelled for this group are SIUI (four cycles) and 1–4 cycles of SIVF. SIUI is the more cost-effective option for this group (Table XGo).


View this table:
[in this window]
[in a new window]
 
Table X. Results of the cost-effectiveness analysis for unexplained infertility
 
Up to two cycles of treatment, SIVF is dominated by SIUI. Expected outcomes are better with SIUI and costs are lower. Beyond two cycles expected outcomes are higher with SIVF, but the additional cost is such that the average cost per pregnancy is always higher. Average costs per pregnancy are £3681 (SIUI) and £9853–£8770 (SIVF, 1–4 cycles). The incremental cost per additional pregnancy implicit in a decision to offer SIVF is £32 848 (after four cycles).

Table XGo also shows break-even values of the key parameters. SIVF could be the most cost-effective option if: (i) the success rate of SIUI is less than 8.1% per cycle (20% in the baseline), or (ii) the success rate of SIVF is more than 60% per cycle (24.0% in the baseline), or (iii) the cost of the full SIUI treatment (including stimulation) exceeds £1517 per cycle (£467 baseline), or (iv) the cost of the full SIVF treatment (including stimulation) is less than £393 per cycle (£1717 baseline).

None of these values appears plausible and the relative cost-effectiveness of SIUI appears to be robust.


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The results of the cost-effectiveness analysis show a high degree of consistency with the treatment guidelines recently produced by the Royal College of Obstetricians and Gynaecologists (RCOG, 1998bGo). This gives us confidence in our own analysis and it also confirms that (in this case) evidence of the relative clinical effectiveness of alternative treatments is broadly consistent with evidence of cost-effectiveness. Results of the modelling exercise suggest that SIVF is the most cost-effective of currently available treatment options in the case of severe tubal factor and severe endometriosis. Surgery is expected to be more cost-effective in cases of mild or moderate disease, although the relative cost-effectiveness in the case of moderate disease is very sensitive to changes in the assumed costs and success rates.

The results for tubal disease are consistent with other studies. Van Voorhis et al. describe how the severity of disease affects outcome, and report that patients with minimal disease have reasonably good pregnancy rates after surgery, while success rates are poor if the patient has severe disease (Van Voorhis et al., 1998Go). Others (Gillett et al., 1995Go) concluded that treatment of tubal disease with a good prognosis is more cost-effective if treated by microsurgery than by IVF.

The most cost-effective treatment for hyperprolactinaemia is bromocriptine because of its lower cost. Medical treatment (anti-oestrogens and gonadotrophin therapy) or medical treatment with LOD is the most cost-effective way to treat ovulatory infertility due to hypothalamic pituitary dysfunction. However, when LOD is not an available treatment option for PCOS-related disorders, SIVF becomes cost-effective, at least for one cycle of treatment. For other causes, SIUI (unexplained, male moderate) and SDIUI (male severe) are cost-effective. In the case of unexplained infertility in particular, other studies have reached similar conclusions (Zayed et al., 1997Go; Guzick et al., 1998Go).

One of the important limitations of this analysis is the fact that outcomes are measured in terms of a clinical pregnancy. It has not been possible to take account of the additional costs resulting from complications such as ectopic pregnancies. Neither has the increased risk of multiple births and low birthweight babies associated with treatment regimes involving ovarian stimulation been taken into account. There is evidence which suggests that the number of higher-order multiple gestations is growing in the Western World and such increase has been attributed mainly to the use of assisted reproduction techniques (Collins, 1994Go). Callahan and colleagues examined the economic impact of multiple gestation pregnancies in Bingham and Women's Hospital, Boston and found that in 1991, the costs associated with triplet deliveries were over 11 times higher than those associated with singleton births (Callahan et al., 1994Go). A more recent cost analysis estimated the cost of an IVF cycle to be $39 000 for singleton deliveries and $340 000 for triplets (US dollars 1992 prices) (Goldfarb et al., 1996Go). The reason these significant factors are excluded from our models is because data are not generally available with the specificity required for all of the parameters and treatments considered. It is important to note that if we had been able to evaluate these factors explicitly, the relative superiority of assisted conception treatments may change.

One possibility for reducing the number of multiple pregnancies in SIVF is to reduce the number of embryos transferred. Such a strategy has the potential to reduce the costs associated with multiple births, but such savings have to be compared against reduced pregnancy rates. UK data published by the Human Fertilisation and Embryology Authority (HFEA) show that the transfer of one embryo results in pregnancy rates of just 8%, whereas transferring three embryos leads to pregnancy rates of 26% (HFEA, 1997). A recent Swedish study (Wølner-Hanssen and Rydhstroem, 1998Go) has attempted to compare the costs associated with the transfer of one or two embryos in IVF treatments. Much of the analyses were based on assumption, in particular success rates relating to one embryo transferred. However, the transfer of a single embryo was found to be more cost-efficient, even when more treatments were required to match the success rate after transfers involving two embryos. Although the likelihood of multiple gestation births may be reduced in IVF treatments through this practice, it is important to remember that any treatment that involves stimulation of the ovaries carries with it an increased risk of multiple birth. It is clear that this is an important area for further research.

Another possibility for reducing the costs associated with IVF treatments is to use frozen embryos as opposed to stimulating and retrieving eggs for each individual treatment cycle. It has been shown (Van Voorhis et al., 1995Go) how such a technique does not reduce on-going pregnancy rates yet saves, per cycle, between 25 and 45% of the cost of any other assisted reproductive procedure. Despite this, other workers (Gillett et al., 1995Go) were unable to demonstrate the relative cost-effectiveness of IVF over surgery for tubal factor infertility even when every second cycle of IVF treatment used frozen embryos. Nevertheless, in cases where the relative superiority of surgery over SIVF is sensitive (e.g. moderate tubal factor and moderate endometriosis), SIVF may become cost-effective if cryopreservation is adopted.

Analysis of cost-effectiveness is designed to promote efficiency in the use of NHS resources. However, to the extent that equity and choice are also important parameters, cost-effectiveness is not the only important determinant of commissioning decisions. For example, there may be good reasons for a couple with severe male factor infertility to prefer ICSI instead of SDIUI because it avoids the use of donor spermatozoa. Our results do not imply that this choice should be prohibited simply because SDIUI is the more cost-effective option.

All of the results reported here are derived from baseline values of the model parameters. Some of these parameters are subject to genuine uncertainty, others are expected to differ substantially between centres. It is important that conclusions derived from the baseline models are validated locally by reference to local costs, success rates and the treatment options available. The sensitivity analysis confirms that in local centres in which the success rate of SIVF is higher than assumed in the baseline, SIVF may become the most cost-effective option in cases of moderate disease. Similarly, SIVF may be cost-effective in centres in which the costs of surgery are higher than assumed or success rates are lower. Where the appropriate surgical procedure is laparoscopic ablation (tubal factor, mild and moderate; endometriosis, mild) we have assumed that appropriate skills and experience are available locally to carry out these procedures. Where this is not the case, costs will be higher and outcomes poorer than assumed in the baseline. A recent study in the USA (Vanderlaan et al., 1998Go) highlights this important clarification. In a comparison of cost and outcome between a Health Maintenance Organization (HMO) and a Preferred Provider Organization (PPO) providing infertility care, the authors describe substantial differences in both cost and outcome which could not be attributed solely to case-mix differences. In addition to the differences in charges between HMO and PPO organisations, the authors remarked how care in the HMO organisation was provided by specialists while in the PPO, treatment was also provided by general gynaecological consultants. The differences in expertise between these groups could have explained the differences in outcome.

Finally, it is also important to note that we have modelled the diagnostic and treatment pathways independently and expected diagnostic costs have been added to overall treatment costs in the evaluation of relative cost-effectiveness. The implicit assumption is that the diagnostic pathway is exogenous to treatment success, or equivalently, that the diagnostic process is efficient: as far as existing procedures allow, the main cause of infertility is accurately determined. In centres that do not follow the diagnostic process recently recommended in UK guidelines (RCOG, 1998aGo) it is possible that couples may be misdiagnosed, with the consequence that treatments may be inappropriate. It is clear that in such cases success rates will be lower than those assumed in the models.

Appendix


View this table:
[in this window]
[in a new window]
 
Baseline pregnancy rates and expected treatment costs*
 

    Acknowledgments
 
This work was commissioned by the National Infertility Awareness Campaign (NIAC), aided by an educational grant from Serono Laboratories (UK) Ltd. We are indebted to all of those who have provided assistance and advice throughout the course of this work. Particular thanks are due to Professor W. Ledger and Doctor E. Lenton. We are also grateful to the three independent clinical experts who have read the full report of this work and who offered useful suggestions for improvement.


    Notes
 
3 To whom correspondence should be addressed at Trent Institute for Health Services Research, University of Nottingham, Nottingham, NG7 2UH, UK Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Acosta, A., Buttram, V., Franklin, R. et al. (1993) A proposed classification of pelvic endometriosis. Obstet. Gynecol., 42, 19–25.

Barraza-Llorens, M. (1998) Effectiveness of Assisted Conception: a Systematic Review of the Literature. York Health Economics Consortium.

British Fertility Society (1995) Diagnostic Classification, Basic Diagnostic Procedures and Basic Diagnostic Data Collection. British Fertility Society.

BNF (1998) British National Formulary, British Medical Association and the Royal Pharmaceutical Society of Great Britain. London.

Callahan, T., Hall, J., Ettner, S. et al. (1994) The economic impact of multiple gestation pregnancies and the contribution of assisted reproductive techniques to their incidence. N. Engl. J. Med., 331, 244–249.[Abstract/Free Full Text]

Collins, J. (1994) Reproductive technology—the price of progress. N. Engl. J. Med., 331, 270–271.[Free Full Text]

Collins, J., Feeny, D. and Gunby, J. (1997) The costs of infertility diagnosis and treatment in Canada in 1995. Hum. Reprod., 12, 951–958.[ISI][Medline]

Corabian, P. (1997) In-vitro Fertilisation and Embryo Transfer as a Treatment for Subfertility. Alberta Heritage Foundation for Medical Research, Canada.

Dawood, M. (1996) In-vitro fertilisation, gamete intrafallopian transfer, and superovulation with intrauterine insemination: efficacy and potential hazards on babies delivered. Am. J. Obstet. Gynecol., 174, 1208–1217.[ISI][Medline]

DoH (1996) Immunisation Against Infectious Diseases. Department of Health, HMSO, London.

Drug Tariff (June 1998) National Health Service, England and Wales. HMSO, London.

EHC (1992) The Management of Subfertility. Effective Health Care Bulletin Number 3. School of Public Health, University of Leeds.

ESHRE Capri Workshop Group (1994) Male sterility and subfertility: guidelines for management. Hum. Reprod., 9, 1260–1264.[ISI][Medline]

Gillett, W., Peek, J. and Lilford, R. (1995) Costs and effectiveness of infertility services in New Zealand: a decision analysis. A Report to the National Advisory Committee on Core Health and Disability Support Services, Wellington, New Zealand.

Goldfarb, J., Austin, C., Lisbona, H. et al. (1996) Cost-effectiveness of in-vitro fertilisation. Obstet. Gynecol., 87, 18–21.[Abstract/Free Full Text]

Guzick, D., Sullivan, M., Adamson, G. et al. (1998) Efficacy of treatment for unexplained infertility. Fertil. Steril., 70, 207–213.[ISI][Medline]

Haan, G. and Van Steen, R. (1992) Costs in relation to effects in in-vitro fertilisation. Hum. Reprod., 7, 982–986.[Abstract]

Holst, N., Maltau, J., Forsdahl, F. et al. (1991) Handling of tubal infertility after introduction of in vitro fertilisation: changes and consequences. Fertil. Steril., 55, 140–143.[ISI][Medline]

Hughes, E., Collins, J. and Vandekerckhove, P. (1997) Clomiphene Citrate vs Placebo for Ovulation Induction in Oligo-amenorrhoeic Women. Subfertility Module of the Cochrane Database of Systematic Reviews. Lilford, R. Hughes E. and Vandekerckhove P. Cochrane Collaboration, Oxford.

Hull, M. (1996) Managed care in infertility. Curr. Opin. Obstet. Gynecol., 7, 160–167.[ISI]

Hull, M. and Fleming C. (1995) Tubal surgery versus assisted reproduction: assessing their role in infertility therapy. Curr. Opin. Obstet. Gynecol., 7, 160–167.[ISI][Medline]

Human Fertility and Embryology Authority (HFEA) (1997) Human Fertilisation and Embryology Authority. Sixth Annual Report. London.

Peterson, C., Poulson, A., Hatasaka, H. et al. (1994) Ovulation induction with gonadotrophins and intrauterine insemination compared with in vitro fertilisation and no therapy: a prospective, nonrandomised, cohort study and meta analysis. Fertil. Steril., 62, 535–544.[ISI][Medline]

RCOG (1998a) The Initial Investigation and Management of the Infertile Couple. Royal College of Obstetricians and Gynaecologists. London.

RCOG (1998b) The Management of Infertility in Secondary Care. Royal College of Obstetricians and Gynaecologists.

Schenker, J. and Ezra Y. (1994) Complications of assisted reproductive techniques. Fertil. Steril., 61, 411–422.[ISI][Medline]

Snick, H., Snick, T., Evers, J. et al. (1997) The spontaneous pregnancy prognosis in untreated subfertile couples: the Walcheren primary care study. Hum. Reprod., 12, 1582–1588.[Abstract]

Van den Eede, B. (1995) Investigations and treatment of infertile couples: ESHRE guidelines for good clinical and laboratory practice. Hum. Reprod., 10, 1246–1271.[ISI][Medline]

Vanderlaan, B., Karande, V., Krohm, C. et al. (1998) Cost considerations with infertility therapy: outcome and cost comparison between health maintenance organization and preferred provider organization care based on physician and facility cost. Hum. Reprod., 13, 1200–1205.[Abstract]

Van Voorhis, B., Sparks, A., Syrop, C. et al. (1995) The efficacy and cost-effectiveness of embryo cryopreservation compared with other assisted reproductive techniques. Fertil. Steril., 64, 647–650.[ISI][Medline]

Van Voorhis, B., Stovall D., Sparks, A. et al. (1997) Cost-effectiveness of infertility treatments: a cohort study. Fertil. Steril., 67, 830–836.[ISI][Medline]

Van Voorhis, B., Stoval D., Allen, B. et al. (1998) Cost-effective treatment of the infertile couple. Fertil. Steril., 70, 995–1005.[ISI][Medline]

Wølner-Hanssen, P. and Rydhstroem, H. (1998) Cost-effectiveness analysis of in-vitro fertilisation: estimated costs per successful pregnancy after transfer of one or two embryos. Hum. Reprod., 13, 88–94.[Abstract]

Zayed, F., Lenton, E. and Cooke, I. (1997) Comparison between stimulated in-vitro fertilisation and stimulated intra-uterine insemination for the treatment of unexplained and mild male factor infertility. Hum. Reprod., 12, 2408–2413.[Abstract]

Submitted on June 11, 1999; accepted on September 28, 1999.