Pulmonary function and sputum characteristics predict CT phenotype and severity of COPD

Abstract

Airway obstruction and parenchymal destruction underlie COPD phenotype and severity. We aimed at predicting by clinical and pulmonary function data the predominant type and the severity of the pathologic changes quantitatively assessed by CT.

Airway wall thickness (AWT-Pi10) and percentage of low attenuation area (%LAA-950) were measured in 100 (learning set) of 473 COPD outpatients undergoing clinical and functional evaluation. Original CT measurements were translated by principal components analysis onto a plane with novel coordinates CT1 and CT2 depending, respectively, on the difference (prevalent mechanism of airflow limitation) and on the sum (severity) of AWT-Pi10 and %LAA-950. CT1 and CT2 estimated in the learning set by cross-validated models of clinical and functional variables were used to classify 373 patients in the testing set.

A model based on DL,CO %, TLC%, purulent sputum predicted CT1 (r=0.64; p<.01). A model based on FEV1/VC, FRC%, purulent sputum predicted CT2 (r=0.77; p<.01). Classification of patients of the testing set obtained by models-predicted CT1 and CT2 reflected, according to correlations with clinical and functional variables, both COPD phenotype and severity.

Multivariate models based on pulmonary function variables and sputum purulence classify patients according to overall severity and predominant phenotype of COPD as assessed quantitatively by CT.