Increased first-trimester CA125 levels associated with the development of preeclampsia

Abstract

Background

Preeclampsia (PE) remains a leading cause of maternal and fetal mortality, necessitating reliable first-trimester biomarkers for early identification and pathogenesis elucidation. This study aimed to evaluate the predictive performance of first-trimester serum CA125 for the PE development and to examine their longitudinal associations with blood pressure trajectories during pregnancy.

Methods

A nested case-control study was conducted within the prospective BiCoS cohort, including 70 PE cases and 70 matched controls. Serum CA125 levels were measured at 12.26 ± 0.44 weeks of gestation and compared using independent t-tests. Logistic regression models were built to assess the predictive performance of CA125 alone, clinical factors, and their combination. Model discrimination was evaluated by the area under the receiver operating characteristic curve (AUC), with internal validation via bootstrapping. Associations between CA125 and clinical parameters, including hepatic and renal function profiles, platelet counts, blood lipid levels and blood pressure, were examined using Pearson correlation analysis. Linear mixed-effects models were used to analyze the association between first-trimester CA125 and subsequent longitudinal blood pressure changes.

Results

First-trimester serum CA125 levels were significantly elevated in the PE group compared to controls (25.44 ± 14.89 vs. 16.17 ± 9.61 U/mL, P < 0.0001), with particularly high levels observed in PE cases complicated by fetal growth restriction (34.94 ± 22.93 vs. 23.27 ± 11.61 U/mL, P = 0.010). Multivariable analysis confirmed that CA125 remained an independent predictor of PE after adjusting for clinical factors (adjusted OR 1.071, 95% CI 1.027–1.118, P < 0.001), along with parity and MAP. The combined model (CA125 + clinical factors) achieved superior discriminatory power (AUC 0.841, 95% CI 0.776–0.906) compared to the model with clinical factors (AUC 0.799), and demonstrated robustness upon internal validation (optimism-corrected AUC = 0.817). Cross-sectional correlation analysis revealed no significant associations between CA125 and a comprehensive panel of first-trimester hepatic, renal, metabolic, hematologic, or blood pressure parameters. Crucially, longitudinal mixed-effects models showed that higher first-trimester CA125 levels were significantly associated with steeper increases in both systolic blood pressure (interaction P = 0.015) and diastolic blood pressure (interaction P = 0.009) throughout gestation. The inclusion of CA125 significantly improved model fit for both SBP (χ² = 7.807, P = 0.005) and DBP (χ² = 7.024, P = 0.008), with interaction models demonstrating the best fit.

Conclusion

First-trimester serum CA125 serves as a robust and independent predictor of preeclampsia. Its integration with established clinical risk factors significantly improves predictive accuracy, facilitating early identification of of PE. The specific association of CA125 with accelerated blood pressure trajectories through gestation suggests its role as an early sentinel of aberrant placental-vascular pathophysiology. These findings position CA125 as a promising and practical biomarker for enhancing early risk stratification.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07565-1.

Introduction

Preeclampsia, a pregnancy-specific multisystem disorder, remains a leading cause of maternal and perinatal mortality worldwide, accounting for an estimated 46,000 maternal deaths and approximately 500,000 fetal and neonatal deaths annually [1], alongside significant long-term health consequences [2]. Since pathophysiological changes precede clinical symptoms, early identification and intervention are critically important [3, 4]. Current risk stratification strategies for preeclampsia rely on clinical risk factors, maternal biomarkers and uterine artery Doppler velocimetry, but exhibit high false-positive rates and uncertain clinical validity [3]. These limitations stem largely from the complex and multifactorial pathophysiology of PE.

Over the past five decades, research has established that placental dysfunction initiates systemic vascular inflammation and endothelial injury [5, 6], pathophysiological pathways also shared with cardiovascular diseases [7]. This overlap has prompted investigation into cardiovascular-related biomarkers as a feasible approach for improving preeclampsia prediction.

Carbohydrate antigen 125 (CA125), also known as MUC16, is a glycoprotein in the mucin family that is released from epithelial cells [8, 9]. While widely recognized as a biomarker for ovarian cancer [10], CA125 is also physiologically expressed during pregnancy [11, 12]. It has been detected in multiple gestational tissues, including fetal coelomic epithelium, amniotic fluid, amniotic epithelium, and decidual tissues [13, 14]. Growing evidence supports the clinical utility of CA125 in cardiovascular diseases (CVDs), where it serves as a biomarker for monitoring both congestion and inflammatory processes [15–19]. Given its established role in CVD prediction, we hypothesized that CA125 might also have prognostic value for preeclampsia. However, current evidence linking CA125 to preeclampsia remains limited. Several studies have reported elevated CA125 levels in women with PE, though these measurements were primarily taken during the second or third trimester [20–24]. These findings are largely derived from cross-sectional studies with insufficient control for confounding variables. A recent meta-analysis further highlighted inherent selection biases in the available literature [25]. Consequently, more compelling evidence from well-designed prospective studies is needed to clarify this association.

In this prospective nested case-control study within the BiCoS cohort, we systematically evaluated the utility of CA125 as a predictive biomarker in PE and its relation with blood pressure changes patterns, providing new insights into PE pathogenesis and contributing to the development of more accurate risk stratification strategies.

Materials & methods

Ethical approval

This study utilized data from the Birth Cohort in Shenzhen (BiCoS), a prospective birth cohort study conducted by Shenzhen Maternity & Child Healthcare Hospital. The study protocol was approved by the Institutional Review Board of Shenzhen Maternity & Child Healthcare Hospital (Approval No. 23) and strictly adheres to the ethical principles of the Declaration of Helsinki. Written informed consent was obtained from all participants prior to enrollment.

Study design and participants

A nested case–control study was designed within the prospective BiCoS cohort. Cases were defined as participants diagnosed with preeclampsia (PE) according to the guidelines established by the American College of Obstetricians and Gynecologists [26]. Controls were healthy pregnant individuals randomly selected and matched to cases in a 1:1 ratio. PE was based on the following criteria: new-onset hypertension (SBP ≥ 140 mmHg or DBP ≥ 90 mmHg) occurring after 20 weeks of gestation, accompanied by proteinuria (≥ 0.3 g/24-hour urine collection or ≥ 1 + on dipstick testing). In the absence of proteinuria, PE was diagnosed when hypertension was accompanied by systemic complications such as, thrombocytopenia (PLT < 100 × 10⁹/L), hepatic impairment (serum transaminases ≥ 2 times the upper limit of normal), renal insufficiency (serum creatinine >1.1 mg/dL or doubling of baseline), pulmonary edema, or new-onset neurological or visual disturbances. In accordance with the contemporary guidelines from the International Society of Ultrasound in Obstetrics and Gynecology (ISUOG) [27] and the International Federation of Gynecology and Obstetrics (FIGO) [28], before 32 weeks of gestation, fetal growth restriction (FGR) was diagnosed if any of the following criteria were met: (1) estimated fetal weight (EFW) or abdominal circumference (AC) below the 3rd percentile for gestational age; or (2) absent end-diastolic flow in the umbilical artery; or (3) EFW or AC below the 10th percentile for gestational age, combined with either uterine artery pulsatility index (UtA-PI) or umbilical artery pulsatility index (UA-PI) above the 95th percentile for gestational age. After 32 weeks of gestation, FGR was diagnosed if any of the following criteria were met: (1) EFW or AC below the 3rd percentile for gestational age; or (2) at least two of the following four criteria were present: (a) EFW or AC below the 10th percentile; (b) a decrease in EFW or AC of more than two quartiles on serial scans; (c) cerebroplacental ratio (CPR) below the 5th percentile for gestational age; or (d) umbilical artery pulsatility index (UA-PI) above the 95th percentile for gestational age.

Exclusion criteria encompassed gestational hypertension, eclampsia, chronic hypertension, HELLP syndrome, and pre-existing comorbidities (diabetes, gestational diabetes, intrahepatic cholestasis, thyroid disorders, autoimmune diseases, cardiovascular disease, renal/hepatic dysfunction, acute/chronic infections) or women with incomplete clinical data. Acute infection was defined as the presence of clinically evident symptoms (e.g., fever > 38.0 °C, productive cough, dysuria, or waist pain) combined with supporting laboratory evidence. The laboratory criteria included: (1) elevated C-reactive protein (CRP) level > 10 mg/L, and/or (2) leukocytosis with a white blood cell count > 12 × 10⁹/L. Chronic infection was defined as a known history of untreated or active chronic infectious diseases, including but not limited to tuberculosis, hepatitis B or C (with signs of active replication or hepatic inflammation), HIV, or syphilis, as confirmed by serological testing and/or clinical diagnosis. Controls were selected using risk-set sampling, matched to cases by age (± 2 years), pre-pregnancy BMI (± 2 kg/m²), and gestational week at sampling (± 1 week). Normotensive status was confirmed through serial blood pressure measurements conducted across all trimesters and postpartum assessments. Finally, 70 PE cases and 70 matched controls were included (Fig. 1).

Fig. 1.

Flowchart of patient selection

Sample and data collection

Peripheral blood samples were collected from participants using 9 mL vacuum tubes containing procoagulant. After centrifugation, serum was aliquoted and stored at − 80℃ until analysis. Samples obtained between 8 and 14 weeks of gestation were included in this study.

Blood pressure measurements were taken at each prenatal visit. Participants rested in a seated position for at least 5 min before measurement. An appropriate sized cuff was selected based on arm circumference. Three consecutive readings were taken at 2-minute intervals, and the average values were used for analysis. Mean arterial pressure (MAP) was calculated as follows: MAP = DBP + ⅓(SBP − DBP).

The timing for both blood pressure and sample collection was defined as gestational age in weeks from the last menstrual period, confirmed and corrected by first-trimester ultrasound where necessary. This approach standardized all measurements to a common clinical timeline from conception. Demographic and clinical data, including maternal age, education level (categorized as ≤ 12, 13–15, or ≥ 16 years), parity (primipara or multipara), height, prepregnancy weight, and prepregnancy BMI, were retrieved from the hospital information system. Hepatic function was assessed using serum biomarkers: aspartate aminotransferase (AST), alanine aminotransferase (ALT), albumin (ALB), total bilirubin (TBIL), and direct bilirubin (DBIL). Renal function was evaluated via serum creatinine (Cr), urea, and the urinary albumin-to-creatinine ratio (UACR). Lipid profiles included triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). Hematologic parameters, including platelet count (PLT), were also analyzed.

Serum CA125 measurement

Serum CA125 levels were quantified using a commercial ELISA kit (R&D Systems, Catalog Number: DCA125, Minneapolis, USA) in strict accordance with the manufacturer’s instructions. To ensure assay accuracy and monitor performance across all runs, we implemented a rigorous quality control procedure. Three levels of commercial QC samples (low, medium, and high concentrations, approximately 1 U/mL, 8 U/mL, and 25 U/mL, respectively) were included on every assay plate. The measured values for all QC samples consistently fell within the acceptable recovery range of 85% to 115% of their expected values throughout the study. The minimum detectable dose (MDD) of the assay was 0.035 U/mL, with a linear range of 0.5–32 U/mL. Intra- and inter-assay coefficients of variation (CV) were < 1.5% and < 7.4%, respectively. All measurements were performed by technicians blinded to the clinical status of the participants. Optical density was read at 450 nm using a microplate reader.

Statistical analysis

Continuous variables are presented as mean ± standard deviation (SD), while categorical variables are expressed as frequencies (percentages). Group comparisons were made using the chi-square test for categorical variables and Student’s t-test for continuous variables, as appropriate. For multiple comparisons, Bonferroni correction was applied for the comparison of PE and normal controls, as well as for PE subgroup analyses. A post-hoc power analysis was conducted using G*Power (version 3.1.9.7) to assess the statistical power of the study based on the observed effect sizes (Cohen’s d), a significance level (α) of 0.05, and the actual sample sizes of the groups. A power value greater than 80% was considered adequate.

Univariate and multivariate binary logistic regression analyses were conducted to predict PE. The modeling strategy was as follows: a univariate model featuring serum CA125 alone (Model 1); a multivariate model adjusted for established clinical risk factors, including maternal age, prepregnancy BMI, primiparity, and MAP (Model 2); and a final combined multivariate model that integrated CA125 with all the clinical covariates from Model 2 (Model 3). The associations are presented as odds ratios (ORs) with their corresponding 95% confidence intervals (CIs). Interaction terms of CA125 with parity and gestational week at sampling were evaluated. The discriminatory ability of each model was evaluated using the area under the receiver operating characteristic curve (AUC) with 95% confidence intervals (CI). Pairwise comparisons of the AUCs were performed using the DeLong test. The optimal cutoff was determined by maximizing Youden’s index (sensitivity + specificity − 1). Key clinical performance metrics, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were computed.

Internal validation was performed to assess model robustness and correct for overfitting. First, 10-fold cross-validation was conducted, reporting the mean AUC across all folds. Subsequently, bootstrap resampling with 1000 iterations was used to calculate the optimism-corrected AUC.

Pearson correlation analysis was employed to examine the relationships between CA125 and blood pressure, platelet count and hepatic/renal function parameters. To account for multiple comparisons in the correlation analyses between CA-125 and 16 clinical factors, we applied the Benjamini-Hochberg false discovery rate (FDR) correction to all P-values. Statistical significance was defined as an FDR-adjusted P-value (Q-value) < 0.05.

Given the longitudinal design of this study, where there is natural variation in the gestational weeks at follow-up measurements and the times of blood pressure measurements per participant, longitudinal relationships between first-trimester CA125 and blood pressure trajectories were analyzed using linear mixed-effects models with an unstructured covariance matrix, fitted via restricted maximum likelihood (REML). This study incorporated intensive longitudinal blood pressure measurements (mean 10.02 measurements per participant, Table S1), thereby providing sufficient temporal resolution to reliably model blood pressure trajectories throughout the entire pregnancy.

A two-sided P < 0.05 was considered statistically significant. Data visualization and basic statistical analyses were performed using GraphPad Prism 7.0 (GraphPad Software, San Diego, CA, USA). Advanced modeling was conducted in R (version 4.4.2, R Foundation for Statistical Computing, Vienna, Austria).

Results

Study population characteristics

As shown in Table 1, the two groups were compared in maternal age (31.47 ± 4.77 vs. 31.96 ± 4.44 years, P = 0.534), prepregnancy BMI (21.84 ± 2.55 vs. 21.01 ± 2.80 kg/m², P = 0.069), and lifestyle factors including smoking (4.29% vs. 2.86%, P = 1.000) and alcohol consumption (5.71% vs. 2.86%, P = 0.681). PE cases exhibited significantly higher first-trimester SBP (117.17 ± 8.61 vs. 110.21 ± 9.42 mmHg, P < 0.001), DBP (70.11 ± 7.93 vs. 64.77 ± 8.41 mmHg, P < 0.001), and mean arterial pressure (85.80. ± 7.33 vs. 79.92 ± 8.26 mmHg, P < 0.001). Additionally, the PE group had a significantly higher proportion of primiparous women (72.86% vs. 42.86%, P < 0.001). No significant differences were observed in education level, family history of hypertension or diabetes, or gestational age at sampling (all P > 0.05).

Table 1.

Baseline characteristics of the study population

Characteristics	PE (n = 70)	Control (n = 70)	P value
Maternal age	31.47 ± 4.77	31.96 ± 4.44	0.534
Prepregnancy BMI	21.84 ± 2.55	21.01 ± 2.80	0.069
Education (≥ 15 years)	48 (68.57)	57 (81.43)	0.079
Family history of hypertension	5 (7.14)	1 (1.43)	0.209
Family history of diabetes	4 (5.71)	0 (0.00)	0.120
Smoking	3 (4.29)	2 (2.86)	1.000
Alcohol consumption	4 (5.71)	2 (2.86)	0.681
Parity (Primipara)	51 (72.86)	30 (42.86)	< 0.001***
Gestational week at sampling (wks)	12.26 ± 0.44	12.26 ± 0.92	1.000
First-trimester BP (mmHg)
SBP	117.17 ± 8.61	110.21 ± 9.42	< 0.001***
DBP	70.11 ± 7.93	64.77 ± 8.41	< 0.001***
MAP	85.80 ± 7.33	79.92 ± 8.26	< 0.001***

Data are presented as mean ± SD or number (percentage), as appropriate. Independent t-tests were used for continuous variables. chi-square or Fisher’s exact tests were used for categorical variables

Significant differences were observed in parity and first-trimester blood pressure parameters (SBP, DBP, MAP) between groups ***P < 0.001

Abbreviations: PE, preeclampsia; BMI, body mass index; BP, blood pressure; SBP, systolic blood pressure; DBP, diastolic blood pressure; MAP, mean arterial pressure

First-trimester CA125 levels in preeclampsia

Serum CA125 levels were significantly elevated in PE cases compared to controls (25.44 ± 14.89 U/mL vs. 16.17 ± 9.61 U/mL, P < 0.0001, Fig. 2A; Table 2), which remained statistically significant following adjustments for multiple comparisons via the Bonferroni method. A post-hoc power analysis based on the observed effect size (Cohen’s d = 0.739, Table 2) confirmed the robustness of this finding, achieving a statistical power of 99.1%. Notably, among PE cases, those without FGR (34.94 ± 22.93 U/mL vs. 23.27 ± 11.61 U/mL, P = 0.010, Bonferroni-corrected P = 0.040, Fig. 1D; Table 2). This comparison yielded a large effect size (Cohen’s d = 0.817) and an achieved power of 74.6%, indicating adequate sensitivity to detect this difference. In contrast, no significant associations were observed between light and severe PE or between early- and late-onset PE, which corresponded to negligible effect sizes and low statistical power (< 10%, Fig. 1B-C; Table 2).

Fig. 2.

First-Trimester Serum CA125 Levels in Women with Preeclampsia and Normal Controls. Independent t-tests were used for comparison of CA125 levels between two groups. (A) CA125 was significantly elevated in the PE group compared to controls. (B–C) No significant differences were found between light and severe PE or early- and late-onset PE subgroups. (D) PE complicated by FGR was associated with higher CA125 levels than PE without FGR. ****P < 0.0001, **P < 0.01, ns: not significant. Abbreviations: PE, preeclampsia; FGR fetal growth restriction

Table 2.

First-trimester CA125 levels in PE and normal controls

Group	CA125 levels(U/ml)	t	P value	Adjusted P value	Significance	Cohen’s d	POWER
First trimester
Control (n = 70)	16.17 ± 9.61	4.374	< 0.0001****	< 0.004***	Yes	0.739	99.1%
PE (n = 70)	25.44 ± 14.89
Severity
Light PE (n = 31)	26.39 ± 14.05	0.476	0.636	1.000	No	0.115	7.6%
Severe PE (n = 39)	24.68 ± 15.67
Onset
Early-onset PE (n = 13)	26.56 ± 19.11	0.298	0.766	1.000	No	0.092	6.0%
Late-onset PE (n = 57)	25.18 ± 13.95
with/ without FGR
With FGR (n = 13)	34.94 ± 22.93	2.659	0.010**	0.04*	Yes	0.817	74.6%
Without FGR (n = 57)	23.27 ± 11.61

CA125 levels are presented as mean ± SD (U/mL)

Group comparisons were performed using independent t-tests

P values were adjusted for multiple testing using the Bonferroni correction procedure for 4 multiple comparisons

Effect sizes are reported as Cohen’s d. Achieved power (1-β) was derived from a post-hoc power analysis using the observed effect size, α level of 0.05, and the actual group sample sizes

First-trimester CA125 levels were significantly elevated in women with PE and were particularly higher in those with coexisting fetal growth restriction, but showed no association with disease severity or time of onset

****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05

Abbreviations: PE, preeclampsia; FGR, fetal growth restriction

Logistic regression analysis and predictive performance of models for preeclampsia

To elucidate the independent and combined contributions of CA125 and established clinical risk factors to the risk of preeclampsia, three logistic regression models were constructed as detailed in Table 3. The univariate model with CA125 alone (Model 1) demonstrated a significant association with PE risk (OR = 1.077, 95% CI: 1.037–1.119, P < 0.001). Model 2, which incorporated clinical factors, identified maternal age (OR 1.139, 95% CI 1.011–1.282, P < 0.05), primiparity (OR 12.526, 95% CI 3.807–41.209, P < 0.0001), and MAP (OR 1.063, 95% CI 1.027–1.101, P < 0.0001) as significant independent predictors. Prepregnancy BMI was not statistically significant in this model. To test for potential effect modification, we included interaction terms between CA125 and both parity and gestational week at sampling in the combined model. The interaction terms between CA125 and both parity (OR = 0.974, 95% CI: 0.895–1.060, P = 0.537) and gestational age (OR = 1.001, 95% CI: 0.978–1.025, P = 0.911) were not statistically significant, indicating no interaction effect between CA125 and these variables, and therefore, the interaction term was not retained in the final combined model. The combined model (Model 3), which integrated CA125 with the clinical factors, demonstrated the robustness of these predictors. Crucially, CA125 remained a significant independent predictor after adjusting for all clinical confounders (OR 1.071, 95% CI 1.027–1.118, P < 0.001), alongside the persistent strong effects of parity (OR 12.947, 95% CI 3.685–45.485, P < 0.0001) and MAP (OR 1.069, 95% CI 1.030–1.110, P < 0.0001), while maternal age remained significant.

Table 3.

Logistic regression analysis of models for preeclampsia

Variable	Model 1 (CA125)	Model 2 (Clinical Factors)	Model 3 (Combined)
Maternal age	-	1.139 (1.011, 1.282)*	1.165 (1.028,1.321)*
Prepregnancy BMI	-	1.137 (0.960, 1.346)	1.105 (0.936,1.305)
Parity	-	12.526 (3.807,41.209)****	12.947 (3.685,45.485)****
MAP	-	1.063 (1.027,1.101)****	1.069 (1.03,1.11)****
CA125	1.077 (1.037,1.119)***	-	1.071 (1.027,1.118)***

The strength of association with variables and PE was assessed using logistic regresion analysis and is expressed as OR (95%CI)

Model 1 included only CA125. Model 2 included clinical risk factors. Model 3 (combined model) included CA125 and clinical factors

The results identified CA125, parity, MAP and maternal age as the most robust and significant independent predictors across models, while CA125 also remained a significant predictor in both Model 1 and the combined Model 3

-, Variable not included in the model

****P < 0.0001, ***P < 0.001, *P < 0.05

Abbreviations: BMI, body mass index; MAP, mean arterial pressure; OR, odds ratio; CI, confidence interval

In terms of discriminatory accuracy, ROC curve analysis was performed, the details of which are presented in Fig. 3 and Table S2. Model 1 showed a moderate discriminative capacity, with an AUC of 0.719 (95% CI: 0.634–0.803; P < 0.0001). The model based on clinical factors (Model 2) demonstrated a significantly higher predictive performance, with an AUC of 0.799 (95% CI: 0.724–0.873; P < 0.0001). Notably, the combined model (Model 3), which included both CA125 and clinical factors, achieved the highest discriminative power, with an AUC of 0.841 (95% CI: 0.776–0.906; P < 0.0001). The increase in AUC from Model 2 to Model 3 was 0.042, indicating a meaningful improvement in predictive accuracy, although this difference did not reach formal statistical significance in this cohort (P = 0.069). Correspondingly, the combined model also yielded the highest maximum Youden’s index (0.557), along with superior sensitivity (75.7%), specificity (81.4%), positive predictive value (76.47%), and negative predictive value (75.0%) compared to the other models. To rigorously assess the combined model performance and adjust for overfitting, we conducted internal validation using 10-fold cross-validation and bootstrapping (Fig. 4, Table S3). The apparent AUC of the combined model on the entire development cohort was 0.841. However, 10-fold cross-validation yielded a mean AUC of 0.828, indicating a robust and generalizable discriminative ability. Furthermore, bootstrap validation with 1000 iterations was performed to calculate the optimism-corrected AUC. The estimated optimism was 0.024, resulting in an optimism-corrected AUC of 0.817. This represents a performance retention of 97.1%, suggesting good performance after accounting for overfitting.

Fig. 3.

Receiver Operating Characteristic (ROC) Curves for Predicting Preeclampsia. ROC curves comparing the predictive performance of three models for preeclampsia: Model 1 (Red line, CA125 alone, AUC = 0.719, 95% CI: 0.634–0.803), Model 2 (Green line, clinical factors, AUC = 0.799, 95% CI:0.724–0.873), and Model 3 (Blue line, combined model of CA125 and clinical factors, AUC = 0.841, 95% CI:0.776–0.906). The dashed grey line represents the reference line of no discriminative ability (AUC = 0.5). Abbreviations: AUC, area under the curve; CI, confidence interval

Fig. 4.

Internal Validation Performance of the Selected Model. The apparent AUC represents performance on the training data, while 10-fold cross-validation and bootstrap correction provide estimates of expected performance on new data. The combined prediction model demonstrated excellent performance retention upon internal validation. The apparent AUC was 0.841. The mean AUC from 10-fold cross-validation was 0.828. Bootstrap validation revealed minimal optimism (0.024), yielding an optimism-corrected AUC of 0.817, corresponding to 97.1% performance retention

Associations of first-trimester CA125 with clinical parameters

The correlation analysis between various clinical parameters and the outcome variable is summarized in Figs. 5, 6 and 7 and Table S4. Among all the parameters assessed, systolic blood pressure (SBP) demonstrated a statistically significant but weak positive univariate correlation (r = 0.179, R² = 0.032, p = 0.034). However, this association did not remain significant after adjusting for multiple comparisons using the False Discovery Rate (FDR) method (Q value = 0.525). No significant correlations were observed between CA125 and hepatic function markers (ALT, AST, TBIL, DBIL, ALB), renal function parameters (Cr, UREA, UACR), platelet counts, metabolic profiles (TG, TC, LDL-C, HDL-C) or BP (DBP and MAP) either before or after FDR correction (all P > 0.05, all FDR Q value > 0.05).

Fig. 5.

Correlation Between First-Trimester CA125 and Hepatic/Renal Function Profiles. Pearson correlation analysis was performed to assess the associations between CA125 and hepatic as well as renal function markers (ALT, AST, TBIL, DBIL, ALB, Cr, Urea and UACR). No significant correlations were observed between CA125 and any hepatic or renal markers. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; TBIL, total bilirubin; DBIL, direct bilirubin; ALB, albumin, Cr, creatinine; UACR, urinary albumin-to-creatinine ratio

Fig. 6.

Correlation Between First-Trimester CA125 and Lipid and Hematologic Profiles. Pearson correlation analysis was performed to assess the associations between CA125 and lipid profiles (TG, TC, HDL-C, LDL-C) as well as PLT. No significant correlations were observed between CA125 and any lipid parameter or PLT. Abbreviations: TG, triglycerides; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; PLT, platelet

Fig. 7.

Correlation Between First-Trimester CA125 and Blood Pressure. Pearson correlation analysis was performed to assess the associations between CA125 and blood pressure (SBP, DBP and MAP). A significant positive correlation was initially found between CA125 and SBP, yet it did not persist after adjustment (Table S4). Additionally, no significant correlations were observed between CA125 and DBP or MAP at any point. *P < 0.05. Abbreviations: SBP, systolic blood pressure; DBP, diastolic blood pressure; MAP, mean arterial pressure

Longitudinal analysis of CA125 and systolic blood pressure trajectories

Longitudinal Analysis of CA125 and Systolic Blood Pressure Trajectories Linear mixed-effects models revealed a significant association between first-trimester CA125 levels and systolic blood pressure (SBP) trajectories throughout pregnancy (Table 4). The base model indicated each gestational week was associated with a 0.167 mmHg increase in SBP (SE = 0.024, P < 0.001). After adding CA125, each 1-SD increase was associated with a 0.153 mmHg elevation in SBP (SE = 0.054, P = 0.005), reducing between-individual variance by 5.3%. After full adjustment for covariates, CA125 remained significantly associated with SBP (β = 0.125, SE = 0.055, P = 0.024), while prepregnancy BMI showed a stronger independent association (β = 0.727, SE = 0.277, P = 0.009). A significant interaction was observed between CA125 and gestational week (β = 0.005, SE = 0.002, P = 0.015), indicating that women with higher CA125 levels experienced a more rapid increase in SBP (0.226 vs. 0.109 mmHg/week). Model comparisons confirmed that incorporating CA125 significantly improved model fit (χ² = 7.807, P = 0.005), with the interaction model demonstrating the best fit (AIC = 10231, BIC = 10294; Table S5).

Table 4.

Longitudinal associations between first-trimester CA125 levels and systolic blood pressure dynamics throughout gestation: results from multilevel mixed-effects models

Variable	Model 1 (Base)	Model 2 (+ CA125)	Model 3 (+ Covariates)	Model 4 (+ Interaction)	Trend
Fixed Effects
Intercept	111.300 (0.985)***	108.100 (1.495)***	85.930 (14.673)***	88.070 (14.69)***	↘↗
Gestational week	0.167 (0.024)***	0.168 (0.024)***	0.168 (0.024)***	0.074 (0.046)	↘
CA125	-	0.153 (0.054)**	0.125 (0.055)*	0.005 (0.075)	↘
Gestational week × CA125	-	-	-	0.005 (0.002)*	-
Maternal age	-	-	0.226 (0.195)	0.228 (0.195)	↗
Smoking	-	-	8.290 (5.530)	8.273 (5.526)
Alcohol use	-	-	-1.664 (5.100)	-1.765 (5.096)	↘
Prepregnancy BMI	-	-	0.727 (0.277)**	0.725 (0.277)**	↘
Parity	-	-	-3.150 (1.659)	-3.148 (1.658)	↗
Blood collection week	-	-	0.099 (1.000)	0.123(0.999)	↗
Random Effects​​
Between-individual variance (σ²)	69.14	65.50	62.62	62.54	↘
Between-individual SD	8.315	8.093	7.913	7.908	↘
Residual variance (σ²)	67.83	67.84	67.84	67.60	↗↘
Residual SD	8.236	8.236	8.236	8.222	↘
ICC	0.505	0.491	0.480	0.481	↘

Fixed effects are expressed as β coefficients (standard error) representing change in SBP (mmHg) per unit increase

Model 1 (Base): unadjusted effect of gestational week. Model 2 (+ CA125): Adds the fixed effect of CA125. Model 3 (+ Covariates): Further adjustment for maternal age, smoking, alcohol use, prepregnancy BMI, parity, and blood collection week. Model 4 (+ Interaction): inclusion of CA125 × gestational week interaction. All models included 1,403 observations from 140 participants. Elevated first-trimester CA125 levels were significantly associated with an accelerated increase in systolic blood pressure throughout pregnancy, and this association remained independent of confounding factors. Trend arrows: ↗= Increasing trend across models,↘ = Decreasing trend across models,→ = Stable/no clear trend ***P < 0.001, **P < 0.01, *P < 0.05

Abbreviations: ICC, Intraclass correlation coefficient (proportion of variance due to individual-level clustering)

Longitudinal analysis of CA125 and diastolic blood pressure trajectories

Mixed-effects models also indicated a significant longitudinal association between CA125 levels and diastolic blood pressure (DBP) trajectories (Table 5). The base model showed that each gestational week was associated with a 0.129 mmHg increase in DBP (SE = 0.018, P < 0.001). After incorporating CA125, each 1-SD increase was associated with a 0.120 mmHg elevation in DBP (SE = 0.045, P = 0.009), reducing between-individual variance by 4.5%. This association remained significant after full adjustment for covariates (β = 0.096, SE = 0.045, P = 0.034). Prepregnancy BMI was independently associated with DBP (β = 0.811, SE = 0.226, P < 0.001). A significant interaction between CA125 and gestational week was observed (β = 0.004, SE = 0.001, P = 0.009), indicating a steeper increase in DBP among women with higher CA125 levels (0.085 mmHg/week vs. 0.019 mmHg/week in the low-CA125 group). Model fit was significantly improved with the inclusion of CA125 (χ² = 7.024, P = 0.008), and the interaction model showed the best fit (AIC = 9485, BIC = 9548; Table S6).

Table 5.

Longitudinal associations between first-trimester CA125 levels and diastolic blood pressure dynamics throughout gestation: results from multilevel mixed-effects models

Variable	Model 1 (Base)	Model 2 (+ CA125)	Model 3 (+ Covariates)	Model 4 (+ Interaction)	Trend
Fixed Effects
Intercept	64.860 (0.773)***	62.370 (1.210)***	50.706 (11.950)***	51.900 (11.970)***	↗
Gestational week	0.129 (0.018)***	0.129 (0.018)***	0.130 (0.018)***	0.052 (0.035)	↗↘
CA125	-	0.120(0.045)**	0.096(0.045)*	-0.000(0.006)	↘
Gestational week × CA125	-	-	-	0.003 (0.001)**	-
Maternal age	-	-	0.079(0.159)	0.083(0.159)	↗
Smoking	-	-	3.994(4.507)	3.927(4.512)	↘
Alcohol use	-	-	-1.827(4.158)	-1.810(4.162)	↗
Prepregnancy BMI	-	-	0.811(0.226)***	0.809(0.214)***	↘
Parity	-	-	-2.022(1.352)	-2.015(1.354)	↗
Blood collection week	-	-	-0.565(0.815)	-0.505(0.816)	↗
Random Effects
Between-individual variance (σ²)	47.28	45.13	42.16	42.27	↘↗
Between-individual SD	6.876	6.718	6.493	6.502	↘↗
Residual variance (σ²)	39.38	39.38	39.38	39.18	↘
Residual SD	6.276	6.276	6.275	6.260	↘
ICC	0.546	0.534	0.517	0.519	↘↗

Fixed effects are expressed as β coefficients (standard error) representing change in SBP (mmHg) per unit increase

Model 1 (Base): unadjusted effect of gestational week. Model 2 (+ CA125): Adds the fixed effect of CA125. Model 3 (+ Covariates): Further adjustment for maternal age, smoking, alcohol use, prepregnancy BMI, parity, and blood collection week. Model 4 (+ Interaction): inclusion of CA125 × gestational week interaction. All models included 1,403 observations from 140 participants. Elevated first-trimester CA125 levels were significantly associated with an accelerated increase in diastolic blood pressure during pregnancy, independent of the covariates. Trend arrows: ↗= Increasing trend across models,↘ = Decreasing trend across models,→ = Stable/no clear trend. ***P < 0.001, **P < 0.01, *P < 0.05

Abbreviations: ICC, Intraclass correlation coefficient (proportion of variance due to individual-level clustering

Discussion

PE seriously threatens the safety of both mothers and infants, underscoring the critical need for early prediction and timely intervention [29, 30]. Pregnancy has long-lasting effects, with molecular rewiring taking place at very early stages, prior to the manifestation of complications at a clinically detectable level [31]. Numerous maternal serum biomarkers have been investigated for PE prediction [32], among which placental growth factor (PlGF) and soluble fms-like tyrosine kinase-1 (sFlt-1) are the most extensively validated [33]. The sFlt-1/PlGF ratio has proven useful as an auxiliary diagnostic tool for monitoring placental function and short-term disease progression [34]. Although PlGF shows some predictive potential in the first trimester, its sensitivity as a standalone marker is limited [35], and sFlt-1 levels typically rise only after 21–24 weeks of gestation [36]. Effective predictive markers for early stages still need to be identified.

To our knowledge, this is the first study to integrate prospective biomarker assessment, longitudinal hemodynamic profiling, and advanced statistical modeling to elucidate the role of CA125 in PE prediction and pathogenesis. We provide confirmation of CA125’s potential as a circulating biomarker for PE, providing a critical earlier window for risk identification and possible preventive strategies. As a glycoprotein secreted from Müllerian duct derivatives, including fetal membranes and decidua, CA125 may play a significant role in placental development and spiral artery remodeling [13]. Elevated CA125 levels may reflect a more active or stressed state of vessel remodeling during placental implantation. This hypothesis is strongly supported by one of our key findings that PE patients complicated by FGR exhibited significantly higher CA125 levels than those without FGR. Given that FGR is typically associated with more severe placental dysfunction and morphological abnormalities [37], this result suggests that elevated CA125 may be linked to the intrinsic severity of placental injury. The absence of significant differences in CA125 levels across PE severity or onset-time subgroups may be attributed to overlapping placental pathogenesis among these phenotypes, or limited statistical power in our current sample size to detect such subgroup differences.

While first-trimester CA125 independently predicts preeclampsia, its sensitivity and specificity as a standalone marker remain moderate, possibly due to limited sample size. Currently, two principal approaches, including PE risk factor-based screening and the multi-marker algorithm developed by the Fetal Medicine Foundation, are internationally employed for early prediction [1, 3, 38]. The core contribution of this study lies in establishing the independent predictive value of CA125 through multivariable logistic regression and comprehensive model comparisons. Importantly, the association between CA125 and PE remained statistically significant even after adjusting for potent clinical factors, including parity and MAP which is consistent with extensive previous research [38], underscoring its capacity to provide incremental predictive information. The resulting integrated model demonstrated excellent discriminatory ability and maintained robust performance upon internal bootstrap validation, supporting its reliability and potential generalizability. Although the incremental improvement in AUC over the clinical-only model did not reach statistical significance, the observed enhancements in sensitivity, specificity, and other comprehensive metrics, coupled with a high performance retention rate of 97.1%. These findings indicate that future risk stratification tools should not rely solely on demographic and clinical parameters. Instead, the integration of circulating biomarkers such as CA125 holds promise for enabling earlier and more accurate risk identification. In the prediction model, we found that parity is a powerful influencing factor. To determine whether the association between first-trimester CA125 levels and PE was modified by parity, interaction analysis was performed. The results showed showed no significant interaction, suggesting that the relationship between CA125 levels and PE risk was consistent in both primipara and multiparous. Thus, parity did not appear to amplify or diminish the predictive utility of CA125.

Of particular importance, our correlation analysis revealed the specificity of CA125. The absence of significant associations between CA125 and a range of first-trimester hepatic, renal, metabolic, PLT and blood pressure, contrasting with reports from the second and third trimester of pregnancy [39, 40], coupled with its marked CA125 elevation in PE complicated with FGR, suggests aplacental rather than a secondary systemic origin. It appears to signal a more placenta-specific pathophysiological mechanism, strengthening its rationale as a disease-specific biomarker for PE.

Furthermore, the most innovative finding of this study comes from the longitudinal analysis. We first employed mixed-effects models to reveal an association between first-trimester CA125 levels and dynamic BP trajectories throughout pregnancy. Not only was the main effect of CA125 statistically significant, but its interaction with gestational week also demonstrated both statistical and clinical relevance. This indicates that pregnant women with elevated CA125 levels exhibit a significantly steeper increase (both ABP and DBP) in blood pressure as gestation progresses.

Through cross-sectional correlation analysis and longitudinal analysis, we were able to obtain a pivotal and nuanced finding that while first-trimester maternal CA125 levels demonstrated no immediate correlation with concurrent blood pressure, they exhibited a significant positive association with the subsequent trajectory of blood pressure increase as pregnancy progressed. This dissociation suggests that CA125 operates as an early harbinger of a pathological cascade that unfolds dynamically as pregnancy progresses. The compelling temporal evidence substantiates a central tenet of preeclampsia pathogenesis that the disease follows a prolonged subclinical course, with placental injury often predating the onset of maternal clinical symptoms [5]. The aberrant placental dysfunction, initiated as early as the first trimester and reflected by elevated CA125 levels, drives progressive maladaptation of the maternal cardiovascular system through continuous release of factors, ultimately manifesting as preeclampsia.

Within this framework, the critical and currently missing link is the direct longitudinal association between CA125 and the key mediators of PE-related endothelial dysfunction, specifically sFlt-1 and PlGF. We hypothesize that the same early placental stress signaled by CA125 later manifests as a dysregulated angiogenic profile. It is plausible that high first-trimester CA125 identifies a subgroup of women destined for a more pronounced “angiogenic switch,” characterized by a steeper rise in sFlt-1 and a steeper decline in PlGF in the second and third trimesters. Our study, while demonstrating the clinical phenotype (BP) of this association, lacks the serial angiogenic biomarker measurements to directly test this hypothesis. Future research must prioritize cohorts with serial biosampling to establish whether CA125 effectively predicts the trajectory of the sFlt-1/PlGF ratio, thereby solidifying its role as an early sentinel within the angiogenic cascade.

Beyond obstetrics, CA125 has emerged as a valuable biomarker for risk stratification in cardiovascular disease [15]. For instance, Núñez et al. demonstrated that CA125 independently predicts adverse outcomes in heart failure, outperforming established risk models and even natriuretic peptides [16]. Collectively, these findings suggest that CA125 may play an active role in cardiovascular function regulation. Mechanistically, CA125 is a high-molecular-weight glycoprotein implicated in fluid homeostasis, inflammation, and tissue repair [41]. Its recognized role as a nexus between congestion and inflammation in heart failure [42] suggests it may function as a key ligand that amplifies inflammatory signaling [43]. Furthermore, its immunomodulatory properties, such as inhibiting NK cell activity and modulating galectin signaling [44], support its potential involvement in the angiogenic imbalance and inflammatory activation central to PE pathogenesis. By analogy, CA125 may contribute to the hemodynamic deterioration in preeclampsia by functioning as a decoy receptor for pro-angiogenic ligands like galectin-1 [44]. This mechanism potentially disrupts angiogenic balance by simultaneously sequestering beneficial factors and amplifying anti-angiogenic signals. However, this proposed mechanism remains speculative, and its precise role in PE awaits direct experimental validation.

Regarding clinical translation, cost-benefit analyses can help guide screening and treatment protocols [45]. CA125 possesses distinct practical advantages over many novel biomarkers still in experimental stages. Its assays are standardized, cost-effective, and widely available due to decades of use in oncology. This makes it particularly suitable for resource-limited settings where advanced angiogenic panels remain inaccessible. A CA125-based screening strategy could improve equity in PE risk stratification and facilitate timely initiation of prophylactic interventions, such as aspirin before 16 weeks of gestation.

Several limitations should be acknowledged. This nested case-control study, though mitigated by rigorous post-hoc power analysis, statistical adjustment and internal validation, requires validation in larger, prospective, multi-center cohorts [46] to enhance generalizability across diverse populations and preeclampsia subtypes. Furthermore, CA125 was measured at a single timepoint, precluding insight into its dynamic changes throughout gestation. Most critically, as noted above, the absence of serial angiogenic factor measurements limits our ability to delineate the precise mechanistic pathway linking CA125 to hypertension.

Future studies should incorporate serial CA125 measurements combined with longitudinal angiogenic profiling to model the temporal relationships among placental markers, anti-angiogenic imbalance, and hemodynamic changes. In addition, multi-omics approaches and experimental models are needed to elucidate the cellular sources and regulatory stimuli of CA125 release in PE placentas. Ultimately, such efforts may help explore the clinical utility of CA125 in guiding targeted preventive strategies, perhaps by identifying a high-risk population that might benefit from closer monitoring or early intervention before the overt angiogenic imbalance and hypertension develop.

Conclusions

In summary, this study reframes first-trimester serum CA125 from a static risk marker into a dynamic predictor of pathological pregnancy adaptation, offering dual utility in both risk prediction and pathophysiological insight. We demonstrated that an elevated CA125 level serves as a robust and independent predictor of PE, particularly in cases complicated by FGR, underscoring its association with the severity of placental dysfunction. The value of CA125 was further highlighted by its ability to significantly enhance the predictive accuracy of a model based on established clinical risk factors.

Critically, the cross-sectional and longitudinal associations provide a new framework for understanding its biological role. The absence of correlation with baseline clinical parameters confirms its specificity as a marker of placental origin, whereas its strong association with accelerated blood pressure trajectories throughout gestation positions CA125 as an early sentinel within the pathological cascade of PE. This quantifiable link between a first-trimester placental signal and subsequent hemodynamic progression provides a crucial bridge between early placental insult and the clinical manifestation of the syndrome. Future research focusing on the temporal relationship between CA125 and key angiogenic pathways will be essential to elucidate the underlying mechanistic pathway and to solidify the potential of CA125 as a cornerstone for early prediction and prevention of preeclampsia.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

PE

Preeclampsia

FGR

Fetal Growth Restriction

CA125

Carbohydrate Antigen 125

SBP

Systolic Blood Pressure

DBP

Diastolic Blood Pressure

MAP

Mean Arterial Pressure

BMI

Body Mass Index

AUC

Area Under the Curve

OR

Odds Ratio

aOR

Adjusted Odds Ratio

CI

Confidence Interval

SD

Standard Deviation

SE

Standard Error

ROC

Receiver Operating Characteristic

LMM

Linear Mixed-effects Model

ICC

Intraclass Correlation Coefficient

ALT

Alanine Aminotransferase

AST

Aspartate Aminotransferase

TBIL

Total Bilirubin

DBIL

Direct Bilirubin

ALB

Albumin

Cr

Creatinine

UACR

Urinary Albumin-to-Creatinine Ratio

TG

Triglycerides

TC

Total Cholesterol

HDL-C

High-Density Lipoprotein Cholesterol

LDL-C

Low-Density Lipoprotein Cholesterol

PLT

Platelet Count

BiCoS

Birth Cohort in Shenzhen

ELISA

Enzyme-Linked Immunosorbent Assay

OD

Optical Density

AIC

Akaike Information Criterion

BIC

Bayesian Information Criterion

PPV

Positive Predictive Value

Author contributions

Each author contributed to this paper. Conceptualization: Xiaoxia Wu, Linlin Wu and Jianmin Niu; Methodology: Aiqi Yin and Yixuan Chen; Data collection: Xiaonian Guan, Pingping Huang, Cuiping Zeng and Kan Liu; Software: Aiqi Yin and Huafan Zhang; Validation: Aiqi Yin and Yixuan Chen; Formal analysis: Aiqi Yin and Yixuan Chen; Writing—original draft preparation: Aiqi Yin; Writing—review and editing, Linlin Wu; Supervision: Jianmin Niu.

Funding

This work was supported by the National Natural Science Foundation of China (72374227), Shenzhen Science and Technology Program (JCYJ20220818103608017, JCYJ20220818103607015) and President Foundation of The Third Affiliated Hospital of Southern Medical University (YQ202410).

Data availability

Data will be made available upon reasonable request.

Declarations

Ethics approval and consent to participate

This study utilized data from the Birth Cohort in Shenzhen (BiCoS), a prospective birth cohort study conducted by Shenzhen Maternity & Child Healthcare Hospital. The study protocol was approved by the Institutional Review Board of Shenzhen Maternity & Child Healthcare Hospital (Approval No. 23) and strictly adhered to the ethical principles of the Declaration of Helsinki. Written informed consent was obtained from all participants prior to enrollment.

Consent for publication

Not applicable.

Conflict of interest

The authors declare no conflicts of interest.