Cardiovascular magnetic resonance patterns of biopsy proven cardiac involvement in systemic sclerosis

Background To determine morphological and functional cardiovascular magnetic resonance (CMR) patterns in histopathologically confirmed myocardial involvement in patients with systemic sclerosis (SSc). Methods Twenty patients (6 females; mean age 41 ± 11 years) with histopathologically proven cardiac involvement in SSc in the years 2008–2016 were retrospectively evaluated. Morphological, functional and late gadolinium enhancement (LGE) images were acquired in standard angulations at 1.5 T CMR. Pathologies were categorized: 1) Pericardial effusion; 2) pathologic left (LV) or right ventricular (RV) contractility (hypokinesia, dyssynchrony, and diastolic restriction); 3) reduced left (LV-EF) and right ventricular ejection fraction (RV-EF); 4) fibrosis and/or inflammation (positive LGE); 5) RV dilatation. 95 % confidence intervals (CI) were calculated for appearance of pathologic EF and RV dilatation. Results Seven patients (35 %) had positive CMR findings in three categories, 9 patients (45 %) in four categories and 4 patients (20 %) in five categories. The distribution of pathologic findings was: minimal pericardial effusion in 7 patients (35 %), moderate pericardial effusion >5 mm in nine patients (45 %); abnormal LV or RV contractility in 19 patients (95 %), reduced LV or RV function in 14 patients (70 %; 95 % CI: 51–88 %), pathologic LGE in all patients, RV dilatation in 6 patients (30 %; 95 % CI: 15–54 %). Conclusions CMR diagnosis of myocardial involvement in SSc requires increased attention to subtle findings. Pathologic findings in at least three of five categories indicate myocardial involvement in SSc. Electronic supplementary material The online version of this article (doi:10.1186/s12968-016-0289-3) contains supplementary material, which is available to authorized users.


Background
Systemic sclerosis (SSc) is a rare chronic systemic autoimmune disease associated with different autoantibodies, the most common are anti-topoisomerase I (Anti-Scl-70) and anti-centromere antibodies (ACA). SSc is associated with high morbidity and mortality [1]. The pathogenesis of SSc is not entirely understood but microvascular changes are suspected to be the main cause of connective tissue infiltration and fibrosis [2]. Besides cutaneous manifestations, visceral involvement especially of bowel, lung, heart and kidney is more common and occurs earlier in patients with diffuse cutaneous subset of SSc (dcSSc) than in patients with limited cutaneous SSc (lcSSc) [3,4]. In general, the prognosis in patients with myocardial involvement is poor [2].
In these patients with cardiac involvement, nonsegmental perfusion defects can be observed in myocardial stress perfusion cardiovascular magnetic resonance (CMR), suggesting resemblances in the pathogenesis of SSc associated Raynaud's phenomenon and transient myocardial ischemia [5]. Cardiac Raynaud's phenomenon may be one reason for subsequent myocardial fibrosis, in addition to non-specific hypertrophy.
Standard treatment of SSc includes cyclophosphamide, which can be cardiotoxic [11]. In our patient group with severe scleroderma, CMR was part of a routine diagnostic work up in the course of preparation and risk stratification for autologous stem cell transplantation and escalation of potentially cardiotoxic therapy. Myocardial histopathological fibrosis and inflammation in SSc is by trend associated with cardiac events [12].
Aim of this study was to systematically determine morphological and functional CMR imaging patterns in systemic sclerosis in patients with histopathologically confirmed myocardial involvement in SSc.
This study retrospectively evaluated a subpopulation of patients with cardiac involvement in SSc as proven by endomyocardial biopsy (EMB) previously undergoing CMR in the years 2008-2016. Clinical findings in a part of the study group (16 patients) were published by Mueller et al. [12]. All patients were clinically diagnosed by an experienced rheumatologist according to the 2013 criteria of the American College of Rheumatology [13]. None of the patients presented with symptoms of cardiac failure. For clinical and laboratory suspicion of myocardial involvement, patients underwent EMB including histopathologic, immunohistologic and molecular pathological evaluation at median 14 days after CMR.

Endomyocardial biopsy and analysis
EMB was sampled from the right ventricular septum. Histopathological and immunohistological analyses were performed by experienced cardiopathologists as described by Mueller et. al. [12]: The degree of fibrosis was rated in percent of the biopsy area. Inflammation was graded 0-4: grade 0 = no inflammation; 1 = single inflammatory cells; 2 = few foci of inflammation; grade 3 = several foci of inflammation; grade 4 = pronounced inflammation. Molecular detection of viral genomes was performed.

CMR acquisition
CMR was performed on a 1.5 T scanner (Magnetom Avanto, Siemens Healthcare, Erlangen, Germany) equipped with a gradient system with a maximum strength of 45 mT/m and a maximum slew rate of 200 T/m/s. A six channel body array coil and six coils of the spine coil were used for reception of MR-signals. All images were acquired in breath hold technique with ECG triggering.
For functional imaging, Steady State Free Precession (SSFP) cine loops (repetition time (TR) 3 ms; echo time (TE) 1.5 ms; flip angle 60°, slice thickness 5 mm; matrix 256x192; 25 frames per cardiac cycle) were acquired in standard angulations: four-chamber view (4CV), twochamber view (2CV) and a stack of short-axis slices (gap 5 mm) covering both entire ventricles from base to apex. For anatomical overview, a T2w Half-fourier Acquisition Single-shot Turbo Spin-Echo (HASTE) in 5 mm gapless axial slices covering the thorax was performed. Optimal inversion time (TI) was determined with an inversion time localizer (TI Scout) SSFP, magnetization preparation: slice selective inversion pulse, 20 ms increment for inversion recovery measurement, (TR 24 ms, TE 1.12 ms, flip angle 60°, slice thickness 8 mm). The inversion time localizer was acquired 9 min after administration of 0.15 mmol Gadobutrol per kg body weight (Gadovist, Bayer Healthcare, Leverkusen, Germany) in a mid-cavity short-axis slice.
LGE imaging was performed 10-15 min after contrast agent application with a 2D T1-weighted inversion recovery spoiled gradient echo sequence (segmented k-space readout, TR 8.0 ms, TE 4.9, TI as predetermined and incrementally adapted, flip angle 30°, slice thickness 8 mm) in standard angulations with corresponding slice positions to the functional imaging sequences. For all sequences, the field of view was 280-340 mm adapted to patient's size, kept to a minimum, respectively.

Image analysis
Image analysis for the retrospective study analysis was performed by two independent readers (eleven and five years of experience in cardiac imaging), aware of the histopathologic biopsy result, using an offline workstation CVI42 (Circle Cardiovascular Imaging, Calgary AB, Canada). Initial reading of the images in clinical routine had been performed by different readers, blinded to the result of the subsequent biopsy. The results of the initial reports were evaluated categorically in 'cardiac SSc involvement' and 'no cardiac SSc involvement'.

Morphology and function
Contractility was evaluated according to the following criteria: Presence of hypokinesia, dyskinesia, intraventricular synchrony, and diastolic function for both ventricles respectively, in a visual qualitative analysis. Functional parameters were determined by volumetry of the left and right ventricle in a stack of short-axis slices according to the modified Simpson rule [14]. Enddiastole and end-systole were determined manually. Endocardial contours were drawn with help of a semiautomated region growing algorithm in both ventricles. Epicardial contours of the left ventricle were drawn manually. Left ventricular end-diastolic volume (LV-EDV) with the corresponding left ventricular enddiastolic volume index (LV-EDVI) normalized to body surface area (BSA), left ventricular ejection fraction (LV-EF) as well as left ventricular myocardial mass (LV-MM) and the respective index (LV-MMI) were calculated. Accordingly, right ventricular end-diastolic volume (RV-EDV), right ventricular end-diastolic volume index (RV-EDVI) and right ventricular ejection fraction (RV-EF) were assessed. Ventricular dilatation was determined according to the recently published criteria by Kawel-Boehm et al. 2015 [15]. Ventricles were classified as dilated if the ventricular end-diastolic volume index was above the upper limit calculated as mean +2 standard deviations (SD) of the normal collective regarding gender and age <60 years: LV-EDVI >95 ml/m 2 for females, >100 ml/m 2 for males; RV-EDVI >96 ml/m 2 for females, >111 ml/m 2 for males. The ventricular function was classified as reduced, if the ejection fraction was below the lower limit calculated as mean -2SD of the normal collective considering gender and age <60 years: LV-EF <58 % for females, <57 % for males; RV-EF <52 % for females, <50 % for males. LV-MM was classified as increased, if LV-MMI was above the upper limit determined as mean +2SD of the normal collective considering gender and age <60 years: LV-MMI >77 g/m 2 for females, >91 g/m 2 for males.
Additionally, the right ventricular diameter was measured in four-chamber view 4CV. RV dilatation was assumed if cross-section dimension in 4CV and basal short-axis was above two standard deviations of the normal value according to Hergan et al. [16].
RV myocardial thickness was measured in the posterior wall in a mid-ventricular short-axis slice. RV hypertrophy was assumed if RV wall thickness was 5 mm or higher [18,19]. Maximum thickness of pericardial effusion was measured in end-diastole in Steady State Free Precession (SSFP) cine loops. Pericardial effusion was classified minimal ≤5 mm pericardial space, and moderate >5 mm [20].

Late Gadolinium enhancement and inversion time localizer
Inversion time localizer nulling patterns of the myocardium and blood pool were categorized by the nulling type patterns 1 (normal) and the abnormal patterns 2-4 as imposed by Pandey et al. [21]: Physiologic type 1 nulling pattern: blood pool nulls first, followed by myocardium and spleen at the same time. Pathologic type 2 nulling pattern: myocardium nulls first, followed by blood pool and spleen at the same time. Pathologic type 3 nulling pattern: myocardium and blood pool null simultaneously, followed by the spleen. Pathologic Type 4 nulling pattern: myocardium, blood pool and spleen all null at different times.
LGE was evaluated qualitatively and semi-quantitatively. Datasets in which the Inversion Time (TI) Scout indicated an insufficient nulling of the myocardium in contrast to the blood pool were excluded from LGE quantification. For quantification of LGE contours in all short axis slices were drawn manually with care to exclude pericardial fat and blood pool at the myocardial borders. For visual reading and semi-automated thresholding, the recommendations of the Society for Cardiovascular Magnetic Resonance Task Force were adopted [22]: signal intensity >5 standard deviations of the normal myocardium was considered as full intensity LGE. In addition, a grayscale analysis of intermediate-signal intensity LGE was performed (≥2SD but <5SD of the normal myocardium). Full intensity and intermediate intensity LGE were recorded in % of the myocardial mass. Inversion time for the first and last LGE image was recorded.

Statistical analysis
Statistical analysis was performed using JMP (Version 12.2, SAS Institute Inc., Cary NC, USA). All continuous variables are expressed as mean value ± standard deviation and range in square brackets. Interobserver-variability for continuous variables was evaluated with Bland-Altman difference plots. Mean difference and the limits of agreement defined as ±1.96 SD were calculated for LV-EDV, LV-EF, RV-EDV and RV-EF. Over all categorical variables in visual analysis interobserver-agreement was measured with Cohen's Kappa. Paired two-sided t-test was applied for null hypothesis if Kappa equals zero. Agreement in Kappa statistic was rated according to Landis and Koch [23]: 'moderate' (0.41-0.6), 'substantial' (0.61-0.8), and 'almost perfect' (0.81-1).
To test if among the study cohort more than the 2.5 % of the normal collective had pathologic values, 95 % confidence interval (CI) was calculated for the quotient pathologic sample/total sample for LV-EF, RV-EF, LV-EDVI, and RV-EDVI. Higher occurrence of pathologic values in the study cohort was assumed if lower 95 % CI was above 2.5 %. Normal distribution for each value was assessed visually in curves.
End-diastolic myocardial thickness of the LV in the mid-ventricular septum (segment 9) was 10.

Late Gadolinium Enhancement
LGE imaging was not evaluable in one patient due to breathing artifacts. In the visual analysis, three of 19 patients (16 %) had intermediate signal-intensity LGE in linear and patchy pattern (Fig. 1). Ten patients (53 %) had diffuse LGE. Two of these patients had generalized LGE with no reasonable nulling of the myocardium (Figs. 2 and 3), they were excluded from semiquantitative analysis.
In the semi-quantitative analysis, pathologic LGE was found in all patients: full intensity LGE was found in 5. All results are summarized by frequency in Table 2.
In evaluation of the initial clinical CMR reports blinded to subsequent EMB, the report was diagnosing 'cardiac SSc involvement' in 6 patients (30 %).

Statistical analysis
The interobserver variability analysis indicated a mean difference for LV-EDV of -1.8 ml with limits of agreement -11.4 ml; 7.9 ml. The mean difference for LV-EF was 0.7 %, limits of agreement were -5.5 %; 7 %. The mean difference for RV-EDV was 1.7 ml, limits of agreement were -7.9 ml; 11.4 ml. The mean difference for RV-EF was 0.2 %, limits of agreement were -5.7 %; 6.1 %. Kappa statistic for interobserver-agreement in categorical variables was 'almost perfect' 0.913, p < 0.0001. LV-EF, RV-EF, LV-EDVI, and RV-EDVI were normally distributed.
In categorical analysis of pathologic CMR findings, seven patients (35 %) had positive findings in three categories, nine patients (45 %) in four categories, and four patients (20 %) in five categories.

Discussion
In the present study we systematically evaluated CMR findings in a rare subgroup of SSc patients with histopathologically confirmed myocardial involvement.
In CMR, pathologic findings were documented in all SSc patients, confirming a high occurrence of abnormal volumetric data as compared to normal healthy collectives (95%CI). Diastolic dysfunction is known to develop early in the course of myocardial involvement in SSc even before the appearance of clinical symptoms [24]. The pericardial effusion found in the majority of our patients is known to be common in SSc [25], and might contribute to the development of diastolic dysfunction [26]. Pericardial effusion is also common in pulmonary arterial hypertension [25], reflecting the complexity of cardiopulmonary involvement in SSc. Notably, a number of patients only had a slight or minimal pericardial effusion, that will commonly be classified as unspecific. Regional or global enhancement by LGE imaging indicating myocardial fibrosis can be regarded as morphologic correlate for restricted diastolic and systolic function as well as for myocardial electrical conduction disorders, potentially resulting in ventricular dyssynchrony. RV dilatation occurred independently from pulmonary hypertension in a relevant number of patients, a phenomenon that has already been described previously [9]. In this context, RV dilatation may indicate RV involvement as a primary symptom in contrast to secondary dilatation.
In visual reading and semi-quantitative evaluation, we found typical non-coronary distribution of full intensity LGE as described before: a linear enhancement [7]; as well as patchy pattern at the insertion points of the right ventricle-which is also typically found in hypertrophy and pulmonary artery hypertension, the latter being frequently co-existent due to severe lung involvement in SSc patients [27]. In the semi-quantitative evaluation, grayscale intermediate intensity LGE was most frequent. We performed semi-quantitative T 1 evaluation with an ordinal scale by means of the inversion time localizer. Pathologic nulling patterns of inversion time localizer were common in our patient group, indicating severe diffuse myocardial fibrosis. Diffuse myocardial fibrosis frequently evades visual detection and is significantly underestimated in LGE quantification, for nulling of the myocardium suppresses general enhancement by shifting the threshold [28]. As a result, diffusely involved myocardium may appear normal. In semi-quantitative LGE evaluation, the region of interest for remote myocardium may be placed in regions with less but still pathologic contrast enhancement, visually not identified as such. Although diffuse general fibrosis can be detected and quantified using T 1 -mapping, the values remain unspecific for diffuse fibrosis [29]. We did not find a 'runaway' TI with initially short TI and large incremental steps during image acquisition, which is known to be typical for amyloidosis [30]. In consequence, non-invasive assessment of the extent of myocardial fibrosis in SSc remains challenging and the implication of imaging findings for prognosis and further treatment remains uncertain. Fibrosis in biopsy may only correlate with T 1 - LGE does shift threshold of healthy towards pathologic when nulling myocardium, and functional parameters are subject to situational adaption. Extent of cardiac involvement in SSc is not necessarily dependent on other objectively verifiable parameters and subject to subclinical periods, primary or secondary involvement and influenced by cardiotoxic therapy [2,12].
CMR is accepted as one of the non-invasive modalities to investigate SSc involvement of the heart [31]. Yet, to determine the degree of cardiac involvement seems to be difficult even for experienced readers. To date, criteria for the CMR diagnosis of myocardial involvement in SSc are not defined. CMR was performed and evaluated before EMB in all patients, so that no influence on imaging and initial blinded reading can be stated.
As systolic dysfunction is a rare finding in SSc, the severity of the involvement tends to be underestimated [32]. Based on the inversion time localizer principle, T 1mapping should be a promising method to further investigate diffuse myocardial fibrosis in SSc involvement [29]. T1-mapping does not require healthy remote myocardium to detect diffuse fibrosis. Extracellular volume determination in T1-mapping has recently shown the potential to correlate with myocardial fibrosis [33].
Our results might contribute to improve the noninvasive diagnosis of cardiac involvement in SSc, using CMR to identify and estimate cardiac involvement. Discrete CMR imaging patterns could be found concurrently, also in severe SSc myocardial involvement. In our retrospective evaluation, myocardial involvement in SSc was systematically underdiagnosed in the initial clinical report blinded to the subsequent histopathologic workup with true positive diagnoses in only 30 % of our SSc cohort. The main reason may be underestimation of subtle morphologic and functional findings in SSc involvement, as this is a rare cardiac diagnosis. To date, mainly CMR findings in the maximum degree of occurrence have been published, potentially leading to a trend to underestimate subtle, yet SSc-typical, findings. Most CMR findings appear to be unspecific and can be found in a variety of diseases. However, the combination of certain findings as described in our work seems to be characteristic for cardiac SSc involvement. For the lack of correlation, CMR cannot replace EMB. EMB itself is rarely performed in a clinical setting. Application of the proposed criteria may improve assessment of clinically relevant SSc involvement. Although CMR findings in patients with cardiac involvement in SSc can easily go underdiagnosed and classified as unspecific, the combination of several CMR findings is frequent and therefore allows the diagnosis of cardiac involvement in patients with SSc by CMR: Besides distinct LGE patterns, most important findings are involvement of both ventricles, pathologic LV/RV contractility, dyssynchrony or restriction. Reading of inversion time localizer for diffuse LGE and can be easily implemented without additional technical requirements.

Study limitations
The study population is comparably small and the study CMR image readers were biased by knowledge of histopathologically proven involvement in SSc in all patients.
Only patients with severe disease were examined. Mild and limited cutaneous SSc might present with less pathologic CMR findings per individual. No T 2 imaging was performed, but non-quantitative T 2 imaging cannot reliably depict diffuse edema [35,36] and edema has not been found in previous reports [10]. Neither T 1 nor T 2mapping were available at the time patients were examined.

Conclusions
CMR seems to be a valuable tool to identify and assess the extent of cardiac involvement and therefore might be useful for risk stratification and making a decision upon therapeutic strategies. Myocardial involvement in systemic sclerosis can be assumed if pathologic findings in three categories are found: 1) pericardial effusion; 2) ventricular kinesic pattern (LV and RV hypokinesia, dyskinesia, dyssynchrony, and diastolic restriction); 3) reduced LV-EF or RV-EF; 4) positive LGE or pathologic inversion time localizer; 5) right ventricular dilatation.