Association between CMR-Derived Pulmonary Artery Pulse Wave Velocity and Pulmonary Risk Factors: the Multi-Ethnic Study of Atherosclerosis COPD Study

Abstract

Purpose

Pulse wave velocity (PWV) is a quantitative marker of arterial stiffness which is widely validated in the systemic circulation. Pulmonary artery (PA) stiffness, as measured by PWV, and its associations with disease, risk factors, and outcomes are largely unestablished. We aimed to assess the association of PA PWV with smoking history.

Methods

This prospective cohort study was conducted among participants enrolled in the Multi-Ethnic Study of Atherosclerosis (MESA) COPD who received cardiac magnetic resonance (CMR). 2-D phase contrast images were acquired to assess flow. PA PWV was assessed using validated semi-automated software for the main, right, and left PA by a single reader. Non-parametric, skewed data were log-transformed as appropriate and univariable, and multivariable linear regression models were used to analyze association of PA PWV and to demographic factors and cardiopulmonary risk.

Results

A total of 166 participants were included in this study after excluding for artifacts or incomplete scans. Participants were 65% male, average age 68±7 years, smoked an average of 34±20 pack-years, and 36% had COPD. A unit increase in pack-year was associated with a higher log average PA PWV of 0.007 (p<0.01). This association persisted after adjustment for demographics, vascular risk factors, COPD stage and ABI. PA PWV was not associated with COPD status.

Conclusions

We demonstrated that a severe smoking history was independently associated with a greater PA PWV. Our study advances the understanding of the mechanisms of cardiopulmonary risk factors and underscores the potential clinical applicability of PA PWV measurement for early disease detection.

Introduction

Age is a potent risk factor for cardiovascular disease (CVD) and among its effects are structural and functional changes in the vasculature, moderated in part through central arterial stiffening [1, 2]. It has profound effects on peripheral vasculature, and in turn various mechanisms have been hypothesized by which it may have deleterious effects on end-organs [3]. A key marker of arterial stiffness is pulse wave velocity (PWV), the rate with which pressure waves move along the vessel walls after ventricular contraction [4]. As measured in the systemic circulation, PWV is well-established as an independent predictor of all-cause mortality, cardiovascular events, hypertension, end-stage renal disease and diabetes.[5–8] It is highly reproducible and has found a role in clinical practice as a marker of end-organ damage [9, 10].

In contrast to arterial stiffness in the systemic circulation, the biological importance of stiffness in the pulmonary circulation remains unestablished. Like the aorta, the pulmonary artery (PA) has a cushioning role in transforming pulsatile flow to a steady flow [11, 12]. Impairment can have several consequences, including an increase in right ventricular afterload with subsequent right-sided heart failure [13, 14]. Anatomical challenges in the non-invasive assessment of PA stiffness, relative to its systemic counterpart, complicate the investigation of the pulmonary circulation. Utilization of transthoracic echocardiography for PA PWV is difficult because of the lack of an acoustic window at the level of the pulmonary arteries and a short penetrating depth [15]. In comparison, cardiac magnetic resonance imaging (CMR) offers superior spatial resolution, excellent reproducibility, and has been widely validated for aortic stiffness [16–19]. The feasibility of CMR for PA PWV measurement has been demonstrated in numerous studies and its performance characteristics are favorable when compared with right heart catheterization [20–25].

Patients with chronic obstructive pulmonary disease (COPD) may benefit from measurement of PA PWV due to the link between heart and lung disease[26, 27]. One of the proposed mechanisms by which COPD is associated with CVD is accelerated arterial aging leading to increased vascular stiffness [28, 29]. It is hypothesized that less compliant arteries cause increased aortic systolic pressure, resulting in increased cardiac afterload and a reduction in diastolic coronary artery filling, ultimately causing left ventricular dysfunction. Quantification of PA stiffness using PWV could in effect measure an intermediary of heart disease and COPD, potentially providing clinicians with a tool for earlier disease detection and tracking of disease progression. This would allow for targeted interventions grounded in knowledge of vascular mechanics and pathophysiology.

In this study, we sought to characterize PA PWV using CMR in individuals with COPD and non-COPD subjects. To further elucidate mechanisms of PA stiffness, we assessed these parameters against established markers of vessel stiffness and traditional cardiovascular risk factors. We hypothesized that cumulative smoking and COPD status are associated with higher PA PWV.

Methods

Study population

Our study population comprised of all participants of the Multi-Ethnic Study of Atherosclerosis (MESA) Chronic Obstructive Pulmonary Disease (COPD) Study, which included 200 participants from MESA follow-up exam (2010–2012) and 125 participants from the Emphysema Cancer Action Project (EMCAP)[30, 31]. All subjects provided written informed consent for study participation, which was reviewed and approved by the study sites’ Institutional Review Board. All participants were 50 to 79 years of age at inclusion, with ≥10 pack years of smoking and without clinically known cardiovascular disease at time of enrollment. Exclusion criteria were asthma before 45 years of age, pulmonary disease besides COPD, cancer, previous lung resection, stages IIIb to V kidney disease, allergy to gadolinium, claustrophobia, metal in the body, pregnancy, or weight >300 lb. Recent COPD exacerbation was a temporary exclusion criterion. Of the 325 subjects, 270 participants underwent MRI. On review of MRI, participants were excluded for aliasing artifacts (n=30), breathing artifacts (n=36) or incomplete data (n=49). One hundred sixty-six subjects were ultimately included in our analysis.

Covariates

Age, sex, race/ethnicity, smoking variables and medical history were self-reported through questionnaires. Medication use was assessed by medication inventory. Height, weight, and blood pressure were measured following standardized MESA protocols [32]. Hypertension was defined as SBP ≥140 mmHg, DBP ≥90 mmHg, or current use of antihypertensive medications. Glucose and lipids were measured after a 12-hour fast. Diabetes mellitus was defined as fasting glucose ≥126 mg/dl or use of insulin or oral hypoglycemic medications. Ankle-brachial index was included as a marker for systemic arterial stiffness. COPD severity was categorized according to Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria [33].

Image acquisition

The method of image acquisition is previously described in depth [34]. In brief, images were acquired with a 1.5T whole-body MRI system (Signa LX, GE Healthcare; Avanto or Espree, Siemens Medical Systems). Breath-holding phase contrast (PC) MRI was obtained perpendicular to the direction of flow in the main pulmonary artery (MPA), left pulmonary artery (LPA) and right pulmonary artery (RPA), with minimized echo and repetition time (approximately 3.1 ms and 6.2 ms respectively), slice thickness 8 mm, and through-plane velocity encoding at 100 cm/s maximum (Figure 1). The RPA was evaluated at a point 2 cm distal to the bifurcation with the plane between the aorta and trachea on axial images. The LPA was evaluated approximately 3 cm distal to the bifurcation, before branching into second-order vessels.

Figure 1. Example of acquisition of right pulmonary artery.

Axial scout image used with slice between aorta and trachea (top left). Slice through LPA, perpendicular to direction of blood flow (top right). Phase contrast cine image displaying LPA (bottom left) with corresponding velocity-encoded image (bottom right).

Anatomical CMR images displaying the pulmonary trunk and arteries were captured using a bright-blood, SSFP, single-shot sequence and were used to calculate path length between the MPA and either branch artery.

Pulmonary artery pulse wave velocity analysis

PA PWV was obtained from the PC images using validated software (ARTFUN; Inserm, Paris, France) [16]. Regions of interest were drawn in the PC images by manually identifying the center of the MPA, LPA, and RPA in each participant, after which software automated tracking of the vessel wall through the cardiac cycle. If necessary for accuracy, the reader manually redrew the vessel contour for each image. The resulting contours were then superimposed on the velocity-encoded images, enabling us to obtain a flow curve for the cross-section (Figure 2).

Figure 2. Tracking region of interest (top) and flow curve analysis (bottom).

Top: manual selection in left panel. Automatic full-cycle tracking of vessel wall (mid panel) with subsequent superimposing on velocity-encoded images (right panel). Bottom: raw data from two flow curves (left panel), which are then normalized (mid panel).

After normalization, the upslopes were estimated from all the datapoints using a least-squares method[16]. Transit time was calculated as the average time difference between the systolic upslope of the MPA flow curve and the flow curve of the LPA or RPA (Figure 2). Path length was defined by creating a three-dimensional curve on anatomical images displaying the pulmonary vasculature, intersecting the aforementioned PC images (Figure 3). PA PWV was then calculated by dividing distance by transit time between the two flow waveforms

Figure 3. Path length determination.

Selection of LPA and MPA cross-section (top left and top right, respectively). Manual tracking of vessel trajectory (bottom left) with resulting curve (bottom right).

Interobserver variability and agreement was assessed in a random subset of analyses (n=38) by two different independent readers of the MRI core lab (R.H. and C.N.).

Statistical analysis

Data were analyzed using STATA software, version 13.0 (StataCorp, College Station, TX, USA). As PA PWV data were skewed, they were logarithmically transformed before analysis. Average PA PWV was calculated as an average of left and right PA PWV data. Continuous data are presented as mean ± SD or median (interquartile range [IQR]) as appropriate. Categorical variables were presented with frequencies and percentages. Linear regression models and Spearman correlation coefficients were used to assess associations. Multivariable linear regression models were used to assess independent relationships of PA PWV and cardiovascular risk factors after adjustment for age, sex, race/ethnicity, height and weight. Subsequent models also adjusted for vascular risk factors and for COPD status and ankle-brachial index (ABI). Interobserver agreement was assessed with concordance correlation coefficients. Bland-Altman plots were used to visualize agreement. A significant difference was defined as p < 0.05.

Results

Participants were 65% male and on average 68±7 years old. They were 39% Caucasian, 8% African-American, 30% Chinese-American, and 22% Hispanic. Participants averaged 34±20 pack-years and 17% were active smokers at time of enrollment. 36% of subjects had prior diagnoses of COPD and 13%, 20% and 2% were GOLD class I, II and III, respectively. Detailed baseline characteristics are shown in table 1.

Table 1.

Baseline characteristics of the study participants.

Variables	No COPD (n = 106)	COPD (n = 60)	P value
Age, y	67 ± 7	70 ± 7	0.06
Male gender	63 (59%)	45 (75%)	0.04
Race
White	35 (33%)	30 (50%)	0.03
African-American	29 (27%)	21 (35%)	0.30
Chinese-American	10 (9%)	4 (7%)	0.54
Hispanic	32 (30%)	5 (8%)	<0.01
BMI, kg·m2	28.5 ± 4.7	27.7 ± 4.9	0.32
Diabetes	17 (16%)	9 (15%)	0.86
Insulin or hypoglycemic medication	14 (13%)	7 (12%)	0.80
Systolic blood pressure, mmHg	119 ± 18	123 ± 17	0.12
Hypertension	51 (48%)	39 (65%)	0.04
Antihypertensive medication	50 (47%)	35 (58%)	0.17
Pack-years	30.0 ± 19.6	40.7 ± 29.9	0.01
Cigarettes per day	11.6 ± 7.3	14.3 ± 8.1	0.34
First cigarette after waking up, mins	70.4 ± 105.1	38.6 ± 52.5	0.01
Current smoker	19 (18%)	12 (20%)	0.74
FEV%/predicted	101.3 ± 15.6	75.6 ± 16.5	<0.001
Cotinine level, ng/ml	1129 ± 2666	2931 ± 5640	0.03
Emphysema, %	2.14 ± 2.1	6.9 ± 6.4	<0.001
CAC score	171.4 ± 345.5	304.2 ± 535.0	0.11
Cholesterol, mg/dl	180.8 ± 36.7	176.7 ± 31.1	0.47
Lipid-lowering medication	35 (33%)	33 (55%)	<0.01
Ankle-brachial index	1.2 ± 0.1	1.1 ± 0.2	0.16
Pulse oxymetry, %	97.1 ± 1.6	96.6 ± 1.9	<0.05
Presence of SOB when walking	5 (5%)	8 (13%)	0.13
Average pulse wave velocity, m/s	4.8 ± 3.4	4.8 ± 3.8	1.0

BMI = body mass index, CAC = coronary artery calcium, FEV = forced expiratory volume, SOB = shortness of breath.

Pulmonary artery pulse wave velocity

Median LPA and RPA PWV were 3.5 (2.1–5.3) and 3.1 (2.1–5.2) m/s, respectively. Data were generally right-skewed: mean PWV for LPA and RPA were 5.2±4.7 and 4.6±3.9 m/s, respectively. Mean global PWV was 5.0±3.8 m/s. Log transforms of global and right, but not left PA PWV were associated with smoking pack-years (r=0.24, p<0.01; r=0.21, p<0.05; r=0.14, p=0.10, respectively, Figure 4). Log average PA PWV was further associated with diabetes and being Hispanic (p<0.05, table 2), and was not significantly different by COPD status or GOLD stage (Figure Supplementary 1). In univariable analysis, a unit increase in pack-year was associated with an increase in log average PA PWV of 0.007 (p=0.002, table 3). This association persisted after adjustment for demographics, vascular risk factors, COPD stage and ABI (0.007, p=0.002).

Figure 4. Scatterplot illustrating relationship between pack-years of smoking versus log mean pulmonary artery pulse wave velocity.

PWV = pulse wave velocity.

Table 2.

Univariable linear regression of demographics, cardiovascular risk factors and smoking parameters versus log PA PWV.

Variable	Coefficients	P value
Age	0.01	0.28
Male gender	0.10	0.34
Race (ref: white)
African-American	0.00	0.97
Chinese-American	0.07	0.72
Hispanic	0.31	<0.05
BMI	0.01	0.47
Diabetes	0.34	<0.05
Insulin or hypoglycemics	0.32	<0.05
Systolic blood pressure	0.005	0.11
Hypertension	0.01	0.94
Antihypertensive medication	0.04	0.67
Pack years	0.01	<0.01
Cigarettes per day	0.01	<0.01
First cigarette after waking up	−0.002	<0.05
COPD	−0.03	0.78
COPD stage	−0.01	0.93
FEV1%/predicted	−0.00	0.29
Emphysema	0.002	0.90
Cotinine level	0.00	0.53
CAC score	0.0003	<0.05
Log(CAC+1)	0.03	0.27
Cholesterol	0.00	0.41
Lipid-lowering medication	0.05	0.67
Pulse oxymetry	0.006	0.84
SOB walking	0.01	0.83

BMI = body mass index, CAC = coronary artery calcium, FEV = forced expiratory volume, PA = pulmonary artery, PWV = pulse wave velocity, SOB = shortness of breath.

Table 3.

Multivariable regression analyses for mean log pulmonary artery pulse wave velocity and demographics, COPD stage and ankle-brachial index.

Variable	Model 1	P value	Model 2	P value	Model 3	P value
Pack years	0.007	<0.01	0.008	<0.01	0.007	<0.01
Age, y			0.007	0.72	−0.003	0.71
Male gender			0.040	0.97	0.042	0.80
Race (ref: white)
Chinese-American			0.23	0.28	0.184	0.23
African-American			0.17	0.21	0.07	0.13
Hispanic			0.63	<0.01	0.60	<0.01
Height, 10 cm			−0.001	0.95	0.01	0.88
Weight, 10 lbs			0.005	0.79	0.001	0.89
COPD Stage (ref: control)
Mild					−0.018	0.18
Moderate					−0.167	0.23
Severe					0.62	0.17
Ankle-brachial index					−0.496	0.45

Model 1: Log pulmonary artery pulse wave velocity versus smoking pack-years. Model 2: Model 1 + demographic factors. Model 3: Model 2 + COPD stage and ankle-brachial index.

Reproducibility

Inter-class correlation coefficient (ICC) for left PWV, right PWV, and global PWV were 0.88, 0.84, and 0.94 respectively. Bland-Altman plots for left, right and mean PWV are shown in Figure 5.

Figure 5. Bland-Altman plots for left, right and mean PWV.

Red middle lines represent the mean difference, upper and lower boundaries represent limits of agreement (±2SD).

Discussion

In a large, multi-ethnic population of elderly subjects with significant smoking history free of clinical CVD at baseline, pulmonary artery pulse wave velocity was significantly associated with pack-years smoked after adjustment for demographics, cardiovascular risk factors, COPD stage and ABI. The method described here for measurement of PA PWV using a CMR-based transit-time method was highly reproducible with excellent inter-observer agreement. The validity of our measurement technique is further supported by values previously described in the literature for PA PWV derived using CMR. Ibrahim et al. reported a mean PA PWV of 2.8±0.85 m/s in subjects without pulmonary arterial hypertension (PAH).[23] In ten healthy subjects averaging 35 years old, Bradlow et al. reported mean LPA and RPA PWV values of 2.1 and 2.3 m/s respectively[23]. In a larger study of 156 participants averaging 37 years old, a mean global PA PWV of 2.2 m/s was obtained [24].

Our study is among the first to demonstrate the association between PA PWV and common risk factors for cardiac and pulmonary disease. It is well-established that aortic PWV is elevated in patients with chronic risk factors including smoking, though the degree of effect modification attributable to smoking may be contended [35, 36]. Likewise, aortic PWV is known to be elevated acutely after smoking, which is believed to be mediated by transient increases in catecholamines and sympathetic tone [37, 38]. In contrast, the effect of such risk factors on PA PWV are limited and to our knowledge, no previous study of PA PWV has assessed risk factors with adjustment as done here. A smaller study of 79 patients did not identify an association of PWV with smoking, though a study of PA strain in MESA COPD identified a significant association with smoking status in univariate analysis[39, 40]. Our finding of a direct, independent relationship of PA PWV with smoking pack-years reinforces that the vascular mechanism of smoking is systemic while implicating the pulmonary vasculature as a parallel driver of disease progression.

It is noteworthy that our population as a whole is significantly older with a much higher rate of co-morbidities than much of the literature, which may explain some of the negative findings in our study as we lacked the participant heterogeneity to demonstrate these features.

For example, PA PWV has been found to be higher in patients with diagnosed PAH, ranging from 5.2 m/s to as high as 10 m/s in one study measuring the velocity intravascularly [23, 41]. While we did not assess PAH status in our participants, this may more broadly suggest an elevation of PA PWV with pulmonary disease. Given the high average pack-years smoked, our participants may have had some degree of preclinical pulmonary disease that also skewed our mean PA PWV to be slightly higher than found in other studies.

We did not identify an association between age and PWV. Similar to our study, Bradlow et al. did not find an association with age and PA PWV [23]. On the other hand, Dawes et al. found an association in a study population averaging 37±12 years, notably much younger than our population[24]. A relationship between age and PA PWV was also described by Laffon et al. [42]. Individual studies of younger populations lend credence to the notion that PA PVW is influenced by age, with a mean PA PWV of 1.31 m/s in healthy children ages 9–12[20] and 1.84 m/s in healthy volunteers age 23.8±1.2 years[21]. We may not have had a sufficiently large population or age range to adequately address this question, particularly if age-related changes in PA PWV are small. It may bear investigating the PA PWV in a longitudinal manner or a more age-diverse population.

Our major goal for future studies is to examine ventricular-arterial coupling—how PA PWV values relate to established markers of both left and right ventricular function. This is largely unexplored and we found only one study, which suggested an association between PA PWV and right ventricular function [24]. Regarding the left ventricle, we postulate that rises in PA pressure due to retrograde heart failure might be similarly reflected in changes in PA PWV. As with measurement of aortic PWV, measurement of PA PWV may potentially play a clinical role in risk stratification in the context of heart failure, and merits further study with consideration of cardiac chamber parameters.

Limitations

We note that our study has several limitations. The COPD subgroup in our study was insufficiently large for us to answer whether PA PWV was associated with COPD stage. More severe COPD was minimally represented, with only four subjects in GOLD 3 and none in GOLD 4. This may be a consequence of the imaging methods, as subjects with severe COPD may be excluded because of difficulty performing breath-hold during acquisition. Additionally, COPD in our study may have been confounded with PAH, a common and interrelated complication associated with an increase in PA PWV [43, 44]. We did not have PAH status in our participants and so were unable to adjust for its effects.

From a technical perspective, a crucial consideration for the transit-time method described in our study is the temporal resolution. The small and densely-branching pulmonary vessels can give rise to early wave reflectance and pseudo pulse waves, which could complicate PA PWV measurement. Given that the flow waves obtained in our study seemed linear in the systolic phase, we are confident that the appropriate pulse waves are indeed represented by our data.

Conclusion

Our study adds to a growing body of evidence regarding the application of CMR in the pulmonary vasculature. We found that smoking pack-years is independently associated with an increase in PA PWV. These findings are congruent with studies identifying an elevation in PA PWV associated with pulmonary disease as well as the established literature demonstrating increases in aortic PWV associated with smoking. In our population, we implicate smoking in the pathophysiological process of PA stiffness thereby elucidating the role of a classical cardio-pulmonary risk factor. Furthermore, we validate that PA PWV measurement using CMR is a feasible and reproducible method which may be used to further investigate the pathophysiological processes of the pulmonary vasculature and has a potential role in the diagnosis and follow-up of pulmonary disease.

Data Statement:

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Supplementary Material

1

Figure Supplemental 1. Boxplots illustrating mean PWV stratified by COPD stage.

Highlights.

• Pulmonary artery pulse wave velocity is independently associated with smoking pack-years even after full adjustment for confounders

• MRI-based measurement of pulmonary artery pulse wave velocity showed excellent reproducibility (ICC>0.84)

• Semi-automated software enabled feasible pulmonary artery pulse wave velocity measurement using MRI

Abbreviations

CVD

cardiovascular disease

COPD

chronic obstructive pulmonary disease

MPA

main pulmonary artery

LPA

left pulmonary artery

RPA

right pulmonary artery

PA

pulmonary artery

PWV

pulse wave velocity

CMR

cardiac magnetic resonance

SD

standard deviation

IQR

inter-quartile range

GOLD

Global Initiative for Chronic Obstructive Lung Disease