This prospective study included the pregnant mother with PE who delivered in the Obstetrics Department of Wuming Hospital from 1 October 2016 to 1 October 2021. According to the relevant criteria of the Expert Consensus on FGR (2019 edition), 303 patients were divided into 62 cases in the group with PE complicated with FGR (restricted group) and 241 cases in the group with PE alone (nonrestricted group) [9]. This study was approved by the Medical Ethics Committee of Wuming Hospital Affiliated to Guangxi Medical University [approval number: WM-2022 (217)].

Inclusion criteria included the women who meet the diagnostic criteria in the Guidelines for the Management of Hypertensive Disorders in Pregnancy (2020) [10] and the Expert Consensus on FGR (2019 Edition) [9]. We excluded the patient with multiple pregnancies, combined history of chronic hypertension, chronic liver and kidney diseases, and other prepregnancy comorbidities or fetal malformations.

Clinical and laboratory data were prospectively collected from patients in the study. Clinical data included age, gestational week, pregestational body mass index (BMI), number of deliveries, number of births, severity of PE, initial systolic blood pressure, initial diastolic blood pressure, diabetes mellitus, thrombophilia, fundus abnormalities, and amniotic fluid status.

Laboratory data included the results for complementary inflammation, coagulation status, hepatobiliary and renal function, lipid profile, and blood counting. Complement inflammatory parameters include complement component 1q (C1q), and C-reactive protein (CRP). The coagulation parameters included prothrombin time activity (PTA), activated partial thromboplastin time (APTT), and prothrombin time (PTA), fibrinogen (Fib) and thrombin time (TT). The main hepatobiliary indicators included albumin (ALB), alanine aminotransferase (ALT), and glutamate transferase (GTR), glutamic oxaloacetic transaminase (AST), total bile acid (TBA), total protein (TP). The renal function tests included uric acid (UA), urea (Urea), creatinine (Cr), and TBA. Lipid profile included apolipoprotein A-I (ApoA1), high-density lipoprotein cholesterol (HDL-C), lipoprotein (a) [Lp (a)], and serum total protein (serum TP), serum total cholesterol (TC), and triglyceride (TG). Blood cell indices mainly included absolute lymphocytosis (Lym), absolute neutrophil (Neu), platelet count (blood platelet, Plt), Neu/Lym ratio, and Plt/Lym ratio.

We used SPSS 22.0 (SPSS Inc., Chicago, IL, USA) and the R (3.6.1) softwares to analyze the data. Continuous data are expressed as the mean $\pm$ standard deviation (SD), and categorical data are presented as the frequency (percentage). The Chi-square test was used to compare the differences. Multivariate analysis and the logistic model were used to validate the XGBoost algorithm model. The accuracy of the XGBoost model was verified using ROC curves. Differences were considered statistically significant when p 0.05.

Prepregnancy BMI, number of births, number of maternity examination, eclampsia staging, and amniotic fluid excess were significantly different between the restricted and nonrestricted groups (p 0.05) (Table 1).

Patients in the restricted group had significantly higher C1q, CRP, PTA, APTT, ALB, ALT, AST, UA, Urea, Cr, ApoA1, low-density lipoprotein cholesterol (LDL-C), TC, TG, Neu, Plt, and Neu/Lym (p 0.05), but they showed lower TBA, P, HDL-C, Lp (a) (p 0.05) than those in the nonrestricted group (Table 2).

Multivariate analysis revealed that higher prepregnancy BMI, number of births, number of maternity examinations, eclampsia staging, C1q, CRP, PTA, APTT, ALB, ALT, AST, UA, Urea, Cr, ApoA1, LDL-C, TC, TG, Neu, Plt, Neu/Lym, presence of low amniotic fluid and lower TBA, P, HDL-C, Lp (a) were risk factors for FGR (Tables 3,4).

The risk factors for FGR were analyzed by the XGBoost algorithm model in order of importance as TG, TC, Lp (a), LDL-C, HDL-C, ApoA1, P, Cr, Urea, UA, TBA, AST, ALT, ALB, APTT, PTA, Neu, CRP, C1q, hypohydramnios, fundus abnormalities, Plt, eclampsia stage, number of births, number of deliveries, prenatal BMI. Plt, eclampsia staging, number of births, number of deliveries, and prepregnancy BMI (Fig. 1). ROC curve analysis revealed that the area under the curve of the XGBoost algorithm model was 0.851 showing a good fit (Fig. 2).

Fig. 1.

ROC curve of the XGBoost algorithm model. ROC, receiver operating characteristic; XGBoost, extreme Gradient Boosting.

Fig. 2.

ROC curve analysis of the XGBoost algorithm model showing the risk factors. ROC, receiver operating characteristic; XGBoost, extreme Gradient Boosting.

The clinical data and laboratory indicators of the two groups of patients showed that prepregnancy BMI, number of births, number of maternity examinations, eclampsia stage, and amniotic fluid were significantly higher in the group with FGR than in the group without FGR. In addition, a study showed that mothers with FGR had a lower BMI compared to mothers with normal fetuses, providing evidence that the maternal BMI plays a crucial role in intrauterine fetal growth and development during pregnancy [11]. A higher number of births can affect body functions and is an important cause of FGR. Regular check-ups during pregnancy promote the health knowledge of the mother and allow her to closely follow the development of the baby so that if any abnormalities occur, timely interventions can be made. Low amniotic fluid compresses the umbilical cord, which can lead to reduced blood circulation to the fetus resulting in hypoxia and asphyxia [12, 13]. Moreover, low amniotic fluid can cause alterations in fetal bowel movements, increased fetal defecation, and a higher risk of amniotic fluid contamination, which is also an significant cause of FGR [14, 15, 16].

In this study, complement inflammation, coagulation, liver, biliary and renal function, lipid profile, and blood cell counts were important factors that affect the growth and development of the fetus. During fetal development, the activation of the complement system is needed to remove the many cellular debris and apoptotic cells derived from the placenta [17]. Therefore, inadequate or excessive activation of the complement system can increase inflammatory response at the local lesion.

In this study, however, in the analysis of blood counts, there were significant abnormalities in Neu, Plt, and Neu/Lym in the patients with FGR. Coagulation disorders are important in activating the platelets and the rise in aggregation capacity, which, along with coagulation disorders, ultimately results in inadequate blood supply to the fetus [18]. Moreover, damage and subtle alterations to the local capillaries of the liver and kidney tissues due to the continuous development of coagulation and blood inflammatory reactions, resulting in abnormal metabolism of local proteins and enzymes as well as alterations in capillary permeability and barrier effects, are important in causing abnormalities in liver and kidney indicators [19]. High lipid levels are also an important cause of local spiral atherosclerosis and reduced vascular smoothness and local thrombosis, and impaired nutrient transport [20].

In this study, the above risk factors were incorporated into the XGBoost algorithm model for further analysis using machine learning, which effectively avoided overfitting in the conventional multifactor analysis [21]. Moreover, its analysis of risk factors was more accurate. The ROC curves analysis revealed a high accuracy of the XGBoost algorithm model for FGR.

In future clinical application, abnormal conditions of TG, TC and Lp (a) can be used as an important basis for the evaluation of fetal growth and development restriction in patients with preeclampsia and the evaluation of therapeutic intervention effects.