What's New

ATS: Pulmonary Arterial Registry Risk Algorithms Improve Upon Clinical Trial Enrichment Methodology

Scott JV, Kanwar M, Garnett C, Stockbridge N, Benza RL. B27. UP-TO-DATE PAH ASSESSMENT AND MANAGEMENT: American Thoracic Society; 2020:A2920-A.

Introduction: Pulmonary arterial hypertension (PAH) is a rare disease with poor prognosis. Clinical trials for PAH now use time to morbidity/mortality event (clinical worsening) as the primary endpoint instead of six-minute walk distance (6MWD). These trials require nearly four times the sample size and 10 times the trial length versus trials with 6MWD as their endpoint. Reduction in trial samples sizes would improve trial efficiency. To this end, registry risk algorithms were investigated as candidates for trial enrichment by identifying patients with high-risk of clinical worsening using baseline characteristics.

Methods: COMPERA, French, REVEAL 2.0 were investigated as trial enrichment candidates. Patient-level data from three PAH phase 3 trials (AMBITION, SERAPHIN, and GRIPHON) were pooled and standardized. Receiver-operating curves (ROC) were generated for each algorithm to determine predictive capability for clinical worsening. Survival tree analysis was performed to identify algorithm cut-points that created unique risk-level groups. Hazard ratios between the pooled treatment and pooled placebo patients were re-calculated for each risk group. Power analysis was conducted by bootstrapping to estimate sample size reductions for multiple enrichment methods. Results: ROC analysis revealed that REVEAL 2.0 and COMPERA performed reasonably (both with AUC 0.70) at predicting clinical worsening, while the French score was less accurate (AUC 0.66). Survival tree analysis demonstrated that all algorithms were able to identify a unique low, intermediate, and high-risk group, but REVEAL 2.0 was more precise, further separating the population into very low, low, intermediate, and high-risk. A treatment effect was found in each unique risk group (all p-values < 0.05). Using power analyses with bootstrapping, enriching trials by enrolling only high-risk patients substantially decreased estimated sample size versus enrichment by NYHA class (Figure 1). Enrollment of intermediate and high-risk patients still lowered estimated sample size versus NYHA enrichment, but only when REVEAL 2.0 was applied in AMBITION and GRIPHON trials. Similar results were found for enrollment of 50% high risk, 50% all other risk levels.

Conclusion: Registry risk algorithms are better predictors of clinical worsening than NYHA class alone. REVEAL 2.0 outperformed all registry risk algorithms, except for in the SERAPHIN trial, which is attributable to SERAPHIN using a non-standard NT-proBNP assay. Further investigations should consider availability of high-risk patients and cost of screening when enrichment is used.

This abstract is funded by: U.S. Food and Drug Administration