Factors that Influence Re-Pregnancy Failure and Prediction Models After Complete Curettage for Missed Abortion

Abstract

Background:

Missed abortion (MA), a type of spontaneous abortion, has become increasingly common in early pregnancy. Retained embryos may lead to dead fetus syndrome or severe hemorrhage, affecting the physical and mental health of women. This study selected MA patients undergoing uterine evacuation to construct a predictive model for factors that influence subsequent pregnancy failure, aiming to improve patient prognosis.

Methods:

A retrospective analysis of 466 women with MA after a complete uterine curettage (May 2021–May 2023) was conducted. Patients were randomly divided into a modeling (326) and a validation (140) group; the modeling group was further classified by re-pregnancy outcome. Logistic regression was used to assess risk factors for re-pregnancy failure after a complete uterine curettage for MA. The nomogram model was constructed in R software. The receiver operating characteristic (ROC) curve was plotted to evaluate the discriminative power of the nomogram model. A decision curve analysis (DCA) was used to assess the clinical value of the model.

Results:

Among 466 women, 88 (18.89%) experienced pregnancy failure. A total of 62 (19.02%) women experienced failure in the modeling group (n = 326). Multivariate logistic regression analysis identified age, prior induced abortions, early uterine fluid accumulation during re-pregnancy, complicated polycystic ovary syndrome, and transforming growth factor beta 1 (TGFβ1) as risk factors for re-pregnancy failure after complete curettage of the uterine cavity for MA (p < 0.05), while matrix metalloproteinase 9 (MMP9) reduced the risk of re-pregnancy failure (p < 0.05). The area under the curve (AUC) of the modeling group was 0.957, and the slope of the calibration curve was close to 1, with a Hosmer-Lemeshow (H-L) test value of χ² = 6.968 and p = 0.696. The AUC in the validation group was 0.990, and the slope of the calibration curve was close to 1, with an H-L test value of χ² = 6.859 and p = 0.676. The DCA curve showed that the high-risk threshold probabilities for the two groups were 0.07–0.78 and 0.08–0.84, respectively. The nomogram model was then used to evaluate the clinical utility of predicting re-pregnancy failure after MA curettage.

Conclusions:

Age, number of previous induced abortions, early uterine fluid accumulation during re-pregnancy, complicated polycystic ovary syndrome, MMP9, and TGFβ1 are influencing factors for re-pregnancy failure after complete curettage of the uterine cavity for MA. A prediction model constructed from these factors accurately estimated the postoperative risk of recurrent pregnancy loss.

Graphical abstract

Keywords

missed abortion / complete curettage of the uterine cavity / re-pregnancy failure / influencing factors / nomogram model

1. Introduction

Missed abortion (MA) is a type of spontaneous abortion where the embryo or fetus ceases development but is not expelled from the uterus in time [1]. MA has no typical symptoms, and most patients are diagnosed upon medical consultation due to symptoms such as lower abdominal pain or irregular vaginal bleeding. Moreover, prolonged retention of MA in the uterine cavity can lead to severe complications such as coagulation dysfunction in the mother. Therefore, it is necessary to remove the dead fetal tissue caused by pregnancy as soon as possible after diagnosis [2, 3]. Curettage is currently an effective clinical treatment for MA, which can effectively remove retained embryonic tissue in the uterine cavity. However, when the pregnant woman becomes pregnant again, there is a risk of failure, which cannot meet the fertility needs of couples of childbearing age [4]. Therefore, in order to effectively reduce the failure rate of subsequent pregnancy after MA curettage, identifying factors that can affect subsequent pregnancy failure after MA curettage in clinical practice can effectively improve pregnancy outcomes. A nomogram can integrate the risk factors screened out in regression analysis to individually predict the risk value of a certain event, thereby quantifying the risk of the event [5, 6]. A study established a predictive model based on the XGBoost algorithm, which can accurately predict the risk of MA in in vitro fertilization and embryo transfer (IVF-ET) patients, and this model outperformed the traditional logistic regression model [7]. This study constructed a nomogram model to enhance clinical applicability and assist clinicians in quantitatively predicting the risk of events. Currently, there are relatively few studies reporting such nomograms. Therefore, this study aims to analyze the influencing factors of subsequent pregnancy failure after MA curettage and construct a prediction model.

2. Materials and Methods

2.1 General Information

A retrospective study included 466 women with missed abortion treated by curettage from May 2021 to May 2023. They were randomly divided into modeling (n = 326) and validation (n = 140) groups (7:3). The modeling group was further classified into pregnancy failure and success groups. Case selection is shown in Fig. 1. Inclusion criterion: (1) diagnosis of MA [8]; (2) undergoing regular prenatal check-ups; (3) confirmed by B-ultrasound and having completed curettage; (4) complete data; (5) subsequent pregnancy more than 1 year after the operation, with early pregnancy failure (Follow-up endpoint: March 2025). Exclusion criteria: (1) the current pregnancy was not ectopic pregnancy, biochemical pregnancy, etc.; (2) those with severe organ dysfunction; (3) those with mental illness who could not communicate normally; (4) those with malignant tumors. The hospital ethics committee approved this study.

2.2 Diagnostic Criteria for Subsequent Pregnancy Failure

Missed abortion [9] was diagnosed by color Doppler ultrasound Voluson E10 (GE Healthcare, Chicago, IL, USA) if: (1) crown-rump length \ge7 mm without heartbeat; (2) mean gestational sac diameter \ge25 mm without embryo; (3) no embryonic heartbeat 2 weeks after a yolk sac-free gestational sac; or (4) no embryonic heartbeat 11 days after a yolk sac-containing gestational sac.

2.3 Clinical Data

Clinical and laboratory data were extracted from medical records, including demographic factors (age, body mass index (BMI), gravidity, parity, education, residence), reproductive history (curettages, induced abortions), lifestyle (smoking, alcohol), gynecologic and endocrine conditions (uterine fibroids, intrauterine fluid, cervical insufficiency, polycystic ovary syndrome, thyroid disorders, pelvic inflammatory disease), and nutritional supplementation (folic acid, calcium). All data were collected by experienced staff and validated for accuracy. Laboratory indicators included white blood cell (WBC), platelet (PLT), interleukin-6 (IL-6), tumor necrosis factor alpha (TNF-\alpha), matrix metalloproteinase 9 (MMP9), transforming growth factor beta 1 (TGF\beta1), and vascular endothelial growth factor (VEGF).

2.4 Statistical Analysis

Data were analyzed using SPSS 25.0 (IBM Corp., Armonk, NY, USA). Categorical variables were compared by \chi² test and continuous variables by t-test. Logistic regression identified risk factors. A nomogram was built with R, and its performance evaluated by receiver operating characteristic (ROC) and decision curve analysis (DCA). Significance was set at p 0.05.

3. Results

3.1 Comparison of Clinical Characteristics in Modeling and Validation Groups

No significant differences in baseline clinical characteristics were observed between the modeling and validation groups (p > 0.05) (Table 1).

3.2 Comparison of Clinical Data Between Subsequent Pregnancy Failure and Success Groups

Of the 466 women, 88 (18.89%) had subsequent pregnancy failure. In the modeling cohort (n = 326), 62 (19.02%) failed to conceive again. Significant differences were observed between the two groups in age, number of prior induced abortions, early intrauterine fluid, presence of polycystic ovary syndrome, MMP9, and TGF\beta1 levels (p 0.05). Other clinical variables showed no significant differences (p > 0.05) (Table 2).

3.3 Analysis of Influencing Factors for Subsequent Pregnancy Failure After MA Curettage

Subsequent pregnancy failure after MA curettage was set as the dependent variable (yes = 1, no = 0), and factors with p 0.05 were used as independent variables (see Table 3 for assignments). Collinearity testing showed all variance inflation factors (VIFs) 10, indicating no collinearity. Multivariable logistic regression (forward stepwise) identified age, prior induced abortions, early intrauterine fluid, polycystic ovary syndrome, and TGF\beta1 as risk factors (p 0.05), while MMP9 was protective (p 0.05) (Table 4).

3.4 Development of a Nomogram for Predicting Subsequent Pregnancy Failure After MA Curettage

In this model, the factors influencing the scoring are, in descending order of impact: MMP9, TGF\beta1, combined polycystic ovary syndrome, early pregnancy intrauterine fluid, age, and number of previous induced abortions. For example, for a pregnant woman aged \ge35 years (25.50 points), with 1 previous induced abortion (0 points), no early pregnancy intrauterine fluid (23.50 points), combined polycystic ovary syndrome (31.50 points), MMP9 (8.01 \pm 1.01) mg/L (58.50 points), and TGF\beta1 (30.21 \pm 3.05) µg/mL (34.50 points). By summing the above scores, a total score of 150 points can be obtained. Drawing a vertical line downward from the total score position shows that the predicted probability of pregnancy failure after MA curettage is 76%. See Fig. 2.

3.5 Nomogram Model in the Modeling Group

The calibration curve of the nomogram for predicting recurrent pregnancy failure after MA uterine evacuation demonstrated good agreement with the observed outcomes in the modeling cohort. The model achieved an area under the curve (AUC) of 0.957 (95% confidence interval (CI): 0.924–0.990), a Brier score of 0.0018, and a Hosmer-Lemeshow test of \chi² = 6.968, p = 0.696, indicating excellent agreement between predicted and observed probabilities (Fig. 3).

3.6 Nomogram Model in the Validation Group

The calibration curve of the nomogram predicting recurrent pregnancy failure after MA uterine evacuation showed good agreement with the observed outcomes in the validation cohort. The validation cohort achieved an AUC of 0.990 (95% CI: 0.977–1.000), a Brier score of 0.0001, and a Hosmer-Lemeshow test of \chi² = 6.859, p = 0.676, indicating excellent agreement between predicted and observed probabilities (Fig. 4).

3.7 Curve of the Nomogram Model

The DCA curve shows that the high-risk threshold probabilities for the two groups were 0.07–0.78 and 0.08–0.84, respectively, where the nomogram provided a higher net benefit. The nomogram model was then used to evaluate the clinical utility of predicting re-pregnancy failure after MA curettage (Fig. 5).

4. Discussion

The etiology of MA remains unclear but may involve chromosomal abnormalities, reproductive tract infections, and immune factors [10, 11]. Most MA pregnant women have no clinical signs and are generally found during prenatal check-ups (based on ultrasound images) that the embryo or fetus has remained in the uterine cavity and has not been expelled in time. Curettage is required to remove the embryo or fetus and reduce adverse effects on the pregnant woman’s body [12]. However, because curettage is an invasive procedure, it may damage the basal layer of the endometrium, and there is a certain risk of subsequent pregnancy failure after the operation [8, 13]. The results of this study found that 88 of the 466 pregnant women experienced subsequent pregnancy failure, with an incidence rate of 18.89%. In the modeling group (n = 326), 62 women experienced subsequent pregnancy failure (19.02%), indicating a high incidence. Therefore, identifying factors that affect subsequent pregnancy failure after MA curettage in clinical practice and intervening in a timely manner can effectively improve pregnancy outcomes for pregnant women.

In this study, multivariate analysis screened out 6 factors (age, number of previous induced abortions, early pregnancy intrauterine fluid, combined polycystic ovary syndrome, MMP9, and TGF\beta1). The reasons are analyzed as follows: (1) As pregnant women age and miss the golden age for childbearing, the body’s tolerance and uterine physiological function decline, and the risk of gestational complications increases, thus increasing the risk of subsequent pregnancy failure [14]. Therefore, prenatal guidance should be provided to older pregnant women to them adjust to the best reproductive state and prevent gestational complications. (2) An increased number of previous induced abortions indicates a thinner endometrium, which is unfavorable for subsequent sperm implantation. It can also lead to intrauterine adhesions and increase the risk of perineal reproductive tract infections and uterine damage, all of which increase the risk of subsequent pregnancy failure [15]. Therefore, it is important to understand the pregnant woman’s abortion history before the operation and ensure complete removal without residue during the curettage. (3) Early pregnancy intrauterine fluid is also a high-risk factor. Intrauterine fluid is generally caused by abnormal hormone levels or external factors and is a manifestation of abnormal pregnancy, significantly increasing the risk of miscarriage [16]. Therefore, pregnant women diagnosed with this condition should pay attention to rest and receive threatened abortion treatment. (4) For pregnant women with combined polycystic ovary syndrome, due to their special endocrine environment (hyperandrogenism, metabolic disorders, etc.), oocyte development is affected, and the receptivity of the endometrium is impacted, thereby affecting early embryonic development and increasing the likelihood of miscarriage. Therefore, these patients need to undergo progesterone treatment to improve uterine endometrial receptivity and reduce the risk of subsequent pregnancy failure [17, 18]. (5) Lower levels of MMP9 reduce proteolytic activity, promote the degradation and deposition of the extracellular matrix, and promote endometrial fibrosis leading to intrauterine adhesions, increasing the risk of subsequent pregnancy failure [19]. (6) High expression of TGF\beta1 is associated with reduced endometrial volume and decreased blood perfusion. Reduced endometrial blood perfusion is not conducive to the self-repair of endometrial damage caused by curettage. Possibly, TGF\beta1 combines with receptor 1 to promote the hydrolysis of the extracellular matrix, promoting the adhesion of endometrial tissue [20, 21].

The nomogram model can integrate multiple influencing factors into a single statistical model, allowing for relatively accurate prediction of patient conditions and demonstrating high clinical utility. Using this intuitive nomogram tool, clinicians can quickly combine multi-factor information to quantitatively predict individual patient outcomes, thereby supporting clinical decision-making and improving the precision and personalization of diagnosis and treatment. A Study has found that predictive models for MA based on coagulation function tests, including D-dimer, AT-III, and PC, have significant predictive value for MA [22]. In this study, the nomogram model yielded AUCs of 0.957 and 0.990 for the modeling and validation groups, respectively, indicating high discrimination. Moreover, the slope of the calibration curve was close to 1, indicating good consistency. Moreover, the DCA curves for both groups showed that when the probability ranged from 0.07–0.78 and 0.08–0.84, respectively, the nomogram model demonstrated high clinical utility. It can help clinicians assess the risk of subsequent pregnancy failure after surgery based on influencing factors, and early intervention can effectively improve patient prognosis. However, this study has several limitations. As a study with inherent limitations, the sample size was relatively small, and there was potential selection bias and uncontrolled confounding factors. Future research will expand the sample size and adopt a prospective, multicenter design to further explore and validate the effects of environmental and genetic factors.

5. Conclusions

Age, number of previous induced abortions, early pregnancy intrauterine fluid, combined polycystic ovary syndrome, MMP9, and TGF\beta1 are influencing factors for subsequent pregnancy failure after MA curettage. The nomogram model constructed based on these factors can better predict the risk of subsequent pregnancy failure after surgery.

Availability of Data and Materials

Data is available from the corresponding author on reasonable request.

References

[1] Jiang WZ, Yang XL, Luo JR. Risk factors for missed abortion: retrospective analysis of a single institution’s experience. Reproductive Biology and Endocrinology. 2022; 20: 115. https://doi.org/10.1186/s12958-022-00987-2.

[2] Tan Y, Li S, Xu H, Wang S. High-risk factors for massive haemorrhage in medical abortion patients with missed miscarriage. BMC Pregnancy and Childbirth. 2024; 24: 521. https://doi.org/10.1186/s12884-024-06682-x.

[3] Xie J, Zhou X, Fang L, Xiong J, Tao X. Evaluation of diagnosis and prognosis with anti-mullerian hormone level in early missed abortion. American Journal of Translational Research. 2021; 13: 5350–5355.

[4] Ryu S, Baek HW, Lee I, Won YB, Kim H, Lee JH, et al. Operative hysteroscopy-assisted pregnancy termination after failed surgical abortion in missed abortion of woman with complete septate uterus. Obstetrics & Gynecology Science. 2020; 63: 102–106. https://doi.org/10.5468/ogs.2020.63.1.102.

[5] Yue M, Ma L, Cao Y, Zhai J. Application of a Nomogram for Predicting the Risk of Subchorionic Hematoma in Early Pregnancy With In Vitro Fertilization-Embryo Transfer/Frozen Embryo Transfer. Frontiers in Endocrinology. 2021; 12: 631097. https://doi.org/10.3389/fendo.2021.631097.

[6] Liu J, Xu M, Zhou L, Yang L, Li H, Li X. Early Magnetic Resonance Imaging Measurements and Prediction of Second Trimester Pregnancy Loss: a Nomogram Model Analysis. International Journal of Women’s Health. 2024; 16: 819–827. https://doi.org/10.2147/IJWH.S453867.

[7] Yuan G, Lv B, Du X, Zhang H, Zhao M, Liu Y, et al. Prediction model for missed abortion of patients treated with IVF-ET based on XGBoost: a retrospective study. PeerJ. 2023; 11: e14762. https://doi.org/10.7717/peerj.14762.

[8] Emmer R, Ajne G, Papadogiannakis N. Missed abortion in the 11-21-week period: Fetal autopsy and placental histopathological analysis of 794 cases. European Journal of Obstetrics, Gynecology, and Reproductive Biology. 2024; 296: 158–162. https://doi.org/10.1016/j.ejogrb.2024.02.054.

[9] Herkiloglu D, Gokce S, Cevik O. Relationship of interferon regulator factor 5 and interferon-γ with missed abortion. Experimental and Therapeutic Medicine. 2022; 23: 356. https://doi.org/10.3892/etm.2022.11283.

[10] Qi Y, Hou Y, Ma M, Li X, Wu J. Circulating levels of Elabela in pregnant women with missed abortion. Gynecological Endocrinology. 2022; 38: 693–696. https://doi.org/10.1080/09513590.2022.2090539.

[11] Chen Y, Chen X, Chen P, Chen X, Pan L, Han L, et al. Alteration of the Gut Microbiota in Missed Abortion. Indian Journal of Microbiology. 2023; 63: 106–119. https://doi.org/10.1007/s12088-023-01063-y.

[12] Luo M, Xiao H, Wang L, Zhao J, Gao J, Ma W. The expression and clinical significance of three lncRNAs in patients with a missed abortion. Experimental and Therapeutic Medicine. 2021; 21: 8. https://doi.org/10.3892/etm.2020.9440.

[13] de Codt M, Balza C, Jadoul P, Forget P, Squifflet JL, Bernard P, et al. Hysteroscopic Resection for Missed Abortion: Feasibility, Operative Technique and Potential Benefit Compared to Curettage. Frontiers in Surgery. 2020; 7: 64. https://doi.org/10.3389/fsurg.2020.00064.

[14] Tang Y, Zhu C, Zhu C, Liang F, Lee A, Yao X, et al. The impact of pre-evacuation ultrasound examination in histologically confirmed hydatidiform mole in missed abortion. BMC Women’s Health. 2020; 20: 196. https://doi.org/10.1186/s12905-020-01064-9.

[15] Gu L, Zhang C, Luo J, Zhou C, Song Y, Huang X. Efficacy and prognostic factors of combined administration of progesterone and estriol valerate tablets for preventing intrauterine adhesions in patients with early missed abortion following dilation and curettage. American Journal of Translational Research. 2024; 16: 3164–3170. https://doi.org/10.62347/AMEB4153.

[16] Yang H, Shi Y, Ji G. Minimally invasive surgical treatment of Robert’s uterus with missed miscarriage: case report. Journal of Obstetrics and Gynaecology: the Journal of the Institute of Obstetrics and Gynaecology. 2024; 44: 2305204. https://doi.org/10.1080/01443615.2024.2305204.

[17] Joó JG, Sulyok E, Bódis J, Kornya L. Disrupted Balance of the Oxidant-Antioxidant System in the Pathophysiology of Female Reproduction: Oxidative Stress and Adverse Pregnancy Outcomes. Current Issues in Molecular Biology. 2023; 45: 8091–8111. https://doi.org/10.3390/cimb45100511.

[18] Khemani S, More A, Nair N, Choudhary N, Shrivastava J, Mamankar D. Comprehensive Management of Polycystic Ovarian Syndrome and Secondary Infertility: Optimizing Fertility Outcomes. Cureus. 2024; 16: e52597. https://doi.org/10.7759/cureus.52597.

[19] Hou T, Zhang P, Tian H, Luo Y, Li J, Zhang K, et al. Semaphorin 4A Maintains Trophoblastic Function via Activating the STAT3 Pathway. Biomolecules. 2024; 14: 826. https://doi.org/10.3390/biom14070826.

[20] Yuan J, Dong J, Li W, Zu Y, Guo Y, Zhang Y, et al. Association between aquaporin 3 and transforming growth factor-beta 1 levels in decidual tissue and serum of patients with missed abortion. Frontiers in Medicine. 2025; 12: 1540257. https://doi.org/10.3389/fmed.2025.1540257.

[21] Jin G, Yang C, He J, Zhang S, Liu H, Li C. Transforming growth factor-beta 1 promotes endometrial fibrosis in rats. Basic & Clinical Medicine. 2023; 43: 1557–1561. https://doi.org/10.16352/j.issn.1001-6325.2023.10.1557. (In Chinese).

[22] Chen D, Wang J, Liu H, Xu T, Qiao J, Liu Q, et al. Coagulation Status in Women with a History of Missed Abortion. Reproductive Sciences. 2024; 31: 480–487. https://doi.org/10.1007/s43032-023-01355-0.