Brain derived neurotrophic factor (BDNF) has been implicated in dementia. Preliminary evidence suggests that BDNF DNA methylation may be a diagnostic biomarker of dementia, but the potential pre-clinical utility remains unclear. Participants in the ESPRIT study were assessed for cognitive function and dementia (DSM-IV criteria) over 14-years. BDNF exon 1 promoter methylation was measured in blood at baseline (n=769), and buccal samples during follow-up (n=1062). Genotyping was carried out for several common BDNF SNPs, including Val66Met (rs6265) and APOE-ɛ4. Multivariable logistic regression analyses determined the association between BDNF methylation and both prevalent and incident dementia. Adjustment for gender, age, education, APOE-ε4 genotype, body mass index, depression, and type 2 diabetes, as well as possible effect modification by gender and genetic variation were also investigated. Weak evidence of an association between lower blood methylation and dementia was observed at one of 11 sites investigated (Δ-0.5%, 95%CI:-0.9,-0.04, p=0.03, p=0.22 adjusted for multiple comparisons). Buccal methylation at two other sites was associated with 14-year incident dementia cases prior to adjustment for multiple comparisons only, and the effect sizes were small (Δ+0.3%, OR:1.57, SE:0.30, p=0.02, p=0.14 adjusted and Δ-1.5%, OR:0.85, SE:0.06, p=0.03, p=0.14 adjusted). Genetic variation in the BDNF gene did not modify these associations, and no gender-specific effects were observed. There was only a weak correlation between blood and buccal BDNF log-methylation at two sites (both r=-0.11). There was no strong evidence that blood or buccal BDNF exon 1 promoter DNA methylation is associated with prevalent or incident dementia, and reported associations would not remain after adjustment for multiple testing.