Effect of Genetic Variants of Gonadotropins and Their Receptors on Ovarian Stimulation Outcomes: A Delphi Consensus

Background A Delphi consensus was conducted to evaluate the influence of single nucleotide polymorphisms (SNPs) in genes encoding gonadotropin and gonadotropin receptors on clinical ovarian stimulation outcomes following assisted reproductive technology (ART) treatment. Methods Nine experts plus two Scientific Coordinators discussed and amended statements plus supporting references proposed by the Scientific Coordinators. The statements were distributed via an online survey to 36 experts, who voted on their level of agreement or disagreement with each statement. Consensus was reached if the proportion of participants agreeing or disagreeing with a statement was >66%. Results Eleven statements were developed, of which two statements were merged. Overall, eight statements achieved consensus and two statements did not achieve consensus. The statements reaching consensus are summarized here. (1) SNP in the follicle stimulating hormone receptor (FSHR), rs6166 (c.2039A>G, p.Asn680Ser) (N=5 statements): Ser/Ser carriers have higher basal FSH levels than Asn/Asn carriers. Ser/Ser carriers require higher amounts of gonadotropin during ovarian stimulation than Asn/Asn carriers. Ser/Ser carriers produce fewer oocytes during ovarian stimulation than Asn/Asn or Asn/Ser carriers. There is mixed evidence supporting an association between this variant and ovarian hyperstimulation syndrome. (2) SNP of FSHR, rs6165 (c.919G>A, p.Thr307Ala) (N=1 statement): Few studies suggest Thr/Thr carriers require a shorter duration of gonadotropin stimulation than Thr/Ala or Ala/Ala carriers. (3) SNP of FSHR, rs1394205 (−29G>A) (N=1 statement): Limited data in specific ethnic groups suggest that A/A allele carriers may require higher amounts of gonadotropin during ovarian stimulation and produce fewer oocytes than G/G carriers. (4) SNP of FSH β-chain (FSHB), rs10835638 (−211G>T) (N=1 statement): There is contradictory evidence supporting an association between this variant and basal FSH levels or oocyte number. (5) SNPs of luteinizing hormone β-chain (LHB) and LH/choriogonadotropin receptor (LHCGR) genes (N=1 statement): these may influence ovarian stimulation outcomes and could represent potential future targets for pharmacogenomic research in ART, although data are still very limited. Conclusions This Delphi consensus provides clinical perspectives from a diverse international group of experts. The consensus supports a link between some variants in gonadotropin/gonadotropin receptor genes and ovarian stimulation outcomes; however, further research is needed to clarify these findings.


INTRODUCTION
Infertility is a significant global health problem and socioeconomic burden, affecting 15% of childbearing-age individuals worldwide and with an increasing prevalence over the last two decades (1). Assisted reproduction technology (ART) has provided a critical tool for addressing reproductive challenges in men and women (2)(3)(4)(5). However, ART remains an area with unmet clinical needs. According to the International Committee for Monitoring Assisted Reproductive Technology (ICMART), the global in vitro fertilization (IVF)/intracytoplasmic sperm injection (ICSI) combined delivery rates per fresh aspiration and frozen embryo transfer cycles in 2013 were 24.2% and 22.8%, respectively, with a cumulative delivery rate per aspiration of 30.4% (6).
Delivery rate is closely associated with the number of oocytes retrieved during ovarian stimulation (7). This relationship is even more evident considering how frozen cycles contribute to cumulative live birth rate (8). In this sense, the goal is to safely retrieve the highest number of mature oocytes in order to get the highest percentage of delivery rate per initiated ovarian stimulation cycle for ART treatment.
A critical step of ART is ovarian stimulation using gonadotropins, the aim of which is to obtain an optimum number of mature oocytes without the risk of ovarian hyperstimulation syndrome (OHSS) (9). Responses to gonadotropin stimulation are highly variable and dependent on individual patient factors. The prediction of ovarian response is critical to enable optimal and individualized management of ovarian stimulation. Current ovarian stimulation protocols use several parameters to predict ovarian response and optimize the dose of gonadotropins accordingly, including age, body mass index, ovarian reserve tests such as anti-Müllerian hormone (AMH), antral follicular count (AFC), endocrine status and baseline serum follicle stimulating hormone (FSH) (10)(11)(12). In particular, AMH and AFC are widely considered the best predictors of ovarian potential (13)(14)(15)(16). However, ovarian reserve cannot fully explain the individual response to ovarian stimulation. For instance, a subgroup of women with normal ovarian reserve but suboptimal or poor response (hyporesponders) have been described (17)(18)(19). These patients have an "unexpected" reduced response to ovarian stimulation and are characterized by low prognosis to ART (17,20,21). The mechanisms underlying this unexpected ovarian resistance to ovarian stimulation are not fully understood, but it is believed that an individual's genetics play a significant role (17,22).
Several gene association studies have identified specific single nucleotide polymorphisms [SNPs (Box 1)] of gonadotropins and their receptors that could influence ovarian response (26)(27)(28). These include SNPs of the FSH receptor (FSHR) gene, FSH bchain (FSHB) gene, luteinizing hormone b-chain (LHB) gene and LH/choriogonadotropin receptor (LHCGR) gene ( Table 1). The identification of genetic variants that are able to predict ovarian response could pave the way to tailor ovarian stimulation on the basis of individual genotype profile. Unfortunately, these data are still controversial and, at times, contradictory or limited.
The aim of the current Delphi Consensus study was to generate a series of literature-supported consensus statements regarding the most relevant genetic variants of gonadotropin and gonadotropin receptors involved in ovarian response.
The consensus concept was initiated and funded by Merck KGaA, Darmstadt, Germany. The sponsor was involved early in the process, defining the overarching topic to be discussed, but did not participate in the development of the statements or in any of the meetings or discussions involved in developing the Delphi consensus. The statements were, therefore, developed independently of the industry sponsor. The authors from Merck KGaA, Darmstadt, Germany, were only involved in the development of the manuscript, critically revising it for important intellectual content, especially in the Introduction, Results and Discussion sections, but could not alter the consensus statements in any way.

Consensus Participants
The Delphi consensus involved a Scientific Board, comprising two Scientific Coordinators (AC and FT) and nine additional experts ( Table 2). Scientific Board members were selected based on their recognized expertise in Reproductive Genetics proven BOX 1 | Single Nucleotide Polymorphisms (SNPs).
• SNPs can occur in both non-coding and coding sequences of genes.
• SNPs in non-coding regions appear to affect transcription as well as non-coding RNAs that can also influence gene expression (23).
• SNPs in coding regions can be synonymous or nonsynonymous.
• Synonymous SNPs do not produce altered coding sequences, since some amino acids are coded for by more than one three-base-pair codon. Although these are often referred to as 'silent' polymorphisms, they can still affect the function of a gene (24). • Nonsynonymous SNPs can be missense (a single change of nucleotide base resulting in a change in amino acid) or nonsense (a point mutation in a sequence of DNA that results in premature termination of protein synthesis and in a truncated and usually non-functional protein product (25). • Specific SNPs are catalogued by researchers and databases according to a non-redundant accession number, the reference SNP cluster identifier (rs).  Table 2.

The Consensus Process
The Delphi consensus comprised three rounds ( Figure 1). During Round 1, statements and supporting references initially developed by the two Scientific Coordinators were discussed and amended by the 11 members of the Scientific Board during two web conferences ( Consensus was considered to be achieved if the proportion of participants either agreeing with a statement (responding "agree" or "absolutely agree") or disagreeing with a statement (responding "disagree" or "absolutely disagree") exceeded 66% (48,49). During Round 3, the consensus results were communicated to the participating experts via two web conferences ( Table 2). Statements that did not achieve consensus in Round 2 were discussed, revised and/or merged by the scientific board. The newly reworded statements were shared with the Extended Panel for voting on their level of

Overall Results
A total of 18 statements with supporting references were proposed by the Scientific Coordinators. Following discussion and refinement of these statements by the Scientific Board during Round 1, a total of 11 statements with supporting references were approved by the Scientific Board and included in the online survey in Round 2 ( Table 3). The Extended Panel who participated in the online survey comprised fertility experts from a number of different regions, including Europe, Asia, and South America. No statements achieved 100% agreement. Consensus was achieved for six statements (Statements 1, 2, 3, 5, 10 and 11). A high level of agreement (≥80% of votes were 'agree' or 'absolutely agree') was achieved for three statements (Statements 1, 10 and 11). Five statements failed to reach consensus (Statements 4, 6, 7, 8, and 9) and were discussed and amended during Round 3. Statements 4 and 7, which had a total agreement level of 50% (i.e. only 50% of votes were 'agree' or 'absolutely agree') and 52%, respectively, after the first vote, were reworded but failed to reach consensus after a second round of voting (57% and 62% total agreement, respectively). Statement 6, which had a total agreement level of 55% after the first vote, was reworded and reached consensus after a second vote (67% total agreement). Statements 8 and 9, which both had a total agreement level of 50% after the first vote, were merged and reworded, and achieved consensus after a second vote (76% total agreement). For two statements (Statements 5 and 7 [revote]) one expert voted 'absolutely disagree'. The reasons that experts provided when voting to disagree with a given statement are shown in Supplementary  Figure 2). The reasons provided by participants for disagreeing with this statement included the absence of data from RCTs of an adequate size and the fact that a meta-analysis failed to show significant differences in gonadotropin consumption. Furthermore, one expert suggested that the extent to which FSHR rs6166 affected gonadotropin consumption depended on the study population (Supplementary Table 2).
A number of studies support a role for the FSHR rs6166 (c.2039A>G, p.Asn680Ser) variant as a prognostic indicator of ovarian response to FSH stimulation (33)(34)(35)(36)(37). The Ser/Ser variant was associated with higher basal levels of FSH (33,34,36), a higher total dose of gonadotropins required during ovarian stimulation (33,34,36), lower peak estradiol levels (34)(35)(36) and fewer retrieved oocytes (36). Collectively these studies suggest that the Ser/Ser variant is associated with a reduced sensitivity of the FSHR to exogenous FSH. A randomized controlled trial (RCT) demonstrated that this reduced sensitivity of the FSHR may be overcome by increasing the FSH dose (35).
Limited data suggest that A/A allele carriers of the FSHR −29 variant (rs1394205, −29G>A) require higher amounts of gonadotropin during ovarian stimulation than G/G carriers 9.
Limited data suggest that A/A allele carriers of the FSHR −29 variant (rs1394205, -29G>A) produce fewer oocytes in response to ovarian stimulation than GG carriers 8/9 merged (revote*) Limited data in specific ethnic groups suggest that A/A allele carriers of the FSHR −29 variant (rs1394205, −29G>A) require higher amounts of gonadotropin and produce fewer oocytes in response to ovarian stimulation than G/G carriers The data on an association of FSHB (FSHB rs10835638, −211G>T) with basal levels of FSH and production of oocytes in response to ovarian stimulation are contradictory Statements related to SNPs of the LHB/LHCGR genes 11.
Limited data suggest that polymorphisms of the LHB/LHCGR genes (V-betaLH rs1800447, c.82T>C, p.Trp8Arg; V-betaLH rs34349826, c.104 A>G, p.Ile35Thr; LHCGR rs2293275, c.935A>G, p.Asn312Ser) can influence ovarian stimulation outcomes and may represent targets for pharmacogenomic research in ART *Any statement that did not achieve consensus in Round 2 was discussed and reworded in Round 3, in order to revote with the Extended Panel. † FSHR rs6166 (c.2039A>G, p.Asn680Ser) and FSHR rs6165 (c.919G>A, p.Thr307Ala) are in linkage disequilibrium, except in some African populations. A, adenine; Ala, alanine; Asn, asparagine; Arg, arginine; ART, assisted reproductive technology; c., coding DNA sequence; C, cytosine; FSH, follicle-stimulating hormone; FSHB, folliclestimulating hormone b-chain; FSHR, follicle-stimulating hormone receptor; G, guanine; Ile, isoleucine; LH, luteinizing hormone; LHB, luteinizing hormone b-chain; LHCGR, luteinizing hormone/choriogonadotropin receptor; SNP, single nucleotide polymorphism; T, thymine; Trp, tryptophan; Thr, threonine.  (35). In addition, a retrospective evaluation in 42 women aged <37 years compared genetic and clinical characteristics in those who required a higher cumulative dose of r-hFSH (>2500 IU; n=17) with those who required a lower cumulative dose of r-hFSH (<2500 IU; n=24) (37). The number of oocytes retrieved (p=0.0005) and embryos transferred (p=0.001) were significantly greater in women with a higher r-hFSH consumption compared with those with a lower r-hFSH consumption. The incidence of the Ser/Ser variant was higher in patients requiring a higher r-hFSH dose (p=0.02), suggesting that this genotype may be associated with a poor response to FSH, that may be overcome by higher FSH doses (37).  Figure 2). The reasons given by participants for disagreeing with this statement included poor replication of data between studies, the fact that most studies were conducted in homogeneous populations and that the extent to which FSHR rs6166 affected basal FSH levels depended on the study population (Supplementary Table 2).
A retrospective study in 1250 Chinese women with poor (≤5 oocytes) or good (>5 oocytes) ovarian response who were undergoing IVF/ICSI reported higher basal FSH levels in women with the Ser/Ser variant compared with the Asn/Asn or Asn/Ser variants (50). In another retrospective study in 136 women with either poor response (n=22), normal response (n=57) or high response (n=57), basal FSH levels (Day 2) were significantly higher in women with the Ser/Ser variant compared with the Asn/Ser or Asn/Asn variants (p<0.05) (53) Figure 2). The main reasons stated by participants for disagreeing with this statement included the fact that most data were from observational studies with a limited number of patients, and the absence of data from RCTs. Furthermore, one participant suggested that the extent to which FSHR rs6166 affected oocyte number was dependent on the study population (Supplementary Table 2).
A number of studies suggest that women with the FSHR rs6166 (c.2039A>G, p.Asn680Ser) Ser/Ser variant produce fewer oocytes in response to ovarian stimulation than women with the Asn/Asn or Asn/Ser variants, despite there being no statistically significant difference in the gonadotropin dose among women with the different polymorphisms (31,36,50,(54)(55)(56)(57)(58)(59).
In a retrospective study in Mexican Mestizo women (n=224) there was a lower distribution for the Ser/Ser variant (9.8%) compared with the Asn/Asn variant (41.9%) or Asn/Ser variant (48.2%) (59). The Ser/Ser variant was associated with a significantly reduced number of retrieved oocytes following ovarian stimulation compared with the Asn/Asn or Asn/Ser variants (p<0.01) in normal oocyte donors. There was also a trend towards lower pregnancy rates in women with the Ser/Ser variant, which was stronger in a separate analysis of women with more Native American ancestry ( Figure 2). The wording of the statement was revised to reflect the fact that there are few large, prospective studies and no RCTs available regarding the influence of genotype on gonadotropin stimulation protocols. Furthermore, there were insufficient data on the comparison of homozygotic carriers of FSHR rs6166 (c.2039A>G, p.Asn680Ser) with grouped heterozygotic carriers. The reworded statement reached 57% agreement during re-voting ( Figure 2); therefore, this statement did not achieve consensus. This statement received 10% disagreement from the Extended Panel, with 33% of experts neither agreeing nor disagreeing with the statement. The main motivation for disagreement was insufficient evidence to support this statement (Supplementary Table 2).
In a study including 455 consecutively enrolled women and 210 unselected women aged <40 years, the number of MII oocytes in a subgroup of women who underwent ICSI (n=317) was significantly lower in Ser/Ser carriers compared with Asn/Ser carriers or Asn/Asn carriers (mean [SD]: 6 (41). However, the differences between the variants were not significant in the study, validation or merged cohorts, suggesting no influence of the polymorphisms on receptor sensitivity for in vitro stimulation response (41).
A meta-analysis of five studies [n=1185 women (also including 41)] reported a numerically lower number of oocytes in women with the Ser/Ser variant compared with women with the Asn/Asn variant, although the difference was not significant after Bonferroni correction (fixed weighted mean difference 1.03, 95% CI 0.01 to 2.05; p=0.05. Bonferroni adjusted p=0.14; I 2 = 0%). No significant differences were observed between women with the Asn/Asn variant and those with the Asn/Ser variant (fixed weighted mean difference 0.79, 95% CI -0.05 to 1.62; I 2 = 0%) or between those with the Ser/Ser variant and those with the Asn/Ser variant (fixed weighted mean difference 0.34, 95% CI -0.57 to 1.26; I 2 = 49%) (31). This statement reached 67% agreement during the first round of voting ( Figure 2). The main reasons stated by participants for disagreeing with this statement included insufficient evidence and the fact that most data were from observational studies, with an absence of data from RCTs, or from studies in specific populations (Supplementary Table 2).
This statement was supported by evidence from three studies (38,58,66). A retrospective study of 586 women undergoing their first IVF treatment determined whether the FSHR rs6166 (c.2039A>G, p.Asn680Ser) predicted the likelihood of developing OHSS (38). In this study, 36 women (6%) developed OHSS, of whom none carried the Ser/Ser variant. FSHR rs6166 was associated with OHSS (P trend = 0.004 and P allele = 0.038), with carriers of Asn having an OR of 1.7 (95% CI 1.0, 2.8; p<0.04), compared with carriers of Ser. Women who developed OHSS were exposed to a lower total hormonal dose, yet produced more oocytes than those without OHSS (16 ± 8 vs 11 ± 6; p=0.001) (38). Conversely, a meta-analysis conducted by Tang et al., comprising 16 cohort studies and a total of 4287 women, reported no evidence for an association between FSHR rs6166 variants and OHSS (OR: 1.58, 95% CI: 0.41, 6.07) (58). Finally, in their retrospective study of 150 Indian women (n=50 in an assisted reproductive technology [ART] program and n=100 with proven fertility [control group]), Achrekar et al. explored the association between FSHR rs6166 variants and variable ovarian response, including the occurrence of OHSS (66). The distributions were 31%, 56%, and 13% in controls and 42%, 46%, and 12% in ART patients, for the Asn/Asn, Asn/Ser and Ser/Ser variants, respectively. (66). Patient age, basal FSH and LH levels, progesterone levels before and on the day of human choriogonadotropin (hCG) administration, number of pre-ovulatory follicles, number of oocytes retrieved, and pregnancy rates showed no statistically significant differences among groups, suggesting that treatment outcome was independent of the FSHR rs6166 variants. There were no statistically significant differences in any of the clinical parameters among women with the different variants, although women with the Ser/Ser variant showed higher serum estradiol levels before or on the day of hCG administration. OHSS developed in 50%, 26%, and 29% of women with the Ser/Ser, Asn/Ser, and Asn/Asn variants, respectively, although these values were not statistically significant (OR: 2.67) (66). The original wording of this statement received 55% agreement during the first round of voting ( Figure 2). Following discussion regarding the lack of prospective studies reporting on this outcome, the wording was revised and received 67% agreement after re-voting ( Figure 2). The reasons given by participants for disagreeing with this statement included the fact that there was a limited number of studies and that the available studies had been poorly designed (Supplementary Table 2).

STATEMENTS RELATED TO SNP OF
In one study including 450 Chinese women undergoing IVF due to male factor, tubal factor, or both, the length of stimulation was significantly different among women with different variants (although only a small absolute clinical difference was observed), with the longest duration of stimulation reported in women with the Ala/Ala variant: Thr/Thr 11.32 (2.15) days, Thr/Ala 12.02 (2.44) days and Ala/Ala 12.62 (2.92) days; p<0.05 (56). A prospective, cross-sectional study of 149 women in Brazil undergoing ART treatment due to male factor (n=93) or tubal factor (n=56) did not report on the stimulation length in women with different variants, but there was no significant difference in mean (SD) basal FSH levels (Thr/Thr 6.0 [2.0] versus Ala/Ala 6.41 [1.96] and Thr/Ala 6.49 [1.73], p=0.402), suggesting a comparable ovarian response for the alleles related to this variant (32).
In a meta-analysis of three studies (679 patients, including Yan 2013), a shorter duration of stimulation that approached statistical significance was reported between women with the Thr/Thr and Ala/Ala variants (random weighted mean difference −0.59 [95% CI −1.24, 0.05], I 2 = 60%, p=0.07). However, the duration of stimulation was significantly shorter in women with the Thr/Thr variants than in those with the Thr/ Ala variant (fixed weighted mean difference −0. 48 Figure 2). The statement was revised to highlight that the FSHR variant (FSHR rs6165, c.919G>A) is in linkage disequilibrium (Box 2) with the FSHR rs6166 (c.2039 A>G, p.Asn680Ser), and following re-voting the revised statement reached 62% agreement ( Figure 2); therefore, this statement did not achieve consensus. This statement received 19% disagreement from the Extended Panel, with an additional 19% of experts neither agreeing nor disagreeing with the statement. The motivations supporting these disagreements are outlined in Supplementary Table 2, with some experts believing there was insufficient data to support this statement and others suggesting that it depended on how the gonadotropin dose was adjusted.
In a meta-analysis of five studies [including (66) and (56) p=0.005). The Ala/Ala variant was threefold higher in poor ovarian responders compared with good ovarian responders, and the presence of a G allele significantly increased the probability of a poor ovarian response (67). Finally, in a study of 450 Chinese women who were categorized according to ovarian response (poor <5 oocytes retrieved; normal 5-14 oocytes retrieved; high >14 oocytes retrieved), the proportion with the Ala/Ala variant was significantly higher in poor ovarian BOX 2 | Linkage Disequilibrium.
• Linkage disequilibrium is defined as the non-random association of alleles of different loci that are inherited co-ordinately.
• Linkage disequilibrium differs between ethnic groups, resulting in various combinations of the different SNPs (30).
• Such genetic distinctions could potentially explain the significant disparities reported in assisted reproductive technology (ART) outcomes according to ethnicity (63) and should be taken into account when assessing studies in different populations (64). • For example, FSHR rs6166 (c.2039A>G, p.Asn680Ser) and FSHR rs6165 (c.919G>A, p.Thr307Ala) are in linkage disequilibrium, except in some African populations (30,31).
responders than the proportions with the Thr/Thr variant or with the Thr/Ala variant (p<0.001) (56).  Figure 2). The same concerns were identified for each statement and so they were softened and merged with the specification that the effect can be seen in 'at least some ethnic groups' for the second round of voting, during which 76% total agreement was reached. The reasons given by participants for disagreeing with this statement are shown in Supplementary  Table 2. One participant explained that ethnicity had been shown to play a fundamental role in SNPs and that 1% of the population would need to carry the FSHR −29 variant in order for it to have a clinical application, and another participant stated that the studies supporting this statement had not been correctly planned. The FSHR −29 variant (rs1394205, −29G>A) has been associated with reduced transcriptional activity, primary or secondary amenorrhea, and poor response to exogenous FSH (31,59,68). The latter has not been found consistently in other studies (59). There is controversial evidence for the influence of this variant on ovarian stimulation. Studies suggest that carriers of the A allele, which has a frequency of 49-55% in a Hispano-American population (59), have reduced FSHR expression, are more likely to have a poor ovarian response (31,68) and may require a higher amount of FSH during OS; exogenous FSH consumption was significantly higher (p<0.001) in A/A carriers than in G/G or G/A carriers in a retrospective study of 50 women undergoing ART (69). A meta-analysis of three studies (n=709 women) also found that A/A carriers needed significantly higher doses of FSH during OS (31). These studies were conducted in only a few homogenous populations where the polymorphism has been found to be highly prevalent (31) and the results may not be generalizable to different populations.

STATEMENTS RELATED TO SNP OF
This variant may be associated with fewer oocytes being retrieved during ART. A review of three studies reporting on this polymorphism showed that a lower number of oocytes were retrieved in A/A carriers than A/G or G/G carriers, although the association was not found to be significant (31). One of these studies reported that the mean (SD) number of oocytes retrieved was significantly lower in A/A carriers (6.00 [1.09]) than in G/G carriers (17.88 [1.75]; P = 0.003) (66). A study by Desai et al. (70) found that the mean (SD) number of oocytes retrieved in A/A carriers (10.50 [1.19]) was significantly lower when compared to G/G carriers (16.43 [1.50]; p=0.046) (70). This association has also been found in another study by the same group (68), who reported that increasing the dose of exogenous FSH did not improve oocyte development, probably due to insufficiency of FSHR expression in granulosa cells (68). These studies from one group were restricted to a small, homogenous population, and further studies in mixed populations are needed to generalize the results.  Table 2).

STATEMENTS RELATED TO SNP OF
This statement was supported by evidence from three studies (45)(46)(47). A cross-sectional study evaluated the potential effects of the FSHB rs10835638 (−211G>T) variant on hormonal profiles and IVF/ICSI outcomes in 140 normo-ovulatory women in Brazil (47). The distributions of the GG (wild-type) variant (n=102) and GT variant (n=38) were 86.4% and 13.6%, respectively; the TT variant was not detected in any women. A poor response to ovarian stimulation was more common in women with the GT variant compared with those with the GG variant (47.4% vs. 26.5%; p=0.010), and fewer oocytes were retrieved from those with the GT variant than from those with the GG variant (3.0 vs. 5.0; p=0.03). However, no difference in pregnancy rates were reported between women with different variants.
Two studies assessing the effects of FSHB rs10835638 on basal FSH levels, in women with known infertility compared with a control group of healthy women, reported that women expressing the T allele showed significantly higher FSH basal levels. One study reported that the T allele was associated with significantly higher basal FSH levels in both non-pregnant healthy women (n=169) and female partners in infertile couples (n=186) (T-allele effect: 0.80 IU/L, p=0.001 after Bonferroni testing) (46). FSHB rs10835638 was estimated to explain 3.5% of the total variance of the measured serum FSH levels in healthy women and 1.6% in the female partners of infertile couples, and could have a diagnostic value in fertility clinics to detect female patients with genetically inherited elevated basal FSH and LH levels (46). In the other study, eumenorrheic women attending an IVF unit for predominantly male-factor infertility (n=365) were compared with a control group of women with proven fertility (n=438) (45). The distribution of the variants was 2.5% for the TT variant (n=9), 23.8% for the GT variant (n=87) and 73 Figure 2). The reason given by participants for disagreeing with this statement was that it had not been proven (Supplementary Table 2).
Four studies supported this statement, the collective results of which suggest that clinicians should be aware of patients with LHB polymorphisms, who may, consequently, fail to respond to ovarian stimulation (43,(71)(72)(73).
Hypo-sensitivity to exogenous FSH was observed in a retrospective study of 220 normogonadotropic Danish women undergoing controlled ovarian stimulation who were carriers of V-betaLH (rs1800447 [c.82 T>C, p.Trp8Arg] and rs34349826 [c.104 A>G, p.Ile135Thr] polymorphisms), which are common genetic variants of LHB (43). Daily doses of r-hFSH were administered on an individualized basis, tailored to age, body mass index, baseline FSH, and antral follicle count. The LHB genotype was assessed by immunofluorometric assay. A total of 24 women carried the V-betaLH variant, of whom 21 were heterozygous and three were homozygous; 196 of the women were wild type. The differences in the mean number of oocytes retrieved and the fertilization and pregnancy rates for each cycle were not statistically significant between the V-betaLH genotypes, but carriers of V-betaLH variant received a significantly higher cumulative dose of r-hFSH compared with women with wild type LH (2435.86 ± 932.8 IU vs. 1959.8 ± 736.45 IU; p=0.048). A within-design one-way ANOVA analysis showed that the V-betaLH variants had a statistically significant effect (p<0.01) on the cumulative dose of r-hFSH, with a mean (SD) increase from 1959.8 (736.45) IU for wild type carriers, to 2267.5 (824.3) IU and 3558.3 (970.9) IU, for heterozygotic and homozygotic carriers, respectively. These results confirm that carriers of V-betaLH variants have hypo-sensitivity to exogenous FSH during controlled ovarian stimulation (43).
In their systematic review of the current status of pharmacogenetic analyses of controlled ovarian hyperstimulation, Altmäe et al. concluded that there is accumulating data to suggest that the ovarian response to COH is mediated by various polymorphisms, including variants of the V-betaLH and LHCGR genes (72). However, further studies investigating the predictive value of such genetic polymorphisms as markers of COH in subgroups of women who may require supplementation with exogenous LH during ovarian stimulation are needed (72). In addition, the results from an observational preliminary trial of 60 normogonadotropic patients undergoing IVF/ICSI suggested that women with the most common polymorphism of LHB (V-betaLH) were hyporesponsive to r-hFSH (72). A greater proportion (31.8%) of carriers of V-betaLH were identified among 22 women who required a cumulative dose of r-hFSH of >3500 IU, relative to those who required between 2000 and 3500 IU r-hFSH (one woman [6.7%] V-betaLH of 15 women). V-betaLH variants were detected in 23 women who required <2000 IU r-hFSH (73). Lastly, sequence analysis indicated that heterozygous point mutations in the LHB gene (Trp8Arg and Ile15Thr) were present in a 35-year-old woman who had failed to conceive after six cycles of human menopausal gonadotrophin (hMG) therapy for ovarian induction (71). The observed LH hypersecretion was likened to that seen in PCOS. A further cycle of ovarian stimulation with hMG, during which estrogen-progesterone replacement therapy effectively controlled basal LH and FSH, led to successful conception and delivery outcomes (71).

DISCUSSION
This Delphi consensus provides clinical perspectives from a diverse international group of experts. It generates a series of literature-supported consensus statements regarding the influence of specific gonadotropin and gonadotropin receptor variants on clinical ovarian stimulation outcomes that will be useful to optimize current stimulation protocols. The consensus results suggest that there is evidence to support a link between SNP variants in gonadotropin and gonadotropin receptors and ovarian stimulation outcomes, although data for some variants are still lacking.
Although there is strong supporting evidence for the impact of polymorphisms of FSHR on several outcomes, this Delphi Consensus shows there are only modest data on the clinical relevance that would support these polymorphisms as the basis for pharmacogenetic approaches to treatment, although we acknowledge that this is an area where new data are being published. A recent multicentre, multinational, prospective study in 368 predicted normal responder women from Vietnam, Belgium, and Spain, published since the selection of the literature to be considered in this Delphi consensus, reported only minimal clinical impact of genotyping for FSHR SNPs rs6165 (c.919G>A, p.Thr307Ala), rs6166 (c.2039A>G, p.Asn680Ser), rs1394205 (−29G>A) and FSHB SNP rs10835638 (−211G>T) prior to initiating ovarian stimulation with r-hFSH (74). Although the study reported a significantly lower number of oocytes retrieved in heterozygous patients for the FSHR variants rs6166 and rs1394205, as well as a significantly higher rate of hyporesponse in heterozygous patients for the FSHR variant rs6166, this resulted in a change of just one or two oocytes in a population of normal responders (74).
Currently, there is contradictory evidence for the clinical impact of polymorphisms in FSHB on ovarian stimulation. Three studies supporting this consensus reported that women expressing the T-allele of the FSHB rs10835638 (−211G>T) polymorphism had fewer retrieved oocytes (47) and significantly higher basal FSH levels compared with women carrying the wild type (GG) variant (45,46). However, in contrast to the aforementioned studies, La Marca et al. observed that FSHB rs10835638 does not affect FSH basal levels per se. The authors observed a significant reduction of FSH basal levels in women expressing the T-allele compared with the wild-type genotype (75).
The inconsistency between some studies, which is also highlighted in the Alviggi 2018 systematic review and metaanalysis (31), is likely due to differences in inclusion criteria between studies, as well as the use of different gonadotropin products and doses, and allowance for dose adjustments during treatment, highlighting the need for new prospective studies in this field.
As already stated in this Delphi consensus, there are limited data on the influence of polymorphisms of the LHB/LHCGR genes on ovarian stimulation outcomes and their usefulness in pharmacogenomic research in ART. While the usefulness of polymorphisms in these genes per se remains to be determined, a recent study (also published since the selection of the literature for consideration in this Delphi Consensus) may indicate some clinical value in determining the need for LH supplementation compared with current supplementation protocols, reporting higher clinical pregnancy rates (p=0.049) and a trend towards improved live birth rates (p=0.082) in 193 women when supplementation was based on a woman's SNP profile compared with conventional methods (76). Furthermore, in a cross-sectional study of LHCGR rs2293275 (c.935 A>G, p.Asn312Ser), Lindgren et al. reported no significant difference in the number of oocytes retrieved or obvious differences in embryo quality between Ser/Ser, Asn/Ser or Asn/ Asn carriers. However, significantly higher clinical pregnancy rates were reported in Ser/Ser carriers compared with Asn/Asn carriers (OR 1.61 [95% CI 1.13, 2.29], p=0.008) (41). These studies suggest that individualising protocols based on specific genotypes, rather than the number and morphological characteristics of the embryos retrieved, may be beneficial in terms of improved pregnancy rates.

Strengths
This consensus has a number of strengths, including the fact that each of the statements in the consensus were supported by a number of peer-reviewed studies. Furthermore, the majority of women included in these studies were aged <35 years (mean/ median age~32 years). As advanced maternal age (>35 years) is associated with a reduction in ovarian reserve, oocyte/embryo competence and cumulative live birth rate (77)(78)(79), the fact that most women in these studies were aged <35 years suggests that age was less likely to impact ovarian response, making it easier to discern the influence of SNP variants. Another strength was that the participants of the consensus were fertility experts from across the globe, representing different regions, including Europe, Asia, and South America, reflecting the quality of healthcare and different approaches to infertility treatment in different parts of the world.

Limitations
The consensus does have some limitations that should be acknowledged. Firstly, the consensus does not represent an exhaustive list of statements referring to all polymorphisms potentially affecting ovarian stimulation. Furthermore, the statements and supporting literature were chosen according to the expert opinion of the Scientific Coordinators and the Scientific Board, and the inclusion of the statements and supporting literature was based on the expert opinion of the Extended Panel. Another limitation was that none of the statements reached 100% agreement, with most statements reaching consensus even though some participants disagreed with them, and only three statements (Statements 1, 10 and 11) achieved a high level of agreement (defined as ≥80% of participants voting 'agree' or 'absolutely agree'). Furthermore, not all statements reached consensus, with Statements 4 and 7 failing to reach consensus even after the statements were revised and revoted. The main reasons stated by participants for disagreeing with a given statement included a limited number of studies, small sample sizes, poor replication of data between studies, and the fact that most data were from observational studies based on homogeneous populations and with homogeneous protocols of ovarian stimulation. Furthermore, a number of participants cited the absence of data from RCTs as a reason for disagreeing with a given statement.

Future Research
Our consensus represents an opportunity to encourage researchers to initiate studies to verify whether a pharmacogenomic approach (i.e. the individualization of treatment based on a patient's genetic profile) could lead to an improved patient-tailored approach to ovarian stimulation. This could be beneficial, as a more unified approach among studies will enable us to reach future clinical decisions. Given the lack of studies on pharmacogenomic approaches, the consensus was unable to include any statement(s) concerning a pharmacogenomic approach to ovarian stimulation. Conversely, we believe that there is sufficient evidence from the results of our consensus to support the fact that specific genetic variants could represent interesting causative factors of impaired ovarian response that cannot be explained by other parameters such as ovarian reserve markers. Gene association studies could focus on all SNPs known to influence ovarian response (including those highlighted in this Delphi consensus), in order to examine whether an individualized treatment approach may significantly improve ovarian stimulation outcomes, compared with a standard approach. Ideally, the ovarian response would be correlated with the "global genetic background" assessed by SNP genotyping, exome sequencing or even whole genome sequencing combined with artificial intelligence in large, multicenter studies. Greater efforts should focus on increasing the number of observations, for example, by utilizing large-scale datasets such as the UKBioBank (80). Hypothetically, the development of a large international registry concerning genome-wide association studies in IVF could also be beneficial. This strategy would lead to "real-world" data that could provide interesting findings concerning the importance of genetics in Reproductive Genetics and ART (81,82). Nonetheless, the quality of association studies in Reproductive Genetics should take into account selection bias, different treatments and differential follow-ups among IVF centers. In addition, the use of check-list proposed STROBE or GRACE guidelines should be encouraged to increase the overall quality of published articles (21,83,84).
Specifically, further research could include investigating an association between FSHR rs6166 (c.2039A>G, p.Asn680Ser) and ovarian morphology, AFC following ovarian stimulation and basal AFC in different age groups. A multivariate analysis assessing the impact of r-hLH supplementation according to SNP status (specifically, FSHR rs6166 [c.2039A>G, p.Asn680Ser] and LHCGR rs2293275 [c.935A>G, p.Asn312Ser]) would also be relevant, as would investigating the effect of follicular fluid steroid hormone levels in different FSHR and LHCGR SNPs and their implication as a biomarker for oocyte preservation in women postponing pregnancy. Furthermore, it would be of interest to test the hypothesis that different r-hFSH doses with or without r-hLH supplementation may improve ongoing pregnancy and live birth rates per started cycle compared with a standard dose, taking into account age and ovarian reserve, and according to SNP status.

CONCLUSIONS
This Delphi consensus supports a link between some genetic variants in gonadotropin and gonadotropin receptors and ovarian stimulation outcomes. The consensus results reinforce the idea that pharmacogenomics may provide a promising new field examining genotype-specific responses to ovarian stimulation medication that may help tailor ovarian stimulation therapies to individual patients, optimizing ART success outcomes.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

ETHICS STATEMENT
Written informed consent was obtained from all Consensus participants for the publication of their name in Table 2.

AUTHOR CONTRIBUTIONS
AC and FT were Scientific Coordinators for the Delphi consensus; AC, FT CA, HMB, RF, LH, NPP, GARR, MS, and SKS were members of the Scientific Board. One Scientific Board member declined authorship of the article owing to personal reasons. All authors made substantial contributions to the concept and design, or the acquisition, analysis or interpretation of data, and to the drafting of the manuscript or revising it critically for important intellectual content. In addition, all authors provided final approval of the manuscript.

FUNDING
The work was funded by Merck KGaA, Darmstadt, Germany.

ACKNOWLEDGMENTS
Medical writing support was provided by Helen Brereton and Nichola Cruickshanks, inScience Communication, and funded by Merck KGaA, Darmstadt, Germany. The Delphi consensus process was coordinated by Sanitanova Srl. The authors would like to thank the non-author members of the Extended Panel listed in Table 2 for their participation in the consensus.