Introduction: Whether depression is a prodromal phase or risk factor for dementia is under debate. We aimed to unveil the nature of depression-dementia association by looking into the time window of depression occurrence.

Methods: Dementia-free twins (n = 41,727) from the Swedish Twin Registry were followed-up for 18 years. Data were analyzed using generalized estimating equation (GEE) for all individuals and conditional logistic regression for co-twin matched pairs.

Results: In the GEE model, multi-adjusted odds ratios (ORs; 95% confidence intervals [CIs]) of dementia were 1.46 (1.09-1.95) for mid-life, 2.16 (1.82-2.56) for late-life, 2.24 (1.49-3.36) for mid- to late-life, and 2.65 (1.17-5.98) for lifelong depression. The ORs in conditional logistic regression and in GEE did not differ significantly (P = 0.60). Education >= 8 years attenuated dementia risk associated with mid-life depression.

Discussion: Not only late-life depression, but also mid-life depression is associated with dementia. Genetic and early-life environmental factors could not account for this association. Education >= 8 years might buffer the impact of mid-life depression on dementia.