Whether late-onset depression is a risk factor for or a prodrome of dementia remains unclear. We investigated the impact of depressive symptoms and early- v. late-onset depression on subsequent dementia in a cohort of elderly general-practitioner patients (n = 2663, mean age = 81.2 years).

Risk for subsequent dementia was estimated over three follow-ups (each 18 months apart) depending on history of depression, particularly age of depression onset, and current depressive symptoms using proportional hazard models. We also examined the additive prediction of incident dementia by depression beyond cognitive impairment.

An increase of dementia risk for higher age cut-offs of late-onset depression was found. In analyses controlling for age, sex, education, and apolipoprotein E4 genotype, we found that very late-onset depression (aged ⩾70 years) and current depressive symptoms separately predicted all-cause dementia. Combined very late-onset depression with current depressive symptoms was specifically predictive for later Alzheimer's disease (AD; adjusted hazard ratio 5.48, 95% confidence interval 2.41–12.46, p < 0.001). This association was still significant after controlling for cognitive measures, but further analyses suggested that it was mediated by subjective memory impairment with worries.

Depression might be a prodrome of AD but not of dementia of other aetiology as very late-onset depression in combination with current depressive symptoms, possibly emerging as a consequence of subjectively perceived worrisome cognitive deterioration, was most predictive. As depression parameters and subjective memory impairment predicted AD independently of objective cognition, clinicians should take this into account.