{"paper_id":"2659fcb7-a7ea-4344-a6fd-60c56bd73a07","body_text":"Controlled ovarian stimulation (COS) is a fundamental component of assisted reproductive technologies (ART) ( 1 ). It maximizes the potential for successful outcomes by stimulating multiple follicles to mature simultaneously, thereby increasing the availability of viable embryos for transfer. However, multiple follicle growth also raises concerns regarding the excessive ovarian response. During the process, follicular development and serum E 2  are closely monitored to justify the decision to trigger oocyte maturation and in the prevention of ovarian hyperstimulation syndrome (OHSS) ( 2 ,  3 ).\nThe importance of E 2  in follicular development in both natural and stimulated cycles has been well-established, as it is secreted by the granulosa cells (GC) in response to endogenous or exogenous follicle-stimulating hormone (FSH) stimulation to support the follicular growth and maturation ( 4 ). During monitoring of stimulated cycles, serum levels of E 2  are expected to increase steadily with follicle growth until final maturation ( 5 ). However, unexpected E 2  decline during COS monitoring is also reported in monitored COS cycles, yet its clinical significance remains controversial. An early study suggested that a decline in serum E 2  before triggering is associated with a dramatically decreased pregnancy rate ( 6 ). However, milder effects or no effect of E 2  decline during COS monitoring are also reported in other studies ( 7 ). The conflicting conclusions surrounding the clinical significance of E 2  decline during COS can be attributed to various factors, such as limited sample size, heterogeneity of patients, or failure to adjust for important confounders. Importantly, previous research has predominantly focused on the outcomes of the fresh transfer, where patients with excessive ovarian response and patients with very poor response may both be excluded from the fresh transfer cycle. Selection bias may occur when paitents being included in the study basing of their exposure or outcomes status ( 8 ). In addition, the role of chance may also be a consideration when only fresh transfer is evaluated. Due to the morphology-based embryo selection having only limited discriminatory power ( 9 ), the “correct” embryos may be selected following multiple transfers.\nEvaluating the cumulative birth rates taking into account all transfer attempts in a complete COS cycle ( 10 ) may minimize the bias associated with fresh transfer and the random effect of embryo selection. Moreover, the majority of the previous studies have associated the E 2  decline with decreased fertilization or reduced embryo yield, suggesting fewer chances of transfer attempts ( 11 – 13 ). We hypothesize that an E 2  decline observed during COS monitoring may compromise the cumulative live birth via the mediation of reduced embryo yield, even if the fresh transfer outcomes are not affected. The present study aims to evaluate the impact of E 2  decline during COS monitoring on the cumulative live birth rate in a large COS cohort, exploring the mediation effect of embryo yield and the contribution of patient heterogeneity.\n\nWe reviewed all patients who underwent ovarian stimulation for assisted reproductive technologies at Xiamen University Affiliated Chenggong Hospital between January 2013 and December 2021 for potential inclusion. The study was approved by the Institutional Review Board (IRB) of Xiamen University Affiliated Chenggong Hospital. Since the research was based on non-identifiable records, as approved by the IRB, obtaining informed consent was not required.\nDue to the primary goal of the study being to evaluate the cumulative live birth following a complete cycle, the inclusion criteria were patients who achieved at least one live birth during the cycle or patients who had all their embryos transferred. The patients with surplus embryos but without a live birth (n=7767) were excluded. The exclusion criteria were patients who received non-conventional ovarian stimulation protocols such as natural, mild, or luteal phase cycles(n=1224), and cycles with errors in data input (n=3). The inclusion/exclusion criteria were detailed in a flowchart ( \n Figure 1 \n ).\nA flow chart for patient inclusion.\nDuring the stimulation cycle, all patients received either the agonist or antagonist protocol and were administered follicle-stimulating hormone (FSH) or human menopausal gonadotropin (hMG), as previously described ( 14 ). The starting dose of gonadotropins (Gn) for ovarian stimulation ranged from 75 to 300 IU, determined according to the patient’s age, BMI, and ovarian reserve. Following the initiation of stimulation, the patient returned for the next visit in 4 to 6 days if the diameter of the follicle was less than 1.2 cm. When the diameter of the follicle was greater than 1.2 cm, the patient returned for monitoring every 1 to 2 days. In each visit during COS, the development of follicles was monitored under transvaginal ultrasonic examination, and the serum levels of serum follicle-stimulating hormone (FSH), estradiol (E2), and luteinizing hormone (LH) were also evaluated. Gn dosage adjustment may occur following a visit on the clinician’s decision, according to the outcomes of monitoring. The occurrence of a decrease in estradiol during the monitoring period was defined as observing an estradiol value lower than that of the previous follow-up.\nOnce ultrasonography confirmed that the average diameter of at least one follicle reached 18 mm or the diameter of two dominant follicles reached 17 millimeters, 200-250 μg of recombinant human chorionic gonadotropin injection (r-HCG, Ovitrelle, Merck Serono, Germany) would be administered subcutaneously to promote the final maturation of the follicles. Transvaginal ultrasound-guided oocyte retrieval was performed 35 to 37 hours after hCG administration. The occurrence of E 2  decline during the monitoring was defined as the observation of a lower E 2  value than the previous visit.\nThe oocytes were inseminated through either conventional  in vitro  fertilization (IVF) or intracytoplasmic sperm injection (ICSI) and cultured in individual droplets with oil overlay (OVOIL, Vitrolife, Göteborg, Sweden) in COOK culture mediums (COOK MEDICAL, Bloomington, IN). Conventional incubators (C200, Labotect, G¨ottingen, Germany) at 37 °C, 6% CO2, and 5% O2 in a humidified atmosphere were used for the  in vitro  culture. On day 3, the quality of embryos was scored manually according to the criteria of the Istanbul consensus ( 12 ). Patients would receive blastocyst culture according to the preference of the patients or clinicians. The blastocysts were scored according to the Gardner criteria ( 15 ).\nFor both cleavages and blastocysts, a vitrification protocol, employing 15% (v/v) dimethyl sulfoxide, 15% (v/v) ethylene glycol, and 0.6 M sucrose as cryoprotectants, was used for potential cryopreservation. A laser system (SATURN, RI, Falmouth, UK) was used for blastocyst collapse before vitrification.\nThe criteria for judging live births in obstetrics follow the definition of the World Health Organization (WHO) ( 16 ). A live birth event is confirmed as a complete expulsion or extraction from its mother of a product of fertilization, irrespective of the duration of the pregnancy, which, after such separation, breathes or shows any other evidence of life, such as heart beat, umbilical cord pulsation, or definite movement of voluntary muscles, irrespective of whether the umbilical cord has been cut or the placenta is attached.\nThe cumulative live birth as the primary outcome of interest was defined as the first live birth event within a complete cycle. A complete cycle was defined as an OPU cycle that achieves at least one live birth event or has all resulting embryos transferred.\nThe association between E 2  decline during ovarian stimulation monitoring and cumulative live birth was evaluated using a generalized linear model (GLM) and propensity score matching (PS-matching). For PS-matching, a MatchIt package in R software was used ( 17 ). The cobalt package ( 18 ) was used to test the balance. Standard differences (D) were calculated to evaluate the balance of the distribution of the baseline characteristics between the groups before and after PS matching. D < 0.1 was used as the threshold to indicate a negligible difference in the mean or prevalence of a covariate ( 19 ). The balance of covariates was also examined by the distribution of propensity score (distance) between matched groups.\nThe covariates and confounders for the analyses were selected based on previous knowledge and clinical experience with the assistance of a direct acyclic graph (DAG). The DAG was created by DAGitty software ( https://dagitty.net/dags.html ) and shown as a supplementary figure ( \n Supplementary Figure S1 \n ). With a hypothesized association between E 2  decline and cumulative pregnancy, the covariates that are associated with both the E 2  decline (P<0.1 in the dataset) and cumulative live birth ( 20 ) were adjusted. These covariates included female age and BMI, fertility-related diagnosis (duration of infertility, tubal factor, endometriosis, PCOS), ovarian reserve markers (basal FSH, LH, and AFC), and ovarian stimulation (protocol and starting dosage). The analyses were also adjusted for potential confounders, including male factors (male age, BMI, total motile sperm count, and sperm normal morphology), insemination protocols (ICSI versus IVF), and clinical decisions (freeze-all and blastocyst culture) ( 20 – 23 ).\nTo test the hypothesis that E 2  decline during monitoring affects the cumulative live birth via decreasing the embryo viability, we used the “mediation” packages to calculate the proportion of the total effect of E 2  decline mediated by the average causal mediation effect (ACME) of the associated decreased embryo number.\nTo investigate whether a dose-dependent association exists between E 2  decline and cumulative live birth, the association was also analyzed using generalized additive models (GAM) adjusted for the aforementioned covariates. The E 2  decline was analyzed as continuous values and natural log-transformed. A “gratia” (Graceful ‘ggplot’-based graphics and utility functions for working with GAMs fitted using the ‘mgcv’) package was used to identify the potential turning point incorporated with the shape of the resulting curves. The package divided the range of E 2  decline to 1000 points, and derivatives were calculated at each point based on the GAM model. Where the derivatives changed significantly (from indistinguishable from 0 to distinguishable from 0 or vice versa), the threshold was defined.\nTo explore the heterogeneity among patients with E 2  decline, we also carried out PS-matching in patients with and without Gonadotropin (Gn) increase following E 2  decrease, patients with and without consecutive decline at two visits, and patients with and without Gn adjustment before E 2  decline.\nE-values were introduced to measure the minimum strength of association that an unmeasured confounder would need to have to fully explain away the association of interest. The E-values were calculated using the R package “EValue”.\nFor descriptive analyses, continuous variables were analyzed using the Wilcoxon test, and categorical variables were analyzed using the chi-square test or Fisher’s exact test; P < 0.05 was considered to be significant. All analyses were performed using R statistical software 4.12 ( 24 ).\n\nIn this study, 30947 ART cycles were reviewed for potential inclusion, and 27487 cycles were finally included. The number of cycles that underwent at least one E 2  decline was 2863 (10.3%). The characteristics of the patients are shown in  \n Table 1 \n . The patients encountering E 2  decline during ovarian stimulation monitoring were associated with older female age, higher BMI, poorer AFC, and a lower proportion of agonist cycle. However, they also have a higher proportion of PCOS diagnoses and a history of delivering live births. Some of the characteristics of male counterparts, including age and Total motile sperm count (TMC), were also significantly different between patients with and without E 2  decline. Following PS-matching, the standardized differences (D) for all the covariates were lower than 0.1, and the distribution of propensity scores (distance) was identical between comparison groups ( \n Supplementary Figure S2 \n ).\nCharacteristics of patients with and without unexpected estradiol decrease during monitoring.\nData were presented as mean ± SD and median [first quartile, third quartile] for continuous variables and n (percentage) for categorical variables. *D: Standardized difference. The absolute value of D is less than 0.1, cohorts can be considered to be balanced concerning the demographics being assessed.\nTable 2 \n  shows the outcomes of patients with or without E 2  decreased. The decrease in E 2  levels during monitoring was associated with poorer ovarian response, which led to a prolonged duration of stimulation, increased total gonadotropin dosage, and a lower oocyte yield in both matched and unmatched cohorts. The number of mature oocytes, embryos, and high-quality embryos also decreased accordingly. These changes ultimately resulted in approximately a 4% difference (59% VS 55%, P = 0.002) in the cumulative live birth rate in the matched cohort. In the multivariate GLM analyses, the adjusted odds ratios (OR) for cumulative live birth were similar in the unmatched and the matched cohort (OR 0.83, 95% CI: 0.76,0.91 for the unmatched cohort; OR 0.84, 95%CI: 0.75,0.94). On the other hand, the live birth following fresh transfer did not significantly differ between patients with and without an E 2  decline after PS-matching.\nClinical outcomes of patients with and without unexpected estradiol decrease during monitoring.\nAll models were adjusted for female and male age, BMI, history of live birth, duration of infertility, tubal factor, PCOS, endometriosis, basic FSH, LH, antral follicle, agonist protocol, Antagonist protocol, Gn start dose, IVF, ICSI, whole embryo blastocyst culture, freeze-all, Sperm normal morphology, total mobil sperm count. The cumulative live birth rate was the dependent variable.\nThe mediation analyses suggested that the decreased cumulative live birth rate following E 2  decline was largely mediated by the reduced number of mature oocytes or decreased embryo availability. The mature oocyte yield mediated 75.3% (P<0.001) of CLBR decline in the unmatched cohort and 78.3% (P = 0.006) in the matched cohort. Embryo number mediated 76.5% (P = 0.002) and 72.5% (P = 0.01) in the unmatched and matched cohorts, respectively.\nThe GAM model suggested a U-shaped association between E 2  decline and cumulative live birth (edf=2, P = 0.01). The cumulative live birth rates were negatively associated with the degree of E 2  decline when the differences between the two visits were less than 108.9 pg/ml (natural log-transformed 4.68) ( \n Figure 2 \n ). However, the negative association diminished in cycles with greater E 2  decline. The OR for the negative association was 0.862(95%CI: 0.765-0.972). When patients were stratified according to COS protocols, the pattern of association between E 2  decline and CLBR was similar between the agonist and antagonist protocols ( \n Supplementary Figure S3 \n ). However, the dose-response lacked statistical significance in patients with the antagonist protocol (p=0.38). We further visualized the difference between the dose-response curves of agonist and antagonist protocols in the GAM model. It shows that the CLBR was significantly lower in the antagonist protocol than the agonist protocol when the E 2  decline is higher than 4.28 pg/ml (log transformed 1.455), and the difference further extends when the degree of E2 decline further increases. Considering a much lower CLBR in the antagonist protocol than that in the agonist protocol (37.3% versus 71.1%), the lack of association in the antagonist protocol would be due to a lack of power.\nAssociation between the degree of E 2  decline and cumulative live birth rate (CLBR).  (A)  Unadjusted GAM models indicate the association between E 2  decline and CLBR and embryo yield. The black solid line indicates the association with CLBR. The Red dashed line indicates the association with embryo yield.  (B)  Adjusted GAM spline for E 2  decline in association with CLBR. The dashed line indicates the inflection point according to the “gratia” package. The model is adjusted for female age and BMI, fertility-related diagnosis (duration of infertility, tubal factor, endometriosis, PCOS), ovarian reserve markers (basal FSH, LH, and AFC), and ovarian stimulation (protocol and starting dosage), male age, BMI, total motile sperm count, sperm normal morphology, insemination protocols (ICSI versus IVF), and clinical decisions (freeze-all and blastocyst culture).\nTable 3 \n  demonstrates the characteristics of the E 2  decline during monitoring. Most of the E 2  decline occurred between the 8th and 13 th  day of stimulation, with a median E 2  decline of 244 pg/ml. The E 2  decline only had a limited association with follicle count and mean diameter at the visit. As shown in a Pearson correlation matrix ( \n Supplementary Figure S4 \n ), the degree of E 2  decline did not correlate with follicle diameter changes in comparison with the previous visit, and only a low correlation with follicle count changes. A modest correlation (r=0.181) between the degree of E 2  decline and the Gn dosage change in the previous visit was also observed. However, only a small proportion of patients with E 2  decline (5.1%, n=146) were associated with a Gn dosage decrease in the previous visit. On the other hand, 18.8% (n=538) of the patients underwent Gn dosage increase following the observation of decline. In addition, about 11.4% of the patients underwent consecutive E 2  decline at two visits. Since the descriptive data suggested that there was considerable heterogeneity in patients with E 2  decline, we further investigated the effects of heterogeneity on cumulative outcomes.\nHeterogeneity of patients with estradiol decline during monitoring.\nBecause the Venn diagram ( \n Supplementary Figure S5 \n ) suggests that there was limited overlapping between patients who underwent Gn dosage decrease before E 2  decline, patients with Gn dosage increase following E 2  decline, and patients with consecutive E 2  decline in two consecutive visits, we analyzed them independently in multivariate models. In multivariate glm models, we compared the aforementioned E 2  decline subgroups (patients with and without Gn increase following E 2  decrease, patients with and without consecutive decline at two visits, and patients with and without Gn decrease before E 2  decline) with patients without E 2  decline ( \n Supplementary Figure S6 \n ). The size of association (ORs) did not significantly differ between the patients with and without Gn increase following E 2  decrease or patients with and without Gn decrease before E 2  decline ( \n Supplementary Figure S6 \n ). However, the OR in patients with two consecutive E 2  declines (0.58, 95% CI: 0.45,0.74) was significantly lower than that in patients without (OR 0.87, 95%CI: 0.79,0.96).\nTo confirm the effect of heterogeneity, we also carried out PS-matching in the subgroups (detailed in  \n Supplementary Table S1 \n - \n 6 \n ,  \n Supplementary Figures S7 \n - \n 9 \n ), and the summarized results are shown in  \n Table 4 \n . Increasing the dosage of Gn following E 2  decline appeared to improve cumulative live birth, but the difference diminished following matching or multivariate analyses. On the other hand, the patients with consecutive E 2  decline in two visits tended to have lower cumulative live birth rates than patients without in both the unmatched and matched cohorts. Although the P values became marginal in the matched cohort due to reduced sample size, the size of the association was robust in both the unmatched and matched cohorts, with or without multivariate adjustment.\nThe effect of heterogeneity in patients with estradiol decline during monitoring on cumulative live birth.\nAll models were adjusted for female and male age, BMI, history of live birth, duration of infertility, tubal factor, PCOS, endometriosis, basic FSH, LH, antral follicle, agonist protocol, Antagonist protocol, Gn start dose, IVF, ICSI, whole embryo blastocyst culture, freeze-all, Sperm normal morphology, total mobil sperm count. The cumulative live birth rate was the dependent variable.\nE-values of these associations ( \n Supplementary Table S8 \n ) suggested that the minimal strength of the association that an unknown confounding needs to explain away the association between E 2  decline and CLBR was 1.43(95%CI: 1.27,1.56), which was greater than that of the well-known predictor, female age (1.25, 95%CI: 1.23,1.27). For patients with a consecutive estradiol decline, the E-value for E 2  decline was even higher (1.95, 95%CI: 1.6,2.35).\n\nIn our study, we observed a decline in E 2  levels in 2,863 out of 27,487 patients undergoing COS, corresponding to an incidence of 10.63%. This incidence indicates that E 2  decline is a moderate occurrence in the context of COS, highlighting the need for further investigation into its clinical significance. However, the data also shows a modest association between the E 2  decline and decreased CLBR following complete cycles. Compared with propensity score-matched controls, a 4% difference in CLBR was observed. Therefore, E 2  decline may be a potential detriment in COS treatment as a part of ART.\nHistorically, the impact of E 2  decline on the ART outcomes is debatable and the majority of the studies focused on the fresh transfer cycles. Some studies reported a dramatic decrease in pregnancy rates following E 2  decline. For instance, Kulshrestha, et al. reported a 14% pregnancy rate in 23 patients undergoing pre-triggering E 2  decline ( 6 ). In a cohort of 78 patients with spontaneous E 2  decline before hCG administration, Fisher et al. found that clinical pregnancy rates were significantly reduced (13% vs. 39%, p = 0.012) ( 11 ). In contrast, Styer et al ( 7 ) reported no significant differences in live birth or pregnancy loss rates between cycles with E 2  decline and controls in GnRH agonist-downregulated protocols (n=65). Other studies demonstrated a modest decrease in pregnancy and live births following E 2  decline ( 12 ,  25 ). This discrepancy may stem from variations in study populations or protocols. Many of the studies have only limited sample sizes (dozens of patients per group), which could be easily affected by random variation and thus give opposite conclusions ( 26 ). In addition, important determinators of embryo transfer outcomes, such as the number of embryos transferred and the stage of the transferred embryo, are adjusted in none of the above-mentioned studies. Our study contributes to the ongoing debate by adding evidence in a patient cohort with a large sample size and confounders adjusted, showing a neutral effect of E 2  decline on fresh transfer.\nDespite conflicting conclusions on the outcomes following the transfer, the majority of the previous studies agree that E 2  decline during COS may decrease embryo yield for subsequent treatment ( 11 – 13 ). It also raises concerns regarding embryo developmental competence. In a donor-recipient cohort, Cobo et al’s study indicates that a fall of ≥30% in serum E2 concentration during ovarian stimulation in donors negatively affects pregnancy rates and embryo quality in recipients ( 27 ). The authors warn against a decreased embryo quality following E 2  decline and suggest considering a cycle cancellation. Corroborating these studies, our study demonstrates that E 2  decline significantly reduced the total number of oocytes, embryos, and high-quality embryos, mediating a decrease in CLBR following complete cycles. However, considering the limited effect size, we do not recommend cycle cancellation due to the E 2  decline.\nThe degree of E 2  decline is another potential concern. Previous studies may use arbitrary thresholds such as 10% or 30% to define E 2  decline ( 7 ,  11 – 13 ,  25 ,  28 ). However, since E 2  is secreted in granulosa cells, the magnitude of both absolute and relative E 2  decline is largely dependent on the size of the patients’ growing follicle cohort. Excessive E 2  decline could only be expected in hyperresponders. On the other hand, the marginal benefit of CLBR per embryo number increase would also be narrowed in hyperresponders ( \n Supplementary Figure S10 \n ). Considering the decrease in CLBR was primarily mediated by embryo yield, it may explain the U-shaped curve observed in GAM analyses. Nevertheless, our dose-response curves still suggested that patients with a suboptimal response (eg. embryo <=4) are more vulnerable to the E 2  decline, and an E 2  decrease of around 100 pg/ml might be a threshold of concern.\nWe also explored the heterogeneity of patients encountering E 2  decline. Several previous studies have distinguished between spontaneous E 2  decline and intended E 2  reductions via Gn dose adjustment ( 7 ,  13 ), showing that suboptimal outcomes are only found in patients with spontaneous E 2  decline. Echoing the studies, we also found patients having a Gn adjustment prior to E 2  decline tended to have an optimal outcome, though the small number of such patients hampered drawing a firm conclusion. On the other hand, we noted that a significant part of the patients underwent consecutive E 2  decline in two visits, which led to a poorer prognosis. Although the mechanism is not known, the decline may reflect compromised granulosa cell function or accelerated follicular atresia in follicles ( 27 ). While COS recruits a cohort of antral follicles by administration of exogenous Gn, follicles in the cohort may have different degrees of atresia which is probably due to unsynchronized growth ( 29 ). The follicles destinated to atresia before triggering produce E 2  in a gradually reduced production rate but resulted in no oocyte. The consecutive decline would suggest a high degree of atresia in the cohort and impaired follicular dynamics.\nWhen encountering a decline in E 2  levels during COS, clinicians may face management decisions, such as maintaining the scheduled dose of Gn or Gn dosage adjustment during COS. Increasing Gn dosage during COS is a commonly used strategy to maximize the ovarian response and oocyte yield ( 30 ,  31 ). However, its usefulness in patients with E 2  decline is less clear. Styer et al. reported no significant differences in live birth rates in patients with E 2  decline, regardless of whether the dose of gonadotropins was adjusted ( 7 ), suggesting increasing the Gn dosage following the occurrence of E 2  decline may not be effective management. Our subgroup analyses also support the point in a larger, multivariate-adjusted cohort. Nevertheless, the evidence so far is based on retrospective studies, and further study is warranted to investigate to optimize the management of these patients.\n\nThe strengths of the study may include a larger sample size than previous studies ( 7 ,  11 – 13 ,  25 ,  28 ), with multivariate adjustment and reporting of CLBR following the complete cycle. By focusing on CLBR, our study bridges a gap in existing literature, statistically linking repeated E 2  declines to diminished reproductive success.\nThe limitations of this study lie in its retrospective research design, which inherently carries the risk of biases related to the selection and evaluation of clinical cases. Although we employed multivariate analysis to adjust for confounding variables, there is still a possibility that unknown or unmeasured factors may have influenced the research results. For instance, although the culture system, equipment, and staff remained stable during the study period, unmeasured variations in the laboratory environment, such as fluctuation in air quality, may contribute to confounding. Nevertheless, the E-values suggest that the minimal strength of the association that an unknown confounding needs to explain away the association between E 2  decline and CLBR was greater than the well-known predictor, female age. These limitations emphasize the need for caution when interpreting the research findings, and also highlight the importance of using prospective research methods in the future to validate our findings and further elucidate the relationship between E 2  levels and reproductive outcomes during the COS process.\nIn addition, due to the study being observational, the study cohort was also involved in interventions such as Gn dose adjustment during COS. Such interventions are based on the clinicians` decisions on individual cases and inevitably introduce bias. Although we also carried out subgroup analyses, they may be underpowered and heterogeneous.\nFinally, the reasons for the occurrence of continuous decreases in E 2  are still unclear. It is possible that patients who underwent E 2  decline for different reasons are misclassified in our study and previous ones, which resulted in a skewed conclusion.\n\nAlthough E 2  is frequently monitored during COS ( 32 ), the value of routine E 2  monitoring during COS has already been questioned ( 33 ). Our data suggest that the decline in E 2  during COS monitoring is associated with the CLBR following a complete cycle, indicating it remains a critical biomarker in predicting the outcomes during COS. However, the overall size of the association is modest, with an E-value (1.43, 95%CI: 1.27,1.56) comparable to that of female age. Further attention should be paid to specific subgroups of patients, such as patients with consecutive E 2  decline.","source_license":"CC-BY-4.0","license_restricted":false}