The Diagnostic and Prognostic Utility of Contemporary Cardiac Magnetic Resonance in Suspected Acute Myocarditis

Cardiovascular magnetic resonance (CMR) is used to investigate suspected acute myocarditis, however most supporting data is retrospective and few studies have included parametric mapping. We aimed to investigate the utility of contemporary multiparametric CMR in a large prospective cohort of patients with suspected acute myocarditis, the impact of real-world variations in practice, the relationship between clinical characteristics and CMR findings and factors predicting outcome. 540 consecutive patients we recruited. The 113 patients diagnosed with myocarditis on CMR performed within 40 days of presentation were followed-up for 674 (504–915) days. 39 patients underwent follow-up CMR at 189 (166–209) days. CMR provided a positive diagnosis in 72% of patients, including myocarditis (40%) and myocardial infarction (11%). In multivariable analysis, male sex and shorter presentation-to-scan interval were associated with a diagnosis of myocarditis. Presentation with heart failure (HF) was associated with lower left ventricular ejection fraction (LVEF), higher LGE burden and higher extracellular volume fraction. Lower baseline LVEF predicted follow-up LV dysfunction. Multiparametric CMR has a high diagnostic yield in suspected acute myocarditis. CMR should be performed early and include parametric mapping. Patients presenting with HF and reduced LVEF require closer follow-up while those with normal CMR may not require it.


Introduction
Cardiovascular magnetic resonance (CMR) is increasingly used to investigate patients with suspected acute myocarditis [1]. However, the majority of data supporting the use of CMR for this indication is retrospective, and few studies have included contemporary parametric mapping techniques [2][3][4][5][6][7]. As a consequence, while CMR is included in guidelines and position statements regarding the investigation of suspected myocarditis, the Class of Recommendation varies and it is with Level of Evidence C, i.e., "consensus of opinion of the experts and or small studies, retrospective studies, registries" [1,8,9].
This study aimed to investigate the clinical utility of CMR in a large prospective cohort of patients with suspected acute myocarditis. The study also aimed to evaluate the impact that real-world variations in practice have on CMR findings, the relationship between clinical characteristics and CMR findings and factors that predict outcome.

Study Design and Participants
This was a prospective longitudinal cohort research study approved by the North West-Greater Manchester West Research Ethics Committee of the UK National Research Ethics Service. Written informed consent was obtained from all participants. The work was conducted according to the Helsinki Declaration.
Consecutive consenting adult patients undergoing clinical CMR for suspected acute myocarditis at Manchester University NHS Foundation Trust, UK, between 1 January 2015 and 31 December 2018, were prospectively recruited. The clinical suspicion of myocarditis was determined from the CMR referral information.
Patients undergoing CMR within 40 days of hospital presentation and without an alternative diagnosis were included in analyses investigating the relationship between baseline factors, CMR findings and outcome. A 40-day cut-off was chosen because this is in keeping with previous studies, and because the study aimed to investigate the impact of a real-world presentation-to-scan interval, which, in the authors' experience, is up to around 6 weeks, even in a healthcare system relatively well served by CMR [10]. Patients diagnosed with acute myocarditis on baseline CMR and who underwent follow-up CMR were included in an additional analysis.

Study Procedures
Data were managed using Research Electronic Data Capture (REDCap) [11]. Acute presentation and baseline comorbidity data were determined from primary and secondary care medical records.

CMR Analysis
CMR image analysis was performed using CVI42 (Circle Cardiovascular Imaging, Calgary, AB, Canada) according to current guidelines [12]. Mean relaxation times and their standard deviation are reported, the latter as an indication of heterogeneity, calculated after dividing regions of interest into 100 radii. Extracellular volume (ECV) was calculated as described previously [13]. Foci of non-ischaemic scar on LGE imaging were included in ECV measurements [14]. LGE was manually quantified from a LV short axis LGE stack [15]. CMR analysis was performed blinded to patient outcomes.

Study Outcomes
CMR diagnosis was made by the clinical reporting physician (Level 3 trained), independent of the research team. The primary endpoint for the outcome analysis was a composite of first hospitalisation for heart failure (HF) after CMR or all-cause mortality. First hospitalisation for HF was recorded from primary and secondary care medical records and determined independently by the clinical team responsible for the patient's care. Mortality status was ascertained from primary and secondary care medical records.

Statistical Analysis
Data are summarised using mean and standard deviation or median and interquartile range (IQR), and were compared using t tests or non-parametric equivalents, as appropriate. χ 2 tests were used to compare categorical variables. Logistic regression models (univariable and backward stepwise multivariable) were used to investigate the relationship between baseline clinical features and a diagnosis of myocarditis on CMR. Linear regression models (univariable and backward stepwise multivariable) were used to investigate the relationship between baseline factors and change in LV ejection fraction on follow-up CMR. The low numbers of patients that experienced the primary endpoint precluded meaningful outcome analysis, but exploratory univariable Cox regression analysis to evaluate the relationship between baseline factors and outcome was performed. Analyses were performed using SPSS (version 22, IBM, Armonk, NY, USA).

Factors Associated with LV Functional Recovery Following Acute Myocarditis
Thirty-nine patients found to have myocarditis on baseline CMR underwent followup CMR at a median 189 (166-209) days following hospital presentation. Left and right ventricular ejection fraction significantly increased and LV mass indexed for body surface area significantly reduced at follow-up compared to baseline (Table 3). Similarly, LGE burden, myocardial T1, T2 and ECV all significantly decreased. Representative parametric mapping is displayed in Figure 2. In univariable analysis, higher baseline LV ejection fraction, lower baseline nonischaemic LGE burden and lower baseline ECV were associated with higher LV ejection fraction at follow-up CMR (Table 4). In multivariable analysis, only baseline LV ejection fraction was independently associated with follow-up LV ejection fraction. Data presented as mean ± standard deviation or median (interquartile range) depending on distribution. * n = 36 for baseline scan and n = 34 for follow up scan. Abbreviations as per Tables 1 and 2.

Factors Associated with Clinical Outcome Following Acute Myocarditis
The 113 patients diagnosed with myocarditis on CMR performed within 40 days of hospital presentation were followed up for a median of 674 (504-915) days. Four (3.5%) patients experienced the primary endpoint: two patients died (1 due to post-myocarditis dilated cardiomyopathy and 1 due to sepsis following non-cardiac surgery) and two patients were hospitalised for HF. The low numbers of patients that experienced the primary endpoint precluded meaningful outcome analysis. In exploratory univariable analysis, lower baseline LV ejection fraction (hazard ratio (HR) 0.87; 95% confidence interval (CI) 0.80-0.94) was associated with the primary endpoint. Presentation with chest pain was associated with a lower rate of the primary endpoint (HR 0.94; 95%CI 0.01-0.91). No patient with normal findings on CMR performed within 40 days of hospital presentation died or were hospitalised for HF during follow-up.

Discussion
This study is the largest prospective evaluation of the clinical utility of CMR in unselected patients referred with suspected acute myocarditis, and the first to include contemporary CMR techniques.
The positive diagnostic yield in our study (72%) is higher than that in earlier studies (e.g., 52% in Biesbroek et al. [16]; 44% in Schumm et al. [17], likely in part reflecting the added diagnostic value of parametric mapping, which permits detection of myocardial oedema in the absence of necrosis and diffuse myocardial injury more effectively than conventional T2-weighted and LGE imaging [18]. Nevertheless, a substantial proportion of patients with suspected myocarditis have normal findings at CMR (28% in this study) [16,17]. Whilst the sensitivity of CMR appears to reduce over time from presentation (see below), there remains a significant number of patients for whom CMR is normal despite it being performed early. For example, Stensaeth et al. found 18% of patients had normal CMR findings despite the CMR being performed within 24 h of admission [19]. Coronary artery spasm, plaque disruption or thromboembolism with too little myonecrosis to be identified using the spatial resolution of CMR and 'non-cardiac' conditions have been proposed as possible mechanisms. Further investigation is required to understand the pathophysiology responsible for the clinical presentation in this group.
Being male and there being a shorter interval between hospital presentation and CMR were independently associated with a diagnosis of myocarditis on CMR. Population studies show myocarditis is more common in males [20]. The mechanism is unclear but preclinical studies have proven that males develop a greater myocardial inflammatory response to Coxsackievirus B infection [21]. Previous CMR studies in suspected myocarditis have shown non-ischaemic LGE to be more common in males [17,22].
Smaller studies have shown that the sensitivity of CMR for detecting myocarditis is higher when performed within 2 weeks of symptom onset compared to beyond 2 weeks, although these studies have generally not corrected for potential confounding factors such as sex [23,24]. Our study advances these earlier findings, demonstrating that shorter presentation-to-scan interval is independently associated with a CMR diagnosis of myocarditis. Indeed, shorter presentation-to-scan interval was associated with multiple metrics of myocardial injury severity, likely reflecting the time course of myocardial injury resolution [25].
This study is the largest to assess the relationship between clinical presentation and myocardial injury. Presentation with chest pain was associated with less myocardial injury and better LV function whereas HF symptoms were associated with greater myocardial injury (higher ECV, more non-ischaemic LGE) and lower LV function. These findings provide mechanistic insight into clinical myocarditis studies, which have shown that people presenting with chest pain have a better prognosis than those presenting with HF [26,27]. The relationship between clinical presentation and CMR findings in previous smaller CMR studies has been conflicting [17,28].
Early studies found endomyocardial biopsy proven-myocarditis to be associated with poor prognosis [29]. The outcome following myocarditis in our study (adverse event rate approximately 2% per year) is much more favourable and is in keeping with other contemporary studies of CMR-diagnosed myocarditis. Indeed, the low event rate precluded multivariable Cox regression analysis. The prognostic value of CMR parameters has varied across previous studies, although lower baseline LV ejection fraction and higher baseline LGE burden are factors that more commonly associate with adverse outcome [17,22,28]. In keeping with these findings, baseline LV ejection fraction was independently associated with follow-up LV ejection fraction in our study. Higher baseline LGE burden and ECV were associated with less LV recovery on univariable analysis, but the associations were no longer present after multivariable adjustment, albeit in the setting of limited statistical power. Further investigation is required.
Putting the findings of this study together with those of other studies, CMR for suspected myocarditis should be performed as early as possible following admission and include parametric mapping. Patients presenting with HF symptoms, demonstrating reduced LV ejection fraction and higher LGE burden require close clinical and imaging follow-up. Those with a normal scan may not require routine follow-up.
Our study provides prospective evidence from a large cohort to add to the already extensive, albeit largely retrospective, observational data to support the use of CMR for suspected acute myocarditis. A positive diagnostic yield of 3 in 4 patients is remarkable by any standard. Indeed, the strength of the observational data means that a higher level of evidence may preclude the field from achieving a higher Level of Evidence; a randomised trial comparing the utility of CMR with other imaging modalities for the diagnosis of, and prognosis following, acute myocarditis would be considered unethical. Endomyocardial biopsy is rarely performed and its own accuracy is limited, thus it is an inadequate comparator. Nevertheless, the consequence of not having randomised data is that CMR may remain as being recommended with only a "moderate level of consensus" and below that of other imaging modalities [8]. The advent of a therapy for myocarditis requiring a positive diagnosis for patient selection, and quantitative monitoring of myocardial response, may drive stronger recommendations. For example, entry into the ongoing trial of an IL-1 receptor antagonist in myocarditis (NCT03018834) requires "myocarditis proven by MRI".

Conclusions
Contemporary multiparametric CMR has a high diagnostic yield in patients with suspected acute myocarditis. Being male and there being a shorter interval between hospital presentation and CMR are associated with a diagnosis of myocarditis on CMR. Presentation with chest pain is associated with less myocardial injury and better LV function than presentation with symptoms of HF. Clinical outcome following CMR-diagnosed myocarditis is favourable. Lower baseline LV ejection fraction predicted less recovery of LV function.

Limitations
Patients did not undergo endomyocardial biopsy, however it is rarely performed in most centres in patients with suspected myocarditis, and due to sampling error, transiency of myocardial injury and variation in histology interpretation, it is limited as a reference standard [30]. It is recognised that using myocardial injury may have lessened or resolved in the 40-day window between presentation and CMR, however this time frame was chosen because it is in keeping with real-world clinical practice, and understanding the impact that presentation-to-scan interval has on diagnostic yield was an important aim of the study. The survival analysis was limited by a low number of adverse events, as discussed. Single centre data may not generalise, but our findings align with prior literature.  Informed Consent Statement: Written informed consent was obtained from all participants.

Data Availability Statement:
The datasets generated and/or analysed during this study are available from the corresponding author on reasonable request.