Modeling Lung Disease and Its Progression in A Chronic Obstructive Pulmonary Disease Cohort

Abstract

<p>Chronic obstructive pulmonary disease (COPD) is an irreversible, slowly progressive lung disease, usually associated with smoking and with respiratory infections. The objective of the study was to investigate the association of lung function and its progression with smoking, infection and inflammation. I discussed the optimal modeling strategy by comparing results from different methods of analysis and methods of handling missing data.</p> <p>It was well documented that smoking significantly accelerated lung function decline at a rate of 89.6 mL/year and was robust to adjustment for age, gender, height and breathlessness. C. pneumoniae was associated with decreased lung function, but was not statistically significant. Log MMP-9, log ratio of MMP-9 to TIMP-1 and log CRP were strongly associated with slope change in FEV 1 and robust to covariate adjustment in weighted least squares models. Thus, they were good biomarkers for COPD progression.</p> <p>There were no significant differences among generalized estimating equations, mixed-effects and robust random-effects models for measuring lung function at multiple-time. Smoking was positively related to lung function, but was not robust to covariate adjustment. C. pneumoniae was not significantly correlated with lung function. Log MMP-9 and log ratio of MMP-9 to TIMP-1 were associated with decreased lung function, but were not statistically significant. However, log CRP was significantly associated with lung function and robust to covariate adjustment. Thus, log CRP was the best biomarker for modeling lung function measured at multiple- time.</p> <p>The GEE modeling results of different data sets imputed by group mean estimation, last observation carried forward, hot-deck and multiple imputation methods were consistent, but multiple imputation was recommended to handle the missing data.</p>