OBJECTIVE To investigate the association between random measured glucose levels in middle and old age and dementia-related death. METHODS Population-based cohort study. METHODS Norwegian Counties Study (middle-aged individuals; 35-49) and Cohort of Norway participants (older individuals; 65-80). METHODS Individuals without (n=74,630) and with (n=3,095) known diabetes mellitus (N=77,725); 67,865 without and 2,341 with diabetes mellitus were included in the complete case analyses (nonmissing for all included covariates), of whom 1,580 without and 131 with diabetes mellitus died from dementia-related causes. METHODS Dementia-related death was ascertained according to the Norwegian Cause of Death Registry. Cox regression was used to assess the relationship between random glucose levels (nonfasting) in individuals without and with diabetes mellitus and dementia-related death. Education, smoking, cardiovascular disease, body mass index, cholesterol, blood pressure, and physical activity were adjusted for. RESULTS Individuals without diabetes mellitus at midlife with glucose levels between 6.5 and 11.0 mmol/L had a significantly greater risk of dementia-related death than those with levels less than 5.1 mmol/L (hazard ratio=1.32, 95% confidence interval=1.04-1.67) in a fully adjusted model. A dose-response relationship (P=.02) was observed. No significant association between high glucose levels in individuals aged 65 to 80 and dementia-related death was detected. CONCLUSIONS High random glucose levels measured in middle-aged but not older age persons without known diabetes mellitus were associated with greater risk of dementia-related death up to four decades later.