Melatonin for Sleep: A Systematic Evidence Review
Does melatonin supplementation improve sleep in adults?
Evidence supports: Sleep Onset Latency, Sleep Quality, Sleep Efficiency, Total Sleep Time +3 more
No clear effect: REM Sleep Latency
Mixed results: Sleep Maintenance
Early data: Insomnia Severity, Next-Day Alertness and Sleep-Related Daytime Function, Daytime Functioning +7 more
Abstract
Melatonin modestly improves sleep overall, and its most dependable benefit is helping people fall asleep sooner.61531 Across pooled trials, the average reduction in sleep onset latency was about 15 minutes, which clears the usual 10 minute threshold for a change people tend to notice, and about 1 in 5 people meaningfully benefit beyond placebo.61531 That is the clearest part of the evidence base.
The rest of melatonin's sleep profile is more limited. Sleep quality improves on average, but by about 1.6 points on the Pittsburgh Sleep Quality Index, a 21 point scale where roughly 3 points is usually considered a clinically meaningful change. That makes the average improvement real but often modest in day to day life.91421 Sleep efficiency also improves consistently across studies, yet the gain is small enough that many people may not feel much difference even when a device can detect it.81526 Total sleep time appears to increase by about 18 minutes on average, but results vary a lot between studies, so some groups gain time while others do not.1726
Melatonin is much less convincing for staying asleep through the night, changing wake time, or reshaping sleep architecture such as REM or deep sleep.71226 Where melatonin often looks strongest is when the problem is circadian mistiming rather than general insomnia, especially delayed schedules, shift related daytime sleep, blindness-related rhythm problems, and jet lag.14101115
Bottom line: the evidence reviewed here shows melatonin as a timing aid with modest sleep benefits, not a broad knockout sedative. If it helps, the change is most likely to be falling asleep earlier and feeling that sleep went somewhat better, not sleeping dramatically longer or more deeply.101521
In Plain Language
Melatonin can help, but mostly in one specific way: it may help you fall asleep sooner. The average benefit is modest, not dramatic. Some people also feel that sleep quality is a bit better and feel less sleepy the next day, but melatonin is much less reliable for staying asleep all night, sleeping much longer, or increasing deep sleep.
It seems to work best when your body clock is off, like jet lag, a delayed sleep schedule, or other timing-related sleep problems. It looks less dependable as a catch-all fix for every kind of insomnia.
If you try melatonin, use it with realistic expectations: think "timing aid" more than "strong sleeping pill."
Introduction
Melatonin matters because many sleep problems are really timing problems. Some people are not unable to sleep so much as unable to sleep at the desired clock time, which is why a supplement that acts like a darkness signal has always been biologically plausible.41531 The practical question is whether that plausibility turns into meaningful sleep improvement in real trials.
The current analysis says yes, but only to a point. Melatonin demonstrates a modest, reliable effect on sleep initiation, and it suggests smaller gains in sleep quality, sleep efficiency, total sleep time, and daytime sleepiness.6141521 It does not show the same clarity for sleep maintenance, wake time, or sleep stage remodeling.71226 The pattern is coherent: melatonin behaves more like a circadian cue than a general-purpose sleep drug.101531
That distinction matters for expectations. If the goal is to fall asleep sooner when sleep timing is off, melatonin has a credible case.61531 If the goal is to stay asleep all night, add large amounts of deep sleep, or transform severe insomnia on its own, the evidence reviewed here is much less reassuring.71228
Evidence 1 of 5
Melatonin helps people fall asleep more reliably than it helps them stay asleep
Melatonin demonstrates its clearest benefit in helping people fall asleep sooner.61531 Across 12 pooled studies, the average reduction in sleep onset latency was 14.8 minutes, which is above the usual 10 minute threshold for a clinically meaningful change, and about 1 in 5 people benefit in a way they likely would not have on placebo alone (NNT 4.7, pooled effect 0.315, 95% CI 0.191 to 0.439).61531 Individual trials make that feel real: adults aged 65 to 80 with insomnia fell asleep about 15.6 minutes sooner after 3 weeks on 2 mg prolonged release melatonin, adults with delayed sleep-wake phase disorder fell asleep about 12 minutes sooner by actigraphy and 18 minutes sooner by diary with 0.5 mg timed before desired bedtime, and a single 2 mg dose in healthy adults shortened latency by about 7 minutes in one night.61531
This benefit is genuine but not equally strong in every setting.21417 Heterogeneity, which measures how much study results disagree with each other, was high for sleep onset latency (I-squared 77.8%). That means the average effect is credible, but the size of benefit depends a lot on who is taking melatonin, why their sleep is off, and how the dose is timed. The prediction interval crossed no effect, so some future studies in new contexts could show little benefit even though the average signal is positive.21417
Melatonin suggests only a modest gain in total sleep time, and that gain is less dependable than the effect on sleep onset.1726 The pooled increase was about 18 minutes, which falls short of the usual 30 minute threshold for a clearly meaningful change (pooled effect 0.281, 95% CI 0.135 to 0.427).1726 Some trials did better than that, such as roughly 24 extra minutes during simulated shift-work daytime sleep, 22 extra minutes by diary in children with neurodevelopmental disorders, and about 25 extra minutes in older adults during a forced-desynchrony lab study.1726 But large adult trials were more restrained, including a long insomnia study where the subgroup signal for total sleep time was only about 13 minutes.6 High heterogeneity again matters here (I-squared 75.2%), meaning extra sleep time is possible, but not something to count on uniformly.
Melatonin does not appear to solve sleep continuity in the same reliable way.71226 Sleep maintenance, usually measured as wake after sleep onset, had a pooled effect that crossed no effect despite a seemingly meaningful average reduction of 22.2 minutes in the small set of available studies (pooled effect 0.174, 95% CI -0.144 to 0.492).71226 One small epilepsy study found about 22 fewer minutes awake after sleep onset, and one laboratory study found shorter awakenings despite slightly more of them, but a larger pediatric trial found no change in night waking frequency or duration.71226 That is a classic mixed outcome: promising in some settings, absent in others, and too thin to treat as established.
Wake-up time is even less interpretable.712 One pediatric study found waking about 30 minutes earlier, while another found waking about 18 minutes later, and the pooled estimate is too contradictory and too underpowered to guide decisions.712 That inconsistency fits the broader picture: melatonin is more reliable at shifting when sleep starts than at determining how the rest of the night unfolds.
What this means
If melatonin helps, the most noticeable change is usually lying awake for less time at the start of the night. Expect a modest chance of sleeping longer, but do not expect dependable protection against middle-of-the-night awakenings.
Sleep Onset Latency
Proven benefit Strong · 76Proven benefit
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 23 papers, majority low risk |
| Inconsistency | Serious | I²=78% (> 75%) |
| Imprecision | No concern | N=3715 meets OIS=400 |
| Publication bias | Serious | Egger's p=0.000, funnel asymmetry detected (k=19) |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Total Sleep Time
Likely helps Good · 51Likely modest benefit
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 25 papers, majority low risk |
| Inconsistency | Serious | I²=75% (> 75%) |
| Imprecision | No concern | N=3815 meets OIS=400 |
| Publication bias | Serious | Egger's p=0.000, funnel asymmetry detected (k=23) |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Sleep Maintenance
Mixed results Very early · 20Studies contradict
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 7 papers, majority low risk |
| Inconsistency | No concern | no concerns (I²=28%, consistency=75%, PI crosses null) |
| Imprecision | No concern | N=415 meets OIS=400 |
| Publication bias | No concern | k=6 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Very low | |
Wake-Up Time
Early data Very early · 19Faint negative signal
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | Serious | 2/3 papers with RoB concerns |
| Inconsistency | No concern | no concerns (no data) |
| Imprecision | Very serious | N=128 far below OIS=400 |
| Publication bias | No concern | k=2 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Very low | |
Evidence 2 of 5
Sleep often feels somewhat better, but usually not dramatically better
Melatonin shows a real but usually modest improvement in overall sleep quality.91421 Across 15 pooled studies, sleep quality improved with a small average effect (0.354, 95% CI 0.159 to 0.549), and about 1 in 6 people may get a meaningful benefit beyond placebo.91421 In native units, the average change was about 1.6 points on the PSQI, a 21 point scale where about 3 points is the usual threshold for a clinically important improvement. That means the average person is likely to notice some easing, but not a dramatic reset of sleep.91421 Individual adult trials reflect that range: breast cancer survivors improved by about 1.7 PSQI points over 4 months, adults with traumatic brain injury improved by 1.79 points over 4 weeks, and patients with cirrhosis improved by 2.18 points over 2 weeks.91421
The strongest sleep quality results often come from smaller, newer trials, so the headline effect should be read with some restraint.193233 A 10 day chemotherapy study and a 1 month COMISA study reported very large between-group differences on the PSQI, and a 14 day cancer insomnia trial showed a striking drop on the Athens Insomnia Scale from 14.8 at baseline to 5.0 by day 14 compared with 13.1 on placebo.193233 Those findings are impressive, but they sit beside much smaller effects in larger studies, including a 791 person insomnia trial where long-term PSQI improvement in low melatonin excretors averaged only 0.66 points, and a 709 person lung cancer trial that found no meaningful sleep benefit at 12 or 24 months.623 High heterogeneity for sleep quality (I-squared 60.7%) means the average benefit is real on paper but not equally visible across all populations.
Sleep efficiency is the most internally consistent nighttime outcome, but it is also one of the least dramatic.81526 The pooled effect was positive and stable across studies (0.359, 95% CI 0.198 to 0.521, I-squared 0%), which means study results lined up unusually well.81526 The catch is magnitude. The average change translates to only about 0.7 points on the PSQI framework, well below the 3 point threshold usually considered meaningfully noticeable.81526 Some device-based studies did show larger shifts, such as a 6.2 percentage point increase in healthy older adults and a 10.9 percentage point difference in the small COMISA trial, but those sit alongside null or borderline findings elsewhere.2632
Insomnia severity scores look promising, but the evidence is still too unstable for a firm claim.15283233 The pooled signal points toward benefit, and the average reduction of about 2.4 points reaches the usual threshold for a meaningful change on the Insomnia Severity Index.152832 But that conclusion rests on only four studies with extreme heterogeneity (I-squared 96.1%), meaning the studies disagree sharply about how much benefit exists. One endometriosis trial showed essentially no effect, while recent trials in cancer insomnia and COMISA reported much larger gains.283233 This is the right place for measured optimism, not certainty.
Melatonin has not shown that it reduces the need for other sleep medication.9 The current analysis has only one usable study for that question, and it found no difference on the PSQI medication-use component after 4 months.9 That is not proof of no effect in every context, only that the evidence reviewed here does not support that claim.
What this means
Sleep may feel smoother and more satisfying on melatonin, but the average improvement is usually modest. It is reasonable to expect better sleep quality scores before expecting a dramatic transformation in severe insomnia.
Sleep Quality
Proven benefit Strong · 76Proven modest benefit
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 28 papers, majority low risk |
| Inconsistency | Serious | I²=61% (> 50%) |
| Imprecision | No concern | N=6035 meets OIS=400 |
| Publication bias | Serious | Egger's p=0.000, funnel asymmetry detected (k=25) |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Sleep Efficiency
Proven benefit Strong · 83Proven but unnoticeable
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 16 papers, majority low risk |
| Inconsistency | No concern | no concerns (I²=0%, consistency=100%, PI does not cross null) |
| Imprecision | No concern | N=1679 meets OIS=400 |
| Publication bias | Serious | Egger's p=0.000, funnel asymmetry detected (k=16) |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Insomnia Severity
Early data Limited · 43Promising early signal
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 4 papers, majority low risk |
| Inconsistency | Serious | I²=96% (> 75%) |
| Imprecision | Serious | N=279 below OIS=400 |
| Publication bias | No concern | k=4 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Sleep Medication Use
Not enough research Limited · 41Not enough research
Single study: W 2014, d=0.00 (n=43+42)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 2 papers, majority low risk |
| Inconsistency | No concern | no concerns (no data) |
| Imprecision | Very serious | N=85 far below OIS=400 |
| Publication bias | No concern | k=1 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Evidence 3 of 5
Some daytime spillover is possible, especially less sleepiness, but the evidence thins out after that
Melatonin likely helps with next-day sleepiness, and this is the clearest daytime payoff in the current analysis.213032 Across eight pooled studies, daytime sleepiness improved with a moderate average effect (0.534, 95% CI 0.248 to 0.821), which works out to about 1.6 points on the Epworth Sleepiness Scale, a 24 point scale where about 2.5 points is usually considered clinically meaningful.213032 That puts the average benefit just below the usual threshold for a clearly felt change, but some settings did better. In cirrhosis, ESS improved by 1.76 points over 2 weeks, and in COMISA the one month endpoint separated much more strongly, with ESS around 6.9 on melatonin versus 13.3 on placebo.2132 About 1 in 8 people may experience a meaningful improvement beyond placebo (NNT 7.8).213032
This daytime sleepiness benefit is not equally secure across all populations.21427 Heterogeneity was high here too (I-squared 78.1%), which means some groups improve clearly while others barely move. The prediction interval crossed no effect, so future trials in different settings could plausibly find little or no benefit even though the average direction remains favorable.21427
Broader daytime functioning shows only a faint signal.91527 The pooled effect was statistically positive (0.407, 95% CI 0.109 to 0.706), but the native-unit change was only about 0.4 points on PSQI daytime dysfunction, far below the 3 point threshold that would count as clearly meaningful.927 Some individual studies were encouraging, including improved PROMIS sleep-related impairment in delayed sleep-wake phase disorder and better PSQI daytime dysfunction in breast cancer survivors, but the average shift is small enough that many people would feel only a slight difference in daily functioning.915
Morning alertness remains an open question rather than an established benefit.31 The current analysis has only a single small dataset for this outcome, and although it leans positive, it is too underpowered to support confident advice.31 That gap matters because feeling less sleepy is not always the same as feeling sharper or more energized on waking.
Fatigue results are also less convincing than the sleepiness data.142330 The current analysis suggests a small overall trend toward less fatigue, but the pooled estimate crossed no effect despite a large combined sample, largely because trials pointed in different directions.142330 Adults with traumatic brain injury had modest gains in fatigue impact and vitality after 4 weeks, chronic pain patients reported a small reduction in aggregated fatigue, but a long cancer trial found only a borderline 12 month change that disappeared by 24 months.142330 The practical reading is simple: melatonin may reduce the sensation of sleepiness more readily than it restores daytime energy in a broad sense.
What this means
A good response to melatonin can carry into the next day, mainly as less sleepiness. Improvements in energy, alertness, and daytime performance are possible, but they are less consistent and usually smaller.
Daytime Sleepiness
Likely helps Strong · 73Likely modest benefit
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 13 papers, majority low risk |
| Inconsistency | Serious | I²=78% (> 75%) |
| Imprecision | No concern | N=890 meets OIS=400 |
| Publication bias | Serious | Egger's p=0.000, funnel asymmetry detected (k=10) |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Next-Day Alertness and Sleep-Related Daytime Function
Early data Limited · 42Faint early signal
Single study: P 2024, d=0.30 (n=64+64)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 2 papers, majority low risk |
| Inconsistency | No concern | no concerns (no data) |
| Imprecision | Very serious | N=128 far below OIS=400 |
| Publication bias | No concern | k=1 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Daytime Functioning
Early data Limited · 44Barely detectable
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 5 papers, majority low risk |
| Inconsistency | No concern | no concerns (I²=36%, consistency=100%, PI crosses null) |
| Imprecision | No concern | N=440 meets OIS=400 |
| Publication bias | No concern | k=5 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
General Fatigue
Likely helps Strong · 73Likely benefit
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 6 papers, majority low risk |
| Inconsistency | Serious | I²=73% (> 50%) |
| Imprecision | No concern | N=2913 meets OIS=400 |
| Publication bias | No concern | k=4 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Evidence 4 of 5
Sleep architecture is not where melatonin's case is strongest
Melatonin does not appear to reliably reshape sleep architecture, and this is one of the weakest parts of the evidence base.1226 The most stable conclusion here is actually a negative one: REM latency probably does not improve in any dependable way.12 The available evidence for REM outcomes is sparse, and the current analysis rates REM latency as probably showing no meaningful benefit overall despite one small study finding a longer delay to REM onset.12
Deep sleep is not supported as a melatonin benefit in the evidence reviewed here.26 In the only study contributing directly to this pooled outcome, 5 mg melatonin in healthy older adults reduced slow-wave sleep by about 5.6 minutes on average, a change that is statistically real within that study but small enough to be hard to feel and far too narrow a base to generalize.26 That same trial also increased stage N2 sleep by about 14 minutes, suggesting a redistribution of sleep stages rather than a simple boost in restorative sleep.26
The epilepsy polysomnography study points in a similar but still uncertain direction.12 After 4 weeks of 9 mg sustained-release melatonin, children fell asleep about 11 minutes faster and spent about 22 fewer minutes awake after sleep onset, but they also had 5.5 percentage points more N3, roughly 58 minutes longer REM latency, and 5.8 percentage points less REM sleep.12 Those are interesting physiology findings, not settled clinical conclusions. They come from a very small sample and have not been replicated at the level needed to say melatonin predictably deepens or reorganizes sleep.
This imbalance between subjective benefit and architecture uncertainty is telling.1426 People often report sleeping better on melatonin even when polysomnography does not show a clean, repeated pattern of stage-by-stage improvement. That suggests melatonin may change sleep timing and continuity enough to improve the experience of sleep without acting like a strong sleep-stage remodeling agent.
What this means
Melatonin should not be chosen because it supposedly increases deep sleep or optimizes REM. The present evidence supports easier sleep initiation much more than any specific improvement in sleep stages.
Deep Sleep / Slow-Wave Sleep
Early data Very early · 38Barely detectable negative
Single study: J 2022, d=-0.21 (n=12+12)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 2 papers, majority low risk |
| Inconsistency | No concern | no concerns (no data) |
| Imprecision | Very serious | N=24 far below OIS=400 |
| Publication bias | No concern | k=1 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Stage N2 Sleep Duration
Early data Very early · 35Large effect, needs confirmation
Single study: J 2022, d=0.68 (n=12+12)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 1 papers, majority low risk |
| Inconsistency | No concern | single study, inconsistency N/A |
| Imprecision | Very serious | single small study (N=24) |
| Publication bias | No concern | k=1 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
REM Sleep Latency
Likely no effect Strong · 67Probably doesn't help
Single study: S 2015, d=-1.05 (n=10+10)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 2 papers, majority low risk |
| Inconsistency | Serious | I²=57% (> 50%) |
| Imprecision | Very serious | N=139 far below OIS=400 |
| Publication bias | No concern | k=2 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Rapid Eye Movement Sleep
Not enough research Very early · 10Not enough research
Single study: S 2015, d=-1.19 (n=10+10)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | Serious | 1/1 papers with RoB concerns |
| Inconsistency | No concern | single study, inconsistency N/A |
| Imprecision | Very serious | single small study (N=20) |
| Publication bias | No concern | k=1 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Very low | |
Evidence 5 of 5
The pattern fits a circadian timing aid more than a broad sleep sedative
The current analysis suggests melatonin works best when the problem is mistimed biology rather than a generic lack of sleepiness.41015 That is why some of the largest and most convincing effects appear in jet lag, delayed sleep-wake phase disorder, shift-related daytime sleep, and blindness-related rhythm disruption rather than in routine sleep duration.1101115
Jet lag is where melatonin looks more powerful than it does in ordinary insomnia.1016 Across two pooled studies, jet lag severity improved with a moderate-to-large average effect (0.991, 95% CI 0.439 to 1.544).1016 In one early controlled trial, none of 8 melatonin users still had clinically significant jet lag at day 7 versus 6 of 9 on placebo. In another, retrospective jet lag scores were lower at day 10, 2.15 versus 3.40 on a 6 point scale, and people re-established a normal sleep pattern faster, about 2.9 versus 4.2 days, with daytime tiredness also resolving sooner, about 3.0 versus 4.6 days.1016 Those are not subtle shifts. They fit the idea that melatonin is especially useful when the clock itself needs help.
Melatonin also clearly raises nighttime melatonin exposure, which confirms that the supplement is biologically active even when sleep outcomes are modest.123135 Across six pooled studies, nighttime melatonin measures rose strongly overall (pooled effect 1.549, 95% CI 0.778 to 2.321), although the size varied widely by assay and design.123135 Single-dose prolonged-release melatonin produced a striking rise in salivary melatonin within an hour, with median 21:00 levels around 891 pg/mL versus 10 pg/mL on placebo, and sustained-release preparations raised salivary or urinary metabolites in several other studies as well.123135 High heterogeneity here is expected because blood, saliva, and urinary assays are not interchangeable, but the directional message is clear: exogenous melatonin does meaningfully increase the body's nighttime melatonin signal.
The small bedtime sleepiness literature points the same way.1 One shift-work laboratory study found greater subjective sleepiness at bedtime after 1.8 mg sustained-release melatonin before daytime sleep, and that coincided with about 24 to 26 more minutes of sleep on the first daytime sleep episode.1 A single study cannot settle the issue, but it fits the broader mechanism: melatonin seems to make the body feel more ready for sleep at the intended time, rather than force sleep regardless of timing.
This circadian interpretation also explains some null results.202325 In populations where sleep problems are driven by alcohol recovery, advanced cancer treatment, or broad medical illness rather than clear clock misalignment, melatonin often shows smaller or absent effects on headline sleep scores despite successful supplementation.202325 Results depend not just on dose, but on whether the biology being targeted is actually circadian.
What this means
Melatonin makes the most sense as a timing tool. It is most compelling when sleep is happening at the wrong biological time, such as jet lag, delayed schedules, or other circadian mismatch states.
Jet Lag Severity
Likely helps Strong · 67Likely benefit
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 3 papers, majority low risk |
| Inconsistency | No concern | no concerns (I²=0%, consistency=100%, PI crosses null) |
| Imprecision | Serious | N=246 below OIS=400 |
| Publication bias | No concern | k=3 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Moderate | |
Bedtime Sleepiness
Early data Very early · 36Early signal of harm
Single study: K 2001, d=-1.04 (n=21+21)
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 1 papers, majority low risk |
| Inconsistency | No concern | single study, inconsistency N/A |
| Imprecision | Very serious | single small study (N=42) |
| Publication bias | No concern | k=1 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Nighttime Melatonin Exposure
Early data Limited · 44Large effect, needs confirmation
▸ GRADE Assessment
| Domain | Rating | Reason |
|---|---|---|
| Risk of bias | No concern | 6 papers, majority low risk |
| Inconsistency | Serious | I²=89% (> 75%) |
| Imprecision | Serious | N=348 below OIS=400 |
| Publication bias | No concern | k=6 usable (< 10), cannot assess per Cochrane 10.4 |
| Indirectness | No concern | deferred to Phase 2 (#1546) |
| Overall certainty | Low | |
Across the Evidence
The main pattern is that melatonin helps start sleep more than it helps build or protect sleep once it has started.61526 That makes biological sense. Melatonin is a darkness signal from the circadian system, not a classic sedative. A signal that says "night has begun" should be expected to move sleep onset and sleep timing more than it increases total sleep need or suppresses all awakenings.
The second pattern is that subjective outcomes look better than architecture outcomes.91426 People often report better sleep quality, less sleep disturbance, or less next-day sleepiness even when polysomnography does not show a clean increase in deep sleep or a stable REM effect. That can reflect real benefit rather than bias alone. Falling asleep sooner, feeling sleepier at the intended bedtime, and sleeping a bit more efficiently can improve how a night feels without dramatically reorganizing sleep stages.
The third pattern is that the average benefit is modest, and some of the largest effects probably overstate what most people should expect.193233 Small recent trials sometimes produced striking changes in PSQI or insomnia scales, while larger and longer studies usually found smaller improvements or none at all.623 Publication bias, which means smaller positive studies are more likely to show up in the literature than small negative ones, was detected for several headline outcomes. That does not erase the effect, but it does argue for sober expectations.
The fourth pattern is that context matters as much as dose.141015 Timing relative to the body clock appears central, yet many studies report dose more clearly than circadian timing. That likely explains part of the heterogeneity. In meta-analysis, heterogeneity means the study results vary more than chance alone would predict. For sleep onset latency, sleep quality, total sleep time, daytime sleepiness, and insomnia severity, the prediction intervals crossed no effect. In plain terms, the average benefit is likely real, but not every future study, clinic, or person will get the same result.
Taken together, the evidence fits a coherent model: melatonin is best viewed as a circadian alignment aid with modest sleep benefits, not as a general sleep-building agent that reliably lengthens, deepens, and stabilizes sleep across all causes of insomnia.101531
Discussion
The evidence reviewed here supports a practical, bounded conclusion: melatonin modestly improves adult sleep, and the most defensible benefit is falling asleep sooner.61531 Sleep quality and sleep efficiency also improve on average, but the average gains are smaller and often below standard thresholds for clearly noticeable change.91421 Total sleep time may increase, but not reliably enough to promise an extra half hour of sleep.1726
Confidence is highest for sleep initiation and lower for everything that happens later in the night.61531 The current analysis shows mixed evidence for sleep maintenance, weak evidence for wake-up time, and little support for claims about deep sleep, REM sleep, or sleep medication reduction.791226 That matters because melatonin is often marketed as a general sleep remedy, while the better-supported picture is narrower.
The evidence is also stronger for circadian use cases than for nonspecific insomnia.4101115 Jet lag results are larger than routine sleep results, delayed sleep-wake phase disorder responds in the predicted direction, and laboratory and biomarker studies show that melatonin clearly changes the nighttime hormonal signal.101531 That is why timing probably matters as much as dose, even though the literature does not report timing well enough to answer that question cleanly.
What would change confidence? Larger preregistered trials with objective sleep measures, explicit circadian timing, and direct comparisons of immediate-release versus prolonged-release formulations would help most.61432 More studies focused on sleep maintenance and sleep architecture would also clarify whether the current uncertainty reflects true lack of effect or simply lack of data.
For now, the supported claim is modest but useful: melatonin can help some adults fall asleep earlier and feel that sleep is better, especially when the problem is one of timing. The current analysis does not support stronger promises than that.
Methodology
We searched PubMed for studies on melatonin and sleep, then included controlled human studies listed in the PRISMA flow, mostly randomized trials, with a smaller number of crossover and controlled studies. We read each study, noted who was studied, what sleep outcomes were measured, how large the trial was, and what it found. We assessed certainty with the GRADE framework and judged clinical importance against published meaningful-change thresholds for scales such as sleep onset latency, PSQI, ESS, and total sleep time. Every cited study is publicly indexed on PubMed.
A key limitation is that GRADE was built for drug trials and tends to rate nutrition and supplement evidence conservatively. It automatically downgrades observational evidence and rarely upgrades unless effects are extremely large, often larger than what is realistic in sleep research. Our separate 0-100 trust score gives a more continuous estimate and also checks whether an effect is large enough to matter clinically, not just statistically. So a result can reasonably read as moderately trustworthy in this review even when a formal GRADE table still says low certainty. Known limitations include inconsistent reporting of dose timing and formulation, frequent reliance on subjective sleep questionnaires, and several outcomes with only a few small studies.
Study Selection
Characteristics of Included Studies
| Study | Design | N | Population | Dose | Duration | RoB |
|---|---|---|---|---|---|---|
| J 1986 FT | controlled trial | 17 | healthy | 5 mg nightly; start 3 days before return and continue 4 days after | 7 days treatment (3 days before return flight + 4 days after return) | Low |
| K 1989 FT | rct | 20 | NR | 5 mg once daily starting 3 days before flight through 3 days after arrival | Measures collected from the day before departure through 10 days after arrival; treatment given 3 days before flight, during flight, and once daily for 3 days after arrival | Some |
| K 2001 FT | rct | 21 | healthy | 1.8 mg sustained‑release, 30 minutes before sleep (two doses) | Two 6-day laboratory sessions (each session: 1 adaptation night, 2 baseline nights, 2 simulated night shifts followed by daytime sleep episodes, 1 recovery night). Sessions separated by ≥1 week. | Some |
| G 2005 FT | rct | 40 | clinical | 3 mg at bedtime for 2 weeks | 2 weeks | Low |
| K 2005 FT | rct | 13 | clinical | 0.3 mg daily for 4 weeks (timed before dim-light melatonin onset) | 4-week period | Low |
| V 2006 FT | controlled trial | 44 | healthy | 0.5 mg once daily in afternoon for 3 days | 14 days total protocol with a 3-day treatment | Some |
| I 2010 FT | rct | 72 | clinical | 0.05 mg/kg daily for 1 week | 2 weeks total (1-week qualification + 1-week treatment) | Some |
| A 2010 FT | rct | 791 | clinical | 2 mg nightly (prolonged-release) for 3 weeks | 33 weeks (overall study: 2-week placebo run-in, 3-week double-blind, 26-week double-blind extension, 2-week run-out) | Low |
| P 2012 FT | rct | 146 | clinical | 0.5–12 mg nightly | 12 weeks | Some |
| M 2013 FT | rct | 67 | clinical | 3 mg nightly for 12 months | 12 months | High |
| W 2014 FT | rct | 95 | clinical | 3 mg nightly for 4 months | 4 months | Low |
| T 2015 FT | rct | 13 | clinical | 2 mg nightly between 9–10 pm for 6 weeks | 2 weeks placebo run-in, 6 weeks randomized treatment, 2 weeks washout (10 weeks total) | Some |
| S 2015 FT | rct | 11 | clinical | 9 mg sustained-release, ~30 min before bedtime for 4 weeks | 4 weeks active treatment, 1 week washout, 4 weeks crossover (total 9 weeks) | Some |
| K 2017 FT | controlled trial | 30 | clinical | 10 mg nightly for 30 days | 30 days | Low |
| N 2018 FT | rct | 33 | clinical | 2 mg nightly for 4 weeks | 10 weeks (2-week baseline run-in, then two 4-week treatment periods separated by a 48-hour washout) | Low |
| T 2018 FT | rct | 116 | clinical | 0.5 mg nightly, 1 hour before desired bedtime for 4 weeks | 4 weeks | Low |
| A 2019 FT | rct | 82 | clinical | 6 mg nightly for 8 weeks | 8 weeks | Some |
| M 2020 FT | rct | 30 | clinical | 4 mg nightly (prolonged-release) for 8 weeks | 4 weeks pre-observation, 8 weeks intervention, 4 weeks post-observation | Low |
| A 2020 FT | rct | 36 | clinical | 20 mg nightly for 10 days | 10 days (started 3 days before chemotherapy and continued 7 days after); baseline and end-of-treatment assessments | Low |
| M 2020 FT | rct | 60 | clinical | 5 mg nightly for 4 weeks | 4 weeks | Low |
| A 2020 FT | crossover trial | 71 | clinical | 3 mg nightly for 2 weeks | 6 weeks (two 2-week intervention periods with a 1-week washout between and a 1-week washout after cross-over) | Low |
| B 2021 FT | rct | 119 | clinical | 2–10 mg nightly for up to 104 weeks | 108 weeks (13-week DB + 91-week OL + 2-week run-out) | High |
| D 2021 FT | rct | 709 | clinical | 20 mg nightly for 1 year | Intervention: 1 year nightly; follow-up for recurrence/mortality up to 60 months | Low |
| A 2021 FT | rct | 40 | clinical | 10 mg nightly for 2 weeks | 6 weeks (2 weeks treatment phase, 2 weeks washout, 2 weeks treatment phase) | Low |
| S 2022 FT | rct | 92 | clinical | 10 mg nightly for 24 weeks | 24 weeks | Some |
| J 2022 FT | rct | 24 | healthy | 5 mg nightly, 30 minutes before bedtime (12-night condition) | Forced desynchrony protocol spanning 30 cycles (~24 calendar days); each condition comprised 12 nights of melatonin or 12 nights of placebo within the FD segment. | Low |
| M 2023 FT | rct | 136 | clinical | 6 mg daily (3 mg morning, 3 mg at bedtime) | 8 weeks | Low |
| L 2023 FT | rct | 40 | clinical | 20 mg nightly for 2 months | Three menstrual cycles: one observational cycle followed by two interventional cycles (total study period) | Low |
| M 2023 FT | rct | 27 | healthy | 10 mg nightly for 10 days | 10 days (each condition) with ≥14-day washout | Low |
| U 2024 FT | rct | 58 | clinical | 2 mg nightly (Circadin™) for 6 weeks per treatment period | Each treatment period 6 weeks with a washout of ≥4 weeks between periods and a 4-week follow-up after the final treatment period. | Low |
| P 2024 FT | rct | 75 | healthy | 2 mg prolonged-release at 20:00 | Single-dose crossover; saliva collections at 20:00 (dose), 21:00, awakening, 30 min after awakening, 10:00 and 12:00 next day. | Low |
| T 2024 FT | rct | 50 | clinical | 10 mg nightly for 1 month | 1 month | High |
| M 2024 FT | rct | 120 | clinical | 3 mg nightly, 2 hours before bedtime | 14 days | Low |
| N 2025 FT | crossover trial | 12 | healthy | 6 mg at 21:30 the night before | Two-day experimental condition per arm (single nocturnal dose at 21:30, performance testing next morning) with a one-week washout between conditions. | Low |
| A 2025 FT | rct | 97 | clinical | 2 mg nightly for 90 days, 1–2 h before bedtime | 90 days supplementation (primary), follow-up to 12 months | High |
Sources
- 1. K 2001. Effects of melatonin administration on daytime sleep after simulated night shift work. (2001) ↑
- 2. G 2005. Melatonin improves abdominal pain in irritable bowel syndrome patients who have sleep disturbances: a randomised, double blind, placebo controlled study. (2005) ↑
- 3. V 2006. Advancing human circadian rhythms with afternoon melatonin and morning intermittent bright light. (2006) ↑
- 4. K 2005. Phase-dependent treatment of delayed sleep phase syndrome with melatonin. (2005) ↑
- 5. I 2010. Dose finding of melatonin for chronic idiopathic childhood sleep onset insomnia: an RCT. (2010) ↑
- 6. A 2010. Nightly treatment of primary insomnia with prolonged release melatonin for 6 months: a randomized placebo controlled trial on age and endogenous melatonin as predictors of efficacy and safety. (2010) ↑
- 7. P 2012. Melatonin for sleep problems in children with neurodevelopmental disorders: randomised double masked placebo controlled trial. (2012) ↑
- 8. M 2013. Long-term effects of melatonin on quality of life and sleep in haemodialysis patients (Melody study): a randomized controlled trial. (2013) ↑
- 9. W 2014. A randomized, placebo-controlled trial of melatonin on breast cancer survivors: impact on sleep, mood, and hot flashes. (2014) ↑
- 10. K 1989. Effect of melatonin on jet lag after long haul flights. (1989) ↑
- 11. T 2015. Prolonged release melatonin for improving sleep in totally blind subjects: a pilot placebo-controlled multicenter trial. (2015) ↑
- 12. S 2015. Melatonin improves sleep in children with epilepsy: a randomized, double-blind, crossover study. (2015) ↑
- 13. K 2017. Melatonin Supplementation Lowers Oxidative Stress and Regulates Adipokines in Obese Patients on a Calorie-Restricted Diet. (2017) ↑
- 14. N 2018. Efficacy of melatonin for sleep disturbance following traumatic brain injury: a randomised controlled trial. (2018) ↑
- 15. T 2018. Efficacy of melatonin with behavioural sleep-wake scheduling for delayed sleep-wake phase disorder: A double-blind, randomised clinical trial. (2018) ↑
- 16. J 1986. Alleviation of jet lag by melatonin: preliminary results of controlled double blind trial. (1986) ↑
- 17. M 2020. Melatonin for rapid eye movement sleep behavior disorder in Parkinson's disease: A randomised controlled trial. (2020) ↑
- 18. B 2021. Sleep, Growth, and Puberty After 2 Years of Prolonged-Release Melatonin in Children With Autism Spectrum Disorder. (2021) ↑
- 19. A 2020. Clinical impact of melatonin on breast cancer patients undergoing chemotherapy; effects on cognition, sleep and depressive symptoms: A randomized, double-blind, placebo-controlled trial. (2020) ↑
- 20. M 2020. Melatonin for Treatment-Seeking Alcohol Use Disorder patients with sleeping problems: A randomized clinical pilot trial. (2020) ↑
- 21. A 2020. Low-dose melatonin for sleep disturbances in early-stage cirrhosis: A randomized, placebo-controlled, cross-over trial. (2020) ↑
- 22. A 2019. Effect of Bedtime Melatonin Administration in Patients with Type 2 Diabetes: A Triple-Blind, Placebo-Controlled, Randomized Trial. (2019) ↑
- 23. D 2021. Adjuvant melatonin for the prevention of recurrence and mortality following lung cancer resection (AMPLCaRe): A randomized placebo controlled clinical trial. (2021) ↑
- 24. A 2021. A Pilot Randomized, Clinical Trial of the Anti-pruritus Effect of Melatonin in Patients with Chronic Liver Disease. (2021) ↑
- 25. S 2022. Melatonin supplementation improves N-terminal pro-B-type natriuretic peptide levels and quality of life in patients with heart failure with reduced ejection fraction: Results from MeHR trial, a randomized clinical trial. (2022) ↑
- 26. J 2022. High dose melatonin increases sleep duration during nighttime and daytime sleep episodes in older adults. (2022) ↑
- 27. M 2023. The effect of melatonin on irritable bowel syndrome patients with and without sleep disorders: a randomized double-blinded placebo-controlled trial study. (2023) ↑
- 28. L 2023. Adjuvant use of melatonin for pain management in endometriosis-associated pelvic pain-A randomized double-blinded, placebo-controlled trial. (2023) ↑
- 29. M 2023. Melatonin supplementation does not alter vascular function or oxidative stress in healthy normotensive adults on a high sodium diet. (2023) ↑
- 30. U 2024. Melatonin treatment has consistent but transient beneficial effects on sleep measures and pain in patients with severe chronic pain: the DREAM-CP randomised controlled trial. (2024) ↑
- 31. P 2024. Exogenous melatonin's effect on salivary cortisol and amylase: A randomized controlled trial. (2024) ↑
- 32. T 2024. Effect of melatonin on insomnia and daytime sleepiness, in patients with obstructive sleep apnea and insomnia (COMISA): A randomized double-blinded placebo-controlled trial. (2024) ↑
- 33. M 2024. Comparison Between Efficacy of Oral Melatonin and Oral L-theanine in Improving Sleep in Cancer Patients Suffering From Insomnia: A Randomised Double-blinded Placebo-controlled Study. (2024) ↑
- 34. N 2025. Melatonin Supplementation Enhances Next-Day High-Intensity Exercise Performance and Recovery in Trained Males: A Placebo-Controlled Crossover Study. (2025) ↑
- 35. A 2025. Melatonin supplementation for quality of life in older patients with advanced cancer: a randomized controlled trial. (2025) ↑