This was a retrospective cohort study performed at the Affiliated Hospital of Kunming University of Science and Technology, Kunming, China, with approval from the Ethics Committee of the First People’s Hospital of Yunnan Province (number: KHLL2020-KY013). Data were collected from patients who visited our hospital from January 1, 2015, to December 31, 2018. Data were de-identified to ensure patient privacy.

To assess the impact of LH on the general population of women, the inclusion criteria for patient enrollment included the following: 1), Normogonadotrophic patients with basal FSH less than 10 mIU/mL on the second day of menstruation received their first cycle of IVF-ET treatment; 2), the pituitary suppression protocol included a long-acting GnRH-agonist; and 3), pituitary suppression was only applied starting at the mid-luteal phase or the early follicular phase. Exclusion criteria included endometriosis, polycystic ovary syndrome (PCOS), family genetic diseases or karyotype abnormalities, immune factors, female reproductive tract malformations, uterine fibroids, no embryo transfer occurred (no oocytes retrieved or no available embryos, available embryos but none transferred), and serious bleeding or complications.

All patients were treated with a GnRH agonist protocol that has been used and described previously (14). Briefly, the long-acting GnRH-agonist triptorelin (Diphereline, Ipsen, Signes, France) was used on the second day of the menstruation or in the middle of the luteal phase (3.75 mg administered for the follicular phase and 1.25 mg for the luteal phase). The dosage of agonist was accordance with the guideline. When pituitary suppression was completed after 14–21 days, serum FSH, LH, E2, and progesterone levels were evaluated. The endometrial thickness was measured by ultrasonography. When the patient had E2 < 50 pg/mL, endometrial thickness < 5 mm, and absence of cysts, a fixed dose of 150–225 IU of exogenous gonadotropin (Gn) (rFSH, Gonal-F, Merck-Serono, Aubonne, Switzerland) was administrated for ovarian stimulation. Ultrasonography was used to monitor follicular growth during the COS process. Based on the patient’s reproductive status, the amount of Gn was adjusted appropriately. When the largest diameter of 2 dominant follicles was ≥ 18 mm, serum levels of LH, E2, and P were determined and 5000 IU of exogenous hCG was injected on the same day. This was followed by ultrasound-guided oocyte retrieval 34–36 hours later. All serum hormone tests were completed using a UniCel DxI 800 Access Immunoassay System (Beckman-Coulter, Brea, CA, USA).

Conventional IVF or ICSI was selected according to male semen assessments. Pronuclear formation of the zygote was observed on the first day after oocyte retrieval to evaluate the normal fertilization rate. The number of available embryos and the proportion of high-quality (defined with 2PN zygote on the 1^st day, 6-8 blastomeres on 3^rd day, and cell debris less than 20%) embryos were evaluated. When the number of available embryos was > 6, blastocyst culture and fresh-cleavage embryo transfer were performed. However, for patients with high ovarian response or insufficient endometrial thickness, or who actively requested frozen embryo transfer (FET), fresh embryo transfer (ET) was canceled, and the cleavage embryos were frozen, and FET was performed in the follow-up cycle. Luteal support was initiated immediately after oocyte retrieval. A biochemical pregnancy test was performed on the 14^th days after ET. If the biochemical pregnancy test was positive, luteal support was sustained until the 8^th weekend.

Patient cycle information was retrieved through review of medical records. Based on the serum concentration of LH on the COS initiation day and tertile of population distribution, patients were divided into 3 groups: low LH (<1 mIU/mL), medium LH (1–1.49 mIU/mL), and high LH (≥1.5 mIU/mL). Primary outcomes were early pregnancy loss rate (ePLR) and live-birth rate (LBR); and secondary outcomes were clinical pregnancy rate (CPR), late pregnancy loss rate (lPLR) and ectopic pregnancy rate (EPR). Biochemical pregnancy was defined by serum β-hCG test positive, and clinical pregnancy was defined by presenting gestational sac using ultrasound scan after 6 weeks of ET. A miscarriage within the first trimester was defined as early pregnancy loss, and a miscarriage after 12 weeks was defined as late pregnancy loss. The PLR refers to the ratio of the number of pregnancy loss cycles to the number of clinical pregnancy cycles, and the CPR, EPR and LBR refers to ratio of corresponding cycles to the number of transfer cycles. The cumulative rate was defined as the proportion of an event that occurred at least once during the entire ovarian stimulation cycle including all fresh and frozen transfer cycles.

All statistical analyses were conducted in SPSS (version 26.0, Armonk, NY, USA). A P < 0.05 in a 2-tailed test was considered statistically significant. We compared the CPR, ePLR, lPLR, EPR, and LBR in three LH groups. For continuous variables, The Shapiro-Wilk’s test combined with P-P plots was used to evaluate normality. Normally distributed data are presented as mean ± SD and were compared using one-way ANOVA. Non-normally distributed data are presented as median and interquartile range (25^th–75^th percentiles) and were compared by the Kruskal-Wallis test. The Bonferroni method was used to correct the P value for multiple comparisons. The categorical variables were analyzed by the Chi-squared test, with the P value for multiple comparison adjusted using the Bonferroni method.

For the multivariate logistic regression, continuous variables, such as age, BMI, serum hormone concentration, endometrial thickness, and number of oocytes or embryos, were transformed into categorical variables based on recognized cut-off values and data distribution. All variables were included for analysis to create the final regression model using the forward likelihood ratio (LR) method. To analyze the correlation between LH on the initiation day and on the hCG trigger day, simple linear regression was used, and Pearson R² and the slope of the regression line equation were calculated. Odds ratios (OR) with 95% confidence intervals (CI) were reported.

Table 1 summarizes the clinical baseline data from the 3 levels of LH on the day of initiation. The number of oocytes retrieved, as well as the number of embryos available for ET in the low and medium LH groups, were significantly lower than those in the high LH group. There was no significant difference among the 3 LH groups with regards to the normal fertilization rate or the high-quality embryo rate. The high LH group had a significantly greater number of embryos available for ET than the low and medium LH groups. Although the median (25^th–75^th percentiles) of some variables was similar, their P value was still <0.05 due to the difference in distribution of large sample size. All variables were ultimately included in the subsequent multivariate binary logistic regression analysis.

We compared the clinical outcome data of fresh ET and the entire stimulation cycle ( Table 2 ). The results indicated that in the fresh ET cycle, the ePLRs in the low and medium LH groups were significantly higher than in the high LH group, and the LBRs in these groups were lower than in the high LH group. However, there was no difference in the ePLR and the LBR between the low and medium LH groups. For the entire ovarian stimulation cycle, a higher LH resulted in a higher rate of fresh ET cancelation, and the cumulative CPR and LBR of the low and medium LH groups were significantly lower than in the high LH group. All rates were calculated based on the clinical pregnancy.

In the multivariate regression model, the main variables affecting early pregnancy loss after fresh ET included age, LH on the day of COS initiation, endometrial thickness on the hCG trigger day, and the number of transferred embryos ( Table 3 ). The model included 3010 clinical pregnancy-positive cycles, 330 of which exhibited early pregnancy loss. Compared with the high LH group, the crude ORs of the low LH and the medium LH groups were 1.518 (95% CI 1.134–2.032, P = 0.005) and 1.436 (95% CI 1.079–1.910, P = 0.013), respectively, with a P value for the trend of 0.011. The adjusted ORs were 1.429 (95% CI 1.065–1.919, P = 0.018) and 1.389 (95% CI 1.041–1.853, P = 0.026), respectively; and the P value for the trend was 0.033. These results indicated that low LH on the COS initiation day might lead to a higher ePLR.

In terms of live-birth analysis for fresh ET, a total of 6279 fresh ET cycles were included, of which 2459 cycles resulted in live births. The univariate results of LH on the day of COS initiation showed that in comparison to the high LH group, the ORs of live birth in the low and the medium LH groups were 0.827 (95% CI 0.729–0.938, P = 0.003) and 0.841 (95% CI 0.744–0.949, P = 0.005), respectively, with a P value for the trend of 0.004. However, in the multivariate model ( Table 3 ). The LH level on the initiation day was excluded from the adjusted model and replaced with LH on the day of hCG trigger, which may be more likely to affect the final outcome of live birth. Thus, with respect to LH on the day of the hCG trigger, for the < 0.5 mIU/mL group, the crude ORs of the 0.5–0.79 mIU/mL and ≥0.8 mIU/mL groups were 1.288 (95% CI 1.125–1.475, P < 0.001) and 1.536 (95% CI 1.342–1.759, P < 0.001), respectively, with a P value for the trend of < 0.001. The adjusted ORs were 1.234 (95% CI 1.074–1.419, P = 0.003) and 1.476 (95% CI 1.284–1.697, P < 0.001), respectively, with a P value for the trend of < 0.001. The analysis results showed that on the day of hCG trigger of GnRH-agonist regimen, a higher serum concentration of LH was beneficial for live birth. The LH level on the day of COS initiation was not an independent factor that affected live births from fresh ET.

There was a linear correlation in the LH concentration between the initiation day and the hCG trigger day (Pearson R² = 0.174, P < 0.001; with a slope of 0.277, 95% CI 0.262–0.292) ( Figure 2 ), which implied that different fresh ET LBR in 3 initiation day LH groups might be caused by affecting the LH level on the day of hCG trigger.

To exclude the influence of advanced age, obesity, history of miscarriage, nulligravidity, poor ovarian function, and low-quality embryos, we performed a subgroup multivariate regression analysis on the fresh ET cycles ( Table 4 ). The results indicated that LH on the day of COS initiation was significantly associated with early pregnancy loss in women of non-advanced age (< 35 years of age), BMI < 25, no history of miscarriage and artificial abortion, history of gravidity, normal ovarian reserve (AFC >5, number of retrieved oocytes >3), and those who experienced the transfer of at least one high-quality embryo or of two embryos. Compared with the high-LH group, the adjusted OR ranges for the low and medium LH groups were 1.404–1.567 and 1.357–1.639, respectively. These results confirmed that the ePLR in the low and medium LH groups was still significantly higher than that in the high LH group in patients with a better reproductive history.

Since different LH levels on the day of initiation of COS correlated with different cumulative CPR and LBR, we analyzed the relationship between LH and them ( Table 5 ). The regression model for clinical pregnancy included 9540 cycles and 6731 cycles of cumulative clinical pregnancy. Univariate analysis showed that with respect to cumulative live births and compared with the high-LH group, the ORs for the low and medium LH groups were 0.825 (95% CI 0.739–0.921, P =0.001) and 0.864 (95% CI 0.777–0.960, P = 0.007), respectively. The regression model for live births included 9540 cycles, and there were 5809 cycles of cumulative clinical pregnancies. Univariate analysis showed that compared with the high LH group in cumulative live births, the ORs for the low and medium LH groups were 0.811 (95% CI 0.732–0.899, P <0.001) and 0.879 (95% CI 0.797–0.971, P = 0.011), respectively. However, multivariate regression results suggested that LH on the initiation day was not an independent variable affecting cumulative clinical pregnancy and live birth.