Heart failure and predictors of outcome in eosinophilic myocarditis: a real-world cohort study

Objective In eosinophilic diseases, myocardial involvement critically impacts prognosis. Given the rarity and heterogeneity of eosinophilic myocarditis (EM), diagnosis and risk-adapted treatment remain challenging, and longitudinal data on cardiac function are scarce. Therefore, this study investigated the course of cardiac function and aimed to identify prognostic biomarkers in a real-world EM cohort across inflammatory disorders. Methods Clinical, laboratory, and imaging data of all patients with eosinophilic myocarditis treated at Ludwig-Maximilians-University Hospital in Munich between January 2010 and May 2025 were reviewed. Descriptive and inferential statistics were applied to delineate disease courses, treatment regimens, and outcomes. Results Of 419 patients screened, 51 had EM (51% female patients, mean age 50 ± 14 years). Underlying diseases included eosinophilic granulomatosis with polyangiitis (EGPA, n = 29), idiopathic hypereosinophilic syndrome (i-HES, n = 16), drug-associated/hypersensitivity myocarditis ( n = 7; 4 with EGPA/i-HES), and idiopathic EM ( n = 3). Mean left-ventricular ejection fraction (LVEF) was impaired at baseline (48 ± 17%) and after a median of 34 months [interquartile range 16–79] ( Δ + 3 ± 12%, p = 0.57). Laboratory parameters improved ( p < 0.001); however, LVEF recovered predominantly in patients with baseline evidence of active inflammation on cardiac magnetic resonance (CMR) or endomyocardial biopsy (EMB) (41 ± 15% to 48 ± 10%; p < 0.05). Median left-ventricular end-systolic/end-diastolic volume indices (LVESVI/-EDVI) increased over time (50 [28–65] to 56 [33–84] and 85 [68–98] to 97 [79–132] mL/m 2 , respectively; p < 0.05), suggesting cardiac remodeling. Glucocorticoids were broadly applied, whereas steroid-sparing drugs were used less frequently. Five patients died, all had reduced baseline LVEF. Age- and sex-adjusted baseline NT-proBNP (log 10 ), LVEF, and the Sartorelli score, a published composite score (incorporating LV dysfunction, arrhythmias, CMR, and serum troponin), were significantly associated with follow-up LVEF, LVEDVI and LVESVI (rho/ r = ±0.44–0.72, R 2 = 0.19–0.64; p < 0.05–<0.0001). Conclusion In this real-world cohort, EM severely impacted multiple domains of cardiac function. Treatment patterns were heterogeneous. Baseline NT-proBNP, LVEF, and the Sartorelli score were associated with long-term LVEF and LV remodeling. Baseline evidence of active myocardial inflammation on CMR and/or EMB was associated with subsequent LVEF improvement. Together, these findings support the utility of multimodal biomarkers for risk stratification in EM but require external validation before informing routine care. Prospective multicenter studies using standardized diagnostics and treatment regimens are needed to translate these findings into biomarker-guided management of eosinophilic cardiomyopathy.