BACKGROUND: Heart rate variability (HRV), a non-invasive measurement of the sympathetic-vagal balance, has been demonstrated as a predictor of long-term survival in various patient populations. However, its predictive value in patients with end-stage renal disease (ESRD) has not been evaluated in a long-term follow-up study. METHODS: Prospective data collected for 41 patients with chronic hemodialysis (age 59 +/- 10 years, men 51.3%, diabetes mellitus 31%, and duration of dialysis 64 +/- 50 months) who underwent a 5-minute electrocardiogram (ECG) recording as a baseline for frequency domain HRV analysis. RESULTS: During a median follow-up of 150.2 months from 2003 to 2014, 15 (35.7%) patients died (3 due to cardiac causes and 12 due to non-cardiac causes). The Cox proportional hazards model suggested that the low frequency versus high frequency signal (LF/HF) of a high ratio for the HRV and diabetes mellitus were two independent predictors of mortality (hazard ratios 3.028 and 3.494; p = 0.033 and 0.022, respectively). Less reduction in MAP during dialysis showed borderline significance of long-term survival than those with larger drop (p = 0.058). CONCLUSION: A short ECG recording and an analysis of the frequency domain of the HRV is clinically predictive of the long-term survival of patients with chronic hemodialysis.