Reduced Total Airway Count and Airway Wall Tapering after Three-Years in Ex-Smokers

Abstract

Computed tomography (CT) total-airway-count (TAC) and airway wall-thickness differ across chronic obstructive pulmonary disease (COPD) severities, but longitudinal insights are lacking. The aim of this study was to evaluate longitudinal CT airway measurements over three-years in ex-smokers. In this prospective convenience sample study, ex-smokers with (n = 50; 13 female; age = 70 ± 9 years; pack-years = 43 ± 26) and without (n = 40; 17 female; age = 69 ± 10 years; pack-years = 31 ± 17) COPD completed CT, ³He magnetic resonance imaging (MRI), and pulmonary function tests at baseline and three-year follow-up. CT TAC, airway wall-area (WA), lumen-area (LA), and wall-area percent (WA%) were generated. Emphysema was quantified as the relative-area-of-the-lung with attenuation < –950 Hounsfield-units (RA₉₅₀). MRI ventilation-defect-percent (VDP) was also quantified. Differences over time were evaluated using paired-samples t tests. Multivariable prediction models using the backwards approach were generated. After three-years, forced-expiratory-volume in 1-second (FEV₁) was not different in ex-smokers with (p = 0.4) and without (p = 0.5) COPD, whereas RA₉₅₀ was (p < 0.001, p = 0.02, respectively). In ex-smokers without COPD, there was no change in TAC (p = 0.2); however, LA (p = 0.009) and WA% (p = 0.01) were significantly different. In ex-smokers with COPD, TAC (p < 0.001), WA (p = 0.04), LA (p < 0.001), and WA% (p < 0.001) were significantly different. In all ex-smokers, TAC was related to VDP (baseline: ρ = –0.30, p = 0.005; follow-up: ρ = –0.33, p = 0.002). In significant multivariable models, baseline airway wall-thickness was predictive of TAC worsening. After three-years, in the absence of FEV₁ worsening, TAC diminished only in ex-smokers with COPD and airway walls were thinner in all ex-smokers. These longitudinal findings suggest that the evaluation of CT airway remodeling may be a useful clinical tool for predicting disease progression and managing COPD.

Clinical trial registration: www.clinicaltrials.gov NCT02279329

Keywords:

• Chronic obstructive pulmonary disease
• ex-smokers
• computed tomography
• disease progression
• airways disease
• longitudinal measurements

Introduction

Chronic obstructive pulmonary disease (COPD) is characterized by chronic airflow obstruction [1], parenchymal lung destruction, and airway abnormalities [2]. For decades, spirometry has provided the key clinical measurements for COPD diagnosis and management [1]. However, spirometry measurements made at the mouth cannot estimate regional airway or parenchymal abnormalities, especially potential abnormalities in the distal, small airways, where disease onset and progression are hypothesized to initiate [3].

Thoracic computed tomography (CT) serves as the clinical mainstay for COPD imaging and imaging phenotypes, including terminal airspace enlargement or emphysema [4] and airways disease [5] and has been utilized in numerous cohort studies, including Multi-Ethnic Study of Atherosclerosis (MESA) [6], Evaluation of COPD Longitudinally to Identify Predictive Surrogate End-points (ECLIPSE) [7], COPD Genetic Epidemiology (COPDGene) [8], SubPopulations and InteRmediate Outcome Measures In COPD (SPIROMICS) [9], and Canadian Cohort Obstructive Lung Disease (CanCOLD) [10]. Segmentation of the CT airway tree from the trachea down to approximately the 6th–10th generation airways provides a way to measure total airway count (TAC) and airway wall and lumen dimensions. CT airway abnormalities may be quantified using a number of different algorithms [11–14] and these abnormalities have been shown to be related to respiratory symptoms [15] and to reflect small airway dimensions measured using histo-pathological approaches [16].

Previous work demonstrated that CT TAC may be quantified in patients with COPD [17]. In the CanCOLD study, TAC was observed to be diminished across COPD grade severity [17] and was associated with the number of terminal bronchioles measured with micro-CT [18], while predictive of incident COPD in at-risk ever-smokers [19]. Airway wall dimension differences were also observed across COPD grade severity in the CanCOLD, SPIROMICS, and MESA cohorts [17, 20] and these were spatially related to missing airways [17]. Furthermore, longitudinal worsening of CT emphysema and air trapping measurements were observed in the COPDGene study [21].

While all of this important information about airways disease and emphysema provide a way to understand differences in patients with different COPD severities, it is still unclear how TAC and airway wall thickness change over time in individual patients or within disease severity subgroups. Longitudinal changes that may occur in the airways of ex-smokers, especially in those with normal pulmonary function, may provide insights into COPD initiation and progression. Longitudinal Thoracic Imaging Network of Canada (TINCan) ³He magnetic resonance imaging (MRI) [22–24] and small vessel density [25] findings were previously described. In a preliminary evaluation, MRI ventilation defect percent (VDP) and apparent diffusion coefficients were measured in 15 participants with COPD [23] and revealed significant worsening after two-years in the absence of forced expiratory volume in 1-second (FEV₁) changes. In another evaluation [24], the power of baseline MRI measurements to predict changes in quality-of-life and spirometry measurements after 30 months was described. Finally, significant CT pulmonary vascular tree measurement changes were described [25] in 90 ex-smokers in whom there was accelerated FEV₁ decline after three-years. Based on cross-sectional findings in a number of cohort studies [17–20], here we hypothesized that, even in the absence of FEV₁ worsening, CT airway measurements would significantly worsen in ex-smokers with and without spirometry or other evidence of COPD. Hence, our primary objective was to evaluate longitudinal CT airway measurements in a relatively large group of ex-smokers in the TINCan cohort study [22], with both CT and hyperpolarized MR imaging acquired at baseline and after three-years.

Methods

Study participants and design

All participants provided written informed consent to an ethics-board (Institutional Ethics Board no. 00000984), Health Canada approved and registered (ClinicalTrials.gov: NCT02279329) protocol. Inclusion criteria included a history of combustible tobacco cigarette smoking ≥10 pack-years and age of 50–85 years at baseline, while exclusion criteria consisted of claustrophobia and any contraindications for MRI or CT and current smokers. Ex-smokers were included who ceased smoking at least one-year prior to the study visit with no maximum cutoff for pack-years. COPD was defined as post-bronchodilator spirometry according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria [1].

This study was prospectively planned and participants for this analysis were enrolled from January 2010 to July 2016. All participants underwent two, 2-h visits for pulmonary function tests, quality-of-life questionnaires, six-minute walk test, MRI, and CT [22]. Follow-up was prospectively planned for 24 ± 6 months after the baseline visit. Novo-Salbutamol HFA was administered in all participants using a metered dose inhaler (four doses of 100 μg, Teva Novopharm Ltd., Toronto, Ontario, Canada) through a spacer (AeroChamber Plus spacer, Trudell Medical International, London, Ontario, Canada). Twenty minutes post-salbutamol, pulmonary function testing, MRI, and CT were performed. The mean elimination half-life of salbutamol is 4–6 h [26], therefore all evaluations were completed within 2-h post-administration. Pulse oximetry estimation of arterial blood oxygen saturation on a fingertip was measured using a Nonin 8500 Series Handheld Pulse Oximeter (Nonin Medical, Inc., Plymouth, Minnesota, USA) upon participant arrival.

Pulmonary function tests and questionnaires

Spirometry, plethysmography, and measurement of the diffusing capacity of the lung for carbon monoxide (DL_CO) were performed according to American Thoracic Society/European Respiratory Society guidelines [27–29] using a body plethysmograph (MedGraphics Elite Series, MGC Diagnostic Corporation, St. Paul, Minnesota, USA) with an attached gas analyzer. St. George’s Respiratory Questionnaire (SGRQ) was administered [30], and a six-minute walk test was also performed [31], under supervision of trained personnel.

Image acquisition

Anatomic ¹H and hyperpolarized ³He ventilation MRI were acquired on a 3 T Discovery MR750 system (GE Healthcare, Milwaukee, Wisconsin, USA) for both baseline and follow-up imaging, as previously described [32]. Anatomic ¹H MRI was acquired using a whole-body radiofrequency coil and a fast-spoiled gradient-recalled echo sequence during inspiration breath-hold from functional residual capacity, as previously described [32]. ³He MRI was acquired using a single-channel rigid elliptical transmit-receive chest coil (RAPID Biomedical, Wuerzburg, Germany) and a fast-spoiled gradient-recalled echo sequence during inspiration breath-hold, as previously described [32].

Within 30 min of MRI, CT was acquired on a 64-slice Lightspeed VCT scanner (GE Healthcare, Milwaukee, Wisconsin, USA) under breath-hold after inhalation of 1 L of N₂ from functional residual capacity for both baseline and follow-up imaging, as previously described [22]. CT images were acquired with the following parameters: beam collimation of 64 × 0.625 mm, 120 kVp, effective mAs of 100, 500 ms tube rotation time, 1.25 pitch, and image reconstruction with a standard convolution kernel to 1.25 mm. Total effective dose was estimated as 1.8 mSv using the ImPACT CT patient dosimetry calculator (based on Health Protection Agency [UK] NRBP-SR250).

Image analysis

³He MRI VDP was measured using a semi-automated segmentation pipeline generated in MATLAB R2019a (MathWorks, Natick, Massachusetts, USA), as previously described [33]. Thoracic CT images at baseline and follow-up were analyzed using VIDAvision2.2 software (VIDA Diagnostics Inc., Coralville, Iowa, USA) to semi-automatically segment the lungs and airway tree by a single trained observer who was blinded to clinical information (PVW, 2 years of experience). As previously described [17], all airway segments in the segmented airway tree were summed to quantify TAC. Anatomically equivalent segmental, subsegmental, and sub-subsegmental airways for all airway paths (third to fifth generation) [20] were used to generate airway lumen area (LA), wall area (WA), wall area percent (WA%), and wall thickness percent (WT%). Emphysema was quantified using the relative area of the segmented lung with attenuation values less than –950 Hounsfield units (RA₉₅₀) and was considered present when RA₉₅₀ was greater than 6.8%, as previously described [4].

Statistics

Statistics were generated using SPSS (ver. 28; IBM Statistics, Armonk, New York, USA). Data were tested for normality using Shapiro–Wilk tests and non-parametric tests were performed when data were not normally distributed. Independent samples t tests and analysis of variance were used to determine significance of between-group differences. Paired samples t tests were used to determine significance between time points. Univariate relationships were evaluated using Pearson (r) correlations for normally distributed variables and Spearman (ρ) correlations for non-normally distributed variables. Variables with significant correlation p values at baseline were used to generate multivariable models, where significant variables included in the model were chosen using the backward approach, to predict TAC at follow-up and the longitudinal change in TAC. The removal criterion for the backwards method included variables with a probability of F ≥ 0.10. Multicollinearity among variables in the multivariable regression models was evaluated using the variance inflation factor and deemed acceptable when less than 10 [34]. Uncorrected p values were reported and results were considered statistically significant when the probability of making a type I error was less than 5% (p < 0.05).

Results

A CONSORT diagram for the TINCan cohort study provided in Figure 1 shows that 266 participants were enrolled and 172 participants completed all imaging examinations at baseline. Of those participants who did not complete Visit 1, some withdrew after consent (n = 43), some were enrolled in a sub-study of oscillatory positive expiratory pressure (n = 33, [35]), and some could not complete imaging measurements (n = 18) due to claustrophobia, poor coil fit, or radiation dose concerns. After 31 ± 7 months, 90 participants returned for a complete follow-up Visit 2. Of those who were excluded from Visit 2 analysis, 64 were lost to follow-up, 14 were deceased, and four had CT data that were not evaluable because of motion artifacts. Supplementary Table S1 shows baseline measurements for ex-smokers who did not return for follow-up (n = 82) compared to those who did (n = 90). Those who did not return for follow-up reported significantly worse body mass index (BMI) (27 kg/m² vs. 28 kg/m²; p = 0.03), FEV₁ (68% vs. 84%; p < 0.001), forced vital capacity (FVC) (84% vs. 95%; p = 0.003), FEV₁/FVC (59% vs. 65%; p = 0.02), residual volume (RV) (152% vs. 127%; p < 0.001), RV to total lung capacity (TLC) ratio (51% vs. 44%; p < 0.001), DL_CO (59% vs. 67%; p = 0.02), six-minute walk distance (6MWD) (366 m vs. 400 m; p = 0.02), and SGRQ (40 vs. 29; p = 0.003) as compared to those that returned.

Figure 1. CONSORT diagram for the Thoracic Imaging Network of Canada (TINCan) cohort study. Of the 266 enrolled, 172 ex-smokers with and without COPD completed the baseline visit (Visit 1), while 90 ex-smokers completed the follow-up visit (Visit 2) and were included in the analysis.

Demographics

Table 1 shows demographics for all 90 participants including ex-smokers with spirometry evidence of COPD (n = 50; age, 70 ± 9 years [mean ± SD]; 37 males, 13 females) and ex-smokers without COPD (n = 40; age, 69 ± 10 years [mean ± SD]; 23 males, 17 females). As shown in Table 1, after three-years, there was a significant but small difference in resting oxygen saturation (95% vs. 94%; p = 0.01) in ex-smokers with COPD, but not in ex-smokers without COPD. Table 2 provides participant demographics by GOLD grade severity and shows that there was a significant but small difference in resting oxygen saturation (95% vs. 94%; p = 0.003) only in ex-smokers with GOLD II grade COPD.

Table 1. Participant demographics at baseline and three-year follow-up for ex-smokers with and without COPD.

	All Ex-smokers n = 90			Without COPD n = 40			With COPD n = 50
Parameter	Baseline	Follow-up	p	Baseline	Follow-up	p	Baseline	Follow-up	p
Age (year)	70 (9)	72 (9)	ND	69 (10)	72 (10)	ND	70 (9)	73 (9)	ND
Female [n (%)]	30 (33)	30 (33)	ND	17 (43)	17 (43)	ND	13 (26)	13 (26)	ND
BMI (kg/m^2)	28 (4)	28 (5)	0.6	30 (4)	30 (4)	0.6	27 (4)	26 (4)	0.1
SaO2 (%)	95 (3)	95 (3)	0.9	95 (5)	96 (2)	0.9	95 (2)	94 (3)	0.01
Pack-years	38 (23)	38 (23)	ND	31 (17)	31 (17)	ND	43 (26)	43 (26)	ND
Follow-up time (m)	–	31 (7)	ND	–	31 (5)	ND	–	31 (8)	ND

Values are reported as mean (SD) unless otherwise indicated. p values represent significance values for paired samples and paired proportions t tests. Bolded values are statistically significant.

BMI, body mass index; COPD, chronic obstructive pulmonary disease; GOLD, Global Initiative for Chronic Obstructive Lung Disease; ND, not done; SaO₂, oxygen saturation.

Table 2. Participant demographics at baseline and three-year follow-up for ex-smokers with COPD according to GOLD grade.

	GOLD I n = 16			GOLD II n = 24			GOLD III/IV n = 10
Parameter	Baseline	Follow-up	p	Baseline	Follow-up	p	Baseline	Follow-up	p
Age (year)	74 (8)	76 (8)	ND	68 (8)	70 (8)	ND	69 (10)	71 (10)	ND
Female [n (%)]	1 (6)	1 (6)	ND	9 (38)	9 (38)	ND	3 (30)	3 (30)	ND
BMI (kg/m^2)	28 (4)	28 (4)	0.3	26 (4)	26 (3)	0.7	27 (6)	26 (5)	0.1
SaO2 (%)	96 (1)	95 (3)	0.3	95 (3)	94 (3)	0.003	95 (2)	95 (3)	0.9
Pack-years	37 (26)	37 (26)	ND	44 (20)	44 (20)	ND	51 (37)	51 (37)	ND
Follow-up time (m)	–	29 (4)	ND	–	33 (9)	ND	–	29 (8)	ND

Values are reported as mean (SD) unless otherwise indicated. p values represent uncorrected significance values for paired samples t tests. Bolded values are statistically significant.

BMI, body mass index; COPD, chronic obstructive pulmonary disease; GOLD, Global Initiative for Chronic Obstructive Lung Disease; ND, not done; SaO₂, oxygen saturation.

Figure 2 shows COPD grade severity progression at follow-up for all participants. Of the 40 ex-smoker participants at baseline, two transitioned to GOLD I at follow-up and one transitioned to GOLD II. The ex-smoker who transitioned to GOLD II had an abnormal FEV₁ (60%) and a preserved ratio of FEV₁ to FVC (77%) at baseline, which was abnormal (63%) at follow-up. Of the 24 GOLD II participants, five transitioned to GOLD III. Of the 10 GOLD III participants, two transitioned to GOLD IV.

Figure 2. Arrows showing progression in severity at follow-up for all participants (n = 90), including ex-smokers without COPD (n = 40) and ex-smokers with COPD (n = 50) within GOLD grade severity subgroups. Of the 40 ex-smoker participants at baseline, two transitioned to GOLD I at follow-up and one transitioned to GOLD II. Of the 24 GOLD II participants, five transitioned to GOLD III. Of the 10 GOLD III participants, two transitioned to GOLD IV.

Longitudinal pulmonary function and imaging measurements

Representative CT emphysema (threshold of –950 Hounsfield units is shown in yellow) and segmented airway tree images at baseline and three-year follow-up for ex-smokers with and without COPD are shown in Figure 3. Based on a qualitative visual inspection of the images and segmented airway trees, in the ex-smoker without COPD, there was no CT evidence of emphysema at both time points, nor diminished TAC at follow-up. In the ex-smoker with GOLD I grade COPD, there was small evidence of emphysema at both time points. In the ex-smoker with GOLD II grade COPD, emphysema was visually obvious in the lower lobes at baseline, whereas at follow-up, emphysema extent was visibly augmented in the left lower lobe and TAC was visually diminished. In the ex-smoker with GOLD III grade COPD, there was widespread emphysema in both upper and lower lobes that was visually obviously increased at follow-up. In addition, TAC was substantially diminished at follow-up.

Figure 3. Baseline and three-year follow-up CT imaging for representative ex-smokers with and without COPD. P37 is a 70-year-old female ex-smoker without COPD, with follow-up time = 31 months (baseline/follow-up: FEV₁%_pred = 93%/93%; RA₉₅₀ = 1%/1%; TAC = 306/297). P10 is a 75-year-old male ex-smoker with GOLD I COPD, with follow-up time = 33 months (baseline/follow-up: FEV₁%_pred = 92%/84%; RA₉₅₀ = 7%/8%; TAC = 265/231). P74 is an 83-year-old male ex-smoker with GOLD II COPD, with follow-up time = 28 months (baseline/follow-up: FEV₁%_pred = 57%/52%; RA₉₅₀ = 20%/23%; TAC = 258/191). P78 is a 67-year-old female ex-smoker with GOLD III COPD, with follow-up time = 26 months (baseline/follow-up: FEV₁%_pred = 37%/33%; RA₉₅₀ = 33%/37%; TAC = 206/174). Left: Coronal CT reconstruction with RA₉₅₀ shown in yellow. Right: Three-dimensional reconstruction of the segmented airway tree.

Table 3 shows pulmonary function, exercise capacity, quality-of-life, and imaging measurements at baseline and three-year follow-up; similar information is provided by GOLD grade severity in Table 4. In ex-smokers without COPD, after three-years, there were significant differences in FEV₁/FVC (81% vs. 79%; p = 0.01), TLC (102% vs. 97%; p < 0.001), DL_CO (79% vs. 87%; p < 0.001), and 6MWD (420 m vs. 407 m; p = 0.049). In addition, RA₉₅₀ (1.5% vs. 2.0%; p = 0.02), airway LA (16.1 mm² vs. 17.0 mm²; p = 0.009), WA% (80.4% vs. 79.5%; p = 0.01), and VDP (6.4% vs. 8.6%; p = 0.002) were different, three-years after the baseline visit. In ex-smokers with COPD, there was a significant difference in FVC (95% vs. 92%; p = 0.04), 6MWD (405 m vs. 387 m; p = 0.049), total lung volume (TLV) (5.8 L vs. 6.1 L; p = 0.001), RA₉₅₀ (9.2% vs. 12.5%; p < 0.001), TAC (241 vs. 217; p < 0.001), airway WA (67.2 mm² vs. 66.9 mm²; p = 0.04), LA (13.3 mm² vs. 14.5 mm²; p < 0.001), WA% (83.5% vs. 82.3%; p < 0.001), WT% (18.0% vs. 17.7%; p < 0.001), and VDP (16.8% vs. 21.3%; p < 0.001).

Table 3. Pulmonary function, exercise capacity, quality-of-life, and imaging measurements at baseline and three-year follow-up for ex-smokers with and without COPD.

	All Ex-smokers n = 90			Without COPD n = 40			With COPD n = 50
Parameter	Baseline	Follow-up	p	Baseline	Follow-up	p	Baseline	Follow-up	p
Lung function
FEV1 (%pred)	84 (27)	84 (30)	0.9	103 (18)	104 (20)	0.5	69 (23)	68 (26)	0.4
FVC (%pred)	95 (17)	94 (20)	0.4	95 (17)	97 (18)	0.2	95 (18)	92 (21)	0.04
FEV1/FVC (%)	65 (17)	64 (16)	0.1	81 (6)	79 (7)	0.01	53 (11)	53 (12)	0.6
RV* (%pred)	126 (40)	126 (40)	0.9	106 (20)	103 (20)	0.4	143 (44)	145 (42)	0.6
TLC* (%pred)	109 (16)	104 (20)	0.01	102 (12)	97 (2)	<0.001	114 (17)	112 (16)	0.4
RV/TLC* (%)	44 (10)	45 (12)	0.4	40 (8)	42 (8)	0.07	47 (11)	48 (12)	0.4
DLCO* (%pred)	68 (21)	72 (25)	0.003	79 (15)	87 (17)	<0.001	60 (20)	60 (24)	0.4
Exercise and QoL
6MWD* (m)	412 (76)	396 (85)	0.006	420 (81)	407 (77)	0.049	405 (72)	387 (91)	0.049
SGRQ*	29 (20)	29 (21)	0.9	22 (20)	21 (19)	0.7	65 (19)	63 (21)	0.7
CT
TLV (L)	5.3 (1.2)	5.5 (1.3)	0.007	4.7 (1.0)	4.8 (1.1)	0.5	5.8 (1.0)	6.1 (1.2)	0.001
RA950 (%)	5.8 (7.6)	7.8 (8.9)	<0.001	1.5 (1.4)	2.0 (1.5)	0.02	9.2 (8.8)	12.5 (9.7)	<0.001
TAC (n)	270 (81)	252 (78)	<0.001	306 (87)	297 (82)	0.2	241 (63)	217 (52)	<0.001
WA (mm^2)	66.6 (1.9)	66.3 (1.7)	0.02	65.8 (2.0)	65.5 (1.7)	0.2	67.2 (1.5)	66.9 (1.5)	0.04
LA (mm^2)	14.6 (3.6)	15.6 (3.6)	<0.001	16.1 (4.0)	17.0 (3.7)	0.009	13.3 (2.7)	14.5 (3.1)	<0.001
WA% (%)	82.2 (4.0)	81.0 (3.9)	<0.001	80.4 (4.4)	79.5 (3.9)	0.01	83.5 (3.0)	82.3 (3.4)	<0.001
WT% (%)	17.8 (0.8)	17.6 (0.8)	<0.001	17.6 (0.8)	17.5 (0.7)	0.07	18.0 (0.7)	17.7 (0.8)	<0.001
MRI
VDP* (%)	12.2 (9.4)	15.8 (11.9)	<0.001	6.4 (3.6)	8.6 (4.8)	0.002	16.8 (10.1)	21.3 (12.9)	<0.001

Values are reported as mean (SD) unless otherwise indicated. p values represent significance values for paired samples t tests. Bolded values are statistically significant.

%_pred, percent of predicted value; 6MWD, six-minute walk distance; COPD, chronic obstructive pulmonary disease; CT, computed tomography; DL_CO, diffusing capacity of the lung for carbon monoxide; FEV₁, forced expiratory volume in 1-second; FVC, forced vital capacity; LA, lumen area; MRI, magnetic resonance imaging; QoL, quality-of-life; RA₉₅₀, relative area of the lung with attenuation less than –950 Hounsfield units; RV, residual volume; SGRQ, St. George’s Respiratory Questionnaire; TAC, total airway count; TLC, total lung capacity; TLV, total lung volume; VDP, ventilation defect percent; WA, wall area; WA%, wall area percent; WT%, wall thickness percent.

RV*, TLC*, RV/TLC*: All ex-smokers n = 89, With COPD n = 49.

DL_CO*: All ex-smokers n = 84, Without COPD n = 37, With COPD n = 47.

6MWD*: All ex-smokers n = 82, Without COPD n = 36, With COPD n = 46.

SGRQ*: All ex-smokers n = 85, Without COPD n = 36, With COPD n = 49.

VDP*: All ex-smokers n = 84, Without COPD n = 37, With COPD n = 47.

Table 4. Pulmonary function, exercise capacity, quality-of-life, and imaging measurements at baseline and three-year follow-up for ex-smokers with COPD according to GOLD grade.

	GOLD I n = 16			GOLD II n = 24			GOLD III/IV n = 10
Parameter	Baseline	Follow-up	p	Baseline	Follow-up	p	Baseline	Follow-up	p
Lung function
FEV1 (%pred)	97 (12)	99 (12)	0.4	62 (8)	59 (13)	0.08	39 (8)	37 (8)	0.2
FVC (%pred)	110 (13)	112 (14)	0.6	94 (13)	89 (12)	0.02	74 (10)	66 (16)	0.1
FEV1/FVC (%)	64 (4)	64 (5)	0.8	50 (9)	50 (10)	0.6	40 (7)	42 (9)	0.3
RV* (%pred)	117 (30)	117 (20)	1.0	138 (28)	144 (32)	0.2	204 (46)	178 (76)	0.3
TLC* (%pred)	110 (13)	109 (11)	0.6	112 (16)	112 (18)	1.0	125 (22)	108 (42)	0.2
RV/TLC* (%)	39 (8)	41 (6)	0.4	47 (9)	49 (8)	0.1	60 (8)	52 (26)	0.3
DLCO* (%pred)	74 (20)	78 (22)	0.2	58 (20)	56 (20)	0.6	43 (10)	42 (20)	0.9
Exercise and QoL
6MWD* (m)	422 (49)	409 (62)	0.4	411 (73)	400 (98)	0.4	362 (89)	322 (92)	0.03
SGRQ*	23 (18)	21 (19)	0.4	35 (15)	37 (16)	0.6	53 (15)	54 (18)	0.6
CT
TLV (L)	5.4 (1.1)	5.9 (1.5)	0.03	5.7 (1.0)	5.9 (1.1)	0.1	6.5 (0.9)	6.8 (1.1)	0.05
RA950 (%)	3.7 (3.2)	6.3 (4.0)	0.001	11.3 (9.7)	14.4 (10.5)	<0.001	13.3 (8.7)	17.7 (9.8)	0.004
TAC (n)	250 (65)	230 (55)	0.002	236 (71)	211 (56)	0.001	237 (42)	209 (38)	0.005
WA (mm^2)	66.7 (1.4)	66.2 (1.4)	0.049	67.6 (1.7)	67.3 (1.4)	0.2	67.1 (1.3)	67.0 (1.6)	0.9
LA (mm^2)	13.8 (2.9)	15.2 (3.2)	0.01	12.9 (2.3)	14.1 (2.6)	0.005	13.4 (3.2)	14.3 (4.2)	0.2
WA% (%)	82.9 (3.2)	81.4 (3.5)	0.01	84.0 (2.7)	82.8 (2.9)	0.009	83.5 (3.5)	82.5 (4.5)	0.3
WT% (%)	17.8 (0.7)	17.5 (0.8)	0.07	18.2 (0.7)	17.7 (0.8)	0.003	17.9 (0.7)	17.9 (0.9)	0.7
MRI
VDP* (%)	7.5 (3.3)	10.9 (4.2)	0.002	17.5 (7.7)	22.0 (10.4)	0.001	29.4 (8.1)	35.8 (13.5)	0.049

Values are reported as mean (SD) unless otherwise indicated. p values represent significance values for paired samples t tests. Bolded values are statistically significant.

%_pred, percent of predicted value; 6MWD, six-minute walk distance; COPD, chronic obstructive pulmonary disease; CT, computed tomography; DL_CO, diffusing capacity of the lung for carbon monoxide; FEV₁, forced expiratory volume in 1-second; FVC, forced vital capacity; GOLD, Global Initiative for Chronic Obstructive Lung Disease; LA, lumen area; MRI, magnetic resonance imaging; QoL, quality-of-life; RA₉₅₀, relative area of the lung with attenuation less than –950 Hounsfield units; RV, residual volume; SGRQ, St. George’s Respiratory Questionnaire; TAC, total airway count; TLC, total lung capacity; TLV, total lung volume; VDP, ventilation defect percent; WA, wall area; WA%, wall area percent; WT%, wall thickness percent.

RV*, TLC*, RV/TLC*: GOLD III/IV n = 9.

DL_CO*: GOLD II n = 22, GOLD III/IV n = 9.

6MWD*: GOLD I n = 15, GOLD II n = 22, GOLD III/IV n = 9.

SGRQ*: GOLD I n = 15.

VDP*: GOLD I n = 14, GOLD III/IV n = 9.

Since FEV₁ did not change significantly over three-years but there were 10 ex-smokers that progressed in GOLD grade, we performed a sub-analysis on the 80 participants that had no GOLD grade progression to determine whether the longitudinal changes shown above would still hold. Table 5 shows pulmonary function, exercise capacity, quality-of-life, and imaging measurements at baseline and three-year follow-up for the 80 ex-smokers with and without COPD who did not progress in GOLD grade over three-years. In the ex-smokers without COPD, after three-years, all significantly different measurements from Table 4 held significance, except for FEV₁/FVC (95% vs. 96%; p = 0.06) and 6MWD (423 m vs. 412 m; p = 0.1). In the ex-smokers with COPD, all significantly different measurements held significance except for FVC (96% vs. 94%; p = 0.3).

Table 5. Pulmonary function, exercise capacity, quality-of-life, and imaging measurements at baseline and three-year follow-up for ex-smokers with and without COPD who did not progress in GOLD grade.

	All Ex-smokers n = 80			Without COPD n = 37			With COPD n = 43
Parameter	Baseline	Follow-up	p	Baseline	Follow-up	p	Baseline	Follow-up	p
Lung function
FEV1 (%pred)	86 (26)	87 (28)	0.3	103 (17)	105 (20)	0.2	72 (23)	72 (25)	0.9
FVC (%pred)	95 (17)	95 (19)	0.9	95 (16)	96 (18)	0.2	96 (18)	94 (20)	0.3
FEV1/FVC (%)	67 (16)	67 (15)	0.5	81 (6)	80 (6)	0.06	54 (11)	55 (10)	0.2
RV (%pred)	124 (41)	122 (38)	0.6	104 (20)	101 (20)	0.4	141 (46)	141 (40)	0.9
TLC (%pred)	108 (16)	104 (17)	0.01	101 (12)	96 (12)	<0.001	113 (18)	111 (17)	0.4
RV/TLC (%)	43 (10)	44 (10)	0.2	39 (7)	41 (7)	0.07	47 (11)	47 (12)	0.7
DLCO* (%pred)	70 (21)	74 (24)	0.003	80 (16)	87 (17)	<0.001	61 (21)	63 (23)	0.3
Exercise and QoL
6MWD* (m)	411 (78)	396 (86)	0.01	423 (82)	412 (76)	0.1	402 (74)	383 (93)	0.04
SGRQ*	27 (19)	27 (20)	0.6	20 (19)	19 (18)	0.7	33 (18)	33 (20)	0.8
CT
TLV (L)	5.2 (1.1)	5.4 (1.3)	0.02	4.6 (0.9)	4.7 (1.0)	0.5	5.7 (1.1)	6.0 (1.3)	0.01
RA950 (%)	5.2 (7.1)	6.9 (8.1)	<0.001	1.4 (1.4)	1.9 (1.4)	0.03	8.4 (8)	11.1 (9)	<0.001
TAC (n)	271 (77)	257 (75)	<0.001	308 (74)	299 (74)	0.3	240 (65)	220 (53)	<0.001
WA (mm^2)	66.6 (1.9)	66.3 (1.7)	0.03	65.7 (1.8)	65.6 (1.7)	0.5	67.3 (1.5)	66.8 (1.4)	0.02
LA (mm^2)	14.6 (3.4)	15.5 (3.4)	<0.001	16.1 (3.6)	16.8 (3.4)	0.03	13.2 (2.7)	14.4 (2.9)	<0.001
WA% (%)	82.1 (3.8)	81.1 (3.7)	<0.001	80.4 (3.9)	79.7 (3.7)	0.04	83.6 (3.0)	82.4 (3.2)	<0.001
WT% (%)	17.8 (0.7)	17.6 (0.7)	<0.001	17.6 (0.7)	17.5 (0.7)	0.1	18.0 (0.7)	17.7 (0.7)	0.001
MRI
VDP* (%)	10.7 (8.2)	13.9 (9.8)	<0.001	5.8 (3.0)	8.2 (4.0)	0.001	14.9 (9.0)	18.8 (10.6)	<0.001

Values are reported as mean (SD) unless otherwise indicated. p values represent significance values for paired samples t tests. Bolded values are statistically significant.

%_pred, percent of predicted value; 6MWD, six-minute walk distance; COPD, chronic obstructive pulmonary disease; CT, computed tomography; DL_CO, diffusing capacity of the lung for carbon monoxide; FEV₁, forced expiratory volume in 1-second; FVC, forced vital capacity; GOLD, Global Initiative for Chronic Obstructive Lung Disease; LA, lumen area; MRI, magnetic resonance imaging; QoL, quality-of-life; RA₉₅₀, relative area of the lung with attenuation less than –950 Hounsfield units; RV, residual volume; SGRQ, St. George’s Respiratory Questionnaire; TAC, total airway count; TLC, total lung capacity; TLV, total lung volume; VDP, ventilation defect percent; WA, wall area; WA%, wall area percent; WT%, wall thickness percent.

DL_CO*: All ex-smokers n = 77, Without COPD n = 35, With COPD n = 42.

6MWD*: All ex-smokers n = 76, Without COPD n = 34, With COPD n = 42.

SGRQ*: All ex-smokers n = 76, Without COPD n = 34, With COPD n = 42.

VDP*: All ex-smokers n = 74, Without COPD n = 34, With COPD n = 40.

CT airway measurements summarized in scatter plots in Supplementary Figure S1 show that TAC significantly decreased in ex-smokers with COPD, but not in ex-smokers without COPD and that WA% decreased in both ex-smoker subgroups. Supplementary Figure S2 provides similar information by COPD grade severity and shows that TAC decreased in all COPD grade severity subgroups, while WA% decreased only in GOLD I and GOLD II subgroups. Figure 4 provides a schematic summarizing the changes observed after three-years for TAC, airway WA, LA, WA%, WT%, and VDP in both ex-smoker subgroups.

Figure 4. Schematic showing decreased TAC in ex-smokers with COPD. Schematic also shows decreased airway WA% and increased LA and VDP in ex-smokers without COPD (ES), as well as decreased airway WA, WA%, and WT%, and increased LA and VDP in ex-smokers with COPD.

Relationships

Figure 5 shows baseline and follow-up correlations for TAC with FEV₁ (baseline: ρ = 0.38, p < 0.001; follow-up: ρ = .48, p < 0.001) and VDP (baseline: ρ = –0.30, p = 0.005; follow-up: ρ = –0.33, p = 0.002). The longitudinal change (Δ) in TAC also correlated with the longitudinal change in FEV₁ (r = 0.38, p < 0.001) and the longitudinal change in VDP (ρ = –0.37, p < 0.001). When adjusting for potential confounding variables, including age, sex, and pack-years, these longitudinal associations continued to be significantly correlated (ΔTAC-ΔFEV₁: r = 0.38, p < 0.001; ΔTAC-ΔVDP: ρ = –0.27, p = 0.01).

Figure 5. Scatter plots showing Spearman (ρ) and Pearson (r) correlations in all ex-smokers at baseline, three-year follow-up, and the longitudinal change between baseline and follow-up (Δ). TAC was correlated with FEV₁ (baseline: ρ = 0.38, p < 0.001; follow-up ρ = 0.48, p < 0.001). ΔTAC was correlated with ΔFEV₁ (r = 0.38, p < 0.001). TAC was correlated with VDP (baseline: ρ = –0.30, p = 0.005; follow-up: ρ = –0.33, p = 0.002). ΔTAC was correlated with ΔVDP (ρ = –0.37, p < 0.001).

We also generated multivariable models to predict TAC at three-year follow-up and the longitudinal change in TAC, as shown in Table 6, while Table 7 shows correlation values for potential predictor variables. The variance inflation factors were acceptable for all variables included in the multivariate models. In the first model (R² = 0.632; p < 0.001), WA% (β = –0.625; p < 0.001) and FEV₁/FVC (β = 0.300; p < 0.001) at baseline were significant predictors of TAC at follow-up, while RV/TLC (β = –0.018; p = 0.8), RA₉₅₀ (β = –0.039; p = 0.7), VDP (β = –0.029; p = 0.8), DL_CO (β = 0.010; p = 0.9), FEV₁ (β = 0.045; p = 0.7), and WT% (β = 0.195; p = 0.2) were excluded. In the second model (R² = .156; p < 0.001), WT% (β = 0.205; p = 0.045) and RA₉₅₀ (β = –0.313; p = 0.003) at baseline were significant predictors of the longitudinal change in TAC, while FEV₁ (β = 0.096; p = 0.5), VDP (β = –0.031; p = 0.8), WA% (β = 0.081; p = 0.7), FEV₁/FVC (β = 0.100; p = 0.5), DL_CO (β = –0.089; p = 0.5), and RV/TLC (β = –0.089; p = 0.4) were excluded.

Table 6. Multivariable linear regression models.

	Unstandardized		Standardized
Parameter	B	SE	β	p
Model 1*: TAC at follow-up
Constant	1,155.783	125.374	–	<0.001
WA%	−12.106	1.400	−0.625	<0.001
FEV1/FVC	1.376	0.331	0.300	<0.001
Excluded
RV/TLC			−0.018	0.820
RA950			−0.039	0.674
VDP			−0.029	0.773
DLCO %pred			0.010	0.898
FEV1 %pred			0.045	0.674
WT%			0.195	0.180
Model 2^†: ΔTAC
Constant	−177.637	82.741	–	0.04
WT%	9.381	4.615	0.205	0.045
RA950	−1.428	0.461	–0.313	0.003
Excluded
FEV1 %pred			0.096	0.497
VDP			−0.031	0.828
WA%			0.081	0.728
FEV1/FVC			0.100	0.505
DLCO %pred			−0.089	0.531
RV/TLC			−0.089	0.447

p values show coefficient significance. Bolded values are statistically significant. All prediction variables represent measurements at baseline.

%pred, percent of predicted value; DL_CO, diffusing capacity of the lung for carbon monoxide; FEV₁, forced expiratory volume in 1-second; FVC, forced vital capacity; RA₉₅₀, relative area of the lung with attenuation less than –950 Hounsfield units; RV, residual volume; TAC, total airway count; TLC, total lung capacity; VDP, ventilation defect percent; WA%, wall area percent; WT%, wall thickness percent.

* R² = .632; p < 0.001.

^† R² = .156; p < 0.001.

Table 7. Correlations for potential predictor variables in linear regression models.

	TAC at baseline
Predictor variables at baseline	ρ	p
Age (year)	0.052	0.6
Sex	–0.13	0.2
BMI (kg/m^2)	0.12	0.3
Pack-years	–0.044	0.7
FEV1 (%pred)	0.38	<0.001
FVC (%pred)	–0.01	0.9
FEV1/FVC (%)	0.47	<0.001
RV (%pred)	–0.35	<0.001
TLC (%pred)	–0.21	0.045
RV/TLC (%)	–0.25	0.02
DLCO (%pred)	0.21	0.046
6MWD (m)	0.09	0.4
SGRQ	–0.18	0.1
VDP (%)	–0.30	0.005
TLV (L)	–0.019	0.9
RA950 (%)	–0.23	0.03
WA (mm^2)	–0.76	<0.001
LA (mm^2)	0.74	<0.001
WA% (%)	–0.75	<0.001
WT% (%)	–0.62	<0.001

p values represent significance values for Pearson and spearman correlations. Bolded values are statistically significant.

%pred, percent of predicted value; 6MWD, six-minute walk distance; BMI, body mass index; DL_CO, diffusing capacity of the lung for carbon monoxide; FEV₁, forced expiratory volume in 1-second; FVC, forced vital capacity; LA, lumen area; ρ, Spearman correlation coefficient; r, Pearson correlation coefficient; RA₉₅₀, relative area of the lung with attenuation less than –950 Hounsfield units; RV, residual volume; SGRQ, St. George’s Respiratory Questionnaire; TAC, total airway count; TLC, total lung capacity; TLV, total lung volume; VDP, ventilation defect percent; WA, wall area; WA%, wall area percent; WT%, wall thickness percent.

Discussion

In this three-year longitudinal study, we quantified for the first time, total airway count, as well as airway wall and lumen measurements in ex-smoker participants in the TINCan cohort. Our findings substantiate and support previous cross-sectional evaluations [17–20]. We observed: (1) significantly decreased TAC in ex-smokers with spirometry evidence of COPD, but no significant TAC changes in ex-smokers with no evidence of COPD, (2) a small but significant change in airway WA% and LA in all ex-smokers, regardless of COPD status, (3) significant relationships between CT TAC and MRI VDP, and, (4) airway wall thickness was predictive of TAC worsening after three-years.

Approximately three-years after a baseline visit, TAC was decreased in ex-smokers with COPD and by GOLD grade, but not in ex-smokers without COPD. We cannot ascertain the cause of this decrease in CT-visible airways, including potential technical issues, airway narrowing, occlusion, obstruction, and/or obliteration. We do know however, that CT TAC was previously shown to be associated with the number of micro-CT terminal bronchioles [18], lending support to the notion that differences in TAC may reflect small airway loss or obliteration. We also know from previous work that TAC explained COPD progression [19], so perhaps by measuring TAC over time, we can better understand or predict which patients will worsen more quickly over time.

Regardless of COPD diagnosis, WA% was diminished in ex-smokers, suggestive of airway wall thinning over time, and this result was consistent with previous cross-sectional findings [20]. It is important to note that in previous work, airways with thinner walls were spatially related to missing airways [17] in COPD, whereas in asthma, missing airways were spatially related to airways with thicker walls [36]. It is possible that CT airway wall thinning stems from progressive airway destruction over time in COPD whereas in asthma, airway inflammation and remodeling play a larger role.

We also observed LA enlarged over time in all ex-smokers, suggestive of luminal dilation. Previous studies have shown that the airway lumen is more narrow in ever-smokers with COPD compared to healthy controls [17]. Here, we also observed a narrower lumen in ex-smokers with COPD compared to those without. We were surprised to observe airway luminal area increasing over time. Perhaps this was simply due to airway remodeling in ex-smokers with COPD, such that as the airway walls are becoming thinner, the lumen is then expanding and widening as a result.

We observed a significant relationship between TAC and FEV₁, consistent with previous findings [17]. TAC was also negatively correlated with VDP, suggesting that for ex-smokers with reduced CT-visible airways, ventilation heterogeneity is greater (worse), which intuitively makes sense. In addition, the longitudinal change in TAC positively correlated with the longitudinal change in FEV₁ and negatively correlated with the longitudinal change in VDP, even when adjusting for potential confounding variables, which suggests that TAC worsening is related to pulmonary function decline and ventilation worsening. Moreover, both MRI VDP and CT wall and lumen measurements changed over three-years in the absence of changes in FEV₁, which suggests that these airway changes occur before they are reflected by spirometry measurements, which are dominated by the large airways. Importantly, previous work demonstrated that TAC was associated with the longitudinal six-year decline in FEV₁ [17]. In agreement with this finding, we generated a multivariable model that identified baseline WA% and FEV₁/FVC as significant predictors of TAC at follow-up. In fact, these two variables together explained more than 60% of the variability of TAC in these participants over time. Thicker airway walls and/or narrower lumen at baseline were associated with lower values for TAC at follow-up. Moreover, ex-smokers with diminished FEV₁/FVC at baseline (who were more obstructed), also had diminished TAC at follow-up. In another multivariable model, we identified baseline WT% and RA₉₅₀ as significant predictors of the longitudinal change in TAC. Thinner airway walls at baseline were associated with a greater decrease in TAC over time. Furthermore, ex-smokers with increased RA₉₅₀ (more emphysematous destruction) also had decreased TAC over time. Thus, for participants in whom TAC significantly worsened after three-years, baseline CT airway wall thickness and emphysema measurements were supportive baseline predictors. We were surprised that baseline MRI VDP, while correlating with TAC, did not significantly contribute to the predictive models. Previous work in the TINCan cohort has shown that baseline VDP was a significant predictor of longitudinal quality-of-life and lung function worsening [24]. Therefore, quantitative CT is more appropriate for monitoring small airway structure in patients with COPD.

We acknowledge a number of study limitations including the relatively small sample size especially compared to other COPD cohort studies [6–10]. In addition, our study lacks a control group of never- and current-smokers, which limits the application of our findings to the general population, and our ability to evaluate the impact of smoking cessation on longitudinal CT airway changes. Previous work has also shown that quantitative CT measurements can be affected by factors such as lung volume [17]. Here, we show CT TLV was increased only in ex-smokers with COPD, where this variation may be due to participant breath-hold accuracy or disease physiology, as lung hyperinflation or air trapping can cause an increase in lung volume in patients with COPD. Thus, future work investigating quantitative CT airway measurements longitudinally in larger COPD cohorts of current- and former-smokers, as well as never-smokers, is needed to improve the generalizability of our results. Furthermore, the study participants were recruited as a convenience and not a random population-based sample that may have biased the results to people who were in better (or worse) health at baseline. This study was also dominated by patients with an absence of, or milder COPD. Participants who attended a baseline visit but who did not return for follow-up reported worse values for pulmonary function, exercise capacity, and quality-of-life compared to those who did return. Taken together, this result suggests that this study provides a relatively conservative estimate of potential longitudinal differences.

Conclusions

In summary, over a relatively short time period of 31 ± 7 months, we observed reduced CT TAC in ex-smokers with COPD, as well as airway wall thinning in all ex-smokers. These longitudinal findings in ex-smokers, in whom there was no FEV₁ worsening, provide insights into mechanisms of COPD progression, while supporting previous cross-sectional evaluations [17, 20], and suggest that the evaluation of airway remodeling on CT in patients with COPD may be an important clinical tool for predicting disease progression and managing COPD in the future.

Author contributions

P.V.W. was responsible for image processing, statistical analyses and interpretation, as well as manuscript preparation and submission. M.S. and V.D. assisted with data analysis and interpretation. D.G.M. was responsible for recruitment of study participants, clinical input in the study design, and clinical interpretation of the data. I.A.C. and M.A.A. assisted with technical and clinical interpretation of the data. M.K. assisted with participant recruitment and data acquisition and interpretation. G.P. was responsible for conception of the study, experimental design, data interpretation, and approval of the final manuscript, as well as being the guarantor of study data integrity. All coauthors had an opportunity to review and revise the manuscript and approved its final submitted version.

Abbreviations:

COPD	=	chronic obstructive pulmonary disease
DLCO	=	diffusing capacity of the lung for carbon monoxide
FEV1	=	forced expiratory volume in 1-second;
FVC	=	forced vital capacity;
GOLD	=	Global Initiative for Chronic Obstructive Lung Disease
LA	=	lumen area;
RV	=	residual volume;
SGRQ	=	St. George’s Respiratory Questionnaire;
TAC	=	total airway count;
TLC	=	total lung capacity;
TLV,	=	total lung volume;
VDP	=	ventilation defect percent;
WA	=	wall area;
WA%	=	wall area percent;
WT%	=	wall thickness percent

Acknowledgment

We thank the participants who volunteered for this study.

Disclosure statement

M.K. disclosed consultant fees from VIDA Diagnostics, not related to this work. Otherwise, no conflicts of interest. P.V.W., M.S., V.D., D.G.M., I.A.C., M.A.A., and G.P. disclosed no potential conflicts of interest.

Data availability statement

Data generated or analyzed during the study are available from the corresponding author upon request.

Additional information

Funding

M.S. is supported by the Natural Sciences and Engineering Research Council (NSERC) of Canada Post-Graduate Doctoral Scholarship. V.D. is supported by the NSERC Undergraduate Student Research Awards program. M.K. is supported by the Parker B. Francis Fellowship Program, NSERC, and holds a Tier 2 Canada Research Chair. G.P. is supported by NSERC, CIHR, the Baran Family Foundation, and holds a Tier 1 Canada Research Chair.