The Cabinet Nobody Reviews
Memory complaints in people over sixty usually trigger a familiar checklist. Thyroid panel. Vitamin B12. Blood pressure. A depression screen, maybe a referral for cognitive testing. All of that is reasonable, and all of it misses a question that costs nothing to ask: what is in the medicine cabinet?
The question matters because a long list of common drugs — some prescription-only, some sitting openly on store shelves — works by blocking acetylcholine, one of the brain's central signaling chemicals. Pharmacologists call these drugs anticholinergic.
The label covers a strange collection of bedfellows: nighttime sleep aids, first-generation allergy tablets, bladder-control medications, tricyclic antidepressants, some antipsychotics, drugs for motion sickness and gut spasms. What unites them is the mechanism. They damp the same neurotransmitter the brain uses, among other jobs, to lay down new memories.
On any given night, one of these drugs produces a fog that lifts by morning. The trouble starts with accumulation. Two anticholinergic drugs do not just add; they stack, because the brain reads the combined blockade, not the ingredient list.
Researchers quantify this with burden scores — a point or two per drug, weighted by potency — and a person taking a sleep aid plus a bladder pill plus an old antidepressant can land in the range the studies flag, without ever taking anything that looks risky.
And there is a structural reason the stack goes unnoticed. Each prescriber sees one drug. The over-the-counter items often go unmentioned at appointments, because patients assume a box labeled for sleep cannot be medicine in the important sense. So the burden builds in the gaps between appointments, in a cabinet everyone opens and nobody audits.
What follows is the evidence that this stack matters — two large studies, one American and one English — and then an older text that worried about exactly this failure mode, two thousand years before pharmacy databases existed.
- Anticholinergic drugs block acetylcholine; common examples include diphenhydramine sleep aids, oxybutynin for overactive bladder, and tricyclic antidepressants like amitriptyline.
- The brain responds to the combined blockade of several such drugs, which is why researchers sum them into a burden score rather than judging each pill separately.
What Anticholinergic Actually Means
Acetylcholine is one of the oldest signaling molecules in the nervous system, and in the brain it does dense work in the hippocampus and cortex — the regions that handle forming and retrieving memories. Drugs with anticholinergic activity do not attack those regions. They just occupy the receptors acetylcholine would bind to, like keys jammed in every lock of a building.
The immediate effects are the ones people notice and shrug off: dry mouth, constipation, drowsiness, blurry vision. The quieter effect is what the jammed locks do upstream — the brain's memory circuits run with the volume turned down.
Clinicians have known about the short-term version for decades. Geriatricians call the acute syndrome by its initials: the anticholinergic toxidrome, with confusion as a headline symptom. What turned into a research program was a different observation — that some people on these drugs long-term seemed to slide, gradually, into something that looked less like a side effect and more like a trajectory.
The scale of routine exposure is easy to underestimate. A box of nighttime cold-and-flu caplets carries a first-generation antihistamine. So does the allergy tablet from the gas station, the 'PM' version of a pain reliever, some nausea pills, a number of sleep supplements that are not supplements at all underneath.
On the prescription side, oxybutynin for overactive bladder, paroxetine among the antidepressants, and older tricyclics are strong anticholinergics. None of these is a poison. Each is a legitimate tool. The problem the research picked up was never a single bottle — it was arithmetic across years.
That word, cumulative, is where the next two studies live. One counts doses over a decade. The other reconstructs eleven years of prescriptions for a quarter-million people. Both asked versions of the same question: does the arithmetic show up in the diagnoses later?
- Check labels for diphenhydramine, doxylamine, chlorpheniramine, oxybutynin, and amitriptyline — five of the most common strong anticholinergics in everyday use.
- The acute effects (dry mouth, drowsiness) and the long-term concern are different questions; the research below concerns years of accumulated exposure, not one rough night.
"The fog from one night's pill lifts by morning. The fog from ten years of nights is the thing the studies went looking for."
Seven Years of Pharmacy Records in Seattle
The Adult Changes in Thought study has been tracking a large random sample of older adults in the Seattle area since the 1990s, and one of its advantages is unglamorous: because the health plan ran its own pharmacies, the researchers hold dispensing records, not memories.
In 2015 a team led by Shelly Gray at the University of Washington used them to follow 3,434 participants aged sixty-five and older, all dementia-free at enrollment, for an average of about seven years.
The exposure measure was blunt and good: total standardized daily doses of strong anticholinergic drugs dispensed over the prior ten years, with the most recent twelve months excluded so that early, undetected disease could not fake the signal. 2 percent — developed dementia. The dose-response line was clean.
Compared with non-users, people whose cumulative exposure stayed under ninety daily doses showed no increased risk. 54, with a confidence interval that excluded zero effect. A trend test across the four exposure bands came back strongly positive.
The most-used drugs in the high-exposure group were exactly the unglamorous ones: tricyclic antidepressants, first-generation antihistamines, bladder antimuscarinics.
One detail in the paper unsettled researchers more than the headline number. The general assumption had been that anticholinergic cognitive effects wash out once the drug stops. In this cohort, the risk did not look like it had fully rinsed out even years after discontinuation — though the authors themselves flagged the limits of what observational data can settle.
Reversal, or the lack of it, needed different methods. And an honest reading of the whole paper requires noticing what it cannot do: pharmacy records show what was dispensed, not what was swallowed, and no amount of statistical adjustment makes a cohort into a trial. The next study tried to shore up the design with sheer size.
What the Classics Knew About Cumulative Dose
Open the Shennong Bencao Jing, the Han-dynasty herbal compiled around two thousand years ago, and the preface sorts every medicine into three grades before a single remedy is listed. ' Middle drugs are used with judgment. ' The organizing idea is not potency. It is how long a substance tolerates continuous use — an explicit ranking of medicines by cumulative-harm risk.
The Suwen devotes a whole chapter's worth of arithmetic to the same worry. '
Read those lines next to a modern anticholinergic burden score and the resemblance is structural, not coincidental. Both systems treat duration and stacking as the hazard, not the individual agent. Both assume the patient will keep taking things unless somebody stands back and looks at the whole shelf.
Both place the duty of withdrawal — deliberately, at a defined point — on the prescriber, because the person inside the treatment rarely feels the accumulation happening.
A boundary is owed here, because it would be easy to oversell the parallel. The Neijing authors knew nothing of receptors or acetylcholine; their toxicology came from observation of poisoning and clinical course, and their dosages were framed for herbal decoctions, not antihistamines. The classical texts do not predict modern pharmacology.
What they supply is a clinical instinct — that the medicine which never gets reviewed becomes the disease's quiet collaborator — that the modern research program spent decades arriving at from the other direction.
- The Bencao Jing's three grades rank drugs by how much continuous use they tolerate — an ancient version of asking how long a medicine should stay on the list.
- Classical doctrine put withdrawal at a defined fraction of the course, deliberately, rather than letting treatment run until something forced a stop.
"Do not overshoot, or you injure the upright qi."
58,769 Case Files and the Classes That Carried the Signal
In 2019, Carol Coupland and colleagues mined the QResearch database of English general practices for a nested case-control study of a scale the Seattle cohort could not reach: 58,769 patients with a dementia diagnosis, each matched to several controls, 225,574 controls in all, everyone fifty-five or older.
Instead of self-report, the team reconstructed eleven years of prescriptions for fifty-six drugs with strong anticholinergic properties and converted them into the same cumulative-dose currency the ACT study used.
The numbers replicated. Highest cumulative exposure — again past about a thousand standardized daily doses — carried roughly fifty percent higher odds of dementia, with a dose-grade from the lowest exposure band upward. The population-attributable fraction, if the association turned out to be causal, came to about ten percent of dementia diagnoses in the window studied.
And the class breakdown was more interesting than the headline, because the signal was not uniform. Strong associations showed up for anticholinergic antidepressants, antipsychotics, bladder antimuscarinics, antiepileptics and antiparkinson drugs.
Antihistamines and gastrointestinal antispasmodics — the classes people would most expect, given the sleep-aid folklore — did not carry a detectable increase in this dataset.
That last finding cuts both ways and is worth sitting with. It weakens the tidy story that any dry-mouth pill rots your memory, and it shows how much still depends on which specific molecules, at what doses, over what windows. It also illustrates the limits of the method: a nested case-control can match, adjust and replicate, but it cannot randomize.
Confounding by indication — the possibility that the conditions being treated, not the drugs, drive some of the risk — is the standing objection, and the authors say so in their own discussion. The editorial that ran alongside was more cautious still: causation, it noted, has not been established by randomized trials, and deprescribing itself can carry costs, from returning symptoms to more acute-care visits.
So the honest summary of two decades of evidence looks like this: a consistent, dose-graded association across two countries and two designs, a plausible mechanism, and no definitive trial saying that removing the drugs removes the risk.
Why You Should Not Quit Cold
Here the article has to slow down, because the most dangerous way to read this evidence is as an instruction to stop.
Abruptly discontinuing an antidepressant can produce withdrawal symptoms that are genuinely miserable. Stopping a bladder antimuscarinic can hand a person back the incontinence the drug was managing — and falls from nighttime bathroom sprints carry their own fracture risk, which is not hypothetical.
Antipsychotics and antiepileptics are in a category of their own; nobody should touch those without their prescriber running the taper. The Gray 2015 paper's senior author said it in plain language when the study was published: no one should stop taking any therapy without consulting their healthcare provider.
There is also an evidence problem on the other side of the ledger. If lowering the burden is supposed to protect cognition, trials of deprescribing should show it. They mostly do not — yet. A 2023 Cochrane review hunted for randomized trials of anticholinergic reduction with cognitive outcomes and found just three, totaling 299 participants, followed for one to three months.
One trial showed a small improvement on a processing-speed test; the others found no difference. A separate meta-analysis of seven intervention trials in Age and Ageing the same year found no significant effect on cognition, falls or quality of life, partly because the interventions barely moved the burden scores at all. The deprescribing literature is a work in progress, not a proven rescue.
What remains defensible is narrower and more useful. A structured medication review — ideally with a pharmacist — is low-risk, free in many health systems, and the mechanism by which inappropriate anticholinergic exposure gets noticed at all. The realistic goal is not to fear every pill.
It is to make sure somebody is reading the whole list, and that anything with anticholinergic weight on it is there because the benefit still clears the bar, at the lowest dose that works.
- Never stop an antidepressant, bladder medication, antipsychotic or antiepileptic abruptly — tapers and substitutions are prescriber decisions, and some conditions return fast without their treatment.
- The practical, evidence-supported step is a full medication review with a prescriber or pharmacist, bringing every bottle including over-the-counter sleep and allergy products.
The Over-the-Counter Shelf Has the Same Chemistry
The prescription side of this story at least has a professional at the counter. The over-the-counter side has a colorfully designed box.
Nighttime formulations deserve special attention, because they are where the exposure hides. A large share of 'PM' pain relievers, nighttime cold caplets, and pharmacy-brand sleep aids get their effect from diphenhydramine or doxylamine — both strongly anticholinergic first-generation antihistamines. In the Seattle cohort, first-generation antihistamines were among the three most-used anticholinergic classes.
A person who takes a PM caplet most nights for three years has, by the study's arithmetic, crossed into the highest exposure band without ever holding a prescription.
The alternatives are often unremarkable, which is the point. Second-generation antihistamines such as loratadine were designed not to cross into the brain in meaningful amounts, and they manage allergy symptoms for most people without the central fog. For sleep, the non-drug basics — fixed wake time, dim light in the evening, a cool room — have trial evidence behind them and zero anticholinergic load.
None of this is exciting advice. Excitement is roughly what the current arrangement lacks.
Two habits close most of the remaining gap. First, read the drug facts panel for the word diphenhydramine or doxylamine before any 'night' or 'PM' product goes in the cart. Second, when a new over-the-counter product enters the house, mention it at the next appointment.
The prescriber cannot adjust for a drug they do not know about, and patients consistently underestimate how interesting an allergy tablet is to a clinician.
The Review
Bring it back to the cabinet. The exercise takes one evening: every bottle and box in the house, prescription and not, onto one table. For each item, write down what it is for and when it was started. Anything whose purpose has been forgotten goes on the questions page.
Then the list goes to a pharmacist or prescriber, with one explicit question — of everything here, which items block acetylcholine, and are they all still earning their place?
A good review ends with fewer items more often than it ends with more. Some drugs stay, and should — the calculus in the Suwen was never zero medicine, it was the right medicine, stopped at the right fraction of the course. Some get swapped for a lower-burden alternative. Some get a taper plan with a date on it. What changes most is not the shelf but the oversight: the stack now has a reader.
Whether clearing the burden preserves cognition over the long run is the open question the trials have not yet answered, and it may take a decade of deprescribing research to settle. In the meantime the cost of the review is one evening and one awkward conversation.
Against a risk that, if causal, would account for roughly one in ten of the dementia diagnoses in a large English dataset, that exchange rate speaks for itself.
The last word belongs, as it often does, to a question the classics would ask without hesitation: what are you still taking that nobody has looked at in years?
"The stack now has a reader. That, before anything else, is the intervention."