Treating Chronic Insomnia

Chronic insomnia (difficulty falling or staying asleep, ≥3 nights/week, ≥3 months) affects 10–15% of adults — about 12% on the best current estimate, from a meta-analysis restricted to random population samples assessed against diagnostic criteria.^[1] The treatment landscape has shifted: CBT-I is now first-line, ahead of any medication. Pharmacotherapy is reserved for cases where CBT-I is inadequate or unavailable.

CBT-I: the first-line treatment — Strong

Cognitive Behavioral Therapy for Insomnia (CBT-I) outperforms hypnotic drugs in head-to-head trials, with effects that persist after treatment ends — unlike medications, which lose effect when discontinued.

How large is the effect? The largest meta-analysis of the field pooled 87 randomised trials (3,724 treated patients versus 2,579 untreated controls) and found CBT-I — whether the full package or its individual components — reduced the Insomnia Severity Index with a standardised effect of g = 0.98, a large effect by conventional benchmarks. Sleep efficiency improved (g = 0.71), as did sleep quality on the Pittsburgh index (g = 0.65), wake after sleep onset (g = 0.63) and sleep-onset latency (g = 0.57). The one thing CBT-I barely changes is total sleep time (g = 0.16) — it consolidates sleep rather than lengthening it, which is why people often feel dramatically better while sleeping roughly the same number of hours.^[2] Face-to-face courses of at least four sessions outperformed self-help and shorter courses.

The 2017 American Academy of Sleep Medicine guideline^[3] and subsequent updates^[4] recommend CBT-I as the first-line treatment for chronic insomnia, ahead of any pharmacotherapy.

CBT-I has five core components:

1. Stimulus control — bed only for sleep and sex; if not asleep within ~20 minutes, get out of bed. Re-associates the bed with sleep, not wakefulness.
2. Sleep restriction therapy — temporarily limit time in bed to consolidate sleep. Counterintuitive, and the component with the strongest standalone evidence: delivered alone by primary-care nurses over four sessions, it cut insomnia severity scores from 13.9 to 10.9 at six months against a sleep-hygiene booklet.^[5] Daytime napping is typically eliminated during this phase because it borrows from the homeostatic pressure that should be driving consolidated night-time sleep — see Daytime naps.
3. Cognitive restructuring — challenges catastrophic thinking about sleep ("If I don't sleep 8 hours, tomorrow will be ruined").
4. Sleep hygiene — the well-known basics. In the component analysis it came out inert on its own, so treat it as a floor rather than a treatment: worth doing, not worth expecting anything from.
5. Relaxation training — for the sleep-anxiety subtype, and the one component that may cut the other way: in the same analysis its point estimate sat slightly below no component at all, which is a reminder that "more techniques" is not automatically better.^[6]

Which component does the most work is genuinely unsettled. Across 241 trials, cognitive restructuring came out on top, with sleep restriction and stimulus control indistinguishable from each other,^[7] while a separate analysis of 80 studies did single out sleep restriction.^[8]

Access to CBT-I

The bottleneck has historically been access — too few trained clinicians. Digital, self-guided CBT-I programs now make it available at scale, and several have been tested in large randomised trials: Sleepio (endorsed for NHS use in the UK), Somryst (FDA-cleared in the US), and SHUTi are the ones with the most published data.

The evidence that digital delivery works is solid rather than merely promotional. A meta-analysis of 11 randomised trials of internet-delivered CBT-I found improvements in insomnia severity, sleep efficiency, sleep quality, wake after sleep onset, sleep-onset latency and total sleep time, with effect sizes from 0.21 to 1.09 that were sustained at 4–48 weeks — a range its authors judged comparable to face-to-face CBT-I.^[9] The largest single trial randomised 1,711 adults with self-reported insomnia symptoms to a digital program or sleep hygiene education, and found large improvements in sleep itself with much smaller gains in daytime well-being and functioning.^[10] It was funded by the maker of the program tested, whose cofounder and chief medical officer is the trial's first author — worth knowing, given the products named above.

The strongest claim — that digital is genuinely as good as a therapist — is the one the evidence does not support. The single head-to-head trial set out to show non-inferiority and did not: among 101 patients referred to a Norwegian sleep clinic, digital came out significantly worse than face-to-face at both time points, by about 4.6 points of Insomnia Severity Index at nine weeks and 2.8 at thirty-three, and the trial's own authors call the non-inferiority question inconclusive because the study was too small to settle it.^[11] One small trial is not a settled literature either. The largest synthesis to date pooled 52 randomised trials across every delivery format, and found that none was statistically superior to any other, with the formats involving a live therapist — in person or over video — sitting at the top of the range and unguided self-help at the bottom.^{[12] [13]} Treat digital CBT-I as clearly effective and a reasonable first move when a therapist isn't available — not as a proven equal.

When pharmacotherapy is appropriate

The 2017 AASM guideline gave only weak (conditional) recommendations for any pharmacotherapy in chronic insomnia, reflecting modest efficacy and meaningful harms. Pharmacotherapy makes sense when:

• CBT-I has been adequately tried and failed
• Comorbid pain conditions are driving insomnia. (Comorbid depression is a weaker reason than it looks: treating the insomnia with CBT-I improves the depression too — across 19 randomised trials, roughly double the depression response rate of control — so it is a reason to deliver CBT-I, not to skip it.^[14])
• Acute, time-limited situations (bereavement, hospitalization, severe transient stress)
• Bridge therapy while CBT-I is being initiated

Specific medications, ranked by how much we know and how long you'd take them

A caution about any ranking, including this one. The largest randomised synthesis — 170 trials, 36 interventions, nearly 48,000 participants, pooling every drug below onto one scale — found sustained benefit beyond placebo for only two: eszopiclone, which sits in Tier 2 here, and lemborexant. It grouped daridorexant and suvorexant with the benzodiazepines on tolerability, and judged the long-term safety data for lemborexant inconclusive.^[15] So read the tiers below as ordered by how well studied a drug is and how long it is meant to be taken, not as a claim that the top of the list is proven safer. The newer drugs mostly lack the long-term harm data the older ones have accumulated — which is not the same as lacking the harms.

Tier 1: Most favorable risk-benefit (where pharmacotherapy is needed)

Dual orexin receptor antagonists (DORAs): suvorexant, lemborexant, daridorexant — Moderate^[16]

• Newer class targeting wakefulness rather than imposing sedation.
• A 2025 network meta-analysis of 8 randomised trials covering 5,198 adults found all three drugs beat placebo on falling asleep, staying asleep and insomnia severity — but the effects are small, in the range where a statistically clear result is not necessarily one a patient would notice, and the trials ran only one to six months.^[17]
• No signal of tolerance, withdrawal or rebound insomnia when the drug is stopped abruptly, in trials running up to six months — a real advantage over benzodiazepines and Z-drugs, and the main reason this class sits at the top of the list. One qualifier: like almost every prescription sleep drug, all three are Schedule IV controlled substances in the US, because in studies of people who use sedatives recreationally they produce measurable drug-liking. At the doses actually prescribed that liking was lower than for zolpidem, and postmarketing surveillance suggests the class carries less abuse risk than the older hypnotics.^[18] Lower risk than a benzodiazepine — not a drug with no controls on it at all.
• The safety trade-off is narcolepsy-like symptoms. A meta-analysis of 11 RCTs and 7,703 patients found DORAs roughly tripled the risk of sleep paralysis (RR 3.40, 95% CI 1.18–9.80) and roughly doubled excessive daytime sleepiness (RR 2.15, 95% CI 1.02–4.52) versus placebo, alongside a small (~9%) excess of treatment-emergent adverse events overall (RR 1.09).^[19] The bracketed 95% confidence interval is the range in which the true effect most plausibly lies; when it sits entirely above 1.0 the increase is real, and a lower bound that nearly touches 1.0 (as sleep paralysis's 1.18 does) means the estimate is imprecise. These are uncommon events — hence the wide intervals — but the direction is consistent.
• Now often preferred over Z-drugs when pharmacotherapy is needed — though it is worth being clear that this preference rests mostly on the absence of the harms Z-drugs are known for, rather than on evidence that this class is safer on hard outcomes. The fracture data are genuinely mixed: one propensity-matched cohort found more hip fractures on an orexin antagonist than on a Z-drug or benzodiazepine,^[20] while a meta-analysis across 46,636 people found no excess of falls or fractures.^[21] This is a newer class with less long-term data, not a proven safer one.

Low-dose doxepin (3–6 mg) — Weak-to-Moderate

• Sedating antidepressant at sub-antidepressant doses.
• A systematic review of randomised placebo-controlled trials found a small-to-medium effect on sleep maintenance and no significant next-day residual sedation, but noted the trials were few and mostly industry-sponsored, leaving short-term benefit-risk and withdrawal effects unsettled.^[22]
• Anticholinergic effects are dose-dependent and minimal at 3–6 mg — worth stating explicitly, because doxepin at higher doses appears in the very anticholinergic-exposure research that underpins the warning against antihistamines below. The dose is doing the work here.

Ramelteon — Weak

• A melatonin-receptor agonist, not a sedative; not a controlled substance.
• Weakly recommended by the sleep-medicine guideline for trouble falling asleep only — it has no recommendation for staying asleep, which makes it a narrow tool rather than a general option.^[23]

Tier 2: Use sparingly, time-limited

Z-drugs (zolpidem, zaleplon, eszopiclone) — Moderate efficacy; Caution (chronic use)

• Effective acutely but with substantial harms with chronic use; the sleep-medicine guideline recommends them only weakly, and only for short periods.^[24]
• Fracture risk is the main one. A systematic review and meta-analysis of observational studies put Z-drug use at a 90% higher relative risk of hip fracture (RR 1.90, 95% CI 1.68–2.13) — higher than benzodiazepines, not lower — with the risk more than doubled in the first weeks of a new prescription (RR 2.39).^[25] These are observational data, so confounding by indication (the people prescribed hypnotics are already frailer and sleeping worse) inflates the number, but the "new user" spike is hard to explain away.
• Limit to four weeks where possible — that is the explicit recommendation of the European insomnia guideline, which endorses these drugs for short-term use only.^[26]
• Complex sleep behaviours — sleep-driving, sleep-eating, sleep-walking — are rare, and serious enough that the US Food and Drug Administration put a boxed warning, its strongest, on zolpidem, eszopiclone and zaleplon, and made a previous episode an outright contraindication.^[27] Rare is not the same as minor: if it has happened once, the drug should not be restarted.
• Stopping is achievable. Across sixteen studies of patient-directed deprescribing — where the patient, not just the prescriber, gets the information and a tapering plan — between roughly a fifth and two-thirds of long-term users in their late sixties and older came off successfully, with more support producing better results. Most of those studies were at high risk of bias, so treat the range as encouraging rather than precise.^[28] A slow taper with clinician support, ideally alongside CBT-I, is the route.

Trazodone — Weak (for sleep efficacy)

• Widely prescribed off-label for sleep, and the efficacy data is genuinely thin: a meta-analysis of 7 placebo-controlled trials (429 patients) found no significant improvement in sleep efficiency, sleep-onset latency, total sleep time or wake after sleep onset. The only significant benefits were fewer early awakenings and better perceived sleep quality.^[29]
• Meaningful next-day sedation.
• Reasonable for patients with comorbid depression — where the antidepressant indication, not the sleep evidence, carries the decision.

Tier 3: Avoid chronic use

Benzodiazepines (temazepam, triazolam, lorazepam, etc.) — Caution

• Effective acutely but with substantial harms: falls, cognitive impairment, dependence, withdrawal.
• Fractures. Pooled observational data give a 52% higher relative risk of hip fracture with benzodiazepine use (RR 1.52, 95% CI 1.37–1.68), rising to more than double (RR 2.40) in the first weeks of a new prescription.^[30] In an older adult, a hip fracture is not a minor adverse event — it carries substantial mortality.
• Not recommended for chronic use; on the Beers Criteria for inappropriate prescribing in older adults.
• Withdrawal can be severe — taper slowly, and it is worth knowing that stopping succeeds far more often than people expect: across sixteen deprescribing studies, between roughly a fifth and two-thirds of long-term users came off.^[31]

Over-the-counter (OTC) antihistamines (diphenhydramine — Benadryl, ZzzQuil) — Caution

• Not recommended chronically.
• High anticholinergic burden, daytime sedation, cognitive effects.
• The sedation also wears off fast. In a randomised crossover trial, healthy volunteers taking diphenhydramine became measurably drowsy on day one and were indistinguishable from placebo by day four.^[32] That was daytime drowsiness rather than sleep itself, and it was fifteen young men — but it is the reason both the American and European sleep guidelines tell clinicians not to use antihistamines for insomnia.^{[33] [34]}
• Anticholinergic medications are associated with dementia. In a prospective cohort of 3,434 adults aged 65+ followed for about seven years, the heaviest cumulative anticholinergic exposure — the equivalent of a standard daily dose for more than three years, accumulated over a decade — carried a 54% higher risk of incident dementia than no use (HR 1.54, 95% CI 1.21–1.96), with a clear dose-response across the four exposure levels.^[35] This is observational and the causal question is not closed — early dementia may itself drive sleep problems and antihistamine use — but the dose-response makes it hard to dismiss. The literature has also moved on since that cohort: a study roughly seventeen times larger, in the same journal, broke the signal down by drug class,^[36] and a pooled analysis across 26 studies and more than 600,000 people puts the class-level association at a more modest level than the figure above.^[37] The direction has held up; the size of it, for antihistamines specifically, is less certain than a single 54% figure suggests.
• Not a good regular sleep aid at any adult age, and from 65 the formal guidance says avoid it. The American Geriatrics Society Beers Criteria — the standard reference for drugs that are usually best avoided in older people — list first-generation antihistamines from age 65.^[38] Below 65 there is no age threshold in the evidence; the reason to skip it earlier is that it works poorly, not that a cut-off exists.

Special populations

Midlife women with vasomotor symptom-driven insomnia — Moderate

When night sweats and hot flushes are what fragment sleep, treating the vasomotor symptoms usually helps the insomnia more than a hypnotic does.

• CBT-I as first-line behavioral treatment (Strong — same evidence base as above).
• Discuss menopausal hormone therapy (MHT) with your clinician — particularly if within 10 years of menopause and without contraindications (Moderate for vasomotor-driven sleep disruption).
• Non-hormonal options where MHT is unsuitable: low-dose paroxetine, gabapentin, venlafaxine (Moderate); and the newer neurokinin antagonists. Elinzanetant has the best sleep-specific data — sleep disturbance was a pre-specified secondary endpoint in two phase 3 trials and improved in both, though both were run by the manufacturer.^[39] Fezolinetant carries a boxed warning for serious liver injury: liver tests before starting, then monthly for three months and again at six and nine months, and it must not be used in anyone with cirrhosis.^[40] That is a drug for someone who needs it, not a casual choice.

Midlife men with low energy/libido

• Get screened for obstructive sleep apnea (OSA) before pursuing testosterone replacement therapy (TRT). Untreated apnea causes the fatigue and low energy that send men looking for testosterone in the first place, so it should be excluded before anyone reaches for a prescription. But be clear about what treating it does and doesn't fix: pooling twelve studies in 388 men, CPAP did not raise total testosterone, and the authors concluded the evidence does not support a direct link between apnea and testosterone at all.^[41] The reason to find and treat the apnea is the apnea — and the fatigue, which testosterone would not have fixed either. See testosterone therapy for the broader framework — lifestyle correction first, who actually benefits from TRT, and why "optimization" toward 25-year-old levels carries a real cardiovascular cost.

Older adults

• DORAs and low-dose doxepin preferred.
• Avoid benzodiazepines, Z-drugs, and antihistamines where possible.

When insomnia and sleep apnea coexist (Moderate)

The two conditions overlap far more often than either is treated as a possibility for the other, and the combination is worse than either alone: in a cohort of 5,236 adults followed fifteen years, having both carried about 47% higher all-cause mortality than having neither, along with roughly double the odds of hypertension.^[42]

The order of treatment matters, and it has been tested. CBT-I works in people with untreated apnea as well as treated apnea,^[43] and doing CBT-I first made people substantially more likely to stick with CPAP afterwards — about an extra hour of use per night in a randomised trial.^[44] So if both are present: treat the insomnia behaviourally first, then the apnea. Reaching for a hypnotic instead risks treating the symptom of the condition you have not diagnosed yet — see Sleep-disordered breathing.

• A 2025 model projected what would happen if the 15 million Americans over 50 who currently take prescription sleep medication stopped: lifetime falls down 8.5%, cognitive impairment down 2.1%, and about 0.11 years — roughly 40 days — of additional remaining life expectancy. Read it as a model rather than a measurement: the fall and dementia risks fed into it come from observational associations, sleep apnea is not represented in it at all, and untreated insomnia is itself a risk factor for falls and cognitive decline — so some of the modelled harm may belong to the insomnia rather than to the drug.^[45]

What about supplements?

Magnesium, melatonin, L-theanine, ashwagandha, glycine, valerian, CBD — covered in detail in the Sleep supplements page. Bottom line: most have modest evidence and are reasonable adjuncts, but none replace CBT-I for chronic insomnia.

Further reading

• Sateia MJ et al. Clinical Practice Guideline for the Pharmacologic Treatment of Chronic Insomnia in Adults: An American Academy of Sleep Medicine Clinical Practice Guideline. J Clin Sleep Med 2017.^[46]
• Edinger JD et al. Behavioral and psychological treatments for chronic insomnia disorder in adults: an American Academy of Sleep Medicine systematic review, meta-analysis, and GRADE assessment. J Clin Sleep Med 2021.^[47]
• van Straten A et al. Cognitive and behavioral therapies in the treatment of insomnia: a meta-analysis. Sleep Med Rev 2018 — 87 RCTs; ISI g = 0.98.^[48]
• Zachariae R et al. Efficacy of internet-delivered cognitive-behavioral therapy for insomnia: a systematic review and meta-analysis of randomized controlled trials. Sleep Med Rev 2016.^[49]
• Espie CA et al. Effect of Digital Cognitive Behavioral Therapy for Insomnia on Health, Psychological Well-being, and Sleep-Related Quality of Life: A Randomized Clinical Trial. JAMA Psychiatry 2019 — n = 1,711.^[50]
• Kallestad H et al. Mode of delivery of Cognitive Behavioral Therapy for Insomnia: a randomized controlled non-inferiority trial of digital and face-to-face therapy. Sleep 2021.^[51]
• Kishi T et al. Comparative efficacy and safety of daridorexant, lemborexant, and suvorexant for insomnia: a systematic review and network meta-analysis. Transl Psychiatry 2025.^[52]
• Na HJ et al. Clinical safety and narcolepsy-like symptoms of dual orexin receptor antagonists in patients with insomnia: a systematic review and meta-analysis. Sleep 2024 — 11 RCTs, 7,703 patients.^[53]
• Donnelly K et al. Benzodiazepines, Z-drugs and the risk of hip fracture: a systematic review and meta-analysis. PLOS ONE 2017.^[54]
• Yi XY et al. Trazodone for the treatment of insomnia: a meta-analysis of randomized placebo-controlled trials. Sleep Med 2018.^[55]
• Yeung WF et al. Doxepin for insomnia: a systematic review of randomized placebo-controlled trials. Sleep Med Rev 2015.^[56]
• Gray SL et al. Cumulative use of strong anticholinergics and incident dementia: a prospective cohort study. JAMA Intern Med 2015.^[57]
• Heun-Johnson H et al. Lifetime burden of prescription medication for insomnia in middle-aged and older adults in the US: a microsimulation study. Lancet Reg Health Am 2025.^[58]
• Simon L et al. Comparative efficacy of onsite, digital, and other settings for cognitive behavioural therapy for insomnia: a systematic review and network meta-analysis. Sci Rep 2023, with author correction 2025.^{[59] [60]}
• De Crescenzo F et al. Comparative effects of pharmacological interventions for the acute and long-term management of insomnia disorder in adults: a systematic review and network meta-analysis. Lancet 2022 — 170 trials, 36 interventions.^[61]
• Furukawa Y et al. Components and Delivery Formats of Cognitive Behavioral Therapy for Chronic Insomnia in Adults: A Systematic Review and Component Network Meta-Analysis. JAMA Psychiatry 2024 — 241 trials.^[62]
• Kyle SD et al. Clinical and cost-effectiveness of nurse-delivered sleep restriction therapy for insomnia in primary care (HABIT). Lancet 2023.^[63]
• Lechat B et al. Comorbid insomnia and sleep apnoea is associated with all-cause mortality. Eur Respir J 2022.^[64]
• Riemann D et al. The European Insomnia Guideline: An update on the diagnosis and treatment of insomnia 2023. J Sleep Res 2023.^[65]
• 2023 American Geriatrics Society Beers Criteria® Update Expert Panel. American Geriatrics Society 2023 updated AGS Beers Criteria® for potentially inappropriate medication use in older adults. J Am Geriatr Soc 2023.^[66]
• Żełabowski K et al. Targeting the Orexin System in the Pharmacological Management of Insomnia and Other Diseases: Suvorexant, Lemborexant, Daridorexant, and Novel Experimental Agents. Int J Mol Sci 2025.^[67]