Semi-quantitative Myocardial Perfusion MRI in Heart Transplant Recipients at Rest: Repeatability in Healthy Controls and Assessment of Cardiac Allograft Vasculopathy

Abstract

Background:

Cardiac Allograft Vasculopathy (CAV) is a major cause of chronic cardiac allograft failure. Invasive coronary angiography (ICA) and intravascular ultrasound (IVUS) are the current diagnostic methods. Myocardial perfusion MRI has become a promising non-invasive method to evaluate myocardial ischemia, but has not been thoroughly validated in CAV. Our objective was to assess the repeatability of myocardial rest-perfusion MRI in healthy volunteers and its feasibility in detecting CAV in transplant patients (Tx).

Methods:

Twelve healthy volunteers and twenty transplant patients beyond the first year post-transplant underwent cardiac MRI at 1.5T at rest including first-pass perfusion imaging in short axis (base, mid, apex) after injection of gadolinium. Volunteers underwent repeated cardiac MRI on different days (interval=15.6 ± 2.4 days) to assess repeatability. Data analysis included semi-automatic contouring of endocardial and epicardial borders of the left ventricle (LV) and quantification of peak perfusion, time-to-peak (TTP) perfusion, and upslope of the perfusion curve.

Results:

Between scans and re-scans in healthy volunteers, peak signal intensity, slope, and TTP demonstrated moderate agreement (ICC=0.53, 0.48, and 0.59, respectively; all, p < 0.001). Peak signal intensity, slope, and TTP were moderately variable with COV values of 23%, 42%, and 35%, respectively. Peak perfusion was significantly reduced in CAV positive (n=9 Tx patients) compared to CAV negative (n=11 Tx patients) groups (90.7 ± 27.0 vs 139.5 ± 30.2, p < 0.001).

Conclusion:

Cardiac MRI is a moderately repeatable method for the semi-quantitative assessment of first-pass myocardial perfusion at rest. Semi-quantitative surrogate markers of LV perfusion could play a role in CAV detection.

1. Introduction:

Cardiac allograft vasculopathy (CAV) is a primary cause of allograft failure leading to significant morbidity and mortality beyond the first year post-heart transplantation (Tx). The 2017 International Society of Heart and Lung Transplantation (ISHLT) registry reported CAV prevalence in transplant patients at 1, 5, and 10 years post-Tx of 8%, 29%, and 47% respectively.[1] Among 85,000 orthotopic heart transplants tracked in one study, CAV accounted for approximately 32% of deaths at 5 year follow up.[2]

CAV manifests at an advanced stage with heart failure secondary to allograft dysfunction, silent myocardial infarctions, and ventricular arrhythmias which can lead to sudden cardiac death.[3] Late stage CAV is often irreversible and portends a poor prognosis, while early detection and immunosuppressive treatment regimens lead to improved outcomes.[4,5] Therefore, regular screening for CAV is important, typically based on invasive coronary angiography (ICA) and intravascular ultrasound (IVUS) as primary and secondary diagnostic tests, respectively.[6]

CAV causes diffuse concentric medial thickening of the vessel wall, leading to minimal and late focal obstruction of the lumen.[7–9] ICA enables detection of obstruction, but not wall thickening. IVUS allows for assessment of intimal medial thickening (IMT) rendering it more sensitive at detecting early CAV[9–11]. In addition, IVUS provides prognostic information as IMT > 0.5 mm measured in the LAD has been associated with a higher incidence of graft loss[11–14]. However, given a lack of widespread availability and experience in the technique, IVUS was designated as a secondary method of screening. Furthermore, CAV affects the epicardial vessels and the myocardial microvasculature separately, usually differing from patient to patient.[15] This complex pathophysiology renders anatomical detection of microvascular disease with either ICA or IVUS difficult. Finally, both methods are invasive, expensive, and associated with complications.[16]

Cardiac MRI and myocardial first-pass perfusion imaging have become promising alternative non-invasive methods to evaluate myocardial ischemia in cardiovascular disease.[17,18] However, the repeatability of semi-quantitative myocardial perfusion cardiac MRI for assessment of myocardial perfusion and its diagnostic performance in the setting of CAV have not been thoroughly evaluated. The objectives of this study were to 1) evaluate the repeatability of myocardial rest-perfusion MRI in healthy volunteers and to 2) assess the feasibility of using semi-quantitative perfusion parameters for the identification CAV in Tx recipients compared to the current gold standards, IVUS and ICA.

2. Material and Methods:

2.1. Study Cohort

Twenty cardiac transplant patients (13 male, age 47.3 ± 17.4 years, mean BMI: 28.3 ± 4.34) were prospectively recruited from August 2014 through March 2017 from a single tertiary care medical center. Eligible patients included any adult cardiac transplant patient who was greater than 1 year post-transplant at the time of enrollment. Exclusion criteria included standard MR contraindications such as history of gadolinium allergy, nephrogenic systemic fibrosis, GFR < 30 ml/min or implanted foreign metal bodies (e.g., pacemakers, aneurysm clips, bullet fragments). Twelve adult healthy volunteers (10 male, age 52 ± 17 years, body weight: 93 ± 20 Kg) were additionally recruited to undergo cardiac MRI. Exclusion criteria included the MR contraindications described above and any history of cardiovascular disease. All study activity was conducted under local IRB approval and all participants provided written informed consent.

2.2. MRI Imaging

Healthy Volunteers

MRI scans were performed at 1.5 T (Aera, Siemens, Erlangen, Germany). Healthy controls underwent two cardiac MRI perfusion scans with the same imaging protocol, separated by an approximately 2-week interval (15.6 ± 2.4 days). Perfusion data were acquired at three short-axis locations (base, mid, apex) of the left ventricle (LV) after injection of a single dose of gadopentetate dimeglumine (Magnevist, 0.1 ml/kg at 4 ml/sec injection rate). Data acquisition was started simultaneously with contrast agent injection and 3 short-axis slices were acquired during each RR-interval for a duration of 60 seconds during partial breath holding. Motion correction was used to account for cardiac and respiratory motion. Pulse sequence parameters were as follows: FOV = 260–290 × 350–360 mm², spatial resolution = 2.8mm × 2.8mm, slice thickness = 8 mm, TR/TE =151.6–172.8/1.1 ms, flip angle = 12°. measurements = 60.[19]

Transplant Patients

MRI scans were performed at 1.5 T (Aera, Siemens, Erlangen, Germany). CMR rest perfusion protocol was different for patients compared to healthy controls. The patient protocol included stress perfusion imaging using 5 ml intravenous (IV) regadenoson (Lexiscan, Astellas US LLC) that preceded rest perfusion imaging. However, stress imaging is not the focus of this study and stress imaging results will not be discussed. The description of initial pharmacologic stress is to provide context to a subsequent discussion regarding differences in perfusion between transplant patients and volunteers, who did not have initial pharmacologic stress. Rest perfusion images were acquired after 15 minutes of recovery and reversal of regadenoson with 50 mg IV aminophylline. After recovery, rest perfusion LV base, mid, and apex short axis cine images were acquired in identical slice positions during first pass (as described above) after injection of gadobutrol (Gadavist, 0.1 ml/kg at 4 ml/sec injection rate).

2.3. Data Analysis

MRI perfusion data were analyzed using dedicated software (ARGUS, Siemens Leonardo Syngo). As summarized in Figure 1, myocardial perfusion data analysis included LV contour segmentation and subsequent quantification of peak perfusion signal intensity, peak slope of the signal change, and time-to-peak perfusion (TTP) at rest. A single reviewer contoured the endocardial and epicardial borders of the LV at each slice for each study participant. Analysis of regional LV perfusion was based on the AHA-16-segment model, which involves evaluating the LV in three slices – base, mid, and apex – and dividing each slice into 6, 6, and 4 segments, respectively (see Figure 1b).

Figure 1 –

Semi-quantitative analysis of myocardial perfusion. (a) Endocardial (red) and epicardial (green) LV borders in an early perfusion image with main signal and contrast in the LV blood pool. (b) Analysis of LV perfusion using the AHA 16 segment model. (c) Graphical representation of LV signal intensity as a function of time; the peak signal intensity corresponding to peak or perfusion, is labeled. (d) Resulting data output including peak perfusion, (Value SI), TTP, and upslope.

Tx patients underwent ICA using standard protocols and the degree of coronary artery narrowing was assessed by the angiographer through visual analysis. These results were then used to classify the degree of CAV according to the 2010 ISHLT CAV Grading Report.[6] Given that IVUS has been shown to be a more sensitive method for CAV detection, an intimal medial thickness (IMT) > 0.5 mm measured by IVUS in the LAD was also used to define CAV. The routine ICA or IVUS examination occurring closest to the date of each patient’s cardiac MRI was used for classification (no greater than one year prior to the cardiac MRI study).

2.4. Statistical Analysis

The repeatability of peak signal intensity, slope, and TTP was assessed by comparing results from the two cardiac MRI exams completed by each healthy control using the intraclass correlation coefficient (ICC) calculated from a two-way mixed effects model and the coefficient of variation (COV). Bland-Altman plots were also applied to examine inter-study variations. Demographics and clinical characteristics of transplant patients without CAV (CAV−) and with CAV (CAV+) were compared using unpaired, two-tailed t-tests for continuous data and chi-square tests for categorical data. Differences in perfusion parameters between CAV− and CAV+ transplant patients were examined both globally and on a per-segment basis using unpaired, two-tailed t-tests. For all group comparisons, normality of the data and homogeneity of the variances were evaluated as appropriate. A p-value of p<0.05 was considered statistically significant.

An ROC analysis was conducted to assess the diagnostic performance of global peak perfusion in relation to CAV status, and using methods described by Liu,[20] an optimal diagnostic cut point was determined. All statistical analyses were conducted using MedCalc (Ostend, Belgium) and Stata 14 (StataCorp, College Station, TX, US).

3. Results:

3.1. Study Cohort

Of the twenty total transplant patients, nine patients had evidence of CAV. Two of the CAV+ patients were diagnosed by IVUS, six were diagnosed by ICA, and one had both ICA and IVUS evidence of disease. Eight patients had ISHLT CAV Stage 1 (mild) disease and one patient had ISHLT CAV Stage 2 (moderate) disease. Select demographics and clinical characteristics of the CAV+ and CAV− groups are described in Table 1. The CAV+ group had a higher proportion of males (p = 0.038) and had higher systolic/diastolic blood pressures (128 ± 10 mmHg vs. 117 ± 11.6 mmHg, p = 0.045) and (82.6 ± 5.64 mmHg vs. 71.4 ± 12.6 mmHg, p = 0.025). In addition, the CAV− group had a higher proportion of hyperlipidemia (p = 0.023).

Table 1 –

Demographics and Clinical Characteristics of Transplant Patients.

Characteristic	CAV− (n=11)	CAV+ (n=9)	p-value
Age at enrollment (mean ±SD)	50.4 ± 18.4	43.4 ± 15.2	0.39
Male – no. (%)	5/11 (45.5)	8/9 (88.9)	0.04
White race – no./total no. (%)	9/11 (81.8)	5/9 (55.6)	0.24
Body Mass Index – kg/m2	27.3 ± 4.08	29.5 ± 4.35	0.28
Systolic Blood Pressure (mmHg)	117 ± 11.6	128 ± 10	0.05
Diastolic Blood Pressure (mmHg)	71.4 ± 12.6	82.6 ± 5.64	0.03
Heart Rate (beats/min)	86.5 ± 14.8	93.1 ± 18.7	0.43
Ejection Fraction – (%)	61.2 ± 3.71	55.3 ± 7.73	0.08
Hypertension – no./total no. (%)	11/11 (100)	8/9 (88.9)	0.35
Hyperlipidemia – no./total no. (%)	8/11 (72.7)	2/9 (22.2)	0.02
Coronary Artery Disease – no./total no. (%)	3/11 (27.3)	1/9 (11.1)	0.38
Diabetes (Type II) – no./total no. (%)	3/11 (27.3)	4/9 (44.4)	0.46
Reason for Transplant
Ischemic Cardiomyopathy – no./total no. (%)	1/11 (9.1)	0/9 (0)	0.34
Non-Ischemic Cardiomyopathy – no./total no. (%)	10/11 (91.9)	9/9 (100)	0.34
Congenital Heart Disease – no./total no. (%)	1/11 (9.1)	1/9 (11.1)	0.89

3.2. Repeatability of Semi-Quantitative First Pass Resting Perfusion

Twelve healthy volunteers successfully completed two serial cardiac MRI examinations within the specified timeframe, resulting in 192 pairs of data points (12 subjects × 16 LV segments) for each of the myocardial perfusion parameters. Average values over all 16 LV segments are summarized in Table 2. There was moderate agreement of peak signal intensity, slope, and TTP on a per-segment basis between the first and second scans, with ICC values of 0.53, 0.48, and 0.59 respectively (all, p < 0.001). In addition, peak signal intensity, slope, and TTP were moderately variable with COV values of 23.01%, 41.96%, and 34.93% respectively. Bland-Altman plots are shown in Figure 2.

Table 2 –

Interstudy variability of rest perfusion Cardiac MRI in healthy volunteers

	Scan 1	Scan 2	ICC (95% CI)	COV (%)
Peak Perfusion	85.3 ± 29.06	86.0 ± 28.75	0.53 (0.43 – 0.63)	23
Slope	4.2 ± 2.5	4.4 ± 2.5	0.48 (0.36 – 0.58)	42
TTP	26.2 ± 13.0	27.3 ± 16.2	0.59 (0.50 – 0.68)	35

TTP – time to peak perfusion; ICC – intraclass correlation coefficient; COV – coefficient of variation

Figure 2 –

Bland-Altman plots showing scan-rescan variability

3.3. Evaluation of CAV with rest perfusion Cardiac MRI

As shown in Table 3, the peak myocardial perfusion parameter significantly differed between the CAV− and CAV+ groups (139.5 ± 30.2 vs 90.7 ± 27.0, p < 0.001). Comparison of peak perfusion on a per-segment basis between CAV− and CAV+ patients demonstrated significant differences in all basal and mid segments (Segments 1–12), but not in the apical segments (Segments 13–16) (Figure 3). ROC analysis for assessment of the diagnostic performance of global peak perfusion in discerning CAV status demonstrated an area under the curve of 0.9. This is depicted below in Figure 4. A cut-point of peak perfusion = 101.5 (95% CI is 80.4–122.7) was estimated to provide optimal diagnostic accuracy, with a resulting sensitivity and specificity of 0.78 and 1.00, respectively.

Table 3 –

Comparison of average global perfusion parameters between CAV positive (CAV+) and CAV negative (CAV−) subjects.

Group	(B) CAV− (n=11)	(C) CAV+ (n=9)	p-value
Mean Peak Perfusion	139.5 ± 30.2	90.7 ± 27.0	0.001
Mean Slope	7.3 ± 4.1	4.5 ± 3.5	0.107
Mean TTP	26.4 ± 11.8	31.7 ± 14.5	0.387

CAV – Cardiac Allograft Vasculopathy; TTP – time to peak perfusion

Figure 3 –

Regional Myocardial Perfusion for CAV− (L) and CAV + (R) Patients in AHA-16 Segment Model

* indicates statistically significant difference between CAV+ and CAV−, p<0.05

Figure 4 –

Parametric ROC curve for global peak perfusion

4. Discussion:

4.1. Discussion Overview

Our feasibility study demonstrates that semi-quantitative surrogate markers obtained by cardiac MRI, particularly global and segmental peak perfusion, are potentially promising non-invasive radiological biomarkers that could be used to detect CAV in patients breathing normally at rest. In addition, we demonstrated that cardiac MRI is moderately repeatable in healthy volunteers. Although the current ISHLT standards recommend cardiac cathfeterization and IVUS as primary and secondary methods of CAV screening, both have limitations previously mentioned. This has led the transplant community to seek a robust, less invasive method with higher diagnostic accuracy. Multiple non-invasive imaging modalities have been investigated in CAV detection, most being able to suggest the lack of significant disease, but unable to detect early disease.[21,22] This is secondary to limited ability in detecting balanced ischemia secondary to the diffuse nature of CAV and the complex interplay between microvascular disease and epicardial disease.[23,24] Therefore, our study addresses an important unmet need in the evaluation of cardiac transplantation health in a more effective manner.

4.2. Cardiac MRI as a non-invasive method to detect CAV

Cardiac MRI is an appealing modality to detect CAV, as it provides a robust comprehensive structural and functional assessment. A number of semi-quantitative parameters obtained by cardiac MRI can be used to estimate myocardial perfusion. Interestingly, in our study, it was the peak perfusion parameter that showed most statistical significance in differentiating CAV+ and CAV− patients.

Our study assessed the potential clinical utility of cardiac MRI as a non-invasive method to detect CAV, which is currently screened for by annual cardiac catheterization. We compared cardiac transplant patients based on CAV status as determined by cardiac catheterization or IVUS. Previous work has described perfusion differences between CAV+ and CAV− patients who were pharmacologically stressed.[25] However, our aim was to assess if perfusion differences could be seen even when CAV+ patients were at rest.

In our study, statistically significant differences were seen with both the average global peak perfusion as well as the segmental peak perfusion values between the CAV+ and CAV− group in the basal and mid segments. The lack of significant findings in the apical segments may be the result of technical difficulties of perfusion measurement in that area. The fact that significant perfusion differences were seen even though the CAV+ population was mostly ISHLT Stage 1 may indicate that CMR is detecting microcirculatory disease not visualized ICA and IVUS. However, the lack of confirmatory invasive metrics measuring microcirculation in our study, such as index of microcirculatory resistance (IMR) or coronary flow reserve (CFR),[26] preclude this conclusion. Overall, our results indicate that semi-quantitative surrogate perfusion markers measured by cardiac MRI may have the ability to detect CAV through global perfusion deficits at rest, which aligned with results of the gold standard ICA and IVUS. This provides support to the potential of cardiac MRI’s use as a non-invasive and accurate imaging modality to assess for CAV.

In previous studies, semi-quantitative estimation of myocardial perfusion reserve index (MPRi) obtained by cardiac MRI have been reported as measures of perfusion deficits associated with CAV. For example, Miller et al demonstrated that MPRi obtained after IVUS and ICA was predictive of both epicardial disease and microvascular disease, even with a higher sensitivity and specificity than ICA.[27] MPRi is a biomarker for global myocardial blood flow, and hence microvascular disease could in theory cause reductions in the MPRi prior to IVUS being able to detect evidence of disease in the larger epicardial vessels. Like most semi-quantitative parameters, there was not an established reference standard for MPRi. Chih et al demonstrated reduced myocardial perfusion index (MPRi) in transplant recipients suffering from CAV. They quantified an MPR < or equal to 1.68 as having a 100 % sensitivity and NPV for diagnosing CAV in the correct clinical context with moderate specificity.[28] Finally, in a study of 43 transplant recipients, Schwid et al demonstrated that semi-quantitative parameters obtained by stress perfusion cardiac MRI (upslope and MPRi) have the ability to detect moderate to severe CAV in cardiac transplant recipients.[25] These promising results all outline the ability of cardiac MRI to potentially play a role in detecting CAV.

The perfusion values obtained in this study were similar to those reported in the literature. Specifically, Tarroni et al reported rest myocardial perfusion values ranging from approximately 60–100 units of signal intensity[29] while Atilli et al indicated rest myocardial perfusion values ranging from 80–150 units of signal intensity.[30] Although Tarroni and Attili et al have reported cardiac MRI myocardial perfusion rest data, these were collected in coronary artery disease (CAD) patients as opposed to CAV patients. Furthermore, the Tarroni and Atilli studies utilized different gadolinium agents which could lead to different results due to varying relaxitivity profiles.

4.3. Cardiac MRI repeatability

We demonstrated that quantification of resting perfusion parameters with cardiac MRI was moderately repeatable between healthy volunteer rest perfusion scans separated by an approximately two-week interval. However, the ICC values ranging from the 0.5 to 0.75 level indicate that although there was a moderate level of correlation between the first and second scans, there is room for improvement. Furthermore, the levels of COV were high, likely secondary to variability in image acquisition and analysis. Optimal COV value would have been less than 20%, although the COV value for global perfusion parameter reaching 23% was encouraging. Technical improvements in image acquisition and processing are likely necessary to improve cardiac MRI’s inter-study variability.

It should be noted that peak perfusion for the healthy volunteers was considerably lower than that of the transplant patients. As noted in the Methods section, the healthy volunteer and transplant patient groups underwent different CMR protocols, and as such, the groups should be considered separately. That being said, a few factors may explain the difference. First, the patients received regadenoson with subsequent aminophylline reversal as pharmacologic-stress images were acquired prior to rest images. Previous authors, including Bhave[31], have noted that regadenson reversal with aminophylline can be incomplete. As such, there may have been residual vasodilation in the patient group that contributed to higher perfusion. Second, Magnevist was used in the volunteer group and Gadavist was used in the patient group. Magnevist’s lower relaxivity would be expected to generate lower peak perfusion.[32] Finally, studies by Kofoed[33] and Wu[34] have shown higher resting myocardial blood flow in transplant patients, which is attributed to vagal nerve denervation.

4.4. Limitations

This study had several limitations and as such, should be viewed as a feasibility study with limited clinical application. First, this was a relatively small, single-center study and its results require confirmation in larger, multi-center studies. Furthermore, we attempted to characterize a global peak value that would be of diagnostic accuracy in our small cohort of patients. We acknowledge that larger multi-center clinical studies will be needed to evaluate the usefulness of resting peak perfusion as a marker of CAV. Second, a control group of patients without cardiovascular disease assessed with the same cardiac MRI protocol was not available. Third, IVUS was not available in all patients, which is a limitation given that IVUS is regarded as a more sensitive measure of CAV relative to ICA. That being said, IVUS was only conducted in the LAD at our institution due in part to the size of the probe, so the IMT of the remaining coronary vessels would be unknown. Fourth, although invasive studies remain the standard, the best-validated non-invasive study is Dobutamine Stress Echocardiography[21], and it would be important to compare the performance of cardiac MR perfusion to this modality. Fifth, perfusion was assessed in patients at rest and it is important to consider stress perfusion or pharmacologically-induced hyperemia given that MPRi may be a more significant measure of CAV progression. The patient protocol included stress imaging, but the analysis of these images was not available at the time of this study. Furthermore, assessment of rest perfusion was complicated by the possibility of residual vasodilation, as described in Section 4.3. Finally, given that the microcirculation is viewed as an important area of CAV pathogenesis, invasive metrics such as IMR or CFR would have served as important comparators to indices of perfusion. Future studies should have larger patient populations with rest and stress perfusion and more complete records regarding IVUS, IMR, and CFR.

5. Conclusion:

Cardiac MRI is a moderately repeatable method of imaging cardiac perfusion, but technical improvements in image acquisition and processing are needed to further lower inter-study variability. In addition, the CAV+ group showed significant differences in average global peak perfusion at rest compared to the CAV− group, providing support for the feasibility of using semi-quantitative perfusion parameters as a potential biomarkers to detect CAV. Further studies are warranted to investigate and confirm the role of these semi-quantitative parameters in detecting CAV.

Highlights:

• Cardiac MR is a moderately repeatable method of imaging cardiac perfusion

• Cardiac MR detected lower perfusion in Cardiac Allograft Vasculopathy (CAV) at rest

• CAV perfusion differences were apparent on global and segmental levels at rest