Materials and methods

Transcription

1 RBMOnline - Vol 19. No Reproductive BioMedicine Online; on web 22 September 2009 Article Low-dose HCG may improve pregnancy rates and lower OHSS in antagonist cycles: a meta-analysis Ioannis Kosmas MD, MSc, PhD, completed his specialty degree in 2002 at the University of Ioannina, Medical School, Greece. After that he spent some time at the Center of Minimal Invasive Surgery at Royal Free Hospital, London (2003) and at the Center of Reproductive Medicine at the Dutch-speaking Free University of Brussels ( ) where he obtained his subspecialty in Reproductive Medicine. His special interest includes meta-analysis and decision sciences, pharmaco-economics and molecular imaging in embryo implantation. Dr Ioannis Kosmas IP Kosmas 1,3, K Zikopoulos 1, I Georgiou 1, E Paraskevaidis 1, C Blockeel 2, H Tournaye 2, J Van der Elst 2, P Devroey 2 1 Centre for Reproductive Medicine, Department of Obstetrics and Gynecology, Ioannina University School of Medicine, Ioannina, Greece; 2 Centre for Reproductive Medicine, University Hospital, Vrije Univestiteit Brussels (Free University of Brussels), Belgium 3 Correspondence: kosmasioannis@gmail.com Abstract Human chorionic gonadotrophin (HCG) may substitute FSH to complete follicular growth in IVF cycles. This may be useful in the prevention of ovarian hyperstimulation syndrome. Relevant studies were identified on Medline. To evaluate outcomes, a meta-analysis of low-dose HCG-supplemented IVF cycles versus non-supplemented ones was performed with data from 435 patients undergoing IVF who were administered low-dose HCG in various agonist and antagonist protocols and from 597 conservatively treated patients who served, as control subjects. Using these published data, a decision analysis evaluated four different management strategies. Effectiveness and economic outcomes were assessed by FSH consumption, clinical pregnancy and incremental cost-effectiveness ratios. Clinical pregnancy and ovarian hyperstimulation were the main outcome measures. Nine trials published in were included. From the prospective studies, in the gonadotrophin-releasing hormone antagonist group, a trend for significance in clinical pregnancy rate was evident (odds ratio [OR], 1.54; 95% confidence interval [CI], ). Ovarian hyperstimulation was less significant in the antagonist low-dose HCG protocol compared with the non-supplemented agonist protocol (OR 0.30; 95% CI ). Less FSH was consumed in the low-dose HCG group but this difference was not statistically significant. Low-dose HCG supplementation may improve pregnancy rates in antagonist protocols. Overall, low-dose HCG-supplemented protocols are a cost-effective strategy. Keywords: FSH, IVF, low-dose HCG, meta-analysis, pregnancy rates Introduction Ovarian stimulation with recombinant follicle-stimulating hormone (rfsh) or human menopausal gonodatrophin is a major part of IVF. FSH acts in an anti-apoptotic way (Ruvolo et al., 2007). Other factors like LH/human chorionic gonadotrophin (HCG), growth hormone, insulin-like growth factor-i and EGF also play the same role (Chun et al., 1996). From animal studies, it has been found that granulosa cells increase their dependence on LH/HCG in the late follicular phase (Mihm et al., 2006). In a clinical study, HCG was proposed as a factor that could complete follicle growth by substituting for FSH (Filicori et al., 1999). Since then, other prospective and retrospective studies have been performed with similar pregnancy rates. From these studies, rfsh replacement by low-dose HCG has emerged as a potential strategy to prevent OHSS because it reduces small, developing ovarian follicles (Filicori et al., 2002) and as a treatment regimen for women with hypogonadotrophic hypogonadism (Filicori et al., 1999). 619 Ó 2009 Published by Reproductive Healthcare Ltd, Duck End Farm, Dry Drayton, Cambridge CB23 8DB, UK

2 620 There is variability on the design of these studies. They have been performed in GnRH antagonist protocols (Koichi et al., 2006; Serafini et al., 2006), and long and short agonist protocols (Berkkanoglu et al., 2007). The main outcomes of the studies are hormonal response after stimulation, embryological characteristics, pregnancy outcome, endometrial thickness and incidence of OHSS (Koichi et al., 2006). Another difference is the ratio of low-dose HCG substitution. In certain studies, there is major substitution of FSH (Filicori et al., 2005b), while in other studies only a small quantity is added with FSH remaining unchanged during stimulation (Berkkanoglu et al., 2007). All studies report similar results between the two groups while in parallel show reduced FSH needs. This decrease is more significant in the studies with a major replacement of FSH (Filicori et al., 2005b; Koichi et al., 2006). This study was undertaken to summarize evidence for clinical pregnancy rates, incidence of OHSS, FSH reduction and endocrinological features when administering low-dose HCG during down-regulation. Different down-regulation protocols were examined and the clinical effect after different levels of FSH substitution were evaluated. A meta-analysis, a decision analysis and a cost-effectiveness analysis of alternative ovarian stimulation strategies for IVF, including the low-dose HCG protocols, were performed to define the extent of cost savings by FSH decrease and to suggest a cost-effective choice. Materials and methods Identification of eligible relevant studies and data extraction Medline searches (until September 2007) with various combinations of terms (low-dose HCG administration, GnRH agonist and antagonists, IVF, gonadotrophin administration) and pregnancy rates were used. The search was completed after examination of the references of recovered papers and review articles. Studies that evaluated the comparison of FSH replacement by low-dose HCG with no supplementation in GnRH agonist and antagonist protocols were included. Although unpublished studies cannot be adequately evaluated for their design and quality, the analysis included meeting proceedings to increase the sample size (Mendes et al., 2005; Davies et al., 2006; Dzik et al., 2006). All studies were included, regardless of the size but whenever possible only prospective controlled studies were synthesized. No language limit was set. For each study, information was obtained on authors, journal and year of publication, country and years of study enrollment, study design and eligibility (inclusion/exclusion) criteria and number of participating centres. Furthermore, withdrawals, power analysis and reports of funding were also mentioned. Finally, the mean age of the subjects, the number of cases and controls, as well as the number of pregnancies per group, were mentioned. Subgroup analysis was performed for the GnRH down-regulation scheme, low-dose HCG substitution and incidence of OHSS. In addition, analysis that included the retrospective trials was performed. Data extraction was performed independently by two investigators and conflicts were solved after discussion. Main outcomes The primary outcome for low-dose HCG replacement was clinical pregnancy rate. As secondary outcomes, incidence of OHSS and the observed endocrinology were considered whenever mentioned. Cost analysis was based on the mean values of FSH consumption for each protocol. Meta-analysis The odds ratios with 95% confidence intervals between the low-dose HCG replacement group and the control group for each study were calculated and the results were combined using the fixed (for prospective studies) and the random effects model. Heterogeneity was assessed with I 2 statistics. The main analysis includes all studies. Subgroup analyses were also performed based on the GnRH downregulation protocol and FSH replacement by low-dose HCG. Analyses were performed by RevMan Analyses ((2003); The Cochrane Information Management System) and using the Statistics Package for Social Sciences (SPSS 15.0; SPSS, Chicago, USA). Decision analysis and cost-effectiveness analysis Using country-specific data, the analysis compared the clinical pregnancy rates, incidence of OHSS and the costs associated with the current stimulation practices. The comparative performance of alternate strategies was described using the incremental cost-effectiveness ratio (ICER), which is defined as the additional cost of a specific strategy divided by its additional clinical benefit compared with the next least expensive strategy. Incidence of OHSS is not associated with pregnancy outcome and is related only to the protocol, thus the OHSS cost was calculated per protocol. Decision analysis and cost analysis were performed with TreeAge Pro 7 software (TreeAge Software, Williamstown, USA). Cost effectiveness was calculated by taking the cost in Euros and dividing it by the effectiveness. Effectiveness is considered as 1 if clinical pregnancy is reported and as 0.5 when stimulation is completed but no clinical pregnancy is reported. Efficiency was measured by the ICER, which is defined as the ratio of the change in costs of a therapeutic intervention (compared with the alternative, using the best available alternative treatment) to the change in effects of the intervention: ICER = IC/IE, where incremental efficiency is health benefit per unit of cost (IE) and incremental cost (IC) is the cost per cycle in alternating arm minus cost per patient against a standard arm.

3 Another more complicated definition of the ICER is [(C 1 C 0 )/(E 1 E 0 )], where 1 denotes the intervention under study and 0 the alternative with which it is compared, C 1 and C 0 are the net present values of costs accrued when the intervention and its alternative are used, and E 1 and E 0 are their respective health outcomes. Thus, C 1 C 0 is the incremental cost and E 1 E 0 is the incremental effectiveness of the intervention over the alternative. Results Eligible studies A total of 28 abstracts were retrieved and screened. From these, thirteen articles that attempted to compare low-dose HCG administration in various protocols with FSH IVF cycles were retrieved for full review. One abstract was excluded because it did not mention the sample size included in the trial (Frattarelli et al., 2007). Another abstract was excluded because it was a preliminary report of a later study (Serafini et al., 2006). The third excluded study was a review (Filicori et al., 2005a) (Figure 1). The studies included in the meta-analysis are summarized in Table 1. Out of nine included studies (Filicori et al., 2005a; Mendes et al., 2005; Davies, 2006; Dzik et al., 2006; Koichi et al., 2006; Serafini et al., 2006; Berkkanoglu et al., 2007; Gomes et al., 2007; Van Horne et al., 2007), three recruited women in Brazil, three recruited in USA and one each recruited in Italy, Japan and Turkey. The only multicentre study covered Brazilian centres. The mean age of participating women ranged between 30 and 40 years of age. None of the studies mentioned ethnic origin of recruited patients. The women participating in the studies were enrolled from 2002 until Four out of nine studies mentioned inclusion and exclusion criteria. Recruitment patterns and eligibility for participants varied, but typically, women with unexplained infer- Figure 1. Flow diagram for meta-analysis: abstract identification and selection. HCG = human chorionic gonadotrophin; RCT = randomized controlled trial. 621

4 Table 1. Studies included in the meta-analysis. Reference Study type Protocol Gonadotrophin used Dose (IU) HCG dose (IU) Intervention Mendes et al., 2005 Filicori et al., 2005a Koichi et al., 2006 Serafini et al., 2006 Davies et al., 2006 Dzik et al., 2006 Gomes et al., 2007 Berkkanoglu et al., 2007 Van Horne et al., 2007 Prospective Prospective randomized Long agonistbromocriptine rfsh-hmg Not 200 Replacement mentioned Long agonist rfsh-hmg Replacement Prospective randomized Long agonistantagonist uhfsh Replacement Prospective Long agonistantagonist rfsh Replacement randomized Prospective Not rfsh Addition randomized mentioned (follitropin a) Retrospective Long agonist rfsh Replacement Prospective Long agonist rfsh-hmg Replacement Prospective Short flare-up randomized agonist Retrospective GnRH antagonist rfsh (follitropin a) Addition rfsh Started (follitropin a concomitantly b) GnRH = gonadotrophin-releasing hormone; HCG = human chorionic gonadotrophin; HMG = human menopausal gonadotrophin; rfsh = recombinant FSH; uhfsh = urinary human FSH. 622 tility, women from couples with mild and severe male infertility and women with tubal factor were eligible. Endogenous FSH concentrations during early follicular phase were varied. Five studies described cancellations and withdrawals. Only three studies mentioned power calculations. More than three outcomes were recorded in all studies. In all of them, interventions were equally experimental. None of the studies reported funding. Meta-analysis Eleven comparisons out of nine studies were performed for clinical pregnancy rates (odds ratio [OR] 1.14; 95% confidence interval [CI] ; random effects model). When only the prospective studies were included, there was no difference in pregnancy rates (OR 1.14; 95% CI ; random effects model). When controlling for OHSS, no difference was observed between the two groups (OR 0.48; 95% CI ; random effects model). Subgroup analyses based on administration protocol Long protocol Four studies (Filicori et al., 2005a; Mendes et al., 2005; Koichi et al., 2006; Serafini et al., 2006) using long agonist protocols included 204 patients. Three of the studies were prospective. There was no significant difference in clinical pregnancy rates when all studies were included (OR 1.15; 95% CI ) or for only the prospective studies (OR 1.43; 95% CI ; random effects model). Three studies (Filicori et al., 2005a; Koichi et al., 2006; Serafini et al., 2006) reported OHSS. When comparing low-dose HCG replacement with no replacement in long protocols there was no difference in the incidence of OHSS (OR 0.70; 95% CI ; random effects model). Short protocol Only one study (Berkkanoglu et al., 2007) reported on supplemented and non-supplemented cycles with the short protocol. There was no significant difference in pregnancy rates between the comparison (OR 0.70; 95% CI ). This study did not report on OHSS. Antagonist protocol Three studies (Koichi et al., 2006; Serafini et al., 2006; Van Horne et al., 2007) including 504 patients were performed in antagonist protocols. Two of them, were prospective. When comparing the prospective ones, there was a trend for a significant difference in clinical pregnancy rates favouring the low-dose HCG-supplemented antagonist protocol (OR 1.54; 95% CI ; fixed effects model; Figure 2). These two studies also reported on OHSS. There was no significant difference in the incidence of OHSS between a supplemented and non-supplemented antagonist protocol (OR 0.69; 95% CI ; random effects model).

5 Figure 2. Comparison of low-dose human chorionic gonadotrophin (HCG) supplemented antagonist cycles versus nonsupplemented cycles and the odds for clinical pregnancy. CI = confidence interval; OR = odds ratio. Subgroup analyses based on FSH substitution by low-dose HCG Low-dose HCG-antagonist protocol versus nonsupplemented agonist protocol Two comparisons, (Koichi et al., 2006; Serafini et al., 2006) exist for the supplemented antagonist protocol with non-supplemented long agonist protocol. There was no significant difference in clinical pregnancy rates (OR 0.87; 95% CI ; random effects model), but when comparing for incidence of OHSS there was a significant difference favouring the low-dose HCGantagonist protocol (OR 0.30; 95% CI ; fixed effects model; Figure 3). From this comparison, similar pregnancy rates were achieved while ovarian hyperstimulation was minimized favouring the supplemented antagonist protocol. Major substitution of FSH by low-dose HCG When comparing clinical pregnancy rates, a significant difference favouring the study group was observed in the prospective studies (Filicori et al., 2005a; Mendes et al., 2005; Koichi et al., 2006; Serafini et al., 2006; Gomes et al., 2007) in which FSH was replaced to a major extent by low-dose HCG. (OR 1.51; 95% CI ; fixed effects model; Figure 4). By excluding the study that added bromocriptine with the low-dose HCG protocol (Mendes et al., 2005), no change was found in the direction of the dichotomous variables (OR 1.59; 95% CI ). Minor substitution of FSH by low-dose HCG There was no significant difference in clinical pregnancy rates in the prospective studies (Davies et al., 2006; Berkkanoglu et al., 2007) where addition of low-dose HCG to the scheme or minor substitution, took place (OR 0.83; 95% CI ; fixed effects model). Subgroup analyses based on endocrinology of the supplemented cycles Oestradiol on the day of the HCG administration Seven studies (Filicori et al., 2005a; Mendes et al., 2005; Koichi et al., 2006; Serafini et al., 2006; Berkkanoglu et al., 2007; Gomes et al., 2007; Van Horne et al., 2007) mentioned comparisons of oestradiol concentrations on the day of HCG administration between different protocols (Table 2). These concentrations were higher in the low-dose HCG group in all protocols. There was no significant difference in these concentrations either between groups or between the different supplemented protocols and between the different non-supplemented protocols (within groups), (one-way analysis of variance (ANOVA)). The short protocol group was excluded from the analysis because only one study (Berkkanoglu et al., 2007) existed. Figure 3. Comparison of low-dose human chorionic gonadotrophin (HCG) supplemented antagonist cycles versus nonsupplemented long agonist cycles and the odds for the incidence ovarian hyperstimulation syndrome. CI = confidence interval; OR = odds ratio. 623

6 Figure 4. Comparison of studies with major replacement of low-dose human chorionic gonadotrophin (HCG) supplemented cycles versus non-supplemented cycles and the odds for clinical pregnancy. Table 2. Oestradiol concentrations on the day of human chorionic gonadotrophin (HCG) administration according to the protocol used. Long agonist (n = 3) Antagonist (n = 3) Antagonist versus long agonist (n =2) Low-dose HCG ± No supplement Low-dose HCG No supplement Low-dose HCG No supplement ± ± ± ± ± Values are mean ± SD. One-way analysis of variance comparisons were performed. No statistically significant differences were observed between the different supplemented protocols or between the different non-supplemented protocols. Progesterone on the day of the HCG administration Only three studies (Filicori et al., 2005a; Serafini et al., 2006; Gomes et al., 2007) mentioned progesterone concentrations on the day of HCG administration. The mean values varied from 1.1 to 2.6 ng/ml for the low-dose HCG group. One study (Gomes et al., 2007), reported that 14 out of 17 patients in the low-dose HCG group, reached progesterone concentrations over 1.5 ng/ml while the mean value of these concentrations was 2.6 ng/ml. No significant difference was observed, for progesterone concentrations, either between groups or within groups. Cost analysis based on FSH consumption The short protocol study was excluded from the comparison between protocols because only one study (Berkkanoglu et al., 2007) existed. Higher consumption of FSH was noted compared with the other protocols ( units of FSH; P = 0.044; one-way ANOVA). Less total FSH (IU) was needed to achieve folliculogenesis in the long protocol (Table 3) while, in parallel, similar pregnancy rates were achieved. Also, less total FSH (IU) was consumed in the antagonist comparison (supplemented versus non-supplemented) and the low-dose HCG antagonist long agonist protocol comparison. Approximately 400 fewer units were consumed, although the difference did not reach statistical significance (one-way ANOVA). In addition, no significant difference was observed in the FSH needed for the low-dose HCG protocols [Tukey Honestly Significant Difference (HSD) test-homogenous subset procedure]. Specifically, in the comparison between supplemented long agonist and antagonist protocols, the mean Table 3. Total FSH (IU) consumed in different protocols. Long agonist (n = 3) Antagonist (n = 3) Antagonist versus long agonist (n = 2) 624 Low-dose HCG ± No supplement ± Low-dose HCG ± No supplement ± 373 Low-dose HCG antagonist ± No supplement agonist ± 142 Values are mean ± SD. HCG = human chorionic gonadotrophin. No statistically significant differences were observed.

7 difference was 297 units (Tukey HSD test-one-way ANOVA post-hoc test); favouring the long agonist protocol. Such a value is the key driver for cost comparisons and treatment choice between supplemented agonist and antagonist interventions. When comparing the incidence of OHSS between the lowdose HCG antagonist protocol and the non-supplemented agonist protocol, it was significantly lower in the HCGantagonist group (odds ratio, 0.30; 95% confidence interval, ; fixed effect model). This is another parameter that contributes to cost minimization. Eighty clinical pregnancies were reported in the supplemented antagonist protocol and 20 in the supplemented agonist protocol. Cost-effectiveness, sensitivity analysis and threshold analysis for supplemented protocol choice according to FSH requirements The cost assumptions used in the model are summarized in Appendix A. Based on the FSH consumption and incidence of OHSS, the expected cost per clinical pregnancy is 957 in the low-dose HCG agonist protocol, 1188 in the low-dose HCG antagonist and 1939 and 1622 for non-supplemented agonist and antagonist protocols respectively (Figure 5). Incremental cost is presented in Table 4. From the cost-effectiveness graph, it is obvious that both supplemented strategies present are possible optimal choices with the low-dose HCG agonist protocol offering the lowest cost option (Figure 6). One-way sensitivity and threshold analysis With regard to incremental net monetary benefits of lowdose HCG agonist FSH while using the low-dose HCG antagonist FSH for comparison, it is evident that at a value of units, the incremental net monetary benefit is 0 (Figure 7). As an extension of one-way sensitivity analysis, threshold analysis was performed by varying the value of FSH requirements for the low-dose HCG agonist protocol. It was found that, between 1244 and units, the option with the highest net benefit was the low-dose HCG agonist strategy, but between and 1960 units, the option with the highest net benefit was the low-dose HCG antagonist strategy. Figure 5. Decision tree and protocol choice. 625

8 Table 4. C ( ) IC ( ) E IE CE ICER Strategy choice Low-dose HCG agonist Low-dose HCG antagonist Non-supplemented antagonist (Dominated) Non-supplemented long agonist ,659 All options referenced to a common baseline Low-dose HCG agonist Low-dose HCG antagonist Non-supplemented antagonist Non-supplemented long agonist Without dominated options (simple or extended) Low-dose HCG agonist Low-dose HCG antagonist Non-supplemented long agonist ,659 Ordered by increasing effectiveness Non-supplemented antagonist Low-dose HCG agonist Low-dose HCG antagonist Non-supplemented long agonist The strategy non-supplemented antagonist is dominated by low-dose HCG antagonist. No strategies were eliminated by extended dominance. C = cost; E = effectiveness; HCG = human chorionic gonadotrophin; IC = incremental cost; ICER = incremental cost-effectiveness ratio; IE = incremental effectiveness. Cost ( ) Non-supplemented long agonist Non-supplemented antagonist Low dose HCG antagonist Low dose HCG agonist Effectiveness NMB Low dose HCG antagonist Low dose HCG agonist Threshold values: ufsh_required_lowdose_agon = EV = ufsh_required_lowdose_agon 626 Figure 6. Cost effectiveness analysis for the different protocols. Figure 7. One-way sensitivity analysis of net monetary benefits (NMB).

9 ufsh_required_lowdose_agon Two-way sensitivity and threshold analysis Simultaneous changes in the values of two independent variables (FSH requirements in the supplemented protocols) were assessed by two-way sensitivity analysis. The resulting graph (Figure 8) identifies which strategy is optimal in regions of values of FSH; thresholds are the border between two regions. Number needed to treat in the antagonist protocol Based on the OR derived from the meta-analysis, the number of patients needed to be treated to prevent one incidence of OHSS in the HCG-supplemented antagonist cycles as compared with the non-supplemented agonist cycles is 19. Discussion Low dose HCG antagonist Low dose HCG agonist ufsh_required_lowdose_antagon Figure 8. Two-way sensitivity analysis of net monetary benefits (NMB). It is obvious from the meta-analysis, that low-dose HCG replacement does not compromise clinical pregnancy rates while in parallel maintains low OHSS rates. Evidence exists that clinical pregnancy rates may be increased in the lowdose HCG-antagonist protocol but this remains to be elucidated in the future. There is no significant difference in the incidence of OHSS between the supplemented long agonist and antagonist protocols. Such substitution achieves the major clinical benefit when there is a major replacement of recombinant FSH. The benefit of HCG supplementation is lower consumption of recombinant FSH in these protocols, thus contributing to a lower cost. It is important to understand that increased effectiveness can be seen when there is major substitution of rfsh with low-dose HCG. When comparing the low-dose HCG-supplemented antagonist protocols with non-supplemented agonist protocols, no difference in pregnancy rates was found. In this way, a supplemented antagonist protocol is equally effective as an agonist protocol. Most importantly, there is significant reduction in OHSS (OR 0.30; 95% CI ; Figure 3), thus favouring the low-dose HCG antagonist protocol. Although, there is no difference in pregnancy rates in the study using the short agonist protocol (Berkkanoglu et al., 2007), no recommendation can be given definitely at this point, because of a lack of other studies. Minor replacement of rfsh was performed in this study, while a study with major replacement is needed. If such a study were to be successful, then FSH consumption could be significantly lowered and could be intensified in the first stages of ovarian stimulation. The limit of this meta-analysis is the small number of prospective studies that were included. Another limitation is the inclusion of meeting proceedings that cannot be adequately evaluated. This inclusion was necessary to increase the sample size and achieve results that are more accurate. In this analysis, 435 patients were included in the low-dose HCG group and 595 in the control group. A sample size of 524 patients is needed (Frattarelli et al., 2007) to prove a significant difference between three groups (rfsh only, rfsh + low-dose HCG, or rfsh + HMG). This metaanalysis approaches these numbers. On the other hand, more prospective studies are needed to confirm these results. The current cost-effectiveness model is based on the results obtained from the meta-analysis. Cancellation rate and miscarriage rate were mentioned separately, in two studies. A detailed Markov chain model needs to be calculated including all pregnancy parameters for more than one treatment cycle (Al-Inany et al., 2006). Results do not come by surprise. From the first studies using the long agonist protocol (Filicori et al., 2002), the cost-benefit and improved safety of this regimen was discussed. Interestingly, clinical pregnancy rates are improved in the supplemented antagonist protocol and, in comparison to the long agonist protocol, a significant difference in the incidence of OHSS was noted. This finding is important and needs to be further evaluated. It is the first randomized evidence between antagonist agonist protocols that combines similar efficacy, increased safety and more patientfriendly IVF. In this meta-analysis, data are pooled for all infertility groups and no results can be drawn specifically for each group. Recombinant FSH was administered in seven trials (Mendes et al., 2005; Davies et al., 2006; Dzik et al., 2006; Serafini et al., 2006; Berkkanoglu et al., 2007; Gomes et al., 2007; Van Horne et al., 2007), urinary human FSH in one trial (Koichi et al., 2006) and in one trial (Filicori et al., 2005a), a combination of rfsh and HMG was used. Seven studies mention individualization of the FSH dose. On the other hand, seven studies (Filicori et al., 2005a; Mendes et al., 2005; Dzik et al., 2006; Koichi et al., 2006; Berkkanoglu et al., 2007; Van Horne et al., 2007; Gomes et al., 2007) mentioned criteria for HCG administration. In these studies, mean follicle diameter during HCG administration ranged from 17 mm to 18 mm. The HCG was administered 627

10 Table 5. Total FSH (IU) consumed in the antagonist protocol. Low-dose HCG supplemented antagonist No-supplement antagonist Prospective studies only ± ± All studies ± ± 373 Values are mean ± SD. One-way analysis of variance comparisons were performed. HCG = human chorionic gonadotrophin. 628 mainly in the urinary form ( ,000 IU) and one study (Mendes et al., 2005) used the recombinant form (250 lg). The underlying mechanism responsible for improved pregnancy rates in GnRH antagonist protocols has yet to be clarified. Possible explanations include the LH effect from the addition of low-dose HCG, the increased oestradiol concentrations on the day of HCG administration and the endometrial effect of low-dose HCG. Controversy still exists when examining current evidence for each mechanism. The effect of added LH in GnRH antagonist cycles is not clear yet. Although some studies present evidence after LH supplementation (Baruffi et al., 2007; Garcia-Velasco et al., 2007), other studies support that no need exists (Kolibianakis et al., 2003, 2006). A meta-analysis (Kolibianakis et al., 2006) found higher serum oestradiol concentrations on the day of HCG administration and a higher number of mature oocytes but no difference in pregnancy rates. The other study (Baruffi et al., 2007) showed that LH add-back strategy (375 IU/day) rescued adverse effects on implantation from high doses of GnRH antagonist. In the current analysis, a trend towards significantly higher clinical pregnancy rates in the low-dose HCG-supplemented versus non-supplemented antagonist protocol was observed, indicating that the LH effect might not be the only responsible factor and that low-dose HCG might also exert an endometrial effect. From luteal phase studies, it is evident (Lovely et al., 2005; Jasinska et al., 2006) that HCG prevents apoptosis by exerting a direct effect on endometrial stromal cells. Another HCG effect in the endometrium can be mimicked by LH. HCG and LH can up-regulate cyclooxygenase-2 gene expression and increase the morphological as well as functional differentiation of human endometrial stromal cells into deciduas (Han et al., 1996, 1999). In general, the role of HCG/LH receptors is currently being examined as a biomarker of uterine receptivity and embryo implantation (d Hauterive et al., 2007). Altered endocrinology has also been observed. Low-dose HCG increased oestradiol concentrations, on the day of HCG administration when compared with non-supplemented protocols. The association between oestradiol concentrations on the day of HCG administration and pregnancy rates is controversial (Kosmas et al., 2004). Also recent evidence in an animal model shows that oestradiol is not compulsory for adequate folliculogenesis (Fatum et al., 2006). Where mentioned, a trend towards higher progesterone concentrations on the day of HCG administration was noted although pregnancy rates did not show a significant difference. In one study in which patients were down-regulated using a long agonist protocol (Gomes et al., 2007), most patients who received low-dose HCG supplementation (14 out of 17) reached over 1.5 ng/ml with a mean concentration of 2.6 ± 0.3 ng/ml. HCG alone stimulates the production of progesterone, oestradiol and testosterone by human luteinized granulosa cells in vitro (Morris et al., 1994).Considering the clinical pregnancy rates, these results contradict previous studies (Fanchin et al., 1996, 1997) with reference to progesterone concentrations 0.9 ng/ml and their effect on pregnancy rates. On the other hand, low-dose HCG might exert a complex role in the endometrium (Licht et al., 2001) thus overcoming the adverse effect from premature luteinization. Evidently, low-dose HCG can be used as a replacement for completing follicular development either with GnRH agonist or antagonist protocol. Benefits include lower FSH consumption and lower OHSS rates. Low-dose HCG-supplemented protocols are a less costly but equally effective method to complete ovarian stimulation and hold the promise of improved clinical pregnancy rates in antagonist cycles. Answering the question of which supplemented protocol to select, both strategies are presented as equally costeffective options. Supplemented agonist and antagonist protocols show small differences in FSH consumption. The choice of the less costly supplemented protocol, eventually, will depend on FSH consumption before low-dose HCG initiation. This is a specific target for future research. Another important point is that low-dose HCG antagonist protocols show less standard deviation on FSH consumption (Table 5) than similar long agonist protocols (Table 3), thus indicating a less risky choice. In a different study, low-dose HCG (2500 IU) used as the ovulation trigger prevented the development of OHSS in high-risk women without compromising success rates (Nargund et al., 2007). This is probably achieved, by reducing the overproduction of vascular endothelial growth factor that in excess increases vascular permeability leading to OHSS. More randomized clinical trials need to be performed to confirm these results. Comparisons must include non-supplemented long agonist protocols to compare the clinical efficacy of a patient-friendly and inexpensive protocol with the so-called golden standard protocol. Adequate reporting of these trials is required.

11 Appendix A. Cost assumptions in the model Therapy was defined as participation to the current cycle. Main outcome of the study was clinical pregnancy rate (intrauterine pregnancy at 6 weeks of gestation confirmed by ultrasound, as derived from the meta-analysis). Spontaneous pregnancies taking place independently from treatment, whenever mentioned, were excluded from the meta-analysis. Ovarian hyperstimulation syndrome hospitalization and work absence: FSH price was considered the same (for either recombinant FSH or human menopausal gonadotrophin) with a cost of 0.63 per unit. Overall agonist price = Overall antagonist price. Direct cost: analysis was performed on the costs of IVF treatment itself, including current IVF cycle. Multiple pregnancy costs, antenatal, perinatal and post-partum care costs in women who became pregnant were not considered. Ultrasound cost considered equal given the fact that patients returned to the clinic every 2 days for ultrasound and blood sampling. No discount rate for costs was applied. Costs were considered at the present value. References Al-Inany HG, Abou-Setta AM, Aboulghar MA et al HMG versus rfsh for ovulation induction in developing countries: a cost-effectiveness analysis based on the results of a recent metaanalysis. Reproductive BioMedicine Online 2, Baruffi RL, Mauri AL, Petersen CG et al Recombinant LH supplementation to recombinant FSH during induced ovarian stimulation in the GnRH-antagonist protocol: a meta-analysis. Reproductive BioMedicine Online 1, Berkkanoglu M, Isikoglu M, Aydin D et al Clinical effects of ovulation induction with recombinant follicle-stimulating hormone supplemented with recombinant luteinizing hormone or low-dose recombinant human chorionic gonadotropin in the midfollicular phase in microdose cycles in poor responders. Fertility and Sterility 3, Chun SY, Eisenhauer KM, Minami S et al Hormonal regulation of apoptosis in early antral follicles: folliclestimulating hormone as a major survival factor. Endocrinology 4, Davies E, Check JH, Brasile JK et al A prospective comparison of in-vitro fertilization (IVF) outcome following controlled ovarian hyperstimulation (COH) regimens using follitropin alpha exclusively or with the addition of low-dose human chorionic gonadotropin (hcg). Fertility and Sterility 85(suppl. 2), 16. d Hauterive SP, Berndt S, Tsampalas M et al Dialogue between blastocyst hcg and endometrial LH/hCG receptor: which role in implantation? Gynecological and Obstetrical Investigation 3, Dzik A, Freitas GC, De Pauwn K et al Efficacy of low-dose HCG alone following Recombinant FSH for controlled ovarian stimulation. Fertility and Sterility 3(suppl. 1), S427. Fanchin R, Righini C, Olivennes F et al Consequences of premature progesterone elevation on the outcome of in-vitro fertilization: insights into a controversy. Fertility and Sterility 5, Fanchin R, Righini C, Olivennes F et al Premature progesterone elevation does not alter oocyte quality in in-vitro fertilization. Fertility and Sterility 6, Fatum M, Gyo Y, Diana P et al Is estradiol mandatory for an adequate follicular and embryo development? A mouse model using aromatase inhibitor (anastrozole). Journal of Assisted Reproduction and Genetics 23, Filicori M, Cognigni GE, Gamberini E et al. 2005a Efficacy of lowdose human chorionic gonadotropin alone to complete controlled ovarian stimulation. Fertility and Sterility 2, Filicori M, Fazleabas AT, Huhtaniemi I et al. 2005b Novel concepts of human chorionic gonadotropin: reproductive system interactions and potential in the management of infertility. Fertility and Sterility 2, Filicori M, Cognigni GE, Tabarelli C et al Stimulation and growth of antral ovarian follicles by selective LH activity administration in women. The Journal of Clinical Endocrinology and Metabolism 3, Filicori M, Cognigni GE, Taraborrelli S et al Low-dose human chorionic gonadotropin therapy can improve sensitivity to exogenous follicle-stimulating hormone in patients with secondary amenorrhea. Fertility and Sterility 6, Frattarelli JL, Miller KA, Kaplan B et al Does leutinizing hormone activity in the form of low-dose HCG or HMG produce better outcomes for GNRH antagonist ART cycles stimulated with rfsh? Fertility and Sterility 88(suppl. 1), S132. Garcia-Velasco JA, Coelingh Bennink HJ, Epifanio R et al High-dose recombinant LH add-back strategy using high-dose GnRH antagonist is an innovative protocol compared with standard GnRH antagonist. Reproductive BioMedicine Online 3, Gomes MK, Vieira CS, Moura MD et al Controlled ovarian stimulation with exclusive FSH followed by stimulation with hcg alone, FSH alone or hmg. European Journal of Obstetrics, Gynecology and Reproductive Biology 1, Han SW, Lei ZM, Rao CV, 1999 Treatment of human endometrial stromal cells with chorionic gonadotropin promotes their morphological and functional differentiation into decidua. Molecular and Cellular Endocrinology 14, Han SW, Lei ZM, Rao CV, 1996 Up-regulation of cyclooxygenase- 2 gene expression by chorionic gonadotropin during the differentiation of human endometrial stromal cells into decidua. Endocrinology 5, Jasinska A, Strakova Z, Szmidt M, Fazleabas AT, 2006 Human chorionic gonadotropin and decidualization in vitro inhibits cytochalasin-d-induced apoptosis in cultured endometrial stromal fibroblasts. Endocrinology 9, Koichi K, Yukiko N, Shima K et al Efficacy of low-dose human chorionic gonadotropin (hcg) in a GnRH antagonist protocol. Journal of Assisted Reproduction and Genetics 5, Kolibianakis EM, Collins J, Tarlatzis B et al Are endogenous LH levels during ovarian stimulation for IVF using GnRH analogues associated with the probability of ongoing pregnancy? A systematic review. Human Reproduction Update 1, Kolibianakis EM, Albano C, Camus M et al Initiation of gonadotropin-releasing hormone antagonist on day 1 as compared to day 6 of stimulation: effect on hormonal levels and follicular development in in-vitro fertilization cycles. The Journal of Clinical Endocrinology and Metabolism 12,

12 Kosmas IP, Kolibianakis EM, Devroey P, 2004 Association of estradiol levels on the day of hcg administration and pregnancy achievement in IVF: a systematic review. Human Reproduction 11, Licht P, Russu V, Wildt L, 2001 On the role of human chorionic gonadotropin (hcg) in the embryo-endometrial microenvironment: implications for differentiation and implantation. Seminars in Reproductive Medicine 1, Lovely LP, Fazleabas AT, Fritz MA et al Prevention of endometrial apoptosis: randomized prospective comparison of human chorionic gonadotropin versus progesterone treatment in the luteal phase. The Journal of Clinical Endocrinology and Metabolism 4, Mendes IR, Kotecki JA, Caldeira G et al Will the administration of low-dose hcg for ovulation induction in association with bromocriptine aid in the prevention of OHSS? Fertility and Sterility 84(suppl. 1), S313. Mihm M, Baker PJ, Ireland Jl et al Molecular evidence that growth of dominant follicles involves a reduction in folliclestimulating hormone dependence and an increase in luteinizing hormone dependence in cattle. Biology of Reproduction 6, Morris RS, Francis MM, Do YS et al Angiotensin II (AII) modulation of steroidogenesis by luteinized granulosa cells in vitro. Journal of Assisted Reproduction and Genetics 3, Nargund G, Hutchinson L, Scaramuzzi R et al Low-dose HCG is useful in preventing OHSS in high-risk women without adversely affecting the outcome of IVF cycles. Reproductive BioMedicine Online 6, Ruvolo G, Bosco L, Pane A et al Lower apoptosis rate in human cumulus cells after administration of recombinant luteinizing hormone to women undergoing ovarian stimulation for in-vitro fertilization procedures. Fertility and Sterility 3, Serafini P, Yadid I, Motta EL et al Ovarian stimulation with daily late follicular phase administration of low-dose human chorionic gonadotropin for in-vitro fertilization: a prospective, randomized trial. Fertility and Sterility 4, Van Horne AK, Bates Jr GW, Robinson RD et al., 2007 Recombinant follicle-stimulating hormone (rfsh) supplemented with low-dose human chorionic gonadotropin compared with rfsh alone for ovarian stimulation for in-vitro fertilization. Fertility and Sterility 4, The data included in this study were presented as a poster at the 24th Annual Meeting of the European Society of Human Reproduction and Embryology, Barcelona, July, Declaration: The authors report no financial or commercial conflicts of interest. Received 9 April 2008; refereed 16 September 2008; accepted 3 June