The number of retrieved oocytes does not decrease during consecutive gonadotrophin-stimulated IVF cycles

Human Reproduction Vol.19, No.4 pp. 899±904, 2004 Advance Access publication February 27, 2004 DOI: 10.1093/humrep/deh178 The number of retrieved oocytes does not decrease during consecutive gonadotrophin-stimulated IVF cycles Evelien J.de Boer 1,2, Isolde Den Tonkelaar 2, Curt W.Burger 3, C.W.N.Looman 4, Flora E.van Leeuwen 1,6, Egbert R.te Velde 5 and on behalf of the OMEGA project group 1 Department of Epidemiology, Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, 2 International Health Foundation, Europalaan 506, 3526 KS, Utrecht, 3 Department of Gynaecology and Obstetrics, Erasmus Medical Center Rotterdam, Postbus 2040, 3000 CA, Rotterdam, 4 Department of Public Health, Erasmus Medical Center, University Medical Center Rotterdam, Postbus 1738, 3000 DR, Rotterdam and 5 Department of Reproductive Medicine, Division of Obstetrics, Neonatology and Gynaecology, University Medical Center Utrecht, Postbus 85500, 3508 GA, Utrecht, The Netherlands 6 To whom correspondence should be addressed. E-mail: f.v.leeuwen@nki.nl BACKGROUND: Our aim was to study whether there is a decreasing trend in the number of retrieved oocytes in women who had all undergone at least seven consecutive IVF cycles. METHODS: A nationwide retrospective cohort study was conducted among women whose rst IVF cycle was stimulated with gonadotrophins in The Netherlands between 1983 and 1995. Among these eligible women, we identi ed all women who had received at least seven consecutive IVF cycles (n = 330). Poisson regression analysis was used to assess the trend in the number of retrieved oocytes over the rst six IVF cycles. RESULTS: The unadjusted results showed a slight but non-signi cant decrease in the number of retrieved oocytes over six IVF cycles. The change in cycle 6 compared with cycle 1 was ±0.06 oocytes (5.8% decrease) (P = 0.21). When adjusting for the number of ampoules and the stimulation protocol (fertility drug used combined with use of GnRH agonists), there was a considerable and highly signi cant decrease from cycle 1 to cycle 6 [change in cycle 6 compared with cycle 1: ±0.19 oocytes (17.4% decrease), (P < 0.0001)]. However, when adjusting for age of the women, this decrease almost completely disappeared [change in cycle 6 compared with cycle 1: ±0.05 oocytes (5% decrease), (P = 0.50)]. CONCLUSION: The results suggest that there is no decrease in the number of retrieved oocytes over subsequent cycles when simultaneously accounting for the increasing age of the women, differences in the number of ampoules of gonadotrophins used, type of stimulation protocol and year of IVF treatment. Key words: consecutive cycles/ivf/ovarian response Introduction Since the introduction of IVF in The Netherlands in 1983, IVF has been widely used for treating sub/infertile couples. In IVF, controlled ovarian stimulation by gonadotrophins is used, aiming at multiple follicle growth. FSH has a predominant role in the development of follicles, the selection of the dominant follicle and further maturation. The role of FSH in the earlier stages of follicular development is less clear. Experiments and observations in humans show that follicles are able to grow into the selectable stage without the presence of gonadotrophins. However, there is some evidence that FSH, although not obligatory, facilitates the initiation of follicular growth or stimulates early follicular development. If so, it can be expected that the high FSH levels reached during treatment may increase the depletion of the follicle pool by stimulating the resting follicles to grow (Gougeon et al., 1994; Te Velde et al., 1998). Moreover, it is possible that the IVF puncture has some destructive effect on the capillaries and follicles of the ovary, the effect of which may accumulate during repeated punctures. Repeated punctures may also induce the release of autoantigens, causing a decline of the follicle pool (Gobert et al., 1992). If this is true, a decrease in the number of retrieved oocytes in women with many IVF cycles is to be expected. Moreover, such women might have an earlier menopause, as has already been mentioned by others (Fauser et al., 1999; Pines et al., 2002). The available literature concerning ovarian response in repeated IVF cycles is limited, based on small numbers of women or a limited number of treatment cycles (Ron-el et al., 1990; Yovel et al., 1994; Caligara et al., 2001; Hoveyda et al., 2002; Kolibianakis et al., 2002). Moreover, the presented evidence is highly controversial; some studies suggest that repeated IVF cycles lead to a decrease of the number of retrieved oocytes (Kolibianakis et al., 2002), while other studies deny such an effect (Caligara et al., 2001; Hoveyda et al., 2002). Human Reproduction vol. 19 no. 4 ã European Society of Human Reproduction and Embryology 2004; all rights reserved 899

E.J.de Boer et al. Figure 1. Description of the recruitment of eligible women. The aim of the present study was to examine whether there is a decreasing trend in the number of retrieved oocytes in women who had undergone at least seven consecutive IVF cycles. Materials and methods Study population and study procedures The study population, study procedures and data collection methods have been described previously in detail (Klip et al., 2001; de Boer et al., 2002, 2003). Brie y, subjects were participants in a nationwide cohort study of 19 242 IVF-treated women in The Netherlands between 1983 and 1995. The main purpose of this study is to examine the risk of hormone-related cancers in IVF-treated women. All institutional ethics committees of the participating clinics approved the study procedures. Between 1997 and 2000, all women received a questionnaire and an informed consent for data abstraction from the medical records. Data collection and selection Research assistants abstracted the medical records to obtain information on the diagnosis and each treatment cycle, including duration of ovarian stimulation, the number of retrieved oocytes, number of embryos transferred and whether the IVF treatment resulted in a pregnancy leading to live birth. Furthermore, data were collected on date of birth, weeks of gestation and the occurrence of multiple births (questionnaire). Figure 1 displays a graphical presentation of the study population with various exclusion criteria. From the 9942 women who 900 returned the questionnaire and whose medical record had been abstracted, we selected those whose rst IVF cycle was stimulated with gonadotrophins only (n = 7960). Women whose rst ovarian stimulation was done with clomiphene citrate, either alone or in combination with gonadotrophins, were excluded. The reason for these exclusions was to reduce a possible bias, which might occur by comparing gonadotrophin-stimulated IVF cycles with non-gonadotrophin-stimulated cycles, and comparability with the literature available. Of the 7960 eligible women, we identi ed all women who had received at least seven IVF cycles (n = 342). From those, women were excluded who had insuf cient data in three or more of their IVF cycles, i.e. unknown date of the IVF treatment, unknown number of retrieved oocytes or unknown number of ampoules of HMG/FSH. Finally, 330 women who underwent a total of 1980 cycles were eligible for the present study. De nitions of variables and methods of analysis We analysed whether there was a decreasing trend in the number of oocytes retrieved over the rst six IVF cycles of women who underwent >7 IVF cycles. We restricted the analyses to the rst six cycles, to make sure that the number of retrieved oocytes during the last cycle in our analysis was not related to the reason for stopping further treatment. Duration of sub/infertility was de ned as the period between the start of involuntary childlessness and the date of the rst visit to the IVF centre.

Oocyte number after successive stimulated IVF cycles Table I. Characteristics of ovarian stimulation and response in women who underwent >7 IVF cycles ( rst six cycles) n = 330 Cycle 1 Cycle 2 Cycle 3 Cycle 4 Cycle 5 Cycle 6 Mean age at IVF treatment (6 SD) 32.8 6 3.7 33.3 6 3.6 33.7 6 3.6 34.3 6 3.6 35.0 6 3.5 35.6 6 3.5 Mean year of IVF treatment (6 SD) 1990.8 6 2.4 1991.3 6 2.5 1991.8 6 2.4 1992.4 6 2.5 1993.0 6 2.5 1993.6 6 2.5 Mean 6 SD number of months between cycles All women ± 5.6 6 6.6 5.6 6 7.8 7.0 6 7.6 7.7 6 9.3 7.0 6 7.9 Women with >1 live births (n = 128) ± 7.0 6 8.2 7.4 6 10.9 9.1 6 9.6 10.6 6 11.2 8.5 6 8.7 Women with no live births (n = 202) ± 4.7 6 5.1 4.5 6 4.4 5.8 6 5.7 5.9 6 7.3 6.0 6 7.2 Mean duration of ovarian stimulation, days 6 SD a 9.0 6 2.6 9.4 6 2.8 9.5 6 3.0 9.9 6 3.0 10.2 6 3.0 10.4 6 3.0 Use of agonists, n (%) No 121 (37) 100 (30) 81 (25) 74 (22) 67 (20) 55 (17) Yes 205 (62) 230 (70) 247 (75) 254 (77) 263 (80) 270 (82) Unkown 4 (1) ± 2 (1) 2 (1) ± 5 (2) Type of gonadotrophin used, n (%) HMG only 269 (82) 271 (82) 264 (80) 261 (79) 257 (78) 257 (78) FSH only 7 (2) 1 (±) 6 (2) 10 (3) 16 (5) 24 (7) HMG and FSH 54 (16) 56 (17) 57 (17) 55 (17) 54 (16) 48 (15) Unknown ± 2 (1) 3 (1) 4 (1) 3 (1) 1 (±) Mean 6 SD number of ampoules HMG/FSH b 22.4 6 8.5 26.1 6 17.3 26.9 6 13.8 28.3 6 14.9 30.8 6 16.8 32.7 6 18.0 Mean 6 SD number of retrieved oocytes c 8.2 6 7.1 7.2 6 5.4 7.8 6 5.5 7.7 6 5.2 7.8 6 6.1 7.7 6 6.0 Reasons for cycle cancellation, n (%) Anticipated poor response 37 (11) 34 (10) 22 (7) 23 (7) 25 (8) 39 (12) Risk of ovarian hyperstimulation syndrome (OHSS) 6 (2) 6 (2) 3 (1) 6 (2) 1 (±) 2 (1) a Known for 1951 IVF treatment cycles. b Known for 1951 IVF treatment cycles. c For cancelled cycles due to anticipated poor response or risk of OHSS, the numbers of retrieved oocytes were set at one and 20, respectively. Analyses were processed with SPSS â (SPSS Inc., Chicago, IL, version 11.0) and GLIM (Generalized Linear Interactive Modelling; NAG, Oxford, UK). Differences between group means of the rst cycle and the sixth cycle were tested by Student's t-test (pairwise) or Wilcoxon signed-rank test, and distributions over categories by c 2 test. A two-sided P-value <0.05 was considered as statistically signi cant. Poisson regression analysis in GLIM was used to assess the trend in the number of retrieved oocytes during the rst six cycles. The cycle number was entered as a categorical predictor with six levels. For cycles which had been cancelled due to an anticipated poor response (insuf cient or no follicle growth) or risk of an ovarian hyperstimulation syndrome (OHSS), the number of retrieved oocytes was set at one or 20, respectively. If the cycle was cancelled for another reason (n = 63), the number of retrieved oocytes was considered to be random missing and the cycle was removed from the data set. In the analyses, women were included as a so-called `nuisance parameter' (Aitkin et al., 1989), meaning that all results were corrected for the variable `woman' as if it was a categorical variable. In this way, the comparisons between cycles are made within the set of observations of a woman. Another consequence was that the linear effects of age at treatment and year of treatment were not identi able. However, possible non-linear effects for age and time still remained distinguishable. To account for dependency of data within a woman and for overdistribution, the results were bootstrapped (Efron and Tibshirani, 1993). A thousand replicas (samples with replacement out of the observed set with `woman' as sampling unit and the same size as the original set) were estimated and con dence intervals were based on percentiles. Poisson regression analyses were performed, with and without adjusting for confounders. In addition to age and year of treatment, other potential confounders were the number of gonadotrophin ampoules used during stimulation (continuous), type of fertility drug (four categories: HMG only, puri ed FSH/recombinant FSH, HMG/ FSH and unknown) use of agonist (yes, no) and year of treatment (continuous). Type of fertility drug and the use of agonists were combined as type of stimulation protocol (three categories; gonadotrophins with agonists, gonadotrophins without agonists, and unknown). If the number of ampoules of HMG/FSH was missing in one or two cycles (n = 29), the number of ampoules was imputed with the predicted value (resulting in a range from 15 to 40) based upon year of IVF treatment, age at IVF treatment, number of IVF cycles and type of fertility drug used. The use of recombinant FSH and puri ed FSH for stimulation was limited and these were therefore combined. Adjusting for confounders was evaluated by adding each confounder separately as splines into the regression model to allow for non-linear effects (Harrell, 2002). Results The sub/infertility diagnoses of the 330 women who underwent at least seven consecutive cycles were tubal (53%), male factor (27%), unexplained (12%) and unknown (9%). The mean duration of sub/infertility was 4.0 years (SD 62.7). Of all IVF treatment cycles included in the analysis (n = 1980), 63 IVF cycles were cancelled due to reasons other than anticipated poor response or risk of OHSS. In the whole group of women, the number of women who gave birth to a child either once or twice during the six consecutive IVF cycles was 124 and four, respectively. The live birth rates per cycle were 5, 5, 8, 10, 8 and 5% from the rst to the sixth cycle, respectively. Table I presents the characteristics of stimulation response in six cycles. The mean age at IVF treatment increased by almost 3 years, from 32.8 in the rst cycle to 35.6 in the sixth cycle. As expected, the same increase from the rst to the sixth IVF cycle was observed for year of treatment. From this, the average time between the cycles could be calculated. This time was very dependent on whether or not a woman had delivered a child following an IVF pregnancy. The mean and median number of oocytes retrieved differed considerably, and were 8.2 and 6.0, respectively in cycle 1, indicating that the number of retrieved oocytes were not normally distributed. This was caused by the relatively few cycles with very many oocytes retrieved and a 901

E.J.de Boer et al. Figure 2. The mean number of retrieved oocytes ( lled circles) and the number of retrieved oocytes per 20 ampoules of HMG/FSH (open squares) per IVF cycle number. relatively high percentage of cycle cancellations due to anticipated poor response. The mean number of ampoules of gonadotrophins, the interval between IVF cycles and the duration of ovarian stimulation also increased over cycles. In cycle 2, the mean number of retrieved oocytes was signi cantly lower than in cycle 1 and cycle 3: 8.2 in cycle 1, 7.2 in cycle 2 and 7.8 in cycle 3. It was expected that the number of ampoules of gonadotrophins used and age at IVF treatment would be important confounders when assessing the trend over cycles in the number of oocytes retrieved. Figure 2 presents the mean number of retrieved oocytes during the six IVF cycles, which decreased from 8.2 in the rst IVF cycle to 7.7 at cycle 6 (P = 0.54). However, when relating the decrease to the number of ampoules used, the decrease appeared to be much larger. For example, the mean number of oocytes retrieved per 20 ampoules of gonadotrophins decreased from 8.0 in cycle 1 to 6.1 in cycle 6 (P < 0.0001). With increasing age, the number of retrieved oocytes more than halved from ages 25 to 41 (Figure 3). Figure 4 presents the results of the Poisson regression analysis for the relative decrease in the number of collected oocytes compared with cycle 1. Without adjustment, there was a slight but non-signi cant decrease. The change in cycle 6 compared with cycle 1 was ±0.06 oocytes (5.8% decrease) (P = 0.21, not shown). When adjusting for the number of ampoules and stimulation protocol, there is a considerable and highly signi cant decrease from cycle 1 to cycle 6 [change in cycle 6 compared with cycle 1: ±0.19 oocytes (17.4% decrease), P < 0.0001]. However, when adjusting for age of the woman, this decrease almost completely disappeared [change in cycle 6 compared with cycle 1: ±0.05 oocytes (5% decrease), P = 0.50]. The results for the cycles 3±5 are completely in line with those in cycle 6. In contrast, results in cycle 2 differed from those in all other cycles in that the number of oocytes retrieved in cycle 2 demonstrated a statistically signi cant decrease, also after all adjustments. Discussion Our study is the rst one examining whether or not repeated stimulation and puncture for IVF leads to a decrease in the number of retrieved oocytes in a large group of women 902 Figure 3. The mean number of retrieved oocytes with age at IVF treatment. Figure 4. The percentage decrease in the number of retrieved oocytes relative to cycle 1: unadjusted model ( lled squares), adjusted for the number of ampoules and the stimulation protocol (open triangles) and adjusted for age, year of the treatment in addition to number or ampoules and the stimulation protocol (crosses). undergoing at least six cycles of IVF, using appropriate methods of analysis. The available literature on this subject is controversial and methods of analysis are limited. Repeated IVF cycles may (Al-Azemi et al., 2000; Kolibianakis et al., 2002) or may not (Ron-el et al., 1990; Caligara et al., 2001; Hoveyda et al., 2002) lead to a decrease in the number of retrieved oocytes. However, with the exception of the study by Kolibianakis et al. (2002), the number of women with more than three consecutive cycles was small. Since ovarian puncture might also contribute to a decrease of the number of retrieved oocytes over time, the ideal study design should include a comparison group of women who had undergone multiple stimulated intrauterine inseminations. Our study includes a homogeneous group of women, meaning that we followed the same women through all of

their six IVF cycles. In the literature available, only Hoveyda et al. (2002) included a homogeneous group of women. However, one limitation of their study was that the analyses included information of the last cycle which is likely to deviate from that in the other cycles. This was re ected in the higher pregnancy rate observed in the third cycle, which was the result of an analysis containing data from both women who stopped and women who continued IVF treatment after the third cycle. In the literature available, little attention has been given to the complicated time dimension in studies with several consecutive cycles in the same women, taking into account the facts that women grow older and that stimulation protocols, but also (other) potentially important effects, may change over time. The average time between the rst and sixth cycle in our study was ~3 years, but exceeded 5 years in 8% of all women. Many changes in IVF procedures take place during such a period of time. For example, since the introduction of IVF in The Netherlands in 1983, IVF regimes have changed profoundly. In the early 1980s, most stimulations were followed by a laparoscopic ovarian puncture. Unfortunately, information on whether the oocyte retrieval per cycle was done laparoscopically or sonographically by the transvaginal route was not available. However, this problem was taken into account by including year of IVF treatment in the multivariate model. The medication used in IVF treatment changed from clomiphene citrate to HMG since 1984, and more recently to puri ed FSH and recombinant FSH. Since the number of oocytes retrieved depends on the type of fertility drug and the dosage used, we selected women whose rst IVF cycle was stimulated with gonadotrophins only. In our data set, agonists were rst introduced in 1985, and a possible effect of this was taken into account by including whether or not agonists were used in the stimulation protocol. Considering these changes over time in an appropriate manner is complicated and demands a sophisticated, multivariate approach to the statistical analysis. By restricting the analyses to the number of retrieved oocytes in the rst six cycles only, the analyses were based on a homogeneous group of women. Information from the seventh cycle was not used since the number of retrieved oocytes during that cycle might deviate from that in the other cycles, and be related to the reason for stopping IVF treatment (e.g. successful treatment or very low oocyte response). A limitation of our study was that many women had to be excluded from the analysis because their stimulation protocol of the rst IVF treatment was unknown (39% of all excluded women) and a possible selection bias might have been introduced. However, of these 773 excluded women, only 28 had undergone >7 IVF cycles. In the Poisson regression analysis, no signi cant decrease emerged in the number of retrieved oocytes during six consecutive IVF cycles of these 28 women, rendering selection bias unlikely. We intentionally selected an extreme subgroup of IVFtreated women undergoing >7 IVF cycles. Even in this group of women, we observed no decreasing trend in the number of retrieved oocytes over the rst six cycles, rendering it highly unlikely that repeated gonadotrophin stimulation leads to a decrease in the number of retrieved oocytes in a less selected Oocyte number after successive stimulated IVF cycles subgroup of IVF-treated women. To investigate a possible selection bias, we compared women who underwent >4, but <7 IVF cycles women with the women included in the present study. In women who all had undergone at least four cycles, but <7, IVF cycles (n = 1833), we also observed no decreasing trend in the number of retrieved oocytes, rendering selection bias unlikely. Moreover, the mean and median numbers of retrieved oocytes in the rst cycle were 8.9 and 7, respectively, which is in line with the results obtained in the present study for women who underwent at least seven cycles. The signi cant decrease in the number of oocytes retrieved in cycle 2 with a subsequent rise again in cycle 3 probably re ects the usual adjustments clinicians make when the numbers of oocytes retrieved are either too high or too low in the rst cycle, in which the standard dose is applied. For example, of the 31 women who had >20 oocytes in cycle 1, the mean dose of ampoules was decreased from 22.9 in cycle 1 to 19.2 in cycle 2, with a subsequent decrease of 24.7 and 13.2 oocytes from cycle 1 to cycle 2, respectively. Without these 31 women, the mean number of oocytes obtained in cycles 1±3 remained stable at ~6.4. Likewise, the rise in the mean number of oocytes in cycle 3 relative to cycle 2 is probably due to increasing the dose for low responders in cycle 2, as a consequence of the lower dose administered in cycle 2. Adjustments of the dose probably occur more often, but are likely to be more common in the second and third cycles. They re ect the `learning curve' of the clinician in his attempts to reach the `optimal' dose in individual women. At rst sight, the data presented in Figure 2 suggest that the lack of decrease in the number of oocytes retrieved from cycle 1 to cycle 6 is related to increasing the doses of gonadotrophins. It seems, therefore, that only when gonadotrophin doses are increased the number of retrieved oocytes can be maintained at a constant level. This impression is in line with the curve resulting after adjustment for the number of ampoules and the stimulation protocol (Figure 4: rst adjustment). After adjustment for the dose of gonadotrophins used and the stimulation protocol applied, a statistically signi cant drop in the number of oocytes occurred. However, the number of follicles remaining in both ovaries strongly depends on the age of the woman. An exponential decline from fetal life onwards is known to occur, with an acceleration after age 35 (Faddy et al., 1992). Most women taking part in this study were treated between the ages of 30 and 40 years. When, in addition to the number of ampoules and stimulation protocol, age and year of treatment were taken into account, no decrease in number of retrieved oocytes over cycles appeared to be present any longer (Figure 4: nal adjustment). Therefore, we conclude that the results of this study demonstrate that there is no decreasing trend in the number of oocytes retrieved in six consecutive IVF cycles. Acknowledgements We are greatly indebted to the participants of the OMEGA project. This study would not have been possible without the efforts of all women who participated. We owe special thanks to H.Klip (PhD) without whose effort the OMEGA project would never been so successful. We are especially grateful to the research assistants 903

E.J.de Boer et al. M.Schippers, I.M.Versteegden, S.Braak, A.H.W.van den Belt- Dusebout, G.M.Plas, I.van Gils and I.Verburg for abstracting data from the medical les in the participating hospitals. We would also like to thank the medical registries of the participating clinics for making patient selection possible; and all attending physicians for providing access to their patients' medical les. The OMEGA project group includes the following persons: M.Kortman, MD (University Medical Center Utrecht), N.Macklon, MD PhD (Erasmus Medical Center, Rotterdam), C.A.M.Jansen, MD PhD (Diaconessenhuis, Voorburg), R.A.Leerentveld, MD PhD (Isala Clinics, Zwolle), W.N.P.Willemsen, MD PhD (Academic Hospital Nijmegen, St Radboud), R.Schats, MD PhD (Academic Hospital Free University, Amsterdam), N.Naaktgeboren, PhD, and F.M. Helmerhorst, MD PhD (Leiden University Medical Center), R.S.G.M.Bots, MD PhD (St Elisabeth Hospital, Tilburg), A.H.M.Simons, MD (Academic Hospital Groningen), H.V.Hogerzeil, MD PhD (Academic Medical Center, Amsterdam), J.L.H.Evers, MD PhD (Academic Hospital Maastricht), and P.A.van Dop, MD PhD (Catharina Hospital, Eindhoven). References Aitkin M, Anderson D, Francis B and Hinde J (1989) Statistical Modelling in GLIM. Clarendon Press, Oxford. Al-Azemi M, Bernal AL, Steele J, Gramsbergen I, Barlow D and Kennedy S (2000) Ovarian response to repeated controlled stimulation in in-vitro fertilization cycles in patients with ovarian endometriosis. Hum Reprod 15,72±75. Caligara C, Navarro J, Vargas G, Simon C, Pellicer A and Remohi J (2001) The effect of repeated controlled ovarian stimulation in donors. Hum Reprod 16,2320±2323. deboer EJ, Den Tonkelaar I, Te Velde ER, Burger CW, Klip H and van Leeuwen FE (2002) Low number of retrieved oocytes at IVF treatment is predictive of early menopause. Fertil Steril 77,978±985. deboer EJ, Den Tonkelaar I, Te Velde ER, Burger CW and van Leeuwen FE (2003) Increased risk of early menopausal transition and natural menopause after poor response at rst IVF treatment. Hum Reprod 18,1544±1552. Efron B and Tibshirani RJ (1993) An Introduction to the Bootstrap. Chapman and Hall, New York. Faddy MJ, Gosden RG, Gougeon A, Richardson SJ and Nelson JF (1992) Accelerated disappearance of ovarian follicles in mid-life: implications for forecasting menopause. Hum Reprod 7,1342±1346. Fauser BC, Devroey P, Yen SS, Gosden R, Crowley WF, Jr, Baird DT and Bouchard P (1999) Minimal ovarian stimulation for IVF: appraisal of potential bene ts and drawbacks. Hum Reprod 14,2681±2686. Gobert B, Barbarino-Monnier P, Guillet-May F, Bene MC and Faure GC (1992) Anti-ovary antibodies after attempts at human in vitro fertilization induced by follicular puncture rather than hormonal stimulation. J Reprod Fertil 96,213±218. Gougeon A, Echochard R and Thalabard JC (1994) Age-related changes of the population of human ovarian follicles: increase in the disappearance rate of non-growing and early-growing follicles in aging women. Biol Reprod 50,653±663. Harrell FE (2002) Regression Modeling Strategies. Springer Verlag, New York. Hoveyda F, Engmann L, Steele J, Lopez Bernal A and Barlow DH (2002) Ovarian response in three consecutive in vitro fertilization cycles. Fertil Steril 77,706±710. Klip H, Burger CW, de Kraker J and van Leeuwen FE (2001) Risk of cancer in the offspring of women who underwent ovarian stimulation for IVF. Hum Reprod 16,2451±2458. Kolibianakis E, Osmanagaoglu K, Camus M, Tournaye H, Van Steirteghem A and Devroey P (2002) Effect of repeated assisted reproductive technology cycles on ovarian response. Fertil Steril 77,967±970. Pines A, Shapira I, Mijatovic V, Margalioth EJ and Frenkel Y (2002) The impact of hormonal therapy for infertility on the age at menopause. Maturitas 41,283±287. Ron-el R, Raziel A, Herman A, Golan A, Nahum H, Soffer Y and Caspi E (1990) Ovarian response in repetitive cycles induced by menotrophin alone or combined with gonadotrophin releasing hormone analogue. Hum Reprod 5,427±430. TeVelde ER, Scheffer GJ, Dorland M, Broekmans FJ and Fauser BC (1998) Developmental and endocrine aspects of normal ovarian aging. Mol Cell Endocrinol 145,67±73. Yovel I, Geva E, Lessing JB, Yaron Y, Botchan A and Amit A (1994) Analysis of the fourth to eighth in-vitro fertilization treatments after three previously failed attempts. Hum Reprod 9,738±741. Submitted on August 11, 2003; resubmitted on October 28, 2003; accepted on December 15, 2003 904