When the Depression Came First

Depression sits on every list of things you can change to lower your risk. The longest study ever run on the question suggests that, much of the time, it is the disease arriving early.

Is depression a cause of dementia, or an early sign of it?

The evidence points substantially toward early sign. Depressive symptoms measured decades ahead of a diagnosis do not predict it. Symptoms measured within about a decade of it do.

This is one of the few places in dementia research where a single well-designed study genuinely changed how the question is understood.

Singh-Manoux and colleagues published it in JAMA Psychiatry in 2017. They used the Whitehall II cohort — 10,189 British civil servants, aged 35 to 55 when they enrolled in 1985 — and measured depressive symptoms on nine separate occasions across 28 years, through to 2015. Three hundred and twenty-two of them developed dementia.

The results, side by side:

• Depressive symptoms recorded in 1985, with a mean follow-up of 27 years: hazard ratio 1.21 (95% CI 0.95–1.54). Not statistically significant.
• Depressive symptoms recorded in 2003, mean follow-up 11 years: hazard ratio 1.72 (95% CI 1.21–2.44). Significant.
• Chronic or recurring symptoms in the early phase, 22-year follow-up: 1.02 (0.72–1.44). Not significant.
• Chronic or recurring symptoms in the late phase: 1.67 (1.11–2.49). Significant.

Then they ran the analysis backward in time, and found that the depressive symptom scores of people who would go on to develop dementia began diverging from everyone else's about 11 years before the diagnosis, with the gap widening steadily from there.

The researchers' own framing is that depressive symptoms are either a prodromal feature of dementia or the two share common causes. What the data does not support is a straightforward causal arrow running from depression in your forties to dementia in your seventies.

Then why is depression on the official risk-factor list?

Because the Commission is working from a different kind of evidence, the effect it assigns is small, and "modifiable risk factor" does not mean "proven cause."

The 2024 report of the Lancet standing Commission on dementia prevention, published in The Lancet on 31 July 2024, identifies 14 modifiable risk factors and calculates that eliminating all of them could potentially prevent or delay about 45 percent of dementia worldwide.

Depression is one of the 14. Its population attributable fraction is 3 percent — the same as traumatic brain injury and untreated vision loss, less than half that of hypertension, hearing loss or low education, each at 7 percent. It is on the list, but it is not near the top of it.

The Commission also moved depression from the late-life category, where the 2020 report placed it, into midlife. That shift matters for exactly the reason the Whitehall data illustrates: the later in life you measure a risk factor, the harder it is to tell a cause from a symptom.

Two things can be true at once here, and both are. Depression is worth treating on its own terms, at any age, for reasons that have nothing to do with dementia. And the specific claim that treating a depressive episode will meaningfully lower a person's dementia risk is not something the evidence currently establishes.

Do antidepressants help someone who already has dementia?

On the best evidence available, no — and the side effects are real. This is one of the widest gaps between routine practice and trial results in the whole field.

The definitive trial is HTA-SADD, published in The Lancet in 2011 by Banerjee and colleagues. It randomized people with Alzheimer's disease and depression, referred to old-age psychiatry services, to sertraline, mirtazapine or placebo, double-blind.

At 13 weeks, the reduction in depression scores did not differ between groups. Sertraline versus placebo: mean difference 1.17 (95% CI −0.23 to 2.58), p = 0.10. Mirtazapine versus placebo: 0.01 (−1.37 to 1.38), p = 0.99. The findings held at 39 weeks.

The harms did differ. Adverse reactions occurred in 29 of 111 on placebo (26%), 46 of 107 on sertraline (43%), and 44 of 108 on mirtazapine (41%). The authors concluded that the practice of using these antidepressants as first-line treatment for depression in Alzheimer's disease should be reconsidered, and noted that many cases resolve with usual care.

The Cochrane review reached the same place. Dudas and colleagues, updating in 2018, pooled ten studies covering 1,592 people. On the one outcome for which they had high-quality evidence — depression rating scale scores — antidepressants showed little or no effect. No effect on ability to manage daily activities. Probably little or no effect on cognition. People taking an antidepressant were probably more likely to drop out and more likely to have at least one unwanted side effect.

There is a further complication in the research itself. The 2025 umbrella review in Molecular Psychiatry that assessed antidepressants as a possible dementia risk factor found the evidence very low certainty and flagged the reason: the comparison group in most studies is people not taking antidepressants, which means people without depression — and depression is itself associated with dementia. The drug and the illness cannot be separated in that design.

None of this means depression in dementia should be ignored. It means the first move is probably not a prescription.

What actually helps, then?

Talking treatments have modest but real evidence. So does finding the thing that is making the person miserable, which is frequently not a mood disorder at all.

The Cochrane review of psychological treatments in this population, updated by Orgeta and colleagues in 2022, found that cognitive behavioral therapy–based treatments probably improve depression symptoms, quality of life, and ability to manage daily activities at the end of the treatment period in people with dementia or mild cognitive impairment. The effects were small, and the review could not establish an effect on anxiety, or say which people benefit most from which approach. Small and real beats large and unsupported.

Before any of that, there is a list worth working through, because in dementia low mood is frequently a symptom of something concrete and fixable:

• Pain. Undertreated pain in someone who can no longer report it reliably presents as withdrawal, irritability and refusal.
• Hearing and vision. Both are on the Lancet list at 7 percent and 3 percent respectively. A person who cannot follow conversation withdraws from it.
• Constipation, a urinary tract infection, poor sleep, a new medication. All of them flatten mood, and all are checkable this week.
• Boredom and lost roles. A person removed from every task they used to do, with nothing put in their place, is having a reasonable response to their circumstances.
• Loneliness. Social isolation carries a 2 percent population attributable fraction in its own right.

If a specific antidepressant is being proposed, the questions from any medication conversation apply: what would we expect to see if it works, how long before we know, and what is the plan if it does not. Several older antidepressants are strongly anticholinergic, which is the one drug property most consistently associated with worsened cognition — that is worth asking about by name.

How do you tell depression from apathy?

Depression involves sadness, guilt, hopelessness. Apathy is loss of motivation and interest without the low mood — the person is not unhappy, they are simply not starting anything.

This distinction is easy to describe and hard to make in a living room, and getting it wrong sends the treatment in the wrong direction.

Apathy is the more common of the two in dementia. The person does not initiate, does not react much to good news or bad, does not seem distressed by any of it. Ask how they feel and they will often say fine, and mean it. Depression looks different from the inside — the person is suffering, and can usually tell you so if asked directly.

There is a further wrinkle, and it is uncomfortable. SSRIs have been associated with apathy and emotional blunting, including in older adults. The published prevalence estimates vary enormously — from single figures to well over half, depending on how it is defined and measured — so no confident number can be given, and the evidence quality is low. But the possibility is real enough that if apathy appears or deepens after an antidepressant is started, it belongs on the list of things to raise with the prescriber rather than being read automatically as the dementia progressing.

The instrument clinicians use for the depression question in this population is the Cornell Scale for Depression in Dementia — nineteen items, each scored absent, mild or severe. Its defining feature is that it is designed to be completed using both an interview with the person and a separate report from someone who knows them well, precisely because self-report becomes unreliable as dementia advances. If a clinician is assessing mood without asking the person who lives with them, they are working with half the information. Offer it.

Could this be dementia being mistaken for depression?

It happens often enough to be worth asking about directly — especially when a first mood problem appears after 50 and does not respond to treatment.

Behavioral-variant frontotemporal dementia commonly begins with apathy, loss of interest, loss of initiative and social withdrawal. Those are the presenting features of depression as well. Published series suggest roughly half of people ultimately diagnosed with bvFTD receive a psychiatric diagnosis first, with average diagnostic delays of five to six years from the first symptoms; in one South American series, 71.4 percent were initially given a psychiatric diagnosis, most often depression in women and anxiety or psychosis in men.

The features that should prompt a neurological opinion rather than another medication trial:

• A first episode of significant mood or personality change after age 50, in someone with no psychiatric history
• Changes in social conduct — tactlessness, loss of empathy, new impulsivity, spending, or disinhibition
• New food preferences, particularly a strong turn toward sweets, or overeating
• Rigid, repetitive routines that were not there before
• Depression that does not respond to two adequate treatment attempts
• A partner saying "this is not who he is" rather than "he seems sad"

Depression that presents for the first time in later life deserves a cognitive assessment alongside the mood assessment, not instead of it. If you have made that request and it has been brushed aside, it is reasonable to ask for it again, in writing, or to ask for a referral.

Untreated depression is dangerous in its own right at any age, and none of this is an argument for leaving it alone. It is an argument for asking a second question at the same time as the first one.