Do sleeping pills cause dementia? What 721,000 people's records actually show

Short answer: Probably a little, probably less than the headlines suggest, and nobody can yet tell you how much of the link is the drug and how much is the insomnia. The largest analysis to date — 13 studies and 721,354 people, published in Drugs in 2026 — found sedative-hypnotic users had about 29% higher odds of an Alzheimer's diagnosis. But when the same researchers looked only at the studies that tracked people forward in time, the association disappeared (hazard ratio 1.17, 95% CI 0.87–1.58 — statistically indistinguishable from no effect). Both numbers come from the same paper. Almost every page you will find quotes the first one.

This is educational content, not medical advice, and it is not a reason to stop anything. Stopping a sleeping tablet abruptly — benzodiazepines especially — can be genuinely dangerous. Every decision about a prescribed medicine belongs with the clinician who prescribed it.

Why this question is so hard to answer

Start with the problem that makes every study on this topic wobble.

Broken sleep is one of the things dementia does on its way in. Years before anyone uses the word, the disease disturbs sleep, stirs up anxiety and flattens mood. So the people who end up on sleeping tablets are disproportionately people who are already in the earliest, unnamed phase of the illness.

That means a study can find that sleeping-pill users get more dementia without the pills having caused any of it. Researchers call this reverse causation or confounding by indication, and on this particular question it is not a technicality — it may be most of the effect.

There is a clean way to test for it: look at when the association appears. A drug effect should still be visible if you go back ten years before diagnosis. Reverse causation should fade the further back you look, because the prodromal symptoms that drove the prescribing have not started yet.

A 2025 Canadian study did exactly that, and the result is the most useful thing in this whole literature — we come to it below.

What the largest analysis found

The best current summary is a systematic review, meta-analysis and meta-regression published in Drugs in 2026 (da Silva and colleagues), which pooled 13 observational studies covering 721,354 people who did not have dementia at the start. It was registered in advance, and the authors rated the certainty of the evidence themselves.

Headline result: sedative-hypnotic use was associated with odds of Alzheimer's about 29% higher (OR 1.29, 95% CI 1.10–1.53). Heterogeneity was very high (I² = 86.5%), meaning the individual studies disagreed with each other a great deal.

Then the part that almost never gets reported. When the analysis was restricted to studies reporting hazard ratios — the design that follows people forward through time rather than looking backwards from a diagnosis — the estimate shrank and lost statistical significance: HR 1.17 (0.87–1.58). The confidence interval comfortably includes "no effect at all."

The authors' own words: residual confounding and reverse causation "limit causal inference," and the certainty of the evidence across their analyses ranged from very low to moderate.

By drug class

Within the same paper:

• Benzodiazepines (diazepam, lorazepam, alprazolam, temazepam) — OR 1.21 (1.07–1.36)
• Z-drugs (zolpidem/Ambien, zopiclone, zaleplon) — OR 1.14 (1.10–1.18), with no heterogeneity at all (I² = 0%), which is unusual and makes this the most internally consistent estimate in the set
• Short-acting agents — OR 1.19 (1.04–1.36)
• Broad-acting benzodiazepines — OR 1.01 (0.98–1.05), i.e. no association
• Long-acting agents — OR 1.44 (0.99–2.09), an interval that crosses 1.0 and so does not reach significance

Two more subgroup results are worth knowing because they cut against the intuitive story. The association was present in people under 75 (OR 1.36, 1.24–1.49) but not in those 75 and over (OR 1.14, 0.61–2.11) — the opposite of what you would expect if the drugs were straightforwardly damaging ageing brains. And studies that defined dementia from administrative diagnosis codes produced a bigger estimate (OR 1.47, 1.16–1.86) than studies that used clinical criteria (OR 1.13, 0.84–1.52). How you define the outcome moves the answer about as much as the drug does.

The half-life question, where the evidence genuinely conflicts

You will read that long-acting sleeping pills are the dangerous ones. Two good studies point different ways, and it would be dishonest to pick one.

A 2025 Canadian case-control study (Legrand and colleagues, Journal of the Neurological Sciences) drew on national health-survey data linked to prescription records: 1,082 people with dementia matched to 4,262 controls, all aged 50 and over. Any benzodiazepine use was associated with dementia (OR 1.65, 1.42–1.93), and the association was strongest for long half-life molecules (OR 2.81) compared with medium half-life ones (OR 1.57).

The 2026 meta-analysis, which pooled far more people, found the reverse emphasis: short-acting agents significant, broad-acting benzodiazepines null, long-acting only borderline.

The honest position is that half-life may matter and the field has not settled which direction. What both papers agree on is the direction of the overall signal — small, real enough to prescribe carefully, and not large.

The finding that changes how you should read all of this

The Canadian study did the test described at the top: it moved the index date backwards, one to ten years before diagnosis, to see when the association existed.

In the fully adjusted model — the one that also accounted for why people were prescribed the drug (insomnia, anxiety, depression) — chronic use of more than 180 days was linked to dementia only within the four years before diagnosis.

The authors are explicit about what that implies: the restriction of the association to that four-year window "suggests confounding by indication or reverse causality."

Put plainly: in the years when someone's sleep and mood were being disturbed by a disease nobody had yet named, they got prescribed sleeping tablets. Go back further, before those symptoms started, and the link largely goes away. That is what a marker looks like. A cause would not switch off.

It is not proof that the drugs are harmless — the same paper still found an overall association, and the authors end by urging "cautious, time-limited" prescribing. But it does mean the frightening version of this story is, at minimum, substantially overstated.

What about over-the-counter sleep aids?

This is where the picture is different, and arguably more concerning.

Many non-prescription sleep aids — ZzzQuil, Unisom SleepTabs, the "PM" versions of Tylenol and Advil, and most supermarket own-brands — are simply diphenhydramine, the antihistamine in Benadryl. It is one of the strongest anticholinergic drugs you can buy without a prescription, and anticholinergic drugs are a separate and somewhat better-evidenced concern than benzodiazepines: they have their own dose-response relationship with dementia risk, and older sedating antihistamines are one of the three drug classes where the signal concentrates.

That evidence has its own page: which medications count as anticholinergic, and what the burden actually means. For diphenhydramine specifically, including how much counts as heavy use, see does Benadryl cause dementia?

The short version: an occasional tablet is not the scenario the research describes. Taking one most nights for years is.

Melatonin is the other common over-the-counter option, and it is not anticholinergic and not sedating in the same way. It has not been linked to dementia risk; it is also, for most adults with chronic insomnia, only modestly effective. It is a reasonable thing to discuss with a pharmacist or GP rather than a solution.

A finding about who was studied, and how to read it

One widely reported 2023 study deserves care, because it is easy to misread.

Leng, Stone and Yaffe followed 3,068 older adults (average age 74, 41.7% Black) for 15 years; 617 developed dementia. People who took sleep medications five or more times a month — compared with once a month or less — were more likely to develop dementia, but the association was seen only among White participants (HR 1.79, 1.21–2.66), not Black participants (HR 0.84, 0.38–1.83).

This does not mean sleeping pills are safe for some people and not others. The authors point to the more mundane explanation: White participants were "almost twice as likely to take all prescription hypnotics," and only 34 Black participants in the entire study were frequent users. With that few people, a study cannot detect a moderate effect even if one exists. "Not statistically significant" here means underpowered, not absent — and the difference in who gets prescribed what is a fact about healthcare access and prescribing patterns, not about brains.

What to do instead — and it is not "just stop"

Do not stop a prescribed sleeping tablet on your own. Benzodiazepine and Z-drug withdrawal can cause rebound insomnia, agitation and, at higher doses, seizures. Any reduction should be planned and tapered with a prescriber.

What the guidelines actually recommend:

1. Cognitive behavioural therapy for insomnia (CBT-I) is the recommended first treatment. The American College of Physicians makes this a strong recommendation based on moderate-quality evidence: all adults with chronic insomnia should be offered CBT-I as the initial treatment, before drugs. It is not counselling about your feelings — it is a short, structured programme that resets sleep timing, cuts time spent awake in bed and unpicks the anxiety loop around sleeping. It works about as well as medication in the short term and better in the long term, and it is available as digital programmes, self-help books and group courses, not just one-to-one therapy.

2. If you are 65 or over, the medication question is worth raising anyway. The American Geriatrics Society's Beers Criteria — the standard US reference for prescribing in older adults — lists benzodiazepines among the medicines to avoid in people 65 and over. That recommendation rests mostly on falls, fractures, confusion and road accidents, which are far better established than the dementia link. Those are reasons enough for a review.

3. Ask the question that actually matters: what is the plan? Sleeping tablets are designed for short, time-limited use. If you have been on one for years without anyone revisiting it, that is worth a conversation regardless of what the dementia research eventually concludes.

4. Treat the sleep problem, not just the sleeplessness. Untreated sleep apnoea, restless legs, chronic pain, depression and alcohol are common causes of persistent insomnia, and each has its own treatment. Poor sleep is itself associated with dementia risk — which is another reason a sleeping tablet that half-works for a decade is not a good outcome. More on that in sleep and dementia risk.

So, do sleeping pills cause dementia?

On the current evidence: no one can say they do, and no one can say they don't.

What can be said fairly:

• The association is real but small, and it shrinks to nothing in the study designs that follow people forward through time.
• A large part of it is probably the insomnia and the illness behind it, not the pill — which is exactly what the four-year prodromal window shows.
• The over-the-counter ones containing older antihistamines are a distinct and somewhat better-evidenced concern than prescription hypnotics.
• The strongest reasons to reduce long-term sleeping-tablet use in older adults have nothing to do with dementia. Falls, confusion and dependence are established. This is not.
• CBT-I is the treatment the guidelines actually recommend first, and most people are never offered it.

If you are worried enough to have searched this, the useful next step is not stopping a tablet. It is booking a medication review — and looking at the risk factors where the evidence is stronger and the actions clearer.

More on this: which medications are anticholinergic? · does Benadryl cause dementia? · sleep and dementia risk · is it anxiety or dementia? · normal forgetfulness vs dementia · check your dementia risk

How we research and review our content →

This article is educational and is not medical advice, diagnosis, or treatment. The studies cited describe findings in general populations, not any individual. Solenna is not a medical device and does not diagnose, prevent, treat, or cure Alzheimer's disease or any other condition. Risk reduction means lowering probability, not eliminating it; individual results vary and no outcome is guaranteed. Do not start, stop or change any prescribed medication — including sleeping tablets — without speaking to the clinician who prescribed it.