Abstract 12424: A Non-Invasive Proteomic Model Predicts Pulmonary Arterial Hypertension Survival

Background: Pulmonary arterial hypertension (PH) is a progressive disease of the pulmonary vasculature. Risk assessment to guide treatment is fundamental to improved outcomes. Current risk models rely on invasive (cardiac cath) and subjective features, prompting need for a non-invasive risk assessment tool in PH. Hypothesis: A multi-marker protein model will accurately predict PH survival Methods: Using a custom multiplex ELISA (based on prior models), we measured proteins targeting pathobiologic pathways in PAH; NTproBNP (cardiac function), ST2 (cardiac fibrosis), IL-6 (inflammation), endostatin (angiogenesis), and HDGF (vascular growth). Subjects were enrolled from the PH Biobank (PAHB, N=2335, age>15) and a validation cohort from Johns Hopkins and Vanderbilt Universities (N=250, age>18). Unsupervised k-means clustering classified subjects by biomarkers. Transplant free survival by cluster were examined using Kaplan-Meier and Cox proportional hazard modeling. Risk by cluster was compared to current clinical models (REVEAL and ESC/ERS Risk) using Harrell’s C Index. Results: The algorithm generated 4 clusters with excellent risk discrimination in both cohorts (Figure 1). The HR for the severe group (vs lowest risk) in the PAHB was 10.1, (C Index 0.72), while in the validation the HR was 8.6 (C Index 0.74). Addition of genetic mutations, age, sex, and comorbidities did not improve model fit. When compared to the REVEAL and ERS/ESC scores (C Index 0.69 and 0.59 respectively), the proteomic model had better discrimination. Conclusions: This multimarker proteomic model non-invasively assessed PH severity and prognosis with improved accuracy compared to current clinical models. A simple blood test may enable frequent assessment of PH patients in a routine clinic setting to support therapeutic decision-making. In addition, this non-invasive model may provide a surrogate endpoint for future clinical trial designs (after prospective validation).