A prediction model of preterm birth in singleton pregnancy after assisted reproduction therapy.

This retrospective study included a total of 268 pregnant women who underwent assisted reproductive technology (ART) and gave birth to singleton neonates at Sun Yat-sen Memorial Hospital, Sun Yat-sen University, between August 2013 and December 2021. Preterm birth was defined as the delivery of a Live neonate between 28 +0 weeks and 36 +6 weeks of gestation. The study cohort was randomly stratified into a training set (160 participants) and an internal validation set (108 participants) using a 6:4 ratio for model development and validation purposes. Inclusion criteria required singleton pregnancies following ART that resulted in live births and had complete medical records available for review. Exclusion criteria included cases with serious underlying medical conditions or prior surgical history, such as cardiovascular disease, hematological disorders, malignancies, or any severe systemic diseases that could influence pregnancy outcomes. Cases involving chromosomal abnormalities in one of the partners, the use of donated sperm for conception, or pregnancies that were terminated due to severe fetal malformations, trauma, exposure to teratogenic substances or radiation, or fetal reduction were also excluded. Data for this study were extracted from the electronic medical records of patients who underwent ART at Sun Yat-Sen Memorial Hospital. The data collected encompassed the following categories: demographic characteristics, medical and obstetric history, ART-related parameters, pregnancy complications, and outcomes related to parturition and newborn conditions. Demographic Characteristics: Information on the maternal and paternal demographics was collected, including maternal age and paternal age at the time of conception, female body mass index (BMI) prior to ART treatment, duration of infertility. Medical and Obstetric History: The patients’ past medical history was documented, including the presence of polycystic ovary syndrome (PCOS), uterine malformation, intrauterine adhesion (IA), and history of gynecological surgeries such as hysteroscopy, conization, and curettage. Obstetric history was also recorded, including the number of previous gravidity, parity, abortion (spontaneous or induced), and ectopic pregnancies and natural childbirth. Since ART is a known risk factor of gestational hypertensive diseases (GHD), the use of low-dose asprin for GHD prevention were also recorded. ART-related Parameters: ART-specific variables was extracted, including types and factors of infertility. We also collected detailed information on the ART protocols used, including the treatment type (IVF or ICSI), as well as the maternal basal hormone levels (Follicle Stimulating Hormone [FSH], Luteinizing Hormone [LH], estradiol [E2]), and the ratio of FSH/LH. Further ART-related data included the basal antral follicle count (AFC), progesterone (P) and E2 levels on the day of human chorionic gonadotropin (hCG) administration. Endometrial thickness (EMT) on the day of hCG administration, the number of oocytes retrieved, the number of embryos implanted, stages of embryo transfer, and number of gestational sacs were also recorded. Pregnancy Complications: Data on pregnancy-related complications were collected, including gestational hypertensive diseases (GHD), gestational diabetes mellitus (GDM), placenta previa (PP), premature rupture of membranes (PROM), and intrauterine infections (II). Parturition and Newborn Outcomes: Data regarding the mode of delivery (vaginal or cesarean section), gestational week at delivery, and any postpartum complications were documented. Newborn outcomes were assessed based on weight at birth, Apgar scores at 1, 5, and 10 min, neonatal complications, and the need for neonatal intensive care unit (NICU) transfer. Demographic Characteristics: Information on the maternal and paternal demographics was collected, including maternal age and paternal age at the time of conception, female body mass index (BMI) prior to ART treatment, duration of infertility. Medical and Obstetric History: The patients’ past medical history was documented, including the presence of polycystic ovary syndrome (PCOS), uterine malformation, intrauterine adhesion (IA), and history of gynecological surgeries such as hysteroscopy, conization, and curettage. Obstetric history was also recorded, including the number of previous gravidity, parity, abortion (spontaneous or induced), and ectopic pregnancies and natural childbirth. Since ART is a known risk factor of gestational hypertensive diseases (GHD), the use of low-dose asprin for GHD prevention were also recorded. ART-related Parameters: ART-specific variables was extracted, including types and factors of infertility. We also collected detailed information on the ART protocols used, including the treatment type (IVF or ICSI), as well as the maternal basal hormone levels (Follicle Stimulating Hormone [FSH], Luteinizing Hormone [LH], estradiol [E2]), and the ratio of FSH/LH. Further ART-related data included the basal antral follicle count (AFC), progesterone (P) and E2 levels on the day of human chorionic gonadotropin (hCG) administration. Endometrial thickness (EMT) on the day of hCG administration, the number of oocytes retrieved, the number of embryos implanted, stages of embryo transfer, and number of gestational sacs were also recorded. Pregnancy Complications: Data on pregnancy-related complications were collected, including gestational hypertensive diseases (GHD), gestational diabetes mellitus (GDM), placenta previa (PP), premature rupture of membranes (PROM), and intrauterine infections (II). Parturition and Newborn Outcomes: Data regarding the mode of delivery (vaginal or cesarean section), gestational week at delivery, and any postpartum complications were documented. Newborn outcomes were assessed based on weight at birth, Apgar scores at 1, 5, and 10 min, neonatal complications, and the need for neonatal intensive care unit (NICU) transfer. Statistical analysis was conducted using R software (version 4.05, http://www.r-project.org/ ). The normality of continuous variables was assessed using the Shapiro-Wilk test. Continuous variables following normal distribution were summarized as Means ± Sandard Deviations (SD). For non-normally distributed data, variables were summarized as medians with interquartile ranges (IQRs). Differences between groups for continuous variables were compared using the t -test for normally distributed data, or the Mann-Whitney U test for non-normally distributed data. Categorical variables were expressed as frequencies and percentages (%), and comparisons were made using the χ 2 test or Fisher’s exact test. A two-sided significance level of 0.05 was used for all tests. To construct the nomogram, the data of training set were used to identify potential risk variables through univariate logistic regression. Multivariable logistic regression was then employed to develop the final predictive model, using the R package “rms.” The model’s performance was subsequently validated using the internal validation set. Discrimination of nomogram was measured using the area under the curves (AUC), which reflects the ability to discriminate the outcomes.

The baseline characteristics were comparable between the training set and the internal validation set (Supplement Tables 1, 2, 3 and 4). Table  1 summarized the comparison of demographic characteristics, ART indicators, and pregnancy complications between the preterm birth group ( N  = 63) and the full-term birth group ( N  = 97) in the training set. The preterm birth group had a higher prevalence of PCOS (15.87% vs. 6.19%, P  < 0.05), intrauterine adhesion (7.94% vs. 1.03%, P  < 0.05), and GSH (58.73% vs. 42.27%, P  < 0.05). The full-term group had significantly higher basal E2 levels (44.34 vs. 36.43 ng/L, P  < 0.05) compared to the preterm group. Furthermore, the number of embryos transferred (7.22% vs. 19.05% for 2 embryos, P  < 0.05) and gestational sacs (5.15% vs. 15.87% for 2 sacs, P  < 0.05) was significantly lesser in the full-term group. The preterm group showed a significantly higher incidence of GHD (15.87% vs. 2.06%, P  < 0.05), PP (17.46% vs. 4.12%, P  < 0.05), and PROM (41.27% vs. 9.28%, P  < 0.05) compared to the full-term group. Variables that were not significantly different between the two groups included female age, male age, female BMI, infertility duration, infertility types, infertility factors, treatment types, basal FSH, basal LH, FSH/LH ratio, basal AFC, hormone levels on HCG day, number of oocytes retrieved, embryo transfer stage, LDS use before 16 weeks, GDM, and intrauterine infection. Table 1 Comparison of demographics, ART indicators, and pregnancy complications between preterm and full-term groups Variables Preterm group ( N  = 63) Full-term group ( N  = 97) P value Female Age 0.734  <35 34 (53.97%) 55 (56.70%)  ≥ 35 29 (46.03%) 42 (43.30%) Male Age 0.453  <35 28 (44.44%) 49 (50.52%)  ≥ 35 35 (55.56%) 48 (49.48%) Female BMI 0.385  <24 47 (74.60%) 78 (80.41%)  ≥ 24 16 (25.40%) 19 (19.59%) Infertility Duration (years) 5 (3, 8) 4 (2, 6) 0.061 Medical history  PCOS 10 (15.87%) 6 (6.19%) 0.046  UM 5 (7.94%) 2 (2.06%) 0.076  IA 5 (7.94%) 1 (1.03%) 0.035  GSH 37 (58.73%) 41 (42.27) 0.042 Obstetric history  G (number) 2 (1, 3) 2 (1, 3) 0.886  P (number) 1 (1, 1) 1 (1, 2) 0.143  A (number) 1 (0, 1) 0 (0, 1) 0.465  E (number) 0 (0, 0) 0 (0, 0) 0.742 Natural childbirth 17 (26.98%) 46 (47.42%) 0.01 Infertility types  Primary 32 (50.79%) 47 (48.45%) 0.772  Secondary 31 (49.21%) 50 (51.55%) Infertility factors  Pelvic and oviduct factors 10 (15.87%) 12 (12.37%) 0.825  Ovulation disorder 11 (17.46%) 21 (21.65%)  Endometriosis 3 (4.76%) 6 (6.19%)  Male factor 9 (14.29%) 16 (16.49%)  Multiple factors 25 (39.68%) 39 (40.21%)  Unknown 5 (7.93%) 3 (3.09%) Treatment Types  ICSI 11 (17.46%) 11 (11.34%) 0.410  IVF 46 (73.02%) 72 (74.23%)  IVF + rICSI 6 (9.52%) 14 (14.43%)  Basal FSH (IU/L) 7.38 (6.11, 9.39) 8.23 (6.83, 9.77) 0.088 Basal LH (IU/L) 4.8 (3.56, 6.85) 4.41 (3.40, 6.14) 0.358 FSH/LH 1.65 (0.99, 2.35) 1.88 (1.34, 2.55) 0.113 Basal E2 level (ng/L) 36.43 (25.93–50.47) 44.34 (34.35–58.63) 0.009 Basal AFC 16 (10, 24) 15 (10, 23) 0.533 P level of HCG Day (ug/L) 1.02 (0.79, 1.27) 0.99 (0.72, 1.28) 0.625 E2 level of HCG Day (ng/L) 3000 (2063, 4529) 3218 (1925, 4836) 0.647 EMT of HCG Day (mm) 12.02 (9.37, 13.75) 11.81 (10.07, 13.14) 0.731 No. of Oocytes Retrieved 11 (8, 15) 11 (7, 16) 0.919 No. of embryo implantation  1 51 (80.95%) 90 (92.78%) 0.024  2 12 (19.05%) 7 (7.22%) Embryo Transfer Stage  D3 50 (79.37%) 85 (87.63%) 0.159  D5 13 (20.63%) 12 (12.37%) No. of Gestational Sacs  1 53 (84.13%) 92 (94.85%) 0.029  2 10 (15.87%) 5 (5.15%) LDS use before 16 weeks 48 (76.19%) 75 (77.32%) 0.869 GHD 10(15.87%) 2(2.06%) 0.001 GDM 21(33.33%) 23(23.71%) 0.183 PP 11(17.46%) 4(4.12%) 0.005 PROM 26(41.27%) 9(9.28%) < 0.001 II 2(3.17%) 3(3.09%) 0.977 Data were shown as Median (Interquartile Range, IQR) or No. (%) A Abortion, AFC Antral Follicle Count, B Body Mass Index, D Day, E Ectopic pregnancy, E2 Estradiol, EMT Endometrial Thickness, F Follicle-Stimulating Hormone, G Gravidity, GDM Gestational Diabetes Mellitus, GHD Gestational Hypertensive Disease, GSH Gynecological Surgical History, HCG Human Chorionic Gonadotropin, IA Intrauterine adhesion, II Intrauterine Infection, ICSI Intracytoplasmic Sperm Injection, IVF In Vitro Fertilization, L Luteinizing Hormone, LDS Low-dose Aspirin, P Parity, P Placenta Previa, PROM Premature Rupture of the Membrane, P Progesterone, PCOS Polycystic Ovary Syndrome, rICSI rescue ICSI, GSH Gynecological Surgical History, UM  Uterine malformation Comparison of demographics, ART indicators, and pregnancy complications between preterm and full-term groups Data were shown as Median (Interquartile Range, IQR) or No. (%) A Abortion, AFC Antral Follicle Count, B Body Mass Index, D Day, E Ectopic pregnancy, E2 Estradiol, EMT Endometrial Thickness, F Follicle-Stimulating Hormone, G Gravidity, GDM Gestational Diabetes Mellitus, GHD Gestational Hypertensive Disease, GSH Gynecological Surgical History, HCG Human Chorionic Gonadotropin, IA Intrauterine adhesion, II Intrauterine Infection, ICSI Intracytoplasmic Sperm Injection, IVF In Vitro Fertilization, L Luteinizing Hormone, LDS Low-dose Aspirin, P Parity, P Placenta Previa, PROM Premature Rupture of the Membrane, P Progesterone, PCOS Polycystic Ovary Syndrome, rICSI rescue ICSI, GSH Gynecological Surgical History, UM  Uterine malformation The comparisons of parturition and newborn conditions were shown in Supplementary Table 5. Gestational week, newborn weight, and 1-minute Apgar score were lower in the preterm group (34.14 vs. 38.68, P  < 0.05); 2.30 vs. 3.31, P  < 0.05); 8.71 vs. 9.42, P  < 0.05). The rate of cesarean section and NICU Transfer were higher in the preterm group (76.19% vs. 59.79%, P  < 0.05; 77.78% vs. 14.43%, P  < 0.05)). The results of univariate and multivariate logistic regression from the training set are shown in Table  2 . In the univariate analysis, PCOS (OR = 1.94, 95% CI: 1.01–3.74, P  = 0.046), IA (OR = 4.83, 95% CI: 1.23–18.98, P  = 0.024), number of embryos transferred (OR = 2.60, 95% CI: 1.14–5.93, P  = 0.023), GHD (OR = 2.46, 95% CI: 1.23–4.92, P  = 0.011), and PROM (OR = 4.18, 95% CI: 1.90–9.19, P  < 0.001) showed significant differences. These factors were then included in the multivariate logistic regression analysis to identify independent predictors of preterm birth after ART. Table 2 Univariate and multivariate logistic analysis of preterm birth after ART Variables Univariate Multivariate P OR (95%CI) P OR (95%CI) Female Age  <35 1.00 (Reference)  ≥ 35 0.55 1.22 (0.64 ~ 2.32) Male Age  <35 1.00 (Reference)  ≥ 35 0.296 1.41 (0.74 ~ 2.70) Female BMI  <24 1.00 (Reference)  ≥ 24 0.569 1.24 (0.59 ~ 2.58) PCOS  No 1.00 (Reference) 1.00 (Reference)  Yes 0.046 1.94 (1.01 ~ 3.74) 0.039 2.27 (1.04 ~ 4.93) IA  No 1.00 (Reference) 1.00 (Reference)  Yes 0.024 4.83 (1.23 ~ 18.98) 0.043 3.32 (0.67 ~ 16.59) UM  No 1.00 (Reference)  Yes 0.066 3.81 (0.91 ~ 15.85) GSH  No 1.00 (Reference)  Yes 0.118 1.68 (0.88 ~ 3.23) LDS use before 16 weeks  No 1.00 (Reference)  Yes 0.219 0.61 (0.28 ~ 1.34) Natural Birth  No 1.00 (Reference)  Yes 0.145 0.58 (0.28 ~ 1.21) Infertility Types  Primary 1.00 (Reference)  Secondary 0.789 1.09 (0.57 ~ 2.08) Infertility Factors  Pelvic and oviduct factors 1.00 (Reference)  Ovulation disorder 0.951 0.95 (0.22 ~ 4.19)  Endometriosis 0.645 0.72 (0.17 ~ 2.97)  Male factor 0.927 0.95 (0.35 ~ 2.58)  Multiple factors 0.076 3.34 (0.88 ~ 12.66)  Unknown 0.817 1.10 (0.49 ~ 2.50) Treatment Types ICSI 1.00 (Reference)  IVF 0.935 0.96 (0.35 ~ 2.62)  IVF + rICSI 0.348 0.62 (0.22 ~ 1.69) No. embryos implantation  1 1.00 (Reference) 1.00 (Reference)  2 0.023 2.60 (1.14 ~ 5.93) 0.014 3.54 (1.29 ~ 9.69) Embryo Day  D3 1.00 (Reference)  D5 0.142 1.78 (0.82 ~ 3.83) No. of Gestational Sacs  1 1.00 (Reference)  2 0.85 1.10 (0.43 ~ 2.82) GDM  No 1.00 (Reference)  Yes 0.109 2.66 (0.80 ~ 8.81) GHD  No 1.00 (Reference) 1.00 (Reference)  Yes 0.011 2.46 (1.23 ~ 4.92) 0.007 3.07 (1.36 ~ 6.93) PROM  No 1.00 (Reference) 1.00 (Reference)  Yes < 0.001 4.18 (1.90 ~ 9.19) < 0.001 4.70 (1.90 ~ 11.61) PP  No 1.00 (Reference)  Yes 0.084 6.39 (1.95 ~ 20.89) II  No 1.00 (Reference)  Yes 0.916 0.88 (0.08 ~ 9.89)  P 0.474 0.74 (0.32 ~ 1.69)  E 0.359 1.95 (0.47 ~ 8.19)  G 0.454 1.09 (0.87 ~ 1.37)  A 0.336 1.14 (0.87 ~ 1.49) Infertility Years 0.229 1.05 (0.97 ~ 1.15)  Basel FSH 0.681 1.02 (0.94 ~ 1.10)  Basel LH 0.736 1.02 (0.93 ~ 1.11)  FSH/LH 0.055 0.74 (0.54 ~ 1.01)  Basel E2 0.508 1.00 (1.00 ~ 1.01)  Basel AFC 0.999 1.00 (0.97 ~ 1.03) P level of HCG Day (ug/L) 0.836 0.94 (0.53 ~ 1.67)  E2 level of HCG Day (ng/L) 0.898 1.00 (1.00 ~ 1.00)  No. of oocytes 0.734 0.99 (0.95 ~ 1.04)  EMT of HCG Day (mm) 0.194 0.92 (0.82 ~ 1.04) Data were shown as odds ratio (95% confidence interval, CI), P value A Abortion, AFC Antral Follicle Count, B Body Mass Index, D Day, E Ectopic pregnancy, E2 Estradiol, EMT Endometrial Thickness, F Follicle-Stimulating Hormone, G Gravidity, GDM Gestational Diabetes Mellitus, GHD Gestational Hypertensive Disease, GSH Gynecological Surgical History, HCG Human Chorionic Gonadotropin, IA Intrauterine adhesion, II Intrauterine Infection, ICSI Intracytoplasmic Sperm Injection, IVF In Vitro Fertilization, L Luteinizing Hormone, LDS Low-dose Aspirin, P Parity, P Placenta Previa, PROM Premature Rupture of the Membrane, P Progesterone, PCOS Polycystic Ovary Syndrome, rICSI rescue ICSI, GSH Gynecological Surgical History, UM Uterine malformation Univariate and multivariate logistic analysis of preterm birth after ART Data were shown as odds ratio (95% confidence interval, CI), P value A Abortion, AFC Antral Follicle Count, B Body Mass Index, D Day, E Ectopic pregnancy, E2 Estradiol, EMT Endometrial Thickness, F Follicle-Stimulating Hormone, G Gravidity, GDM Gestational Diabetes Mellitus, GHD Gestational Hypertensive Disease, GSH Gynecological Surgical History, HCG Human Chorionic Gonadotropin, IA Intrauterine adhesion, II Intrauterine Infection, ICSI Intracytoplasmic Sperm Injection, IVF In Vitro Fertilization, L Luteinizing Hormone, LDS Low-dose Aspirin, P Parity, P Placenta Previa, PROM Premature Rupture of the Membrane, P Progesterone, PCOS Polycystic Ovary Syndrome, rICSI rescue ICSI, GSH Gynecological Surgical History, UM Uterine malformation In the multivariate logistic regression analysis, the number of embryos transferred (OR = 3.54, 95% CI: 1.29–9.69, P  = 0.014), GHD (OR = 3.07, 95% CI: 1.36–6.93, P  = 0.007), PROM (OR = 4.70, 95% CI: 1.90–11.61, P  < 0.001), PCOS (OR = 2.27, 95% CI: 1.04–4.93, P  = 0.039), and IA (OR = 3.32, 95% CI: 0.67–16.59, P  = 0.043) were identified as independent prognostic factors for preterm birth after ART. These five risk factors were incorporated to build the predictive model, and then the model was visualized using a nomogram (Fig.  1 A). The prediction score of preterm birth can be calculated by the nomogram. The performance of this nomogram predict model was evaluated by AuC, calibration plots, and DCA. The AuC was 0.77 (0.69–0.84) and 0.71 (0.61–0.81) in the training set and validation set respectively, which denoted an acceptable performance (Fig.  1 B-C). Besides, the calibration curve also demonstrated good agreement between prediction and observation both in the training set and the internal validation set (Fig.  2 A-B). Furthermore, the decision curve analysis showed all curves were above the reference Line if the threshold probabilities were between 0.00 and 0.83 (Fig.  2 C-D). Fig. 1 Nomogram to predict preterm birth after ART.  A  Nomogram for predicting risk, including variables such as PCOS, intrauterine adhesion (IA), number of embryos transferred (NO. Embryo), gestational hypertensive diseases (GHD), and premature rupture of membranes (PROM). B  Receiver operating characteristic (ROC) curve for the training cohort, with an area under the curve (AUC) of 0.77 (95% CI: 0.69–0.84). C  ROC curve for the validation cohort, with an AuC of 0.71 (95% CI: 0.61–0.81) Nomogram to predict preterm birth after ART.  A  Nomogram for predicting risk, including variables such as PCOS, intrauterine adhesion (IA), number of embryos transferred (NO. Embryo), gestational hypertensive diseases (GHD), and premature rupture of membranes (PROM). B  Receiver operating characteristic (ROC) curve for the training cohort, with an area under the curve (AUC) of 0.77 (95% CI: 0.69–0.84). C  ROC curve for the validation cohort, with an AuC of 0.71 (95% CI: 0.61–0.81) Fig. 2 Calibration and threshold analysis of the predictive model.  A  Calibration curve for the training set. B  Calibration curve for the validation set. C  The decision curve analysis (DCA) of the nomogram for probability of prterm delivery in the training set. D  DCA of the nomogram for probability of prterm delivery in the validation set Calibration and threshold analysis of the predictive model.  A  Calibration curve for the training set. B  Calibration curve for the validation set. C  The decision curve analysis (DCA) of the nomogram for probability of prterm delivery in the training set. D  DCA of the nomogram for probability of prterm delivery in the validation set Before 16 weeks of gestation, the difference in cervical length (Δl) between preterm birth group and the term birth group was < 2 mm, with Δl beginning to increase gradually from 16 weeks of pregnancy. The growth rate of Δl was the highest between 20 and 28 weeks of pregnancy, indicating that the cervical length was most significantly shortened in the preterm group in the late second trimester. The growth rate of Δl slowed down after 28 weeks of pregnancy and reached its maximum value at 36 weeks of pregnancy (12.93 mm). These results are shown in Fig.  3 . Fig. 3 Comparison of cervical length between preterm group and full-term group. Red line: cervical length change curve with gestational week in the preterm group (Light red area is 95% confidence interval); blue line: cervical length change curve with gestational week in the full-term group (Light blue area is 95% confidence interval); dotted line: average cervical length difference curve with gestational week between the full-term group and the preterm group Comparison of cervical length between preterm group and full-term group. Red line: cervical length change curve with gestational week in the preterm group (Light red area is 95% confidence interval); blue line: cervical length change curve with gestational week in the full-term group (Light blue area is 95% confidence interval); dotted line: average cervical length difference curve with gestational week between the full-term group and the preterm group

ART has become a crucial approach in the treatment of infertility, offering hope to millions of couples worldwide who struggle with conception [ 20 ]. While ART has significantly improved the chances of pregnancy for infertile couples, it is associated with an increased risk of adverse pregnancy outcomes, particularly preterm birth [ 20 ]. Our study aimed to develop and validate a nomogram to predict the individual probability of preterm birth in women who conceive through ART. We identified five independent risk factors for preterm birth in this population: the number of embryo implantations, GHD, PCOS, PROM and IA. The nomogram we constructed offers a user-friendly and personalized tool for the early identification of women at high risk for preterm birth after ART, enabling clinicians to provide timely and targeted medical interventions to improve pregnancy outcomes. Our study contributes to the comprehensive analysis of ART-related parameters and pregnancy complications, the development of an accessible nomogram with robust predictive performance, and the novel insights into cervical length dynamics throughout gestation in ART pregnancies. The association between the number of embryo implantations and preterm birth is a significant finding in our study. The results of this study indicated that the proportion of women who had two embryos transferred was significantly higher in the preterm group compared to the full-term group. Additionally, the transfer of two embryos emerged as an independent risk factor for preterm birth. Although the number of gestational sacs was not identified as a predictive factor for preterm birth in this study, similar to the results regarding the number of embryos transferred, the proportion of women with two gestational sacs in early pregnancy was significantly higher in the preterm group than in the full-term group. These findings suggested a higher likelihood of vanishing twin syndrome (VTS) in the preterm group. VTS refers to the phenomenon in multiple pregnancies where one or more embryos die naturally and are absorbed by the maternal body or surviving fetuses, typically occurring within the first trimester [ 21 ]. A study estimated that VTS occurs in approximately 21% to 30% of twin pregnancies [ 22 ]. In ART treatment, more than one embryos are often transferred to improve pregnancy success rates, leading to a significantly higher incidence of VTS in patients undergoing ART. Previous studies have shown that VTS is a risk factor for singleton preterm birth [ 23 ]. ART singletons who experienced VTS were found to have lower birth weights compared to controls without VTS, with 1.84 times risk of being small for gestational age (SGA) [ 23 , 24 ]. This may be related to embryos that cease to develop in the uterus, undergo necrosis, release inflammatory factors that disrupt the uterine microenvironment, stimulate contractions, and lead to cervical relaxation and preterm birth [ 17 , 25 ]. These findings are consistent with previous studies that have demonstrated the association between multiple embryo transfer and increased preterm birth risk. A cohort study by Yan et al. reported similar increased odds of preterm delivery with double embryo transfer compared to single embryo transfer, supporting our observations regarding the impact of embryo number on pregnancy outcomes [ 26 ]. This study also confirm that PCOS and IA are risk factors for preterm birth in ART singletons. Oligo-ovulation or anovulation is one of the primary causes of infertility in women with PCOS, leading to their increased use of ART. Women with PCOS often experience insulin resistance and hyperandrogenism, which impair endometrial blood flow and vascular integrity. This results in increased endothelial oxidative stress, contributing to preterm delivery [ 27 , 28 ]. In addition, IA also is one of the major factors prompting patients to seek ART for infertility treatment. IA can lead to abnormal uterine cavity morphology and impaired endometrial function [ 29 ]. Studies have shown that patients with IA who undergo IVF have a significantly higher incidence of placental abnormalities, such as placental adhesions and placenta previa, compared to control groups without IA [ 30 , 31 ]. These placental abnormalities are important risk factors for preterm birth. These results aligned with previous studies indicated that women with PCOS have higher risk of preterm birth [ 17 , 32 ]. Moreover, although there is no literature currently reporting IA as a risk factor of preterm in ART, meta-analyses have reported that IA increases the risk of preterm birth, intrauterine growth restriction, and fetal malformations, which is consistent with the findings of this study [ 33 ]. Pregnancy complications are also one of the important causes of preterm birth. This study showed that GHD and PROM were independent risk factors of preterm in ART singleton. Many studies have demonstrated that the risk of GHD is elevated in pregnancies conceived through ART compared to those conceived naturally [ 34 , 35 ]. The high levels of estrogen during ovarian stimulation in ART procedures may contribute to an increased risk of GHD in pregnancies achieved with ART assistance. Additionally, patients who undergo ART-assisted pregnancy are typically elder, and as age increases, the degree of vascular endothelial damage also escalates, leading to a higher incidence of gestational hypertensive diseases [ 36 – 38 ]. Moreover, research indicated that luteal support during ART-assisted pregnancy and the administration of human chorionic gonadotropin can activate the renin - aldosterone system, potentially contributing to an increased incidence of hypertensive disorders during pregnancy [ 39 ]. Elevated blood pressure during pregnancy can cause spasms in small blood vessels throughout the body, which may affect blood flow to the uterus and placenta, potentially leading to premature birth [ 40 ]. PROM compromised the protective function of the amniotic sac, leading to an increased risk of infection, which may result in chorioamnionitis [ 41 ]. This condition can trigger uterine contractions and ultimately lead to preterm birth. Additionally, the reduction of amniotic fluid can increase the pressure of the fetus on the uterine wall and the risk of umbilical cord compression, further affecting the blood supply and oxygen delivery to the fetus, thus causing preterm birth [ 41 ]. In addition, although PP did not included as a risk factor of preterm in this study, the preterm group had a significant higher proportion than the full-term group, which indicated the important role of PP in ART singleton preterm. Recent meta-analyses of observational studies have demonstrated a positive association between ART conception and PP [ 42 ]. The ART process may contribute to abnormalities not only in trophoblastic invasion but also in the epigenetic regulation of placental formation and function. This occurs through the disruption of endometrial receptivity and alterations in the embryonic environment. Antepartum and postpartum hemorrhage are among the most concerning obstetric complications associated with placenta previa, significantly impacting both maternal and fetal outcomes [ 43 ]. Recent studies suggest that patients with PCOS and uterine malformations who conceive through ART are at an increased risk for cervical insufficiency (CI) [ 44 ]. Mechanical cervical injury resulting from ART procedures during pregnancy examinations or treatments, such as tubal flushing, hysterosalpingography, or hysteroscopy, along with the use of medications in ART and the higher incidence of multiple pregnancies, may contribute to the development of CI [ 45 ]. This study revealed that the preterm group exhibited significantly shortened cervical lengths in the late second trimester, suggesting an association between short cervical length and preterm delivery. Our cervical length findings are consistent with established literature demonstrating that cervical shortening becomes more pronounced in the second trimester among women destined for preterm delivery [ 46 , 47 ]. Therefore, we recommend that cervical length monitoring be initiated between 16 and 20 weeks of gestation for pregnant women undergoing ART, to detect and diagnose CI as early as possible. Timely interventions, such as cervical cerclage or contraction inhibition, should be implemented to prolong gestation and improve pregnancy outcomes. However, there are several limitations to our study. It is a single-center, retrospective analysis with a relatively small sample size, and the results have only been internally validated at our center without external validation, which may introduce some bias when applied to other institutions. Additionally, some patients did not returned to our hostpital for further antenatal care and delivery after the second trimester, resulting in missing data on cervical length, as other medical institutions rarely monitor this parameter. Future research should focus on external validation of our prediction model in multi-center cohorts and different populations to confirm its generalizability. Additionally, prospective studies incorporating cervical length monitoring protocols and exploring the molecular mechanisms underlying preterm birth in ART pregnancies would further enhance our understanding and clinical management strategies.

This study developed a predictive model for preterm birth in singleton pregnancies following ART, identifying key risk factors such as the number of embryos implanted, GHD, PROM, PCOS, and IA. The model demonstrated acceptable predictive performance and can help clinicians identify high-risk pregnancies early, enabling timely interventions. Cervical length monitoring was also found to be a useful predictor.

Preterm birth is a major global health issue, which is the leading cause of death in neonates and children under the age of five worldwide [ 1 , 2 ]. A recent study reported that approximately 13.4 million preterm births occurring worldwide in 2020, accounting for more than 1 in 10 live births [ 3 ]. Although improvements in perinatal care, the global incidence of preterm birth has shown a declining trend overall, though the age-standardized incidence rate (ASIR) began to increase since 2016, indicating emerging challenges in certain regions [ 3 , 4 ]. Compared to term newborns, preterm neonates are at a higher risk of experiencing developmental delays, cognitive impairments, and respiratory issues [ 5 , 6 ]. They are also more likely to face long-term health complications, including neurodevelopmental disorders, cerebral palsy, and lung diseases [ 7 , 8 ]. Therefore, identifying the risk factors for preterm birth is critical to improving perinatal and neonatal management. Infertility is defined as the inability to achieve pregnancy after 12 months of regular unprotected sexual intercourse [ 9 ]. The estimated Lifetime and period prevalence of 12-month infertility are 17.5% and 12.6%, respectively [ 10 ]. Assisted reproductive technology (ART), which has become a common solution for infertility, refers to treatments that include procedures such as in vitro fertilization (IVF), intracytoplasmic sperm injection (ICSI), and intrauterine insemination (IUI) [ 11 ]. It has been reported that over 10 million children have been born through ART, accounting for up to 7.9% of births in Europe and 5.1% in the United States [ 12 ]. While increased access to reproductive therapies has helped reduce the global burden of infertility, studies have found that ART is associated with a higher risk of adverse pregnancy outcomes [ 13 , 14 ]. Even in singleton pregnancies, preterm birth is more prevalent among neonates conceived through ART compared to those conceived naturally [ 15 ]. Previous study reported that IVF/ICSI pregnancies were associated with a slight increase in preterm birth before 37 weeks (adjust-OR 1.72) and iatrogenic (induced) preterm birth (adjust-OR 1.41) [ 16 ]. The causes of preterm birth in singleton pregnancies following ART are likely multifactorial. Several inherent risk factors, including the duration and cause of infertility, as well as the use of donor oocytes, contribute to an increased likelihood of preterm birth [ 17 ]. Blastocyst transfer and frozen-thawed embryo transfer have been identified as additional risk factors for preterm birth in a study involving over 20,000 singleton newborns [ 18 ]. However, another study concluded that while most risk factors for preterm and early preterm birth in IVF/ICSI pregnancies are similar to those in natural pregnancies [ 19 ]. This discrepancy highlights the need for further research into the risk factors associated with preterm birth in ART singletons. Developing a prediction model based on these factors is essential, enabling clinicians to identify at-risk pregnancies early and implement timely interventions to prevent preterm birth. Our study aimed to identify the risk factors associated with preterm birth in singleton pregnancies following ART and to develop an effective risk assessment model that can optimize post-perinatal management and reduce the incidence of preterm birth.

Supplementary Material 1. Supplementary Material 1.