Effect of the Timing of Frozen Embryo Transfer on Pregnancy Outcomes in High Ovarian Response Patients Undergoing Freeze-All Strategy

Background: Frozen embryo transfer (FET) can greatly improve the pregnancy outcomes for high ovarian response (HOR) population. However, it is not known whether the impaired endometrial receptivity derived from controlled ovarian hyperstimulation (COH) can be fully recovered in the rst menstrual cycle after oocyte retrieval, and whether the timing of FET is a risk factor on pregnancy outcomes in HOR population undergoing freeze-all strategy. Methods: A retrospective cohort study to compare the pregnancy outcomes of the immediate and delayed FET groups in HOR population undergoing freeze-all strategy. Propensity score matching was used to make the potential risk factors of the immediate and delayed FET groups comparable. Multivariable regression analysis was used to study the effect of the timing of FET on pregnancy outcomes in the entire cohort and propensity score-matched cohort, even in different COH protocol cohorts as subgroup analysis. Results: We showed that the immediate FET group were no worse than delayed FET group in the entire cohort [clinical pregnancy rate (CPR), adjusted odd ratio (OR), 0.942, 95% condence interval (CI), 0.784-1.133; spontaneous abortion rate (SAR), adjusted OR, 1.118, 95% Cl (0.771-1.623); live birth rate (LBR), adjusted OR, 1.060, 95% Cl (0.886-1.267)]. The same results were obtained by χ 2 test in the propensity score-matched cohort (CPR, 60.5% versus 63.5%; SAR, 11.6% versus 12.3%; LBR, 48% versus 49.3%) (P > 0.05). Subgroup analysis indicated that pregnancy outcomes of immediate FET were non-inferior to delayed FET in short-acting gonadotropin-releasing hormone agonist (GnRH-a) long protocol (P > 0.05). The SAR of the immediate FET group were lower than that of the delayed FET group in GnRH antagonist protocol (adjusted OR, 0.646, 95% CI, 0.432-0.966) and insemination method, number of embryo transferred, embryo stage and cause of infertility.


Background
Controlled ovarian hyperstimulation (COH) is the key step in in vitro fertilization and embryo transfer (IVF-ET). High ovarian response (HOR) refers to the abnormal sensitivity of the ovary to gonadotropin, which leads to simultaneous development of multiple follicles and increases the risk of ovarian hyperstimulation syndrome (OHSS) [1]. Supraphysiological steroid hormones during COH affect endometrial receptivity by changing the endometrial immune environment and gene expression [2][3][4], resulting in poor pregnancy outcomes [5]. Adopting the freeze-all strategy in HOR patients can greatly reduce the risk of OHSS and avoid the in uence of COH on endometrial receptivity [6]. An increasing number of studies have con rmed that frozen embryo transfer (FET) has better pregnancy and perinatal outcomes than fresh embryo transfer [7][8][9].
However, the best time to perform FET following COH in HOR patients is controversial in clinical work.
Postponement of FET would increase the anxiety of patients [10]; in the immediate FET cycle, poor endometrial receptivity or physical condition may not be fully recovered to the pre-stimulation state, which may affect pregnancy outcomes [11]. It is not clear whether the detrimental effects on endometrial receptivity caused by COH could sustain over a long period of time, up to the subsequent menstrual cycle, especially in patients with HOR who are most affected by COH. Moreover, the use of different gonadotropin-releasing hormone (GnRH) analogues in the process of COH disturb the reproductive endocrine physiology by inhibiting pituitary function in different ways and degrees [12], and it is controversial whether the timing of FET affects pregnancy outcomes in different COH protocols [11,13].
Thus, this study aimed to investigate whether the FET timing affect the pregnancy outcomes in patients with HOR undergoing freeze-all strategy, and whether different COH protocols affect pregnancy outcomes in the subsequent FET cycle, and to provide reference results for the HOR population to choose the optimal time to start FET.

Study population and design
We conducted a retrospective cohort study including all patients from January 2015 to March 2019 at our reproductive medicine center and the study was conducted in accordance with ethical standards (2020PS006F), informed consent was obtained from all subjects. The inclusion criteria were as follows: (1) patients on their rst IVF or intracytoplasmic sperm injection (ICSI) cycle who were diagnosed with HOR and adopted freeze-all strategy, and had at least one embryo that could be used for FET. The diagnosis criteria of HOR is more than 5000 pg/ml estradiol on human chorionic gonadotropin (HCG) day or more than 15 oocytes retrieved [14,15]; (2) we only selected three different GnRH analogues stimulation protocols, including short-acting GnRH agonist (GnRH-a) long protocol, long-acting GnRH-a long protocol, and GnRH antagonist (GnRH-ant) protocol; (3) women aged 20-45 years old; and (4) hormone replacement therapy (HRT) for endometrial preparation in FET cycle. The exclusion criteria were as follows: (1) presence of uterine abnormalities; (2) patients with endometriosis and adenomyosis; (3) presence of autoimmune, endocrine and metabolic diseases; (4) previous diagnosis of uterine adhesion; (5) patients with chromosomal abnormalities; (6) patients who underwent blastocyst biopsy for preimplantation genetic testing; (7) patients using frozen donor semen; (8)patients using long-acting GnRH-a as pretreatment for FET after freeze-all cycle; and (9) patients with ectopic pregnancy as pregnancy outcome.
Patients were divided into immediate FET group and delayed FET group, which were de ned as that the FET took place either within the rst menstrual cycle following oocyte retrieval or afterwards.
Ovarian stimulation protocol, endometrial preparation protocol, and luteal support According to age, anti-Müllerian hormone, body mass index (BMI), number of antral follicles in bilateral ovaries, and prior response to stimulation, we can predict the HOR population and determine the initial dose of gonadotropin to prevent the occurrence of OHSS [16,17]. All patients were treated with the following three COH protocols: the short-acting GnRH-a long protocol involved daily subcutaneous injection of 0.05 mg short-acting GnRH-a triptorelin (Diphereline, 0.1 mg, IPSEN, France) at the middle luteal phase of the menstrual cycle as pituitary down-regulation for 14 days, introducing gonadotropin at the subsequent menstruation; the long-acting GnRH-a long protocol involved injection of a quarter to full dose (0.75-3.75 mg) of long-acting GnRH-a (Diphereline, 3.75 mg, IPSEN, France) intramuscularly in a single dose on the second day of menstruation, gonadotropin was given 20-30 days later when the follicle diameter reached 3-5 mm; the exible GnRH-ant protocol involved starting gonadotropin on the second day of menstruation, and GnRH antagonist (Cetrotide, Merck Serono, France) was added when the lead follicle reached 13-14 mm in diameter or the estradiol was more than 300 pg/ml. Follicles development were detected by transvaginal ultrasonography, and the dosage of gonadotropin was adjusted according to different ovarian responses.
When the follicles reached more than 17 mm in diameter by mean, trigger was performed for nal oocyte maturation. HCG or triptorelin was used alone or in combination in the GnRH-ant protocol. Only HCG was used for trigger in GnRH-a long protocol. Oocyte retrieval was performed 36 hours after triggering by transvaginal ultrasound-guided aspiration.
Hormone replacement therapy (HRT) was used for endometrial preparation, 4-8 mg of estradiol valerate (Progynova, Bayer, Germany) orally for at least ten days from the 3rd to 5th day of menstruation to promote the growth of endometrium. Ultrasonic examination should be completed before medication, when the thickness of endometrium is less than 6 mm, the drug can be used, otherwise the FET in this cycle will be cancelled. The luteal phase was supported by vaginal progesterone gel (Crinone, Fleet Laboratories Ltd., UK) 90 mg per day for vaginal administration, and estradiol was maintained at the original dose. Luteal support continued to use until 11 weeks of gestational age.

Statistical analysis
The outcomes of our study were clinical pregnancy rate (CPR), spontaneous abortion rate (SAR), and live birth rate (LBR). Clinical pregnancy was de ned as detection of a gestational sac on ultrasound image at 7 weeks of gestational age [18]. In China, spontaneous abortion was de ned as loss of pregnancy spontaneously after clinical pregnancy and before 28 weeks of gestational age, and live birth was de ned as the survival delivery after 28 weeks of gestational age.
As an observational study, multiple maternal and IVF characteristics were considered as potential risk factors that could moderate pregnancy outcomes, and the potential risk factors between the immediate and delayed FET groups were actual unbalanced distribution (Table 1). Thus, we used propensity score matching to make the potential risk factors between the immediate and delayed FET groups balanced and comparable. We used 1:1 nearest-neighbor matching without replacement to compare the variables and tried the match tolerance value from 1 to 0 until P values of the variable between the two groups were 1.000. χ 2 test was performed for comparison of the categorical variables and the pregnancy outcomes of the immediate and delayed FET group in propensity score-matched cohort ( Table 1).
As the effect of FET timing on different COH protocols were controversial, subgroup analysis were performed. Multivariable logistic regression models were calculated on each COH cohort, with the timing of FET as the main exposure of interest. Potential risk factors entered into the multivariable regression model were those that showed clinical relevance or showed a univariate relationship with pregnancy outcomes. The included variables were carefully selected based on the number of events available to ensure the stability of the regression equation. Adjusted odds ratio (OR) and their 95% con dence interval (CI) were calculated to analyze the independent effect of immediate and delayed FET on the pregnancy outcomes. P < 0.05 was considered as statistically signi cant. All statistical analyses were performed using IBM SPSS 26.0 (IBM Corp., Armonk, NY, USA).

Results
Potential risk factors between immediate and delayed FET groups in the entire and propensity scorematched cohort A total of 2128 HOR patients adopting freeze-all strategy underwent their rst IVF/ICSI cycle (Fig. 1).
There are 1130 patients in the immediate FET group and 998 patients in the delayed FET group. Patients' and IVF characteristics in the immediate and delayed FET groups, which as potential risk factors, were presented in Table 1. Before matching, the distribution of these risk factors were not absolutely balanced. The distribution of COH protocol, number of embryo transferred, embryo stage and multiple pregnancies were signi cantly different between the two groups (P < 0.05). No signi cant differences were found in maternal age, body mass index (BMI), insemination method, and infertility causes (P > 0.05). We obtained 1366 patients by propensity score matching, and all potential risk factors and the pregnancy outcome of multiple pregnancies among them are balanced and comparable (Table 1).  Immediate FET versus delayed FET cycles on pregnancy outcomes in propensity score-matched cohort The CPR (60.5% versus 63.5%), SAR (11.6% versus 12.3%), LBR (48.0% versus 49.3%) had no signi cant differences between the immediate and delayed FET groups in the propensity score-matched cohort (P > 0.05), which were detailed in Table 3.    "-" means that the number of miscarriages in the group over 38 years old is 0.

Discussion
In clinical work, when to start the FET cycle after COH is controversial, especially for patients with HOR, who are most affected by COH. As known, in the COH process, increased levels of supraphysiological steroid hormones and premature progesterone affected the gene expression and immune environment of the endometrium, which altered the embryo-endometrium asynchrony and negatively affected endometrial receptivity, reducing the CPR and LBR [5,19]. Moreover, the secretory activity factors produced by the residual luteal cysts derived from COH would last longer after oocyte retrieval in HOR patients. However, no study has investigated the speci c duration of these adverse effects. To avoid this concern, some clinicians recommended the conservative scheme to start the FET cycle at the second or third withdrawal bleeding after oocyte retrieval, which would undoubtedly prolong the IVF treatment period and increase the anxiety of patients who had experienced infertility for many years, increasing their mental and economic loses and affecting their pregnancy outcomes [10,20]. In our retrospective study, we only selected the HOR population who adopted HRT as endometrial preparation protocol in the past 5 years, and showed no signi cant differences on CPR, SAR, and LBR between the immediate and delayed FET groups. We believe that it is not accurate to assume that COH will still affect endometrial receptivity in the rst withdrawal bleeding cycle after oocyte retrieval and that the HOR population do not have to wait for several menstrual cycles to FET after freeze-all strategy.
In the process of COH, different GnRH analogues have different degrees and properties of inhibition effects on hypothalamic-pituitary-ovarian axis, it also has different effects on corpus luteum [21], which might have impact on endometrial receptivity and pregnancy outcomes [22]. A retrospective study have showed that immediate FET had similar CPR to delayed FET in patients with GnRH-ant protocol [13], which is in agreement with our subgroup results. However, our results are contrary to a population-based study on short-acting GnRH-a long protocol, which found that delayed FET was better for pregnancy outcomes, they believed that the initial are up effect of short acting GnRH-a during the down regulation period caused an early rise of progesterone, which affected the outcomes in immediate FET cycle [11]. An important limitation of that study is the small sample size, 67 patients in immediate FET group and 62 in delayed FET group. In this study, 1000 patients with short acting GnRH-a long protocol were studied (434 in immediate FET group and 566 in delayed FET group), and we found that the timing of FET did not affect pregnancy outcomes in short-acting GnRH-a long protocol. However, in the GnRH-ant and longacting GnRH-a long protocol, we found that the SAR in delayed FET group was signi cantly higher than that in immediate FET group. Among all the COH protocols, the GnRH-ant protocol has the shortest treatment period, which oocyte retrieval take place after 8-10 days of ovarian stimulation by mean; the short-acting GnRH-a long protocol needs 14 days of down regulation on that basis; while the long-acting GnRH-a long protocol needs a down regulation for more than 20 days. The immediate FET cycle in GnRHant protocol can obtain the shortest treatment period, while the delayed FET cycle in long-acting GnRH-a long protocol has the longest treatment period. We know that psychological factors are important factors leading to infertility and spontaneous abortion, its potential impact on neuroendocrine and immune changes could affect early pregnancy risk [23,24]. We consider that the longer the treatment period, the more anxious the patients are, which could be the cause of the increase in the SAR. Moreover, long acting GnRH-a can effectively improve endometrial receptivity [22], which may be the other reason for the lower SAR in the immediate FET group in long acting GnRH-a long protocol. It is worth mentioning that one research have shown that residual luteal cysts may increase the expression of relaxin in circulation [25], which is related to the endometrial angiogenesis and prevent recurrent abortion [26]. Therefore, the effects of residual luteal cysts in immediate FET cycles that we had previously worried about may be bene cial to endometrial receptivity and pregnancy outcomes.
The limitation of this study lies in the retrospective nature as well as the possibility of unmeasured confounding factors such as smoking habits and alcohol consumption. Although we have obtained many cases with HOR and made the immediate and delayed FET groups comparable by propensity score matching, we have lost 762 cases without successful matching in this process. However, we do not know whether these cases will affect the actual situation. Notably, the pregnancy outcomes we have studied were not the only endpoint, other obstetric outcomes and neonatal outcomes should also be noticed.
In summary, this study implied that immediate FET may not affect the pregnancy outcomes in HOR population undergoing freeze-all strategy although their pelvic environment were not fully recovered.
Intended to delay FET in HOR population may increase anxiety of infertility couples and increase the SAR. Clinicians and patients with HOR do not have to worry about FET on the immediate cycle, and they can arrange the start time of FET cycle on their own convenience and desire, making the IVF treatment process more relax and e cient. However, more studies are needed to further investigate the effect of FET timing on other obstetric and neonatal outcomes in the HOR population.