Table 1

Associations of k-means clusters with absolute and relative annualised FEV₁ and emphysema changes

	K-means cluster	Univariable models		Multivariable models
	K-means cluster	Beta (SE)	P value	Beta (SE)	Beta (SE)
Absolute annualised FEV₁ (mL/year) changes	ULE	3.07 (2.21)	0.2	3.63 (2.23)	0.1
	AD	11.64 (1.86)	<0.0001	8.84 (1.90)	<0.0001
	SE	2.84 (2.17)	0.2	4.98 (2.25)	0.03
Relative annualised FEV₁ changes (percent changes from baseline)	ULE	−0.27 (0.11)	0.01	−0.18 (0.11)	0.1
	AD	0.09 (0.09)	0.3	0.004 (0.09)	1.0
	SE	−1.35 (0.10)	<0.0001	−1.24 (0.11)	<0.0001
Absolute annualised emphysema (adjusted Perc15 density) changes	ULE	−0.99 (0.11)	<0.0001	−0.89 (0.11)	<0.0001
	AD	−0.69 (0.09)	<0.0001	−0.82 (0.09)	<0.0001
	SE	−0.89 (0.11)	<0.0001	−0.77 (0.11)	<0.0001
Relative annualised emphysema (adjusted Perc15 density) changes (percent changes from baseline)	ULE	−1.15 (0.16)	<0.0001	−1.01 (0.16)	<0.0001
	AD	−0.82 (0.13)	<0.0001	−0.99 (0.14)	<0.0001
	SE	−1.19 (0.16)	<0.0001	−1.07 (0.16)	<0.0001

Absolute annualised changes were computed by subtracting Visit 1 values from Visit 2 values and dividing by the time in years between both visits for each subject. Relative annualised changes were calculated by dividing absolute annualised changes by Visit 1 values and multiplying by 100. Negative values indicate worsening of the disease between visits. Univariable linear regression models included only Visit 1 k-means cluster assignment. Multivariable models also included adjustments for age, sex, race, BMI and smoking pack-years. The reference group was the RSS cluster. P<0.05 are italicised.
AD, airway-predominant disease; BMI, body mass index; FEV₁, forced expiratory volume in 1 s; Perc15, 15th percentile; RSS, relatively resistant smoker; SE, severe emphysema; ULE, upper lobe-predominant emphysema.