528
Left ventricular dysfunction, not myocardial injury, drives use of cardioprotective medications in acute myocarditis: insights from machine learning

Funding Acknowledgements: U.S. National Center for Advancing Translational Sciences (NCATS) UL1TR000090, UL1TR000124

Background: Myocarditis may be self-limited or precede cardiomyopathy (DCM) and heart failure (HF). Medications such as angiotensin-converting-enzyme inhibitors (ACEI), angiotensin II receptor blockers (ARB), beta-blockers (BB), and mineralocorticoid antagonist (MRA) can mitigate adverse remodeling. Myocardial injury assessed by troponin (Tn) and LGE-CMR are associated with adverse remodeling and DCM/HF risk, but their impact on anti-remodeling drug prescription in myocarditis is unknown.

Purpose: We investigated imaging and clinical parameters in building a model to predict cardioprotective medication prescription for acute myocarditis.

Methods: The electronic health record at a large referral center was queried using International Classification of Disease codes to retrospectively identify patients without prior DCM/HF hospitalized for a first episode of acute myocarditis. Myocarditis was confirmed by requiring: 1) cardiac symptoms, 2) myocardial injury by Tn/EKG, and 3) coronary artery disease exclusion. Clinical and imaging data were recorded. Thirty features were evaluated by logistic regression (LR) and least absolute shrinkage and selection operator (LASSO) analyses.

Results: Of 1744 encounters screened, 199 patients were included (age 43 ± 16 years, 43% female; Table), with 32 deaths at follow-up. At discharge, new prescription of ACEI/ARB, BB, MRA, or any combination, were 27%, 48%, 8%, and 55%, respectively. Myocardial injury was evident by LGE in 91 of 139 CMR exams. Both LR and LASSO identified features related to LV size and systolic function as well as BNP elevation as the main predictors, rather than degree of Tn elevation or myocardial damage by LGE, of new cardioprotective medication prescription at discharge (accuracy = 0.87-0.95, recall = 0.72-0.85, f1 = 0.75-0.91). LASSO modeling including LV volumes, BNP elevation and T2 elevation strongly predicted mortality (accuracy = 0.95, recall = 0.66, f1 = 0.79).

Conclusions: In a large cohort of subjects affected by first episode of acute myocarditis, cardioprotective medications prescription appear mostly driven by evident adverse remodeling and HF rather than myocardial injury by blood or CMR biomarkers. Myocardial injury by T2 elevation and LGE warrant further investigation as a target for cardioprotective medication therapies.

Subjects with confirmed myocarditis

N	199
BMI Kg/m2	29.7 ± 7.4
HTN, N (%)	68 (34)
DM, N (%)	24 (12)
Smoking, N (%)	48 (24)