Insomnia treatment has settled into two tracks that are rarely compared in the same place.
One is a structured behavioural programme that changes when you get into bed and how long you stay there. The other is a newer class of sleep drug that blocks the brain signal keeping you awake.
A network meta-analysis published in Translational Psychiatry on 16 March 2026 pooled ten double-blind, placebo-controlled trials covering 5,597 adults with insomnia disorder.[1]
It is the largest pooled picture of that drug class so far, and the effect sizes are more modest than the marketing around new sleep drugs suggests.
Why insomnia treatment changed direction
For three decades the default answer to chronic insomnia was a sedative.
Benzodiazepines and the so-called Z-drugs work by amplifying the brain chemistry that dampens neural activity, which pushes the whole nervous system towards unconsciousness rather than towards sleep specifically.
That mechanism carries known costs: daytime grogginess, falls in older adults, tolerance, and dependence.
Guidelines in the United States and Europe responded by moving structured behavioural therapy to first line and reserving drugs for people in whom therapy is unavailable or insufficient.
The newer drugs take a different route. Orexin is a signal the brain releases to keep you awake; blocking both of its receptors removes the wake signal instead of adding a sedative one. Three of these are now licensed in multiple countries: daridorexant, lemborexant and suvorexant.
Insomnia treatment, what the 2026 network meta-analysis found
The updated analysis added two new Chinese phase 3 trials to the eight already pooled in 2025, bringing the total to ten trials and 5,597 participants.[1] Average age was 55.5 years and 68 per cent were women. Every active dose beat placebo on every efficacy measure.
The largest reported effect on time to fall asleep at one month was lemborexant 10 mg, at a standardised mean difference of 0.40 in the direction of improvement.
The smallest was suvorexant 20 mg, at 0.16. For total sleep time, daridorexant 50 mg led at 0.44 and lemborexant 5 mg trailed at 0.20.
Read plainly, these are small to moderate effects. They are real and consistent across ten trials, but a person expecting a drug to turn a broken night into an unbroken one is expecting more than the data promise.
Where the three drugs actually differ
Lemborexant 10 mg came out ahead of daridorexant 25 mg and suvorexant on time to fall asleep, and ahead of lemborexant 5 mg on total sleep time. Daridorexant 50 mg beat both daridorexant 25 mg and lemborexant 5 mg on total sleep time.[1]
On wake after sleep onset and on the Insomnia Severity Index, no active drug beat any other. In ranking terms lemborexant 10 mg topped every measure except total sleep time, where daridorexant 50 mg led. The gaps between drugs are narrower than the gap between any drug and placebo.
One caution the authors raise themselves: for lemborexant 10 mg on time to fall asleep, the indirect and direct evidence disagreed. They advise using the head-to-head figure of 0.40 rather than the network estimate.
Side effects in the pooled data
Daytime sleepiness was reported more often than with placebo for daridorexant at both doses, lemborexant 10 mg and suvorexant.[1] That is the expected trade-off for a drug that suppresses the wake signal.
Beyond that, nothing separated the drugs from placebo. Dropout for any reason, dropout because of side effects, dizziness, falls, headache and upper respiratory infection all came out similar.
Sleep paralysis and vivid or disturbing dreams were rare and no more common than on placebo.
The analysis also looked at withdrawal symptoms during the run-out phase after stopping. Daridorexant and lemborexant showed none. The suvorexant trials did not report this outcome, so nothing can be said about it either way.
Insomnia treatment, why behavioural therapy comes first
The largest component analysis of cognitive behavioural therapy for insomnia pooled 241 randomised trials and 31,452 participants in JAMA Psychiatry.[2] It took the therapy apart and asked which pieces carry the benefit.
Four components mattered: cognitive restructuring, third-wave approaches such as mindfulness and acceptance, sleep restriction, and stimulus control.
Sleep hygiene education, the advice most people have actually received, contributed nothing measurable. Relaxation training trended in the wrong direction.
The full package delivered face to face raised remission by 33 percentage points over in-person education alone, with a number needed to treat of 3. That is a larger and more durable effect than any of the drug figures above, and it is the reason guidelines put therapy first.
Our earlier piece on sleep apnea treatment in 2026 covers a different cause of broken nights that is often mistaken for insomnia, and that no insomnia drug will fix.
What sleep restriction actually involves
The name misleads people. Sleep restriction does not mean sleeping less. It means spending less time in bed, so that the time you do spend there is mostly spent asleep.
You start by logging two weeks of actual sleep. If you lie down for eight hours and sleep five, the prescribed time in bed becomes about five and a half hours. Outside that window you stay up.
The first few nights feel worse. That extra tiredness is the point: it raises the pressure to sleep, and as the tossing time shrinks, the learned link between the bed and being awake weakens.
Once sleep fills 85 to 90 per cent of the time in bed, the window widens by 15 minutes. In the component analysis this was the piece that improved sleep quality, sleep efficiency and wake after sleep onset together.[2]
Daytime sleepiness does spike early on, so anyone who drives or operates machinery should time the start with a clinician. It is not recommended in uncontrolled epilepsy or bipolar disorder.
Delivery format changed the result
The 241-trial analysis also separated how the therapy was delivered. In-person, therapist-led programmes performed best. Adding that one element widened the ratio of people who reached remission to those who did not by a factor of 1.83.[2]
Self-guided app and web programmes scored lower. That does not make them useless. Far more people can reach a phone than can reach a trained insomnia therapist.
A workable order is this: try in-person first if you can get it, start with a validated digital programme if you cannot, and move to in-person or a medication discussion if eight weeks bring no change.
When a sleep drug is worth considering
Therapy being first line does not mean drugs are never appropriate. For a short acute stretch, after a bereavement or during pain, a drug works within days where therapy takes weeks.
Access is the other real reason. Shift workers and people far from a clinic often cannot attend a fixed weekly session, and a programme you cannot attend has no effect size at all.
If eight weeks of properly delivered therapy leave the Insomnia Severity Index roughly where it started, adding a drug is a reasonable next step. The two are not mutually exclusive.
If you have been on a sleep drug for years, plan the taper together with behavioural therapy. Stopping abruptly produces a stretch of worse sleep, and that experience is what pushes people back onto the drug.
| Aspect | Finding |
|---|---|
| Evidence base | 10 double-blind placebo-controlled drug trials, 5,597 adults, mean age 55.5 years, 68 per cent women |
| Time to fall asleep | Lemborexant 10 mg largest at SMD 0.40; suvorexant 20 mg smallest at 0.16, both versus placebo at one month |
| Total sleep time | Daridorexant 50 mg largest at SMD 0.44; lemborexant 5 mg smallest at 0.20 |
| Insomnia Severity Index | Range 0.35 (lemborexant 10 mg) down to 0.22 (daridorexant 25 mg) at one month |
| Commonest side effect | Daytime sleepiness, raised versus placebo for daridorexant 25 and 50 mg, lemborexant 10 mg and suvorexant |
| Withdrawal | No signal for daridorexant or lemborexant after stopping; suvorexant trials did not report it |
| Therapy components that work | Cognitive restructuring, third-wave methods, sleep restriction, stimulus control, delivered in person |
| Therapy components that do not | Sleep hygiene education added nothing; relaxation training trended towards harm |
| Certainty of drug evidence | Rated low to very low overall; all trials ran three months or less |
Insomnia treatment, what the evidence does not settle
Every trial in the drug analysis ran for three months or less, and most outcomes were measured at one month. Chronic insomnia lasts years. Nothing here tells you what happens in year two.
The authors rated overall confidence in the network evidence as low to very low, mostly because of how few head-to-head comparisons exist. Almost every drug was tested against placebo rather than against another drug, so the rankings rest on indirect arithmetic.
All outcomes were self-reported rather than measured in a sleep laboratory. That is arguably the right choice, since how rested a person feels is what matters, but it is not the same as objective sleep architecture.
Talk to a doctor before starting or stopping any sleep medication. This article summarises published trial results and is not medical advice for any individual.
References
- Kishi T, Ikuta T, Hatano M, et al. Optimal use of dual orexin receptor antagonists for insomnia: a network meta-analysis perspective. Translational Psychiatry, 16 March 2026; 16:149.
- Furukawa Y, Sakata M, Yamamoto R, 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, April 2024; 81(4):357-365.
- Kishi T, Ikuta T, Citrome L, et al. Comparative efficacy and safety of daridorexant, lemborexant, and suvorexant for insomnia: a systematic review and network meta-analysis. Translational Psychiatry, 2025; 15:211.
