Abstract
Background: The Global initiative for chronic Obstructive Lung Disease (GOLD) ABCD groupings were recently modified. The GOLD 2011 guidelines defined increased risk as forced expiratory volume in 1 second (FEV(1)) < 50% predicted or ≥ 2 outpatient or ≥ 1 hospitalized exacerbation in the prior year, whereas the GOLD 2017 guidelines use only exacerbation history. We compared mortality and exacerbation rates in the Genetic Epidemiology of COPD Study cohort (COPDGene(®)) by 2011 (exacerbation history/FEV(1) and dyspnea) versus 2017 (exacerbations and dyspnea) classifications. Methods: Using data from COPDGene(®), we tested associations of ABCD groups with all-cause mortality (Cox models, adjusted for age, sex, race and comorbidities) and longitudinal exacerbations (zero-inflated Poisson models). Results: In 4469 individuals (mean age 63.1 years, 44% female), individual distributions in 2011 versus 2017 systems were: A, 32.0% versus 37.0%; B, 17.6% versus 36.3%; C, 9.4% versus 4.4%; D, 41.0% versus 22.3%; (observed agreement 76% [expected 27.8%], Kappa 0.67, p<0.001). Individuals in group D-2011 had 1.1 ± 1.6 exacerbations/year (mean ± standard deviation [SD]) versus 1.4 ± 1.8 for D-2017 (median follow-up 3.7 years). Using group A as reference, for both systems, mortality (median follow-up 6.8 years) was highest in group D (D-2011, [hazard ratio] HR 5.2 [95% confidence interval (CI) 4.2, 6.4]; D-2017, HR 5.5 [4.5, 6.8]), lowest for group C (HR 1.9 [1.4, 2.6] versus HR 1.9 [1.3, 2.8]) and intermediate for group B (HR 2.6 [2.0, 3.4] versus HR 3.4 [2.8, 4.1]). GOLD 2011 had better mortality discrimination (area under the curve [AUC] 0.68) than GOLD 2017 (AUC 0.66, p<0.001 for comparison) but similar exacerbation rate prediction. Conclusions: Relative to the GOLD 2011 consensus statement, discriminate predictive power of the 2017 ABCD classification is similar for exacerbations but lower for survival.