Fauvel, C., Liu, Z., Lin, S., Correa-Jaque, P., Webb, A., Vanderpool, R., Kanwar, M., Kraisangka, J., Mathur, P., Perer, A., Everett, A., & Benza, R. L. “Comparison between PAH risk assessment methods including PHORA”.

BACKGROUND. Risk assessment is critical to predict survival and then guide therapeutic management for pulmonary arterial hypertension (PAH) patients. Several risk assessment methods have been developed but the comparison of their performances using the same dataset of PAH patients remain to be investigated.
AIM. In a same patients dataset, to compare the performance of the main PAH risk assessment methods used nowadays.
METHODS. We used a harmonized dataset of patients from several PAH trials. First, we built a new version of Pulmonary Hypertension Outcome and Risk Assessment (PHORA) model, (PHORA 2.1) using Random Forrest and Bayesian network analysis as previously published. In the same dataset, the Registry to Evaluate Early and Long-term PAH Disease Management (REVEAL) 2.0 and REVEAL Lite 2.0, 3- and 4-strata Comparative Prospective Registry of Newly Initiated Therapies for Pulmonary Hypertension (COMPERA) 2.0 and noninvasive French Pulmonary Hypertension Registry (FPHR) risk assessment methods were applied. The discrimination performance to predict 1-year survival was compared between all these methods using the area under the curve (AUC) after 5-fold cross validation, and Wilcoxon signed rank test.
RESULTS. A dataset of 494 PAH patients was used for this analysis (mean age 47±18 years-old, 77% of female, 62% of idiopathic or heritable PAH, mean PAP 53 mmHg). PHORA 2.1 method included 13 variables, with an AUC of 0.84 compared to 0.71, 0.78, 0.72, 0.81 and 0.80 for noninvasive FPHR, 4-stata and 3-strata COMPERA, REVEAL 2.0 and REVEAL Lite 2.0, respectively after the 5-fold cross validation. In general, the 95%CI AUC of PHORA 2.0 covers higher values in general with a Wilcoxon signed rank test significantly higher compared to all the other methods (p<0.001).
CONCLUSION. The PHORA 2.1 model, based on a Bayesian Network analysis, depicted the highest AUC to predict the outcome, compared to others PAH risk assessment methods widely used.

