The Wrong Reason to Stop Taking Unisom

The dementia headline is the weakest part of the case against this drug. The stronger case against it is the risk of falls.

What is actually in Unisom?

Two different drugs, sold under one brand name, on the same shelf. Unisom SleepTabs contain doxylamine succinate. Unisom SleepGels, SleepMinis and PM Pain contain diphenhydramine.

This matters more than it sounds. A person who has taken SleepTabs for years and picks up SleepGels because the store was out has changed active ingredient entirely, with nothing on the front of the box to tell them so.

According to the current US labels on DailyMed, the NIH's drug label database:

• Unisom SleepTabs — doxylamine succinate, 25 mg per tablet
• Unisom SleepGels — diphenhydramine hydrochloride, 50 mg per softgel
• Unisom SleepMinis — diphenhydramine hydrochloride, 25 mg per capsule, with a labeled dose of two capsules, so 50 mg
• Unisom PM Pain — acetaminophen 325 mg plus diphenhydramine hydrochloride 50 mg

The warnings are drug-specific, which produces a real gap. The SleepTabs label warns against combining with other doxylamine products. The SleepGels label warns against combining with other diphenhydramine products, "even one used on skin." Neither warns you about the other. Someone taking SleepTabs at night and a diphenhydramine allergy pill during the day is not warned by either box.

The brand also now includes melatonin gummies and pills. Those are dietary supplements rather than drugs, they are not in the FDA's label database, and their contents are not verifiable the same way. So "Unisom" is not a single thing.

Does Unisom cause dementia?

The two largest studies ever run on this question both found no significant dementia association for antihistamines specifically — and both were clearly able to detect associations for other drugs in the same analysis.

This is not the answer most articles give, so here is the evidence in full.

Coupland and colleagues, JAMA Internal Medicine, 2019. A nested case-control study in English primary care: 58,769 people diagnosed with dementia, matched to 225,574 controls, aged 55 and over. Overall, the highest category of anticholinergic exposure carried an adjusted odds ratio of 1.49 (95% CI 1.44–1.54). But broken out by drug class, the authors wrote: "There were no significant increases in risk associated with antihistamines, skeletal muscle relaxants, gastrointestinal antispasmodics, antiarrhythmics, or antimuscarinic bronchodilators." For antihistamines at the highest exposure level, the adjusted odds ratio was 1.14 (0.98–1.34), and there was no dose-response pattern. The classes that did show a clear association were antidepressants, bladder antimuscarinics, antiparkinson drugs and antiepileptics.

Richardson and colleagues, BMJ, 2018. A separate UK database: 40,770 dementia cases aged 65 to 99, median age 83, with 283,933 controls. Same result. Antidepressants, urological drugs and antiparkinson drugs showed significant associations. Antihistamines did not — odds ratio 1.03 (1.00–1.07), and null at every time window they examined, including 15 to 20 years before diagnosis. Their own summary: "there was a significant association between dementia incidence and any prescription of antidepressant, antiparkinson, or urological drugs with an ACB score of 3, but no association with antispasmodic, antipsychotic, antihistamine, or other drugs with an ACB score of 3."

Two independent databases, roughly 100,000 dementia cases between them, both null for antihistamines while both detected real signals elsewhere. A null result in a study demonstrably powered to find effects is the most informative kind of null there is.

Then where does the 54% figure come from?

From a 2015 study of cumulative anticholinergic exposure whose heavy-use group was driven mainly by antidepressants — not by sleep aids.

The number is real and it is quoted correctly as far as it goes. Gray and colleagues, publishing in JAMA Internal Medicine in 2015, followed 3,434 adults aged 65 and over in the Adult Changes in Thought study in Seattle. People in the highest cumulative-exposure group had an adjusted hazard ratio of 1.54 (95% CI 1.21–1.96) for dementia — the 54%.

The part that rarely travels with the number is what made up that exposure. Antihistamines were the most widely used class in the cohort, taken by 64.8 percent of participants — but they accounted for only 17.2 percent of the total standardized daily doses. Antidepressants, used by fewer people, accounted for 63.1 percent. The heavy-exposure category that produced the 1.54 was mostly antidepressants by dose. Gray and colleagues published no antihistamine-specific hazard ratio at all.

That last sentence is true, but it is not the whole story. Gray's team anticipated this objection and tested it. They ran a sensitivity analysis splitting the exposure into antidepressants versus every other anticholinergic class, entered both into one model, and found the heaviest users at similarly elevated risk in both groups. Their discussion says so plainly: the increased risk was "consistent across anticholinergic subclasses, with increased risk found for people with high use of anticholinergic medications other than antidepressants, such as first generation antihistamines and bladder antimuscarinics."

So Gray is not silent on antihistamines. It is a fair reading of that paper that the signal was not confined to antidepressants.

What decides it is what came next. Gray's subclass finding is a grouped analysis — "everything that is not an antidepressant" — in a cohort of 3,434 people in one Seattle health plan. Coupland and Richardson measured antihistamines as their own class, separately, in more than 600,000 people between them, and both found nothing while both found clear signals for other classes in the same analysis. A specific measurement in 600,000 outweighs a grouped one in 3,434. It does not erase it.

There is a further honest gap, and it cuts both ways. All three of these studies read prescription records. Unisom is bought off the shelf. Richardson's team said so directly: "Drugs obtained over-the-counter are not recorded in our data, hence we will have likely underestimated the use of certain antihistamines." Coupland's team raised the same point, noting that this kind of misclassification would tend to push odds ratios toward the null and might be part of why antihistamines showed no signal.

So the accurate statement is narrower than either the scare version or the all-clear version: there is no study of over-the-counter doxylamine and dementia risk. What exists is prescription data, and in prescription data antihistamines do not show the signal that other anticholinergic classes do.

So is there any reason to stop?

Yes — several, and they are better documented than the dementia question. Falls, next-day impairment, delirium, and a benefit small enough to disappear into its own margin of error.

Falls and fractures. A 2018 systematic review and meta-analysis in Osteoporosis International by Cho and colleagues pooled five observational studies in older adults and found first-generation antihistamine use associated with an odds ratio of 2.03 (95% CI 1.49–2.76) for injurious falls or fracture. Heterogeneity was zero — the studies agreed with each other. Roughly doubled odds of the fall that ends up in an emergency room.

Delirium. Agostini and colleagues, in Archives of Internal Medicine in 2001, followed 426 hospitalized patients aged 70 and over. Diphenhydramine exposure carried a relative risk of 1.7 (95% CI 1.3–2.3) for delirium symptoms, with a dose-response relationship. A large 2025 study in the Journal of the American Geriatrics Society, covering 328,140 admissions of adults 65 and over across 17 Ontario hospitals, found that patients treated by physicians in the highest antihistamine-prescribing quartile had higher odds of delirium — adjusted odds ratio 1.41 (1.28–1.56). Delirium is not dementia; it is sudden, it is usually reversible, and confusing the two does families no favors.

Next-day carryover. In a study published in Clinical Pharmacokinetics in 1989, a single 25 mg dose of doxylamine — the exact SleepTabs dose — had an elimination half-life of about 10 hours in younger adults, 12.2 hours in older women, and 15.5 hours in older men, whose drug clearance was significantly reduced. A half-life roughly the length of a night's sleep means a meaningful share of the peak level is still circulating at breakfast.

For diphenhydramine, the evidence is more direct and more alarming. A randomized crossover trial in Annals of Internal Medicine in 2000 put 40 licensed drivers through the Iowa Driving Simulator after diphenhydramine 50 mg, after alcohol dosed to an estimated blood alcohol concentration of 0.1 percent, and after placebo. Driving was worse after diphenhydramine than after the alcohol. Lane excursions over the center line more than doubled. The authors concluded that diphenhydramine "may have an even greater impact than does alcohol on the complex task of operating an automobile."

The finding inside that study that matters most for a caregiver: self-reported drowsiness barely correlated with actual impairment. You cannot tell from how you feel whether you are safe to drive. Two limits worth stating plainly — those drivers were aged 25 to 44, not older adults, and the trial was funded by a company that makes a competing antihistamine, which came out looking identical to placebo.

How well does it work? The American Academy of Sleep Medicine's 2017 clinical practice guideline pooled the trials. Diphenhydramine 50 mg reduced time to fall asleep by about 8 minutes compared with placebo — with a 95% confidence interval running from 2 minutes worse to 17 minutes better. Total sleep time was 12 minutes longer, again with a confidence interval crossing zero. Quality of sleep: no improvement. The guideline suggests clinicians not use diphenhydramine for insomnia, though it grades that recommendation weak, on low-quality evidence, with benefits about equal to harms.

What does geriatric medicine actually say about this drug?

Avoid it in anyone 65 or older. That is a strong recommendation, and doxylamine is named specifically.

The American Geriatrics Society Beers Criteria are the reference list of medications that are potentially inappropriate in older adults. The 2023 update — still the current version as of August 2026 — lists ten first-generation antihistamines to avoid, and doxylamine and oral diphenhydramine are both on it.

The stated rationale: "Highly anticholinergic; clearance reduced with advanced age, and tolerance develops when used as hypnotic; risk of confusion, dry mouth, constipation, and other anticholinergic effects or toxicity." The recommendation is graded strong, on moderate-quality evidence, and there are separate strong recommendations to avoid these drugs in anyone with delirium, with dementia or cognitive impairment, with a history of falls or fractures, and in men with urinary symptoms or an enlarged prostate.

Note what the rationale rests on: confusion, falls, tolerance, toxicity. Not a dementia claim.

Is it safe in pregnancy?

Yes, and this is a genuinely different situation. Doxylamine combined with vitamin B6 is an FDA-approved prescription treatment for nausea and vomiting in pregnancy, first approved in 1976.

If you have seen a scary headline about doxylamine and are pregnant or taking Diclegis or Bonjesta, none of the research described above applies to you.

Those products contain doxylamine succinate plus pyridoxine, at a maximum of 40 mg of doxylamine per day. The FDA-approved labels summarize the safety evidence directly: a meta-analysis of 16 cohort and 11 case-control studies published between 1963 and 1991 "reported no increased risk for malformations from first trimester exposures," and a second meta-analysis of 17 more studies found no statistically significant relationship between fetal abnormalities and first-trimester use. The American College of Obstetricians and Gynecologists names pyridoxine, with or without doxylamine, as first-line drug treatment for nausea and vomiting in pregnancy.

The dementia studies enrolled people aged 55 and over, 65 and over, and 65 to 99. They measured exposures accumulated over years to decades. A course of treatment for pregnancy nausea is, at most, a few hundred days, in a population thirty to fifty years younger, using the one drug class that came back null in both large studies that reported it separately. These are not the same question.

What has better evidence for sleep?

Cognitive behavioral therapy for insomnia. It is the only insomnia treatment that carries a strong recommendation from both the sleep medicine and internal medicine guideline bodies.

The American Academy of Sleep Medicine's 2021 behavioral treatment guideline puts it plainly: "We recommend that clinicians use multicomponent cognitive behavioral therapy for insomnia for the treatment of chronic insomnia disorder in adults. (STRONG)" — with strong defined as "almost all patients should receive the recommended course of action." It is the only approach in that guideline to earn a strong recommendation, and the evidence base is 49 studies.

The American College of Physicians reached the same conclusion in Annals of Internal Medicine in 2016: CBT-I as initial treatment for all adults with chronic insomnia, strong recommendation, moderate-quality evidence, with medications ideally used no longer than four to five weeks.

CBT-I is usually four to eight sessions. It is now widely delivered by telehealth and by app, which removes the "there is nobody near me" problem that used to end the conversation. Two honest caveats: the guideline notes that out-of-pocket cost can be higher than a bottle of pills, and one component of CBT-I — getting out of bed when you cannot sleep — needs adapting for anyone at high risk of falling.

One thing that does not work on its own: sleep hygiene advice. The 2021 guideline specifically recommends against using it as a standalone treatment. If that is all anyone has offered you, ask for more.

If someone in your house has been taking a nightly sleep aid for years, the conversation to have with a pharmacist is not "does this cause dementia." It is: this barely outperforms placebo, it roughly doubles the odds of a fall, it is still half on board in the morning, and geriatrics' own guideline says avoid it. What could we use instead, and how do we taper off this one without a week of rebound insomnia convincing us we need it forever?