Diagnostic accuracy of cardiovascular magnetic resonance imaging of right ventricular morphology and function in the assessment of suspected pulmonary hypertension results from the ASPIRE registry

Pulmonary Hypertension (PH) is defined as a mean pulmonary artery pressure (mPAP) greater than or equal to 25 mmHg measured at cardiac catheterisation [1]. It ranges from a rare, progressive condition characterised by a vasculopathy, pulmonary arterial hypertension (PAH), to mild elevations of pulmonary artery pressure commonly seen in association with respiratory and cardiac disease. In patients with PAH an increase in mPAP and pulmonary vascular resistance (PVR) results in right ventricular failure and death with a median survival in untreated patients of less than 3 years [2, 3].

Right heart catheterisation (RHC) is the gold standard test used to assess for the presence or absence of PH by directly measuring mPAP and allows measurement of cardiac output (CO) and index (CI), right atrial pressure, mixed venous oxygen saturation and PVR, which are used as markers of disease severity. However, although safe in expert hands it is an invasive test and its role is limited to confirming a diagnosis of PH and in selected cases assessing the response to treatment. Echocardiography is currently the most commonly used non-invasive test in patients with suspected PH, however, this technique has a number of limitations and does not perform well for certain aetiologies of PH [4, 5], significantly over and underestimating invasively measured mPAP [6]. Consequently, there is increasing interest in developing other non-invasive imaging tools.

Several studies have shown significant relationships between cardiovascular magnetic resonance (CMR) derived measurements with mPAP and PVR in patients with PH; including ventricular mass index (VMI) [7–11], inter-ventricular septal configuration [12–14], pulmonary arterial blood flow quantification and pulsatility [10, 15–18]. However, these studies have primarily been limited to patients with PAH and the diagnostic utility of CMR measurements in a large population of unselected patients with PH has not been evaluated.

The aim of this study was to compare the diagnostic accuracy of a variety of CMR parameters to identify PH confirmed at cardiac catheterisation in unselected patients with suspected PH attending a referral centre.

548 incident, treatment naïve patients who had undergone investigation for suspected PH with a right heart catheter based approach were identified from our PH database between January 2008 and March 2010. 244 patients underwent CMR and RHC within 48 h of each other. 11 patients (5%) were excluded as their CMR scans were degraded by artefact and hence cardiac measurements and volumetric analysis could not be reliably performed in these patients. Of the remaining 233 patients, 194 had PH as defined by mPAP ≥ 25 mmHg, the remaining 39 were labelled as ‘no PH’. Patients with PH were divided into subgroups according to the Dana Point classifications, see Table 1 [24]. 106 patients underwent CMR including phase contrast imaging and RHC between May 2009 and March 2010, and 159 patients underwent late gadolinium enhancement imaging. Echocardiography was performed in 195 of 233 patients at our institution. Scans in the remaining patients were performed at other institutions prior to referral and were excluded from analysis. Echocardiography was in most cases performed prior to RHC with a mean interval 35 ± 50 days. A flow chart of the study profile is shown in Figure 2.

Clinical classification of PH patients as per Dana Point[[24]]
Gp1 PAH (n = 83)
Idiopathic	28
PAH Connective tissue diseases	39
PAH Portal hypertension	5
PAH Congenital heart disease	11
Gp2 PH due to left heart disease (n = 21)
LV Diastolic dysfunction	21
Gp3 PH due to lung diseases and/or hypoxia (n = 29)
Chronic obstructive pulmonary disease	19
Interstitial lung disease	5
Mixed obstructive and restrictive	7
Gp4 CTEPH	59

Correlations with invasive haemodynamics

Ventricular mass index demonstrated the strongest correlation of all quantitative MR measurements with mPAP (r = 0.78; p < 0.0001) and PVR (r = 0.74; p < 0.0001). Echo predicted mPAP correlated well with mPAP (r = 0.74; p < 0.0001), and with PVR (r = 0.74; p < 0.0001), Figure 3. sEI demonstrated a strong correlation with mPAP (r = 0.71; p < 0.0001) and PVR (r = 0.66; p < 0.0001), dEI demonstrated a significant though weaker correlations with mPAP, (r = 0.45; p < 0.0001) and PVR (r = 0.43; p < 0.0001). PA average velocity and PA relative area change demonstrated moderate negative linear correlations with mPAP r = −0.55, p < 0.0001 and r = −0.54, p <0.0001 respectively. Table 4 presents the correlations between all right ventricular morphological and functional MR indices with invasive haemodynamic metrics of mPAP and PVR.

	mPAP r value	p-value	PVR r value	p-value
Cardiac morphology
RVEDVI	0.39	<0.0001	0.42	<0.0001
RV mass index	0.74	<0.0001	0.65	<0.0001
VMI	0.78	<0.0001	0.74	<0.0001
Right heart functional indices
RVEF	−0.43	<0.0001	−0.45	<0.0001
RVSVI	−0.24	0.001	−0.21	0.003
TAPSE	−0.40	<0.0001	−0.46	<0.0001
f-TAAD	−0.53	<0.0001	−0.50	<0.0001
SFD	−0.46	<0.0001	−0.40	<0.0001
f-SFD	−0.55	<0.0001	−0.42	<0.0001
RVRAC	−0.59	<0.0001	−0.56	<0.0001
sEI	0.71	<0.0001	0.66	<0.0001
dEI	0.45	<0.0001	0.43	<0.0001
Phase contrast CMR
Average velocity (cm/s)	−0.55	<0.0001	−0.56	<0.0001
Retrograde flow (L/min/m2)	−0.33	<0.0001	0.25	0.029
Percentage retrograde flow (%)	−0.34	<0.0001	−0.31	0.001
PA relative area change (%)	−0.54	<0.0001	−0.54	<0.0001
Systolic PA area (cm2)	0.28	0.001	0.17	0.054
Diastolic PA area (cm2)	0.35	<0.0001	0.26	0.003

Diagnostic accuracy

VMI and RV mass index were the CMR measurements with the highest diagnostic accuracy for the identification of PH from ROC curve analysis (AUC 0.91 for both), Figure 4. Identification of late gadolinium enhancement at the inter-ventricular hinge points was sensitive (83%) and specific (94%) for the identification of PH (AUC 0.89). Measures of right ventricular function were of only modest diagnostic accuracy for determining the presence of PH.

PA relative area change (AUC 0.87), PA retrograde flow (AUC 0.84), diastolic PA area (AUC 0.82) and average PA velocity (AUC 0.80) were the measurements made from phase contrast sequence images with the highest diagnostic accuracy for the detection of PH. PA relative area change < 15% was sensitive (86%) and specific (70%) for the detection of the presence of PH. PA retrograde flow greater than 0.3 L/min/m² demonstrated sensitivity (82%) and specificity (71%) for the detection of pulmonary hypertension, more accurate than percentage retrograde flow with lower sensitivity (73%) and specificity (56%). PA area measured in systole (AUC 0.77) was a weaker marker of PH than diastolic PA area (AUC 0.82). Table 5 presents sensitivity, specificity, positive predictive value and negative predictive value and ROC analysis for the MRI measurements. Table 5 presents the CMR indices between subgroups of PH.

	Sensitivity	Specificity	PPV	NPV	AUC
Cardiac morphology
RVEDV index ≥ 75 ml/m2	67	50	89	20	0.72*
RV mass index ≥ 20 g/m2	83	84	96	50	0.91*
VMI ≥ 0.4	81	88	97	50	0.91*
Late gadolinium enhancement	83	94	98	58	0.89*
Right heart functional indices
RVEF ≤ 35%	67	71	93	28	0.72*
RVSVI ≤ 30 ml/m2	70	47	89	21	0.68*
TAPSE ≤ 2 cm	76	64	92	32	0.75*
f-TAAD ≤ 25%	88	56	92	44	0.78*
SFD ≤ 1 cm	77	67	93	34	0.73*
f-SFD ≤ 25%	79	64	92	35	0.78*
RVRAC ≤ 30%	61	81	95	28	0.79*
sEI ≥ 1.2	65	61	90	24	0.72*
dEI ≥ 1.17	62	54	88	20	0.57
Phase contrast CMR
Average velocity ≤ 10 (cm/s)	82	62	94	32	0.80*
Retrograde flow ≥ 0.3 (L/min/m2)	83	71	95	38	0.84*
Retrograde flow (%)	73	56	87	34	0.75*
PA relative area change ≥ 15%	86	70	94	48	0.87*
Systolic PA area ≥ 8 cm2	74	67	92	32	0.77*
Diastolic PA area ≥ 6 cm2	88	66	94	52	0.82*

AUC (area under the curve) *p < 0.05.

The bias between observers for CMR derived RV mass was −2.1 g (standard deviation 10.8) with limits of agreement −23.3 to 19.1 g. For VMI measurements the bias between observers was 0.02 (standard deviation 0.1) with limits of agreement of −0.21 to 0.24. See Figure 5.

The diagnostic accuracy of TRJV and echo derived mPAP had good diagnostic accuracy for predicting the presence of PH, AUC 0.86, and AUC 0.88 respectively,

This study has demonstrated that VMI and sEI were the CMR metrics with the strongest correlation with mPAP and PVR in the largest population of patients with suspected pulmonary hypertension studied to date undergoing CMR as part of their routine diagnostic work-up. VMI also demonstrated the highest diagnostic accuracy for the diagnosis of PH of all measured MR indices. In addition, late gadolinium enhancement at the inter-ventricular hinge points was sensitive and specific for the identification of PH and was present in 95% of patients with IPAH and 97% of patients with CTPEH, supporting a role for the routine measurement of these MR metrics in patients undergoing diagnostic CMR evaluation for suspected pulmonary hypertension.

CMR provides accurate and reproducible measurements of RV morphology and function, including mass, RVEDVI, RVEF, and RVSVI [25]. Currently manual RV volume and mass measurements are preferred to semi-automated methods with a manual technique showing stronger inter-observer agreement [7]. This study supports previous work assessing RVEDVI,,RV mass index and VMI which were all found to be elevated in PH when compared to patients with ‘No PH’. RVEF and RVSVI were significantly reduced compared to patients with ‘No PH’ [26, 27]. TRJV and estimated mPAP measured at echocardiography had a good correlation with invasively measured mPAP and had high diagnostic accuracy for the detection of PH confirming that both echocardiography and CMR have diagnostic utility in patients with suspected PH, However, echocardiography and CMR may have limited value in an individual patient due for example to difficult acoustic windows or claustrophobia, respectively.

In the work-up of patients with suspected PH CMR may be particularly useful in patients in whom echocardiography is suboptimal, and this study identifies VMI and late gadolinium enhancement as the metrics of greatest diagnostic value. However, there is currently no reliable fully automated method for calculating RV mass, therefore RV muscle wall has to be manually segmented on each SA image slice. This is a time consuming process and is difficult to perform in patients with low RV mass as the wall is thin and is difficult to accurately measure, further work refining methods of determining RV mass are required [8, 9]. RV mass intuitively increases over time due to elevated RV after load in patients with PH. We postulate that RV hypertrophy is an adaption mechanism of the RV in order to preserve function. The temporal relationship between RV mass and pulmonary artery pressure has not been fully described in the literature. The length of time an individual is exposed to high pulmonary artery pressure and their response to this increased after load will influence the degree of hypertrophy. Studying the temporal relationship between RV mass and mPAP would be of particular value when assessing RV responses to therapy.

Simple functional measurements such as left ventricular eccentricity indices have the advantage over other more labour intensive MR indices that they are easy and quick to measure, and can be performed in routine clinical practice whilst reviewing the SA images. Although these measurements have been investigated with echocardiography, there is little published evidence for their clinical utility using MRI. Given that the reproducibility of right ventricular measurements from echocardiography is accepted as challenging, validation of such measures with MRI is desirable. sEI assesses the effect of the ventricular pressure differential on the shape of the left ventricle, with values increasing as the septum bows further to the left side deforming the LV cavity and has been shown to correlate with pulmonary arterial and RV pressure using echocardiography in patients with PH [28]. Our MR derived sEI measurement correlated strongly with mPAP and PVR as shown with echocardiography [23, 29, 30]. We demonstrated similar results using MRI to those found using echocardiography with a sEI index >1.2 suggesting the presence of PH [23]. In patients with mild elevations in mPAP, the ventricular pressure gradient is not significant enough to cause paradoxical motion, this in part explains the poor sensitivity of sEI to detect the presence of PH and caution should be taken when interpreting sEI in patients with low mPAP.

Patients with PH have significantly more retrograde PA flow and significantly reduced average PA velocity than patients without PH. Retrograde PA flow >0.3 L/min/m² and average velocity (10 cm/s) had reasonable diagnostic accuracy for the detection of PH in this studyl. Sanz et al. suggest that average velocity is the most robust phase contrast diagnostic measurement at the pulmonary artery [15], with reduced average velocity and increased pulmonary artery area occurring simultaneously as a response to increased pressure and resistance. However, retrograde flow outperformed average velocity as a diagnostic tool in our study, and our finding supports a recent study showing that early retrograde flow is a characteristic feature in patients with PH [31].

We have also summarised the differences in haemodynamic and CMR parameters between the larger sub-groups of patients in this study, Table 3. Patients with IPAH and CTEPH had evidence of severe pulmonary hypertension at right heart catheter and 95% and 97% of patients respectively had delayed myocardial enhancement demonstrating that this metric is highly sensitive for the detection of severe increases in RV after load seen IPAH and CTEPH. Importantly qualitative assessment of late gadolinium enhancement can be performed rapidly. Patients with IPAH and CTEPH had more severe disease at cardiac catheter than patients with SSc-PAH and this is reflected in CMR morphological findings with increased RV mass index and increased VMI and more severe functional impairment than in IPAH and CTEPH compared to PAH-CTD. There is little published data of CMR RV findings in patients with LV diastolic dysfunction. Importantly this study demonstrates that VMI is low in these patients and they have much better preserved cardiac function than patients with pre-capillary disease such as IPAH and CTEPH where pulmonary vascular resistance is significantly elevated. This is likely to reflect the site of disease with patients with post-capillary disease such as LHD only having minimally elevated pulmonary vascular resistance despite high pressures. The low prevalence of late gadolinium enhancement in 27% in patients with PH-LHD is consistent the primarily post-capillary nature of pulmonary vascular disease in the majority of these patients. Patents with PH-RESP had moderate elevation of mPAP and PVR and elevated VMI and impaired RV function consistent with a significant increase in RV after load.

MRI is a robust alternative to echocardiography in the evaluation of patients with suspected PH. VMI showed a linear relationship with mPAP and had the highest accuracy of all CMR measurements for the identification of PH. Late gadolinium enhancement, PA relative area change and retrograde flow are other useful markers that should be evaluated in patients with suspected pulmonary hypertension who undergo CMR.