Pulmonary hypertension (PH) is an independent risk factor for adverse outcome across the spectrum of medical diseases. However, there is wide variability in clinical event rates within PH populations: the estimated 5-year mortality rate following PH diagnosis is ~48% in right heart catheterization (RHC) registries, whereas normal lifespan is reported for a sizeable subgroup of patients with severe PH. Understanding the clinical profiles corresponding to PH risk and resilience has important implications on prognosis and individualized treatment planning. Traditional methods for classifying and prognosticating patients use reductionist methods, such as linear regression-based models, which provide important information on individual risk predictors. However, such probabilistic methods do not consider interdependent relationships between variables for classifying patients, and, thus, provide limited knowledge on key clinical parameters that integrate to determine phenotypes. In this proposal, we present novel preliminary data to illustrate the utility of network (Bayesian) methods for enhancing classification and risk stratification of patients. Specifically, we developed a correlation network based on data from a large cohort of patients with unexplained exercise intolerance undergoing cardiopulmonary exercise testing. Our findings focused on a novel collection of 10 interrelated exercise variables that identified 4 patient groups, which were independent of traditional exercise diagnoses and associated with distinctly different clinical profiles and clinical event rates. An overarching goal of this project is to apply this network methodology to PH. The Veterans Affairs Clinical Assessment, Reporting and Tracking (VA-CART) program includes RHC, clinical, and outcome data from a national patient cohort. We recently analyzed data from RHC patients in VA- CART (2007-2012; N=21,727), and observed that the prevalence of PH was 57%. The adjusted hazard ratio for PH was 2.16 (95% confidence interval, 1.96-2.38, P<0.001) compared to non-PH patients. These data affirm the importance of PH on survival; however, the factors influencing outcome within PH patients remain unknown. Thus, the central hypothesis of this proposal is: Analyzing the VA-CART database using network methods will discover novel PH subgroups defined by a unique collection of variables, distinct clinical profiles, and significant outcome differences. The study aims are (1) develop a correlation network using RHC and other data to discover novel PH subgroups defined by unique clinical and outcome profiles, and (2) use Betweeness centrality analyses to determine clinical variables that differentiate survivors vs. non-survivors (e.g., resilience vs. risk, respectively) in PH. Our network findings from VA-CART will also be validated in a second, gender-balanced and outcome-linked RHC database (Vanderbilt University). Collectively, these studies leverage network methods for optimally phenotyping and prognosticating PH patients, and aim to provide a basis for individualizing care for patients affected by this highly morbid disease.
Pulmonary hypertension is a medical disorder characterized by high blood pressure in the lung artery circulation, and is an important cause of death across the spectrum of medical diseases. However, agreement on the factors that are important for determining risk and survival in pulmonary hypertension is lacking. This project aims to apply new analytical methodologies focusing on networks to identify collections of clinical variables that optimally classify and prognosticate pulmonary hypertension patients. In doing so, findings from this study will establish a key path toward individualizing care for patients affected by this highly morbid disease.
