Cardiac magnetic resonance in histologically proven eosinophilic myocarditis

The digital pathology reports of Helsinki University Hospital spanning the time from January 2000 to September 2019 (n = 268,341) were screened for potential cases of EM. A total of 497 reports of myocardial histology included findings considered diagnostic of myocarditis or showing inflammatory changes or eosinophilic infiltration. Of these, reports of 34 different cases were identified as likely representing EM. Their histologic specimens were retrieved from Helsinki Biobank and re-analyzed in detail. Ultimately, 28 cases of histologically proven EM were ascertained based on microscopy of myocardial samples taken on EMB (n = 25) or at autopsy (n = 3) showing myocyte damage, interstitial edema, and abundant degranulated and/or intact eosinophils with or without areas of necrosis and/or fibrosis (Fig. 1). Among the 28 histologically proven cases, 27 patients had been hospitalized and one died prior to admission and was diagnosed at autopsy. Fifteen had undergone early CMR for diagnosis. The CMR studies were re-evaluated for the present work, and the hospital charts of all 27 patients were reviewed for clinical data from presentation to end of follow-up in May 2020.

The CMR studies were performed with 1.5T or 3T scanners (Avanto, Avanto Fit, Verio and Sonata; Siemens, Erlangen, Germany) at or shortly after presentation using phased-array receiver coils and standard protocols according to contemporary hospital routines [17]. Not all CMR techniques were available in all studies done over the 2-decade study period.

To assess left ventricular (LV) and right ventricular (RV) volumes and ejection fraction (EF), breath-hold cine studies were done using electrocardiographically gated steady‐state free-precession. Cine images were obtained in long-axis (2-, 3- and 4-chamber view) and short-axis planes covering both ventricles (typical slice thickness, 6–8 mm; and interslice gap, 20%). One patient had a temporary pacemaker with active fixation pacing lead during scanning [18].

LGE images were obtained 10 to 15 min after intravenous injection of contrast agent (Dotarem®, Guerbet, Aulnay-sous-Bois, France; 0.15 mmol/kg) using inversion‐recovery gradient echo technique in views identical to the cine studies (inversion time 240–360 ms). First-pass rest perfusion imaging was performed in three short-axis slices (basal, mid, and apical) and an apical four-chamber view. T2-weighted (T2W) fat saturation images/short tau inversion recovery images were regularly obtained. Quantitative myocardial T1 mapping was performed in 1–3 short-axis slices (basal, mid, and apical) using a shortened Modified Look-Locker Inversion-recovery (MOLLI) sequence before and 15 min after gadolinium injection. Hematocrit from the same day was used for extracellular volume fraction measurements. T2 mapping was performed in a breath-hold fashion using three T2-prepared balanced steady state free precession images (T2-preparation times 0 ms, 25 ms, and 55 ms) before gadolinium injection in views identical to T1 mapping.

Images were analyzed by a single CMR-trained cardiologist (P.P.) blinded to clinical data and read for consensus in a virtual meeting with another experienced CMR cardiologist (C.S.). Image analysis was performed using QMass MR software® (version 8.1, Medis Medical Imaging Systems, Leiden, the Netherlands). Ventricular volumes, mass and EF were evaluated using standard protocols [19] and compared to UK biobank reference values [20], with abnormal values in women (men) as LV end-diastolic volume > 101 (117) ml/m², LV mass > 55 (72) g/m², LVEF < 50 (47) %, RV end-diastolic volume > 110 (128) ml/m², RVEF < 45 (40) %. Pericardial effusion was defined as > 5 mm pericardial space anterior to the RV wall [21]. Presence of LGE was identified visually. The number of LV segments with LGE was counted according to the AHA 17-segment model [22]. LGE pattern in each segment was classified as subendocardial (including LGE on the RV side of the septum), mid-wall/subepicardial (myocarditis-like LGE) [23], or transmural. Wide-spread subendocardial LGE was defined as ≥ 4 adjacent segments [24]. The extent of LGE as a percentage of LV mass was assessed using the full width at half-maximum method [25]. Multifocality of LGE was defined as more than one discrete lesion within the LV. Insertion site LGE was not counted in multifocality unless clearly continuing into the LV. Finally, in each case, the congruence of the composite CMR findings with the 2018 Lake Louise Criteria for myocarditis [4] was assessed.

Continuous variables are presented as mean ± standard deviation for normally distributed data and median (interquartile range) for skewed data, respectively. Categorical variables are presented as frequency (%). Group comparisons were performed with Student’s t test, Mann–Whitney U test, Chi-square test with continuity correction, or Fisher’s exact test, as appropriate. A p-value of < 0.05 was considered statistically significant and all tests were 2-sided. Statistical analysis was performed on R [RStudio, version 4.1.2, The R Foundation; (https://​www.​r-project.​org/​)].

Table 1 shows the presenting clinical characteristics and the key laboratory and echocardiographic observations in all 27 patients with EM comparing cases with and without CMR. The study population had a mean age of 52 years. Altogether 14 of the 15 patients undergoing CMR were hospitalized and diagnosed after 2010 while most cases without CMR (8 of 12) presented before 2010 (p = 0.003). The histologic diagnosis of EM was obtained by EMB in all except 2 cases that were detected at autopsy.

The presenting signs and symptoms of EM included fever (53% in patients with CMR vs 58% in those without), dyspnea (47% vs 58%), chest pain (53% vs 25%), overt heart failure (33% vs 58%), symptomatic 3rd degree atrio-ventricular block (20% vs 17%) and ventricular tachycardia (7% vs 8%). As shown in Table 1, hypersensitivity reaction was the most common etiology, its causes comprising exposure to drugs including mesalazine (4 cases), amoxicillin, cephalexin, doxycycline, and gefitinib, with in vitro fertilization and alcohol suspected in the remainder of the cases.

On laboratory examinations, blood eosinophil count was abnormally high (> 0.44 × 10⁹/L) in 60% of patients on admission and in 80% during the entire hospitalization. Circulating cardiac troponins were abnormally elevated in all patients, natriuretic peptides being elevated in 82%. A total of 64% of patients had ST segment or T wave abnormalities on 12-lead electrocardiogram. In general, as Table 1 shows, patients with and without CMR for diagnosis had similar characteristics at presentation except that severe LV dysfunction on echocardiography was less prevalent in patients undergoing CMR.

All studies were performed between March 2004 and August 2019. The median delay from hospital admission to CMR was 1 (range: 1–7) day. The heart rate during cine and LGE imaging averaged 92 ± 20 beats/min and 85 ± 19 beats/min, respectively. Figure 2 is a collage of typical CMR findings in EM, and Table 2 summarizes the CMR data comparing patients with (n = 6) and without (n = 9) overt myocardial necrosis by histopathology.

Table 3 summarizes the treatment and outcome data for all EM patients and compares the subgroups with and without CMR for diagnosis. The data highlight the clinical seriousness of EM in showing that as many as 13 out of 27 patients (48%) needed circulatory or mechanical cardiac support. Among the 15 patients undergoing CMR, the 7 needing circulatory support, compared to the rest 8 patients, had lower LV ejection fraction (32 ± 14% vs 48 ± 11%, p = 0.024), higher maximal LV wall thickness [14 (13–19) mm vs 11 (11–13) mm, p = 0.043], lower RV ejection fraction (35 ± 14% vs 49 ± 7.5%, p = 0.037) and trends toward higher LV mass [83 (72–99) g vs 69 (58–73) g, p = 0.072] and higher LGE mass [28 (22–35) % vs 14 (9.1–18) %, p = 0.101]. Altogether 8 patients (30%) either suffered cardiac death (n = 5), underwent transplantation (n = 1), or experienced life-threatening ventricular arrhythmia (n = 2), of which 4 cardiac deaths in patients on inotropic or mechanical support and 1 ventricular fibrillation occurred during hospitalization.

The main strength of our study is its design—a systematic analysis of CMRs in a 20-year cohort of patients cared in an academic hospital due to histologically proven EM. Although our findings likely are less influenced by selection bias than solitary case reports, comparison of cases with and without CMR (Table 1) suggest that patients with poorest LV function and most severe cardiovascular compromise could not undergo CMR imaging. Thus, some of the sickest patients with EM may have been missed for this reason. The median time from symptom onset to hospitalization was only few days in this series. More chronic forms of EM may not fulfill all of the Lake Louise Criteria for myocarditis. Our patients did not receive anti-interleukin-5 therapies. Other limitations of our work include its retrospective design, the small number of patients, lack of follow-up CMR examinations, and lack of control groups of other types of fulminant or non-fulminant myocarditis. LGE with long inversion time, optimized for the detection of small intra-cavitary thrombi, was not used in this study.