Clinical predictors for ventricular arrhythmia and heart failure hospitalisation in cardiac sarcoidosis

Introduction Ventricular arrhythmia (VA) and heart failure (HF) hospitalisation are major adverse outcomes in patients with cardiac sarcoidosis (CS). We aimed to identify clinical, laboratory and imaging predictors on the outcomes of these patients. Methods 572 patients with a definite or probable diagnosis of CS according to the 2014 Heart Rhythm Society criteria were enrolled from 6 tertiary centres in Europe and North America. A multi-step statistical approach was employed to identify independent predictors of the outcomes (VA, n = 557;HF hospitalisation, n = 571). A model including 21 clinical, laboratory, and imaging variables was analyzed using categorical regression with LASSO (Least Absolute Shrinkage and Selection Operator) regularization to perform feature selection and minimize overfitting. Following this initial screening and including clinically relevant parameters, a refined multivariate model was constructed. For sustained VA, candidate predictors included renal function (eGFR), de/repolarization abnormalities, sustained VA at presentation, age at first presentation, sex, late gadolinium enhancement (LGE) burden and distribution (LV or RV), and renal-adjusted biomarkers (high-sensitivity troponin T and natriuretic peptides). For HF hospitalisation, predictors included LVEF%, eGFR, renal-adjusted natriuretic peptides, LGE burden, sustained VA at presentation, age, sex, and use of TNF-α therapy. Results Median age was 54 [46-61] years, 43.5% were female. Forty-eight patients (9%) suffered from sustained VA and n=251 (44%) from HF hospitalisation (median follow-up 4.7 [2.2-7.8]years). An ICD was placed within 3 months from first presentation in n=189 (33%) and a pacemaker in n=47 (8%). For sustained VA, the model showed modest explanatory power (R²=0.081; adjusted R²=0.048; p=0.001). Younger age at first presentation was the strongest independent predictor (p=0.003). LV-LGE was also independently associated with VA risk (p=0.001). Total LGE burden showed only a statistical trend (p=0.071). Renal function, hsTnT, RV LGE, de/repolarization abnormalities, and sex were not independently associated after adjustment. Overall, myocardial tissue characteristics were more informative for VA risk stratification than electrocardiographic measures. For HF hospitalisation (n=251, 44%) the final model demonstrated modest discrimination (R²=0.103; adjusted R²=0.078, p<0.001). Independent predictors of HF hospitalisation were use of TNF-α inhibitors (p<0.001), sustained VA at presentation (p<0.001), impaired renal function (p=0.006), lower LVEF (p=0.036), and male sex (p=0.024). Conclusions In patients with CS, younger age and LV-LGE were independent predictors for VA whereas risk factors for HF hospitalisation included the use of TNF-α therapy, sustained VA at presentation, renal dysfunction, low LVEF and male sex. A multi-faceted approach is essential for optimal management of CS patients.