The polygenic score for age at menopause is independently associated with the response of controlled ovarian hyperstimulation in women undergoing IVF.

The data generated and/or analysed during the current study are not publicly available due to legal and ethical regulations. The data will be made available to qualified researchers upon reasonable request to the corresponding author. Available data include de-identified individual participant data and key variables underlying the analyses presented in this manuscript. Access will be granted to researchers whose proposed use of the data is consistent with the aims of the original study and complies with institutional and ethical guidelines. Data will be shared via secure file transfer after signing a data transfer agreement. The analytical code used to calculate the polygenic scores and perform the Mendelian Randomization analysis is publicly available on GitHub ( https://github.com/slamballais/anm_ivf2026 ).

We performed a cohort study to investigate whether an ANM PGS based on a large-scale ANM GWAS is predictive of the outcome of IVF stimulation. 13 After inclusion, a blood sample was taken for genetic analysis and patient files were used to collected information on covariates and the outcome the IVF treatment. When women started their IVF procedure, AMH is always measured and this single blood draw was also used to collect the necessary sample for genetic analysis. In addition to AMH, we also tested for luteinising hormone (LH), follicle-stimulating hormone (FSH), oestradiol, and progesterone only from the blood samples taken between day 2 and day 6 of the menstrual cycle. Inclusion in this study did not alter the IVF treatment. This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Erasmus Medical Centre (The Netherlands) (protocol code NL75062.078.20; November 2020). Written informed consent was obtained from all participants included in the study. The included women were between the age of 18 and 45 years, had a regular menstrual cycle, and had undergone their first IVF treatment in the Erasmus Medical Centre from August 2017 until July 2024. Women were excluded if they had a history of ovarian surgery, chemotherapy, or radiation therapy, or if they lacked a Northern European ancestry, given the source GWAS for the PGS. All included women were treated with follitropin delta and the dose was personalised using a standardised dosing algorithm as described by the manufacturer based on body weight and AMH serum levels. All included women received follitropin delta (Rekovelle, 72 μg/2.16 ml) via a daily subcutaneous injection in the abdomen. The personalised dose was determined by the patient's body weight and their serum AMH level, which was sampled at the Erasmus MC laboratory as part of the standard treatment, no more than one year before the start of IVF. Specifically, women with AMH ≥2.10 μg/L, the daily follitropin delta dose ranged continuously from 0.19 to 0.10 μg/kg, thus depending on both AMH and body weight. Women with AMH <2.10 μg/L received a fixed daily dose of 12 μg, regardless of body weight and women with AMH <1.00 μg/L received a fixed daily dose of 18 or 24 μg. The extra blood sample was drawn and stored at −80 °C until genotyping was conducted. DNA was extracted from peripheral blood and genotyped using the Illumina Global Screening Array (Multi-Disease v3). Quality control consisted of filtering for SNP call rate (first <90%, then <97.5%), Hardy–Weinberg equilibrium (p < 0.00001), excess heterozygosity, sex mismatch, and genetic duplicates. The data was phased with ShapeIT v2 r904 and imputed with Minimac4 v1.0.2 against the HRC r1.1 reference panel. Post-imputation, variants with an imputation quality score (R 2 ) > 0.3 were retained. Ten genomic components (GCs) of ancestry were calculated based on a pruned subset of independent markers. We calculated the PGS for ANM using the Bayesian regression method PRS-CS. This method estimates posterior effect sizes for SNPs using a continuous shrinkage prior, accounts for local linkage disequilibrium and the estimated proportion of causal variants, and assigns weights to variants when calculating the PGS. We utilised summary statistics from a previous large-scale GWAS of ANM (∼200 K samples). 13 The 1000 Genomes Phase 3 European subset was used as the LD reference panel. The global shrinkage parameter was learnt automatically from the data. The procedure resulted in 554,777 variants with non-zero weights included in the final score. The PGS was calculated for each individual by summing the dosage of risk alleles weighted by their posterior effect sizes using PLINK 2.0. 22 To assess the quantity of the ovarian response we used the following definitions. First, the ‘Number of follicles 2 days before ovum pick-up (OPU)’ which were >11 mm, this outcome includes women with cancellation of the cycle. Second, the ‘Number of follicles aspirated at OPU’. Third, the ‘Number of oocytes’ that were aspirated during OPU. Finally, ‘Diminished ovarian response’ (DOR) was defined by a cycle cancellation due to low response (<3 follicles of at least 14 mm) or less than 3 oocytes aspirated during OPU. The DOR group was further categorised into ‘Expected DOR’ and ‘Unexpected DOR’. Patients were classified as ‘Expected DOR’ based on clinical markers of low ovarian reserve (AMH <1 ng/ml or AFC <5), which dictated a treatment protocol with follitropin delta dosages exceeding the standard maximum. To assess the quality of the ovarian response we used the following definitions. First, the ‘Number of good quality oocytes’, which was defined as the number of metaphase 2 oocytes in a subset of assessed oocytes (in ICSI, social freezers and in IVF with TTF or low fertilisation rate). Second, the ‘Number of formed embryos’ defined as the number of 2 PN (pronuclei). Third, the ‘Total number of embryos’ defined as all embryos which were used for embryo transfer (fresh or frozen). Finally, ‘Ongoing pregnancy’ defined as any pregnancy exceeding 12 weeks of gestation resulting from a fresh or frozen embryo transfer. While data regarding ‘Number of follicles 2 days before OPU’ was available for all included participants, a natural loss to follow-up occurred for later outcome measures as a result of cycle cancellations and failure of embryo development. Sample size calculations were performed with the ‘pwrss’ R package (pwrss:pwrss.f.reg). In the UK Biobank, the median of explained variance for the ANM PGS onto 539 traits was 4.4%. 23 The link between the ANM PGS and ovarian response outcomes may be partly mediated by AMH, which is already accounted for in the Rekovelle concentration. Therefore, we conservatively estimated that we would want to find an explained variance of the ANM PGS of 1.5–2% with power = 0.80 and alpha = 0.05, leading to a sample size of 387–518. For interpretation purposes, we standardised the PGS using a z-score (zPGS). Baseline characteristics were shown in mean with standard deviation (SD) or median with interquartile range (IQR) based on the distribution. For continuous outcomes we used a linear regression model to calculate the R square, beta of the standardised PGS (zPGS) and its p-value. Model 1 included the zPGS and ten GCs. Model 2 included the zPGS, ten GCs and age. For categorical outcomes we used an analysis of covariance (ANCOVA) for both models. The statistical analyses were performed using the Statistical Package of Social Sciences version 24.0 for Windows (SPSS Inc., Chicago, IL, USA). To validate the PGS findings, we performed a two-sample Mendelian Randomisation (MR) analysis exploring the causal relationship between ANM (exposure) and number of follicles 2 days before OPU (outcome). Genetic instruments were obtained from the 290 top independent hits reported by Ruth et al. (2021). 13 We extracted 233 of these variants in our cohort and performed linear regressions against the number of follicles 2 days before OPU, adjusting for age and the top 10 genomic components, to generate outcome-specific summary statistics ( Supplemental Table S1 ). Palindromic variants with a minor allele frequency >0.42 in our cohort (n = 6) were excluded. The MR analysis was conducted using the TwoSampleMR R package. We used the inverse variance weighted (IVW) method as the primary test. Sensitivity analyses included the weighted median and MR-Egger regression to assess robustness to horizontal pleiotropy. We also investigated if the PGS offered additional value beyond AMH. To do this, we first verified the association between AMH and the outcomes, therefore we created model 3, which solely included AMH. Then, we developed model 4, which incorporated zPGS, ten GCs, age and AMH. Furthermore, we conducted causal mediation analyses to determine if AMH levels mediated the association between the PGS and the outcomes: ‘Number of follicles 2 days before OPU’ and ‘Number of oocytes’. We selected these outcomes due to their high statistical power and clinical relevance, assuming that similar results would apply to other outcomes. Causal mediation analysis helps us understand the separate contributions of the PGS's direct effect on the outcomes and its indirect effect through a mediator (i.e. AMH). We used non-parametric bootstrapping to estimate the confidence intervals for these effects. A key assumption of causal mediation is that all confounding factors for both the direct and indirect effects are accounted for. To address this, we corrected for age and ten genomic components when assessing the exposure-outcome association, these components were included to control for population stratification and account for ancestral genetic background. For the mediator–outcome association, we additionally adjusted for known confounders such as BMI, 24 smoking, 24 cycle duration, 25 stimulation dose, 26 socioeconomic status (SES), 27 education, 28 and age of menarche. 29 All analyses were conducted using the ‘mediation’ package (mediation:mediation) in R v4.5.0 (Tingley et al., 2014). 30 Funders had no role in study design, data collection, data analyses, interpretation, or writing of the report.

Between August 2017 and July 2024, 490 women were invited to participate. Of these women 55 were excluded because they had not started the IVF cycle at the time of the analysis (n = 3), were non-northern European based on DNA (n = 6), had an irregular cycle (n = 2), the stimulation was done with another medication protocol (n = 16), they had a history of IVF treatment somewhere else (n = 20) or a history of an operation to a part of the ovary (n = 8), resulting in 435 women included for analyses. The included women did not differ significantly from the excluded women on baseline characteristics. The baseline characteristics of the included women are shown in Table 1 . The mean age was 33.9 years (standard deviation (SD) = 4.2). Most women used the oral contraceptive pill before (91.8%) and were nulligravida (74%). The indications for IVF were male factor (61.1%), unexplained subfertility (20.7%), tubal factor (2.5%), social freezer (6.9%), endometriosis (5.3%) and donor semen and oocyte donation (3.4%). Table 1 Baseline characteristics. Total (n = 435) Age, mean (SD) 33.9 (4.2) BMI, mean (SD) 24.4 (3.6) Education, n (%)  Low 4 (1.0)  Medium 134 (32.1)  High 279 (66.9) Social economic status, n (%)  Low 108 (24.8)  Medium 202 (46.4)  High 123 (28.3) History of oral contraceptive pill use, n (%) 394 (91.8) Total duration OCP use in years, mean (SD) 10.1 (5.8) History of intrauterine device use, n (%) 104 (24.4) Other contraceptive use, n (%) 70 (16.4) Total duration contraceptive use in years, mean (SD) 11.7 (5.7) Smoking, n (%)  Never 299 (69.9)  History 129 (30.1) Packyears, median (IQR) 0 (0–1) Alcohol use, n (%) 218 (50.3) Number of alcohol consumptions, median (IQR) 1 (0–2) Drugs use, n (%) 6 (1.3) Menarche age in years, mean (SD) 12.8 (1.5) Menopausal age mother in years, mean (SD) 50.3 (4.5) Gravida, median (IQR) 0 (0–1) Para, median (IQR) 0 (0–0) Spontaneous abortion, median (IQR) 0 (0–0) Cycle duration in days, mean (SD) 28.0 (2.0) LH in IU/L, median (IQR) 4.2 (3.0–5.2) FSH in IU/L, mean (SD) 7.4 (2.6) Oestradiol in pmol/L, median (IQR) 140.0 (113.0–183.0) Progesterone in nmol/L, median (IQR) <1.5 (<1.5–<1.5) AMH in μg/L, median (IQR) 2.5 (1.5–3.8) Co-treatment GnRH agonist, n (%) 34 (7.8) Co-treatment GnRH antagonist, n (%) 401 (92.2) Follitropin delta dose, median (IQR) 12.0 (9.0–12.0) Gonadotrophin trigger, n (%) 311 (82.7) GnRH agonist trigger, n (%) 65 (17.3) Indication, n (%)  Male factor 266 (61.1)  Unexplained subfertility 90 (20.7)  Tubal factor 11 (2.5)  Social freezer 30 (6.9)  Endometriosis 23 (5.3)  Other 15 (3.4) Normospermia, n (%) a 133 (30.6) Values are shown in means, medians or numbers depending on their distribution. BMI, body mass index; LH, luteinizing hormone; FSH, follicle-stimulating hormone; AMH, anti-Müllerian hormone; GnRH, gonadotropin-releasing hormone. a Since not all participants had a male partner, semen samples were tested from a subset of 401 individuals. Baseline characteristics. Values are shown in means, medians or numbers depending on their distribution. BMI, body mass index; LH, luteinizing hormone; FSH, follicle-stimulating hormone; AMH, anti-Müllerian hormone; GnRH, gonadotropin-releasing hormone. Since not all participants had a male partner, semen samples were tested from a subset of 401 individuals. Table 2 shows the mean values of the ovarian response outcomes and the associations between the PGS and these outcomes. A mean of 10.8 follicles were aspirated at OPU with a wide range of 2–33 follicles. A mean of 3.2 embryos were available and used for embryo transfer. The zPGS was significantly associated with all ovarian response outcomes, with and without adjusting for age. For every SD increase of the PGS (indicative for later menopause), we found a 0.950 increase in the number of follicles 2 days before OPU (p < 0.001) and 0.918 increase in the number of oocytes (p = 0.001) (linear regression model). Table 2 Association between polygenic score for age at natural menopause (PGS) and ovarian response outcomes—continuous variables. Mean (range) Model 1 Model 2 R square Beta zPGS p-value R square Beta zPGS p-value Beta age p-value Number of follicles 2 days before OPU a 10.5 (1–40) 0.031 0.854 0.002 0.083 0.950 <0.001 −0.302 <0.001 Number of follicles aspirated at OPU b 10.8 (2–33) 0.043 0.796 0.003 0.088 0.875 <0.001 −0.263 <0.001 Number of oocytes c 10.0 (1–29) 0.040 0.844 0.003 0.072 0.918 0.001 −0.237 <0.001 Number of good quality oocytes d 7.4 (0–24) 0.049 0.546 0.051 0.065 0.591 0.034 −0.156 0.020 Number of formed embryos e 5.0 (0–20) 0.042 0.395 0.050 0.060 0.433 0.031 −0.125 0.009 Total number of embryos f 3.2 (0–13) 0.040 0.249 0.049 0.059 0.273 0.030 −0.078 0.009 Model 1 included the standardised polygenic score for age at natural menopause (zPGS) and ten genetic components (linear regression), model 2 included the zPGS, ten genetic components and age (linear regression). a Data on the outcomes was available for n = 432. b Data on the outcomes was available for n = 386. c Data on the outcomes was available for n = 389. d Data on the outcomes was available for n = 322. e Data on the outcomes was available for n = 388. f Data on the outcomes was available for n = 362. Association between polygenic score for age at natural menopause (PGS) and ovarian response outcomes—continuous variables. Model 1 included the standardised polygenic score for age at natural menopause (zPGS) and ten genetic components (linear regression), model 2 included the zPGS, ten genetic components and age (linear regression). Data on the outcomes was available for n = 432. Data on the outcomes was available for n = 386. Data on the outcomes was available for n = 389. Data on the outcomes was available for n = 322. Data on the outcomes was available for n = 388. Data on the outcomes was available for n = 362. Table 3 shows the associations between the PGS and the categorical ovarian response outcomes. In both models we found that the PGS was not statistically significantly different in women with or without DOR (p = 0.19 for model 1; p = 0.15 for model 2) (ANCOVA). We did find a significant difference between the PGS of women with no DOR and expected DOR (p = 0.01) and between unexpected DOR and expected DOR (p = 0.03) in model 2 (ANCOVA). Women with Expected DOR had a lower PGS (indicative for earlier menopause) compared to no DOR and unexpected DOR. When we looked at the follicle numbers that were aspirated at OPU, we found that retrieving less than 8 follicles is significantly associated with a lower PGS (indicative for earlier menopause) than 8–14 follicles and more than 14 follicles (p = <0.01 and p = 0.02, respectively) (ANCOVA). Finally, no significant difference was found between the PGS of women with or without an ongoing pregnancy. Table 3 Association between polygenic score for age at natural menopause (PGS) and ovarian response outcomes—categorical variables. Model 1 Model 2 Mean difference in zPGS p-value Mean difference in zPGS p-value DOR a −0.172 0.19 −0.188 0.15 Expected and unexpected DOR b  No DOR vs Expected DOR 0.527 0.02 0.592 0.01  Unexpected DOR vs expected DOR 0.497 0.06 0.568 0.03  No DOR vs unexpected DOR 0.030 0.85 0.024 0.87 Follicle number c  <8 vs 8–14 −0.301 0.01 −0.330 14 0.024 0.86 0.009 0.95   14 −0.277 0.04 −0.320 0.02 Ongoing pregnancy d 0.028 0.78 0.059 0.57 Model 1 is the standardised polygenic score for age at natural menopause (zPGS) corrected for ten genetic components (ANCOVA), model 2 is the zPGS corrected for ten genetic components and age (ANCOVA). DOR, Diminished ovarian response. a 69 of 422 women (16.4%) had DOR. b 21 of 69 women (30.4%) had expected DOR and 48 of the 69 women (69.6%) had unexpected DOR. c 146 of 389 women (37.5%) had less than 8 follicles, 163 of the 389 women (41.9%) had 8–14 follicles and 8 of the 389 women (20.6%) had more than 14 follicles. d 174 of 397 women (43.8%) had an ongoing pregnancy. Association between polygenic score for age at natural menopause (PGS) and ovarian response outcomes—categorical variables. Model 1 is the standardised polygenic score for age at natural menopause (zPGS) corrected for ten genetic components (ANCOVA), model 2 is the zPGS corrected for ten genetic components and age (ANCOVA). DOR, Diminished ovarian response. 69 of 422 women (16.4%) had DOR. 21 of 69 women (30.4%) had expected DOR and 48 of the 69 women (69.6%) had unexpected DOR. 146 of 389 women (37.5%) had less than 8 follicles, 163 of the 389 women (41.9%) had 8–14 follicles and 8 of the 389 women (20.6%) had more than 14 follicles. 174 of 397 women (43.8%) had an ongoing pregnancy. Fig. 1 shows a boxplot of the zPGS corrected for GC and age for every number of follicles that was aspirated during OPU. We found a positive trend, i.e. a negative z-score of the PGS for the lowest number of follicles and a z-score of the PGS above 0 for the higher number of follicles that were aspirated (p = 0.012) (ANCOVA). Fig. 1 Boxplots of polygenic score for age at natural menopause corrected for genetic components and age of response to ovarian hyperstimulation during in vitro fertilisation. Boxplots of polygenic score for age at natural menopause corrected for genetic components and age of response to ovarian hyperstimulation during in vitro fertilisation. Two-sample MR identified a significant causal association, where genetic liability for later ANM was associated with a higher number of follicles 2 days before OPU (IVW β = 0.47, p = 0.03) ( Supplemental Figure S1 ). Although sensitivity analyses were limited by lower statistical power, the effect estimates were highly consistent across methods (weighted median β = 0.57; MR-Egger β = 0.45), and the MR-Egger intercept indicated no evidence of directional pleiotropy (p = 0.96). These results support a causal dose–response relationship, reinforcing that the association observed with the PGS is driven by ovarian ageing biology. Supplemental Table S2 shows the association between AMH and the ovarian response outcomes, AMH was significantly associated with all ovarian response outcomes. In model 4 we found that the zPGS was significantly associated with some of the ovarian response outcomes, when we additionally corrected for AMH. For every SD increase of the PGS, we found a 0.489 increase in the number of follicles 2 days before OPU (p = 0.041), 0.484 increase in number of follicles aspirated at OPU (p = 0.041) and 0.567 increase in the number of oocytes (p = 0.032) (linear regression). Supplemental Table S3 shows the associations between AMH and the categorical ovarian response outcomes, AMH was significantly associated with almost all ovarian response outcomes. In model 4 we found a significant mean difference of −0.259 in the zPGS when we compare <8 follicles vs 8–14 follicles 2 days before OPU, when we additionally corrected for AMH. To assess the relative contribution of AMH to the association between the PGS and ovarian response outcomes we conducted a causal mediation model. We found that AMH partially mediated the effect of the PGS on the number of follicles 2 days before OPU (proportion mediated = 31%, 95% confidence interval (CI) = [11%–71%], p = 0.002). This means that 31% of the total effect of the PGS on the number of follicles 2 days before OPU can be explained by changes in AMH levels. Similarly, AMH partly mediated the effect of the PGS on the number of oocytes (proportion mediated = 29%, 95% CI = [10%–74%], p = 0.001).

This study provides compelling initial evidence that a PGS derived from genetic variants associated with ANM is significantly associated with quantitative outcomes of ovarian response during the first IVF cycle. Notably, the observed R-squared values for predicting the number of follicles 2 days before OPU, follicles aspirated on the day of OPU, and retrieved oocytes are remarkably encouraging, suggesting that genetic predisposition for the timing of menopause may indeed play a role in ovarian responsiveness to exogenous FSH. This finding is particularly noteworthy as it demonstrates the potential for ANM-related genetic architecture to influence a seemingly distinct, but related, reproductive trait. In addition, we found that the PGS is independent of AMH associated with IVF outcomes. Suggesting that the PGS for ANM has the potential to serve as a novel and complementary predictor to AMH, offering additional insights into a woman's ovarian responsiveness and further optimising IVF stimulation protocols. While the PGS did not significantly predict the binary outcome of diminished ovarian response (DOR), a closer examination of the continuous follicle count reveals a significant association. Women with a low number of aspirated follicles ( 14 follicles. This suggests that the PGS may capture a gradient of ovarian reserve and responsiveness. Even if we could accurately predict conditions like DOR, clinicians would likely still proceed with treatment for these patients. Therefore, viewing ovarian response as a continuous measure offers a more positive and nuanced outcome for patients, and could also open doors for improvement of the treatment. Furthermore, our sample size calculation was primarily based on the expected effect size for continuous outcomes like the number of aspirated follicles, drawing parallels from well-powered ANM GWAS, and was not specifically powered to assess oocyte quality or pregnancy outcomes. Additionally, pregnancy outcomes depend on numerous factors beyond oocyte quality alone, including semen quality and endometrial receptivity. Consequently, the non-significant association between the PGS and ongoing pregnancy should be interpreted with caution and warrants further investigation in larger studies specifically designed to assess these endpoints and DOR. While our study focused on the polygenic architecture of ANM and its association with IVF outcomes, it is important to consider the potential contribution of specific monogenic variants. For instance, the FMR1 gene has been implicated in premature ovarian insufficiency and altered ovarian reserve. 18 Furthermore, the study by Anagnostou et al. (2012) highlighted the potential of combining polymorphisms in ESR1, ESR2, and the FSH receptor gene to predict poor responders. 31 Future research could explore the combined effects of our ANM-derived PGS and specific monogenic variants, including those related to FSH signalling and ovarian function, to potentially improve the predictive power for IVF outcomes and provide a more comprehensive genetic assessment for women undergoing fertility treatment. Our study population underwent a standardised, algorithm based, IVF stimulation protocol with follitropin delta dosed based on individual body weight and AMH levels. Despite AMH being included in the algorithm, we still observe a significant association between AMH and IVF outcomes. This indicates that the algorithm does not fully account for the effect of AMH, suggesting that correction for AMH remains warranted. Furthermore, we demonstrated that AMH acts as a mediator; however, the proportion of this mediation is only around 30%, implying that the remaining association between PGS and IVF outcomes is attributable to PGS itself and thus due to the genetic make-up of the patient herself. This is further supported by Model 4, where a significant association with PGS persists even after correcting for AMH. We acknowledge that not all IVF clinics utilise standardised IVF stimulation protocols with follitropin delta. Our hypothesis is that in settings without AMH personalised dosing, the effect of the PGS will be even more pronounced because the influence of AMH has not yet been incorporated into treatment decisions. Further research in more heterogeneous treatment settings or studies specifically designed to disentangle the interplay between genetic factors, AMH, and ovarian response are needed to fully elucidate this relationship. The standardisation of our PGS into a z-score (zPGS) facilitates interpretability within our study population, allowing for a clear understanding of the direction and magnitude of the genetic effect on ovarian response. However, for future clinical translation, it will be crucial to establish a reference group and develop a clinically meaningful conversion of the PGS into a clinically actionable metric. This would involve genotyping a larger, representative population of women undergoing IVF to define the distribution of the PGS and establish clinically relevant cut-offs for predicting different levels of ovarian response. In conclusion, our findings provide evidence supporting a link between the genetic predisposition for the timing of menopause and ovarian response to FSH during IVF, independent of AMH. While further research is needed to validate these findings in larger and more diverse cohorts, and to explore the clinical utility of this PGS, our study represents a significant step towards understanding the complex interplay of genetic factors influencing ovarian ageing. Future studies should focus on refining the PGS, investigating its predictive power for various IVF outcomes including oocyte quality and pregnancy, and exploring its interaction with other genetic and clinical markers to ultimately personalise fertility treatments.

Conceptualisation: J. Laven, Y. Louwers, Data curation: C. van Zwol—Janssens, S. Lamballais; Formal analysis: C. van Zwol—Janssens, S. Lamballais, E. Loehrer; Funding acquisition: J. Laven; Methodology: C. van Zwol—Janssens, S. Lamballais, E. Loehrer, Y. Louwers; Project administration: C. van Zwol—Janssens, Y. Louwers; Supervision: T. Kleefstra, J. Laven, Y. Louwers; Raw data validation: C. van Zwol—Janssens, S. Lamballais; Writing—original draft: C. van Zwol—Janssens, S. Lamballais; Writing—review & editing: E. Loehrer, T. Kleefstra, J. Laven, Y. Louwers. All authors have read and approved the final version of the manuscript.

While in vitro fertilization (IVF) is a common treatment for many infertile couples, approximately 40% of IVF treatments do not result in a live birth. 1 A significant challenge lies in achieving an optimal ovarian response, defined as retrieving 8–14 follicles. 2 , 3 , 4 Suboptimal responses, including diminished ovarian response (DOR) where too few follicles develop, can lead to cycle cancellation and emotional distress. Conversely, an excessive response may lead to ovarian hyperstimulation syndrome (OHSS), this complication poses serious health risks. Both scenarios underscore the need for precise and personalised stimulation protocols. To mitigate these risks and enhance IVF outcomes, personalised stimulation protocols have emerged. Anti-Müllerian hormone (AMH) is a biomarker for ovarian reserve and has become a cornerstone in this approach, guiding individualised gonadotropin dosing. Studies like the PROFILE study, the ESTHER and GRAPE trial have shown that using AMH significantly improves safety compared to a “one-size-fits-all” approach, leading to similar stimulation outcomes and reduced complications. 5 , 6 , 7 Despite these advancements, 25.6%–56.7% of women still do not produce an optimal number of follicles. 5 , 6 , 7 This suggests that AMH alone may not fully capture the complexity of ovarian response, highlighting a clear need for improvement. Menopause, the permanent cessation of menstrual cycles, is marked by the depletion of ovarian follicles and loss of ovulation. While the average age is 51 years, there is significant individual variability in its timing. 8 This timing profoundly impacts a woman's reproductive lifespan and overall health, leading to increased attention on the underlying physiological mechanisms. The variation in the age at natural menopause (ANM) is hypothesised to stem from a complex interplay of environmental factors like stress, nutrition, and socioeconomic status, alongside hormonal aspects and strong genetic influences, with heritability estimates ranging from 31% to 87%. 9 , 10 , 11 , 12 Genome-wide association studies (GWAS) have identified around 290 genetic variants in 200,000 women of Northern European descent associated with ANM, revealing its polygenic nature and highlighting genes involved in genome stability (DNA repair and maintenance), immune function, and mitochondrial function. 13 The peri-menopausal stage is characterised by a gradual decline in fertility, culminating in natural subfertility about ten years before menopause. 14 This subfertility is due to a decrease in both oocyte quality and follicle numbers, leading to diminished ovarian reserve and reduced responsiveness to exogenous FSH. 15 , 16 , 17 , 18 Importantly, women who respond poorly to ovarian stimulation with gonadotropins are likely to experience earlier menopause, underscoring the connection between ovarian reserve, fertility treatment outcomes, and menopausal timing. 19 , 20 , 21 Critically, we only know a woman's ANM after she has already reached menopause, making it impossible to use as a predictive tool at the time of IVF stimulation. This limitation prevents its direct application in guiding real-time treatment decisions. Therefore, we hypothesised that the solution to this dilemma lies in genetics. We know that ANM is highly heritable and polygenic, which allows for the construction of a polygenic score (PGS) to predict timing of ANM. In this context, a lower PGS would predict earlier menopause along with decreased oocyte quality and follicle numbers, whereas a higher PGS indicates later menopause. Given that GWASs on ANM have identified genetic variants involved in pathways distinct from those directly reflected by AMH (such as DNA repair, immune, and mitochondrial function), it's highly probable that an ANM PGS captures different aspects of ovarian biology. 13 Therefore, a PGS for ANM has the potential to serve as a novel and complementary predictor to AMH, offering additional insights into a woman's ovarian responsiveness and further optimising IVF stimulation protocols.

Y.V. Louwers received an internal research grant from the Erasmus MC (The Synergy grant) and she received fees from Ferring and Merck for presentations. J.S.E. Laven has received grant support during the last 3 years from the following organisations (in alphabetic order); Ansh Labs, Ferring, Merck, National Institute of Health (NIH). He also received consultancy fees from the following companies Ansh Labs, Ferring, Gedeon Richter, Roche Diagnostics and Titus Health Care. C. van Zwol- Janssens, S. Lamballais, E. Loehrer and T. Kleefstra have no disclosures.