Infectious diseases have been hypothesised to increase the risk of dementia. However, the evidence is sparse, captures only a limited range of infectious diseases, and relies on short follow-up periods. We assessed a wide range of severe (hospital-treated) bacterial and viral infections and their subtypes as risk factors for dementia in three large cohorts followed up for almost two decades and replicated the main findings in the UK Biobank. Methods: For primary analysis, we pooled

more »

... vel data from three prospective cohort studies with a median follow up of 19 years (from 1986-2005 to 2012-2016) and a total of 273 125 dementia-free community-dwelling participants. The replication analysis with the UK Biobank was based on 492 146 individuals (median follow-up 9.0 years from 2006-2010 to 2018). We ascertained exposure to infectious diseases and their subtypes before dementia onset using linked records from national hospital inpatient registers. Incident dementia was identified from linked hospital inpatient and outpatient records, medication reimbursement entitlements, and death certificates. Findings: In the primary analysis based on 5.3 million person-years at risk, 88 099 participants had a hospital-treated infection before dementia onset and 3064 developed dementia. Gram-negative bacterial infections (hazard ratio [HR] 1.64; 95% confidence interval [CI] 1.25-2.14) and herpesvirus infections (HR 1.96; 95% CI 1.31-2.93) were robustly associated with an increased risk of dementia. For these infections, the relative risk of dementia remained similar when reverse causation and ascertainment biases were minimised by assessing only new dementia cases that occurred more than 10 years after the infection, and when comorbidities and potential confounders were considered. The associations were replicated in the UK Biobank with stronger relations observed for vascular dementia than Alzheimer's disease. In contrast to gram-negative bacterial and herpesvirus infections, the hazard ratio for all bacterial infections combined attenuated from 1.60 (95% CI 1.48-1.72) to 1.27 (95% CI 1.14-1.41) when bias was minimised. For all viral infections, the corresponding attenuation was from HR 1.63 (95% CI 1.32-2.00) to 1.19 (95% CI 0.87-1.62). Interpretation: Gram-negative bacterial infections and herpesvirus infections were associated with a moderately increased risk of dementia both in the short- and long-term. Hospital-treated bacterial and viral infections in general had only modest long-term associations with incident dementia.