Modified letrozole vs GnRH antagonist protocols in ovarian aging women for IVF: an open-label, multicenter, randomized controlled trial.

We conducted a multicenter, RCT across six academic fertility centers in China. This was an open-label trial; therefore, participants and treating clinicians were not blinded to the treatment allocation. However, outcome assessors and the statistician remained blinded throughout the study to minimize assessment and analytical bias. All data provided to them was coded to conceal group identities. The trial is registered at www.chictr.org.cn (ChiCTR2000029272) (Original study protocol V1.0, 2019-12-20; Final study protocol V2.0, 2023-12-01; Supplementary material). Approval was granted by the ethics committees of Sun Yat-sen Memorial Hospital (2019-KY-067), Guangdong Provincial People’s Hospital (2020062H (R1)) and other participating hospitals including Affiliated Hospital of Guizhou Medical University (2022[269]), The Tenth Affiliated Hospital of Southern Medical University (DRYA2021-016-A1), Shenzhen Maternal and Child Health Hospital (SFYLS[2020]063) and Zhejiang Provincial People’s Hospital (2020-KY-001). An Independent Data Monitoring Committee (IDMC) supervised by Sun Yat-sen University’s Office of Scientific Research and Development monitored the study. The trial was funded by the 5010 clinical research grants from Sun Yat-sen University and inspected annually by IDMC (2021-2022). Participants were infertile women under 40 years of age with DOR with anti-mullerian hormone (AMH) < 1.2 ng/mL or antral follicle count (AFC) < 5, or aged 40–45 years, scheduled for less than four IVF or intracytoplasmic sperm injection (ICSI) cycles. Exclusion criteria were a history of ovarian surgery, significant male factor infertility (e.g., a low male spermatogenic function, non-obstructive azoospermia, or AZF gene microdeletion), unilateral or bilateral hydrosalpinx, uterine abnormalities (e.g., single horn uterus, uterine adenomyosis, intrauterine adhesions, untreated endometrial polyps, and submucous myoma), and abnormal parental karyotype. After written informed consent, participants were randomly assigned using a block randomization scheme with 1:1 allocation to the mLP and GnRH-ant group with predefined stratification for age (<40 years vs ≥40 years) and study center (six academic fertility centers). Within each stratum, a separate block randomization sequence was generated with a block size of 4 and an allocation ratio of 1:1, with block sequences concealed from study recruiters. The allocation sequence was implemented via a centralized computer-generated randomization system, developed and managed by an independent biostatistics team from the School of Public Health at Sun Yat-sen University. This system ensured strict allocation concealment, wherein treatment assignment was only revealed after participant enrollment. Recruitment occurred from March 2020 to February 2023, with follow-up completed in November 2023. Participants provided written informed consent prior to randomization. Standardized procedures were implemented per predefined standard operating procedures (SOPs) (Supplementary Table  1 ) to minimize interoperator variability. The mLP group initiated letrozole (5 mg/day) on menstrual day 2 for five days alongside gonadotropins (Gn) (300 IU/day from day 5); and the GnRH-ant group received Gn (300 IU/day from day 2). Both groups were added daily antagonist (0.25 mg) from day 6 until the trigger. The trigger timing of mLP is determined by follicular maturity (≥50% ratio of ≥18 mm to ≥14 mm follicles); and that of the GnRH-ant protocol is activated upon achieving ≥ 2 follicles ≥ 17 mm or 1 follicle ≥ 18 mm. Dual trigger 10,000 IU human chorionic gonadotropin (HCG) and 0.1 mg GnRH-agonist were given on the trigger day. All cycles underwent daily ultrasound monitoring using 3D automated follicle tracking (Voluson E8), with hormonal assessments (FSH, luteinizing hormone [LH], estradiol [E 2 ], progesterone [P], and total testosterone [TT] measured by chemiluminescent immunoassay) regularly. Oocyte pick-up (OPU) was conducted 34–36 h post-trigger, with embryo transfer mainly with one or two cleavage embryos, a few with blastocyst. Luteal-phase support was provided using vaginal progesterone gel at a dosage of 90 mg daily, or Utrogestan at 0.2 g administered three times daily, starting immediately post-OPU and continuing until the day serum HCG levels were measured. Subject to cancellation under certain conditions including the absence of available embryos, a thin endometrium (endometrial thickness < 6 mm), ongoing endometrial diseases, needle passage through the endometrium during OPU, abnormal vaginal bleeding, a high risk of ovarian hyperstimulation syndrome (OHSS) indicated by the retrieval of ≥20 oocytes, elevated progesterone levels (>2.0 ng/mL) on the trigger day, spontaneous pregnancy, fever (body temperature ≥ 37.5 °C), and personal reasons. The endometrial preparation methods for frozen-thawed embryo transfer (FET) included hormone replacement therapy (HRT) cycle, GnRH-a + HRT cycle, natural cycle (NC), and ovulation induction cycle. FET was scheduled with the same luteal-phase support. Two weeks after fresh ET or FET, serum hCG levels were assessed. For those with positive hCG (≥20 IU/mL on the 14th day) tests, luteal-phase support was extended through the 10th week of gestation. Detailed follow-ups were conducted for all participants until the completion of IVF treatment or determination of the pregnancy outcome. In case of conception, transvaginal ultrasonography was conducted two weeks later to verify clinical pregnancy, with a follow-up scan at 12 weeks to confirm the ongoing pregnancy. AEs were systematically recorded and defined as any undesirable medical occurrence observed in trial participants from the time of informed consent signing through the study completion. The primary outcomes were initially set as the CCPR of one started IVF cycle. Since live birth represents the gold-standard outcome in IVF, the CCPR (CLBR) was added as a co-primary outcome. The CCPR and CLBR, analyzed through two metrics: (1) per initiated COS cycle, defined as the proportion of women achieving ≥1 clinical pregnancy or live birth from fresh and subsequent FETs within a single COS cycle; and (2) per embryo-available COS cycle, calculated as the proportion of women attaining ≥1 clinical pregnancy or live birth following any embryo transfer (fresh or FET). Secondary outcomes included live birth rate (LBR, the proportion of women with viable delivery ≥28 gestational weeks per fresh transfer), clinical pregnancy rate (CPR, the proportion of women with intrauterine gestational sac confirmed by ultrasound 28–30 days post-transfer), and pregnancy loss rate stratified as pregnancy loss among biochemical pregnancy (the percentage of women with serum hCG ≥ 20 IU/mL at 14 days post-transfer) or miscarriage (with pregnancy loss ≤ 12 weeks post-confirmation in clinical pregnancy) (Definition of outcomes; Supplementary Table  2 ). For the initial sample size calculation, the expected CCPR in one COS was 44.44% in the mLP group and 27.70% in the GnRH-ant group. The superiority margin was set at 5%. A 1:1 allocation ratio between the two groups was used (statistical power 0.8; two-sided significance level 0.025), 261 cases were required per group. Anticipating 10% loss to follow-up or dropout, 580 participants were required. Due to the COVID-19 pandemic, recruitment was much slower than anticipated, and in February 2023, the trial was halted after recruitment of 318 participants. The post-hoc power analysis indicated that this sample size provided 81.1% power to detect the 16.7% difference that was previously determined to be clinically relevant if a superiority margin of 0% is used. The FAS was derived by including all subjects who underwent randomization and completed ≥1 post-randomization follow-up assessment. The PPS constituted a subset of the FAS comprising subjects who satisfied all inclusion/exclusion criteria, demonstrated protocol compliance without major deviations, completed the full treatment regimen, maintained adequate study adherence, and had fully documented case report forms (CRFs) without critical omissions in primary efficacy data. Primary analyses were conducted on the FAS, with per-protocol analyses serving as secondary evaluations. Both FAS and per-protocol approaches were applied to primary outcomes (CCPR and CLBR). Post-hoc subgroup analyses were performed exclusively using per-protocol data for hypothesis generation. For fresh embryo transfer outcomes (LBR, CPR, and pregnancy loss rate), per-protocol analyses (PPS) were conducted as exploratory subgroup assessments. Continuous variables were expressed as mean ± standard deviation and compared using Wilcoxon rank-sum tests. Categorical variables were summarized as counts (percentages) and analyzed with χ² tests or Fisher’s exact tests when expected cell counts fell below five. Treatment effects were reported as rate ratios (RRs) and risk differences (RDs) with 95% confidence intervals calculated via the Wald method. Multiple comparisons adjustment for FAS analyses employed the Benjamini–Hochberg procedure. In addition, a prespecified sensitivity analysis was performed using GLMM with random intercepts for clinical centers. This model was added to address a reviewer’s comment on accounting for center effects and clustering by study site. Between-center variability was quantified through intraclass correlation coefficients (ICCs) with 95% CIs derived from parametric bootstrapping (1000 replicates). Stratified analyses examining treatment effect modification by clinical site and age category were pre-specified. We also performed stratified analysis by DOR with age <40 years and age of 40–45 years, age (<35 years, 35–39 years, ≥40 years) based on AMH concentration (<1.2 ng/mL, ≥1.2 ng/mL), and the number of IVF cycles (none vs ≥1). And the corresponding stratified analyses were included to evaluate heterogeneity in protocol effects. The pre-specified Statistical Analysis Plan is provided in the supplementary material (Final Statistical Analysis Plan V2.0, dated 2023-12-05). During manuscript preparation and revision, the following changes to the statistical analyses were introduced: Benjamini–Hochberg procedure for multiplicity, GLMM to further estimate the treatment effect, and the ICC calculation to quantify between-center variability. Except for those analyses adjusted by the BH procedure, all other statistical tests were two-sided with α = 0.05. Analyses were performed using R version 4.4.2. Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

The clinical trial has been completed. Between March 2020 and February 2023, we screened a total of 437 women. Of these women, 318 who met the inclusion criteria were provided informed consent. Then they underwent randomization (with 159 women assigned to each group). The trial profile is shown in Fig.  1 . Follow-up for all pregnancies was finalized by November 2023. Among them, eight participants stopped halfway since preovulation or absent follicular development, and seven participants withdrew or deviated from the protocols. Throughout the study duration, no adverse events (AEs) were observed. 153 women in the mLP group and 150 women in the GnRH-ant group who finished the study were secondary analyzed by per protocol. Baseline characteristics between the two groups were comparable, as analyzed by the full analysis set (FAS) (Table  1 ). Ovarian stimulation outcomes and laboratory parameters for both groups are detailed in Table  2 analyzed by per-protocol set (PPS), and in Supplementary Table  3 by FAS, with the detailed data regarding the cancellation of oocyte retrieval or fresh embryo transfer presented in Supplementary Table  4 . The generalized linear mixed model (GLMM) yielded consistent results with the primary analysis and estimated treatment effects as odds ratios (ORs) with 95% CIs. Fig. 1 Flow chart of the trial enrollment. Between March 2020 and February 2023, we screened a total of 437 women. Of these women, 318 who met the inclusion criteria were provided informed consent. Then they underwent randomization (with 159 women assigned to the mLP group and the GnRH-ant group). Follow-up for all pregnancies was finalized by November 2023. Among them, 6 participants in the mLP group withdrew or deviated from the protocols, while 9 did so in the GnRH-ant group. One hundred fifty-three women in the mLP group and 150 women in the GnRH-ant group who finished the study were secondary analyzed by per protocol. mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol. Table 1 The baseline characteristics of the mLP and GnRH-ant group (FAS) Baseline characteristics mLP ( n  = 159) GnRH-ant ( n  = 159) Age, years a 37.0 (4.9) 37.9 (4.2) Body mass index, kg/m 2b 22.9 (3.2) 22.7 (2.8) AMH, ng/mL 0.9 (0.8) 0.9 (0.8) FSH, IU/L 10.3 (5.9) 10.5 (4.5) Luteinizing hormone, IU/L 4.7 (3.2) 4.4 (2.0) Estradiol, pg/mL 109.2 (109.4) 165.7 (384.3) Total testosterone, nmol/mL 1.1 (0.8) 1.2 (1.3) Prolactin, ng/mL 22.2 (43.0) 26.6 (65.7) Thyroid-stimulating hormone, mU/L 2.3 (1.3) 2.1 (1.5) Fasting glucose, mmol/L 5.1 (0.6) 5.1 (0.6) Fasting insulin, mU/L 15.6 (20.1) 13.3 (16.8) Antral follicle count in both ovaries 6.1 (4.0) 6.2 (3.4) Infertile duration, years 4.4 (3.9) 4.8 (4.4) Primary infertile, % c 33.3 (53/159) 26.4 (42/159) Age with 40–45 years, % 40.3 (64/159) 44.0 (70/159) DOR, % d 59.8 (95/159) 56.0 (89/159) Source data are provided as a Source Data file in the “Data availability” section. FAS full analysis set, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, AMH anti-mullerian hormone, DOR diminished ovarian reserve, IVF in vitro fertilization, ICSI intracytoplasmic sperm injection. a Values are means (standard deviations); difference tested using the Wilcoxon rank sum test. b The body-mass index is the weight in kilograms divided by the square of the height in meters. c Values are % (no./total no.); difference tested using Chi-square test. d DOR is defined as AMH < 1.2 ng/mL or AFC < 5 in women under 40 years old. Table 2 Cycle and IVF laboratory parameters of the mLP and GnRH-ant group (PPS) mLP ( n  = 153) GnRH-ant ( n  = 150) p value No. of days of Gonadotropin stimulation a 6.6 (2.8) 9.4 (2.1) <0.0001 Total gonadotropin dose, IU b 2006.0 (797.3) 2652.8 (657.7) <0.0001 No. of oocytes retrieved 4.0 (2.7) 4.7 (3.2) 0.054 No. of MII oocytes retrieved 2.4 (2.3) 3.1 (2.6) 0.028 Follicle output rate (FOR), % 83.1 (66.1) 82.6 (50.7) 0.936 The endometrial thickness on trigger day, mm 9.2 (2.5) 10.3 (3.0) 0.001 No. of available embryos 2.0 (1.4) 2.7 (1.9) <0.001 No. of high-quality embryos 1.0 (1.1) 1.2 (1.3) 0.076 Fresh embryo transfer rate, % c 77.7 (94/121) 71.7 (99/138) 0.341 No. of embryos transferred 1.3 (0.9) 1.5 (0.7) 0.014 Proportion of rest embryos, % ( n /total no.) 26.1 (70/268) 30.7 (121/394) 0.233 Endometrium 6–7 mm on trigger day, % 17.0 (26/153) 6.7 (10/150) 0.009 Endometrium <6 mm on trigger day, % 4.6 (7/153) 4.0 (6/150) 1.000 Source data are provided as a Source Data file in the “Data availability” section. PPS per-protocol analysis, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, MII meiosis II. a Values are means (standard deviations); difference tested using the Wilcoxon rank sum test. b Within each group of 159 participants, 6 in the mLP group and 9 in the GnRH-ant group discontinued before undergoing ovarian stimulation or deviated from the study protocols. A total of 153 patients in the mLP group and 150 female patients in the GnRH antagonist group who strictly adhered to the study protocol were analyzed by per-protocol. In the mLP group, 7 participants canceled their oocyte retrieval due to preovulation or lack of follicular development, whereas only 1 participant in the GnRH-ant group did so. c Values are % (no./total no.); difference tested using Chi-square test. In the mLP group, 121 out of 153 patients had available embryos, with 94 undergoing fresh embryo transfer. In the GnRH-ant group, out of 150 cases, 138 patients had available embryos, and 99 of these patients underwent fresh embryo transfer. Detailed data regarding the cancellation of oocyte retrieval or fresh embryo transfer is presented in Supplementary Table  4 . Between March 2020 and February 2023, we screened a total of 437 women. Of these women, 318 who met the inclusion criteria were provided informed consent. Then they underwent randomization (with 159 women assigned to the mLP group and the GnRH-ant group). Follow-up for all pregnancies was finalized by November 2023. Among them, 6 participants in the mLP group withdrew or deviated from the protocols, while 9 did so in the GnRH-ant group. One hundred fifty-three women in the mLP group and 150 women in the GnRH-ant group who finished the study were secondary analyzed by per protocol. mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol. The baseline characteristics of the mLP and GnRH-ant group (FAS) Source data are provided as a Source Data file in the “Data availability” section. FAS full analysis set, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, AMH anti-mullerian hormone, DOR diminished ovarian reserve, IVF in vitro fertilization, ICSI intracytoplasmic sperm injection. a Values are means (standard deviations); difference tested using the Wilcoxon rank sum test. b The body-mass index is the weight in kilograms divided by the square of the height in meters. c Values are % (no./total no.); difference tested using Chi-square test. d DOR is defined as AMH < 1.2 ng/mL or AFC < 5 in women under 40 years old. Cycle and IVF laboratory parameters of the mLP and GnRH-ant group (PPS) Source data are provided as a Source Data file in the “Data availability” section. PPS per-protocol analysis, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, MII meiosis II. a Values are means (standard deviations); difference tested using the Wilcoxon rank sum test. b Within each group of 159 participants, 6 in the mLP group and 9 in the GnRH-ant group discontinued before undergoing ovarian stimulation or deviated from the study protocols. A total of 153 patients in the mLP group and 150 female patients in the GnRH antagonist group who strictly adhered to the study protocol were analyzed by per-protocol. In the mLP group, 7 participants canceled their oocyte retrieval due to preovulation or lack of follicular development, whereas only 1 participant in the GnRH-ant group did so. c Values are % (no./total no.); difference tested using Chi-square test. In the mLP group, 121 out of 153 patients had available embryos, with 94 undergoing fresh embryo transfer. In the GnRH-ant group, out of 150 cases, 138 patients had available embryos, and 99 of these patients underwent fresh embryo transfer. Detailed data regarding the cancellation of oocyte retrieval or fresh embryo transfer is presented in Supplementary Table  4 . The mLP group exhibited lower endometrial thickness on hCG trigger day (9.2 ± 2.5 vs 10.3 ± 3.0 mm, p  = 0.001), fewer metaphase II (MII) oocytes retrieved (2.4 ± 2.3 vs 3.1 ± 2.6, p  = 0.028), and fewer available embryos (2.0 ± 1.4 vs 2.7 ± 1.9, p  < 0.001). The mLP group demonstrated significantly shorter ovarian stimulation duration (6.6 ± 2.8 vs 9.4 ± 2.1 days, p  < 0.0001) and less total Gn consumption (2006.0 ± 797.3 vs 2652.8 ± 657.7 IU, p  < 0.0001) compared to the GnRH-ant group, with statistically significant differences (Table  2 ). But no significant intergroup differences were observed in follicle output rate (FOR) or the number of high-quality embryos ( p  > 0.05). There is no significant difference regarding the fertilization method between the two groups (Supplementary Table  5 ). A formal protocol amendment was implemented during the trial. The original primary outcome was the cumulative clinical pregnancy rate (CCPR) in one COS. Analysis of the FAS ( n  = 159 per group) showed comparable CCPR between the mLP and GnRH-ant groups (32.1% vs 34.0%; RR 0.94, 95% CI: 0.69–1.29; p  = 0.283). This finding was consistent with the PPS analysis per COS cycle (33.3% vs 36.0%; RR 0.93, 95% CI: 0.68–1.27; p  = 0.275) and the CCPR per embryo-available cycle (45.5% vs 41.2%; RR 1.10, 95% CI: 0.82–1.47; p  = 0.794). The amended primary outcomes included both the CCPR and the Cumulative live birth rate (CLBR) in one COS. FAS analysis demonstrated no significant difference in CLBR between the mLP and GnRH-ant groups (24.5% vs 22.6%; RR 1.08, 95% CI: 0.73–1.61; p  = 0.991). Similarly, the PPS analysis (mLP: n  = 153; GnRH-ant: n  = 150) showed comparable results (25.5% vs 24.0%; RR 1.06, 95% CI: 0.72–1.57; p  = 0.991). For CLBR per embryo-available COS cycle, the mLP group also showed numerically elevated rates compared with the GnRH-ant group (34.8% vs 27.5%; RR 1.27, 95% CI: 0.87–1.85; p  = 0.508), though these differences lacked statistical significance (Table  3 ). Sensitivity analyses using age and center-adjusted ORs corroborated the primary findings (Supplementary Table  6 ). Sensitivity analyses adjusting for clinical center effects through GLMMs (ICC  0.20 for center interaction) (Supplementary Table  7 ). Table 3 Main outcomes of the mLP and GnRH-ant group (FAS and PPS) Analysis population Outcomes mLP ( n  = 159) GnRH-ant ( n  = 159) Absolute difference between groups (95% CI) RR for mLP vs GnRH-ant (95% CI) p value p adjusted * FAS CLBR, % 24.5 (39/159) 22.6 (36/159) 1.89 (−7.44 to 11.22) 1.08 (0.73–1.61) 0.991 0.991 CCPR, % 32.1 (51/159) 34.0 (54/159) −1.89 (−12.23 to 8.45) 0.94 (0.69–1.29) 0.283 0.566 PPS CLBR per COS cycle, % a 25.5 (39/153) 24.0 (36/150) 1.49 (−8.23 to 11.21) 1.06 (0.72–1.57) 0.991 0.991 CCPR per COS cycle, % b 33.3 (51/153) 36.0 (54/150) −2.67 (−13.38 to 8.04) 0.93 (0.68–1.27) 0.275 0.991 CLBR per embryo-available COS cycle, % c 34.8 (39/112) 27.5 (36/131) 7.34 (−4.33 to 19.01) 1.27 (0.87–1.85) 0.508 0.991 CCPR per embryo-available COS cycle, % d 45.5 (51/112) 41.2 (54/131) 4.31 (−8.18 to 16.80) 1.10 (0.82– 1.47) 0.794 0.991 Live birth rate in fresh ET, % e 34.0 (32/94) 22.2 (22/99) 11.82 (−0.78 to 24.42) 1.53 (0.96–2.43) 0.164 0.488 Clinical pregnancy rate in fresh ET, % f 42.6 (40/94) 30.3 (30/99) 12.25 (−1.24 to 25.74) 1.40 (0.96–2.05) 0.753 0.753 Pregnancy loss rate in fresh ET, % Among biochemical pregnancy g 22.0 (9/41) 34.3 (13/35) −15.19 (−35.61 to 5.23) 0.59 (0.29–1.21) 0.244 0.488 Among clinical pregnancy h 20.0 (8/40) 26.7 (8/30) −6.67 (−26.77 to 13.43) 0.75 (0.32–1.77) 0.665 0.753 Source data are provided as a Source Data file in the “Data availability” section. FAS full analysis set, PPS per-protocol analysis, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, COS controlled ovarian stimulation, ET embryos transfer. Values are % (no./total no.); difference tested using the Chi-square test. The 95% CIs of the absolute difference and the RR are two-sided. a CLBR per COS cycle: Defined as the proportion of women achieving ≥ 1 live birth from at least one embryo transfer (fresh + subsequent frozen-thawed) originating from a single COS cycle.Calculated as: (number of women with ≥1 live birth from index COS-derived embryos)/(total initiated COS cycles) × 100%. b CCPR per COS cycle: defined as the proportion of women with ≥1 clinical pregnancy (gestational sac confirmed by ultrasound) from at least one embryo transfer (fresh + subsequent frozen-thawed) originating from a single COS cycle. Calculated as: (number of women with ≥1 clinical pregnancy from index COS-derived embryos)/(total initiated COS cycles) × 100%. c CLBR per embryo-available COS cycle: defined as the proportion of women achieving ≥1 live birth after at least one embryo transfer derived from a single COS cycle. Calculated as: (number of women with ≥1 live birth from index COS-derived embryos)/(total number of women with available embryos) × 100%. d CCPR per embryo-available COS cycle: defined as the proportion of women achieving ≥1 clinical pregnancy after at least one embryo transfer derived from a single COS cycle, limited to cycles yielding ≥1 transferable embryo. Calculated as: (number of women with ≥1 clinical pregnancy from index COS-derived embryos)/(total number of women with available embryos) × 100%. In the mLP group, out of 153 ovarian stimulation cycles, only 112 patients underwent at least one fresh or frozen embryo transfer (including 94 patients who underwent one fresh transfer and 28 patients who underwent one or two frozen transfers). In the GnRH-ant group, out of 150 ovarian stimulation cycles, only 131 cases underwent at least one fresh or frozen embryo transfer (including 99 patients who underwent one fresh transfer and 32 patients who underwent one or two frozen transfers). e Live birth was defined as a viable delivery ≥28 gestational weeks after embryo transfer. Live birth rate in fresh ET was defined as the number of women with delivery of at least one live birth per fresh embryo transfer. f Clinical pregnancy was defined as the presence of intrauterine gestation sacs at 28-30 days after embryo transfer. Clinical pregnancy rate in fresh ET was defined as the percentage of women achieving a clinical pregnancy after fresh embryo transfers. g Pregnancy loss rate in fresh embryo transfer (ET) among biochemical pregnancies was defined as the percentage of pregnancy loss in biochemical pregnancies, which are characterized by a serum level of human chorionic gonadotropin (hCG) ≥ 20 IU/mL on day 14 following the fresh embryo transfer. h Pregnancy loss rate in fresh ET among clinical pregnancies was defined as the percentage of miscarriage before the 12th week of pregnancy after fresh embryo transfer. * Utilizing the Benjamini–Hochberg procedure for multiple testing correction. The BH procedure was applied separately to each pre-specified family of tests (e.g., all primary endpoints in FAS analysis, all primary endpoints in PPS analysis, and endpoints in subgroup analysis). Main outcomes of the mLP and GnRH-ant group (FAS and PPS) Source data are provided as a Source Data file in the “Data availability” section. FAS full analysis set, PPS per-protocol analysis, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, COS controlled ovarian stimulation, ET embryos transfer. Values are % (no./total no.); difference tested using the Chi-square test. The 95% CIs of the absolute difference and the RR are two-sided. a CLBR per COS cycle: Defined as the proportion of women achieving ≥ 1 live birth from at least one embryo transfer (fresh + subsequent frozen-thawed) originating from a single COS cycle.Calculated as: (number of women with ≥1 live birth from index COS-derived embryos)/(total initiated COS cycles) × 100%. b CCPR per COS cycle: defined as the proportion of women with ≥1 clinical pregnancy (gestational sac confirmed by ultrasound) from at least one embryo transfer (fresh + subsequent frozen-thawed) originating from a single COS cycle. Calculated as: (number of women with ≥1 clinical pregnancy from index COS-derived embryos)/(total initiated COS cycles) × 100%. c CLBR per embryo-available COS cycle: defined as the proportion of women achieving ≥1 live birth after at least one embryo transfer derived from a single COS cycle. Calculated as: (number of women with ≥1 live birth from index COS-derived embryos)/(total number of women with available embryos) × 100%. d CCPR per embryo-available COS cycle: defined as the proportion of women achieving ≥1 clinical pregnancy after at least one embryo transfer derived from a single COS cycle, limited to cycles yielding ≥1 transferable embryo. Calculated as: (number of women with ≥1 clinical pregnancy from index COS-derived embryos)/(total number of women with available embryos) × 100%. In the mLP group, out of 153 ovarian stimulation cycles, only 112 patients underwent at least one fresh or frozen embryo transfer (including 94 patients who underwent one fresh transfer and 28 patients who underwent one or two frozen transfers). In the GnRH-ant group, out of 150 ovarian stimulation cycles, only 131 cases underwent at least one fresh or frozen embryo transfer (including 99 patients who underwent one fresh transfer and 32 patients who underwent one or two frozen transfers). e Live birth was defined as a viable delivery ≥28 gestational weeks after embryo transfer. Live birth rate in fresh ET was defined as the number of women with delivery of at least one live birth per fresh embryo transfer. f Clinical pregnancy was defined as the presence of intrauterine gestation sacs at 28-30 days after embryo transfer. Clinical pregnancy rate in fresh ET was defined as the percentage of women achieving a clinical pregnancy after fresh embryo transfers. g Pregnancy loss rate in fresh embryo transfer (ET) among biochemical pregnancies was defined as the percentage of pregnancy loss in biochemical pregnancies, which are characterized by a serum level of human chorionic gonadotropin (hCG) ≥ 20 IU/mL on day 14 following the fresh embryo transfer. h Pregnancy loss rate in fresh ET among clinical pregnancies was defined as the percentage of miscarriage before the 12th week of pregnancy after fresh embryo transfer. * Utilizing the Benjamini–Hochberg procedure for multiple testing correction. The BH procedure was applied separately to each pre-specified family of tests (e.g., all primary endpoints in FAS analysis, all primary endpoints in PPS analysis, and endpoints in subgroup analysis). In fresh embryo transfer per-protocol analyses, secondary outcomes showed no statistically significant differences: LBR (34.0% vs 22.2%; RR 1.53, 95% CI: 0.96–2.43, p  = 0.164), CPR (42.6% vs 30.3%; RR 1.40, 95% CI: 0.96–2.05, p  = 0.753), biochemical pregnancy loss rates (22.0% vs 34.3%; RR 0.59, 95% CI: 0.29–1.21, p  = 0.244) and miscarriage rate (20.0% vs 26.7%; RR 0.75, 95% CI: 0.32–1.77); which remained non-significant after multiplicity adjustment ( p adjusted > 0.05) (Table  3 ). Both groups maintained comparable embryo transfer profiles, with cleavage-stage transfers predominating (mLP: 97.9% [92/94] plus 2.1% D5 blastocysts vs GnRH-ant: 99.0% [98/99] plus 1.0% D5 blastocysts, p  = 0.615) and similar single/double cleavage embryo utilization rates (single: 23.4% vs 25.3%, p  = 0.873; double: 72.3% vs 73.7%, p  = 0.892). Table 4 Outcomes of the mLP and GnRH-ant group in fresh embryo transfer stratified by age (PPS) Analysis subgroup Outcomes b mLP ( n  = 153) GnRH-ant ( n  = 150) Absolute difference between groups (95% CI) RR for mLP vs GnRH-ant (95% CI) p value p adjusted * DOR a Live birth rate in fresh ET, % c 45.1 (23/51) 29.1 (16/55) 16.01 (−2.17 to 34.19) 1.55 (0.93–2.59) 0.132 0.440 Clinical pregnancy rate in fresh ET, % d,e 54.9 (28/51) 34.5 (19/55) 20.36 (1.80–38.92) 1.59 (1.02–2.47) 0.056 0.280 1 cleavage embryo transferred 25.0 (3/12) 27.3 (3/11) −2.27 (−38.23 to 33.69) 0.92 (0.23–3.64) 1.000 1.000 2 cleavage embryos transferred 65.8 (25/38) 36.4 (16/44) 29.43 (8.70–50.16) 1.81 (1.15–2.85) 0.015 0.150 Pregnancy loss rate among clinical pregnancies in fresh ET, % f 17.9 (5/28) 15.8 (3/19) 2.07 (−19.61 to 23.75) 1.13 (0.31–4.18) 1.000 1.000 Age 40–45 years Live birth rate in fresh ET, % 20.9 (9/43) 13.6 (6/44) 7.29 (−8.54 to 23.12) 1.53 (0.60–3.93) 0.408 0.816 Clinical pregnancy rate in fresh ET, % g 27.9 (12/43) 25.0 (11/44) 2.91 (−15.63 to 21.44) 1.12 (0.55–2.25) 0.949 1.000 1 cleavage embryo transferred 10.0 (1/10) 21.4 (3/14) −11.43 (−39.85 to 16.99) 0.47 (0.06−3.86) 0.615 0.914 2 cleavage embryos transferred 36.7 (11/30) 27.6 (8/29) 9.08 (−14.63 to 32.79) 1.33 (0.63–2.83) 0.640 0.914 Pregnancy loss rate among clinical pregnancy in fresh ET-% 25.0 (3/12) 45.5 (5/11) −20.45 (−58.74 to 17.84) 0.55 (0.17–1.78) 0.400 0.816 Source data are provided as a Source Data file in the “Data availability” section. PPS per-protocol analysis, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, DOR diminished ovarian reserve, ET embryo transfer. P interaction  > 0.05, there is no interaction effect between age and embryo types. * Utilizing the Benjamini–Hochberg procedure for multiple testing correction. The BH procedure was applied separately to each pre-specified family of tests (e.g., all primary endpoints in FAS analysis, all primary endpoints in PPS analysis, and endpoints in subgroup analysis). a DOR is defined as AMH < 1.2 ng/mL or AFC < 5 in women under 40 years old. b Values are % (no./total no.); difference tested using Chi-square test. The 95% CIs of the absolute difference and the RR are two-sided. c Live birth was defined as a viable delivery ≥ 28 gestational weeks after embryo transfer. Live birth rate in fresh ET was defined as the number of women with delivery of at least one live birth per fresh embryo transfer. d Clinical pregnancy was defined as the presence of intrauterine gestation sacs at 28–30 days after embryo transfer. Clinical pregnancy rate in fresh ET was defined as the percentage of women achieving a clinical pregnancy after fresh embryo transfers. e In the DOR subgroup, the mLP group had a single case of double blastocyst transfer that failed to result in pregnancy, while no instances of double blastocyst transfers or higher-order embryo transfers were recorded in the GnRH-ant subgroup. f Pregnancy loss rate in fresh ET among clinical pregnancy was defined as the percentage of miscarriages considering all clinical pregnancies from fresh embryo transfer. g Within the 40–45 years age subgroup, the mLP group included one case of single blastocyst transfer and two cases of triple cleavage-stage embryo transfers, while the GnRH-ant group contained a single case of double blastocyst transfer. None of the blastocyst transfers in either subgroup resulted in a clinical pregnancy. In the mLP group, 97.9% (92/94) of transfers were cleavage-stage embryos and 2.1% (2/94) were D5 blastocysts; in the GnRH-ant group, these proportions were 99.0% (98/99) and 1.0% (1/99), respectively. For cleavage-stage transfers, 23.4% (22/94) in the mLP group and 25.3% (25/99) in the GnRH-ant group involved single embryos ( p  = 0.873), while 72.3% (68/94) and 73.7% (73/99) involved double embryos, respectively ( p  = 0.892). Twin pregnancy rates after double transfers showed no significant between-group differences (8.0% vs 12.5%, RR 0.64, 95% CI: 0.14–2.88, p  = 0.615). Outcomes of the mLP and GnRH-ant group in fresh embryo transfer stratified by age (PPS) Source data are provided as a Source Data file in the “Data availability” section. PPS per-protocol analysis, mLP modified letrozole protocol, GnRH-ant gonadotropin-releasing hormone antagonist protocol, DOR diminished ovarian reserve, ET embryo transfer. P interaction  > 0.05, there is no interaction effect between age and embryo types. * Utilizing the Benjamini–Hochberg procedure for multiple testing correction. The BH procedure was applied separately to each pre-specified family of tests (e.g., all primary endpoints in FAS analysis, all primary endpoints in PPS analysis, and endpoints in subgroup analysis). a DOR is defined as AMH < 1.2 ng/mL or AFC < 5 in women under 40 years old. b Values are % (no./total no.); difference tested using Chi-square test. The 95% CIs of the absolute difference and the RR are two-sided. c Live birth was defined as a viable delivery ≥ 28 gestational weeks after embryo transfer. Live birth rate in fresh ET was defined as the number of women with delivery of at least one live birth per fresh embryo transfer. d Clinical pregnancy was defined as the presence of intrauterine gestation sacs at 28–30 days after embryo transfer. Clinical pregnancy rate in fresh ET was defined as the percentage of women achieving a clinical pregnancy after fresh embryo transfers. e In the DOR subgroup, the mLP group had a single case of double blastocyst transfer that failed to result in pregnancy, while no instances of double blastocyst transfers or higher-order embryo transfers were recorded in the GnRH-ant subgroup. f Pregnancy loss rate in fresh ET among clinical pregnancy was defined as the percentage of miscarriages considering all clinical pregnancies from fresh embryo transfer. g Within the 40–45 years age subgroup, the mLP group included one case of single blastocyst transfer and two cases of triple cleavage-stage embryo transfers, while the GnRH-ant group contained a single case of double blastocyst transfer. None of the blastocyst transfers in either subgroup resulted in a clinical pregnancy. In the mLP group, 97.9% (92/94) of transfers were cleavage-stage embryos and 2.1% (2/94) were D5 blastocysts; in the GnRH-ant group, these proportions were 99.0% (98/99) and 1.0% (1/99), respectively. For cleavage-stage transfers, 23.4% (22/94) in the mLP group and 25.3% (25/99) in the GnRH-ant group involved single embryos ( p  = 0.873), while 72.3% (68/94) and 73.7% (73/99) involved double embryos, respectively ( p  = 0.892). Twin pregnancy rates after double transfers showed no significant between-group differences (8.0% vs 12.5%, RR 0.64, 95% CI: 0.14–2.88, p  = 0.615). The prespecified subgroup analyses indicated numerical, though not statistically significant, improvements in LBR with the mLP among women with DOR under 40 years (45.1% vs 29.1%; RR 1.55, 95% CI: 0.93–2.59; p  = 0.132) and those aged 40–45 years (20.9% vs 13.6%; RR 1.53, 95% CI: 0.60–3.93; p  = 0.408). Notably, DOR patients receiving two cleavage-stage embryos by using mLP achieved significantly higher CPR compared to GnRH-ant (65.8% vs 36.4%; RR 1.81, 95% CI: 1.15–2.85; p  = 0.015) (Table  4 ), though blastocyst transfers were infrequent (mLP: 2/94; GnRH-ant: 1/99) with no successful pregnancies. Safety analyses showed numerically lower multiple pregnancy rates with mLP when transferring two cleavage-stage embryos (8.0% vs 12.5%; RR 0.64, 95% CI: 0.14–2.88; p  = 0.615), with no significant interaction effect among the age and embryo types detected ( p interaction  > 0.05) (Table  4 ). Stratified analyses by serum AMH thresholds (<1.2 vs ≥1.2 ng/mL) and maternal age deciles revealed no statistically significant intergroup disparities in CPR, LBR, or CLBR (Fig.  2A–C ). However, all ART outcome metrics exhibited a non-significant numerical decline paralleling advancing age, with CLBR decreasing from 28.4% (age < 35 years) to 14.7% (age ≥ 40 years), independent of ovarian reserve status. In addition, the impact of the mLP among the women with repeated IVF cycles was also evaluated, presented in Supplementary Table  8 . For those undergoing more than one cycle, the mLP group demonstrated a non-significantly higher LBR (41.4% vs 16.7%), for a RR of 2.48 (95% CI: 1.0–6.17; p  = 0.071), and a significantly higher CPR (51.7% vs 23.3%), for a RR of 2.22 (95% CI: 1.06–4.64; p  = 0.047). Regarding those women with the endometrial thickness of 6-7 mm, there were 9 and 4 cases in the mLP and GnRH-ant groups, respectively. There is no live birth among them (Supplementary Table  9 ). Fig. 2 The pregnancy outcomes in a fresh IVF cycle and the dynamic changes of hormone levels with follicular growth during COS in mLP and GnRH-ant groups. Total n  = 303, including mLP Group ( n  = 153) and GnRH-ant Group ( n  = 150). Patients were divided into 4 subgroups based on age and AMH concentrations (Age < 35& AMH < 1.2 ng/mL [ n  = 82], Age 35–39& AMH < 1.2 ng/mL Goup [ n  = 86], Age ≥ 40& AMH < 1.2 ng/mL Goup [ n  = 81], Age ≥ 40& AMH ≥ 1.2 ng/mL Goup [ n  = 46]). From A – C , values are %; differe n ce tested using the Chi-square test. The dashed line shows the linear trend of pregnancy outcomes among subgroups. A Clinical pregnancy rate (CPR). B Live birth rate (LBR). C Cumulative live birth rate (CLBR). From ( D – F ), values are means (standard deviations); difference tested using the Wilcoxon rank sum test. Error bars represent standard deviation. D Dynamic changes in serum TT levels with follicle development during COS. Data points show individual hormone levels. E Dynamic changes in serum E 2 levels with follicle development during COS. Data points show individual hormone levels. F Dynamic changes in serum LH levels with follicle development during COS. Data points show individual hormone levels. G The serum TT levels per follicle with a diameter ≥ 18 mm on trigger day ( p  = 0.0017). Violin plots depict data distributions, with center lines at medians and the dotted lines at 25th and 75th percentiles. H The serum E 2 levels per follicle with a diameter ≥ 18 mm on trigger day ( p  < 0.001). Violin plots depict data distributions, with center lines at medians and the dotted lines at 25th and 75th percentiles. FSH follicle-stimulating hormone, LH luteinizing hormone, E 2 estradiol, TT total testosterone. A – H p values indicated by asterisks ( * p  < 0.05; ** p  < 0.01; *** p  < 0.001; **** p  < 0.0001). Source data are provided as a Source Data file in the “Data availability” section. Statistical significance was calculated using a two-sided significance level of 0.025. Total n  = 303, including mLP Group ( n  = 153) and GnRH-ant Group ( n  = 150). Patients were divided into 4 subgroups based on age and AMH concentrations (Age < 35& AMH < 1.2 ng/mL [ n  = 82], Age 35–39& AMH < 1.2 ng/mL Goup [ n  = 86], Age ≥ 40& AMH < 1.2 ng/mL Goup [ n  = 81], Age ≥ 40& AMH ≥ 1.2 ng/mL Goup [ n  = 46]). From A – C , values are %; differe n ce tested using the Chi-square test. The dashed line shows the linear trend of pregnancy outcomes among subgroups. A Clinical pregnancy rate (CPR). B Live birth rate (LBR). C Cumulative live birth rate (CLBR). From ( D – F ), values are means (standard deviations); difference tested using the Wilcoxon rank sum test. Error bars represent standard deviation. D Dynamic changes in serum TT levels with follicle development during COS. Data points show individual hormone levels. E Dynamic changes in serum E 2 levels with follicle development during COS. Data points show individual hormone levels. F Dynamic changes in serum LH levels with follicle development during COS. Data points show individual hormone levels. G The serum TT levels per follicle with a diameter ≥ 18 mm on trigger day ( p  = 0.0017). Violin plots depict data distributions, with center lines at medians and the dotted lines at 25th and 75th percentiles. H The serum E 2 levels per follicle with a diameter ≥ 18 mm on trigger day ( p  < 0.001). Violin plots depict data distributions, with center lines at medians and the dotted lines at 25th and 75th percentiles. FSH follicle-stimulating hormone, LH luteinizing hormone, E 2 estradiol, TT total testosterone. A – H p values indicated by asterisks ( * p  < 0.05; ** p  < 0.01; *** p  < 0.001; **** p  < 0.0001). Source data are provided as a Source Data file in the “Data availability” section. Statistical significance was calculated using a two-sided significance level of 0.025. Hormonal profiling revealed distinct pharmacodynamic responses between protocols (Fig.  2D-H ). The mLP group exhibited sustained TT elevation during folliculogenesis (1.3 → 1.8 nmol/L vs GnRH-ant 1.0 → 1.3 nmol/L; p  < 0.05), with per-follicle TT concentrations at terminal maturation (≥16–18 mm) tripling controls (1.5 vs <1.0 nmol/L; p  < 0.001). This androgen surge paralleled mid-cycle LH spikes at critical follicular milestones (11–12 mm and ≥18 mm diameters), while concurrently suppressing E 2 production (trigger-day: 516.2 vs 1428.7 pg/mL; per-follicle: 200–300 vs >600 pg/mL; p  < 0.0001). FSH trajectories remained comparable between groups despite these divergent steroidogenic patterns.

This multi-center RCT provides the first prospective validation of the mLP for ovarian stimulation of reproductive aging women, extending prior observational evidence through rigorous methodological design. While FAS analyses showed no statistically significant difference between-group differences in CLBR (24.5% vs 22.6%; RR 1.08, 95% CI: 0.73–1.61), both mLP and GnRH-ant protocol achieved unprecedented outcomes exceeding historical benchmarks (<15% per cycle) in this challenging cohort 16 . These results were achieved through implementation of two synergistic strategies: preferential utilization of fresh day-3 cleavage-stage embryo transfers (97.9% and 99.0% of cycles, respectively) to bypass resource-intensive blastocyst culture in patients with limited available embryos (2.0 ± 1.4 vs 2.7 ± 1.9 embryos/cycle), and strict adherence to SOP across six centers, as evidenced by minimal between-center variability (ICC < 0.1). Prespecified subgroup analyses suggested a potentially favorable outcome with mLP in dual cleavage embryo transfers for women aged less than 40 years with DOR (CPR 65.8% vs 36.4%; RR 1.81, 95% CI: 1.15–2.85, p  = 0.015), aligning with the clinical imperative to prioritize expedited fresh transfers over extended interventions in embryo-scarce populations. By harmonizing rapid cycle efficiency with protocol standardization, this framework demonstrates that operational precision can recalibrate success thresholds in ovarian aging management, offering a scalable model to reduce time-to-pregnancy while mitigating socioeconomic burdens. This study cohort consists of women with DOR, including 40.0% who are over 40-45 years old and have an urgent need for pregnancy after more than 4 years of infertility. The results of the current RCT indicated that the mLP demonstrated comparative CLBR with GnRH-ant protocol (25.5% vs 24.0% per COS cycle and 34.8% vs 27.5% per embryo transfer evaluations); and the observed non-significant numerical improvement in CPR (42.6%) and LBR (34.0%) in fresh embryo transfer by mLP with limited embryo availability (2.0 ± 1.4) holds clinical relevance, particularly given the population’s severely compromised ovarian reserve. These findings align with our prior retrospective data and surpass outcomes from other letrozole-based protocols, notably Shapira et al.’s study 15 , 17 , which reported lower CPR (31.5%) without LBR using divergent stimulation timing (initiation on cycle day 2-3 vs mLP’s day 2 start). This discrepancy underscores the criticality of protocol standardization, as our SOP-mandated later Gn initiation (day 5) and adjusted trigger thresholds (the ratio of number of follicles ≥ 18 mm to number of follicles ≥14 mm ≥1/2) likely optimized follicular maturation. Our findings challenge the prevailing reliance on blastocyst culture and preimplantation genetic testing (PGT-A) in advanced-age/DOR populations. The inherent inefficiency of these approaches is magnified in ovarian aging contexts, where women >40 years require ≥12.2 mature oocytes per euploid blastocyst 18 , inevitably prolonging treatment through repeated stimulations. The mLP’s superior clinical pregnancy rate with fresh dual cleavage-stage transfers (65.8% vs 36.4%; RR 1.81, p  = 0.015) validates the clinical advantage of prioritizing immediate embryo utilization over extended culture or genetic screening in embryo-scarce scenarios. These findings advocate for the clinical utility of this treatment paradigm, which needs to be widely spread first and prioritize patient-centered outcomes by shortening time-to-pregnancy and containing financial burdens. The mechanism of the novel mLP may be due to its therapeutic effect through orchestrated endocrine modulation, simultaneously enhancing follicular competence while optimizing endometrial receptivity 19 – 22 . Emerging evidence delineates a critical therapeutic window for androgen receptor (AR)-mediated follicular maturation, wherein LE administration in our RCT precisely augmented serum testosterone (TT: 1.3 → 1.8 nmol/L, p  < 0.001) during the follicular growth phase. This pharmacodynamic profile paralleled mid-cycle LH surges (peak at 11–12 mm follicular diameter) and sustained elevation through terminal folliculogenesis (≥18 mm), demonstrating LE’s unique capacity to recapitulate physiological steroidogenic rhythms. Mechanistically, we posit that LE-induced HPO axis resensitization operates through: (i) AR-dependent potentiation of FSH receptor trafficking via PI 3 K/AKT phosphorylation cascades; (ii) dynamic LH-TT crosstalk stabilizing androgen/estrogen equilibrium during follicular selection; (iii) estrogenic attenuation maintaining physiological E 2 levels (trigger-day: 516.2 vs 1428.7 pg/mL in controls), preventing premature endometrial maturation. In our team’s new basic science research, single-cell transcriptomic profiling of peripheral blood mononuclear cells (PBMCs) on the day of fresh embryo transfer revealed that the mLP upregulates folate metabolism (DHFR) and AP-1 signaling (FOS/JUN), inducing glycolytic reprogramming to polarize Treg/NK cell function and suppress neutrophil hyperactivity (unpublished data). This immunometabolic remodeling fosters a dual-optimized endometrial-immune microenvironment, potentially explaining its superior fresh embryo transfer pregnancy rates compared to GnRH-ant protocols (with a CPR of 60.0% vs 27.3% in this new study, unpublished data). FET strategies face dual challenges in ovarian aging populations: inherent cryodamage risks counterbalance theoretical endometrial synchronization benefits 23 , 24 . In this trial, FET outcomes showed no significant inter-protocol differences (mLP vs GnRH-ant: LBR 34.8% vs 27.5%, p  = 0.508), aligning with evidence that freeze-all approaches may paradoxically reduce success rates in poor responders 25 , 26 . While FET theoretically optimizes embryo-endometrial dialog, its clinical implementation requires cost-utility analyses given prolonged treatment timelines and cumulative financial burdens. This equipoise underscores the need for individualized decision-making that weighs cryopreservation risks against socioeconomic constraints. Protocol selection should therefore prioritize holistic parameters-embryo quality metrics, endometrial receptivity biomarkers, and patient-centered logistics-over singular outcome comparisons. This study had some limitations. Firstly, the comparable outcomes between initial and repeat IVF cycles (ΔLBR < 5%) suggest mLP’s efficacy may be cycle-number agnostic, though dedicated trials should clarify its role in recurrent implantation failure populations. Secondly, while the COVID-19 pandemic necessitated protocol adaptations (e.g., intermittent enrollment pauses), even with extended recruitment, we could not achieve the sample size of 580 participants. Therefore, the study’s power remains limited for detecting smaller but clinically meaningful effects, particularly in subgroup analyses. Larger-scale multicenter clinical trials are deserved in the future to further validate the findings of this research. In the future, the economic benefit ratio pending further investigation and more comprehensive basic researches are needed to dig into the underlying causes of DOR and validate the current hypothesis. This multicenter RCT demonstrated comparable CLBR and CCPR between the mLP and GnRH-ant groups, establishing mLP as an effective standardized strategy for ovarian aging populations. Although the mLP has not displayed superiority in LBR and CLBR, by prioritizing fresh day-3 dual embryo transfers over resource-intensive blastocyst/PGT-A approaches, mLP achieved superior CPR in women < 40 years with diminished ovarian reserve while halving time-to-pregnancy. Rigorous protocol standardization ensured cross-center reproducibility. These findings redefine clinical priorities, demonstrating that streamlined fresh embryo utilization aligns with optimal outcomes in ovarian aging. Multi-ethnic trials should validate mLP’s generalization and optimize protocols for recurrent IVF failures.

The progressive decline in oocyte quality and quantity associated with diminished ovarian reserve (DOR) poses significant challenges in assisted reproduction 1 , 2 , affecting 25–50% of couples undergoing in vitro fertilization-embryo transfer (IVF-ET) 3 . Women under 40 years with DOR experience 10–15% reductions in clinical pregnancy rates (CPR) and 5–17% live birth rates (LBR) per cycle compared to those with normal ovarian reserve 3 – 8 . This challenge intensifies in women aged 40–45 years, resulting in LBRs below 10% in many cohorts 9 . While diverse controlled ovarian stimulation (COS) protocols have been explored, no consensus exists due to inconsistent efficacy reports, particularly for advanced maternal age populations 10 – 13 . The lack of evidence-based protocols tailored to women with DOR or advanced age, with 40–45 years—a demographic experiencing both accelerated ovarian aging and time-sensitive fertility decline-underscores the critical need for optimized COS strategies that balance ovarian response efficiency with minimized treatment burden. Letrozole, a third-generation aromatase inhibitor, enhances antral follicle recruitment through androgen priming, elevating intraovarian androstenedione and testosterone levels, thereby increasing follicle-stimulating hormone (FSH) sensitivity. This pharmacodynamic profile positions letrozole as a promising agent for optimizing COS in DOR populations. Our prior clinical investigations demonstrated that follicular-phase serum testosterone surges correlate with 2.3-fold higher CPR 14 . And our retrospective cohort study revealed that the modified letrozole protocol (mLP), which strategically combines letrozole administration with gonadotropin supplementation, significantly improved both CPR (42.1% vs 28.3%, p  = 0.023) and LBR (31.6% vs 18.9%, p  = 0.038) in women ≥ 40 years or <40 years with DOR 15 . These results suggest that mLP’s dual mechanism, amplifying androgen-mediated follicular recruitment while maintaining physiological estrogen synthesis, may address the dual challenges of poor ovarian response and endometrial receptivity in DOR patients. To establish robust evidence for this innovative approach, we conducted the first multicenter randomized controlled trial (RCT) comparing mLP with conventional gonadotropin-releasing hormone antagonist (GnRH-ant) protocols in infertile patients <40 years with DOR, or advanced age, 40–45 years. This study not only evaluates mLP’s efficacy among these populations but also pioneers the development of a standardized, patient-tailored COS protocol for this clinically challenging cohort.

Source data Source data