Skip to main content

Table 4 Multivariate regression analysis for longitudinal change in a CT index

From: Fractal analysis of low attenuation clusters on computed tomography in chronic obstructive pulmonary disease

Dependent variable

Independent variables

Current smoker (Ref. former smoker)

ΔCT-TLV

Baseline CT index

ΔLAA%

β* = 0.08

p = 0.68

β* = 0.72

p < 0.0005

β* = 0.14

p = 0.14

ΔD950

β* = − 0.16

p = 0.58

β* = − 0.34

p = 0.009

β* = 0.01

p = 0.92

ΔD’15

β* = − 0.59

p = 0.03

β* = 0.31

p = 0.01

β* = − 0.19

p = 0.12

ΔD’25

β* = − 0.71

p = 0.01

β* = 0.27

p = 0.03

β* = − 0.05

p = 0.68

ΔD’35

β* = 0.06

p = 0.84

β* = 0.41

p = 0.002

β* = − 0.07

p = 0.57

  1. Five models were constructed. Each model included a longitudinal change in either LAA%, D950, D’15, D’25, or D’35 between baseline and follow-up CT scans (ΔLAA%, ΔD950, ΔD’15, ΔD’25, or ΔD’35) as a dependent variable. Smoking status (current vs former smoker), a longitudinal change in CT-derived total lung volume (ΔCT-TLV), and a corresponding baseline CT index (for example, the model for ΔLAA% used baseline LAA%) were included as independent variables for all models. D950, D’15, D’25, and D’35 are exponents that characterize a power law that governs the cumulative frequency size distribution of LAA clusters that are identified using a threshold of − 950 HU and the 15th, 25th, and 35th percentiles of a CT density histogram, respectively