All images were acquired on a 1.5-T clinical scanner (Sonata, Siemens Medical Solutions) with prospective ECG gating during repeated breath-holds of 6-15 seconds, depending on the patient's heart rate.

After scout imaging was performed, a steady-state free precession sequence (true fast imaging with steady-state free precession [FISP]) was used to acquire functional images of the heart. Sequence parameters were TR/TE, 3.1/1.6; flip angle, 60°; slice thickness, 5.0 mm; slice spacing, 5.0 mm; 13-15 segments per heart beat, giving a temporal resolution of 40-46 milliseconds. The matrix was 256 × 156, and the field of view was altered (300-400 × 204-352 mm) according to individual patient body size to minimize infolding, giving an in-plane spatial resolution of 1.2-1.6 × 1.3-2.2 mm.

Standard long-axis views—four-chamber, two-chamber, and three-chamber (left ventricular outflow tract)—and right ventricular two- and three-chamber views were obtained. A stack of short-axis slices, planned on the four-chamber view, perpendicular to the interventricular septum and parallel to the plane of the mitral valve, was acquired, with complete coverage from the base to the apex. The stack of short-axis slices allows calculation of left ventricular and right ventricular mass, volume, and ejection fraction without any geometric assumptions [17]. The extent of pericardial effusion was graded according to maximum diameter of fluid seen in any view (0, none; 1, trace; 2, ≤ 10 mm; and 3, > 10 mm) [2]. The left ventricular eccentricity index was calculated at mid ventricular level in the short-axis orientation [2] at end-diastole and during maximal septal displacement during early diastole. Tricuspid regurgitation was graded semiquantitatively by assessing the area of signal void in the right atrium.

Delayed contrast-enhanced images were acquired by copying the image position for each of the short-axis slices and long-axis views as for cine images, 10-20 minutes (mean, 14 minutes) after the administration of 0.2 mmol/kg of gadopentetate dimeglumine (Magnevist, Schering) using a 2D inversion recovery fast gradient-echo sequence [18] with triggering on alternate cycles to late diastole (≈ 100 milliseconds before the next Q wave on the ECG). Sequence parameters were 9.6/4.4; matrix, 256 × 208; flip angle, 25°; slice thickness, 5.0 mm; and slice spacing, 5.0 mm. Meticulous attention was paid to the inversion time to null normal myocardium with typical values of 250-300 milliseconds. Patients with severe tricuspid regurgitation and right ventricular dysfunction tended to require a lower inversion time—approximately 200-220 milliseconds—for optimal nulling. In two patients a single-shot sequence rather than a segmented sequence was used to minimize breathing artifacts [19]. If doubt existed about the presence of hyperenhancement, the phase-encoding direction was switched to enable differentiation from image artifacts.

Because hyperenhancement was not usually seen in the standard long-axis views, a modified two-chamber view planned perpendicularly through the inferior right ventricular insertion point was performed in several patients to show enhancement in a long-axis view. The presence of hyperenhancement was agreed on by two experienced observers. Endocardial and epicardial borders were traced manually, excluding trabeculations and papillary muscles. Hyperenhancement was defined by visual analysis, during which the window setting could be freely adjusted to the personal preference of the observer. Hyperenhanced regions were manually contoured on each short-axis slice, and the total volume of hyperenhancement was calculated by the software package MASS (Medis) and converted to mass by multiplying by 1.05. Delayed contrast-enhanced mass was expressed as a percentage of the total (left ventricular plus right ventricular) myocardial mass. Semiautomatic analysis, by setting a window threshold for signal intensity, was not performed because doing so can lead to large overestimations of hyperenhancement compared with manual contouring [20]. Interobserver agreement for the presence of abnormal delayed contrast enhancement (κ, 0.76) [18] and interobserver (5% ± 18%) and intraobserver (4% ± 7%) variability for quantification of delayed contrast-enhanced mass [20] were excellent for our group of patients.

Diagnostic right heart catheterization was performed with a balloon-tipped, flow-directed 7-French Swan-Ganz catheter (131HF7, Baxter Health Care) within 2 days of cardiac MRI measurements in 14 patients. The other patient underwent catheterization at diagnosis but it was not repeated at the time of cardiac MRI, and hemodynamic data are not included for this patient. Right atrial, right ventricular, pulmonary artery, and pulmonary capillary wedge pressures were measured. Blood was sampled with the catheter positioned in the main pulmonary artery. Arterial oxygen saturation was measured from blood sampled from the radial artery. Cardiac output was assessed by the Fick oxygen method, and pulmonary vascular resistance was calculated using the standard formula.

The N-terminal pro brain natriuretic peptide (N-BNP) was measured using a commercially available assay within 48 hours of cardiac MRI.

Analyses were performed using the computer software package Minitab, version 11.21 (Minitab). All ventricular volumes and masses were indexed by body surface area. The relationship between the weight of hyperenhanced mass and clinical, hemodynamic, and cardiac MRI variables was assessed using linear regression analysis for continuous and logistic regression analysis for categoric variables. Variables entered into the regression analyses were the clinical variables age, sex, time since diagnosis, N-BNP (log transformed to normalize data), and the 6-minute walking distance; the hemodynamic variables right atrial pressure, mean PAP, right ventricular end-diastolic pressure, pulmonary vascular resistance, and cardiac index; and the cardiac MRI variables right ventricular and left ventricular end-diastolic and end-systolic volumes, masses, and ejection fractions, left ventricular eccentricity index, tricuspid regurgitation grade, and pericardial effusion grade. A formal power calculation was not performed at the time of study planning because to our knowledge there had been no previous publications on delayed contrast enhancement in PAH to guide sample size requirements. Results are expressed as mean ± SD or median, and interquartile range is given where stated. A p value of less than 0.05 was considered statistically significant.

All subjects completed the imaging protocol, and no adverse reactions to contrast injection occurred. All patients had evidence of right ventricular dilatation (Fig. 1A, 1B, 1C, 1D and Table 2) and tricuspid regurgitation (grade 1, n = 3; grade 2, n = 4; grade 3, n = 2; and grade 4, n = 5). Mean ventricular masses, volumes, and ejection fractions are shown in Table 3. Mean right ventricular volumes and masses were significantly increased and systolic function decreased compared with normal ranges [5]. Mean left ventricular volumes were toward the lower limit of the normal range with a normal ejection fraction [5]. Thirteen of 15 subjects had pericardial effusion (grade 1, n = 4; grade 2, n =4; and grade 3, n = 5). The mean left ventricular eccentricity indexes were 1.4 ± 0.2 in end-diastole and 2.1 ± 0.2 at maximal septal bowing.

All patients showed hyperenhancement of ventricular myocardium (mean weight, 3.1 ± 1.9 g; size range, 0.3-3.9% of total myocardial mass) in the interventricular septum. Hyperenhancement occurred almost exclusively at the right ventricular insertion points (15 inferior, 13 anterior) and was more prominent toward the base of the heart (Figs. 1A, 1B, 1C, 1D and 2A, 2B, 2C, 2D). No subendocardial or transmural enhancement and no hyperenhancement were seen in the right ventricular or left ventricular free walls. The extent of hyperenhancement was not correlated with any clinical or hemodynamic variable but was inversely correlated with stroke volume index (r = -0.67, p = 0.006) and right ventricular ejection fraction (r = -0.63, p = 0.01) measured on cardiac MRI. A trend was seen toward significance for the correlation with the right ventricular end-systolic volume index (r = -0.51, p = 0.052).

Our findings are similar to the those in the only other report of delayed contrast enhancement in PAH, a study by Blyth et al. [13], in which 23 of 25 patients had hyperenhancement affecting the right ventricular insertion points. The extent of hyperenhancement was similar, as were the correlations with right ventricular volumes and the inverse correlation with the right ventricular ejection fraction [13]. We did not find a significant correlation with invasively measured PAP, probably because of the different populations in the two studies. The subjects in our study tended to have more severe disease, as manifested by greater PAPs, right ventricular volumes, reduced systolic function, and longer duration of illness. In the study by Blyth et al., patients were not receiving any specific treatment for pulmonary hypertension, whereas most of our patients were being reassessed for response to treatments that reduce PAP. This fact may explain why delayed contrast enhancement did not correlate with invasive hemodynamics in our study.

Delayed contrast enhancement of the septal right ventricular insertion point has also been reported in congenital heart disease with right ventricle overload [21, 22]. In transposition of the great arteries, in which right ventricular pressures may be similar to those in severe PAH, the extent of hyperenhancement also correlates inversely with right ventricular ejection fraction [21].

Delayed contrast enhancement has been characterized best in relation to ischemic heart disease. The typical pattern is of transmural or subendocardial enhancement following the territory supplied by one or more coronary arteries [7, 18, 23]. In this situation, hyperenhancement represents a collagenous scar formation [7]. The extent of delayed enhancement on cardiac MRI has also correlated significantly with myocardial collagen, but not with myocyte disarray, in a single patient with hypertrophic obstructive cardiomyopathy who underwent heart transplantation [24]. In other forms of cardiomyopathy, late enhancement is also likely to represent interstitial fibrosis [9, 11]. Enhancement at the right ventricular insertion points has been seen, particularly in patients with hypertrophic cardiomyopathy [10, 25], and may reflect focal fibrosis [26]. Myocardial disarray is frequently seen at the right ventricular junctions, even in healthy subjects [27], but disarray was not related to enhancement in an individual patient with hypertrophic obstructive cardiomyopathy [24].

Histologic studies of PAH have focused on changes in the pulmonary vasculature [28]. We, like other authors [13], failed to identify any reports of histologic changes in the right ventricular myocardium in patients with PAH despite the importance of right ventricular function to prognosis [3]. Given that all the patients in our study and all but two (with mild disease and normal right ventricular function) of 25 in a previous report [13] showed delayed contrast enhancement at the right ventricular insertion points, we think that histologic results obtained in two patients with PAH who died and underwent postmortem autopsy, but who did not undergo cardiac MRI during the study period, are likely to be representative of PAH patients in general. In these two patients, fibrosis was present at the right ventricular insertion points (Appendix 1 and Fig. 3A, 3B, 3C) suggesting that this is the causal mechanism of delayed contrast enhancement in our study. However, we also saw evidence of interstitial space expansion, which may increase the volume of distribution of gadopentetate dimeglumine and contribute to hyperenhancement in these areas. Indeed, inflammation is an early response to pressure overload leading to interstitial fibrosis [29] and is also associated with delayed contrast enhancement in acute myocarditis [30]. The small increase in fat, seen at a distance from the perivascular regions, may also contribute to hyperenhancement in these areas because fat also has a high signal intensity on the inversion recovery gradient-echo sequence used for delayed contrast enhancement.

The most surprising aspect of our study is that no delayed contrast enhancement was seen in the right ventricular free wall, even though most of the patients had evidence of severe systolic dysfunction. Contrast enhancement in thin-walled ventricles, such as in arrhythmogenic right ventricular dysplasia, may be difficult to detect because of partial volume effects, but the patients in our study had marked right ventricular hypertrophy, and we did not have any difficulty nulling the right ventricular myocardium (Figs. 1A, 1B, 1C, 1D and 2A, 2B, 2C, 2D). It is unlikely that diffuse homogeneous fibrosis affects the right ventricle because the inversion time to null the right ventricle would have been different from that of the left ventricle, and we did not see any fibrosis in the interventricular septum on histology other than at the right ventricular insertion points.

These findings suggest that the insertion points are particularly prone to developing fibrosis. These junctional areas have considerable forces acting on them—the circumferential strain of left ventricular myocardial shortening and right ventricular longitudinal shortening—which may explain why myocardial disarray and some fibrosis not extending into the mid myocardium may be observed in healthy individuals [27]. In PAH, the force generation required for right ventricular systolic emptying is greatly increased and there is the added abnormal interventricular septal motion, which may create further shear stress at the interventricular septal insertion points. However, the pattern of delayed contrast enhancement is specific neither to the underlying cause [13] nor to PAH [10, 25].

The extent of delayed contrast enhancement in this study was moderately correlated with cardiac MRI parameters of right ventricular dysfunction. Right ventricular systolic dysfunction is closely related to right heart failure, the commonest cause of death in PAH [1, 31], but right ventricular ejection fraction is difficult to measure with echocardiography, and there is a lack of outcome studies with cardiac MRI-measured ejection fraction. No significant correlation was seen with any other hemodynamic or clinical variable that has been shown to predict outcome in PAH [3, 15]. Therefore, the significance of contrast enhancement is unclear and remains to be elucidated in a longitudinal study of PAH. Delayed contrast enhancement in hypertrophic cardiomyopathy correlates with risk factors for sudden death [10] and, in transposition of the great arteries, with clinical events, including syncope and nonsustained ventricular tachycardia [21]. One may speculate that the focal nature of the delayed contrast enhancement in PAH, which probably represents fibrosis, may act as a trigger for potentially fatal ventricular tachyarrhythmias contributing to the risk of sudden cardiac death.

The patients in this study were highly selected, and therefore our results may not be applicable to patients with less severe PAH. The cause and significance of delayed contrast enhancement in PAH need to be determined in future studies. Because this is a cross-sectional study, we are uncertain whether delayed contrast enhancement precedes right ventricular dysfunction, and we cannot say that they are causally related. As in other conditions, the extent of delayed contrast enhancement may vary with time [32, 33], although because the PAP remains elevated we assume the stimuli causing hyperenhancement will also remain, making regression of hyperenhancement less likely. The severity of tricuspid incompetence was assessed semiquantitatively on steady-state free precession imaging and may be underestimated in comparison with fast gradient-echo imaging.

Delayed enhancement on cardiac MRI occurs frequently in patients with severe symptomatic PAH and is inversely related to measures of right ventricular systolic function. The significance of hyperenhancement in this population needs to be determined in a longitudinal study.