Introduction: Although pre-treatment with a GnRH agonist can reduce the size of adenomyosis lesions, the supra-physiological hormone level induced by controlled ovarian hyperstimulation (COH) may negate the usefulness of the GnRH agonist in patients with adenomyosis lesions, leading to continued poor outcomes in fresh embryo transfer cycles during in vitro fertilization (IVF). It is unclear whether GnRH agonist pre-treatment before starting the long GnRH agonist protocol for IVF/ICSI (intracytoplasmic sperm injection) can improve cumulative live birth rate (CLBR) of infertile women with adenomyosis. Method: In this retrospective cohort study, a total of 374 patients diagnosed as adenomyosis (477 cycles) underwent IVF/ICSI with long GnRH agonist protocol with or without GnRH agonist pre-treatment between January 2009 and June 2018. Logistic regression was used to assess the association between GnRH agonist pre-treatment and pregnancy outcome after adjusting for confounding factors. Results: The live birth rate in fresh embryo transfer cycles was higher in the non-pre-treatment group than in the GnRH agonist pre-treatment group (37.7 vs. 21.2%, P = 0.028); the adjusted odds ratio (OR) for the long agonist protocol without pre-treatment was 1.966 (95% CI: 0.9–4.296, P = 0.09). The CLBR was higher in the non-pre-treatment group than in the GnRH agonist pre-treatment group (40.50 vs. 27.90%, P = 0.019); the adjusted OR for the long agonist protocol without pre-treatment was 1.361 (95% CI: 0.802–2.309, P = 0.254). Conclusion: Our results indicated that GnRH agonist pre-treatment before starting the long GnRH agonist protocol does not improve the live birth rate in fresh embryo transfer cycles or CLBR in infertile women with adenomyosis after IVF/ICSI treatment when compared to that in non-pre-treated patients. A subsequent prospective randomized controlled study is needed to confirm these results.

STUDY QUESTION: What are the unrecognized risk factors for recurrent implantation failure (RIF) as defined in the ESHRE recommendation? SUMMARY ANSWER: Anti-Müllerian hormone (AMH) is the strongest predictor for RIF, followed by chronic endometritis (CE), intrauterine adhesions, and BMI. WHAT IS KNOWN ALREADY: Advanced age is a well-known risk factor for implantation failure, and the definition of RIF was stratified by age in the 2023 ESHRE recommendation. However, the literature identifies other risk factors, including CE, endometriosis, BMI, endometrial polyps, intrauterine adhesions, hydrosalpinx, uterine malformation, submucosal myoma, polycystic ovary syndrome, thyroid dysfunction, rheumatic diseases, and hyperprolactinemia, to be associated with implantation failure. In addition, our clinical experience suggests AMH and a history of previous livebirth affect RIF. It remains unclear which of these factors are the best predictors of RIF. STUDY DESIGN, SIZE, DURATION: A cohort study drawn from ART cycles between June 2019 and June 2022. PARTICIPANTS/MATERIALS, SETTING, METHODS: Two hundred and ninety-eight RIF patients and 2056 controls (women who achieved successful embryo implantation within 1–2 transfer cycles) were identified from 15 329 ART cycles at the Reproductive Medical Center at the First Affiliated Hospital of Sun Yat-sen University. RIF was defined according to the recommendation of ESHRE 2023. Basic characteristics, reproductive history, laboratory indicators (autoantibodies and endocrine factors), ultrasound, laparoscopy, hysteroscopy, hysterosalpingography, biopsy, and immunohistochemistry results were collected from the electronic medical record system. The Random Forest procedure was applied to build a machine learning model for predicting RIF. Overall predictive accuracy was assessed by using the AUC of receiver–operator characteristic curve and calibration plots. The SHapley Additive exPlanations (SHAP) framework was used to interpret the model. MAIN RESULTS AND THE ROLE OF CHANCE: From 32 variables, elevated AMH level and greater number of live births were associated with lower risk of RIF, while CE, intrauterine adhesions, high FSH level, high testosterone level, advanced female age, polyps, history of recurrent pregnancy loss, history of cesarean section, polycystic ovary syndrome, and rheumatic diseases were associated with higher risk of RIF according to the established random forest model. The predictive model yielded AUCs of 0.83 (95% CI: 0.80–0.86) in training dataset and 0.78 (95% CI: 0.73–0.84) in testing dataset. The calibration curve indicated good predictive performance in both training and testing datasets. SHAP values indicated that AMH had the greatest influence on the RIF risks, whereas CE, intrauterine adhesions, and BMI were the second, third, and fourth most significant risk factors for predicting RIF, respectively. LIMITATIONS, REASONS FOR CAUTION: This research was limited by its retrospective design from a single reproductive medical center. Moreover, some diseases, such as polyps, submucosal myomas, and rheumatic diseases, had been treated before ART, which indicates that these factors impact RIF even after treatment. WIDER IMPLICATIONS OF THE FINDINGS: In addition to age, certain high-risk factors, such as AMH, should also be included in the considerations for RIF. Patients with a combination of these high-risk factors may require more attempts to achieve a successful pregnancy. STUDY FUNDING/COMPETING INTEREST(S): This work was supported by the Guangzhou Municipal Science and Technology Project under Grant (202206010003) and the National Natural Science Foundation of China under Grant (81871159). None of the authors has any conflict of interest. TRIAL REGISTRATION NUMBER: N/A.

Previous studies have suggested that the trophoblast cells inhibit the proliferation of peripheral natural killer (pNK) cells and the level of pNK cells decrease in middle and late pregnancy stage among healthy women. The change of pNK cells level during early pregnancy and the relationship between the change of pNK level and pregnancy outcomes in women with unexplained recurrent pregnancy loss (URPL) has not been sufficiently explored. This study aims to characterize the level of pre-pregnancy pNK to early pregnancy among women with URPL and whether the change in the level of pNK cells in early pregnancy from pre-pregnancy can predict pregnancy outcomes. In this prospective cohort study, 1758 women with recurrent pregnancy loss were recruited between January 2017 and December 2021, of which 252 URPL women had pre-pregnancy and early pregnancy (4-6 weeks gestation) pNK measurements. These 252 women were divided into two groups: those with a lower gestational pNK levels (Group 1) compared to pre pregnancy, and those without (Group 2). Their respective outcomes on live birth and pregnancy loss were comparatively analyzed using Chi-square and Student's t test. Candidate influence factors for live birth were selected using the Akaike information criterion (AIC). Then the participates were randomly divided into training and testing groups. Multivariable logistic regression model was performed, and nomogram was calculated to assess the possibility of live birth. Predictive accuracy was discriminated by area under the receiver operating characteristic curve (ROC), validated by plotting the predicted probabilities and the observed probabilities. Hosmer-Lemeshow test was used to assess the goodsess of fit. When early gestational pNK cells levels were compared with pre pregnancy pNK cells levels, 61.5% (154) women had a comparatively lower early gestational pNK cells levels versus 38.9% (98) women with increase or no change of their pNK cells levels. The live birth rate in Group 1 was 89.0% (137/154), which was significantly higher than 49.0% (48/98) of Group 2 (p<0.001). Decrease of pNK cells level (odds ratio [OR]=1.36, 95% CI 1.22-1.55, p<0.001) and anti-Muellerian hormone (AMH) (OR=1.41, 95% CI 1.14-1.81, p=0.003) were important predicting factors for higher live birth. Female BMI (OR=0.97, 95% CI 0.82-1.15, p=0.763) and parity (OR=1.61, 95% CI 0.71-4.12, p=0.287) were also predicting factors. Furthermore, the area under the ROC curve of the model to diagnose of live birth was 0.853 with a sensitivity of 81.6% and specificity of 78.0% using the training dataset. And the Hosmer-Lemeshow test showed that the model was a good fit (p = 6.068). We report a comparative decrease in pNK cells levels in over 60% URPL women at 4-6 weeks of gestation, when compared to their pre-pregnancy pNK cells level. Compared to pre pregnancy pNK cells levels, a decrease pNK cells level during early pregnancy might be a useful predictor for LBR in women with URPL.

Can blastocyst aneuploidy be predicted for patients with previous aneuploid pregnancy loss (PAPL) and receiving preimplantation genetic testing for aneuploidy (PGT-A)? Multivariable logistic regression models were established to predict high risk of blastocyst aneuploidy using four identified factors, presenting good predictive performance. Aneuploidy is the most common embryonic chromosomal abnormality leading to pregnancy loss. Several studies have demonstrated a higher embryo aneuploidy rate in patients with PAPL, which has suggested that PGT-A should have benefits in PAPL patients intending to improve their pregnancy outcomes. However, recent studies have failed to demonstrate the efficacy of PGT-A for PAPL patients. One possible way to improve the efficacy is to predict the risk of blastocyst aneuploidy risk in order to identify the specific PAPL population who may benefit from PGT-A. We conducted a multicenter retrospective cohort study based on data analysis of 1119 patients receiving PGT-A in three reproductive medical centers of university affiliated teaching hospitals during January 2014 to June 2020. Acohort of 550 patients who had one to three PAPL(s) were included in the PAPL group. In addition, 569 patients with monogenic diseaseswithout pregnancy loss were taken as the non-PAPL group. PGT-A was conducted using single nucleotide polymorphism microarrays and next-generation sequencing. Aneuploidy rates in Day 5 blastocysts of each patient were calculated and high-risk aneuploidy was definedas a rate of ≥50%. Candidate risk factors for high-risk aneuploidy were selected using the Akaike information criterion andwere subsequently included in multivariable logistic regression models. Overall predictive accuracy was assessed using the confusionmatrix, discrimination by area under the receiver operating characteristic curve (AUC), and calibration by plotting the predictedprobabilities versus the observed probabilities. Statistical significance was set at P < 0.05. Blastocyst aneuploidy rates were 30 ± 25% and 21 ± 19% for PAPL and non-PAPL groups, respectively. Maternal age (odds ratio (OR) = 1.31, 95% CI 1.24–1.39, P < 0.001), number of PAPLs (OR = 1.40, 95% CI 1.05–1.86, P = 0.02), estradiol level on the ovulation trigger day (OR = 0.47, 95% CI 0.30–0.73, P < 0.001), and blastocyst formation rate (OR = 0.13, 95% CI 0.03–0.50, P = 0.003) were associated with high-risk of blastocyst aneuploidy. The predictive model based on the above four variables yielded AUCs of 0.80 using the training dataset and 0.83 using the test dataset, with average and maximal discrepancies of 2.89% and 12.76% for the training dataset, and 0.98% and 5.49% for the test dataset, respectively. Our conclusions might not be compatible with those having fewer than four biopsied blastocysts and diminished ovarian reserves, since all of the included patients had four or more biopsied blastocysts and had exhibited good ovarian reserves. The developed predictive model is critical for counseling PAPL patients before PGT-A byconsidering maternal age, number of PAPLs, estradiol levels on the ovulation trigger day, and the blastocyst formation rate. This prediction model achieves good risk stratification and so may be useful for identifying PAPL patients who may have higher risk of blastocyst aneuploidy and can therefore acquire better pregnancy outcomes by PGT-A. This work was supported by the National Natural Science Foundation of China under Grant (81871159). No competing interest existed in the study. N/A.