Pulmonary Artery Stiffness by Cardiac Magnetic Resonance Imaging Predicts Major Adverse Cardiovascular Events in patients with Chronic Obstructive Pulmonary Disease

In this prospective pilot study, we aimed to evaluate the ability of cardiac magnetic resonance imaging (CMR) parameters of right ventricular function and pulmonary artery stiffness to identify pulmonary hypertension (PH), predict major adverse cardiovascular events (MACEs) in patients with secondary PH due to chronic obstructive pulmonary disease (COPD), and to estimate a prospective sample size necessary for a reliable power of the study. Thirty consecutive patients with COPD and suspected secondary PH were assessed by clinical examination, the six minute walk test, echocardiography, right heart catheterization and CMR, and followed–up for a mean period of 16 months to identify MACEs (cardiac death, ventricular tachyarrhythmia, and heart failure). Among CMR parameters of pulmonary artery stiffness, pulse wave velocity (PWV) yielded the best sensitivity (93.5%) and specificity (92.8%) for identifying PH, as diagnosed by cardiac catheterization. Moreover, PWV proved to be a valuable predictor of MACEs (HR = 4.75, 95% CI 1.00 to 22.59, p = 0.03). In conclusion, PWV by phase-contrast CMR can accurately identify PH in patients with COPD and may help stratify prognosis.

stiffness by right heart catheterization (RHC) is currently the standard for diagnosis, prognosis assessment and follow-up after treatment in PH patients. In several studies, invasively measured pulmonary artery pressure (PAP) was shown to associate with increased mortality 16 . Non-invasive methods, such as echocardiography 17 and, more recently, cardiac magnetic resonance imaging (CMR) 10,18,19 were used for PH assessment in combination with invasive modalities, leading to promising results. However, echocardiography can only be used for assessing the probability of PH, and studies that would endorse the use of CMR alone are still scarce.
On these considerations, we aimed to conduct a prospective study that allowed us to evaluate RV function and relationship between PA stiffness indices using CMR and adverse long-term outcome in patients with COPD and suspected secondary PH.

Methods
Patient selection and study design. We conducted a prospective pilot study in which we included 30 consecutive patients previously diagnosed with COPD and suspected of secondary PH (test group) who were hospitalized for acute respiratory failure in the 2 nd Department of Internal Medicine in Cluj-Napoca between October 2011 and October 2016.
Patients were included in the study provided they fulfilled the following criteria: (1) confirmed diagnosis of COPD, according to the criteria of the Global Initiative for Chronic Obstructive Lung Disease 20 ; (2) confirmed diagnosis of acute exacerbation defined as aggravated dyspnoea, cough, or sputum production warranting a change in therapy; 3) high probability of PH by echocardiographic criteria (tricuspid regurgitation velocities >3.4 m/s); 4) absence of other factors that might have induced PH (e.g. pulmonary embolism, significant valvular heart disease, known coronary heart disease and left cardiomyopathies). Exclusion criteria were: 1) life expectancy of less than a year due to PH or other diseases; 2) contraindications for CMR (incompatible implanted devices, severe renal impairment, severe obesity); 3) time interval between echocardiography, RHC and CMR > 10 days. Each enrolled patient was submitted to a standard investigation protocol which included history taking, physical examination, chest radiographs, pulmonary function tests and arterial blood gases, echocardiography, six minutes walk test, and Holter ECG monitoring for 24 hours. All tests were performed within 10 days after admission, provided that the patients had been stabilized by standard therapy with bronchodilators (β2 agonists, n = 22; phosphodiesterase inhibitors, n = 7; tiotropium, n = 15), short courses of oral or intravenous corticosteroids, n = 29; oxygen therapy, n = 8; antibiotics, n = 17; and diuretics, n = 11). The cardiovascular risk profile was established in each case, based on personal and family history, laboratory findings and known exposure to toxins (active and passive smoking, other professional or environmental toxins). The investigators analysing CMRs were blinded to RHC results.
The test group was compared against a control group of 30 healthy volunteers, matched for age and sex, who were submitted to a query excluding any history of chronic disease, any current symptoms or ongoing treatment for a chronic ailment, with normal clinical examination and normal electrocardiogram, chest radiographs, pulmonary function tests, echocardiography, Holter ECG monitoring for 24 hours and CMR. The healthy volunteers were enrolled within the same time span as test group patients.
The study was approved by the Ethics Committee of the Iuliu Hatieganu University of Medicine and Pharmacy, Cluj-Napoca and conducted in accordance with the Declaration of Helsinki. All patients and healthy volunteers signed an informed consent at enrolment.
All data have been collected by and are the intellectual property of the authors. Clinical follow-up was accomplished by clinical evaluation during hospital visits, telephone house-calls, or both at six month intervals until October 2016. In order to analyse event-free survival, we recorded the first occurring event from the following: death or aborted death from cardiac cause, sustained ventricular tachyarrhythmia, or hospitalization for heart failure. Hospitalizations due to non-cardiac causes were not considered significant events.

Pulmonary function measurements. Pulmonary function was assessed using a water-sealed Collin
Survey II volume displacement spirometer (Collins Medical, Inc.). Forced expiratory volume in 1 second (FEV1) and forced vital capacity (FVC) were measured with the purpose of diagnosing and grading COPD. Persistent airflow limitation and COPD were diagnosed in the presence of a FEV1/FVC ratio of less than 0.70. COPD was graded as mild (FEV1 >80% of the normal predicted value), moderate (50-79%), severe (30-49%), and very severe (<30%) 20 . Echocardiography. All images were acquired on an Aloka ProSound 5500 (Aloka Co. Ltd, Tokyo, Japan) echocardiograph with a wide-angle phased-array 2.5 MHz transducer. Measurements were performed according to previously published guidelines 21 . Images were recorded and analysed off-line to accurately quantify left ventricular (LV) and RV systolic and diastolic function. LV ejection fraction was calculated based on Simpson's modified single plane method from the apical 4-chamber view and considered normal if ≥55%. Systolic PAP was calculated as the sum between the RV-right atrium (RA) gradient, as measured on the tricuspid regurgitation flow, and the RA pressure; the latter was estimated by measuring the inferior vena cava diameter and collapse with a sniff. The probability of PH by echocardiography was considered high for tricuspid regurgitation velocities >3.4 m/sec 21 . Right heart catheterization. RHC  Cardiac magnetic resonance imaging. Image Acquisition. CMR ECG-gated images were acquired during apnoea with a 1.5-Tesla magnet (Magnetom Symphony, Siemens Medical Solutions, Germany) using a dedicated surface coil for radio frequency signal detection. Patients were required to hyperventilate briefly, then exhale and hold their breaths during image acquisition.
Firstly, standard localized views and contiguous short-axis and 4-chamber cine views covering both ventricles from base to apex were obtained by standard steady-state free precession (SSFP) cine CMR sequences in order to evaluate ventricular volumes and function. Pre-contrast imaging parameters were recorded by a standard protocol: repetition time/echo time 3.6/1.8 ms; flip angle 60°; slice thickness 6 mm; field of view 360 × 360 mm 2 ; matrix 192 × 192 pixels; voxel size 1.9 × 1.9 × 6 mm; 25-40 ms temporal resolution reconstructed to 25 cardiac phases. Standard late gadolinium enhancement (LGE) images were acquired 10 minutes after the intravenous injection of a contrast agent (Gadoterated imeglumine or Dotarem, Guerbet) -0.2 mmol/kg. Images were acquired in long and short axis views using a segmented inversion-recovery gradient-echo sequence after adjusting inversion time to optimize nulling of apparently normal myocardium. Brachial systolic and diastolic blood pressures were measured during the CMR SSFP acquisitions.
Phase-contrast sequence acquisition at the level of the main PA was made using a SSFP sequence of the RV outflow tract following an orthogonal plane to optimally visualise the main PA and pulmonary valve, provided that the imaging plane remained between the pulmonary valve and the PA bifurcation for the entire duration of the cardiac cycle 23 . To record PA stiffness parameters, the following specifications were applied to a free-breathing phase-contrast sequence: repetition time/echo time 7/5 ms; flip angle 15°; slice thickness 6 mm; field of view 360 × 360 mm 2 ; matrix 256 × 256 pixels; voxel size 2.2 × 1.8 × 6 mm; <35 ms temporal resolution; 20 reconstructed cardiac phases; velocity encoding 150 cm/sec 10 . For the free-breathing sequences, the average of 2 measurements was used Image Analysis. Cine and phase contrast images were evaluated by two experienced observers (one cardiologist and one radiologist) blinded to all clinical data, using specialized software (Argus, Siemens Healthineers Global).

Ventricular quantification.
Endocardial and epicardial borders were traced semi-automatically on short-axis cine images covering the entire LV and RV in order to measure end-diastolic and end-systolic LV and RV volumes, ejection fractions and end-diastolic LV mass. End-diastolic LV and RV mass to volume ratios were calculated and used to assess concentric remodelling. The maximum left atrium (LA) and RA areas were measured in all patients from the 4-chamber view. All volumes were indexed to body surface area. Tricuspid annular plane systolic excursion (TAPSE) was measured from the mid-four-chamber cardiac view to assess RV longitudinal motion. RV remodelling was defined as one or more of the following: the RV mass to RV end-diastolic volume ratio, the RV mass to RV end-systolic volume ratio and the RV/LV mass ratio. The presence/absence and location of LGE in both the LV and RV were also assessed and recorded.
Estimation of PA stiffness. PA stiffness assessment included the following parameters 10 : • PA elasticity (%) -the variation in luminal area during the cardiac cycle: (maximal area − minimal area/ minimal area). • PA distensibility (%/mmHg) -the variation in luminal area during the cardiac cycle for a given change in pressure: (PA elasticity/systolic PA pressure -diastolic PA pressure). • PA compliance (mm 2 /mmHg) -absolute changes in luminal area for a given pressure variation: (maximal area -minimal area)/(PA pulse pressure). • PA capacitance (mm 3 /mmHg) -the variation in volume (calculated by phase-contrast imaging) during the cardiac cycle for a given pressure variation: (RV stroke volume/(PA pulse pressure). • Elastic modulus, reflects the changes in luminal areas as a consequence of pressure variations: (PA pulse pressure) × minimal area/(maximal area − minimal area). • Stiffness index beta: the ln (systolic PAP/diastolic PAP)/PA elasticity.
• Pulse wave velocity (PWV) was calculated as described by Davies et al. 24 : PWV = √∆Q 2 /∆A 2 ; Q is the flow and A is the area through the PA. The contours for the main PA cross-section were manually traced on the magnitude images, with the purpose of determining the maximal and minimal PA areas, as well as peak and average velocities of the PA flow using specialized software (Syngo.Via VB20A). The program then automatically calculated flow and velocity data. Main PA and AA diameters were measured at the level of the PA bifurcation.
Statistical analysis. Data was tested for normality by the The Kolmogorov-Smirnov test. Data with normal distribution were expressed as mean ± standard deviation (SD), data with non-normal distribution -as median value ± interquartile range (IQR), and relative frequencies for categorical variables were expressed in percentages. Variables with normal distribution were processed using the Student's t test, while variables with non-parametric distribution were analysed by the ANOVA test. Baseline patient characteristics, echocardiography, CMR measures, and invasive hemodynamics were compared between patient groups using Fisher's exact test or a Kruskal-Wallis rank sum test. Spearman's coefficient was used to analyse correlations between RV function, PA stiffness and PAP. Pending the results of univariable analysis, multivariable analysis was performed to assess independent correlations between PAP, RV remodelling parameters and PA stiffness indices. The relationship between the six-minute walk test distance and PA stiffness was tested by linear regression analysis. Areas under the curves (AUC) were calculated by receiver-operator characteristic (ROC) curve analysis with the purpose of determining the sensitivity, specificity, positive predictive value and negative predictive value of PA stiffness parameters for During a mean follow-up period of 16 (6 to 38) months, 7 patients had majors cardiovascular events, as further described: cardiac death due to atrioventricular dissociation (n = 1); Holter ECG-documented ventricular tachyarrhythmia (n = 1), and hospitalization for heart failure (n = 5). Table 2. No significant differences in LV end-diastolic volumes have been recorded between the two groups. However, patients in the test group had lower LV ejection fraction. LV and RV wall mass indexes were significantly higher in the test group. Right chamber structural parameters (RA area, RV end-systolic volume, as well as the RV mass index) were higher in COPD patients, who also had poorer RV systolic function, assessed by the lower RV ejection fraction and TAPSE.

CMR characteristics. CMR characteristics are detailed in
LGE was evident in 7 test group patients (6 patients with PH and 1 patient without PH) and strictly distributed at mid-myocardial level in the basal segments of the RV septal insertion points. Phase-contrast examination revealed significantly increased PA maximal and minimal areas, PA elastic modulus, and stiffness indexes beta, while PA elasticity, distensibility, compliance and capacitance were diminished in test group patients (p < 0.001).

Correlation Coefficients between PA Stiffness Parameters and Pulmonary Pressures and
Resistance. Spearman's coefficient analysis revealed significant associations between stiffness indices assessed by CMR, and PAP and PVR determined by RHC. Results are detailed in Table 3. PA distensibility, compliance and capacitance were inversely and strongly correlated to mean and systolic PAP, while associations to PVR index were only low to moderate.
PA elastic modulus was strongly and directly correlated to mean PAP and systolic PAP, as was minimal PA area. The PA:AA diameter ratio was also strongly correlated to the PVR index, although associations to mean and systolic PAP were only moderate. The strongest associations emerged between PWV and mean and systolic PAP, with reasonably strong associations to the PVR index.

Relationship between Pulmonary Artery Stiffness Indices and the Six-Minute Walk Test
Distance. Positive correlations were obvious between the six-minute walk test distances and PA distensibility, compliance and capacitance. In addition to that, the six-minute walk test distance was negatively correlated to the PWV, elastic modulus and index beta after linear regression analysis (Fig. 1). Multivariable analysis was used to explore the association between RV and PA morphological and functional parameters, as assessed by CMR, and mean and systolic PAP, in the presence of some potentially confounding factors, such as age or gender. To that purpose, four different models were constructed: model A included age, gender and RV mass index; model B -age, gender and PA elastic modulus; model C -age, gender and PWV; model D -age, gender, RV mass index, PA elastic modulus and PWV. Results are detailed in Table 4. Among all studied parameters, PWV performed best, being correlated to both mean and systolic PAP independently of age, gender, RV mass index, and PA elastic modulus (R 2 = 0.89, p < 0.001; and R 2 = 0.89, p < 0.001, respectively). The predictive abilities of different PA stiffness indices for detecting PH as evaluated by ROC curve analysis have been compared (Fig. 2). For PWV, ROC curve analysis showed an AUC of 0.982 (95%CI: 0.948 to 0.999, p < 0.0001). The optimal cut-off value for PWV was established at 7.8 m/s, yielding 93.5% sensibility, 92.8% specificity, 93.7% positive predictive value and 92.9% negative predictive value for identifying PH. Associations between CMR-derived PWV as marker of PA stiffness and the overall occurrence of major cardiovascular events were assessed by Kaplan-Meier analysis. Pending this analysis, patients who exhibited higher PWV were more likely to experience MACEs by comparison with patients with low PWV [HR 4.75 (95% CI 1.00 to 22.59), p = 0.03] (Fig. 3). Sample size estimation. For assuring a 95% confidence, and 80% power of the study and based on the difference in MACE between groups, we have calculated a minimum required sample size of 40 patients in the test group for testing the ability of cardic MRI derived parameters to identify patients at risk for MACEs.

Discussion
Currently, data regarding PWV measurements by CMR in patients with COPD are scarce. In our study, significant inverse correlations occurred between CMR-measured PWV and the indices for compliance and distensibility of the pulmonary vasculature, thus indicating that CMR could be used for early non-invasive identification of PH. As shown in previous studies, the assessment of PWV in major arteries provides crucial information regarding PA stiffness 23 . Interestingly, in our study, PWV performed best among all PA stiffness parameters in identifying PH, as ROC curve analysis yielded reasonable performance indices.
In addition to that, the results of our research highlighted the presence of right chamber remodelling assessed by echocardiography and CMR in patients with secondary PH due to COPD. Right chamber dimensions and RV mass were considerably elevated in test group patients by comparison with healthy volunteers, which was expected, considering the fact that the RV responds to elevated pulmonary pressures by progressive hypertrophy, followed by dilatation 25 . A previous CMR-based research on patients with COPD and no or mild hypoxemia revealed that RV mass was increased in test patients by comparison with healthy volunteers despite normal RV systolic function, suggesting that RV hypertrophy is the earliest sign of RV pressure overload, occurring as an adaptation process 26 . Another study comparing healthy volunteers to COPD patients with or without PH supported the same hypothesis; the study showed that changes suggesting RV hypertrophy were more obvious in COPD patients without PH vs. healthy controls, with little increase in patients with COPD and PH 27 . RV dilatation is also an essential component of RV remodelling. In our study, CMR measurements of RV dimensions, as well as RV end-diastolic volumes were significantly more elevated in COPD patients. Previous research has shown that larger RV volumes measured by CMR were an independent predictor of higher overall and cardiovascular mortality 28 .
Patients in our study were only admitted to the test group provided they had evidence of PH. In our research, CMR examination revealed impaired RV systolic function, as assessed by TAPSE, the RV ejection fraction and RV stroke volume in COPD patients by comparison with healthy volunteers. RV stroke volume reduction was also obvious in a previous study by Roeleveld et al. who also showed improvement of this parameter after specific vasodilator treatment 29 , while van Wolferen et al. demonstrated that reduced RV stroke volume was an independent predictor of increased mortality and treatment failure 30 .
In a study by Kjaergaard et al., diminished TAPSE was shown to be highly predictive of increased mortality in patients with COPD and PH 31 ; similar results were obtained by Forfia et al. who demonstrated significant correlations between TAPSE and right chamber remodelling, as well as other indices of impaired RV systolic function, and showed that patients with PH and decreased TAPSE had lower survival 32 .
Our CMR-based examinations also showed significant differences in PA stiffness indices between test group patients and healthy volunteers. In addition to that, statistical analysis revealed important correlations between PA stiffness indices, such as PA elasticity, distensibility, capacitance, compliance, elastic modulus, stiffness indices alpha and beta, and RV remodelling and function parameters, such as RV ejection fraction, RV mass indexed, RV stroke volume indexed, TAPSE, and RV end-systolic volume; these findings are consistent with results from other studies 33 . Interestingly, there was barely any correlation to RV end-diastolic volumes, an observation which also emerged from the study of Stevens et al. 19 . These results may suggest that RV dilatation with no impairment in RV systolic function is less likely to be accompanied by significantly increased PA stiffness. In fact, RV systolic impairment may be a consequence of pressure overload triggered by persistent severely increased PA stiffness, since a previous study showed that PA stiffness indices became impaired before significant increases in resting systolic PAP 10 , and may, therefore, have a significant contribution to disease progression. Accordingly, we intended to establish correlations between systolic PAP and PA stiffness indices; the strongest linear correlations were obvious between systolic PAP and PA distensibility, compliance, capacitance and elastic modulus, which is consistent with the results of previous studies in which either CMR 10 or intravascular ultrasound were used for determining pulmonary stiffness 34,35 . Interestingly, low PWV values at CMR were shown to predict survival, a novel discovery, undocumented, to our knowledge in previous research. However, results should be interpreted with caution, considering the low number of patients, and implicitly the low number of individuals who experienced MACEs.
Indeed, the main limitations of the current study are related to the rather small number of participants. We aimed to diminish the impact of this limitation on our results by carefully selecting the study population using well established criteria and following a standard detailed examination protocol. Also, the study needs to be continued until the minimal estimated sample size is reached.
CMR examinations were carefully performed, yielding good quality images. 2D-velocity-encoded CMR allowed accurate non-invasive measurements of aortic blood flow parameters with good temporal and spatial resolutions. Although 4D phase-contrast CMR was not available for the assessment of the patients in this study, the limited temporal and spatial resolutions that have been reported for such techniques 36 rendered this examination superfluous. Further research on a larger number of patients may help consolidate the results of the current study.

Conclusions
CMR phase-contrast imaging may provide an extensive and accurate assessment of right chambers remodelling and PA stiffness in patients with COPD. In the future, it may become a valuable non-invasive tool for quantifying PH, provided that its value is further confirmed by more extensive studies. PA stiffness assessment by CMR seems to be promising in terms of early identification of pulmonary vascular disease, which may result in earlier diagnosis and treatment onset, thus diminishing arterial remodeling progression. Moreover, CMR-derived PA stiffness indices, particularly PWV, may be useful for predicting MACEs. However, our research is only hypothesis generating and larger studies are necessary to confirm these findings.