Chapter  108:  Melatonin for Treatment of Sleep Disorders

Chronotherapy

Chronotherapy may be used for delayed or advanced sleep phase syndrome and usually involves application of a series of consecutive shifted 24-hour days, thus phase-delaying the sleep cycle three hours per sleep-wake cycle, until the desired bedtime is reached.⁷ Light therapy may be used alone or in conjunction with chronotherapy and is often a treatment of choice for circadian rhythm disorders, since light is the principal synchronizer of circadian timing.⁸ For delayed sleep phase syndrome, bright light exposure in the morning will lead to a phase advance, leading to an earlier time of rising,⁹ while for advanced sleep phase syndrome, bright light exposure in the evening is effective in re-synchronizing the circadian rhythm.^{10 11}

Light can shift the timing of the melatonin rhythm in a dose-response manner, with early night exposure resulting in phase delays and late night exposures resulting in phase advances.^{12 13} Light can also suppress endogenous melatonin levels in a dose-response manner.¹²

Pharmacotherapy

Pharmacotherapy with sedative/hypnotic drugs is also widely used in the treatment of sleep disorders. Hypnotics promote drowsiness and facilitate the onset of sleep, while sedatives induce a calming effect. Ideally, a hypnotic should be rapidly absorbed into the bloodstream, display specific receptor binding, and induce sleep quickly, without causing side effects, buildup of tolerance, physical dependence, and respiratory or central nervous system depression.¹⁴ However, no hypnotic is perfect. Benzodiazepines are the most commonly prescribed hypnotics and act on the inhibitory neurotransmitter receptors that are directly activated by the amino acid gamma-aminobutyric acid. Drugs of this class can be effective in treating transient insomnia and, due to anxiolytic as well as sedative/hypnotic properties, may be useful in the management of insomnia associated with select psychiatric disorders.⁶ Although there are several chemical classes of benzodiazepines, some benzodiazepines are metabolized in the body to N-desmethyldiazepam (nordiazepam), an active metabolite with a long elimination half-life and sedative effects. Unwanted effects of the benzodiazepines include daytime sedation, respiratory depression, dependence, and rebound insomnia.⁶ Non-benzodiazepine hypnotics include zolpidem and zaleplon. Drugs of this class are more specific, display more rapid onset and shorter duration of action as well as fewer negative effects on memory and motor coordination, and are less likely to result in rebound insomnia, compared to benzodiazepines.⁶ Antidepressants such as tricyclic antidepressants, trazodone and mirtazapine may also have sedative/hypnotic properties. Alternatives to traditional hypnotics include herbal and “natural” products such as St. John's Wort, valerian, kava kava, and melatonin.¹⁵ These products have been reported to be effective, however, the methodological quality of studies on effectiveness of these products as well as their inconsistent findings, necessitate further research in this area.

Discovery and History of Melatonin

Melatonin was discovered as a result of the observation that bovine pineal extracts caused blanching of the skin of tadpoles when it was added to swimming water.²² Aaron Lerner, an American dermatologist, isolated and characterized the hormone from beef pineal extracts in 1958, naming it melatonin based on its ability to lighten melanocytes.²³

Melatonin is present in a number of organisms such as bacteria, algae, fungi, plants, insects and vertebrates, including humans.²⁴ Melatonin is also found in foodstuffs such as vegetables, fruits, rice, wheat and herbal medicines.²⁴

Early research involving melatonin was conducted on animals and examined its effects on gonadal maturation and circadian systems. These early animal experiments provided evidence for chronobiologic and sleep-inducing effects of melatonin,²² suggesting a role for this hormone in sleep and behavior in humans. The first experiments of melatonin on humans were conducted in the early 1970s, which provided evidence of a sleep inducing effect of melatonin in humans.^{25 26} The first study involving administration of chronic small doses of melatonin in human volunteers was conducted in 1984 and this study found that melatonin increased self-rated tiredness.²⁷ Sedative-hypnotic effects of melatonin were also noted in a study examining the behavioral effects of melatonin.²⁸ In 1984, melatonin was tested for its ability to alleviate the symptoms of jet lag,²⁷ and this stimulated further trials of melatonin for the treatment of sleep disorders.

Physiology of Endogenous Melatonin

Melatonin secretion follows a circadian rhythm and is entrained to the light/dark cycle; light suppresses the production of melatonin, and with the onset of darkness, melatonin is produced and secreted from pinealocytes.²⁹ Light input is transmitted from the photic receptors in the retina through the retinohypothalamic tract to the suprachiasmatic nucleus (SCN), which is located in the anterior hypothalamus and functions as the central circadian pacemaker of the body.³⁰ During the dark period, the SCN stimulates the release of norepinephrine from the superior cervical ganglion; activation of pinealocyctes by norepinephrine results in production and release of melatonin.³¹

Melatonin is not stored in the pineal gland, but is secreted upon production. The hormone is likely secreted into the bloodstream before entering the cerebrospinal fluid (CSF) of the third ventricle, although it may also be secreted directly into cerebrospinal fluid.³² Evidence for direct secretion of melatonin into CSF has been provided by findings that melatonin levels in CSF are substantially higher than in plasma.³³ Melatonin can also be measured in saliva, where levels are about 70 percent of plasma levels. The onset of melatonin secretion occurs at approximately 2200–2300 hours and maximal plasma concentrations occur at about 0300–0400 hours for a regular sleep cycle.³⁴ The offset of melatonin secretion occurs at approximately 0700–0900 hours.³⁴ The levels of metabolite in urine correlate positively with plasma levels of the hormone²⁹ and provide a non-invasive method of measuring melatonin levels in the body.³⁵

Although melatonin is present in plasma of newborns, the circadian rhythm of melatonin does not exist at birth, but appears at 9–12 weeks of age and is fully established by 5–6 months of age.³⁵ Melatonin reaches high values at 1–3 years of age, with plasma levels peaking at approximately 250 pg/ml. Melatonin levels in plasma begin to decrease just prior to puberty to peak values of less than 100 pg/ml in adulthood.³⁶ There are, however, marked individual differences in the levels of melatonin that are produced by the pineal gland.²⁹

Effects of Exogenous Melatonin

Melatonin has several effects on the body. It is best known as an entrainer of the circadian rhythm.³⁷ In mammals, removal of the pineal gland abolishes melatonin secretion.¹⁹ Exogenous melatonin will cause a phase advance of the melatonin rhythm if given at dusk, and a phase delay if given in the morning.³⁸ The constant lag time between the onset of melatonin secretion and the onset of sleep suggests that exogenous melatonin could promote sleep.³⁹ Administration of exogenous melatonin to healthy volunteers has been shown to increase sleep propensity, reduce sleep onset latency and decrease REM sleep latency.⁴⁰

The secretion of melatonin is also associated with the thermoregulatory cycle. The circadian rhythm of melatonin inversely correlates with the temperature rhythm in humans; melatonin levels in blood increase as core body temperature decreases.⁴¹ Administration of pharmacologic, as opposed to physiologic, doses of exogenous melatonin, has been reported to cause a reduction in core body temperature.^{42 43}

The secretion of melatonin secretion is also associated with the reproductive rhythm. In humans, melatonin secretion is inversely correlated with gonadal development; peak melatonin levels fall just prior to the onset of puberty.⁴⁴ In addition, higher levels of plasma melatonin have been noted in women with amenorrhea.⁴⁵ Taken together, these findings suggest an inhibitory effect of melatonin on the reproductive rhythm.

Melatonin is also involved in immune function, and evidence suggests an immunoenhancing function for melatonin, via stimulation of natural killer cell activity, regulation of cytokine expression and inhibition of apoptosis in immune cells.⁴⁶ In support of such a function, high affinity melatonin receptors have been detected in human T lymphocytes.⁴⁷ Melatonin has also been shown to have oncostatic effects; it reduces tumor growth in animals and humans,^{48 49} may reduce angiogenesis, protects DNA from mutation, and may also decrease tumor initiation.²⁹

Melatonin Receptors

Melatonin has endocrine, autocrine and paracrine actions,²⁹ and some of these actions are receptor-mediated, while others are direct. There are three classes of melatonin receptors, MT1, MT2, and MT3.⁵⁰ In mammalian tissues, the distribution of melatonin receptors appears to be widespread.²⁹ The receptors are most consistently found in the SCN and the pars tuberalis of the adenophysis, although current research suggests that few tissues are devoid of melatonin receptors.²⁹ MT1 receptors are high affinity receptors that fall into the G-protein coupled receptor superfamily, and binding of melatonin to these receptors results in inhibition of adenylate cyclase activity in target cells.⁵¹ There are two subgroups of the ML1 receptors, ML1a receptors and ML1b receptors.⁵² The ML1 receptors are likely involved in regulation of retinal function, circadian rhythms and reproduction.³¹ The ML2 receptors are low affinity receptors that are coupled to phosphoinositol hydrolysis.³¹ Activation of MT3 receptors inhibits leukotriene B4-induced leukocyte adhesion and decreases intraocular pressure.⁵⁰

Clinical Trials of Melatonin for Sleep Disorders

The circadian phase modulating effects of melatonin point to its potential use in the treatment of circadian rhythm disorders, while the hypnotic/soporific effects of melatonin suggest its potential use in the treatment of insomnia. The use of melatonin in the elderly is considered a potential treatment for sleep disturbances in this population. Similarly, sleep disorders secondary to other medical conditions, such as depression or neurological disorders, may involve circadian rhythm abnormality, and thus could be mitigated by melatonin. A number of randomized controlled trials have been conducted to examine the effect of melatonin in the treatment of various types of insomnia^53–58 such as sleep maintenance insomnia,³⁷ terminal insomnia,⁵⁷ sleep onset insomnia,⁵⁹ psychophysiological insomnia⁶⁰ as well as circadian rhythm disorders such as time zone change (jet lag) syndrome,^61–65 shift work sleep disorder,^66–70 delayed sleep phase syndrome,^71–75 and non-24-hour sleep wake disorder (associated with blindness).^{73 76 77} Randomized controlled trials have also been conducted to examine the effect of melatonin in the treatment of sleep disorders secondary to neurological conditions such as dementia,⁷⁸ Alzheimer's syndrome,⁷⁹ Rett syndrome,⁸⁰ tuberous sclerosis⁸¹ and various other developmental disabilities⁸² as well as disorders secondary to psychiatric conditions such as depression, bipolar disorder⁸³ and seasonal affective disorder.^{84 85} Many case studies have also been conducted, particularly in children, on the use of melatonin for sleep difficulties secondary to neurological syndromes such as Rett syndrome,⁸⁶ Smith-Magenis syndrome,⁸⁷ Angelman syndrome,⁸⁸ autism⁸² and epilepsy.⁸³ Randomized controlled trials have also been conducted to examine the effect of melatonin in the treatment of parasomnias and REM sleep behavior disorder.⁸⁹

Formulation and Dosage of Melatonin Used in Clinical Trials

The trials on melatonin for the treatment of sleep disorders vary in the formulation, timing of administration, frequency and duration of melatonin administration. The providers of melatonin for the various clinical trials on melatonin are diverse; Nestle, Sigma, Neurim, and Regis formulations are common providers. Melatonin products vary from fast-release to sustained release formulations. The formulation of melatonin used in the treatment of sleep disorders may have an effect on sleep outcomes, for example, in one trial, constant-release melatonin improved sleep quality in elderly insomniacs,⁹⁰ but in another trial, fast-release melatonin did not improve sleep quality in elderly insomniacs.⁹¹ By far the most common method of melatonin administration is orally by capsule; the capsule usually consists of melatonin and lactose in a gelatin capsule. However, melatonin has also been administered by a sublingual tablet route, in patch format, and has also been tested intravenously. Commercially available agents are even more variable; melatonin products available include capsule, tablet (oral or sublingual), lozenge, liquid or spray forms. In trials of melatonin, the hormone has been administered orally or by transbuccal patch in dosages between 0.1 and 10 mg. The duration of melatonin administration in these trials varied from a single, one-time dose of melatonin⁹² to multiple doses of melatonin administered for several months.⁸⁶ In most studies, melatonin is administered thirty minutes to two hours before usual bedtime or desired sleep time. The sleep outcomes analyzed in these studies include such measures as sleep onset latency, total sleep time, sleep duration, quality of sleep, number of awakenings, wake time after sleep onset, sleep efficiency as well as alertness, mood and performance. Some of these studies have found a positive effect of melatonin on these outcomes in people with sleep disorders, whereas some have shown no benefit of melatonin administration, that is, no improvement in sleep quality.

Adverse Effects of Melatonin

Compared to some pharmacological treatments for sleep disorders, melatonin has a very short half-life and its effects are short-lived.⁹³ There have been some side effects of melatonin reported, such as drowsiness and headache.^{65 94 95} In general, most trials have not reported any hangover effects of melatonin, although some trials have reported adverse effects of melatonin on performance.⁹⁶ Melatonin administration in epileptic children has been associated with increased seizure activity.⁹⁷ Melatonin has also been associated with deterioration of mood in depression,^{83 98} and has been reported to be associated with development of autoimmune hepatitis in one case.⁹⁹

Status in the United States

Currently, melatonin falls under the Food and Drug Administration's (FDA's) Dietary Supplement Health and Education Act¹⁰⁵ in the category “other dietary supplements”. Melatonin is not considered a drug, since it is a naturally occurring substance¹⁰⁶ and it is designated “generally recognized as safe” (GRAS). Recognizing the lack of a common framework for evaluating the safety of dietary supplements, the Institute of Medicine Food and Nutrition Board has proposed a framework accompanied by six prototype monographs for the evaluation of various dietary supplements, including melatonin.¹⁰⁷

Status in Canada

The Natural Health Products Directorate (NHP) of Health Canada has been re-evaluating natural health products such as melatonin. New NHP regulations have come into effect as of January 1st, 2004, which permit natural health products to be sold in Canada if they meet specific licensing, manufacturing, labelling, and safety standards. Melatonin is now available for sale in Canada.¹⁰⁸

Status in Europe

In the European Union, melatonin is not considered as a foodstuff but rather a medicine or hormone. It is available by prescription only.¹⁰⁹

Status in Australia

Melatonin is an unregistered good under the Therapeutic Goods administration. However, it can be imported for use under the Personal Import Scheme with a prescription.¹¹⁰

Topic Area 1: Physiology and Pharmacology of Melatonin

1. What are the various formulations of melatonin? How are the formulations different in terms of content and quality as well as safety and effectiveness? What is the clinical importance of any observed differences?

2. What is the pharmacology of exogenous melatonin (including pharmacokinetics and pharmacodynamics)? How is it absorbed, distributed, metabolized and excreted? What blood levels are achieved? Does it penetrate the blood/brain barrier?

3. What is the evidence linking endogenous melatonin to sleep cycles?

4. What are the basic mechanisms by which melatonin produces sleepiness?

5. What is the effect of exogenous melatonin on sleep latency, sleep efficiency, and REM latency in normal sleepers?

6. How is endogenous melatonin involved in circadian rhythms?

Topic Area 2: Population at Risk

1. Which sleep disorders would be most effectively managed by treatment with melatonin?

2. Which populations, based on gender, age, ethnicity, genetic factors and co-morbid conditions, would benefit most from treatment with melatonin?

Topic Area 3: Effectiveness of Melatonin

1. What is the effect of exogenous melatonin on people with sleep disorders?

2. What is the appropriate dosage/duration of melatonin for the treatment of sleep disorders? Does the appropriate dosage depend on patients' gender, age, and/or ethnicity?

3. What is the timing of melatonin administration during the sleep/wake cycle that would produce optimum treatment effects?

Topic Area 4: Safety of Melatonin

1. What are the adverse effects of short and long-term use of exogenous melatonin?

2. How do the benefits and harms of exogenous melatonin vary based on dose, timing of administration, and patient factors such as gender, age and ethnicity?

3. How do the benefits and harms of melatonin compare to those of other approved pharmacological treatments for sleep disorders?

Overview

The methods of the University of Alberta Evidence-based Practice Centre (UAEPC) were used to conduct a systematic review and synthesis of evidence relevant to the questions of the review. A number of steps were followed in producing this Evidence Report:

• Comprehensive Search

• Development of Inclusion Criteria

• Study Selection

• Assessment of Study Quality

• Data Extraction

• Data Analysis

Comprehensive Search

Table 2. Biomedical Databases Searched

Table 2. Biomedical Databases Searched

Database	Platform	Dates of Search
MEDLINE®	Ovid	1982 to June Week 4, 2003
Cochrane Central Register of Controlled Trials	Ovid	3rd Quarter, 2003
Science Citation Index	ISI Web of Knowledge	July 4, 2003
Biological Abstracts	SilverPlatter version 4.3	July 4, 2003
International Pharmaceutical Abstracts	OVID	1970 to August, 2003
NLM® Gateway	http://gateway.nlm.nih.gov/gw/Cmd	August 13, 2003
OCLC Papers First and Proceedings First	OCLC FirstSearch	July 11, 2003
TOXLINE	CSA Internet Database Service	July 4, 2003

Table 3. Keywords and Subject Headings used in Searches (more...)

Table 3. Keywords and Subject Headings used in Searches

melatonin	restless legs syndrome
melatonine	nocturnal eating (drinking) syndrome
5-methoxy-N-acetyltryptamine	time-zone change syndrome
N-(2-(5-methoxy-1H-indol-3-yl)ethyl)acetamide	Jet lag
N-acetyl-5-methoxytryptamine	parasomnias
3-(2-acetamidoethyl)-5-methoxyindole	confusional arousals
Acetamide, N-(2-(5-methoxy-1H-indol-3-yl)ethyl)-(9CI)	rhythmic movement disorder
Acetamide, N-(2-(5-methoxyindol-3-yl)ethyl)-, N-(2-(5-methoxyindol-3-eyl)ethyl)acetamide, CAS Reg No: 73-31-4	nocturnal leg cramps
luzindole	nightmares
sleep	nocturnal paroxysmal dystonia
sleep disorders	sudden unexplained nocturnal death syndrome
dyssomnias	snoring
insomnia	congenital central hypoventilation syndrome
narcolepsy	sudden infant death syndrome
hypersomnia	subwakefulness syndrome
central alveolar hypoventilation syndrome	fragmentary myoclonus
periodic limb movement disorder	terrifying hypnagogic hallucinations
circadian

Literature searches were limited to English-language reports of studies on human subjects, with no restrictions applied for age, gender or ethnicity. We searched for reports of phase 1 and 2 clinical trials; phase 3 and 4 randomized clinical trials; quasi-randomized controlled trials; prospective cohorts; case series; registry data; as well as narrative and systematic reviews. In addition to these initial searches, similar searches of MEDLINE® and EMBASE were conducted periodically for more recently published studies that were potentially relevant to the review.

In addition to the electronic searches described above, the reference lists of a random sample of reports, encompassing half of all studies included in the review, were reviewed. The reference lists of narrative and systematic reviews related to melatonin and sleep disorders were also reviewed. We also reviewed the reference list of a Health Canada document on the use of melatonin for the treatment of various disorders as well the reference list of a document from Natural Standard Research Collaboration on the use of melatonin for the treatment of sleep disorders. Lastly, we hand-searched Associated Professional Sleep Society (APSS) Abstracts of 1999 to 2003.

As mentioned above, searches were limited to English-language reports. We sought to avoid the inclusion of non-English language reports in the review, unless deemed necessary, as a means of containing resource requirements for this review, which was already large in scope. The Core Team, in consultation with NCCAM and AHRQ, devised a strategy for inclusion of non-English language reports in the review. Our approach was to evaluate the presence of publication bias across studies relevant to the question of the review pertaining to the effectiveness of melatonin in the treatment of sleep disorders. If publication bias were found, we would expand our search to include non-English language data. We would also expand our search to include non-mainstream data sources. Publication bias refers to a bias in the literature whereby the publication of research is dependent upon the results of research. In Western medical journals, this phenomenon is reflected in the fact that results indicating no effect of an intervention are less likely to be published.¹¹¹ The problem is reversed for CAM-related research, such that studies with negative results are more likely to be published in mainstream Western medical journals (e.g. “MEDLINE®”), and CAM studies with positive results are more likely to be published in smaller journals that may not be accessible on usual search engines.¹¹² Thus, it was necessary to assess this bias across studies related to the effectiveness of melatonin included in this review. We considered the use of non-English language reports and expansion of data sources only for the latter question of the review, since this question related to the main thesis of the report.

Development of Inclusion Criteria

Specific inclusion criteria were developed for each question of the review. In general, only controlled clinical trials were included for each question of the review, except for questions pertaining to the pharmacology of exogenous melatonin and the basic mechanism by which melatonin produces sleepiness. For the latter questions, uncontrolled clinical trials, case-series, cohort, cross-sectional and case-control studies were also included. For all questions of the review, the population of the study could include individuals of any age, gender, ethnicity and socioeconomic status; however, these individuals were required to be free of any type of sleep disorder in the case of the question relating to the effect of melatonin on normal sleepers, and to suffer from a sleep disorder in the case of the question relating to the effect of melatonin on people with sleep disorders. For questions pertaining to the administration of exogenous melatonin to a study population, any formulation, dosage, timing, frequency and duration of melatonin administration was acceptable; however, melatonin was required to be the primary intervention, and in the case of controlled trials, compared to placebo. In addition, a study was included for a particular question of the review if it analyzed at least one of the pre-determined outcomes relevant to that question. Only English-language reports were included in the review.

Question-Specific Inclusion Criteria

What are the various formulations of melatonin? How are the formulations different in terms of content, quality as well as safety and effectiveness?

A study was considered relevant to the portion of this question that pertains to the differences in the safety of various formulations of melatonin if it met inclusion criteria for the question relating to the safety of melatonin, and the formulation of melatonin used in the study was specified in the report. A study was considered relevant to all other portions of this question if it met inclusion criteria for the question relating to the effectiveness of melatonin, and the formulation of melatonin used in the study was specified in the report.

What is the pharmacology of exogenous melatonin, including pharmacokinetics and pharmacodynamics? How is it absorbed, distributed, metabolized and excreted? What blood levels are achieved? What is its half-life? Does it penetrate the blood-brain barrier?

A study was considered relevant to this question of the review if it met the following inclusion criteria:

• it involved human participants

• melatonin was administered to a group of participants

• at least one of the following outcomes was assessed in participants' serum/plasma/blood within hours of melatonin administration and a value was ascribed to it in the text of the report:

  • - half-life of melatonin (t_1/2)

  • - time to reach peak concentration of melatonin (T_max)

  • - peak concentration of melatonin (C_max)

  • - area under the melatonin versus time curve (AUC)

What is the evidence linking endogenous melatonin to sleep cycles?

A study was considered relevant to this question of the review if it met the following inclusion criteria:

• it was a controlled clinical trial

• it involved human participants

• it involved an intervention that altered either endogenous melatonin or the sleep cycle, such as manipulation of light/dark exposure or manipulation of the sleep schedule, respectively. If the intervention involved light administration, a lower intensity light condition was required as a control. If the intervention involved manipulation of the sleep schedule, a normal sleep schedule condition was required as a control.

• it involved only one intervention; a constant routine was not considered a secondary intervention if it was applied to both the experimental and control groups.

• it assessed the levels of melatonin and/or the phase of the melatonin rhythm in participants' blood, urine, saliva or cerebrospinal fluid in the case where the intervention altered the sleep cycle, or it assessed an aspect of participants' sleep cycle in the case where the intervention altered endogenous melatonin.

What are the basic mechanisms by which melatonin produces sleepiness?

Initially, a study was considered relevant to this question of the review if it met the following inclusion criteria:

• it involved human participants

• it involved administration of exogenous melatonin or an intervention that manipulated endogenous melatonin levels

• it characterized a mechanism by which alterations in endogenous melatonin levels affect sleep propensity

Given the lack of studies that met these inclusion criteria, the latter criteria were revised. A study was considered relevant to this question of the review if it met the inclusion criteria of the question relating to the effectiveness of melatonin in normal sleepers or the question relating to the effectiveness of melatonin in people with a sleep disorder, and the report provided a proposed mechanism by which melatonin produces sleepiness based on findings of the study.

What is the effect of exogenous melatonin on sleep latency, sleep efficiency and REM latency in normal sleepers?

A study was considered relevant to this question of the review if it met the following inclusion criteria:

• it was a controlled clinical trial

• it involved participants that did not have a sleep disorder

• melatonin was administered to a group of participants and placebo was administered to a group of participants

• at least one of the following outcomes was assessed:

• sleep onset latency

• sleep efficiency

• REM latency

How is endogenous melatonin involved in circadian rhythms?

The scope of this question was limited to an analysis of how endogenous melatonin is involved in the temperature rhythm. A study was considered relevant to this question of the review if it met the following inclusion criteria:

• it was a controlled clinical trial

• it involved human participants

• it involved an intervention that altered endogenous melatonin or the temperature rhythm, such as manipulation of light/dark exposure or temperature exposure, respectively. If the intervention involved light administration, a lower intensity light condition was required as a control. If the intervention involved manipulation of the temperature rhythm, a normal temperature condition was required as a control.

• it involved only one intervention; a constant routine was not considered a secondary intervention if it was applied to both the experimental and control groups.

• it assessed the levels of melatonin and/or the phase of the melatonin rhythm in participants' blood, urine, saliva or cerebrospinal fluid in the case where the intervention altered the temperature rhythm, or it assessed an aspect of participants' temperature rhythm in the case where the intervention altered endogenous melatonin.

What is the effect of exogenous melatonin on people with sleep disorders?

A study was considered relevant to this question of the review if it met the following inclusion criteria:

• it was a randomized controlled clinical trial

• it involved human participants who suffer from a sleep disorder and this condition was explicitly mentioned in the report

• melatonin was administered to a group of participants and placebo was administered to a group of participants

• at least one of the following outcomes was assessed:

  • - sleep onset latency

  • - sleep efficiency

  • - sleep quality

  • - wakefulness after sleep onset

  • - total sleep time

  • - percent time in REM sleep

Which sleep disorders would be most effectively managed by treatment with melatonin? Which populations based on gender, age, ethnicity, genetic factors and co-morbid conditions, would benefit most from treatment with melatonin?

What is the appropriate dosage/duration of melatonin for the treatment of sleep disorders? Does the appropriate dosage depend on patients' gender, age, and/or ethnicity? What is the timing of melatonin administration during the sleep/wake cycle that would produce optimal treatment effects?

A study was considered relevant to these questions of the review if it met the inclusion criteria for the question relating to the effectiveness of melatonin in people with sleep disorders, and the report provided the information necessary for the study to be incorporated into a subgroup analysis related to at least one variable specified in the question.

What are the adverse effects of short and long-term use of exogenous melatonin?

A study was considered relevant to this question of the review if it met the following inclusion criteria:

• it included human participants

• melatonin was administered to a group of participants and placebo was administered to a group of participants

• it reported on adverse events and/or adverse effects of the interventions

For this question of the review, short-term melatonin use was defined as less than three months duration and long-term melatonin use was defined as three months or greater duration.

How do the benefits and harms of exogenous melatonin vary based on dose, timing of administration, and patient factors such as gender, age and ethnicity?

A study was considered relevant to this question of the review if it met the inclusion criteria for the question relating to the safety of melatonin, and the report provided the information necessary for the study to be incorporated into a subgroup analysis related to at least one variable specified in the question.

How do the benefits and harms of melatonin compare to those of other approved pharmacological treatments for sleep disorders?

A study was considered relevant to this question of the review if it met the inclusion criteria for the question relating to the effectiveness of melatonin in normal sleepers and the question relating to the effectiveness of melatonin in people with sleep disorders, except that melatonin and another pharmacological treatment for sleep disorders, instead of placebo, were administered to groups of participants.

Study Selection

The librarian removed all duplicates of the initial search results. In the first stage of study selection, the titles and abstracts of all potentially relevant articles were screened, independently, by two reviewers and classified as “relevant”, “clearly irrelevant” and “unclear”. A given article was considered “relevant” to the review if it was relevant to at least one key question of the review. The full text of all articles deemed “relevant” or “unclear” by each reviewer was retrieved. In the second stage of screening, the reviewers independently appraised the manuscripts using pre-determined inclusion criteria for each key question of the review. Only studies that met all inclusion criteria for a given question of the review, as determined by both reviewers, were considered relevant to that question. Disagreements among reviewers were resolved by discussion and consensus.

Assessment of Study Quality

For the question pertaining to the effect of melatonin on people with sleep disorders, only randomized controlled trials were used as a source of evidence. Therefore, the Jadad Scale¹¹³ was used to assess the quality of studies relevant to this question. The Jadad Scale assigns studies a quality score of zero to five, with a score of five indicating high quality. The scale assesses the components of randomization, blinding and reporting of dropouts and withdrawals. To our knowledge, neither this scale nor any other has been validated for the quality assessment of crossover trials. However, this scale has been validated for the quality assessment of randomized-controlled trials, and thus, was considered an appropriate quality assessment tool for this review. The concealment of allocation in the randomized-controlled trials was assessed as “adequate”, “inadequate” and “unclear”.¹¹⁴ For all other questions of the review, which relied on evidence from studies of other designs in addition to randomized controlled trials, the Downs and Black Checklist¹¹⁵ was used to assess the quality of studies relevant to these questions. This checklist is partially validated and assesses a number of design components including reporting, internal and external validity, and the statistical power of a study to detect a clinically important difference. Two reviewers assessed study quality, independently, and disagreements were resolved by discussion and consensus. The overall quality of the evidence regarding the safety and effectiveness of melatonin in the treatment of sleep disorders was assessed using the framework developed by the Oxford centre for Evidence-Based Medicine. See Appendix B ^* for Quality Assessment Forms.

Data Extraction

Data were extracted from all reports of studies that were included in the review using a standardized Data Extraction Form. The type of information extracted from reports included details of study design and inclusion/exclusion criteria; details of the population such as gender, age, ethnicity and type of sleep disorder; the number of individuals that were eligible for, and enrolled in, the study; the number of comparison groups and participants allocated to each group; the number of participants who withdrew from the study; details of the intervention such as the formulation, dosage, timing, frequency and duration of melatonin administration as well as the type and frequency of usage of concurrent medication; and results obtained for pre-determined, question-specific outcomes.

Additional information that was extracted from reports included the name of the first author of the report and year of publication of the report; the country where the study took place; the source of funding for the study; authors' objectives and conclusions; and whether an intention-to-treat analysis was planned or performed. A trained reviewer extracted relevant data from a given report and a second reviewer verified the data that were extracted for that article for accuracy and completion. Disagreements between reviewers were resolved by discussion and consensus. See Appendix B ^* for the Data Extraction Form.

Data Analysis

Table 4. Questions of the Review and Type of Analysis (more...)

Table 4. Questions of the Review and Type of Analysis Applied to Data Relevant to these Questions

Questions	Type of Analysis Applied to Data Relevant to Question
Formulations of melatonin	Qualitative and Quantitative
Pharmacology of melatonin	Qualitative
Endogenous melatonin and the sleep cycle	Qualitative
Mechanism of action of melatonin	Qualitative
Effect of melatonin on normal sleepers	Quantitative
Endogenous melatonin and circadian rhythms	Qualitative
Effectiveness of melatonin among types of sleep disorders	Quantitative
Effectiveness of melatonin among types of populations	Quantitative
Effect of melatonin on people with sleep disorders	Quantitative
Appropriate dosage of melatonin for treatment of sleep disorders	Quantitative
Appropriate timing of melatonin administration for treatment of sleep disorders	Quantitative
Adverse effects of melatonin	Quantitative
Adverse effects of melatonin as a function of dose, timing, and patient factors	Quantitative
Melatonin and other pharmacological treatments for sleep disorders	Qualitative

Quantitative Analysis

For all continuous outcomes (e.g. sleep onset latency, sleep efficiency) studies were combined using a Weighted Mean Difference (WMD) with the exception of sleep quality where studies were combined using a Standardized Mean Difference (SMD). Due to the large number of studies with a crossover design, the Inverse Variance Method¹¹⁶ was used to weight the studies. An effectiveness estimate with corresponding 95 percent confidence interval was computed for each outcome.

We were usually able to calculate the effectiveness estimates for each study exactly (i.e. weighted mean difference, standardized mean difference, risk difference), but occasionally, estimates had to be made by extracting from graphs or using medians. Standard errors of the differences were calculated exactly from available data (i.e. individual patient data or exact p-values) whenever possible. For studies with a parallel design, this calculation was usually accomplished with the standard formula for variance of difference of independent variables: var(A-B) = var(A) + var(B). For studies with a crossover design, the standard error was estimated using the formula for variance of difference of dependant variables: var(A-B) = var(A) + var(B) -2ρ(var(A)var(B))^½ and using a correlation estimate of 0.5. In cases where this calculation could not be done, standard errors were estimated using conservative p-values (i.e. p < 0.05), inter-quartile ranges, and extracting from graphs. As a last resort, an average of standard deviations of other studies was used to impute standard deviations of a study.

For studies with a parallel design, change from baseline data was used if available, otherwise final data were used. For studies with a crossover design, final data were always used.

When continuous data were presented for multiple conditions, which we wished to combine, a new mean and standard deviation were computed. If the study had a parallel design, the new mean and standard deviation could be computed exactly using the formula:

where is the mean of the newly formed combined arm, g is the number of groups combined, _i are the means of each group (i may take the value of 1 through g), s _y is the standard deviation of the newly formed combined arm, n _i are the sample sizes of each group, s _i are the standard deviations of each group and N is the total new sample size (the sum of the n _i). If the study had a crossover design, we treated the data as we would a repeated measures experiment. The formula for the mean was the same, but the following formula was used for the standard deviation with the within subject correlation (ρ) being estimated as 0.5.

Dichotomous outcomes (i.e. safety outcomes) were combined using a Risk Difference with corresponding 95 percent confidence interval. Many studies stated that there were no reported adverse events. These were included in the analysis, but a sensitivity analysis excluding them was also performed, since the lack of reporting on adverse events does not necessarily indicate that they did not occur in the study.

All meta-analyses were performed using a Random Effects Model. Bailey¹¹⁷ suggests that the Random Effects Model is more appropriate when making recommendations for management and treatment of the next given patient. Fixed effects were considered in a sensitivity analysis.

All estimates of effectiveness (weighted mean differences, standardized mean differences, and risk differences) were assessed for heterogeneity using the I-squared statistic.¹¹⁸ Based on this statistic, heterogeneity for each outcome was classified as negligible (I² = 0 percent), minimal (I² < 20 percent), moderate (20 percent < I² < 50 percent), or substantial (I² > 50 percent). For our primary outcomes, heterogeneity was explored in subgroup analyses using a number of variables. These variables were: age, gender, ethnicity, use of concurrent medication, formulation, dosage, duration of study, method of measurement, study design, and study quality. For patients with sleep disorders, we also examined type of disorder and allocation concealment, while for subjects with normal sleep patterns, we also examined patient description, time of sleep, and use of multiple sleep onset techniques. Deeks' chi-square statistic¹¹⁹ was used to test for significant heterogeneity reduction in partitioned subgroups.

We tested for publication bias visually using the Funnel Plot and quantitatively using the Rank Correlation Test,¹²⁰ the Graphical Test,¹²¹ and the Trim and Fill Method.¹²²

Qualitative Analysis

What are the various formulations of melatonin? How are the formulations different in terms of content, quality as well as safety and effectiveness? What is the clinical importance of any observed differences?

In order to obtain more detailed information regarding the content and quality of the melatonin formulations that were used in the studies relevant to this question of the review, the corresponding authors of these studies were contacted and asked to respond to a short questionnaire. The following is a list of questions that were posed:

1. What are the constituents of the formulation and the relative proportion of each constituent?

2. Was the melatonin component natural or synthetic and what was the purity of this component?

3. Do you have information on the pharmacology of this formulation in humans?

The information provided by corresponding authors was used to supplement information that was provided in the report of these studies. It was used to answer the portion of the question regarding the differences in the content and quality of melatonin formulations that have been used in relevant studies.

What is the pharmacology of exogenous melatonin, including pharmacokinetics and pharmacodynamics? How is it absorbed, distributed, metabolized, excreted? What blood levels are achieved? What is its half-life? Does it penetrate the blood brain barrier?

A detailed Evidence Table was created outlining the design details and results of studies relevant to this question of the review. The results of the studies were also summarized in a Summary Table. The key elements of these tables were summarised.

What is the evidence linking endogenous melatonin to sleep cycles? AND How is endogenous melatonin involved in circadian rhythms?

The studies relevant to these questions of the review were categorized according to details of study design. First, studies were categorized according the type of intervention that was employed; in the case of the question relating to the link between endogenous melatonin and the sleep cycle, these interventions were either alterations in lighting, alterations in the sleep schedule or exposure to another intervention that altered either endogenous melatonin or the sleep cycle; in the case of the question relating to the link between endogenous melatonin and the temperature rhythm, these interventions were either alterations in lighting or temperature. The studies were further subdivided into studies involving participants with or without a sleep disorder or with a disorder other than a sleep disorder. Each of these categories of studies was further categorized according to timing of the intervention. The analysis of data pertaining to these sub-categories of studies began with a summary of the conditions of the intervention and characteristics of the population and continued with a synthesis of the results of each study as they pertain to the question being addressed.

What are the basic mechanisms by which melatonin produces sleepiness?

The studies relevant to this question of the review were categorized as involving participants with or without a sleep disorder and further grouped according to the proposed mechanism by which melatonin produces sleepiness. The findings upon which the proposed mechanisms were based were described for each category of studies.

How do the benefits and harms of melatonin compare to those of other approved pharmacological treatments for sleep disorders?

The studies relevant to this question of the review were categorized as involving participants with or without a sleep disorder and the effects of melatonin and another pharmacological treatment for sleep disorders were compared in terms of their effects on one or more of the following outcomes: sleep onset latency, sleep efficiency, sleep quality, wakefulness after sleep onset, total sleep time and percent time in REM sleep. The adverse events accompanying both treatments were also compared.

Sleep Onset Latency

Primary Analysis

Figure 1. Meta-Graph: Sleep Onset Latency in Normal (more...)

   Figure 1. Meta-Graph: Sleep Onset Latency in Normal Sleepers

There were a total of twenty studies with data on sleep onset latency for normal sleepers. The combined estimate comparing melatonin to placebo showed that melatonin caused a statistically significantly reduction in sleep onset latency (weighted mean difference (WMD): -3.9 minutes (min.); 95 percent confidence interval (CI): -5.3 min, -2.6 min). This effect appears to be clinically insignificant. Heterogeneity among the studies was moderate (I²: 47.1 percent). Nineteen of the twenty studies had a point estimate that favoured melatonin (Figure 3-1).

Subgroup and Sensitivity Analyses

Table 6. Subgroup and Sensitivity Analysis: Sleep Onset (more...)

Table 6. Subgroup and Sensitivity Analysis: Sleep Onset Latency in Normal Sleepers

Subgroup	Categorization	Number of studies	Point Estimate (min)	95 percent Confidence Interval (min)	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	14	-4.4	-6.3, -2.5	Moderate (I2:46.5 percent)	0.24
	Mixed	6	-3.2	-5.4, -1.0	Substantial (I2: 51.1 percent)
Use of Concurrent Medication	Yes	1	-0.5	-2.5, 1.5	NA	NA
	No	10	-4.0	-5.3, -2.6	Minimal (I2: 11.0 percent)
Dosage of Melatonin Administration	< 1 mg	5	-7.6	-11.7, -3.5	Moderate (I2: 37.9 percent)	NA
	1–3 mg	10	-6.1	-9.1, -3.2	Substantial (I2: 54.0 percent)
	4–5 mg	6	-2.6	-4.2, -1.1	Substantial (I2: 57.5 percent)
	6–10 mg	7	-6.1	-8.9, -3.3	Moderate (I2: 30.6 percent)
	> 10 mg	2	-3.4	-7.6, 0.8	Negligible (I2: 0 percent)
Timing of Melatonin Administration	< 1800h	11	-4.6	-6.0, -3.2	Moderate (I2: 29.4 percent)	0.002
	> 1800h	10	-3.2	-5.5, -1.0	Moderate (I2: 41.9 percent)
Duration of Melatonin Administration	< 1 week	14	-4.2	-5.6, -2.8	Moderate (I2: 43.7 percent)	0.03
	1–2 weeks	1	-6.3	-14.3, 1.7	NA
	3–4 weeks	5	-2.5	-6.9, 2.0	Moderate (I2: 27.9 percent)
Method of Measurement of Sleep Outcomes	Polysomnography	14	-3.7	-5.0, -2.4	Moderate (I2: 25.0 percent)	0.001
	Actigraphy	4	-2.1	-4.6, 0.4	Moderate (I2: 34.5 percent)
	Questionnaire	2	-8.8	-12.5, -5.2	Negligible (I2: 0 percent)
Explicit Statement in Report that Subjects did not Suffer from a Sleep Disorder	Yes	8	-3.9	-6.5, -1.3	Substantial (I2: 55.1 percent)	0.35
	No	12	-4.1	-5.8, -2.4	Moderate (I2: 43.5 percent)
Time of Sleep	Daytime	9	-4.6	-6.0, -3.2	Moderate (I2: 30.9 percent)	NA
	Night time	13	-3.0	-4.9, -1.0	Moderate (I2: 32.5 percent)
Use of Multiple Sleep Onset Latency Test	Yes	3	-4.2	-5.7, -2.6	Negligible (I2: 0 percent)	0.16
	No	17	-4.0	-5.7, -2.2	Substantial (I2: 51.1 percent)
Study Design	Parallel	2	-4.5	-10.6, 1.6	Negligible (I2: 0 percent)	0.67
	Crossover	18	-3.9	-5.4, -2.5	Substantial (I2: 51.8 percent)
Quality Score	10–15 (low)	4	-3.8	-4.9, -2.7	Negligible (I2: 0 percent)	0.01
	16–20 (moderate)	15	-5.0	-7.4, -2.7	Moderate (I2: 44.3 percent)
	21–25 (high)	1	-0.5	-2.5, 1.5	NA

Abbreviations: NA: not applicable, min: minutes

The results of all the subgroup and sensitivity analyses are summarized in Table 6. Subgroups that could not be analyzed for sleep onset latency included age (all studies involved adult participants), ethnicity (not specified in any study), and formulation of melatonin (all studies used fast-release melatonin).

As can be seen from Table 6, none of the subgroups produced markedly different results from those obtained from the primary analysis. All of the subgroup point estimates favoured melatonin and this result was significant in most cases; the only exceptions were cases with only a small number of studies. Four of the partitions showed a significant Deeks' chi-square value, indicating that heterogeneity was significantly reduced by the partition. Method of measurement of sleep outcomes and timing of melatonin administration had the largest effect. Method of measurement of sleep outcomes showed the questionnaire group to have a larger effect than the other methods, but this could be due to the smaller sample size of only two studies. For all other subgroup analyses, the subgroups had overlapping confidence intervals.

Using the Fixed Effects Model instead of the Random Effects Model tightened the confidence interval but did not greatly change the point estimate (WMD: -3.2; 95 percent CI: -4.0, -2.5).

Assessment of Publication Bias

Figure 2. Funnel Plot: Sleep Onset Latency in Normal (more...)

   Figure 2. Funnel Plot: Sleep Onset Latency in Normal Sleepers

Abbreviations: _ES: effect size, _seES: standard error of effect size; note that the smaller studies are associated with a larger standard error.

There was evidence of asymmetry in the funnel plot, indicating possible publication bias (Figure 3-2). In the numerical tests, the results of Egger's Graphical Test showed a p-value that was borderline significance, indicating possible bias (p-value: 0.049). Begg's Test (Kendall's S=-39 giving a p-value of 0.217 for 20 studies) and Duvall's Trim and Fill Test did not indicate publication bias.

Sleep Efficiency

Primary Analysis

Figure 3. Meta-Graph: Sleep Efficiency in Normal Sleepers (more...)

   Figure 3. Meta-Graph: Sleep Efficiency in Normal Sleepers

There were a total of thirteen studies with data on sleep efficiency for normal sleepers. A statistically significant increase in sleep efficiency with melatonin was calculated when comparing melatonin to placebo (WMD: 2.3 percent; 95 percent CI: 0.7 percent, 3.9 percent). This effect appears to be clinically insignificant. Heterogeneity among the studies was substantial (I²: 53.9 percent). Eleven of the thirteen studies showed a point estimate indicating increased sleep efficiency with melatonin, one study indicated neutrality between melatonin and placebo, and one study indicated increased sleep efficiency with placebo (Figure 3-3).

Subgroup and Sensitivity Analyses

Table 7. Subgroup and Sensitivity Analysis: Sleep Efficiency (more...)

Table 7. Subgroup and Sensitivity Analysis: Sleep Efficiency in Normal Sleepers

Subgroup	Categorization	Number of studies	Point Estimate (percent)	95 percent Confidence Interval (percent)	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	11	2.8	0.6, 4.9	Substantial (I2: 58.1 percent)	0.73
	Mixed	2	1.8	-0.9, 4.5	Substantial (I2: 52.2 percent)
Use of Concurrent Medication	Yes	6	2.2	0.1, 4.3	Moderate (I2: 24.4 percent)	NA
	No	1	0.9	-0.5, 2.2	NA
Dosage of Melatonin Administration	< 1 mg	4	3.4	0.6, 6.1	Negligible (I2: 0 percent)	NA
	1–3 mg	6	5.1	2.9, 7.3	Minimal (I2: 12.7 percent)
	4–5 mg	2	0.7	-0.4, 1.9	Negligible (I2: 0 percent)
	6–10 mg	5	4.8	-1.1, 10.7	Substantial (I2: 76.6 percent)
	> 10 mg	1	1.8	-1.5, 5.1	NA
Timing of Melatonin Administration	< 1800h	6	1.0	-6.0, 2.5	Moderate (I2: 43.2 percent)	0.01
	> 1800h	7	4.4	1.5, 7.4	Moderate (I2: 43.4 percent)
Duration of Melatonin Administration	< 1 week	9	3.3	-0.9, 5.7	Substantial (I2: 65.5 percent)	0.46
	1–2 weeks	0	NA	NA	NA
	3–4 weeks	4	1.1	-0.2, 2.3	Negligible (I2: 0 percent)
Method of Measurement of Sleep Outcomes	Polysomnography	10	3.0	0.6, 5.3	Substantial (I2: 62.9 percent)	0.78
	Actigraphy	3	1.3	0.0, 2.5	Neglibible (I2: 0 percent)
	Questionnaire	0	NA	NA	NA
Explicit Statement in Report that Subjects do not Suffer from a Sleep Disorder	Yes	6	1.5	-0.2, 3.1	Moderate (I2: 33.8 percent)	0.41
	No	7	3.7	0.5, 6.9	Substantial (I2: 66.4 percent)
Time of Sleep	Daytime	5	8.0	1.0, 15.0	Substantial (I2: 70.6 percent)	NA
	Night time	10	1.2	-0.0, 2.4	Moderate (I2: 20.2 percent)
Study Design	Parallel	1	0.0	-4.1, 4.1	NA	0.50
	Crossover	12	2.6	0.9, 4.3	Substantial (I2: 57.0 percent)
Quality Score	10–15 (low)	2	7.5	-5.4, 20.3	Substantial (I2: 82.2 percent)	0.44
	16–20 (mod.)	10	2.4	0.4, 4.5	Substantial (I2: 52.1 percent)
	21–25 (high)	1	0.9	-0.5, 2.2	NA

Abbreviations: NA: not applicable

The same subgroups that were analysed for sleep onset latency were also analysed for this outcome except for the use of the multiple sleep onset latency test (there were no studies employing this test) and study design (all studies had a crossover design). The results are summarized in Table 7.

While most of the subgroups produced estimates there were not markedly different from the primary analysis, some differences are noteworthy. The most striking difference was in time of sleep; the efficiency effect of melatonin was much more prominent in the daytime sleepers. Timing of melatonin administration was the only partitioned result that showed a significant reduction in heterogeneity, but the confidence intervals of the two groups overlapped.

Using the Fixed Effects Model in place of the Random Effects Model gave a slightly lower effectiveness estimate of sleep efficiency but remained significant (WMD: 1.38; 95 percent CI: 0.5, 2.3).

Assessment of Publication Bias

Figure 4. Funnel Plot: Sleep Efficiency in Normal Sleepers (more...)

   Figure 4. Funnel Plot: Sleep Efficiency in Normal Sleepers

Abbreviations: _ES: effect size, _seES: standard error of effect size; note that the smaller studies are associated with a larger standard error.

All four tests of asymmetry indicated the possible presence of publication bias (Figure 3-4). The funnel plot showed a marked asymmetry that was confirmed by both Begg's Rank Correlation Test (Kendall's S=41—with 13 studies giving a continuity corrected p-value of 0.014) and Egger's Regression Test (bias p-value of 0.005). Duval's Trim and Fill Algorithm added 5 studies to the analysis and gave a modified estimate that was not significant (WMD: 0.9; 95 percent CI: -0.8, 2.7).

REM Latency

Primary Analysis

Figire 5/ Meta-Graph: REM Latency in Normal Sleepers (more...)

   Figire 5/ Meta-Graph: REM Latency in Normal Sleepers

Eleven studies had data on REM latency for normal sleepers. The point estimate showed that REM latency was slightly higher with melatonin but the difference was not significant (WMD: 2.6 min.; 95 percent CI: -4.1 min, 9.2 min). Heterogeneity among the studies was substantial (I²: 55.2 percent). Six of the 11 studies had a point estimate that showed an increase in REM latency for melatonin, while five studies showed a decrease (Figure 3-5).

Sensitivity and Subgroup Analysis

Table 8. Subgroup and Sensitivity Analysis: REM Latency (more...)

Table 8. Subgroup and Sensitivity Analysis: REM Latency in Normal Sleepers

Subgroup	Categorization	Number of studies	Point Estimate (min)	95 percent Confidence Interval (min)	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	9	-1.2	-6.2, 3.9	Minimal (I2:17.8 percent)	0.06
	Mixed	2	17.7	-8.0, 43.5	Substantial (I2: 80.6 percent)
Use of Concurrent Medication	Yes	0	NA	NA	NA	NA
	No	6	-4.0	-8.1, 0.1	Negligible (I2: 0 percent)
Dosage of Melatonin Administration	< 1 mg	4	-7.1	-18.9, 4.7	Moderate (I2: 34.2 percent)	NA
	1–3 mg	6	12.7	6.8, 18.6	Negligible (I2: 0 percent)
	4–5 mg	3	15.2	-7.2, 37.5	Substantial (I2: 92.0 percent)
	6–10 mg	3	3.8	-7.9, 15.5	Negligible (I2: 0 percent)
	> 10 mg	1	13.6	-45.9, 18.7	NA
Timing of Melatonin Administration	< 1800h	4	4.3	-7.0, 15.5	Mimimal (I2: 13.9 percent)	0.21
	> 1800h	7	2.6	-5.6, 10.7	Substantial (I2: 65.3 percent)
Duration of Melatonin Administration	< 1 week	8	0.4	-5.3, 6.2	Moderate (I2: 39.2 percent)	0.04
	1–2 weeks	0	NA	NA	NA
	3–4 weeks	3	11.2	-13.5, 35.9	Substantial (I2: 69.2 percent)
Explicit Statement in Report that Subjects do not Suffer from a Sleep Disorder	Yes	6	2.5	-7.5, 12.5	Substantial (I2: 68.7 percent)	0.09
	No	5	4.8	-2.2, 11.9	Negligible (I2: 0 percent)
Time of Sleep	Daytime	2	12.4	-0.0, 24.9	Negligible (I2: 0 percent)	NA
	Night time	10	0.9	-5.9, 7.6	Substantial (I2: 50.0 percent)
Use of Multiple Sleep Onset Latency Test	Yes	1	12.6	-0.3, 25.5	NA	0.04
	No	10	1.2	-5.6, 7.9	Substantial (I2: 50.3 percent)
Quality Score	10–15 (low)	2	3.9	-20.3, 28.1	Substantial (I2: 54.0 percent)	0.11
	16–20 (mod.)	9	1.8	-5.2, 8.8	Substantial (I2: 54.5 percent)
	21–25 (high)	0	NA	NA	NA

Abbreviations: REM: rapid eye movement, NA: not applicable, min: minutes

The subgroups analysed for REM latency in normal sleepers were the same as those analysed for sleep onset latency with the exception of method of measurement (all studies used polysomnography to measure sleep outcomes) and study design (all studies had a crossover design). The results are summarized in Table 8.

All of the subgroup point estimates, except one, showed a non-significant difference in REM latency between melatonin and placebo; the 1–3 mg dosage subgroup, showed increased REM latency compared to both higher and lower doses. Although three of the partitioned subgroups indicated significant reduction in heterogeneity, besides the group already mentioned, there was no indication that REM latency is affected by melatonin.

The point estimate using the Fixed Effects Model in place of the Random Effects Model favoured placebo rather than melatonin, but was not statistically significant (WMD: -0.4, 95 percent CI: -3.9, 3.1).

Assessment of Publication Bias

Figure 6. Funnel Plot: REM Latency in Normal Sleepers (more...)

   Figure 6. Funnel Plot: REM Latency in Normal Sleepers

Abbreviations: _ES: effect size, _seES: standard error of effect size; note that the smaller studies are associated with a larger standard error.

While a visual inspection of the funnel plot seems to indicate some minor asymmetries (Figure 3-6), there was no evidence of publication bias in any of the quantitative tests. Begg's Test (Kendall's S=2; with 11 studies this is a p-value of 0.938), Egger's Test (bias p-value: 0.257) and Duval's Trim and Fill test (no studies added) did not give any indication of any asymmetry in the funnel plot.

What is the effect of exogenous melatonin on people with sleep disorders?

Thirty randomized controlled trials were relevant to this question. The quality of relevant studies was assessed using the Jadad Scale.¹¹³ The overall quality scores ranged from two to five on a five-point scale. Three studies had a quality score of two,^{125 137 138} eight studies had a quality score of three,^{66 70 80 132 139–142} 12 studies had a quality score of four^{54 60 63 69 71 73 74 81 82 90 131 135} and seven studies had a quality score of five.^{57 61 67 68 78 91 143} It was unclear whether there was adequate concealment of treatment allocation in all studies except six.^{67–69 78 90 143} All studies were described as randomized and double-blind and a description of withdrawals and dropouts was provided in all reports except six.^{80 125 137–139 141} The method of randomization was described and was appropriate in nine studies,^{57 61 63 67 68 78 91 135 143} while the method of randomization was not described in all other studies. All reports except 11^{60 63 66 70 125 132 135 137 138 140 142} provided a description of an appropriate method of double-blinding; the method of double-blinding was not described in all other reports. These studies differed from studies involving normal sleepers mentioned above in that participants were randomized to intervention groups. However, other aspects of study design such as allocation concealment and blinding were similar to studies involving normal sleepers. See Evidence Table C-2 for a description of design characteristics and overall quality scores of studies relevant to this question of the review.

The following six outcomes were examined with respect to effectiveness of melatonin in people with sleep disorders, the first two being the primary outcomes:

• Sleep Onset Latency: Defined as the amount of time between the subject laying down to sleep, and the onset of stage 1 sleep.

• Sleep Efficiency: Defined as the amount of time the subject spent asleep as a percentage of the total time spent in bed.

• Sleep Quality: Defined as the overall quality of sleep attained. This outcome was measured differently across studies and was thus combined using a Standardized Mean Difference.

• Wakefulness After Sleep Onset (WASO): This is the amount of time spent awake in bed following the first attainment of stage one sleep.

• Total Sleep Time: Defined as the total time spent asleep while in bed.

• Percentage Time in REM Sleep: Defined as the total time spent in REM sleep as a percentage of total sleep time.

People with a Primary Sleep Disorder

Sleep Onset Latency

Figure 7. Meta-Graph: Sleep Onset Latency in People (more...)

   Figure 7. Meta-Graph: Sleep Onset Latency in People with a Primary Sleep Disorder

Primary Analysis. There were 12 studies that examined sleep onset latency in patients with a primary sleep disorder. The combined weighted mean difference (WMD) of the studies showed that those in the melatonin group had a statistically significant shorter sleep onset latency period than those in the placebo group (WMD: -10.7 min.; 95 percent CI: -17.6 min., -3.7 min.), although there was substantial heterogeneity among the studies (I²: 81.5 percent). This effect appears to be clinically insignificant. Nine of the 12 studies showed a difference that favoured melatonin (Figure 3-7). These results are based on trials of four weeks or less in duration.

Table 9. Subgroup and Sensitivity Analyses: Sleep Onset (more...)

Table 9. Subgroup and Sensitivity Analyses: Sleep Onset Latency in People with a Primary Sleep Disorder

Subgroup	Categorization	Number of studies	Point Estimate (min)	95 percent Confidence Interval (min)	Heterogeneity	Deeks' Chi-Square p-value
Age	Children	1	-17.0	-33.5, -0.5	NA	0.002
	Adult	7	-11.2	-27.7, 5.4	Substantial (I2: 84.0 percent)
	Elderly	4	-7.8	-17.4, 1.7	Substantial (I2: 69.6 percent)
Ethnicity	Caucasian	2	-17.5	-33.9, -1.2	Negligible (I2: 0 percent)	NA
Use of Concurrent Medication	Yes	1	-14.0	-28.7, 0.7	NA	NA
	No	1	1.7	-30.6, 34.0	NA
Dosage of Melatonin Administration	< 1 mg	2	-0.9	-5.4, 3.6	Negligible (I2: 0 percent)	NA
	1–3 mg	5	-6.0	-12.9, 0.8	Moderate (I2: 28.0 percent)
	4–5 mg	6	-13.3	-30.3, 3.7	Substantial (I2: 90.0 percent)
Duration of Melatonin Administration	< 1 week	1	-9.7	-20.5, 1.1	NA	0.07
	1–2 weeks	5	-7.9	-17.5, 1.6	Negligible (I2: 0 percent)
	3–4 weeks	6	-12.4	-21.9, -2.8	Substantial (I2: 90.3 percent)
Method of Measurement of Sleep Outcomes	Polysomnography	5	-14.2	-27.9, -0.5	Substantial (I2: 89.5 percent)	0.001
	Actigraphy	3	-8.1	-21.3, 5.0	Substantial (I2: 70.0 percent)
	Questionnaire	4	-2.3	-23.5, 18.9	Negligible (I2: 0 percent)
Primary Diagnosis	Insomnia	10	-4.3	-8.4, -0.1	Moderate (I2: 44.9 percent)	< 0.00001
	Delayed Sleep-Phase Syndrome	2	-38.8	-50.3, -27.3	Negligible (I2: 0 percent)
Study Design	Parallel	2	-17.1	-32.4, -1.8	Negligible (I2: 0 percent)	0.03
	Crossover	10	-9.9	-17.2, -2.5	Substantial (I2: 83.8 percent)
Quality Score	Moderate (2–3)	4	-5.4	-11.8, 0.9	Moderate (I2: 37.2 percent)	0.13
	High (4–5)	8	-13.1	-28.9, 2.8	Substantial (I2: 86.7 percent)
Allocation Concealment	Unclear	10	-9.6	-17.2, -2.0	Substantial (I2: 82.7 percent)	0.007
	Adequate	2	-15.3	-26.3, -4.4	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable, min: minutes

Sensitivity and Subgroup Analysis. Table 9 summarizes the results of the subgroup and sensitivity analyses of sleep onset latency in patients with a primary sleep disorder. Subdivision by gender was not possible as all studies were mixed gender. Only two studies specified ethnicity—both were Caucasian. For use of concurrent medications, all but two studies either did not specify or had a mixture of patients on and off medications. For dosage, three studies used multiple doses and were included in two or more groups, while two studies were excluded since they did not specify dosage.

Many of these sub groupings significantly reduced heterogeneity despite retaining a substantial heterogeneity statistic in at least one subgroup. The one subgroup that is noteworthy is that of primary diagnosis, which substantially reduced the heterogeneity and is the only sub grouping that gave results with non-overlapping confidence intervals. This variable appears to explain much of the heterogeneity in the primary analysis.

Using the Fixed Effects Model rather than Random Effects Model greatly changes the results of the primary analysis as well as the conclusion. One study⁹¹ received nearly 92 percent of the weight and the new difference was not significant (WMD = -0.32; 95 percent CI -1.3, 0.6).

Figure 8. Funnel Plot: Sleep Onset Latency in People (more...)

   Figure 8. Funnel Plot: Sleep Onset Latency in People with a Primary Sleep Disorder

Abbreviations: _ES: effect size, _seES: standard error of effect size; note that the smaller studies are associated with a larger standard error.

Assessment of Publication Bias. Assessment of publication bias for sleep onset latency of patients with primary sleep disorders was performed. No obvious asymmetry was evident in the funnel plot (Figure 3-8). Begg's test gave a Kendal's score of S=6—with 12 studies, this constitutes a continuity corrected p-value of 0.732 indicating no publication bias. Performing Duval's Trim and Fill Method, no studies were added to the meta-analysis, and the final results were identical to the primary analysis. The only test that showed any indication of publication bias was Egger's Graphical test, which had a p-value of 0.027 on the bias of the funnel plot.

Sleep Efficiency

Figure 9. Meta-Graph: Sleep Efficiency in People with (more...)

   Figure 9. Meta-Graph: Sleep Efficiency in People with a Primary Sleep Disorder

Primary Analysis. Nine trials were included in the analysis of sleep efficiency for people with primary sleep disorders. Although the WMD did favour melatonin, the difference was not significant (WMD: 1.5 percent; 95 percent CI: -0.7 percent, 3.6 percent) and the heterogeneity among the studies was substantial (I²: 62.8 percent). Six out of the nine studies showed a point estimate that favoured melatonin (Figure 3-9). These results are based on trials of four weeks or less in duration.

Table 10. Subgroup and Sensitivity Analyses: Sleep (more...)

Table 10. Subgroup and Sensitivity Analyses: Sleep Efficiency in People with a Primary Sleep Disorder

Subgroup	Categorization	Number of studies	Point Estimate (percent)	95 percent Confidence Interval (percent)	Heterogeneity	Deeks' Chi-Square p-value
Age	Adult	6	-0.0	-1.6, 1.5	Minimal (I2: 16.1 percent)	0.004
	Elderly	3	3.6	-0.8, 8.0	Substantial (I2: 73.0 percent)
Use of Concurrent Medication	Yes	1	8.0	4.1, 11.9	NA	NA
	No	1	-1.4	-4.3, 1.5	NA
Dosage of Melatonin Administration	< 1 mg	2	3.4	-4.6, 11.3	Substantial (I2: 78.8 percent)	NA
	1–3 mg	5	2.4	-1.7, 6.5	Substantial (I2: 77.9 percent)
	4–5 mg	4	-0.0	-1.4, 1.3	Negligible (I2: 0 percent)
Duration of Melatonin Administration	< 1 week	1	0.3	-4.4, 5.0	NA	0.91
	1–2 weeks	5	0.8	-2.1, 3.8	Moderate (I2: 45.6 percent)
	3–4 weeks	3	2.6	-2.1, 7.2	Substantial (I2: 85.7 percent)
Method of Measurement of Sleep Outcomes	Polysomno-graphy	5	0.2	-2.1, 2.6	Moderate (I2: 31.1 percent)	0.48
	Actigraphy	3	3.1	-1.2, 7.5	Substantial (I2: 85.9 percent)
	Questionnaire	1	-3.0	-18.3, 12.3	NA
Primary Diagnosis	Insomnia	8	1.7	-0.8, 4.1	Substantial (I2: 67.3 percent)	0.75
	Sleep-Phase Syndrome	1	0.2	-3.7, 4.1	NA
Quality Score	Moderate (2–3)	4	1.7	-1.2, 4.6	Moderate (I2: 34.0 percent)	0.39
	High (4–5)	5	1.3	-1.9, 4.5	Substantial (I2: 75.3 percent)
Allocation Concealment	Unclear	8	0.3	-0.9, 1.5	Minimal (I2: 6.3 percent)	0.0002
	Adequate	1	8.0	4.1, 11.9	NA

Abbreviations: NA: not applicable

Sensitivity and Subgroup Analysis. Table 10 summarizes the results of the subgroup analysis of sleep efficiency in patients with a primary sleep disorder. Subdivision by gender and ethnicity was not possible since all studies were of mixed gender and none of the studies specified ethnicity. For use of concurrent medications, all studies either did not specify or had a mixture of patients on and off medications. For subdivision by dosage, three studies used multiple doses and were in multiple groups; one study was excluded, as it did not specify dosage.

The only sub-grouping that is noteworthy is that of allocation concealment. Removing the study by Garfinkel et al.,⁹⁰ which was considered to have adequate allocation concealment, removed the vast majority of the heterogeneity from the analysis and the result based on this study was significant. This finding is likely due to chance, since allocation concealment has been associated with smaller, rather than larger, effect sizes.¹¹⁴ The remaining eight studies showed no effect. The only other sub grouping that significantly reduced heterogeneity was age, but the resulting confidence intervals were non-overlapping and non-significant.

If we use a fixed effects model instead of a random effects model in our analysis, our point estimate decreases slightly and our confidence interval tightens, but the result is still non-significant (WMD: 0.8; 95 percent CI: -0.3, 1.8).

Figure 10. Funnel Plot: Sleep Efficiency in People (more...)

   Figure 10. Funnel Plot: Sleep Efficiency in People with a Primary Sleep Disorder

Abbreviations: _ES: effect size, _seES: standard error of effect size; note that the smaller studies are associated with a larger standard error.

Assessment of Publication Bias. The funnel plot showed some asymmetry (Figure 3-10), however all other tests did not indicate publication bias. Begg's test gave a Kendal's score of S=7—with nine studies, this constitutes a continuity corrected p-value of 0.529. Egger's Graphical Test had a p-value of 0.441 on the bias of the funnel plot. Finally, Duval's Trim and Fill Method had no studies added to the meta-analysis, and the final results were identical to the primary analysis.

People with a Secondary Sleep Disorder

Sleep Onset Latency

Figure 11. Meta-Graph: Sleep Onset Latency in People (more...)

   Figure 11. Meta-Graph: Sleep Onset Latency in People with a Secondary Sleep Disorder

Primary Analysis. There were six trials involving melatonin administration to individuals with a secondary sleep disorder that reported on sleep onset latency. Their combined estimate favoured melatonin but was non-significant (WMD: -13.2 min.; 95 percent CI: -27.3 min., 0.9 min.). Heterogeneity among the studies was substantial (I²: 79.2 percent) due primarily to one study¹³² that had a very small standard deviation and an estimate very different from the other five studies. This study had a point estimate that favoured placebo, while the other five studies had point estimates that favoured melatonin (Figure 3-11). These results are based on trials of four weeks or less in duration.

Table 11. Subgroup and Sensitivity Analyses: Sleep (more...)

Table 11. Subgroup and Sensitivity Analyses: Sleep Onset Latency in People with a Secondary Sleep Disorder

Subgroup	Categorization	Number of studies	Point Estimate (min)	95 percent Confidence Interval (min)	Heterogeneity	Deeks' Chi-Square p-value
Age	Children	3	-18.1	-29.4, -6.8	Negligible (I2: 0 percent)	0.0001
	Adult	3	-6.6	-24.6, 11.4	Substantial (I2: 79.2 percent)
Gender	Female	1	-12.9	-27.6, 1.8	NA	NA
Dosage of Melatonin Administration	1–3 mg	2	-4.6	-29.8, 20.6	Substantial (I2: 78.1 percent)	NA
	4–5 mg	1	-23.4	-45.2, -1.6	NA
	6–10 mg	1	-13.5	-32.5, 5.5	NA
Duration of Melatonin Administration	1–2 weeks	2	-25.7	-43.3, -8.0	Negligible (I2: 0 percent)	< 0.00001
	3–4 weeks	2	-4.6	-29.8, 20.6	Substantial (I2: 78.1 percent)
	> 4 weeks	2	-13.1	-24.8, -1.5	Negligible (I2: 0 percent)
Method of Measurement of Sleep Outcomes	Polysomnography	1	5.8	2.5, 9.1	NA	< 0.00001
	Actigraphy	3	-14.5	-25.0, -4.1	Negligible (I2: 0 percent)
	Questionnaire	2	-25.7	-43.3, -8.0	Negligible (I2: 0 percent)
Co-Morbidity	Schizophrenia	2	-4.6	-29.8, 20.6	Substantial (I2: 78.1 percent)	NA
Study Design	Parallel	1	-13.5	-32.5, 5.5	NA	0.08
	Crossover	5	-13.5	-29.7, 2.8	Substantial (I2: 81.0 percent)
Allocation Concealment	Unclear	5	-17.4	-26.4, -8.4	Negligible (I2: 0 percent)	< 0.00001
	Adequate	1	5.8	2.5, 9.1	NA

Abbreviations: NA: not applicable, min: minutes

Sensitivity and Subgroup Analysis. Table 11 summarizes the results of the subgroup and sensitivity analyses of sleep onset latency for patients with a secondary sleep disorder. Subgroups not analysed below were ethnicity (not specified in any study), concurrent medications (all patients taking additional medications in all studies), and Jadad score (same for all studies). For gender, all study populations were mixed except for a study by McArthur et al.,⁸⁰ which was all female. For dosage, only four studies are listed since one study did not specify dosage,⁸² and another gave varying doses based on weight.⁸⁰ Primary disorder was different for all studies except for two studies that both examined patients with schizophrenia.^{131 132}

One study by Shamir et al.¹³² completely dictated the results of the subgroup analysis. Subgroups that omitted this study, showed a significant result in favour of melatonin with negligible heterogeneity, while subgroups that did include this study were non-significant with substantial heterogeneity. The two sub groupings in which this study stood alone do not shed much light on the reason for the difference. The Shamir et al. study¹³² was the only study that used polysomnography to measure sleep outcomes and also the only study that was considered to have adequate allocation concealment.

Using the Fixed Effects Method instead of the Random Effects Method drastically changed the results due to the vast majority of the weight being assigned to the study by Shamir et al.¹³² The point estimate for sleep onset latency favoured placebo and was non-significant (WMD: 3.0; 95 percent CI: -0.1, 6.1).

Assessment of Publication Bias. With only six studies analysing sleep onset latency in patients with a secondary sleep disorder, the number of studies was deemed too few to do any meaningful tests for publication bias.

Sleep Efficiency

Figure 12. Meta-Graph: Sleep Efficiency in People with (more...)

   Figure 12. Meta-Graph: Sleep Efficiency in People with a Secondary Sleep Disorder

Primary Analysis. There were six trials for which data were available for comparing melatonin to placebo in sleep efficiency. The WMD of the six studies showed a statistically significant effect that favoured melatonin (WMD: 1.9 percent; 95 percent CI: 0.5 percent, 3.3 percent). This effect appears to be clinically insignificant. Heterogeneity among the studies was negligible (I²: 0 percent). Five of the six studies had point estimates that favoured melatonin while one study had a point estimate that favoured neither melatonin nor placebo (Figure 3-12). These results are based on trials of four weeks or less in duration.

Table 12. Sensitivity and Subgroup Analyses: Sleep (more...)

Table 12. Sensitivity and Subgroup Analyses: Sleep Efficiency in People with a Secondary Sleep Disorder

Subgroup	Categorization	Number of studies	Point Estimate (percent)	95 percent Confidence Interval (percent)	Heterogeneity	Deeks' Chi-Square p-value
Age	Children	1	3.4	-3.9, 10.7	NA	0.89
	Adult	3	2.6	-1.3, 6.4	Substantial (I2: 52.9 percent)
	Elderly	2	2.0	0.2, 3.8	Negligible (I2: 0 percent)
Use Concurrent Medication	Yes	3	2.3	-1.4, 6.0	Moderate (I2: 48.9 percent)	NA
Dosage of Melatonin Administration	1–3 mg	3	1.9	-0.5, 4.3	Moderate (I2: 47.4 percent)	NA
	6–10 mg	3	2.2	0.1, 4.3	Negligible (I2: 0 percent)
Duration of Melatonin Administration	1–2 weeks	1	2.0	-4.1, 8.1	NA	0.99
	3–4 weeks	4	2.5	-0.5, 5.4	Moderate (I2: 32.6 percent)
	> 4 weeks	1	2.0	0.1, 3.9	NA
Method of Measurement of Sleep Outcomes	Polysomography	1	0.0	-2.7, 2.7	NA	0.28
	Actigraphy	5	2.6	1.0, 4.2	Negligible (I2: 0 percent)
Co-Morbidity	Schizophrenia	2	2.3	-2.9, 7.4	Substantial (I2: 72.5 percent)	NA
Study Design	Parallel	2	2.2	0.3, 4.0	Negligible(I2: 0 percent)	0.93
	Crossover	4	2.0	-0.7, 4.7	Moderate (I2: 23.8 percent)
Allocation Concealment	Unclear	4	2.6	1.0, 4.3	Negligible (I2: 0 percent)	0.33
	Adequate	2	0.3	-2.2, 2.8	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable

Sensitivity and Subgroup Analysis. Table 12 summarizes the results of the subgroup analysis of sleep efficiency on patients with a secondary sleep disorder. Subgroups that were not possible were gender, ethnicity, timing, and Jadad score. Only three studies could be classified by use of concurrent medications (others were mixed or not specified). For dosage, one study was excluded since it gave varying doses based on weight and another gave multiple doses and was included in multiple categories. Primary disorder was different for all studies except for two studies that both examined patients with schizophrenia.

The only study that did not show a positive effect for sleep efficiency used polysomnography instead of actigraphy as its method to measure sleep outcomes. The two studies considered to have adequate allocation concealment had the lowest sleep efficiency estimates. However, the differences were not enough to significantly reduce heterogeneity in either of the subgroups.

Due to the negligible amount of heterogeneity, using the Fixed Effects Model instead of the Random Effects Model did not change the estimate in the primary analysis. The WMD and confidence interval were identical.

Assessment of Publication Bias. With only six included studies analysing sleep efficiency in patients with a secondary sleep disorder, the number of studies was deemed too few to do any meaningful tests for publication bias.

People Suffering from Sleep Restriction

Sleep Onset Latency

Figure 13. Meta-Graph: Sleep Onset Latency in People (more...)

   Figure 13. Meta-Graph: Sleep Onset Latency in People Suffering from Sleep Restriction

Primary Analysis. There were a total of nine studies that provided data on sleep onset latency for patients suffering from sleep restriction. Despite a tight confidence interval, the nine studies did not show a significant effect for melatonin on sleep onset latency (WMD: -1.0 min.; 95 percent CI: -2.3 min., 0.3 min.). Heterogeneity among the studies was minimal (I²: 4.0 percent). Six of the nine studies had a point estimate that favoured melatonin (Figure 3-13).

Table 13. Subgroup and Sensitivity Analyses: Sleep (more...)

Table 13. Subgroup and Sensitivity Analyses: Sleep Onset Latency in People Suffering from Sleep Restriction

Subgroup	Categorization	Number of studies	Point Estimate (min)	95 percent Confidence Interval (min)	Heterogeneity	Deeks' Chi-Square p-value
Use of Concurrent Medication	No	2	-3.4	-10.4, 3.7	Substantial (I2: 56.7 percent)	NA
Dosage of Melatonin Administration	< 1 mg	1	-11.8	-23.6, -0.0	NA	NA
	1–3 mg	2	-4.5	-17.3, 8.3	Substantial (I2: 75.3 percent)
	4–5 mg	5	-1.0	-4.0, 2.1	Minimal (I2: 18.2 percent)
	10–20 mg	1	-2.0	-7.5, 3.5	NA
Method of Measurement of Sleep Outcomes	Polysomnography	2	-6.6	-14.7, 1,5	Negligible (I2: 0 percent)	0.24
	Actigraphy	1	0.8	-2.7, 4.3	NA
	Questionnaire	6	-1.1	-2.2, 0.1	Negligible (I2: 0 percent)
Type of Sleep Restriction	Jet Lag	3	-4.7	-12.6, 3.1	Minimal (I2: 16.9 percent)	0.17
	Shift Work	5	-0.8	-1.9, 0.3	Negligible (I2: 0 percent)
	Deprivation	1	-9.0	-19.2, 1.2	NA
Study Design	Parallel	4	-6.1	-11.9, -0.2	Negligible (I2: 0 percent)	0.08
	Crossover	5	-0.8	-1.9, 0.3	Negligible (I2: 0 percent)
Quality Score	High (4–5)	5	-1.2	-4.6, 2.3	Minimal (I2: 18.6 percent)	1.00
	Moderate (2–3)	4	-0.9	-2.7, 0.8	Minimal (I2: 12.2 percent)
Allocation Concealment	Unclear	6	-1.4	-3.8, 1.1	Moderate (I2: 26.2 percent)	0.73
	Adequate	3	-0.5	-3.7, 2.7	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable, min: minutes

Sensitivity and Subgroup Analyses. Many of the planned subgroup analyses of sleep onset latency in patients suffering from sleep restriction could not be performed. The subgroups include gender (all studies involved both males and females and did not provide a breakdown by gender), age (subjects' age was similar across studies), ethnicity (not specified in any study), timing (all patients took melatonin before bed), and duration (the duration of melatonin administration was similar across studies). The results of the subgroup analyses that could be performed are summarized in Table 13. Note that only two studies could be classified in terms of concurrent medication use.

Using the Fixed Effect Model in place of the Random Effects Model does not change the conclusions. The point estimate and confidence interval (WMD: -1.0; 95 percent CI: -2.1, 0.1) are comparable to the random effects estimate.

Figure 14. Funnel Plot: Sleep Onset Latency in People (more...)

   Figure 14. Funnel Plot: Sleep Onset Latency in People Suffering from Sleep Restriction

Abbreviations: _ES: effect size, _seES: standard error of effect size; note that the smaller studies are associated with a larger standard error.

Assessment of Publication Bias. The funnel plot showed no obvious signs of asymmetry (Figure 3-14). There were also no indications of publication bias by any of the tests conducted. Begg's Test gives a Kendall's score of -10, which gives a continuity corrected p-value of 0.348 with nine studies. Egger's Test has a p-value of 0.479 on the bias of the funnel plot. Duval's Trim and Fill Method added no new studies and thus had the same effectiveness estimate.

Sleep Efficiency

Figure 15. Meta-Graph: Sleep Efficiency in People Suffering (more...)

   Figure 15. Meta-Graph: Sleep Efficiency in People Suffering from Sleep Restriction

Primary Analysis. Data on sleep efficiency were available for only five studies that examined patients suffering from sleep restriction. The combined estimate of the studies showed no significant difference between melatonin and placebo with respect to sleep efficiency (WMD: 0.5 percent; 95 percent CI: -0.6 percent, 1.6 percent). Heterogeneity among the studies was moderate (I²: 20.9 percent). Four of the five studies had point estimates that favoured melatonin (Figure 3-15).

Table 14. Subgroup and Sensitivity Analyses: Sleep (more...)

Table 14. Subgroup and Sensitivity Analyses: Sleep Efficiency in People Suffering from Sleep Restriction

Subgroup	Categorization	Number of studies	Point Estimate (percent)	95 percent Confidence Interval (percent)	Heterogeneity	Deeks' Chi-Square p-value
Method of Measurement of Sleep Outcomes	Polysomnography	2	1.8	0.1, 3.5	Negligible (I2: 0 percent)	0.11
	Actigraphy	1	0.2	-3.9, 4.3	NA
	Questionnaire	2	-0.2	-1.1, 0.6	Negligible (I2: 0 percent)
Type of Sleep Restriction	Jet Lag	1	2.9	-1.4, 7.2	NA	0.10
	Shift Work	3	-0.2	-1.1, 0.6	Negligible (I2: 0 percent)
	Deprivation	1	1.6	-0.2, 3.4	NA
Study Design	Parallel	2	1.8	0.1, 3.5	Negligible (I2: 0 percent)	0.11
	Crossover	3	-0.2	-1.1, 0.6	Negligible (I2: 0 percent)
Quality Score	High (4–5)	2	-0.2	-1.1, 0.6	Negligible (I2: 0 percent)	0.14
	Moderate (2–3)	3	1.6	0.0, 3.1	Negligible (I2: 0 percent)
Allocation Concealment	Unclear	2	-0.2	-1.1, 0.6	Negligible (I2: 0 percent)	0.14
	Adequate	3	1.6	0.0, 3.1	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable

Sensitivity and Subgroup Analyses. Many subgroup analyses could not be performed, including gender (all studies involved both males and females and did not provide a breakdown by gender), age (subjects' age was similar across studies), ethnicity (not specified in any study), timing (all patients took melatonin before bed), duration (the duration of melatonin administration was similar across studies), use of concurrent medication (not stated in any of the studies) and dosage (different across studies). The partitions that could be performed are outlined in Table 14.

Interestingly, all four subdivisions above gave negligible heterogeneity in all of their respective subgroups, although it was never a significant reduction in overall heterogeneity.

Using the Fixed Effects Model in place of the Random Effects Model to obtain the estimate of sleep efficiency for patients suffering from sleep restriction did not differ substantially from the primary analysis. The effectiveness estimate slightly favoured melatonin but was non-significant (WMD: 0.2; 95 percent CI: -0.6, 0.9).

Assessment of Publication Bias. There were an insufficient number of studies that involved subjects suffering from sleep restriction that examined sleep efficiency to justify performing tests for publication bias.

Duration of Administration

When studies involving sleep disorders were subdivided by duration of administration (i.e., <1 week, 1–2 weeks, 3–4 weeks), there was no apparent melatonin effect with respect to either sleep onset latency or sleep efficiency. The results were generally the same (i.e. overlapping confidence intervals) regardless of the duration.

There were no data for subjects suffering from sleep restriction, as all studies were approximately the same duration.

What is the timing of melatonin administration during the sleep/wake cycle that would produce optimal treatment effects?

Without exception, every sleep disorder study administered melatonin to its patients just before they went to bed. As a result there is no information on effect of timing of melatonin administration.

How do different formulations of melatonin differ with respect to effectiveness?

There was insufficient information on melatonin formulations in the sleep disorder studies to allow us to do any subgroup analysis by formulation.

What are the adverse effects of short and long-term use of melatonin?

Thirty-four studies were relevant to this question of the review. The overall quality of these studies was assessed using the Downs and Black Checklist.¹¹⁵ The overall quality of studies ranged from 12 to 26 on a 29-point scale; one study had a score of 12,¹⁴⁴ 13 studies had a score between 16 and 20,^{27 60 61 63–65 81 92 94 124 145–147} 19 studies had a score between 21 and 25^{54 66 68 69 71 74 76 78 80 82 90 95 129 130 143 148–151} and one study had a score of 26.⁶⁷ The quality of reporting ranged from five to 11 on an 11-point scale; two studies had a score of five or six,^{63 144} 15 studies had a score between seven and nine^{27 61 64–66 68 76 80–82 92 94 145–147} and 17 studies had a score between 10 and 11.^{54 60 67 69 71 74 78 90 95 124 129 130 143 148–151} The external validity of studies ranged from zero to three; most studies had a score of one, eight studies had a score of two,^{63 66 74 80 82 130 145 151} four studies had a score of three,^{54 67 90 143} and six studies had a score of zero.^{78 94 95 146 147 149} The quality of internal validity ranged from five to 12 on a 13-point scale; most studies had a score between nine and 11; one study had a score of five,¹⁴⁴ one study had a score of eight,²⁷ one study had a score of 10,¹⁵¹ and two studies had a score of 12.^{54 130} A power calculation for the primary outcome was reported for eight studies;^{66–68 76 78 82 143 150} four of these had sufficient power to detect a clinically important effect where the probability value for a difference being due to chance is less than 5 percent,^{66–68 150} while four others did not.^{76 78 82 143} A power calculation was not provided for one study, but it had sufficient power to detect a clinically important effect where the probability value for a difference being due to chance is less than 5 percent.⁶³ See Evidence Table C-3 for a description of design characteristics and overall quality scores of studies relevant to this question of the review.

Primary Analysis. There were few reports of adverse events accompanying melatonin administration. The most common adverse events reported were headaches, dizziness, nausea and drowsiness. In all cases there was no significant differences found between melatonin and placebo despite tight confidence intervals.

Figure 16. Meta-Graph: Headaches

   Figure 16. Meta-Graph: Headaches

Figure 17. Meta-Graph: Dizziness

   Figure 17. Meta-Graph: Dizziness

Figure 18. Meta-Graph: Nausea

   Figure 18. Meta-Graph: Nausea

Figure 19. Meta-Graph: Drowsiness

   Figure 19. Meta-Graph: Drowsiness

There were a total of 33 studies with information on headaches. There was no difference between placebo and melatonin (Risk Difference: 0.00; 95 percent CI: -0.02, 0.02) and heterogeneity was minimal (I²: 0 percent) (Figure 3-16). The results were similar for dizziness (32 studies; RD: 0.00; 95 percent CI: -0.02, 0.02; I²: 0 percent) (Figure 3-17), nausea (33 studies; RD: 0.00; 95 percent CI: -0.02, 0.01; I²: 0 percent) (Figure 3-18) and drowsiness (34 studies; RD: 0.00; 95 percent CI: -0.01, 0.02; I²: 0 percent) (Figure 3-19).

Subgroup and Sensitivity Analyses. Analyses were performed on the safety outcomes for the following subgroups: gender, age, use of concurrent medications, dosage, duration, patient category, study design, quality score, and allocation concealment score. The homogeneity of the comparisons made all of these analyses irrelevant. For all outcome measures, the calculated risk difference was not significant and the point estimate was never more than a few percentage points from zero. Thus, no change was observed between melatonin and placebo in terms of headaches, dizziness, nausea or drowsiness for different doses of melatonin, type of sleep disorder (or lack thereof), duration of melatonin treatment, or for any other subgroup mentioned above.

Using the Fixed Effect Model in place of the Random Effects Model did not change any of the results due to the homogeneity of studies for all four outcomes.

A sensitivity analysis using only the studies where the specific outcome was mentioned was also performed, and still none of the results were significant. The point estimate for risk difference of headaches between melatonin and placebo included 14 studies and was just a fraction above zero and not significant (RD: 0.00; 95 percent CI: -0.04, 0.05). This estimate had moderate heterogeneity (I²: 33.9 percent). For dizziness, only four studies were included, and again the results were just above zero and non-significant (RD: 0.03; 95 percent CI: -0.03, 0.09) with moderate heterogeneity (I²: 36.8 percent). The nausea estimate included five studies and actually favoured melatonin (RD: -0.02; 95 percent CI: -0.05, 0.02), but was also not significant; the heterogeneity was negligible (I²: 0 percent). Finally, the drowsiness estimate included nine studies with non-significant results that favoured placebo (RD: 0.03; 95 percent CI: -0.05, 0.11). The heterogeneity for this last estimate was substantial (I²: 57.0 percent).

How do the harms of exogenous melatonin vary based on dose, timing of administration, and patient factors such as gender, age, and ethnicity? How do different formulations of melatonin differ with respect to safety?

There was insufficient information to answer this question in terms of timing of melatonin administration (all studies involved melatonin administration just before bed) and ethnicity (no studies reported information on ethnicity). The sub groupings for dose, gender, age, and formulation are provided in Tables 3-11, 3-12, 3-13 and 3-14 below. It is clear from these tables that there is no evidence that the adverse effects of melatonin change by gender, age, dose or formulation. This was to be expected as the primary analysis showed that the adverse effects of melatonin were quite minimal.

Intervention: manipulation of light/dark exposure

Manipulation of light/dark exposure is designed to alter endogenous melatonin levels. The studies that employed this intervention can be categorized as those involving normal sleepers, people with a sleep disorder and people with a disorder that may or may not be accompanied by a sleep disorder. For most studies, a comparison was made between the effects of light of different intensities on endogenous melatonin and the sleep cycle, with the light of lower intensity serving as a control. The levels of light intensity varied widely across studies, such that a comparison of “bright” and “dim” light involved very different light levels across studies. Here, we use “brighter” to denote the light of higher intensity and “dimmer” to denote the light of lower intensity, for studies in which a comparison was made between the effects of light of different intensities, in order to highlight the fact that these light levels were relative and do not necessarily indicate “bright” or “dim” light in absolute terms.

Table 19. Effect of Manipulation of Light/Dark Exposure (more...)

Table 19. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in Normal Sleepers: Manipulation during Evening or Night

Study	Endogenous Melatonin	Sleep Cycle	Assessment of Correlation
Bunnell, 1992	↓ endogenous MLT levels	↑ REM latency and NREM period length, ≠ REM cycle and period length	Not conducted
Burgress, 2001	↓ endogenous MLT levels	↑ SOL	Not conducted
Cajochen, 2000	↓ endogenous MLT levels	↑ alertness and performance	Positive correlation was found between changes
Daurat, 1996	↓ endogenous MLT levels	↑ alertness and performance	Not conducted
Dollins, 1993	↓ endogenous MLT levels	≠ alertness and performance	Not conducted
Higuchi, 2003	↓ endogenous MLT levels	≠ alertness and performance	Not conducted
Horne, 1991	↓ endogenous MLT levels	↑ alertness and performance	Not conducted
Kubota, 2002	Delayed MLT rhythm	Delayed sleep onset	No correlation was found between changes
Lavoie, 2003	↓ endogenous MLT levels	≠ SOL, ≠ alertness and performance	None of the vigilance variables were found to correlate to endogenous MLT levels.

Abbreviations: MLT: melatonin, SOL: sleep onset latency, REM: rapid eye movement, NREM: non-REM, ↑: increased ↓: decreased, ≠: no change in

Table 20. Effect of Manipulation of Light/Dark Exposure (more...)

Table 20. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in Normal Sleepers: Manipulation During Morning or Daytime

Study	Endogenous Melatonin	Sleep Cycle	Assessment of Correlation
Danilenko, 2000	Advance of MLT rhythm, ↓ endogenous MLT levels	↑ alertness,	Phase of MLT rhythm was correlated to sleepiness and mid-point of sleep
Dijk, 1989	Advance of MLT rhythm	↓ sleep duration and REM sleep, ≠ REM latency, percent time spent in various sleep stages and sleep quality	No correlation between phase of MLT rhythm, and sleep duration
Wakamura, 2000	↓ endogenous MLT levels	↑ alertness	Not conducted

Abbreviations: MLT: melatonin, REM: rapid eye movement, ↑: increased, ↓: decreased, ≠: no change in

Table 21. Effect of Manipulation of Light/Dark Exposure (more...)

Table 21. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in Normal Sleepers: Manipulation Involves Unique Conditions

Study	Endogenous Melatonin	Sleep Cycle	Assessment of Correlation
Daurat, 1997	≠ endogenous MLT levels	≠ TST, REM latency, WASO and REM sleep	Not conducted
Gordijn, 1999	↓ endogenous MLT, ≠ phase of MLT rhythm	↑ movement time, ↓ duration of first REM episode, delayed sleep termination, ≠ sleep latency and REM latency	Not conducted
Lushington, 2002	≠ endogenous MLT levels or phase of MLT rhythm	↑ wakefulness	Not conducted
Wehr, 1991	↓ duration of nocturnal endogenous MLT	↓ sleep period	Not conducted

Abbreviations: MLT: melatonin, TST: total sleep time, REM: rapid eye movement, WASO: wakefulness after sleep onset, ↑: increased ↓: decreased, ≠: no change in

In a study by Gordijn et al. in which the effects of morning and evening ocular light exposure were compared, evening light exposure resulted in suppression of endogenous melatonin levels compared to morning light exposure, although the phase of endogenous melatonin was not affected.¹⁹⁹ The changes in endogenous melatonin with evening light exposure were accompanied by greater “movement time”, shorter duration of the first REM episode, later time of sleep termination and no change in sleep latency and REM latency, compared to morning BL exposure.¹⁹⁹ In a study by Wehr et al., exposure to a longer photoperiod resulted in a reduction in the duration of the nocturnal endogenous melatonin rhythm, duration of the sleep period and the nocturnal phase of increasing sleepiness, compared to exposure to a shorter photoperiod.¹⁸⁷ Similarly in a study by Daurat et al., light exposure or a light/dark cycle were administered for 36 hours during sleep deprivation and no difference was found in endogenous melatonin levels or in total sleep time, REM latency, WASO and REM sleep.²⁰⁹ In a study by Lushington et al. in which light stimuli were administered behind the knee, BL did not affect endogenous melatonin, however, it resulted in increased wakefulness, relative to DL.¹²⁸ (Table 13-17)

Table 15. Subgroup Analysis: Headaches

Table 15. Subgroup Analysis: Headaches

Subgroup	Categorization	Number of studies	Risk Difference	95 percent Confidence Interval	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	3	0.04	-0.13, 0.19	Negligible (I2: 0 percent)	NA
	Female	2	0.00	-0.07, 0.07	Negligible (I2: 0 percent)
Age	Children	6	-0.02	-0.08, 0.03	Negligible (I2: 0 percent)	0.74
	Adult	22	0.00	-0.02, 0.03	Negligible (I2: 0 percent)
	Elderly	5	0.00	-0.06, 0.06	Negligible (I2: 0 percent)
Dosage	1–3 mg	8	0.00	-0.05, 0.04	Negligible (I2: 0 percent)	NA
	4–5 mg	14	0.00	-0.03, 0.04	Minimal (I2: 18.9 percent)
	6–10 mg	9	0.00	-0.04, 0.04	Negligible (I2: 0 percent)
	> 10 mg	2	0.00	-0.20, 0.20	Negligible (I2: 0 percent)
Formulation	Fast Release	3	-0.06	-0.14, 0.02	Negligible (I2: 0 percent)	NA
	Slow Release	4	0.00	-0.05, 0.05	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable

Table 16. Subgroup Analysis: Dizziness

Table 16. Subgroup Analysis: Dizziness

Subgroup	Categorization	Number of studies	Risk Difference	95 percent Confidence Interval	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	2	0.00	-0.19, 0.19	Negligible (I2: 0 percent)	NA
	Female	2	0.00	-0.07, 0.07	Negligible (I2: 0 percent)
Age	Children	6	0.02	-0.04, 0.08	Negligible (I2: 0 percent)	1.00
	Adult	21	0.00	-0.02, 0.02	Negligible (I2: 0 percent)
	Elderly	5	0.00	-0.06, 0.06	Negligible (I2: 0 percent)
Dosage	1–3 mg	8	0.00	-0.04, 0.05	Negligible (I2: 0 percent)	NA
	4–5 mg	14	0.00	-0.02, 0.03	Negligible (I2: 0 percent)
	6–10 mg	7	0.00	-0.04, 0.04	Negligible (I2: 0 percent)
	> 10 mg	2	0.00	-0.20, 0.20	Negligible (I2: 0 percent)
Formulation	Fast Release	3	0.05	-0.04, 0.15	Minimal (I2: 14.6 percent)	NA
	Slow Release	4	0.00	-0.05, 0.05	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable

Table 17. Subgroup Analysis: Nausea

Table 17. Subgroup Analysis: Nausea

Subgroup	Categorization	Number of studies	Risk Difference	95 percent Confidence Interval	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	2	0.00	-0.19, 0.19	Negligible (I2: 0 percent)	NA
	Female	2	0.00	-0.07, 0.07	Negligible (I2: 0 percent)
Age	Children	6	-0.02	-0.08, 0.03	Negligible (I2: 0 percent)	0.81
	Adult	21	0.00	-0.02, 0.02	Negligible (I2: 0 percent)
	Elderly	5	0.00	-0.06, 0.06	Negligible (I2: 0 percent)
Dosage	1–3 mg	8	0.00	-0.03, 0.03	Negligible (I2: 0 percent)	NA
	4–5 mg	14	-0.01	-0.04, 0.02	Negligible (I2: 0 percent)
	6–10 mg	9	0.00	-0.04, 0.03	Negligible (I2: 0 percent)
	> 10 mg	2	0.00	-0.20, 0.20	Negligible (I2: 0 percent)
Formulation	Fast Release	3	-0.02	-0.08, 0.03	Negligible (I2: 0 percent)	NA
	Slow Release	4	0.00	-0.04, 0.05	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable

Table 18. Subgroup Analysis: Drowsiness

Table 18. Subgroup Analysis: Drowsiness

Subgroup	Categorization	Number of studies	Risk Difference	95 percent Confidence Interval	Heterogeneity	Deeks' Chi-Square p-value
Gender	Male	3	0.19	-0.21, 0.60	Substantial (I2: 76.6 percent)	NA
	Female	2	0.00	-0.07, 0.07	Negligible (I2: 0 percent)
Age	Children	5	0.00	-0.06, 0.06	Negligible (I2: 0 percent)	1.00
	Adult	24	0.00	-0.02, 0.02	Negligible (I2: 0 percent)
	Elderly	6	0.01	-0.04, 0.05	Negligible (I2: 0 percent)
Dosage	1–3 mg	8	-0.01	-0.06, 0.04	Negligible (I2: 0 percent)	NA
	4–5 mg	13	0.00	-0.03, 0.03	Negligible (I2: 0 percent)
	6–10 mg	9	0.01	-0.04, 0.05	Negligible (I2: 0 percent)
	> 10 mg	2	0.32	-0.37, 1.01	Substantial (I2: 86.4 percent)
Formulation	Fast Release	2	-0.07	-0.19, 0.06	Negligible (I2: 0 percent)	NA
	Slow Release	4	-0.01	-0.06, 0.04	Negligible (I2: 0 percent)

Abbreviations: NA: not applicable

People with a Sleep Disorder

Table 22. Effect of Manipulation of Light/Dark Exposure (more...)

Table 22. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in People with Sleep Disorders

Study	Endogenous Melatonin	Sleep Cycle	Assessment of Correlation
Ando, 1999	≠ phase of MLT rhythm	≠ total sleep period, total sleep time and sleep quality	Not conducted
Bougrine, 1995	≠ phase of MLT rhythm	≠ sleep quality, performance and subjective feelings of tiredness	Not conducted
Boulos, 2002	Delay in MLT rhythm	≠ sleep efficiency, sleep quality, daytime sleepiness, jet-lag severity or mood	No correlation was found between phase of MLT rhythm and performance
Budnick, 1995	↓ endogenous MLT	≠ total sleep time, ↑ alertness and performance	Not conducted
Cole, 2002	≠ phase of MLT rhythm	≠ mood, total sleep time, sleep quality, morning sleepiness	Not conducted
Costa, 1997	≠ endogenous MLT levels	≠ alertness and performance	Not conducted
Horowitz, 2001	Delay in MLT rhythm	≠ sleep start time and wake time, ↑ alertness and performance	Not conducted
Kelly, 1997	Delay in MLT rhythm	↑ sleep time and continuity, ≠ sleep latency, ↑ alertness and performance	Not conducted
Ross, 1995	Not explicitly stated	↓ sleep latency, ≠ sleep duration, sleep quality, night awakenings and mood	Not conducted
Yoon, 2000	Delay in MLT rhythm	↑ alertness and performance	Not conducted

Abbreviations: MLT: melatonin, ↑: increased ↓: decreased, ≠: no change in

People with a Disorder that may or may not be Accompanied by a Sleep Disorder.

Table 23. Effect of Manipulation of Light/Dark Exposure (more...)

Table 23. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in People with a Disorder that may or may not be Accompanied by a Sleep Disorder

Study	Endogenous Melatonin	Sleep/Wake Cycle	Assessment of Correlation
Gordijn, 1998	Advance of MLT rhythm	Earlier tendency for sleep termination	No correlation was found between phase of MLT rhythm and wake-up time
Koorengevel, 2001	≠ phase of MLT rhythm	≠ mood, alertness, total sleep duration, time of awakening and sleep onset	Not conducted
Partonen, 1996	≠ endogenous MLT levels	≠ sleepiness	Not conducted

Abbreviations: MLT: melatonin, ↑: increased, ↓: decreased, ≠: no change in

Intervention: manipulation of light/dark exposure

The studies that employed this intervention can be categorized as those involving normal sleepers, people with a sleep disorder and people with a disorder that may or may not be accompanied by a sleep disorder. For most studies, a comparison was made between the effects of light of different intensities on endogenous melatonin and the temperature rhythm, with the light of lower intensity serving as a control. The levels of light intensity varied widely across studies, such that a comparison of “bright” and “dim” light involved very different light levels across studies. Here, we use “brighter” to denote the light of higher intensity and “dimmer” to denote the light of lower intensity, for studies in which a comparison was made between the effects of light of different intensities, in order to highlight the fact that these light levels were relative and do not necessarily indicate “bright” or “dim” light in absolute terms.

Normal Sleepers. The nature, magnitude, duration and timing of the light stimulus intervention varied across studies. Most studies involved direct application of brighter (BL) or dimmer (DL) stimuli,^{128 176 183 191 194 199 204–206 208 209 223–225 228} while others involved application of these signals within the context of dawn simulation¹⁷⁹ or a video display terminal.¹⁸² Most studies compared the effect of BL to DL,^{128 176 179 182 183 191 194 199 204–206 208 223–225 228} while one study compared the effect of evening versus morning BL administration.¹⁹⁹ The magnitude of the light stimuli ranged from 0.1 lux to 200 lux for DL administration and 45 lux to 13 000 lux for BL administration and the duration of light stimuli application ranged from 2 to 36 hours, over hours to weeks. The timing of light administration varied from early morning to late night. Most studies involved ocular light administration,^{176 179 182 183 191 194 199 204–206 208 209 223 225 228} while two studies involved extra-ocular light administration applied to the bend of the knee.^{128 224} Some studies involved application of light stimuli under conditions of prolonged sleep deprivation.^{183 194 204 206 209 224 225 228} (Evidence Table C-8)

Table 24. Effect of Manipulation of Light/Dark Exposure (more...)

Table 24. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in Normal Sleepers: Manipulation During Evening or Night

Study	Endogenous Melatonin	Temperature Rhythm	Assessment of Correlation
Bunnell, 1992	↓ endogenous MLT levels	↑ core body temperature, ≠ tympanic temperature	Not conducted
Burgress, 2001	↓ endogenous MLT levels	↑ core body temperature	Not conducted
Cagnacci, 1993	Delayed MLT rhythm, ↓ endogenous MLT levels	≠ value or timing of core body temperature minima	Not conducted
Cajochen, 2000	↓ endogenous MLT levels	≠ core body temperature	Not conducted
Daurat, 1996	↓ endogenous MLT levels	↑ core body temperature, reduced and delayed temperature minima	Not conducted
Eastman, 2000	≠ MLT rhythm	≠ core body temperature	Not conducted
Higuchi, 2003	↓ endogenous MLT levels	↑ core body temperature	Not conducted
Horne, 1991	↓ endogenous MLT levels	≠ oral temperature	Not conducted
Kubota, 2002	Delayed MLT rhythm, ↓ endogenous MLT levels	Delay in core body temperature minima	No correlation was found between the change in phase of MLT rhythm and temperature rhythm
Lavoie, 2003	↓ endogenous MLT levels	↑ core body temperature	Not conducted
Lushington, 2002	≠ endogenous MLT levels or rhythm	≠ nocturnal core body temperature rhythm	Not conducted
Strassman, 1991	↓ endogenous MLT levels	↑ minimal rectal temperature, ≠ maximal rectal temperature	Not conducted

Abbreviations: MLT: melatonin, ↑: increased ↓: decreased, ≠: no change in

Most studies involved application of light stimuli during the evening or night,^{128 176 182 183 191 194 204–206 223–225} while some studies involved light exposure during the morning or night,¹⁹⁹ morning,¹⁷⁹ daytime²⁰⁸ or during a prolonged period of sleep deprivation.^{209 228} Of the studies involving application of ocular light stimuli during the evening or night, some studies found that BL delayed the onset of endogenous melatonin secretion, relative to DL^{204 223} and all studies found that BL suppressed endogenous melatonin levels, relative to DL. The delay in the phase of the melatonin rhythm was accompanied by a delay in the core body temperature rhythm in one study,²⁰⁴ while the change in phase of the melatonin rhythm was not accompanied by a change in the value or timing of the core body temperature minima in another study.²²³ Of the studies that found a suppression of endogenous melatonin levels with BL, many studies found an accompanying increase in core body temperature,^{176 182 194 205 206} while one study found no change in core body temperature.¹⁹¹ In a study by Strassman et al., the suppression of endogenous melatonin levels was accompanied by an increase in minimum rectal temperature and no change in the maximum rectal temperature,²²⁵ while in a study by Daurat et al., the suppression of endogenous melatonin levels was accompanied by a reduction and delay in the temperature minimum.²⁰⁶ In studies by Horne et al. and Bunnell et al., the suppression of endogenous melatonin was not accompanied by any change in oral temperature¹⁸³ or tympanic temperature.¹⁷⁶ In a study by Wright et al., whereby light exposure occurred during a 45-hour period of sleep deprivation, BL resulted in suppression of endogenous melatonin levels accompanied by increased body temperature, relative to DL.²²⁸ In a similar study by Daurat et al., whereby either BL or a light/dark cycle were imposed on subjects for 36 hours during sleep deprivation, there were no differences in endogenous melatonin levels between conditions. However, during one of the follow-up nights, the BL group showed increased rectal temperature compared to the light/dark group, although the phase of the temperature rhythm was not different.²⁰⁹ In a study by Lushington et al., BL administered behind the knee during the daytime did not have an effect on endogenous melatonin or the phase of nocturnal core body temperature, relative to DL administration.¹²⁸ In a similar study, extra-ocular light exposure in the evening did not affect either the phase of the melatonin rhythm or core body temperature.²²⁴ In a study by Gordijn et al., the application of light stimuli in the evening resulted in the suppression of endogenous melatonin levels during the early evening, without a change in the phase of the melatonin rhythm, compared to morning light exposure. This effect was accompanied by increased body temperature without a change in the phase of the temperature rhythm.¹⁹⁹ In a study by Wakamura et al., ocular light stimuli were administered during the daytime and resulted in a suppression of endogenous melatonin levels, which was accompanied by a reduction in minimum core body temperature, no change in maximum core body temperature and an advance of the core body temperature rhythm.²⁰⁸ (Table 24)

Table 25. Effect of Manipulation of Light/Dark Exposure (more...)

Table 25. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in Normal Sleepers: Manipulation Involved Unique Conditions

Study	Endogenous Melatonin	Temperature rhythm	Assessment of Correlation
Danilenko, 2000	↓ endogenous MLT levels, shift in MLT rhythm	Shift in temperature rhythm	Shifts in MLT rhythm and temperature rhythm were correlated
Daurat, 1997	≠ endogenous MLT	↑ rectal temperature, ≠ phase of temperature rhythm	Not conducted
Eastman, 2000	≠ phase of MLT rhythm	≠ core body temperature	Not conducted
Gordijn, 1999	↓ endogenous MLT levels, ≠ phase of MLT rhythm	↑ body temperature, ≠ phase of temperature rhythm	Not conducted
Lushington, 2002	≠ endogenous MLT levels	≠ phase of core body temperature rhythm	Not conducted
Wakamura, 2000	↓ endogenous MLT levels	↓ minimum core body temperature, ≠ maximum core body temperature, advance of the core body temperature rhythm	Not conducted
Wright, 1997	↓ endogenous MLT levels	↑ body temperature	Not conducted

Abbreviations: MLT: melatonin, ↑: increased, ↓: decreased, ≠: no change in

In a study by Danilenko et al., morning BL exposure suppressed endogenous melatonin levels, and the phase shift in the melatonin rhythm (DLMOn and DLMOff) was correlated with the phase shift in the temperature rhythm.{Danilenko, Wirz-Justice, et al. 2000 #22510} By contrast, in a study by Kubota et al., a correlation was not found between the change in phase of the melatonin rhythm and that of the temperature rhythm, following BL administration.²⁰⁴ (Table 25)

People with a Sleep Disorder. A number of studies in this category involved shift workers,^{178 193 210} while another study involved people with delayed sleep phase syndrome.¹⁸⁹ In the case of all studies involving night-shift workers, light stimuli were administered during the night-shift,^{178 189 193 210} and in the case of the study involving people with delayed sleep phase syndrome light stimuli were administered in the early morning.¹⁸⁹ The magnitude of the light stimuli ranged from 0.1 lux to 300 lux for dim/control light administration and 400 lux to 4300 lux for BL administration. The duration of light stimuli application ranged from 4 periods of 40-minute exposures to an entire night-shift, over hours to days. All studies involved ocular light administration. (Evidence Table C-8)

Table 26. Effect of Manipulation of Light/Dark Exposure (more...)

Table 26. Effect of Manipulation of Light/Dark Exposure on Endogenous Melatonin and the Sleep Cycle in People with a Sleep Disorder

Study	Endogenous Melatonin	Temperature rhythm	Assessment of Correlation
Ando, 1999	≠ MLT rhythm	≠ temperature rhythm	Not conducted
Costa, 1997	≠ endogenous MLT levels	≠ temperature rhythm	Not conducted
Horowitz, 2001	Delay MLT rhythm	Delay of core body temperature rhythm	Not conducted
Kelly, 1997	Delay MLT rhythm	≠ temperature rhythm	Not conducted

Abbreviations: MLT: melatonin, ↑: increased, ↓: decreased, ≠: no change in

Of the studies involving night-shift work,^{178 193 210} BL delayed the melatonin rhythm in two studies^{193 210} and had no effect on endogenous melatonin levels in another study,¹⁷⁸ relative to DL. Moreover, while the delay in the melatonin rhythm was accompanied by a phase delay in core body temperature in one study,¹⁹³ it was not accompanied by a change in the temperature rhythm in another study.²¹⁰ The lack of effect of BL on endogenous melatonin levels, relative to DL, in the study by Costa et al., was paralleled by a lack of effect on the temperature rhythm.¹⁷⁸ (Table 26)

In a study by Ando et al., patients with delayed sleep phase syndrome received either 500 lux for three hours over 12 days prior to awakening or 0.1 lux of the same timing. BL had no significant effect on the phase of either the melatonin or temperature rhythm, compared to DL.¹⁸⁹

People with a Disorder that may or may not be Accompanied by a Sleep Disorder. One study involved application of either morning or evening light to people with non-seasonal depression. Morning administration of light resulted in a phase advance of both the melatonin and temperature rhythms compared to evening light administration.²¹² Moreover, no correlation was found between the shifts in the phase of the melatonin and temperature rhythms.²¹²

Summary: Endogenous melatonin and the temperature rhythm

To summarize, our literature review indicated evidence of a link between endogenous melatonin and the temperature rhythm. Specifically, a reduction in endogenous melatonin levels was often accompanied by an increase in core body temperature, and a shift in the rhythm of endogenous melatonin was often accompanied by a similar shift in the rhythm of core body temperature.

How do the benefits and harms of melatonin compare to those of other approved pharmacological treatments for sleep disorders?

Only four studies compared the effects of melatonin to other pharmacological treatments for sleep disorders in terms of sleep variables; three of these involved normal sleepers^{123 222 229} and one of these involved people suffering from jet-lag.⁶¹ The overall quality score, according to the Down's and Black Checklist, ranged from 11 to 20 on a 29-point scale and the quality of reporting ranged from four to eight on an 11-point scale. The external validity of studies ranged from zero to one on a three-point scale; two studies had a score of zero^{123 229} and two studies had a score of one.^{222 230} The internal validity of studies ranged from seven to 11 on a 13-point scale; two studies had a score of seven,^{123 229} one study had a score of eight²²² and one study had a score of 11.⁶¹ For none of the studies was a power calculation reported. See Evidence Table C-9 for a description of design characteristics and overall quality scores of studies relevant to this question of the review.

Of the studies involving normal sleepers, two studies compared the effects of melatonin and triazolam,^{123 229} while another study compared the effect of melatonin and zopiclone,²²² on sleep variables. In a study by Satomura et al., participants received either 1, 3 or 6 mg melatonin or 0.125 mg triazolam at 13:30h; there were no differences in the effect of the two agents on total sleep time, sleep efficiency, and REM latency. Melatonin decreased sleep onset latency to a greater extent than triazolam, however, the effect of these compounds on sleep onset latency was not significantly different from placebo. The authors did not report on adverse events or adverse effects of the two agents, precluding a comparison of the harms of these agents.¹²³ In a similar study by Ferini-Strambi et al., participants received either 100 mg melatonin or 0.125 mg triazolam at 22:30h; there were no differences in the effect of the two agents on total sleep time, sleep onset latency, wakefulness after sleep onset, sleep efficiency, number of awakenings, percent time spent in the various sleep stages, REM latency and REM periods. As in the study by Satomura et al., the authors did not report on adverse effects of the two agents, precluding a comparison of the harms of these agents.²²⁹ In a study by Holmes et al., normal sleepers received either 5mg melatonin or 7.5mg zopiclone at 14:00h and sleep onset latency was assessed hourly from 11:00h to 20:00h using a modified multiple sleep latency test. Zopiclone reduced sleep onset latency to a greater extent than melatonin at 15:00h and from 17:00h to 19:00h. The authors did not report on adverse events or adverse effects of the two agents.²²²

In a study comparing melatonin and zolpidem, air travellers, crossing six to nine time-zones, received either 5mg melatonin or 10mg zolpidem on an eastbound return flight to Switzerland and once daily at bedtime on four consecutive days after the flight.⁶¹ The agents did not differ in their effects on total sleep time, sleep latency, number of awakenings, wakefulness after sleep onset and overall sleep quality during the flight. Moreover, the agents did not differ in overall sleep quality, sleep onset latency, number of awakenings and wakefulness after sleep onset across the four nights following the flight. When subjects were asked to rate the effectiveness of their study medication in alleviating jet-lag, the responses did not differ between agents. In general, out of the 35 people taking melatonin, only one person reported adverse events, while six people out of 34 people taking zolpidem reported adverse events. The individual taking melatonin who reported adverse events suffered from insomnia and palpitations. Of the individuals taking zolpidem who reported adverse events, four people reported nausea, two people reported vomiting, two people reported confusion, one person reported dizziness, two people reported headache, one person reported lack of concentration, one person reported amnesia, one person reported trembling, one person reported agitation, one person reported palpitation, one person reported difficulties in articulation and one person reported dry mouth.⁶¹

Overall Grade of Evidence Pertaining to Effectiveness and Safety of Melatonin

Table 27. Oxford Centre for Evidence-based Medicine (more...)

Table 27. Oxford Centre for Evidence-based Medicine Levels of Evidence

Grade of Recommendation	Level of Evidence	Therapeutic Use
A	1a	SR (with homogeneity) of RCTs
	1b	Individual RCT (with Narrow Confidence Interval)
	1c	All or none
B	2a	SR (with homogeneity) of cohort studies
	2b	Individual cohort study (including low quality RCT; e.g., .80 percent follow-up)
	2c	“Outcomes Research”
	3a	SR (with homogeneity) of case-control studies
	3b	Individual case control study
C	4	Case series (and poor quality cohort and case-control studies)
D	5	Expert opinion without explicit critical appraisal, or based on physiology, bench research or “first principles”

Abbreviations: SR: systematic review, RCT: randomized controlled trial

Adapted from http://minerva.minervation.com/cebm/documents/levels_cebm_23nov99.rtf

The source of funding was either not reported or unclear for the majority of studies relevant to the questions relating to the effectiveness of melatonin in people with sleep disorders and the safety of melatonin. Of the studies for which the source of funding was described, most studies received public sources of funding. For studies in which there was a discrepancy in the number of participants enrolled in the study and the number of participants for whom data was analyzed, the planning or conduct of an intention-to-treat analysis was reported in only two studies relevant to the question relating to the effectiveness of melatonin in people with sleep disorders ^{78 144} and two studies relevant to the question relating to the safety of melatonin.^{78 135} The overall evidence surrounding the effectiveness of melatonin for treatment of sleep disorders and the safety of melatonin was graded using the framework of the Oxford Centre for Evidence-based Medicine (Table 27). The evidence surrounding the effectiveness of melatonin for the treatment of sleep disorders receives a grade of “A” and a level of evidence designation of “1b”. The evidence surrounding the safety of melatonin receives a grade of “B” and a level of evidence designation of “2b”.

Effectiveness of Exogenous Melatonin in Normal Sleepers

Effectiveness of Exogenous Melatonin in People with Sleep Disorders

People Suffering from Sleep Restriction

• Melatonin did not have an effect on sleep onset latency in people suffering from sleep restriction; the effect of melatonin did not vary by dose or type of sleep restriction disorder i.e. shift-work and jet lag

• Melatonin did not have an effect on sleep efficiency in people suffering from sleep restriction; the effect of melatonin did not vary by dose

• Melatonin did not have an effect on sleep quality, WASO and percent time spent in REM sleep in people suffering from sleep restriction, but significantly increased total sleep time in this population

• Generally, these studies were of moderate to high quality.

Table 28. Summary of the Evidence Surrounding the Effect (more...)

Table 28. Summary of the Evidence Surrounding the Effect of Melatonin on Sleep in Various Populations

	Normal Sleepers	Primary Sleep Disorder	Secondary Sleep Disorder	Sleep Restriction
Sleep Onset Latency	Decreased	Decreased	No Effect	No Effect
	WMD: -3.9 min; 95 percent CI: -5.3 min., -2.6 min.	WMD: -10.7 min.; 95 percent CI: -17.6 min., -3.7 min.
	N=20	N=12	N=6	N=9
Sleep Efficiency	Increased	No Effect	Increased	No Effect
	WMD: 2.3 percent; 95 percent CI: 0.7 percent, 3.9 percent	N=9	WMD: 1.9 percent; 95 percent CI: 0.5 percent, 3.3 percent	N=5
	N=13		N=6

Abbreviations: WMD: weighted mean difference, CI: confidence interval

See Table 28 for a summary of the evidence surrounding the effect of melatonin on sleep in various populations.

Safety of Exogenous Melatonin

• The most commonly reported adverse effects of melatonin were nausea (incidence: ~ 1.5 percent), headache (incidence: ~ 7.8 percent), dizziness (incidence: 4.0 percent) and drowsiness (incidence: 20.33 percent); however, these effects were not significant compared to placebo. This result did not change by dose, the presence or absence of a sleep disorder, type of sleep disorder, duration of treatment, gender, age, formulation of melatonin, use of concurrent medication, study design, quality score and allocation concealment score.

• Generally, these studies were of moderate to high quality.