Evaluation of the European Society of Cardiology Risk Assessment Score in Incident Systemic Sclerosis‐Associated Pulmonary Arterial Hypertension

Patients with pulmonary arterial hypertension (PAH) may be stratified as low, intermediate, or high risk of 1‐year mortality. In 2022, the European Society of Cardiology (ESC) updated and simplified its risk stratification tool, based on three variables: World Health Organization functional class, serum N‐terminal pro‐brain type natriuretic peptide and six‐minute walk distance, applied at follow‐up visits, intended to guide therapy over time.

3][4] The French PH Registry counted the number of variables meeting the low-risk criteria definition at baseline (time of PAH diagnosis) and at first follow-up (within 1 to 12 months of diagnosis).The number of low-risk criteria at both baseline and follow-up (median 4.4 months) accurately predicted the risk of death. 3The Comparative Prospective Registry of Newly Initiated Therapies for PH (COMPERA) group and the Swedish Pulmonary Arterial Hypertension Register assigned an integer value of 1, 2, or 3 to each variable corresponding with low, intermediate, or high risk cut points, herein after referred to as the COMPERA method.The average value of available variables at baseline and follow-up was calculated for each patient (Supplementary Table 2).Both registry groups confirmed that the 2015 ESC risk stratification strategy was a valid method for predicting outcome at baseline and follow-up (4 to 7.2 months) assessments. 2,4][4][5][6] Among SSc-PAH cohorts, the 2015 ESC risk tool has been applied according to both the French PH Registry method (stratified by number of low-risk criteria fulfilled) and the COMPERA method (stratified by the average integer value of risk criteria). 5,6n SSc-PAH, the 2015 ESC risk criteria, according to the French method, accurately predicted survival from baseline and more significantly when applied at follow-up.The COMPERA method accurately predicted mortality at follow-up in those with at least three variables recorded. 5,6The French registry validation of the 2015 ESC risk criteria in SSc-PAH went further to evaluate the determinants of transplant-free survival in SSc-PAH, highlighting the importance of hemodynamic variables at follow-up within the first year of diagnosis.4][5] The latest 2022 ESC/ERS guidelines recommend a more nuanced risk stratification assessment to better characterize those in the intermediate risk profile during follow-up (Supplementary Table S2). 7,8he 2022 ESC risk criteria were proposed and validated using data from the French PH registry and were determined using cut points derived from the COMPERA 2.0 analysis in a three-strata approach at baseline and a four-strata model at follow-up (3 to 24 months after PAH diagnosis).These criteria showed better discrimination for mortality at 1 year and longer term. 7There were more patients who changed risk category between baseline and follow-up using the four-strata approach, suggesting that this may be a more sensitive tool to assess change in risk status.This analysis included patients with SSc-PAH and defined PAH as mean pulmonary arterial pressure (mPAP) greater than or equal to 25 mmHg. 7otable changes in the 2022 ESC risk criteria include updated risk of mortality thresholds and refined cut points for serum (NT-proBNP).Although the three-strata (low, intermediate, high) risk model is maintained for baseline assessments, the fourstrata model assesses prognosis at follow-up.The four-strata model does not include hemodynamic variables and increases granularity of the intermediate group by subdividing this group into intermediate-low and intermediate-high categories. 8The new 2022 ESC guidelines elaborate that the optimal timing of follow-up right heart catheterization (RHC) remains unknown, and practices vary internationally. 8As few patients undergo repeat RHC following PAH diagnosis, the new noninvasive fourstrata risk assessment may be more clinically applicable.The objective of this study was to validate the 2015 and 2022 ESC risk prediction models in Singaporean and Australian patients with SSc with newly diagnosed PAH, for the first time defined by the 2022 definition (mPAP >20 mmHg, pulmonary arterial wedge pressure ≤ 15mmHg and pulmonary vascular resistance > 2 Wood Units [WU]) and by evaluating associations between risk score and all-cause and cardiac-cause mortality. 9

MATERIALS AND METHODS
Study design and patient population.This was a retrospective analysis of prospectively collected data from patients with SSc undergoing annual PAH screening.We included patients with SSc fulfilling the American College of Rheumatology/EULAR criteria who were diagnosed with PAH by RHC, enrolled in the Systemic Sclerosis Cohort, Singapore (SCORE) and the Australian Scleroderma Cohort Study (ASCS). 10rom 2008, the SCORE cohort has recruited patients from three of the following tertiary hospitals in Singapore: the Singapore General Hospital, Tan Tock Seng Hospital, and the Singapore National Hospital, whereas ASCS has collected

SIGNIFICANCE & INNOVATIONS
and according to the 2022 ESC cut points for three-stratum risk assessment at baseline and four-stratum risk assessment at follow-up. 8or the 2015 French PH method, we applied the methods described by Boucly et al in which patients are classified according to the number of low-risk criteria present at baseline out of WHO FC I-II, 6MWD greater than 440 m, right atrial pressure (RAP) less than 8mm Hg, and cardiac index (CI) greater than or equal to 2.5L.m −1 .m−2 . 3Only patients with all four variables recorded were included.
For the 2015 COMPERA registry method, we applied the methods described by Mercurio et al 6 in which a grade of 1 to 3 is given to each of the following variables: WHO FC, 6MWD, RAP, CI, and NT-proBNP according to the 2015 ESC guideline cut offs for low, intermediate, and high risk.The average of the sum obtained after grading each variable was calculated to provide a composite score of 1 (low risk), 2 (intermediate risk), or 3 (high risk) at baseline and first follow-up.This is based upon the COMPERA registry method and included patients with missing variables.We also applied the 2022 ESC risk assessment model three-strata score at baseline and the four-strata assessment model at follow-up using the specified cut offs and included patients with missing variables. 8dditional variables included in the ESC risk assessment tool that were also available in these datasets included clinical signs of right heart failure and progression of symptoms.Clinical signs of right heart failure were defined as clinician-reported presence of peripheral edema and/or raised jugular venous pressure (JVP) and were graded 1 if there was absence of peripheral edema and raised JVP or three if there was presence of peripheral edema or raised JVP.Progression of symptoms was defined as patient-reported worsening of breathlessness in the 6 months before review and was graded 1 if there was no worsening of breathlessness was reported or three if worsening breathlessness was reported.
An echocardiographic variable for risk based on the ESC risk assessment tool was based on right atrial area and presence of pericardial effusion according to the 2015 and 2022 cut offs.Hemodynamic risk variables included RAP and CI according to the 2015 and 2022 cut offs.
Statistical methods.Sex-and age-adjusted Cox proportional hazards multivariate analysis was used to calculate hazard ratios (HRs) for all-cause and cardiac-cause mortality of each parameter at baseline, first follow-up, and change from baseline to follow-up.Patients who maintained a low-risk score or improved from intermediate or high risk were defined as "Improved/Maintained Low Risk," and those who maintained a high risk score or increased risk score to intermediate or high risk were defined as "Increased/Maintained High Risk." Survival from PAH diagnosis was assessed using Kaplan-Meier survival curves for all-cause and cardiac-cause mortality.Accelerated failure time (AFT) models were used to parametrically assess overall survival according to the risk score and compared with low-risk patients.An AFT model assumes that any time spent in a higher risk profile accelerates time to mortality.The data underlying this article will be shared on reasonable request to the corresponding author.

RESULTS
Patient characteristics.From 2007, there were 260 patients with incident SSc-PAH and sufficient variables to evaluate the ESC risk assessment tool at diagnosis with PAH, 54 patients from the SCORE cohort, and 206 from the ASCS cohort (baseline characteristics in Supplementary Table 3).There were more patients classified as having the diffuse cutaneous SSc in the SCORE cohort compared with the ASCS cohort (15 [33.3%] vs 43 [21.6%];P = 0.095).There were significant differences in autoantibody profile between the two SSc cohorts, with more patients with positive anticentromere antibody (105 [52.0%] vs 11 [25.0%];P = 0.001) and fewer patients with positive antitopoisomerase (anti-Scl70), anti-Ro52, and anti-U1RNP antibody in the ASCS compared with the SCORE cohort (Supplementary Table 3).
In terms of patient symptomatology and PAH severity, there were no significant differences in 6MWD, NT-proBNP, mean RAP, mPAP, pulmonary arterial wedge pressure, or CI between the two cohorts.However, the baseline WHO FC was better in the SCORE cohort compared with the ASCS cohort with more patients with WHO FC I in SCORE compared with ASCS (12 [23.5%] vs 9 [5.3%];P < 0.001).Hemodynamic differences at baseline were observed including lower cardiac output in the SCORE cohort compared with the ASCS cohort (3.90 L/min [3.50-4.50]vs 4.90 L/min [4.00-6.10];P < 0.001) and higher BROWN ET AL pulmonary vascular resistance in the SCORE cohort compared with the ASCS (5.72 WU [3.74-8.00]vs 4.00 WU [2.88-6.60];P = 0.008) (Supplementary Table 3).
The mean time from PAH diagnosis to first follow-up database visit in the SCORE cohort was 322.0 days (277.0-412.0)and ASCS was 375.0 days (336.0-434.0)(P = 0.010).The mean duration of follow-up from registry enrollment to death or censoring in the SCORE cohort was 4.3 years (1.7-7.8) and ASCS was 4.3 years (2.4-6.8)(P = 0.907) (Supplementary Table 3).The frequency of recorded and missing variables at baseline and followup in the combined cohorts is included in Supplementary Table 4.
There were significant differences in treatment strategy between the SCORE and ASCS cohorts.Significantly more patients in ASCS compared with SCORE received treatment before first follow-up with any endothelin receptor antagonist (65 [56.7%] vs 0 [0%]; P < 0.001) or with dual therapy (14 [  5).In the whole cohort, we used Sankey diagrams to represent the change in risk scores from baseline to most complete follow-up within 2 years according to the 2015 COMPERA risk score and the 2022 ESC three-and four-strata model (Supplementary Figure 1).
Cox proportional hazards regression models.2015 ESC risk assessment individual variables.In Cox proportional hazards regression models of the individual variables included in the 2015 ESC risk assessment score (Table 2), there was an increasing risk of all-cause death for patients with high-risk WHO FC (FC IV) at baseline and follow-up (HR 2.11 [95% confidence interval 1.10-4.02];P = 0.024 and HR 3.88 [95% confidence interval 1.89-7.96];P < 0.001), compared with low-risk groups (FC I, II).There was increased risk of cardiac-cause mortality for those with high-risk WHO FC at baseline, follow-up, and maintenance of high-risk status at follow-up, compared with low-risk groups.
Those classified as high risk 6MWD at follow-up had increased risk of all-cause mortality (HR 9.01 [95% confidence interval 2.97-27.29];P < 0.001) compared with those classified as low risk.There was increased risk of cardiac-cause mortality for those with high risk 6WMD at baseline and follow-up, compared with low-risk groups.
Serum NT-proBNP, if high risk at baseline, was associated with increased risk of all-cause mortality (HR 2.90 [95% confidence interval 1.32-6.39];P = 0.008) compared with those classified as low risk.Serum NT-proBNP did not reach significance as a predictor for cardiac-cause mortality at baseline, follow-up, or if change in value.Echocardiographic parameters (right atrial area and presence of pericardial effusion) were not associated with all-cause or cardiac-cause mortality at baseline or follow-up in this cohort.
Those classified as intermediate and high risk by RHC parameters (RAP and CI) at baseline had increased all-cause mortality (HR 1.64 [95% confidence interval 0.94-2.89;P = 0.084] and HR 1.96 [95% confidence interval 1.02-3.77];P = 0.045, respectively).There were only four patients in this analysis who had an RHC repeated within 2 years.
2015 French PH Registry composite score.Patients with at least one low-risk criterion at baseline had reduced allcause (HR 0.43 [95% confidence interval 0.24-0.77];P = 0.005) and cardiac-cause mortality (HR 0.46 [95% confidence interval 0.24-0.88];P = 0.019) at baseline compared with those with no low-risk criteria (Table 2).Those with two low-risk criteria at baseline had reduced cardiac-cause mortality (HR 0.16 [95% confidence interval 0.04-0.67];P = 0.012) but did not reach significance for all-cause mortality (HR 0.57 [95% confidence interval 0.27-1.22];P = 0.148) compared with those with no low-risk criteria.The French score relies on the number of low-risk criteria present; therefore, the authors did not include patients with missing data for this analysis and were not able to evaluate the French risk score at follow-up because of the large numbers of missing RHC data.
2015 COMPERA composite score.The 2015 COMPERA composite score applied at baseline and follow-up was associated with all-cause and cardiac-cause mortality (  high risk had an HR of 4.00 (95% confidence interval 1.82-8.78;P = 0.001) for all-cause mortality and an HR of 5.61 (95% confidence interval 1.75-18.02;P = 0.004) for cardiac-cause mortality, compared with low-risk groups.
2022 ESC three-and four-strata models.Cox regression models, there was an increased risk of mortality at baseline and follow-up with increasing 2022 ESC risk scores (Table 2).Those with a high-risk score at baseline and follow-up compared with those with a low-risk score had increased all-cause mortality (HR 2.05 [95% confidence interval 1.01-4.19];P = 0.048 and HR 5.31 [95% confidence interval 2.54-11.11];P < 0.001, respectively).Those with an intermediate risk score at baseline compared with low risk had increased cardiac-cause mortality (HR 3.40 [95% confidence interval 1.20-9.63];P = 0.021), although intermediate risk profile did not reach significance for all-cause mortality.
At follow-up, those with an intermediate-high risk score had an increased all-cause mortality (HR 3.17 [95% confidence interval 1.65 -6.08]; P = 0.001) compared with low-risk groups.There was no significant association with mortality by intermediate-low risk at baseline and follow-up.
Change in 2022 ESC risk score between baseline and followup was associated with all-cause and cardiac-cause mortality.Those who increased risk score or who maintained intermediate or high risk compared with those who maintained low risk or improved to low or lower risk had an HR of 3.04 (95% confidence interval 1.91-4.83;P < 0.001) for all-cause and an HR of 2.39 (95% confidence interval 1.25-4.58;P = 0.008) for cardiac-cause mortality.
Kaplan-Meier survival curve analysis.2015 COMPERA composite score.Overall survival, including both allcause and cardiac-cause mortality, using the 2015 COMPERA risk method at baseline, follow-up, and according to change in risk profile, is shown in Supplementary Figure 2.There were significant differences in overall survival according to the 2015 COMPERA risk method at baseline and at follow-up (P = 0.014 and P = 0.001, respectively).There were also significant differences in cardiac-cause mortality by risk score at baseline and ii.

Kaplan-Meier Survival estimates
Survival by change in risk scores follow-up assessments (P = 0.017 and P = 0.018, respectively).Change in 2015 COMPERA score was not associated with overall or cardiac survival (P = 0.266 and P = 0.455, respectively).
2022 ESC three-and four-strata models.Using the 2022 three-strata and four-strata models, overall survival at baseline, follow-up, and according to change in scores is shown in Supplementary Figure S3 and Figure 1, with survival to cardiac-cause mortality shown in column B. There were significant differences in overall survival and cardiac mortality using the three-and four-strata approach at baseline, follow-up, and change in risk profile.
AFT model for 2022 ESC three-and four-strata risk models.Using the 2022 ESC three-and four-strata risk score, AFT models demonstrated that time ever spent in higher risk profiles significantly accelerated time to all-cause mortality compared with low-risk groups.Time ratio to mortality for those in an intermediate-low risk profile was 0.50 (95% confidence interval 0.27-0.89;P = 0.020) and 0.15 (95% confidence interval 0.08-0.30;P < 0.001) for those in a high risk profile (Table 3).
Time to death was attenuated by treatment with endothelin receptor antagonist, phosphodiesterase-5 inhibitor, or both when compared with those in low-risk profile according to the 2022 ESC method who were receiving treatment.For example, for those who were classified as intermediate-low risk and receiving treatment, the time ratio to death was 0.51 (95% confidence interval 0.25-1.04;P = 0.062) compared with those who were intermediate-low risk not receiving treatment in whom time ratio to mortality was 0.29 (95% confidence interval 0.12-0.72;P = 0.007) (Table 4) meaning that PH treatment increased the time to death or reduced the impact of increasing risk score on accelerating time to mortality.Analysis using Cox proportional hazards model supported the AFT analysis for the 2022 ESC risk scores with and without treatment (Tables 3 and 4).Subgroup analysis of those with mPAP greater than or equal to 20 mmHg who received PH-specific therapy.Supplementary Kaplan-Meier survival curve analyses were undertaken only including those patients with PAH who received any PH-specific therapy at any time.Survival to all-cause and cardiac-cause mortality were evaluated according to the 2022 ESC risk models (Supplementary Figures 4 and 5), and the risk tools maintained the ability to discriminate survival between risk strata.To further characterize those with mPAP 20 to 25 mm Hg compared with those with mPAP greater than or equal to 25 mm Hg, a supplementary comparison of characteristics was undertaken (Supplementary Table 6).

DISCUSSION
In this large multinational cohort of patients with SSc with incident PAH, we evaluated the 2015 and 2022 ESC risk assessment criteria and composite scoring methods.Our main findings were that the majority of SSc-PAH (72.2%) were graded as intermediate mortality risk at baseline according to the 2022 ESC three-stratum risk assessment and at follow-up, approximately half (55.5%) of the cohort were classified as low or intermediatelow risk.
The 2022 ESC three-strata risk model at baseline was associated with all-cause and cardiac-cause mortality.At follow-up, the 2022 ESC four-strata risk model demonstrated statistically significant differences in all-cause and cardiac survival between the four-strata risk groups with higher risk groups having worse survival.Risk assessment according to the 2022 ESC guidelines in this analysis provided better characterization of those classified in the intermediate-risk profile by identifying those at intermediatelow or intermediate-high risk at follow-up categories in this analysis who had different outcomes.
Compared with the 2015 COMPERA score, change in 2022 ESC score was associated with significant changes in risk of mortality.In this analysis, those with an improvement in 2022 ESC risk score between baseline and follow-up or maintenance of low risk at follow-up had better survival compared with those who maintained high-risk status or increased from low risk to intermediate or high risk; whereas, improvement in 2015 COMPERA score was not associated with improvement in survival.
Our results, strengthened by the inclusion of AFT modeling, confirm the importance of achieving a low-risk status in SSc-PAH in order to improve survival.This can be done through regular risk assessments, which enable the appropriate guidance on the use and escalation of PAH-specific therapy in accordance with the 2022 ESC guidelines. 7,8For rheumatologists who may be the primary physicians in the care of patients with SSc, the noninvasive variables required to complete the 2022 four-strata risk assessment at follow-up provide an accessible means to guide treatment and predict outcome.
6][7] In this analysis, WHO FC, NT-proBNP, and 6WMD were associated with mortality.High riskserum NT-proBNP levels at baseline were significantly associated with all-cause mortality, which has not been shown previously in SSc-PAH, though is a well-established predictor of prognosis in group 1 PAH. 5 Transthoracic echocardiograph parameters at baseline and follow-up were not significantly associated with outcome.
With respect to outcomes in SSc-PAH, compared with the French registry's combined group 1 PAH cohort (including patients with idiopathic, heritable, drug-induced, and SSc-PAH), the proportions of patients in each 2022 risk category at baseline in this analysis were similar (low risk 16% vs 15.7%, intermediate risk 67% vs 72.2%, high risk 16% vs 12.1%, respectively). 7At follow-up, fewer patients in our cohort compared with the French group 1 PH registry were classified as low risk according to the 2022 ESC four-strata assessment (19.1% vs 39%, respectively). 7his supports the idea that patients with SSc-PAH are a unique population with worse outcomes compared with those with PAH because of other causes. 2,5nalyses of PH treatments and their impact on mortality in this cohort did show attenuation of mortality risk and increased time to mortality among those who received any PH treatment in all risk categories.This analysis should be interpreted with the understanding that these registries include patients who were diagnosed and treated with PAH from 2007 onwards and received treatments according to criteria for access to reimbursed therapy appropriate to that era.We included patients meeting the current definition of PAH, whereas before 2019, those with mPAP between 20 and 25 mmHg may not have been eligible to receive PH therapies in the time they were being managed; this may affect the treatment data and outcomes in this analysis.
In Australia, government-funded PH-specific therapy is available for patients with mPAP greater than 25 mmHg.As we defined PAH according to mPAP greater than or equal to 20mmHg, not all of the patients included in this analysis received PH-specific therapies.To evaluate the prognostic accuracy of the 2022 ESC risk stratification models only among treated patients, supplementary analyses were undertaken, and we found that the 2022 ESC risk models maintained significant discrimination of risk strata.
Additionally, there were SSc-specific phenotypic differences noted between the Australian and Singaporean cohorts, which have been previously described. 11Of relevance to this analysis, there was a higher frequency of patients classified as having the diffuse cutaneous subtype and more severe ILD with a significantly lower median forced vital capacity percentage predicted in the SCORE cohort compared with the ASCS.Previous analyses of SCORE participants with concomitant ILD and PH (ILD-PH) have shown that this group has worse outcomes than those with PAH alone. 12

Figure 1 .
Figure 1.Kaplan-Meier survival curves according to the 2022 European Society of Cardiology risk score from PAH diagnosis to mortality applied at first follow-up visit (top left and right) and by change in risk scores (bottom left and right).(A) All-cause mortality (B) Cardiac-cause mortality.PAH, pulmonary arterial hypertension.

Table 1 .
Patient characteristics at baseline by risk score determined using WHO FC, 6MWD, NT-proBNP, and hemodynamic scores according to the 2015 and 2022 ESC criteria* Variables were assessed according to the cut points in the 2015 ESC risk assessment table individually, 1 in composite scoring methods according to the 2015 French PH method and the 2015 COMPERA method,

Table 3 )
. Characteristics at baseline according to the 2015 COMPERA risk assessment score and the 2022 ESC risk assessment composite score are shown in Table 1.Using the 2015 COMPERA risk score at baseline, most patients (134 [69.8%]) were classified as intermediate risk at baseline; there were 34 patients (17.7%) classified as high risk and 24 (12.5%) as low risk.Using the 2022 ESC three-strata risk assessment at baseline, 31 patients (15.7%) were low risk, 143 (72.2%) were intermediate risk, and 24 (12.1%) were high risk.Using the 2015 COMPERA risk score at follow-up, most patients were classified as intermediate risk (82 [68.9%]), with 22 (18.5%)classified as low risk and 15 (12.6%) as high risk.Using the 2022 ESC risk score at follow-up, most patients were classified at intermediate-low risk (64 [32.0%]), with 58 (29.0%) classified as intermediate-high risk, 47 (23.5%) as low risk, and 31 (15.5%) at high risk (Supplementary Table

Table 2 .
HRs for all-cause and cardiac-cause mortality according to individual risk variable and composite scores in combined ASCS and SCORE cohort* BROWN ET ALhigh-risk score at baseline compared with those in the low-risk category had an HR of 2.34 (95% confidence interval 1.11-4.95;P= 0.026) for all-cause mortality and an HR of 7.39 (95% confidence interval 1.62-33.64;P= 0.010) for cardiac-cause mortality.Intermediate score at baseline was associated with cardiaccause mortality (HR 4.87 [95% confidence interval 1.17-20.30];P = 0.030) but did not reach significance for all-cause mortality, compared with low-risk groups.At follow-up, those classified as

Table 4 .
AFT model and corresponding Cox proportional hazards model for all-cause mortality according to treatment by 2022 ESC Risk Stratification score* * 95% CI, 95% confidence interval; AFT, accelerated failure time; ESC, European Society of Cardiology.

Table 3 .
AFT model and corresponding Cox proportional hazards model for all-cause mortality by 2022 ESC Risk Stratification score* ESC 2022 RISK ASSESSMENT MODEL IN SSc-PAH