Validation of GAP score in Korean patients with Idiopathic Pulmonary Fibrosis

Abstract

Introduction: The GAP model has been validated in independent cohorts in western countries. However, no study has assessed whether the risk of mortality predicted by GAP model matches the observed mortality in different populations. We evaluated the clinical course of IPF and validated the GAP model in Korean IPF patients.

Methods: We included 268 patients who had been diagnosed with IPF according to established clinical and histologic criteria in Seoul National University Hospital between 2005 and 2009. For each patient, demographics, and lung physiologic parameters such as percent predicted functional vital capacity (FVC), percent predicted carbon monoxide diffusing capacity (DLco) at the diagnosis of IPF were evaluated. The occurrence of respiratory hospitalization, acute exacerbation of IPF, mechanical ventilator care, and death were also evaluated. Finally, we validated the GAP model using discrimination and calibration to predict the risk of death in Korean IPF patients.

Results: The study population consisted of 181 men and 87 women, with a mean age of 65.9 year (SD = 9.6). Mean baseline of percent predicted FVC was 77.8 (SD = 18.8) and percent predicted DLco was 65.9 (SD = 21.7). 54 (20.1%) patients underwent surgical lung biopsy to confirm the diagnosis, and 10 (3.7%) were diagnosed with lung cancer. 157 (58.6%) deaths occurred during the follow-up period, and median time to death was 4.64 years. Observed cumulative mortality at 1, 2, and 3 years were 10.4%, 20.9%, and 31.0%, respectively and cumulative mortality incidence differed substantially among GAP stages (p < 0.001). The GAP model produced estimates of 1-year mortality risk consistent with observed data (c-statistics: GAP calculator 0.74 and GAP index and staging system 0.72, p < 0.29). However, Calibration (c-statistics: GAP calculator 0.68 and GAP index and staging system 0.69) and discrimination (p < 0.001) of GAP model were compromised with under-prediction of 3-year risk of death.

Conclusions: The GAP model did not predict the 3-year risk of death accurately in Korean IPF patients. Further external validation or modification of the GAP model is needed before using it in a clinical setting in Korea.