BACKGROUND: Uncertainty remains concerning the association of blood cholesterol with the risk of subsequent dementia. Using data from people with lipid measurements in the UK Clinical Practice Research Datalink (CPRD), we examined the association between blood lipid levels and dementia (both vascular and non-vascular, including Alzheimer's disease) by age at first measurement of blood lipids and duration of follow-up. METHODS: We studied a cohort from the UK CPRD of people aged 40 years or older with a first total cholesterol recording between Jan 1, 1992, and Dec 31, 2009. Follow-up was until the first record of dementia, the last data collection date, patient death or transfer out of the practice, or Jan 5, 2015, whichever was earliest. We excluded individuals with a record of dementia before the total cholesterol measurement. We used Poisson regression to examine the association between baseline total cholesterol, LDL cholesterol and HDL cholesterol, and triglycerides and incident dementia diagnosis. Analyses were stratified by age at first measurement (<65 years or ≥65 years) and duration of follow-up (<10 years or ≥10 years). Our primary focus was LDL cholesterol. We adjusted for age, sex, calendar year, country within the UK, socioeconomic status, ethnicity, smoking, alcohol, body-mass index, comorbidities, and prescriptions. FINDINGS: 1 853 954 people had a first total cholesterol recording (dementia diagnosis in 49 416 [2·7%] people), including 953 635 [51·4%] people with LDL cholesterol values for analysis (dementia diagnosis in 21 602 [2·3%] people). Overall, we found a modest positive association between LDL cholesterol and dementia, with an adjusted rate ratio (RR) of 1·05 (95% CI 1·03-1·06) per SD increase in LDL cholesterol (1·01 mmol/L or 39 mg/dL increase). Adjusted RRs per 1-SD increase in LDL cholesterol in people younger than 65 years at baseline (n=636 262) were 1·10 (95% 1·04-1·15) for dementia diagnosed in the first 10 years after measurement and 1·17 (1·08-1·27) for dementia diagnosed more than 10 years after measurement. Associations for LDL cholesterol in people aged 65 years or older at baseline (n=317 373) were weaker compared with people younger than 65 years (RR 1·03 [95% CI 1·01-1·05] for dementia diagnosed during the first 10 years of follow-up and 1·07 [1·03-1·13] for dementia diagnosed after 10 years). We observed a weaker association between total cholesterol and dementia incidence and no consistent associations for HDL cholesterol and triglycerides. INTERPRETATION: LDL cholesterol measured in mid-life (<65 years) is modestly associated with dementia risk more than 10 years later. LDL cholesterol should be added to the list of modifiable risk factors for dementia. FUNDING: The Alzheimer's Society.