Associations between variability of glomerular filtration rate (GFR), death, and cardiovascular events have not been reported among patients with chronic kidney disease (CKD). In order to evaluate this, we retrospectively analyzed the risk of death and de novo heart failure as a function of variability in estimated GFR among a cohort of 3361 patients with stage 3 CKD. At baseline, patients with greater variability were younger, more likely to have diabetes, hypertension, and other comorbid conditions, and were more likely to have proteinuria and higher estimated GFR. In multivariate-adjusted Cox proportional hazard models over a median follow-up of 3.9 years, the risk of death associated with the highest relative to the lowest quartile of variability was 1.40 (95% confidence interval 1.05-1.87); there was no association with new-onset heart failure. The mortality association was independent of serum albumin, proteinuria, baseline estimated GFR, and the slope of the estimated GFR. Thus, variability in estimated GFR predicts death among patients with stage 3 CKD independent of previously reported risk factors. The prognostic utility of complementing existing risk stratification metrics with dynamic changes in GFR among patients with CKD warrants investigation.