Higher small pulmonary artery and vein volume on computed tomography is associated with mortality in current and former smokers

Summary

Background

In chronic obstructive pulmonary disease (COPD), vascular alterations have been shown to contribute to hypoxia and pulmonary hypertension, but the independent contribution of small vessel abnormalities to mortality remains unclear.

Methods

We quantified artery and vein dimensions on computed tomography (CT) down to 0.2 mm. Small vessel volumes (<1 mmᴓ) were normalized by body surface area. In 7903 current and former smokers of the COPDGene study (53.2% male) the independent contribution of small artery and small vein volume to all-cause mortality was tested in multivariable Cox models. Additionally, we calculated the 95^th percentile of small arteries and veins in 374 never smokers to create two groups: normal and high small artery or vein volume. We describe clinical, physiological and imaging characteristics of subjects with a high small artery and high small vein volume.

Findings

Both high small artery and high small vein volumes were independently associated with mortality with an adjusted hazard ratio of 1.07 [1.01, 1.14] and 1.34 [1.21, 1.49] per mL/m² increase, respectively. In COPDGene, 447 (5.7%) had high small artery volume and 519 (9.1%) subjects had high small vein volume and both had more emphysema, more air trapping and more severe coronary calcium.

Interpretation

In smokers, abnormally high volumes in small arteries and veins are both relevant for mortality, which urges investigations into the aetiology of small pulmonary vessels and cardiac function in smokers.

Funding

Award Number U01-HL089897 and U01-HL089856 from the NHLBI. COPD Foundation with contributions from AstraZeneca, Boehringer Ingelheim, Genentech, GlaxoSmithKline, Novartis, Pfizer, Siemens, and Sunovion.

Research in context.

Evidence before this study

We performed a PubMed search for all studies about small pulmonary vessels/arteries/veins on imaging or histology, in relation to smoking, chronic obstructive pulmonary disease or pulmonary hypertension.

No language constrictions were applied. Established post-mortem studies have described increased small vein dimensions in subjects with emphysema, with several explanations including collaterals with the bronchial veins. Most recent studies on pulmonary vasculature have either focused on the small arteries or small vessels without separation of the arteries and veins. Some studies have used methods to separately quantify artery and veins, but did not investigate the independent contribution to mortality.

Added value of this study

This study uses a novel method to investigate small pulmonary artery and vein volumes (<ᴓ1mm) on non-contrast CT scans. Survival models corrected for clinical and technical confounders show that both small arteries and veins contribute independently to mortality in smokers.

Implications of all the available evidence

Our findings show that small pulmonary arteries and veins (<ᴓ1mm) have an independent effect on mortality in smokers, and urges further investigations into their aetiology.

Introduction

In 1959 Liebow stated: “Of greater interest than the arterial is the venous side of the collateral circulation in emphysema, which appears to be expanded far beyond what is observed on other types of pulmonary disease … it exists in the absence of heart failure”.¹ Since then, we have learned much about the biology and consequences of smoking induced injury to the airways and subsequent development of emphysema by using imaging and physiological tests.² Nevertheless, our knowledge on small pulmonary arteries and veins in smoking related lung disease, chronic obstructive pulmonary disease (COPD) and association with poor outcomes such as mortality remained limited.

The small pulmonary arteries run in the centre of the secondary pulmonary lobule where smoking-related destruction of the bronchioles and alveoli begins.² Post-mortem vessel casts and histological analysis demonstrated that small pulmonary arteries remodel and thicken their wall and thus increase in volume. On the other hand, hypoxic vasoconstriction leads to a reduction in small artery volume, at least initially, and it has been suggested that small arteries may disappear in smokers’ lungs, or at least decrease in density due to hyperinflation.3, 4, 5, 6, 7 One in vivo study showed that lower small artery volume divided by total artery volume predicted emphysema progression and pulmonary function decline.⁸ However, this lower ratio could be explained either by enlargement of the central pulmonary arteries and/or by less small artery volume. Indeed, another study showed an increase in small artery volume with increasing COPD severity.⁹ Etiologically it was suggested that small artery remodelling is a consequence of bronchiolar obliteration and emphysema and not a cause, although definite proof is lacking.¹⁰ In short, the effect of remodelling on small artery volumes in the smokers lungs is unclear–remodelling may lead to an increase in volume and obliteration to a decrease.

The pulmonary veins are located in the interstitium at the outer border of the lobule, about 1 cm away from where bronchiolitis and emphysema start. Injected cast studies demonstrated an increase in small vein dimensions in COPD.^1,3,4 There are several possible explanations for this observation. Emphysema development and hyperinflation are thought to stretch the border of the lobule which may lead to a larger venous blood content.³ Also, lower blood-pumping by reduced lung movement could lead to venous stasis.³ Interestingly, the casts demonstrated that collaterals exist with the bronchial veins, proposing that shunting could be present.⁴ Venous remodelling (intima fibrosis and arterialization) has also been shown to occur in advanced COPD in combination with worse haemodynamics.¹¹ Another extra pulmonary cause for venous congestion could be impaired left ventricular function due to coronary artery disease, arrhythmia and/or diastolic dysfunction, and is highly prevalent especially in advanced COPD.¹²

We hypothesized that due to the abovementioned phenomena of remodelling and hemodynamic changes in advanced COPD, small artery and vein volumes would be associated with mortality. We separately quantified volume distribution of the pulmonary arteries and veins on non-contrast CT. In 7903 current and former smokers we did a preliminary investigation to test whether small artery and vein volumes were related to higher all-cause mortality independent of emphysema, airway disease and coronary calcium as a surrogate marker of heart disease. We subsequently describe the clinical, imaging and physiological characteristics of those with increased small pulmonary artery and small pulmonary vein volume.

Methods

Study population

We included two cohorts in this study. Cohort one (“Reference cohort”) consisted of never smokers derived from Phase 1 and 2 of the COPDGene study (ClinicalTrials.gov: NCT00608764),¹³ and was used to determine cut-offs for increased high small artery and vein volumes. Cohort two (“COPDGene cohort”) was a cohort of current and former smokers with and without COPD derived from Phase 1 of the COPDGene study. The COPDGene study is a longitudinal multi-centre study, consisting of participants between the ages of 45–80 years, with at least 10 pack-years of smoking history (except for the never smokers). Among the exclusion criteria were a history of lung disease except asthma, active cancer treatment and suspected lung cancer. The full study protocol including inclusion and exclusion criteria is available at www.copdgene.org.

Physiological, clinical, imaging data and mortality

Spirometry data was collected using a standardized protocol and spirometer (ndd EasyOne™ Spirometer, Zurich, Switzerland). Details are described in a previous paper.¹³ Forced expiratory volume in 1 s (FEV₁) and forced vital capacity (FVC) are presented as a percentage of their predicted values (%predicted), post-bronchodilator.¹⁴ The pulmonary function test results were categorized in the Global initiative for Obstructive Lung Disease stage 0–4 or Preserved Ratio Impaired Spirometry (PRISm).

COPD was defined if FEV₁/FVC <0.7 and classified according to the GOLD criteria (GOLD I: FEV₁pred ≥80%; GOLD II: 50% ≤ FEV₁pred <80%; GOLD III: 30% ≤ FEV₁pred <50%; GOLD IV: FEV₁pred <30%). Subjects were defined as ‘PRISm’ if they a preserved ratio (FEV₁/FVC ≥0.7), but impaired spirometry (FEV₁pred <80%).¹⁵

Race, sex and packyears (including current and former smokers) were self-reported. Height and weight were measured and Body Mass Index (BMI) was calculated. Functional exercise performance was assessed with the 6-min Walking Distance (6 MWD). Activity-related dyspnoea was self-assessed using the modified Medical Research Council (mMRC) dyspnoea scale.¹⁶

CT scan protocols have been described previously.¹³ Briefly, multidetector CT scanners (>16 detector channels) were used to acquire non contrast-enhanced volumetric CT scans at inspiration (200 mAs) and at the end of expiration (50 mAs). Image reconstruction utilizes sub-millimeter slice thickness. Coronary artery calcification was quantified using the Agatston algorithm on ungated CT scans, therefore referred to as the ‘modified Agatston’ (mAgatston).¹⁷ Severe to extensive coronary artery disease was defined as >300 Agatston Units similar to the CAD-RADS™ 2.0.¹⁸ Emphysema was expressed as percentage of voxels with an attenuation lower than −950 Hounsfield Units (HU) on inspiratory CT scans (%LAA950). Air trapping was quantified on expiratory CT scans and expressed as percentage of voxels with an attenuation lower than −856 HU. Airway wall thickness was expressed as the square root wall area standardized to a ‘theoretical’ airway with a diameter of 10 mm (Pi10).¹⁹ All quantitative CT analyses were performed using Thirona’s lung quantification platform LungQ (Thirona, The Netherlands, http://www.thirona.eu). Emphysema was also visually assessed by expert readers according the Fleischner Society recommendations as: none, trace, mild, moderate, confluent or advanced destructive.²⁰

Subjects were followed until death if possible and cause of death was determined by an adjudication committee.²¹

CT pulmonary artery and vein quantification

The artery-vein phenotyping analysis (AVX) was performed using the lung quantification platform LungQ (Thirona, Nijmegen, The Netherlands). AVX consists of three main components: 1) voxel-wise segmentation of pulmonary arteries and veins, 2) separation of the identified vascular tree into individual branches, and 3) quantification of vascular diameters and volumes per branch.22, 23, 24, 25 In short, AVX starts by automatically segmenting the arterial and venous trees on non-contrast CT. The vascular segmentation component is a deep learning-based algorithm which was specifically designed to identify vascular structures and separate them into arteries and veins. An example of an artery-vein segmentation as overlay on a CT scan is provided in the online supplement. Next, the voxel-wise segmentation of the arteries and veins is split into individual branches based on detection of branching points within the vascular tree. The algorithm measures the vascular outer diameter of each centreline point in each branch, and these diameters are combined to obtain the branch diameter. The final step consists of quantifying all arterial and venous branches. For each branch, the vascular volume is quantified in relation to the outer vascular diameter, allowing for a detailed characterization of the pulmonary vascular volume distribution. Thirona uses a proprietary vessel intensity profile quantification algorithm that allows for sub-resolution quantification of vessel diameters up to 0.2 mm. For each subject, we quantified small artery volume as the accumulated volume of all pulmonary arterial branches with a diameter <1 mm (AVX_SA), and small vein volume as the accumulated volume of all pulmonary venous branches with a diameter <1 mm (AVX_SV). AVX_SA and AVX_SV were extracted as absolute pulmonary vascular volumes. However, to accommodate for inherent differences in body size, AVX_SA and AVX_SV were normalized by dividing it by body surface area (AVX [BSA] _SA, AVX [BSA] _SV), which was calculated according to the DuBois method (BSA [m2] = 0.007184 ∗ Height [cm]^0.725 ∗ Weight [kg]^0.425).²⁶

Statistical analysis

Continuous data are presented as mean ± standard deviation or median (25^th-75^th percentile), depending on normality. Normality was checked with histograms and Q–Q plots. All variables were normally distributed, except for emphysema (%LAA950), air trapping and mAgatston score.

Frequency and percentages were presented for nominal variables.

Missing values were present in eight variables. In six variables, the percentage of missing values was <1%. In two variables, the percentage missing data was 8.8% and 14.7%.

Cox proportional hazard models were performed to investigate the association of small artery and vein dimensions (cut-offs at diameters of <1 mm, <2 mm and <3 mm) with all-cause mortality accounting for age, sex, race, BMI, smoking status, packyears smoked, FEV₁ %predicted, mMRC Dyspnoea Score, 6 MWD, %LAA950, Pi10, mAgatston score, severe exacerbations, pixel spacing and scanner model. We explored this association separately for arteries and veins. The proportional hazard assumptions were evaluated using the Schoenfeld residuals. We investigated via visual inspection and the likelihood ratio test whether there are non-linear effects of small artery and small vein volume. In particular, for the non-linear effect we used natural cubic splines basis with 2 or 3 degrees of freedom. Adding non-linear effects did not seem to improve the model. Data was considered statistically significant with a p value < 0.05.

We determined the 95^th percentile for small artery volume and for small vein volume in male and female non-smokers in the Reference cohort. In the COPDGene cohort, we subsequently described the clinical, physiological and imaging characteristics from smokers with high small artery and high small vein volume. Data were analysed using IBM SPSS Statistics 26.0.0.1 (IBM Corp., Armonk, New York, USA) and R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria).

Ethics

The COPDGene study was approved by the institutional review boards at each of the 21 clinical sites (Supplementary Table S5). All subjects have provided written informed consent.

Role of funders

The funding sources had no role in the study design, data collection and analysis, interpretation of the data, or writing of the manuscript.

Results

COPDGene and reference cohort

The COPDGene cohort consisted of 7903 current and former smokers classified according to the Global Initiative on Obstructive Pulmonary Disease (GOLD) severity stages (3313 GOLD 0; 947 PRISm; 608 GOLD 1; 1541 GOLD 2; 947 GOLD 3; 498 GOLD 4). Their baseline characteristics are provided in Table 1. The reference cohort consisted of 374 never smoking males (43%) and females (57%). The mean age was 59.9 ± 9.4 years and 26.2% were found to be obese (BMI>25 kg/m²). Of all adjudicated deaths in the COPDGene cohort, 388 (36.2%) were due to respiratory cause and 178 (16.6%) due to a cardiovascular cause. Further baseline characteristics are shown in Table 1.

Table 1.

Clinical characteristics of the study population.

	Reference cohort (N = 374)	Copdgene cohort (N = 7903)
Age (yr)	59.9 ± 9.4	60.1 ± 9.0
Male sex (N, %)	161 (43.0%)	4206 (53.2%)
Bmi (kg/m2)	27.2 ± 4.5	28.8 ± 6.3
Obesity (BMI > 30)	98 (26.2%)	2937 (37.2%)
Packyears (yr)	N/A	44.7 ± 25.2
No. of non-hispanic white subjects (N, %)	307 (82.1%)	5388 (68.2%)
Current smokers (n,%)	N/A	4018 (50.8%)
Spirometry
FEV1 (% predicted)	103.3 ± 13.0	75.5 ± 25.7
FVC (% predicted)	99.2 ± 12.8	86.5 ± 18.2
CT quantified parameters
Emphysema (%)	0.9 (0.3–2.5)	2.3 (0.6–7.7)
Air trapping (%)	7.5 (3.9–12.7)	15.3 (7.0–33.3)
Pi10 (mm)	1.82 ± 0.35	2.36 ± 0.61
mAgatston score	N/A	17 (0–190)
mAgatston score > 300	N/A	1335 (18.5%)
95th pctl of small arterial volume–males (mL/m2)	5.52	N/A
95th pctl of small arterial volume–females (mL/m2)	5.27	N/A
95th pctl of small venous volume–males (mL/m2)	3.58	N/A
95th pctl of small venous volume–females (mL/m2)	3.45	N/A
6 MWD (ft)a	1725 ± 333	1338 ± 396
mMRC dyspnoea scale	0 (0–0)	1 (0–3)
Emphysema visual (N, %)
0—none	N/A	2619 (33.2%)
1—trace	N/A	1409 (17.8%)
2—mild	N/A	1520 (19.2%)
3—moderate	N/A	1262 (16.0%)
4—confluent	N/A	808 (10.2%)
5—advanced destructive	N/A	282 (3.6%)
COPD GOLD stage (N, %)
Prism	N/A	947 (12.1%)
0	N/A	3313 (42.2%)
1	N/A	608 (7.7%)
2	N/A	1541 (19.6%)
3	N/A	947 (12.1%)
4	N/A	498 (6.3%)
5-year mortality	8 (2.4%)	923 (12.6%)
10-year mortality	14 (19.4%)	1964 (30.7%)
Total mortality	20 (5.3%)	2419 (30.6%)

Legend: Data given are mean ± standard deviation, median and interquartile range in parentheses, or number and percentage in parenthesis, depending on the data distribution. BMI is body mass index, FEV₁ is Forced Expiratory Volume in 1 s, FVC is Forced Vital Capacity, CT is computed tomography, Pi10 is square root of wall area for a theoretical airway with a lumen of 10 mm, pctl is percentile, 6 MWD is 6 min walking distance, mMRC is Modified Medical Research Counsel Dyspnoea scale, COPD is Chronic Obstructive Lung Disease, GOLD is Global Initiative for Chronic Obstructive Lung Disease, PRISm is preserved ratio impaired spirometry.

^aOne foot is 0.3048 m.

High small artery and vein volume are independently associated with mortality

During a follow-up of 15 years, mortality was 30.6% (2419 deaths in the COPDGene cohort) (Table 1). In the groups with high small artery and high small vein volume, mortality was 34.0% and 55.1%, respectively. The proportional hazard assumptions were not violated. High small artery volume and high small vein volume were both associated with higher mortality independent of confounders (Table 2). The multivariable adjusted hazard ratio was 1.07 [1.01, 1.14] for each 1 mL/m² increase in small arterial volume and 1.34 [1.21, 1.49] for each 1 mL/m² increase in small vein volume (Table 2). For small artery volume, the effect was stronger in females than in males (Supplementary Tables S1 and S2). Larger arterial and venous volumes (cut-offs at 2 mm and 3 mm diameter) showed lower hazard ratios (Supplementary Tables S3 and S4). The adjusted association between the hazard for all-cause mortality and small artery and vein volume is presented in Fig. 1. Adjusted Kaplan–Meier curves show a lower survival probability for the 90^th percentile of small artery and vein volume, compared to the 10^th percentile (Supplementary Figs. S3 and S4). Illustrative examples of a subject with high small artery volume and a subject with high small vein volume, both of whom died from respiratory disease are presented in Fig. 2.

Table 2.

Hazard ratios with confidence intervals of multivariable adjusted Cox regression models.

Determinant	Unit change	Small veins		Small arteries
		Hazard ratio	CI	Hazard ratio	CI
Small vein/artery volume	Per 1 mL/m2	1.341	[1.207, 1.489]	1.072	[1.009, 1.139]
Age	Per year	1.041	[1.034, 1.047]	1.041	[1.034, 1.048]
Sex	For females	0.786	[0.707, 0.875]	0.755	[0.679, 0.839]
BMI	Per kg/m2	0.979	[0.970, 0.988]	0.974	[0.965, 0.983]
Race	For African Americans	1.170	[1.033, 1.325]	1.117	[0.984, 1.269]
Packyears smoked	Per packyear	1.003	[1.001, 1.005]	1.003	[1.002, 1.005]
Smoking status	For current smokers	0.628	[0.564, 0.699]	0.621	[0.557, 0.693]
FEV1 %predicted	Per %	0.988	[0.985, 0.991]	0.988	[0.985, 0.991]
Emphysema	Per %	1.007	[1.002, 1.012]	1.010	[1.005, 1.015]
Pi10	Per mm	1.035	[0.943, 1.135]	1.042	[0.950, 1.143]
Coronary calcium	Per Agatston unit	1.038	[1.019, 1.058]	1.040	[1.021, 1.060]
6 MWD	Per ft.	0.999	[0.999, 0.999]	0.999	[0.999, 0.999]
Severe exacerbations	For severe exacerbations	1.272	[1.132, 1.430]	1.266	[1.127, 1.423]
mMRC Dyspnoea Score 1a	For score 1	1.000	[0.861, 1.162]	1.003	[0.863, 1.166]
mMRC Dyspnoea Score 2a	For score 2	1.246	[1.075, 1.445]	1.262	[1.088, 1.462]
mMRC Dyspnoea Score 3a	For score 3	1.469	[1.278, 1.689]	1.498	[1.304, 1.721]
mMRC Dyspnoea Score 4a	For score 4	1.645	[1.393, 1.943]	1.676	[1.419, 1.978]

Legend: BMI is body mass index, FEV₁ is Forced Expiratory Volume in 1 s, Pi10 is square root of wall area for a theoretical airway with a lumen of 10 mm, 6 MWD is 6 min walking distance, mMRC is Modified Medical Research Counsel Dyspnoea scale. Pixel spacing and CT scanner models were also added as confounders.

^amMRC Dyspnoea Score 0 as reference.

Fig. 1.

Small artery and vein volumes and the increased hazard ratio for all-cause mortality adjusted for confounders. Legend: Effect plots for small artery volume (top) and small vein volume (bottom) corrected for age, sex, race, BMI, smoking status, packyears, FEV1 %predicted, mMRC Dyspnoea Score, 6 MWD, %LAA950, Pi10, mAgatston score, severe exacerbations, pixel spacing and scanner model.

Fig. 2.

Results of the artery-vein phenotyping analysis (AVX, Thirona, The Netherlands). Examples of quantified pulmonary arteries and veins on non-contrast CT of a surviving female from the reference cohort with normal small artery and vein volume (top figure), a deceased female from the COPDGene cohort with high small vein volume (middle figure) and a deceased male from the COPDGene cohort with high small artery volume (bottom figure). The AVX analysis quantified the volume of the small pulmonary arteries and veins (diameter <1 mm) and are presented as the lightest shade of blue and red. The remaining larger arteries and veins are presented in the two darker shades of blue and red.

Characteristics of smokers with high small pulmonary artery volume

In the reference cohort, the 95^th percentile cut-off for abnormally high small artery volume was 5.52 mL/m² (males) and 5.27 mL/m² (females). Based on this cut-off, high small artery volume occurred in the COPDGene cohort in 219 (5.2%) males and 228 (6.2%) females (Table 3). Compared to subjects with normal small artery volumes, those with a high small artery volume were older and had more packyears smoked. Their lungs were somewhat more diseased as reflected by more emphysema and air trapping. Clinically, dyspnoea and 6 MWD were similar to those with a normal small artery volume. In participants with confluent emphysema, 9.5% of subjects had a high small artery volume. Severe coronary calcium was slightly higher in those with high small artery volume. Of all adjudicated deaths in those with a high small artery volume, 33 (50.0%) were due to respiratory cause and 7 (10.6%) due to a cardiovascular cause.

Table 3.

Clinical, physiological and imaging characteristics of current and former smokers with normal and high small pulmonary arterial volumes, corrected for BSA (mL/m²).

	Normal small artery volume (N = 7456)	High small artery volume (N = 447)
Age (yr)	60.0 ± 9.1	62.1 ± 8.1
Male sex (N, %)	3987 (53.5%)	219 (49.0%)
BMI (kg/m2)	29.0 ± 6.3	25.6 ± 5.3
Obesity (BMI > 30)	2863 (38.4%)	74 (16.6%)
Packyears (yr)	44.3 ± 25.1	52.1 ± 27.0
No. of non-hispanic white subjects (N, %)	4968 (66.6%)	420 (94.0%)
Current smokers (n,%)	3738 (50.1%)	280 (62.6%)
Spirometry
FEV1 (% predicted)	75.5 ± 25.6	76.1 ± 26.5
FVC (% predicted)	86.2 ± 18.1	91.7 ± 18.6
CT quantified parameters
Emphysema (%)	2.2 (0.6–7.4)	5.2 (1.7–12.3)
Air trapping (%)	15.1 (6.9–32.9)	20.6 (9.2–39.7)
Pi10 (mm)	2.36 ± 0.61	2.29 ± 0.62
mAgatston score	15 (0–186)	47.5 (0–253)
mAgatston score >300	1237 (18.3%)	98 (23.1%)
Mean small arterial volume (mL/m2)	3.72 ± 0.74	5.91 ± 0.52
Total arterial volume (mL/m2)	26.23 ± 6.49	36.27 ± 7.12
6 MWD (ft)a	1332 ± 395	1444 ± 393
mMRC dyspnoea scale	1 (0–3)	1 (0–2)
Emphysema visual (N, %)
0—none	2529 (33.9%)	90 (20.1%)
1—trace	1352 (18.1%)	57 (12.8%)
2—mild	1424 (19.1%)	96 (21.5%)
3—moderate	1153 (15.5%)	109 (24.4%)
4—confluent	731 (9.8%)	77 (17.2%)
5—advanced destructive	264 (3.5%)	18 (4.0%)
COPD GOLD stage (N, %)
PRISm	920 (12.3%)	27 (6.0%)
0	3154 (42.3%)	159 (35.6%)
1	537 (7.2%)	71 (15.9%)
2	1437 (19.3%)	104 (23.3%)
3	897 (12.0%)	50 (11.2%)
4	464 (6.2%)	34 (7.6%)
Medical history
Coronary artery disease	528 (7.2%)	32 (6.2%)
Stroke	199 (2.7%)	6 (1.3%)
Cancer	379 (5.1%)	24 (5.4%)
5-year mortality	858 (12.4%)	65 (15.0%)
10-year mortality	1838 (30.6%)	126 (31.7%)
Total mortality	2267 (30.4%)	152 (34.0%)

Legend: Data given are mean ± standard deviation or median and interquartile range in parentheses or number and percentage in parenthesis, depending on the data distribution. BMI is body mass index, FEV₁ is Forced Expiratory Volume in 1 s, FVC is Forced Vital Capacity, CT is computed tomography, Pi10 is square root of wall area for a theoretical airway with a lumen of 10 mm, 6 MWD is 6 min walking distance, mMRC is Modified Medical Research Counsel Dyspnoea scale, COPD is Chronic Obstructive Lung Disease, GOLD is Global Initiative for Chronic Obstructive Lung Disease, PRISm is preserved ratio impaired spirometry.

^aOne foot is 0.3048 m.

Characteristics of smokers with high small pulmonary vein volume

In the reference cohort, the 95^th percentile cut-off for abnormally high small vein volume was 3.58 mL/m² (males) and 3.45 mL/m² (females). Based on this cut-off, high small vein volume was observed in 265 (6.3%) males and 254 (6.9%) females (Table 4). Compared to participants with normal small vein volumes, those with high small vein volume were older and had smoked more. In addition, their lungs were much more diseased as reflected in worse spirometry, more emphysema, more air trapping, and thicker airway walls. In subjects with advanced destructive emphysema, 28.0% of subjects had a high small vein volume. With increasing severity of COPD, the number of subjects with a high small vein volume increased up to 23.5% of subjects in GOLD 4. Severe coronary calcification (mAgatston score >300) was more common. Of all adjudicated deaths in those with a high small vein volume, 88 (58.7%) were due to a respiratory cause and 14 (9.3%) due to a cardiovascular cause.

Table 4.

Clinical, physiological and imaging characteristics of current and former smokers with normal and high small pulmonary venous volumes, corrected for BSA (mL/m²).

	Normal small vein volume (N = 7384)	High small vein volume (N = 519)
Age (yr)	59.9 ± 9.0	63.9 ± 8.3
Male sex (N, %)	3941 (53.4%)	265 (51.1%)
BMI (kg/m2)	29.2 ± 6.2	24.0 ± 5.1
Obesity (BMI > 30)	2888 (39.1%)	49 (9.4%)
Packyears (yr)	44.0 ± 24.8	55.3 ± 28.5
No. of non-hispanic white subjects (N, %)	4919 (66.6%)	469 (90.4%)
Current smokers (n,%)	3795 (51.4%)	223 (43.0%)
Spirometry
FEV1 (% predicted)	76.7 ± 24.9	59.3 ± 30.5
FVC (% predicted)	86.6 ± 17.9	85.1 ± 22.2
CT quantified parameters
Emphysema (%)	2.0 (0.6–6.7)	13.6 (4.2–27.8)
Air trapping (%)	14.6 (6.7–30.7)	43.2 (19.9–60.7)
Pi10 (mm)	2.35 ± 0.61	2.50 ± 0.63
mAgatston score	15 (0–182)	86 (0–372)
mAgatston score > 300	1193 (17.8%)	142 (29.2%)
Mean small venous volume (mL/m2)	2.60 ± 0.43	3.83 ± 0.31
Total venous volume (mL/m2)	22.39 ± 5.06	30.87 ± 5.65
6 MWD (ft)a	1341 ± 393	1296 ± 427
mMRC dyspnoea scale	1 (0–3)	2 (0–3)
Emphysema visual (n, %)
0—none	2571 (34.8%)	48 (9.2%)
1—trace	1371 (18.6%)	38 (7.3%)
2—mild	1457 (19.7%)	63 (12.1%)
3—moderate	1122 (15.2%)	140 (27.0%)
4—confluent	657 (8.9%)	151 (29.1%)
5—advanced destructive	203 (2.7%)	79 (15.2%)
COPD GOLD stage (N, %)
PRISm	939 (12.7%)	8 (1.5%)
0	3211 (43.5%)	102 (19.7%)
1	547 (7.4%)	61 (11.8%)
2	1426 (19.3%)	115 (22.2%)
3	833 (11.3%)	114 (22.0%)
4	381 (5.2%)	117 (22.5%)
Medical history
Coronary artery disease	528 (7.2%)	32 (6.2%)
Stroke	197 (2.7%)	8 (1.5%)
Cancer	363 (4.9%)	40 (7.7%)
5-year mortality	787 (11.5%)	136 (26.9%)
10-year mortality	1714 (28.9%)	250 (53.1%)
Total mortality	2133 (28.9%)	286 (55.1%)

Legend: Data given are mean ± standard deviation or median and interquartile range in parentheses or number and percentage in parenthesis, depending on the data distribution. BMI is body mass index, FEV₁ is Forced Expiratory Volume in 1 s, FVC is Forced Vital Capacity, CT is computed tomography, Pi10 is square root of wall area for a theoretical airway with a lumen of 10 mm, 6 MWD is 6 min walking distance, mMRC is Modified Medical Research Counsel Dyspnoea scale, COPD is Chronic Obstructive Lung Disease, GOLD is Global Initiative for Chronic Obstructive Lung Disease, PRISm is preserved ratio impaired spirometry.

^aOne foot is 0.3048 m.

Discussion

In this preliminary study we quantified small pulmonary arteries and veins in non-contrast CT down to 0.2 mm in diameter. Results showed that both high artery and high small vein volumes were significantly associated with increased mortality in the COPDGene cohort, independent of clinical and radiological markers of lung disease and cardiovascular disease. We found that subjects with high small artery and vein volumes are characterized by a higher age, more packyears smoked, more emphysema and more air trapping. Their coronary calcium score was also higher. This study cannot provide evidence on causality, but poses that small vessels in smokers’ lungs may require more attention as a hallmark of disease progression. It illustrates the notion that COPD is fundamentally also a perfusion-related disease, where the impaired function and structural changes in small pulmonary vessels lead to alterations in blood flow and gas exchange within the lungs, and ultimately, clinical outcomes.

Previous literature shows different views on the role of small arteries in the initiation and progression of lung disease in smokers. It is difficult to compare our findings to other studies, as most in vivo studies did not separate arteries from veins, but instead measured total small vessel volume (normalized by total vessel volume) on imaging.27, 28, 29, 30 An increase in small vessel volume in subjects with severe COPD has been described,³¹ as well as associations between higher small vessel volumes and cigarette smoke exposure.³² Additionally, a recent abstract that separated arteries and veins found an increase in small arterial volume in subjects with increasing pulmonary hypertension severity in patients with COPD.³³ This suggests that smoking-related lung disease does not necessarily lead to pruning or loss of arteries as has often been suggested. On the contrary, some studies have suggested that lower small to total artery ratio was associated with emphysema severity or right ventricular enlargement and an increased mortality.^8,27,34 A possible explanation could be a drop in arterial density relative to lung volume, which is a likely finding in hyper inflated lungs. Although this was named pruning of the small arteries, the observation could also be explained by a volume increase in the larger, central arteries. For example, in pulmonary hypertension the central arteries may dilate more than the small arteries, which could lead to a relative decrease in small artery volume.

Nevertheless, some investigators did find lower absolute small blood volumes in subjects with more emphysema or increased COPD severity, which may be due to arterial loss related to centrilobular emphysema,^27,35 but a definitive answer is lacking at this moment. Interestingly, it has also been suggested that damaged pulmonary vascular endothelium is involved in the formation of emphysema.³⁶ Possible explanations for our high small artery volume are increased blood volume due to arterial hypertension or due to tobacco-induced vasodilation (nitric oxide mediated) in current smokers, or an increased wall thickness due to arterial remodelling.³⁴ How these small arteries are related to mortality and whether such mechanisms could be targeted requires further investigation.

There has been limited interest in small pulmonary veins in smokers, although there is more data on non-pulmonary veins in smokers. We did observe a stronger mortality relation for small pulmonary veins compared to small pulmonary arteries. Our findings on high small vein volume are in line with established post-mortem studies.^3,4 The investigators hypothesized that dilatation of small pulmonary veins was caused by extensive collaterals between dilated bronchial veins and pulmonary veins, but clinical relevance was never further tested as these studies were post-mortem. For high pulmonary small vein (or vessel) volume, several factors were hypothesized as an explanation: vasodilation by toxins, increased vessel wall thickness (intima thickening and arterialization),¹¹ perivascular fibrosis, impaired left ventricular function, venous stretching around bulla, hyperinflation leading to traction or altered vascular appearance, stronger breathing efforts leading to recruitment, weaker breathing efforts leading to stasis and vascular thrombosis.³² To this we re-introduce the hypothesis raised from cast studies: venous shunting from the bronchial veins (mostly oxygen poor) to the pulmonary veins (mostly oxygen rich) and venous valve deficiency. Shunting could contribute to (local) hypoxia, potentially leading to remodelling of the small pulmonary veins. The remodelling of pulmonary veins has also been shown to correlate strongly with the severity of heart-failure related pulmonary hypertension,³⁷ suggesting that cardiac dysfunction could be involved. There is also accumulating evidence that venous changes are more systemic in nature. Multiple cohort studies consistently demonstrated that smokers have larger retinal venules.³⁸ It was suggested that veins dilate as a response to chronic hypoxia and hypoperfusion, polycythaemia, nicotine³⁸ or toxins in cigarettes,³² and it is increasingly recognized that veins play an important role in neuroinflammation.³⁹ Additionally, chronic venous insufficiency in the legs and varices of the tongue appear to be associated with smoking.^40,41 In short, the observed high small pulmonary venous volume is either related to lung disease, left-sided cardiac failure, or systemic effects of smoking. Its independent association with mortality suggests that it is relevant to further investigate the role of small pulmonary veins in smokers.

A strength of our study is that we used a large cohort of well characterized and carefully followed non-smokers and smokers. We used software that separates and quantifies pulmonary arteries and veins without the use of contrast media. This enabled us to shed light on small vessels in current and former smokers, with sufficient power for extensively adjusted mortality analyses.

Our study has several limitations. First, we do not provide histological proof for our findings. Further histological assessment of wall and lumen dimensions would be valuable, especially given the many hypotheses that exist around small vessel disease in smokers.^1,3,4,42 A possibility remains that the observed enlargement not only includes vessel volume, but also surrounding tissue, e.g., lymph fluid or interstitial fibrosis and therefore measurement error might play a role. To ensure the accuracy of the automatic artery-vein separation, experienced chest radiologists have visually assessed several CT scans and validated that the anatomical locations of the segmented arteries and veins were correct within the secondary pulmonary lobule. Second, different normalization methods are used, but we think that dividing vessel volumes by body surface area is a reasonable approach, which is more often used in vascular research.⁴³ Third, we chose a diameter of 1 mm as a cut-off for small vessels, which is smaller than previous investigators who chose a cross-sectional cut-off of 5 mm²,^27,32,44 which would correspond to a diameter of 2.52 mm for a round vessel. For better comparison with previous papers we also provide the analysis for arteries and veins with cut-offs at 2 mm and 3 mm (Supplementary Tables S3 and S4). This makes it possible to better compare with previous literature on BV5, pulmonary vessels with an area of 5 mm². We prespecified our analysis for the cut-off at 1 mm, as we hypothesized that most signal would be in the smallest vessels. Although the segmentation is less accurate due to the CT resolution, the number of branches increases sharply. Fourth, unobserved technical and clinical confounders remain a possibility, although we adjusted our analysis for pixel spacing (voxel size), CT scanner model and multiple clinical parameters. Fifth, our study does not translate directly to therapeutic interventions, although it is of interest that lung volume reduction surgery does improve venous pressures.⁴⁵ Sixth, because we could only include smokers who could undergo a CT scan successfully and had mortality data available after 15 years, a selection bias might have been introduced by the absence of smokers who were ill or possibly deceased. Finally, the quantitative AVX measurements are based on hundreds of thousands of data points per scan. There will be situations in which isolated small vessels may not identified. We think this does not influence our associations given the large sample size. For individual patients a visual check maybe needed.

In conclusion, the contribution of small pulmonary arteries and veins to smoking-related lung disease has remained largely obscure, especially on the venous side. Our data show that in smokers, abnormally high volumes in small arteries and small veins are both relevant for mortality, with the strongest signal observed in the veins. This independent effect on mortality urges further investigations into the aetiology of small pulmonary vessels and cardiac function in smokers.

Contributors

Study design: AK, DL, PdJ, EP.

Data acquisition: DL.

Data analysis (CT scans): JPC and LGE.

Data interpretation: AK, EA, TvdV, LGE, JPC, SH, DL, HT, PdJ and EP.

Verification of underlying data: AK, EP, PdJ.

Statistical analysis: AK, EA.

Manuscript writing: AK, LGE, JPC, PdJ and EP.

Manuscript revisions: AK, EA, TvdV, LGE, JPC, SH, DL, HT, PdJ and EP.

Final responsibility for the decision to submit for publication: all authors.

All authors read and approved the final version of the manuscript.

Data sharing statement

All publications based on COPDGene work will be subject to the NIH Public Access Policy described in Notice NOT-OD-08-033. We will submit to the NIH National Library of Medicine’s PubMed Central an electronic version of the final, peer-reviewed manuscripts upon acceptance for publication, to be made publicly available no later than 12 months after the official date of publication. Novel research resources created or to be created under this project include: (1) Whole genome sequencing analysis for the entire COPDGene cohort (completed through TOPMed); (2) RNA-Seq, Metabolon metabolomics, DNA methylation arrays, and SOMAScan proteomics on blood samples from COPDGene Phase 2 samples and Phase 3 samples; (3) Clinical assessments of the COPDGene cohort (without jeopardizing confidentiality of individuals); (4) Chest CT scans from the initial clinical visit, five-year follow-up, ten-year follow-up and fifteen-year follow-up of the COPDGene cohort; and (5) Serum, plasma, and DNA samples for the initial clinical visit, five-year follow-up, ten-year follow-up and fifteen-year follow-up of the COPDGene cohort. Open access to all of the genetic and phenotypic data for COPDGene will be provided through dbGaP. Serum, plasma, DNA, and RNA samples will also be shared in response to investigator-initiated proposals. Proposals for access to these biological samples will be reviewed by the COPDGene Executive Committee and will be approved for qualified applicants provided the request for sharing does not overlap with previously approved projects. The serum, plasma, RNA, and DNA samples will be limited in volume and thus requests for access to these samples will be evaluated for scientific merit in competition with other proposed uses of these biological samples. All shared data will be de-identified with respect to specific research subjects.

Declaration of interests

AK, EA, TV, EP: no conflicts relevant to this manuscript. DL and SH: the department of radiology of National Jewish Health received grant support from the NHLBI. PdJ: The Department of Radiology from the UMC Utrecht receives research support from Philips Healthcare. JPC: employee and shareholder of Thirona. LGE, HT: employee of Thirona.