Abstract 4143964: Combined electrocardiography artificial intelligence and echocardiography model to characterize precapillary versus postcapillary pulmonary hypertension in end stage chronic kidney disease.

Introduction: Pulmonary hypertension (PHTN) is common in patients with chronic kidney disease (CKD), however categorization as precapillary or postcapillary is challenging due to the need for invasive right heart catheterization (RHC). Studies have validated transthoracic echocardiography (TTE) to characterize PHTN and ECG-artificial intelligence (AI) to predict diastolic dysfunction risk. The utility of these noninvasive tools in patients with CKD is unknown. Methods: Retrospective analysis of patients with end stage CKD who underwent RHC between 2011 and 2021. Patients with PHTN (pulmonary artery pressure >20mmHg on RHC) were categorized as having precapillary (pulmonary capillary wedge pressure, PCWP, ≤15mmHg and pulmonary vascular resistance, PVR, >2WU) or postcapillary PHTN (PCWP >15mmHg and PVR ≤2WU for isolated postcapillary or PVR >2WU for combined). TTE and ECG within 3 months of RHC were obtained. TTE measurements were used to calculate ePLAR (echocardiographic pulmonary to left atrial ratio); tricuspid regurgitant peak velocity (TRV max ) divided by mitral E/e’. ECGs were analyzed with an AI risk tool for diastolic dysfunction (score 0-3, increasing with increasing risk). Receiver operating curves (ROC) evaluated the performance of these tests. Results: 60 patients were included; 16 (27%) with precapillary and 44 (73%) with postcapillary PHTN. ROC analysis found good accuracy for both ECG-AI diastolic dysfunction risk score to predict precapillary PHTN with AUC 0.724 (sensitivity=60% and specificity=88% for an ECG score value ≤1) and ePLAR with AUC 0.715 (sensitivity=75% and specificity=70% for ePLAR >0.17m/sec). Highest accuracy for detection of precapillary PHTN was reached by combination of ePLAR and ECG-AI tool, with AUC 0.756 (sensitivity=73%, accuracy of 86% for a value ≤1.17). Conclusions: ECG-AI generated diastolic risk score of 0-1 (low risk) and high ePLAR showed good accuracy in characterizing precapillary PHTN in patients with end stage CKD. The combination of these noninvasive tools further increased the diagnostic accuracy, reflecting its potential utility in classification of patients with CKD and PHTN, a crucial determinant of the ideal therapeutic plan in kidney transplant candidates.