Abstract 17819: Predicting the Proportional Risk of Sudden Cardiac Death in a Multicenter Heart Failure Cohort

Introduction: Many patients with heart failure die suddenly. Implantable cardiac defibrillators (ICDs) can reduce the rate of sudden death. Strategies that help identify patients who have most to gain from an ICD could reduce morbidity and maximize cost-effectiveness of ICDs. Methods: We used data from 9,885 heart failure patients, of whom 2,549 died during an average follow-up of 2.3 years. Using commonly available baseline clinical and demographic information, we developed a mode-specific multinomial logistic regression model to identify variables associated with a disproportionate fraction of total mortality due to sudden death. Results: In multivariable-adjusted models, we confirmed that commonly utilized variables for ICD implantation decisions, such as lower ejection fraction (OR for sudden death vs. other deaths: 1.15 per 10% decrease, p=0.005) and better New York Heart Association functional class (OR 1.30 for class 2 versus 3/4, p=0.004) were associated with greater proportion of deaths due to sudden death. We also demonstrated that younger age (OR 1.25 per decade younger, p<0.001), male sex (OR 1.40, p=0.002), and higher body mass index (BMI) (OR 1.24 per 5 kg/m 2 , p<0.001) independently increased the chance that if a patient died, it would be due to sudden death. Patients with diabetes (OR 0.75, p=0.001), systolic BP deviating above or below 140mmHg (OR 0.92 per 10mmHg, p=0.005), higher creatinine (OR 0.65 per mg/dL, p<0.001), or lower serum sodium (OR 0.96 per mEq below 138, p=0.029) had a lower proportional risk of sudden death. The use of several heart failure medications, left ventricular end-diastolic dimension, or plasma NT-proBNP levels was not independently associated with risk of sudden versus non-sudden death. Conclusions: Several demographic and clinical variables beyond ejection fraction and functional class are independently associated with higher or lower proportional risk for sudden death. Our findings support the need for further investigation whether this novel predictive model, in the context of absolute mortality rates, can be used to target the use of ICDs and other life-saving therapies to populations who will derive greatest benefit from their application.