Cumulative probability of achieving an ongoing pregnancy after in-vitro fertilization and intracytoplasmic sperm injection according to a woman's age, subfertility diagnosis and primary or secondary subfertility

A.M. Stolwijk1,3, A.M.M. Wetzels2 and D.D.M. Braat2

1 Department of Epidemiology, University of Nijmegen and 2 Department of Obstetrics and Gynaecology, University Hospital Nijmegen, Nijmegen, The Netherlands


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The aim of this study was to estimate reliable cumulative probabilities of achieving an ongoing pregnancy after successive in-vitro fertilization or intracytoplasmic sperm injection (IVF/ICSI) cycles, according to a woman's age, subfertility diagnosis and primary or secondary subfertility. Therefore reasons for quitting treatment without achieving an ongoing pregnancy were taken into account. Moreover, we studied whether there were trends in cumulative probabilities after adjustment for potential confounding effects of the other two characteristics, duration of subfertility, year of first treatment and reason for quitting treatment. In total, 2984 IVF/ICSI cycles were performed in 1315 couples at the University Hospital Nijmegen, The Netherlands, between 1991 and 1998. The `realistic' cumulative probability of achieving an ongoing pregnancy was 54.5% after five consecutive IVF/ICSI cycles, which was about 10% lower (absolute value) than the optimistic probability calculated by life-table analysis and about 10% higher (absolute value) than the most pessimistic estimate. Women of 35 years or younger had a higher probability of achieving an ongoing pregnancy than the older women. As ICSI is now an option, there were no obvious differences between the subfertility diagnosis subgroups. The cumulative probability after the first two IVF/ICSI cycles was higher in women with secondary subfertility than in those with primary subfertility; this advantage disappeared after further treatment. These trends remained valid after adjustment for confounding factors.

Key words: ICSI/IVF/pregnancy/probability/subfertility


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
It is important for patients who are considering (the next) treatment with in-vitro fertilization (IVF) to know the probability of success after several IVF treatments. Various studies have considered this cumulative probability of success, which was often defined as a clinical pregnancy or live birth. As some people may quit treatment because of poor results, and others may quit for reasons that are unrelated to the IVF results, the reason for quitting treatment should be taken into account when calculating the cumulative probability. A method used frequently is life-table analysis. This method assumes inherently that those who quit treatment have the same probability of success as those who continue treatment. This may overestimate the cumulative probability. In a previous study, we showed that the cumulative probability of achieving a pregnancy in IVF patients who were treated at the University Hospital Nijmegen, The Netherlands, between 1988 and 1993 was greatly overestimated by life-table analysis when the reason for quitting treatment was not taken into account (Stolwijk et al., 1996Go). These findings were confirmed at another Dutch IVF centre (Land et al., 1997Go).

The success of IVF is expected to depend on, amongst other things, the age of the woman, the subfertility diagnosis and primary or secondary subfertility. This information is known for each couple that considers IVF treatment. Therefore separate figures are warranted for subgroups. Several papers report the cumulative probability of achieving a pregnancy based on life-table analysis for age groups and subfertility diagnosis (Tan et al., 1992Go, 1994Go; Check et al., 1994Go; Alsalili et al., 1995Go; Dor et al., 1996Go; Kodama et al., 1996Go; De Mouzon et al., 1998Go; Prietl et al., 1998Go). In general, the younger the woman, the higher the cumulative probability of success. The differences in results per subfertility group are less obvious. Male subfertility showed somewhat poorer results than other subfertility diagnoses; intracytoplasmic sperm injection (ICSI) as an option during IVF treatment was not included in these studies. It has been observed that the cumulative probability of achieving clinical pregnancy in women with primary subfertility problems was lower than that in women with secondary subfertility problems, including those who became pregnant after some kind of fertility treatment (54 versus 79% after eight IVF cycles with follicle puncture; P < 0.01) (Prietl et al., 1998Go). Another indication for differences in results between primary and secondary subfertility can be derived from a study that compared patients with and without previous IVF pregnancies; the group with former IVF pregnancies showed the best results (Tan et al., 1994Go). As far as we know, no cumulative probabilities of success are described in the literature for IVF treatments in which ICSI was used if necessary. In several of the studies mentioned above which used life-table analysis, the authors noticed that the cumulative probability of success after consecutive cycles was much higher than the total percentage of success achieved in couples who started IVF treatment.

In this study we aimed to estimate realistic cumulative probabilities of achieving ongoing pregnancy after successive IVF cycles, on the basis of a woman's age, subfertility diagnosis and primary or secondary subfertility. Therefore we took into account the reason for quitting treatment without achieving an ongoing pregnancy. We included the results from ICSI treatments, which was offered to patients if considered necessary. In addition, we studied whether there were any differences in the cumulative probabilities of achieving ongoing pregnancy between subgroups after adjustment for the effects of other variables, including the reason for quitting treatment.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Data were used from 1315 couples who underwent their first IVF treatment during the period March 1991 to December 1997 at the University Hospital Nijmegen, The Netherlands. Women who were <=41 years and did not have very high basal follicle stimulating hormone (FSH) concentrations (<20 IU/l) and older women with normal basal FSH levels (<=10 IU/l) could be treated with IVF. Out of the 1315 women, 14 were >=42 years at their first IVF cycle. Data were included from the treatments that were performed during the period March 1991 to December 1998. This allowed most of the couples to complete their desired number of IVF treatments and incorporated a follow-up that was long enough to know the outcome of the last treatment. Since September 1994, ICSI has been available at this hospital in the case of severe male subfertility (<106 motile spermatozoa) or after total fertilization failure twice in couples with normal semen characteristics.

Success after IVF/ICSI was defined as achieving an ongoing pregnancy, i.e. a pregnancy that continued for at least 12 weeks after embryo transfer. Because of incomplete follow-up data after that time, we could not estimate the cumulative probabilities of live births.

All cycles were included in which ovarian stimulation was started, whether or not oocyte aspiration and/or embryo transfer was performed. Data from transfers of cryopreserved embryos were also included. Data from treatments were used up to and including the one in which an ongoing pregnancy was achieved. Cycles were excluded if donor spermatozoa were used; donor oocytes were never used. Cycles were also excluded if the spermatozoa were obtained by means of microepididymal sperm aspiration (MESA) or testicular sperm extraction (TESE), because in The Netherlands these procedures have been allowed only for a very short period. In the population under study, MESA or TESE was performed from the first IVF cycle onwards in 14 couples; as stated above data from these couples were excluded. In another four couples MESA or TESE was performed during the second and/or third IVF cycle; the results of their first cycles without MESA or TESE are included in this study. The maximum number of IVF/ICSI cycles per couple included in this study was five, as only six couples underwent six or more cycles. During most IVF/ICSI cycles (93.8%) ovarian stimulation was performed by means of a long protocol of gonadotrophin-releasing hormone agonist, which was started on day 21 of the previous cycle, followed by human menopausal gonadotrophin (HMG).

The cumulative probability of achieving pregnancy after x cycles can be calculated by: [1 – {pi} (1 – number of pregnant women in cycle x/number of women at risk in cycle x)]x100%, where {pi} indicates the product of the terms specified between round brackets (Kalbfleisch and Prentice, 1980Go). Several assumptions have been made to estimate the cumulative probability of pregnancy; these assumptions influence the number of women `at risk' in each cycle: (i) optimistic assumption (conform to life-table analysis): women who stopped treatment without achieving an ongoing pregnancy had the same probability of achieving an ongoing pregnancy after IVF/ICSI as those who continued; (ii) realistic assumption: women who stopped treatment without achieving an ongoing pregnancy because of a medical indication had no chance of achieving an ongoing pregnancy after IVF/ICSI, while those who stopped treatment for other reasons had the same probability of achieving an ongoing pregnancy after IVF/ICSI as those who continued; (iii) pessimistic assumption: women who stopped treatment without achieving an ongoing pregnancy had no chance of achieving an ongoing pregnancy after IVF/ICSI. A medical indication to quit treatment was given in the case of either a poor ovarian response defined as three or fewer oocytes retrieved after a high dose (>=4 ampoules per day) of HMG or recombinant follicle stimulating hormone, or total fertilization failure (TFF) defined as <5% of the oocytes fertilized by ICSI while five or more oocytes were present. Analyses were stratified according to the women's age (<=30, 31–35, >=36 years), subfertility diagnosis [tubal pathology alone, male factor alone, a combination of tubal pathology and male factor (no other subfertility diagnosis), other diagnoses (i.e. cervical factor and/or endometriosis, whether or not in combination with tubal pathology or male factor), and idiopathic subfertility], and primary or secondary subfertility of the woman. These three variables were measured at the start of the first IVF/ICSI cycle.

Chi-square ({chi}2) tests were used to study whether there were any associations between variables. The log-rank test and the Wilcoxon test were used to test whether there were any differences in `survival' (i.e. not achieving an ongoing pregnancy) between subgroups. The former test emphasizes differences between the tails, the latter between the beginnings of the curves. If there was more than one independent variable in the survival model, Cox proportional hazard models or Cox extended models were used.

Survival analysis was performed to derive hazard ratios that compared the cumulative probability of achieving ongoing pregnancy during successive cycles between subgroups while adjusting for the effects of the other variables. An estimation of the cumulative probability of achieving a pregnancy cannot be derived from this analysis. Cox proportional hazard models were applied if only time independent variables were included in the model, whereas Cox extended models were applied if at least one time dependent variable was included. Time was defined in the model by the IVF/ICSI cycle number. Ties in failure time were handled by using the `discrete' value in the PHREG procedure of the statistical software of SAS. An event was defined as the occurrence of an ongoing pregnancy (in the IVF/ICSI cycle itself or in any cryopreserved embryo transfer following that IVF/ICSI cycle). As we were interested in the impact of a woman's age, subfertility diagnosis and primary or secondary subfertility of the woman on the cumulative probability of achieving an ongoing pregnancy, these variables were forced into the model in subgroups as described above. Adjustments were made for potential confounding effects of the duration of subfertility and the year of first treatment. All these variables were measured at the start of the first IVF/ICSI cycle. Moreover, as we presume that the reason for censoring (i.e. not achieving an ongoing pregnancy) may not have occurred at random, or in other words, that those censored and not censored might not have the same chance of achieving an ongoing pregnancy, survival analysis cannot be used without taking this information into account. Therefore we included an indicator for the expected outcome in the model: a medical indication to quit treatment, as defined above. As this variable can change at every IVF/ICSI cycle, a Cox extended model was used in which a medical indication was included as a time dependent variable. Cox extended models were also used if any of the other variables in the model showed time dependency.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The 1315 couples underwent a total of 2984 IVF/ICSI cycles: 2324 regular IVF and 660 ICSI cycles. In addition, 512 cryopreserved embryo transfers were initiated; in 55 of these no embryos were transferred.

The mean age of the women at the start of the first IVF cycle was 32.8 years (SD = 4.1, range 20–44, median 33). The most common subfertility diagnosis was male factor alone (33.7%). In 20.9% of the couples tubal pathology alone was diagnosed, in 4.5% tubal pathology and male factor; 24.8% of the couples had other diagnoses and 16.0% experienced idiopathic subfertility. Seventy-two percent of the women had primary subfertility problems. The mean duration of subfertility was 3.6 years (SD = 2.6, range 0–20.5, median 3).

More of the women aged <=30 years were primary subfertile than the women of 31–35 years and the women of 36 years and older (80.3, 69.7 and 68.0% respectively; {chi}2 = 17.45, df = 2, P = 0.000). The age of the woman and the subfertility diagnosis were only related for male factor and idiopathic subfertility. Male factor was present in 38.6% of the couples in which the woman was <=30 years, in 30.5% in which the woman was 31–35 years of age and in 33.9% of the couples in which the women was >=36 years ({chi}2 = 6.67, df = 2, P = 0.03). Idiopathic subfertility occurred in 11.5% of the couples in which the woman was <=30 years, in 18.6% of those aged 31–35 years and in 16.7% of those aged >=36 years ({chi}2 = 8.90, df = 2, P = 0.01). The other subfertility diagnoses were not related to the age of the woman (tubal pathology: {chi}2 = 3.01, df = 2, P = 0.22; tubal pathology and male factor: {chi}2 = 1.21, df = 2, P = 0.55; other diagnoses: {chi}2 = 2.77, df = 2, P = 0.25). Fewer of the women with primary subfertility had tubal pathology or idiopathic subfertility than the women with secondary subfertility (tubal pathology: 17.3 versus 30.4%; {chi}2 = 27.13, df = 1, P = 0.000; idiopathic subfertility: 14.0 versus 21.3%, {chi}2 = 10.21, df = 1, P = 0.001); the subfertility diagnosis `male factor' was more common in the primary cases than in the secondary cases (38.0 versus 22.7%; {chi}2 = 27.51, df = 1, P = 0.000). This was also valid for `other diagnoses' (26.5 versus 20.4%; {chi}2 = 5.11, df = 1, P = 0.02); there was no difference in the prevalence of the combination of tubal pathology and male factor ({chi}2 = 0.65, df = 1, P = 0.42).

In Figure 1Go and Tables I–IVGoGoGoGo, optimistic, realistic and pessimistic cumulative probabilities of achieving an ongoing pregnancy are presented for the total population and for the subgroups: woman's age, subfertility diagnosis and primary or secondary subfertility. In the total population as well as in the subgroups, the optimistic cumulative probability of achieving an ongoing pregnancy within five IVF/ICSI cycles was about 20% higher (absolute value) than the pessimistic cumulative probability. The realistic cumulative probability was about 10% lower (absolute value) than the optimistic one, except in the subgroup `tubal pathology and male factor', because none of the subjects in that subgroup received a medical indication to quit treatment. When log rank tests and Wilcoxon tests were used to study differences between the cumulative probabilities of achieving an ongoing pregnancy during consecutive IVF/ICSI cycles, the application of the three assumptions showed the same trends (Tables II–IVGoGoGo). The three age groups differed in the probability of achieving an ongoing pregnancy (Table IIGo). Further analysis revealed that the youngest two age groups did not differ in the probability of achieving an ongoing pregnancy (P values for the log rank test varied from 0.29 to 0.45, and for the Wilcoxon test from 0.58 to 0.70, using the optimistic, realistic or pessimistic assumption). The two youngest groups differed from the oldest age group (P values from the log rank test and the Wilcoxon test for differences between youngest versus oldest, or middle group versus oldest, were all 0.0001, using the optimistic, realistic or pessimistic assumption). There were no differences in `survival' between the five subfertility diagnosis subgroups (Table IIIGo). Women with secondary subfertility problems had a higher cumulative probability after the two first IVF cycles than those with primary subfertility. However, this advantage disappeared gradually after further treatment with IVF/ICSI (Table IVGo).



View larger version (13K):
[in this window]
[in a new window]
 
Figure 1. Cumulative probability of achieving an ongoing pregnancy after IVF or ICSI cycles with 95% confidence interval for the total population. – – – Optimistic, ——— realistic and • • • pessimistic cumulative probability.

 

View this table:
[in this window]
[in a new window]
 
Table I. The cumulative probability (CP) of achieving an ongoing pregnancy after IVF/ICSI cycles with 95% confidence interval (95% CI) for the total population
 

View this table:
[in this window]
[in a new window]
 
Table II. The cumulative probability (CP) of achieving an ongoing pregnancy after IVF/ICSI cycles with 95% confidence interval (95% CI) according to a woman's age*
 

View this table:
[in this window]
[in a new window]
 
Table III. The cumulative probability (CP) of achieving an ongoing pregnancy after IVF/ICSI cycles with 95% confidence interval (95% CI) according to subfertility diagnosis*
 

View this table:
[in this window]
[in a new window]
 
Table IV. The cumulative (CP) probability of achieving an ongoing pregnancy after IVF/ICSI cycles with 95% confidence interval (95% CI) according to primary and secondary subfertility of the woman*
 
The effects of a woman's age, subfertility diagnosis and primary or secondary subfertility on the probability of achieving an ongoing pregnancy after adjustment for confounding was studied by means of survival analysis. In Table VGo, the results are presented. The unadjusted hazard ratios (HR) confirm the results of analysis described above using the optimistic assumption. By forcing the three variables `woman's age', `subfertility diagnosis' and `primary or secondary subfertility' into the model, the effects of each variable were adjusted for the effects of the other two variables. The advantage of a relatively young age at the first IVF cycle remained valid after adjustment for subfertility diagnosis and primary or secondary subfertility: the women aged <=35 years had a 1.8 times higher probability of achieving an ongoing pregnancy than the older women. The effect of the subfertility diagnosis on the probability of achieving an ongoing pregnancy was negligible. Using Cox extended models, women with secondary subfertility problems had a 1.4 times higher probability of achieving an ongoing pregnancy during the first two IVF/ICSI cycles than women with primary subfertility; after further treatment the differences disappeared. The hazard ratios changed a little after adjustment for the effects of the other two patient characteristics, the year of first treatment, the duration of subfertility at the first treatment and the time dependent variable `medical indication to quit treatment', but the trends in the probabilities of achieving an ongoing pregnancy between the subgroups woman's age, subfertility diagnosis and primary or secondary subfertility of the woman, remained the same.


View this table:
[in this window]
[in a new window]
 
Table V. Hazard ratios (HR) and 95% confidence intervals (CI) for the probability of achieving an ongoing pregnancy after successive IVF/ICSI cycles between woman's age, subfertility diagnosis and primary or secondary subfertility, unadjusted and adjusted for potential confounding effects
 

    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Data were used from couples who underwent their first IVF/ICSI treatment during the period March 1991 to December 1997. This data consisted of all the treatments that were performed during the period March 1991 to December 1998. The extra follow-up period of 1 year allowed most of the couples to finish their desired number of IVF/ICSI treatments. Our pessimistic cumulative probabilities might be a little too pessimistic, because not all the couples finished their desired number of cycles. During this period, IVF treatment hardly changed at the University Hospital Nijmegen, except for the introduction of ICSI in September 1994. Many couples who had discontinued IVF treatment because of poor fertilization after regular IVF in the preceding period, returned to the hospital to continue treatment when ICSI became available. Restricting the data to those from couples who started treatment after August 1994 did not alter the results fundamentally (results not presented). In The Netherlands, up to three IVF/ICSI cycles are paid for by health insurance companies. Therefore a financial reason to quit treatment might only play a role after the third cycle.

We adjusted the cumulative probability of achieving an ongoing pregnancy by taking the reason for quitting into account: if couples quit treatment because of a medical indication, i.e. because of poor ovarian response or total fertilization failure after ICSI, we assumed that the probability of success was equal to zero. However, some couples continued treatment despite having such a medical indication. Although they were less successful than those without a medical indication, their probability of achieving an ongoing pregnancy was greater than zero. For instance, an ongoing pregnancy was achieved in seven out of the 56 couples (12.5%) who continued treatment despite a medical indication to quit, versus 551 out of the 1194 couples (46.1%) who did not have a medical indication to quit. Our realistic cumulative probability will therefore have slightly underestimated the real cumulative probability.

At another Dutch IVF centre, no differences were found in the woman's age, fertilization rate, two or fewer oocytes and less than two embryos transferred between those who quit and those who continued treatment (Roest et al., 1998Go). It was therefore assumed that the cumulative probability of achieving pregnancy was not overestimated with life-table analysis. In contrast, we found differences in the woman's age, two or fewer oocytes and in less than two embryos transferred between those who quit and those who continued treatment. This indicates a less favourable situation in those who had quit treatment (results not presented). More importantly, in the population described (Roest et al., 1998Go), 30% of the couples had quit treatment after the first cycle and 41% after the second cycle. In our population, these rates were only 15 and 26% respectively. These differences suggest that in a large proportion of the couples in the study described above (Roest et al., 1998Go), the reason for quitting treatment without achieving pregnancy was probably not poor results. Therefore the assumption that those who had quit treatment had the same probability of success as those who continued might have slightly overestimated the real cumulative probability in that population. In general, our realistic assumption method will estimate the real cumulative probability better than the optimistic one with life-table analysis.

In this study, we first estimated the cumulative probabilities that separate subgroups would achieve pregnancy by using a so-called `stratified analysis'. This method has some drawbacks. Because of the small number of couples per subgroup, estimates can become imprecise, which is reflected in the width of the 95% confidence intervals. Moreover, the effects of other characteristics are not taken into account. For instance, in this population primary subfertility was more common in the youngest age group than in the oldest age group. As the cumulative probability of achieving an ongoing pregnancy in the youngest age groups was higher than in the oldest age group, the cumulative probability that women with primary subfertility or secondary subfertility problems presented in Table IVGo would achieve pregnancy may have been distorted by the unequal age distribution between the groups. By using survival analysis, it is possible to take the confounding effects of these and other variables into account and to test whether there are differences between groups; these differences can then be expressed in terms of a hazard ratio. Cumulative probabilities cannot be calculated, however.

In our study, the interpretation of the differences in cumulative probability of achieving an ongoing pregnancy on the basis of a woman's age, subfertility diagnosis and primary or secondary subfertility did not alter fundamentally after the confounding effects of other variables and a medical indication for quitting treatment had been taken into account. In agreement with other studies, e.g. those mentioned above in the Introduction, we found that women <36 years had a higher probability of achieving an ongoing pregnancy than the older women. Although several studies indicated that the cumulative probability of success after regular IVF was lower in couples with a male factor than in those with other subfertility diagnoses (e.g. Tan et al., 1992; Alsalili et al., 1995; Prietl et al., 1998), we did not find any differences in cumulative probabilities between the subfertility diagnosis subgroups after IVF cycles in which ICSI treatment had been an option. Although we found that women with secondary subfertility had an initial advantage in the probability of becoming pregnant after IVF/ICSI treatment in comparison with women with primary subfertility problems, we could not confirm that there were any differences in the cumulative probability of achieving pregnancy after more than two cycles between women with primary subfertility and those with secondary subfertility as found previously (Prietl et al., 1998Go).

The cumulative probabilities of achieving an ongoing pregnancy presented here and those presented in other studies cannot be compared easily for several reasons. First, differences in success rates between IVF centres cannot be fully explained by population characteristics (Haan et al., 1991Go). Second, there are differences in the definitions of success between studies, e.g. we defined it as ongoing pregnancy, whereas others studied clinical pregnancies or live births. Third, the denominators can differ between studies: we included all started cycles, whereas others include only cycles in which follicle puncture was performed, or only cycles in which at least one embryo was transferred. Moreover, treatments can differ: e.g. in our study ICSI was an option during IVF treatment, whereas in other studies it is not, or there may be differences in the maximum number of embryos transferred. Our cumulative probabilities of achieving an ongoing pregnancy can be used for counselling purposes in couples who want to be treated with IVF/ICSI in Nijmegen, although they will not be valid for every IVF centre. However, the presence or absence of differences in trends between subgroups does form clear indicators of relevant qualitative differences between subgroups for other centres as well.


    Acknowledgments
 
We are grateful to H.Straatman for his statistical advice and to G.A.Zielhuis for the constructive criticism and suggestions. This study was conducted on data made available by the Department of Obstetrics and Gynaecology of the University Hospital Nijmegen, The Netherlands.


    Notes
 
3 To whom correspondence should be addressed at: Department of Medical Affairs (code 111), University Hospital of Nijmegen,PO Box 9101, 6500 HB Nijmegen, The Netherlands Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Alsalili, M., Yuzpe, A., Tummon, I. et al. (1995) Cumulative pregnancy rates and pregnancy outcome after in-vitro fertilization: >5000 cycles at one centre. Hum. Reprod., 10, 470–474.[Abstract]

Check, J.H., Baker, A., Lurie, D. et al. (1994) Comparison of the cumulative probability of pregnancy after in vitro fertilization–embryo transfer by infertility factor and age. Fertil. Steril., 61, 257–261.[ISI][Medline]

De Mouzon, J., Rossin-Amar, B., Bachelot, A. et al. (1998) Influence du rang de la tentative en FIV. (Role of attempt rank in in vitro fertilization). Contracept. Fertil. Sex., 26, 466–472.[ISI][Medline]

Dor, J., Seidman, D.S., Ben-Shlomo, I. et al. (1996) Cumulative pregnancy rate following in-vitro fertilization: the significance of age and infertility aetiology. Hum. Reprod., 11, 425–428.[Abstract]

Haan, G., Bernardus, R.E., Hollanders, J.M.G. et al. (1991) Results of IVF from a prospective multicentre study. Hum. Reprod., 6, 805–810.[Abstract]

Kalbfleisch, J.D. and Prentice, R.L. (1980) The statistical analysis of failure time data. John Wiley & Sons, Inc., New York.

Kodama, H., Fukuda, J., Karube, H. et al. (1996) Benefit of in vitro fertilization treatment for endometriosis-associated infertility. Fertil. Steril., 66, 974–979.[ISI][Medline]

Land, J.A., Courtar, D.A. and Evers, J.L.H. (1997) Patient dropout in an assisted reproductive technology program: implications for pregnancy rates. Fertil. Steril., 68, 278–281.[ISI][Medline]

Prietl, G., Engelberts, U., Maslanka, M. et al. (1998) Kumulative Schwangerschaftsraten der konventionellen In-vitro-Fertilisation in Abhängigkeit der Diagnose und des Alters der Patientinnen. Ergebnisse des Bonner IVF-Programms. (Cumulative rates of conception following conventional in vitro fertilization as a function of patient age and diagnosis: Results of the Bonn IVF programme). Geburtshilfe. Frauenheilkd., 58, 433–439.[ISI]

Roest, J., Van Heusden, A.M., Zeilmaker, G.H. and Verhoeff, A. (1998) Cumulative pregnancy rates and selective drop-out of patients in in-vitro fertilization treatment. Hum. Reprod., 13, 339–341.[Medline]

Stolwijk, A.M., Hamilton, C.J.C.M., Hollanders, J.M.G. et al. (1996) A more realistic approach to the cumulative pregnancy rate after in-vitro fertilization. Hum. Reprod., 11, 660–663.[Abstract]

Tan, S.L., Royston, P., Campbell, S. et al. (1992) Cumulative conception and livebirth rates after in-vitro fertilisation. Lancet, 339, 1390–1394.[ISI][Medline]

Tan, S.L., Doyle, P., Maconochie, N. et al. (1994) Pregnancy and birth rates of live infants after in vitro fertilization in women with and without previous in vitro fertilization pregnancies: a study of eight thousand cycles at one center. Am. J. Obstet. Gynecol., 170, 34–40.[ISI][Medline]

Submitted on April 19, 1999; accepted on September 16, 1999.