Validation of the ILD-GAP Model and a Local Nomogram in a Singaporean Cohort

Abstract

Background: The ILD-GAP model was developed and validated in a Western cohort to predict 1-, 2- and 3-year mortality in chronic interstitial lung disease (ILD). Objectives: We aimed to validate the ILD-GAP model and identify predictors of mortality to derive a nomogram to predict mortality in our local Asian population. Methods: Characteristics of patients on follow-up in a tertiary ILD referral center were retrospectively reviewed. Results: There were 181 patients and 48 mortalities. 29.8% had idiopathic pulmonary fibrosis, 2.8% unclassifiable ILD, 33.1% connective tissue disease-associated interstitial lung disease (CTD-ILD), 28.7% idiopathic nonspecific interstitial pneumonia and 5.5% chronic hypersensitivity pneumonitis. Univariable analysis showed that a higher ILD-GAP index, unclassified ILD, males, older age, higher pulmonary artery systolic pressure, lower forced vital capacity percent predicted and carbon monoxide diffusion capacity (DLCO) correlated with increased mortality, and CTD had lower mortality. Multivariable analysis utilizing Akaike’s information criterion stopping rule showed males and a lower DLCO predicted increased mortality, while CTD predicted lower mortality. These were used to generate a nomogram which predicted overall mortality better (C index 0.817, adequacy index 99.5%) than ILD-GAP (C index 0.777, adequacy index 60.7%) and provided superior estimates based on likelihood ratio testing. Calibration plots showed the nomogram predicted 1-year mortality better, whilst the ILD-GAP model predicted 2- and 3-year mortality closer to actual mortality rates but underpredicted 1-year mortality. Conclusion: The nomogram performed better than ILD-GAP in predicting overall mortality and 1-year mortality. Both demonstrated good performance in predicting mortality risk.