Airway Tapering in Chronic Obstructive Pulmonary Disease

Abstract

Background

Luminal narrowing is a hallmark feature of airway remodeling in COPD, but current measures focus on airway wall remodeling. Quantification of the natural increase in cumulative cross-sectional area along the length of the human airway tree can facilitate assessment of airway narrowing.

Methods

We analyzed the airway trees of 7641 subjects enrolled in the multicenter COPDGene cohort. Airway luminal tapering was assessed by estimating the slope of the change in cumulative cross-sectional area along the length of the airway tree over successive generations (T-Slope). We performed multivariable regression analyses to test the associations between T-Slope and lung function, St. George’s Respiratory Questionnaire (SGRQ), modified Medical Research Council (mMRC) dyspnea score, 6-minute walk distance (6MWD), FEV₁ change, exacerbations, and all-cause mortality after adjusting for demographics, %CT emphysema, and total airway count.

Results

The T-Slope decreased with increasing COPD severity: 2.69 (0.70) in nonsmokers and 2.33 (0.70), 2.11 (0.65), 1.78 (0.58), 1.60 (0.53), and 1.57 (0.52) in GOLD stages 0 through 4 respectively (Jonckheere-Terpstra p=0.04). On multivariable analyses, the T-Slope was independently associated with FEV₁ (β = 0.13 L, 95%CI 0.10 to 0.15, p<0.001), 6MWD (β = 15.0 m, 95%CI 10.8 to 19.2, p<0.001), change in FEV₁ (β = −4.50 ml/year, 95%CI −7.32 to −1.67; p=0.001), exacerbations (IRR = 0.78, 95%CI 0.73 to 0.83, p<0.001), and mortality (HR = 0.79, 95%CI 0.72 to 0.86, p<0.001).

Conclusion

T-Slope is a measure of airway luminal remodeling and is associated with respiratory morbidity and mortality.

Introduction

Airway remodeling is a hallmark pathophysiologic feature in chronic obstructive pulmonary disease (COPD) (1). Airway changes can be visualized on chest computed tomography (CT), and several metrics exist for their quantification (2). Most of these, including airway wall thickness, the wall area percentage, and the Pi10, are estimates of changes in the airway wall(2, 3). Although airway wall changes are associated with several important clinical outcomes in COPD (4), they do not completely describe the resulting changes in the airway lumen and they may also be influenced by adventitial thickening. In contrast, the airway lumen diameter directly impacts the resistance to airflow(5).

Airway luminal narrowing and loss are characteristic features of the airway remodeling process associated with COPD(1, 6). The airway lumens are often narrowed and distorted due to an admixture of chronic inflammation, fibrosis, and mucus plugging (1). In contrast to measurements of the airway wall, studies of quantification of changes in the airway lumen have been minimal due to the complex recurring branching patterns and varying airway lumen sizes over successive generations (2, 7). In addition, metrics of lumen size are reliant on the airway generation that is measured. Tanabe and colleagues utilized the volume of the airway tree adjusted for the size of the lungs as a measure of luminal remodeling, but this measure is disproportionately influenced by the size of larger airways (8). The total airway count (TAC) is the number of visualized airway branches and thereby an estimate of missing branches(7). TAC does not, however, indicate the degree of luminal narrowing nor account for the size and generation of missing branches.

According to the classic Weibel model, the airway tree is fractal in nature and is designed to facilitate the most efficient air flow (9). We recently showed that such repetitive airway fractal nature can be quantified, and the airway fractal dimension can help phenotype structural differences (10). The parallel arrangement of branches at each generation results in an exponential increase in the total cross-sectional area of the conducting airways over successive generations (9). A quantification of the increase in cumulative cross-sectional area of the lumen from central to distal airways (the tapering slope or T-Slope) should therefore provide a measure of luminal remodeling. We hypothesized that any deviations from the expected increase in luminal cross-sectional area across successive airway generations will be progressively greater with worsening disease, and that the T-Slope will be associated with lung function, dyspnea, respiratory quality of life, exacerbations, lung function change, and mortality.

Methods

Study Population

We included participants enrolled in the Genetic Epidemiology of COPD (COPDGene) study (CONSORT diagram in Figure 1) (11). Participants with bronchiectasis were excluded from enrollment. The enrolled subjects were current and former smokers between the ages of 45 and 80 years and with at least 10 pack-year smoking history. The eligibility criteria and study protocol of COPDGene have been previously published (11). Smoking status was classified as current if participants smoked within 30 days of the study visit. The diagnosis of COPD was based on post bronchodilator FEV₁/FVC <0.70 and disease severity was determined according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) recommendations. We also classified participants by symptom assessment into A, B, and E categories according to the 2023 GOLD document (12, 13). Participants without airflow obstruction (FEV₁/FVC ≥0.7) were considered as GOLD stage 0. We excluded participants with Preserved Ratio Impaired Spirometry or PRISm (FEV₁/FVC ≥0.70 and FEV₁ %predicted <80). All participants performed a six-minute walk test to assess functional capacity (14). The St. George’s Respiratory Questionnaire (SGRQ) score was used to assess respiratory quality of life and the modified Medical Research Council (mMRC) dyspnea scale was used to measure shortness of breath (12, 15).

Figure 1:

Consort Diagram

Exacerbations were defined as increase in cough, sputum, and/or dyspnea that lasted at least 48 hours and required the use of antibiotics and/or systemic steroids. Exacerbations were prospectively ascertained every 6 months using an automated telephony system, or by telephone by research coordinators using a validated questionnaire(16). Participants returned for in-person follow-up approximately 5 years after the initial visit. Change in lung function between visits was assessed by calculating the annualized difference in FEV₁ between enrollment and the 5-year follow-up visit. Vital status was assessed at the 6-monthly follow-up phone calls, and deaths were confirmed via a search of the Social Security Death Index. All participants provided written informed consent and the study protocol was approved by the Institutional Review Board (IRB) at each of the 21 participating centers.

CT Imaging

High-resolution volumetric CT scans of the chest were acquired at full inspiration (total lung capacity, TLC) at each visit using multi-detector scanners. The lungs and airway trees were segmented using Thirona lung quantification software (LungQ, Thirona, Netherlands)(17). Emphysema was defined as the voxels within the lung mask with attenuation < −950 Hounsfield Units (HU). The total airway count (TAC) was estimated from the airway tree after skeletonization and object counting using in-house code developed in MATLAB (MathWorks, Natick, MA).

Airway Tapering Slope

The initial tracheal inlet was labelled automatically by identifying an airway object in a slice with maximum diameter (Figure 2A). Skeletonization of the airway tree was performed by binary thinning of the airway segmentations to identify the center line using ‘bwskel’ function in MATLAB (18). The centroid of the trachea in the inlet slice was then used as the starting point to compute the three-dimensional (3D) geodesic distance along the airway centerline from proximal to distal airway tree along each visualized branch using ‘bwdistgeodesic’ function in MATLAB (Figure 2B) (19). The distance from each centerline point to the closest surface of the airway tree was computed and used to approximate the lumen radius r along the length of the airway tree (Figure 2C). These radii were then used to estimate cross-sectional areas at each level using the expression πr². We then calculated the cumulative cross-sectional area at each point by sequentially summing the cross-sectional areas moving from the proximal to the distal end of the airway tree. The cumulative cross-sectional area was then plotted against the corresponding geodesic path distance from the top of the airways with 0 being the geodesic distance value at the tracheal inlet. We used the Levenberg-Marquardt algorithm to fit a two-degree exponential curve (V_estimated = Ae^-−Bt + Ce^−Dt ) to the cumulative airway tapering curve where D represents the rate of rise of the airway cross-sectional area over the length of the airway tree (20). We termed this measure the airway tapering slope (T-Slope) (Figure 2D). The faster the rate of rise of the slope, the more positive D is. The more intact the airway branches are, the faster the rate of rise of the cumulative cross-sectional area and therefore the steeper the T-slope is. With progressive airway narrowing and/or loss, the T-slope flattens. Therefore, the lower the value of D, the flatter the slope and the greater the extent of airway tapering. In sensitivity analyses, we also computed T-Slope by quantifying airway cross sectional area using the formula A = C²/4π, where C is the circumference or perimeter of the airway. The results were similar and therefore we used πr² for all calculations. All computations were performed using in-built image processing and curve fitting functions in MATLAB software (MathWorks, Natick, MA).

Figure 2: Assessment of tapering in an airway tree extracted from chest computed tomography scan acquired at end-inspiration.

(A) 3D rendering of airway tree with identified tracheal inlet (B) center line of airway tree extracted by skeletonization process (C) Geodesic distance-based color map with tracheal inlet as the starting node. (D) Centerline-based distance map representing approximate diameter of the airway along the length of the airway tree (E) Tapering slope estimation from cumulative lumen cross-sectional area vs. geodesic distance with tracheal inlet as the starting point.

Statistical Analysis

Pearson’s test was used to assess the correlation between T-Slope and lung function measures. Multivariable regression analyses were performed to estimate associations between the T-Slope, FEV₁, and FEV₁/FVC with adjustment for age, race, sex, body mass index (BMI), smoking status, pack-years of smoking, TAC, %CT emphysema, and CT scanner type. Models for associations between T-Slope and six-minute walk distance, SGRQ scores, and change in FEV₁ were additionally adjusted for baseline FEV₁. Negative binomial regression was used to estimate the incidence risk ratio (IRR) to evaluate the association between T-Slope and the total number of exacerbations on follow-up. This model was additionally adjusted for the number of exacerbations in the 12 months prior to enrollment. Cox proportional hazards analysis was used to calculate adjusted hazards ratios (HRs) for T-Slope with all-cause mortality as the outcome, adjusting for age, race, sex, BMI, smoking status, pack-years of smoking, TAC, %CT emphysema, and CT scanner type. We categorized T-Slope into 4 quartiles (normal, mild, moderate, and severe) and performed Kaplan-Meier survival analysis with log-rank test to compare survival for individuals in each quartile. The quartile with the highest tapering slope was considered the reference group for these analyses. To compare the relative utility of the lumen measures T-Slope and TAC, we stratified participants into four categories based on median estimates of T-Slope and TAC, (1) Normal: Subjects with T-Slope > median and TAC > median, (2) Abnormal Airway Tapering Only: Subjects with T-Slope ≤ median and TAC > median, (3) Abnormal TAC Only: Subjects with T-Slope > median and TAC ≤ median, and (4) Both: Subjects with T-Slope ≤ median and TAC ≤ median. We compared all-cause mortality in these groups using multivariable Cox proportional hazards models. A 2-sided P value less than 0.05 was considered significant for all analyses. All analyses were performed using R statistical software (V 3.2).

Results

Subject Characteristics

Of the 10,198 participants who completed a baseline visit, we excluded 1675 participants within the PRISm category, 495 with incomplete CT imaging data, and 387 participants with incomplete airway segmentations (Figure 1). Of the remaining 7641 participants, 106 (1.3%) were lifetime never smokers, and 3767 (49.2%), 686 (8.9%), 1623 (21.2%), 964 (12.6%), and 495 (6.4%) participants had GOLD disease severity stages 0 through 4, respectively.

The mean (SD) age of the cohort was 59.9 (9.0) years. 3573 (46.7%) were females and 2292 (29.9%) were of African American race. 3851 (50.3%) were current smokers (Table 1). The mean (SD) T-Slope was 2.0 (0.7). The mean (SD) T-Slope in lifetime nonsmokers was 2.69 (0.70). The T-Slope decreased with increasing COPD severity: 2.33 (0.70), 2.11 (0.65), 1.78 (0.58), 1.60 (0.53), and 1.57 (0.52) in GOLD stages 0 through 4 respectively (p<0.001). (Figure 3).

Table 1:

Baseline Demographics of Participants (n=7641)

Parameters	Results
Age, years	59.9 (9.0)
Sex, Female, n (%)	3573 (46.7%)
African American race, n (%)	2292 (29.9%)
Body Mass Index, kg/m2	28.2 (5.8)
Current Smokers, n (%)	3851 (50.3%)
Pack years of smoking	43.7 (25.2)
FEV1, L	2.2 (0.9)
FEV1 %predicted	78.1 (26.8)
FEV1/FVC	0.65 (0.10)
GOLD Stage, n (%)
Nonsmokers:	106 (1.3%)
0:	3767 (49.2%)
1:	686 (8.9%)
2:	1623 (21.2%)
3:	964 (12.6%)
4:	495 (6.4%)
%CT Emphysema	6.8 (10.2)
Total Airway Count	146 (73)
T-Slope	2.0 (0.7)

All values expressed as mean (SD) unless specified otherwise.

FEV₁: Forced Expiratory Volume in the First Second

FEV₁ % predicted estimated according to National Health and Nutrition Examination Survey III guidelines.

FVC: Forced Vital Capacity

CT = Computed tomography

Emphysema measured as the percentage of voxels with density < −950 Hounsfield Units on inspiratory CT scans

T-Slope represents the airway tapering slope

Figure 3: Airway tapering slope across COPD severity stages and by symptom burden.

The panels show differences in T-slope by GOLD stage (left) and by symptom burden (right).

Airway Tapering and Lung Function

On univariate analysis, the T-Slope was significantly associated with FEV₁ (β = 0.46 L, 95%CI 0.43 to 0.49, p<0.001), FEV₁ %predicted (β = 16.64, 95%CI 15.89 to 17.39, p<0.001), and FEV₁/FVC (β = 0.09, 95%CI 0.08 to 0.09, p<0.001). In multivariable analyses after adjusting for age, race, sex, BMI, current smoking status, pack years of smoking, TAC, %CT emphysema, and CT scanner type, T-Slope was significantly associated with FEV₁ (adjusted β = 0.13 L, 95%CI 0.10 to 0.15, p<0.001), FEV₁ %predicted (adjusted β = 5.96, 95%CI 5.22 to 6.69, p<0.001), and FEV₁/FVC (adjusted β = 0.029, 95%CI 0.025 to 0.033, p<0.001), and FEF_25-75% (adjusted β = 0.19, 95%CI 0.15 to 0.23, p<0.001)

Airway Tapering and Respiratory Morbidity

On univariate analysis, the T-Slope was significantly associated with six-minute walk distance (β = 52.38 m, 95%CI 48.76 to 56.01, p<0.001), total SGRQ score (β = −11.40, 95%CI −12.06 to −10.74, p<0.001), and mMRC dyspnea score (β = −0.59, 95%CI −0.63 to −0.54, p<0.001). after adjusting for age, race, sex, BMI, current smoking status, pack years of smoking, FEV₁, TAC, %CT emphysema, and CT scanner type, the T-Slope was significantly associated with six-minute walk distance (adjusted β = 15.01 m, 95%CI 10.81 to 19.22, p<0.001), total SGRQ score (adjusted β = −2.48, 95%CI −3.23 to −1.72, p<0.001) and mMRC dyspnea score (adjusted β = −0.05, 95%CI −0.09 to −0.01, p=0.014).

Airway Tapering and Lung Function Decline

Of 7641 participants included in the study, 4306 subjects returned for a second visit at 5 years after the baseline visit. The change in FEV₁ was annualized between the baseline and 5-year visits. After adjusting for age, race, sex, BMI, current smoking status, pack years of smoking, baseline FEV₁, TAC, %CT emphysema, and CT scanner type, the baseline T-Slope was significantly associated with change in FEV₁ (adjusted β = −4.50 ml/year, 95%CI −7.32 to −1.67; p=0.001).

Airway Tapering and Exacerbations

Of 7641 participants included in the study, exacerbation data was available for 6802 participants. The baseline T-Slope was inversely associated with exacerbation frequency (adjusted IRR = 0.78, 95%CI 0.73 to 0.83, p<0.001), after adjusting for age, race, sex, BMI, current smoking status, %CT emphysema, CT scanner type, and frequency of exacerbations in the year prior to enrollment. When estimated by quartiles with the highest baseline T-slope quartile as the reference group, higher rate of exacerbations was observed with increase in tapering severity: adjusted IRR was 1.13, (95%CI 0.99 to 1.28, p=0.05) in mild, 1.42 (95%CI 1.25 to 1.61, p<0.001) in moderate, and 1.64 (95%CI 1.44 to 1.87, p<0.001) in the severe tapering quartile.

Airway Tapering and Mortality

Participants were followed for a median of 10.9 (IQR 7.4) years. Of 7641 participants, 2180 (28.5%) participants died. After adjusting for age, sex, race, BMI, smoking status, pack years, %CT emphysema, TAC, and CT scanner-type, baseline T-Slope was significantly associated with mortality (adjusted HR = 0.79, 95%CI 0.72 to 0.86, p<0.001). When categorized into quartiles, the hazards ratio for mortality significantly increased with increased severity of tapering. With the highest T-slope quartile as reference, the adjusted HR for mortality was 1.09 (95%CI 0.94 to 1.26, p=0.23), 1.27 (95%CI 1.09 to 1.48, p=0.001), and 1.55 (95%CI 1.32 to 1.82, p<0.001) for mild, moderate, and severe tapering, respectively (Figure 4A).

Figure 4:

Adjusted survival plots for (A) Tapering categories based on quartiles and (B) Groups categorized by abnormal T-slope and TAC based on median cutoffs

Patient Characterization by T-Slope Quartiles

Table 2 shows the patient characteristics in each tapering quartile. Among 1910 subjects in the severe airway tapering category, 482 (25.2%) had no airflow obstruction but were symptomatic (GOLD 0). Compared to the normal tapering category (4.1%), the %CT emphysema on average increased with worsening quartiles of T-Slope, and was 5.4(8.7), 8.2(9.6), and 9.6(12.3) respectively in the mild, moderate, and severe quartiles, respectively. Similarly, FEV₁ decreased with severity of tapering (mild=2.4 L (0.8), moderate=2.1 L (0.9), severe=1.8 L (0.9)) as compared to subjects with normal tapering (2.7L (0.8)). Participants had lower number of airway branches with increase in tapering: mild (154±55 branches), moderate (121±43 branches), and severe tapering (92±34 branches).

Table 2:

Comparison of Patient Characteristics by Airway Tapering Quartiles.

Parameters	Normal Tapering (n=1911)	Mild Tapering (n=1910)	Moderate Tapering (n=1910)	Severe Tapering (n=1910)
Age, years	60.7 (9.0)	59.1 (8.9)	60.0 (9.2)	59.6 (9.0)
Sex, Female, n	937 (49.0%)	862 (45.1%)	878 (45.9%)	896 (46.9%)
African American, n	476 (24.9%)	620 (32.4%)	600 (31.4%)	596 (31.2%)
Body Mass Index, kg/m2	27.7 (5.4)	28.3 (5.6)	28.2 (5.8)	28.7 (6.3)
Current Smokers, n	810 (42.3%)	1020 (53.4%)	971 (50.8%)	1050 (54.9%)
Pack years of smoking	38.2 (22.7)	42.0 (24.6)	45.5 (24.8)	49.0 (27.4)
FEV1, L	2.7 (0.8)	2.7 (0.8)	2.1 (0.9)	1.8 (0.9)
FEV1 %predicted	94 (18.9)	84.2 (23.3)	71.9 (26.0)	62.2 (26.9)
FEV1/FVC	0.74 (0.11)	0.69 (0.14)	0.62 (0.16)	0.56 (0.17)
GOLD Stage, n (%)
Nonsmokers:	62 (3.2%)	24 (1.2%)	16 (0.8%)	4 (0.2%)
0:	1445 (75.6%)	1100 (57.5%)	740 (38.7%)	482 (25.2%)
1:	162 (8.4%)	220 (11.5%)	182 (9.5%)	122 (6.3%)
2:	164 (8.5%)	367 (19.2%)	513 (26.8%)	579 (30.3%)
3:	52 (2.7%)	135 (7.6%)	312 (16.3%)	465 (24.3%)
4:	26 (1.3%)	64 (3.3%)	147 (7.6%)	258 (13.5%)
%CT Emphysema	4.0 (6.6)	5.4 (8.7)	8.2 (11.3)	9.6 (12.3)
Total Airway Count	215 (82)	154 (55)	121 (43)	92 (34)
T-Slope	3.0 (0.4)	2.2 (0.1)	1.7 (0.1)	1.1 (0.3)

All values expressed as mean (SD) unless specified otherwise.

FEV₁: Forced Expiratory Volume in the First Second

FEV₁ % predicted estimated according to National Health and Nutrition Examination Survey III reference equation.

FVC: Forced Vital Capacity

CT = Computed tomography

Emphysema measured as the percentage of voxels with density < −950 Hounsfield Units on inspiratory CT scans

T-Slope represents airway tapering slope

Comparison of T-Slope and TAC

The cohort was comprised of 2884 (37.7%) with normal T-Slope and TAC with median estimate for both parameters as the cutoff. 890 (11.6%) had abnormal airway tapering only. 936 subjects had abnormal TAC only and 2931 (38.3%) had both significant tapering and airway loss. Survival analysis, adjusted for potential confounders (age, race, sex, smoking status, body mass index, pack-years of smoking, CT emphysema %, and scanner type), revealed significant differences between the groups. Compared to the normal group, the abnormal airway tapering only group exhibited significantly worse survival (adjusted HR = 1.25, 95% CI: 1.07-1.46, p = 0.004). The hazards for the abnormal TAC group (adjusted HR = 1.28, 95% CI: 1.10-1.49, p = 0.001) and subjects with both abnormal (adjusted HR = 1.74, 95% CI: 1.56-1.94, p < 0.001), were also high compared to normal. Furthermore, the airway tapering only group comprised a higher proportion of subjects in early disease stage (GOLD 0) compared to the abnormal TAC only group (54% vs. 43%, respectively). Similarly, in GOLD 1 (11.1% vs. 8.9%). In contrast, the abnormal TAC only group had more participants in GOLD stages 2-4 as compared to abnormal tapering only (47.5% vs. 33.4%), suggesting that significant airway tapering may precede substantial airway loss in the progression of COPD.

Discussion

In a multicenter cohort of current and former smokers, we demonstrated that a new CT-based measure of airway luminal remodeling, the T-Slope, decreases with increase in COPD severity, and is associated with lower FEV₁, lower six-minute walk distance, worse respiratory quality of life, more dyspnea, greater frequency of exacerbations, faster FEV₁ decline, and higher mortality. We also found that a significant number of smokers without airflow obstruction have evidence of moderate to severe airway tapering.

Substantial changes in the airway have been documented to occur in individuals at risk for COPD well before significant airflow obstruction has set in (2, 21). We observed that 25% of subjects with a smoking history but without spirometric evidence of airflow limitation (GOLD 0) exhibited pronounced airway tapering as quantified by the worst quartile of the T-slope. This finding underscores the presence of structural alterations in airway luminal morphology occurring in pre-clinical stages of chronic obstructive pulmonary disease (COPD). The T-slope metric offers potential for earlier phenotyping of at-risk individuals and may contribute to improved prognostication of disease trajectories. Existing measures of airway remodeling are predominantly those that measure changes in the walls of the airways. Airway wall thickening is associated with a faster decline in FEV₁ and with increased incidence of COPD(22). Changes in the airway wall reflect a combination of concentric and eccentric thickening, and therefore, airway wall measures may not always impact airway resistance and therefore there has been considerable interest in the development of luminal metrics for quantifying airway disease in COPD. TAC is a measure of the number of visualized airways, which may be impacted by both airway narrowing to the extent that they are no longer visualized, or by true obliteration of airways. Compared to lifetime non-smokers, those with GOLD stage 1 have 17% to 40% lower TAC (7, 23). These findings were first observed in explanted lungs (1, 24)and these reductions are also observable in human chest CT scans where only medium and large size airways can be visualized (7, 23). Lower TAC is associated with lung function decline and incident COPD (7, 25). However, it is important to note that TAC does not provide a direct measure of airway narrowing in those airways that remain visible on CT.

The airway tree has been quantified using several models. The classic Weibel “A” model is based on the tracheobronchial tree structure as a hierarchical network of tubes with symmetrical branching (9). Weibel’s “B” model and subsequent models by Horsfield and others (26) provided more realistic reflections of the division of each parent branch into asymmetric major and minor daughter branches. Regardless of the exact proportions, airway branching follows a pattern whereby there is progressively greater cross-sectional area as air flows from proximal to distal airway branches, and the relationship is approximately exponential. This results in an exponential decrease in airway resistance with successive airway generations. Quantifying this exponential increase in airway cross-sectional area (T-Slope) is a structural correlate of the functional reduction in airway resistance from proximal to distal airway generations. Several disease processes in COPD can impact this slope. The mechanisms are likely multifactorial and include airway narrowing due to submucosal gland hyperplasia, intraluminal mucus plugs, airway fibrosis, and a loss of tethering of airways from the adjacent parenchyma due to emphysema. Some of these factors are potentially reversible.

Several attempts have been made in the past to quantify airway lumen tapering. Most of the studies were performed for the purpose of quantifying airway lumen dilatation and for the detection of bronchiectasis. Kuo and colleagues measured intra-branch tapering as the percentage reduction in airway diameter per millimeter along the centerline of a branch, and inter-branch tapering as the reduction in average diameter of the daughter branch compared with its parent (27). A similar method of inter-branch tapering was used to quantify traction bronchiectasis in adults with pulmonary fibrosis (28). Odry et al. computed the slope of the airway diameters along several airway paths to determine if there was an unexpected reduction in the expected narrowing from proximal to distal airways (29). Oguma et al. calculated the internal radii of the airway lumen from the carina distally to the right posterior basal bronchus and computed the fluctuation of radii from the regression line of these radii to calculate airway irregularity (30). They found that airways of patients with COPD are more irregular than healthy controls and those with asthma. Weinheimer and colleagues developed a taper index by converting airway trees into an acyclic graph to label individual branches and further utilizing an integral-based airway wall quantification method to measure the tapering index in CT scans of patients with cystic fibrosis (31). They found moderate correlations with the visual Brody bronchiectasis scores. Quan and colleagues used exponential regression to model the relationship between the airway length and the cross-sectional area along individual airway segments to identify bronchiectasis (32). Such methods, although useful for identifying bronchiectasis, have a major limitation in that airway tapering identified along individual segments does not indicate airway tapering along the entire tracheobronchial tree.

Our study has several strengths. We extend the literature by showing that the cumulative slope of the airway lumens of the tracheobronchial tree can be quantified and that this slope is associated with important clinical outcomes in COPD. We analyzed participants enrolled in the COPDGene study, who underwent CT and spirometry acquisition according to standardized protocols. All images were subject to stringent quality control. We included individuals with a broad spectrum of disease severity. The study also has a few limitations. First, the current methodology for estimating tapering through the radius of airway branches at each centerline point may inadequately address the intricate morphology of tortuous and non-circular airway cross-sections. Enhancing the area estimation techniques in future research may improve the accuracy of tapering slope assessment. Second, COPDGene included current and former smokers and hence these findings should be replicated in individuals with COPD resulting from causes other than exposure to cigarette smoke. Third, adults with clinically significant and severe bronchiectasis were excluded from the cohort. Although bronchiectatic airways may affect the slope, the impact of isolated dilated airways may not be significant because our slopes were cumulative across multiple airway segments at any given generation. More research is needed to validate this metric in individuals with severe bronchiectasis.

In conclusion, we demonstrate that airway tapering can be quantified on chest CT and this metric is associated with important clinical outcomes in COPD.