Prediction and prognostication in interstitial lung disease associated pulmonary hypertension using baseline and longitudinal trends in non-invasive variables

Pulmonary hypertension (PH) commonly occurs in patients with interstitial lung disease (ILD) and is associated with worsening of symptoms and an adverse prognosis. The onset of PH is extremely difficult to predict due to the very similar symptomology of the two conditions and confounding of common screening tests for PH in patients with ILD. It is not clear what invasive and non-invasive variables predict mortality in ILD-PH patients, or whether existing mortality prediction tools used in ILD patients are valid in ILD-PH. I hypothesised that the prediction of PH occurring in ILD patients was possible using non invasive screening tests, and that baseline and longitudinal change in non-invasive variables would predict mortality in ILD-PH patients. The integration of echocardiographic, brain natriuretic peptide (BNP), pulmonary function tests and CT variables showed that prediction of PH occurring in ILD patients was possible, although false positives were common. Echocardiographic variables best correlated with invasive right heart catheter (RHC) pressures. A score was developed to predict severe PH using echocardiographic variables, and was effective even when blinding the most powerful predictor which is commonly unavailable in patients with ILD-PH. CT is commonly employed in suspected ILD-PH patients to exclude co-existent pulmonary emboli and assess parenchymal disease progression. The right ventricle to left ventricle measured at CT pulmonary angiography was superior to both echocardiographic and RHC derived variables at predicting mortality. The presence of PH confounds commonly used mortality prediction tools in ILD. A multi modality mortality prediction model was developed to predict mortality using baseline demographics, lung function and ILD diagnosis. Longitudinal change in pulmonary function tests and BNP were shown to predict mortality. A longitudinal model using demographics and change in gas transfer was developed. External validation of the mortality prediction tools is necessary before its utility is demonstrated.