The added value of echocardiography in PAH risk assessment an AI machine learning-derived analysis of the ULTRA RIGHT VALUE registry

Abstract Background The right ventricle (RV) significantly influences outcomes in pulmonary arterial hypertension (PAH), yet its assessment is not included in current mortality risk scores. Purpose This study aimed to evaluate the added prognostic value of RV echocardiographic parameters (including RV strain analysis) when integrated with established risk scores in a large multicenter European cohort of prevalent PAH patients enrolled in the ULTRAsound RIGHt venTricular eVALUation in PAH (ULTRA RIGHT VALUE) study. Methods A total of 401 prevalent PAH patients, enrolled between January 2022 and December 2023, underwent comprehensive echocardiographic evaluation and risk assessment using the REVEAL 2.0 and COMPERA 2.0 scoring systems. The primary outcome was a composite of all-cause mortality and heart failure hospitalizations. The secondary outcome was the maintenance or achievement of low-risk status according to ESC/ERS and REVEAL 2.0 at the latest follow-up. Results During a median follow-up of 177 days (IQR 126–241), 54 patients (13.5%) experienced adverse events, including heart failure hospitalization (44, 11.0%) or death (10, 2.5%). In univariate Cox regression analysis adjusted for age and PAH duration, increased right heart dimensions and reduced left ventricular (LV) dimensions were significantly associated with adverse outcomes. Markers of impaired RV-pulmonary artery (RV-PA) coupling showed the strongest associations with the primary outcome, while LV eccentricity index and pericardial effusion were most predictive of the secondary outcome. Missing data on functional class, 6-minute walking distance, or brain natriuretic peptide levels reduced the discrimination capacity of risk scores. Incorporating RV echocardiographic parameters into the COMPERA and REVEAL 2.0 models in case of one key missing data improved the c-index. Enhanced risk stratification was also observed when integrating RV echocardiographic variables with the complete ESC/ERS and REVEAL 2.0 scores. Random survival forest analysis and permutation importance analysis identified systolic dimensional metrics of the RV and LV (i.e., RV end-systolic area, LV end-systolic area, and LV eccentricity index), along with RV-PA coupling markers, as those with the highest added prognostic value. Conclusion Integrating echocardiographic parameters of the right heart enhances the predictive accuracy of existing risk scores in PAH, thereby improving risk stratification and potentially guiding clinical decision-making. Figure legend: The plot shows the relative importance of echocardiographic variables when added to the ESC/ERS and REVEAL 2.0 risk models (y-axis) by randomly shuffling their values and observing the change in the c-index (x-axis). For each variable in the model the c-index drop and the relative 95% C.I. are reported in the corresponding row.Permutation importance analysis