Cumulative live birth rates in low-prognosis women.

In the OPTIMIST study, 985 women received a fixed FSH dose of 150 IU/day in the first cycle. A total of 551 (55.9%) women met the POSEIDON criteria and were categorized in the pre-defined groups ( Fig. 1 ). These women underwent 1128 fresh and 329 FET cycles during the 18 months of follow-up. Additionally, 164 younger and 78 older normal responders were included for supplemental comparison. Table II shows the baseline and treatment characteristics. By definition, POSEIDON groups 2 and 4 had a higher age than group 1 and 3, and AMH levels were higher in POSEIDON groups 1 and 2 compared to groups 3 and 4. Younger unexpected poor responders (subgroup 1a) had a higher body weight than the other (sub)groups. Primary infertility occurred more often in the younger POSEIDON groups (1 and 3), and they were most often treated for male factor infertility, whereas unexplained infertility occurred more frequently in the older groups (2 and 4). The majority of low-prognosis women (75%) were treated with a GnRH agonist, and ICSI was most often performed in the younger POSEIDON groups (1 and 3). Unexpected suboptimal responders (subgroups 1b and 2b) had the lowest number of fresh and highest number of FET cycles. Unexpected poor responders (subgroup 1a and 2a) had the highest cancellation rates, and the FSH dose was increased (~60 IU/day) between cycle 1 and 2 in the majority of the expected or unexpected poor responders (subgroups 1a and 2a, groups 3 and 4). It should be noted that these features are likely to be related to the characteristics that determined the assignment to the POSEIDON (sub)groups. In low-prognosis women, the average CLBR over 18 months of IVF/ICSI treatment was between 54% (conservative) and 57% (optimistic) ( Table III ). Figure 2 shows the cumulative incidence curves for each of the POSEIDON groups, and Table IV presents the results of the pairwise log-rank tests. The younger groups (1 and 3) had the highest CLBR over 18 months of treatment ( Table III ), and these groups also had the highest LBR per stimulation, oocyte retrieval, and embryo transfer. Within group 1, small differences were observed between the unexpected poor (subgroup 1a) and suboptimal responders (subgroup 1b). The older groups (2 and 4) had lower CLBR over 18 months of treatment. Within group 2, unexpected suboptimal responders (subgroup 2b) seemed to have higher CLBR than unexpected poor responders (subgroup 2a), although this difference was not statistically significant ( Table IV ). Older women with an impaired ovarian reserve (group 4) had the lowest CLBR, but still reached a rate between 37% (conservative) and 41% (optimistic) over 18 months of treatment ( Table III ). During the 18 months of follow-up, there were no large differences in the number of fresh treatment cycles between the POSEIDON groups with an average of 2 cycles per woman ( Table II ), yet women with the lowest prognosis (subgroup 2a and group 4) had a slightly higher number of fresh cycles (2.5 and 2.3 fresh cycles, respectively ( Table II )). Cumulative ongoing pregnancy results within 18 months of IVF/ICSI treatment, resulting in a live birth for low-prognosis women stratified according to the POSEIDON criteria ( Poseidon group et al ., 2016 ). Data are presented as number (%), unless stated otherwise. CLBR, cumulative live birth rate; LBR, live birth rate; FET, frozen-thawed embryo transfer. a Includes the results of subsequent fresh and frozen-thawed embryo transfers. Cumulative live birth curves for low-prognosis women over 18 months of IVF/ICSI treatment. Women were stratified according to the POSEIDON criteria ( Poseidon group et al ., 2016 ), and the curves were calculated by using ( a ) the life table analysis (optimistic approach) and ( b ) the competing risk method (conservative approach). Pairwise log-rank comparisons of the ‘optimistic’ and ‘conservative’ cumulative incidence curves. * A p-value of <0.05 is considered to indicate a statistically significant difference in CLBR over 18 months of IVF/ICSI treatment Supplementary Figure S1 shows the cumulative incidence curves of the younger (<35 years) and older (≥35 years) normal responders. The CLBR for the younger normal responders was ~72%, and for the older normal responders ~58% over 18 months of treatment ( Supplementary Table S1 ).

Data of a recent Dutch multicenter prospective cohort study (OPTIMIST study), which included 1515 women between 2011 and 2014, were used for the analyses (NTR2657). Participants were aged <44 years, had regular menstrual cycles, and no significant abnormalities on transvaginal ultrasound. Women with polycystic ovarian syndrome, metabolic or endocrine abnormalities, or undergoing oocyte donation were excluded. All participants had their first IVF/ICSI cycle, or the first after a previous live birth. A more detailed study description was reported previously ( van Tilborg et al ., 2017a ). For the current study, we included low-prognosis women, who used a fixed FSH dose of 150 IU/day in the first cycle. Small dose adjustments between cycles were permitted, based on the response in the preceding cycle ( van Tilborg et al ., 2012 ). We categorized all women in the POSEIDON groups by using age, AMH, and the ovarian response in the first cycle ( Poseidon group et al ., 2016 ). We used AMH, as recent studies indicate that it may be a more accurate and robust biomarker than the AFC ( Fleming et al ., 2015 ; Iliodromiti and Nelson, 2015 ; Nelson et al ., 2015a ). Women with an adequate ovarian reserve and a normal response to stimulation (defined as 10–15 retrieved oocytes), whom are generally considered to have an optimal prognosis ( Sunkara et al ., 2011 ; Polyzos et al ., 2018 ), were added to compare the CLBR to low-prognosis women. In the OPTIMIST study, blood sampling was performed prior to the start of stimulation in the early follicular phase, and AMH levels were determined in one batch by using the fully automatic Elecsys assay (Roche Diagnostics, Germany). As automated assays produce substantially lower values than the pre-existing enzyme linked immunosorbent assays (ELISA) ( Gassner and Jung, 2014 ; Nelson et al ., 2015b ), and as the POSEIDON cut-off value of 1.2 ng/ml is based on studies evaluating the pre-existing assays ( Humaidan et al ., 2016 ), we adjusted the cut-off value to 0.96 ng/ml using the formula Elecsys = 0.087 + (0.729 * Gen II ELISA) ( Nelson et al. , 2015b ). This formula corresponds with our internal laboratory comparison of the Gen II ELISA with the Elecsys assay, which was carried out when the latter was implemented in our hospital at the beginning of 2018 (unpublished data). The proportion of missing AMH values was 11.6%. As the missing values were related to logistic issues, they were considered to be missing completely at random and multiple imputation was performed ( Sterne et al ., 2009 ; Janssen et al ., 2010 ) In this process, hundred imputed datasets were created using a multivariate imputation by chained equations algorithm ( van Buuren and Groothuis-Oudshoorn, 2011 ). In each of the imputed datasets, women were classified in one of the POSEIDON groups, and results were pooled by assigning the women into the group that occurred in more than half (i.e. at least 51 out of 100) of the imputed datasets. The primary outcome was the CLBR of the POSEIDON groups over multiple complete IVF/ICSI cycles, including all subsequent fresh and FET cycles, within 18 months of treatment. Additionally, we calculated the live birth rate (LBR) per consecutive cycle, per started stimulation, per oocyte retrieval, and per embryo transfer. All live births, irrespective of the mode of conception, were taken into account. Time to ongoing pregnancy leading to live birth was depicted by cumulative incidence curves, for which we used two approaches. First, a life table analysis (optimistic) assumed that the chances for couples who discontinue treatment would have been equal to couples who continue. Second, a competing risk approach (conservative) assumed that couples who discontinue treatment would have had zero chances of conceiving. The realistic curve is considered to lie between these two curves ( Stolwijk et al ., 1996 ). To measure whether significant differences exist between the POSEIDON groups, a (pairwise) log-rank test was performed. P -values were adjusted using the Hommel correction for multiple testing ( Hommel, 1988 ). A P -value of <0.05 was considered to indicate a statistically significant difference. Statistical analyses were performed using R for Windows (version 3.3.2; R Foundation for Statistical Computing, Vienna, Austria). Ethical approval was obtained by the Institutional Review Board of the University Medical Centre (MEC 10-273), and by the board of directors of the participating centres. All participants provided written informed consent.

In conclusion, the CLBR of low-prognosis women is on average ~56% over 18 months of IVF/ICSI treatment, and varies considerably between the POSEIDON groups. The variation is primarily determined by female age, which reflects the importance of oocyte quality. In the younger groups, relatively high CLBRs are reached over 18 months of treatment, despite reduced quantitative parameters. In the older groups, the CLBRs are substantially lower, and as no effective interventions exist to counteract the reduced oocyte quality, expectations should be managed before initiating treatment.

This multicenter observational cohort study evaluated the CLBR of low-prognosis women according to the POSEIDON criteria and reveals that ~56% has a live birth after 18 months of IVF/ICSI treatment. A considerable variation is seen between the POSEIDON groups, which is primarily attributable to a woman’s age. Younger women had the highest CLBR, without a large impact of the first cycle ovarian response on the prognosis over 18 months. The CLBRs of older women were lower, especially for those with an impaired ovarian reserve, but still exceeded ~39%. These findings are in line with several studies that demonstrated female age to be the main predictor of pregnancy in IVF/ICSI treatment ( van Loendersloot et al ., 2010 ; Broer et al ., 2013 ; McLernon et al ., 2016 ). The distinct role of a woman’s age on the reproductive capacity is explained by the age-related decline in oocyte quality, which coincides with a progressive decrease in the primordial follicle number ( Broekmans et al ., 2009 ; Cimadomo et al. , 2018 ). As a consequence, the number of euploid embryos in IVF/ICSI treatment rapidly decreases after the age of 35 ( Franasiak et al ., 2014 ; Demko et al ., 2016 ), which most likely explains the substantially lower CLBR in the older subgroups. The variation in CLBR between the POSEIDON subgroups was secondarily attributable to the quantitative parameters. This is in line with studies that show that, within specific age categories, lower AMH levels and a reduced ovarian response are associated with a decreased probability of a live birth ( Sunkara et al ., 2011 ; Hamdine et al ., 2015 ; Polyzos et al ., 2018 ). Yet, female age had a much more significant impact on the CLBR than the quantitative parameters, which is probably explained by the higher importance of the quality of the oocyte, as opposed to their number, in order to obtain a good quality embryo with a high implantation capacity ( Baart et al ., 2007 ; Arce et al ., 2014 ). Not all low-prognosis women had substantially reduced pregnancy prospects. The 18-month CLBR of the younger unexpected poor and suboptimal responders approached those of normal responders, who are generally considered to have optimal prospects in IVF/ICSI treatment. These findings are comparable to previous studies that evaluated CLBR of unexpected poor responders over multiple cycles ( Klinkert et al ., 2004 ; Hendriks et al ., 2008 ; Oudendijk et al ., 2012 ; Moolenaar et al ., 2013 ). Although the pathophysiologic mechanism of the hypo-responsiveness is not fully understood in these younger women ( Alviggi et al ., 2018 ), it is unlikely to be related to a reduced oocyte quality ( Morin et al ., 2018 ), which probably explains the relatively high CLBR over multiple IVF/ICSI cycles. As the CLBR is calculated over 18 months of treatment, the success rates over consecutive cycles determine the prognosis of each of the subgroups. Variation exists in the success rates of subsequent treatment cycles between the subgroups, which may be partly related to differences in the effect of therapeutic adjustments between cycles. Still, the differences in baseline characteristics, including female age and ovarian reserve status, will mainly determine the LBR in the subsequent treatment cycles, as is illustrated by the persisting low LBR in subsequent cycles in older women with an impaired ovarian reserve. This study initiates the essential validation of the POSEIDON criteria and provides valuable information on long-term pregnancy prospects of the proposed groups. In recent years, embryo cryopreservation has become an integral part of IVF/ICSI treatment, and many couples have more than one fresh treatment cycle ( Wong et al ., 2014 ; McLernon et al ., 2016 ). Therefore, evaluating CLBR over multiple complete cycles, instead of studying single fresh cycle results, provides a more comprehensive overview of the chance of success over an entire treatment period. CLBRs are often overestimated due to the use of optimistic analytic approaches ( Stolwijk et al ., 1996 ). In this study, both an optimistic and a conservative approach were applied. This assured the robustness of the findings and carefully addressed the issue of treatment discontinuation, which is of particular importance in low-prognosis women. The prospective design of the OPTIMIST study ensured reliable data collection with relatively low rates of missing values. Multiple imputation was applied to handle missing data, which is considered to be the preferred strategy for ‘missings (completely) at random’ to prevent biased estimates, to increase precision, and to avoid the waste of resources ( Sterne et al ., 2009 ; Janssen et al ., 2010 ). The primary limitation of this study was the relatively small numbers in some of the subgroups, limiting the power to detect statistically significant differences and decreasing the precision of the estimates. Although this hindered the drawing of firm conclusions, our findings still provide the first meaningful indication of the proportion, characteristics, and prognosis of women in the POSEIDON groups. Second, the majority of blood samples were obtained during downregulation with a GnRH agonist, which may have slightly affected serum AMH levels ( Wang et al ., 2007 ; Jayaprakasan et al ., 2008 ; Su et al ., 2013 ). However, as such a change most likely reflects a change in the follicle number and follicle size distribution, the accuracy to predict the ovarian response is unlikely to be compromised, as was confirmed by a previous study ( Wang et al ., 2007 ; Cai et al ., 2018 ). In the POSEIDON classification, AMH is used as an ovarian response predictor to categorize women into expected and unexpected poor responders. As AMH maintains its predictive accuracy when measured during downregulation, the AMH values in the current study allowed for a valid and accurate classification of the low-prognosis women, and no large impact on the CLBR of the POSEIDON groups is expected. Furthermore, all women started with a fixed FSH dose of 150 IU/day, which may be considered as a low dose for women with an expected poor response. However, as previous studies revealed no beneficial impact of increased FSH doses on CLBR, it is unlikely that a higher starting dose would have altered our findings ( van Tilborg et al ., 2017b ; Lensen et al ., 2018 ). Also, small dose adjustments between cycles were permitted, which could have induced therapeutic differences between the subpopulations. Yet, as such dose adjustments closely reflect current practice, this allows for a greater generalizability of our findings. Finally, the inclusion of multiple centers in the OPTIMIST study resulted in some between-center variation in treatment protocols among the included women, which may have influenced the success rates of treatment. Yet, as such variation mirrors the actual differences between infertility clinics, this also increases the representability of the results. The recently introduced POSEIDON criteria identify low-prognosis women in IVF/ICSI treatment and combine quantitative and qualitative parameters to provide a more detailed stratification into homogenous groups ( Poseidon group et al ., 2016 ). This validation study shows the variation in CLBR between the proposed groups and reveals a primary role of female age, reflecting the importance of oocyte quality in the probability of a live birth. For younger low-prognosis women, who generally have high-quality oocytes, the findings suggest that the quantitative parameters are of limited importance for their pregnancy prospects over multiple treatment cycles. Therefore, the question rises whether these women should be considered to have a low prognosis in clinical practice, especially as the present results suggest that current clinical management achieves relatively high CLBR over 18 months of treatment. For older low-prognosis women, a higher oocyte yield may be needed to compensate for the decreased oocyte quality. However, the age-related decline is generally accompanied by a decreased size of the primordial follicle pool, which hinders the retrieval of a high number of oocytes ( Broekmans et al ., 2007 ). Therapeutic interventions that aim to improve the ovarian response, such as the use of increased doses of gonadotropins or co-treatment with growth hormone, dehydroepiandrosterone, or testosterone, have all failed to improve clinical outcomes in these women ( Pandian et al ., 2010 ; Nagels et al ., 2015 ; Lensen et al ., 2018 ). Also, no treatment options are available that target oocyte quality. Therefore, the medical management of the older low-prognosis women remains particularly difficult and forms a challenge in IVF/ICSI treatment. Until new therapeutic interventions become available for this group, increasing awareness about the age-related decline in reproductive chances is needed to manage expectations and to inform younger women about fertility preservation options such as oocyte cryopreservation.

In IVF/ICSI treatment, one of the main challenges is the management of women with an impaired ovarian reserve or a reduced response to exogenous gonadotropins. These ‘poor responders’ generally have lower live birth rates and higher treatment discontinuation rates ( Olivius et al ., 2004 ; Busnelli et al ., 2015 ; Polyzos et al ., 2018 ). The definition of the poor responder has been standardized in the Bologna criteria ( Ferraretti et al ., 2011 ). However, questions have been raised about the capacity of these criteria to select homogenous populations for clinical trials ( Ferraretti and Gianaroli, 2014 ; Papathanasiou, 2014 ). Considerable variation is seen in baseline characteristics and prognosis due to the several ways the Bologna criteria can be fulfilled. This heterogeneity is associated with differences in the underlying etiology, and may cause variation in the effectiveness of interventions ( Papathanasiou, 2014 ). Therefore, analysis of the poor responder population as a whole as defined by the Bologna criteria might dilute potential treatment effects and could prevent the progress in clinical management for specific subpopulations. In 2016, the POSEIDON group proposed a more subtle stratification of ‘low-prognosis women’ ( Poseidon group et al ., 2016 ). In this concept, women are categorized into four groups based on female age, ovarian reserve tests (anti-Müllerian hormone (AMH) or antral follicle count (AFC)), and the ovarian response in case of a previous stimulation ( Table I ). The proposed classification identifies women with an adequate ovarian reserve and a poor or suboptimal response to standard stimulation (unexpected poor or suboptimal responders) and women with an impaired ovarian reserve (expected poor responders). It attempts to differentiate between relevant subpopulations of women, in whom specific interventions might be beneficial. The POSEIDON criteria could thereby improve the homogeneity and comparability of clinical trials, decrease the dilution of potential treatment effects, and guide a more patient-tailored approach for low-prognosis women ( Humaidan et al ., 2016 ; Poseidon group et al ., 2016 ). Although a recent trial already used the POSEIDON criteria to select their study population ( Xu et al ., 2018 ), the actual prognosis of the low-prognosis women has not yet been properly investigated. Such information could help to validate the new POSEIDON concept and provides an initial insight in the necessity of new interventions for each group. Therefore, the current study aims to evaluate the cumulative live birth rate (CLBR) of the POSEIDON groups over multiple complete IVF/ICSI cycles, including all subsequent fresh and frozen-thawed embryo transfers (FET), within 18 months of treatment. The proposed POSEIDON groups of women with a low prognosis in IVF/ICSI treatment based on quantitative and qualitative parameters. AFC, antral follicle count; AMH, anti-Müllerian hormone; adapted from Poseidon group et al . (2016) . POSEIDON group 1 and 2 are each divided in two subgroups (a and b), based on the first cycle ovarian response to standard FSH stimulation. Flowchart of the study population of low-prognosis women according to the POSEIDON criteria ( Poseidon group et al ., 2016 ) . * These twelve women were not assigned to the same group in more than half of the hundred imputed datasets. a Hyper response, >15 retrieved oocytes or cycle cancellation for too many follicles according to the POSEIDON criteria. b Normal response, 10–15 retrieved oocytes according to the POSEIDON criteria. c Poor response, <4 retrieved oocytes or cycle cancellation for insufficient follicular growth according to the POSEIDON criteria. d Suboptimal response, 4–9 retrieved oocytes according to the POSEIDON criteria. AMH, anti-Müllerian hormone. Baseline and treatment characteristics of low-prognosis women stratified according to the POSEIDON criteria ( Poseidon group et al ., 2016 ). Data are presented as mean (SD) or number (%) unless otherwise specified. AFC, antral follicle count (2–10 mm); AMH, anti-Müllerian hormone (ng/ml); IQR, interquartile range; FET, frozen-thawed embryo transfer.

Click here for additional data file. Click here for additional data file.