Skip to main content
Download PDF

Predicting risk of the subsequent early pregnancy loss in women with recurrent pregnancy loss based on preconception data

BMC Women's Health volume 24, Article number: 381 (2024) Cite this article

Abstract

Background

For women who have experienced recurrent pregnancy loss (RPL), it is crucial not only to treat them but also to evaluate the risk of recurrence. The study aimed to develop a risk predictive model to predict the subsequent early pregnancy loss (EPL) in women with RPL based on preconception data.

Methods

A prospective, dynamic population cohort study was carried out at the Second Hospital of Lanzhou University. From September 2019 to December 2022, a total of 1050 non-pregnant women with RPL were participated. By December 2023, 605 women had subsequent pregnancy outcomes and were randomly divided into training and validation group by 3:1 ratio. In the training group, univariable screening was performed on RPL patients with subsequent EPL outcome. The least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression were utilized to select variables, respectively. Subsequent EPL prediction model was constructed using generalize linear model (GLM), gradient boosting machine (GBM), random forest (RF), and deep learning (DP). The variables selected by LASSO regression and multivariate logistic regression were then established and compared using the best prediction model. The AUC, calibration curve, and decision curve (DCA) were performed to assess the prediction performances of the best model. The best model was validated using the validation group. Finally, a nomogram was established based on the best predictive features.

Results

In the training group, the GBM model achieved the best performance with the highest AUC (0.805). The AUC between the variables screened by the LASSO regression (16-variables) and logistic regression (9-variables) models showed no significant difference (AUC: 0.805 vs. 0.777, P = 0.1498). Meanwhile, the 9-variable model displayed a well discrimination performance in the validation group, with an AUC value of 0.781 (95%CI 0.702, 0.843). The DCA showed the model performed well and was feasible for making beneficial clinical decisions. Calibration curves revealed the goodness of fit between the predicted values by the model and the actual values, the Hosmer–Lemeshow test was 7.427, and P = 0.505.

Conclusions

Predicting subsequent EPL in RPL patients using the GBM model has important clinical implications. Future prospective studies are needed to verify the clinical applicability.

Trial registration

This study was registered in the Chinese Clinical Trial Registry with the registration number of ChiCTR2000039414 (27/10/2020).

Peer Review reports

Background

Pregnancy loss (PL) is defined as the spontaneous demise of a pregnancy before the fetus reaches viability (from the time of conception until 24 weeks of gestation), also referred to as a miscarriage or spontaneous abortion, is one of the common health problems in childbearing women, which impacts 10–15% of clinically recognized pregnancies [1]. And early PL (EPL) is the loss pregnancy before 10 weeks of gestational age, which accounts for more than 80% of all PLs [2, 3]. 1-3% of women of childbearing age will have two or more PLs, known as recurrent pregnancy loss (RPL). Studies have identified various risk factors being related to PL, including age, BMI, ethnicity, previous miscarriages, uterine anatomy abnormalities, chromosomal abnormalities, infection, immune dysfunction, endocrine disturbance, and unhealthy lifestyle [4, 5], but there still approximately 60% of RPL cases remains unexplained, and these cases are referred to as unexplained RPL (URPL) [3].

RPL greatly affects the physical and mental health of couples of childbearing ages [6]. Women with a history of RPL showed more psychological problems during their subsequent pregnancy. Notably, couples must deal with the cumulative emotional effects by the subsequent RPL [7]. Whereas, some patients come to the hospital seek for help when they had a pregnancy-related concern (bleeding, abdominal pain, or worsening anxiety due to prior miscarriages or ectopic pregnancy), at which point the risk of pregnancy loss is increased. In addition, clinicians use clinical and demographic information to predict pregnancy outcomes after a patient becomes pregnant, but some early pregnancy loss is inevitable. For women who have experienced RPL, it is not only important to diagnose and treat them, but also to evaluation of the risk for recurrence which can reduce the subsequent PL rates.

The risk of RPL increased with the number of times of pregnancy loss, and the incidence of pregnancy loss was only 11.6% in women without a history of pregnancy loss. In women with a history of one, two, and three or more pregnancy losses, the probability of subsequent pregnancy loss was 19.8%, 27.7%, and 41.9%, respectively [8]. A population-based study has found that the lowest risk of PL (9.8%) in women aged 25–29 years and the risk of PL increases in women aged 30–35 years, then rises steeply to 33.2% in women aged 40–44 year [8]. The number of previous PLs was another independent risk factor for RPL [9]. Genetic factors may also be involved in the risk of miscarriage. One large genome-wide association study identified four distinct susceptibility loci for sporadic and RPL that have a role in progesterone production, placentation, and gonadotropin regulation [10]. There is research found that women experiencing bleeding without nausea between 6- and 8-weeks’ gestation had an increased risk of clinical pregnancy loss, but bleeding and nausea were not predictive risk factors of clinical pregnancy loss prior to 6 weeks’ gestation [11]. A meta-analysis found that early pregnancy ultrasound markers, including fetal bradycardia, crown rump length, intra uterine hematoma, and mean gestational sac diameter minus could predicting miscarriage in women with diagnosed viable intrauterine pregnancy [12]. In recent years, some novel markers of immune tolerance and angiogenesis in maternal blood have been reported as potential RPL predictors, including immune tolerance proteins galectin-9 (Gal-9) and interleukin (IL)-4, and angiogenesis proteins (vascular endothelial growth factors (VEGF) A, C, and D) [13].

Among these risk factors, some of them are limited by their late appearance or poor temporary availability and it is difficult to comprehensive assessment of the risk of subsequent RPL. Furthermore, most RPL risk assessment use traditional regression models (e.g., logistic regression), which make an implicit assumption that each risk factor is linearly related to PL [14], this may ignore the complex relationships of many risk factors with non-linear interactions, and the predictive performance is always suboptimal. Therefore, exploring an effective prediction model to predict the subsequent EPL for RPL patients is necessary.

Recently, machine learning (ML) methods, has been reported to demonstrate a powerful self-learning ability with improved prediction accuracy [15, 16], and it has been successfully applied to diagnosing diseases and predicting clinical outcomes, such as for in vitro fertilization treatment [15], RPL [16], postpartum hemorrhage [17] and other pregnancy pathological events [18]. Furthermore, the nomogram as a simple statistical visual tool, which is widely used to predict the occurrence of diseases. In this study, we develop and validate a prediction machine learning model based on the preconception demographic information, reproductive history, and clinical blood parameters of admission to identify the risk of subsequent EPL for RPL patients.

Materials and methods

Participants

The study population was drawn from a prospective, dynamic cohort, which was carried out at the Department of Reproductive Medicine, Second Hospital of Lanzhou University [19]. The cohort began in September 2019 and enrolled 1050 nonpregnant RPL patients through December 2022. The inclusion criteria were: (1) Have experienced at least two history of PL that meets the diagnostic criteria of the ESHRE; (2) aged 18–42 years. The follow-up period ended in December 2023 and the exclusion criteria were: (1) Patients who were lost to follow-up and who were not yet pregnant; (2) Subsequent pregnancy outcomes are ectopic pregnancy, hydatidiform mole, dysplasia, and current pregnancy < 10 weeks; (3) Subsequent pregnancies were assisted reproductive technology and twin pregnancies. This study was approved by the Ethics Committee of Lanzhou University Second Hospital (Ethical Approval Number: 2019 A-231). All patients provided written informed consent.

Predictive variables

Demographic information (including age, height, weight, education, ethnic, menarche, menstrual cycle, and pelvic surgery), reproductive history (including total pregnancy numbers, pregnancy loss numbers, induced abortion, live birth, and pregnancy type) were obtained from outpatient medical records and body mass index (BMI) was calculated as weight in kilograms divided by the square of body height in meter. Preconception treatments and subsequent pregnancy outcome was obtained from the follow-up. Each patient was followed up through electronic medical record system and telephone every 6 months after the first visit to track the patient’s pregnancy status, most recently in December 2023. Blood samples obtained at the initial visit in a nonpregnant state in in the morning, when the patient was underwent overnight fasting, and was tested according to the standard manufacturer’s protocols within an hour at our hospital and the blood parameters including 50 indicators. The demographic information and blood test indicators are presented in Table 1.

Table 1 The baseline demographic information and blood test indicators of RPL patients between the training and validation group
Full size table

Outcomes

The primary outcomes of this analysis included subsequent EPL and ongoing pregnancy (OP). EPL was defined as pregnancy less than 10 weeks of gestational age, including biochemical pregnancy. OP was defined as pregnancy beyond 10 weeks of gestational age.

Statistical analyses

Continuous variables are described by mean ± standard deviation or median (interquartile range), categorical variables are described using percentages. Independent sample T test, Mann–Whitney–Wilcoxon test, chi-square test and Fisher’s exact test were used appropriate. Multiple imputations were performed for a few missing variables (details of statistical methods for each variable, the number and percentage of missing data are presented in Supplementary Table 1). We generated five data sets by multiple imputations, and sensitivity analysis showed that these five data sets were not significantly different from the original data, the results are presented in Supplementary Table 2. All analyses were performed using Empower(R) (www.empowerstats.com, X&Y solutions, inc.Boston MA) and R (http://www.R-project.org).

Table 2 The baseline demographic information and blood test indicators of RPL patients between subsequent EPL and OP in the training group
Full size table

Variable selection

The dataset of the RPL women was randomly split into the development (75%) and validation (25%) groups. Data for 63 variables during pre-pregnancy were obtained from the patient self-reports and electronic medical records. First, in order to create an efficient approach for clinical practice with fewer redundant variables, we performed independent sample t test, Mann–Whitney–Wilcoxon test, chi-square test or Fisher’s exact test at the appropriate time, and then the variables with P < 0.05 were used the least absolute shrinkage and selection operator (LASSO) logistic regression algorithm and multivariate regression analysis with the training group to select related features.

Model training, evaluation, and validation

The features selected by LASSO regression were performed on the training group using the four models including generalize linear model (GLM), gradient boosting machine (GBM), random forest (RF), and deep learning (DP). After the model was established, we used area under the ROC curve (AUC), area under the precision-recall curve (AUCPR), logloss, mean per-class error, root mean square error (RMSE) and mean square error (MSE) to compare the models. And the model with the largest AUC value was selected as the best model. Next, the variables selected by LASSO regression and multivariate logistic regression were established and compared using the best prediction model in the training group. The AUC, calibration curve (Hosmer–Lemeshow test), and clinical decision curve (DCA) were performed to assess the prediction performances and clinical utility of the best model. We further performed an internal validation for the developed prediction model using the validation group. Finally, based on the best predictive features, a nomogram was established to take advantage of fitting a line with a non-linear relationship for the prediction of subsequent EPL.

Results

Baseline characteristics

Finally, 605 eligible RPL patients were enrolled in this stud, they were randomly divided into training group and validation group by 3:1 ratio, with 454 patients in the training group and 151 patients in the validation group. A flow chart of the process is represented in Fig. 1. Subsequent EPL occurred on 23.14% (140/605) in all patients and 23.35% (106/454) in the training group, 22.52% (34/151) in the validation group. Women with 423 (69.92%), 145 (23.97%), 37 (6.12%) had experienced 2, 3 or ≥ 4 prior pregnancy losses, respectively. No statistically significant difference between training group and validation group (P > 0.05) (Table 1).

Fig. 1
figure 1

Flow chart for RPL patient selection. RPL: recurrent pregnancy loss; EPL: early pregnancy loss; OP: ongoing pregnancy

Full size image

Screening of predictors

Sixty-three variables underwent preliminary univariable screening and 18 variables were found to be statistically significant in the training group were based on independent sample t test, Mann–Whitney–Wilcoxon test, chi-square test or Fisher’s exact test at the appropriate time with P < 0.05 for EPL vs. OP, including age, BMI, IgM, C4, TPO-Ab, ACA, β2GP1, LA, HCY, 2 h-BG, 2 h-INS, HDL, CHR, LHR, PNR, induced abortion, TPs, and PLs (Table 2). Then, the 18 variables move to the LASSO logistic regression model in the training set which found 16 variables (age, BMI, IgM, C4, TPO-Ab, ACA, β2GP1, LA, HCY, 2 h-BG, HDL, CHR, PNR, induced abortion, TPs, and PLs) to be predictable (Fig. 2A and B).

Fig. 2
figure 2

Clinical features selection using the least absolute shrinkage and selection operator (LASSO). A Tuning parameter(lambda) selection in the LASSO model used 10-fold cross-validation via minimum criteria. B LASSO coefficient profiles of the 16 clinical features. A coefficient profile plot was produced against the log (λ) sequence

Full size image

Subsequently, the 18 variables selected by univariate screening move to multivariate logistic regression which found that 9 EPL–related features [Age (OR: 1.07; 95% CI: 1.01 to 1.15), BMI (OR: 1.17; 95% CI: 1.00 to 1.37), PLs (OR: 3.56; 95% CI: 1.31 to 9.69), IgM (OR: 1.92; 95% CI: 1.13 to 3.26), HCY (OR: 1.20; 95% CI: 1.06 to 1.36), LHR (OR: 1.47; 95% CI: 1.03 to 2.11), PNR (OR: 1.02; 95% CI: 1.01 to 1.04), induced abortion (OR: 6.64; 95% CI: 1.75 to 25.23), and ACA (OR: 5.11; 95% CI: 1.40 to 18.67)] were identified as independent predictors of subsequent EPL in women with prior RPL (Table 3).

Table 3 A multivariate logistic regression found the risk factors for subsequent EPL in women with RPL
Full size table

Construction and validation of prediction models

Subsequent EPL prediction model was constructed in the training group using GLM, GBM, RF, and DP with 16 variables selected by LASSO logistic. The four models’ performance results including AUC, AUCPR, logloss, mean per-class error, RMSE and MSE are shown in Table 4. Overall, the GBM model achieved the best performance, with the highest AUC (0.805) and AUCPR (0.783).

Table 4 Performance results of four models in the training group based on 16 variables
Full size table

Then we used the 9 variables screened again by multivariate regression to construct GBM prediction models in the training group, and compared the prediction performance using 16 and 9 variables, respectively. The results showed that the use of 9 variables did not significantly reduce the prediction performance in the training group, the AUC in 16-variable and 9-variable models were 0.805 (95%CI 0.716, 0.878) and 0.777 (95%CI 0.690, 0.853), P > 0.05 (Table 5; Fig. 3A-C). The threshold probability of the DCA is 28% and the corresponding net benefit is 0.44 in 16-variable model, 23% and the corresponding net benefit is 0.48 in 9-variable model. It indicates that two models improve the benefit, and there was no significant difference between the two models (Fig. 3D). Figure 3E-F shows the calibration curve, which suggested that subsequent EPL by 16-variable and 9-variable model were essentially accurate, the Hosmer–Lemeshow test p-value of the two models were 0.607 and 0.559. Figure 3G-H shows the 95% CI in the training group for 16-variable and 9-variable model.

Table 5 Comparative analyses in models constructed using 16-variable and 9-variable in training group with GBM
Full size table
Fig. 3
figure 3

The ROC, DCA, calibration plots, and 95% confidence intervals in the training group for the 16-variables and 9-variables model using GBM. Model 1: 16 variables prediction model; Model 2: 9 variables prediction model. A-B ROC curves for models 1 and 2. C Comparison of ROC curves between model 1 and model 2. D DCA for model 1 and model 2. The y-axis represents the standardized net benefit (sNB), the X-axis represents the threshold probability. The cost-benefit ratio is also shown below the DCA. The black line represents the net benefit when all subjects not occured EPL, and the gray line is the net benefit at each risk threshold when all subjects occurred EPL. The blue and red lines are the net benefits of the risk probabilities estimated by models 1 and 2 at the risk threshold. E-F Calibration curves for model 1 and model 2. G-H The 95% confidence intervals for model 1 and model 2

Full size image

Because the 16-variable model did not significantly improve predictive power over the 9-variable model, we used the 9-variable model for internal validation based on the validation group. The predictive model displayed a well discrimination performance in the validation group, with an AUC value of 0.781 (95%CI 0.702, 0.843) (Fig. 4A). The DCA showed the model performed well and was feasible for making beneficial clinical decisions (Fig. 4B). Calibration curves revealed the goodness of fit between the predicted values by the model and the actual values, the Hosmer–Lemeshow test was 7.427, and P = 0.505 (Fig. 4C).

Fig. 4
figure 4

The ROC, DCA, and calibration plots in the validation group for the 9-variables using GBM. A ROC curves for GBM model in the validation group. B DCA for GBM model in the validation group. The y-axis represents the standardized net benefit (sNB), the X-axis represents the threshold probability. The cost-benefit ratio is also shown below the DCA. The black line represents the net benefit when all subjects not occurred EPL, and the gray line is the net benefit at each risk threshold when all subjects occurred EPL. The red line is the net benefits of the risk probabilities estimated at the risk threshold. C Calibration curves for GBM model in the validation group

Full size image

Finally, 9 variables were finally selected for nomogram presentation in Fig. 5. A total score was obtained by adding matching points for each parameter in the nomogram to evaluate subsequent EPL possibility. The prediction model calculation formula was as follows: subsequent EPL prediction model score = -10.0695 + 0.03738*age + 0.12240*BMI + 0.90452*induced abortion + 1.05194*ACA + 0.62670*IgM + 0.86737*LHR + 0.15993*HCY + 0.02297*PNR + 0.51565*PLs.

Fig. 5
figure 5

Nomogram for the subsequent EPL prediction of RPL with 9 variables, which including age, BMI, PLs, induced abortion, ACA, HCY, IgM, LHR, and PNR

Full size image

Discussion

The factors affecting the pregnancy outcome of RPL patients are complex and diverse, but it is worth mentioning that a comprehensive review of guidelines states that genetic thrombophilia, vaginal infections, and immunologic and male factors of infertility are not recommended as part of routine RPL investigations and there is also some controversy about the need for ovarian reserve testing, thyroid disease, screening for diabetes or hyperhomocysteinemia, measurement of prolactin levels, and endometrial biopsy [20]. In our study, we first compared the accuracy of multiple machine learning algorithms (GLM, GBM, RF and DP) in predicting subsequent recurrent EPL in patients with RPL, using demographic information and multiple clinical parameters of women with RPL before pregnancy. Subsequently, the best prediction model was used to find that there were no detailed differences in the prediction models constructed with 16 variables and 9 variables. Meanwhile, the 9-variable model displayed a well discrimination performance in the validation group, and the DCA showed the model performed well and was feasible for making beneficial clinical decisions. Calibration curves revealed the goodness of fit between the predicted values by the model and the actual values, the Hosmer–Lemeshow test was 7.427, and P = 0.505. The 9 variables include age, BMI, PLs, induced abortion, ACA, HCY, IgM, LHR, and PNR. Our study brings forward the risk assessment of subsequent EPL in women with RPL before pregnancy, which has very important clinical implications.

The association between female age and RPL has been consistently demonstrated in several studies. Studies have shown that couples should start trying to conceive when the woman is 31 or less to have at least a 90% chance of having a two-child family, and if IVF is not feasible, couples should start planning no later than 27. In order to achieve a one-child family, couples should start trying before the age of 32, or 35 if IVF is an option [21]. There are also variations in the threshold of BMI for pregnancy. Zhang et al. reported that, a BMI of 24.0 kg/m2 or greater was associated with an increased risk of RPL, but Lo and colleagues demonstrated that maternal obesity (BMI ≥ 30.0 kg/m2) significantly increased the risk of miscarriage in couples with unexplained RPL and there was no increased risk in women with overweight and underweight [22, 23]. The conflicting results may due to differences in study design, varying definitions of RPL and BMI ranges. A systematic review and meta-analysis found that the maternal BMI of women with a history of RPL is significantly higher than the BMI of controls, mean difference 0.7 kg/m2 [95% CI 0.2–1.3]. It is recommended that BMI be discussed as part of preconception and abortion counseling [24].

We found that for patients with RPL, previous induced abortion increased the risk of RPL recurrence, however previous studies have found that the risk of spontaneous abortion decreases with the increase in the number of induced abortions. This is not consistent with our results. The possible reason is the reference population was derived from all female workers in the Jinchang cohort in China, most of whom had normal reproductive function [25]. In addition, recent studies have found that for IVF patients, termination, miscarriage, ectopic pregnancy, or prior live birth does not compromise subsequent live birth and perinatal outcomes [26]. We also found that the risk of RPL recurrence increased with the number of previous miscarriages. Some studies have found that ≥ 4 previous miscarriages increase the cumulative clinical pregnancy loss rate and reduce the cumulative live birth rate in young women [27], however other studies found that the risk of further miscarriage following two or three RPLs is similar [28]. In a nested cohort, it was demonstrated that the number of prior miscarriages was a determinant both for time to live birth and cumulative incidence of live birth [29, 30]. It is worth noting that for secondary unexplained RPL, only consecutive pregnancy losses after the birth influenced the subsequent prognosis, while the number of losses prior to the birth did not affect the prognosis in the next pregnancy [31].

The negative effects of elevated HCY levels on pregnancy are well known, which is associated with a variety of pregnancy complications, such as preeclampsia (PE), early PL (EPL), placental abruption (PA), intrauterine growth restriction (IUGR) and venous thrombosis [32]. Approximately one third of spontaneous abortion before 20 weeks’ gestation are associated with elevated HCY levels [33]. A longitudinal study based on Chinese population has explored the reference intervals of HCY in three periods of pregnancy, which provides a basis for the management and detection of HCY in Chinese women during pregnancy [34]. However, most studies on the relationship between HCY and pregnancy diseases have focused on the first trimester, ignoring the effect of HCY before pregnancy. Research from our team found that for women with a previous miscarriage, HCY can increase the uterine artery resistance in the non-pregnant state and is associated with the abortion rate of subsequent pregnancy [35]. The present study found that pre-pregnancy HCY plays an important role in the recurrence of RPL in women based on pre-pregnancy data, suggesting that pre-detection of HCY levels in women trying to become pregnant has a positive effect on preventing the recurrence of RPL.

Antiphospholipid antibodies (aPLs) are the leading causes of adverse pregnancy outcomes (APOs). A cluster analysis found that patients with triple antibodies or high-risk lupus characteristics were prone to occurred gestational hypertension and premature delivery. Isolated LA or ACA/aβ2GPI positivity were found to be more frequently associated with early-stage fetal loss [36]. Takeshita et al. found that the only risk factor for persistently positive ACA antibodies is a high antibody titer during the initial test. When the ACA antibody titer in the initial test exceeds the cut-off value (ACA - IgG antibodies > 15 U/mL and ACA - IgM antibodies > 11 U/mL), treatment can be initiated immediately [37]. The complement system has attracted attention as a potential mediator of pathogenic mechanisms induced by aPL. Complement C3 and C4 serum levels were assessed in several cohorts of pregnant patients with APS and/or aPL, these studies have yielded inconsistent results, while some studies have come to find a correlatio, other studies have not revealed a prognostic role for the complement in relation to pregnancy morbidity among aPL-positive women [38,39,40]. Our study reconfirmed the important effect of positive aPLs and C4 on the outcome of the next pregnancy of RPL patient. But a meta-analysis found that the presence of positive aPL neither decreased clinical pregnancy rate and live birth rate, nor increased miscarriage rate in women undergoing IVF, which is differed from the opinion of clinical practice [41].

Glucose and lipid metabolism levels are not included in the routine screening program of RPL patients in the current guidelines. In this study, 2-hour postprandial glucose and 2-hour postprandial insulin were significantly elevated in women with subsequent EPL, however, there was no significant difference in FBG and FINS between the EPL and OP groups. Study have found that 2-h postprandial glycemia level is more precise than fasting glycemia for type 2 diabetes [42]. As early as 10 years ago, researchers had found that the 1-, 2-, and 3-hour plasma glucose and insulin levels were significantly higher in women with RPL (more than 2 PLs) as compared to controls [43]. Numerous studies associate abnormal glucose metabolism in the endometrium with a higher risk of adverse pregnancy outcomes [44]. Furthermore, several studies have linked altered levels of lipids and a higher risk of adverse pregnancy outcomes, which is visible in patients with recurrent miscarriage (RM) [45, 46], HDL concentrations were lower in women with RM and together with the abdominal obesity were the most frequent components of the RM profile [47]. Subsequently, the study by Depciuch et al. reconfirmed that changes in the metabolomic and lipidomic pathways may be potential risk factors as well as therapeutic targets for RM [48]. In addition, a retrospective cohort study found serum lipid levels were associated with treatment outcomes in women undergoing assisted reproduction, higher HDL-C was associated with greater numbers of oocytes retrieved, higher live birth rates, and lower miscarriage rates [49]. Besides, in long-term follow-up, the researchers found that females with history of PL were experienced more prediabetes (50% vs. 45.5%), diabetes (28.9% vs. 21.3%), and metabolic syndrome (70% vs. 60.1%) than females without such history [50]. This also revealed the interaction between metabolism and RPL.

The role of systemic inflammatory reactions in the pathogenesis of EPL has been confirmed in several studies [51,52,53,54]. Inflammatory markers from complete blood count (CBC), such as platelet lymphocyte ratio (PLR), neutrophil-lymphocyte ratio (NLR), neutrophil to monocyte ratio (NMR), lymphocyte to monocyte ratio (LMR) and platelet to neutrophil ratio (PNR) are readily available [55]. This study we found that elevated PNR was a risk factor for EPL in RPL patients in their next pregnancy, and no statistically significant differences were found in the remaining inflammatory markers between the EPL and OP groups. However, the levels of PNR in different diseases are found to be inconsistent. In reproductive events, lower PNR were associated with early natural menopause [56]. For mothers with hypertensive disorders of pregnancy, neonatal Apgar scores at 1 and 5 min of birth were positively associated with PNR [57]. For patients with ST-elevation myocardial infarction, a nonlinear relationship was found between the PNR and major adverse cardiovascular events, which was positively associated with the PNR when the PNR exceeded threshold [58]. For patients with ovarian cancer, PNR were independent prognostic indicators of poor relapse-free survival [59].

However, our study still has some limitations. First, there are some missing data and it is necessary to evaluate the characteristics of RPL women more comprehensively, such as blood pressure, waist circumference and hip circumference, and register more detailed reproductive history and examination results. Second, the inflammatory and immune status of the endometrium are thought to be closely related to RPL, but such data were lacking in this study. Finally, this study is a single-center study and lacks external validation, the model presented here needs further study with more multi-center clinical data. Based on the above shortcomings, our team is carrying out a cohort study of RPL patients to observe and record the whole process of RPL patients from the first visit to the subsequent pregnancy outcome in detail, to provide more accurate clinical treatment for RPL patients.

Conclusions

Our study innovated the use of pre-pregnancy demographic data and clinical laboratory indicators to predict subsequent EPL in RPL patients, which has important clinical implications. Age, BMI, PLs, induced abortion, ACA, HCY, IgM, LHR, and PNR are the key factors affecting subsequent EPL.

Availability of data and materials

No datasets were generated or analysed during the current study.

References

  1. Atik RB, Christiansen OB, Elson J, Kolte AM, Lewis S, Middeldorp S, et al. ESHRE guideline: recurrent pregnancy loss. Hum Reprod open. 2018;2018:hoy004.

    Article  Google Scholar 

  2. Kolte AM, Bernardi LA, Christiansen OB, Quenby S, Farquharson RG, Goddijn M, et al. Terminology for pregnancy loss prior to viability: a consensus statement from the ESHRE early pregnancy special interest group. Hum Reprod (Oxford England). 2015;30:495–8.

    Article  CAS  Google Scholar 

  3. Dimitriadis E, Menkhorst E, Saito S, Kutteh WH, Brosens JJ. Recurrent pregnancy loss. Nat Rev Dis Primers. 2020;6:98.

    Article  PubMed  Google Scholar 

  4. Coomarasamy A, Dhillon-Smith RK, Papadopoulou A, Al-Memar M, Brewin J, Abrahams VM, et al. Recurrent miscarriage: evidence to accelerate action. Lancet (London England). 2021;397:1675–82.

    Article  CAS  PubMed  Google Scholar 

  5. Quenby S, Gallos ID, Dhillon-Smith RK, Podesek M, Stephenson MD, Fisher J, et al. Miscarriage matters: the epidemiological, physical, psychological, and economic costs of early pregnancy loss. Lancet (London England). 2021;397:1658–67.

    Article  CAS  PubMed  Google Scholar 

  6. Tavoli Z, Mohammadi M, Tavoli A, Moini A, Effatpanah M, Khedmat L, et al. Quality of life and psychological distress in women with recurrent miscarriage: a comparative study. Health Qual Life Outcomes. 2018;16:150.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Geller PA, Kerns D, Klier CM. Anxiety following miscarriage and the subsequent pregnancy: a review of the literature and future directions. J Psychosom Res. 2004;56:35–45.

    Article  PubMed  Google Scholar 

  8. Magnus MC, Wilcox AJ, Morken NH, Weinberg CR, HÃ¥berg SE. Role of maternal age and pregnancy history in risk of miscarriage: prospective register based study. BMJ (Clinical Res ed). 2019;364:l869.

    Google Scholar 

  9. Maconochie N, Doyle P, Prior S, Simmons R. Risk factors for first trimester miscarriage–results from a UK-population-based case-control study. BJOG: Int J Obstet Gynecol. 2007;114:170–86.

    Article  CAS  Google Scholar 

  10. Laisk T, Soares ALG, Ferreira T, Painter JN, Censin JC, Laber S, et al. The genetic architecture of sporadic and multiple consecutive miscarriage. Nat Commun. 2020;11:5980.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. DeVilbiss EA, Naimi AI, Mumford SL, Perkins NJ, Sjaarda LA, Zolton JR, et al. Vaginal bleeding and nausea in early pregnancy as predictors of clinical pregnancy loss. Am J Obstet Gynecol. 2020;223:e5701–14.

    Article  Google Scholar 

  12. Pillai RN, Konje JC, Richardson M, Tincello DG, Potdar N. Prediction of miscarriage in women with viable intrauterine pregnancy-A systematic review and diagnostic accuracy meta-analysis. Eur J Obstet Gynecol Reprod Biol. 2018;220:122–31.

    Article  PubMed  Google Scholar 

  13. Wyatt MA, Baumgarten SC, Weaver AL, Van Oort CC, Fedyshyn B, Ruano R, et al. Evaluating markers of immune tolerance and angiogenesis in maternal blood for an association with risk of pregnancy loss. J Clin Med. 2021;10:3579.

  14. Barnett-Itzhaki Z, Elbaz M, Butterman R, Amar D, Amitay M, Racowsky C, et al. Machine learning vs. classic statistics for the prediction of IVF outcomes. J Assist Reprod Genet. 2020;37:2405–12.

    Article  PubMed  PubMed Central  Google Scholar 

  15. Curchoe CL, Bormann CL. Artificial intelligence and machine learning for human reproduction and embryology presented at ASRM and ESHRE 2018. J Assist Reprod Genet. 2019;36:591–600.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Bruno V, D’Orazio M, Ticconi C, Abundo P, Riccio S, Martinelli E, et al. Machine learning (ML) based-method applied in recurrent pregnancy loss (RPL) patients diagnostic work-up: a potential innovation in common clinical practice. Sci Rep. 2020;10:7970.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Venkatesh KK, Strauss RA, Grotegut CA, Heine RP, Chescheir NC, Stringer JSA, et al. Machine learning and statistical models to predict postpartum hemorrhage. Obstet Gynecol. 2020;135:935–44.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Ye Y, Xiong Y, Zhou Q, Wu J, Li X, Xiao X. Comparison of machine learning methods and conventional logistic regressions for predicting gestational diabetes using routine clinical data: a retrospective cohort study. J Diabetes Res. 2020;2020:4168340.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Yang X, Mu F, Zhang J, Yuan L, Zhang W, Yang Y, et al. Reproductive factors and subsequent pregnancy outcomes in patients with prior pregnancy loss. BMC Pregnancy Childbirth. 2024;24:219.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Giouleka S, Tsakiridis I, Arsenaki E, Kalogiannidis I, Mamopoulos A, Papanikolaou E, et al. Investigation and management of recurrent pregnancy loss: a comprehensive review of guidelines. Obstet Gynecol Surv. 2023;78:287–301.

    Article  PubMed  Google Scholar 

  21. Habbema JD, Eijkemans MJ, Leridon H, te Velde ER. Realizing a desired family size: when should couples start? Human reproduction (Oxford. England). 2015;30:2215–21.

    Google Scholar 

  22. Zhang BY, Wei YS, Niu JM, Li Y, Miao ZL, Wang ZN. Risk factors for unexplained recurrent spontaneous abortion in a population from southern China. Int J Gynaecol Obstet. 2010;108:135–8.

    Article  PubMed  Google Scholar 

  23. Lo W, Rai R, Hameed A, Brailsford SR, Al-Ghamdi AA, Regan L. The effect of body mass index on the outcome of pregnancy in women with recurrent miscarriage. J Fam Commun Med. 2012;19:167–71.

    Article  Google Scholar 

  24. Eapen A, Hayes ET, McQueen DB, Beestrum M, Eyck PT, Boots C. Mean differences in maternal body mass index and recurrent pregnancy loss: a systematic review and meta-analysis of observational studies. Fertil Steril. 2021;116:1341–8.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Hu X, Miao M, Bai Y, Cheng N, Ren X. Reproductive factors and risk of spontaneous abortion in the Jinchang Cohort. Int J Environ Res Public Health. 2018;15:2444

  26. Chen D, Xu Q, Mao X, Zhang J, Wu L. Reproductive history does not compromise subsequent live birth and perinatal outcome following in-vitro fertilization: analysis of 25 329 first frozen-thawed embryo transfer cycles without preimplantation genetic testing for aneuploidy. Ultrasound Obstet Gynecology: Official J Int Soc Ultrasound Obstet Gynecol. 2023;62:430–8.

    Article  CAS  Google Scholar 

  27. Zhao ST, Liu YB, Zhang Q, Ni TX, Yan JH. [The impact of the number of previous miscarriages on embryo euploid rate and pregnancy outcomes in patients with unexplained recurrent pregnancy loss]. Zhonghua Yi Xue Za Zhi. 2023;103:2874–80.

    CAS  PubMed  Google Scholar 

  28. Bashiri A, Ratzon R, Amar S, Serjienko R, Mazor M, Shoham-Vardi I. Two vs. three or more primary recurrent pregnancy losses–are there any differences in epidemiologic characteristics and index pregnancy outcome? J Perinat Med. 2012;40:365–71.

    Article  PubMed  Google Scholar 

  29. Kling C, Magez J, Hedderich J, von Otte S, Kabelitz D. Two-year outcome after recurrent first trimester miscarriages: prognostic value of the past obstetric history. Arch Gynecol Obstet. 2016;293:1113–23.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Kaandorp SP, van Mens TE, Middeldorp S, Hutten BA, Hof MH, van der Post JA, et al. Time to conception and time to live birth in women with unexplained recurrent miscarriage. Hum Reprod (Oxford England). 2014;29:1146–52.

    Article  CAS  Google Scholar 

  31. Egerup P, Kolte AM, Larsen EC, Krog M, Nielsen HS, Christiansen OB. Recurrent pregnancy loss: what is the impact of consecutive versus non-consecutive losses? Human reproduction (Oxford. England). 2016;31:2428–34.

    CAS  Google Scholar 

  32. Gaiday AN, Tussupkaliyev AB, Bermagambetova SK, Zhumagulova SS, Sarsembayeva LK, Dossimbetova MB, et al. Effect of homocysteine on pregnancy: a systematic review. Chemico-Biol Interact. 2018;293:70–6.

    Article  CAS  Google Scholar 

  33. Nasiri M, Arsanjani Shirazi A, Sadeghi O, Bagheri Bidakhavidi M. The relationship between homocysteine levels and spontaneous abortion in Iranian women with migraine. Iran J Public Health. 2017;46:1149–51.

    PubMed  PubMed Central  Google Scholar 

  34. Liu H, Li Q, Liu T, Tang Y, Yu F. Trimester-specific reference intervals of serum homocysteine levels for pregnant women: a longitudinal study in China. Gynecol Endocrinology: Official J Int Soc Gynecol Endocrinol. 2023;39:2242974.

    Article  Google Scholar 

  35. Yang X, Tian X, Liu H, Wang J, Wang F. Homocysteine increases uterine artery blood flow resistance in women with pregnancy loss. J Gynecol Obstet Hum Reprod. 2023;52:102533.

    Article  PubMed  Google Scholar 

  36. Long Y, Huang C, Cui Y, Xie Z, Zhou Y, Shi X, et al. Cluster analysis of antiphospholipid antibodies-associated adverse pregnancy outcome patients: based on a 13-years cohort study. Clinical and experimental medicine.; 2023.

    Google Scholar 

  37. Yokote R, Kuwabara Y, Kasano S, Yonezawa M, Ouchi N, Ichikawa T, et al. Risk factors for persistent positive anticardiolipin antibodies in women with recurrent pregnancy loss. J Reprod Immunol. 2023;156:103920.

    Article  CAS  PubMed  Google Scholar 

  38. Reggia R, Ziglioli T, Andreoli L, Bellisai F, Iuliano A, Gerosa M, et al. Primary anti-phospholipid syndrome: any role for serum complement levels in predicting pregnancy complications? Rheumatology (Oxford. England). 2012;51:2186–90.

    CAS  Google Scholar 

  39. Deguchi M, Yamada H, Sugiura-Ogasawara M, Morikawa M, Fujita D, Miki A, et al. Factors associated with adverse pregnancy outcomes in women with antiphospholipid syndrome: a multicenter study. J Reprod Immunol. 2017;122:21–7.

    Article  PubMed  Google Scholar 

  40. Nalli C, Lini D, Andreoli L, Crisafulli F, Fredi M, Lazzaroni MG, et al. Low preconception complement levels are Associated with adverse pregnancy outcomes in a Multicenter Study of 260 pregnancies in 197 women with antiphospholipid syndrome or carriers of antiphospholipid antibodies. Biomedicines. 2021;9:671.

  41. Tan XF, Xu L, Li TT, Wu YT, Ma WW, Ding JY, et al. Serum antiphospholipid antibody status may not be associated with the pregnancy outcomes of patients undergoing in vitro fertilization. Medicine. 2022;101:e29146.

    Article  CAS  PubMed  Google Scholar 

  42. García Nava RE, Rodríguez Moctezuma R, Munguía Miranda C, Hernández Santiago JL, Casas de la Torre E, Martínez Bermúdez M, et al. Comparison of fasting and 2 hours postprandial blood glucose in the control of patients with type 2 diabetes. Revista de investigacion clinica; organo del Hospital de Enfermedades de la Nutricion. 2003;55:26–30.

    PubMed  Google Scholar 

  43. Wang Y, Zhao H, Li Y, Zhang J, Tan J, Liu Y. Relationship between recurrent miscarriage and insulin resistance. Gynecol Obstet Invest. 2011;72:245–51.

    Article  CAS  PubMed  Google Scholar 

  44. Vrhovac Madunić I, Karin-Kujundžić V, Madunić J, Šola IM, Šerman L. Endometrial glucose transporters in health and disease. Front cell Dev Biology. 2021;9:703671.

    Article  Google Scholar 

  45. Baig S, Lim JY, Fernandis AZ, Wenk MR, Kale A, Su LL, et al. Lipidomic analysis of human placental syncytiotrophoblast microvesicles in adverse pregnancy outcomes. Placenta. 2013;34:436–42.

    Article  CAS  PubMed  Google Scholar 

  46. Jiang S, Jiang J, Xu H, Wang S, Liu Z, Li M, et al. Maternal dyslipidemia during pregnancy may increase the risk of preterm birth: a meta-analysis. Taiwan J Obstet Gynecol. 2017;56:9–15.

    Article  PubMed  Google Scholar 

  47. Hilali NG, Sak S, Incebiyik A, Uyanikoglu H, Sak ME, Incebiyik H, et al. Recurrent pregnancy loss and metabolic syndrome. Ginekologia Polska. 2020;91:320–3.

    Article  PubMed  Google Scholar 

  48. Guleken Z, Bahat PY, Toto ÖF, Bulut H, Jakubczyk P, Cebulski J, et al. Blood serum lipid profiling may improve the management of recurrent miscarriage: a combination of machine learning of mid-infrared spectra and biochemical assays. Anal Bioanal Chem. 2022;414:8341–52.

    Article  CAS  PubMed  Google Scholar 

  49. Cai WY, Luo X, Chen E, Lv H, Fu K, Wu XK, et al. Serum lipid levels and treatment outcomes in women undergoing assisted reproduction: a retrospective cohort study. Front Endocrinol. 2021;12:633766.

    Article  Google Scholar 

  50. Rahmati M, Saei Ghare Naz M, Azizi F, Ramezani Tehrani F. Pregnancy loss and subsequent risk of prediabetes, diabetes and metabolic syndrome in couples: Tehran lipid and glucose study. J Translational Med. 2022;20:372.

    Article  CAS  Google Scholar 

  51. Kalagiri RR, Carder T, Choudhury S, Vora N, Ballard AR, Govande V, et al. Inflammation in complicated pregnancy and its outcome. Am J Perinatol. 2016;33:1337–56.

    Article  PubMed  Google Scholar 

  52. Oğlak SC, Aydın MF. Are neutrophil to lymphocyte ratio and platelet to lymphocyte ratio clinically useful for the prediction of early pregnancy loss? Ginekologia Polska. 2020;91:524–7.

    Article  PubMed  Google Scholar 

  53. Yazdizadeh M, Hivehchi N, Ghaemi M, Azizi S, Saeedzarandi M, Afrooz N, et al. Platelet to lymphocyte and neutrophil to lymphocyte ratio in the first trimester of pregnancy, are they useful for predicting spontaneous miscarriage? A case-control study. Int J Reproductive Biomed. 2023;21:463–70.

    CAS  Google Scholar 

  54. Soysal C, Sarı H, Işıkalan MM, Özkaya EB, Ulaş Ö, Taşçı Y, et al. Role of the systemic immune-inflammation index in threatened abortion patients and predicting of abortion. J Obstet Gynaecol Res. 2023;49:1723–8.

    Article  CAS  PubMed  Google Scholar 

  55. Kutlay Ö, Yalım Z, Aktan AK. Inflammatory biomarkers derived from whole blood cell count in atrial fibrillation patients. Kardiologiia. 2023;63:50–5.

    Article  PubMed  Google Scholar 

  56. Cramer DW, Vitonis AF. Signatures of reproductive events on blood counts and biomarkers of inflammation: implications for chronic disease risk. PLoS ONE. 2017;12:e0172530.

    Article  PubMed  PubMed Central  Google Scholar 

  57. Okoye HC, Madu AJ, Korubo K, Efobi C, Eze OE, Obodo O, et al. Correlates of neutrophil/lymphocyte, platelet/lymphocyte, and platelet/neutrophil ratios of neonates of women with hypertensive disease of pregnancy with neonatal birth outcomes. Hypertens Pregnancy. 2019;38:105–10.

    Article  PubMed  Google Scholar 

  58. Wang H, Qing X, Wang H, Gu Y. Association between platelet to neutrophil ratio (PNR) and clinical outcomes in STEMI patients after successful pPCI: a secondary analysis based on a Cohort Study. Cardiovasc Ther. 2022;2022:2022657.

    Article  PubMed  PubMed Central  Google Scholar 

  59. Song Q, Xu SX, Wu JZ, Ling L, Wang S, Shu XH, et al. The preoperative platelet to neutrophil ratio and lymphocyte to monocyte ratio are superior prognostic indicators compared with other inflammatory biomarkers in ovarian cancer. Front Immunol. 2023;14:1177403.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

Not applicable.

Funding

This work was granted by the Medical Innovation and Development Project of Lanzhou University (Grant No. lzuyxcx-2022-137) and the real-world study of recurrent pregnancy loss in China (Grant No. YJS-BD-19).

Author information

Author notes

  1. Xin Yang and Ruifang Wang contribute equally to this research.

Authors and Affiliations

  1. Department of Reproductive Medicine, Lanzhou University Second Hospital, Lanzhou, 730030, China

    Xin Yang, Ruifang Wang, Wei Zhang, Yanting Yang & Fang Wang

  2. Department of Obstetrics and Gynecology, The First Affiliated Hospital, College of Clinical Medicine of Henan University of Science and Technology, Luoyang, 471003, China

    Ruifang Wang

Authors
  1. Xin Yang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Ruifang Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Wei Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Yanting Yang
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Fang Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Xin Yang: data analysis and manuscript writing. Ruifang Wang, Wei Zhang and Yanting Yang: collecting and preparing the clinical data. Fang Wang: designed the study. All authors contributed to the article and approved the submitted version.

Corresponding author

Correspondence to Fang Wang.

Ethics declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Lanzhou University Second Hospital (Ethical Approval Number: 2019 A-231). All subjects gave informed consent before participation. All methods were performed in accordance with the relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Yang, X., Wang, R., Zhang, W. et al. Predicting risk of the subsequent early pregnancy loss in women with recurrent pregnancy loss based on preconception data. BMC Women's Health 24, 381 (2024). https://doi.org/10.1186/s12905-024-03206-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12905-024-03206-9

Keywords

BMC Women's Health

ISSN: 1472-6874

Contact us