The Delphi process

During Round 1, two of the initial 16 statements proposed by the Scientific Coordinators were approved without changes and 14 statements were approved after changes. One new statement (Statement 15) was included following discussion (Fig. 2).

Fig. 2

Results of voting on statements in Round 2

Results of the online survey distributed to the 53 members of the extended panel who voted anonymously on their level of agreement with the statements using a six-point Likert-type scale (1 = Absolutely agree; 2 = More than agree; 3 = Agree; 4 = Disagree; 5 = More than disagree; 6 = Absolutely disagree). Consensus was achieved if the proportion of participants either agreeing with a statement (responding “absolutely agree”, “more than agree” or “agree”) or disagreeing with a statement (responding “absolutely disagree” or “more than disagree” or “disagree”) exceeded 66% during Round 2 (red dotted line)

During the first round of voting (Round 2), three statements (statements 3, 5 and 7) achieved 100% agreement, eight statements (statements 1, 2, 6, 8, 10, 11, 13 and 14) achieved > 90% agreement, three statements (statements 4, 9 and 15) achieved > 80% agreement, and two statements (statements 12 and 17) achieved > 70% agreement (Fig. 2). For statements 1, 9, 11, 12, 13, 14 and 17, 3% of the panel (n = 1 in each case) ‘more than disagreed’ with the statement; for one statement (the original wording of statement 16), 8% (n = 3) of the panel ‘more than disagreed’ with the statement. One member of the panel (3%) absolutely disagreed with statement 15. Statements with more than 20% disagreement (Statements 12 and 16) and the reasons for disagreement are shown in Supplementary Table 2. Statement 16 achieved only 61% agreement during the first round of voting. The statement was reworded and achieved 84% agreement after a second round of voting. Thirty-two of the 36 panel members voted for the inclusion of the new statement 15, which achieved a consensus agreement of 84%.

The agreed statements were subcategorized according to overall relevance, impact of female age, biomarkers’ relevance and thresholds, oocyte number, prevalence, low prognosis validation, primary outcome in POSEIDON studies, ART Calculator importance, role of the Follicle-to-Oocyte Ondex, management of POSEIDON Group 2, and management of POSEIDON patients.

StatementsStatements relating to overall relevance

Statement 1: The POSEIDON criteria to identify and classify patients with low prognosis in ART is timely and clinically sound because responses to gonadotropin stimulation and ART outcomes are highly variable, depending on individual patient characteristics.

This statement received 92% agreement from the extended panel. Before POSEIDON, the ESHRE Bologna criteria for POR included a heterogeneous population and, therefore, could not support the routine use of any particular intervention for particular subsets of patients with POR, resulting in a lack of recommendations for the clinical handling of patients [4, 8]. The POSEIDON criteria introduced a more nuanced picture of the low prognosis patient in ART (Groups 1 & 2 supported the inclusion of variables that are not normally considered) and can guide the clinician to optimally manage patients with low prognosis through including a pragmatic endpoint for patient management and prediction of success (i.e., of the concept of oocyte quality) [4, 8, 9].

Statements relating to impact of female age

Statement 2: 35 years represents an acceptable female age threshold to distinguish young and older low-prognosis patients.

This statement achieved 97% agreement from the extended panel. Maternal age can impact oocyte and embryo competence through several mechanisms, including reduced ovarian reserve and decreased oocyte/embryo competence due to age-related chromosomal abnormalities (e.g., increased incidence of aneuploidies and possibly decreased mitochondrial activity, shortening of telomeres [14], cohesin dysfunctions, meiotic impairments during oogenesis, and flawed chromosome segregation patterns, such as non-disjunction, premature separation of sister chromatids, or reverse segregation), with 35 years defined as the lowest age threshold to define advanced maternal age (AMA) [15].

In a study reporting the results from chromosomal analysis performed on trophoectoderm biopsies of blastocysts, the lowest risk for embryonic aneuploidy was between ages 26 and 30. Both younger and older age groups had higher rates of aneuploidy and an increased risk for more complex aneuploidies. The overall risk did not measurably change after age 43 years [16].

In a retrospective study using logistic regression analysis of array comparative genomic hybridization (CGH) results of 7753 embryos from 990 patients, Ata et al. determined that increasing female age was associated with a significant decrease in euploidy rate in Day 3 and Day 5 embryos (p < 0.001 for both groups), and the probability of having at least one euploid embryo in an assisted cycle decreased by increasing female age (p < 0.01 for both Day 3 and Day 5 embryos) and was significantly increased by every additional embryo available (p < 0.01 for both Day 3 and Day 5 embryos) [6]. Equivalence of mean proportion of euploid versus aneuploid embryos (50%) was observed at age 35 years.

Furthermore, using a logistic regression model derived from the retrospective analysis of 1296 trophoectoderm biopsies (436 couples undergoing intracytoplasmic sperm injection [ICSI] and preimplantation genetic testing for aneuploidy) by next-generation sequencing (NGS), Esteves et al. reported on the age-related decrease in blastocyst euploid and the number of embryos needed to have at least one euploid blastocyst as a function of age [17]. This model predicted that the decrease in the probability of blastocyst euploidy followed an age-related binomial distribution. The geometric mean of the yearly variation was 13.6% and the decrease was progressive with every year of female age (decreasing by 1.2% at age 28, and by 2.0%, 3.5%, 6.7%, 9.8%, 13.6%, 17.9%, and 24.5% at ages 30, 32, 35, 37, 39, 41 and 44, respectively; p < 0.0001). The number of blastocysts required to obtain one euploid embryo began to increase progressively from the age of 35 years (from three at 28 years, to four at 35 years, and five, six, nine, sixteen and twenty nine for ages 37, 39, 41, 43, and 45, respectively) [10, 17].

Fecundity in women decreases gradually from the age of 32 years, with accelerated decrease after 37 years [18]. In light of the anticipated age-related decline in fertility, the increased incidence of disorders that impair fertility, and the higher risk of pregnancy loss, the American College of Obstetricians and Gynecologists and the American Society for Reproductive Medicine recommend that women older than 35 years should receive expedited evaluation and treatment within 6 months [19], whereas the joint committee of the American College of Obstetricians and Gynecologists and the American Society for Reproductive Medicine recommend women older than 40 years should receive immediate treatment and evaluation [18].

Statements relating to biomarkers’ relevance and thresholds

Statement 3: AMH and AFC thresholds as per POSEIDON criteria are probably sufficient for identifying patients with low ovarian reserve. However, they cannot fully predict the responses to ovarian stimulation due to variation in ovarian sensitivity and clinical protocols.

This statement achieved 100% agreement from the extended panel. AMH and AFC have a fundamental role in the prediction of POR in the POSEIDON criteria [20]. In their systematic review of tests predicting ovarian reserve and IVF outcome using receiver operating characteristic (ROC) analysis, Broekmans et al. concluded that ovarian reserve tests (including AMH) have only modest predictive value for poor response to hyperstimulation, whereas AFC could be considered as a clinically adequate test for poor response prediction at a low threshold level in normal cycling women [21]. However, in a meta-analysis of 13 studies reporting on AMH and 17 studies reporting on AFC, the ROC curves for poor response and non-pregnancy showed no significant difference between the two markers, leading the authors to conclude that AMH has at least the same level of accuracy and clinical value for the prediction of poor response and non-pregnancy as AFC [22]. Nevertheless, they acknowledged that clinical applicability ultimately depends on the way that abnormal test results might alter patient management. A more recent systematic literature review (2013) demonstrated that AFC (41 studies) and AMH (25 studies) were the most sensitive markers of ovarian reserve at the time of the review and were pivotal in planning personalized ovarian stimulation protocols (based on the selective use of gonadotropin hormone-releasing hormone [GnRH] analogs and gonadotropin starting dose adjustment), owing to their reliable accuracy across the extremes of ovarian response and their interchangeability [23].

Notably, the above analyses predate the development of the POSEIDON criteria and, accordingly, AMH and AFC were investigated in isolation in these studies. By contrast, the POSEIDON criteria consider these in the context of the age of the patient, previous response to ovarian stimulation and other risk factors for POR, providing a substantial advancement in the diagnosis and clinical management of these patients, although without relevant improvements in IVF outcomes [20]. A population-based cohort study of 9484 consecutive patients treated at three fertility centres between 2015 and 2017, assessed the agreement between AFC and AMH levels in the context of patient classification using the POSEIDON criteria. The study showed that for low oocyte yield, the optimal AFC and AMH cutoff values were five and 1.27 ng/ml, with sensitivities of 0.61 and 0.66, specificities of 0.81 and 0.72, and AUC receiver operating characteristics of 0.791 and 0.751, respectively, which were similar to the thresholds included in the Poseidon criteria. However, owing to the low predictive value of POSEIDON for low oocyte yield, the authors recommended that clinicians should adopt the biomarker that is most reflective of each clinical setting [24]. Furthermore, in the development of the concept of Follicle-to-Oocyte Index (FOI), Alviggi et al. defined these biomarkers as a static snapshot of an individual’s ovarian reserve that is not reflective of the dynamic nature of follicular growth in response to exogenous ovarian stimulation when used in isolation [25].

Statement 4: AFC and AMH may be used interchangeably for patient classification under the POSEIDON criteria. Combining both biomarkers for that purpose brings little additional information. Clinicians should adopt the biomarker that best suits their clinical setting.

This statement achieved 89% agreement from the extended panel. In a real-world study of 9484 consecutive patients by Esteves et al., the degree of agreement in classifying patients according to POSEIDON groups was strong overall (kappa = 0.802; 95% CI 0.792; 0.811). Furthermore, the study showed that for low oocyte yield, the optimal AFC and AMH cutoff values were five and 1.27 ng/ml, with sensitivities of 0.61 and 0.66, specificities of 0.81 and 0.72, and AUC receiver operating characteristics of 0.791 and 0.751, respectively, which were similar to the thresholds included in the POSEIDON criteria [24].

Statements relating to oocyte number

Statement 5: Number of oocytes retrieved following ovarian stimulation is a prognostic factor of live birth in fresh cycles and cumulative delivery rates per initiated or aspirated IVF/ICSI cycle.

This statement achieved 100% agreement from the extended panel. When utilizing all fresh and frozen embryos in 1099 women undergoing their first ovarian stimulation cycles followed by single embryo transfer (ET) according to ovarian response category (poor: 1–3 oocytes; suboptimal: 4–9 oocytes; normal: 10–15 oocytes; high: >15 oocytes), low responders had a lower live birth rate (LBR) in fresh transfer cycles compared with suboptimal, normal and high responders (p < 0.05) and multivariable logistic regression analysis showed this was an independent predictive factor (p < 0.001) for CLBR [26]. In another retrospective multicenter analysis of individual patient data in 14,469 patients undergoing their first cycle of IVF/ICSI between 2009 and 2014, CLBR steadily increased with the number of oocytes, reaching 70% when ≥ 25 oocytes were retrieved and more modest increases above 27 oocytes; live-birth probability in fresh transfer cycles increased up to seven oocytes and then levelled off between seven and 25 oocytes, and decreased thereafter, which could be attributed to an increase in freeze-all cycle rate in patients with > 20 oocytes retrieved [27]. Notably, maximum CLBR was observed up to 25 oocytes in a large retrospective study of 221,221 autologous cycles between January 2009 and December 2015 but only in women aged between 18 and 35 years, showing that the maximum CLBR observed during ART is dependent on female age. In older women (36–44 years) the maximum CLBR was achieved beyond 30 oocytes, dropping to nine oocytes in women ≥ 45 years [28].

Furthermore, in a systematic review and meta-analysis, which included the study by Drakopoulos and Blockeel, 28 studies (three prospective and 25 retrospective) reporting on 291,752 ART cycles confirmed a positive correlation between oocyte number and the number of top/good quality embryos (p < 0.001 for correlations with Day 2/3 embryos, Day 5/6 embryos metaphase II [MII] oocytes, oocytes with two pronuclei and euploid embryos), suggesting that increased oocyte yield from a single cycle stimulated cycle may maximize outcomes [29].

Statements relating to prevalence

Statement 6: POSEIDON patients represent approximately 40–60% of patients treated in fertility clinics.

This statement achieved 95% agreement from the extended panel. Of 13,146 patients included in a multicentre population-based cohort study in fertility clinics in Brazil, Turkey, and Vietnam, POSEIDON patients represented 43.0% (95% confidence interval [CI] 42.0–43.7) of the studied population, and the prevalence rates varied across study centers (range: 38.6–55.7%) [30]. However, this range may be an over- or under-estimation, and more data are needed, particularly from Europe and the USA, to gain insight into the true number. For example, in a retrospective cohort study of 62,749 women (97,388 cycles) who underwent ART treatment at the Reproductive and Genetic Hospital of CITIC-XIANGYA, China, between January 2014 and June 2017, 19,781 (31.52%) women fulfilled the POSEIDON criteria (26,697 cycles) [31]. Lastly, of a total of 32,128 fresh IVF cycles from January 2014 to October 2018 at a single IVF clinic in Xi’an, China, 6383 (19.9%) were low prognosis, based on the POSEIDON criteria [32].

Statements relating to low prognosis validation

Statement 7: POSEIDON patients exhibit lower [cumulative live birth rate] CLBR per aspirated IVF/ICSI cycle than normal responders, and CLBRs vary across POSEIDON groups. The differences are mainly determined by female age and the number of oocytes retrieved, reflecting the importance of combining oocyte quality and quantity.

This statement received 100% agreement from the extended panel. In a multicenter population-based retrospective cohort study involving 9073 patients treated in three fertility clinics in Brazil, Turkey and Vietnam between 2015 and 2017, the CLBR in POSEIDON patients was between 33.7% vs. 50.6% lower than in normal responders (p < 0.001) and varied across POSEIDON groups (Group 1 [n = 212] 27.8%; Group 1b [n = 1785] 47.8%; Group 2a [n = 293] 14.0%; Group 2b [n = 1275] 30.5%; group 3 [n = 245] 29.4%; Group 4 [n = 623] 12.5%). In POSEIDON Groups 1 and 2, the CLBR was twice as high in suboptimal responders (4–9 oocytes retrieved) than in poor responders (< 4 oocytes retrieved) (p = 0.0004). Predictors of CLBR as ascertained using logistic regression analysis were POSEIDON grouping, number of embryos obtained, number of ET cycles per patient, number of oocytes collected, female age, duration of infertility and body mass index (p < 0.001) [33].

In a retrospective cohort study of 62,749 women (97,388 cycles) undergoing ART treatment at the Reproductive and Genetic Hospital of CITIC-XIANGYA between January 2014 and June 2017, > 30% of women undergoing IVF/ICSI may be classified as low prognosis. Different reproductive outcomes were observed among the four POSEIDON groups; after three successive cycles of treatment, the most optimal outcomes were observed in Groups 1, 2 and 3. The variables associated with live birth in the first cycle were POSEIDON stratification and ovarian stimulation protocol; for the second stimulation cycle, POSEIDON stratification (except Group 3) and ovarian stimulation protocol were associated with live birth [31]. In another retrospective cohort study of 10,615 women who underwent IVF treatment at the Peking University between January and December 2017, the CLBR in the first cycle in each of the POSEIDON groups was lower (p > 0.001) than in non-POSEIDON patients. In the second cycle, CLBR was lower in older patients (Groups 2b [p = 0.001] and 4 [p < 0.001]) and in younger patients with poor ovarian response (Group 3 [p = 0.019]) compared with non-POSEIDON patients; younger patients had higher CLBR than older patients in both cycles (p < 0.001) [34]. Furthermore, data from a multicenter observational cohort study of 551 low-prognosis women age < 44 years initiating IVF/ICSI treatment with fixed-dose 150 IU/day follicle-stimulating hormone (FSH) between 2011 and 2014 reported a mean CLBR of 56% over 18 months of treatment. CLBR varied among POSEIDON groups, primarily determined by age (younger unexpected poor responder ∼ 65%, younger unexpected suboptimal responders ∼ 68%, younger expected poor responders ∼ 59%, older unexpected poor responders 42%, older unexpected suboptimal responders 54%, older expected poor responders ∼ 39%) compared with younger normal responders (∼ 72%) and older normal responders (∼ 58%) [35]. Finally, using statistical modelling, a retrospective cohort study conducted at McGill University Health Center on 459 patients who underwent IVF treatment between 2011 and 2014 showed that age and ovarian response were both predictors of pregnancy and live birth in women with poor response (one or two follicles ≥ 14 mm) and grouped according to AFC (0–5, 6–10 and ≥ 11), in concert with the POSEIDON criteria classifications. The likelihood of live birth as a function of age and AFC showed that a 1-year increase in age reduces the likelihood of live birth by 11%, whereas a one unit increase in AFC lead to a 9% increase in the odds of a live birth (p < 0.05 for both). AFC had a significant effect in the youngest age group: women with AFC > 11 had a 56% LBR compared with 6% LBR in those with AFC ≤ 11 [36].

Statements relating to primary outcome in POSEIDON studies

Statement 8: CLBR per started ovarian stimulation cycle is the best primary outcome to characterize differences in prognosis among the four individual POSEIDON groups or between POSEIDON and non-POSEIDON patients.

This statement received 95% agreement from the extended panel. A conservative estimate of a couple’s chance of a live birth over an entire treatment course (CLBR) was reported as 51% (95% CI 49–52) after six cycles in 614 patients (14,248 cycles); the optimistic estimate was 72% (95% CI 70–74), with higher rates in those aged < 35 years than in those aged ≥ 40 years. CLBR declined with increasing age, and age-stratified curves for women < 35 years versus ≥ 40 years were significantly different (p < 0.001) [37].

In their literature review, Moragianni et al. (2010) emphasized the need to highlight CLBR during counselling for couples with infertility as a more realistic estimate of success, when taking maternal age and genetic factors into consideration [38]. This point was reiterated by Maheshwari et al. (2015) in their later review of the literature. They recommended that CLBR is generally perceived to be the preferred reporting system in IVF and called for an international consensus on how this is calculated, reported and interpreted across the world [39]. CLBR was further defined in the International Glossary on Infertility and Fertility Care (2017) as ‘The number of deliveries with at least one live birth resulting from one initiated or aspirated ART cycle, including all cycles in which fresh and/or frozen embryos are transferred, until one delivery with a live birth occurs or until all embryos are used, whichever occurs first’ [40] and in the ESHRE guideline: ovarian stimulation for IVF/ICSI (2020) [41], which provided guidelines for efficacy in terms of CLBR (or LBR) per cycle. Using these criteria in their multicenter population-based retrospective cohort study of 9073 patients, Esteves et al. reported that the CLBR of POSEIDON patients was on average 50% lower than in normal responders and varied across POSEIDON groups. The differences were primarily determined by female age, number of embryos obtained, number of ET cycles per patient, number of oocytes retrieved, duration of infertility, and BMI [33].

Notably, although CLBR is recognized as a suitable way to report the success of IVF treatment, calculation of this outcome varies between studies. While some studies report CLBR as all live births per the number of women who attempted stimulation, CLBR has also been reported as at least one live birth episode per the number of women who had oocyte collection. New recommendations consider the reporting of CLBR in the short term (first live birth per women in the 2 years after one oocyte retrieval), medium term (all live births per women in the 5 years after one oocyte retrieval) and long term (all live births per women over 10 years after three oocyte retrievals) [39].

Statements relating to ART Calculator importance

Statement 9: The ART Calculator is a helpful predictive model to estimate the number of MII oocytes needed to obtain at least one euploid blastocyst for transfer and can be used for counselling and treatment planning.

This statement received 86% agreement from the extended panel. Maternal age can impact oocyte and embryo competence through several mechanisms, including reduced ovarian reserve and decreased oocyte/embryo competence due to aging-related chromosomal abnormalities (as discussed in Statement 2) [15]. Using a logistic regression predictive model derived from the retrospective analysis of 1296 trophoectoderm biopsies (436 couples undergoing ICSI and preimplantation genetic testing for aneuploidy) by next-generation sequencing the probability of blastocyst euploidy decreased according to an age-related binomial distribution, progressing with each additional year (from 1.2% at age 28 years to 24.5% at age 44 years; p < 0.0001) [17]. The number of blastocysts required to obtain one euploid embryo began to increase progressively from the age of 35 years (from three at 28 years, to four at 35 years); therefore, more oocytes and embryos will be needed with increasing maternal age to counteract this decrease.

In a randomized controlled trial (RCT) of 205 infertile couples with a female partner < 43 years and with AMH ≥ 1.2 ng/mL and Day 3 FSH < 12 IU/L/, Forman et al. reported that during embryo selection using comprehensive chromosome screening, the transfer of a single euploid blastocyst results in ongoing pregnancy rates that are the same as transferring two untested blastocysts while dramatically reducing the risk of twins [42]. However, while PGT-A may reduce the time to live birth by decreasing the risk of pregnancy loss in certain populations, the available evidence is insufficient to support the use of preimplantation genetic testing for aneuploidy in routine clinical practice [43, 44].

To estimate the number of oocytes needed to achieve at least one euploid embryo for transfer and provide a revised estimate when fewer than the predicted number of embryos were obtained after one or more oocyte retrieval cycle, Esteves et al. assessed the factors that influenced embryo ploidy and estimated the predicted probability of blastocyst euploidy as a function of each mature oocyte retrieved [10]. Using a negative binomial distribution to model the number of euploid blastocysts and the adaptive LASSO (Least Absolute Shrinkage and Selection Operator) method for variable selection, the fitted model identified female age, sperm source used for ICSI, and the number of mature (metaphase II) oocytes as predictive factors (p < 0.0001). In the final predictive model, developed using logistic regression analysis, and internally validated by the holdout method, the estimated predicted probabilities of a mature oocyte developing into a euploid blastocyst decreased progressively with female age and was negatively modulated overall by use of testicular sperm across age (p < 0.001) [10].

Using this model, the ART Calculator was developed to make two types of predictions automatically: pretreatment information to estimate the minimum number of oocytes to achieve ≥ 1 euploid blastocysts; and pretreatment information and the actual number of mature oocytes collected or accumulated to provide a revised estimate of the probability of achieving the aforesaid outcome when fewer than the predicted number of mature oocytes are obtained after one or more oocyte retrieval cycles [10]. The ART Calculator was validated in a multicenter study using retrospective clinical and embryonic data from 1464 consecutive infertile couples who had IVF/ICSI with the intention to have preimplantation genetic testing for aneuploidy. This analysis showed that the fitting between the ART Calculator and the validation model were sufficiently close for both the estimated probabilities of a euploid blastocyst per MII oocyte (r = 0.91) and the minimum number of MII oocytes (r = 0.88). The frequency of patients with at least one euploid blastocyst among those who achieved the estimated minimum number of MII oocytes were 84.8% (70% predicted probability of success), 87.5% (80% predicted probability of success) and 90% (90% predicted probability of success) [11].

Recently, the ART Calculator was rebranded as ‘ONE – Oocyte Number Estimator’ and is freely available online (https://art-one.merckgroup.com/art) for professionals and patients) [45].

Statements relating to the Follicle-to-Oocyte Index

Statement 10: The Follicle-to-OocyteIndex (FOI) is helpful to assess the dynamic nature of follicular growth in response to ovarian stimulation (OS) and trigger and may be particularly informative in patients with unexpected suboptimal or poor responses. Low FOI values might indicate a hypo-response to gonadotropin stimulation, implying that only a fraction of available antral follicles was adequately recruited.

This statement received 94% agreement among the extended panel. The FOI, the ratio between the number of oocytes collected at ovum pick up and the number of antral follicles at the beginning of ovarian stimulation, expressed as a range from 0 to 100, was proposed by Alviggi et al. as a novel parameter to assess hypo-response. The value of FOI compared with traditional ovarian reserve markers is that it might optimally reflect the dynamic nature of follicular growth in response to exogenous gonadotropins in light of the many non-mutually exclusive factors that influence ovarian resistance to ovarian stimulation (low gonadotropin starting dose, genetic or environmental factors, asynchronous follicle development, timing and mode of final follicular maturation trigger, and technical issues related to oocyte retrieval) [25].

Importantly, genotype is implicated in the outcome of ovarian stimulation, as reported in a systematic review and meta-analysis of 33 studies, in which ovarian stimulation outcomes related to seven polymorphisms (FSH receptor [FSHR; rs6165, FSHR rs6166, FSHR rs1394205]; luteinizing hormone beta subunit [LHB; rs1800447, LHB rs1056917]; luteinizing hormone chorionic gonadotropin receptor [LHCGR; rs2293275] and LHCGR [rs13405728]) were evaluated. Higher FSH consumption is expected in homozygotes for the A allele of the FSHR (rs1394205) polymorphism than carriers of the G allele. Moreover, FSHR (rs6166) GG homozygotes have fewer oocytes than AA and AG carriers. It is feasible that the effect of these polymorphisms on ovarian stimulation may partially explain the phenomenon of ‘hypo-response’ [46,47,48,49].

To investigate ovarian sensitivity in subgroups of patients with low prognosis, as defined by the POSEIDON criteria, using FOI and the follicle output rate (FORT), Chen et al. performed a retrospective cohort study of 32,128 treatment cycles from a single IVF clinic between January 2014 and October 2018. Using FORT as a marker for ovarian sensitivity, this was highest in POSEIDON Group 3, followed by Group 4, Group 1 and Group 2, and the trend in FOI values was consistent with those for FORT. Adjustment of the ovarian stimulation protocol was recommended for patients with poor ovarian sensitivity, whereas an adjustment to the gonadotropin starting dose was preferred for those with normal ovarian sensitivity [32]. Furthermore, adjustment of time and mode of final follicular maturation trigger may be of benefit for patients with normal FORT and low FOI [50].

Finally, a retrospective analysis of 264 IVF cycles after the first and cumulative ETs investigated the relation between oocyte yield (total retrieved oocytes [Oc] and total mature oocytes ([MII]) relative to the antral follicle count (Oc/AFC and MII/AFC) and AMH (Oc/AMH and MII/AMH). Oc/AMH and MII/AMH ratios had no effect on the occurrence of LBR or on implantation rate after the first or cumulative ET when stratified by age < 36 years and > 39 years, whereas the ratios of Oc/AFC and MII/AFC seemed promising indicators to assess ovarian response [51].

Statements related to management of POSEIDON group 2

Statement 11: POSEIDON Group 2 could benefit from r-hLH supplementation at the dose of 75–150 IU given from stimulation day one.

This statement achieved 92% agreement by the extended panel. A single-center randomized, parallel group, comparative study aimed to identify potential benefits of mid-follicular (from Day 6) r-hLH supplementation in 131 women (68 allocated to recombinant human FSH [r-hFSH] and 63 allocated to r-hFSH plus r-hLH) aged 35–39 years undergoing ovarian stimulation for ICSI. No differences were observed in oocyte or embryo quality or quantity; however, higher implantation rates (11.3% vs. 18.1%; p = 0.049) and live births per started cycle (7.4 vs. 19.0; p = 0.047) were observed with r-hLH supplementation [52]. Similarly, in a randomized open-label controlled trial of r-hFSH versus r-hFSH plus r-hLH from Day 1 in two age groups (≤ 35 years [380 women] and 36–39 years [340 wom