A health-economic decision-analytic model comparing double with single embryo transfer in IVF/ICSI

Human Reproduction A health-economic decision-analytic model comparing double with single embryo transfer in IVF/ICSI Paul De Sutter 1 Jan Gerris 0 Marc Dhont 1 0 Centre for Reproductive Medicine, Middelheim Hospital , Antwerpen , Belgium 1 Infertility Centre, University Hospital Gent BACKGROUND: Single embryo transfer (SET) is the sole strategy with which to reduce the incidence of twins following assisted reproductive technology (ART), but SET may increase the number of ART cycles needed per live-born child. Its cost-effectiveness compared with double embryo transfer (DET) is therefore unknown. METHODS: A decision-analytic model comparing SET with DET was developed. Estimates were obtained from literature, national pregnancy registers and local hospital records. A sensitivity analysis was performed, using pregnancy rates from four published studies. The outcome measure was the cost per child born, calculated from IVF procedure-related, pregnancy-related and neonatal care costs. Neonatal mortality and long-term morbidity costs were not taken into account. RESULTS: Independently of the pregnancy rates used, the SET cost per child born was in all instances the same as with DET, varying from C 9 520 (SET) versus C 9 511 (DET) to C 12 254 (SET) versus C 12 934 (DET). CONCLUSIONS: More ART cycles are required to obtain the same numbers of children born following SET compared with DET. Because SET allows the avoidance of twins and thus diminishes pregnancy-related and neonatal care costs, there is no difference in the cost per child born between SET and DET. The real advantage of SET is the avoidance of the very high long-term costs resulting from the increased morbidity of twins after birth. cost-effectiveness/decision-analytic model/health-economics/IVF-ICSI/single embryo transfer Introduction Infertility is an important issue for health-economic evaluation because of the high impact on society (Philips et al., 2000). The demand for infertility treatment is still increasing and both patients and insurance companies want to know how much a given treatment will cost. A less efficacious but relatively cheap treatment (e.g. no treatment or intrauterine inseminations) may be preferred over a more efficacious but more expensive technique (e.g. IVF, ICSI) (Mol et al., 2000). New and more successful drugs may be more expensive than older, less successful ones and the decision for reimbursement undoubtedly requires an economic evaluation. The studies comparing urinary and recombinant gonadotrophins are a recent example of the importance of health-economics in assisted reproduction. Several meta-analyses, published a couple of years ago, have demonstrated the superiority of recombinant gonadotrophins over the older urinary products (Out et al., 1996; Daya and Gunby, 1999). Recently the same groups published a cost-effectiveness study showing that recombinant gonadotrophins may be more expensive, but due to their superiority, are more cost-effective than the urinary products (Daya et al., 2001; Sykes et al., 2001; Silverberg et al., 2002). The cost-effectiveness studies of recombinant gonadotrophins illustrate the methodology used in health-economics. Since a prospective economic evaluation of a sufficiently large group of real patients is almost impossible to perform in practice, a mathematical simulation is used. A decision-analytic model, called a Markov model (Briggs and Sculpher, 1998) consists of a tree structure in which each arm corresponds to a certain outcome occurring with a certain probability. Probabilities for each arm as well as estimates of costs for each particular outcome are obtained from meta-analyses, randomized trials, national registries, insurance data and expert opinions. A computer program allows a high number of virtual patients enter the tree model and calculates the final outcomes and corresponding costs. Since the input parameters can be varied, the impact of each individual parameter on the output can be studied. This so-called sensitivity analysis has the limitation that only one parameter can be varied at a time, in contrast to the Monte Carlo method (Doubilet et al., 1985) where distributions of all parameters are taken into account at once. We developed a simple reproducible decision-analytic model in Microsoft Excel to investigate the cost-effectiveness of single embryo transfer (SET). Elective SET unquestionably is the only effective measure to reduce the incidence of twins following assisted reproduction techniques (ART) (Dhont, 2001; Gerris et al., 2001). Although this principle has already been acknowledged both from a medical and an ethical point of view (ESHRE Campus Course Report, 2001), the costeffectiveness of SET has not yet been established using real pregnancy rates. Wlner-Hanssen and Rydhstroem have calculated the cost-effectiveness of SET using hypothetical pregnancy rates, no exact data on the success of SET being available at that time (Wlner-Hanssen and Rydhstroem, 1998). It can be anticipated that by performing SET the number of ART cycles needed to obtain a pregnancy will be increased. Of paramount importance, therefore, is the question to what extent SET would influence the success rate of ART. Only a health-economic analysis taking into account all possible variables can answer this question (Meltzer, 2001). Another aspect is that many infertile couples deliberately opt for a twin pregnancy to short-cut their costly and unpredictable efforts to establish a family. In comparing SET with double embryo transfer (DET), there are various costs to consider: direct and indirect; short and long-term; and those that are measurable and non-measurable. These latter costs are especially difficult to calculate. Materials and methods For our model we developed a spreadsheet tree model as depicted in Figure 1. A probability must be attributed to each branch of the tree and looping must be possible. After each failed treatment cycle we allowed each patient to stop or to continue, with a fresh or thaw cycle, depending on whether embryos remained frozen. A pregnancy may end in a miscarriage and in our model the most important outcome events are of course premature birth and the costs of neonatal intensive care. When a computer simulation is used one has the choice to define the end-point as the birth of two healthy children and the model could then calculate how many cycles of SET as compared with DET are required and at what overall cost. Of course the results of this modelling exercise strongly depend on the estimates used, and the value of the conclusions therefore depends on the reliability of the data used for input. Pregnancy rates in particular may vary widely between centres and therefore influence the analysis. Also the cost of the IVF procedure itself may vary extensively and finally the real costs of premature twin births are difficult to estimate correctly. In our simulation we had to agree on certain presumptions. Every (...truncated)