Diagnosing pulmonary hypertension due to left heart disease using diastolic echo markers: The National Echo Database of Australia (NEDA) PH-LHD predictive formula

Aims: Pulmonary hypertension (PH) is commonly due to left heart disease caused by ischaemic heart disease, hypertension and valvular heart disease. It is under diagnosed and associated with a high mortality. PH diagnosed using echo requires a measurable tricuspid regurgitation velocity (TRV) to estimate the pulmonary artery systolic pressure (PH = PASP \u3e40mmHg). However, up to 40% of studies have insufficient TRV to calculate a PASP. This can lead to significant delays in the diagnosis of pulmonary hypertension, increased morbidity and delays in the initiation of treatment. This thesis seeks to determine the prevalence of PH and the diastolic echo markers related to the development of PH in left heart disease (PH-LHD) and create a predictive model using diastolic echo markers to diagnose PH in the absence of a TRV. Methods: This study is a retrospective observational cohort study with data derived from the National Echo Database of Australia (NEDA). Using PH as the dependent variable and markers of diastolic function as the independent variables we performed univariate and multivariate analysis on the entire cohort to identify predictive diastolic markers that correlates with PH. To create a predictive formula to diagnose PH-LHD, the entire cohort was randomised 1:1 into a development (DD) and validation database (VD). Using logistic regression analysis on diastolic markers and the presence of PH in the DD, we derived a constant (con) that could be used to predict the probability of PH. Using probability analysis, the Receiver Operating Characteristic (ROC) curve was generated using a 0.5 cut off to evaluate the accuracy of the model. The accuracy of the model was then tested using the VD. Results: Of the 174,229 patients in the NEDA, 75,204 (43%) had insufficient TRV to calculate a PASP. Of the 99,025 patients with a PASP, 19,767 (20%) had PH. Patients with PH were older (76 vs 62 yr) (p = The DD (150,979 echos) had 5,181 valid studies to create the NEDA PH-LHD Constant (Con) = -6.649 + (0.035 x Age) + (0.072 x E’) + (0.077 x E/E’) + (0.509 x E/A) + (0.03 x LAVI) to predict the probability of PH. The DD model AUC ROC is 75% accurate in 14 diagnosing PH-LHD. Applying our formula to the VD (151,767 echos), the AUC of the ROC curve is 0.742. Conclusion: Using the NEDA, 20% of patients were diagnosed with PH. Using Age, E’, E/E’ ratio, E/A ratio and LAVI, the NEDA PH-LHD formula can diagnose PH-LHD in 75% of cases in the absence of TRV