BACKGROUND: Impaired cerebral blood flow has been associated with an increased risk of falls. Mean arterial pressure (MAP) and variability in MAP have been reported to affect cerebral blood flow but their relationships to the risk of falls have not previously been reported. METHODS: Utilising data from the Aspirin in Reducing Events in the Elderly trial participants, we estimated MAP and variability in MAP, defined as within-individual SD of MAP from baseline and first 2 annual visits. The relationship with MAP was studied in 16 703 participants amongst whom 1539 falls were recorded over 7.3 years. Variability in MAP was studied in 14 818 of these participants who experienced 974 falls over 4.1 years. Falls were confined to those involving hospital presentation. Cox regression was used to calculate hazard ratio and 95% CI for associations with falls. RESULTS: Long-term variability in MAP was not associated with falls except amongst frail or prefrail participants using antihypertensive medications. Within this group each 5 mm Hg increase in long-term variability in MAP increased the risk of falls by 16% (hazard ratio, 1.16 [95% CI, 1.02-1.33]). Amongst the antihypertensive drugs studied, beta-blocker monotherapy (hazard ratio, 1.93 [95% CI, 1.17-3.18]) was associated with an increased risk of falls compared with calcium channel blockers. CONCLUSIONS: Higher levels of long-term variability in MAP increase the risk of serious falls in older frail and prefrail individuals taking antihypertensive medications. The observation that the relationship was limited to frail and prefrail individuals might explain some of the variability of previous studies linking blood pressure indices and falls.