Luteinizing hormone change trajectories in gonadotropin-releasing hormone antagonist protocols are associated with assisted reproductive outcomes: results from a retrospective cohort study

Abstract

Context

Luteinizing hormone (LH) plays an integral role in follicular development. In gonadotropin-releasing hormone (GnRH) antagonist protocols, LH levels have been demonstrated to affect assisted reproductive technology (ART) outcomes. However, the exact nature of this relationship remains unclear, and LH measurements at a single time point may not be sufficient to capture it fully.

Methods

We conducted a retrospective cohort study on 4,953 infertile women who underwent their first ART cycle with the GnRH antagonist protocol. Four LH measurements (basal status, antagonist administration day and 24 h later, and ovulation trigger day) were obtain for all participants. We used group-based trajectory modeling (GBTM) to characterize the trajectories of LH level changes during antagonist protocol implementation. Factors influencing the different LH trajectories were explored. Associations between LH levels at different time points and their trajectories with the number of obtained oocytes and ART outcomes in embryo transfer (ET) cycles were analyzed.

Results

Based on the statistical criteria, the model with 4 trajectories were selected as the best-fitting model, and the LH change trajectories during antagonist protocol can be grouped into stable (77.0%), rising (11.0%), decreasing (6.2%) and fluctuating (5.8%) types. Different trajectory groups exhibited significant differences in the number of retrieved oocytes. Compared with that in the stable trajectory group, the oocyte yield in the rising trajectory group decreased and that in the fluctuating trajectory group increased. Of these participants, 884 underwent fresh ET cycles, with no significant differences in pregnancy outcomes between the different trajectory groups. On ovulation trigger day, high LH levels predicted better pregnancy outcomes, indicating significant positive associations with the odds of biochemical pregnancy, clinical pregnancy and live birth.

Conclusion

Oocyte yield differed significantly among the LH trajectory groups, suggesting that LH dynamics during antagonist controlled ovarian stimulation may be relevant to treatment outcomes. For individuals with risk factors for a rising trajectory (e.g., advanced age, higher body mass index), closer monitoring of LH profiles may be warranted. For fresh ET cycles, trigger-day LH levels may merit heightened clinical attention.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13048-026-02140-8.

Introduction

Infertility is a globally prevalent condition that affects an estimated 8%~12% of women of reproductive age worldwide [1]. In vitro fertilization/intracytoplasmic sperm injection (IVF/ICSI) technology, an effective tool for treating infertility, has brought hope to millions of infertile couples. Controlled ovarian stimulation (COS) is a crucial step in IVF/ICSI cycles [2]. In recent decades, the effect of gonadotrophin-releasing hormone (GnRH) antagonists in blocking the luteinizing hormone (LH) surges prior to the planned oocyte retrieval, thereby reducing the risk of cycle cancellation, has been widely recognized in clinical practice [3, 4].

Antagonists bind directly and efficiently to pituitary GnRH receptors to inhibit gonadotropin release competitively and efficiently, within 1 ~ 2 h. Therefore, it is not difficult to understand that GnRH antagonist protocols have the following advantages over GnRH agonist protocols: the absence of a prolonged downregulation period and the consequent side effects due to estrogen deficiency. Furthermore, several studies have shown that GnRH antagonist protocols can shorten the stimulation period with less Gn use and decrease the risk of early ovarian hyperstimulation syndrome (OHSS) [5].

One of the problems associated with adding GnRH antagonists during COS appears to be a dramatic decrease in LH levels [6, 7]. During follicular development, the growth of the antral follicle requires the coregulation of follicle-stimulating hormone (FSH) and LH [8], and the LH surge plays an important role in oocyte meiotic maturation by activating oocyte maturation promoting factors [8, 9]. After GnRH antagonists are added during COS, a functional state of hypogonadotropic hypogonadism may occur, and folliculogenesis, steroidogenesis and oocyte competence may be compromised because of LH deficiency [10]. Specific subgroups of infertile women (those with poor ovarian response or advanced age) may benefit from LH supplementation [7]. Moreover, the manipulation of the LH signaling pathway has been shown to be closely related to human oocyte development and embryo quality [11]. The concept of an "LH therapeutic window" has been developed by researchers, in which estradiol (E₂) production is not adequate when LH levels are below the "threshold", and follicular development may be impaired when LH levels are above the "ceiling" [12]. However, the "threshold" and "ceiling" discussion of LH levels remains unabated. Whether serum LH dynamic changes or a single serum LH assessment should be used as an indicator for COS is a question that should be investigated.

Because of the conflicting and limited data available, we designed this retrospective cohort study. Here, we not only focused on specific LH values at specific time points but also described the distinct trajectories of LH changes based on group-based trajectory modeling (GBTM), a commonly used statistical method to describe the course of repeated measures and identify groups of individuals with similar trajectories. Next, we explored the relationships between different LH change trajectories and IVF/ICSI outcomes, including the number of oocytes retrieved, the rates of metaphase 2 (MII) oocyte, fertilization, blastocyst formation, and clinical pregnancy rate and the live birth rate resulting from the following embryo transfer (ET) cycle, with the expectation of identifying the optimal LH monitoring protocols and appropriate management strategies.

Materials and methods

Study design and participants

We conducted a single-center retrospective cohort study at the Reproductive Centre of the First Affiliated Hospital of Anhui Medical University (Hefei, China). Women were included if they met the following criteria: (i) age < 40 years; (ii) having undergone their first GnRH antagonist cycle between June 1, 2019, and January 1, 2024; (iii) had at least one oocyte retrieved during the COS cycle; and (iv) had LH levels measured at a minimum of four time points during the COS cycle, specifically, at baseline, on the day of antagonist initiation, 24 hour post-antagonist administration, and on the day of ovulation trigger.

Women with the following conditions were excluded from the study: (i) congenital Mullerian tract anomalies; (ii) uncontrolled endocrine disorders (hypogonadotropic hypogonadism, hyperprolactinaemia and uncontrolled thyroid disorders); (iii) complex immune system disorders and a history of tumors; (iv) chromosomal or genetic abnormalities and a preimplantation genetic testing (PGT) treatment cycle; and (v) missing key covariate or outcome data.

All patients underwent basic investigations, including transvaginal ultrasonography and basic sex hormone testing, before ovarian stimulation. Age, body mass index (BMI), education level and infertility type were also recorded. In this study, polycystic ovary syndrome (PCOS) was diagnosed according to the Rotterdam criteria [13]. Given that consensus criteria are lacking for diminished ovarian reserve (DOR), in this study, women who met the following criteria were defined as having DOR: (ⅰ) a basal AFC < 7; (ⅱ) a basal FSH level > 10 IU/L; or (ⅲ) an AMH level < 1.1 ng/mL [14]. Women who had neither PCOS nor DOR were classified as having a normal ovarian reserve (NOR).

The data were extracted from the hospital information system (HIS) and desensitized by the staff. Lastly, 4,953 women were enrolled in the study, and the selection process is detailed in Fig. 1. The present study was approved by the Ethics Committee of Anhui Medical University (No. 20160270).

Fig. 1.

The flow diagram

GnRH antagonist protocol

At our clinic, flexible GnRH antagonist protocols for COS are routinely performed. COS was started from day 2 or 3 of menstruation using Gn (recombinant FSH and/or human menopausal gonadotrophin (hMG)), with the starting dose being determined by a professional reproductive physician depending on the ovarian reserve and BMI, and the subsequent doses of Gn were adjusted according to ovarian responses. GnRH antagonist were added daily when any of the following conditions were met: (ⅰ) at least one follicle ≥ 13 mm in diameter; (ⅱ) serum E₂ levels ≥ 1884 pmol/L; and (ⅲ) LH levels ≥ 10 IU/L [15]. Next, ovarian development was continued to be monitored by ultrasound, and when standards were reached (at least three follicles 17 mm in diameter or larger are visible), ovulation was triggered and the oocytes were extracted after 36 hour. Retrieved oocytes were fertilized by IVF or ICSI depending on sperm quality and then routinely cultured to the blastocyst stage for ET or freezing.

For all included ET cycles, serum human chorionic gonadotropin (hCG) levels were assessed 14 days post-ET. If the test was positive (hCG levels > 25 mIU/ml), a vaginal ultrasound was performed 30 days post-ET. Clinical pregnancy was defined as the presence of a gestational sac, and live birth was defined as the delivery of at least one liveborn infant at ≥ 24 weeks of gestation. Luteal phase support (LPS) was provided for all the ET cycles, which continued until 12 weeks of gestation if pregnancy was achieved.

Serum LH monitoring

LH levels were measured using the UniCel DxI 800 Access chemiluminescence immunoassay system with the Access hLH reagent kit (Beckman Coulter, Inc., USA). Calibration was performed using Access^® hLH Calibrators (2, 10, 25, 100, 250 IU/L). Blood samples were collected before 9:00 AM to minimize circadian variation, and serum was separated by centrifugation at 3,000 g for 10 min.

IVF/ICSI outcome measurement

IVF/ICSI outcomes, including the number of retrieved oocytes, embryo quality indicators (oocyte maturation rate (MII rate, calculated only in ICSI cycles), fertilization rate, and blastocyst formation rate), as well as the odds of biochemical pregnancy, clinical pregnancy and live birth in ET cycles, were evaluated.

Covariates

We carefully reviewed the relevant literature and constructed a directed acyclic graph (DAG) to guide covariate selection (Supplementary Fig. 1). Based on the DAG and prior evidence, the following factors were included as covariates: age, BMI, education level, ovarian reserve, Gn starting dose, total Gn dose, antagonist dose, endometrial thickness, among others. Age and BMI are important factors influencing IVF outcomes. Previous studies have shown that both advanced age and high BMI are clearly associated with adverse results [16–18]. Specifically, obese infertile women face higher Gn doses, increased cycle cancellation rates and higher miscarriage rates. Infertile women of advanced age often exhibit decreased ovarian reserve, which in turn impairs the developmental competence of oocytes and embryos [16]. In addition, previous studies have indicated that increased doses of Gn, particularly in women with higher ovarian reserve, result in higher oocyte yields, which are associated with higher cumulative live birth rates [19–22]. Data on educational level, tobacco use, and type of infertility were collected via questionnaires developed on the basis of medical records from the Anhui assisted reproduction study and described in our previous report [23]. Given that elevated serum P levels and a thinner endometrium are associated with lower implantation rates, these factors were also included as covariates when the relationship between LH levels and ET cycle outcomes was analyzed [24, 25]. As previously reported, we also considered the number of embryos transferred as a covariate [26]. In this study, we did not identify a correlation between passive smoking status and LH trajectories (Table S3).

Statistical analysis

SPSS (IBM, version 23.0) was used for the statistical analyses and StataMP 18 (64-bit) software was used to fit the LH trajectories in GnRH antagonist cycles.

For participant baseline characteristics and hormonal levels (continuous variables), normally distributed variables are expressed as the means ± standard deviation (SD), and nonnormally distributed variables are expressed as the median (quartiles). For categorical variables, the data are expressed as percentages. To assess basic demographic differences between the different ovarian reserve groups and the different LH trajectory groups, the Kruskal-Wallis test was used to compare quantitative variables, and chi-square tests were used to compare categorical variables.

GBTM was used to fit LH Z-scores trajectories during GnRH antagonist cycles [27]. The maximum likelihood method was adopted for model parameter estimation and model fitting. Based on the selection criteria for best-fitting models reported in previous studies [28–30], the number of trajectories was determined by exploring the models with the lowest Bayesian information criterion (BIC) and the number of individuals in each group (> 5%). Basic characteristics that differed between subgroups of LH trajectories were included in a multinomial logistic regression to explore factors associated with LH trajectories.

The associations between serum LH levels (both as trajectory groups and as tertiles of LH levels at each time point) and reproductive outcomes were estimated by marginal structural models with inverse probability weighting (MSM-IPW) with robust standard errors [31]. Different outcome-specific link functions were used: (ⅰ) for oocyte count (count outcome): negative binomial regression; (ⅱ) for embryo quality indicators: fractional logit regression; (ⅲ) for ET outcomes: logistic regression. Marginal contrasts (group comparisons) are presented as contrasts with 95% confidence intervals on the original outcome scale. A two-tailed p value < 0.05 was considered to indicate statistical significance.

Results

Patient demographics and characteristics

This study included 4953 patients aged between 20 and 40 years, with a mean (SD) BMI of 23.17 (3.43) kg/m². These patients underwent their first IVF/ICSI cycles using GnRH antagonist protocols at our reproductive center. Their baseline characteristics are presented in Table 1.

Table 1.

Demographic and clinical characteristics of infertile women in this study

Variables		Variables
Age (year)	30.34 ± 4.12	Trigger method (%)
BMI (Kg/m2)	23.17 ± 3.43	HCG	4149 (83.77)
Education levels (%)		GnRH Agonist or Dual Trigger	804 (16.23)
< 9 years	1781 (35.96)	No. of oocytes retrieved	13.00 (8.00, 19.00)
9 ~ 12 year	2005 (40.48)	No. of blastocysts formed	5.00 (3.00, 8.00)
> 12 year	1167 (23.56)	ICSI MII rate	0.81 (0.69, 0.92)
Passive smoking (%)		Fertilization rate	0.86 (0.73, 1.00)
No	2992 (60.41)	Embryo formation rate	0.57 (0.42, 0.73)
Yes	1961 (39.59)	Fertilization type
Duration of infertility (year)	2.68 ± 2.18	IVF	3359 (67.82)
Infertility type (%)		ICSI	1283 (25.90)
Primary infertility	2889 (58.33)	Rescue ICSI	135 (2.73)
Secondary infertility	2064 (41.67)	Combined (IVF+ICSI) type or oocyte freezing	176 (3.55)
AFC	16.00 (11.00, 24.00)	Fresh/frozen-embryo transfer cycles	n = 4715
Basal FSH (IU/L)	7.14 ± 2.19	Biochemical pregnancy
Basal LH (IU/L)	4.86 (3.47, 6.76)	No	1515 (32.13)
Basal E2 (pmol/L)	151.00 (105.00, 211.00)	Yes	3200 (67.87)
E2 on HCG day (pmol/L)	12754.48 (7322.90, 17846.00)	Clinical pregnancy
LH on HCG day (IU/L)	2.53 (1.58, 4.01)	No	1761 (37.36)
P on HCG day (nmol/L)	3.84 (2.61, 5.51)	Yes	2953 (62.64)
Gn initiating dose (IU)	187.50 (150.00, 225.00)	Early spontaneous miscarriage
Gn total does (IU)	1875.00 (1500.00, 2325.00)	No	2574 (88.94)
alone rFSH (%)	3279 (66.20)	Yes	320 (11.06)
HMG alone or combined (%)	1674 (33.80)	Live births
Antagonist dose (mg)	1.25 (1.00, 1.50)	No	2162 (46.33)
		Yes	2505 (53.67)

BMI Body Mass Index, AFC Antral Follicle Count, FSH Follicle-Stimulating Hormone, rFSH Recombinant Follicle-Stimulating Hormone, LH Luteinizing Hormone, E₂ estradiol, HMG Human Menopausal Gonadotropin, HCG Human Chorionic Gonadotropin, Gn Gonadotropin, GnRH Gonadotropin-releasing Hormone, P progesterone

The average duration of infertility was 2.68 years, and the median AFC, as a predictor of ovarian reserve, was 16.00 (11.00, 24.00). In this study, 39.59% of the participants were passive smokers, and 3172 (64.04%) of the participants had received at least 9 years of education (40.48% for 9~12 years and 23.56% for > 12 years). For the basal hormone profile, the average FSH concentration was 7.14 (SD = 2.19) IU/L, the median LH concentration was 4.86 (3.47, 6.76) IU/L, and the median E₂ concentration was 151.00 (105.00, 211.00) pmol/L. On hCG day, the medium level of E₂ concentration was 12754.48 (7322.90, 17846.00) pmol/L, the medium level of LH level was 2.53 (1.58, 4.01) IU/L, and the medium level of progesterone (P) level was 3.84 (2.61, 5.51) nmol/L.

In this study, the median initial dose of Gn was 187.50 (150.00, 225.00) IU, and the median total dose of Gn was 1875.00 (1500.00, 2325.00) IU. The median number of total retrieved oocytes was 13.00 (8.00, 19.00), and the median number of available embryos was 5.00 (3.00, 8.00). We also calculated the embryo quality indicators, yield a median MII rate (only ICSI cycle) of 0.81 (0.69, 0.92), a median fertilization rate of 0.86 (0.73, 1.00), and a median blastocyst formation rate of 0.57 (0.42, 0.73).

Here, 586 patients were diagnosed with DOR, and 960 were diagnosed with PCOS. As expected, patients with PCOS were younger (28.79 ± 3.64 years), had a higher average BMI (24.25 ± 3.58 kg/m²), had a greater tendency toward primary infertility (68.12%), and had higher median basal LH (7.05 (4.59, 10.31) IU/L) levels. For the patients with PCOS, we used a lower median starting dose of Gn (150.00 (150.00, 200.00) IU), and total dose (1687.50 (1350.00, 2150.00) IU) but obtained more median number of oocytes (17.00 (11.00, 25.00)) and blastocysts (7.00 (4.00, 11.00)), although the MII rate was lower (0.79 (0.67, 0.88)). In contrast, patients with DOR showed the trend opposite to those with PCOS. They had an older average age (33.51 ± 4.15 years), a lower median AFC (6.00 (4.00, 7.00)), a higher average basal FSH level (8.98 ± 3.27 IU/L), and fewer median oocytes (6.00 (4.00, 9.00)) and blastocysts (2.00 (1.00, 4.00)) (Table S1).

Notably, 4715 participants underwent ET cycles: the biochemical pregnancy rate was 67.87% (3200/4715), and the clinical pregnancy rate was 62.64% (2953/4714, with one patient lost to follow-up after biochemical pregnancy confirmation). Among these 2953 clinical pregnancies, 12 were ectopic pregnancies, and 389 were pregnancy losses (320 spontaneous abortions before 12 gestational weeks (GW), 37 spontaneous abortions after 12 GW, and 32 medically induced abortions due to specific medical indications), 47 patients were lost to follow-up, and 2505 had live births (Table 1).

LH trajectories and associated factors

In this study, GBTM analyses were performed based on changes in LH levels after Z-scores standardized during COS, and the best model was selected from 2 to 5 types of potential trajectory models. The results of the fitting indices for each model are detailed in Table S2. Lastly, according to the selection principles described above [27–30], the model with 4 trajectories was chosen as the optimal potential model and the trajectories were grouped into stable (77.0%), rising (11.0%), decreasing (6.2%) and fluctuating (5.8%) types (Fig. 2A). We also used box plots to represent the changes in raw LH concentrations for different trajectories (Fig. 2B).We further analyzed the differences in demographic characteristics between trajectory subgroups. The results revealed that age, BMI, education level, duration of infertility, type of infertility, ovarian reserve, antagonist dose, Gn starting dose and total Gn dose were statistically significant, except for passive smoking status (Table S3). Characteristics that differed were subsequently included in the multinomial logistic regression, and the results are shown in Table S4. Specifically, women with advanced age (OR = 1.04; 95% CI: 1.01, 1.06) and high BMI (OR = 1.05; 95% CI: 1.02, 1.08) were more likely to experience the “rising trajectory”. And, those with higher education levels (9 ~ 12 years vs. < 9 years: OR = 1.39; 95% CI: 1.03, 1.88; >12 years vs.< 9 years: OR = 1.45; 95% CI: 1.03, 2.05) were more susceptible to LH fluctuations. In addition, patients with PCOS were more likely to have LH instability during COS, with a significantly increased risk of experiencing rising (OR = 2.17; 95% CI: 1.71, 2.74), decreasing (OR = 11.09; 95% CI: 8.38, 14.67), and fluctuating (OR = 2.24; 95% CI: 1.68, 2.98) types of LH trajectories.

Fig. 2.

LH trajectories during controlled ovarian stimulation (COS) in antagonist protocol. A, LH Z-scores trajectories were fited using Group-based trajectory modeling (GBTM); B, boxplot shows median and quartiles of raw LH levels during COS in different trajectory subgroups, with connecting lines showing changes in median levels). Note: Time 1 represents the basal status; Time 2 represents the antagonist day; Time 3 represents 24 hours after antagonist administration; and Time 4 represents the ovulation trigger day

Associations of LH trajectories with the number of oocytes obtained

The data in Table 2 reveal significant differences in the number of oocytes obtained and embryos formed between LH trajectories. To determine whether the trajectory of change in LH was more predictive of the ovarian response, we used the MSM-IPW. The results are presented in Table 3. Compared with the stable trajectory, the rising trajectory had lower oocyte yield (difference = -1.47, 95% CI: -2.29, -0.65) and the fluctuating trajectory had higher yield (difference = 4.22, 95% CI: 2.82, 5.62). Stratified analyses revealed heterogeneity by ovarian reserve status. In the NOR group, the results mirrored the overall population. In the DOR group, no significant associations emerged. Notably, compared with the overall population, the decreasing trajectory showed a reversed direction for the effect (difference = -2.80, 95% CI: -5.84, 0.24). In the PCOS group, the fluctuating trajectory remained significantly associated with increased oocyte yield (difference = 6.22, 95% CI: 3.47, 8.97). To account for the potential impact of Gn medications, we further restricted the analysis to rFSH-only cycles. The results were consistent with those observed in the overall population, with adjusted mean differences (95% CI) relative to the stable trajectory as follows: rising trajectory: -1.29 (-2.34, -0.24); decreasing trajectory: 2.13 (0.42, 3.85); and fluctuating trajectory: 4.04 (2.79, 5.30) (Table S5).

Table 2.

IVF/ICSI outcomes in different LH trajectories

Variables	Stable trajectory	Rising trajectory	Decreasing trajectory	Fluctuating trajectory	p-value
No. of oocytes retrieved	12.00 (8.00, 18.00)	10.00 (7.00, 15.00)	17.00 (10.00, 24.00)	19.00 (13.00, 26.00)	< 0.001
No. of blastocysts formed	5.00 (3.00, 8.00)	4.00 (2.00,7.00)	6.50 (4.00, 10.00)	7.00 (4.00, 11.00)	< 0.001
ICSI MII rate	0.81 (0.69, 0.92)	0.78 (0.67, 0.93)	0.79 (0.66, 0.90)	0.80 (0.65, 0.89)	0.285
Fertilization rate	0.86 (0.73, 1.00)	0.86 (0.71, 1.00)	0.88 (0.77, 0.96)	0.86 (0.75, 0.94)	0.111
Embryo formation rate	0.57 (0.42, 0.72)	0.57 (0.40, 0.75)	0.60 (0.43, 0.75)	0.58 (0.44, 0.72)	0.381
Fresh/frozen-embryo transfer cycles
Biochemical pregnancy					0.439
No	1188 (32.41)	162 (33.26)	80 (28.27)	85 (30.47)
Yes	2478 (67.59)	325 (66.74)	203 (71.73)	194 (69.53)
Clinical pregnancy					0.413
No	1371 (37.41)	185 (37.99)	94 (33.22)	111 (39.78)
Yes	2294 (62.59)	302 (62.01)	189 (66.78)	168 (60.22)
Early spontaneous miscarriage					0.660
No	2000 (89.09)	258 (86.87)	169 (89.89)	147 (89.63)
Yes	245 (10.91)	39 (13.13)	19 (10.11)	17 (10.37)
Live births					0.338
No	1682 (46.40)	233 (48.14)	117 (41.49)	130 (47.10)
Yes	1943 (53.60)	251 (51.86)	165 (58.51)	146 (52.90)

IVF in vitro fertilization, ICSI intracytoplasmic sperm injection, LH Luteinizing Hormone, MII metaphase II

Table 3.

Marginal structural models with inverse probability weighting for total oocytes retrieved: a stratified analysis by ovarian reserve status

	Overall population		NOR		DOR		PCOS
	Difference (95% CI)	p	Difference (95% CI)	p	Difference (95% CI)	p	Difference (95% CI)	p
Basal LH level
Tertile 1 (Low LH)	Ref		Ref		Ref		Ref
Tertile 2 (Middle LH)	0.52 (−0.10, 1.13)	0.099	0.45 (−0.16, 1.06)	0.151	−0.35 (−1.07, 0.38)	0.350	1.21 (−0.67, 3.08)	0.208
Tertile 3 (High LH)	0.98 (0.34, 1.63)	0.003	1.08 (0.39, 1.78)	0.002	−0.62 (−1.55, 0.31)	0.192	1.20 (−0.50, 2.90)	0.166
LH level on antagonist day
Tertile 1 (Low LH)	Ref		Ref		Ref		Ref
Tertile 2 (Middle LH)	0.73 (0.17, 1.29)	0.011	0.66 (0.09, 1.23)	0.023	−0.21 (−0.91, 0.48)	0.545	1.53 (−0.22, 3.29)	0.087
Tertile 3 (High LH)	3.58 (2.94, 4.23)	< 0.001	4.02 (3.36, 4.68)	< 0.001	−0.67 (−1.73, 0.39)	0.217	4.46 (2.82, 6.10)	< 0.001
LH levels after antagonist days
Tertile 1 (Low LH)	Ref		Ref		Ref		Ref
Tertile 2 (Middle LH)	−0.50 (−1.09, 0.09)	0.099	−0.64 (−1.27, 0.00)	0.049	−0.34 (−1.10, 0.42)	0.376	−0.17 (−1.83, 1.48)	0.839
Tertile 3 (High LH)	−0.37 (−0.96, 0.23)	0.228	−0.41 (−1.05, 0.23)	0.211	−0.28 (−1.10, 0.54)	0.499	−0.25 (−1.84, 1.34)	0.758
LH level on HCG day
Tertile 1 (Low LH)	Ref		Ref		Ref		Ref
Tertile 2 (Middle LH)	−1.97 (−2.56,−1.38)	< 0.001	−2.24 (−2.85,−1.62)	< 0.001	−0.91 (−1.76,−0.05)	0.039	−1.79 (−3.44,−0.13)	0.034
Tertile 3 (High LH)	−2.86 (−3.47,−2.25)	< 0.001	−3.18 (−3.83,−2.53)	< 0.001	−1.99 (−2.76,−1.21)	< 0.001	−2.21 (−3.85,−0.57)	0.008
Trajectory models
Stable trajectory	Ref		Ref		Ref		Ref
Rising trajectory	−1.47 (−2.29,−0.65)	< 0.001	−1.60 (−2.58,−0.62)	0.001	−0.88 (−1.81, 0.04)	0.062	−1.32 (−3.16, 0.51)	0.157
Decreasing trajectory	1.77 (0.21, 3.33)	0.026	2.27 (0.52, 4.01)	0.011	−2.80 (−5.84, 0.24)	0.071	−0.29 (−1.94, 1.36)	0.731
Fluctuating trajectory	4.22 (2.82, 5.62)	< 0.001	3.25 (1.86, 4.64)	< 0.001	1.50 (−2.17, 5.16)	0.424	6.22 (3.47, 8.97)	< 0.001

Model: Marginal structural models with inverse probability weighting (MSM-IPW) using negative binomial regression with robust standard errors. Difference refers to the difference in oocyte count between comparison groups. Weight calculation: Stabilized IPW weights derived from multinomial logistic regression adjusted for age, BMI, infertility duration, infertility type, education level, ovarian reserve, Gn starting dose, Gn total dose, antagonist dose, HMG use, E2/P levels, and trigger method. Stratified analyses: Stratified analyses were performed by ovarian reserve

LH Luteinizing Hormone, HCG Human Chorionic Gonadotropin

In addition, we analyzed the relationship between LH levels at each time point and the number of oocytes acquired. We found that higher basal LH (Tertile 3 vs. 1: adjusted difference = 0.98, 95% CI: 0.34, 1.63) and higher LH on the antagonist administration day (Tertile 2 vs. 1: adjusted difference = 0.73, 95% CI: 0.17, 1.29; Tertile 3 vs. 1: adjusted difference = 3.58, 95% CI: 2.94, 4.32) were positively associated with oocyte yield, whereas higher LH on the HCG day was negatively associated with yield (Tertile 2 vs. 1: adjusted difference = -1.97, 95% CI: -2.56, -1.38; Tertile 3 vs. 1: adjusted difference = -2.86, 95% CI: -3.47, -2.25).

In terms of oocyte maturation rates, fertilization rates, and blastocyst formation rates, we did not observe statistically significant differences between the four trajectory groups (Table 2). However, further analysis using MSM-IPW revealed that the decreasing trajectory was positively associated with fertilization rates (adjusted difference = 0.04, 95% CI: 0.01, 0.06). According to the subgroup analyses, this association was significant for IVF cycles but not for ICSI cycles. In terms of ovarian reserve status, a positive association was observed in the NOR and PCOS groups, whereas no significant association was found in the DOR group.

Associations of LH trajectories with reproductive outcomes

Table 4 shows the association between LH levels and the pregnancy outcomes in ET cycles. No significant differences were observed in fresh cycle transfer outcomes across the four LH trajectory groups. However, we observed that moderately higher basal LH levels were associated with favorable pregnancy outcomes, but in a nonlinear manner. Compared with the lowest tertile, the second tertile of basal LH was significantly associated with biochemical pregnancy (adjusted difference = 0.04, 95% CI: 0.00, 0.07), clinical pregnancy (adjusted difference = 0.04, 95% CI: 0.00, 0.08), and live birth (adjusted difference = 0.06, 95% CI: 0.02, 0.10). In contrast, the associations were attenuated to nonsignificant for the highest tertile. In fresh ET cycles, the LH level on the ovulation trigger day was positively associated with pregnancy outcome. Compared with the lowest tertile, the adjusted differences (95% CI) for the second and third tertiles were as follows: biochemical pregnancy: 0.12 (0.04, 0.21) and 0.16 (0.08, 0.25); clinical pregnancy: 0.12 (0.03, 0.20) and 0.17 (0.08, 0.26); and live births: 0.10 (0.02, 0.19) and 0.16 (0.07, 0.25).

Table 4.

Marginal structural models with inverse probability weighting for pregnancy outcomes in fresh and frozen-thawed ET cycles

	Biochemical pregnancy		Clinical pregnancy		Early spontaneous miscarriage		Live birth
	Difference (95% CI)	p	Difference (95% CI)	p	Difference (95% CI)	p	Difference (95% CI)	p
Overall population
Basal LH level
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.04 (0.00, 0.07)	0.032	0.04 (0.00, 0.08)	0.035	-0.03 (-0.06, 0.00)	0.059	0.06 (0.02, 0.10)	0.003
Tertile 3	0.02 (-0.02, 0.05)	0.390	0.02 (-0.02, 0.06)	0.302	-0.02 (-0.05, 0.01)	0.272	0.04 (0.00, 0.08)	0.078
LH level on antagonist day
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.02 (-0.01, 0.06)	0.243	0.02 (-0.02, 0.06)	0.286	0.00 (-0.03, 0.03)	0.902	0.02 (-0.02, 0.06)	0.362
Tertile 3	0.00 (-0.03, 0.04)	0.839	0.00 (-0.04, 0.04)	0.885	-0.01 (-0.05, 0.02)	0.410	0.00 (-0.04, 0.04)	0.901
LH levels after antagonist days
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.03 (-0.01, 0.06)	0.109	0.03 (-0.01, 0.06)	0.139	-0.01 (-0.04, 0.02)	0.494	0.03 (0.00, 0.07)	0.087
Tertile 3	0.00 (-0.03, 0.04)	0.777	-0.01 (-0.04, 0.03)	0.677	-0.01 (-0.04, 0.03)	0.720	0.00 (-0.04, 0.04)	0.948
LH level on HCG day
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.02 (-0.01, 0.06)	0.182	0.02 (-0.02, 0.05)	0.321	0.00 (-0.03, 0.03)	0.971	0.01 (-0.02, 0.05)	0.459
Tertile 3	0.04 (0.01, 0.08)	0.018	0.03 (-0.01, 0.07)	0.115	0.01 (-0.02, 0.04)	0.617	0.02 (-0.02, 0.06)	0.362
Trajectory models
Stable trajectory	Ref		Ref		Ref		Ref
Rising trajectory	0.01 (-0.04, 0.05)	0.831	0.00 (-0.05, 0.05)	0.925	0.01 (-0.03, 0.06)	0.489	0.00 (-0.05, 0.05)	0.910
Decreasing trajectory	-0.02 (-0.10, 0.07)	0.692	0.00 (-0.09, 0.08)	0.966	-0.03 (-0.08, 0.02)	0.279	0.02 (-0.07, 0.11)	0.662
Fluctuating trajectory	0.00 (-0.08, 0.07)	0.943	-0.04 (-0.12, 0.05)	0.390	-0.02 (-0.08, 0.03)	0.421	-0.01 (-0.09, 0.07)	0.810
Fresh ET cycles
Basal LH level
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.03 (-0.05, 0.11)	0.510	0.03 (-0.05, 0.11)	0.431	0.01 (-0.05, 0.08)	0.757	0.02 (-0.06, 0.10)	0.634
Tertile 3	0.02 (-0.07, 0.11)	0.675	0.03 (-0.06, 0.12)	0.457	0.03 (-0.04, 0.10)	0.430	0.03 (-0.06, 0.12)	0.571
LH level on antagonist day
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.03 (-0.04, 0.11)	0.393	0.02 (-0.05, 0.10)	0.569	0.02 (-0.04, 0.09)	0.516	0.02 (-0.06, 0.09)	0.640
Tertile 3	0.07 (-0.02, 0.17)	0.131	0.04 (-0.06, 0.14)	0.421	-0.02 (-0.10, 0.07)	0.694	0.06 (-0.04, 0.16)	0.266
LH levels after antagonist days
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.10 (0.02, 0.18)	0.012	0.12 (0.04, 0.20)	0.002	0.00 (-0.07, 0.06)	0.896	0.11 (0.03, 0.19)	0.007
Tertile 3	0.05 (-0.03, 0.14)	0.215	0.03 (-0.05, 0.12)	0.432	0.03 (-0.05, 0.11)	0.427	0.04 (-0.05, 0.12)	0.390
LH level on HCG day
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.12 (0.04, 0.21)	0.005	0.12 (0.03, 0.20)	0.007	0.01 (-0.07, 0.09)	0.801	0.10 (0.02, 0.19)	0.020
Tertile 3	0.16 (0.08, 0.25)	< 0.001	0.17 (0.08, 0.26)	< 0.001	0.00 (-0.08, 0.07)	0.966	0.16 (0.07, 0.25)	< 0.001
Trajectory models
Stable trajectory	Ref		Ref		Ref		Ref
Rising trajectory	0.08 (-0.03, 0.19)	0.138	0.06 (-0.05, 0.17)	0.312	0.00 (-0.10, 0.10)	0.944	0.05 (-0.06, 0.17)	0.345
Decreasing trajectory	-0.18 (-0.39, 0.03)	0.096	-0.16 (-0.37, 0.05)	0.140	0.06 (-0.21, 0.33)	0.649	-0.14 (-0.34, 0.05)	0.158
Fluctuating trajectory	0.07 (-0.18, 0.31)	0.595	0.09 (-0.15, 0.33)	0.453	0.00 (-0.15, 0.16)	0.991	0.06 (-0.19, 0.31)	0.617
Frozen ET cycles
Basal LH level
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.04 (0.00, 0.08)	0.035	0.04 (0.00, 0.08)	0.047	-0.04 (-0.07, 0.00)	0.035	0.07 (0.03, 0.11)	0.002
Tertile 3	0.02 (-0.03, 0.06)	0.443	0.02 (-0.03, 0.06)	0.423	-0.03 (-0.06, 0.01)	0.139	0.04 (-0.01, 0.08)	0.092
LH level on antagonist day
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.02 (-0.02, 0.06)	0.354	0.02 (-0.02, 0.06)	0.351	-0.01 (-0.04, 0.03)	0.723	0.02 (-0.03, 0.06)	0.415
Tertile 3	-0.01 (-0.05, 0.03)	0.586	-0.01 (-0.06, 0.03)	0.570	-0.01 (-0.05, 0.02)	0.475	-0.01 (-0.05, 0.04)	0.680
LH levels after antagonist days
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.01 (-0.03, 0.05)	0.579	0.00 (-0.03, 0.04)	0.846	-0.01 (-0.04, 0.02)	0.518	0.01 (-0.03, 0.05)	0.515
Tertile 3	-0.01 (-0.04, 0.03)	0.733	-0.02 (-0.06, 0.02)	0.380	-0.01 (-0.05, 0.02)	0.465	-0.01 (-0.05, 0.03)	0.730
LH level on HCG day
Tertile 1	Ref		Ref		Ref		Ref
Tertile 2	0.00 (-0.04, 0.04)	0.932	0.00 (-0.04, 0.03)	0.836	0.00 (-0.03, 0.03)	0.938	-0.01 (-0.05, 0.03)	0.789
Tertile 3	0.02 (-0.02, 0.05)	0.435	0.00 (-0.04, 0.04)	0.900	0.01 (-0.02, 0.04)	0.553	-0.01 (-0.06, 0.03)	0.509
Trajectory models
Stable trajectory	Ref		Ref		Ref		Ref
Rising trajectory	-0.01 (-0.06, 0.04)	0.651	-0.01 (-0.07, 0.05)	0.723	0.02 (-0.03, 0.06)	0.458	-0.02 (-0.07, 0.04)	0.586
Decreasing trajectory	0.01 (-0.08, 0.10)	0.900	0.02 (-0.07, 0.11)	0.650	-0.04 (-0.09, 0.01)	0.124	0.05 (-0.05, 0.14)	0.349
Fluctuating trajectory	-0.02 (-0.10, 0.06)	0.628	-0.06 (-0.15, 0.02)	0.132	-0.03 (-0.09, 0.03)	0.365	-0.03 (-0.11, 0.06)	0.528

Marginal structural models with inverse probability weighting (MSM-IPW) using logistic regression with robust standard errors. Difference refers to the risk difference for binary outcomes between comparison groups. Weights: Stabilized IPW weights from multinomial logistic regression adjusted for age, BMI, duration of infertility, infertility type, education level, ovarian reserve, Gn starting dose and Gn total dose, antagonist dose, use of HMG, trigger method, number of embryos transferred, oocyte retrieved, cycle type, endometrial thickness, and E2/P level. Stratified by: Fresh vs. frozen-thawed cycles

LH Luteinizing Hormone, HCG Human Chorionic Gonadotropin

Discussion

Most previous studies have focused on the relationship between LH levels at specific time points during COS and ART outcomes [32–40]. It is well known that COS is an evolving process. Assessing the impact of LH levels on ART outcomes merely from a single measurement may not be adequate. Therefore, in this study, we focused not only on the serum LH levels at specific time points (basal status; the antagonist administration day and 24 h later; ovulation trigger day) but also on the changes in LH levels during COS, and we constructed LH trajectories based on the GMBT. To our knowledge, this is the first time the GMBT was used to explore the relationship between changes in LH and reproductive outcomes.

Follicular development is a highly intricate process that depends on the secretion and precise regulation of FSH and LH. It is currently well known that the secretion of Gn is regulated by sex steroids, hypothalamic hormones and feedback provided by themselves ( known as internal or autofeedback) [41]. During follicular development, LH promotes follicular maturation by acting through specific receptors [42], and LH fluctuations may adversely affect follicular development. However, this hypothesis remains unsubstantiated by evidence. We innovatively fitted the LH change trajectory using GBTM and explored the associated factors in the hope of providing personalized LH monitoring protocols during COS.

Our results indicated that the LH trajectory was associated with age, BMI, ovarian reserve, and Gn dosage. The "rising trajectory" implied that pituitary function was not effectively suppressed after the addition of the antagonist; thus, it is not surprising that the number of oocytes obtained from this group was significantly lower. This finding may be due to the late or insufficient addition of antagonists or individual differences in LH sensitivity. In addition, women with advanced age and a higher BMI were prone to a "rising trajectory" and patients with PCOS were also more likely to experience LH instability during COS. Identifying these factors may help in the selection of appropriate LH monitoring and management options.

The "decreasing trajectory" indicated that basal LH was at a high level, but with follicular development and increased estrogen production, LH levels were suppressed until the ovulation trigger day. Elevated basal LH means that the body may be exposed to high LH levels for prolonged periods, which can lead to the desensitization of the LH receptor and thus affect follicular maturation [43]. High basal LH levels are likely to indicate stronger follicular activity, especially in patients with PCOS [44]. It is well known that hypersecretion of LH is an important feature of PCOS [45–47]. Our results also suggested that the presence of PCOS was a positive factor for the "decreasing trajectory" relative to the "stable trajectory". The effects of higher basal LH levels on IVF outcomes remain controversial. Several studies have not reported an adverse effect of elevated basal LH levels on IVF outcomes [44, 45, 48]. However, a retrospective cohort study on 1011 PCOS patients treated with a progestin-primed ovarian stimulation (PPOS) protocol revealed an increase in oocyte production and blastocyst count in women with higher basal LH levels (LH ≥ 10 IU/L) than in a group of patients with lower basal LH levels [45]. Popovic-Todorovic B et al. reported that higher LH levels at the start of ovarian stimulation were favorable for oocyte yield following a GnRH agonist trigger [44]. Our findings also suggested that elevated basal LH levels were associated with increased oocyte yield. Moreover, moderately elevated basal LH levels (i.e., the second tertile) were associated with favorable pregnancy outcomes, whereas high basal LH levels (the third tertile) did not significantly improve outcomes.

The "fluctuating trajectory" showed that the LH level was significantly greater than the basal status on the day of antagonist administration, whereas it was suppressed immediately and decreased sharply after the addition of the antagonist. In our study, associated factor analysis revealed that education level, PCOS status, and Gn dose were relevant to LH fluctuations. We presume that these factors are associated with the recruitment and maturation of more follicles and that excess estrogen secretion has a positive feedback effect on transient LH release. Little attention has been gaven to this phenomenon. Zhang D et al. reported that a transient premature LH surge can cause detrimental effects on embryo development and implantation in fresh ET cycles [40]. In contrast, the results reported by Huirne J A et al. suggested that transient elevations in LH do not affect clinical pregnancy rates and that an early surge appears to be associated with DOR [37]. In our study, we found that the "fluctuating trajectory" group had a greater number of oocytes and that there was no significant association with embryo quality indicators. Although we did not report the incidence of OHSS in this study, previous studies have shown that the number of retrieved oocytes is a positive predictor of OHSS [49, 50]. If the risk factors for LH fluctuation are recognized and an individualized COS protocol, such as a lower exogenous Gn dose, is applied appropriately, this action contributes to increased cost efficiency and improved safety. In addition, we reported that LH levels on the day of antagonist administration were positively correlated with oocyte yield, and similar results were reported by Pang C et al. [38]. This finding indicated that more follicles were recruited , which is easily explained.

The trigger day is a critical stage of oocyte maturation during COS, and the association between LH levels on this day and ART outcomes is controversial. A recent meta-analysis revealed that the long GnRH agonist protocol was more favorable than the GnRH antagonist protocol in terms of sustained pregnancy rates [51]. In other words, antagonists may have adverse effects on pregnancy outcomes. Although the underlying mechanism has not been fully elucidated, low implantation rates in fresh ET cycles are thought to be related to impaired endometrial receptivity, given that this problem is not seen in frozen ET cycles [15]. Whether the reduction in endometrial receptivity is due to a decrease in LH levels following the addition of antagonists has not been adequately investigated. In the present study, we did not find differences in pregnancy outcomes between different LH trajectory groups; however, high LH levels on HCG days were evidence of favorable pregnancy chances even after adjustment. Other retrospective studies have reported that higher LH on trigger days may improve pregnancy outcomes [32, 52]. In the present study, we did not attempt to determine the most appropriate threshold for serum LH levels on the trigger day to predict pregnancy outcomes, and previous studies have recommended inconsistent thresholds [32, 53, 54]. As a next step, we will expand the study sample size to review all fresh ET cycles at our center, with the aim of identifying LH thresholds that can be used to guide clinical work. In addition, we observed a negative correlation between LH levels on the HCG day and the number of oocytes retrieved. This observation aligns with prior research findings [39, 55]. Notably, a decrease in the number of retrieved oocytes implies a reduction in the number of embryos, which subsequently affects the cumulative live birth rate. Therefore, a comprehensive decision to perform fresh ET should be made based on the number of embryos available and the LH level on the HCG day. In this study, LH trajectory patterns were not adversely associated with embryo development indicators. Our interpretation is that the amount of LH required for follicular growth is minimal and still has sufficient biological potency after pituitary desensitization. In particular, most of our study participants were young adults who were much less likely to experience LH deficiency. Therefore, changes in LH should be monitored during COS to obtain an optimal number of oocytes.

In this study, we introduced GBTM to identify subgroups of LH level changes during GnRH antagonist cycles and analyzed differences in IVF outcomes between the subgroups. In addition, for data quality control, all 4 measurements of LH levels in the subjects were performed at our clinic. However,several limitations should be noted. A main limitation of our study is its retrospective design, which introduced various confounding factors. Although we adjusted for a range of relevant covariates, we could not control for all potential confounders. Residual confounding and selection bias inherent to retrospective data may still affect our findings. Moreover, measurement variability—particularly for hormone levels—cannot be fully excluded, which may have introduced misclassification. Furthermore, the data-driven GBTM approach may carry a risk of overfitting. Additionally, the decision to proceed with fresh embryo transfer is not randomized but influenced by multiple clinical factors, which may introduce selection bias in the assessment of embryo implantation outcomes. Another limitation is that this study included data from only one center in Hefei, Anhui Province, and the number of subjects who underwent fresh ET was relatively small. Thus, our findings should be considered hypothesis-generating and require confirmation in prospective studies.

Conclusion

In this study, differences in the number of retrieved oocytes were observed among the different LH trajectory groups, suggesting that monitoring LH changes during COS with antagonist protocols may help optimize oocyte yield. Additionally, our results suggest a possible association between low LH levels on the trigger day and poorer embryo implantation in fresh ET cycles, suggesting that LH levels at this point may warrant closer clinical attention. Nevertheless, specific LH thresholds remain to be determined, and future prospective studies with larger sample sizes are warranted to confirm these findings.

Authors' contributions

Xun Su: Writing - original draft, Formal analysis, Data curation. Qi Zheng: Writing - original draft, Visualization, Formal analysis. Dongyang Zhang: Methodology, Investigation. Qian Li: Acquisition, Analysis. Hua Zhang: Data curation. Yiqing Fan: Data curation. Tao Yin: Data curation. Weiwei Zou: Data curation. Lei Zhang: Acquisition, Analysis. Jie Zhang: Acquisition, Analysis. Zhaolian Wei: Acquisition. Yunxia Cao: Supervision, Conceptualization. Chunmei Liang: Supervision, Funding acquisition. Dongmei Ji: Design of the work, Funding acquisition.

Funding

This work was supported by the National Natural Science Foundation of China (NSCF-82173532; NSCF-82471896), Anhui Provincial Universities Fund for Distinguished Young Scholars(2022AH020072), Research Fund of the Institute of Health Research, Hefei Comprehensive National Science Center (JKS2023015).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval for this study was obtained from the Ethical Committee of Anhui Medical University (No. 20160270). The study was conducted in accordance with the ethical standards of the Declaration of Helsinki and its later amendments. Given the retrospective nature of the study and the use of de-identified patient data, the requirement for informed consent was waived.

Competing interests

The authors declare no competing interests.