Pulmonary vascular features on chest computed tomography differentially associate with adverse outcomes in smokers in COPDGene

Abstract

Rationale:

In smokers with and without chronic obstructive pulmonary disease (COPD), the differential strengths of association between chest computed tomography (CT)–based metrics of pulmonary vascular disease and adverse outcomes are unknown.

Objectives:

We aimed to quantify the differential strengths of association of CT features, from the distal pulmonary arteries to the central great vessels and cardiac chambers, with acute respiratory exacerbations (AREs) and mortality in smokers with and without COPD.

Methods:

Smokers with and without COPD with pulmonary vascular morphology and outcomes data were identified in COPDGene. Negative binominal and multivariable Cox proportional hazard models were used to investigate the association of CT features, including volume of the distal pulmonary arterial vasculature or pruning (<5 mm² normalized to total arterial blood vessel volume [aBV5/aTBV]), preacinar vessels (5–20 mm²), and pulmonary artery to aorta (PA/Ao) and right to left ventricular epicardial volume (RV/LV) ratios, with outcomes. Kaplan–Meier curves were used to describe pruning risk on mortality.

Results:

A total of 3169 smokers with COPD and 2530 smokers without COPD were analyzed. Among smokers with COPD, PA/Ao was the only imaging feature significantly associated with AREs (incidence rate ratio, 1.08 [95% CI, 1.04–1.12]), even after adjusting for aBV5/aTBV. Conversely, pruning demonstrated the strongest association with mortality, even in smokers without COPD (hazard ratio, 1.22 [95% CI, 1.14–1.30] and 1.26 [95% CI, 1.11–1.42], respectively). The association of preacinar vessels with mortality in smokers with COPD and in those without COPD, but with significant emphysema on imaging (≥5%), was novel.

Conclusions:

Pruning is significantly associated with mortality risk in smokers with and without COPD; however, PA/Ao selectively associates with AREs in COPD, even when accounting for distal vasculopathy.

Introduction

Chest computed tomography (CT)–based metrics of pulmonary vasculopathy inform clinical risk in subjects with chronic obstructive pulmonary disease (COPD).^1–6 For example, in COPD, loss of small pulmonary arterial vascular volume (<5 mm²) or pruning is associated with disease progression, severity, and death.^1,3,7 Similarly, an increased pulmonary artery to aorta (PA/Ao) ratio >1 is associated with increased risk of COPD exacerbations and treatment failure.^4,5 However, the relative associations of these vascular features with adverse outcomes remains unknown. Furthermore, strength of association of these CT metrics with clinical endpoints in smokers without COPD is not known. Differential strength of vascular features’ association with adverse outcomes, from the central hilum to the distal vasculature, may suggest differential pathology across the vascular tree.

Presently, noninvasive CT markers of vasculopathy in smokers are underutilized to predict adverse outcomes. This is underscored by the presence of quantitative CT abnormalities in young smokers compared to control despite preserved pulmonary function.⁸ In fact, imaging data from a generally healthy population identified pruning as a predictor of all-cause mortality.⁹ The high prevalence symptom burden among these subjects suggests an opportunity to expand our present understanding of imaging marker association with outcomes in smokers with and without COPD.^10,11

COPDGene is a longitudinal multicenter observational study seeking to understand the clinical, epidemiologic, and biologic risk factors for the development of COPD in smokers.¹² Over 10000 ever-smokers were enrolled across a range of lung function from normal to advanced COPD. In this manuscript we explore how vascular features on CT differ in smokers with and without COPD and the relative prognostic value of these imaging markers for acute respiratory exacerbations (AREs) and mortality. We hypothesized that loss of small pulmonary arterial volumes or pruning would be most strongly associated with these outcomes owing to the early effects of tobacco exposure on the distal pulmonary vasculature.^13,14

Methods

Study population

The COPDGene study has been described in detail previously.¹² In brief, 10199 current and former smokers, aged 45–80 years, with at least a 10 pack-year history of smoking were recruited at 21 clinical centers in the United States. An additional 107 never-smoking controls were recruited. People underwent detailed baseline characterization, including demographics, anthropometrics, medical and smoking history, pre- and post-bronchodilator spirometry, 6-minute walk distance (6MWD) testing according to American Thoracic Society (ATS) standards, questionnaire-based self-assessments including the modified Medical Research Council dyspnea scale (mMRC), and inspiratory and expiratory CT scanning using a protocol standardized across the research network. Enrollment in phase 1 of the COPDGene Study took place between 2008 and 2011, and study participants were followed prospectively as part of the longitudinal follow-up program (LFU).¹⁵ The COPDGene study was approved by the institutional review board at each center, and all participants provided written informed consent.

For the current investigation, we included current and former smokers who had valid spirometry, complete baseline quantitative CT scan–measured pulmonary vascular morphology, and available vital status and acute respiratory exacerbation data. Separate modeling was done on smokers with and without COPD. Pre- and post-bronchodilator spirometry was performed according to ATS/European Respiratory Society criteria. Smoking status was self-reported. COPD was defined as those with a post-bronchodilator forced expiratory volume in 1 second (FEV₁) to forced vital capacity (FVC) ratio <0.7. We defined normal spirometry as a post-bronchodilator FEV₁/FVC ≥0.7 and FEV₁ ≥80% according to the reference equations. Never-smokers and those with preserved ratio impaired spirometry were excluded. The COPDGene study was approved by the institutional review board at each center, and all participants provided written informed consent.

CT analysis

Noncontrast chest CT were obtained at full inspiration on CT scanners. Objective assessments of pulmonary vascular morphology were obtained as described previously.^1,16 After lung segmentation, 2 intraparenchymal vascular features (pruning and preacinar dilation) and 2 extraparenchymal vascular features (pulmonary artery to aorta diameter ratio [PA/Ao] and right ventricular to left ventricular epicardial volume ratio [RV/LV]) were included. In brief, to obtain the intraparenchymal features, the vasculature was arterial-venous segmented from the surrounding parenchyma and metrics of blood vessel volume were obtained.¹⁶ Arterial pruning was defined as the ratio of volume of small pulmonary arterial vessels (cross-sectional area ≤5 mm²) to the total pulmonary arterial blood vessel volume (aBV5/aTBV); a lower ratio indicates a smaller volume of small vessels. Preacinar intraparenchymal arterial dilation was calculated as the pulmonary arterial vascular volume 5–20 mm² in cross-sectional area normalized to total arterial volume as previously described.² The PA/Ao ratio as well as the ratio of the RV to LV epicardial volumes were calculated as described previously.³ All metrics were previously validated in a reproducibility study.¹⁷ In addition, our segmentation approach has been validated to achieve high accuracy for the range of vessel sizes typically detectable within the contemporary CT resolution limits.¹⁸ Densitometric assessments of the lung parenchyma were additionally performed to quantify any degree of radiographic emphysema present in the study population, including in smokers without spirometric COPD. Low-attenuation areas representative of emphysema were defined as areas on lung CT below −950 Hounsfield units relative to the total lung volume (%LAA-950).¹⁹ To determine the interaction between the percentage of emphysema in smokers without COPD and vascular feature association with outcomes, a cut-off of 5% was used based on prior associations with mortality and exacerbation risk.²⁰

Primary endpoints

Acute respiratory exacerbations

Exacerbations were tallied prospectively from baseline using the LFU data. Number of AREs from baseline as well as years followed were collected from the LFU dataset. Exacerbations are defined in COPDGene as a new onset of or increase in cough, phlegm, or dyspnea. An episode that requires antibiotics and/or steroids is counted as an exacerbation. Participants were asked every 3–6 months about exacerbation episodes through the COPDGene LFU program.¹⁵

Mortality

Vital status was determined through the Social Security Death Index (SSDI) and the COPDGene LFU program.¹⁵ Deaths ascertained from the SSDI were back-censored 3 months to account for the expected lag time between a death and its reporting in the SSDI. Vital status, for those in whom follow-up was supplied via the LFU program, was back-censored 6 months prior to the mortality dataset pull to ensure subjects were being actively followed with biannual phone calls. Mortality refers to all-cause mortality.

Statistical methods

Data are presented as mean ± SD, median and IQR, number of participants (percentages), hazard ratios (HRs) and 95% CI, or incidence rate ratios (IRRs) and 95% CI where appropriate. Negative binomial modeling with an offset term to account for duration of observation was used to evaluate the association of each vascular feature with acute respiratory exacerbation risk. Cox proportional hazards models were used to evaluate the differential association of vascular features and mortality; the proportional hazards assumption was checked for all models. All vascular features were standardized or rescaled to a mean of 0 and an SD of 1 to allow comparison of risk prediction across continuous variables. In the absence of normative vascular data from healthy controls to use as a reference, the analysis of smokers with COPD was performed using standardization of vascular features to smokers without COPD to inform clinical insight.²¹ Owing to violation of proportional hazards in the models of aBV5/aTBV and PA/Ao in smokers with COPD and a significant interaction between age and pruning, risk for vascular features in smokers with COPD was modeled within the top 3 age quartiles. In addition, PA/Ao was modeled across clinically relevant tertiles of PA/Ao (<0.8, 0.8–1.1, and >1.1) to account for ongoing proportional hazards violation when modeled as a continuous variable.²² Vascular variable modeling was replicated in the parallel analysis of the smoking cohort without COPD to aid in comparative interpretation across groups; in this group, PA/Ao tertiles were determined based on distribution. Akaike information criterion was used to compare the Cox proportional hazard model fit of each vascular feature within the 2 cohorts of interest. Kaplan–Meier curves and log-rank tests were used to visualize quartiles of pruning risk on mortality.

Covariates for respiratory exacerbation modeling were selected based on their established relevance for COPD exacerbation risk and included gastrointestinal reflux disease, St Georges Respiratory Questionnaire score, and FEV₁ (% predicted).⁴ Exacerbation models were additionally adjusted for age, sex, race, FEV₁ (% predicted), history of prior respiratory exacerbations, and congestive heart failure (CHF), as well as smoking status and smoking pack-years. Secondary analyses investigated the isolated roles of the pulmonary artery and aorta diameters for predicting exacerbations as well as additional adjustment for small pulmonary venous volumes (vBV5/vTBV). Covariates for mortality modeling (all measured at baseline) were selected on the basis of their biological plausibility to confound the relationship between vascular features and mortality and included age, sex, race, body mass index, smoking status (current or former), pack-years of smoking, FEV₁ (% predicted), 6MWD, supplemental oxygen use, mMRC ≥2, %LAA-950, and CHF.

Statistical analysis was performed with SAS version 9.4 and Python version 3.11 software (https://www.python.org/down-loads/). A 2-sided P value <.05 was considered significant.

Results

Clinical and radiological findings

A total of 7585 subjects from the COPDGene phase 1 visit had available mortality and acute respiratory exacerbation data. Of these with complete vascular imaging data, 3169 had evidence of COPD based on an FEV₁/FVC <0.7 and 2530 had normal spirometry (Figure 1). Baseline characteristics of both groups are shown in Table 1. People with COPD were significantly older than those without COPD (median age, 64 [IQR, 57.5–70.2] years vs 57 [IQR, 50.9–64.1] years); the COPD group had a higher proportion of males (55.9% vs 51.3%). Approximately 50% of those with normal lung function and 40% of those with COPD were current smokers. Last, people with COPD had more quantitative emphysema on CT (%LAA-950) than smokers without COPD (median, 7.6 [IQR, 2.5–18.6] vs 1.1 [IQR, 0.4–3.0]).

Figure 1.

Cohort flow diagram. Abbreviations: COPD, chronic obstructive pulmonary disease; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; GOLD, Global Initiative for Chronic Obstructive Lung Disease.

Table 1.

Demographics and radiographic vascular features of smokers with and without COPD from the COPDGene Cohort.

Characteristic	COPD (n = 3169)	Non-COPD (n = 2530)	P value
Demographics
Age, y	64.0 [57.7–70.2]	57.0 [50.9–64.1]	<.001
Male sex, No. (%)	1771 (55.9)	1299 (51.3)	<.001
Race, No.			<.001
White	2605 (82.2)	1747 (69.1)
Black	564 (17.8)	783 (30.9)
BMI, kg/m2	27.4 [24.1–31.7]	28.7 [25.6–32.8]	<.001
Smoking status, No. (%)			<.001
Current	1239 (39.1)	1279 (50.6)
Former	955 (30.1)	1251 (49.4)
Smoking pack-years	46.5 [34.5–66.5]	34.4 [22.5–46.0]	<.001
On oxygen, No. (%)	762 (24.1)	35 (1.4)	<.001
FEV1 (% predicted)	56.7 [39.6–73.4]	95.7 [88.3–104.2]	<.001
Emphysema (%LAA-950)	7.6 [2.5–18.6]	1.1 [0.4–3.0]	<.001
6-minute walk distance (ft)	389.9 [306.3–465.1]	464.8 [398.9–536.5]	<.001
mMRC ≥2, No. (%)	1825 (57.6)	530 (20.9)	<.001
No. with exacerbations in the previous year, No. (%)	1084 (34.2)	237 (9.4)	<.001
Total score on SGRQ	35.3 [17.6–53.7]	9.6 [3.1–23.9]	<.001
History of CHF	135 (4.3)	35 (1.4)	<.001
Radiographic vascular features
aBV5/aTBV	0.56 [0.49–0.62]	0.61 [0.56–0.66]	<.001
Preacinar dilation	0.24 [0.21–0.28]	0.21 [0.19–0.24]	<.001
PA/Ao ratio	0.87 [0.78–0.97]	0.81 [0.74–0.87]	<.001
RV/LV ratio	0.52 [0.45–0.59]	0.50 [0.43–0.56]	<.001

Continuous variables are presented as median [IQR] if at least 1 of the variables is nonnormally distributed. Group comparisons were performed using Mann-Whitney U test. Categorical variables are presented as counts (percentages) and compared using the Fisher exact test.

Abbreviations: aBV5/aTBV, ratio of volume of distal pulmonary arterial vessels whose cross-sectional area was <5 mm² to the total pulmonary arterial blood vessel volume; BMI, body mass index; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; FEV₁, forced expiratory volume in 1 second; %LAA- 950, percentage of low-attenuation areas with <950 Hounsfield units; mMRC, modified Medical Research Council dyspnea scale; SGRQ, St George's Respiratory Questionnaire; PA/Ao, pulmonary artery to aorta diameter ratio; preacinar dilation, the pulmonary arterial vascular volume 5–20 mm² in cross-sectional area normalized to total arterial volume; RV/LV, right to left ventricle epicardial volume ratio.

The distributions of our radiologic features are provided in Figure 2. Compared to smokers without COPD, smokers with COPD had on average, more pruning of the distal arterial pulmonary vasculature (aBV5/aTBV, 0.55 vs 0.61; P <.0001), more dilation of the proximal preacinar arteries (0.24 vs 0.22, P <.0001), a greater PA/Ao ratio (0.88 vs 0.81, P <.0001), and a greater RV/LV ratio (0.52 vs 0.50, P <.0001).

Figure 2.

Vascular reconstructions and distribution of pulmonary and extrapulmonary vascular metrics in smokers with and without chronic obstructive pulmonary disease (COPD). (A) Representative 3-dimensional reconstructions of the pulmonary arterial tree in a subject without COPD (left) and a subject with COPD (right). Vessels are color-coded by cross-sectional area (CSA): red indicates distal vasculature with CSA <5 mm², yellow represents preacinar vasculature with CSA between 5 and 20 mm², and blue corresponds to proximal vasculature with CSA >20 mm². (B) CSA of small pulmonary arterial vessels (<5 mm², aBV5/aTBV). (C) CSA of preacinar pulmonary arteries (5–20 mm², aBV5–20/aTBV). Extrapulmonary vasculature features, including (D) the pulmonary artery to aorta diameter ratio (PA:A) and (E) right to left ventricular volume ratio (RV/LV) derived from epicardial ventricular volumes. (F–I) Distributions of these computed tomography–derived vascular metrics highlighting differences between smokers with and without COPD.

Acute respiratory exacerbations

The adjusted association of CT vascular features with ARE was assessed in 3168 people with COPD and 2530 people with normal lung function. For COPD, the PA/Ao ratio was the only vascular feature predictive of future ARE (IRR, 1.08 [95% CI, 1.04–1.12]; P = .02) (Table 2). The association of PA/Ao with acute exacerbation risk remained true when adjusting for pruning (IRR, 1.07 [95% CI, 1.04–1.11], P = .02) and when modeling PA/Ao as a categorical variable of PA/Ao >1 (IRR, 1.31 [95% CI, 1.17–1.47], P <.01). To further interrogate this association, we repeated the analysis with pulmonary artery (PA) and aorta (Ao) as unique predictors; PA, rather than Ao diameter, was the primary driver of this association with clinical risk (IRR_PA, 1.09 [95% CI, 1.06–1.13], P = .02; IRR_Ao, 1.04 [95% CI, 0.99–1.09], P = .11). Adjustment for small venous volume as a surrogate for left heart failure or postcapillary pulmonary hypertension did not change the PA/Ao signal (IRR, 1.08 [95% CI, 1.041.12], P = .02; venous BV5/TBV IRR, 1.03 [95% CI, 0.98–1.09], P = .29) (Table S1). For smokers without COPD, neither the intra- nor extraparenchymal vascular features were significantly associated with acute respiratory exacerbations (Table 2).

Table 2.

Associations of chest computed tomography-based pulmonary vascular tree features with acute respiratory exacerbations in smokers with and without COPD.

Feature	Non-COPD (n=2530)				COPD (n=3168)
	IRR	(95% CI)	P value	AIC	IRR	(95% CI)	P value	AIC
aBV5/aTBV	1.06	(0.97–1.17)	.21	9131.953	1.05	(0.99–1.10)	.07	20081.177
Preacinar dilation	1.10	(0.99–1.22)	.09	9130.582	1.01	(0.96–1.07)	.63	20084.149
PA/Ao ratio	1.03	(0.95–1.12)	.43	9132.878	1.08	(1.04–1.12)	.02	20068.069
RV/LV	1.01	(0.93–1.10)	.87	9133.479	1.04	(1.00–1.09)	.05	20080.474

All vascular variables are standardized to allow comparison of IRR (see Methods). The multivariable COPD models are adjusted for age, sex, race, gastrointestinal reflux disease, St Georges Respiratory Questionnaire score, forced expiratory volume in 1 second (% predicted), history of prior COPD exacerbations and history of congestive heart failure, smoking status, and smoking pack-years.

Abbreviations: aBV5/aTBV, ratio of volume of distal pulmonary arterial vessels whose cross-sectional area was <5 mm² to the total pulmonary arterial blood vessel volume; AIC, Akaike information criterion; CI, confidence interval; COPD, chronic obstructive pulmonary disease; IRR, incidence rate ratio; PA/Ao, pulmonary artery to aorta diameter ratio; preacinar dilation, the pulmonary arterial vascular volume 5–20 mm² in cross-sectional area normalized to total arterial volume; RV/LV, right to left ventricle epicardial volume ratio.

Mortality

All-cause mortality was higher in those with COPD as compared to those with normal lung function: 1337 smokers with COPD (42.2%) died over a median follow-up of 10.3 years while 334 (13.2%) ever-smokers with normal lung function died over a median follow-up of 12.4 years. At 15-year follow-up, overall survival in the quartile of smokers with COPD with the least pruning (Q4) was significantly higher (83.4%) than the quartile with the most pruning (Q1) (44.3%) (P <.0001) (Figure 3A). Survival in smokers without COPD was similarly higher among those with the least pruning (93.4% vs 86.4%; P <.0001) (Figure 3B).

Figure 3.

Mortality risk by quartiles of pruning, “defined as the ratio of volume of distal pulmonary arterial vessels with cross-sectional area, 5 mm2 to total pulmonary arterial blood vessel volume” in smokers with (A) and without (B) chronic obstructive pulmonary disease. For smokers with and without obstruction airway disease, loss of small vascular volume is significantly associated with worse all-cause mortality.

As compared to vascular volumes of smokers without COPD, for each standard deviation decrease in small vessel volume (more pruning), smokers with COPD experienced a 22% increased risk of death (HR, 1.22 [95% CI, 1.14–1.30]; P <.01). Smokers without COPD experienced a 26% increased risk of death for each SD increase in pruning (HR, 1.26 [95% CI, 1.11–1.42]; P = .04) (Table 3). In smokers without COPD, pruning risk was independent of presence of significant emphysema (Table S1). For each SD increase in preacinar arterial dilation there was a 61% increased risk of death in people with COPD (HR, 1.61 [95% CI, 1.09–1.24]; P <.01). Preacinar dilation was also predictive of death in people with normal lung function (HR, 1.14 [95% CI, 1.00–1.29]; P = .04). However, this preacinar dilation effect was attenuated by the extent of emphysema; smokers without COPD with <5% emphysema did not experience mortality risk whereas those with ≥5% emphysema experienced an even greater risk (HR, 1.09 [95% CI, 0.87–1.06], P = .19 and 1.51 [95% CI, 1.04–2.21], P = .03, respectively) (Table S2). As compared to a PA/Ao ratio <0.8, a ratio >1.1 was associated with a 43% increased risk of death in people with COPD (HR, 1.43 [95% CI, 1.12–1.83]). This association was not observed in people with normal lung function. The RV/LV ratio did not predict mortality in either cohort. In a model inclusive of arterial pruning (aBV5/aTBV) and PA/Ao ratio tertiles, only pruning remained significant for COPD (HR, 1.24 [95% CI, 1.14–1.35], P <.0001 and 1.07 [95% CI, 0.95–1.21], P = .24, respectively) (Table S1).

Table 3.

Associations of chest computed tomography-based pulmonary vascular tree features with mortality in smokers with and without COPD.

Feature	No COPD (n=2530)				COPD (n=3169)
	Hazard ratio	(95% CI)	P value	AIC	Hazard ratio	(95% CI)	P value	AIC
aBV5/aTBV	1.26	(1.11–1.42)	<.01	4849.196	1.22	(1.14–1.30)	<.0001	14162.618
Preacinar dilation	1.14	(1.00–1.29)	.04	4857.736	1.61	(1.09–1.24)	<.0001	14204.783
PA/Ao a	1.12	(0.27–4.59)	.88	4858.507	1.43	(1.12–1.83)	<.01	14222.680
RV/LV	1.01	(0.90–1.13)	.85	4861.648	0.99	(0.94–1.05)	.78	14222.920

All vascular variables are standardized to allow comparison of hazard ratio (HR) (see Methods). Multivariable models are adjusted for age, sex, race, body mass index, smoking status (current or former), pack-years of smoking, forced expiratory volume in 1 second (% predicted), 6-minute walk distance, supplemental oxygen use, modified Medical Research Council dyspnea scale ≥2, percentage of low-attenuation areas with <950 Hounsfield units, and congestive heart failure.

Abbreviations: aBV5/aTBV, ratio of volume of distal pulmonary arterial vessels whose cross-sectional area was <5 mm² to the total pulmonary arterial blood vessel volume; AIC, Akaike information criterion; CI, confidence interval; COPD, chronic obstructive pulmonary disease; PA/Ao, pulmonary artery to aorta diameter ratio; preacinar dilation, the pulmonary arterial vascular volume 5–20 mm² in cross-sectional area normalized to total arterial volume; RV/LV, right to left ventricle epicardial volume ratio.

^aRepresents tertiles of PA/Ao defined by clinically significant groups as <0.8, 0.8–1.1, and >1.1 where PA/Ao <0.8 is referent.

Discussion

In this investigation we explored a series of CT-based metrics of pulmonary vasculopathy in ever-smokers with and without COPD. Our goal was not only to describe differences in vascular changes among smokers with and without COPD, but also to determine the association of these imaging features for specific clinical outcomes. These features ranged from pruning of the distal intraparenchymal arterial vessels and dilation of the more proximal preacinar arteries to increases in the pulmonary artery to aorta and epicardial right to left ventricular volume ratios. The clinical outcomes of interest included both prospectively ascertained acute respiratory exacerbations and all-cause mortality. Our multivariable modeling demonstrated unique differences in the clinical significance of these vascular imaging variables based on the outcome of interest. In smokers with COPD, PA/Ao was the only metric significantly associated with future acute respiratory exacerbations. This effect could not be explained by pruning of the small arterial vessels. In both groups of people with and without COPD, arterial pruning was most strongly associated with all-cause mortality. The distribution of these measures and an interpretation of their performance in multivariable modeling merits further discussion.

We endeavored to examine these 4 features as they represent the most “upstream” (pruning) and “downstream” (RV/LV) processes associated with pulmonary vasculopathy evident on volumetric CT scan of the chest. Figure 2 provides the histogram distributions of the 4 imaging features ascertained from the COPDGene CT scans. Visual inspection of these plots suggests that people with COPD have on average more vascular pruning, more preacinar dilation, and a greater PA/Ao ratio. Differences in the distributions of the RV/LV volume between those with and without COPD is less evident subjectively; however, objective analyses demonstrate a small but statistically significant increase in the RV/LV ratio in those with COPD (Table 1). Based upon the data in this figure as well as a belief that vasculopathy is first detectable in the small vessels, the lack of associations between the RV/LV and clinical outcomes in our models was not surprising. We do not discount the clinical utility of such measures obtained from CT. Rather we merely note that in the specific modeling we did to facilitate a more direct comparison of features, the RV/LV was not prognostic and is likely a less sensitive radiologic feature of early or more subtle disease.

The first clinical outcome of interest in our investigation was acute respiratory exacerbations, also referred to as acute exacerbations of COPD in those people with expiratory airflow obstruction. Unlike the modeling results for mortality, the only vascular feature associated with these prospectively ascertained respiratory events was the PA/Ao ratio. While this recapitulates earlier work demonstrating the ability of this measure to predict such outcomes in this same cohort,^4,5 the lack of the capacity for arterial pruning to modulate this effect was surprising. For example, Wells and colleagues previously hypothesized that the risk of respiratory exacerbations associated with an increased PA/Ao ratio may reflect hemodynamically significant pulmonary vasculopathy.⁴ In the absence of invasive hemodynamics, the data to definitively determine this is not available in our study cohort but one conclusion of our results may be that distal pruning of the pulmonary arteries does not predict pulmonary vascular hemodynamics in people with COPD. In fact, studies to date correlating small vessel volume, expressed as a percentage of total cross-sectional area of small vessels <5 mm², have been done with imaging data that lack arterial-venous segmentation.²³ Instead, the association of PA/Ao, and more specifically PA dilation, with acute respiratory exacerbations may reflect an inflammatory process at the level of the great vessels.²⁴ Alternatively, this association may reflect an early sign of RV/PA decoupling that is only evident in periods of pulmonary vascular system stress, including during ARE. Future work in cohorts of people with COPD who have undergone CT imaging and right heart catheterization should interrogate to what extent hemodynamically confirmed pulmonary vasculopathy mediates acute respiratory exacerbation risk.

Pulmonary arterial vascular pruning is indicative of pulmonary hypertension in pulmonary arterial hypertension (PAH); however, the sensitivity of this metric for acute events in COPD may be outweighed by the effect of emphysema.²⁵ Emphysematous remodeling is characterized by the local destruction of the parenchyma in a process that likely mirrors what is believed to be occurring in the airway tree. As the distal small vessels are destroyed, the remaining small vessels evident on CT scan shrink and by our volumetric assessment, become the new distal vasculature. In this fashion, vessel destruction only partially diminishes the BV5 blood vessel volume and the sensitivity and specificity of the BV5/TBV ratio as an indicator of hemodynamically significant pulmonary vasculopathy may be less than the PA/Ao.

Arterial pruning was the only measure associated with mortality in both study groups that remained significant even after adjustment for the PA/Ao ratio. The association of pruning with mortality in smokers without COPD is not surprising in the setting of our prior work establishing the predictive capacity of pruning for spirometric decline and emphysema progression.⁷ This likely reflects the nature of small vessel injury that acts as a precursor of emphysema, possibly due to endothelial dysfunction with increased recruitment of inflammatory cells, slower transit, and therefore, persistent exposure to inflammation.^26,27 Our association of preacinar dilation with mortality in smokers with and without COPD was novel. Preacinar dilation is hypothesized to represent a later vascular adaptation to pruning with an intraparenchymal response to distal vasculopathy. Prior work associates this vascular feature with early interstitial changes on CT.² Similarly, our sensitivity analysis among smokers without COPD with and without clinically meaningful emphysema highlights the power of pruning as an early disease marker and preacinar dilation in smokers once significant emphysema is present. The interplay of these aggregate pathologies cannot be fully accounted for through simple adjustment of CT emphysema or spirometric impairment. In fact, these data emphasize a strong vascular-mediated mortality risk that precedes parenchymal injury and may potentiate airflow obstruction and impairment in gas exchange.^6,13,28 These findings are not surprising as unlike the PA or cardiac morphology, pruning of the distal vasculature and associated proximal dilation is the result of multiple disease processes including a primary vasculopathy due to noxious exposures, such as tobacco smoke.²⁹ Future work should investigate the drivers of this volumetric increase, including vessel wall thickness and the extent to which dilation occurs as a structural response to increased pulmonary vascular resistance.

There are limitations to this work that must be considered. First, the lack of normative vascular CT values from a healthy control population limit immediate clinical applicability of our results. However, our standardization of vascular features in the smokers with COPD to those without COPD provides insight into the relative increase in mortality risk associated with a 1-SD change compared to values identified in smokers without COPD. To enhance the clinical relevance of these findings and expand future work, studies with protocolized CT data that allow arterial/venous segmentation in normal controls are needed.^9,30 In addition, the lack of longitudinal vascular imaging data limit our ability to correlate endpoints with vascular feature trajectories; future work may also focus on the capacity for the most predictive vascular features identified in this manuscript to prospectively predict outcomes. In addition, invasive hemodynamic data to inform the presence or absence of pulmonary hypertension would be informative. It is also important to note that use of SSDI and LFU data to determine vital status may introduce misclassification due to underreporting in SSDI and subjects lost to follow-up in the LFU data. Finally, the precision of the effect estimates may be limited by lack of time-varying covariates and therefore inability to capture dynamic changes in risk factors over time. Therefore, our conclusions regarding the association of each radiographic feature with pulmonary vascular disease are limited.

Prior to this investigation, we believed that pruning would be the best predictor of clinical outcomes as it generally precedes dilation of the preacinar vessels and intraparenchymal processes generally precede dilation of the PA relative to the aorta or RV relative to the LV. This belief was only partially borne out as it was true for our mortality analyses but did not align with our findings for modeling acute respiratory exacerbations. While further work is needed, our findings suggest that the strongest vascular metric to predict mortality is pruning of the small intraparenchymal pulmonary arteries while the strongest predictor of acute respiratory exacerbations in ever-smokers with COPD is the PA/Ao.

Supplementary Material

Supplementary material

Supplementary material is available at Annals of the American Thoracic Society online.

Data availability

This article has a data supplement, which is accessible at the Supplements tab. The data underlying this article will be shared on reasonable request to the corresponding author pending IRB eligibility and approval of the COPDGene study.