Cumulative live birth after unstimulated in vitro maturation versus conventional in vitro fertilization (IVF) in women with polycystic ovary syndrome: a 36-month retrospective matched cohort study

Abstract

Background

Polycystic ovary syndrome (PCOS) is characterized by heterogeneous ovarian responsiveness and a heightened risk of ovarian hyperstimulation syndrome (OHSS) during gonadotropin stimulation. Strictly unstimulated in vitro maturation (IVM) avoids ovarian stimulation and may reduce treatment burden; however, long-term cumulative live birth outcomes remain insufficiently defined, especially across baseline ovarian phenotypes.

Methods

This retrospective matched cohort study included women aged 20–42 years with PCOS, diagnosed by the Rotterdam criteria, treated at a university-affiliated fertility center between January 2010 and December 2022 (follow-up through December 2023). Women undergoing unstimulated IVM (n = 545) were matched 1:3 by age and BMI to women undergoing conventional in vitro fertilization (IVF) using a GnRH antagonist protocol (n = 1,635). The primary outcome was the 36-month cumulative live birth rate (CLBR) per initiated oocyte retrieval cycle. Secondary outcomes included implantation and live birth after the first transfer (fresh vs. frozen), OHSS, and obstetric/perinatal outcomes. Odds ratios (ORs) were calculated for the freeze-all subset.

Results

The 36-month CLBR per initiated retrieval cycle was significantly lower with IVM than IVF (22.20% vs. 47.95%; P < 0.001). CLBR/embryo transfer cycles was also lower with IVM (31.43% vs. 50.19%; P < 0.001). Implantation was lower in the IVM group after fresh transfer (22.94% vs. 29.45%; P = 0.005) but higher after frozen embryo transfer (43.86% vs. 32.62%; P = 0.007). Live birth after the first fresh transfer was lower with IVM (25.96% vs. 33.20%; P = 0.029), whereas after the first frozen transfer, live birth was comparable (33.33% vs. 36.61%; P = 0.486). No severe OHSS occurred in the IVM group, compared to 0.98% in IVF cycles. In women with AFC ≥ 24, live birth after the first frozen transfer was similar between groups (OR 0.99, 95% CI 0.57–1.72). Obstetric and neonatal outcomes were broadly comparable.

Conclusions

Unstimulated IVM in PCOS women yielded a lower 36-month CLBR compared to IVF, primarily due to lower embryo yield. IVM eliminated OHSS and achieved comparable live birth rates following the FET, particularly in women with high AFC. These findings advocate for a ‘freeze-all’ IVM strategy to optimize both safety and reproductive outcomes in this high-risk population.

Clinical trial number

Not applicable.

Introduction

Controlled ovarian hyperstimulation (COH) is central to assisted reproductive technology (ART), aiming to obtain an adequate number of mature oocytes to maximize cumulative live birth rates (CLBRs) [1]. However, gonadotropin stimulation increases the risk of ovarian hyperstimulation syndrome (OHSS), a concern that is particularly relevant in women with polycystic ovary syndrome (PCOS) [2]. In clinical practice, PCOS is characterized by heterogeneous ovarian responsiveness: intensive stimulation may precipitate OHSS, whereas conservative dosing can lead to suboptimal follicular recruitment or mono-follicular development, potentially compromising treatment efficiency [3, 4]. These challenges underscore the need for alternative approaches that reduce iatrogenic risk while maintaining reproductive efficacy. In vitro maturation (IVM) offers such an approach by retrieving immature cumulus–oocyte complexes (COCs) from unstimulated or minimally stimulated antral follicles, followed by maturation in vitro. By reducing or avoiding gonadotropin exposure, IVM may lower treatment burden and improve safety for women with PCOS [5].

Although successful human IVM leading to live birth was first reported in the 1990s, its clinical adoption has progressed slowly [6]. In 2021, the American Society for Reproductive Medicine (ASRM) recognized IVM as a non-experimental procedure [7]. Nevertheless, uptake remains limited, in part due to variable maturation and fertilization outcomes and the absence of universally standardized protocols [8]. A major contributor to heterogeneity is the evolving definition of IVM in clinical practice. While IVM is broadly defined as the use of oocytes retrieved before in vivo maturation, protocols differ substantially, particularly regarding the use of hormonal priming (e.g., FSH or hCG) versus strictly unstimulated approaches [9–11].

The clinical value of hormonal priming remains debated. While some studies suggest that FSH priming may improve oocyte yield, evidence regarding its clinical benefit is inconsistent. Similarly, hCG priming is controversial; although intended to trigger maturation, it may result in partial in vivo maturation, leading to a heterogeneous cohort of oocytes at retrieval [12]. Notably, a randomised trial demonstrated that adding FSH to hCG priming failed to improve maturation, fertilization, or pregnancy outcomes in women with PCOS [13]. Furthermore, the use of gonadotropins blurs the distinction between IVM and conventional IVF, raising concerns regarding reproducibility and obscuring the specific benefits of IVM technology itself [14].

To address these heterogeneity concerns, our center implements a strictly unstimulated IVM protocol, avoiding both FSH and hCG priming to standardize outcomes and eliminate iatrogenic OHSS risk. Building on our previous randomized controlled trial (RCT) [15, 16], which demonstrated comparable blastocyst implantation rates, the present study evaluates long-term clinical outcomes in a large cohort of 2,180 women. We specifically focus on the 36-month CLBR and stratify outcomes by antral follicle count (AFC) to identify the ideal patient phenotype for this strategy.

Materials and methods

Study design and study population

This retrospective cohort study was conducted at the Reproductive Center of Peking University Third Hospital. Data were analyzed for women undergoing ART treatment between January 1, 2010, and December 31, 2022, with clinical follow-up extending through December 2023. The cumulative observation period was defined as 36 months following the initiation of the oocyte retrieval cycle. The study protocol was approved by the Reproductive Medicine Ethics Committee of Peking University Third Hospital (Reference number: SZ-066). Written informed consent was obtained from all participants.

Participants and matching procedure

Eligible participants were women aged 20–42 years diagnosed with PCOS according to the Rotterdam consensus criteria [17]. The study group comprised 545 women who underwent unstimulated IVM cycles. To minimize selection bias, a control group was established using a 1:3 matching ratio. Controls were selected from women undergoing conventional IVF with a gonadotropin-releasing hormone (GnRH) antagonist protocol, matched strictly for female age and body mass index (BMI). IVM cases without suitable matches were excluded. Consequently, 1,635 matched IVF cycles were included for analysis.

Exclusion criteria

Exclusion criteria were applied uniformly to both groups: (i) cycles involving preimplantation genetic testing (PGT); (ii) abnormal chromosome karyotypes; (iii) oocyte donation cycles; (iv) congenital or acquired uterine abnormalities; (v) use of controlled ovarian stimulation protocols other than the standard GnRH antagonist protocol (e.g., GnRH agonist long protocols, minimal stimulation protocols); and (vi) patients who failed to return for the first embryo transfer (including fresh or frozen transfers).

Clinical procedures

Throughout the 12-year study period (2010–2022), laboratory protocols for Intracytoplasmic sperm injection (ICSI), embryo culture, and vitrification remained consistent, adhering to established Standard Operating Procedures (SOPs) within our ISO-certified center. Although minor updates to consumables were made, the core culture systems and vitrification kits remained stable.

Unstimulated IVM protocol

Women in the IVM group underwent a strictly unstimulated protocol without gonadotropin or hCG priming. Retrieval was scheduled when endometrial thickness reached ≥ 6 mm and the leading follicle remained < 10 mm. Based on our center’s data, the diameter of the punctured follicles ranged from approximately 6 to 10 mm. Oocytes were retrieved under intravenous anesthesia using a 19-gauge single-lumen needle (K-OPS-7035-RWH-B-ET; Cook Medical, Australia) at a reduced aspiration pressure of 90 mmHg. Immature COCs were cultured in IVM medium (Sage, CT, USA) supplemented with 0.075 IU/mL FSH and 0.075 IU/mL LH (Menopur; Ferring, Germany) for 28–32 h. This duration was selected based on our established laboratory protocols for unstimulated cycles, ensuring optimal nuclear maturation (MII) prior to ICSI while minimizing the risk of oocyte aging. Metaphase II (MII) oocytes were fertilized via ICSI.

Conventional IVF protocol

Women in the control group underwent controlled ovarian stimulation (COS) using a standard GnRH antagonist protocol. Recombinant FSH (Gonal-F; Serono, Germany) was administered daily (112.5–225 IU), with a GnRH antagonist (Cetrotide; Serono) added when the leading follicle reached ≥ 12 mm. Triggering was performed with recombinant hCG (250 µg Ovidrel; Serono) or a GnRH agonist (0.2 mg Triptorelin; Ipsen, France) for high-risk OHSS cases. Oocyte retrieval was performed 36 h post-trigger using a 17-gauge single-lumen needle (Cook Medical) at 120 mmHg aspiration pressure. Fertilization was achieved via conventional IVF or ICSI.

Embryo culture and transfer

Zygotes were cultured in cleavage medium (G-1 PLUS; Vitrolife, USA). Good-quality cleavage-stage embryos were defined as those with 7–9 cells (Day 3), < 10% fragmentation, and no multinucleation. Blastocyst culture was performed using G-2 PLUS medium (Vitrolife), with grading based on the Gardner system [18, 19].

Fresh embryo transfer

One or two cleavage-stage embryos or a single blastocyst were transferred. Luteal support (vaginal progesterone gel 90 mg/day + oral dydrogesterone 10 mg bid) commenced on retrieval day.

Frozen-Thawed Embryo Transfer (FET)

Endometrium was prepared with oral estradiol valerate (Progynova). Luteal support (vaginal Crinone 8% or oral dydrogesterone) began when endometrial thickness was appropriate, with transfer scheduled 5–7 days later. Support continued until 10 weeks of gestation.

Outcome measures

The primary outcome was the CLBR per initiated cycle, defined as the probability of achieving at least one live birth following the transfer of all fresh and frozen embryos derived from a single retrieval cycle.

Secondary reproductive outcomes included: embryological parameters, fertilization rate (2PN/MII), cleavage rate, good-quality embryo rate, and blastocyst formation rate; clinical pregnancy (gestational sac on ultrasound at 7 weeks), ongoing pregnancy (fetal heart activity at ≥ 12 weeks), and live birth (delivery of a live infant at ≥ 28 weeks); biochemical pregnancy (positive hCG without ultrasonographic sac), and ectopic pregnancy.

Safety and Perinatal Outcomes Safety outcomes focused on OHSS. Diagnosis and severity classification were performed in accordance with the 2024 ASRM guidelines [20]. Critical OHSS was defined as the presence of life-threatening complications, such as thromboembolism, acute respiratory distress syndrome (ARDS), or severe ascites/hydrothorax, accompanied by extreme laboratory abnormalities, including hematocrit > 55%, WBC > 25,000/mL, or renal failure (creatinine ≥ 1.6 mg/dL).

Obstetric and neonatal complications were defined as follows: gestational diabetes mellitus (GDM) and pregnancy-induced hypertension (PIH) were diagnosed according to international guidelines; preterm birth was defined as delivery < 37 weeks of gestation. Miscarriage was categorized as early (< 12 weeks) or late (12–28 weeks), and birth defects.

Statistical analysis

Continuous variables were assessed for distributional assumptions using the Kolmogorov-Smirnov test and visual inspection. Normally distributed data are presented as mean ± standard deviation (SD), while non-normally distributed data are expressed as median (interquartile range [IQR]). Categorical variables are presented as frequencies and percentages [n (%)].

Baseline and Outcome Comparisons Between-group comparisons were conducted using the Student’s t-test for normally distributed continuous variables or the Mann–Whitney U test for skewed distributions. Categorical variables were compared using the Pearson Chi-square test; Fisher’s exact test was utilized when expected cell counts were < 5.

Survival Analysis Kaplan-Meier survival analysis was employed to estimate the cumulative probability of live birth over the 36-month follow-up period. The differences between the survival curves of the IVM and IVF groups were compared using the log-rank test.

Subgroup Analysis To evaluate the consistency of treatment outcomes across different clinical phenotypes, subgroup analyses were performed stratified by female age, BMI, and ovarian reserve (AFC 12–23 vs. High AFC ≥ 24). For these comparisons, unadjusted odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to quantify the association between treatment modality (IVM vs. IVF) and live birth outcomes.

Software All statistical analyses were performed using SPSS software version 27.0 (IBM Corp., Armonk, NY, USA). A two-tailed P-value of < 0.05 was considered statistically significant.

Results

Study population and baseline characteristics

Of the 185,624 women scheduled for ART during the study period, 6,521 met the inclusion criteria (Fig. 1). A total of 545 women underwent strictly unstimulated IVM, and 1,635 matched controls underwent conventional IVF.

Fig. 1.

Flowchart of patient selection and study design. The diagram illustrates the inclusion and exclusion process for women with PCOS undergoing unstimulated in vitro maturation (IVM) or conventional in vitro fertilization (IVF) from 2010 to 2022. Abbreviations: BMI, body mass index; ET, embryo transfer; GnRH, gonadotropin-releasing hormone; IVF, in vitro fertilization; IVM, in vitro maturation; PCOS, polycystic ovary syndrome; PGT, preimplantation genetic testing

While demographic characteristics—including female age, BMI, and the number of prior ART attempts—were matched and comparable (Table 1), the basal endocrine profiles exhibited significant differences, reflecting the distinct clinical phenotypes managed by each strategy (Table 2). Specifically, women in the IVM group presented with a more severe PCOS phenotype, characterized by significantly higher basal LH levels, lower FSH levels, and markedly elevated anti-Müllerian hormone (AMH) levels compared to the IVF group (13.28 ± 6.59 vs. 6.59 ± 3.97 ng/mL; P < 0.001). Consistent with this, the baseline antral follicle count (AFC) was significantly higher in the IVM group (P < 0.001). Indeed, 68.26% of women in the IVM cohort had a high AFC (≥ 24), whereas the IVF cohort predominantly consisted of women with an intermediate AFC (12–23; 76.15%) (Table 2).

Table 1.

Baseline demographic and clinical characteristics of the study participants

Characteristics	IVM (n = 545)	IVF (n = 1635)	P-value
Female age, years	29.88 ± 3.46	29.87 ± 3.68	0.951
< 30, n(%)	259 (47.52%)	777 (47.52%)	1.000
30 ~ 34, n(%)	230 (42.20%)	690 (42.20%)
35 ~ 38, n(%)	50 (9.17%)	150 (9.17%)
39 ~ 42, n(%)	6 (1.10%)	18 (1.10%)
Female BMI, kg/m2	25.64 ± 4.48	25.47 ± 4.20	0.421
< 22	113 (20.73%)	339 (20.73%)	1.000
22 ~ 24, n(%)	93 (17.06%)	279 (17.06%)
24 ~ 28, n(%)	186 (34.13%)	558 (34.13%)
≥ 28, n(%)	153 (28.07%)	459 (28.07%)
Male age, years	31.60 ± 4.45	31.59 ± 4.67	0.956
Male BMI, kg/m2	25.86 ± 4.15	25.87 ± 4.33	0.947
Duration of infertility, years	4.18 ± 2.76	4.46 ± 3.03	0.058
Primary infertility (%)	402 (73.76%)	1271 (77.74%)	0.057
Secondary infertility (%)	143(26.24%)	364(22.26%)
Nulliparous (%)	534 (97.98%)	1601 (97.92%)	0.931
Number of prior IVF treatments, n(%)			0.726
< 3	485 (88.99%)	1446 (88.44%)
≥ 3	60 (11.01%)	189 (11.56%)

Table 2.

Baseline characteristics and Embryology results in Unstimulated IVM versus Conventional IVF Treatment

	IVM (n = 545)	IVF (n = 1635)	P-value
Basic Endocrinology
FSH, (mIU/mL)	5.57 ± 1.68	5.99 ± 1.93	< 0.001
LH, (mIU/mL)	8.22 ± 5.04	6.08 ± 4.85	< 0.001
E2, (pmol/L)	184.59 ± 96.07	182.90 ± 133.57	0.802
P, (mIU/mL)	1.60 ± 3.59	1.93 ± 8.05	0.547
AMH, (ng/ml)	13.28 ± 6.59	6.59 ± 3.97	< 0.001
Number of AFC	29.87 ± 12.05	18.03 ± 6.36	< 0.001
PCOS AFC, n(%)			< 0.001
AFC 12–23	173(31.74%)	1245(76.15%)	< 0.001
AFC ≥ 24	372(68.26%)	390(23.85%)
Duration of stimulation, days	0	10.56 ± 2.43	< 0.001
Total dose of gonadotropins used, IU	0	1918.69 ± 862.48	< 0.001
Average number of AFC oocytes retrieved	16.82 ± 10.49	14.25 ± 7.75	< 0.001
Oocytes retrieved	9165	23,291	——
MⅡ, n(%)	3909 (42.65%)	18,270 (78.44%)	< 0.001
Normal fertilized oocytes, n(%)	2411 (61.68%)	12,337 (67.53%)	< 0.001
Fertilized oocytes, n(%)	2922 (74.75%)	13,712 (75.05%)	0.693
Cleavage embryos	2852 (97.60%)	13,548 (98.80%)	< 0.001
D3 available embryos, n(%)	2161 (75.77%)	10,119 (74.69%)	0.226
Cycle with available embryos	510 (93.58%)	1606 (98.23%)	< 0.001

Cycle characteristics and embryologic outcomes

As per protocol, the IVM group received no ovarian stimulation, whereas the IVF group required an average of 10.56 days of stimulation with a mean total gonadotropin dose of 1,918.69 IU (P < 0.001). Despite the lack of stimulation, the mean number of oocytes retrieved was significantly higher in the IVM group (16.82 ± 10.49 vs. 14.25 ± 7.75; P < 0.001), consistent with their high ovarian reserve (Table 2).

In the IVM group, 9,165 oocytes were retrieved, achieving MII rate of 42.65% (3,909/9,165). Following ICSI, the normal fertilization rate (2PN) was lower in the IVM group compared to IVF (61.68% vs. 67.53%; P < 0.001). However, a notable finding was that the rate of Day 3 available embryos per fertilized oocyte was comparable between the two groups (75.77% vs. 74.69%; P = 0.226). Ultimately, due to the initial attrition during maturation, the proportion of cycles resulting in at least one transferable embryo was lower in the IVM group (93.58% vs. 98.23%; P < 0.001).

Reproductive outcomes of the first embryo transfer

In late 2017, our clinical strategy for IVM shifted to prioritize FET over fresh transfer. This evolution in practice and the corresponding annual distribution of study cases are illustrated in Fig. 2.

Fig. 2.

Annual distribution of initiated cycles in the IVM and IVF groups (2010–2022). This butterfly chart presents the number of initiated cycles for each year to illustrate the temporal distribution of the cohort. The orange bars on the left side represent the unstimulated IVM group, and the blue bars on the right side represent the conventional IVF group

Outcomes of the first embryo transfer are presented in Table 3. The proportion of cycles utilizing a “freeze-all” strategy was significantly higher in the IVM group (27.52% vs. 20.55%; P < 0.001). For fresh transfers, the live birth rate (LBR) was significantly lower in the IVM group compared to IVF (25.96% vs. 33.20%; P = 0.029). In contrast, for the first frozen embryo transfer, the LBR was comparable between the IVM and IVF groups (33.33% vs. 36.61%; P = 0.486). Furthermore, the implantation rate in frozen cycles was significantly higher in the IVM group (43.86% vs. 32.62%; P = 0.007).

Table 3.

Pregnancy outcomes following the first embryo transfer in women with Polycystic Ovary Syndrome (PCOS)

	Fresh transfer first			Frozen transfer first
	IVM (n = 235)	IVF (n = 1226)	P-value	IVM (n = 150)	IVF (n = 336)	P-value
Number of cycles	43.12% (235/545)	74.98% (1226/1634)	< 0.001	27.52% (150/545)	20.55% (336/1634)	< 0.001
Number of embryos transferred mean (SD)	1.97 ± 0.51	1.96 ± 0.39	0.797	1.14 ± 0.35	1.68 ± 0.47	< 0.001
Stage of embryos transferred			0.150			< 0.001
Day 3	230 (97.87%)	1176 (95.92%)		23 (15.33%)	221 (65.77%)
Day 5	5 (2.13%)	50 (4.08%)		127 (84.67%)	115 (34.23%)
The incidence rate of Critical OHSS	0 (0.00%)	12 (0.98%)	0.233*	0 (0.00%)	0 (0.00%)	——
Positive beta hCG	100 (42.55%)	590 (48.12%)	0.117	79 (52.67%)	166 (49.40%)	0.506
Total number of embryos transferred (n)	462	2401	——	171	564	——
Implantation	106 (22.94%)	707 (29.45%)	0.005	75 (43.86%)	184 (32.62%)	0.007
Clinical pregnancy, n(%)	86 (36.60%)	509 (41.52%)	0.160	69 (46.00%)	153 (45.54%)	0.924
Miscarrige, n(%)	19 (22.09%)	79 (15.52%)	0.129	19 (27.54%)	26 (16.99%)	0.071
Early miscarriage, n(%)	12 (13.95%)	45 (8.84%)	0.136	13 (18.84%)	19 (12.42%)	0.207
Late miscarriage, n(%)	7 (8.14%)	34 (6.68%)	0.621	6 (8.70%)	7 (4.58%)	0.231*
Live birth, n(%)	61 (25.96%)	407 (33.20%)	0.029	50 (33.33%)	123 (36.61%)	0.486
Singleton	54 (88.52%)	281 (69.04%)	0.002	45 (90.00%)	104 (84.55%)	0.347
Twins	7 (11.48%)	126 (30.96%)		5 (10.00%)	19 (15.45%)

*Fisher’s exact test

Cumulative live birth rate and safety outcomes

The CLBR per initiated cycle at 36 months was significantly lower in the IVM group compared to the IVF group (22.20% vs. 47.95%; Table 4). Survival analysis using Kaplan-Meier curves further demonstrated this difference in time to cumulative clinical pregnancy (Fig. 3) and cumulative live birth (Fig. 4) (Log-rank P < 0.001).

Table 4.

Clinical outcomes after the complete treatment cycle

	IVM (n = 545)	IVF (n = 1635)	P-value
Type of first transfer			< 0.001
Fresh ET	235 (43.12%)	1226 (74.98%)
Frozen ET	150 (27.52%)	336 (20.55%)
Untransplanted cycle	6 (1.10%)	24 (1.47%)	0.524
Cumulative embryo transfer cycles, n(%)	385 (70.64%)	1562 (95.54%)	< 0.001
Cumulative clinical pregnancy/ ET, n(%)	174 (45.19%)	926 (59.28%)	< 0.001
Cumulative live birth/ ET, n(%)	121 (31.43%)	784 (50.19%)	< 0.001
Cumulative clinical pregnancy, n(%)	174 (31.93%)	926 (56.64%)	< 0.001
Cumulative Miscarrige, n(%)	49 (28.16%)	170 (18.36%)	< 0.001
Cumulative live birth, n(%)	121 (22.20%)	784 (47.95%)	< 0.001
Gestational age(week)			0.078
≥37	92 (76.03%)	631 (80.48%)
34 ~ 36	21 (17.36%)	106 (13.52%)
32 ~ 33	1 (0.83%)	26 (3.32%)
<32	7 (5.79%)	21 (2.68%)
Delivery type			0.901
Caesarean Section	92 (76.03%)	592 (75.51%)
Vaginal Delivery	29 (23.97%)	192 (24.49%)
Pregnancy complications	22 (18.18%)	106 (13.52%)	0.171
Gestational diabetes mellitus	15 (12.40%)	58 (7.40%)	0.060
Hypertensive disorders	7 (5.79%)	48 (6.12%)	0.390
Birth defects	1 (0.83%)	3 (0.38%)	0.437*

*Fisher’s exact test

Fig. 3.

Kaplan-Meier analysis of the 36-month cumulative clinical pregnancy rate (CCPR). The curves illustrate the time to the first clinical pregnancy over a 36-month follow-up period for women with polycystic ovary syndrome (PCOS). The orange line represents the conventional in vitro fertilization (IVF) group, and the blue line represents the unstimulated in vitro maturation (IVM) group. The cumulative clinical pregnancy rate was significantly higher in the IVF group compared to the IVM group (Log-rank test, P < 0.001)

Fig. 4.

Kaplan-Meier analysis of the 36-month cumulative live birth rate (CLBR). The graph displays the cumulative probability of achieving a live birth over the 36-month follow-up period. The orange line represents the IVF group, while the blue line represents the IVM group. Consistent with clinical pregnancy outcomes, the cumulative live birth rate was significantly higher in the IVF group (Log-rank test, P < 0.001)

Regarding safety, no cases of OHSS occurred in the IVM group. In the IVF group, critical OHSS was observed exclusively in women who underwent fresh embryo transfer, with an incidence of 0.98% (12/1226).

Subgroup analysis: the “freeze-all” strategy

To further explore the efficacy of the “freeze-all” approach, we performed a subgroup analysis on cycles where the first transfer was FET (Table 5). Stratification by female age, BMI, and ovarian reserve (AFC) revealed that the LBR per first frozen transfer did not differ significantly between IVM and IVF across any subgroup. Notably, in women with a high antral follicle count (AFC ≥ 24)—the primary target population for IVM—the live birth rate after the first frozen transfer was nearly identical between the two groups (IVM 34.29% vs. IVF 34.43%; Odds Ratio 0.99, 95% CI 0.57–1.72). This indicates that when a “freeze-all” strategy is employed, the developmental competence of IVM-derived embryos is comparable to that of conventional IVF embryos in this high-risk population.

Table 5.

Live birth after the first embryo transfer and cumulative live birth in the freeze-all/FET subset, stratified by female age, BMI and AFC

Subgroup	No. IVM	Live birth / First ET (IVM)	No. IVF	Live birth / First ET (IVF)	OR (95% CI)	P-value	Cumulative LBR (IVM)	Cumulative LBR (IVF)	OR (95% CI)	P-value
Female age, years
< 30	77	25 (32.47%)	179	70 (39.11%)	0.75 (0.43–1.32)	0.313	30 (38.96%)	102 (56.98%)	0.48 (0.28–0.83)	0.008
30 ~ 34	63	22 (34.92%)	147	51 (34.69%)	1.01 (0.54–1.88)	0.975	25 (39.68%)	90 (61.22%)	0.42 (0.23–0.76)	0.004
35 ~ 38	9	2 (22.22%)	10	2 (20.00%)	1.14 (0.13–10.39)	1	2 (22.22%)	3 (30.00%)	0.67 (0.08–5.30)	1
39 ~ 42	1	1 (100.00%)	0	0 (0.00%)	N/A	-	1 (100.00%)	0 (0.00%)	N/A	-
Female BMI, kg/m²
< 22	34	13 (38.24%)	49	20 (40.82%)	0.90 (0.37–2.20)	0.813	16 (47.06%)	32 (65.31%)	0.47 (0.19–1.15)	0.098
22 ~ 24	26	12 (46.15%)	52	26 (50.00%)	0.86 (0.33–2.20)	0.749	13 (50.00%)	36 (69.23%)	0.44 (0.17–1.17)	0.098
24 ~ 28	53	14 (26.42%)	134	44 (32.84%)	0.73 (0.36–1.49)	0.392	17 (32.08%)	72 (53.73%)	0.41 (0.21–0.79)	0.008
≥ 28	37	11 (29.73%)	101	33 (32.67%)	0.87 (0.38–1.98)	0.742	12 (32.43%)	55 (54.46%)	0.40 (0.18–0.89)	0.022
PCOS Phenotype
AFC 12–24	45	14 (31.11%)	214	81 (37.85%)	0.74 (0.37–1.48)	0.394	16 (35.56%)	128 (59.81%)	0.37 (0.19–0.72)	0.003
High AFC (≥ 24)	105	36 (34.29%)	122	42 (34.43%)	0.99 (0.57–1.72)	0.982	42 (40.00%)	67 (54.92%)	0.55 (0.32–0.93)	0.025

Odds ratios (ORs) are presented as IVM vs. IVF and were calculated from the corresponding 2 × 2 tables within each stratum; 95% confidence intervals (CIs) were derived using the Wald method. Where cell counts were small, Fisher’s exact test was used for P values. ORs were not estimable for strata with zero events in one group

Discussion

Principal findings

This large retrospective cohort study provides robust evidence regarding the long-term efficacy and safety of strictly unstimulated IVM for women with PCOS. While the 36-month CLBR per initiated cycle was lower for IVM compared to conventional IVF (22.20% vs. 47.95%; P < 0.001), the safety profile was significantly superior, with no cases of OHSS in the IVM group. Notably, our data align closely with findings from our previous RCT, which reported a 22.3% cumulative ongoing pregnancy rate at six months [16]. The variance in success rates compared to IVF is largely attributable to the difference in oocyte yield and subsequent embryo availability, rather than the developmental competence of individual embryos.

The “freeze-all” imperative in unstimulated IVM

The embryo transfer strategy evolved over the study period. In the IVM group, practice shifted from fresh transfers (2010–2017) to a “freeze-all” strategy (2018–2022). In the IVF group, fresh transfers remained standard unless contraindicated (e.g., OHSS risk), in which case a “freeze-all” approach was adopted. A critical observation in our study, consistent with findings by Walls [21], is the distinct divergence in outcomes based on the transfer strategy. While fresh embryo transfers in the IVM group yielded significantly lower success rates compared to IVF, outcomes from FET cycles were fully comparable. Specifically, the live birth rate per first frozen transfer did not differ significantly between the two groups (33.33% vs. 36.61%). Moreover, the implantation rate in frozen cycles was significantly higher in the IVM group (43.86% vs. 32.62%).

Our data strongly suggest that the lower efficacy observed in fresh IVM transfers is attributable to endometrial asynchrony caused by the shortened follicular phase and inadequate endogenous estrogen exposure in unstimulated cycles [22, 23]. Interestingly, a recent study by Vuong et al. (2025) reported no significant difference in live birth rates between fresh and frozen transfers following CAPA-IVM [24]. However, a critical methodological distinction must be noted. In their “fresh” ET group, patients received intensive exogenous estrogen priming (8 mg/day for 8 days) before oocyte retrieval to artificially prepare the endometrium. Vuong’s artificial endometrial preparation effectively circumvents the natural asynchrony, validating our hypothesis that the endometrial window is indeed significantly compromised in true unstimulated fresh cycles. Consequently, for standard monophasic IVM protocols lacking profound exogenous endometrial priming, decoupling oocyte retrieval from embryo transfer via a “freeze-all” strategy remains essential to unleash the full implantation potential of the embryos.

Influence of age on IVM outcomes

Our subgroup analysis (Table 5) indicated that in younger women (< 30 years), the cumulative LBR was significantly lower in the IVM group compared to IVF, whereas this gap narrowed in older age groups. This disparity in younger patients is likely driven by the sheer quantity of oocytes retrieved in stimulated IVF cycles, which maximizes cumulative potential. Younger women with PCOS are typically “high responders” to stimulation, yielding a high number of blastocysts. In contrast, unstimulated IVM relies on the intrinsic antral follicle pool without recruitment, naturally limiting the cohort size. However, for patients > 35 years (though sample size is limited), the difference was less pronounced, suggesting that as ovarian reserve and oocyte quality naturally decline, the “quantity advantage” of stimulated IVF diminishes.

Ovarian reserve markers and phenotype-based selection

Our findings should be interpreted in the context of baseline ovarian reserve. In our cohort, women allocated to IVM had markedly higher AMH and AFC, reflecting the real-world clinical rationale of triaging predicted hyper-responders to a safer, stimulation-free approach. This aligns with recent evidence from Mostinckx et al. (2024), who demonstrated that IVM performance is strongly modulated by ovarian reserve, with very high AMH and AFC levels serving as strong independent predictors of live birth [25]. Our subgroup analysis confirms this clinical paradigm: within the “high-responder” phenotype (AFC ≥ 24), the live birth rate per frozen transfer in the IVM group was nearly identical to that of conventional IVF (OR 0.99, 95% CI 0.57–1.72). These data emphasize that AMH and AFC are not merely confounders but crucial effect modifiers in ART. Ultimately, this reinforces a phenotype-based selection strategy, confirming that high-AFC PCOS patients are the ideal candidates for IVM, converting their severe risk of OHSS into a safe and highly efficient reproductive outcome.

Safety profile in high-risk phenotypes

Regarding safety, no cases of OHSS occurred in the IVM group. In the IVF group, critical OHSS requiring hospitalization was observed exclusively in women who underwent fresh embryo transfer, with an incidence of 0.98% (12/1226). This finding is consistent with our previous RCT, which reported a moderate-to-severe OHSS rate of 6.3% in the IVF arm versus 0% in the IVM [16]. The total absence of OHSS in the current IVM cohort is particularly striking given the baseline heterogeneity: women in the IVM group had significantly higher AFC and AMH levels, reflecting real-world clinical triage where patients with the highest intrinsic risk are directed toward IVM. The fact that no OHSS occurred in this high-risk phenotype confirms IVM as the safest theoretical approach [26].

Treatment burden and patient-centered care

Crucially, given that our retrospective data collection was restricted to “critical” cases requiring hospitalization, the statistical focus on severe outcomes likely underestimates the broader disparity in treatment burden. It is clinically recognized that for women with PCOS, the absence of documented complications does not equate to a lack of suffering. Although GnRH antagonist protocols and GnRH agonist trigger have reduced severe OHSS, women with PCOS often endure a “universal discomfort” from ovarian stimulation—characterized by prolonged injections, abdominal distension, and inevitable ovarian enlargement—which significantly impacts daily functioning. Strictly unstimulated IVM circumvents this entirely, offering a physically gentler experience. This improved tolerability is a vital component of patient-centered care, particularly for high-responders who prioritize safety and quality of life over maximizing oocyte yield [26].

Embryo quantity versus quality

The lower overall CLBR in the IVM group was primarily driven by the “quantity” rather than the “quality” of embryos. IVM cycles yielded significantly fewer available embryos per cycle compared to IVF, a direct consequence of the expected attrition during the in vitro maturation and fertilization phases. However, our data indicate that once an embryo reaches the cleavage/blastocyst stage and is transferred in a frozen cycle, its potential for live birth is equivalent to that of an IVF-derived embryo [16]. This highlights that the limiting factor for IVM is the efficiency of oocyte maturation and blastocyst formation, pointing to the need for continued optimization of IVM culture systems (e.g., CAPA-IVM) to improve embryo yield [12, 27].

Future perspectives: optimization of culture systems

Finally, the relatively low post-IVM MII rate in our strictly unstimulated protocol underscores the biological challenge of maturing less-advanced oocytes retrieved from small antral follicles. In our practice, oocyte retrieval is triggered when the leading follicle remains < 10 mm; thus, most aspirated follicles are typically within the small antral range, and the exact average follicle diameter at puncture was not systematically recorded over the full study period. While the 28–32 h maturation window reflects our standardized laboratory workflow for standard monophasic IVM, we fully acknowledge the emerging efficacy of alternative biphasic strategies. Systems such as CAPA-IVM aim to mimic physiological meiotic arrest, promoting cytoplasmic maturation before nuclear maturation. Indeed, as recently noted by Vuong et al. (2025), the application of a biphasic CAPA-IVM method in hormone-free cycles yielded a substantially higher ongoing pregnancy rate (38.3%) than the 22.3% achieved in our previous RCT using a standard serum-free monophasic system [27]. This direct comparison highlights that optimizing the in vitro culture environment is the critical next step. Given the lower fertilization and blastocyst formation rates observed in our current monophasic IVM cohort compared to IVF, adopting such a two-step culture system could potentially bridge the gap in embryo yield. Alongside these biphasic systems, further improvements in granulosa cell co-culture platforms and enhanced oocyte/cumulus-oocyte complex (COC) selection methods remain essential strategies to maximize the developmental competence of retrieved immature oocytes [28]. Future research in our center will aim to integrate these biphasic protocols to evaluate whether they can further enhance clinical efficiency while preserving the absolute safety advantages for high-risk PCOS phenotypes.

Strengths and limitations

A key strength of this study is the extended 36-month follow-up, providing one of the first long-term assessments of CLBR in strictly unstimulated IVM. However, several limitations must be acknowledged. First, the retrospective, non-randomized design is vulnerable to selection bias and residual confounding. To rigorously address this, we employed strict 1:3 exact matching for age and BMI. While this approach maximizes internal validity by creating highly comparable cohorts—thereby avoiding residual confounding that can occur in standard multivariable regression models—it inherently necessitates excluding unmatched cycles and those cancelled before retrieval. Consequently, our findings reflect the outcomes of this rigorously matched subset, which might slightly overestimate per-start success rates compared to the general population. Furthermore, baseline ovarian phenotypes (e.g., AFC and AMH) inevitably differed between cohorts, reflecting the real-world clinical triage of predicted hyper-responders to IVM. While our matching isolated the primary treatment effect, future studies utilizing complementary multivariable regression or propensity-score approaches incorporating these ovarian reserve covariates could further evaluate the robustness of our findings. Second, clinical practice evolved over the 12-year study period (e.g., increasing adoption of freeze-all strategies, embryo culture preferences, and incremental laboratory updates), introducing potential time-related confounding. Third, transfer practices, including the number of embryos transferred and embryo stage, differed between groups and across time, potentially influencing pregnancy and multiple-gestation outcomes. Fourth, key cycle-level variables, such as exact follicle size distribution at retrieval, were not uniformly captured over the full study period, limiting deeper mechanistic inferences regarding endometrial asynchrony in fresh transfers. Finally, incomplete utilization of all cryopreserved embryos within the 36-month window may lead to an underestimation of the true CLBR, particularly in conventional IVF strategies that generate a larger cohort of embryos.

Conclusion

In conclusion, strictly unstimulated IVM is a safe and viable alternative to conventional IVF for women with PCOS, particularly those with a high AFC phenotype. Although the cumulative live birth rate is lower due to reduced embryo yield, the “freeze-all” strategy achieves implantation and live birth rates per transfer that are comparable to or even better than conventional IVF. For selected high-risk patients, IVM offers a unique balance: it eliminates the risk of OHSS and reduces treatment burden, while maintaining high per-transfer efficacy when coupled with modern cryopreservation strategies.

Abbreviations

AFC

Antral follicle count

AMH

Anti-Müllerian hormone

ASRM

American Society for Reproductive Medicine

ART

Assisted reproductive technology

BMI

Body mass index

COH

Controlled ovarian hyperstimulation

COS

Controlled ovarian stimulation

CLBR

Cumulative live birth rate

CLBRs

Cumulative live birth rates

COCs

Cumulus–oocyte complexes

FET

Frozen-Thawed Embryo Transfer

GnRH

Gonadotropin-releasing hormone

IVM

In vitro maturation

IVF

In vitro fertilization

ICSI

Intracytoplasmic sperm injection

LBR

Live birth rate

MII

Metaphase II

ORs

Odds ratios

OHSS

Ovarian hyperstimulation syndrome

PCOS

Polycystic ovary syndrome

PGT

Preimplantation genetic testing

RCT

Randomised controlled trial

SOPs

Standard Operating Procedures

Authors’ contributions

Jie Qiao and Rong Li conceived of the study. Jie Qiao, Rong Li, and Xiaoying Zheng designed the study. Wei Guo, Lixue Chen, and Zi Yang wrote the manuscript. Wei Guo, Lixue Chen, Zi Yang, Linlin Wang, Hongping Wu, Rui Yang, and Shuo Yang coordinated the data collection and performed the analysis. Hsun-Ming Chang, and Xiumei Zhen reviewed and edited the manuscript. All the authors critically reviewed and approved the final manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (82201805); Peking University Clinical Scientist Training Program supported by “the Fundamental Research Funds for the Central Universities” (BMU2025PYJH029); 2025 Beijing Physician-Scientist Training Program (BJPSTP-2025-04); the Beijing Municipal Science & Technology Commission (Z191100006619073); the National Science Fund for Distinguished Young Scholars (81925013); and the National Key Research and Development Project of China (2022YFC2702500).

Data availability

The datasets used in the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Reproductive Medicine Ethics Committee of Peking University Third Hospital (Reference number: SZ-066), in accordance with the principles of the Declaration of Helsinki. All participants received written and oral information and signed informed consent before any examination.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.