Multimodal Model for Predicting Exercise-induced pulmonary hypertension Validated by Invasive Exercise Hemodynamics: A Prospective Study

Background Exercise-induced pulmonary hypertension (EiPH) represents an early stage of pulmonary vascular disease that remains challenging to identify noninvasively, particularly in patients with borderline resting haemodynamics. We aimed to develop and validate a multimodal non-invasive model integrating clinical characteristics, exercise echocardiography, and cardiopulmonary exercise testing (CPET) to accurately predict invasively confirmed EiPH. Methods In this prospective cohort study, consecutive adults with exercise intolerance after chronic thromboembolic disease, increased tricuspid regurgitation velocity on transthoracic echocardiography, or previously documented mildly elevated pulmonary artery pressure were enrolled. All participants underwent comprehensive clinical assessment, resting and exercise echocardiography, CPET, and invasive exercise right heart catheterization as the reference standard. Feature selection was performed using Spearman’s correlation analysis and regression with the Least Absolute Shrinkage and Selection (LASSO) operator. Single-modality and multimodal prediction models based on clinical, echocardiographic, and CPET variables were constructed. Model performance was evaluated using receiver operating characteristic (ROC) analysis, net reclassification improvement and integrated discriminant improvement. Prespecified subgroup analyses were performed. Associations between selected predictors and invasive hemodynamic parameters were analyzed to explore underlying pathophysiology. Results This study included a total of 86 participants, comprising 45 patients with EiPH and 41 patients without EiPH. Compared with the non-EiPH group, EiPH patients had a poorer WHO functional class; during exercise, their peak mPAP was significantly elevated, accompanied by impaired right ventricular reserve and reduced RV-PA coupling efficiency. Six predictors were selected by correlation analysis and LASSO regression: peak heart rate, age, peak tricuspid annular plane systolic excursion to pulmonary artery systolic pressure ratio, ventilatory equivalent for oxygen, maximal oxygen uptake, and change in tricuspid regurgitation velocity. The combined Clinical+Echo+CPET model achieved the best diagnostic performance (AUC: 0.906), with significant improvement in reclassification compared with single- or dual-modality models. Model performance remained consistent across subgroups stratified by resting mPAP, age, and sex. Mechanistic analysis demonstrated strong correlations between selected predictors and invasive haemodynamic parameters. Conclusions A noninvasive multimodal model integrating clinical variables, exercise echocardiography, and cardiopulmonary metabolic parameters enables robust identification of exercise-induced pulmonary hypertension and reflects underlying abnormalities in RV–PA coupling and ventilatory efficiency. This framework may facilitate early detection and risk stratification of occult pulmonary vascular disease.