Objectives: To explore the association of different types of blood pressure (BP) variability measures estimated from either short-term ambulatory reading-to-reading or long-term clinic visit-to-visit BP records with long-term survival in an elderly treated hypertensive population. Methods: A subset of patients (n = 508) aged at least 65-years was studied from the Second Australian National Blood Pressure study. We estimated SBP and DBP BP variability as the SD of ambulatory (24-h, daytime, night-time) and clinic visit-to-visit BP directly from all corresponding on-treatment within-individual BP records. Ambulatory 'weighted day-night' variability was calculated as a weighted mean of daytime and night-time SD. Cox-proportional hazard models adjusted for baseline risk factors (Model 1) and corresponding on-treatment BP (Model 2) or average night-time SBP (best predictive BP measure for outcome) (Model 3) were used to determine the relationship between long-term outcome and BP variability. Results: Over a median of 10.6 years, 101 patients died from any cause, of which 51 deaths were cardiovascular. We observed increase in 'daytime' and 'weighted day-night' SBP/DBP variability was significantly associated with increased all-cause mortality in all models. For cardiovascular mortality, only 'weighted day-night' SBP variability significantly predicted risk in all models (Model 3 hazard ratio: 1.09, 95% confidence interval: 1.00-1.19, P = 0.04). Long-term BP variability was not associated with any outcome. On direct comparison, both 'daytime' and 'weighted day-night' BP variability measures provided similar prognostic information. Conclusion: Short-term 'daytime' and 'weighted day-night' SBP variability from ambulatory BP recordings was a better predictor of mortality in elderly treated hypertensive patients than long-term BP variability from visit-to-visit BP recordings.