Risks III: Cognitive Decline

17 hours ago

15 min read

Part III of the Risks series. Part I covered cardiovascular disease. Part II covered cancer.

The short version

Lifetime risk is higher than most people think, and later than they fear. Roughly 42% of Americans who reach 55 will develop dementia by 95. But the risk is about 4% by 75 and 20% by 85. Most of it lands after 85.
Midlife is where the risk gets set. About 45% of dementia traces to 14 modifiable factors, and most of the leverage sits before 65. The two largest single contributors are ones nobody expects: untreated hearing loss and high LDL.
We can now find Alzheimer's pathology with a blood test. That doesn't mean you should look. Four blood tests are FDA-cleared, and all four are for people who already have symptoms. Finding something earlier only helps if doing something earlier changes the outcome. For the worried well, it doesn't yet.

Nobody in my practice lies awake worrying about their vitamin D status. Plenty of them lie awake after watching a parent stop recognizing them. Or after they walk into the kitchen and can't remember why they're there.

The kitchen one first, because it's the most common version of this question I get from people in their forties and fifties. Walking into a room and forgetting why is normal. So is blanking on an actor's name, or losing a word mid-sentence when you're tired and running on four hours of sleep. Forgetting how to get home from a place you've driven to a hundred times is not normal. Neither is a spouse who has quietly started paying the bills because you missed three of them.

Of everything in this series, this is the one people fear most. More than heart disease, and for a lot of people more than cancer. I understand why. Cancer threatens your life. Dementia threatens the person living it.

But the fear is aimed at the wrong decade. The disease mostly shows up after 80. The risk mostly gets built between 40 and 65. If you're reading this, you are almost certainly living in the decade where the decisions get made.

In the cancer article I laid out three places to intervene: lifestyle on the front end, screening in the middle, treatment on the back end. Same frame here, with one difference that runs through this entire article. For cancer, all three buckets have real tools in them. For cognitive decline, the middle and back buckets are thin. Almost all of the leverage is on the front end.

What we're actually talking about

Dementia is not a disease. It's a syndrome, a description of what's happening rather than why. Alzheimer's disease is the most common cause. Vascular dementia, Lewy body dementia, and frontotemporal dementia are the others you'll hear about. In older brains, the most common finding at autopsy is a mix of more than one. That matters for this article, because a large share of what we call dementia has a vascular component. Which means the cardiovascular article already did half the work.

There's also a spectrum to keep straight. Normal aging slows processing speed and makes names harder to retrieve, but you still function the way you did. Mild cognitive impairment is measurable decline on testing while daily life stays intact. Dementia is decline that interferes with living independently. The dividing line between those categories is function, not a test score. That's the same line I drew in the health and physical walkthrough of the 5/2: function is the goal, and the numbers are the instrument panel.

One more complication, and it's one I've written about before. We are unreliable narrators of our own state. In my experience, the person sitting across from me who is worried about their memory often isn't the one I'm worried about. The person whose spouse made the appointment often is. I want to be careful with that, because subjective memory complaints do carry some added risk and shouldn't be waved off. But if the people who know you best have started noticing, that deserves more weight than your own impression. You're the last one who gets to see it clearly.

What actually drives the risk

The number one risk factor for dementia is age, the same as cancer. A large 2025 analysis of more than 15,000 Americans put the lifetime risk of dementia after age 55 at 42%. That's more than double the older estimates, and it's a scary number until you look at when it arrives. The risk was about 4% by 75 and 20% by 85. Most of it lands after 85. Women carry more lifetime risk than men, 48% versus 35%, largely because women live long enough to get there.

So the honest framing is not that four in ten of us are headed for dementia in our sixties. It's that the longer you live, the more this becomes the thing you're managing. The better you do at not dying of heart disease and cancer, the more this article applies to you.

Genetics

The gene everyone asks about is APOE. Carrying one copy of the ε4 variant raises lifetime risk to roughly 48%. Two copies push it close to 60%. No copies, about 39%. A lot of you already know your status because a consumer genetics test told you, whether you wanted to know or not.

For the record, I've never checked mine. That isn't avoidance. It's that I can't find a decision that would change. If I'm an ε4 carrier, the plan is to keep my blood pressure, lipids, and metabolic health in range, train, sleep, stay connected, and keep learning. If I'm not, the plan is exactly the same. Where APOE status does matter is narrower: it changes the side-effect risk if you ever become a candidate for the newer Alzheimer's drugs, and it can matter for research trials. That's a conversation for when you have symptoms, not before. If knowing would motivate you, that's a legitimate reason to find out. If it would just give you something new to lie awake about, it's a legitimate reason not to.

Family history works the way it did in the cancer article. Know who, what kind, and at what age. A parent diagnosed at 58 means something different from a grandparent who got confused at 91.

The part you control

In 2024, the Lancet Commission on dementia updated its list of modifiable risk factors to fourteen. Across early life, midlife, and late life: less education, hearing loss, high LDL cholesterol, depression, traumatic brain injury, physical inactivity, diabetes, smoking, hypertension, obesity, excessive alcohol, social isolation, air pollution, and untreated vision loss. Their estimate is that about 45% of dementia could theoretically be prevented or delayed by addressing all fourteen.

If that split sounds familiar, it should. In the cancer article it was roughly 40% in your control and 60% not. Here it's 45 and 55. Say both halves out loud again. Almost half is a lot. It is also not most.

I want to be careful with that 45%, the same way I was careful with the insulin resistance verb in the cancer article. It's a population attributable fraction. It assumes every one of those associations is causal, and it treats the factors as if they don't overlap. They overlap heavily. Inactivity, obesity, diabetes, and hypertension are mostly one metabolic story told four ways. Treat 45% as a ceiling, not a promise.

The two largest single factors on the list surprise almost everyone: hearing loss and high LDL, each credited with about 7% of cases.

Hearing

This is the most neglected lever in the whole article. The proposed mechanisms are reasonable ones. A brain straining to decode speech has less left over for everything else, people who can't hear withdraw from conversation, and a quieter world is a less stimulating one.

The best trial we have is ACHIEVE, published in 2023. It randomized about 1,000 adults in their 70s and 80s with untreated hearing loss to hearing aids and audiology support, or to health education. Across the whole group, hearing intervention did not slow cognitive decline over three years. But in the prespecified subgroup of 238 people at higher risk, the ones who were declining fastest, it slowed decline by 48%.

That's a subgroup result, and I'd hold it loosely. I'd also point out what it costs to act on it. A hearing test is cheap. Hearing aids are now sold over the counter. And the people who need them most are the ones least likely to wear them, because they don't think they need them. See the insight problem above.

Blood pressure and LDL: the vascular half

This is where the cardiovascular article pays off. The SPRINT MIND trial showed that more intensive blood pressure control reduced mild cognitive impairment. A large cluster-randomized trial in China, roughly 34,000 people with hypertension, found that lowering blood pressure cut all-cause dementia by 15%. LDL is new to the Lancet list as of 2024, and it came in tied for the largest single factor.

Same numbers. Second organ. Your blood pressure and your LDL are brain numbers. We just file them under heart.

Metabolic health

In the cancer article I said that fixing metabolic health is one of the few interventions that lowers risk across cardiovascular disease, cognitive decline, and cancer at once. This is where that promise gets cashed. Diabetes, obesity, and inactivity are all on the list, and they mostly travel together.

You'll hear Alzheimer's called type 3 diabetes. It's a catchy phrase and an overclaim. Insulin resistance and dementia are clearly connected, and the brain is an energy-hungry organ that doesn't do well in a metabolically sick body. That's enough to act on. It isn't the same as saying Alzheimer's is diabetes of the brain, and I'd rather not borrow certainty the evidence hasn't earned.

Infection and the shingles vaccine

This is the most interesting finding of the last two years, and it comes from an accident of bureaucracy. When Wales rolled out its shingles vaccine program in 2013, eligibility was set by exact date of birth. People born one week too early were essentially shut out. That created something close to a randomized trial, without anyone planning one.

Researchers at Stanford followed about 280,000 Welsh adults. Those who were eligible and got the vaccine were roughly 20% less likely to receive a new dementia diagnosis over the next seven years. A follow-up study found fewer diagnoses of mild cognitive impairment and fewer deaths from dementia among people who already had it. A separate analysis found the current recombinant vaccine, Shingrix, associated with even lower risk than the older live vaccine studied in Wales.

The caveats are real. The Welsh data used the older live vaccine. The effect was clear in women and not statistically significant in men. And this is still observational data, even if it's unusually clean observational data. But for a cheap vaccine that's already recommended for everyone over 50, I don't need certainty to act. I'll spend more time on the infection side of this in Risks VI.

Things people worry about that don't deserve it, and one that does

Aluminum cookware, memory supplements advertised during the evening news, and brain-training apps take up a lot of the public conversation about dementia. None of them deserve it. The supplements are mostly unproven. The apps mostly make you better at the app.

What deserves far more attention is your medicine cabinet. A class of drugs called anticholinergics has been linked to cognitive decline with long-term use. The most common one is diphenhydramine, which is Benadryl, and it's also the sleep ingredient in Tylenol PM and most over-the-counter sleep aids. Several older bladder medications and some antidepressants carry the same burden. Long-term benzodiazepine use belongs on this list too. Almost no wellness content talks about this, because there's nothing to sell. It's one of the simplest conversations you can have with your doctor.

One more that comes up constantly in my practice: hormone therapy. The timing of when estrogen is started seems to matter, and starting combined hormone therapy after 65 was associated with harm in the Women's Health Initiative memory study. Current guidance doesn't support starting hormone therapy for the purpose of preventing dementia. There are good reasons to use it. This isn't one of them yet.

Sleep, alcohol, exercise, and social connection all belong in this section too. I've written about each of them at length, and alcohol gets its own article in November. The short version for exercise: the VO2 max and grip strength on the health score card both track with lower dementia risk. The score card is a brain score card too. We just didn't call it that.

What's worth measuring

You're already measuring most of what matters. You just file it under heart.

Blood pressure, LDL, A1c, and fasting insulin are on the score card for cardiovascular and metabolic reasons, and every one of them is also a brain number. So are VO2 max and grip strength. The two things almost nobody measures are the ones tied for the top of the Lancet list: hearing and vision. A hearing test takes about twenty minutes and costs little or nothing. An eye exam is the same. Both belong on the errands list I argued for in the 5/2 health walkthrough, right next to the dental cleaning you've been putting off.

If there's a real symptom, the workup changes. That means a cognitive screening test, a medication review, and a look for the reversible causes that masquerade as decline: low B12, thyroid disease, depression, untreated sleep apnea, and too many sedating medications. More often than people expect, the answer is in that list.

The blood test question

For most of my career, confirming Alzheimer's pathology meant an amyloid PET scan or a spinal tap. That changed fast. The FDA cleared the first blood test for Alzheimer's pathology in May 2025, and as of last month there are four. The newest, cleared in August, is a single pTau217 test that's meant to be usable in primary care, both to rule the disease in and to rule it out.

That's genuine progress. For a patient with symptoms, a blood draw is a far better first step than a spinal tap. If they become more cost effective we can consider these blood tests, if there are treatments developed and found to be effective for managing.

Notice who they're cleared for, though: people with signs or symptoms of cognitive decline. Not people who feel fine and want to know. That's not a regulatory technicality. It's the whole argument.

Amyloid in the brain is not dementia. Plenty of people carry amyloid for years and never decline. There is no approved treatment for someone who is amyloid-positive with no symptoms. So if you feel fine and test positive, you've bought anxiety, possible insurance questions, and nothing new to do, because the action list is the same one you'd follow with a negative result.

This is the ovarian cancer lesson from the last article, with a cleaner answer. Finding a disease earlier only helps if there's a treatment at that earlier stage that changes the outcome. Colorectal screening works because you can remove the polyp. Screening healthy people for amyloid doesn't work yet, because there's no polyp to remove. I'll come back to this in December when I write about the harms of screening.

If you have symptoms, I'll order the test and walk you through what the result means. If you don't, and you still want it, I'll have that conversation too. What I won't do is sell it to you as prevention.

Treatment, and where the money goes

The back end of this disease got its first real drugs in the last three years, and they're worth understanding even if you never need them.

Lecanemab (Leqembi) and donanemab (Kisunla) are antibodies that clear amyloid from the brain. In their trials, people with early Alzheimer's declined roughly a quarter to a third more slowly over 18 months than people on placebo. On the rating scale the trials used, which runs from 0 to 18, the difference was well under one point. They require regular infusions or injections and repeated MRIs, because the main side effect is brain swelling and small bleeds called ARIA. That showed up on scans in roughly one in eight people on lecanemab and one in four on donanemab in the trials, and it's more common in APOE ε4 carriers. Most cases were found on routine scans with no symptoms. A small number were serious.

In April, a Cochrane review pooled 17 trials of seven anti-amyloid antibodies and concluded they provide no clinically meaningful benefit. The pushback was immediate and fair: five of those seven drugs had already failed, and averaging two drugs that work with five that don't is a good way to make everything look like it doesn't work. Both sides have a point. The drugs clearly do something to biology. Whether a few months of slower decline is worth the infusions, the scans, and the risk is a values question. Different families will answer it differently, and I don't think either answer is wrong.

The other result worth knowing came from the GLP-1 world. Observational data had suggested semaglutide might protect the brain, and Novo Nordisk ran two large trials, EVOKE and EVOKE+, in about 3,800 people with early Alzheimer's. Biomarkers improved. The disease didn't slow. That was treatment of established disease, not prevention, so the prevention question is still open. But it's a clean example of why I don't take a promising association to the bank until the trial reads out.

The money

The Alzheimer's Association projects that health care and long-term care for people with dementia will cost $409 billion this year. That doesn't count unpaid family caregiving, which is where most of the real cost lands. Drug development spending is climbing fast. And the most encouraging randomized result in this field last year didn't come from a drug at all. It came from a lifestyle trial.

There's a financial way to think about this that I find useful. Dementia is a tail risk: unlikely in any given year, catastrophic if it arrives, and concentrated late. You hedge tail risk early and cheaply, not late and expensively. That applies to your health, and it applies to your finances, which is why long-term care planning will show up when I get to finances in the 5/2 series. The family side of this, the caregiving, gets its own article in March.

What's in your control

That lifestyle trial is U.S. POINTER, published in JAMA in July 2025. It randomized about 2,100 older adults at elevated risk of cognitive decline to one of two programs for two years. The self-guided group got education and encouragement. The structured group got a prescription: aerobic exercise four times a week plus strength and flexibility work, the MIND diet, cognitive and social challenge, and regular reviews of their health numbers with someone holding them accountable.

Both groups improved. The structured group improved more. Later analyses found the structured group also slept better and moved other risk markers in the right direction.

Two honest things about it. First, it's the first large American randomized trial to show that a lifestyle program protects cognition, and that's a big deal. Second, the effect was modest, and the trial wasn't designed to show fewer dementia diagnoses. It showed better thinking over two years. We don't yet know how that translates to twenty.

The finding I care most about is the gap between the two groups. Structure beat willpower. The people who were told what to do, and had someone checking, did better than the people who were given the same information and trusted to figure it out. That's the entire argument for the 5/2, for the pirate map, and for paying yourself first before the day gets a vote.

I should say plainly that I have a conflict here. Structure and accountability are what I sell. In the GLP-1 article I told you to ask who profits from the recommendation, and that applies to me too. So discount my enthusiasm accordingly, and notice that the trial doesn't require me. It requires structure. A training partner, a standing walk with a friend, a calendar you actually keep, or a spouse who asks how the week went all count.

The first item on my own daily list is to learn something. I didn't put it there for dementia prevention. But the research on cognitive reserve, the idea that a brain that keeps learning has more room to lose before it shows, is a decent reason to keep it there.

Not in your control

Age, which is the largest risk factor and only moves one direction
Your genes, including APOE
Your family history

In your control

Your blood pressure, lipids, and metabolic health
Whether your hearing and vision problems get treated
Whether you move, sleep, drink, and smoke
Whether you stay connected to people and keep learning
What's in your medicine cabinet
Whether you get the shingles vaccine
Whether you get a real workup when something changes, instead of waiting a year

That second list should look familiar. It's nearly the same list from the cardiovascular article and the cancer article. This is the third disease in a row where the upstream work turns out to be the same work. That isn't a coincidence, and it's the best news in this series. You don't need a separate plan for your heart, your cancer risk, and your brain. You need one plan, done consistently, for a long time.

Three things to do

1. Get your hearing and vision checked this year. Together they account for roughly 9% of preventable dementia. Both are cheap to test, and both are the items nobody puts on their list. If you need hearing aids, wear them. If the people you live with keep asking you to turn the TV down, you probably need them.

2. Treat your midlife numbers as brain numbers. Blood pressure, LDL, and A1c aren't only about your heart. The years between 40 and 65 are when this risk gets set, and those are the years you're living now. If one of those numbers has been "borderline" for five years, that's a decision you've been deferring, not a decision you've made.

3. If you're 50 or older, get the shingles vaccine, and skip the amyloid test. The vaccine is already recommended, it's inexpensive, and it has the strongest dementia signal of any single intervention we have right now. An amyloid blood test with no symptoms buys you a number and nothing to do with it.

Next in this series: Risks IV — Bones, Joints, and Falling Apart.