Doxepin

Silenor (low dose)

Doxepin (Silenor (low dose)) is a cognitive & mood research compound. At LOW dose it's a near-pure H1 histamine antagonist, which is what quiets nocturnal wakefulness. At the antidepressant doses (75–300mg) it becomes a messy tricyclic with anticholinergic effects. The dose changes the drug.

How Doxepin works

At LOW dose it's a near-pure H1 histamine antagonist, which is what quiets nocturnal wakefulness. At the antidepressant doses (75–300mg) it becomes a messy tricyclic with anticholinergic effects. The dose changes the drug.

Proposed benefits

Researched for focus, memory, neuroprotection, mood and stress resilience.

Where to get Doxepin

The evidence for Doxepin

Graded by what exists behind each claim.

✅ Clinically validated

• Approved twice at very different doses — as a tricyclic antidepressant at 75–300 mg, and as Silenor at 3–6 mg for sleep maintenance insomnia, with randomized trials supporting both.
• The low-dose insomnia data is notably good: improved sleep maintenance without next-day impairment or dependence, which is unusual in that class.

📊 Correlative data

• Long clinical use. The reported experience splits cleanly by dose — low doses are sedating and clean, antidepressant doses bring the full tricyclic anticholinergic burden.

🧪 Theoretical / extrapolated

• At low doses it is an almost pure H1 histamine antagonist — one of the most potent known. At higher doses it additionally inhibits serotonin and noradrenaline reuptake and blocks muscarinic and alpha-adrenergic receptors.
• The dose determines which drug you are taking. That is the cleanest example of receptor-affinity ordering in this Vault: selectivity at low concentration, promiscuity at high.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What Doxepin actually does

Doxepin at 6 mg and doxepin at 150 mg are not the same drug, and the number that makes that true is published. Richelson 1984 measured equilibrium dissociation constants for antidepressants at receptors in normal human brain tissue obtained at autopsy — histamine H1, muscarinic acetylcholine, α1- and α2-adrenergic, and dopamine D2. Doxepin came out as the most potent H1 antagonist in the series, with KD = 0.24 nM, against fluvoxamine at the other end of the same table at 109 µM. That is a span of roughly a million-fold inside one drug class, at one receptor. The Silenor label says the same thing in one sentence and names nothing else: “Doxepin binds with high affinity to the histamine H1 receptor (Ki < 1 nM) where it functions as an antagonist” U.S. Food and Drug Administration 2010.

Here is the arithmetic that turns that affinity into a dose, done out loud from the label’s own numbers. The mean apparent volume of distribution after a single 6 mg oral dose is 11,930 liters U.S. Food and Drug Administration 2010. Six milligrams spread through 11,930 liters is about 0.5 ng/mL, and doxepin’s molecular weight of about 279 makes that roughly 1.8 nM total. Plasma protein binding is approximately 80% U.S. Food and Drug Administration 2010, so the free concentration is on the order of 0.4 nM. Against an H1 KD of 0.24 nM, occupancy — concentration divided by concentration plus KD — lands somewhere near 60%. That is an average rather than a peak and it ignores bioavailability, so treat it as an order-of-magnitude calculation and not a measurement. But the conclusion it supports is robust to all of that: a few milligrams is enough to occupy most H1 receptors and nowhere near enough to touch anything else. The label states the consequence in its own language — doxepin is an antidepressant at 10- to 100-fold higher doses than Silenor U.S. Food and Drug Administration 2010.

What the other receptors are, and why they only turn on higher. At 75–300 mg doxepin is an ordinary tricyclic: serotonin and noradrenaline reuptake inhibition, muscarinic blockade (dry mouth, constipation, urinary retention, cognitive fog), α1-adrenergic blockade (orthostatic hypotension) and sodium channel effects (the cardiac risk in overdose). Richelson 1984 measured the receptor half of that list in the same human brain preparation — H1, muscarinic, α1, α2 and D2 — and the muscarinic and adrenergic affinities are orders of magnitude weaker than the 0.24 nM at H1, so at 25× the dose you are not getting “more sleep drug”. You are getting a different pharmacology bolted onto a receptor that was already saturated.

And blocking H1 is a structurally different way to produce sleep than every GABA drug in the same drawer. The histaminergic neurons of the tuberomammillary nucleus fire to sustain wakefulness, tonically, through the night. A benzodiazepine or a Z-drug adds inhibition at the GABA-A receptor — it pushes the brain down. An H1 antagonist removes a wake-promoting drive — it stops the brain being held up. That is not a rhetorical distinction: it predicts the drug should act on staying asleep rather than falling asleep, and it predicts no tolerance, no dependence and no rebound, because nothing is being potentiated. The label reports exactly that: no evidence of rebound insomnia, no symptoms indicative of a withdrawal syndrome, and not associated with abuse potential U.S. Food and Drug Administration 2010.

Cell, rodent, human — and where it stops

Step one, human receptors, no living subject. Radioligand binding on post-mortem human brain gives the affinity table the whole argument rests on Richelson 1984. Human tissue, which matters here — the same paper notes antidepressants had lower affinities at human receptors than in animal preparations, so a rodent number would have overstated the case.

Step two, humans, oral, with a sleep laboratory attached. Scharf 2008 ran a randomized, double-blind, placebo-controlled crossover of doxepin 1 mg, 3 mg and 6 mg in 76 elderly patients with primary insomnia. Wake time after sleep onset, total sleep time and sleep efficiency all improved at all three doses, each at p < 0.0001. The dose-response detail is the useful part: sleep efficiency improved significantly across all thirds of the night only at 3 mg and 6 mg. Even 1 mg moved the objective measures — which is the strongest possible statement that this is a high-affinity, low-occupancy-requirement mechanism.

Step three, human pharmacokinetics. Peak concentration at 3.5 hours fasted, terminal half-life 15.3 hours for doxepin and 31 hours for the active metabolite nordoxepin, clearance through CYP2C19 and CYP2D6 primarily with CYP1A2 and CYP2C9 secondary, and 80% protein binding U.S. Food and Drug Administration 2010.

Step four, the independent synthesis, and it is more sober than the trials. Everitt 2018 is a Cochrane review of antidepressants for insomnia covering 23 randomized trials and 2,806 participants, five of them using doxepin. Its conclusion is carefully hedged: there may be a small improvement in sleep quality with short-term use of low-dose doxepin compared with placebo, with four studies (518 participants) showing moderate improvement in subjective sleep quality; and the tolerability and safety of antidepressants for insomnia is uncertain because of limited reporting of adverse events. Read the two steps together: the polysomnogram separates cleanly at p < 0.0001 and the pooled subjective benefit is described as small and low-certainty. Both are true.

The obstacle, stated exactly. The endpoint that moves is wake after sleep onset — minutes of electroencephalographic wakefulness — and a minute recovered on a polysomnogram is not the same thing as a night that felt different. That gap is why Cochrane’s language is weaker than Scharf’s p-values. The second obstacle is duration: the trials are weeks, Cochrane is explicit that the evidence is short-term, and insomnia is a condition people treat for years. Nobody has published a controlled trial of low-dose doxepin past a few months, and nobody has run it against a Z-drug with next-day function as the primary endpoint — which is the comparison the whole argument for this drug depends on.

Doxepin pharmacokinetics — how much of it actually gets in

Route: oral, once at bedtime, and the food rule on the label is not decoration. A high-fat meal raises AUC by 41% and Cmax by 15% and delays Tmax by about 3 hours U.S. Food and Drug Administration 2010 — which is why the label says to take it within 30 minutes of bedtime and not within 3 hours of a meal. Taken after a late dinner, a 6 mg tablet behaves like a larger dose peaking several hours later, and that is a mechanistic account of next-morning grogginess rather than an idiosyncrasy.

What degrades it. CYP2C19 and CYP2D6 are the primary routes, with CYP1A2 and CYP2C9 secondary U.S. Food and Drug Administration 2010. The two interactions the label quantifies show how much that matters: cimetidine roughly doubles both Cmax and AUC, and sertraline raises mean AUC by about 21% and Cmax by about 32%. A doubling of exposure on a drug whose entire selling point is that the dose is small enough to hit only one receptor is not a trivial interaction — it is the mechanism of turning a sleep drug back into a tricyclic.

The half-life arithmetic, and why it does not mean what people assume. Doxepin’s terminal half-life is 15.3 hours and nordoxepin’s is 31 hours U.S. Food and Drug Administration 2010. Across an eight-hour night that is barely half of one half-life, so roughly 70% of the parent and almost all of the metabolite are still present at wake-up. The drug does not clear before morning, and it was never supposed to. What limits daytime effect is not elimination but concentration: at 3–6 mg the level is high enough to hold H1 and too low to engage anything else, so the residue is pharmacologically quiet. That also explains why raising the dose is disproportionately costly — you keep the long tail and add the receptors.

The oral barrier, and what is not available. Absorption is adequate by mouth and the label reports dose-proportional peak concentrations, with a mean apparent volume of distribution of 11,930 liters after 6 mg U.S. Food and Drug Administration 2010 — an enormous number that says the drug is overwhelmingly in tissue rather than in plasma, which is what a lipophilic amine does. There is no injectable, sublingual or transdermal low-dose doxepin with published pharmacokinetics, so no route bypasses first-pass metabolism and no route escapes the CYP2C19/2D6 interaction surface.

What would have to be true, and how you would know it was not

Four predictions. The first two are how you would know it is working, the third is the one that argues against escalating the dose, and the fourth is the one that argues against the whole framing.

1. The signal is wake time in the second half of the night, and it needs a written baseline. Keep a sleep diary, or run actigraphy, for 14 nights before the first dose, logging lights-out, estimated minutes awake after sleep onset, and number of awakenings. Then 14 nights on. The endpoint that moved in the laboratory was wake after sleep onset, and sleep efficiency improved across all thirds of the night only at 3 mg and above Scharf 2008. Without the pre-written baseline the effect size here — which Cochrane calls small Everitt 2018 — is below what memory can resolve.

2. It should work on the first night, and if it takes two weeks it is not this mechanism. H1 antagonism is receptor occupancy, not adaptation. There is no downregulation step, no delayed neuroplasticity, nothing that needs to accumulate — unlike the same molecule’s antidepressant effect, which takes weeks. So the falsifiable claim is that night one should look like night fourteen. A gradual onset over a fortnight is the signature of expectation or of a changed routine, not of a 0.24 nM H1 antagonist Richelson 1984.

3. The prediction that cuts against escalating: above 6 mg you buy anticholinergic burden, not sleep. Scharf 2008 found 1 mg, 3 mg and 6 mg all significant at p < 0.0001 on the objective measures — the curve is close to flat over a sixfold range because the receptor is nearly saturated. Doxepin is also sold as 10, 25, 50, 75, 100 and 150 mg capsules and as an oral concentrate, so moving up is trivially easy. The measurable consequence, if somebody does it, is a Trail Making Test taken at the same hour each morning: the prediction is that it is unchanged from baseline at 3–6 mg and gets worse at tricyclic doses, because muscarinic blockade is a known cause of next-day cognitive slowing and H1 blockade at low occupancy is not.

4. The prediction that cuts against the drug: if the effect is real it should survive being measured, and much of the reported benefit will not. Cochrane found low-certainty evidence of a small improvement and flagged that safety reporting was inadequate Everitt 2018. So run the honest version: 14 nights baseline, 14 nights on, then stop for 7 nights. The label reports no rebound insomnia and no withdrawal syndrome U.S. Food and Drug Administration 2010, which means the stop week is safe to run and interpretable — if sleep after stopping is no worse than the original baseline, and the on-drug fortnight was not clearly better than either, the drug is not doing the work. A GABA drug cannot be tested this way, and that is the one genuine practical advantage of this mechanism.

What nobody has tested yet

Nobody has published H1 receptor occupancy at 3 or 6 mg in a living person. Everything on this page argues from an affinity measured in autopsy tissue Richelson 1984 plus arithmetic on a volume of distribution U.S. Food and Drug Administration 2010. Receptor occupancy in a living brain is measurable by positron emission tomography with an H1 radioligand, at a handful of doses, in a handful of people. That single experiment would convert the central claim of this drug — that a few milligrams saturates H1 and touches nothing else — from inference into measurement, and it has not been done at the insomnia doses.

Nobody has genotyped CYP2C19 in a low-dose doxepin trial. The label names CYP2C19 as a primary clearance route and shows that cimetidine doubles exposure U.S. Food and Drug Administration 2010. A poor metabolizer is therefore taking, in effect, a larger dose — which on a drug sold specifically because the dose is small is the whole ballgame. Neither Scharf 2008 nor the trials pooled by Everitt 2018 reported genotype. A trial that stratified by CYP2C19 and 2D6 phenotype and looked for anticholinergic symptoms in the slow clearers would answer it.

Nobody has run it against a Z-drug on next-day performance. The argument for this molecule is not that it is a stronger hypnotic — it plainly is not — but that it carries no dependence, no tolerance and no rebound U.S. Food and Drug Administration 2010. That claim is only worth anything against a comparator. A randomized crossover against zolpidem with Trail Making and a driving simulator the following morning, plus a discontinuation week at the end, is an entirely feasible trial that nobody has published.

And nobody knows what a year of it does. Cochrane is explicit that the evidence base is short-term Everitt 2018, and insomnia is chronic. Whether H1 receptors upregulate over months — the mechanism by which tolerance would appear if it were going to — has never been looked for in people taking this dose.

Doxepin — its own safety story, not its class's

The contraindications, and the reason two of them look oversized for a 6 mg tablet. Silenor is contraindicated with a monoamine oxidase inhibitor currently or within the past two weeks, in untreated narrow-angle glaucoma, and in severe urinary retention U.S. Food and Drug Administration 2010. The last two are anticholinergic contraindications, and at 3–6 mg the muscarinic occupancy is negligible — they are inherited from the 150 mg drug and kept out of caution. Worth knowing which of the warnings on this label describe the tablet in front of you and which describe its parent.

The adverse-event table is unusually small, and that is the finding. At the doses studied the only reaction with a clear dose signal was somnolence or sedation: 4% on placebo, 6% at 3 mg, 9% at 6 mg. Upper respiratory infection, nausea and hypertension appeared at 2–4% with no dose pattern U.S. Food and Drug Administration 2010. For comparison the same molecule at antidepressant doses carries dry mouth, constipation, urinary retention, orthostatic hypotension and cardiac conduction effects in overdose. Cochrane’s caveat still stands over all of it: tolerability and safety are uncertain because adverse events were poorly reported across this literature Everitt 2018.

This drug’s own biggest risk is a dosing error, and it is specific to it. Doxepin exists as 3 mg and 6 mg sleep tablets and as 10, 25, 50, 75, 100 and 150 mg capsules and as an oral concentrate. Nothing else in the sleep drawer has a twenty-five-fold strength range under one name. Somebody who runs out of the small tablets and substitutes a capsule from a different prescription has taken 4 to 25 times the intended dose, and the label itself frames the antidepressant range as 10- to 100-fold higher U.S. Food and Drug Administration 2010. And because cimetidine roughly doubles exposure and sertraline raises it by a fifth to a third U.S. Food and Drug Administration 2010, the same conversion can happen without changing the tablet at all.

What it does not carry, stated plainly because it is the reason people choose it. No GABA-A activity, therefore no rebound insomnia, no withdrawal syndrome and no abuse potential in the label’s own words U.S. Food and Drug Administration 2010. It also has no interaction with alcohol at the GABA receptor, which does not make the combination sensible — two sedating drugs are still two sedating drugs — but does mean the specific respiratory risk of stacking a benzodiazepine on alcohol is not the risk here — and the label is unambiguous anyway: patients should not consume alcohol with Silenor, with additive effects also flagged for other CNS depressants and sedating antihistamines U.S. Food and Drug Administration 2010. Two warnings on it belong to the hypnotic class rather than to this molecule and are worth naming: 5.2 Abnormal Thinking and Behavioral Changes, which cites complex behaviors such as sleep-driving reported with hypnotics, and 5.3 Suicide Risk and Worsening of Depression, which records that in primarily depressed patients, worsening of depression including suicidal thoughts and actions has been reported in association with the use of hypnotics U.S. Food and Drug Administration 2010. There is no boxed warning on this label.

Sources read for this page

• U.S. Food and Drug Administration. SILENOR (doxepin) tablets, for oral use - full prescribing information (NDA 022036).. FDA approved labeling, revision 2010
• Richelson E. Antagonism by antidepressants of neurotransmitter receptors of normal human brain in vitro.. J Pharmacol Exp Ther 1984 · PMID 6086881
• Scharf M. Efficacy and safety of doxepin 1 mg, 3 mg, and 6 mg in elderly patients with primary insomnia: a randomized, double-blind, placebo-controlled crossover study.. J Clin Psychiatry 2008 · PMID 19192438
• Everitt H. Antidepressants for insomnia in adults.. Cochrane Database Syst Rev 2018 · PMID 29761479

Doxepin — safety, predicted from mechanism

Doxepin — interference & stacking

Bloodwork to run alongside Doxepin

Baseline first, then again at 8–12 weeks.

The Brain Fog & Cognition panel covers these in one order — 12 markers, $233.06 with the discount applied.

Check results you already have → · All 103 markers A–Z

Doxepin — frequently asked questions

What Doxepin is used for

Doxepin appears under 2 goals in the goal router.

Related Cognitive & Mood compounds

← Explore the full Protocol Vault

↑ Back to on this page