This observational retrospective cohort study was conducted at Fuwai Hospital, Chinese Academy of Medical Sciences (Beijing, China). We screened all patients with idiopathic pulmonary arterial hypertension (IPAH) and chronic thromboembolic pulmonary hypertension (CTEPH) who underwent right heart catheterization (RHC) from January 1, 2014, to December 31, 2018. Patients with CA 125 data and multiple clinical visit/hospitalization records were enrolled as long as they had a minimum of 1 year of follow-up data for outcomes. In addition, echocardiography-suspected PH patients with normal invasive pulmonary arterial pressure and CA 125 data were also recruited as the control group. The establishment of IPAH and CTEPH was based on the 2009 (before January 2016) or 2015 European Society of Cardiology/European Respiratory Society (ERS) guidelines (18, 19). Normal pulmonary arterial pressure was defined as the mean pulmonary arterial pressure (mPAP) <25 mm Hg (18, 19). By design, patients were excluded if they had (1) any malignancy, (2) inflammatory diseases, or (3) active infection. The following clinical data were collected via an electronic medical record system by two independent reviewers: demographics, etiology of PH, 6-minute walk distance (6MWD), N-terminal pro-brain natriuretic peptide (NT-proBNP) levels, smoking history, alcohol consumption, World Health Organization functional class (WHO-FC), PH-specific medication, history of balloon pulmonary angioplasty/pulmonary endarterectomy, comorbidities, parameters derived from echocardiography and RHC, serum CA 125 levels, and follow-up data. The study protocol was approved by the Ethics Committee of Fuwai Hospital. Written informed consent was obtained from each patient.

Fasting venous blood samples were collected for CA 125 measurement on the first day of admission. Serum levels of CA125 were measured using a chemiluminescent microparticle immunoassay (product name: Access OV Monitor; Cat. No. 386357; Beckman Coulter Inc., Brea, CA, USA). Please refer to the manufacturer's website for the detailed methodology of the Access OV Monitor (https://mms.mckesson.com/product/586335/Beckman-Coulter-386357). The upper limit of normal for CA 125 was 35 U/ml with the Access OV Monitor. Accordingly, the included patients were divided into either the CA 125 > 35 U/ml group or the CA 125 ≤ 35 U/ml group.

The detailed protocol for RHC has been provided in our previous publications (20–23). Briefly, with local anesthesia under continuous electrocardiographic monitoring, a 6 French pigtail catheter or 7 French Swan-Ganz catheter (Edwards Lifesciences World Trade Co., Ltd, Irvine, CA, USA) was advanced into the pulmonary artery through the right femoral vein or right internal jugular vein by placement of a 6 or 7 French vascular sheath. Correct catheter positioning was verified by fluoroscopy. Transducers were positioned at the midaxillary line and zeroed at atmospheric pressure. Transthoracic echocardiography was performed by experienced ultrasonologists in the Department of Echocardiography under the current guidelines (24).

We considered the cumulative 1-year clinical worsening-free survival rate as the primary endpoint. Clinical worsening was defined as the occurrence of any of the following events: deteriorated WHO-FC, escalation of PH-specific therapy and rehospitalization due to heart failure or progression of PH. End-point events were adjudicated by two senior clinicians. Any discordance was resolved by the supervisors (QL and ZHL).

Continuous variables are presented as the mean ± standard deviation. Categorical variables are given as counts. Comparisons between two groups were performed using an independent-sample t-test, the Mann–Whitney U-test or the chi-square test, as appropriate. Correlations between CA 125 and other variables were examined using the Spearman correlation coefficient. To adjust for potential confounding factors, associations with P < 0.100 were further assessed using multivariate linear regression analysis (enter method).

The Kaplan–Meier method was used to assess differences in the rate of 1-year clinical worsening events between patients with values above or below 35 U/ml; curves were compared with the log-rank test. The association between serum CA 125 levels and clinical worsening events was evaluated by a Cox proportional hazards model. Univariate Cox analysis was first performed to screen all prognostic factors. Variables with clinical significance or P < 0.100 in univariate analysis were selected for multivariate Cox analysis (enter method). We tested the Cox proportional hazards assumption for each covariate using Schoenfeld residuals. The linearity assumption for CA 125 was evaluated by restricted cubic splines with four knots. Collinearity diagnostics were examined for the potential presence of collinearity between independent variables in multivariate linear regression analysis and multivariate Cox analysis. Receiver operating characteristic (ROC) curve analysis was performed to assess the predictive performance of CA 125. Internal validation was performed using 500 bootstrap resamples (25, 26).

Values of CA 125 and NT-proBNP were logarithmically transformed (ln) and then used in correlation analysis, linear regression and the Cox proportional hazards model. No single missing value was replaced. A two-sided P < 0.05 was considered indicative of statistical significance. Data analysis was performed using SPSS (version 23.0), R-studio (version 1.4.1106), R (version 4.0.5), and MedCalc (version 19.7.2).

We identified 231 (45.9%) eligible records for IPAH/CTEPH patients from the 503 records assessed; of the patients, 164 were IPAH and 67 were CTEPH. Furthermore, 84 patients with normal invasive pulmonary arterial pressure and CA 125 data were included as controls. A flow chart displaying the enrolment process is shown in Figure 1.

The baseline characteristics of all included patients are presented in Table 1. Among 231 patients with PH, 191 were categorized into the CA 125 ≤ 35 U/ml group and 40 into the CA 125 > 35 U/ml group. At baseline, 111 (48.1%) of 231 patients presented with WHO-FC III/IV, and 40 (17.3%) patients did not receive PH-specific medication. During the follow-up period, 73 (31.6%) patients experienced clinical worsening events. More specifically, 20 patients had deteriorated WHO-FC, 16 patients escalated their PH-specific therapy, and 37 patients were rehospitalized due to heart failure or progression of pulmonary hypertension. Among patients who experienced clinical worsening, 23 were in the CA 125>35 U/ml group, and 50 were in the CA 125 ≤ 35 U/ml group.

Compared to those in the control group, patients with PH were younger (40.0 ± 15.3 vs. 50.4 ± 16.2 years, P < 0.001), had worse WHO-FC, 6MWD, echocardiographic and haemodynamic parameters, and had higher serum levels of NT-proBNP and CA 125 [17.3 (11.3, 25.8) vs. 9.5 (6.1, 18.2) U/ml, P < 0.001].

Compared to those with CA 125 ≤ 35 U/ml, patients with CA 125 > 35 U/ml had lower mixed venous oxygen saturation (S_VO₂), lower cardiac index (CI) values, a larger right ventricular end-diastolic diameter (RVED), higher prevalence rates of hyperlipidaemia and pericardial effusion, higher mean right atrial pressure (mRAP), and higher serum levels of NT-proBNP and CA 125 [55.9 (43.7, 83.0) vs. 14.5 (10.6, 21.1) U/ml, P < 0.001]. In addition, patients with CA 125 > 35 U/ml tended to have a smaller left ventricular end-diastolic diameter (LVED) (36.2 ± 6.6 vs. 38.2 ± 6.2 mm, P = 0.064).

As shown in Table 2, ln(CA 125) was weakly correlated with 6MWD, WHO-FC, ln(NT-proBNP), echocardiographic parameters (LVED, RVED, and pericardial effusion), and haemodynamic parameters [S_vO₂, mRAP, CI and pulmonary vascular resistance (PVR)]. In addition, ln(CA 125) tended to correlate with pulmonary arterial wedge pressure (PAWP) (r = 0.123, P = 0.062). However, no correlations were observed between ln(CA 125) and left atrial dimension (r = −0.066, P = 0.318), left ventricular ejection fraction (r = 0.034, P = 0.605), systolic pulmonary arterial pressure (r = −0.037, P = 0.585), or mPAP (r = 0.106, P = 0.109). Similar results were observed in the CTEPH and IPAH subgroups (Supplementary Tables 1, 2).

In multivariate linear regression analysis (enter method), we further assessed correlations between CA 125 and functional status (6MWD, WHO-FC, and NT-proBNP) and echocardiographic (LVED, RVED and pericardial effusion) and haemodynamic (S_vO₂, mRAP, CI, PVR, and PAWP) parameters by adjusting for age, sex, and body mass index. The results showed that ln(CA 125) was still positively correlated with WHO-FC, ln(NT-proBNP), RVED, pericardial effusion, mRAP, and PAWP and negatively correlated with 6MWD, LVED, S_vO₂, and CI (Table 2). Similar results were observed in the CTEPH and IPAH subgroups (Supplementary Tables 1, 2). No problems with collinearity were detected in multivariate linear regression analysis (variance inflation factor <5).

Kaplan–Meier analysis showed that IPAH/CTEPH patients with CA 125 > 35 U/ml had a lower cumulative one-year clinical worsening-free survival rate than those with CA 125 ≤ 35 U/ml (42.5 vs. 73.8%, P < 0.0001; Figure 2).

In univariate Cox analysis, 6MWD, ln(NT-proBNP), S_VO₂, mRAP, PAWP, and CA 125 > 35 U/ml had a P < 0.100 (Table 3). Considering their clinical importance, age, WHO-FC and pericardial effusion were also selected for multivariate Cox analysis (enter method). Events per variable are often used to estimate the sample size needed in multiple Cox analyses, and the lowest acceptable number of events per variable is usually considered to be 10 (27). Given that 73 patients reached the primary endpoint in the present study, it was relatively safe for us to include a maximum of 7 independent variables into multivariate Cox analysis. Model 1 was adjusted for S_vO₂, 6MWD, ln(NT-proBNP), mRAP, and PAWP. Model 2 was adjusted for the variables in model 1 plus age. Model 3 was adjusted for the variables in model 1 plus hyperlipidemia. Model 4 was adjusted for the variables in model 1 plus WHO-FC. Model 5 was adjusted for the variables in model 1 plus pericardial effusion. In all 5 Cox models, CA 125 was found to be an independent predictor of clinical worsening (Table 4). The C statistic was 0.648 [95% CI: 0.577–0.718] for model 1, 0.652 [95% CI: 0.582–0.722] for model 2, 0.649 [95% CI: 0.578–0.720] for model 3, 0.649 [95% CI: 0.579–0.720] for model 4 and 0.647 [95% CI: 0.576–0.718] for model 5. Using bootstrap validation, the optimism-corrected C statistic was 0.609 for model 1, 0.606 for model 2, 0.603 for model 3, 0.604 for model 4 and 0.602 for model 5, indicating that the predictive ability of the models is relatively stable in future patients. We did not observe statistically significant deviations from the proportional hazards assumption in any of the Cox models (Supplementary Table 3). When modeled as restricted cubic splines, CA 125 showed a linear association with the HR for clinical worsening (Supplementary Figure 1). No problems with collinearity were detected in multivariate Cox analysis (variance inflation factor <5). Subgroup analysis also showed that CA 125 was an independent predictor of clinical worsening in CTEPH and IPAH (Supplementary Tables 4–7).

To provide better insight into the predictive value of CA 125 for clinical worsening, we benchmarked it against the established PH biomarker NT-proBNP (18). The areas under the curve for CA 125, NT-proBNP, and combined CA 125 and NT-proBNP were 0.604 (95% CI 0.537–0.667), 0.573 (95% CI 0.507–0.638), and 0.637 (0.571–0.699), respectively. The area under the curve of CA 125 + NT-proBNP was significantly higher than that of NT-proBNP alone (P = 0.0233), as shown in Figure 3. No significant differences were observed between CA 125 and NT-proBNP (P = 0.5108) or CA 125 and CA 125 + NT-proBNP (P = 0.2710).

Compared to patients with normal invasive pulmonary arterial pressure, patients with PH had higher serum CA 125 levels, which was consistent with the results seen in chronic obstructive pulmonary disease (16) and congenital heart disease (17).

WHO-FC, 6MWD, NT-proBNP, pericardial effusion, mRAP, CI, and S_vO₂ are well-established prognostic markers of IPAH (18). We demonstrated that CA 125 was positively correlated with WHO-FC (9), NT-proBNP (28), RVED (10), pericardial effusion (10), mRAP (9, 11), and PAWP (9, 11) in PH, which was consistent with the results seen in heart failure. Additionally, we also found that CA 125 was negatively correlated with 6MWD, LVED, S_vO₂, and CI in PH. Therefore, CA 125 may serve as a novel biomarker of severity in PH.

To date, it remains unclear what leads to CA125 overproduction in heart failure (29, 30). It has been hypothesized to correlate with so-called “stressed” mesothelial cells: (1) mesothelial cells are stimulated by tissue stretching/mechanical stress induced by fluid overload due to heart failure. (2) mesothelial cells are stimulated by inflammatory cytokine network activation (interleukin-1, tumor necrosis factor-α, lipopolysaccharides) (29, 31). PH is characterized by increased mPAP and high PVR, which cause right ventricular hypertrophy, and finally result in right-sided heart failure. Based on our results, we hereby offered a hypothesis to explain the relationship between CA 125 elevation and right heart failure. Elevated PVR increased the afterload of right ventricle, which would further cause right ventricular dilation and elevation of right atrial filling pressure, leading to elevated hydrostatic pressure and congestion, which would further cause both serosal mechanical stretch and third space fluid retention with resultant inflammation and cytokines release (8), ultimately resulting in the elevation of CA 125 (29).

Compared to those with CA 125 ≤ 5 U/ml, patients with CA 125 > 35 U/ml had higher serum levels of NT-proBNP and worse echocardiographic and haemodynamic parameters at baseline. In all 5 Cox models we constructed, CA125 > 35 U/ml was associated with an over 2-fold increased risk of 1-year clinical worsening, which was similar to the results seen in heart failure (32, 33). Furthermore, ROC analysis showed that CA 125 provided additional predictive value in addition to the established PH biomarker NT-proBNP (18). Due to its close relationship with congestion (8), CA 125 should be considered a final organ damage marker in cardiovascular diseases. This may explain why we found that CA 125 was a severe and prognostic marker in PH.

Based on the current knowledge, CA 125 has several merits for use as a biomarker in clinical practice: (1) it is inexpensive, widely available, and measurable with standard methods and has a relatively long half-life (5–7 days) (34–36). (2) It correlates with the severity and prognosis of PH, providing additional information in combination with established risk factors. Therefore, CA 125 may become a valuable tool in the management of PH in the near future. Further studies are needed to evaluate its capacity to monitor response to treatment, serving as a therapeutic target and predicting hard outcomes (such as mortality).

As a retrospective cohort study, follow-up bias is our biggest concern. Two hundred nine patients with CA 125 data were excluded for not having a minimum of 1 year of follow-up data for outcomes (Figure 1). Among these 209 patients, 29 had CA 125 > 35 U/ml, and 180 had CA 125 ≤ 35 U/ml. In other words, these two groups had similar rates of loss to follow-up [29/(40 + 29), 42.0% for CA 125 > 35 U/ml; 180/(191 + 180), 48.5% for CA 125 ≤ 35 U/ml]. Moreover, the percentage of patients who reached the endpoint was higher in the CA 125 > 35 U/ml group (57.5 vs. 26.2%). Taken together, these results indicate that the effect size (e.g., hazard ratio) in the present study would be underestimated rather than exaggerated. Follow-up bias should not undermine our conclusion. The present study included only patients with IPAH/CTEPH, which may limit its generalizability to other etiologies of PH. We planned to conduct a comprehensive retrospective study to investigate the correlations between CA 125 and functional status and echocardiographic and hemodynamic parameters in all five groups of PH patients.