What Sleep Medicine Guidelines Actually Say About Cannabis, and About the Pills It Gets Compared To

The American Academy of Sleep Medicine’s 2017 clinical practice guideline for the pharmacologic treatment of chronic insomnia issued fourteen recommendations, every one of them graded WEAK. Eight agents were suggested for use. Six were suggested against, including two of the most widely taken sleep aids in the United States: trazodone and diphenhydramine.

Cannabis appears nowhere in that guideline. It was not recommended, and it was not recommended against, because the evidence base was too thin to evaluate. Separately, the same organization issued a position statement that medical cannabis should not be used for obstructive sleep apnea.

The comparison that circulates in cannabis marketing, and in a fair amount of well-meaning patient education, sets cannabis against Ambien, Xanax, and Benadryl and declares cannabis the safer choice. That framing quietly assumes the comparison drugs are a reasonable standard of care. For several of them, sleep medicine says they are not.

A clinician who accepts the framing ends up defending cannabis against a straw man. The more useful conversation starts somewhere else: with what is causing the insomnia, whether cognitive behavioral therapy has been offered, and only then with what, if anything, should be taken at night.

The American College of Physicians made the first move in 2016. In Annals of Internal Medicine, Qaseem and colleagues issued a STRONG recommendation, on moderate-quality evidence, that every adult with chronic insomnia disorder receive cognitive behavioral therapy for insomnia as initial treatment. Medication was addressed separately and weakly, as a shared decision for patients in whom behavioral therapy alone had not worked.

The American Academy of Sleep Medicine followed in 2017 with a drug-by-drug guideline in the Journal of Clinical Sleep Medicine. Sateia and colleagues used GRADE methodology and suggested use of suvorexant, eszopiclone, zaleplon, zolpidem, triazolam, temazepam, ramelteon, and doxepin, each for specified insomnia patterns and each as a WEAK recommendation. The task force then suggested clinicians not use trazodone, tiagabine, diphenhydramine, melatonin, tryptophan, or valerian.

Two features of that list matter for the cannabis conversation. First, every recommendation was weak, including the favorable ones, because the trial evidence for hypnotics is smaller and more industry-funded than most patients assume. Second, two agents that patients commonly reach for, trazodone and diphenhydramine, were specifically recommended against for chronic insomnia.

Cannabis, THC, CBD, nabilone, and dronabinol are not among the agents the 2017 AASM guideline evaluated. That absence is not a verdict. It reflects a literature that, at the time, contained too few adequately designed randomized trials in diagnosed insomnia disorder to support a GRADE assessment in either direction.

Where the same organization did take a position, the position was restrictive. In a 2018 position statement in the Journal of Clinical Sleep Medicine, Ramar and colleagues concluded that medical cannabis and its synthetic extracts should not be used for the treatment of obstructive sleep apnea, citing unreliable delivery methods and insufficient evidence on effectiveness, tolerability, and safety. The statement went further and recommended that obstructive sleep apnea be excluded from the list of qualifying conditions in state medical cannabis programs.

This distinction gets flattened constantly. Sleep apnea is not insomnia. A patient who snores heavily, wakes unrefreshed, and has been told cannabis will help their sleep may in fact have an undiagnosed breathing disorder that cannabis does not treat and that a sedating agent could plausibly worsen. Sorting that out before prescribing anything is the actual clinical priority.

The strongest single trial is small. Walsh and colleagues published a randomized, double-blind, placebo-controlled crossover study in Sleep in 2021, testing two weeks of nightly sublingual cannabinoid extract in 24 adults with chronic insomnia symptoms. Compared with placebo, the extract lowered Insomnia Severity Index scores by 5.07 points, increased self-reported total sleep time by roughly 65 minutes, and improved actigraphy-derived sleep efficiency. Forty mild adverse events were reported, 36 of them during active treatment.

That result is real and it is limited. Twenty-four participants, two weeks, one proprietary formulation. A systematic review and meta-analysis by Bhagavan and colleagues in CNS Drugs in 2020 located only five studies with 219 participants total across the entire field and judged every one of them poor quality, concluding that the evidence does not reliably inform practice.

The signal that has held up best is not for THC. A 2023 randomized placebo-controlled trial of 293 adults with poor self-rated sleep found that 20 mg of cannabinol nightly reduced the number of nighttime awakenings and overall sleep disturbance, with the effect on global sleep quality falling short of significance. Adding CBD did not improve on cannabinol alone. The study was conducted by employees of a cannabis company, which is a relevant limitation rather than a disqualifying one. A 2026 systematic review reached a compatible conclusion: formulations containing THC did not show significant sleep improvement and carried more adverse effects.

Diphenhydramine. Recommended against by AASM for chronic insomnia. It is also a strong anticholinergic, and a prospective cohort of 3,434 older adults published in JAMA Internal Medicine in 2015 found a dose-response relationship between cumulative strong anticholinergic exposure and incident dementia, with a hazard ratio of 1.54 at the highest exposure category. First-generation antihistamines were among the three most commonly used classes in that cohort. This is the one comparison where the popular framing is broadly correct: nightly diphenhydramine is a poor long-term choice.

Zolpidem and the other Z-drugs. Weakly recommended by AASM for the specific insomnia patterns studied. The complex sleep behavior risk that patients hear about is real enough that the FDA added a boxed warning, but the guideline still places these agents in the suggested-use column, which cannabis has not earned.

Benzodiazepines. Temazepam and triazolam are weakly suggested for defined short-term use. The familiar concerns about tolerance, dependence, falls, and cognitive effects in older adults are well documented and are the reason these are not first-line for most patients.

Trazodone. Widely prescribed off-label for sleep and specifically recommended against by AASM. A patient told to prefer cannabis over trazodone is being offered a choice between an agent the guideline rejects and an agent the guideline never assessed.

Every argument against long-term hypnotic use applies to cannabis with roughly equal force, and the withdrawal data are not hypothetical. Bolla and colleagues recorded polysomnography in 18 heavy cannabis users during abrupt abstinence and documented falling total sleep time, falling sleep efficiency, and rising wake after sleep onset across two weeks, with periodic limb movements increasing in proportion to prior use.

Vandrey and colleagues went further in a crossover study of 20 daily users, showing that abstinence degraded objectively measured sleep architecture and that extended-release zolpidem attenuated the disruption. Read in one direction, that is a treatment finding. Read in the other, it is a clear demonstration that nightly cannabis produces a sleep pattern that depends on continued use.

This is the part of the comparison that marketing leaves out. Cannabis is not exempt from the tolerance-and-rebound problem that makes chronic hypnotic use unattractive. A patient who has taken an edible every night for two years and cannot sleep without one is in a recognizable situation, whatever the substance.

The guideline evidence is strong in structure and modest in content. GRADE methodology applied by an expert task force to the available randomized trials is the right process, and the fact that every recommendation came out weak tells you something honest about hypnotic research generally rather than something disparaging about the panel.

The cannabinoid evidence is weaker still. One adequately controlled two-week crossover trial in 24 people, a handful of poor-quality studies, and two industry-conducted trials of cannabinol do not add up to a body of evidence that could support a guideline recommendation. Anyone claiming otherwise is arguing past the literature.

The 2017 AASM guideline is now nine years old, and its silence on cannabis reflects the literature as it stood then. Several trials have appeared since. None of them individually would change a GRADE assessment, but the absence of cannabis from that document should be read as a snapshot rather than a permanent judgment.

The cannabinol findings deserve a specific caution. The 293-participant trial was designed, conducted, and authored by employees of a cannabis company, and its primary endpoint of global sleep quality did not reach significance. The secondary findings on awakenings and sleep disturbance are worth taking seriously and worth replicating independently before they become a clinical recommendation.

None of this shows that cannabis is safer than zolpidem, eszopiclone, or temazepam for sleep. No head-to-head randomized trial of cannabis against a guideline-recommended hypnotic in diagnosed chronic insomnia exists. The single published comparison in this space, nabilone versus amitriptyline, was conducted in fibromyalgia rather than insomnia disorder.

It also does not show that cannabis preserves sleep architecture better than prescription agents, that it avoids next-day impairment, or that it lacks a dependence liability at nightly doses. Those are claims the literature has not tested at the standard required to make them.

The practical consequence of the guideline structure is that the first question is not pharmacologic at all. Cognitive behavioral therapy for insomnia carries the only strong recommendation in this entire field, and it is the one intervention with durable effects after treatment stops. Most patients asking about cannabis for sleep have never been offered it.

The second question is diagnostic. Persistent insomnia is frequently a symptom of something else, including sleep apnea, restless legs, thyroid disease, pain, depression, or medication timing. A sedating agent layered on top of an unidentified cause tends to blunt the signal without fixing the problem, which is a general principle rather than a cannabis-specific one.

When a patient tells me cannabis is safer than Ambien, I ask what they are comparing it to and why they are taking either. Most of the time the honest answer is that nobody has worked out why they cannot sleep, and a substance is standing in for a diagnosis. That is the conversation worth having, and it is more useful than a scorecard.

I will say what I think the evidence supports. For a patient with situational sleep difficulty, a low nightly cannabinoid dose is a defensible short-term option, and cannabinol looks more promising than THC on the data we have. For a patient who has taken an edible every night for three years, the relevant question is no longer whether cannabis works. It is what happens on the night they run out, and whether anyone has looked for an underlying cause. I would rather find the cause.

Sleep medicine guidelines recommend cognitive behavioral therapy first, weakly support eight specific medications, weakly recommend against six others including diphenhydramine and trazodone, and do not evaluate cannabis at all. Cannabis is not the safe alternative to a good option; it is an unevaluated option next to a set of mediocre ones. Find out why the patient is not sleeping before choosing anything.

The finding to carry forward is that the popular comparison is rigged by its choice of comparator. Cannabis usually gets measured against the two agents sleep medicine explicitly rejects. Measured against the agents sleep medicine weakly supports, or against cognitive behavioral therapy, it has not been measured at all.