Pulmonary artery compliance is associated with mortality but lacks predictive utility

Background: Pulmonary artery compliance has physiologic appeal in pulmonary hypertension, but its clinical relevance remains in question. Methods: 3 mL/mmHg) versus high (>3 mL/mmHg) pulmonary artery compliance were analyzed in multiple hemodynamic subgroups. For prediction analysis, resampling-based model comparisons were used to develop training (for model fitting) and testing (for model performance evaluation) datasets. Full models including hemodynamic and clinical variables were compared to "handicapped" models in which a variable of interest (i.e., pulmonary artery compliance) was removed using the Harrell's Concordance index and Bayesian methods. Results: Among 12,866 patients with 68,444 person-years of follow-up, there were 4946 deaths. Mortality risk progressively increased with declining pulmonary artery compliance. Low pulmonary artery compliance was associated with a 54% increased hazard of death (HR 1.54; 95%CI 1.45 - 1.61, p<0.001) in adjusted Cox models. However, removal of pulmonary artery compliance from full predictive Cox models did not substantially diminish predictive performance for mortality on hold-out testing data (mean Harrell's Concordance Index 0.703 versus 0.704, p=0.541). Conclusions: Although low pulmonary artery compliance is associated with increased mortality, inclusion of pulmonary artery compliance in multivariate predictive models does not improve mortality prediction. Pulmonary artery compliance does not substantially contribute to predictive performance for mortality.