Does the 2013 GOLD classification improve the ability to predict lung function decline, exacerbations and mortality: a post-hoc analysis of the 4-year UPLIFT trial

BACKGROUND: The 2013 GOLD classification system for COPD distinguishes four stages: A (low symptoms, low exacerbation risk), B (high symptoms, low risk), C (low symptoms, high risk) and D (high symptoms, high risk). Assessment of risk is based on exacerbation history and airflow obstruction, whatever results in a higher risk grouping. The previous system was solely based on airflow obstruction. Earlier studies compared the predictive performance of new and old classification systems with regards to mortality and exacerbations. The objective of this study was to compare the ability of both classifications to predict the number of future (total and severe) exacerbations and mortality in a different patient population, and to add an outcome measure to the comparison: lung function decline.METHODS: Patient-level data from the UPLIFT trial were used to analyze 4-year survival in a Weibull model, with GOLD stages at baseline as covariates. A generalized linear model was used to compare the numbers of exacerbations (total and severe) per stage. Analyses were repeated with stages C and D divided into substages depending on lung function and exacerbation history. Lung function decline was analysed in a repeated measures model.RESULTS: Mortality increased from A to D, but there was no difference between B and C. For the previous GOLD stages 2-4, survival curves were clearly separated. Yearly exacerbation rates were: 0.53, 0.72 and 0.80 for stages 2-4; and 0.35, 0.45, 0.58 and 0.74 for A-D. Annual rates of lung function decline were: 47, 38 and 26 ml for stages 2-4 and 44, 48, 38 and 39 for stages A-D. With regards to model fit, the new system performed worse at predicting mortality and lung function decline, and better at predicting exacerbations. Distinguishing between the sub-stages of high-risk led to substantial improvements.CONCLUSIONS: The new classification system is a modest step towards a phenotype approach. It is probably an improvement for the prediction of exacerbations, but a deterioration for predicting mortality and lung function decline.TRIAL REGISTRATION: ClinicalTrials.gov NCT00144339 (September 2, 2005)