Clinical management of low ovarian response to stimulation for IVF: a systematic review

Transcription

1 Human Reproduction Update, Vol.9, No.1 pp. 61±76, 200 DOI:10.109/humupd/dmg007 Clinical management of low ovarian response to stimulation for IVF: a systematic review B.C.Tarlatzis 1, L.Zepiridis, G.Grimbizis and J.Bontis Unit for Human Reproduction, 1st Department of Obstetrics and Gynaecology, Aristotle University of Thessaloniki, Greece 1 To whom correspondence should be addressed at: IVF Center, Geniki Kliniki Thessalonikis, 2 Gravias str., 56 5 Thessaloniki, Greece. tarlatzis@hol.gr Poor response is not a rare occurrence in ovarian stimulation. Although not fully accepted, the most dominant criteria for poor ovarian response are small numbers of follicles developed or oocytes retrieved, and low estradiol (E 2 ) levels after the use of a standard stimulation protocol. There is no ideal predictive test as the poor responder is revealed only during ovulation induction; however, increased levels of day FSH and E 2 as well as decreased levels of inhibin B can be used to assess ovarian reserve. Several protocols have been proposed for clinical management of low ovarian response in IVF. Although high doses of gonadotrophins have been used by the vast majority of authors, results have been controversial and prospective randomized studies have shown little or no bene t. The few available relevant studies do not indicate that recombinant FSH improves outcome. Flare-up GnRH agonist protocols (including all dosage varieties) produce better results than standard long luteal protocols. Luteal initiation GnRH agonist `stop' protocols were shown to improve ovarian response according to prospective studies with historical controls, but this was not con rmed by well-designed prospective, randomized, controlled studies. The few available data obtained with GnRH antagonists have not shown any bene ts. Adjuvant therapy with growth hormone (GH) or GH-releasing factors results in no signi cant improvement. The use of corticosteroids reduces the incidence of poor ovarian response in women undergoing IVF treatment. The limited data obtained with nitric oxide donors are encouraging. Pretreatment with combined oral contraceptives prior to stimulation may help ovarian response. No bene t was observed with standard use of ICSI or assisted hatching of zona pellucida. Finally, natural cycle IVF has produced results which are comparable with those obtained with stimulated cycles in true poor responders. Well-designed, large-scale, randomized, controlled trials are needed to assess the ef cacy of these different management strategies. Keywords: clinical management/ivf/ovarian reserve/ovarian stimulation protocols/poor ovarian response Introduction A reasonable percentage of women undergoing infertility treatment respond poorly to the usual gonadotrophin stimulation protocol applied. Garcia et al. (198) rst described as a poor responder the patient with peak estradiol (E 2 ) levels <00 pg/ml and decreased follicular response, expressed as fewer retrieved and fertilized oocytes and also fewer transferred embryos. These patients have lower pregnancy rates compared with normal responders, following a standard ovarian stimulation protocol with human menopausal gonadotrophins (hmg). Recent studies have shown convincingly that poor ovarian response is a rst sign of ovarian ageing (early ovarian failure or early menopause) (Beckers et al., 2002; De Boer et al., 2002; Nikolaou et al., 2002). Among patients undergoing IVF treatment, the prevalence of low ovarian response is 9 to 2% (Keay et al., 1997). Methodology This systematic review aimed to identify and evaluate qualitatively all the existing protocols applied to previous poor responders to ovarian stimulation. The literature search was performed using PubMed and hand-searching relevant journals. The basic keywords were `poor' or `low' or `bad' `responders' and `IVF'. Among the studies identi ed, those not involving IVF treatment were discarded. Similarly, studies presented at meetings or congresses, with only abstracts available, were excluded. This search revealed 16 prospective randomized trials, 2 prospective non-randomized or studies with historical controls, and seven retrospective studies. All of the studies were evaluated and consequently ranked in order to reveal the speci c power of each, based on certain criteria (Sackett et al., 1991; Von Korff, 199; Geurts et al., 2001). This Ó European Society of Human Reproduction and Embryology 61

2 B.C.Tarlatzis et al. was done solely to help the reader assess objectively the value of each study. Therefore, each parameter tested was graded with `1' if true or with `0' if not true. These graded sets of criteria were: d Study design: ± Prospective studies were graded with `1' ± Randomized studies were graded with `1' ± The existence of a simultaneous control group was graded with `1' ± Placebo-controlled studies were graded with `1' d The sample size. If >25 per comparison group, the study was graded with `1' ± The existence of statistical analysis was graded with `1' ± The existence of standardized outcome measures was graded with `1' Thus, a maximum total score of 7 was achieved if a study ful lled all of the above criteria. Studies accumulating a score > were considered as trials with powerful evidence of their results, but those with a score < were considered as poor. Studies with a score of were considered to be of moderate quality. De nition Unfortunately, there is no universally accepted de nition for the `low', `poor', `bad' or `non'-responder, although these patients have de nitely lower pregnancy rates compared with `normal' responders. Numerous criteria have been used to characterize poor response. The number of developed follicles and/or number of oocytes retrieved after a standard-dose ovarian stimulation protocol are, hierarchically, two of the most important criteria for de ning poor ovarian response. The proposed number varies among different authors, and ranges from less than three to less than ve dominant follicles on the day of hcg administration (Land et al., 1996; Fridstrom et al., 1997; Raga et al., 1999), and/ or less than three to less than ve retrieved oocytes (Chong et al., 1986; Rombauts et al., 1998; Surrey et al., 1998). Peak E 2 level is another criterion used, as it correlates with the number of developing follicles. A peak E 2 level of <00 to <500 pg/ml has been proposed as being crucial for de ning poor response (Garcia et al., 198; Brzyski et al., 1988; Raga et al., 1999), although a level <100 pg/ml on day 5 of stimulation has also been suggested (Schoolcraft et al., 1997). An elevated day FSH level ranging from >7 to >15 miu/ml has been proposed as an additional criterion to de ne poor ovarian response (Droesch et al., 1989; Feldberg et al., 199; Karande et al., 1997; Faber et al., 1998). Likewise, an advanced patient age of >0 years (Karande et al., 1997), disappointing or no response to the clomiphene challenge test (Navot et al., 1987) and a failed `Lupron screening test' (Katayama et al., 1988) may be predictive of poor response. More logically, but not necessarily more accurately, criteria such as at least one cancelled IVF cycle (Schachter et al., 2001), increased number of hmg or FSH ampoules used (>) (Shaker et al., 1992), increased (>00 IU/day) gonadotrophin dose used (Faber et al., 1998) and prolonged duration of gonadotrophin stimulation (Toth et al., 1996) have been suggested. It should be noted that, in a variety of studies, these criteria have been used either alone or in combination, thereby highlighting the complexity, the lack of uniformity in de nitions and also the major dif culties encountered when comparing the different strategies proposed. Some authors have suggested a classi cation of poor responders. Thus, according to Gorgy and Taramissi, 2001) there are two subgroups of poor responders (Gorgy and Taramissi (2001). The rst subgroup includes the young (aged <7 years) and slim (body weight <70 kg), who developed less than ve follicles following 9 days of ovarian stimulation with 225 IU/day and did not reach oocyte retrieval, or required >600 IU of gonadotrophin per retrieved oocyte if they did reach that stage. The second subgroup included patients aged >7 years and weighing >70 kg who had their cycles cancelled due to less than ve developed follicles following 9 days of ovarian stimulation with 00 IU/day of gonadotrophins. Similarly, others (Barri et al., 1998), proposed that three subgroups of low responders should be identi ed: (i) patients with a low response to previous IVF attempts but normal basal FSH levels; (ii) young patients with persistently high FSH levels; and (iii) older patients with an abnormal endocrinological pro le. Clinical experience has not shown any real advantage in subcategorizing patients however, as no signi cant differences regarding response to ovarian hyperstimulation protocols have been observed (Wang et al., 2002). In fact, low ovarian response is con rmed only after the patient has failed ovarian stimulation following an accepted as `normal' or `standard' ovarian stimulation regimen. Although several possible aetiologies have been suggested, it seems that a diminished ovarian reserve is the principal factor of poor ovarian response (Pellicer et al., 1998). Alternatively, a decreased number of FSH receptors available in granulosa cells (Zeleznik et al., 1981), defective signal transduction after FSHreceptor binding (Hernandez et al., 1992), the presence of a special FSH receptor-binding inhibitor in the follicular uid (Lee et al., 199), an inappropriate local vascular network for the distribution of gonadotrophins (Pellicer et al., 199), the presence of autoantibodies against granulosa cells (Pellicer et al., 1998) and lowered circulating gonadotrophin surge-attenuating factor (GnSAF) bioactivity (Martinez et al., 2002) have each also been proposed. Prediction As currently there is no accurate and 100% predictive test available to assess ovarian response, there is likewise no screening test to detect poor ovarian response, and only ovarian reserve can be assessed. Several tests have been suggested for this purpose, but at present there is no de nitive evidence for their predictive value. Clearly, the ideal test is the response of the ovaries to ovarian stimulation. High levels of serum FSH (>12 or >15 miu/ml) on cycle day 2 or (Cameron et al., 1988; Scott et al., 1989; Toner et al., 1991), elevated levels of serum E 2 (>0 or >75 pg/ml) on cycle day 2 or (Licciardi et al., 1995), and decreased levels of serum inhibin B (<5 pg/ml) on cycle day 2 or (Seifer et al., 1997) seem to be the main predictive tests. Recently, the predictive value of the reduced GnSAF production and bioactivity (Martinez, et al., 2002) and low insulin-like growth factor (IGF)-I concentrations in follicular uid (Oosterhuis et al., 1998) were evaluated in poor responders. Furthermore, it has been proposed that decreased serum concentrations of antimuèllerian hormone may be a novel marker for ovarian ageing. This correlates with the number of 62

3 Low ovarian response to stimulation for IVF antral follicles and age, but less strongly with FSH levels, while its decrease occurs earlier than do changes in other markers associated with ovarian ageing (de Vet et al., 2002). Sonographic tests have also been suggested to predict ovarian response. Thus, decreased ovarian volume was rst proposed (Lass et al., 1997), followed by decreased basal antral follicle count (Chang et al., 1998) and later a signi cantly decreased ovarian stromal blood ow (Engmann et al., 1999). Recently, it was proposed that the antral follicle count provides better prognostic information on the occurrence of poor response during hormone stimulation for IVF than does the patient's chronological age and available endocrine markers (Bancsi et al., 2002). Multivariate analyses involving basal FSH and inhibin B levels to a logistic model with the antral follicle count, may signi cantly improve the prediction of poor ovarian response in IVF (Bancsi et al., 2002). More classically applied tests include the clomiphene challenge test (Navot et al., 1987), the `Lupron' screening test (Padilla et al., 1991), the exogenous FSH ovarian reserve test (Fanchin et al., 199), the elevated ratio of FSH/LH in day blood tests (Mukherjee et al., 1996), and the GnRH stimulation test (Karande and Gleicher, 1999). However, the response of some patients will be poor despite their predictive tests not being suggestive of low ovarian reserve. Management High doses of gonadotrophins The initial unresponsiveness to gonadotrophin stimulation unavoidably leads the clinician to increase the dosage of medication for ovarian hyperstimulation, and indeed this action forms part of the de nition of the low responder. According to most authors, the common initial dose for poor responders is at least 00 IU/day. Few studies have been conducted on this issue however, as high doses of gonadotrophins have been used in the large majority of regimens employed (see Table I). One group (Cedrin-Durnerin et al., 2000), in a prospective randomized study, compared a high xed versus a step-down dose of gonadotrophins on a minidose are-up GnRH agonist regimen. Patients were pretreated for 1 days with 10 mg/day norethisterone, ceasing at cycle day 0, followed by 100 mg triptorelin s.c. from day 1, reduced to 25 mg s.c. from day of the cycle. Puri ed FSH at a dose of 50 IU/day i.m. was administered from day either in a xed manner or decreasing to 00 IU/day, and nally to 150 IU/day. Thus, these authors showed, by applying this decremental are-up GnRH agonist protocol, that there were increased cancellation rates and similar pregnancy rates but, as expected, a signi cantly reduced number of ampoules of GnRH was used. Doubling of the starting hmg dose, from 225 IU to 50 IU/day from day 5 of ovarian stimulation onwards, has been evaluated in a prospective randomized study (Van Hooff et al., 199). These authors concluded that such an approach was ineffective in enhancing ovarian response in low responders, this being in accordance with the hypothesis that follicular recruitment occurs only during the late luteal and early follicular phases of the menstrual cycle. Similar doses (50 IU/day) were also used in a retrospective study (Karande et al., 1990), where in poor responders the mean number of retrieved oocytes (2.67) was still low, as was the pregnancy rate (12%). The latter group concluded therefore that there was no bene t from the incremental increase in FSH dose. The same conclusions were reached in another retrospective study (Land et al., 1996), where 126 previously poor responders were given 50 IU/day hmg. The women showed an increased number of oocytes retrieved, but the pregnancy rate remained low (.2%). Conversely, in a prospective study with historical controls, the effects of the increased gonadotrophin doses in poor responders were evaluated as improved (Hofmann et al., 1989). The authors showed increased pregnancy rates (. versus 6.7%) and lower cancellation rates (9 versus 5%), by using 50 IU/day of puri ed FSH (in step-down fashion), rather than 00 IU/day. The bene ts of using a very high dose of puri ed FSH were also reported in a retrospective study (Crosignani et al., 1989) which included 116 poor responders and resulted in satisfactory follicular growth in 70% of cases. In general, high doses of gonadotrophins have been used by the vast majority of authors and form a clear part of all protocols for ovarian stimulation in poor responders. None the less, the results of studies evaluating the use of gonadotrophins in these patients remains controversial, though the true prospective randomized studies have shown either minimal or no bene t. Use of recombinant FSH versus puri ed urinary FSH The use of recombinant FSH (rfsh) appears to be associated with better assisted reproduction results than do either puri ed urinary FSH or hmg (Out et al., 1996; Daya and Gunby, 2002). As patients included in these studies were not poor responders, these results stimulated the concept that rfsh might also improve oocyte and embryo quality in poor-response patients (see Table I). The use of rfsh versus puri ed FSH in poor responders was evaluated in a small (15 versus 15 patients), prospective randomized study (Raga et al., 1999). The authors found an increased mean number of oocytes collected (7.2 versus 5.6), improved pregnancy rates ( versus 6%) and lower cancellation rates (1 versus 0%). Similarly, another prospective study, albeit with historical controls (De Placido et al., 2000), assessed the ef cacy of 00 IU rfsh versus the same dose of puri ed FSH in the are-up protocol involving 28 cycles of poor-responder patients in each group. These authors suggested that rfsh was associated with a signi cantly larger number of oocytes retrieved (2. versus 1.7) and signi cantly increased pregnancy rates (1. versus 0%). It seems therefore, that there is no evidence that rfsh produces better results in poor responders, though larger prospective randomized trials are needed to elucidate this issue further. Luteal initiation of FSH This interesting approach was rst suggested in the late 1980s (Rombauts et al., 1998). Healthy, small, antral follicles are present in late follicular phase, and initiation of their further development occurs under action of the premenstrual FSH rise (Gougeon, 1996). The hypothesis was that earlier administration of FSH might increase the number of recruited follicles by opening the recruitment window earlier, in the late luteal phase of the preceding cycle. In a prospective randomized controlled study in two patient groups (n = 20 each), these authors administered 6

4 B.C.Tarlatzis et al. Table I. Various protocols Reference Study design Criteria for low response Regimen No. of cycles Retrieved oocytes/cycle Cancellation rate No. of clinical pregnancies/et Comments Score Keay et al. (2001) Battaglia et al. (1999) Cedrin-Durnerin et al. (2000) Van Hoof et al. (199) Hofmann et al. (1989) Karande et al. (1990) Land et al. (1996) Rombauts et al. (1998) De Placido et al. (2000) Raga et al. (1999) Moreno et al. (1998) Feldman et al. (2001) Multicentre prospective randomized, double-blind, placebo-controlled trial (long luteal GnRH agonist + dexamethasone versus long luteal GnRH agonist) Prospective randomized trial ( are-up GnRH + L-arginine versus are-up GnRH) Prospective randomized study (high xed versus step-down dose of gonadotrophins in a are-up regimen) Prospective randomized study (doubling hmg dose on day 5 of COH versus unchanged hmg dose) Prospective study with historical controls (6 FSH versus FSH) Retrospective study (6 FSH from day 1 versus 6 FSH from day 2) Retrospective study ( hmg ampoules/day versus 6 hmg ampoules/day) Prospective randomized controlled study (rfsh from cycle day (±5) versus rfsh from cycle day ()) Prospective study with historical controls (rfsh versus puri ed FSH) Prospective randomized study (rfsh versus puri ed FSH) Prospective randomized study (ICSI versus IVF) Prospective study with historical controls (natural cycle versus stimulated cycles) Normal responders Pretreatment from day 19 with norethisterone 10 mg/day for 7 days, long luteal GnRH agonist buserelin 0.5mg on day 21, hmg 150±00 IU/day i.m. from day of down-regulation plus 1 mg/day dexamethasone per os up to day of hcg < follicles and E2 <1100 pmol/l on day 8 of COH in previous cycle <5 oocytes retrieved in previous cycle, increased basal FSH levels < follicles and E 2 <00 pg/ml on day 5 of stimulation in previous cycle < follicles and peak E 2 <00 pg/ml in previous cycle, FSH >1 miu/ml Poor response in previous cycle with ampoules FSH <5 dominant follicles on hcg day ±6 retrieved oocytes in previous IVF attempt Peak E2 <500 pg/ml, < oocytes retrieved in previous cycles < follicles in previous IVF attempt, at least two cancelled IVF cycles, normal FSH levels 0.1 mg triptorelin s.c. from cycle day 1 plus 50 IU/day i.m. puri ed FSH plus 16 g/day L-arginine orally 1 days of pretreatment with 10 mg/day norethisterone ceasing at cycle day 0, 100 mg triptorelin s.c. from day 1, reduced to 25 g s.c. from day. 50 IU/day i.m. puri ed FSH from day xed or decreasing to 00 IU/day and to 150 IU/day 225 IU/day i.m. hmg from day for 5 days, increasing to 50 IU/day 15 versus versus 17 8 versus 8 25 versus versus 10.1 versus 1.6 (P = 0.09) 6. versus IU/day puri ed FSH (step-down) versus 2 50 IU/day puri ed FSH from cycle day 1 GnRH long protocol, 50 IU/day hmg after down-regulation Standard are-up leuprolide s.c.; 150 IU/day rfsh from day 25 of previous cycle Flare-up decapeptyl 0.1 mg/day, 00 IU rfsh from cycle day versus 12.% (P = 0.001) 11 versus 76% (P = 0.001) versus 29.16% 26.9 versus 17.2% (P = 0.07) 17.6 versus 0% 12.9 versus 10.7% Improved pregnancy rates, lower cancellation rates Increased no. of oocytes collected, lower cancellation rates, improved pregnancy rates Signi cantly increased number of ampoules used, decreased cancellation rates, similar pregnancy rates No change No change No change No change 6 9 versus 5% (P = 0.0). versus 6.7% (P = 0.02) Improved pregnancy rates, lower cancellation rates 2.67 ± 12.0% No change 126 versus versus versus IU/day rfsh IU/day hmg 15 versus 15 <6 retrieved oocytes Long luteal leuprolide s.c. 1 mg/day, hmg + FSH 00±600 IU/day i.m. from day of down-regulation, standard ICSI treatment 52 versus 52 ± Natural cycle versus versus 7.5 (P < 0.05).5 versus 6 2. versus 1.7 (P = 0.01) 7.2 versus 5.6 (P < 0.05) 70.6 versus.7% versus 25%.2% Increased no. of oocytes collected Not mentioned 7.10% 15. versus 0% 1 versus 0% (P < 0.05). versus versus 11.5% At least 1 in 85% versus 6% (P = 0.01) 21.1 versus 17.% Essentially no change 5 Improved ovarian outcome Increased no. of oocytes collected, improved pregnancy rates, lower cancellation rates No change % 9.0% Comparable results with those of stimulated cycles

5 Low ovarian response to stimulation for IVF Table I. Continued Comments Score Cancellation rate No. of clinical pregnancies/et Retrieved oocytes/cycle Regimen No. of cycles Reference Study design Criteria for low response Improved results 18.8 versus 0% 0.9 versus versus 8% Natural cycle 16 versus 25 2 cancelled IVF cycles due to <200 pg/ml E2 and <2 follicles on hcg day Prospective study with historical controls (natural cycle versus stimulated cycles) Bassil et al. (1999) 7.5% 2.08% Similar results with those of stimulated cycles Age > years Natural cycle 8 At least 1 in 8.5% Prospective study with no controls (natural cycle versus stimulated cycles) Bar Hava et al. (2000) COH = controlled ovarian hyperstimulation; ET = embryo transfer. leuprolide according to a standard short or long follicular protocol. The control group received 150 IU/day rfsh from cycle day, while the study group received the same dose from day 25 of the previous cycle. Unfortunately, the results were not encouraging, with increased cancellation rates ( versus 25%), decreased number of oocytes collected (.5 versus 6) and lower pregnancy rates (0 versus 5%); in addition, increased FSH doses (1950 versus 1500 IU) were needed for a longer stimulation period (15 versus 11 days) (see Table I). Flare-up GnRH agonist regimens: short and ultra-short protocols The are-up regimens involve early follicular phase commencement of the GnRH agonist, with minimal delay before the onset of gonadotrophin administration (Howles et al., 1987; Marcus et al., 199). There are two theoretical advantages with this approach: rst, the ovarian suppression is not excessive; and second, the initial stimulation of the GnRH receptors and consequent secretion of endogenous gonadotrophins enhances the effects of the exogenously administered gonadotrophins. By contrast, it has been proposed that by decreasing the GnRH agonist dose, a lighter ovarian suppression could be obtained and, hence a better response to gonadotrophin stimulation could be achieved. Furthermore, several microdoses of GnRH agonists in the areup protocols have been tested, the aim being to achieve gonadotrophin release while eliminating the phenomena of increased LH, androgen and progesterone secretion noted in the classic are-up protocols (Scott et al., 199; Deaton et al., 1996). Thus, at least in theory, these regimens would be suited to patients with low ovarian response. Although widely used, there are no prospective randomized controlled trials of are-up protocols which can be used to assess their ef cacy compared with standard protocols. The relevant studies are summarized in Table II. Standard-dose are-up regimens In a prospective study with historical controls and using an ultrashort protocol (Howles et al., 1987), according to which seven patients were administered 0.5 mg/day buserelin during only the rst three days of the cycle a 0% cancellation rate and 2.9% pregnancy rate were found. Similarly, in a prospective study with no controls, the same are-up regime was used in three categories of various responders after a Lupron screening test (Padilla et al., 1996). The group of 5 poor-responding patients had a low cancellation rate (11.%) and a good pregnancy rate (29%) despite the low number of oocytes retrieved. In accordance with these data, others (Toth et al., 1996) compared retrospectively the are-up versus the luteal GnRH agonist regimen, and observed higher pregnancy rates (20. versus 11.7%) and lower cancellation rates with the are-up protocol. By contrast, other authors failed to con rm any substantial bene t of using a classic are-up protocol. In a prospective study with historical controls (Karande et al., 1997), 80 poor responders were treated using a classic are-up GnRH agonist regimen with leuprolide 0.5 mg/day from cycle day 2, and 50±600 IU/day hmg from cycle day. The authors found an increased number of retrieved oocytes (10 per cycle), but otherwise no improvement, with a high cancellation rate (2.%) and a low pregnancy rate (1.%). 65

6 B.C.Tarlatzis et al. Table II. GnRH agonist are-up protocols Reference Study design Criteria for low response Regimen No. of cycles Retrieved oocytes/cycle Cancellation rate No. of clinical pregnancies/et Comments Score Surrey et al. (1998) Howles et al. (1987) Schoolcraft et al. (1997) Scott et al. (199) Karande et al. (1997) Padilla et al. (1996) Leondires et al. (1999) Toth et al. (1996) Prospective non-randomized trial with historical controls (microdose are-up versus long luteal GnRH agonist) Prospective analysis with historical controls Prospective analysis with historical controls (microdose are-up GnRH + GH + COC pretreatment + FSH versus long luteal GnRH + GH + FSH) Prospective analysis with historical controls (microdose are-up GnRH + FSH versus long luteal GnRH + FSH) Prospective analysis with historical controls Prospective study, no controls (The same regime in three categories of various responders after Lupron screening test) Retrospective analysis (microdose are-up versus long luteal with decreasing dose of GnRH agonist) Retrospective study ( are-up versus luteal GnRH agonist regime) Peak E2 <500 pg/ml, < oocytes retrieved in previous cycle, < follicles >15 mm (all had FSH <15 miu/ml) Poor response to prior CC plus hmg stimulation E2 <100 pg/ml on day 5 of stimulation with FSH, < follicles, cancelled cycle Peak E2 levels <500 pg/ml during stimulation with FSH 21 days of pretreatment with COC, 80 mg leuprolide s.c. from day. 00±50 IU/day i.m. puri ed FSH from day mg/day buserelin nasally from day 1 to day. hmg from day 21 days of COC, 80 mg/day leuprolide s.c. plus IU/day GH i.m. from day post-coc, 50 IU/day i.m. puri ed FSH from day 5 post-coc 0 mg/day leuprolide s.c. from cycle day ; 50 IU/day i.m. puri ed FSH from cycle day 5 Day FSH >7 and <12 miu/ml 0.5 mg leuprolide s.c. from cycle day 2, plus 6±8 ampoules hmg from cycle day No are-up response in GnRH (leuprolide acetate) screening test Peak E 2 <500 pg/ml, < oocytes retrieved in previous cycle, one cancelled IVF attempt Flare-up leuprolide, high doses of puri ed FSH and hmg 21 days of pretreatment with COC, 0 mg/day leuprolide s.c. from day, puri ed FSH 00 IU/day i.m. from day 5 Basal FSH levels >15 miu/ml Flare-up GnRH agonist + high doses of puri ed FSH and hmg 15 versus 15 and 19 versus 19 (total number of patients) 7. versus 6. and.8 versus 6.7 versus 5.% (P < 0.05) and 1.6 versus 7.7% (P < 0.05) 1.7 versus 0% and 27. versus 0% Signi cantly decreased cancellation rates 7 0% 2.85% Decreased cancellation rates, increased pregnancy rates % 50% Signi cantly decreased cancellation rates, increased pregnancy rates, and number of retrieved oocytes 1.8 versus 5.1 (P < 0.05) 0.00% 0 versus 11.8% Signi cantly increased number of retrieved oocytes, increased peak E 2 levels % 1.% Increased no. of aspirated oocytes, otherwise no improvement 5 Signi cantly lower 71 versus versus % 29.0% Low no. of aspirated oocytes, increased cancellation and pregnancy rates 22.5 versus 8.2% (P = 0.02) Decreased (P < 0.0) 7. versus 60% 20. versus 11.7% (P < 0.01) Signi cantly increased cancellation rate Increased pregnancy rates, decreased cancellation rates COC = combined oral contraceptives. 66

7 Low ovarian response to stimulation for IVF Hence, most studies conducted to assess the standard dose are-up protocols (that are non-randomized) demonstrated a degree of improvement. Reduced-dose GnRH agonist are-up regimens The idea of minimizing the dose of the GnRH agonist created the so-called `mini' and `micro' dose are-up GnRH agonist regimens. Improved outcome was observed in a prospective, non-randomized trial with historical controls (Surrey et al., 1998), who treated patients who had poor response and no pregnancies in a previous IVF attempt with the long luteal regimen, with a microdose are-up GnRH agonist regimen (leuprolide 80 mg/day s.c. from day ). The cancellation rate was signi cantly decreased in the are-up microdose protocol, and the clinical pregnancy rate was increased. Impressive results using the same microdose leuprolide protocol were also reported in a prospective study with historical controls (Schoolcraft et al., 1997). Here, 2 poor responders were pretreated for 21 days with a combined oral contraceptive (COC). On day post-coc, each patient received leuprolide 0 mg b.i.d. and GH IU/day (i.m.), followed on day 5 post-coc by a high dose of gonadotrophins (50 IU puri ed FSH). The cancellation rate was 12.5%, a mean of 10.9 oocytes per patient was retrieved, and a very optimistic pregnancy rate of 50% was obtained. Even lower doses were used in a prospective study with historical controls (Scott and Navot, 199), wherein leuprolide was given to 2 women at a dose of 20 mg b.i.d. from cycle day, followed by high doses of FSH from cycle day 5. The authors found higher peak E 2 levels and a higher number of retrieved oocytes. Despite these ndings, the initial optimism for the microdose are-up regimen was not supported by the results of other studies. Hence, in a retrospective analysis one group (Leondires et al., 1999) compared a microdose are-up regimen with a long luteal with decreasing dose of GnRH agonist, and observed signi cantly higher cancellation rates (22.5 versus 8.2%, P = 0.02), lower clinical pregnancy rates (7. versus 60%, P = NS) and a decreased number of oocytes retrieved per cycle (1. versus 16.5, P = NS) with the microdose are-up regimen. The results derived from the use of reduced-dose GnRH agonist are-up regimens are controversial. There is a trend towards improvement, but larger controlled and randomized studies are needed to support this issue. Luteal initiation of GnRH agonist regimens These regimens are characterized by the use of relatively low doses of GnRH agonists commencing in the mid-luteal phase of the cycle and usually ending at the time of menses or shortly thereafter, in combination with high doses of gonadotrophins. The possible mechanism of action is the reduced effect of the GnRH agonists on their ovarian receptors (Latouche et al., 1989; Kowalik et al., 1998), that results in reduced ovarian suppression and consequently, in increased ovarian response. Despite the early discontinuation of the agonist, the incidence of premature LH surge is low but the results are quite contradictory. Only two published prospective randomized controlled trials showed no statistically signi cant increase in pregnancy rates, whereas seven prospective trials with historical controls and one retrospective study demonstrated improved outcome. These studies are summarized in Table III. In a prospective randomized, controlled trial involving 78 cycles, a `stop agonist' regimen was compared with a standard long luteal protocol (Dirnfeld et al., 1999). Thus, the use of GnRH agonist (buserelin 1 mg/day intranasally or triptorelin 0.1 mg/day, s.c.) was initiated on day 21 of the pre-stimulation cycle and ceased on the day of con rmed pituitary suppression (E 2 level <10 pmol/l). Ovarian stimulation was induced with the use of 225 to 75 IU/day hmg or puri ed FSH (i.m.), commencing on the day of down-regulation. No improvement was found, as the mean number of the retrieved oocytes remained unchanged and no signi cant increase in the pregnancy rate was noted. Similar results were observed in another well-designed, prospective, randomized, controlled trial (Garcia-Velasco et al., 2000), where the `stop' versus `non-stop' protocol of GnRH agonist, plus high doses of gonadotrophins were compared. The authors used leuprolide (1 mg, s.c.) from day 21 of the cycle, ceasing on the day of menses, followed by 75±50 IU hmg and/or puri ed FSH daily (i.m.). A signi cantly higher number of aspirated follicles was found (8.7 versus 5. per cycle, P = 0.027), but there was no signi cant difference in either cancellation rate (5.7 versus 2.8%) or pregnancy rate (18.7 versus 1.%). These results are not supported by other studies which, although prospective, had historical controls. Thus, in a prospective analysis involving 22 cycles (Faber et al., 1998), a low-dose midluteal GnRH agonist (leuprolide 0.5 mg, s.c.) was administered but then discontinued with the onset of menses. All of the patients had received the GnRH agonist for at least 7 days, after which 50±600 IU of puri ed FSH or hmg were administered (i.m.) daily. The dose of gonadotrophins was decreased 2 days prior to hcg administration, and this was referred to as the `stop-lupron protocol'. The authors reported a low cancellation rate of 12.5%, a high number of oocytes aspirated per cycle (11.1), and an impressive clinical pregnancy rate per transfer (2%). Interestingly, among the poor responders the authors did not nd any statistically signi cant difference when comparing the use of hmg/puri ed FSH with puri ed FSH alone. In accordance with this approach, another group (Pu-Tsui et al., 2002), in a 52 cycle, non-randomized prospective study, administered 0.5 mg leuprolide s.c. from day 21 to the next cycle's day 2. After this, 00±50 IU/day hmg and puri ed FSH were used for controlled ovarian stimulation. The patients showed a good response to stimulation (mean 7.5 oocytes per cycle) and encouraging pregnancy rates (20.5% per embryo transfer). The same protocol was also assessed by another group in an 82-cycle prospective analysis with historical controls; both high pregnancy rates (.%) and cancellation rates (1.6%) were observed (Karande et al., 1997). Switching between various GnRH agonists did not seem make any difference. Thus, in a prospective study with historical controls and involving 6 poor responders, the use of nafarelin (0.6 mg/day, commenced in the midluteal phase and discontinued on day 5 of ovarian stimulation) resulted in an enhanced ef cacy of the gonadotrophin treatment (Schachter et al., 2001). Herein, the number of retrieved oocytes was increased by 28%, cancellation rates were decreased (to 8.%) and pregnancy rates increased (to 19.%). In another prospective study with historical controls, and using the same GnRH agonist at the same dosage but discontinuing it on day 1 of the next cycle, 9 poor responders were treated and showed a positive effect on the number of 67

8 B.C.Tarlatzis et al. Table III. Luteal-onset GnRH agonist protocols Reference Study design Criteria for low response Regimen No. of cycles Retrieved oocytes/cycle Cancellation rate No. of clinical pregnancies/et Comments Score Dirnfeld et al. (1999) Juan Garcia-Velasco et al. (2000) Prospective, randomized, controlled trial (luteal GnRH- agonist stop versus non-stop protocol) Prospective, randomized, controlled trial (luteal GnRH agonist stop versus non-stop protocol) Karande et al. (1999) Prospective analysis with historical controls Sera ni et al. (1988) Prospective analysis with historical controls (GnRHagonist COH versus COH without it) Olivennes et al. (1996) Prospective analysis with historical controls (decreasing low-dose luteal GnRH agonist versus long luteal) Schachter et al. (2001) Prospective study with historical controls (discontinuation of GnRHagonist on day 5 of ovarian stimulation) Pinkas et al. (2000) Prospective study with historical controls (discontinuation of GnRHagonist on cycle day 1 + hmg versus standard long luteal + hmg) Peak E2 <2000 pmol/l, < oocytes retrieved in previous cycle, day FSH >9 and <12 miu/ml < oocytes retrieved in previous cycle, day FSH <12 miu/ml Peak E2 <185.5 pmol/l, < oocytes retrieved in previous cycle, day FSH >7 and <12 miu/ml, one cancelled cycle, age >9 < oocytes retrieved in previous cycle Day FSH >6.5 and <9.5 miu/ml Midluteal GnRH agonist until pituitary desensitization, then ±5 ampoules hmg or puri ed FSH per day, i.m. Midluteal 1 mg leuprolide s.c. to day of menstruation, 5±6 ampoules hmg or puri ed FSH Midluteal 0.5 mg leuprolide s.c. to day of menstruation, 6±8 ampoules hmg or puri ed FSH Leuprolide 1 mg s.c. from day 2, 150±00 IU/day i.m. hmg plus 150±50 IU/day i.m. FSH from day of down-regulation (DR) 0.1 mg leuprolide s.c. from midluteal phase to day of downregulation, then 0.05 mg/day plus ampoules hmg or puri ed FSH One cancelled IVF attempt Midluteal 0.6 mg nafareline s.c. to day 5 of ovarian hyperstimulation, 00 IU/day hmg from day of DR < oocytes retrieved in previous cycle, two cancelled IVF attempts Faber et al. (1998) Prospective analysis Peak E2 <600 pmol/ml, < oocytes retrieved in previous cycle, day FSH >9 IU/l Pu-Tsui et al. (2002) Prospective, non-randomized study <2 oocytes retrieved in previous cycle, day FSH >15 miu/ml, age >0 years Feldberg et al. (199) Retrospective study (comparison between three long luteal GnRH agonist regimens) Day FSH >15 IU/l and two cancelled IVF attempts Midluteal 0.6 mg nafareline s.c. to cycle day 1, 00 IU/day hmg from cycle day Midluteal 0.5 mg leuprolide s.c. to cycle day 2; 6±8 ampoules hmg or puri ed FSH Midluteal 0.5 mg leuprolide s.c. to cycle day 2; ±6 ampoules hmg or puri ed FSH (A) Midluteal 0.1 mg triptorelin s.c. to day of DR, then 0.05 mg/day plus ampoules hmg. (B).75 mg triptorelin on day 21 once. (C) Midluteal 0.55 mg triptorelin s.c. reduced to 0.1 mg on day of down-regulation 0 versus versus versus 5% (P < 0.08) versus versus 6.2 (P = 0.027) 5.9 versus 2.8% 11.1 versus 10.% 18.7 versus 1.% No improvement 6 Increased no. of aspirated oocytes %.% Increased no. of aspirated oocytes 27 versus versus.8 (P < 0.01) versus versus 82.% (P < 0.01) 11 versus 2% 5 versus 7.7% (P < 0.01) Signi cantly increased pregnancy rates, oocytes received 16.% Decreased no. of ampoules used; increased no. of oocytes collected 6 Increased by 28% 8.% 19.% Increased no. of oocytes collected, pregnancy rates, decreased cancellation 9 versus 60 Increased ± 10. versus 2.8% rates Increased no. of oocytes collected, pregnancy rates % 2.0% Improvement % 20.5% Encouraging results 106 (6/29/1).1/1.2/. 11.1/.8/ 21.9% 28.1/6.2/15.6% Increased no. of oocytes collected, lower cancellation rates for group A 6 68

9 Low ovarian response to stimulation for IVF retrieved oocytes and the pregnancy rates (10.7 versus 2.8%) (Pinkas et al., 2000) (see Table I). Subsequently, reducing by a factor of two an already low dose of the GnRH agonist had encouraging results (increased number of oocytes collected, decreased total gonadotrophins used), as shown in another prospective study with historical controls (Olivennes et al., 1996). These authors used 0.1 mg/day leuprolide s.c. from cycle day 21, reducing it to 0.05 mg/day on down-regulation day in 98 cycles. However, the cancellation rate remained high (2%) and the pregnancy rate relatively low (16.%). The above approach was evaluated retrospectively in another study of 106 cycles (Feldberg et al., 199). Triptorelin was used in the same step-down fashion (from 0.1 to 0.05 mg/ day), whereupon a higher number of oocytes and improved pregnancy rates (28.1%) were observed, compared with the higher dose step-down triptorelin (from 0.5 to 0.1 mg/day) or single dose depot triptorelin (.75 mg). Administration of a single dose depot GnRH agonist preparation (leuprolide.75 mg) on day 21 of the pre-stimulated cycle was also assessed in a prospective study with historical controls including 27 cycles (Sera ni et al., 1988). These authors observed a signi cantly increased pregnancy rate and number of oocytes retrieved, while the cancellation rate was decreased. GnRH antagonist regimens The relatively new GnRH antagonist regimens aim to avoid the premature LH surge and, at the same time, to utilize the maximum of the ovarian oocyte cohort by minimizing the suppressing effects of the GnRH analogues on the ovarian receptors, thus avoiding ovarian suppression at the stage of the follicle recruitment (Kenigsberg et al., 198). Only three published studies have addressed this issue. One prospective randomized controlled study comparing the antagonist versus the are-up agonist protocol, reported better results with the are-up regimen. The other two studies (one randomized and one with historical controls) demonstrated an improvement (albeit statistically non-signi cant) with the use of GnRH antagonists. These studies are summarized in Table IV. One group (Craft et al., 1999) used clomiphene citrate (100 mg/ daily, from cycle days 2 to 5) combined with the appropriate dose of gonadotrophins (mean 75 IU/day), in 2 cycles of poorresponding patients. The GnRH antagonist cetrorelix was started on cycle day 6 at a dose of 0.25 mg/day. Compared to previous results with GnRH agonists in the same patients, fewer abandoned cycles (29.2 versus 56.5%), increased number of retrieved oocytes per cycle (6. versus.7) and increased pregnancy rates per transfer (2.5 versus 10%) were observed. There was also a reduction in the amount of gonadotrophin injections used. Nevertheless, the aforementioned results did not reach statistical signi cance. Another prospective randomized study (Akman et al., 2000) reported that the use of GnRH antagonists, together with high doses of gonadotrophins (00 IU/day hmg + 00 IU/ day puri ed FSH from cycle day 2) in previous poor responders, was associated with lower cancellation rates (20 versus 25%) and increased pregnancy rates (20 versus 6.25%), as compared with gonadotrophins alone. However, these differences were not statistically signi cant and no change in the number of the oocytes retrieved was observed. Table IV. GnRH antagonist protocols Comments Score Cancellation rate No. of clinical pregnancies/et Reference Study design Criteria for low response Regimen No of cycles Retrieved oocytes/cycle 5 Better results with the are-up regimen 22.2 versus 26.% 25 versus 20.8% 2 versus 2.5 versus 5.5 (P = 0.02) COC pretreatment, 00 IU/day of hmg + 00 IU/day puri ed FSH from cycle day 2, Cetrorelix 0.25 mg s.c./day when the leading follicle size >1 mm Peak E 2 <500 pg/ml, < oocytes retrieved in previous cycle, day FSH >15 miu/ml, two or more cancelled IVF cycles Prospective randomized trial (antagonist versus are-up agonist) Akman et al. (2001) Non-signi cant improvement 2.5 versus 10% 29.2 versus 56.5% (P = 0.06) 2 versus 2 6. versus.7 Clomiphene citrate 100 mg/day on days 2±6 of cycle; gonadotrophins at a mean dose of 75 IU/day from cycle day 2; Cetrorelix 0.25 mg s.c./day from cycle day 6 One cancelled cycle of IVF, >600 IU FSH per retrieved oocyte, <5 oocytes retrieved in previous cycle, day FSH >18 miu/ml Prospective study with historical controls Craft et al. (1999) 5 Non- signi cantly improved pregnancy rates 20 versus 6.25% 25 versus 20% 20 versus versus.6 00 IU/day hmg + 00 IU/day puri ed FSH from cycle day 2, Cetrorelix 0.25 mg s.c./day when the leading follicle size >1 mm Peak E2 <500 pg/ml, < oocytes retrieved in previous cycle, day FSH >15 miu/ml Prospective randomized trial (GnRH antagonist versus no GnRH analogues) Akman et al. (2000) 69

10 B.C.Tarlatzis et al. The same authors (Akman et al., 2001) in a subsequent prospective, randomized controlled trial, compared the multidose GnRH antagonist protocol with the are-up GnRH- agonist regimen in poor responders (2 cycles in each group). These authors observed signi cantly less oocytes retrieved per cycle (.5 versus 5.5) in the antagonist group (P = 0.02), but no signi cant difference was seen in either the cancellation or pregnancy rates (25 versus 20.8% and 22. versus 26.%, respectively). The available limited data, derived from small or preliminary studies, do not show any advantage from the use of GnRH antagonists. Therefore larger, controlled, prospective randomized trials using GnRH antagonists are necessary to investigate this issue. Adjunctive use of GH or GH-releasing factor or pyridostigmine The hypothesis that GH stimulates ovarian steroidogenesis, follicular development and enhances the ovarian response to FSH was proposed in 1986 (Jia et al., 1986). This action of GH is believed to be mediated via the IGF-1 that acts in synergy with FSH, amplifying its effects on granulosa cells (Adashi and Rohan, 199). This was the theoretical basis for the introduction of GH or GH-releasing factor (GH-RF) in the IVF treatment of poor responders. Usually, to 12 IU GH are administered s.c., commencing on the day of ovarian stimulation with gonadotrophins. A total of nine prospective trials have been reported; four of these were non- randomized but had historical controls. Seven of the studies revealed essentially no change or no signi cant improvement in the clinical results. The studies are summarized in Table V. In one a large multicentre prospective randomized, doubleblind, placebo-controlled trial (Howles et al., 1999), GH-RF was administered and caused an increase in endogenous levels of GH. However, the nal cancellation and pregnancy rates were similar to those found for the protocol without GH-releasing hormone (12.5 versus 16% and 8. versus 8% respectively). No statistically signi cant improvement in cancellation and pregnancy rates were reported in another double-blind, placebo-controlled trial (Suikkari et al., 1996) in which IU/day of GH was used as adjuvant therapy. Increasing the GH dose to 12 IU/day in a long luteal GnRH agonist regimen led to similar results in a prospective study with historical controls (Shaker et al., 1992). The above discouraging ndings were also reported by others (Hughes et al., 199), who treated 21 previous poor responders with 12 IU/day GH in a prospective double-blind, placebocontrolled study and found no signi cant differences in serum E 2 levels, duration of the follicular phase, total hmg dose and number of oocytes between the placebo or GH cycles. Likewise, in a small, prospective, randomized, double-blind, placebocontrolled study, another group administered 18 IU GH on alternative days in a classic are-up triptorelin protocol with 00 IU/day of hmg, in 1 cycles of poor responders (Dor et al., 1995). Unfortunately, the results were also disappointing. The same conclusion was also reached in two additional small studies (Hugues et al., 1991; Levy et al., 199). Optimistic results were reached in a small prospective study with historical controls (Ibrahim et al., 1991), where 10 patients had higher numbers of oocytes collected (7.5 versus.5, P < 0.001) and improved pregnancy rates (60%). GH secretion can also be increased by acetylcholine, which inhibits somatostatin secretion at the hypothalamic level (Delitala et al., 1988). Pyridostigmine is an acetylcholinesterase inhibitor which, by enhancing the action of acetylcholine, can increase GH secretion. This approach was evaluated in a randomized, doubleblind, placebo-controlled study (Chung-Hoon et al., 1999) which included 70 poor responders who were given 120 mg/day pyridostigmine orally, from the day of down-regulation until the day of hcg, along with a long luteal GnRH agonist regimen (triptorelin 0.1 mg on day 21, hmg/fsh 00 IU/day, i.m.). Compared with placebo, pyridostigmine was associated with a signi cantly lower number of ampoules used (8. versus 8.), a higher number of oocytes collected (5.9 versus.7) and improved (but not signi cantly so) pregnancy rates (25.7 versus 11.%). In a recent Cochrane Review, a meta-analysis was conducted of the trials assessing the effectiveness of GH adjuvant therapy in women undergoing ovulation induction (Kotarba et al., 2002). In previous poor responders, the common odds ratio for pregnancy per cycle instituted was 2.55 (95% CI 0.6±10.12). No signi cant difference was noted in either the number of follicles and oocytes, or gonadotrophin usage. Therefore, these published data do not support any bene t from the use of GH as adjuvant therapy in poor responders. Adjunctive use of glucocorticosteroids (dexamethasone) It has been suggested that dexamethasone may directly in uence follicular development and oocyte maturation via its isoform (11- bhsd) in the granulosa cells (Smith et al., 2000) or indirectly, by increasing serum GH and consequently intrafollicular IGF-1 (Miell et al., 199). In addition, it may provoke immunosuppression within the endometrial microenvironment (Polak de Fried et al., 199). To date, no studies have been reported involving poor responders. In one double-blind, placebo-controlled prospective randomized study in 290 cycles of normal responders (aged <1 years), dexamethasone was administered at 1 mg/day in the long luteal protocol until the day prior to oocyte retrieval (Keay et al., 2001), and the authors found a signi cantly lower cancellation rate (2.8 versus 12.%, P = 0.001) (see Table I). These ndings provided great encouragement, as they reveal a very low incidence of poor response with the use of corticosteroids; however, the data are limited and can only be considered as preliminary. Consequently, further trials are needed to support the role of corticosteroids in patients con rmed as poor responders. Adjunctive use of nitric oxide (NO)-donors (L-arginine) In humans, increased vascularization appears to play a critical role in the selection, growth and maturation of follicles in both natural and IVF cycles (Weiner et al., 199). L-Arginine is a potential vasodilator as a NO-donor; in fact, NO is derived in vivo from L-arginine by a NO-synthase enzyme that is either cytokineinducible or calcium-dependent (Moncada et al., 1991). It is also thought that NO is involved in follicular maturation and selection (Anteby et al., 1996), possibly due to its participation in periovulatory vasodilatory modulation, as proven in a rat model (Ben-Shlomo et al., 199). Promising results were presented by one group (Battaglia et al., 1999) in a prospective randomized study in which two groups of 17 poor responders were compared, each of which was treated 70