BACKGROUND AND PURPOSE: Elevated blood pressure (BP) is prevalent and modifiable and has been hypothesized to lead to increased risk of dementia. DATA: Data on 2 593 629 people from the UK Clinical Practice Research Database aged ≥40 years with a BP measurement between 1992 and 2011 and no prior record of dementia were collected. METHODS: Poisson regression models were used to study the association between BP and physician-diagnosed dementia. BP is believed to fall during the prodromal phase of dementia development, so associations were investigated by categories of time since BP measurement (<5, 5-10, >10 years) and by subtypes of dementia. RESULTS: During a median follow-up of 8.2 years, 65 618 cases of dementia were observed: 49 161 Alzheimer's, 13 816 vascular dementia and 2541 other subtypes. For each 10 mmHg higher systolic BP, the future dementia risk was 9.2% (95% confidence interval 8.4%-10.0%) lower, but this association varied markedly by time since BP measurement. Short-term associations with dementia were inverse with a 15.8% (15.5%-17.0%) lower risk 0-5 years after BP measurement and a 5.8% (7.7%-4.4%) lower risk 5-10 years after BP measurement. During the period >10 years after BP measurement, dementia risk was only 1.6% (0.1%-3.0%) lower, with a 4.3% (2.5%-6.0%) lower risk of Alzheimer's disease and a 7.0% (3.8%-10.2%) higher risk of vascular dementia. CONCLUSIONS: Elevated BP is associated with decreased risk of dementia in the short term, possibly due to reverse causation. Long-term associations of BP with dementia are less marked and differ by dementia subtype.