Hemodynamic Parameters in Predicting Survival in Pulmonary Arterial Hypertension

Abstract

<h3>Purpose</h3> Hemodynamics are considered to be extremely relevant when clinically assessing patients with pulmonary arterial hypertension (PAH), especially for estimating prognosis. Yet, inclusion of hemodynamic data in PAH risk-stratification remains inconsistent and controversial, having been derived from limited data-sets. <h3>Methods</h3> We analyzed hemodynamic variables obtained at baseline in 4 PAH clinical trials: SERAPHIN, BREATHE-1, PATENT-2 and the subcutaneous treprostinil study; and at follow-up from SERAPHIN (week 26), PATENT-2 (week12), and the subcutaneous treprostinil study (week12). Primary outcome analyzed was mortality at <i>time to event</i> noted in each individual trial. Only the significant p-values (p-value<0.05) of each hemodynamic variable were aggregated across the clinical trials using Stouffer's method, weighed by their ‘n' in clinical trials. Then, we converted the aggregated p-value for each hemodynamic into a prioritization matrix (a sig. Stouffer's w) using the logarithmic function, for visualization. A sig. Stouffer's w of 1.3 is equivalent to a p-value = 0.05, with higher values being increasingly significant. <h3>Results</h3> Collective data from 2,211 patients were available at baseline and for 1,487 patients at follow-up. 30 hemodynamic variables (a combination of calculated and derived data) were considered for this analysis. At baseline, the most significant variables predictive of primary outcomes were PA sat, stroke work, stroke volume, stroke volume index and pulmonary arterial elastance (more reflective of output). At follow-up, the most predictive variables were PA sat, total pulmonary resistance, pulmonary vascular resistance, pulmonary arterial elastance, and right atrial pressure (more reflective of RV afterload) <h3>Conclusion</h3> To our knowledge, this is the largest aggregated attempt to analyze hemodynamic parameters in predicting survival. This analysis will be used to select features for derivation of machine-learned PAH patient risk stratification models.