New Treatment Pathways, Drug Development and Clinical Trial Designs: Incorporation of Risk Stratification

Pulmonary arterial hypertension (PAH) is a rare disease characterized by remodeling of the pulmonary vasculature leading to increases in pulmonary vascular resistance and pulmonary pressures and ultimately right ventricular (RV) failure and death.^1,,–3 While significant advances have been made in the evaluation and management of patients with PAH, the mortality for these patients remains high.³ To better tailor therapy and predict clinical course, several risk-stratification tools have been developed. Incorporating these tools into routine clinical practice has enabled practitioners to determine the severity of disease at the time of diagnosis, select an appropriate initial therapeutic regimen, and modify treatment based upon subsequent serial risk reassessments. Unfortunately, despite the availability of these risk tools, use remains suboptimal with a reliance on physician gestalt which often underestimates patients’ objective risk.^4,5 In this review, we discuss the incorporation of risk stratification in the treatment of PAH and its use in clinical trials.

The risk-stratification tools in current clinical use have been developed through data generated by large PAH registries. The currently available risk-stratification tools include the US-based Registry to Evaluate Early and Long-Term PAH Disease Management (REVEAL) 2.0 and REVEAL Lite 2, the Swedish/Comparative Prospective Registry of Newly Initiated Therapies for Pulmonary Hypertension (COMPERA) score, the French Pulmonary Hypertension Network Registry (FPHN) score, the Swedish PAH registry (SPAHR) risk score, and the ESC/ERS risk assessment tool.^6,–12 These multiparametric risk scores have been validated and include variables for patient demographics, functional capacity, biomarkers, imaging parameters, and hemodynamics.

Current guidelines recommend performing risk stratification at the time of initial diagnosis and at follow-up evaluations.¹² Of the available risk scores, each differs in terms of the variables incorporated, variables required for score calculation, and their applicability ­(Table 1). The ESC/ERS guidelines score and the REVEAL 2.0 score include the most variables for assessing initial risk. In comparison with the European-based risk scores, the REVEAL 2.0 score has been validated in different PAH patient populations and can act as a continuous score that includes nonmodifiable variables such as sex, age, and World Health Organization (WHO) group subtype.^6,7 The REVEAL Lite 2.0 is an abridged version of the REVEAL 2.0 score that includes 6 modifiable variables to risk-stratify patients.⁸ The FPHN and COMPERA 2.0 risk scores are the most abbreviated by including only the most prognostic variables in assessing PAH risk, which are WHO-FC, 6-minute walk distance (6MWD), and Brain natriuretic eptide (BNP)/N-terminal prohormone brain natriuretic peptide (NT-proBNP) and classifying patients into 3 or 4 risk strata based on various cutoffs for these variables.^9,,–11 

A notable issue with current risk stratification in PAH is the overcategorization of patients into the intermediate risk group. Authors of previous studies have showed that, when using the 2015 ESC-based guidelines for risk stratification, up to 60%–70% of patients are classified as intermediate risk.¹² To help address this, the updated 2022 ESC guidelines used BNP and 6MWD to further subclassify patients into intermediate low- and intermediate high-risk strata with the recommendation that a 4-strata risk tool be used for follow-up evaluation of PAH patients.^12,13 A recent analysis of PAH patients from the GoDeep meta-registry evaluated the predictive value of several PAH risk scores. While the various risk scores have their strengths and weaknesses, the authors of this study found the REVEAL risk scores to have superior and more granular risk prediction than the ESC/ERS and COMPERA risk scores.¹⁴ 

Use of risk stratification in PAH has led to development of evidence-based treatment pathways for patients based on initial and follow-up risk assessment. The 2022 ESC guidelines recommend those at low/intermediate risk at initial evaluation be started on dual combination therapy with endothelin receptor antagonists and phosphodiesterase inhibitor (class I) and those at high risk be started on triple therapy with addition of prostacyclin-based therapy (class IIa).¹² Subsequent titration of therapy is based on risk reassessment at follow up. Since our treatment of PAH is guided by risk assessment, being able to risk-stratify patients at the time of diagnosis and over time is critical for titration of medical therapy and early referral for lung transplantation consideration.

Despite improvements in our treatment algorithm of PAH patients, our therapies still fail to substantially change patient risk scores over time,¹² suggesting room for improvement in our current risk-stratification systems. To improve our risk-stratification tools, several variables are under investigation. Notably, risk scores are currently lacking an assessment of RV function as a part of their risk assessment. Authors have shown that the addition of RV assessment to current risk scores has increased the prognostic impact of these risk scores.^15,,–17 In a recent retrospective analysis of the single-center registry, it was found that the addition of echocardiographic parameters such as left ventricular end diastolic eccentricity index improved the prognostic power of the REVEAL Lite 2.0 score to predict outcomes of disease progression.¹⁵ This was taken a step further by El-Kersh et al. with the creation of a REVEAL-ECHO score, which retrospectively evaluated the ability of 4 variables (RV size, RV function, severity of tricuspid regurgitation, and the presence of a pericardial effusion) to further risk-stratify PAH patients.¹⁸ This score was found to appropriately reclassify patients’ risk, particularly for those presenting for follow up in the low and intermediate REVEAL risk score classification.¹⁶ Another promising risk-stratification parameter is the tricuspid annular plane systolic excursion/systolic pulmonary artery pressure (TAPSE/SPAP) ratio, a measure of RV-arterial uncoupling¹⁷ that has been found to be prognostic in PAH.^19,20 Other novel blood biomarker, genetic, functional, and hemodynamic parameters are being evaluated for the ability to contribute to risk-stratification assessments.²¹ 

Another challenge relates to the fact that PAH patient demographics have changed since risk-stratification tools were first introduced. For example, the mean age of diagnosis continues to increase, and patients diagnosed with PAH now have more cardiopulmonary comorbidities, both of which worsen their prognosis.²² Further confounding the situation is the finding that older patients with PAH are less likely than the younger cohort to improve their risk profile or be on more aggressive treatment.^22,23 Based upon ESC guidelines, PAH patients with cardiopulmonary comorbidities are treated with initial oral monotherapy regardless of risk, given their decreased tolerance of PAH-based therapies.¹² Developing risk scores that better account for this new demographic of PAH patients is critical to tailor therapy for these patients.

Current initiatives in the field of PAH risk stratification include efforts to better understand interrelatedness between multiple variables as well as how best to give weight to risk variables. Newer analytical models based upon Bayesian networks have shown promise as tools capable of providing a more personalized risk assessment.^21,24 A Bayesian-network-derived prediction model called PHORA was recently validated in PAH patients from the REVEAL and COMPERA registries and showed better discriminatory power than either score alone. Studies to validate this model of risk stratification across a broader range of PAH patient data are ongoing.²⁴ 

Clinical trials in PAH have historically incorporated 6MWD as the primary endpoint. While 6MWD is a marker used for risk stratification and prognostication in PAH, it can be affected by other patient factors and comorbidities, and its change over time is not clearly associated with mortality outcomes.^25,26 Clinical trial design in PAH subsequently shifted to looking at composite endpoints reflecting time to clinical worsening (CW). Though more clinically useful, this type of trial design requires long follow-up periods and larger patient populations to find meaningful differences in outcomes; these limitations add complexity and cost to trials. The composite metrics were also challenging to interpret, as many patients enrolled in these trials were already on a background of PAH therapy. Composite endpoints looking at clinical improvement offer a more meaningful outcome for patients that is easier to evaluate over time but are not currently endorsed by the FDA.^27,28 Recently, risk scores have been explored as surrogates for clinical trial endpoints in multiple post hoc analyses of previous clinical trial data (Table 2).

A novel approach to evaluating the ability of a risk score to predict survival and CW-free survival in a clinical trial involved a post hoc analysis of the PATENT 1 and PATENT 2 trials involving riociguat. This examination of the effect of riociguat versus placebo in PAH patients demonstrated that baseline and 12-week REVEAL risk score as well as change in risk score were significantly associated with both survival and CW-free survival.²⁹ 

A post hoc analysis of the FREEDOM-EV trial, which evaluated the effect of oral treprostinil therapy versus placebo on 690 patients with PAH, evaluated the therapy responsiveness of 4 risk scores (noninvasive French risk assessment, 4-strata COMPERA, REVEAL 2.0, and REVEAL Lite 2.0) at baseline and 12-week follow up. It was demonstrated that follow-up risk scores at 12 weeks were able to predict CW better than baseline risk assessment. In addition, the risk scores improved with treatment.³⁰ The authors concluded that risk scores may be surrogate markers for CW.

Sitbon et al. explored the prognostic and predictive value of the REVEAL 2.0 risk calculator for morbidity and mortality in the GRIPHON study. This clinical trial randomized 1156 patients with PAH to selexipag versus placebo. In the post hoc analysis, patients were evaluated for their number of low-risk criteria (WHO-FC, 6MWD, and N-terminal proBNP) as well as their REVEAL 2.0 risk category at baseline and serially through the study. It was shown that patients treated with selexipag were more likely to increase their number of low-risk criteria and improve their REVEAL 2.0 risk score over the course of the study. These findings suggest that clinical improvement in risk scores could be a surrogate marker for trial endpoints.³¹ 

A recent meta-analysis of PAH trials by Blette et al. provides a different perspective.³⁰ The authors included 3 large, randomized PAH clinical trials in their post hoc analysis (AMBITION, GRIPHON, and SERAPHIN), using time to CW as the primary outcome and time to all-cause mortality as the secondary outcome. Importantly, they assessed the surrogacy of 5 risk scores (COMPERA, COMPERA 2.0, ­REVEAL 2.0, REVEAL Lite 2, and the noninvasive FPHR) for improvement in long-term CW and survival. While this analysis showed an improvement in risk scores and delay in time to CW in the treatment arms, weak correlation was found between the treatment effect and the attainment of low-risk status. The ordinal improvement in risk scores correlated somewhat better with treatment effect but not enough to infer surrogacy. While limitations to this analysis exist, including generalizability to a larger PAH patient cohort and missing data in the compilation of some risk scores, the author suggested further interrogation is needed before risk scores can be consider as surrogates for endpoints in clinical trials.³⁰ 

While the post hoc design of these 4 innovative studies must be acknowledged, they suggest that risk scores offer an appealing option as clinical trial endpoints in future PAH clinical trials.

PAH is a complex disease, and our understanding of the various phenotypes and factors that affect prognosis in this patient populations continues to evolve. Current risk tools have advanced our ability to predict clinical outcomes, and post hoc analyses suggest risk scores hold promise as clinical trial endpoints. However, current risk tools have limitations, in part because they are based on assumptions including a linear relationship between variables and outcomes.³¹ Further refinement of risk models perhaps based upon novel Bayesian and/or machine-learning-based approaches may provide more accurate, individualized prognosis by considering nonlinear relationships between variables and outcomes as well as a wider variety of parameters. This next generation of risk tools will no doubt improve the effectiveness of treatment pathways and clinical trial design. In the meantime, essential next steps include validation of current risk models as valuable clinical trial outcomes using a prospective trial design.