BACKGROUND: Determining a specific death cause may facilitate individualized therapy in patients with heart failure (HF). Cardiac resynchronization therapy (CRT) decreased mortality in the Cardiac Resynchronization in Heart Failure trial by reducing pump failure and sudden cardiac death (SCD). This study analyzes predictors of specific causes of death. METHODS AND RESULTS: Univariate and multivariate analyses used 8 baseline and 3-month post-randomization variables to predict pump failure and SCD (categorized as "definite," "probable," and "possible"). Of 255 deaths, 197 were cardiovascular. There were 71 SCDs with a risk reduction by CRT of 0.47 (95% confidence interval 0.29-0.76; P = .002) with similar reductions in SCD classified as definite, probable, and possible. Univariate SCD predictors were 3-month HF status (mitral regurgitation [MR] severity, plasma brain natriuretic peptide [BNP], end-diastolic volume, and systolic blood pressure), whereas randomization to CRT decreased risk. Multivariate SCD predictors were randomization to CRT 0.56 (0.53-0.96, P = .035) and 3-month MR severity 1.82 (1.77-2.60, P = .0012). Univariate pump failure death predictors related to baseline HF state (quality of life score, interventricular mechanical delay, end-diastolic volume, plasma BNP, MR severity, and systolic pressure), whereas randomization to CRT and nonischemic cardiomyopathy decreased risk; multivariate predictors of pump failure death were baseline plasma BNP and systolic pressure and randomization to CRT. CONCLUSION: CRT decreased SCD in patients with systolic HF and ventricular dyssynchrony. SCD risk was increased with increased severity of MR (including the 3-month value for MR as a time-dependent covariate) and reduced by randomization to CRT. HF death was increased related to the level of systolic blood pressure, log BNP, and randomization to CRT. These results emphasize the importance and interdependence of HF severity to mortality from pump failure and SCD.