Modeling survival using decision tree analysis in pulmonary hypertension due to left heart disease. the prognostic significance of PVR <3WU when TAPSE <16mm

Abstract

Abstract Aims In 2019, PVR<3 WU was adopted to stratify patients at low risk of pulmonary hypertension due to left heart disease (PH-LHD) as well those with isolated PH-LHD. We sought to evaluate whether the supervised machine learning with Decision Tree analysis that provides more information than Cox Proportional analysis by forming a hierarchy of multiple covariates, confirms this risk stratification. Methods 202 consecutive patients (mean age: 69±11 y, females 42%) with mean pulmonary artery pressure (mPAP)≥20mmHg and wedge pressure>15mmHg were recruited. Transpulmonary pressure gradient ≥12mmHg, pulmonary vascular resistance (PVR) ≥3WU, diastolic pressure gradient ≥7mmHg, pulmonary arterial capacitance<1.1 ml/mmHg, TAPSE<16 mm, peak systolic tissue Doppler velocity<10cm/s and right ventricular end-diastolic area ≥25 cm2 were the seven categorical values to enter the model. To predict the mortality from the Decision Tree, we used the CHAID method. Each node and branch were compared using survival analysis at 6-year follow-up. Results Mean PAP, wedge pressure, cardiac index, and PVR were 40.3±10.0mmHg, 22.3±7.1mm Hg, 2.9±0.8L/min/m2, and 3.6±2.1WU, respectively. Among the seven dichotomous values linked to the prognosis in PH-LHD, only 2 variables entered the model. To predict the mortality, TAPSE was first selected following by PVR. Compared to patients with PVR<3WU and TAPSE ≥16mm, patients with PVR ≥3WU and TAPSE ≥16mm or patients with PVR ≥3WU and TAPSE <16 mm has significant increased mortality (HR=3,0, 95% CI: [1,4–6,4], p=0.006 and HR=3,3, 95% CI: [1,6–6,9], p=0.002, respectively), while patients with PVR <3WU and TAPSE <16 mm exhibiting the worst prognosis (HR=7,2, 95% CI: [3,3–15,9], p=0.0001). Conclusion Used for solving regression and classification problems, decision tree analysis indicates that among 7 prognostic factors, TAPSE and PVR have to be interpreted altogether and simultaneously in PH-LHD for mortality assessment. Therefore, in future research, PVR <3 WU should be understood primarily based on right ventricular systolic function assessed by echocardiography whether TAPSE is or not ≥16 mm. Funding Acknowledgement Type of funding source: None