What Benzodiazepines Actually Do — And Why Stopping Is the Risky Part

If you are currently taking a benzodiazepine, do not stop or reduce it based on anything in this article. Abrupt discontinuation is the one genuinely dangerous thing you can do with this class of drug, and the published cases of withdrawal delirium below are what that looks like. Any change belongs to you and your prescriber, together, on a plan.

Benzodiazepines are among the most effective drugs ever made for acute anxiety and acute insomnia, and among the most difficult to stop. Both halves of that are true at once, which is why the public conversation about them tends to collapse into either "miracle" or "scandal." This article is what the evidence actually shows — including the places where the evidence is weaker or more surprising than the standard warnings imply. We sell sleep supplements, one of which acts at the same receptor site, so the last section is about that.

What a benzodiazepine does mechanically

GABA is the brain's main inhibitory neurotransmitter — the brake pedal. When GABA binds the GABAA receptor, a channel opens, chloride ions flow into the neuron, and the neuron becomes harder to fire.

A benzodiazepine does not open that channel itself. It binds a separate site on the same receptor and makes the receptor more responsive to the GABA already there — the technical term is positive allosteric modulation. Diazepam, lorazepam, alprazolam, clonazepam, temazepam and chlordiazepoxide all work this way; they differ mainly in how fast they arrive, how long they last, and which receptor subtypes they favour.

Two consequences follow from that mechanism, and they explain nearly everything else on this page. First, because the drug amplifies an existing brake rather than creating a new one, it works fast — within an hour, often less. There is no multi-week wait for an effect, which is exactly why it feels so unlike an antidepressant. Second, because the brain regulates its own receptors, sustained amplification invites the system to adapt — fewer or less sensitive receptors, so the same dose does less. That adaptation is tolerance, and the same adaptation running in reverse, with the drug removed, is withdrawal.

The "Z-drugs" are not a different thing

Zopiclone, zolpidem and zaleplon are marketed as a separate, gentler category. Chemically they are distinct from benzodiazepines, but they act on the same GABAA receptor complex, at the same site, as positive modulators — which is why the research literature usually calls them "benzodiazepine receptor agonists" and studies them alongside benzodiazepines rather than separately — the 2026 deprescribing review discussed below treats both as one problem in a single analysis.3 They are more selective for certain receptor subtypes, which plausibly explains a somewhat different side-effect profile. They are not a mechanistically different answer, and the dependence and withdrawal literature treats them as part of the same problem.

What the long-term evidence actually shows — including one surprise

Guidelines across jurisdictions converge on short-term use, typically two to four weeks. The evidence base for what happens beyond that is thinner than the confidence of those guidelines suggests. A 2026 systematic review and network meta-analysis in the Journal of Sleep Research looked specifically at use beyond three months in adults with chronic insomnia, covering 27 studies published between 1987 and 2023.1 Five outcomes had enough data to pool: insomnia severity, sleep quality, falls, depression and anxiety. What it found:

  • People on long-term benzodiazepines had more severe insomnia than healthy sleepers — but less severe than people with insomnia taking nothing. Read carefully, that is neither a vindication nor an indictment; it is consistent with the drugs still doing something, and with the people taking them having been worse off to begin with.
  • They had significantly higher levels of depression and anxiety than healthy sleepers.
  • Fall risk was similar between users and non-users. This one runs against the standard warning, and it is worth sitting with rather than skipping past. It does not mean benzodiazepines don't cause falls — the comparison here is within an insomnia population, not against the general public, and the review's own headline is how few long-term studies exist.

The authors' conclusion is that extended use may worsen sleep quality, mental health and quality of life, and that the guidance to limit duration and prioritise behavioural therapy stands. The honest gloss: the strongest argument against long-term use is not a dramatic measured harm. It is that the benefit erodes, the drug becomes hard to leave, and the research needed to say more than that has not been done.

Stopping them, and why "just do CBT instead" is harder than it sounds

Cognitive behavioural therapy for insomnia (CBT-I) is the standard recommendation for getting off sleeping tablets. The evidence for that has moved in an uncomfortable direction, and both studies deserve printing.

In 2019 a meta-analysis in Sleep Medicine Reviews pooled eight randomised trials and found CBT-I plus gradual tapering clearly beat tapering alone for discontinuing benzodiazepine hypnotics in the short term — risk ratio 1.68 (95% CI 1.19–2.39, p = 0.003) — along with improved insomnia symptoms. At twelve months, the discontinuation advantage was no longer significant (RR 1.67, 95% CI 0.91–3.07, p = 0.10).2

A 2026 review in the Journal of International Medical Research revisited the question with thirteen randomised trials and tighter methodology — GRADE certainty ratings, RoB 2, and a deliberate refusal to pool outcomes measured on incompatible thresholds.3 For early complete discontinuation it found a risk ratio of 1.35 (95% CI 0.87–2.11) across six trials and 298 participants: pointing the right way, but no longer statistically significant. At longest follow-up the estimate was 1.64 with a confidence interval running from 0.45 to 5.89 — which is to say, unusable. Dropout showed no clear difference. Insomnia severity scores did improve (mean difference −4.73, 95% CI −8.46 to −0.99), but from only three small trials. Its conclusion is explicit: there is no high-certainty evidence that CBT-I-assisted interventions reliably achieve complete discontinuation.

What to take from the pair: CBT-I is still the right thing to offer, it demonstrably helps people sleep and cope during a taper, and it is not the reliable exit from benzodiazepines that the 2019 figure implied. Anyone presenting deprescribing as a solved problem is overselling it. If you want what the behavioural evidence does support on its own terms, we covered it in CBT-I: the sleep treatment with better evidence than any supplement, and the wider ranking of what actually works is in sleep hygiene tips vs real treatment.

Withdrawal: the part that is genuinely dangerous

Most benzodiazepine withdrawal is unpleasant rather than hazardous — rebound insomnia and anxiety worse than the original, tremor, sweating, sensory sensitivity. The serious end is real, though, and it is worth being specific rather than vague about it.

A 2026 systematic review in the Journal of the Academy of Consultation-Liaison Psychiatry gathered every published case of benzodiazepine withdrawal delirium it could find — 24 cases described between 1977 and 2025.4 The pattern:

  • Ages 27 to 84 (mean 55.8), 63% male.
  • Alprazolam, lorazepam and clonazepam were the most common drugs beforehand, at a median of 8.3 mg daily in lorazepam-equivalent terms.
  • Prior exposure ranged from six weeks to fifty years. Six weeks. This is not exclusively a decades-of-use phenomenon.
  • Delirium began 12 hours to 13 days after the dose changed, and most often the change was a taper or abrupt stop carried out by an inpatient healthcare team — not by the patient.
  • Every case involved disorientation and cognitive dysfunction, frequently with agitation, hallucinations and persecutory delusions.
  • The treatment was restarting a benzodiazepine, which fully resolved symptoms in 22 of the 24.

Two honest caveats. These are case reports, so they establish that this happens and roughly what it looks like, not how often. And no case used a standardised delirium scale, which the authors flag. But the practical message is unambiguous and it is the reason for the warning at the top of this page: the danger with benzodiazepines is in stopping them badly, and that can happen at six weeks of use as well as fifty years.

The dementia question, and a finding that counts against us

You have probably seen the claim that benzodiazepines cause dementia. The careful evidence does not support it as confidently as the headlines did.

A 2025 systematic review and meta-analysis in Harvard Review of Psychiatry pooled five studies on chronic use and found a hazard ratio for all-cause dementia of 1.17 (95% CI 0.96–1.43) — not statistically significant. For Alzheimer's disease specifically, across three studies, the hazard ratio was 1.00 (95% CI 0.87–1.15): no association at all.5

A large South Korean cohort study illustrates why this literature keeps producing signals that shrink on inspection. Among 616,256 propensity-matched patients aged 50 and over, benzodiazepine users showed a 23% higher dementia risk than non-users (HR 1.23, 95% CI 1.14–1.32). When the authors extended the lag period — ignoring dementia diagnoses in the first seven years after starting the drug — the association fell towards null (HR 1.17, 95% CI 1.04–1.30).6 That lag matters because disturbed sleep and anxiety are early symptoms of neurodegenerative disease, often years before diagnosis. People in the earliest stages of dementia get prescribed sedatives because of the dementia. The drug then appears to have caused what prompted it.

Here is the part that counts against our own commercial interest, and we are printing it anyway. A 2026 study in Scientific Reports used the TriNetX network — 189,858 propensity-matched pairs — to compare people with sleep disorders newly prescribed melatonin against those prescribed benzodiazepines.7 Melatonin came out worse: all-cause dementia HR 2.09 (95% CI 2.01–2.16) versus benzodiazepines, and 1.74 versus zolpidem. Also elevated for vascular dementia, Alzheimer's and Parkinson's disease. The direction held across sensitivity analyses including lag times up to four years.

We sell melatonin, in two products. So read what the authors themselves say, because it is the correct reading: these findings should not be interpreted as evidence of a causal pharmacologic effect, and likely reflect residual confounding, confounding by indication, and prodromal neurodegenerative disease. Melatonin is what clinicians reach for in exactly the older patients whose sleep disturbance is itself an early neurological sign — REM sleep behaviour disorder being the clearest example, since it is strongly associated with later Parkinson's and Lewy body dementia. That is almost certainly what a hazard ratio of 2.09 is detecting.

But "almost certainly confounding" is not "nothing," and a supplement company quietly declining to mention a 2.09 on its own ingredient would be doing the thing this blog exists not to do. So: we have never claimed melatonin is neuroprotective, we are not going to start, and we checked the rest of this library before publishing this page to confirm no post of ours makes that claim. If you want the dose-response evidence on what melatonin actually does for sleep — which is real and modest — it is in the real dose-response evidence.

The combination that actually kills people

Benzodiazepines alone are difficult to fatally overdose on. Combined with an opioid, they are not, because both suppress breathing and the effect is synergistic.

A 2026 study in Academic Emergency Medicine analysed serum from patients arriving at ten emergency departments with confirmed opioid overdose, screening for over 1,200 substances.8 Among 1,427 patients with opioids present, 29.0% also had detectable benzodiazepines — 20.5% prescription, 8.5% "novel" (illicitly manufactured designer benzodiazepines that are not prescribed at all). Alprazolam was the most common prescription one; bromazolam accounted for 46.3% of the novel ones.

The severity finding needs stating precisely, because it is easy to garble. It was the novel benzodiazepines, not the prescription ones, that carried increased odds of needing mechanical ventilation compared with opioids alone — adjusted odds ratio 2.14 (95% CI 1.07–4.05) — and higher rates of not responding to naloxone. People with novel benzodiazepines were also younger (median age 34 versus 40–41). The prescription-benzodiazepine group did not show that elevated intubation risk in this analysis.

That nuance does not soften the clinical rule, which remains that combining a prescribed benzodiazepine with a prescribed opioid is a decision for a prescriber who knows both. It does tell you something about where the sharpest end of this risk currently sits: in an unregulated drug supply, not in a pharmacy.

Where our products come into this

Directly, as it happens, and more so than with most medication classes.

Valerian acts at the benzodiazepine binding site. Not a similar site — the same one. Laboratory work shows valerian extracts act at the GABAA receptor's benzodiazepine binding site, with a biphasic effect: enhancing binding at low concentrations, inhibiting it at higher ones.9 Valerian is in Sleep Formula, at a disclosed 150mg — and worth saying plainly, most valerian trials use more than that, so treat 150mg as a supporting amount rather than a studied one. It is not in Sleep Support.

Valerian's own measured sedative effect in people is modest. In a head-to-head trial before third-molar surgery it produced less drowsiness than midazolam,10 and a meta-analysis of 18 randomised placebo-controlled trials found it improved subjective sleep quality while having no measurable effect on how long people took to fall asleep.11 So the practical interaction risk is lower than the receptor story alone suggests. The mechanistic overlap is still real, and it is why we flag valerian rather than treat it as inert botanical filler. The drug-by-drug detail sits in our medication interactions guide, and the honest read on valerian itself is in valerian root for sleep.

Sleep Support contains other sedating botanicals, at undisclosed amounts. Its 14-ingredient blend includes hops, Chinese skullcap, ashwagandha, L-theanine, passion flower and lemon balm — several of which have a calming or sedative reputation — and not one of them carries an individual milligram amount on the label. Our manufacturer confirmed in September 2026 that those amounts cannot be published, and there is no timeline on that. If you are taking a benzodiazepine or a Z-drug, that is the honest problem with Sleep Support: neither you nor we can work out how much of each sedating ingredient you would be adding.

GABA itself is in all three of those products — disclosed at 100mg in Sleep Formula, and undisclosed inside the blends in Sleep Support and Cognitive Support — and whether oral GABA meaningfully reaches the receptors benzodiazepines act on is an open question, not a settled one. We wrote that up in what oral GABA supplements can and can't do. The honest position is that we cannot tell you it is safe to stack with a benzodiazepine, because the research to say either way does not exist.

And the thing we most want to say. The people most likely to read a page like this are people trying to get off a benzodiazepine, or trying to avoid starting one. If that is you: a sleep supplement is not a substitute for a tapering plan, and self-medicating a taper with sedating botanicals is a way to make an already difficult process harder to read — because when symptoms shift you will not know which thing moved. Tell your prescriber what you are taking. If you want the honest version of what our sleep products do and don't do, the Sleep Support breakdown and how the two compare are the places we are most direct about the limits.

The short version

  • Benzodiazepines make the GABAA receptor more responsive to the GABA already present. That is why they work in under an hour, and why the brain adapts.
  • Z-drugs act on the same receptor complex. They are not a mechanistically different category.
  • Long-term use is associated with worse depression and anxiety than healthy sleepers; in one 2026 review, fall risk was similar to non-users; and the biggest finding is how few long-term studies exist.
  • CBT-I helps with tapering but the 2026 evidence does not support it as a reliable route to complete discontinuation.
  • Withdrawal delirium is documented after as little as six weeks of use, usually following a taper or abrupt stop. Never stop without a plan.
  • The dementia link does not survive careful analysis — 1.17 (0.96–1.43) for dementia, 1.00 for Alzheimer's. Prodromal sleep disturbance confounds this whole literature, including for melatonin, where the observed hazard ratio against benzodiazepines is higher still and almost certainly for that reason.
  • With opioids, the risk is real and synergistic. The sharpest signal in current ED data comes from illicit designer benzodiazepines, not prescriptions.
  • Valerian acts at the benzodiazepine binding site, though its measured effect in people is modest. It is in Sleep Formula at 150mg, not in Sleep Support — whose sedating ingredients carry no disclosed doses at all. If you take a benzodiazepine or a Z-drug, our sleep products are a prescriber conversation, not a self-service decision.

The rest of this series

This is one of four explainers on the medication classes that come up most often alongside what we sell. Each one is the mechanism, the best current evidence, and the places where that evidence is weaker or stranger than the standard account — including where it counts against us.

  • What SSRIs actually do — the serotonin transporter, why the "chemical imbalance" account is disputed in print, and why published withdrawal estimates range from about 15% to 55%.
  • What ADHD stimulants actually do — dopamine and noradrenaline transporters, where efficacy stops climbing with dose, and the placebo study that should unsettle every focus supplement on the market.
  • What antipsychotics actually do — four dopamine pathways and the side effects that follow, tardive dyskinesia rates, and the one prescribed off-label as a sleeping tablet.

This article is for informational purposes only and has not been evaluated by the FDA. It is not intended to diagnose, treat, cure, or prevent any disease. We are not doctors, pharmacists or prescribers. Nothing here is a reason to start, stop, change or skip a prescribed medication. Benzodiazepine withdrawal can be medically serious, including seizures and delirium — any reduction should be planned and supervised by the clinician who prescribed it. If you take a benzodiazepine, a Z-drug, or an opioid, speak to your doctor or pharmacist before taking any supplement.

References

  1. Riemann D, Baglioni C, Nissen C, Nevière A, Irfan O, Le Nouveau P, Leontiou C, Léger D. "Risks Associated With Benzodiazepine Long-Term Use in Chronic Insomnia: A Systematic Review and (Network) Meta-Analysis." Journal of Sleep Research, 2026, 35(5):e70352. doi:10.1111/jsr.70352
  2. Takaesu Y, Utsumi T, Okajima I, Shimura A, Kotorii N, Kuriyama K, Yamashita H, Suzuki M, Watanabe N, Mishima K. "Psychosocial intervention for discontinuing benzodiazepine hypnotics in patients with chronic insomnia: A systematic review and meta-analysis." Sleep Medicine Reviews, 2019, 48:101214. doi:10.1016/j.smrv.2019.101214
  3. Zhang C, Yang L, Zhang J. "Cognitive behavioral therapy for insomnia-assisted discontinuation or reduction of benzodiazepine receptor agonists and Z-drugs in chronic insomnia: A systematic review and meta-analysis." Journal of International Medical Research, 2026, 54(9). doi:10.1177/03000605261487272
  4. Al-Soleiti M, Park JH, Kummerlowe MN, Kinzelman-Vesely E, Gerberi DJ, Philbrick KL, Fipps DC. "Clinical Characteristics and Treatment of Benzodiazepine Withdrawal Delirium in Adults: An Illustrative Case and Systematic Review of Published Cases." Journal of the Academy of Consultation-Liaison Psychiatry, 2026, 67(4):304–319. doi:10.1016/j.jaclp.2026.06.001
  5. Rivas J, Hernández M, Erazo JM, Martínez MJ, González C, Cortés MP, Muñoz J, Miranda C. "Chronic Use of Benzodiazepine in Older Adults and Its Relationship with Dementia: A Systematic Review and Meta-Analysis." Harvard Review of Psychiatry, 2025, 33(1):1–7. doi:10.1097/HRP.0000000000000414
  6. Baek YH, Lee H, Kim WJ, Chung JE, Pratt N, Kalisch Ellett L, Shin JY. "Uncertain Association Between Benzodiazepine Use and the Risk of Dementia: A Cohort Study." Journal of the American Medical Directors Association, 2020, 21(3):380–387. doi:10.1016/j.jamda.2019.08.017
  7. Chang HC, Su YJ, Yang SC, Chen CC, Wu MC, Gau SY. "Comparative dementia and Parkinson's disease risk associated with melatonin, benzodiazepine, or zolpidem use in patients with sleep disorders: a retrospective cohort study." Scientific Reports, 2026. doi:10.1038/s41598-026-58529-4
  8. Hughes A, Spungen H, Culbreth R, Aldy K, Krotulski A, Hendrickson RG, et al. "Benzodiazepine Co-Exposure Among Patients Presenting to the Emergency Department With a Confirmed Opioid Overdose." Academic Emergency Medicine, 2026, 33:e70104. doi:10.1111/acem.70104
  9. Ortiz JG, Nieves-Natal J, Chavez P. "Effects of Valeriana officinalis extracts on [3H]flunitrazepam binding, synaptosomal [3H]GABA uptake, and hippocampal [3H]GABA release." Neurochemical Research, 1999. doi:10.1023/a:1022576405534
  10. Farah GJ, Ferreira GZ, Danieletto-Zanna CF, Luppi CR, Jacomacci WP. "Assessment of Valeriana officinalis L. (valerian) for conscious sedation of patients during the extraction of impacted mandibular third molars: a randomized, split-mouth, double-blind, crossover study." Journal of Oral and Maxillofacial Surgery, 2019. doi:10.1016/j.joms.2019.05.003
  11. Fernández-San-Martín MI, et al. "Effectiveness of valerian on insomnia: a meta-analysis of randomized placebo-controlled trials." Sleep Medicine, 2010. doi:10.1016/j.sleep.2009.12.009