Heart Failure And Cardiomyopathies

Prognostic value of multimodality imaging in the contemporary management of cardiac sarcoidosis

Abstract

Background Echocardiography, cardiac magnetic resonance and cardiac 18fluorodeoxyglucose positron emission tomography (FDG-PET) imaging play key roles in the diagnosis and management of cardiac sarcoidosis (CS), but the relative value of each modality in predicting outcomes has yet to be determined. This study sought to determine the prognostic importance of multimodality imaging data over and above demographic characteristics and left ventricular ejection fraction (LVEF).

Methods Consecutive patients newly diagnosed with CS were included. Parameters evaluated included echocardiographic regional wall motion abnormality (RWMA), myocardial strain, LVEF, right ventricular ejection fraction (RVEF), late gadolinium enhancement (LGE) extent, SUVmax and RV FDG uptake. The primary endpoint was a composite of all-cause mortality and serious ventricular arrhythmia.

Results The study population consisted of 208 patients with mean age of 55±13 years and LVEF of 55±12%. During a median follow-up period of 46 (IQR: 18–55) months, 14 patients died and 28 suffered serious ventricular arrhythmias. On multivariable analysis, RWMA (HR for RWMA presence 2.55, 95% CI 1.27 to 5.28, p=0.008), LGE extent (HR per 1% increase 1.02, 95% CI 1.00 to 1.04, p=0.018), RVEF (HR per 1% decrease 0.97, 95% CI 0.94 to 0.99, p=0.008) and RV FDG uptake (HR for RV FDG presence 2.48, 95% CI 1.15 to 5.33, p=0.020) were independent predictors of the primary endpoint, while LVEF was not predictive. The risk of adverse events was significantly greater in those with LGE extent ≥15% (HR for ≥15% presence 3.96, 95% CI 2.17 to 7.23, p<0.001).

Conclusion In our CS population, RWMA, LGE extent, RVEF and RV FDG uptake were strong independent predictors of an adverse outcome. These findings offer an important insight into the key multimodality imaging parameters that may be used in a future risk stratification model of patients with CS.

What is already known on this topic

  • Multimodality imaging is paramount for the diagnosis and management of patients with cardiac sarcoidosis.

  • While previous studies have identified risk biomarkers from individual imaging modalities, the relative prognostic importance of each modality during each phase of the management pathway has yet to be delineated.

What this study adds

  • This study is the first to demonstrate the prognostic value of multimodality imaging using a stepwise clinical algorithm-based approach with echocardiography for screening, and cardiac magnetic resonance and 8fluorodeoxyglucose positron emission tomography imaging for diagnosis and treatment of cardiac sarcoidosis.

  • This study underlines the importance of regional wall motion abnormality, late gadolinium enhancement presence, right ventricular ejection fraction and right ventricular FDG uptake in the prediction of clinical outcome in cardiac sarcoidosis.

How this study might affect research, practice or policy

  • Awareness of such high-risk imaging features should enable earlier identification of at-risk patients and expedite the use of disease-modifying treatments.

Introduction

Cardiac sarcoidosis (CS) is an inflammatory disease characterised by patchy areas of myocardial inflammation and/or fibrosis due to non-caseating granuloma formation in the myocardium.1 In recent years, it has become clear that up to 30% of patients with systemic sarcoidosis may have myocardial involvement as detected by cardiac magnetic resonance (CMR) and cardiac 18fluorodeoxy glucose positron emission tomography (FDG-PET) imaging to detect myocardial fibrosis and inflammation, respectively.2 Cardiac involvement of sarcoidosis confers an adverse outcome with potential complications including advanced atrioventricular (AV) node disease, ventricular arrhythmias (VA), heart failure (HF) and sudden cardiac death.

As outlined in the HRS consensus statement, the development of presumed cardiac symptoms, ECG changes or echocardiographic abnormalities in patients with extracardiac sarcoidosis warrants further investigation with CMR and FDG-PET imaging.3 Consequently, echocardiography may be used as a screening tool to detect regional wall motion abnormalities (RWMA), measure left ventricular ejection fraction (LVEF), assess right ventricular (RV) function, and estimate pulmonary artery pressure. A unique strength of CMR imaging is the detection of late gadolinium enhancement (LGE) which may represent myocardial oedema, necrosis or fibrosis as part of the granulomatous disease process.4 CMR imaging also allows an accurate measurement of right ventricular ejection fraction (RVEF). This information is complemented by FDG-PET imaging which plays a key role in detecting active myocardial inflammation and helps determine the need for immunosuppressive therapy.5

Previous studies have evaluated the prognostic significance of imaging data in CS populations, but these have either been limited by small sample sizes, the study of a single imaging modality or an analysis of limited imaging parameters between modalities.4–16,17 Among the available data, the relative prognostic value of the CMR and FDG-PET parameters in CS remains unclear. Moreover, echocardiographic data including the presence of RWMA and myocardial strain data have rarely been incorporated. During the screening, diagnostic and treatment phases, patients with CS tend to have all three imaging modalities performed. We sought to determine the incremental prognostic value of multimodality imaging parameters using a stepwise clinical algorithm-based approach starting with echocardiography for screening, followed by CMR imaging for diagnosis and FDG-PET imaging for the detection and treatment of active inflammation.

Methods

Patient population

The study population was derived from a retrospective database of consecutive patients referred to our institution for evaluation of suspected CS between January 2015 and February 2020. Patients either had known extracardiac sarcoidosis and had abnormal screening investigations or presented with unexplained cardiac manifestations (figure 1). As part of our clinical algorithm, all patients had a 12-lead ECG and comprehensive two-dimensional transthoracic echocardiography (TTE), followed by CMR imaging and a cardiac 18FDG-PET scan within 3 months. The diagnosis of CS was established in a multidisciplinary team setting guided by the diagnostic criteria set out in the HRS consensus statement.3 Care was taken to exclude other potential causes of the cardiac manifestations and take into account normal variations in the imaging findings such as RV/LV insertion point fibrosis on CMR, or artefactual FDG-PET findings due to dietary failure. An abnormal ECG was defined as presence of left or right bundle branch block, pathological Q-waves, VA, ventricular ectopy, any degree of AV block or supraventricular arrhythmia. Patients with significant structural heart disease including coronary artery disease (previous coronary revascularisation, ST-elevation myocardial infarction or existing coronary stenosis ≥75% in one or more vessel), at least moderate valvular disease or known congenital heart disease were excluded.

Figure 1
Figure 1

Patient pathway for the diagnosis or exclusion of cardiac sarcoidosis (CS). Echocardiography, cardiac magnetic resonance (CMR) and 18fluorodeoxy glucose positron emission tomography (FDG-PET) used in all patients at various diagnostic phases. LVEF, left ventricular ejection fraction.

Echocardiographic protocol

TTE was performed using a Philips EPIQ or IE33 ultrasound system (Philips Medical Systems, Eindhoven, Netherlands). Standard views were obtained and all echocardiographic parameters were measured as per guidelines.18 Presence of RWMA was determined in the standard views. For speckle-tracking strain analysis, 2D images of the LV were obtained in the parasternal short-axis (apical, mid-papillary, basal) and apical 2-chamber, 3-chamber and 4-chamber views. Images were recorded with frame rates between 40 and 90 frames/s. Offline analysis was performed using commercially available software (TomTec Imaging System, 2D-Cardiac Performance Analysis module version 1.4) by an independent expert (JO) blinded to other imaging data. Tracking quality was visually verified and if >2 segments could not be adequately tracked, the strain data were excluded from the final analysis. Global longitudinal strain (GLS) and global circumferential strain (GCS) at peak systole were assessed by averaging the values using a 17-segment model.

CMR data acquisition, analysis and LGE quantification

All CMR scans were performed on a 1.5-Tesla system (Magnetom Sonata, Avanto, or Aera; Siemens, Erlangen, Germany). Imaging protocols included steady-state free precession breath-hold cines for the assessment of ventricular volumes, function and morphology, and LGE sequences for the detection of myocardial fibrosis, as per recommendations.19 Additional tissue characterisation images of T2-weighted imaging were added to the protocol. CMR analyses were performed by an expert reader (AA) blinded to all other patient information using semiautomated software (CMR tools; Cardiovascular Imaging Solutions, London). LGE was identified visually and the location, distribution (solely epicardial, solely midwall, subepicardial+midwall, subendocardial+midwall or transmural in any segments) and focality (unifocal=1 lesion, multifocal ≥1 discrete lesion) within the LV noted using a 16-segment model. The presence of RV LGE and its location (RV side of septum and/or free wall) were noted. LGE extent was assessed semiquantitatively from the area of hyperenhanced myocardium on a 5-point scale. Segments were scored 0=no hyperenhancement, 1=1%–25%, 2=26%–50%, 3=51%–75% and 4=76%–100%. The segment scores were weighted by the midpoint of the range of hyperenhancement, such that 1=13, 2=38, 3=63 and 4=88. The LGE extent was quantified as a percentage of the LV myocardium by adding the weighted scores and dividing by the total number of segments.

FDG-PET acquisition and analysis

Our FDG and Rubidium-82 PET protocol is summarised in online supplemental file 1. Images were acquired using the Siemens mCT flow system (Siemens Healthcare, Germany). A specialist (KW) blinded to patient data analysed attenuation-corrected inflammatory and perfusion PET images. Focal or focal-on-diffuse FDG uptake was considered compatible with active CS. The presence of RV FDG uptake was noted. The SUVmax of the myocardium was determined automatically. Active inflammation was defined as SUVmax ≥2.5.

Follow-up and endpoints

The clinical endpoint was the first occurrence of a major adverse cardiovascular event (MACE), consisting of all-cause mortality or serious VA. The latter was defined as appropriate device shock, resuscitated cardiac arrest with documented ventricular tachycardia (VT) or ventricular fibrillation (VF), or clinical presentation with symptomatic sustained VT lasting ≥30 s. Stored electrograms documenting arrhythmias were reviewed to confirm appropriateness of device therapy. Follow-up data were collected from electronic medical records. Time to event was calculated as the period between MDT diagnosis of CS and the first event.

Statistical analysis

Continuous variables were expressed as mean±SD when normally distributed or median with IQR. Differences in continuous variables between two groups were tested using Student’s t-test or Mann-Whitney U test accordingly. Categorical variables between two groups were compared using χ2 test. Univariable and multivariable Cox proportional hazard regression analysis was performed to assess parameters associated with MACE. Care was taken to avoid model overfitting and so, as an initial step, multivariate models were created within each imaging domain, to determine the strongest echocardiographic, CMR and FDG-PET parameter, respectively. A multivariate model including significant (p<0.05) imaging parameters from each domain was then devised. Variance inflation factor was used to ensure the absence of multicollinearity within models. Model discrimination was evaluated by Harrell’s C-index. MACE-free survival was depicted in Kaplan-Meier curves with subgroups compared using the log-rank test. For all tests, a p value <0.05 was considered statistically significant. Statistical analyses were conducted using Stata software V.18.0 (StataCorp).

Results

Study population

A total of 572 patients were referred for evaluation of suspected CS of whom 364 were excluded either because CS was ruled out (n=251); other significant structural heart disease was present (n=25), or echocardiography, CMR and FDG-PET imaging were not performed within 3 months (n=88). Therefore, the study population consisted of 208 patients with newly diagnosed CS. Of these, 108 patients had previously known extracardiac sarcoidosis and the remaining 98 presented with overt cardiac manifestations subsequently considered to be due to CS associated with previously undetected extracardiac sarcoidosis.

Baseline characteristics

The mean age of the study population was 55±13 years, of whom 67% were male, 81% were Caucasian, 37% had hypertension, 15% had diabetes mellitus and 35% were current smokers (table 1). Palpitations were reported in 56%, chest pain in 22% and syncope in 19% of patients. An abnormal ECG was noted in 53% of cases. The mean LVEF was 55±12.1% by echocardiography and 58±13% by CMR.

Table 1
|
Baseline demographic characteristics

Online supplemental table 1 lists the criteria fulfilled for CS diagnosis. Only two patients had endomyocardial biopsy-proven CS while 206 had biopsy-proven extracardiac sarcoidosis with diagnostic cardiac manifestations. CMR imaging demonstrated LGE in 93% of patients. FDG-PET imaging showed active inflammation in 73% and perfusion defects in 50% of patients. All patients had either LGE presence, FDG uptake or both.

Drug and device therapy

At initial presentation, 51% of patients were on steroid therapy, 38% on beta-blockers and 30% on ACE-I or ARB medication (table 1). Forty-four (21%) patients had an existing pacing device and a further 24 underwent device implantation following CS diagnosis. A total of 70 (34%) patients had dedicated device monitoring for VA consisting of 34 ICDs, 19 cardiac resynchronisation therapy—defibrillators (CRT-D), 15 permanent pacemakers (PPM) and 2 internal loop recorders. Three patients with PPM were upgraded to CRT-D and three were upgraded to ICD.

Events data

During a median follow-up of 45 (IQR 18–55) months, 42 (20%) patients experienced at least 1 MACE. The first events consisted of 14 deaths and 28 serious VA events (17 appropriate ICD shocks, 7 sustained VT requiring electrical cardioversion and 4 resuscitated cardiac arrest with documented VT/VF). Three patients who initially suffered VA subsequently died. Kaplan-Meier analysis of all patients with CS showed that death-free survival was 98% at 1 year and 88% at 5 years. MACE-free survival was 91% at 1 year and 74% at 5 years.

Demographic data and events

Among the demographic data, only non-Caucasian ethnicity (36% vs 15%, p=0.002) and an abnormal ECG (69% vs 49%, p=0.019) were more common in those with events (table 1).

Echocardiographic data and events

As shown in table 2, those with MACE had lower LVEF and higher LV dimensions and LAVi than those without MACE. The prevalence of RWMA was 2.5 times higher in those with events compared with those without events (67% vs 27%, respectively; p<0.001). Strain analysis was feasible in 187 patients. Mean GLS and GCS were reduced in the whole cohort at −14.5±4.4% and −16.7±5.9%, respectively, and significantly lower in those with MACE.

Table 2
|
Multimodality imaging characteristics

CMR data and events

All patients underwent CMR imaging; LGE extent quantification was feasible in 200 (96%) patients and inhibited by device artefact in 8 patients. Those with MACE had higher LV and RV end-systolic volumes, and LVMi than those without events (table 2). LVEF and RVEF were lower in MACE patients. RWMA was present in 69% and 37% of patients with and without events, respectively (p <0.001).

Greater LGE extent (19.4% vs 4.8%, p <0.001) was noted in those with events, and this group was more likely to have a multifocal rather than unifocal LGE distribution (p=0.011). Those with events were more likely to have LGE in the basal septum, basal and mid-anterior, mid-apical inferior and apical lateral regions (online supplemental table 2). RV LGE was more common in those with events.

FDG-PET data and events

All patients had evidence of either inflammation, perfusion defect or both. As shown in table 2, there was no significant difference in the proportion of patients with active inflammation or perfusion defects among the two groups. However, SUVmax was greater and the prevalence of RV FDG uptake was nearly fivefold higher in those with MACE (29% vs 6%; p <0.001).

Univariable analysis

As shown in table 3, non-Caucasian ethnicity, syncope and abnormal ECG had significant univariable associations with MACE. For the echocardiographic data, LVEF, RWMA, LAVi, GLS and GCS had significant associations. On CMR, LVEF, RWMA, LGE extent, RV LGE and RVEF were associated with MACE. For FDG-PET imaging, SUVmax and RV FDG uptake were associated with MACE.

Table 3
|
Univariable association with MACE

Multivariable analysis

Multivariable analysis within each imaging modality showed that RWMA and LAVi were the strongest echocardiographic predictors, LGE extent and RVEF the best CMR predictors and RV FDG uptake the most robust FDG-PET predictor of MACE (table 4). A primary multivariable model consisting of these five parameters was then created (table 4). In this model, TTE RWMA (HR 2.55, 95% CI 1.27 to 5.28; p=0.008), LGE extent (HR 1.02, 95% CI 1.00 to 1.04; p=0.018), RVEF (HR 0.97, 95% CI 0.94 to 0.99; p=0.039) and RV FDG uptake (HR 2.48, 95% CI 1.15 to 5.33; p=0.02) were independent predictors of MACE, whereas LAVi was not. Substitution of TTE RWMA with CMR RWMA resulted in a similar C-index (0.84 vs 0.85; p=0.548). Among the demographic characteristics, ethnicity was the strongest univariate predictor of MACE. When added to a model containing TTE RWMA, LGE extent, RVEF and RV FDG uptake, non-Caucasian ethnicity (HR 2.18, 95% CI 1.13 to 4.18; p=0.019) was also an independent predictor of MACE (χ2=55.21 and C-statistic=0.86).

Table 4
|
Multivariable Cox regression analysis for MACE

Risk stratification based on CMR RVEF, LGE extent, TTE RWMA and RV FDG uptake

In our patient population, LGE extent had a good ability to predict MACE (C-index=0.73, 95% CI 0.64 to 0.82). The optimal cut-off of LGE extent of ≥15% best predicted MACE, with sensitivity of 62% and specificity of 79%. The positive and negative predictive values were 43% and 89%, respectively. Kaplan-Meier survival curves for TTE RWMA, LGE extent, RVEF and RV FDG uptake are shown in online supplemental figure 1). Among the 208 study patients, we identified three risk categories based on TTE RWMA presence, CMR RVEF, LGE extent and RV FDG uptake presence. Patients at ‘high risk’ of developing MACE were all those with RVEF <50%, LGE extent ≥15% and presence of either TTE RWMA, RV FDG uptake or both. Patients at ‘low risk’ were all those with RVEF ≥50% and LGE extent <5% and absence of both TTE RWMA and RV FDG uptake. The ‘intermediate risk’ category consisted of those with either RVEF <50% and LGE extent <15%, or RVEF ≥50% and LGE extent ≥15%, irrespective of the TTE RWMA and RV FDG uptake findings. Based on this risk classification, a total of 21 (10%) patients were in the high risk, 122 (59%) in the intermediate risk and 65 (31%) in the low-risk categories. Survival curves for the respective subgroups are shown in figure 2. The event rate was 57% (28.7% per person year) within the high-risk category and 24% (6.9% per person year) within the intermediate-risk category. Only 2% (0.4% per person year) of patients in the low-risk category experienced a MACE during the follow-up period.

Figure 2
Figure 2

Risk stratification of patients with cardiac sarcoidosis (CS) using multimodality imaging to predict future all-cause mortality and ventricular arrhythmia (VA). The Kaplan-Meier curves reveal the frequency of all-cause mortality or VA-free survival according to low-risk, intermediate-risk and high-risk groups depending on the presence of cardiac magnetic resonance (CMR) right ventricular systolic dysfunction, extent of late gadolinium enhancement (LGE), presence of regional wall motion abnormality (RWMA) on transthoracic echocardiography (TTE) and presence of right ventricular (RV) 18fluorodeoxy glucose (FDG)-uptake on PET. Patients with CS in the low-risk, intermediate-risk and high-risk categories experienced an event rate of 2%, 24% and 57%, respectively.

Figure 3 depicts the stepwise incremental predictive value of TTE RWMA, LGE extent, RVEF and RV FDG uptake. Although the latter parameters were independent predictors of MACE, the addition of RV FDG uptake to the CMR data did not increase the C-index value, because of the overall low prevalence of RV FDG uptake in the study population.

Figure 3
Figure 3

Multimodality imaging predictors in cardiac sarcoidosis and stepwise incremental value to predict major adverse cardiovascular event (MACE). CMR, cardiac magnetic resonance; FDG, 18fluorodeoxy glucose; LGE, late gadolinium enhancement; RVEF, right ventricular ejection fraction; RWMA, regional wall motion abnormalities; TTE, transthoracic echocardiography.

Discussion

We have previously shown LVEF and SUVmax to be predictors of adverse events in a CS cohort, but a detailed analysis of the echocardiographic, CMR and FDG-PET data was not performed.15 In this study of a more recent dataset, we incorporated a detailed and comprehensive prognostic evaluation of demographic, clinical and multimodality imaging data in a large cohort of patients with CS. We performed a stepwise analysis, based on our clinical diagnostic algorithm, evaluating the echocardiographic, CMR and FDG-PET data in sequence. The most robust imaging markers of future risk were RWMA, LGE extent, RVEF and RV FDG uptake. The use of LGE extent and RVEF thresholds, in particular, enabled risk stratification into low-risk and high-risk categories. The 5-year survival rate of our population was 88%, similar to that reported in a recent registry.20 Despite advances in HF therapies, CS remains a disease process associated with significant adverse outcomes.

Regional wall motion abnormality

About one-third of our CS cohort had RWMA on the initial TTE, conferring a 2.5-fold increased risk of MACE. This suggests that TTE may provide useful prognostic information prior to CMR and FDG-PET imaging and serve as a triaging tool when access to advanced imaging is limited. Interestingly, LVEF was not a predictor of outcome once other imaging data were taken into account. The majority of patients with CS tend to have a focal, patchy distribution of active myocardial inflammation, fibrosis, infiltration or necrosis. These changes are usually not extensive enough to lead to a reduction in LVEF but may cause RWMA which are typically discrete, usually occur in a non-coronary distribution and are more readily detectable when there is subendocardial or transmural involvement of the myocardium.21 Although a small study has shown basal septal thinning on echocardiography to be an independent predictor of events,11 a systematic analysis of multiterritory RWMA has not previously been performed. Fibrotic changes confined to the subepicardium or midwall may not be associated with overt RWMA, but can be readily identified by LGE on CMR imaging, explaining the strength of LGE extent as a prognostic marker.

LGE extent

The presence of LGE per se is not predictive of outcome in CS as the vast majority of patients have LGE on CMR,13 but an assessment of LGE extent has major prognostic relevance. Crawford et al showed that in 51 patients with CS, LGE exceeding 6% of the LV mass was associated with a 75% sensitivity and 82% specificity for identifying those at risk of death or VA.4 Ise et al found in 43 patients with CS that LGE mass ≥20% correlated with death, VA, HF admission and lack of LVEF recovery following corticosteroids.9 LGE mass ≥22% provided a 75% PPV for death or VA in 59 patients with CS studied by Ekström et al.10 Our findings support these datasets showing that LGE extent >15% led to a significant step up in the incidence of death and VA beyond this threshold. On the basis of these findings, serious consideration should be given to integrating LGE extent data in selecting patients for primary prevention ICD therapy. Non-standardisation of LGE quantification methods across studies limits the ability to identify an accurate LGE extent cut-off point, but likely approximates to 15%–20% of the myocardium. Athwal et al presented a novel method of pathology-frequent LGE, independent of LGE extent, that identified patients with CS at high risk.22 Among the 504 patients with suspected CS, 32% had any LGE and 20% had subepicardial, septal, RV free wall or multifocal LGE. This pattern was associated with arrhythmia and HF hospitalisation while LGE extent was only associated with arrhythmia. Our study did not include an HF endpoint, but more detailed LGE phenotyping may have added to the predictive ability of CMR.

RV abnormalities

Previous studies have noted an association between RV FDG uptake and an adverse outcome in patients with CS. Blankstein et al studied 118 patients with suspected or confirmed (n=32) CS for a median of 1.5 years.5 In those with an abnormal FDG-PET scan, 15% had RV FDG uptake and this was associated with a fivefold higher rate of death and VT. Similarly, Tuominen et al identified RV FDG uptake as a marker of death, VT and LVEF decline in 137 patients with suspected CS.6 Finally, Bekki et al studied 44 patients with steroid-naïve CS and identified RV FDG uptake as a predictor of adverse events after adjusting for LVEF.7 However, none of these studies evaluated the prognostic value of RV FDG uptake in the context of CMR data. Our study adds to the literature by showing that RV FDG uptake was an independent predictor of outcome even after correcting for LGE extent. We hypothesise that RV involvement signifies a proarrhythmic substrate of disease which predisposes to scar and VA circuit formation. RV LGE has also been associated with increased arrhythmogenicity and death23 but was not an independent predictor of events in our study. The reason for the predictive value of RVEF is unclear as the prevalence of pulmonary hypertension or advanced lung disease in our study population was low. It may be hypothesised that the inflammatory milieu of systemic sarcoidosis might have a direct effect on RV myocardial function.

Relative values of multimodality imaging parameters

Our findings support the central role of CMR imaging in the risk stratification of patients with CS. The bulk of the prognostic information was provided by RWMA (whether TTE or CMR), LGE extent and RVEF. Although RV FDG uptake was present in only 11% of patients, when present it conferred an additional 2.5-fold increased risk of events. Those at the greatest risk demonstrated LGE extent ≥15%, RVEF <50% and at least one of TTE RWMA or RV FDG-uptake. Conversely, the low-risk category had limited LGE extent <5% and no other abnormalities, conferring a very good prognosis. While the addition of FDG-PET data increased the C-index, the increase was modest and not statistically significant. Nevertheless, most patients with CS undergo FDG-PET imaging to determine the presence of active inflammation as part of the diagnostic, treatment and disease-monitoring pathways. In this setting, analysis of RV inflammation could complement the prognostic data provided by the other imaging modalities.

Few studies have attempted to determine the complementary value of CMR and FDG-PET in predicting outcomes among patients with CS. A recent pooled analysis found that higher LGE extent and RV FDG uptake (but not LV FDG) to be predictive of MACE.24 Our study adds to the strength of this finding as we uniquely adjusted for other multimodal imaging parameters. Only one prior study has incorporated all three imaging modalities for the prediction of VA and death in CS.14 Hutt et al examined 234 patients with CS and found TTE LVEF and any scar on CMR or FDG-PET to be predictive of adverse outcomes. However, the multimodality nature of the study was limited as only 54% of patients had all three imaging modalities performed and only a limited number of imaging parameters were examined. In addition, LGE quantification was not performed.

Ethnicity

Among the demographic data, non-Caucasian ethnicity was associated with future adverse events. Previous studies have rarely addressed the prognostic significance of ethnicity in CS. One study found that in those with LGE, black patients had a higher risk of death or VT than Caucasians.4 Further research is required to fully understand the mechanisms that underpin any racial differences.

Limitations

This was a retrospective study performed in a single tertiary centre, so referral bias is likely. Due to the retrospective nature of the study, an evaluation of the impact of immunosuppressive and HF treatment regimens on outcome was not possible. However, all management decisions were guided by multidisciplinary CS team discussions. While only 42% of patients had device monitoring for VA, the majority underwent ambulatory Holter monitoring as per routine clinical practice. The cause of death was not always known but the endpoint of all-cause death tends to be less vulnerable to selection bias than cardiac death and allows comparison with previous studies. Finally, studies have found brain natriuretic peptide to be a prognostic marker of adverse events in CS.15 However, BNP was not systematically performed in our patient population and so could not be included in the analysis.

Conclusions

Multimodality imaging modalities consisting of echocardiography, CMR and FDG-PET play an important role in the diagnostic and prognostic evaluation of CS. In our CS population, TTE RWMA, LGE extent, RVEF and RV FDG uptake were robust independent predictors of clinical outcome. CMR imaging data using LGE extent and RVEF thresholds, in particular, plays a central role in identifying not only low-risk patients with a good clinical outcome, but also high-risk patients who require risk-modifying therapeutic interventions including immunosuppressive agents, HF medications and pacing device therapy.