Background Depression has been noted as one of the modifiable risk factors for dementia. However, since dementia has a long prodromal period, studies with short follow-up are susceptible to reverse causation bias. To determine whether depression is a cause or consequence of dementia, we investigated its association with incident dementia in separate follow-up periods over a maximum of 23 years.

Methods The Million Women Study is a population-based prospective cohort study of 1.3 million UK women who were recruited during 1996-2001. Women reported on current treatment for depression or anxiety and recent medication use on a re-survey in median year 2001 (IQR 2001-2003). All participants were followed for incident dementia through electronic linkage to NHS hospital records. Cox regression yielded adjusted hazard ratios (aHR) and 95% confidence intervals (CIs) for the first hospital record mentioning dementia in separate follow-up periods (≤5 years, >5-≤10 years, >10-≤15 years, >15 years), comparing those currently treated for depression/anxiety (with and without psychotropic medication use) to those without depression/anxiety. Adjustment was made for age, region, deprivation, education, body mass index, menopausal hormone therapy, physical activity, smoking, alcohol intake and comorbidities.

Results 805,673 women (mean age 60.0 years) without any prior dementia were included in the analysis. During a mean follow-up of 18.1 years, there were 45,909 incident cases of dementia. Compared to those with no depression/anxiety (n=748,142; 93%), the aHR for dementia for those with depression/anxiety and taking medication (n=41,130; 5%) was statistically significantly high during the first five years of follow-up (aHR 3.03, 95% CI 2.45-3.73). The aHRs diminished over the subsequent follow-up periods, and during >15 years of follow-up the aHR was 1.64 (95% CI 1.57–1.73). The association between depression/anxiety and vascular dementia was slightly stronger than the association with Alzheimer’s disease (aHR 1.76, 95% CI 1.57–1.97 vs 1.46, 95% CI 1.34–1.59), during >15 years of follow-up. Results were similar for the 16,401 women (2%) who reported being treated for depression/anxiety but were not taking medication.

Conclusion This study found evidence of reverse causation bias in the association between depression/anxiety and dementia. However, even after excluding 15 years of follow-up, depression/anxiety was significantly associated with a long-term higher risk of dementia. The difference in risk of the two main subtypes of dementia requires further investigation.