OBJECTIVES: To estimate associations between attributes of COPD and to develop a model that predicts economic and health outcomes associated with COPD progression. METHODS: We utilized data from ECLIPSE (clinicaltrials.gov identifier: NCT00292552), a three year cohort study of COPD patients to estimate the associations between central COPD attributes (exacerbations, lung function, exercise capacity, and signs/symptoms) while adjusting for co-morbidities, body composition (BMI), biomarkers, smoking history, age, and gender.  As disease progression endpoints we used the total score of the St. George's Respiratory Questionnaire (SGRQ) and mortality. We applied random coefficient models to assess the relationships between the central COPD attributes longitudinally and thereby describe patient trajectories over time. As appropriate, non-linear functional forms were explored to characterize the nature of the data.  Endogeneity among the central attributes of COPD was addressed by time-lagging in the regression models.  RESULTS: Severe exacerbations in the preceding 12-months were associated with an average decline in lung function (FEV) of up to 10 ml (P<0.05) and with a reduced exercise capacity (6 minute walk test) of 13 meters (P<0.0001).  A 1% increase in FEV% predicted was also associated with a 5% reduction in the probability of experiencing dyspnoea on most days/week (P<0.0001).  All central attributes were found to significantly impact disease progression, measured by the SGRQ, with the largest estimated effect for dyspnoea on most days/week (18 point increase in the SGRQ score; P<0.0001). Lung function and exercise capacity, however, were the only central attributes that were significant predictors of mortality (P<0.05). CONCLUSIONS: The use of appropriate analytical techniques to account for the longitudinal nature and endogeneity of COPD attributes enables the estimation of their impact on important health outcomes. Our results confirm the expected associations between the central attributes of COPD and their effect on patient health status (SGRQ) and mortality.