Individualized cost-effective conventional ovulation induction treatment in normogonadotrophic anovulatory infertility (WHO group 2)

Transcription

1 Human Reproduction Page 1 of 8 Hum. Reprod. Advance Access published July 8, 2005 doi: /humrep/dei164 Individualized cost-effective conventional ovulation induction treatment in normogonadotrophic anovulatory infertility (WHO group 2) Marinus J.C.Eijkemans 1,2,4, Suzanne Polinder 1, Annemarie G.M.G.J.Mulders 2, Joop S.E.Laven 2, J.Dik F.Habbema 1 and Bart C.J.M.Fauser 2,3 1 Department of Public Health and 2 Division of Reproductive Medicine, Department of Obstetrics and Gynaecology, Erasmus MC University Medical Center, Rotterdam and 3 Department of Reproductive Medicine, University Medical Center, Utrecht, The Netherlands. 4 To whom correspondence should be addressed at: Department of Public Health, Erasmus MC University Medical Center, Dr Molewaterplein 50, PO Box 1738, 3000 DR Rotterdam, The Netherlands. m.eijkemans@erasmusmc.nl BACKGROUND: Conventional treatment in normogonadotrophic anovulatory infertility (WHO 2) consists of clomiphene citrate (CC), followed by exogenous gonadotrophins (FSH) and IVF. Response to these treatments may be predicted on the basis of individual patient characteristics. We aimed to devise a patient-tailored, cost-effective treatment algorithm involving the above-mentioned treatment modalities, based on individual patient characteristics. METHODS: Sixteen prognostic groups are defined, according to the presence or absence of: age >30 years, amenorrhea, elevated androgen levels and obesity. The chances of response with each of the three treatments were calculated using prediction models. Treatment costs were based on the data of 240 patients visiting a specialist academic fertility unit. Outcome was an ongoing pregnancy within 12 months after initiation of treatment. The costs per pregnancy of three different strategies were compared, with a threshold for cost-effectiveness of RESULTS: The strategy CC + FSH + IVF compared with FSH + IVF generated more pregnancies against lower costs. Compared with CC + IVF, it also produced more pregnancies, but at higher costs. For <30 years of age with normal androgen levels, costs per pregnancy were less than For women >30 years old, costs per pregnancy were and over , when presenting with normal or elevated androgen levels, respectively. CONCLUSIONS: The conventional treatment protocol is efficient for women aged <30 years with normal androgen levels. For women >30 years old with elevated androgen levels, FSH may be skipped. Key words: cost-effectiveness/ivf/ovulation induction Introduction The most prevalent form of anovulatory infertility is characterized by normal FSH together with normal estrogen serum levels (WHO group 2) (ESHRE Capri Workshop Group, 1995). Recently introduced treatment options for this form of cycle disturbances include laparoscopic ovarian surgery (Cohen, 1996; Farquhar et al., 2001; Bayram et al., 2004), insulinsensitizing agents (Nestler et al., 2002; Harborne et al., 2003; Lord et al., 2003) and aromatase inhibitors (Mitwally and Casper, 2001). However, most women still undergo the conventional treatment algorithm, involving three main treatment options. These treatments exhibit different burden, risks, costs and chances of ongoing pregnancy. Almost all patients suffering from WHO 2 anovulatory infertility receive the anti-estrogen clomiphene citrate (CC) as the first-line treatment modality. CC is relatively cheap, safe and has relatively few side-effects. Approximately 25% of patients show no ovulation, even after the maximal dose of CC (clomiphene-resistant anovulation; CRA) (Imani et al., 1998). Of the women who do ovulate, 40% do not achieve a pregnancy within 6 9 months (clomiphene failures) (Imani et al., 1999). Therefore, CC fails in 55% (= 25% + 75% 0.4) of patients, despite having been treated for half a year (Gorlitsky et al., 1978; Hammond et al., 1983; Eijkemans et al., 2003). The women who do not become pregnant with CC are subsequently treated with exogenous FSH, which is more burdensome, more expensive and also involves increased chances for complications. Intense ovarian response monitoring is required to avoid the risk of ovarian hyperstimulation syndrome (OHSS) or multiple gestation resulting from multiple rather than single dominant follicle development. Despite intensive monitoring during treatment, these complications can never be ruled out, and some patients may be more complication-prone than others. The percentage of ovulatory cycles following FSH ovulation induction is 70% and the cumulative pregnancy rate after seven cycles of treatment is at least 55% (van Santbrink et al., 1995; White et al., 1996; Mulders et al., 2003a). Multiple The Author Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved. 1

2 M.J.C.Eijkemans et al. pregnancy rates are usually reported to be 10% (Fauser and Van Heusden, 1997). The last step in the conventional treatment sequence for ovulatory disorders is IVF, representing the most burdensome and costly treatment. The risk of serious complications, however, may be less compared with FSH ovulation induction. In IVF, the chance of a multiple pregnancy is strongly determined by the number of embryos transferred. In case of imminent OHSS, all embryos may be cryopreserved, clearly reducing the risk of OHSS. Overall, IVF success rates in anovulatory women are similar compared with other indications for IVF (Mulders et al., 2003c), with reported pregnancy rates 20% per started cycle and 50 60% per started treatment (dependent on the number of cycles involved) (Kremer et al., 2002; Mulders et al., 2003c). It seems self-evident to employ the least burdensome and cheapest treatment as first-line, and only turn to more burdensome and expensive treatments in case of failure. However, it may be postulated that for some patient groups, a different treatment regimen from the standard one could offer an improved balance between pregnancy chances and costs. This approach would render ovulation induction more patient tailored. In order to explore this further, we need to know the costs of each treatment separately, as well as the chances of pregnancy, based on prognostic characteristics assessed at initial examination. For example, if we can predict beforehand that a patient has a small chance to conceive with CC (i.e. a high chance of being CRA or of clomiphene failure), it might be more efficient to start immediately with FSH. In women of more advanced reproductive age this means that precious time with declining pregnancy chances is saved. This also applies for IVF treatment only after unsuccessful FSH. In some patient groups CC and/or FSH treatment could be omitted, based on initial screening characteristics. Alternatively, patient groups might be identified (for instance with mild cycle disturbances) whose chances for spontaneous pregnancy are so high that treatment is not yet indicated. Materials and methods Previous prognostic studies In a series of previous studies, on a cohort of 240 consecutive patients with WHO 2 anovulatory infertility attending the outpatient clinic of Erasmus MC Rotterdam, we identified predictors of response to treatment. Detailed descriptions of patients and treatment regimen are given elsewhere (Eijkemans et al., 2003). Multivariable prediction models were presented for the chance of (persistent absence of) ovulation with CC (Imani et al., 1998; 2000), for the chance of pregnancy in CC-ovulatory patients (Imani et al., 1999) and the combination of these two (Imani et al., 2002). In addition, predictors could be identified for the chance of pregnancy or complications both with FSH (Mulders et al., 2003a) and with IVF (Mulders et al., 2003c), and for the overall prognosis of conventional ovulation induction (CC + FSH combined) (Eijkemans et al., 2003). Next to these patient cohort studies, a systematic review has been conducted concerning predictors of outcome of FSH ovulation induction (Mulders et al., 2003b). The chance of ovulation with CC was negatively influenced by an elevated free androgen index, a higher body mass index (BMI), amenorrhea instead of oligomenorrhea and larger mean ovarian volume (Imani et al., 1998). In case of an ovulatory response to CC, women of 2 older age and oligomenorrhea instead of amenorrhea had lower pregnancy chances (Imani et al., 1999). In the study concerning the prediction of pregnancy chances following FSH (Mulders et al., 2003a), a multivariable prediction model was constructed based on the initial serum insulin-like growth factor-i (IGF-I), testosterone levels and the woman s age. The response during the CC treatment (CRA versus clomiphene failure) was not associated with the cumulative chance of ongoing pregnancy (P = 0.6). In the Rotterdam cohort, only 26 patients started IVF after having received foregoing CC and FSH treatments without success (Mulders et al., 2003c). No predictors for pregnancy after IVF could be identified within this group of patients. We therefore based our calculations of the prognosis with IVF on the Templeton model (Templeton et al., 1996). That study did not find an effect of diagnostic category, and we therefore assumed that these predictions could also be applied to anovulatory patients. Female age was the only factor in this model that was relevant to the current study. Details are given in the Appendix. Prognostic groups An individualized treatment protocol based on all the characteristics mentioned would be too complex, and in case of IGF-I not feasible. Therefore, we restricted the number of predictors to be used in the cost-effectiveness analysis to the four most important ones: age, the severity of cycle disturbances (oligo- or amenorrhea), androgen levels expressed as normal or elevated (androstenedione 16.3 nmol/l or testosterone 3.2 nmol/l), and BMI (= weight in kg/height in m 2 ). Age and BMI were dichotomized, using a cut-off at 30 years and 27 kg/m 2 respectively, so that 16 ( ) different patient groups resulted. The patient groups are presented in Table I together with their frequency of occurrence in the Rotterdam cohort of 240 patients. Next, the prediction models for the chance of being CRA, the pregnancy chance with CC in non-cra patients and the pregnancy chance with FSH were re-estimated with the dichotomized predictors as defined here, using the data from Eijkemans et al., (2003). These data are the accumulation of the data from the previous studies and form the main empirical basis of the calculations in the present study. Shrinkage of regression coefficients was applied (van Houwelingen and Le Cessie, 1990) as had been done in the original models. Table I. Definition of the 16 patient groups, and prevalence among the 240 normogonadotropic anovulatory infertile patients (data from Eijkemans et al., 2003) Age Androgens a BMI Cycle duration % of patients <30 years Normal <27 Oligomenorrhea 18.8 Amenorrhea Oligomenorrhea 13.3 Amenorrhea 9.2 Elevated <27 Oligomenorrhea 10.0 Amenorrhea Oligomenorrhea 10.0 Amenorrhea years Normal <27 Oligomenorrhea 5.4 Amenorrhea Oligomenorrhea 2.9 Amenorrhea 0.8 Elevated <27 Oligomenorrhea 5.0 Amenorrhea Oligomenorrhea 3.3 Amenorrhea 0.8 a Elevated androgen levels are defined as androstenedione 16.3 nmol/l and/or testosterone 3.2 nmol/l.

3 Individualized cost-effective ovulation induction Costs of treatment For the three treatment options, the costs within the health-care sector were calculated from a hospital perspective. Prices were calculated from tariffs in the year The direct medical costs per cycle were determined by the use of resources in the data of the 240 patients that were used in the overall analysis of classical ovulation induction (Eijkemans et al., 2003). The total costs of treatment depend on the number of treatment cycles, which is related to prognosis: the higher the pregnancy chances, the sooner the patient will become pregnant, needing fewer cycles. Therefore, the costs per cycle were integrated into total costs together with the assessment of the prognosis in the final cost-effectiveness analysis. The costs of a CC treatment cycle were calculated separately for patients who had an ovulatory response to CC and patient who did not (CRA). The reason for doing this was that patients who were CRA (no ovulatory response after the maximal dose of 150 mg/day) had on average three progesterone assessments to detect ovulation, while patients who ovulated in the first cycle had only one. We further assumed that a pregnancy test was performed after every second cycle on average. In calculating the costs of a FSH ovulation induction cycle, we distinguished the costs of medication, recombinant FSH (rfsh), ultrasonographic monitoring during the cycle, injection of HCG to induce ovulation and a pregnancy test after each cycle. For the costs of IVF we used data from the literature, because our own data were too limited to provide reliable estimates. The costs of IVF in The Netherlands have been determined in the study by Goverde et al. (2000), using the price level of The results of that study were used, after updating to the price level of 2002 [the accumulated inflation over the period was 20.4% (source: Dutch central bureau of statistics index.htm)]. Comparison of treatment strategies The costs and ongoing pregnancy chances of the three treatments were calculated for the 16 prognostic groups, using the cost estimates from Table III and the chances from the prognostic models. Next, these results were incorporated in the treatment strategies that we wanted to compare, using direct calculations. We performed a cost effectiveness analysis of a strategy in which the CC step is skipped and treatment starts with FSH followed by IVF (FSH + IVF), and a strategy in which the FSH step is skipped and treatment after unsuccessful CC is IVF (CC + IVF). These two alternative treatment strategies will be compared with the reference strategy CC + FSH + IVF. The decision to stop further treatment, or even not commence with treatment, will be analysed by comparing the prognosis after treatment with the spontaneous pregnancy prognosis. Figure 1 displays the possible pathways that will be considered. The costs and the ongoing pregnancy chances of the various treatment strategies are compared with those of the reference strategy, and the incremental cost-effectiveness ratio (extra costs per extra ongoing pregnancy) was calculated. In case the costs are higher and at the same time the pregnancy chances are lower, the alternative strategy is unacceptable (it is dominated by the reference strategy). In the reverse case, with lower costs and higher ongoing pregnancy chances, we have a win-win situation (the alternative dominates the reference strategy). In the two other possible cases (higher costs together with higher ongoing pregnancy chances, or when both are lower) the costeffectiveness ratio determines whether the alternative is acceptable. If the cost-effectiveness ratio is below a threshold value of per extra ongoing pregnancy, we consider the alternative acceptable. We also explored a lower and a higher value for the threshold cost-effectiveness ratio of 5000 and , respectively. Figure 1. (A) Schematic representation of the applied conventional, linear ovulation induction protocol with the sequence of CC as firstline followed by FSH as second-line and IVF as final and last treatment. Patients proceed to the next treatment when no pregnancy is achieved. (B) Possible modifications of the treatment sequence, consisting of stopping further treatment or skipping a treatment step. For the comparison of the alternative treatment strategies with the strategy CC + FSH + IVF we had to make a few additional assumptions. The previously published flow chart (Eijkemans et al., 2003) showed that not all patients continue with FSH after unsuccessful CC treatment. We have seen that 81% of the CRA patients, but only 54% of the clomiphene failures, continued treatment with FSH. Furthermore, only 59% of the patients who were not pregnant after FSH continued with IVF. We will assume that in the alternative treatment strategies the same percentage of non-pregnant patients move to the next treatment step after unsuccessful CC and FSH, respectively. A further implicit, but important, assumption is that overall efficacy of individual treatment options would not change in relation to the order of treatment, e.g. we assumed that pregnancy chances with FSH are the same in a strategy in which FSH is the second-line treatment modality after unsuccessful CC and a strategy in which FSH would be given in the first-line. Apart from skipping treatment steps, we also consider stopping treatment in case of very poor prognosis, or not starting at all in cases 3

4 M.J.C.Eijkemans et al. where the spontaneous pregnancy chances are still high enough. These strategies may be evaluated by calculating the extra costs per extra ongoing pregnancy of continuing treatment compared with trying to become pregnant spontaneously. We make the simplifying assumption that trying to become pregnant spontaneously does not require extra costs. To calculate the spontaneous pregnancy chances, we used the Snick model (Snick et al., 1997). This model provides estimates dependent on age, duration of infertility, type of infertility (primary or secondary) and diagnosis, see the Appendix. Results Table II shows the results of the cost calculations per cycle for CC and FSH. The costs per cycle of CC are higher for CRA patients than for non-cra patients, mainly due to a higher requirement of progesterone assessment to detect ovulation. The average costs per cycle were 53. The costs of FSH ovulation induction were 1108 per cycle, which is more than 20 times higher than the costs per CC cycle. The estimate of the costs per IVF cycle was DFL 3350 (Goverde et al. 2000; price level of 1995), i.e (price level of 2002), 60% higher than the price per cycle of FSH. In Table III, the average costs per treatment and the costs per ongoing pregnancy are shown. With CC, a total of 103 ongoing pregnancies occurred, all in non-cra patients (Eijkemans et al., 2003), and the costs per ongoing pregnancy were 544. The subsequent FSH treatment produced 44 ongoing pregnancies in 84 patients (52%) (Eijkemans et al., 2003), against costs of 8584 per ongoing pregnancy. The costs per ongoing pregnancy with IVF were estimated to be Table II. Prices per unit and cost per cycle of CC and for FSH treatment Price per unit ( ) Units per cycle Costs per cycle ( ) a Cost volumes per cycle are based on the data described in Eijkemans et al. (2003). Non-CRA CC a Outpatient clinic visit 75 (15 min) 2, before and after 1st cycle Clomid 4.62/10 tablets of 50 mg 5 days, dose 50 mg/day increasing to max mg/day in 3rd cycle Progesterone assessment st cycle and after increase of daily dose Pregnancy test 10 1 per two cycles 5.00 Total FSH a First outpatient clinic visit 75 (15 min) 1 before 1st cycle rfsh 33.54/75 IU Injection HCG 4.63/5000 IU Only in non-cancelled cycles 3.80 Lesson self-injection 25 1 before 1st cycle 6.25 Outpatient clinic visit 75 (15 min) Ultrasonography Pregnancy test 10 Every cycle 10 Total 1108 Table III. Costs per ongoing pregnancy Not pregnant Pregnant Average Non-CRA CRA Non-CRA CC a n = 80 n = 57 n = 103 n = 240 Number of cycles Costs per cycle ( ) (see Table II) Costs per treatment ( ) Costs per ongoing pregnancy ( ) 544 FSH a n = 40 n = 44 n = 84 Number of cycles Costs per cycle ( ) (see Table II) Costs per treatment ( ) Costs per ongoing pregnancy ( ) 8584 IVF Number of cycles b Costs per cycle ( ) c Costs per treatment ( ) Costs per ongoing pregnancy ( ) 7686 CRA a Data on CC and FSH are based on Eijkemans et al., (2003) or additionally calculated from those data. b Data on number of IVF cycles from Kremer et al., (2002). c Costs per cycle of IVF from Goverde et al. (2000), updated to the price level of

5 Individualized cost-effective ovulation induction Table IV. Costs and effects per patient of two alternative treatment strategies (FSH + IVF and CC + IVF) compared to the reference strategy CC + FSH + IVF Age Androgens BMI Cycle duration Reference strategy Alternative 1 Alternative 2 CC + FSH + IVF FSH + IVF CC + IVF Pregnancy chance (%) Costs ( ) Costs/preg. ( ) Difference from reference strategy Difference from reference strategy chance (%) costs ( ) C/ E ratio a chance (%) costs ( ) C/ E ratio a <30 years Normal <27 Oligomenorrhea b Amenorrhea Oligomenorrhea Amenorrhea Elevated <27 Oligomenorrhea Amenorrhea Oligomenorrhea Amenorrhea years Normal <27 Oligomenorrhea Amenorrhea Oligomenorrhea Amenorrhea Elevated <27 Oligomenorrhea Amenorrhea Oligomenorrhea Amenorrhea Average a C/ E ratio is the cost-effectiveness ratio: the costs per ongoing pregnancy of the respective alternative treatment strategy compared with the reference strategy CC + FSH + IVF. b With, the cases are indicated in which the alternative strategy is dominated by the reference strategy (i.e. the alternative strategy has both higher costs as well as lower ongoing pregnancy chances than the reference strategy). Owing to rounding, the cost-effectiveness ratios in the table may differ from those calculated directly from the costs and pregnancy chances in the table. 5

6 M.J.C.Eijkemans et al. The results of the individualized cost-effectiveness analysis are presented in Table IV. The extra costs (or savings) of the alternative strategies compared with the reference strategies are related to the gain (or loss) in ongoing pregnancy chance. The treatment strategy FSH + IVF (i.e. skip CC) is either dominated by the reference strategy or produces more ongoing pregnancies against costs per ongoing pregnancy that vary between and , which is far above the threshold value of Therefore, this strategy would not be acceptable for any patient group. Under the lower ( 5000) as well as under the higher ( ) threshold value, these conclusions remain the same. The other alternative treatment strategy, CC + IVF (i.e. skip FSH), produces less pregnancies but also lower costs than the reference strategy, for all patient groups. In this case, the costeffectiveness ratio represents the cost savings per ongoing pregnancy that is missed by performing the inferior strategy CC + IVF instead of the superior reference strategy. This is the same as saying that the reference strategy produces more ongoing pregnancies against costs per pregnancy determined by this cost effectiveness ratio. The ratio is above the threshold for acceptability of , except for women under 30 years of age who have normal androgen levels. For these women, FSH after CC gives extra ongoing pregnancies against extra costs of This concerns 40% of the group of patients who are not pregnant after CC and who would continue treatment. For all other women, FSH should be skipped. Under the lower threshold value ( 5000), FSH is no longer acceptable for any patient group. On the contrary, under the higher alternative threshold value ( ) FSH becomes acceptable for all women below 30 years of age. Above age 30 years, women with normal androgen levels (15% of women) are just above this threshold value, while for women with elevated androgen levels (14% of women) the cost-effectiveness ratio exceeds per ongoing pregnancy. In addition, three alternative strategies were evaluated in which no further treatment was started at some point in the sequence (data are not shown in Table IV). The extra costs per extra ongoing pregnancy of the reference strategy (CC + FSH + IVF) compared with no treatment at all vary between 2490 and The cost-effectiveness ratio of continuing with FSH + IVF after CC compared with trying to become pregnant spontaneously varied between 9830 and per ongoing pregnancy, assuming that spontaneous pregnancy chances are unaffected after unsuccessful CC treatment. Finally, the costeffectiveness ratio of continuing with IVF after unsuccessful CC + FSH compared with trying to become pregnant spontaneously varies between 7690 and The highest values suggest that indeed there are patient groups for whom continuing treatment is not efficient under the threshold value of per extra ongoing pregnancy. 6 Discussion A cost-effectiveness analysis was performed comparing several alternative treatment strategies within the framework of the conventional linear ovulation induction reference strategy CC + FSH + IVF, for 16 different patient groups. The alternative strategies were FSH + IVF (i.e. skip CC), CC + IVF (i.e. skip FSH), CC (i.e. stop after unsuccessful CC), CC + FSH (stop after unsuccessful FSH) and finally no treatment at all. In the latter three strategies it was assumed that patients would try to become pregnant spontaneously. This analysis was performed with the assumption that overall efficacy of individual treatment options would not change in relation to the order of treatment. It seems likely that the inclusion of more recently introduced ovulation induction strategies such as insulin-sensitizing agents or laparoscopic ovarian cautery would give rise to different outcomes. The current analysis, however, is based on a long-term follow-up study in normogonadotropic anovulatory infertile women performed in our unit, in which individual patient characteristics associated with ovulation induction outcomes could be identified. The current standard treatment protocol (start with CC and proceed with FSH followed by IVF, in case no pregnancy has been achieved) is acceptable with a maximal acceptable threshold for the costs per ongoing pregnancy of , for women below 30 years of age presenting with normal androgen levels. A robust finding was further that for women above 30 years of age with elevated androgen levels, FSH treatment could be skipped: only a small increase in pregnancy chances is observed against costs of over per extra ongoing pregnancy in these women. For the remaining women (age <30 years with elevated androgens or age >30 years with normal androgens), the costs per extra ongoing pregnancy achieved by performing FSH are slightly above The strong association between outcome of treatment and obesity (for CC from our own analysis: Imani et al., 1998; 2002; for gonadotrophins mainly from the literature: Mulders et al., 2003b) implies that chances could be improved by reducing overweight (Norman et al., 2004). This concerns 30 40% of the WHO 2 group. Three strategies in which either no treatment was started or further treatment was stopped after CC or FSH were compared with the reference strategy, assuming that patients who stopped would have spontaneous pregnancy chances equal to patients who were never treated. The cost-effectiveness ratio of continuing treatment compared with stopping reached values of , which is above our limit for acceptance. This would imply that continuing treatment would not be worthwhile given the costs for some patient groups. However, we note that it is likely that the spontaneous pregnancy chances of patients who did not become pregnant with CC or FSH are lower than assumed here (among other things, because the couples with better chances will already have become pregnant). Therefore, the cost-effectiveness ratios are probably lower than we have calculated and we cannot conclude for any patient group that stopping treatment is efficient. The literature with regard to the economic evaluation of treatment in anovulatory infertility has been scarce so far. In 1999, Fridstrom et al. (1999) performed a cost-effectiveness analysis based on the data of a small randomized controlled trial comparing ovulation induction with IVF, in clomipheneresistant polycystic ovary syndrome. They concluded that IVF seemed to be cost-effective compared with ovulation induction. In our analysis, this comparison was not made directly, but could be made by comparing the strategies CC + FSH (i.e.

7 Individualized cost-effective ovulation induction stop after FSH) with CC + IVF (i.e. skip FSH). It appears that CC + IVF gives rise to more pregnancies against acceptable cost per extra pregnancy (less than ). However, our analysis also shows that CC + FSH + IVF is the most efficient strategy, with the exceptions described above (see Table IV). George et al. (2003) calculated the costs per pregnancy in a trial comparing metformin followed by CC (metformin + CC) with ovulation induction using exogenous gonadotrophins. The costs per pregnancy were US$71 for the metformin strategy, which is lower than our estimates for CC. Furthermore, their estimated US$277 per pregnancy for exogenous gonadotrophins is much lower than our estimate for FSH, differences for which we have no explanation. Our analysis is restricted to what may currently be referred to as the conventional treatment sequence of ovulation induction and IVF. Modern treatment options in WHO 2 patients include, among others, laparoscopic ovarian surgery. A recent study has shown that this treatment is comparable to FSH in costs per pregnancy (van Wely et al., 2004). The authors imply that the lower number of twin pregnancies may be decisive in favour of the electrocautery strategy, when the costs of deliveries are included. In their study, the electrocautery strategy had only one multiple in 56 ongoing pregnancies (1.8%), whereas the FSH strategy produced nine multiples in 57 ongoing pregnancies (15.8%) (Bayram et al., 2004). In our own data, focusing on live birth as primary endpoint, FSH generated six multiples in 44 ongoing pregnancies (13.6%) (Eijkemans et al., 2003), which is not significantly different. We did not include the costs made during pregnancy or delivery. The balance between FSH and IVF treatment would certainly shift in favour of FSH with the inclusion of these costs, given that currently in Europe 25 30% of all IVF pregnancies are multiples (Nyboe Andersen et al., 2004). On the other hand, IVF is a more controlled treatment than FSH, and a tendency is present towards single embryo transfer in IVF, which will almost eliminate multiple pregnancies after IVF. Including these costs would require us to make additional assumptions regarding the effectiveness measurement: it is unclear how a twin pregnancy should be weighted compared with a singleton pregnancy. Twins will have a higher incidence of pre-term birth, more complications, higher neonatal costs and a higher incidence of birth defects. Yet, with one treatment, two children are produced, which might be regarded as two effectiveness units. There is no self-evident way to assess the outcome in this context, so that sensitivity or scenario analyses would be required to show the impact of different assumptions. We have not performed such analyses, since the degree of complexity of the current analysis is already high. In the comparison between FSH and IVF, multiple pregnancies may in the future be a decisive argument against FSH. Our estimates of the direct medical costs of FSH were much lower compared with recently published data (van Wely et al., 2004) ( 4497 versus 5418). In our data, patients received less ampoules of rfsh per cycle (17.7 versus 26.9) and slightly more ultrasonography (4.4 versus 3.8 per cycle). To explore the impact of these differences, we repeated our analysis with the published direct medical costs of FSH (van Wely et al., 2004) instead of our own. The treatment strategy CC + IVF (skip FSH) now became optimal for all 16 patient profiles, with extra costs per extra pregnancy of CC + FSH + IVF between and Only under a critical value of the cost-effectiveness ratio of per extra pregnancy would CC + FSH + IVF be indicated, for women <30 years of age with normal serum androgen levels. Apparently, the direct medical costs of FSH play a decisive role in this cost-effectiveness analysis. So far, none of the prediction models has been externally validated. Nevertheless, we used the models in a cost-effectiveness analysis to formulate recommendations for the treatment protocol of WHO 2 anovulatory infertile patients. Because only internal validity has been assessed, we may only rely on the results of this cost-effectiveness analysis for patients who are similar and from a similar setting. Future studies should be designed to perform external validation of the prediction models in independent patient groups. We conclude that the use of individual patient information may lead to a more patient-tailored and more efficient treatment of WHO 2 anovulatory patients. The classical sequence of treatment is efficient for young patients with normal serum androgen levels, but the second-line FSH treatment step could be skipped in older patients with elevated serum androgen levels. Depending on assumptions concerning the costs of FSH, the indication to skip FSH could be widened to younger patients with elevated serum androgen levels. Further health economic studies in the area of ovulation induction should also include alternative treatment strategies, and costs related to multiple gestations should be taken into consideration. References Bayram N, van Wely M, Kaaijk EM, Bossuyt PM and van der Veen F (2004) Using an electrocautery strategy or recombinant follicle stimulating hormone to induce ovulation in polycystic ovary syndrome: randomised controlled trial. BMJ, 328, 192. Cohen J (1996) Laparoscopic procedures for treatment of infertility related to polycystic ovarian syndrome. Hum Reprod Update 2, Eijkemans MJ, Imani B, Mulders AG, Habbema JD and Fauser BC (2003) High singleton live birth rate following classical ovulation induction in normogonadotrophic anovulatory infertility (WHO 2). Hum Reprod 18, ESHRE Capri Workshop Group (1995) Anovulatory infertility. Hum Reprod 10, Farquhar C, Vandekerckhove P and Lilford R (2001) Laparoscopic drilling by diathermy or laser for ovulation induction in anovulatory polycystic ovary syndrome. Cochrane Database Syst Rev 4 CD Fauser BC and Van Heusden AM (1997) Manipulation of human ovarian function: physiological concepts and clinical consequences. Endocr Rev 18, Fridstrom M, Sjoblom P, Granberg M and Hillensjo T (1999) A cost comparison of infertility treatment for clomiphene resistant polycystic ovary syndrome. Acta Obstet Gynecol Scand 78, George SS, George K, Irwin C, Job V, Selvakumar R, Jeyaseelan V and Seshadri MS (2003) Sequential treatment of metformin and clomiphene citrate in clomiphene-resistant women with polycystic ovary syndrome: a randomized, controlled trial. Hum Reprod 18, Gorlitsky GA, Kase NG and Speroff L (1978) Ovulation and pregnancy rates with clomiphene citrate. Obstet Gynecol 51, Goverde AJ, McDonnell J, Vermeiden JP, Schats R, Rutten FF and Schoemaker J (2000) Intrauterine insemination or in-vitro fertilisation in idiopathic subfertility and male subfertility: a randomised trial and cost-effectiveness analysis. Lancet 355, Hammond MG, Halme JK and Talbert LM (1983) Factors affecting the pregnancy rate in clomiphene citrate induction of ovulation. Obstet Gynecol 62, Harborne L, Fleming R, Lyall H, Norman J and Sattar N (2003) Descriptive review of the evidence for the use of metformin in polycystic ovary syndrome. Lancet, 361,

8 M.J.C.Eijkemans et al. Imani B, Eijkemans MJ, te Velde ER, Habbema JD and Fauser BC (1998) Predictors of patients remaining anovulatory during clomiphene citrate induction of ovulation in normogonadotropic oligoamenorrheic infertility. J Clin Endocrinol Metab 83, Imani B, Eijkemans MJ, te Velde ER, Habbema JD and Fauser BC (1999) Predictors of chances to conceive in ovulatory patients during clomiphene citrate induction of ovulation in normogonadotropic oligoamenorrheic infertility. J Clin Endocrinol Metab 84, Imani B, Eijkemans MJ, de Jong FH, Payne NN, Bouchard P, Giudice LC and Fauser BC (2000) Free androgen index and leptin are the most prominent endocrine predictors of ovarian response during clomiphene citrate induction of ovulation in normogonadotropic oligoamenorrheic infertility. J Clin Endocrinol Metab 85, Imani B, Eijkemans MJ, te Velde ER, Habbema JD and Fauser BC (2002) A nomogram to predict the probability of live birth after clomiphene citrate induction of ovulation in normogonadotropic oligoamenorrheic infertility. Fertil Steril 77, Kremer JA, Beekhuizen W, Bots RS, Braat DD, van Dop PA, Jansen CA, Land JA, Laven JS, Leerentveld RA, Naaktgeboren N et al. (2002) Resultaten van in-vitro fertilisatie in Nederland, Ned Tijdschr Geneeskd 146, Lord JM, Flight IH and Norman RJ (2003) Metformin in polycystic ovary syndrome: systematic review and meta-analysis. BMJ, 327, Mitwally MF and Casper RF (2001) Use of an aromatase inhibitor for induction of ovulation in patients with an inadequate response to clomiphene citrate. Fertil Steril, 75, Mulders AG, Eijkemans MJ, Imani B and Fauser BC (2003a) Prediction of chances for success or complications in gonadotrophin ovulation induction in normogonadotrophic anovulatory infertility. Reprod Biomed Online, 7, Mulders AG, Laven JS, Eijkemans MJ, Hughes EG and Fauser BC (2003b) Patient predictors for outcome of gonadotrophin ovulation induction in women with normogonadotrophic anovulatory infertility: a meta-analysis. Hum Reprod Update 9, Mulders AG, Laven JS, Imani B, Eijkemans MJ and Fauser BC (2003c) IVF outcome in anovulatory infertility (WHO group 2) including polycystic ovary syndrome following previous unsuccessful ovulation induction. Reprod Biomed Online 7, Nestler JE, Stovall D, Akhter N, Iuorno MJ and Jakubowicz DJ (2002) Strategies for the use of insulin-sensitizing drugs to treat infertility in women with polycystic ovary syndrome. Fertil Steril 77, Norman RJ, Noakes M, Wu R, Davies MJ, Moran L and Wang JX (2004) Improving reproductive performance in overweight/obese women with effective weight management. Hum Reprod Update 10, Nyboe Andersen A, Gianaroli L and Nygren KG (2004) Assisted reproductive technology in Europe, Results generated from European registers by ESHRE. Hum Reprod 19, Snick HK, Snick TS, Evers JL and Collins JA (1997) The spontaneous pregnancy prognosis in untreated subfertile couples: the Walcheren primary care study. Hum Reprod 12, Templeton A, Morris JK and Parslow W (1996) Factors that affect outcome of in-vitro fertilisation treatment. Lancet 348, van Houwelingen JC and Le Cessie S (1990) Predictive value of statistical models. Stat Med 9, van Santbrink EJ, Hop WC, van Dessel TJ, de Jong FH and Fauser BC (1995) Decremental follicle-stimulating hormone and dominant follicle development during the normal menstrual cycle. Fertil Steril 64, van Wely M, Bayram N, van der Veen F and Bossuyt PM (2004) An economic comparison of a laparoscopic electrocautery strategy and ovulation induction with recombinant FSH in women with clomiphene citrate-resistant polycystic ovary syndrome. Hum Reprod 19, White DM, Polson DW, Kiddy D, Sagle P, Watson H, Gilling-Smith C, Hamilton-Fairley D and Franks S (1996) Induction of ovulation with low-dose gonadotropins in polycystic ovary syndrome: an analysis of 109 pregnancies in 225 women. J Clin Endocrinol Metab 81, Submitted on December 2, 2004; Resubmitted on May 13, 2005; accepted on May 17, 2005 Appendix Prediction of outcome Pregnancy chances with IVF. The Templeton model (Templeton et al., 1996) was used to predict the chances of ongoing pregnancy with IVF in the two age groups considered. This model was developed on the data of patients from the period , and the overall livebirth rate in the first cycle was 14%. We adjusted the model with recently published data from the national Dutch IVF registration (Kremer et al., 2002), also available from the internet at dynamic/voorlichting.asp?maingrp=vl.ivfresultaten&statgrp=vl. ivfresultaten.static, which had an ongoing pregnancy rate per cycle of 19% during the period We calculated the chances of ongoing pregnancy with IVF for ages under and above 30 years as 24% and 19%, respectively, per treatment cycle, assuming that the duration of infertility was 1 3 years and that the couple had had no previous pregnancy or live birth, either with or without IVF. With on average 2.1 cycles performed (Kremer et al., 2002), the overall pregnancy chances with IVF are 50% and 40%, respectively. Spontaneous pregnancy chances. We used the Snick model (Snick et al., 1997) to predict the spontaneous pregnancy chances for our WHO 2 patients. For women with an ovulation disorder and age under 30 years old, the model predicts a spontaneous pregnancy chance within 1 year of 20%, assuming shorter than 2 years duration of infertility and that infertility is primary. Since amenorrheic women do not ovulate, and therefore have a zero chance of spontaneous pregnancy, oligomenorrheic women must have a pregnancy chance that is above the average for women with an ovulation disorder. In our group of 240 patients, 78% had oligomenorrhea. They alone account for the group total of 20% pregnancies, and therefore their pregnancy chance is 0.20/0.78 = 26%. For women of 30 years or older the calculation is similar: the Snick model predicts 15% for the whole group of women with an ovulation disorder. In case of oligomenorrhea, the prognosis is then equal to 0.15/0.78 = 19%. 8