Risk prediction in medically treated chronic thromboembolic pulmonary hypertension

In this national multicenter prospective registry study, patients with CTEPH were consecutively recruited from 18 participating medical centers throughout China. We performed retrospective data analysis using the abovementioned prospective CTEPH registry data in the current study. The study protocol was approved by the Institutional Review Board (IRB) of Fuwai Hospital (Approval No. 2009-208), complies with the Declaration of Helsinki, and is registered on ClinicalTrials.gov (Identifier: NCT01417338). Written informed consent was obtained from all enrolled patients.

Patients were enrolled in the registry according to the following criteria: 1) right heart catheterization (RHC) performed within one month before enrollment between August 2009 and July 2018; 2) PH confirmed by RHC with a pulmonary arterial pressure (mPAP) ≥ 25 mmHg and pulmonary arterial wedge pressure (PAWP) ≤ 15 mmHg at rest; 3) CTEPH diagnosed based on mismatch on ventilation/perfusion (V/Q) scintigraphy with at least one large perfusion defect in one segment or two subsegments or evidence of pulmonary vascular lesions on computed tomography (CT)/magnetic resonance imaging/pulmonary angiography; 4) administration of at least 3 months of effective anticoagulation; and 5) age ranging from 14 to 85 years. Patients complicated with systemic vasculitis, severe pulmonary disease such as interstitial fibrosis or other comorbid conditions that could have caused nonthromboembolic pulmonary hypertension or those with a life expectancy of less than half a year were excluded. In the current study, we selected medically treated patients who had not undergone PEA or BPA before enrollment or during follow-up. The validation cohort consisted of medically treated CTEPH patients retrospectively enrolled from four centers across China from October 2006 to July 2009 according to the same criteria above.

Electrocardiography (ECG), chest X-ray, transthoracic echocardiography, pulmonary function tests, V/Q scintigraphy lung scan, high-resolution CT, pulmonary angiography (if necessary), RHC and laboratory tests were performed to evaluate cardiac and pulmonary function, aid in the diagnosis and guide the treatments of CTEPH. Operability for PEA was assessed by an experienced multidisciplinary team (MDT) in the operation centers. Surgically inoperable CTEPH was defined as CTEPH in which the thrombus was in a peripheral location. For enrolled, medically treated CTEPH patients, the following data were collected: (1) demographics, medical history, clinical symptoms and vital signs; (2) examination results; and (3) treatments.

The primary endpoint of this study was all-cause mortality. Overall survival was measured from the date of RHC to the date of death from any cause. Follow-up was performed by telephone calls, outpatient visits or inpatient admissions every 6 months ± 2 weeks. At each follow-up, vital status was confirmed, as well as surgical events, interventions, and instances of cardiac hospitalization. Patients were followed until death or until the cutoff date of the current study (March 2019).

Continuous variables are presented as the mean ± standard deviation or median [interquartile range (IQR)]. Differences were compared by Student’s t test or the Mann–Whitney U test for two groups and 1-way analysis of variance or the Kruskal–Wallis test for multiple groups, as appropriate. Categorical variables are shown as frequencies and percentages and were compared with the chi-square test. Multiple imputation was used to replace missing values for corresponding variables. Cox proportional hazards analyses were performed to compute hazard ratios (HRs) with 95% confidence intervals (CIs). The proportional hazards assumption was examined by the Schoenfeld residuals method. Univariable Cox analyses were first conducted to screen candidate variables, which were based on the literature and clinical expertise and included demographics, clinical assessments, comorbidities, clinically assessed biomarkers, and variables obtained from pulmonary function tests, echocardiography and RHC. In the univariable analyses, 6MWD, WHO-FC, NT-proBNP, cardiac index, RAP and mixed venous oxygen saturation (SvO₂), which were included as variables in the Swedish/COMPERA risk stratification method, showed significant prognostic value for mortality [15, 16]. Because the Swedish/COMPERA risk stratification method has previously been validated in CTEPH patients [10], we integrated the six parameters into a composite variable according to the specified rules. Hence, we graded the risks for the six variables from 1 to 3 (1 = low, 2 = intermediate, 3 = high) using the recommended thresholds in the guidelines. The rounded means of these grades were then used to define the risk stratum, which was further included in the multivariable model as a priority. Because categorical specifications are favored in daily practice, other continuous risk factors were dichotomized according to the optimal thresholds determined by maximally selected rank statistics before being entered in the multivariable Cox model [18]. Thereafter, stepwise variable selection with entry and exit criteria (P < 0.05) was used to obtain the final model. For routine prognostic assessments, a simplified risk score was then derived by assigning integer numbers to each variable according to the adjusted HRs, and the overall risk score was defined as the sum of the risk points.

Model performance was evaluated for discrimination and calibration. Harrell’s C-index was used to assess the discriminatory power, while the calibration of the 5-year risk prediction was visually evaluated by plotting and comparing the predicted and observed risk. For model comparison, net reclassification improvement (NRI) and integrated discriminatory index (IDI) were used. For NRI, we classified the 5-year mortality into low (< 25%), intermediate (25–50%) and high (> 50%) based on previous literature [11]. Survival was estimated by means of Kaplan–Meier analysis, and difference were compared by the log-rank test. Sensitivity analyses were performed in the following subgroups: (1) newly diagnosed patients, (2) surgically inoperable patients, and (3) a subset that excluded patients with chronic liver disease. Internal validation was evaluated with bootstrapping. External validation was performed to justify whether the derived model and the risk score were also predictive of death in a validation cohort. Differences were considered statistically significant when the two-sided P value was < 0.05. All analyses were performed with the R statistical package (version 4.0.0, R Foundation for Statistical Computing, Vienna, Austria).

A total of 432 medically treated CTEPH patients were enrolled in the current study (Fig. 1). The mean age of the cohort was 53.54 ± 12.25 years, and 52.3% of the patients were males. A total of 59.3% of these patients were newly diagnosed, and 54.2% were surgically inoperable. A total of 251 (58.1%) patients had a history of pulmonary embolism, and chronic kidney disease, defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m², was the most common comorbidity (21.3%). In addition to conventional medical therapies, 55.3% of the patients received at least one PH-targeted drug, while only 9.3% of them received combination therapy. However, as riociguat, the only approved targeted drug for CTEPH, was not available in China before June 2018, it was not in used by any of the patients in the current study cohort. Baseline patient characteristics are presented in Table 1.

During a median follow-up time of 38.73 months (IQR: 20.79, 66.10), 94 patients (21.8%) died. The leading cause of death was right heart failure, and other frequent causes included infection, hemoptysis, and respiratory failure; sudden death was also commonly reported in the cohort. The survival estimates at 1, 3 and 5 years were 95.5% (95% CI: 93.5–97.5%), 83.7% (79.9–87.8%), and 70.9% (65.5–76.7%), respectively.

In univariable Cox analysis, WHO-FC, 6MWD, SvO₂, RAP, cardiac index, mPAP, PVR, NT-proBNP, total bilirubin (TBIL), creatinine, uric acid, blood urea nitrogen, coronary heart disease and CKD were significant predictors for survival (Additional file 1: Table S1). In addition, the Swedish/COMPERA risk stratum also showed significant predictive value for survival: compared with the low-risk group, the intermediate- (HR 4.008; 95% CI 1.746–9.199) and high-risk (HR 9.740; 95% CI 3.537–26.820) groups demonstrated increased risks for mortality (Additional file 1: Table S1 and Figure S1).

According to multivariable Cox proportional hazards analysis, the final model included the following variables: the Swedish/COMPERA risk stratum (low-, intermediate- or high-risk stratum), PVR (≤ or > 1600 dyn s/cm⁵), TBIL (≤ or > 38 µmol/L) and CKD (no or yes). The adjusted HRs and CIs are presented in Table 2. The model showed a C-index of 0.715 (95% CI 0.660–0.770) with good calibration (Fig. 2a).

To simplify its use in clinical practice, a risk score was then derived based on the adjusted coefficients of the above model (Table 2). The overall risk score, which ranged from 0 to 13, demonstrated good discriminatory power similar to that of the model, with a C-index of 0.713 (95% CI 0.658–0.768). Figure 2b shows that the risk score was also well calibrated.

Compared with the model consisting only of the Swedish/COMPERA risk stratum, both the full model [NRI 0.300 (0.100–0.542); IDI 0.095 (0.041–0.191)] and the risk score [NRI 0.300 (0.091–0.480); IDI 0.093 (0.044–0.169)] demonstrated improved discriminatory performance (Table 3). Based on the distribution of the risk score, we divided the patients into 3 risk groups, with low-, intermediate- and high-risk groups determined by risk scores of 0–3, 4–5 or ≥ 6 points, respectively.

The characteristics of the three subgroups are shown in Additional file 1: Table S2. The variables included in the derived risk model differed significantly between the three risk score groups; furthermore, the survival differences between the three risk categories were statistically significant, with P < 0.0001 for all comparisons, P = 0.006 for the 0–3 versus 4–5 points group comparison, P = 0.007 for the 4–5 versus ≥ 6 points group comparison and P < 0.001 for the 0–3 versus ≥ 6 points group comparison (Fig. 3). The estimated 5-year survival rates for the three groups were 91.9% (84.9–99.5%), 79.0% (71.9–868%) and 49.3% (40.0–60.7%), respectively (Table 4).

Sensitivity analyses were performed in three subgroups of patients: patients who were newly diagnosed, those who were surgically inoperable and those without chronic liver disease (excluding 5 patients with chronic hepatitis). The baseline characteristics are shown in Additional file 1: Tables S3-S5. Both the derived model [C-index in newly diagnosed patients: 0.709 (95% CI 0.638–0.780), in surgically inoperable patients: 0.737 (0.649–0.825), in patients without chronic liver disease: 0.718 (0.663–0.773)] and the risk score [C-index: 0.709 (0.640–0.778), 0.743 (0.655–0.831), 0.716 (0.661–0.771), respectively] showed consistent significant predictive value for survival in subgroup analyses. The three risk groups defined by the new risk score also showed significant discriminatory power (Additional file 1: Tables S6-S8 and Figures S2-S4).

The validation cohort consisted of 84 medically treated CTEPH patients with a mean age of 52.01 ± 12.94 years. Twenty-nine patients (34.5%) died during a median follow-up time of 103.56 months (IQR 48.70, 120.00). The survival estimates at 1, 3 and 5 years were 96.4%, 84.2%, and 75.7, respectively. The baseline data of this cohort are presented in Additional file 1: Table S9. Both the derived model [C-index: 0.707 (95% CI: 0.623–0.791)] and the risk score [C-index: 0.721 (0.633–0.809)] demonstrated consistent performance in the validation cohort (Additional file 1: Figure S5). Due to the limited sample size, no subgroup analyses were further performed in the validation cohort.