Local heterogeneity of normal lung parenchyma and small airways disease are associated with COPD severity and progression

Table 2 Multivariable regression for COPD subset

Performance	FEV₁% predicted	FEV₁/FVC	FEF_{25 − 75} (L)	V^Emph
Adjusted R²	0.516	0.602	0.526	0.778
SE	15.8	0.084	0.331	0.057
Age (yrs)	0.085**	0.021 (0.06)	-0.184**	0.035**
Sex (M/F)		0.018 (0.08)	-0.283**	-0.035**
BMI (kg/cm²)	-0.110**	0.033*	0.024**	-0.232**
Smoking (Pack Years)	-0.046**	-0.013 (0.22)	-0.051**	-0.015 (0.06)
CT vendor				0.111**
Race		0.113**	-0.033*
V^Norm	0.727**	0.668**	0.688**	-1.01**
V^fSAD	0.065*		0.138**	-0.408**
χ^Norm		-0.120**	0.134**	0.150**
χ^fSAD	0.106**		0.175**	0.118**

Notes Multivariable regression modelling using volume density (V) and Euler-Poincaré Characteristic (χ) for PRM-derived Normal and fSAD (introduced stepwise) to model pulmonary function testing measures in the COPD subset. Each column presents results for a different regression model. FEV₁, forced expiratory volume in one second; FVC, forced vital capacity; FEF_{25 − 75}, forced mid-expiratory flow; Emph, emphysema; SE, standard error of the estimate; BMI, body mass index; Norm, Normal; fSAD, functional small airways disease. Model performance is reported as adjusted R² and standard error of the estimate. Feature association is reported as standardized beta coefficients (β); cells for stepwise variables removed from final model. All regression models were controlled for age, sex, race, BMI, pack years and CT vendor. P values ≥ 0.01, < 0.01 and ≥ 0.001, and < 0.001 are presented as values in parentheses, *, and **, respectively

ISSN: 1465-993X