Non-linear impact of abnormal fertilization on sibling 2PN embryo development: a 107,527-embryo cohort study defining a novel clinical tipping point

Abstract

Background

Abnormal fertilization is a frequent occurrence in in vitro fertilization (IVF) cycles, yet the impact of abnormally fertilized zygotes on their sibling normally fertilized (2PN) embryos remains a subject of ongoing debate. This study aimed to characterize the non-linear relationship between the abnormal fertilization rate (AFR) and sibling 2PN blastocyst developmental potential to identify a clinically actionable biomarker for oocyte cohort quality.

Methods

We conducted a large-scale retrospective cohort study analyzing 107,527 sibling 2PN-derived embryos from 13,167 conventional IVF cycles (March 2019–January 2023). AFR was defined as the proportion of non-2PN zygotes among all fertilized oocytes. Restricted cubic splines (RCS) were employed to explore potential non-linearity, and multivariable generalized estimating equation (GEE) models were used to assess the independent impact of AFR on blastocyst formation, adjusting for 22 clinical and embryological confounders while accounting for embryonic clustering.

Results

RCS analysis identified a critical biological tipping point at 25% AFR. Beyond this threshold, the odds of blastocyst formation for sibling 2PN embryos significantly declined (adjusted odds ratio (aOR) 0.933, 95% CI 0.883–0.986, P = 0.014). Subgroup analysis demonstrated that this developmental impairment was robustly preserved in patients with adequate ovarian reserve (AMH ≥ 1.2 ng/mL: aOR 0.934, P = 0.018) and standard oocyte yields (5–14 oocytes: aOR 0.913, P = 0.017). Conversely, the association was not statistically significant in ultra-high-yield cycles (≥ 15 oocytes) or diminished ovarian reserve (AMH < 1.2 ng/mL), which may be primarily attributed to the truncated sample sizes and diminished statistical power within these specific sub-cohorts.

Conclusions

Our findings establish AFR as a sensitive, independent biomarker of oocyte cohort integrity. The identification of the 25% threshold provides a high-level evidence base for Day-1 risk stratification, personalized prognostic counseling, and optimized laboratory quality management. Integrating this tool into ART workflows can enhance the precision of clinical outcome expectations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-08636-7.

Introduction

In modern assisted reproductive technology (ART), extended embryo culture to the blastocyst stage has become a cornerstone strategy for improving live birth rates and facilitating elective single embryo transfer (eSET) [1, 2]. The ability of a zygote to reach the blastocyst stage is not only a hallmark of its developmental competence [3] but also serves as a critical performance indicator for both ovarian stimulation efficacy and laboratory quality control [4]. Despite advancements in standardized protocols, significant inter-cycle variability in blastocyst development persists [5, 6], even after accounting for established predictors such as maternal age and oocyte yield. This variability underscores the necessity of identifying novel, early-stage biomarkers that can accurately reflect the physiological “health” of the oocyte cohort.

The fertilization process represents the first critical transition in embryonic life [7, 8]. Abnormal fertilization—characterized by the formation of single pronuclear (1PN) or multi-pronuclear (e.g., 3PN) zygotes—is a frequent occurrence in conventional in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) cycles. While abnormally fertilized zygotes are routinely discarded in clinical practice due to their high risk of chromosomal aberrations and poor developmental potential [9], a fundamental biological question persists: does a high rate of abnormal fertilization (AFR) reflect a compromised microenvironment or systemic cohort-wide impairment that independently impairs the developmental capacity of its normally fertilized siblings? This potential “bystander effect” remains a subject of ongoing debate with significant implications for personalized reproductive medicine.

To date, extensive research has predominantly focused on the morphokinetics, ploidy status, and limited clinical viability of the abnormally fertilized zygotes themselves [10, 11]. While some studies suggest that an elevated AFR might reflect a global impairment of oocyte cohort quality [12], the use of AFR as an independent, quantitative biomarker for sibling 2PN blastocyst competence remains largely underexplored. This gap in the literature likely stems from two major methodological limitations. First, there is a lack of consensus regarding the calculation of AFR [13], many studies utilize inconsistent denominators (e.g., including or excluding unfertilized/immature oocytes) [14], which introduces significant biological noise and prevents a standardized assessment of the true fertilization landscape. Second, most existing predictive models assume a strictly linear relationship between embryological parameters and developmental outcomes, thereby overlooking the non-linear “tipping points” or potential threshold effects commonly observed in complex biological systems.

Leveraging a high-granularity clinical database of over 13,000 IVF cycles involving 107,527 sibling embryos, we sought to investigate the independent impact of AFR on blastocyst formation among sibling 2PN embryos. By refining the AFR denominator to minimize confounding biological noise and employing restricted cubic splines (RCS) to explore non-linearity alongside generalized estimating equations (GEE) to adjust for embryonic clustering, we aimed to identify a definitive clinical threshold for AFR. Furthermore, we evaluated the robustness of this association across different clinical strata, particularly focusing on how maternal age, ovarian reserve (AMH), and oocyte yield modulate the vulnerability of sibling embryos to high AFR. Through this large-scale analysis, we aimed to provide a high-level evidence base for more accurate prognostic counseling and laboratory quality management in clinical practice.

Materials and methods

Ethical approval

Ethical approval for this protocol was granted by the Institutional Review Board (IRB) of the Ethical Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (TJ-IRB20211280). All clinical and embryological data were systematically extracted from the center’s electronic medical record (EMR) system, with meticulous double-checking by two independent researchers to ensure data integrity. At the initiation of their ART treatment, all patients provided written informed consent specifically authorizing the use of their de-identified clinical data for future retrospective academic research. To protect patient privacy, all datasets were fully anonymized prior to statistical analysis, and the study was conducted in full compliance with the Declaration of Helsinki.

Study design and population

This was a large-scale retrospective cohort study conducted at Tongji Hospital. Data were extracted from the electronic clinical database for IVF cycles performed between March 2019 and January 2023. The primary objective was to investigate the association between the AFR and the blastocyst formation potential of sibling normally fertilized (2PN) embryos. To facilitate clinical translation, our subgroup stratification was aligned with the internationally recognized Patient-Oriented Strategies Encompassing Individualized Oocyte Number (POSEIDON) criteria [15, 16], which categorize patient prognoses based on maternal age, oocyte yield, and ovarian reserve markers. Specifically, an anti-Müllerian hormone (AMH) threshold of 1.2 ng/mL was utilized to differentiate between standard ovarian reserve (≥ 1.2 ng/mL) and diminished ovarian reserve (< 1.2 ng/mL) cohorts. Furthermore, in strict accordance with POSEIDON criteria requirements, a baseline minimum of 5 retrieved oocytes was established for cycle inclusion to ensure cohort evaluation validity. Consistent with previous studies establishing that live birth rates plateau and safety risks escalate beyond 15 oocytes [17, 18], a retrieval of 5–14 oocytes was defined as a standard oocyte yield, whereas ≥ 15 oocytes denoted an ultra-high-yield cohort. The clinical flow of participant enrollment, screening, and the structured comparative framework of this study are illustrated in Fig. 1. To ensure the highest level of methodological transparency and reporting quality, this retrospective cohort study was conducted and reported in strict accordance with the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) guidelines [19]. Specifically, we utilized these guidelines to refine the selection of confounding variables, standardize the reporting of our statistical models (GEE), and ensure the clear presentation of subgroup analyses across diverse clinical strata.

Fig. 1.

Study flow chart of the participant recruitment and analytical framework. Notes: IVF, in vitro fertilization; n, number; ICSI, intracytoplasmic sperm injection; PGT, preimplantation genetic testing; AFR, abnormal fertilization rate; 2PN, two pronuclei

Inclusion and exclusion criteria

The study population was strictly screened to ensure data homogeneity and biological relevance. Cycles were eligible for inclusion if they met the following criteria: (i) autologous IVF cycles using fresh oocytes; and (ii) a minimum of five retrieved oocytes per cycle. This oocyte threshold was specifically implemented to ensure a robust denominator for the calculation of the AFR, thereby minimizing potential skewness in proportions associated with low oocyte yields. Exclusion criteria were applied as follows: (i) cycles involving ICSI or split insemination (IVF/ICSI), or rescue ICSI (r-ICSI) to eliminate confounding variables associated with micromanipulation or delayed insemination windows; (ii) use of donor gametes (oocytes or sperm); (iii) cycles undergoing preimplantation genetic testing (PGT); and (iv) cycles with incomplete core clinical or embryological datasets.

Controlled ovarian stimulation and laboratory protocols

Ovarian stimulation and laboratory procedures were conducted according to the established protocols of our center, which have been described in detail previously [20, 21]. Briefly, ovarian stimulation was performed using individualized standard clinical protocols, including the GnRH agonist long or short protocols, the GnRH antagonist protocol, and progestin-primed ovarian stimulation (PPOS). The selection of the specific stimulation regimen and the initial dose of gonadotropin (Gn) were tailored to each patient based on maternal age, body mass index (BMI), and markers of ovarian reserve, such as anti-Müllerian hormone (AMH) and basal antral follicle count (AFC).

Final oocyte maturation was triggered when at least two or three follicles reached a diameter of ≥ 18 mm, and oocyte retrieval was performed via transvaginal ultrasound-guided aspiration 34–36 h after hCG or dual trigger. For conventional IVF, oocytes were cultured in fertilization medium and inseminated 39–40 h post-trigger. The concentration of the progressively motile spermatozoa used for insemination was strictly standardized to approximately 1 × 10⁵ to 2 × 10⁵ motile spermatozoa spermatozoa/mL, depending on the semen quality assessment. The gametes were co-incubated for a standard duration of 16–18 h. Fertilization was assessed 16–18 h post-insemination. Normal fertilization was defined by the presence of two pronuclei (2PN), whereas abnormal fertilization was characterized by the presence of a 1PN, multiple pronuclei (≥ 3PN), or other non-2PN configurations.

All embryos were cultured in G-1™ PLUS (Vitrolife) in a benchtop incubator system maintained in a humidified atmosphere of 5% O2 and 6% CO2 at 37 °C. Embryo development was monitored daily, and blastocyst quality was rigorously assessed on Day 5 and Day 6 according to the Gardner and Schoolcraft grading system [22]. Only blastocysts meeting the criteria for utilization (transfer or cryopreservation) were included in the final analysis of the blastocyst formation rate.

Outcome measures and definition

The primary outcome of this study was the blastocyst formation rate, defined as the number of blastocysts formed divided by the total number of normally fertilized (2PN) embryos available for extended culture.

In addition, the AFR was calculated as the number of abnormally fertilized zygotes divided by the total number of fertilized zygotes (Total fertilized = 2PN + 1PN + ≥ 3PN). This specific denominator was selected to strictly isolate the fidelity of the fertilization process itself. By excluding unfertilized or immature oocytes, this mathematical approach filters out upstream confounding variables such as male-factor sperm penetration failures or primary oocyte non-activation, thereby ensuring that the AFR accurately reflects the systemic cytoplasmic maturity and cortical integrity of the oocyte cohort. Good-quality D3 embryo rate was the proportion of Day 3 embryos with 6–10 cells, < 20% fragmentation, and even blastomere symmetry among the total number of 2PN embryos. Maturation Rate was defined as the number of metaphase II (MII) oocytes divided by the total number of retrieved oocytes. Normal Fertilization Rate was defined as the number of oocytes displaying two pronuclei divided by the total number of fertilized oocytes (2PN + non-2PN).

Statistical analysis

Continuous variables were expressed as mean ± standard deviation (SD) and compared using one-way ANOVA or the Kruskal-Wallis test, depending on the distribution of the data. Categorical variables were presented as frequencies (percentages) and analyzed using the Chi-square test or Fisher’s exact test where appropriate. For multiple pairwise comparisons between the three AFR groups, post-hoc analyses were systematically performed using the Bonferroni correction to control the family-wise error rate.

To facilitate clinical translation and characterize the potentially non-linear relationship between the AFR and the probability of blastocyst formation, AFR was modeled using RCS [23] to identify optimal thresholds for risk stratification. And RCS were fitted with four knots positioned at the 5th, 35th, 65th, and 95th percentiles. Based on the biological inflection point identified by the RCS curve, AFR was further analyzed as a categorical variable (0%, 1–24%, and ≥ 25%). To account for the inherent intra-class correlation (clustering effect) of sibling embryos within a single stimulation cycle, multivariable analysis was performed using GEE with a logit link function and an exchangeable working correlation structure. This robust statistical approach ensures that the standard errors are correctly estimated despite the non-independence of embryos from the same patient.

Multivariable GEE models were adjusted for a comprehensive set of potential confounders, including maternal age, BMI, infertility duration, infertility type, cause of infertility, ovarian reserve markers (AMH, basal AFC, basal FSH), ovarian stimulation protocol, total Gn dose, Gn duration, and the number of retrieved oocytes. All statistical analyses were conducted using SPSS 26.0 (IBM Corp., Armonk, NY, USA) and R software (version 4.2.2; R Foundation for Statistical Computing, Vienna, Austria). All tests were two-sided, and a P-value < 0.05 was considered statistically significant.

Results

Study population and threshold determination

The process of patient inclusion and screening in this study is illustrated in the Fig. 1. A total of 13,167 IVF cycles involving 107,527 sibling 2PN-derived embryos were included in this study. To determine the specific biological relationship between the AFR and blastocyst development, RCS were utilized. A significant non-linear association was observed, revealing a distinct clinical threshold at an AFR of 25% (Fig. 2). Based on this cut-off, the study population was categorized into three groups: 0% AFR (n = 3,831), 1–24% AFR (n = 5,718), and ≥ 25% AFR (n = 3,618).

Fig. 2.

RCS curve for the non-linear association between AFR and blastocyst development. Notes: The RCS model represents the association between the AFR (expressed as a percentage) and the predicted probability of blastocyst formation, adjusted for maternal age, ovarian reserve, and stimulation protocols. The solid blue line represents the adjusted odds ratio (aOR), and the shaded area represents the 95% CI. The vertical dashed line at 25% AFR indicates the identified clinical threshold. Beyond this inflection point, a significant downward trend in blastocyst developmental potential is observed, signaling the onset of cohort-wide quality degradation. The reference point (aOR = 1.0) is set at 0% AFR. This non-linear visualization confirms the existence of a threshold effect rather than a simple linear decline

Baseline characteristics

Baseline maternal and clinical characteristics across the three AFR groups are summarized in (Table 1). Importantly, because these cohorts exhibit substantial unadjusted variations in baseline profiles (such as ovarian reserve markers and stimulation protocols), these univariate comparisons are highly susceptible to confounding and are presented strictly for descriptive purposes; definitive conclusions regarding independent associations are derived solely from the subsequent multivariable GEE models. While maternal age exhibited a statistically significant difference among the three groups (31.19 ± 3.93, 31.15 ± 3.85, and 31.52 ± 3.93 years, respectively; P < 0.001), the absolute numerical variance was marginal, suggesting that the cohorts were clinically comparable in terms of age distribution. BMI remained remarkably consistent across all groups (P = 0.623). Regarding the etiology of infertility, significant heterogeneity was observed (P < 0.001), and the high AFR group (≥ 25%) was characterized by a higher prevalence of isolated female factors (46.0%) and secondary infertility (36.5%) compared to the 0% and 1–24% AFR groups.

Table 1.

Baseline characteristics of patients according to abnormal fertilization rate (AFR)

Characteristics	Overall (N = 13167)	0% AFR (n = 3831)	1–24% AFR (n = 5718)	≥ 25% AFR (n = 3618)	P-value
Age (years)	31.27 ± 3.90	31.19 ± 3.93	31.15 ± 3.85	31.52 ± 3.93	< 0.001
BMI (kg/m²)	22.09 ± 3.07	22.05 ± 3.06	22.08 ± 3.09	22.12 ± 3.04	0.623
Infertility duration (y)	3.21 ± 2.24	3.35 ± 2.28	3.17 ± 2.20	3.12 ± 2.26	< 0.001
Infertility type, n (%)					< 0.001
Primary	8792 (66.8%)	2721 (71.0%)	3774 (66.0%)	2297 (63.5%)	/
Secondary	4375 (33.2%)	1110 (29.0%)	1944 (34.0%)	1321 (36.5%)	/
Infertility cause, n (%)					< 0.001
Female factors only	5418 (41.1%)	1508 (39.4%)	2246 (39.3%)	1664 (46.0%)	/
Male factors only	3523 (26.8%)	1105 (28.8%)	1609 (28.1%)	809 (22.4%)	/
Combined factors	2980 (22.6%)	842 (22.0%)	1363 (23.8%)	775 (21.4%)	/
Unexplained	1246 (9.5%)	376 (9.8%)	500 (8.7%)	370 (10.2%)	/
Ovarian Reserve	-
AMH (ng/mL)	4.73 ± 3.53	4.61 ± 3.52	5.01 ± 3.58	4.42 ± 3.44	< 0.001
Basal AFC	14.85 ± 7.00	14.59 ± 6.83	15.52 ± 7.11	14.07 ± 6.90	< 0.001
Basal FSH (IU/L)	7.23 ± 2.12	7.33 ± 2.11	7.15 ± 2.22	7.24 ± 1.97	< 0.001
Ovarian Stimulation
Ovarian stimulation protocol, n (%)					< 0.001
GnRH agonist protocols	6855 (52.1%)	1900 (49.6%)	3109 (54.4%)	1846 (51.0%)	/
GnRH antagonist protocol	5870 (44.6%)	1778 (46.4%)	2450 (42.8%)	1642 (45.4%)	/
PPOS	390 (3.0%)	137 (3.6%)	135 (2.4%)	118 (3.3%)	/
Other protocols	52 (0.4%)	16 (0.4%)	24 (0.4%)	12 (0.3%)	/
Total Gn dose (IU)	2270.96 ± 849.10	2293.74 ± 843.50	2216.64 ± 843.22	2332.70 ± 859.11	< 0.001
Gn duration (days)	9.95 ± 1.80	9.90 ± 1.77	9.97 ± 1.82	9.97 ± 1.79	0.095
E2 on hCG day (pg/mL)	3073.51 ± 1803.50	2834.49 ± 1587.39	3277.40 ± 1947.79	3004.38 ± 1744.67	< 0.001
P on hCG day (ng/mL)	0.94 ± 0.68	0.90 ± 0.67	0.97 ± 0.74	0.93 ± 0.59	< 0.001
LH on hCG day (mIU/mL)	3.04 ± 2.35	3.04 ± 2.34	3.01 ± 2.36	3.09 ± 2.36	0.311
Endometrial thickness (mm)	11.35 ± 2.57	11.40 ± 2.56	11.41 ± 2.55	11.18 ± 2.62	< 0.001

Notes: Values are expressed as mean ± standard deviation (SD) for continuous variables or number (percentage) for categorical variables. The abnormal fertilization rate (AFR) was defined as the number of abnormally fertilized oocytes (1PN, 3PN, and poly-PN) divided by the total number of fertilized oocytes (2PN + non-2PN). P-values represent the statistical comparison across the three AFR groups (0%, 1–24%, and ≥ 25%) using one-way ANOVA or Kruskal-Wallis test for continuous variables and Chi-square test or Fisher’s exact test for categorical variables, with the Bonferroni correction applied for post-hoc pairwise comparisons. Abbreviations: BMI, body mass index; AMH, anti-Müllerian hormone; AFC, antral follicle count; FSH, follicle-stimulating hormone; Gn, gonadotropin; E2, estradiol; P, progesterone; LH, luteinizing hormone; PPOS, progestin-primed ovarian stimulation

In terms of ovarian reserve, patients in the 1–24% AFR group demonstrated the most favorable profiles, significantly higher AMH levels (5.01 ± 3.58 ng/mL) and basal AFC (15.52 ± 7.11) compared to the other two cohorts (P < 0.001). Conversely, the ≥ 25% AFR group showed a relatively diminished ovarian reserve, alongside a shorter infertility duration (3.12 ± 2.26 years).

Regarding ovarian stimulation, the GnRH agonist long/short protocols were the most frequently utilized across the cohorts (52.1%), followed by the GnRH antagonist protocol (44.6%). Although stimulation protocols differed significantly in distribution (P < 0.001), the majority of patients in all groups underwent standard pituitary suppression. The total dosage of Gn was significantly higher in the ≥ 25% AFR group (2332.70 ± 859.11 IU) compared to the 1–24% group (2216.64 ± 843.22 IU, P < 0.001), while the duration of stimulation remained similar (P = 0.095). On the day of hCG trigger, the 1–24% AFR group exhibited the highest peak E2 levels (3277.40 ± 1947.79 pg/mL) and slightly higher P levels (0.97 ± 0.74 ng/mL; P < 0.001). No significant differences were observed in LH levels on the trigger day (P = 0.311).

Embryological outcomes

The embryological outcomes, stratified by AFR categories, revealed significant variances in both early fertilization success and late-stage developmental competence (Table 2).

Table 2.

Embryological outcomes and sibling embryo development

Characteristics	0% AFR (n = 3831)	1–24% AFR (n = 5718)	≥ 25% AFR (n = 3618)	P-value
No. of oocytes retrieved	13.99 ± 5.25	16.61 ± 6.69	14.59 ± 6.03	< 0.001
No. of MII oocytes	11.36 ± 4.42	14.36 ± 5.90	12.36 ± 5.49	< 0.001
Maturation rate (%)	82.07 ± 14.15	86.98 ± 11.66	84.89 ± 14.02	< 0.001
No. of 2PN embryos	8.92 ± 3.92	11.04 ± 4.79	7.31 ± 3.97	< 0.001
No. of D3 embryos	8.74 ± 3.89	10.79 ± 4.73	7.11 ± 3.92	< 0.001
No. of good-quality D3	4.34 ± 2.98	5.33 ± 3.56	3.58 ± 2.76	< 0.001
D3 good-quality embryo rate	49.19 ± 25.57	49.01 ± 23.73	49.31 ± 27.54	0.847
Blastocyst formation rate (%)	68.31 ± 26.42	68.77 ± 25.21	60.34 ± 32.77	< 0.001
No. of frozen D5 blastocysts	3.58 ± 2.83	4.81 ± 3.54	3.07 ± 2.88	< 0.001

Notes: Values are presented as mean ± SD. The maturation rate was calculated as the number of MII oocytes divided by the total number of retrieved oocytes. The blastocyst formation rate was defined as the number of blastocysts formed divided by the total number of normally fertilized (2PN) embryos available for extended culture. Statistical significance (P < 0.05) is observed across all embryological parameters among the three cohorts. Abbreviations: MII, metaphase II; 2PN, two pronuclei; D3, day 3; D5, day 5

Consistent with their robust ovarian reserve, patients in the 1–24% AFR group yielded the highest number of retrieved oocytes (16.61 ± 6.69), MII oocytes (14.36 ± 5.90), and the highest maturation rate (86.98 ± 11.66) across all cohorts (P < 0.001). However, as the AFR reached the ≥ 25% threshold, laboratory efficiency significantly diminished. Despite an adequate oocyte yield (14.59 ± 6.03), the ≥ 25% AFR group generated the lowest number of normal 2PN embryos per cycle (7.31 ± 3.97) and Day 3 embryos (7.11 ± 3.92) compared to the other cohorts (P < 0.001).

A pivotal finding was the discrepancy between early cleavage morphology and subsequent blastocyst formation. While the absolute number of good-quality Day 3 embryos was lowest in the ≥ 25% AFR group (3.58 ± 2.76, P < 0.001), the Good-quality Day 3 embryo rate remained remarkably consistent across the 0%, 1–24%, and ≥ 25% groups (49.19%, 49.01%, and 49.31%, respectively; P = 0.847). This suggests that the impact of chaotic fertilization may not be fully apparent during early morphological assessment.

In contrast, the subsequent developmental potential was severely compromised beyond the 25% threshold. The blastocyst formation rate in the ≥ 25% AFR group (60.34 ± 32.77%) was significantly lower than that observed in the 0% AFR (68.31 ± 26.42%) and 1–24% AFR groups (68.77 ± 25.21%, P < 0.001). This cumulative attrition ultimately resulted in the fewest number of frozen Day 5 blastocysts in the high AFR cohort (3.07 ± 2.88), representing a substantial reduction in the available embryo pool for transfer.

Independent impact of AFR on blastocyst formation

To account for the inherent intra-class correlation of sibling embryos within individual stimulation cycles, multivariable GEE models were employed to evaluate the independent association between AFR and blastocyst formation potential (Table 3).

Table 3.

Multivariable analysis of the impact of AFR on blastocyst formation using GEE

Variables	Unadjusted OR (95% CI)	P-value	Adjusted OR (95% CI)	P-value
Abnormal Fertilization Rate
AFR (Continuous, %)	0.997 (0.995–0.998)	< 0.001	0.996 (0.995–0.998)	< 0.001
AFR Group
0% (Reference)	1	/	1	/
1–24%	1.058 (1.009–1.109)	0.021	1.042 (0.993–1.094)	0.094
≥ 25%	0.947 (0.897-1.000)	0.052	0.933 (0.883–0.986)	0.014
Maternal Age (years)	/	/	0.992 (0.987–0.998)	0.010
BMI (kg/m²)	/	/	1.002 (0.994–1.009)	0.641
Infertility duration (years)	/	/	0.987 (0.978–0.996)	0.004
Infertility type
Primary (Reference)	/	/	1	/
Secondary	/	/	0.994 (0.953–1.038)	0.798
Cause of Infertility
Female factors only (Reference)	/	/	1	/
Male factors only	/	/	0.850 (0.809–0.893)	< 0.001
Combined factors	/	/	0.937 (0.888–0.987)	0.015
Unexplained	/	/	0.881 (0.819–0.947)	0.001
Ovarian Reserve
AMH (ng/mL)	/	/	0.990 (0.982–0.998)	0.010
Basal AFC	/	/	1.000 (0.996–1.005)	0.878
Basal FSH (IU/L)	/	/	1.004 (0.993–1.016)	0.455
Ovarian Stimulation & Response
Protocol	/	/	/	/
GnRH agonist (Reference)			1	/
GnRH antagonist	/	/	1.058 (1.006–1.112)	0.027
PPOS	/	/	0.875 (0.759–1.008)	0.064
Others	/	/	0.770 (0.545–1.087)	0.138
Total Gn dose (IU)	/	/	1.000 (1.000–1.000)	< 0.001
Gn duration (days)	/	/	1.033 (1.015–1.052)	< 0.001
No. of retrieved oocytes	/	/	0.996 (0.991–1.000)	0.050
No. of large follicles	/	/	1.009 (1.002–1.015)	0.009
E2 on trigger day (pg/mL)	/	/	1.000 (1.000–1.000)	0.112
P on trigger day (ng/mL)	/	/	1.015 (0.983–1.047)	0.364
LH on trigger day (mIU/mL)	/	/	1.007 (0.997–1.018)	0.145
Endometrial thickness (mm)	/	/	1.001 (0.992–1.010)	0.829

Notes: The GEE model utilized a logit link function and an exchangeable working correlation structure to account for the intra-class correlation of sibling embryos within a single cycle. Unadjusted ORs were derived from univariate GEE models for AFR. Adjusted ORs (aORs) were calculated after adjusting for all clinical and embryological confounders listed in the table, including maternal age, BMI, infertility duration/type/cause, ovarian reserve markers, and stimulation parameters. Abbreviations: OR, odds ratio; aOR, adjusted odds ratio; CI, confidence interval

When analyzed as a continuous variable, AFR exhibited a highly significant inverse association with the odds of blastocyst formation (unadjusted OR 0.997, 95% CI 0.995–0.998; P < 0.001). This negative correlation remained robust after adjusting for 22 clinical and embryological confounders, including maternal age, ovarian reserve, and stimulation protocols (adjusted OR (aOR) 0.996, 95% CI 0.995–0.998; P < 0.001).

The categorical GEE model further substantiated the non-linear threshold effect identified by the RCS curve. Compared to the reference group (0% AFR), a low level of abnormal fertilization (1–24% AFR) did not significantly impair sibling embryo development after multivariable adjustment (aOR 1.042, 95% CI 0.993–1.094; P = 0.094). However, crossing the identified biological inflection point (≥ 25% AFR) resulted in a significant reduction in developmental competence. Specifically, an AFR ≥ 25% independently reduced the odds of blastocyst formation by 6.7% (aOR 0.933, 95% CI 0.883–0.986; P = 0.014) relative to the 0% reference group.

Other independent negative predictors for blastocyst formation identified in the multivariable GEE model included advanced maternal age (aOR 0.992, 95% CI 0.987–0.998; P = 0.010), isolated male factor infertility (aOR 0.850, 95% CI 0.809–0.893; P < 0.001), and a longer duration of infertility (aOR 0.987, 95% CI 0.978–0.996; P = 0.004). Conversely, the GnRH antagonist protocol was associated with a higher likelihood of blastocyst formation compared to the GnRH agonist long/short protocols (aOR 1.058, 95% CI 1.006–1.112; P = 0.027).

Subgroup and sensitivity analyses

To evaluate the robustness of the identified 25% AFR threshold and its consistency across diverse clinical contexts, subgroup analyses were performed according to maternal age, ovarian reserve (AMH), and oocyte yield (Fig. 3).

Fig. 3.

Subgroup analysis of the association between AFR ≥ 25% and blastocyst formation potential. Notes: Forest plot illustrating the independent association between an AFR ≥ 25% and the odds of blastocyst formation across various clinical strata. The dots represent the adjusted odds ratios (aORs), and the horizontal lines represent the 95% CIs. The overall effect and all subgroup analyses were adjusted for maternal age, infertility factors, ovarian reserve, and stimulation protocols where applicable. The vertical solid line indicates an OR of 1.0 (no effect). Statistical significance was reached for the "Overall," "AMH ≥ 1.2 ng/mL," and "Oocytes retrieved 5–14" cohorts (P < 0.05). Abbreviations: y, years; n, number; aOR, adjusted odds ratio; CI, confidence interval

The detrimental impact of an AFR ≥ 25% on sibling blastocyst formation demonstrated a highly consistent downward trend (all adjusted odds ratios (aORs) < 1.0) across all evaluated strata, indicating the pervasive nature of this biological phenomenon. Notably, the statistical significance of this developmental impairment was robustly preserved in patients with adequate ovarian reserve (AMH ≥ 1.2 ng/mL: aOR 0.934, 95% CI 0.882–0.988, P = 0.018). In contrast, while a similar magnitude of effect was observed in the diminished ovarian reserve cohort (AMH < 1.2 ng/mL: aOR 0.947, 95% CI 0.744–1.207), the association did not reach statistical significance (P = 0.666). Importantly, this lack of significance must be interpreted with caution, as the baseline exclusion of cycles with < 5 oocytes inherently selects a specific subpopulation of low-reserve patients who exhibited an unexpectedly preserved oocyte yield, thereby limiting the generalizability of this specific subgroup finding to standard poor responders. A distinctive shift in the vulnerability profile was also observed when stratifying by oocyte yield. In the standard oocyte yield cohort (5–14 oocytes), an AFR ≥ 25% remained a significant independent risk factor for impaired blastocyst formation (aOR 0.913, 95% CI 0.847–0.984, P = 0.017). Conversely, this association lost statistical significance in the extreme high-yield cohort (≥ 15 oocytes: aOR 0.957, 95% CI 0.882–1.038, P = 0.290), suggesting that high embryo redundancy may partially mitigate the cohort-wide quality decline signaled by high AFR.

Furthermore, although the aOR for advanced maternal age (≥ 35 years: aOR 0.900, 95% CI 0.794–1.020) showed a pronounced downward trend, it did not reach statistical significance (P = 0.100). These findings collectively suggest that while an elevated AFR is universally indicative of suboptimal embryological outcomes, its clinical impact is most acutely manifest in cycles with standard ovarian responses and adequate embryo pools.

Discussion

This large-scale cohort study demonstrates a significant non-linear relationship between the abnormal fertilization rate and the blastocyst developmental potential of sibling normal zygotes. To our knowledge, this is the first study to characterize a critical clinical threshold of 25% for AFR using RCS. Beyond this limit, the developmental competence of sibling 2PN embryos significantly declines, independent of maternal age, ovarian reserve, and stimulation protocols. Furthermore, our subgroup analysis revealed that this detrimental “bystander effect” is particularly pronounced in patients with a standard oocyte yield (5–14 oocytes) or adequate ovarian reserve (AMH ≥ 1.2 ng/mL), suggesting that oocyte cohort resilience plays a pivotal role in modulating the impact of chaotic fertilization.

Consistent with the knowledge gap identified earlier, while the developmental arrest and reduced clinical viability of abnormally fertilized zygotes themselves (1PN/3PN) have been extensively documented [24–27], research investigating their broader implications as a predictive metric for sibling 2PN embryos remains scarce. Existing literature has predominantly focused on the morphokinetics, chromosomal integrity, and clinical outcomes of non-2PN embryos themselves [10, 11, 28]. Furthermore, while it is established that the developmental speed or quality of sibling embryos can serve as a surrogate marker for cycle potential and live birth rates [29, 30], the quantitative impact of AFR on the developmental competence of the 2PN cohort has often been overlooked. Some investigations suggest that an elevated incidence of abnormal fertilization reflects a systemic decline in oocyte cohort quality [31, 32], whereas others argue that such events are isolated stochastic occurrences with limited impact on the implantation or live birth rates of sibling 2PN embryos [24, 33].These discrepancies likely stem from the methodological limitations prevalent in prior research. First, the definition of AFR has historically lacked consensus. leading to significant biological noise that masks the true fertilization landscape [14]. Second, previous analyses predominantly assumed a monotonic linear relationship, thereby overlooking the non-linear “tipping points” inherent in complex biological systems. By refining the AFR denominator to strictly reflect the fertilized cohort and utilizing GEE models to adjust for embryonic clustering across over 100,000 embryos, our study resolves these controversies and provides a high-granularity understanding of how the fertilization landscape independently predicts sibling blastocyst competence.

The identification of a 25% threshold, rather than a continuous linear decline, strongly suggests that the human oocyte cohort possesses an intrinsic biological buffering capacity. When AFR remains below this inflection point, sibling 2PN embryos appear to maintain a robust developmental trajectory. From a mechanistic perspective, low-level abnormal fertilization may simply reflect localized, random events during spindle assembly or second polar body extrusion [34, 35]. However, once the 25% threshold is crossed, it likely signifies a systemic compromise of the oocyte cohort’s integrity. This temporal discrepancy—where developmental impairments are masked on Day 3 but acutely manifest during blastocyst—is closely linked to the timeline of embryonic genome activation (EGA) and the maternal-to-zygotic transition (MZT) [36, 37]. Early embryonic cleavages up to Day 3 are sustained primarily by maternally accumulated transcripts and nutrients within the oocyte [38], rendering early morphology relatively resilient to cohort-wide anomalies. Conversely, major EGA in humans occurs around the 4-to-8-cell stage, shifting developmental control to the newly formed embryonic genome [39]. The energy-intensive metabolic transitions required for blastocyst cavitation, expansion, and hatching demand flawless functional synergy across the embryonic apparatus [40]. Consequently, an elevated AFR (≥ 25%) reflects subtle, widespread cytoplasmic or metabolic deficiencies across the entire gamete pool that remain phenotypically silent during maternally driven cleavages but culminate in failure post-EGA. Importantly, we remain cautious in ascribing a direct causal link. High AFR does not necessarily cause sibling embryo failure, but rather acts as a highly sensitive surrogate biomarker for the underlying molecular “health” of the gamete pool.

A key highlight of our findings is the heterogeneity of the AFR impact across different ovarian response categories, particularly when stratified by the POSEIDON criteria [15, 16]. In ultra-high-yield cycles (≥ 15 oocytes), which typically represent optimal ovarian responders, the abundance of the embryo pool may confer a “dilution effect.” In these scenarios, developmental redundancy ensures that a sufficient number of high-competence 2PN embryos can still reach the blastocyst stage despite the presence of abnormal neighbors. In contrast, patients with standard oocyte yields (5–14 oocytes) or those meeting the POSEIDON criteria for diminished ovarian reserve (AMH < 1.2 ng/mL) exhibit a lower tolerance for fertilization errors. In our study, while the negative trend of AFR ≥ 25% was observed across all AMH strata, the statistical significance was robustly preserved only in the AMH ≥ 1.2 ng/mL cohort. Crucially, the lack of statistical significance within the low AMH (< 1.2 ng/mL) sub-cohort must first be attributed to the substantially reduced sample size in this specific stratum, which inherently diminished the statistical power to declare a significant difference despite the downward trend in the point estimate. Beyond this statistical limitation, the lack of significance may also stem from the inherent biological heterogeneity and reduced embryo redundancy characteristic of low prognosis groups. This distinction emphasizes the necessity of a nuanced, strata-specific interpretation of embryological data. This suggests that in patients with adequate ovarian reserve, an elevated AFR serves as a powerful and sensitive signal of cohort-wide quality degradation.

Conversely, for patients with AMH < 1.2 ng/mL, the lack of statistical significance—despite a downward trend in aOR—may stem from the inherent heterogeneity and reduced embryo redundancy characteristic of low prognosis groups. This distinction emphasizes the necessity of a nuanced, strata-specific interpretation of embryological data. This finding is of high clinical relevance, as it allows for more sophisticated, evidence-based prognostic counseling and laboratory expectation management. It emphasizes that AFR is a sensitive biomarker of the underlying molecular “health” of the gamete pool, particularly for the vast majority of patients who fall into the standard ovarian response category. Ultimately, our data advocate for the integration of AFR into routine laboratory quality control and clinical decision-making algorithms to enhance the precision of individualized ART outcomes.

Strengths and limitations

The primary strengths of this study lie in its large sample size and the corresponding methodological rigor. By leveraging a high-granularity database of 13,167 IVF cycles and 107,527 sibling embryos, this study provides a highly powered framework that allowed for robust, stratified non-linear estimations. Furthermore, this extensive dataset permitted the use of GEE models to account for the clustering effect of sibling embryos, a statistical rigor essential for embryo-level analyses. By refining the AFR denominator to exclude immature oocytes, we minimized biological noise and provided a more accurate assessment of the fertilization environment. Unlike previous reports that focused on clinical outcomes, our findings establish AFR as a sensitive laboratory biomarker for blastocyst prognosis, and the 25% threshold offers a reference for embryologists and clinicians.

When AFR reaches or exceeds this level, expectations for blastocyst yield should be adjusted downward, even if the absolute number of 2PN embryos appears adequate.

Despite the large sample size, several limitations must be acknowledged. First, as a retrospective analysis, we cannot definitively establish a causal link between abnormal fertilization events and sibling embryo failure. Second, because fertilization assessment was based on a single static observation at 16–18 h post-insemination rather than continuous time-lapse monitoring, we cannot completely exclude the possibility that a small fraction of annotated 1PN zygotes actually resulted from asynchronous pronuclear fading. Nevertheless, within the framework of routine IVF laboratory assessment, these zygotes still represented non-2PN fertilization outcomes at the standard clinical assessment time point and were therefore included in the operational definition of AFR. Although their precise biological origin may differ from true monopronuclear fertilization or parthenogenetic activation, they may still indicate atypical fertilization dynamics under conventional IVF conditions. Given the rarity of such events in conventional IVF, this classification limitation is unlikely to affect the overall stability of the identified 25% AFR threshold. Third, although we adjusted for a comprehensive array of confounders, the impact of specific laboratory variables, such as incubator types or culture media, over 4 years could not be fully eliminated. Furthermore, a selection bias exists within our low ovarian reserve subgroup analysis. To secure a robust denominator for AFR calculation, we excluded all cycles with < 5 retrieved oocytes. While necessary to prevent statistical skewness, this conditioning means that our low AMH (< 1.2 ng/mL) subgroup exclusively comprises individuals who managed to achieve a standard oocyte yield despite depleted reserves. This specific casuistry prevents the direct extrapolation of our AFR threshold findings to classical poor ovarian responders or patients undergoing mild stimulation protocols. Future prospective studies utilizing time-lapse morphokinetic data or transcriptomic profiling of sibling 2PN embryos from high-AFR cycles are warranted to further elucidate the underlying molecular mechanisms.

Clinical implications

The clinical implications of this large-scale study are twofold: First, we identify the abnormal fertilization rate (AFR) as a powerful, non-invasive “Day-1 biomarker” for assessing oocyte cohort integrity. By establishing a non-linear threshold of 25%, our findings provide embryologists and clinicians with a definitive, evidence-based tipping point to identify compromised cycles within 24 h of fertilization—long before traditional morphokinetic markers become informative.

Second, this study facilitates precision medicine in ART through targeted prognostic counseling. Our data demonstrate that the detrimental “bystander effect” of high AFR is most pronounced in standard ovarian responders (AMH ≥ 1.2 ng/mL). Conversely, for the depleted ovarian reserve cohort (< 1.2 ng/mL), the independent association did not reach statistical significance. This finding must be interpreted with strict caution and should not be definitively viewed as a biological absence of effect, given the diminished statistical power from a smaller sample size and the unavoidable selection bias introduced by excluding cycles with < 5 oocytes. Therefore, we advocate for integrating the 25% AFR threshold into routine laboratory standard operating procedures (SOPs) primarily for the early stratification of standard-prognosis patients. For those exceeding this threshold, clinicians can deliver more accurate, individualized expectations regarding blastocyst yields, thereby reducing patient anxiety and optimizing downstream laboratory resource allocation.

Conclusion

Our study establishes that the abnormal fertilization rate (AFR) is an independent negative predictor of sibling 2PN embryo blastocyst formation. We identified a critical clinical threshold at 25% AFR, beyond which the developmental competence of the entire oocyte cohort appears systematically compromised. While this detrimental ‘bystander effect’ of chaotic fertilization was robustly preserved in cycles with a standard oocyte yield (5–14 oocytes) or adequate ovarian reserve (≥ 1.2 ng/mL), our statistical constraints preclude a definitive extension of this conclusion to the lowest ovarian reserve strata due to diminished sample sizes and a restricted oocyte threshold. Recognizing the 25% AFR threshold allows for precise, strata-specific prognostic counseling and facilitates individualized laboratory management, ultimately enhancing the accuracy of outcome expectations in modern reproductive medicine.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution

All authors met the criteria for authorship. L.W. and M.R.L.: Conceptualization, Methodology, Software, Formal analysis, Investigation, Writing – original draft. X.N.: Formal analysis, Investigation, Writing – original draft. Y.Q.Y.: Investigation. J.L.: Supervision, Funding acquisition, Writing – review & editing. Z.B.: Supervision, Writing – review & editing. All authors contributed to Validation, Writing – review & editing, and approved the final manuscript.

Funding

This work was supported by National Key Research and Development Program of China (No. 2025YFC3508002-01) and the National Natural Science Foundation of China (No. 82501968). The funders had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Data availability

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

Declarations

Ethical approval

Ethical approval for this protocol was granted by the Institutional Review Board (IRB) of the Ethical Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (TJ-IRB20211280).

Consent for publication

All authors have read and approved the final version of the manuscript, agree to submit it for publication in this journal, and consent to the publication of the manuscript, including all tables, figures and supplementary materials.

Conflict of interest

The authors declare that there were no competing interests in this study.