Risk Factor Assessment for Acute Pulmonary Hypertension Crisis

Overview

The goal of this observational study is to learn about pulmonary hypertension crisis (PHC) - a severe, and often fatal complication - in patients with pulmonary hypertension (PH). The main questions this study aims to answer are: What are the clinical and hemodynamic features of PHC, and what underlying pathophysiological mechanisms cause it to develop? Can these features be used to diagnostic PHC, predict who is at risk of developing it or dying from it, and develop targeted prevention and treatment of PHC? A multi-center registry platform and biobank will be established to enroll and follow up patients with PH. Clinical data, hemodynamic measurements, and biological specimen will be collected. Risk prediction and early warning tools of PHC will be developed.

Pulmonary hypertension crisis (PHC) is a severe and fatal complication of pulmonary hypertension (PH), characterized by the sudden elevation of pulmonary vascular resistance, acute right hear failure and collapse of systemic circulation. However, PHC currently lacks any internationally accepted definition or diagnostic criteria, and it is still unclear which are the clinical, pathophysiological and molecular risk factors of PHC. This study aims to establish a large clinical cohort of patients with PH, validate the definition of PHC, identify risk factors of PHC occurrence and mortality, and develop tools for the risk-prediction and early-warning of PHC. This is a multi-center, prospective cohort study with a retrospective component for model training. Following national health-information standards, a standardized registry dataset will be established to cover demographics, clinical features, laboratory tests, imaging, and hemodynamic profile. Existing PH cohort databases will be integrated and expanded to a target of 5,000 patients. A subset of cases with right heart catheter parameters (RHC) will be extracted. A multi-center biospecimen management platform will be established. For PHC onset, the investigators will assemble a clinical-molecular-imaging feature set and use convolutional neural networks and tree-based algorithms for feature extraction, then build a cross-modal prediction model using multi-task learning, neural networks, and a Transformer architecture. The model will be prospectively validated and tuned against real-world performance. For PHC mortality, models will be built on patients meeting the new PHC criteria, using 28-day attributable death, all-cause death, and long-term death as endpoints. Models will be trained on a retrospective cohort and then evaluated for accuracy and generalizability on both an internal validation set and an external prospective-cohort validation set.