What's New

AHA 2023: Comparison of Pulmonary Arterial Risk Assessment Tools using a harmonized FDA dataset

Fauvel, C., Lin, S., Correa-Jaque, P., Everett, A., Perer, A., Liu, Y., Kanwar, M., Vanderpool, R., Kraisangka, J., Benza, R. “Comparison of Pulmonary Arterial Risk Assessment Tools using a harmonized FDA dataset”.

Introduction: 

One-year mortality risk for a pulmonary arterial hypertension (PAH) patient is the most substantial. Accordingly, the initial risk stratification should be the most accurate to ensure appropriate aggressiveness of the primary treatment choices. According to PAH guidelines, the European 3-strata system is recommended at baseline for risk assessment. However, the US REVEAL2.0 and REVEAL Lite risk scores also provide high performances. Yet, comparison of risk assessment performances at baseline in the same dataset of patients is still missing.

Hypothesis: 

REVEAL 2.0 provide higher discrimination performances than the European 3-strata system at baseline. We aim to compare the discrimination power at baseline of the currently available tools.

Methods: 

Treatment naïve PAH patients from 6 PAH trials were provided by the FDA and harmonized (AMBITION n=500, ARIES n=126, PATENT n=221, PHIRST n=77, GRIPHON n=231, and SERAPHIN n=270). Kaplan-Meier and log-rank tests were performed for each of the following tools: noninvasive FRENCH method (FR), European 3-strata system (E3), COMPERA2.0 4-strata (CO), REVEAL2.0 (RE), and REVEAL Lite (RL). Their C-indices were compared. A total of 100 bootstrap samples were obtained from the original data providing a robust method for comparing across the different tools.

Results: 

Among the 1,425 patients, 78% were female, 54% NYHA 3-4 with a median age of 51(38, 64) yo, 6-min walking distance of 365(295, 418)m, mPAP of 49(38, 59) mmHg, PVR 9.3(6.2, 13.5) WU. At baseline, the C-index for FR, CO, E3, RL, and RE were 0.64, 0.67, 0.68, 0.69, 0.73, respectively. Out of the 100 bootstrap samples, 99% led to a better C-index for RE compared to E3, with a statistically significant paired t-test (p<e-40) as well as RL compared to E3 (p<e-8). Forrest plots of 95%CI for the C-index differences among the 100 bootstrap samples showed that RE was the best performer, followed by RL, while the E3 and FR models were comparable and worse than the others.

Conclusions: 

REVEAL 2.0 and REVEAL Lite risk scores perform better than the European-3 strata, COMPERA-4 strata, and the noninvasive FRENCH models in predicting 1-year survival at baseline and should be considered as the initial stratification scheme in future guidelines.

https://doi.org/10.1161/circ.148.suppl_1.16736