A cycle-based model to predict no usable blastocyst formation following cycles of in vitro fertilization in patients with normal ovarian reserve.

The study included a total of 7,901 patients, of whom 446 (5.64%) were in the no available blastocyst group. A total of 4,090 fresh embryo transfer (ET) cycles were performed. Baseline characteristics: Basal follicle-stimulating hormone (bFSH), basal luteinizing hormone (LH), anti-Müllerian hormone (AMH), antral follicle count (AFC), mean gonadotropin dosage, human chorionic gonadotropin (hCG) trigger day estradiol (E2) level, progesterone (P) level, and number of dominant follicles were significantly different between the two groups ( p   0.05) (Table  1 ). In the blastocyst group, the number of oocytes retrieved, mature (MII) oocytes, 2-pronuclear (2PN) embryos, cleavage-stage embryos, Day 3 (D3) embryos, D3 embryos available for blastocyst culture, 2PN rate, cleavage rate, D3 high-quality embryo rate, and fresh cycle clinical pregnancy rate were all significantly higher than the non-blastocyst group ( p   0.05) (Table  2 ). Table 1 Patient demographics and baseline characteristics AVAILABLE.BLASTOCYST group NO.AVAILABLE.BLASTOCYST group P -value N 7455 446 Female Age(Years) 30.4 ± 3.9 30.7 ± 4.2 0.132 Male Age(Years) 31.2 ± 4.5 31.6 ± 4.8 0.141 BMI 23.3 ± 3.7 23.4 ± 3.7 0.493 Infertile Duration 3.4 ± 2.6 3.6 ± 2.6 0.123 Basal FSH(IU/L) 6.2 ± 1.6 6.4 ± 1.7 0.001 Basal LH(IU/L) 6.2 ± 4.0 5.7 ± 3.4 0.009 AMH(ng/ml) 5.0 ± 3.5 4.5 ± 3.1 0.001 AFC 16.5 ± 5.8 15.0 ± 5.6 < 0.001 Gonadotropin Dosage(IU) 2268.0 ± 945.9 2385.1 ± 934.3 0.011 Gonadotropin Days(days) 11.7 ± 2.5 11.5 ± 2.6 0.173 E2 on hCG trigger day(pg/ml) 2519.0 ± 1499.6 1942.3 ± 1112.5 < 0.001 LH on hCG trigger day(IU/L) 1.0 (0.5–1.8) 1.0 (0.5–1.8) 0.352 P on hCG trigger day(ng/ml) 0.6 (0.4–0.9) 0.5 (0.3–0.8) 0.023 Dominant follicles on hCG trigger day 12.1 ± 5.1 9.4 ± 4.6 < 0.001 Infertility Type 0.28  Primary 3665 (49.2%) 231 (51.8%)  Secondary 3790 (50.8%) 215 (48.2%) Infertility Factor 0.74  Tubal factor 5099 (68.4%) 309 (69.3%)  Ovulation disturbance 1378 (18.5%) 73 (16.4%)  Endometriosis 411 (5.5%) 29 (6.5%)  Male factor 567 (7.6%) 35 (7.8%) Values are the mean ± standard deviation or number (percentage). hCG, human chorionic gonadotropin; BMI, body mass index; FSH, follicle-stimulating hormone; LH, luteinizing. hormone; PRL, prolactin; E2, estrogen; P, progesterone; AFC, total antral follicle count; Gn, gonadotropin; CI, confidence interval Patient demographics and baseline characteristics Values are the mean ± standard deviation or number (percentage). hCG, human chorionic gonadotropin; BMI, body mass index; FSH, follicle-stimulating hormone; LH, luteinizing. hormone; PRL, prolactin; E2, estrogen; P, progesterone; AFC, total antral follicle count; Gn, gonadotropin; CI, confidence interval Table 2 Patients laboratory index and clinical outcomes AVAILABLE.BLASTOCYST group NO.AVAILABLE.BLASTOCYST group P -value N 7455 446 Oocyte 14.3 ± 6.3 9.9 ± 4.8 < 0.001 No. of MII 12.4 ± 5.6 8.0 ± 3.9 < 0.001 No. of 2PN 9.1 ± 4.3 5.1 ± 2.7 < 0.001 No. of Cleavage 8.9 ± 4.2 4.9 ± 2.6 < 0.001 No. of D3 embryo 7.8 ± 3.9 4.0 ± 1.7 < 0.001 No. of high quality D3 embryo 4.0 (2.0–6.0) 1.0 (0.0–2.0) < 0.001 No. of embryo used for blastocyst culture 7.8 ± 4.3 3.6 ± 2.5 < 0.001 2PN rate 0.7 ± 0.2 0.6 ± 0.2 < 0.001 Cleavage rate 1.0 ± 0.1 0.9 ± 0.2 < 0.001 Top day 3 embryo rate 0.5 (0.3–0.7) 0.2 (0.0-0.5) < 0.001 Embryo type transfered in fresh cycle(%) < 0.001  D2 1 (0.0%) 0 (0.0%)  D3 2340 (61.3%) 269 (99.3%)  D4 2 (0.1%) 2 (0.7%)  D5 1351 (35.4%) 0 (0.0%)  D6 126 (3.3%) 0 (0.0%) Embryo number transfered in fresh cycle < 0.001  single embryo transfer 2853 (74.7%) 79 (29.2%)  double embryo transfer 967 (25.3%) 192 (70.8%) Clinical pregnancy rate of fresh cycle(%) 2458 (64.3%) 138 (50.9%) < 0.001 Fresh embryo transfer cycles 3820 271 Live birth rate of fresh cycles(/ET cycles)* 55.80%(1972/3534) 43.04% (102/237) 0.15 Values are the mean ± standard deviation or number (percentage). ET, embryo transfer. 2PN Rate = Number of 2PN Oocytes/Number of MII Oocytes; Cleavage Rate = Number of Cleaved Embryos/Number of Fertilized Oocytes; Top Day 3 Embryo Rate = Number of Top Quality Embryos on Day 3/Total Number of Embryos on Day 3. * Patients who were more than 3 months pregnant but had not yet delivered were not included (14 patients in the No Avilable Blastocyst group and 146 patients in the Avilable Blstocyst group) Patients laboratory index and clinical outcomes 43.04% (102/237) Values are the mean ± standard deviation or number (percentage). ET, embryo transfer. 2PN Rate = Number of 2PN Oocytes/Number of MII Oocytes; Cleavage Rate = Number of Cleaved Embryos/Number of Fertilized Oocytes; Top Day 3 Embryo Rate = Number of Top Quality Embryos on Day 3/Total Number of Embryos on Day 3. * Patients who were more than 3 months pregnant but had not yet delivered were not included (14 patients in the No Avilable Blastocyst group and 146 patients in the Avilable Blstocyst group) The candidate predictors included in the original model were FemaleAge, MaleAge, Fertilizationtype, BMI, Infertility type, Infertility factor, Infertile time, bFSH, bLH, HCGdayE2, HCGdayLH, HCGdayP, HCGdayLargeFollicle, AMH, AFC, Gndosage, Gndays, oocytesNumber, MII, 2PN, cleavagenumber, D3embryo, HQ-D3embryo, and embryoforBC. Using LASSO regression analysis performed on the entire patient cohort, these were reduced to the three most significant predictors: D3embryo, HQ-D3embryo, and embryoforBC. The LASSO regression path diagram and coefficient profiles of these characteristics are illustrated in Fig.  2 A and B. The most regularized and parsimonious model, with a cross-validated error within one standard error of the minimum, included these three variables. Further multivariate logistic analyses were conducted across the entire cohort, with results shown in Table  3 . Fig. 2 Variable selection using the least absolute shrinkage and selection operator (LASSO) regression algorithm. (A) Lasso regression path diagram. (B) LASSO coefficient profiles of the characteristics. Parameters were screened out by 10-fold cross-validation and using lambda.1se as the criteria Variable selection using the least absolute shrinkage and selection operator (LASSO) regression algorithm. (A) Lasso regression path diagram. (B) LASSO coefficient profiles of the characteristics. Parameters were screened out by 10-fold cross-validation and using lambda.1se as the criteria Table 3 Results of Multivariate Logistic regression Characteristic N Event N OR 95% CI p -value No. of D3 embryo 7901 446 0.86 0.78, 0.95 0.004 No. of high quality D3 embyo 7901 446 0.58 0.53, 0.65 < 0.001 No. of embryo used for blastocyst culture 7901 446 0.78 0.72, 0.84 < 0.001 OR = Odds Ratio, CI = Confidence Interval Results of Multivariate Logistic regression OR = Odds Ratio, CI = Confidence Interval The developed model for estimating the lack of usable blastocysts used the selected variables, incluiding three independent predictors (D3embryo, HQ-D3embryo, and embryoforBC) as indicators. The nomogram for prediction is depicted in Fig.  3 . Each parameter was assigned a vertical extension (shown in the top points bar) individually. The total score was determined by summing up the scales for each factor. The overall point projected on the bottom scale suggests the likelihood of lack of available blasyst. The equation for the nomogram is as follows: logit (NO. AVAILABLE.BLASTCYST.RCD) = 0.59400-0.11383*D3EMBRYO-0.57934*HQ.D3EMBRYO − 0.28251*EMBRYOFORBC. Fig. 3 Nomogram to predict the risk of having no available blastocysts for transfer. The points for each variable were determined by drawing a vertical line from the value of the variable to the point scale. The total points were calculated by summing the individual points, and a downward line was drawn from the “Total Points” axis to intersect with the “Probability of No Available Blastocyst” axis, which provided the estimated probability Nomogram to predict the risk of having no available blastocysts for transfer. The points for each variable were determined by drawing a vertical line from the value of the variable to the point scale. The total points were calculated by summing the individual points, and a downward line was drawn from the “Total Points” axis to intersect with the “Probability of No Available Blastocyst” axis, which provided the estimated probability To investigate the forecast value of the model, we performed analyses of the ROC curves. The results indicated that our model demonstrated good discriminative potential, with an AUC of 0.879 (95%CI = 0.862–0.892). The AUC confidence interval and significance test were obtained using the bootstrap method (bootstrap resampling times = 500) (Fig.  4 A and B). The calibration plots of the nomogram model are plotted in Fig.  4 C, which demonstrate a good correlation between the observed and predicted no usable blastocyst. The calibration curve was relatively close to the ideal curve, which indicates that the predicted results were consistent with the actual findings. The Fig.  4 D displays the Decision Curve Analysis (DCA) curves related to the nomogram. A high-risk threshold probability indicates the chance of significant discrepancies in the model’s prediction when clinicians encounter major flaws while utilizing the nomogram for diagnostic and decision-making purposes. This research shows that the nomogram offers substantial net benefits for clinical application through its DCA curve. In addition, we compared the predictive potential of the single variables and the predictive model using the ROC curves. The AUC of the nomogram was significantly. (Table  4 ) Fig. 4 Model evaluation and validation results. The AUC confidence interval and significance test were obtained using the bootstrap method (bootstrap resampling times = 500) (Fig.  4A and B ). The calibration plots of the nomogram model are plotted in Fig.  4C . The Fig.  4D displays the Decision Curve Analysis (DCA) curves related to the nomogram Model evaluation and validation results. The AUC confidence interval and significance test were obtained using the bootstrap method (bootstrap resampling times = 500) (Fig.  4A and B ). The calibration plots of the nomogram model are plotted in Fig.  4C . The Fig.  4D displays the Decision Curve Analysis (DCA) curves related to the nomogram Table 4 Predictive accuracy of the model and various variables for no available blastocyst Variable AUC 95%CI Best threshold Specificity(%) Sensitivity(%) PPV(%) NPV(%) No. of D3 embryo 0.8232 0.796–0.857 5.5 0.6943 0.8408 0.1413 0.9865 No. of high quality D3 embyo 0.8431 0.822–0.8609 2.5 0.719 0.8251 0.1494 0.9857 No. of embryo used for blastocyst culture 0.827 0.807–0.848 4.5 0.7657 0.7332 0.1577 0.9796 Model 0.8794 0.8612-0.890 -2.5961 0.7902 0.8251 0.1905 0.9869 Predictive accuracy of the model and various variables for no available blastocyst Fig. 5 A worked example for how to use the nomogram—A patient obtained 5 Day 3 embryos, among which 1 was high quality, 4 embryos were used for blastocyst culture A worked example for how to use the nomogram—A patient obtained 5 Day 3 embryos, among which 1 was high quality, 4 embryos were used for blastocyst culture The nomogram is used to calculate the probability of no blastocyst development. The predictor score that corresponds to each variable (black arrow) is read on the upper scale. The total points (196 points) were calculated by summing the individual points and a downward line was drawn from the “Total Points” axis to intersect with the “Probability of No Available Blastocyst” axis, which provided the estimated probability of no available blastocyst (15.65%). The equation to use linear predictor to predict no available blastocyst is as follows: P (NO.AVAILABLE.BLASTCYST) = 1/ (1 + e− (linear Predictor) (Fig.  5 ).

A retrospective cohort study was conducted to analyze the clinical data of patients who had undergone IVF treatment at the Reproductive Medicine Center of the Henan Provincial People’s Hospital, Henan, China from September 1st, 2016 to March 1st, 2024. For this study, patients who underwent blastocyst culture after oocyte retrieval in all fresh cycles were included. Patients were excluded if they had received cycles with preimplantation genetic testing (PGT), sperm/oocyte donation, surgically retrieved sperm or had chromosomal abnormalities. Figure  1 shows a flowchart depicting the selection process and the inclusion criteria. All participants provided informed consent. This study was approved by the Reproductive Ethics Committee of Henan Provincial People’s Hospital (SYSZLL-2019110401). Fig. 1 Flowchart of the data collection process. Notes: AMH, anti-Müllerian hormone; AFC, antral follicle count Flowchart of the data collection process. Notes: AMH, anti-Müllerian hormone; AFC, antral follicle count Controlled ovarian stimulation and blastocyst culture based on the patient’s age, ovarian function, and other factors, a suitable controlled ovarian stimulation protocol was selected. On the day of oocyte retrieval, sperm and oocytes were inseminated. On day 2 post-retrieval, the fertilization status of the embryos was observed. On day 3 post-retrieval, the embryos were scored based on parameters such as cell number, embryo fragmentation, and blastomere uniformity. Considering the embryo growth, the patient’s physical condition, and their preference, the decision was made whether to proceed with blastocyst culture. Embryos selected for blastocyst culture were transferred to blastocyst culture dishes and cultured until days 5 and 6 post-retrieval, when blastocyst scoring was performed. The Gardner scoring system was used to evaluate the blastocysts [ 8 , 9 ]. Blastocysts scored 4BB or higher (with neither inner cell mass nor trophectoderm scored as C) were regarded as high-quality blastocysts. Blastocysts that could not be used for transfer or cryopreservation because of poor quality were identified as failures of blastocyst formation. There was no change in our laboratory protocols during our study period. The endpoint of our study was that no blastocyst formation is available. Data analysis was implemented using EmpowerStats statistical software (X&Y Solutions). Non-normal data were presented as median (interquartile ranges). In the univariate analysis, chi-square test or Fisher’s exact test was used to analyze the categorical variables, while the Student’s t-test or rank-sum test was used to examine the continuous variables. The least absolute shrinkage and selection operator (LASSO) binary logistic regression analyse was used for multivariate analysis to screen the independent risk factors and build a prediction nomogram for no available blastcoyst. LASSO was implemented by the “glmnet” package in R for variable selection and nomogram was created with the “rms” package. The “glmnet” package was utilized for efficient variable selection using LASSO and ridge regression techniques, helping simplify the model and improve prediction accuracy. The “rms” package was then employed to create a prediction nomogram, translating complex statistical models into user-friendly graphical tools to estimate probabilities of events like no blastocyst formation. The prediction nomogram can offer graphical representations of the selected factors in the model and to facilitate users in calculating probabilities [ 10 ]. For validation, we used internal validation with bootstrapping (500 repetitions) to reduce the overfit of the model and to obtain relatively unbiased estimates [ 11 ]. The same “rms” analysis was performed in the 500 data sets and a shrinkage factor was calculated by analyzing the variability of the models. It was used to correct the final model and the prediction formula was extracted from the data. The performance of the nomogram was assessed using the receiver operating characteristic (ROC) curve and calibration curve, with the area under the ROC curve (AUC) ranging from 0.5 (no discriminant) to 1 (complete discriminant). Calibration curves were also constructed to validate the model. A decision curve analysis (DCA) was also performed to determine the net benefit threshold of prediction [ 12 ].

We have successfully devised a cycle-specific predictive model for assessing the risk of unsuccessful blastocyst formation in women undergoing IVF. Despite our model’s promising calibration and performance metrics, further validation is essential. Once validated externally, our model could substantially aid clinical decision-making, potentially recommending day 3 embryo transfer for patients at elevated risk of blastocyst formation failure. This approach promises to enhance the outcomes of assisted reproductive technologies by personalizing patient care.

The lack of available embryos imposes not only a financial burden on patients undergoing IVF but also induces substantial psychological stress and can lead to strained doctor-patient relationships. In this study, we identified key clinical and laboratory factors influencing blastocyst formation through robust data analysis, and developed a predictive model for patients under 40 with normal ovarian reserve post-oocyte retrieval (Fig.  5 ). Our model primarily incorporates predictors such as the number of day 3 embryos, top-quality day 3 embryos, and the number of embryos used for blastocyst culture, with these variables proving statistically significant in multivariate logistic regression analysis. Embryo quality assessment and blastocyst culture are key laboratory routines. Current methods include traditional morphological assessment [ 8 , 13 ], non-invasive oocyte and embryo viability assessment via metabolite measurement [ 14 , 15 ], and time-lapse imaging [ 16 , 17 ]. These focus on individual embryos, not entire retrieval cycles and they incur additional costs. Predictive models based on clinical data could advise on blastocyst culture necessity, reducing IVF costs and aiding clinician decisions. We utilized LASSO regression for effectively managing variable selection and interaction testing without overfitting. This ensured the inclusion of clinically relevant variables, aligning well with established literature. Specifically, factors like the quantity and quality of day 3 embryos were emphasized due to their documented impact on blastocyst formation, supported by studies [ 18 , 19 ]. This methodological advancement results in a more streamlined and concise model that maintains high predictive accuracy evidenced by an AUC of 0.88. (Researchs before reported an AUC of 0.80) [ 20 ]. The resulting nomogram provides a graphical synthesis of the predictive model, facilitating intuitive calculations for clinicians to estimate specific probabilities of outcomes. This is especially beneficial given the increased accessibility and application of nomograms in reproductive medicine [ 21 , 22 ]. Our research also reflects real-world scenarios by allowing patients the option to select some or all embryos for blastocyst culture, which is more consistent with actual clinical practices compared to the more rigid criteria in Dessolle et al. Their models predict day 5 blastocyst formation [ 20 ], assuming least five zygotes on Day 1 after fertilization and all were used for blastocyst culture. Embryonic transfer decisions are complex, and our model can support the implementation of individualized transplantation strategies in clinical practice. Our model demonstrated robust performance upon testing through repeated sampling, evidenced by strong calibration and discrimination metrics. However, calibration plots suggested a slight overestimation of results, which we acknowledge as a potential area for refinement. Furthermore, our study evaluates the model’s clinical utility through internal validation and the incorporation of decision curve analysis (DCA). This analysis helps clinicians determine the most beneficial threshold for treatment decisions, balancing the benefits of true positives against the costs of false positives. While retrospective analysis naturally carries inherent biases, we meticulously controlled selection bias by setting stringent inclusion criteria related to ovarian reserves and COS protocols. These measures aimed to ensure population homogeneity and enhance the clinical applicability of our findings. Acknowledging that our model’s validation was restricted to internal data, we recognize the need for external validation across multiple centers. Such validation is crucial before clinical adoption, given the common performance variances seen in models tested outside their original population [ 23 , 24 ]. Future studies should thus focus on broader data collection to validate and generalize our model’s utility.

In vitro fertilization (IVF) has emerged as a widely adopted assisted reproductive technology (ART) for addressing infertility in couples. A prevalent protocol in IVF involves culturing embryos to the blastocyst stage prior to uterine transfer. The selection of high-quality blastocysts for transfer is critical, as it can significantly enhance the success rate of IVF treatment [ 1 , 2 ]. These optimal blastocysts demonstrate superior implantation potential [ 3 ] and are associated with improved live birth outcomes. Contemporary trends in ART are increasingly favoring elective single embryo transfer (eSET) and extended culture blastocyst transfer [ 4 ]. These approaches are being implemented to mitigate the incidence of multiple gestations, a common complication associated with assisted reproductive techniques. The shift towards eSET and blastocyst transfer represents a strategic evolution in IVF protocols, aimed at optimizing treatment outcomes while minimizing associated risks. However, not all embryos are able to reach the blastocyst stage, and some may arrest at earlier stages of development. This can lead to the situation where after blastocyst culture, there are no viable blastocysts available for transfer. Recent studies have demonstrated that the rate of cultured embryos progressing to the blastocyst stage exhibits considerable variation, with reported ranges spanning from 28 to 55% [ 5 ]. Notably, at the individual patient level, blastocyst formation rates display even greater heterogeneity, ranging from complete failure (0%) to near-perfect success (approaching 100%). This marked inter-patient variability underscores the complex interplay of factors influencing embryonic development and highlights the need for personalized approaches in assisted reproductive technologies. Thus, understanding the factors that influence embryo development and predicting the likelihood of unsuccessful blastocyst formation is crucial for optimizing IVF outcomes. Currently, reliable predictors of blastocyst development remain a subject of debate in the field of reproductive medicine [ 6 , 7 ]. Limited efforts have been made to establish cycle-specific predictive models that assess the risk of blastocyst formation failure in individual patients. In clinical practice, such a model is highly desirable, particularly for patients with good ovarian reserve and heightened expectations regarding blastocyst culture outcomes. A robust prediction model would serve multiple purposes: Provide early warning of potential blastocyst culture failure, optimize cost-effectiveness of IVF procedures, mitigate psychological stress for patients, and facilitate personalized clinical management strategies. In this study, we conducted a comprehensive analysis of patients undergoing IVF with normal ovarian reserve, incorporating a wide range of factors associated with blastocyst formation. We set out to establish a well-calibrated nomogram model to predict the likelihood of failed blastocyst culture following oocyte retrieval. This model could provide an accurate reference for both patients and clinicians, facilitating more informed decision-making and optimized management of IVF-ET cycles.