Variability of ventricular rate was quantified by two measures of heart rate variability: the SD of the mean R-R interval (SDNN) and the SD of the 5-minute mean R-R interval (SDANN). In 35 patients with atrial fibrillation and advanced heart failure (left ventricular ejection fraction 20% +/- 9%, cardiac index 2.4 +/- 0.7 L/min/m2), SDNN and SDANN were compared to 13 preselected clinical and hemodynamic variables for prediction of outcome. During a 12-month follow-up period, 8 (23%) patients deteriorated clinically; 3 (9%) died, and 5 (14%) underwent heart transplantation. SDNN and SDANN correlated to the difference of the mean R-R interval between night (2 AM to 3 PM) and day (11 AM to noon) with r values of 0.62 and 0.77, respectively. From 15 preselected variables, only SDANN (chi 2 = 6.7, p = 0.01) was independently associated with survival on multivariate analysis. Dichotomized SDANN at 100 msec accurately predicted 12-month survival in 28 (80%) patients (relative risk = 9.77, p = 0.001). In conclusion, analysis of heart rate variability is useful in quantifying diurnal variation of ventricular rate in atrial fibrillation and might be useful in predicting survival in patients with advanced heart failure.