This is a retrospective, single-center, cohort study that includes IVF–intracytoplasmic sperm injection (ICSI) cycles (n = 1,077) in the National Taiwan University Hospital from May 2012 to December 2022. All cycles were only triggered with GnRH agonist, and all embryos were frozen to prevent OHSS. Patients with hypogonadotropic hypogonadism were excluded. Women with prior long-term oral contraceptive pill (OCP) usage before starting the COS cycle were also excluded.

This study was approved by the Institutional Review Board (IRB) of the National Taiwan University Hospital (Taipei City, Taiwan; serial number 202112224RINA). The IRB waived the requirement for informed consent owing to the retrospective nature of the study.

Ovarian stimulation was started on day 2 of the menstrual cycle with 100–300 IU/day of recombinant follicle-stimulating hormone (FSH) (rFSH: Gonal-F, Merck Pharmaceuticals, Darmstadt, Germany; Elonva, Merck Sharp & Dohme, Rahway, NJ, USA) or highly purified human menopausal gonadotropin (HMG) (HPHMG; Menopur, Ferring Pharmaceuticals, Saint-Prex, Switzerland). The dosage was adjusted according to the response of women undergoing COS by serial transvaginal ultrasound monitoring and blood estradiol level, which has been published previously (13, 14). Pituitary suppression was achieved with a daily 0.25 mg GnRH antagonist injection of ganirelix (Orgalutran, Merck Pharmaceuticals) starting on day 6 of ovarian stimulation or medroxyprogesterone acetate (MPA) 10 mg/day from the start of ovarian stimulation (3, 15). When at least three follicles of 17 mm or more were visible, triggering of final oocyte maturation was carried out with a single subcutaneous dose of GnRH agonist, 0.2 mg of triptorelin (Decapeptyl, Ferring Pharmaceuticals). Oocyte retrieval using the flushing method was performed at 35 to 36 h following the trigger. All follicles with a mean diameter of >12 mm were aspirated.

After oocyte pick-up (OPU), fertilization of the oocytes was performed by all IVF, 1/2 ICSI, or all ICSI, depending on semen parameters and physician preference. The freeze-all strategy was performed for all cycles.

In all agonist-triggered cycles, blastocysts were vitrified on day 5 or 6. The vitrification/thawing protocol was Cryotop method (Kitayazo, Japan) based on the method described by Kuwayama (16). Subsequently, thawed embryo transfer was scheduled based on the physician’s preference and the patient’s condition, using either a natural ovulatory cycle or a hormonal replacement cycle. For the hormone replacement cycle, endometrial preparation was initiated by stepwise increasing doses of oral estradiol valerate (Estrade 2 mg, Synmosa, Hsinchu, Taiwan) followed by the addition of 90 mg of vaginal progesterone (Crinone 8%; Merck Serono, Hertfordshire, UK) twice daily after at least 10–14 days of estrogen administration, as described previously (15). Blastocyst transfer was performed on the sixth day of progesterone administration. Clinical pregnancy was defined as the presence of fetal cardiac activity by transvaginal ultrasound at 7 weeks of gestation. Ongoing pregnancy was defined as progression beyond 12 weeks of gestation. Live birth rate was defined as the delivery of a live child after 24 weeks of gestation. The calculation of clinical pregnancy, ongoing pregnancy, and live birth rates were specifically focused on the first thawed embryo transfer cycle, where the highest-graded embryo was transferred.

Several studies have indicated that a low basal LH level is a highly valuable marker for identifying suboptimal responders. Lu et al. (7) found that a basal LH level below 2.27 IU/L was the most significant marker for identifying suboptimal responders. Furthermore, Popovic-Todorovic et al. (17) reported that basal LH levels below 2 IU/L are associated with a 25.2% risk of suboptimal response.

To figure out whether LH max during COS plays a role in evaluating the optimal oocyte yield in the GnRH agonist trigger cycle, subjects were divided into three groups, depending on basal LH level (10th percentile in our cohort, 2.2 IU/L) and LH max level.

Group 1 included a cycle where basal LH level and LH max were both ≤2.2 IU/L. Group 2 included a cycle where basal LH level ≤2.2 IU/L and LH max level >2.2 IU/L. Group 3 included a cycle where basal LH level and LH max were both >2.2 IU/L. The schematic diagram is shown in Figure 1 .

Serum estradiol, progesterone, FSH, and LH concentrations were measured for all patients using chemiluminescence assays (IMMULITE^® 2000; Siemens Healthineers, East Walpole, MA, USA) in every visit during ovarian stimulation, including day 2, day 6, day 8, and trigger day. Intra-assay and inter-assay coefficients of variation were 3.6% and 4.3%, respectively, for FSH; 4.8% and 10.7%, respectively, for LH; 6.7% and 9.7%, respectively, for E2; and 9.7% and 12.2%, respectively, for P4.

The primary outcome measure was oocyte yield rate, defined as the ratio of the total number of collected oocytes to the number of follicles (diameter > 12 mm) on the day of the trigger. The secondary outcome measures include the number of collected oocytes, suboptimal response to GnRH agonist trigger (defined as an oocyte yield below the <5th percentile (60.7% in our study)), oocyte maturity rate (the ratio of metaphase II oocytes to the number of denuded oocytes), fertilized rate (the ratio of normally fertilized oocytes (two distinct pronuclei (2PNs)) to the number of oocytes used for fertilization), clinical pregnancy rate (detection of a gestational sac during ultrasound), ongoing pregnancy rate (presence of fetal cardiac activity by transvaginal ultrasound at 9 weeks of gestation), and live birth rate (delivery of a live child after 24 weeks of gestation). The pregnancy outcomes were calculated for the first FET cycle in which the highest-graded embryo was transferred.

SAS Version 9.4 (SAS Institute Inc.) was used for data analysis. For normally distributed data, the mean ± standard deviation was calculated for each variable. Non-normally distributed continuous data were reported as medians and interquartile ranges. Comparisons among groups were analyzed by analysis of variance (ANOVA) followed by an appropriate post-hoc test. Categorical variables were presented as frequencies, using percentages.

Multivariable logistic and linear regression analyses were conducted for adjustment of covariates to evaluate the suboptimal response, oocyte yield rate, mature oocyte rate, total collected oocyte, and clinical pregnancy. A p-value of less than 0.05 was statistically significant.

Baseline characteristics, hormone profiles, and laboratory parameters are shown in Tables 1 , 2 . The mean age, BMI, and anti-Mullerian hormone (AMH) were 34.8 ± 4.3 years, 22.1 ± 3.5 kg/m², and 6.2 ± 4.1 ng/ml, respectively. The mean serum FSH and LH levels at the start of stimulation were 6.6 ± 2.8 and 5.3 ± 2.7 IU/L, respectively ( Table 2 ).

Regarding the indication of enrolled participants for COS in this study, oocyte freezing accounted for 190 cycles (19.4%) and male infertility for 178 cycles (18.1%), while preimplantation genetic screening (PGS) was performed in 48 cycles (4.9%). The remaining indications for COS are presented in Table 1 .

On average, patients had 24.5 ± 10.2 follicles >12 mm in diameter on the day of trigger, and the mean number of retrieved oocytes was 23.7 ± 9.9. Oocyte maturity was calculated only for patients undergoing ICSI or 1/2 ICSI. The mean oocyte yield rate, maturity, and fertilization rate were 89.4% ± 13.7%, 83.8% ± 13.3%, and 64.7% ± 17.9%, respectively. Thirty-nine patients exhibited a suboptimal response to GnRH agonist trigger, defined as an oocyte yield rate below the <5th percentile (60.7% in our study). For the first frozen embryo following embryo freezing, clinical pregnancy, ongoing pregnancy, and live birth rates were 42.7%, 42.2%, and 37.9%, respectively ( Table 2 ).

Following stratification by basal LH and LH max, 17 women (1.8%) had an LH level ≤2.2 IU/L throughout the entire stimulation cycle before the trigger, 31 women (3.2%) had basal LH ≤2.2 IU/L with LH max >2.2 IU/L, and 933 women (95.1%) had an LH level >2.2 IU/L during the entire stimulation cycle before the trigger ( Table 3 ).

Women with LH levels ≤2.2 IU/L during the whole stimulation cycle (Group 1) had the youngest age at retrieval and the lowest basal FSH level (p < 0.05), but no significant differences were observed in AMH (6.1 ± 3.7 vs. 5.2 ± 3.7 vs. 6.2 ± 4.2), basal estradiol level (49.4 ± 59.4 vs. 45.1 ± 27.6 vs. 40.6 ± 22.8), or progesterone level (0.50 ± 0.28 vs. 0.51 ± 0.27 vs. 0.42 ± 0.31) among the three groups.

Group 1 had a significantly lower oocyte yield rate compared to the other two groups. However, oocyte yield rates did not differ significantly between Group 2 and Group 3. The suboptimal response rate was significantly higher in Group 1 (23.5%) compared to Group 2 (6.5%) and Group 3 (3.6%).

There were no significant differences in the COS protocols, maturity rate, fertilization rate, fertilizing method, clinical pregnancy rate, ongoing pregnancy rate, and live birth rate among the groups, and the OHSS rate did not differ significantly among the three groups.

We conducted a multiple regression analysis to further investigate the association between the three groups and the impact on oocyte yield rate, maturity rate, and fertilization rate. After adjustment for all potential explanatory variables, such as age at retrieval, BMI, basal FSH, and E2 level, Group 1 (b = −12.723; 95% CI, −20.9 to −4.55; p = 0.002) and BMI (b = −0.254; 95% CI, −0.5 to −0.004; p = 0.46) were still significantly negatively correlated with oocyte yield rate compared to Group 2. As shown in Table 4 , none of the groups were associated with the number of retrieved oocytes. Only age at retrieval (b = −0.583; 95% CI, −0.72 to −0.44; p < 0.0001) and basal FSH level (b = −0.709; 95% CI, −0.92 to −0.49; p < 0.0001) remained significantly associated with the number of retrieved oocytes in the multiple regression model. The maturity rate and fertilization rate did not differ between the three groups after adjusting for all confounding variables.

We utilized multiple logistic regression to compare the effects among three groups and investigate the impact of LH max and other critical variables on the suboptimal response of oocyte pick-up, clinical pregnancy, ongoing pregnancy, and live birth rates. After adjustment for all variables that may be related to suboptimal response, Group 1 still had a significantly higher risk of suboptimal response when compared to Group 2 (OR: 6.99; 95% CI, 1.035–47.274; p = 0.046). Groups 2 and 3 continued to have no difference in any clinical outcomes Table 5 .

These results demonstrate that in patients undergoing GnRH antagonist or PPOS protocol with GnRH agonist trigger only, LH max level and BMI should be considered to achieve optimal oocyte yield outcomes, but they are not associated with clinical pregnancy, ongoing pregnancy, and live birth rates.

Suboptimal response to GnRH agonist trigger mainly caused by suboptimal trigger was a frustrating phenomenon during COS, leading to low oocyte yield or even empty follicle syndrome. Recently, one systemic review reveals that the incidence of empty follicle syndrome and suboptimal response (post-trigger LH <15 IU/L) in cycles triggered with GnRH agonists is approximately 0.5%–3.4% (17, 19, 20). In our study, a suboptimal response is 3.9% defined as an oocyte yield rate <60% (the lower 5th percentile of overall oocyte yield rate), which is comparable to that reported in other studies (18, 20). The risk factors for suboptimal response to GnRH agonist trigger had been reported as low or high BMI, low basal LH level or low trigger day LH level, prolonged stimulation, higher gonadotropins dose, long-term use of OCP, and hypogonadotropic hypogonadism (20). However, the determination of GnRH agonist or a dual trigger remains elusive.

Several studies attempted to explore the efficacy of GnRH agonist trigger, which focused on baseline LH or trigger day LH (7, 10, 17–19, 21). Chen et al. (18) reported that serum LH level <15 IU/L at 12 h post-trigger with GnRH agonist was associated with a dramatically lower oocyte yield (38.3%), compared to those with >15 IU/L (62.5%). Chang et al. (21) disclosed that patients with a low baseline LH <1 IU/L had a 13.3% risk of failure to GnRH agonist trigger, while that risk was reduced to 1.8% if the LH level was ≥2 IU/L. A lower LH level on the trigger day was also associated with an increased incidence of post-trigger serum LH <15 IU/L (10, 19). A low post-trigger LH level or low oocyte yield from the GnRH agonist trigger could be compensated by additional exogenous hCG. Meyer et al. (10) suggested a dual trigger for patients with low LH (<0.5 IU/L) on the day of the trigger. However, frequent administrations of hCG in the GnRH triggering cycles may increase the risk of OHSS.

Recent studies indicated that basal LH level was a better predictor for the efficacy of the GnRH agonist trigger than the trigger day LH (7, 17). Lu et al. (7) found that basal LH <0.5, 1.0, 1.5, 2.0, 2.5, 3.0, and 4.0 IU/L was associated with a 45.4%, 29.3%, 18.3%, 10.2%, 6.7%, 4.98%, and 3.5% of the risk for suboptimal response, respectively. They indicated that a cutoff basal LH level (<2.27 IU/L) was the single most valuable predictor (with area under the curve of 0.805) of a suboptimal response to the GnRH agonist trigger. They proposed a dual trigger for those patients with a cutoff basal LH level <2.27 IU/L. In our study, we defined the cutoff value by the 10th lower percentile of all basal LH levels, which was 2.2 IU/L, which was similar to the cutoff value by Lu et al. (7). We suggest maximum LH level during COS, representative of endogenous LH reserve, as a reference to determine agonist trigger or a dual trigger. Whether this strategy will achieve optimal oocyte yield and reduce OHSS deserves more prospective studies.

In our center, oocyte pick-up is routinely performed by junior fellows using follicle flushing and curetting. The oocyte yield rate is also a criterion of junior fellow evaluations. The value of follicular flush in oocyte retrieval remains controversial (22, 23). Some investigators demonstrated that follicular flush increased oocyte yield (22, 24). However, some investigators found that follicular flushing failed to increase the oocyte yield rate (23). In addition, follicular curettage during oocyte retrieval may increase oocyte yield without sacrificing oocyte quality (25, 26). The average oocyte yield rate (89.4%) in the present study was higher than that of previous other reports (68.26%) (7). These may partly explain our relatively high oocyte yield rate in GnRH agonist-trigger patients. The value of flushing and curettage in the GnRH agonist trigger cycles may deserve further investigation.

The number of cases in both Group 1 and Group 2 was relatively low in our cohort. In our center, it is not our routine practice to administer OCPs for withdrawal of bleeding prior to COS. Additionally, long-term OCP users were excluded from our study. The use of oral contraceptive pills as pretreatment prior to COS appeared to be more common in a previous article by Meyer et al. (2019). It is worth noting that the use of OCP may result in lower basal LH levels.

In our retrospective study, in order to account for confounding factors, we performed a multiple regression analysis to examine the relationship between the three groups (taking into account the impact of LH max) and their effects on oocyte yield rate, maturity rate, and fertilization rate.

The primary weakness of our study is its retrospective design, which introduced heterogeneity and confounders due to the varied implementation of stimulation protocols by different providers. To better evaluate the LH max, a randomized prospective trial comparing Group 1, Group 2, and Group 3 should be conducted in the future.

The post-trigger LH level was not routinely measured in our center. Our primary outcome was essential for oocyte yield rate, but no sufficient data on post-trigger LH levels were available. Conversely, serum LH profile was obtained in every revisit during ovarian stimulation and could be more comprehensive to evaluate endogenous LH reserve. Furthermore, most subjects did not undergo preimplantation genetic testing (PGT), which would control for aneuploidy, and the embryo transfer number and quality were not reported in this study.

Future research should investigate the relationship between post-trigger LH levels and LH max levels. Additional data are necessary to determine a definitive serum LH max cutoff level for optimal GnRH agonist triggering, which may vary based on a patient’s baseline LH level and BMI. It would also be valuable to analyze a cohort of patients with a euploid embryo confirmed by PGS.