This observational study was conducted at Shanghai Pulmonary Hospital using data collected from January 2012 to April 2023. A total of 875 patients diagnosed with pulmonary hypertension (PH) and 182 age- and gender-matched controls were included. The diagnostic criteria for PH were defined as mean pulmonary artery pressure (mPAP) >20 mmHg, pulmonary vascular resistance (PVR) ≥3 Wood units, and pulmonary artery wedge pressure (PAWP) ≤15 mmHg. All PH patients were diagnosed by right heart catheterization according to the ESC/ERS Guidelines for the Diagnosis and Treatment of Pulmonary Hypertension available in enrollment period (19–21). Individuals without PH or chronic lung disease were included as controls. The study protocol was approved by the Ethics Committee of Shanghai Pulmonary Hospital, and written informed consent was obtained from all participants.

Demographic data, including age and gender, were collected from all participants. For PH patients, body mass index (BMI), 6-minute walking distance (6MWD), and World Health Organization functional class (WHO-FC) were assessed. Venous blood samples were collected for laboratory examinations, including interferon-α (IFN-α), interferon-γ (IFN-γ), N-terminal pro-brain natriuretic peptide (NT-proBNP), uric acid (UA), C-reactive protein (CRP), and lactic acid (LA). Hemodynamic parameters obtained by right heart catheterization included mean right atrial pressure (mRAP), mean pulmonary arterial wedge pressure (mPAWP), mPAP, PVR, cardiac output (CO), cardiac index (CI), and pulmonary arterial oxygen saturation (PA-SaO₂). The primary endpoint of this study was all-cause mortality.

Peripheral blood samples (10 ml) were collected from all participants via the femoral vein and stored in both anticoagulant and procoagulant tubes. The samples were used to assess peripheral blood immunophenotypes and to isolate plasma for further analysis. Plasma concentrations of IFNs were measured using a Luminex-based immunoassay. Human IFN-α ProcartaPlex Simplex (EPX01A-10216-901, Thermo Fisher Scientific) and Human IFN-γ ProcartaPlex Simplex (EPX01A-10228-901, Thermo Fisher Scientific) kits were used following the manufacturer's instructions.

Briefly, capture beads were added to 96-well plates, followed by the addition of samples and standards. Plates were sealed and incubated for 2 h at room temperature (RT), then washed twice using a handheld magnetic plate washer (EPX-55555-000, Thermo Fisher Scientific). Detection antibodies were added and incubated for 30 min at RT, followed by additional washing. Streptavidin-PE was then added and incubated for 30 min at RT. After a final wash, the beads were resuspended, and data were acquired using a Luminex 200 system (Thermo Fisher Scientific).

The Shapiro–Wilk test was used to assess the normality of continuous variables. Normally distributed data were presented as mean ± SD, while non-normally distributed data were presented as median (25th–75th percentile). Differences between groups were analyzed using unpaired Student's t-tests with Welch correction or one-way ANOVA with post hoc Tukey's tests. Categorical variables were presented as frequencies (percentages) and compared using the Pearson χ² test.

Logistic regression models were applied to evaluate the association between IFN levels and PH risk. Receiver operating characteristic (ROC) analysis was used to determine optimal cutoff values. Correlations between variables were assessed using Spearman's rank correlation coefficient (r_s). Cox proportional hazards regression was performed to analyze the relationship between variables and patient outcomes, while Kaplan–Meier survival curves were used to evaluate survival differences according to IFN-α levels. The predictive performance of prognostic models was assessed using ROC analysis and decision curve analysis. A nomogram was constructed to estimate individual overall survival probabilities. A two-tailed P-value <0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 30.0 (IBM Corp.) and RStudio version 4.4.2.

The detailed baseline characteristics of the study population are summarized in Table 1, including 875 PH patients (mean age: 53.1 ± 17.2 years; males: 36.1%) and 182 age- and gender-matched controls (mean age: 51.0 ± 14.5 years; males: 42.3%). PH patients were classified into 4 subgroups (G1/G2/G3/G4: 448/116/236/75, respectively), and 161 patients reached the primary endpoint during follow-up. PAH patients were further classified into IPAH, CHD and CTD (180/88/180, respectively). The mean 6MWD among PH patients was 345.8 ± 134.8 meters. The median NT-proBNP concentration was 1,680.7 pg/mL (interquartile range: 218.0–1,983.5 pg/mL). Most patients were classified into WHO functional class II (n = 284, 32.5%) or III (n = 508, 58.1%).

To examine the relationship between interferon levels and PH, plasma concentrations of IFN-α and IFN-γ were compared between PH patients and controls. Both plasma IFN-α and IFN-γ levels were significantly elevated in PH patients (Figures 1A,D). IFN-α levels were markedly increased across all four PH subgroups and all three PAH subtypes (Figures 1B,C). In contrast, IFN-γ levels were elevated only in G1–G3 PH and in CHD-PAH and CTD-PAH subtypes (Figures 1E,F).

Logistic regression and receiver operating characteristic (ROC) analyses were performed to evaluate the association between IFN levels and PH risk (Table 2, Figure 2). Both IFN-α and IFN-γ were associated with increased disease risk, with IFN-α showing stronger predictive performance, particularly in patients with G4-PH and IPAH. The area under the curve (AUC) values for IFN-α among G1-PH, G2-PH, G3-PH, and G4-PH were 0.66 (sensitivity 54%, specificity 70%, P < 0.0001), 0.69 (sensitivity 89%, specificity 41%, P < 0.0001), 0.67 (sensitivity 64%, specificity 64%, P < 0.0001), and 0.61 (sensitivity 47%, specificity 70%, P < 0.001), respectively (Figure 2A). In comparison, AUC values for IFN-γ were 0.62 (sensitivity 64%, specificity 55%, P < 0.0001), 0.63 (sensitivity 51%, specificity 70%, P < 0.0001), 0.68 (sensitivity 77%, specificity 54%, P < 0.0001), 0.62 (sensitivity 64%, specificity 55%, P < 0.0001), and 0.51 (sensitivity 57%, specificity 52%, P > 0.05) when compared with controls (Figure 2B). Notably, within the PAH subgroup, IFN-α demonstrated superior predictive ability compared with IFN-γ, particularly in CTD-PAH patients (Figures 2C,D).

During follow-up, 161 PH patients died, including 87 with G1-PH, 18 with G2-PH, and 56 with G3-PH. Among PAH patients, 44 IPAH, 18 CHD-PAH, and 25 CTD-PAH patients did not survive. Plasma IFN-α levels were significantly higher in nonsurvivors compared with survivors (P < 0.01; Figure 3A), whereas IFN-γ levels were not significantly different (P = 0.34; Figure 3B).

Cox regression and Kaplan–Meier survival analyses were conducted to evaluate the relationship between IFN levels and overall survival. Univariate Cox regression revealed that IFN-α levels were significantly associated with mortality in PH patients (HR: 1.122, 95% CI: 1.032–1.219, P = 0.007), while IFN-γ was not (HR: 1.087, 95% CI: 0.990–1.193, P = 0.081; Table 3). Pearson's correlation analysis showed that IFN-α levels were modestly correlated with IFN-γ (r_s = 0.15, P < 0.0001) but not with CRP (P = 0.87; Figure 3C), another classical inflammatory marker.

In multivariate Cox regression, IFN-α remained a significant independent predictor of mortality (HR: 1.120, 95% CI: 1.001–1.253, P = 0.048), together with gender (HR: 0.549, 95% CI: 0.375–0.805, P = 0.002), WHO-FC (HR: 1.407, 95% CI: 1.009–1.963, P = 0.044), 6MWD (HR: 0.576, 95% CI: 0.383–0.868, P = 0.008), and NT-proBNP (HR:1.709, 95% CI: 1.398–2.089, P < 0.001; Table 3). However, age had no significance in the multivariate Cox regression model (HR: 1.007, 95% CI: 0.995–1.020, P = 0.247). ROC analysis between survivors and nonsurvivors identified an optimal IFN-α cutoff value of 2.131 pg/mL (AUC: 0.61, 95% CI: 0.56–0.65, P < 0.0001). Kaplan–Meier analysis demonstrated that patients with IFN-α >2.131 pg/mL had significantly lower overall survival (Figure 3D).

To evaluate the prognostic performance of multivariate Cox regression model, we compared it with a clinical model that only included clinical variables (WHO-FC, 6MWD and NT-proBNP). Decision curve analysis indicated that both models provided positive clinical net benefits for intervention thresholds between 8% and 30%, with the Cox regression model offering additional net benefit within the 20%–30% high-risk range (Figure 3E).

Finally, a prognostic nomogram incorporating gender, WHO-FC, 6MWD, NT-proBNP, and IFN-α was constructed to predict individual overall survival (Figure 4). In the nomogram, female patients were assigned a score of 0, while male patients were assigned 13 points. WHO-FC contributed 0–28 points (from class I to IV), 6MWD contributed 0–33 points (decreasing with distance), and IFN-α contributed 0–44 points (increasing with higher concentrations).