Melatonin for preoperative and postoperative anxiety in adults

Summary of findings 1. Summary of findings

Outcomes	*Illustrative comparative risks (95% CI)**		No. of participants (studies)	Quality of the evidence (GRADE)	Comments
Melatonin compared with placebo
Patient or population: patents undergoing elective surgery Setting: hospital Intervention: melatonin Comparison: placebo
	Assumed risk	Corresponding risk
	Placebo	Melatonin
Preoperative anxiety (VAS) VAS (0 to 100 mm) measured approximately 50 to 120 minutes after premedication 0: no anxiety 100: maximum anxiety possible	Mean VAS total ranged across control groups from 22.7 to 66.5, and mean change in VAS ranged across control groups from 4 to ‐22	Mean VAS in intervention groups was 11.69 lower (13.80 lower to 9.59 lower) Lower score indicated less preoperative anxiety compared to placebo	1264 (18 studies)	⊕⊕⊕⊖ Moderate^a	Melatonin most likely decreases preoperative anxiety compared with placebo
Preoperative anxiety (STAI) STAI (20 to 80) measured approximately 120 minutes after premedication 20: no anxiety 80: maximum anxiety possible	Mean STAI in control group measured just before entrance to the operating room was 39.73	Mean STAI in intervention group measured just before entrance to the operating room was 41.18	44 (1 study)	⊕⊖⊖⊖ Very low^b	Because only 1 study examined preoperative anxiety using an STAI, no meta‐analysis was performed
Preoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured approximately 90 minutes after premedication 6: no anxiety 24: maximum anxiety possible	Mean STAI in control group measured at patient arrival to the operating room was 13.5	Mean STAI in intervention group measured at patient arrival to the operating room was 11.6	36 (1 study)	⊕⊕⊖⊖ Low^c	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
Immediate postoperative anxiety (VAS) VAS (0 to 100 mm) measured after surgery, in recovery, or at discharge from recovery room 0: no anxiety 100: maximum anxiety possible	Mean VAS total ranged across control groups from0 to 48, and mean change in VAS ranged across control groups from ‐4.7 to ‐6.5	Mean VAS in intervention groups was 5.04 lower (9.52 lower to 0.55 lower) Lower score indicated less postoperative anxiety compared to placebo	524 (7 studies)	⊕⊕⊝⊝ Low^d	Melatonin may have an effect on postoperative anxiety compared with placebo; however, this effect was below the minimum clinical effect
Delayed postoperative anxiety (STAI) STAI (20 to 80) measured 6 hours after surgery 20: no anxiety 80: maximum anxiety possible	Mean STAI ranged across control groups from 42.2 to 42.5	Mean STAI in intervention groups was 5.31 lower (8.78 lower to 1.84 lower) Lower score indicated less postoperative anxiety compared to placebo	73 (2 studies)	⊕⊕⊝⊝ Low^e	Melatonin may have an effect on postoperative anxiety compared with placebo; however, this effect was below the minimum clinical effect
Postoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured 1 hour and 6 hours after surgery 6: no anxiety 24: maximum anxiety possible	Mean STAI value in control group 1 hour after surgery was 11 Mean STAI value in control group 6 hours after surgery was 11.6	Mean STAI value in melatonin group 1 hour after surgery was 8 Mean STAI value in melatonin group 6 hours after surgery was 7.9	36 (1 study)	⊕⊕⊝⊝ Low^f	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; STAI: State‐Trait Anxiety Inventory; VAS: visual analogue scale.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
^aThe certainty of evidence was downgraded by one level due to unclear overall risk of bias and the presence of substantial heterogeneity. We chose not to downgrade by two levels because sensitivity analysis excluding all studies with high risk of bias showed a similar effect estimate; we therefore concluded that high risk of bias in the included studies did not affect conclusions. ^bWe chose to downgrade the evidence by three levels due to imprecision and high risk of bias: only one study with 44 participants examined preoperative anxiety using STAI; this study also had overall high risk of bias. ^cWe chose to downgrade the evidence by two levels due to imprecision: only one study with 36 participants examined preoperative anxiety using a six‐item STAI. ^dThe certainty of evidence was downgraded by two levels due to large heterogeneity of the studies (I² = 89%) and overall high risk of bias. Several of the included studies had overall high risk of bias, making the overall risk of bias for the outcome high. When all studies with high risk of bias were excluded from the sensitivity analysis, the effect of the intervention was lost, which is why we suspect that inclusion of studies with overall high risk of bias may alter conclusions. ^eThe certainty of evidence was downgraded by two levels due to the small numbers of participants. ^fThe certainty of evidence was downgraded by two levels due to the small numbers of participants.

See Characteristics of included studies, Characteristics of excluded studies, Characteristics of studies awaiting classification, and Characteristics of ongoing studies.

Summary of findings 2. Summary of findings

Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
Melatonin compared with benzodiazepine
Patient or population: patients undergoing elective surgery Setting: hospital Intervention: melatonin Comparison: benzodiazepine (midazolam, alprazolam)
	Assumed risk	Corresponding risk
	Benzodiazepine	Melatonin
Preoperative anxiety (VAS) VAS (0 to 100 mm) measured approximately 90 minutes after premedication 0: no anxiety 100: maximum anxiety possible	Mean VAS total ranged across control groups from 3.6 to 16.7, and mean change in VAS ranged across control groups from ‐7.7 to ‐50	Mean VAS in intervention groups was 0.78 higher (2.02 lower to 3.58 higher) A higher score indicated greater preoperative anxiety compared to benzodiazepine		409 (7 studies)	⊕⊕⊕⊝ Moderate^a	Melatonin most likely has little or no effect on preoperative anxiety compared with benzodiazepines
Preoperative anxiety (STAI) STAI (20 to 80) 20: no anxiety 80: maximum anxiety possible	No studies available	No studies available	‐	‐	‐	‐
Preoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured approximately 90 minutes after premedication 6: no anxiety 24: maximum anxiety possible	Mean STAI in benzodiazepine group measured at patient arrival to the operating room was 10.5	Mean STAI in melatonin group measured at patient arrival to the operating room was 11.6		35 (1 study)	⊕⊕⊖⊖ Low^b	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
Immediate postoperative anxiety (VAS) VAS (0 to 100 mm) measured approximately 90 minutes after surgery or in recovery room 0: no anxiety 100: maximum anxiety possible	Mean VAS in control group was 7.4 and mean change in VAS ranged across control groups from ‐5.3 to ‐6.4	Mean VAS in intervention groups was 2.12 lower (4.61 lower to 0.36 higher) Lower score indicated less postoperative anxiety compared to benzodiazepine		176 (3 studies)	⊕⊕⊝⊝ Low^c	Melatonin had little or no effect on postoperative anxiety compared with benzodiazepines
Postoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured 1 hour and 6 hours after surgery 6: no anxiety 24: maximum anxiety possible	Mean STAI value in benzodiazepine group 1 hour after surgery was 10.4 Mean STAI value in benzodiazepine group 6 hours after surgery was 9.3	Mean STAI value in melatonin group 1 hour after surgery was 8 Mean STAI value in melatonin group 6 hours after surgery was 7.9		35 (1 study)	⊕⊕⊖⊖ Low^d	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; STAI: State‐Trait Anxiety Inventory; VAS: visual analogue scale.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
^aThe certainty of evidence was downgraded by one level due to high overall risk of bias for the outcome and substantial heterogeneity. We decided not to downgrade the evidence by two levels because sensitivity analysis excluding all studies with overall high risk of bias showed a similar effect estimate; we therefore concluded that high risk of bias in the included studies did not alter conclusions. ^bWe chose to downgrade the evidence by two levels because only one study with 35 participants examined preoperative anxiety using a six‐item STAI. ^cThe certainty of evidence was downgraded by two levels due to the small numbers of participants. One study had overall high risk of bias, but sensitivity analysis excluding this study showed a similar effect, which is why we chose not to downgrade by another level. ^dThe certainty of evidence was downgraded by two levels due to the small numbers of participants.

Background

Description of the condition

Anxiety is a human reaction to any unknown situation and is defined as a state of uneasiness and apprehension (Jellish 2012). Anxiety frequently occurs in patients throughout the perioperative period and has been described as the worst aspect of the perioperative experience (Jellish 2012; Johnston 1980; Walker 2016).

Preoperative anxiety is described as an unpleasant state of tension that occurs secondary to a patient being concerned about a disease, hospitalization, incapacitation, anaesthesia, or surgery, or his or her anticipation of postoperative pain and the unknown (Caumo 2001a; Ramsay 1972). In clinical studies, the prevalence of preoperative anxiety has varied widely, from 11% to 80%, depending on the methods used to assess it (Aust 2018; Corman 1958; Johnston 1980; Norris 1967; Wallace 1984). High levels of anxiety can occur for at least five or six days before admission to hospital; for some patients, anxiety remains high for several days after surgery (Johnston 1980). Risk factors for preoperative anxiety include female sex, high trait anxiety (tendency to experience anxiety), negative future perception, history of cancer and smoking, previous psychiatric disorder, moderate to intense depressive symptoms, and higher educational level (> 12 years) (Caumo 2001a). Previous surgery reduces the risk of preoperative anxiety (Caumo 2001a). Furthermore, preoperative anxiety has been found to correlate with high postoperative anxiety (Caumo 2001).

Historically, postoperative anxiety has received less attention than preoperative anxiety; however, recent evidence suggests that postoperative anxiety may have adverse effects on postoperative outcomes (Jellish 2012). Risk factors shown to be associated with postoperative anxiety are moderate to intense postoperative pain, preoperative state anxiety, history of smoking, negative future perception, and minor psychiatric disorder (Jellish 2012). Systemic multi‐modal analgesia has been shown to be protective for postoperative anxiety (Jellish 2012).

According to the literature, medical interventions including the most widely used anxiolytic‐sedatives (benzodiazepines), effective communication strategies in the perioperative period, cognitive‐behavioural therapy (CBT), perioperative education, music therapy, massage therapy, and psychological preparation can be used successfully to reduce anxiety among surgical patients (Bailey 2010; Bradt 2013; Dao 2011; Jellish 2012, Kesanen 2017; Powell 2016; Wentworth 2009; Wilson 2016). Other drugs such as alpha‐ and beta‐adrenoceptor blockers have been used to reduce preoperative anxiety, but they may result in cardiovascular complications (Blessberger 2019a; Blessberger 2019b; Duncan 2018). Other drugs (e.g. lidocaine (Weibel 2018; Weinstein 2018)) used to treat pain may also have calming or euphoric effects but usually are given postoperatively.

Description of the intervention

Melatonin (N‐acetyl‐5‐methoxytryptamine) is synthesized from tryptophan and is secreted principally by the pineal gland. It has an endogenous circadian rhythm of secretion induced by the suprachiasmatic nuclei of the hypothalamus that is entrained to the light and dark cycle (Claustrat 2005). Melatonin has several putative functions including regulation of circadian rhythm, as well as sedative, analgesic, anxiolytic, anti‐inflammatory, antioxidative, and oncostatic effects (Brzezinski 1997; Ebadi 1998; Maestroni 1993; Reiter 1995).

Exogenous melatonin is produced synthetically from reacting chemical compounds (Jarratt 2011). Synthetic melatonin is produced from pharmacy‐grade ingredients under strict laboratory conditions in the form of tablets, capsules, liquids, or powder.

Although synthetic melatonin is molecularly identical to endogenous melatonin, its bioavailability varies widely (Harpsoe 2015). Oral doses (1 to 5 mg) result in serum melatonin concentrations that are 10 to 100 times higher than the usual night‐time peak within one hour after ingestion, followed by a decline to baseline values in four to eight hours (Brzezinski 1997). Very low oral doses (0.1 to 0.3 mg) given in the daytime result in peak serum concentrations that are within the normal night‐time range (Dollins 1994).

How the intervention might work

Anxiety is considered to be a multi‐factorial phenomenon with genetic, biochemical, humoral, neurophysiological, and psychological factors (Nolte 2011).

Autoradiographic studies and receptor assays in humans have demonstrated the presence of melatonin receptors in various regions of the central nervous system (CNS) and other tissues (Stankov 1991). In addition, both experimental ‐ Tian 2010 ‐ and clinical studies ‐ Acil 2004; Caumo 2007; Caumo 2009; Dianatkhah 2015; Ionescu 2008; Ismail 2009; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Torun 2019; Turkistani 2007 ‐ have shown an anxiolytic effect of melatonin. Exogenous administration of melatonin has been found to facilitate the onset of sleep and to improve its quality (Wurtman 1995). As premedication, compared to widely used benzodiazepines, melatonin produces no residual effects or suppression of rapid eye movement sleep (Zhdanova 1995). Therefore, it could be a worthy alternative.

Due to various effects of melatonin (regulation of circadian rhythm, and sedative, analgesic, anti‐inflammatory, antioxidative, and oncostatic effects (Brzezinski 1997; Ebadi 1998; Maestroni 1993; Reiter 1995)), it is not possible to distinguish the direct anxiolytic effect because it may occur as an interaction of several of these mechanisms.

Melatonin is considered a drug of low toxicity. A safety study done with very high oral doses of melatonin (50 mg/kg body weight orally) showed no serious adverse events (Nickkholgh 2011). In addition, a non‐systematic review reported headache, dizziness, nausea, and sleepiness as the most common adverse effects (Andersen 2016). These review authors concluded that short‐term use of melatonin is safe even in large doses. A systematic review assessed adverse effects of melatonin reported in 50 studies (Foley 2019). These review authors concluded that melatonin supplement in humans appears relatively safe, and that reported adverse events are generally minor, short‐lived, and easily managed, with some exceptions in particular populations such as patients with Huntington's chorea.

Why it is important to do this review

Patients' preoperative anxiety influences their postoperative anxiety (Caumo 2001), pain (Bayrak 2019; Doleman 2018; Gorkem 2016; Kain 2000; Thomas 1995), analgesic requirements (Thomas 1995), length of hospital stay (Caumo 2001), and satisfaction with perioperative care and treatment (Ali 2017; Caumo 2001a; Jamison 1993). Perioperative anxiety can lead to aggressive reactions that result in an increase in distress experienced by the patient and can make management and control of postoperative pain more difficult (Caumo 2001a). In addition, psychological distress, including preoperative and postoperative anxiety, may lead to more frequent demands for analgesics in patient‐controlled analgesia, as well as increased intraoperative analgesic requirements (Ip 2009; Pan 2006). Overall, it appears that patients with a high level of anxiety or a high level of distress preoperatively may experience higher rates of postoperative complications and may have impaired wound healing (Britteon 2017; Mavros 2011). Furthermore, preoperative anxiety has been shown to be a predictor of mortality and major morbidity in older patients (> 70 years) undergoing cardiac surgery (Williams 2013). Overall, treating anxiety in the perioperative period can improve the perioperative experience of the patient (Jellish 2012).

It is common practice in some day‐case surgical units to use benzodiazepines, opioids, or beta‐blockers as anxiolytic premedication when needed (Walker 2009). Their known adverse effects limit the safe use of these drugs. In particular, use of benzodiazepines can result in psychomotor impairment, cognitive impairment, daytime sleepiness, and sedation ('hang‐over effect'), even after single‐dose administration (Ashton 1994; Edwards 1981; Gudex 1991; Woods 1992).

Potential clinical benefits of new therapeutic options in this setting have been only sparsely investigated. Several studies have investigated the perioperative anxiolytic effects of melatonin (Capuzzo 2006; Dianatkhah 2015; Hoseini 2015; Pokharel 2014), and some have found positive results (Acil 2004; Caumo 2007; Caumo 2009; Ionescu 2008; Ismail 2009; Jain 2019; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Torun 2019; Turkistani 2007). Furthermore, melatonin is a non‐toxic drug with no reports of serious adverse events with short‐term use (less than three months) (Andersen 2016; Buscemi 2006; Nordlund 1977; Seabra 2000).

The hypnotic, antinociceptive, and anticonvulsant properties of melatonin endow this neurohormone with the profile of a novel hypnotic‐anaesthetic agent (Naguib 2007). Melatonin administration is also associated with a tendency towards faster recovery and a lower incidence of postoperative excitement than are seen with midazolam (Naguib 2007). Thus, we found it important and relevant to investigate whether melatonin can provide the preoperative and postoperative anxiolytic effects sometimes needed in day‐case and in‐patient surgery.

This is the first update of a previously published review (Hansen 2015). The purpose of updating this review was to explore if new trials have been published that would either alter or support the conclusions made in the previous review (Hansen 2015).

Objectives

To assess the effects of melatonin on preoperative and postoperative anxiety in adults compared to placebo or benzodiazepines.

Methods

Criteria for considering studies for this review

Types of studies

We included randomized controlled‐ and cluster‐randomized studies that were placebo‐controlled or standard treatment‐controlled, or both, that evaluated the effects of melatonin on preoperative or postoperative anxiety.

We included studies irrespective of language and publications status. We excluded quasi‐randomized and cross‐over studies.

Types of participants

We included adult patients of both sexes (15 to 90 years of age) undergoing any kind of surgical procedure for which it was necessary to use general, regional, or topical anaesthesia.

Types of interventions

To be included, patients had to receive melatonin, placebo, or a benzodiazepine administered on the day before surgery or immediately before surgery.

The intervention group (melatonin) was compared with a group receiving placebo or was compared with a group receiving benzodiazepines.

Types of outcome measures

Primary outcomes

Preoperative anxiety measured by a visual analogue scale (VAS), State‐Trait Anxiety Inventory (STAI), or any other validated assessment tool. We regarded the preoperative period as the two hours leading up to either surgery or induction of anaesthesia. There were no restrictions regarding how long after premedication preoperative anxiety had to be assessed.

The STAI is a validated questionnaire used to assess anxiety. The scale is divided into two subscales: the Trait‐Anxiety subscale consists of 20 questions focusing on a person's general level of fearfulness, whereas the State‐Anxiety subscale measures immediate situational anxiety. The range of scores is 20 to 80 per subscale, with higher scores indicating greater anxiety. Trait‐anxiety is a constant, whereas State‐anxiety can differ according to the situation. The two subscales are not combined but are viewed separately.

VAS is a 100‐mm scale, ranging from 0 to 100, whereby the extremes are marked "no anxiety" and "worst anxiety ever".

Both the simple VAS and the STAI have proved to be useful and valid measures of preoperative anxiety, and they are equivalent in terms of the assessment of preoperative anxiety (Kindler 2000; Millar 1995).

A minimal clinically important difference for preoperative and postoperative anxiety has not yet been fully established; however, for acute pain assessment, a difference of 9 to 14 on a VAS has previously been estimated to be the minimal clinically significant difference (Kelly 1998; Kelly 2001). Therefore, we regarded a difference in preoperative and postoperative anxiety of 9 to 14 mm on a 0 to 100 mm VAS as clinically important.

To our knowledge, no minimal clinically important difference in STAI for preoperative and postoperative anxiety has been established. We viewed a difference of 10% (8 points on the STAI) as the minimal clinically important difference.

Secondary outcomes

Postoperative anxiety measured by VAS or STAI. Postoperative anxiety was divided into immediate postoperative anxiety (measured after surgery in the recovery room or at discharge from the recovery room) and delayed postoperative anxiety (measured six hours after surgery).

In addition, harms reported in the included studies were summarized qualitatively.

Search methods for identification of studies

This review is the first update of a previously published review (Hansen 2015). The search strategy has been updated to include additional search terms in the interest of improving the sensitivity of the search. Searches were conducted and reported as outlined in the Cochrane Handbook for Systematic Review of Interventions (Higgins 2019b). We did not impose any language or publication restrictions.

Electronic searches

We searched the following databases.

Cochrane Central Register of Controlled Trials (CENTRAL; Issue 7 of 12; July 2020), in the Cochrane Library.
MEDLINE ALL (Ovid SP, 1966 to July 2020).
Embase (Ovid SP, 1980 to July 2020).
Cumulative Index to Nursing and Allied Health Literature (CINAHL; EBSCOhost; 1982 to July 2020).
Web of Science (SCI‐EXPANDED 1945 to July 2020).

We searched CENTRAL using the terms found in Appendix 1. We adapted the search strategy for MEDLINE (Appendix 2), Embase (Appendix 3), CINAHL (Appendix 4), and Web of Science (Appendix 5). We combined the MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE (Lefebvre 2019). When appropriate, we used similar search strategies for identifying RCTs in the other databases. We searched the bibliographic references and citations of relevant studies and systematic reviews for further potentially relevant studies. We searched the following trial registries for unpublished and ongoing studies.

ClinicalTrials.gov (https://www.clinicaltrials.gov/).
World Health Organization International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/Default.aspx).

Searching other resources

We screened the reference lists of all eligible trials and reviews. The lead review author (BKM) contacted the authors of published trials to request additional information when necessary.

Data collection and analysis

Selection of studies

Using results of the above searches, we screened all titles and abstracts for eligibility and excluded the ones that clearly did not meet the inclusion criteria. Two review authors (BKM and DZ) independently performed this screening. For the remaining studies, we read the full manuscript or trial register entry to assess whether they should be included. If a trial was excluded, the reason for exclusion was documented (see Excluded studies).

In the case of insufficient published information to make a decision about inclusion, we contacted the corresponding author of the relevant trial (BKM). If a study was reported in a foreign language not understandable to the present review author group, a suitable translator was found.

Details on the included studies can be seen in the Characteristics of included studies tables.

Data extraction and management

One review author (BKM) independently extracted data twice using a standard form and looked for discrepancies before entering data into RevMan. Any discrepancies in the extracted data were resolved by discussion (BKM and DZ).

In the case of additional information being required, BKM contacted the corresponding author of the relevant trial. If a study was reported in a foreign language, a translator was found to help with extraction of data.

Data extracted included information on study design, country of origin, number of participants and demographic details, type of surgery and anaesthesia, intervention and dosing regimen, preoperative anxiety outcome measures, and postoperative anxiety outcome measures.

Assessment of risk of bias in included studies

One review author (BKM) independently assessed the methodological quality of the included trials. If a study was reported in a foreign language, we found a translator to assist with assessment of bias.

We performed the assessment as suggested in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions ‐ Higgins 2011 ‐ because we assessed risk of bias before the release of version 6.0 (Higgins 2019). See the 'Risk of bias' table in Characteristics of included studies.

Review authors assessed the risk of bias for the following domains.

Random sequence generation.
Allocation concealment.
Incomplete outcome data.
Selective reporting.
Blinding of participants and personnel.
Blinding of outcome assessment.
Other potential sources of bias.

Review authors reviewed the aforementioned domains to perform an overall risk of bias assessment.

Review authors judged each of the above domains to have low (adequate), high (inadequate), or unclear risk of bias. If there were any doubts about the judgement, two review authors (BKM and DZ) resolved this uncertainty by discussion.

Random sequence generation (checking for possible selection bias)

We considered random sequence generation adequate if it was generated by a computer or by a random number table algorithm. We judged other processes ‐ such as tossing a coin ‐ to be adequate if the whole sequence was generated before the start of the trial, and if it was performed by a person not otherwise involved in patient recruitment.

We considered random sequence generation unclear if insufficient information was provided about the sequence generation process to permit judgement.

We considered random sequence generation inadequate if a non‐random system, such as dates, names, or identification numbers, was used.

Allocation concealment (checking for possible selection bias)

We considered concealment adequate if the process used prevented patient recruiters, investigators, and participants from knowing the intervention allocation of the next participant to be enrolled in the study. Acceptable systems included a central allocation system, sealed opaque envelopes, or an on‐site locked computer.

We considered allocation concealment unclear if the method of concealment was not described.

We considered concealment inadequate if the allocation method that was used allowed patient recruiters, investigators, or participants to know the treatment allocation of the next participant to be enrolled in the study. For example, alternate medical record numbers, reference to case record numbers or date of birth, an open allocation sequence, or unsealed envelopes.

Incomplete outcome data (checking for possible attrition bias)

We considered dropout or missing data reported as adequate if studies had no dropouts or missing data. We also considered the domain adequate if studies described reasons for dropouts, and if there were balanced numbers of participants dropping out across intervention groups.

Selective reporting (checking for possible reporting bias)

We considered selective reporting adequate if the study protocol was available, and if all of the study's pre‐specified outcomes were reported in the article.

We considered selective reporting unclear if a study protocol was referred to but was not obtainable, or if no study protocol was available.

We considered selective reporting inadequate if one or more outcomes reported in the article were not pre‐specified in the study protocol.

Blinding of participants and personnel (checking for possible performance bias)

We considered blinding adequate if participants and personnel were each blinded to the intervention. With regards to the intervention, we deemed blinding to be adequate if the melatonin, placebo, or benzodiazepines had an identical appearance.

We considered blinding unclear if there was insufficient information to permit judgement.

We considered blinding inadequate if participants and personnel were not blinded to the intervention.

Blinding of outcome assessment (checking for possible detection bias)

We considered blinding of outcome assessors adequate if the blinding was sufficiently described.

We considered blinding of outcome assessors unclear if there was insufficient information to permit judgement.

We considered blinding of outcome assessors inadequate if outcome assessors were not blinded.

Other potential biases

We considered other sources of bias not covered in the above domains.

Overall risk of bias

We assessed the domains blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting for each outcome. We performed an overall risk of bias assessment for each outcome, which we used to assess the certainty of evidence. Based on this assessment, we defined the included trials and each outcome result as showing low, unclear, or high risk of bias. An outcome was regarded to have low risk of bias if the included studies had an overall low risk of bias. If studies had low or unclear risk of bias in all domains, it was regarded as overall unclear risk of bias and if it was considered plausible that bias might raise some concerns about the results. If studies had high risk of bias for one or several domains, the overall risk of bias for the outcome was regarded as high, and it was interpreted that bias might seriously weaken confidence in the results.

Measures of treatment effect

We extracted VAS or STAI data for our primary outcome as the mean (standard deviation (SD)) or median (interquartile range (IQR) or range). We chose to analyse VAS or STAI data as continuous data and presented these as the mean difference (MD) when outcome measures were on the same scale. We expressed the overall results for our primary outcome as mean difference with 95% confidence intervals (CIs).

Unit of analysis issues

We included only randomized placebo‐controlled, standard treatment‐controlled, single‐ or double‐blinded trials, and we excluded quasi‐randomized and cross‐over trials. We separated comparisons (benzodiazepines vs melatonin and placebo vs melatonin) into two separate forest plots; hence there were no unit of analysis issues.

No cluster‐randomised trials were found, but we had planned to include them in the meta‐analysis. If such trials had been identified, we would have adjusted the sample size according to the method described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019), using an estimate of the intracluster correlation coefficient (ICC) derived from the trial or using external estimates obtained from similar studies or populations. Furthermore, a sensitivity analysis would be performed to investigate the robustness of conclusions.

Dealing with missing data

Whenever possible, we contacted the original investigators to request missing data.

We converted standard error of the mean (SEM) to standard deviation (SD), and we converted median (IQR or range) to mean (SD), using the methods outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

Assessment of heterogeneity

We assessed the clinical heterogeneity of included studies, assessed as clinical diversity (e.g. different types of anaesthesia (regional, general, topical), differences in patient characteristics, variable melatonin doses, differences in analgesics) and as methodological diversity (variability in study design and in risk of bias).

We assessed statistical heterogeneity with the I² statistic, thereby estimating the percentage of total variance across studies that was due to heterogeneity rather than to chance (Higgins 2019).

The authors interpreted values of the I² statistic as follows (Higgins 2019).

0% to 40%, might not be important.
30% to 60%, may represent moderate heterogeneity.
50% to 90%, may represent substantial heterogeneity.
75% to 100%, considerable heterogeneity.

Assessment of reporting biases

We assessed publication bias and small‐study effects in a qualitative manner using a funnel plot in Review Manager 5.3 (RevMan 5.3). Because we had 27 included studies, we planned to look at whether the largest studies were near the average and small studies were spread on both sides of the average.

Data synthesis

We performed data synthesis and statistical analysis using Review Manager software (RevMan 5.3). Because the population was varied, we included all types of anaesthesia and surgery, adult participants of both sexes between the ages of 15 and 90 years, dosing regimens, and study sizes. Due to this variation, a random‐effects model was deemed suitable for the meta‐analysis.

As some studies used several different doses of melatonin or benzodiazepines, we chose to combine the groups receiving different doses of either melatonin or benzodiazepine into one melatonin or benzodiazepine group, respectively.

Studies reported our primary outcome as mean (SD) or median (IQR or range). For all studies reporting median, we assumed symmetrical distribution of data and used the median value directly in the meta‐analyses as the mean. However, we decided to perform sensitivity analysis without studies reporting outcomes using a median (IQR), because the use of interquartile ranges rather than standard deviations can sometimes indicate that the outcome distribution is skewed (Higgins 2019). If the studies did not present data in a tabular fashion, we read the values directly from the graphs. If the studies reported changes from baseline (VAS change scores), we used corresponding negative or positive values. Both studies reporting VAS and those reporting VAS change scores were entered in the same meta‐analysis as subgroups, and the results of both subgroups were pooled.

We converted SEM to SD using the method presented in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). For all other studies, we converted median (IQR or range) to mean (SD) using the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

We analysed continuous data using an inverse variance method. We performed the analysis using Review Manager software (RevMan 5.3).

We chose to perform five meta‐analyses.

Primary outcome: melatonin versus placebo (VAS) preoperatively.
Secondary outcome: melatonin versus placebo (VAS) postoperatively.
Secondary outcome: melatonin versus placebo (STAI) postoperatively.
Primary outcome: melatonin versus benzodiazepine (VAS) preoperatively.
Secondary outcome: melatonin versus benzodiazepine (VAS) postoperatively.

Two studies measured preoperative anxiety using STAI, where one study used a modified version of STAI (STAI‐S). Due to the limited number of studies, we chose not to perform meta‐analysis on this primary outcome, and instead to provide a narrative description of study findings.

Subgroup analysis and investigation of heterogeneity

Studies that used VAS to measure anxiety reported the outcome either as VAS total or as change in VAS from baseline. We used a random‐effects model when both subgroups (VAS total and change in VAS from baseline) were included in the same meta‐analysis.

We performed three additional subgroup analyses to explore heterogeneity.

Anaesthetic modality (regional or general).
Participants' age (≤ 60 or > 60 years).
Melatonin dose (anticipated range 1 to 20 mg).

We decided to divide studies into two groups according to the dose of melatonin administered (< 6 mg or ≥ 6 mg).

We explored if heterogeneity disappeared when the data were divided into subgroups depending on anaesthesia modality, participant age, or dose of melatonin administered.

Sensitivity analysis

We performed sensitivity analysis whereby we repeated the meta‐analysis for preoperative anxiety (VAS) after excluding studies reporting only the median (IQR or range) for VAS data on preoperative anxiety. We did this because the use of interquartile ranges rather than standard deviations can sometimes be taken as an indicator that the outcome distribution is skewed (Higgins 2019).

We also performed sensitivity analysis for postoperative anxiety (VAS) after excluding studies reporting the median (IQR or range) for VAS data on postoperative anxiety, or studies reporting SD values of zero, which made us suspect that outcome was skewed.

We also performed a separate sensitivity analysis of our primary and secondary outcomes. We excluded all studies with an overall high risk of bias to be able to explore if studies with high risk of bias affected conclusions.

Summary of findings and assessment of the certainty of the evidence

We used the principles of the GRADE system (Guyatt 2008) to assess the certainty of the body of evidence associated with our primary outcome (preoperative anxiety) and secondary outcome (postoperative anxiety) and constructed "Summary of findings" (SoF) tables using Review Manager 5.3 (RevMan 5.3).

The GRADE approach appraises the certainty of a body of evidence‐based on the extent to which one can be confident that an estimate of effect or association reflects the item being assessed. The certainty of a body of evidence considers five domains: study limitations (risk of bias), inconsistent results, indirectness of evidence, imprecision, and publication bias. The certainty of evidence can be downgraded if a reason in the above‐mentioned domains is found. If a serious reason was found the certainty of evidence was downgraded by one level if a very serious reason was found, the certainty of evidence was downgraded by two levels. We downgraded evidence if the outcome had an overall high risk of bias. However, if sensitivity analysis excluding studies with an overall high risk of bias showed a similar effect estimate, we decided not to downgrade evidence since we concluded that the inclusion of studies with a high risk of bias did not alter conclusions. We downgraded evidence if the outcome had high heterogeneity expressed with a high I² value. Evidence was downgraded if there were indirectness of evidence or high probability of publication bias. Also, evidence was downgraded if there were few participants and thus wide confidence intervals.

SoF tables were created for both intervention comparisons, including all primary and secondary outcomes:

melatonin versus placebo: melatonin versus placebo (VAS) preoperatively, melatonin versus placebo (STAI) preoperatively, melatonin versus placebo (six‐item STAI) preoperatively, melatonin versus placebo (VAS) immediate postoperative anxiety, melatonin versus placebo (STAI) delayed postoperative anxiety, melatonin versus placebo (six‐item STAI) postoperatively
melatonin versus benzodiazepine: melatonin versus benzodiazepine (VAS) preoperatively, melatonin versus benzodiazepine (STAI) preoperatively, melatonin versus benzodiazepine (six‐item STAI) preoperatively, melatonin versus benzodiazepine (VAS) immediate postoperative anxiety, melatonin versus benzodiazepine (six‐item STAI) postoperatively.

Results

Description of studies

Twenty‐seven studies, published between 1999 and 2019, met the inclusion criteria.

Twenty‐four studies compared melatonin with placebo. Twelve studies compared only melatonin with placebo (Abbasivash 2019; Capuzzo 2006; Caumo 2007; Ismail 2009; Jain 2019; Khezri 2013; Khezri 2016; Mowafi 2008; Naguib 2006; Norouzi 2019; Seet 2015; Turkistani 2007). In addition to placebo, six studies compared melatonin with the benzodiazepine midazolam (Acil 2004; Ionescu 2008; Naguib 1999; Naguib 2000; Patel 2015; Torun 2019), two compared melatonin with the benzodiazepine alprazolam (Khare 2018; Pokharel 2014), three compared melatonin with gabapentin (Hoseini 2015; Javaherforooshzadeh 2018; Khezri 2013b), and two compared melatonin with clonidine (Caumo 2009; Hoseini 2015). One study compared only melatonin with the benzodiazepine oxazepam (Dianatkhah 2015), and one compared melatonin with both pregabalin and alprazolam (Khanna 2019). Another study compared melatonin with gabapentin and placebo; however, the placebo group received 1 mg of midazolam intravenously during operation; hence, we decided to classify this study as having melatonin, gabapentin, and midazolam groups (Marzban 2016).

Results of the search

We identified 2881 references in primary electronic databases in July 2020 through our search strategy. We searched clinical trial registration databases and identified 1386 trial register records. We searched bibliographic references and citations of relevant studies and systematic reviews and identified two potential references.

Out of the total of 2881 database references, we removed 748 duplicates. In total, together with the 1386 records identified through other sources, we screened 3519 records and excluded 3424 records because they clearly did not meet the eligibility criteria (Figure 1).

Figure 1

Study flow diagram.

We conducted more in‐depth screening of 95 records (including full‐text reports and trial register records). We obtained full‐text reports for 55 references found through our electronic database searches and through searching of reference lists, to check if they strictly fulfilled the inclusion criteria. We excluded 29 studies due to irrelevant interventions or outcomes, lack of an appropriate comparison group, or irrelevant study design, or because the report was a review article, PhD thesis summary, or conference abstract (Characteristics of excluded studies). In addition, we thoroughly read the full trial register records of 40 records found through our clinical trial registration database searches. Two studies were awaiting classification when we updated this review (Characteristics of studies awaiting classification). We retrieved the published manuscript for one study and therefore included it in our review (Marzban 2016). Furthermore, we found 10 ongoing studies (Characteristics of ongoing studies).

We found 27 studies that completely fulfilled the inclusion criteria for this review; 12 of these were included in the former review (Acil 2004; Capuzzo 2006; Caumo 2007; Caumo 2009; Ionescu 2008; Ismail 2009; Khezri 2013; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Turkistani 2007), and 15 were new additions (Abbasivash 2019; Dianatkhah 2015; Hoseini 2015; Jain 2019; Javaherforooshzadeh 2018; Khanna 2019; Khare 2018; Khezri 2013b; Khezri 2016; Marzban 2016; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019).

Included studies

See Characteristics of included studies for a description of the methods, participants, interventions, and outcomes of the individual studies.

A total of 2319 patients were randomized in the included studies, of whom 2227 patients (25 studies) had data concerning preoperative anxiety (Abbasivash 2019; Acil 2004; Capuzzo 2006; Dianatkhah 2015; Hoseini 2015; Ionescu 2008; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khanna 2019; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Marzban 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019; Turkistani 2007), and 1354 patients (15 studies) had data concerning postoperative anxiety (Acil 2004; Capuzzo 2006; Caumo 2007; Caumo 2009; Dianatkhah 2015; Ionescu 2008; Javaherforooshzadeh 2018; Khanna 2019; Khezri 2013; Khezri 2013b; Khezri 2016; Marzban 2016; Naguib 1999; Naguib 2000; Norouzi 2019). The age of included patients ranged from 17 to 85 years. No cluster‐randomized studies were found.

The number of participants in the included studies varied from 33 to 200. Of the 27 studies, five included only women (Caumo 2007; Caumo 2009; Naguib 1999; Naguib 2000; Khezri 2016), three included more females (Khare 2018; Pokharel 2014; Torun 2019), three included more men (Dianatkhah 2015; Khezri 2013b; Seet 2015), and 12 had a close to equal distribution of males and females (Abbasivash 2019; Capuzzo 2006; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khezri 2013; Marzban 2016; Mowafi 2008; Naguib 2006; Norouzi 2019; Patel 2015; Turkistani 2007), with the exception of two studies that had a greater number of females in the placebo group ‐ Naguib 2006 ‐ and in the midazolam group ‐ Patel 2015 ‐ respectively. Three studies did not provide information on distribution of sex (Acil 2004; Hoseini 2015; Khanna 2019), and the remaining study did not define which group was female or male (Ionescu 2008). Seventeen of the 27 studies were carried out in Middle East countries (Saudi Arabia, Turkey, and Iran), one in Italy, one in Romania, two in Brazil, four in India, one in Singapore, and one in Nepal.

Three studies used STAI to measure anxiety (Caumo 2007; Caumo 2009; Hoseini 2015), one used a modified version of STAI (STAI‐S) consisting of six items from the STAI questionnaire producing a total score between 6 to 24 (Ionescu 2008), one used the Hamilton Anxiety Rating Scale (HAM‐A) (Dianatkhah 2015), one used a numerical rating scale (NRS) (Capuzzo 2006), one used the Beck Anxiety Inventory (BAI) (Khanna 2019), and the remaining studies assessed anxiety using a visual or verbal anxiety scale (VAS) (Abbasivash 2019; Acil 2004; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Marzban 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019; Turkistani 2007).

Twenty‐two studies compared melatonin with placebo for preoperative anxiety (Abbasivash 2019; Acil 2004; Capuzzo 2006; Hoseini 2015; Ionescu 2008; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019; Turkistani 2007). Eighteen studies used a VAS to measure anxiety (Acil 2004; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019; Turkistani 2007), and two studies used STAI and a six‐item STAI, respectively (Hoseini 2015; Ionescu 2008). Because these two studies were not comparable to the rest, we decided not to include these two studies in the meta‐analysis. One study used an NRS (Capuzzo 2006), and we assumed that this was comparable to the VAS (Hjermstad 2011).

Eleven studies compared melatonin with placebo in the recovery room, in the recovery room at discharge, or 90 minutes postoperatively (Acil 2004; Capuzzo 2006; Ionescu 2008; Javaherforooshzadeh 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Marzban 2016; Naguib 1999; Naguib 2000; Norouzi 2019). Of these, 10 studies used VAS or NRS to measure anxiety (Acil 2004; Capuzzo 2006; Javaherforooshzadeh 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Marzban 2016; Naguib 1999; Naguib 2000; Norouzi 2019), whereas one study used a six‐item STAI and therefore was not included in the meta‐analysis (Ionescu 2008).

Four studies compared melatonin with placebo six hours postoperatively (Caumo 2007; Caumo 2009; Javaherforooshzadeh 2018; Ionescu 2008). Three studies measured postoperative anxiety using the STAI (Caumo 2007; Caumo 2009; Ionescu 2008); however, Ionescu 2008 used a six‐item state anxiety scale (STAI‐S) and therefore was not included in the meta‐analysis. One study measured postoperative anxiety using a VAS (Javaherforooshzadeh 2018). This study was not included in the meta‐analysis for six hours postoperative because it was not comparable with the remaining two studies, which measured delayed postoperative anxiety using the STAI (Caumo 2007; Caumo 2009).

Eleven studies compared melatonin with a benzodiazepine for preoperative anxiety (Acil 2004; Dianatkhah 2015; Ionescu 2008; Khanna 2019; Khare 2018; Marzban 2016; Naguib 1999; Naguib 2000; Patel 2015; Pokharel 2014; Torun 2019). Eight studies used a VAS to measure anxiety (Acil 2004; Khare 2018; Marzban 2016; Naguib 1999; Naguib 2000; Patel 2015; Pokharel 2014; Torun 2019), one study used a six‐item STAI (Ionescu 2008), one study used HAM‐A (Dianatkhah 2015), and one study used BAI (Khanna 2019). Dianatkhah 2015 Ionescu 2008, and Khanna 2019 were not included in the meta‐analysis because the scales used were not comparable to the visual or verbal anxiety scale.

Seven studies compared melatonin with a benzodiazepine 60 to 90 minutes postoperatively (Acil 2004; Dianatkhah 2015; Ionescu 2008; Khanna 2019; Marzban 2016; Naguib 1999; Naguib 2000). Four studies used a VAS to measure anxiety (Acil 2004; Marzban 2016 Naguib 1999; Naguib 2000). One study used the HAM‐A (Dianatkhah 2015), one study used the BAI (Khanna 2019), and one study used the six‐item STAI (Ionescu 2008); these studies were not included in the meta‐analysis.

For eight studies (Acil 2004; Caumo 2007; Caumo 2009; Ismail 2009; Khezri 2016; Naguib 1999; Naguib 2000; Pokharel 2014), data were presented only graphically. For one study (Naguib 2000), the graph for melatonin and midazolam was difficult to read, and study authors were contacted, but we received no answer. From the graph, the mean and the SD had to be measured with a ruler to interpret the VAS score. In Naguib 1999, it was straightforward to measure mean and SD for both placebo and melatonin arms. In Naguib 2000, the mean of the placebo arm could be read, and we assumed that the bar with the highest value indicated the SD of the placebo group. For melatonin and midazolam arms, three doses were used, and the means of doses were pooled, as they had equal numbers of participants. We did not find it possible to read the SD of the six arms, as we could not distinguish the error bars. Therefore, we chose to impute the SD for both melatonin and midazolam arms from the SD of the placebo arm. For Khezri 2016, it is not clear if the graph presented the outcome as mean (SD) or median (IQR or range). In the methods section of the manuscript, the author of the study stated that normally distributed data would be presented as mean (SD). Still, we were not able to verify if the data were normally distributed. Khezri 2016 also measured anxiety on a scale from 0 to 10 and graphically illustrated a scale going from 0 to 18. Therefore, we chose not to include the study in our meta‐analysis, because we could not with certainty conclude what the graph showed (we contacted the study author by email but received no reply). Acil 2004 did not report an SD for preoperative or postoperative anxiety (we contacted the study author but received no answer), so we did not include this study in the meta‐analysis as the conversion from P value to SD was not possible. Khanna 2019, using the BAI, also provided no SD values or range. Study authors provided no contact information, so we were unable to contact them.

One study did not report how outcomes were presented (Marzban 2016), but we assumed they were presented as mean SD (the study author was contacted by email and confirmed this). The study compared melatonin to a placebo. However, the placebo group was given midazolam before preoperative anxiety was measured; therefore, we chose to consider the placebo group as a midazolam group and included the study in the meta‐analysis comparing melatonin to a benzodiazepine. The study reported SD values of 0, indicating that distribution was skewed, and melatonin and the comparator (midazolam) were given at different times via different administration routes. Therefore, we decided not to include this study in the sensitivity analysis.

Eight studies reported preoperative anxiety as median (range or IQR) (Capuzzo 2006; Ismail 2009; Khezri 2013; Khezri 2013b; Mowafi 2008; Naguib 2006; Pokharel 2014; Turkistani 2007), and two studies reported postoperative anxiety as mean (SEM) (Caumo 2007; Caumo 2009). We converted these to mean (SD) using the method provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). However, because this method was not robust, we performed sensitivity analysis on preoperative anxiety while excluding the eight studies (Capuzzo 2006; Ismail 2009; Khezri 2013; Khezri 2013b; Mowafi 2008; Naguib 2006; Pokharel 2014; Turkistani 2007). Three additional studies reported medians of zero (Capuzzo 2006; Khezri 2013; Khezri 2013b), thereby violating the assumption of symmetry; they were not included in the sensitivity analysis.

We decided to perform an additional sensitivity analysis whereby we excluded all studies with overall high risk of bias from our meta‐analysis. Eight studies were assessed as having overall high risk of bias, and of these, we excluded seven studies from the sensitivity analysis (Hoseini 2015; Javaherforooshzadeh 2018; Khezri 2013; Khezri 2013b; Marzban 2016; Norouzi 2019; Pokharel 2014). The remaining study was not included in any meta‐analysis (Dianatkhah 2015).

Type of surgery and anaesthesia

Two studies were performed in patients undergoing abdominal hysterectomy (Caumo 2007; Caumo 2009), four studies in patients undergoing cataract surgery (Ismail 2009; Khezri 2013; Khezri 2013b; Marzban 2016), four studies in patients undergoing laparoscopic cholecystectomy (Acil 2004; Ionescu 2008; Pokharel 2014; Hoseini 2015), two studies in patients undergoing gynaecological laparoscopic procedures (Naguib 1999; Naguib 2000), one study in patients undergoing caesarean section (Khezri 2016), two studies in patients undergoing elective hand surgery (Abbasivash 2019; Mowafi 2008), five studies in patients undergoing different surgical procedures (not specified) (Capuzzo 2006; Jain 2019; Khare 2018; Patel 2015; Turkistani 2007), one study in patients undergoing laparoscopic surgery (Khanna 2019), one study in patients having coronary artery bypass surgery (CABG) (Dianatkhah 2015), one study in patients having spinal surgery at two or three levels of laminectomy (Javaherforooshzadeh 2018), one study in patients undergoing elective extraction of all four wisdom teeth (Seet 2015), one study in patients undergoing impacted mandibular third molar surgery (Torun 2019), one study in patients undergoing non‐emergency abdominal surgery (Norouzi 2019), and one study in which researchers did not specify the type of surgery performed (Naguib 2006).

Seventeen studies used general anaesthesia (Acil 2004; Caumo 2007; Caumo 2009; Hoseini 2015; Ionescu 2008; Jain 2019; Javaherforooshzadeh 2018; Khanna 2019; Khare 2018; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Turkistani 2007), four used topical anaesthesia (Ismail 2009; Khezri 2013; Khezri 2013b; Marzban 2016), three used regional or local anaesthesia (Abbasivash 2019; Mowafi 2008; Torun 2019), one used spinal anaesthesia (Khezri 2016), and one did not specify the type of anaesthesia given (Dianatkhah 2015). One study used both general and spinal anaesthesia (Capuzzo 2006).

Interventions

Melatonin doses varied from 3 mg to 10 mg in the majority of studies. However, four studies administered melatonin as mg/kg ranging from 0.05 to 0.4 mg/kg (Naguib 2000; Naguib 2006; Patel 2015; Torun 2019). Seventeen studies administered melatonin orally (Abbasivash 2019; Capuzzo 2006; Caumo 2007; Caumo 2009; Hoseini 2015; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khanna 2019; Khare 2018; Marzban 2016; Mowafi 2008; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019; Turkistani 2007), nine studies administered melatonin sublingually (Acil 2004; Ionescu 2008; Khezri 2013; Khezri 2013b; Khezri 2016; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019), and one study did not describe the administration route (Dianatkhah 2015).

Melatonin was administered approximately 20 to 120 minutes before either surgery or induction of anaesthesia. Three studies also administered one dose of melatonin the evening before surgery (Caumo 2007; Caumo 2009; Ionescu 2008). One study administered only melatonin one hour before assigned sleep time the night before surgery (Dianatkhah 2015), and one study administered melatonin on the day of surgery but did not provide more detail (Capuzzo 2006).

Midazolam was administered in doses ranging from 1 to 15 mg or from 0.05 to 0.2 mg/kg. Alprazolam was administered in doses ranging from 0.25 to 0.5 mg. One study compared melatonin to 10 mg oxazepam (Dianatkhah 2015). Some studies also compared melatonin to clonidine or gabapentin (Caumo 2009; Hoseini 2015; Javaherforooshzadeh 2018; Khezri 2013b; Marzban 2016). Clonidine was administered in doses of 0.1 to 0.2 mg, and the gabapentin dose was 600 mg in all studies. One study also compared melatonin to 75 mg pregabalin (Khanna 2019)

Adverse effects

See Table 1 for more information on adverse effects described in primary study reports.

Table 1. Harms reported in primary study reports

Author, year	Comparison	Harms
Abbasivash 2019	Melatonin, placebo	No harms reported
Acil 2004	Melatonin, midazolam, placebo	"The melatonin group showed increased levels of sedation 90 min after premedication with respect to placebo... This group showed decreased levels of sedation with respect to midazolam..." (page 555) "Furthermore, in the preoperative period, impairment in psychomotor performance was more significant in the midazolam group. In the Trail Making A and B test....the melatonin and midazolam groups exhibited a significantly poorer performance compared with placebo. However, in the Word Fluency test, the midazolam group showed a significant impairment...whereas there was no difference between the scores of the melatonin and placebo groups... The placebo group showed better postoperative performance on the Word Fluency test. Amnesia was only significant in the midazolam group..." (page 556) No harms reported
Capuzzo 2006	Melatonin, placebo	No harms reported
Caumo 2007	Melatonin, placebo	No harms reported
Caumo 2009	Melatonin, clonidine, placebo	No harms reported
Dianatkhah 2015	Melatonin, oxazepam	"A smaller proportion of the participants experience delirium in the melatonin group (n=4, 0.06%) than in the oxazepam group (n=9, 0.12%), but this difference was not statistically significant (P value = 0.187)" (page 125) No harms reported
Hoseini 2015	Melatonin, clonidine, gabapentin, placebo	No harms reported; however, the frequency of vomiting and the severity of nausea were measured, and no differences between groups were observed (Table 3) (page 123)
Ionescu 2008	Melatonin, midazolam, placebo	"Amnesia scores, assessed as the number of remembered pictures, were significantly better (the score of the remembered pictures was greater) in the melatonin group in comparison to the midazolam group at every evaluation time, whereas there were no significant difference between the melatonin and placebo groups" (page 11) "No side effects of melatonin were noted" (page 10)
Ismail 2009	Melatonin, placebo	"Contraray to the control group, MAP decreased significantly after melatonin premedication. No incidence of hypotension or bradycardia requiring intervention was reported in groups...One patient in the melatonin group complained of dizziness, and another patient in control group suffered nausea" (page 1148)
Jain 2019	Melatonin, placebo	"In our study, there were no untoward incidences of bradycardia, cardiac arrhythmias, respiratory depression, nausea, hypotension, anaphylaxis, and drug interactions, in any of the groups" (page 20)
Javaherforooshzadeh 2018	Melatonin, gabapentin, placebo	"In this study, a single dose of gabapentin was used, thus, patients did not report any side effects. Ismail et al. found that MAP was significantly reduced after melatonin premedication, although it was described that this difference, at some points, was unimportant between the groups and was consistent with our results" (page 5)
Khanna 2019	Melatonin, pregabalin, alprazolam	"In or study we found that patients in group M were more sedated as compared to group P or group A at all intervals, and the difference at all intervals was statistically significant, whereas sedation score in patients of group P and group A was comparable at all intervals, and the difference at all intervals was statistically insignificant" (page 70) "Side effects like headache, dizziness were comparable in all groups..." (page 70)
Khare 2018	Melatonin, alprazolam, placebo	"Our results showed that both melatonin and alprazolam caused significant sedation in patients as compared to placebo. Among Group M and Group A, melatonin caused less sedation than alprazolam" (page 661) "In our study, alprazolam caused change in orientation score in patients when compared to melatonin and placebo groups...There was a decline in cognitive function in Group A as compared to Group P, whereas the cognitive function was enhanced or maintained in Group M..." (pages 661‐662) No harms reported
Khezri 2013	Melatonin, placebo	"No patient developed hypoxia, hypotension, or bradycardia. Only one patient in the melatonin group complained of a mild headache" (page 322)
Khezri 2013b	Melatonin, gabapentin, placebo	"Significant differences were observed between sedation scores during RBB placement in gabapentin and placebo groups. The difference in sedation scores during RBB placement in melatonin versus gabapentin and placebo was insignificant" (page 584) "No patient developed hypoxia, hypotension, bradycardia, excessive drowsiness (or sleepiness), nausea, and vomiting during surgery. One patient in the melatonin group complained of mild headache, and one in the gabapentin group of severe dizziness while staying in the ward" (page 584)
Khezri 2016	Melatonin, melatonin, placebo	"As shown in Table 3, apart of headache, no significant differences were found in the three groups in terms of other intraoperative and postoperative side effects including pruritus, nausea, vomiting, and respiratory depression. The incidence of headache in group M₆ was significantly higher than other groups" (page 966) "All newborns in our study were free of any adverse effect" (page 967)
Marzban 2016	Melatonin, gabapentin, placebo (midazolam)	Views sedation No harms reported
Mowafi 2008	Melatonin, placebo	"Melatonin premedication reduced MAP compared to control group... No incidence of hypotension or bradycardia requiring intervention was reported in either group" (page 1424) One patient complained of dizziness and two in the melatonin group had excessive sleepiness" (page 1424)
Naguib 1999	Melatonin, midazolam, placebo	"Patients who received midazolam and melatonin showed increased levels of sedation at 60 and 90 min... Furthermore, patients in the midazolam group showed significantly (P<0.05) higher levels of sedation compared with the melatonin group at 30 and 60 min after premedication..," (page 877) "However, in the preoperative period only patients in the midazolam group experienced significant impairment of psychomotor skills. After operation, patients who received midazolam or melatonin had increased levels of sedation at 30 min and impairment of performance of the DSST... Amnesia was notable only in the midazolam group for one preoperative event" (pages 878‐879) "No side effects were noted" (page 878)
Naguib 2000	Melatonin, midazolam, placebo	"...patients who received premedication with 0.05, 0.1 or 0.2 mg/kg sublingual midazolam or melatonin had a significant decrease in anxiety levels (Figure 1) and increase levels of sedation preoperatively... After operation, patients who received 0.2 mg/kg midazolam premedication had increased levels of sedation at 90 min compared with the 0.05 and 0.1 mg/kg melatonin groups" (pages 877‐878) "However, in the preoperative period, only patients in the three midazolam groups experiences significant impairment in psychomotor skills... In addition, patients in the three midazolam groups had impairment of performance on the DSST at 15, 30, 60, and 90 minutes postoperatively... Amnesia was notable only with the 0.2 mg/kg midazolam group for two preoperative events" (pages 477‐478) "No side‐effects were noted" (page 477)
Naguib 2006	Melatonin, placebo	"Here, oral premedication with 0.2 mg/kg melatonin approximately 50 min before induction of anaesthesia significantly reduced preoperative anxiety and increased sedation without impairment of orientation..." (page 1450) No harms reported
Norouzi 2019	Melatonin, placebo	"In addition, no significant difference was found in orientation between both before melatonin administration and in recovery (P>0.05), while it was statistically significant before anesthesia induction (P=0.44) and lower in the melatonin group before induction... In addition, there was no significant difference in sedation between the two groups..." (pages 64‐65) "The results of this double‐blinded clinical trial showed that MAP was lower in the melatonin group..." (page 65) No harms reported
Patel 2015	Melatonin, midazolam, placebo	"This showed that psychomotor and cognitive functions were not affected in melatonin group patients whereas they were significantly affected in midazolam group patients... This showed that midazolam produced the maximum derangement in both psychomotor and cognitive functions after premedication and before surgery" (pages 39‐40) "The intergroup comparison of sedation scores showed that midazolam produced the highest degree of sedation when compared to melatonin and placebo. Melatonin also showed sedative properties when compared with placebo" (page 41) No harms reported
Pokharel 2014	Melatonin, alprazolam, melatonin + alprazolam, placebo	"In our patients, alprazolam produced more sedation scores than placebo at 60 min after premedication, but the difference was not statistically significant. However, our patients who received alprazolam got sedated half an our earlier than placebo... We too found that the melatonin administration was associated with earlier onset of sleep than placebo" (pages 3‐4) "More number of patients in groups receiving the combination drugs and alprazolam (9 each) did not recognize the picture shown at 60 min after premedication...Amnesia for two events was notable in maximum number of patients in the group receiving the combination of alprazolam and melatonin. However the difference was statistically significant only between groups receiving combination drugs (5 (26%)) and placebo (0) for only one event" (page 3) "There was no statistical difference between the groups in the number of people reporting occurrence of nausea, vomiting, dizziness, headache, or restlessness (Table 1)" (page 3)
Seet 2015	Melatonin, placebo	No harms reported
Torun 2019	Melatonin, midazolam, placebo	"Although sedation levels were considerably higher in the melatonin group than in the placebo group at 25, 30, and 35 minutes, during this increase, patient RSS scores did not exceed 3 and did not affect cognitive or psychomotor functions. No side effects were encountered" (page 6)
Turkistani 2007	Melatonin, melatonin, no premedication (placebo)	No harms reported

DSST: Digit Symbol Substitution Test.

MAP: mean arterial pressure.

RBB: retrobulbar block.

RSS: Ramsey Sedation Scale.

Fourteen studies did not report on adverse effects (Abbasivash 2019; Acil 2004; Capuzzo 2006; Caumo 2007; Caumo 2009; Dianatkhah 2015; Hoseini 2015; Khare 2018; Marzban 2016; Naguib 2006; Norouzi 2019; Patel 2015; Seet 2015; Turkistani 2007). However, six of these studies examined psychomotor and cognitive function (Acil 2004; Dianatkhah 2015; Khare 2018; Naguib 2006; Norouzi 2019; Patel 2015). Three studies found that benzodiazepines impaired psychomotor and cognitive function compared with melatonin and placebo (Acil 2004; Khare 2018; Patel 2015). Dianatkhah 2015 examined incidences of delirium and found that a smaller proportion experienced delirium in the melatonin group compared with the oxazepam group; however, this difference was not statistically relevant. Norouzi 2019 found that orientation was impaired in the melatonin group compared with the placebo group for one preoperative event, and Naguib 2006 found no difference in orientation scores between melatonin and placebo groups. Seven studies evaluated sedation (Acil 2004; Khare 2018; Marzban 2016; Naguib 2006; Norouzi 2019; Patel 2015; Seet 2015). Three studies found that benzodiazepines produced the highest degree of sedation but melatonin also showed sedative properties (Acil 2004; Khare 2018; Patel 2015). Three studies found no difference in sedation between melatonin and placebo (Naguib 2006; Norouzi 2019; Seet 2015). The remaining study found that melatonin and midazolam produced a higher degree of sedation compared with gabapentin (Marzban 2016). One study explored the severity of nausea and vomiting and found no differences between melatonin, gabapentin, clonidine, and placebo groups (Hoseini 2015). One study reported that mean arterial pressure (MAP) was lower in the melatonin group at all times compared with the placebo group (Norouzi 2019).

Six studies specifically reported that no side effects were observed (Ionescu 2008; Jain 2019; Javaherforooshzadeh 2018; Naguib 1999; Naguib 2000; Torun 2019). However, four of these studies assessed psychomotor and cognitive function as well as sedation (Ionescu 2008; Naguib 1999; Naguib 2000; Torun 2019). Two studies reported amnesia in the benzodiazepine groups (Ionescu 2008; Naguib 1999), three studies found that benzodiazepines impaired psychomotor and cognitive function (Naguib 1999; Naguib 2000; Torun 2019); one of these studies also found that melatonin caused impairment on the Digit‐Symbol Substitution Test (DSST) postoperatively (Naguib 1999), and one study found that DSST scores were lower in the melatonin group compared with the placebo group after administration of medication (Torun 2019). The remaining study found no difference in orientation score (Naguib 2000). Four studies reported that benzodiazepines caused the highest sedation score compared with melatonin and placebo (Ionescu 2008; Naguib 1999; Naguib 2000; Torun 2019); however, melatonin also showed sedative properties.

The remaining seven studies reported adverse effects (Ismail 2009; Khanna 2019; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Pokharel 2014). Cases of headache in the melatonin group were described in three studies (Khezri 2013; Khezri 2013b; Khezri 2016), a case of dizziness in the melatonin group was described in one study (Ismail 2009), and one study described two cases of excessive sleepiness in the melatonin group (Mowafi 2008). One study reported that no difference in occurrence of vomiting, headache, dizziness, and restlessness was seen between groups (Pokharel 2014). Another study reported that side effects, such as headache and dizziness, were similar in melatonin, pregabalin, and alprazolam groups (Khanna 2019). Two studies reported a decrease in MAP after melatonin administration (Ismail 2009; Mowafi 2008). Three studies viewed sedation, and one of these studies found that gabapentin increased sedation (Khezri 2013b), one study found that combination drugs of alprazolam and placebo or alprazolam and melatonin increased levels of sedation (Pokharel 2014), and one study found that melatonin produced the highest degree of sedation compared with alprazolam and pregabalin (Khanna 2019).

Missing information and unspecified issues

In the case of any missing information or unspecified issues, we contacted the study authors to clarify these issues. Details are available in the "notes" in the Characteristics of included studies section. Khanna 2019, however, provided no contact information, and we were unable to contact these authors to clarify unspecified issues.

One study was written in Farsi (Marzban 2016), and we contacted a suitable translator to help with extracting data and assessing bias.

Excluded studies

We excluded 36 studies; for detailed reasons, see Characteristics of excluded studies.

Awaiting classification

Two studies are awaiting classification (IRCT20160430027677N8; CTRI/2017/08/009245); see Characteristics of studies awaiting classification.

Ongoing studies

Ten studies are ongoing (CTRI/2018/02/011895; CTRI/2018/04/012960; CTRI/2018/08/015192; CTRI/2018/08/015537; CTRI/2018/10/015917; CTRI/2019/12/022358; CTRI/2020/02/023330; IRCT20100707004345N6; IRCT20190120042432N1; NCT02386319); see Characteristics of ongoing studies.

Risk of bias in included studies

We assessed each study using the Cochrane risk of bias tool presented in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Overall findings are presented in the 'Risk of bias' graph (Figure 2), which shows the review authors' judgements about each risk of bias item presented as percentages across all included studies; and in the 'Risk of bias' summary (Figure 3), which shows the review authors' judgements about each risk of bias item for each included study. We produced an overall risk of bias judgement for each outcome and for each study.

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Ten studies adequately described the method used to generate the random sequence and conceal the allocation (Capuzzo 2006; Caumo 2007; Caumo 2009; Dianatkhah 2015; Hoseini 2015; Khezri 2013; Khezri 2013b; Khezri 2016; Naguib 2006; Seet 2015), whereas eight studies did not describe it adequately (Acil 2004; Ionescu 2008; Marzban 2016; Naguib 1999; Naguib 2000; Norouzi 2019; Patel 2015; Turkistani 2007). Eight studies described the method used to generate the random sequence adequately but did not describe how the allocation was concealed sufficiently (Abbasivash 2019; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khare 2018; Mowafi 2008; Pokharel 2014; Torun 2019). One study provided no information regarding randomization or allocation concealment (Khanna 2019). No studies had high risk of selection bias .

Blinding

Blinding of participants and personnel was adequately described in 18 studies (Capuzzo 2006; Caumo 2007; Caumo 2009; Dianatkhah 2015; Hoseini 2015; Ionescu 2008; Ismail 2009; Jain 2019; Khezri 2013; Khezri 2013b; Khezri 2016; Naguib 1999; Naguib 2000; Naguib 2006; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019), and 21 studies adequately described blinding of outcome assessors (Abbasivash 2019; Acil 2004; Capuzzo 2006; Caumo 2007; Caumo 2009; Dianatkhah 2015; Hoseini 2015; Ionescu 2008; Ismail 2009; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019). Two studies had high risk of bias in both blinding of participants and personnel and blinding of outcome assessors (Javaherforooshzadeh 2018; Marzban 2016). Seven studies had unclear risk of performance bias (Abbasivash 2019; Acil 2004; Khanna 2019; Khare 2018; Mowafi 2008; Norouzi 2019; Turkistani 2007), and three studies had unclear risk of detection bias (Khanna 2019; Khare 2018; Turkistani 2007).

Incomplete outcome data

All studies included in this review, except two (Caumo 2007; Pokharel 2014), had low risk of attrition bias. Studies adequately accounted for their dropouts and reported reasons for attrition and exclusion.

Selective reporting

We noted unclear risk of reporting bias as no study protocol was available for 19 studies. Seven studies were assessed as having high risk of bias because protocols were not consistent with what was reported in the articles (Dianatkhah 2015; Hoseini 2015; Javaherforooshzadeh 2018; Khezri 2013; Khezri 2013b; Norouzi 2019; Pokharel 2014).

Other potential sources of bias

None of the studies reported receipt of funding from drug manufacturers or agencies with commercial interests. Eleven studies reported funding or grants received from academic, institutional, or departmental sources; however, this was not seen as a basis for bias.

Thirteen studies had low risk of bias for this domain (Abbasivash 2019; Capuzzo 2006; Caumo 2007; Caumo 2009; Hoseini 2015; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Norouzi 2019). Eight studies had unclear risk of bias because of an uneven distribution of sex; however, this uneven distribution was similar in all treatment groups, or there was no mention of sex distribution (Acil 2004; Dianatkhah 2015; Ionescu 2008; Khanna 2019; Khezri 2013b; Pokharel 2014; Seet 2015; Torun 2019). The remaining six studies had high risk of bias based on an uneven distribution of females and males in some treatment groups, an uneven distribution of age, or both (Khare 2018; Khezri 2013; Marzban 2016; Naguib 2006; Patel 2015; Turkistani 2007).

Summary assessments of risk of bias

No study had low risk of bias for all domains (allocation, blinding, incomplete outcome data, selective reporting). Most studies had unclear risk of bias for one or more domains and were regarded as having overall unclear risk of bias (Abbasivash 2019; Acil 2004; Capuzzo 2006; Caumo 2007; Caumo 2009; Ionescu 2008; Ismail 2009; Jain 2019; Khanna 2019; Khare 2018; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Patel 2015; Seet 2015; Torun 2019; Turkistani 2007). Eight studies had high risk of bias in one or more domains, and the overall bias assessment for these studies was that they were at high risk of bias (Dianatkhah 2015; Hoseini 2015; Javaherforooshzadeh 2018; Khezri 2013; Khezri 2013b; Marzban 2016; Norouzi 2019; Pokharel 2014). We did not regard "other sources of bias" as a key domain; therefore, we did not include this domain in our overall risk of bias assessments.

Our overall risk of bias assessment for our primary and secondary outcomes can be seen in the summary of findings tables (summary of findings Table 1; summary of findings Table 2). We performed sensitivity analysis while excluding all studies with overall high risk of bias to assess if inclusion of these studies would alter our conclusions (Table 2).

See: Summary of findings 1 Summary of findings; Summary of findings 2 Summary of findings

Table 2. Sensitivity analysis ‐ primary and secondary outcomes ‐ exclusion of studies with an overall high risk of bias

Outcomes	Statistical method	Studies	Participants	Effect estimate (I²)
Preoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	13	936	‐11.20 (‐13.87 to ‐8.53) (54%)
Final VAS scores	MD (IV, Random, 95% CI)	10	778	‐10.49 (‐13.97 to ‐7.00) (65%)
Change VAS scores	MD (IV, Random, 95% CI)	3	158	‐12.59 (‐16.23 to ‐8.95) (0%)
Postoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	3	236	‐0.79 (‐3.67 to 2.09) (0%)
Final VAS scores	MD (IV, Random, 95% CI)	1	138	0.00 (‐4.94 to 4.94) (‐)
Change VAS scores	MD (IV, Random, 95% CI)	2	98	‐1.20 (‐4.75 to 2.35) (0%)
Preoperative anxiety VAS [mm] ‐ melatonin vs benzodiazepine ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	5	315	0.85 (‐3.01 to 4.72) (66%)
Final VAS scores	MD (IV, Random, 95% CI)	2	133	‐0.95 (‐7.97 to 6.07) (55%)
Change VAS scores	MD (IV, Random, 95% CI)	3	182	2.49 (‐3.68 to 8.66) (79%)
Postoperative anxiety VAS [mm] ‐ melatonin vs benzodiazepine ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	2	122	‐2.02 (‐5.82 to 1.78) (0%)

CI: confidence interval.

IV: inverse variance.

MD: mean difference.

SD: standard deviation.

VAS: visual analogue scale.

Effects of interventions

See "Summary of findings" tables (summary of findings Table 1; summary of findings Table 2), "Additional" tables (Table 2; Table 3; Table 4; Table 5), and "Data and analyses" tables (Data and analyses).

Table 3. Primary and secondary outcomes as reported in the primary study reports

Author, year	Preoperative VAS	Preoperative STAI	Preoperative anxiety HAM‐A	Preoperative BAI	Postoperative VAS	Postoperative STAI	Postoperative HAM‐A	Postoperative BAI
Abbasivash 2019	￬ (90 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Acil 2004	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	￬ (90 min postop) compared to placebo ￬ (90 min postop) compared to midazolam	NM	NM	NM
Capuzzo 2006	→ (90 min after premed) compared to placebo	NM	NM	NM	→ (in recovery room) compared to placebo	NM	NM	NM
Caumo 2007	NM	NM	NM	NM	NM	￬ (6 h postop) compared to placebo	NM	NM
Caumo 2009	NM	NM	NM	NM	NM	￬ (6 h postop) compared to placebo → (6 h postop) compared to clonidine	NM	NM
Dianatkhah 2015	NM	NM	→ (before surgery) compared to oxazepam	NM	NM	NM	￬ (after surgery) compared to oxazepam	NM
Hoseini 2015	NM	→ (120 min after premed) compared to placebo → (120 min after premed) compared to clonidine → (120 min after premed) compared to gabapentin	NM	NM	NM	NM	NM	NM
Ionescu 2008	NM	→ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	￬ (1,6 and 24 h postop) compared to placebo) ￬ (1 h and 24 h postop) compared to midazolam → (6 h postop) compared to midazolam	NM	NM
Ismail 2009	￬ (90 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Jain 2019	￬ (120 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Javaherforooshzadeh 2018	￬ (85 min after premed) compared to placebo → (85 min after premed) compared to gabapentin	NM	NM	NM	￬ (1 h after arrival to recovery room) compared to placebo ￬ (6 h after arrival to recovery room) compared to placebo → (6 h after arrival to recovery room) compared to gabapentin	NM	NM	NM
Khanna 2019	NM	NM	NM	→ (60 min after premed) compared to pregabalin → (60 min after premed) compared to alprazolam	NM	NM	NM	→ (1, 2, 6, 12 hours after surgery) compared to pregabalin → (1, 2, 6, 12 hours after surgery) compared to alprazolam
Khare 2018	￬ (120 min after premed) compared to placebo → (120 min after premed) compared to alprazolam	NM	NM	NM	NM	NM	NM	NM
Khezri 2013	￬ (60 min after premed) compared to placebo	NM	NM	NM	￬ (before discharge from recovery room) compared to placebo	NM	NM	NM
Khezri 2013b	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to gabapentin	NM	NM	NM	￬ (postoperative before discharge) compared to placebo → (postoperative before discharge) compared to gabapentin	NM	NM	NM
Khezri 2016	￬ (20 min after premed) compared to placebo	NM	NM	NM	→ (in recovery room) compared to placebo	NM	NM	NM
Marzban 2016	→ (90 min after premed) compared to placebo/midazolam → (90 min after premed) compared to gabapentin	NM	NM	NM	→ (in recovery room) compared to placebo/midazolam → (in recovery room) compared to gabapentin)	NM	NM	NM
Mowafi 2008	￬ (90 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Naguib 1999	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	→ (90 min postop) compared to placebo → (90 min postop) compared to midazolam	NM	NM	NM
Naguib 2000	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	→ (90 min postop) compared to placebo → (90 min postop) compared to midazolam	NM	NM	NM
Naguib 2006	￬ (50 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Norouzi 2019	￬ (50 min after premed) compared to placebo	NM	NM	NM	￬ (in recovery room) compared to placebo	NM	NM	NM
Patel 2015	￬ (60 to 90 min after premed) compared to placebo → (60 to 90 min after premed) compared to midazolam	NM	NM	NM	NM	NM	NM	NM
Pokharel 2014	→ (60 to 90 min after premed) compared to placebo → (60 to 90 min after premed) compared to alprazolam	NM	NM	NM	NM	NM	NM	NM
Seet 2015	→ (30 to 60 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Torun 2019	￬ (60 min after premed) compared to placebo → (60 min after premed) compared to midazolam	NM	NM	NM	NM	NM	NM	NM
Turkistani 2007	￬ (approximately 100 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM

→: no difference between groups.

￬: lower, difference compared to placebo or midazolam.

BAI: Beck Anxiety Inventory.

HAM‐A: Hamilton Anxiety Rating Scale.

NM: not measured.

STAI: State Trait Anxiety Inventory.

VAS: visual analogue scale.

Table 4. Sensitivity analysis ‐ primary and secondary outcomes

Outcome	Statistical method	Studies	Participants	Effect estimate (I²)
Preoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies not reporting outcome in mean (SD)	MD (IV, Random, 95% CI)	10	621	‐11.90 (‐14.24 to ‐9.55) (34%)
Final VAS scores	MD (IV, Random, 95% CI)	7	463	‐11.34 (‐14.62 to ‐8.06) (55%)
Change VAS scores	MD (IV, Random, 95% CI)	3	158	‐12.59 (‐16.23 to ‐8.95) (0%)
Postoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies not reporting outcome in mean (SD) or reporting SD values of zero	MD (IV, Random, 95% CI)	4	246	‐4.31 (‐7.18 to ‐1.44) (39%)
Final VAS scores	MD (IV, Random, 95% CI)	2	148	‐6.09 (‐8.74 to ‐3.44) (0%)
Change VAS scores	MD (IV, Random, 95% CI)	2	98	‐1.20 (‐4.75 to 2.35) (0%)
Preoperative anxiety VAS [mm] ‐ melatonin vs benzodiazepine ‐ excluding studies not reporting outcome in mean (SD) and an additional study due to lack of blinding	MD (IV, Random, 95% CI)	5	315	0.91 (‐3.02 to 4.38) (67%)
Final VAS scores	MD (IV, Random, 95% CI)	2	133	‐0.95 (‐7.97 to 6.07) (55%)
Change VAS scores	MD (IV, Random, 95% CI)	3	182	2.61 (‐3.68 to 8.90) (80%)

CI: confidence interval.

IV: inverse variance.

MD: mean difference.

SD: standard deviation.

VAS: visual analogue scale.

Table 5. Subgroup analysis ‐ preoperative anxiety ‐ melatonin vs placebo

Outcome	Statistical method	Studies	Participants	Effect estimate (I²)	Test for subgroup differences (P)
Anaesthetic modality	MD (IV, Random, 95% CI)	17	1136	‐12.13 (‐14.00 to ‐10.26) (31%)	0.52
General anaesthesia	MD (IV, Random, 95% CI)	11	796	‐12.25 (‐14.85 to ‐9.64) (51%)
Spinal, regional, or topical anaesthesia	MD (IV, Random, 95% CI)	6	340	‐10.97 (‐13.91 to ‐8.02) (0%)
Age of participants	MD (IV, Random, 95% CI)	17	1184	‐11.78 (‐13.99 to ‐9.85) (50%)	0.16
Age > 60 years	MD (IV, Random, 95% CI)	3	258	‐8.04 (‐13.58 to ‐2.50) (0%)
Age ≤ 60 years	MD (IV, Random, 95% CI)	14	946	‐12.36 (‐14.62 to ‐10.09) (50%)
Dose of melatonin	MD (IV, Random, 95% CI)	17	1216	‐11.71 (‐13.91 to ‐9.50) (52%)	0.54
Melatonin dose ≥ 6 mg	MD (IV, Random, 95% CI)	10	735	‐12.28 (‐15.21 to ‐9.35) (57%)
Melatonin dose < 6 mg	MD (IV, Random, 95% CI)	7	481	‐10.98 (‐13.88 to ‐8.09) (22%)

CI: confidence interval.

IV: inverse variance.

MD: mean difference.

ST: standard deviation.

VAS: visual analogue scale.

We assessed preoperative anxiety between 20 and 120 minutes after premedication to enable data extraction from all studies. If studies applied different doses of melatonin or benzodiazepines, we pooled the reported results.

We did not include Acil 2004 in meta‐analysis because no standard deviation (SD) was reported, and we did not include Khezri 2016 in meta‐analysis because we were unable to extract data from the graph presented in the study.

In total, we excluded nine studies from sensitivity analysis (Capuzzo 2006; Ismail 2009; Khezri 2013; Khezri 2013b; Marzban 2016; Mowafi 2008; Naguib 2006; Pokharel 2014; Turkistani 2007). We made these exclusions either because these studies reported only median (interquartile range (IQR) or range) for visual analogue scale (VAS) data on preoperative anxiety (we contacted the corresponding author of these studies to retrieve more detailed data, but we received no response or we encountered email delivery failure) or, in the case of one study (Marzban 2016), because benzodiazepine and melatonin were distributed at different times, and we, therefore, suspected that the study was not sufficiently blinded.

We assessed postoperative anxiety at two different time points. We chose to group results obtained while in the recovery room, at recovery room discharge, and 90 minutes postoperatively as one group, and six hours postoperatively as another, to explore immediate and delayed anxiety, respectively.

We did not include Acil 2004 in meta‐analysis because no SD was reported, and we did not include two other studies in meta‐analysis because they used the Beck Anxiety Inventory (BAI) and a modified version of State‐Trait Anxiety Inventory (STAI‐S) (Ionescu 2008; Khanna 2019), respectively, which were not comparable to the VAS. We performed sensitivity analysis after exclusion of three studies (Capuzzo 2006; Khezri 2013; Marzban 2016). We excluded these because they reported medians of zero, thereby violating the assumption of symmetry.

Melatonin versus placebo

Preoperative anxiety

The meta‐analysis comparing melatonin with placebo showed a reduction in preoperative anxiety measured by a VAS (mean difference (MD) ‐11.69, 95% confidence interval (CI) ‐13.80 to ‐9.59; P < 0.00001, I² = 49%; 18 studies, 1264 participants; moderate‐certainty evidence; Analysis 1.1; Figure 4). The 95% CI is relatively narrow, making us certain that melatonin reduces preoperative anxiety compared with placebo. When performing a sensitivity analysis of only studies that reported the outcome using the mean (SD), we showed a reduction in preoperative anxiety (MD ‐11.90, 95% CI ‐14.24 to ‐9.55; P < 0.00001, I² = 34%; 10 studies, 671 participants; Table 4).

Figure 4

Forest plot of comparison: 1 Melatonin versus placebo, outcome: 1.1 Preoperative anxiety (VAS) (mm) with subgroup 1.1.1 Final VAS scores and subgroup 1.1.2 Change VAS scores.

Based on individual primary study results (Table 3), 17 studies (Abbasivash 2019; Acil 2004; Ismail 2009; Jain 2019; Javaherforooshzadeh 2018; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Norouzi 2019; Patel 2015; Torun 2019; Turkistani 2007) reported a reduction in preoperative anxiety measured by VAS when comparing melatonin with placebo. When results (median (IQR) to mean (SD)) from one of these studies were extracted and converted (Ismail 2009), the effect of this study was lost because the 95% CI included the value of zero. Capuzzo 2006 showed no difference between melatonin and placebo in preoperative anxiety measured by a numerical rating scale (NRS). Two studies had wide 95% CIs and included the value of zero; hence these authors reported no difference between melatonin and placebo in preoperative anxiety measured by VAS (Pokharel 2014; Seet 2015). Two studies not included in the meta‐analysis showed no difference between melatonin and placebo in preoperative anxiety measured by a modified six‐item STAI (STAI‐S) or a standard STAI, respectively (Hoseini 2015; Ionescu 2008).

We prepared a funnel plot for our primary meta‐analysis (Figure 5; Analysis 1.1), which appeared symmetrical, indicating that publication bias was unlikely.

Figure 5

Funnel plot of comparison: 1 Melatonin versus placebo, outcome: 1.1 Preoperative anxiety (VAS) [mm].

We performed a separate sensitivity analysis when we excluded all studies that were assessed as having an overall high risk of bias. We excluded from meta‐analysis four studies with an overall high risk of bias (Javaherforooshzadeh 2018; Khezri 2013; Norouzi 2019; Pokharel 2014). We found that melatonin reduced preoperative anxiety compared to placebo (MD ‐10.20, 95% CI ‐13.87 to ‐8.53; P < 0.00001, I² = 54%; 13 studies, 936 participants; Table 2).

To explore heterogeneity, we performed subgroup analysis based on anaesthetic modality, participants' age, and the dose of melatonin administered. It is not recommended to perform subgroup analysis if fewer than 10 studies are identified (Higgins 2019); hence, we performed subgroup analysis only on our primary outcome: preoperative anxiety melatonin versus placebo.

Anaesthetic modality

Eleven studies used general anaesthesia and showed a reduction in preoperative anxiety (MD ‐12.25, 95% CI ‐14.85 to ‐9.64; P < 0.00001, I² = 51%; 11 studies, 796 participants; Table 5). Six studies used topical, regional, or spinal anaesthesia and showed a reduction in preoperative anxiety (MD ‐10.97, 95% CI ‐13.91 to ‐8.02; P < 0.00001, I² = 0%; 6 studies, 340 participants; Table 5). Capuzzo 2006 used general and spinal anaesthesia but did not provide information regarding anxiety measurements for each group; hence, this study was not included in the analysis. The test for subgroup differences indicated no statistically significant subgroup effect (P = 0.52; Table 5). It does not appear that anaesthetic modality alters effects of an intervention; however, fewer trials and participants contributed data to one subgroup (topical, regional, or spinal anaesthesia), meaning that the analysis might not be able to detect subgroup differences.

Age of participants (≤ 60 or > 60 years)

Three studies included only participants older than 60 years and showed a reduction in preoperative anxiety (MD ‐8.04, 95% CI ‐13.58 to ‐2.50; P = 0.004, I² = 0%; 3 studies, 258 participants; Table 5). Ismail 2009 included only patients over the age of 60, and Capuzzo 2006 included only patients over the age of 65. The remaining study included patients between 35 and 85 years of age (Khezri 2013b), but mean age was above 70 years, which is why this study was included in the > 60 years of age group. Khezri 2013 included patients 25 to 80 years of age. Mean age in the melatonin group was 63.50 ± 15.28; we decided to not include this study in subgroup analysis because it did not fit into either group (≤ 60 or > 60 years of age). Two studies also included patients over 60 years of age (Pokharel 2014; Seet 2015), but the mean age was way below 60 years, which is why these studies were included in the ≤ 60 years group.

Fourteen studies included patients younger than 60 years and also showed a reduction in preoperative anxiety when comparing melatonin to placebo (MD ‐12.36, 95% CI ‐14.62 to ‐10.09; P < 0.02, I² = 50%; 14 studies, 946 participants; Table 5). The test for subgroup differences did not reach statistical significance (P = 0.16; Table 5).

Dose of melatonin (< 6 mg or ≥ 6 mg)

We decided to divide studies into two groups depending on the dose of melatonin administered (< 6 mg or ≥ 6 mg).

Ten studies administered melatonin doses ≥ 6 mg and showed a reduction in preoperative anxiety compared to placebo (MD ‐12.28, 95% CI ‐15.21 to ‐9.35; P < 0.00001, I² = 57%; 10 studies, 735 participants; Table 5). Naguib 2006 administered 0.2 mg/kg melatonin, and Patel 2015 administered 0.4 mg/kg melatonin, but when the dose of melatonin was calculated based on mean weight in the melatonin groups, doses were above 6 mg of melatonin, which is why these studies where included in the ≥ 6 mg group. Torun 2019 administered melatonin at a dose of 0.4 mg/kg; however, this study provided no information regarding the mean weight of participants. We assumed that doses given were above 6 mg if participants weighed between 40 and 80 kg, which is why the study was included in the ≥ 6 mg group. Naguib 2000 administered different doses of melatonin (0.05, 0.1, 0.2 mg/kg); however, this study reported outcomes graphically, and it is not possible to distinguish the groups from one another, which is why this study was not included in subgroup analysis.

Seven studies administered < 6 mg of melatonin and showed a reduction in anxiety compared with placebo (MD ‐10.98, 95% CI ‐13.88 to ‐8.09; P < 0.00001, I² = 22%; 7 studies, 481 participants; Table 5).

The test for subgroup differences did not reach statistical significance (P = 0.16; Table 5).

Postoperative anxiety

Immediate postoperative anxiety (recovery room discharge to 90 minutes postoperatively)

The meta‐analysis showed a difference in postoperative anxiety between the two groups (MD ‐5.04, 95% CI ‐9.52 to ‐0.55; P = 0.03, I² = 89; 7 studies, 524 participants; low‐certainty evidence; Analysis 1.2); however, the 95% confidence interval was wide, which limited the certainty of evidence. When a sensitivity analysis was performed while excluding studies that did not report an SD or reported an SD value of 0, there was still a difference; however, the 95% confidence interval was wide (MD ‐4.31, 95% CI ‐7.18 to ‐1.44; P = 0.003, I² = 39; 4 studies, 246 participants; Table 4).

We performed an additional sensitivity analysis from which we excluded studies with an overall high risk of bias (Javaherforooshzadeh 2018; Khezri 2013; Khezri 2013b; Norouzi 2019; Pokharel 2014), and we found no difference in postoperative anxiety (MD ‐0.79, 95% CI ‐3.67 to 2.09; P = 0.70, I² = 0; 3 studies, 236 participants; Table 2).

Delayed postoperative anxiety (6 hours postoperatively)

The meta‐analysis (excluding two studies ‐ Ionescu 2008 and Javaherforooshzadeh 2018) showed a reduction in postoperative anxiety (MD ‐5.31, 95% CI ‐8.78 to ‐1.84; P = 0.003, I² = 0; 2 studies, 73 participants; low‐certainty evidence; Analysis 1.3). Ionescu 2008 also showed a reduction in postoperative anxiety measured six hours postoperatively in the melatonin group compared with the placebo group. Javaherforooshzadeh 2018 showed a reduction in postoperative anxiety measured six hours postoperatively.

Melatonin versus benzodiazepine

Preoperative anxiety

The meta‐analysis showed no difference in preoperative anxiety between the two groups (MD 0.78, 95% CI ‐2.02 to 3.58; P = 0.59, I² = 55%; 7 studies, 409 participants; moderate‐certainty evidence; Analysis 2.1; Figure 6). When performing sensitivity analysis excluding all studies reporting outcomes using IQR or range, while excluding one additional study because the study was not sufficiently blinded (Marzban 2016), we also found no difference in preoperative anxiety between the two groups (MD 0.91, 95% CI ‐3.02 to 4.83; P = 0.65, I² = 67%; 5 studies, 350 participants; Table 4).

Figure 6

Forest plot of comparison: 2 Melatonin versus benzodiazepine ‐ preoperative anxiety, outcome: 2.1 Preoperative anxiety (VAS) [mm].

Based on individual primary study results (Table 3), none of the studies showed a difference between melatonin and benzodiazepine. Khare 2018 found that melatonin reduced anxiety more than alprazolam, but we could not reproduce this result from data extracted from the study. Marzban 2016 also stated that melatonin reduced anxiety more than midazolam; however, we could not reproduce this result. Khanna 2019 reported no difference in preoperative anxiety measured on the BAI when comparing melatonin with alprazolam.

We performed an additional sensitivity analysis from which we excluded all studies with an overall high risk of bias (Marzban 2016; Pokharel 2014). We found that melatonin did not decrease preoperative anxiety compared with benzodiazepines (MD 0.85, 95% CI ‐3.01 to 4.72; P = 0.67, I² = 66%; 5 studies, 315 participants; Table 2).

Postoperative anxiety

Immediate postoperative anxiety (recovery room discharge to 90 minutes postoperatively)

The meta‐analysis showed no difference in postoperative anxiety between the two groups (MD ‐2.12, 95% CI ‐4.61 to 0.36; P = 0.09, I² = 0%; 3 studies, 176 participants; very low‐certainty evidence; Analysis 2.2).

Ionescu 2008, using a modified version of STAI (the State scale ‐ S‐STAI) (which was not included in the meta‐analysis because the scale used was not comparable to the VAS), reported a difference between melatonin and benzodiazepine one hour postoperatively. Khanna 2019 using the BAI showed no difference between alprazolam and melatonin one hour postoperatively.

Sensitivity analysis from which all studies with an overall high risk of bias were excluded ‐ Marzban 2016 ‐ showed no difference in postoperative anxiety (MD ‐2.02, 95% CI ‐5.82 to 1.78; P = 0.30, I² = 0%; 2 studies, 122 participants; Table 2).

Delayed postoperative anxiety (6 hours postoperatively)

Ionescu 2008 measured postoperative anxiety using a modified version of S‐STAI six hours postoperatively and showed no difference between the two groups (Table 3). Dianatkhah 2015, using the Hamilton Anxiety Rating Scale (HAM‐A) to measure anxiety after surgery, showed a difference between melatonin and oxazepam. Khanna 2019, using the BAI, showed no difference between alprazolam and melatonin six hours postoperatively.

Discussion

Summary of main results

This systematic review identified 27 randomized controlled trials (RCTs) assessing melatonin for treating preoperative anxiety, postoperative anxiety, or both. Twenty‐four of the 27 studies compared melatonin with placebo, and 11 studies compared melatonin with a benzodiazepine.

Melatonin versus placebo

We found moderate‐certainty evidence, based on meta‐analysis of 18 studies, to show that melatonin likely reduces preoperative anxiety compared to placebo. The previous version of this review found that melatonin decreased preoperative anxiety compared to placebo by 13 points on a visual analogue scale (VAS) (Hansen 2015). In this review update, we found a slightly smaller decrease of 12 points on a VAS.

Results of individual primary studies indicate that 16 of the 21 studies that assessed effects of melatonin on preoperative anxiety showed a reduction compared to placebo. Six studies did not show any differences between melatonin and placebo.

Fifteen studies assessed effects of melatonin on postoperative anxiety. Of these, 11 studies compared melatonin with placebo, and seven studies compared melatonin with a benzodiazepine.

We found low‐certainty evidence, based on meta‐analysis of seven studies, suggesting that melatonin may reduce immediate postoperative anxiety compared to placebo. However, the result was below our minimum clinically important difference. Low‐certainty evidence, based on meta‐analysis of two studies, suggests that melatonin may reduce delayed postoperative anxiety compared to placebo.

The previous review found no evidence of a decrease in postoperative anxiety when melatonin was compared to placebo (Hansen 2015). In contrast, in this review update, we found a decrease in immediate and delayed postoperative anxiety of 5 points on a VAS, but evidence was uncertain in the analysis, and the result was below our estimated minimum clinically important difference.

Melatonin versus benzodiazepine

We found moderate‐certainty evidence, based on meta‐analysis of seven studies, to show that melatonin may result in no difference in preoperative anxiety compared to benzodiazepines. None of the 11 studies that assessed effects of melatonin compared with a benzodiazepine on preoperative anxiety showed a difference.

Evidence on the effect of melatonin on immediate postoperative anxiety compared to benzodiazepines was very uncertain. No difference was seen in the meta‐analysis of three studies.

Overall completeness and applicability of evidence

A minimum difference in preoperative and postoperative anxiety VAS score has not been fully established. However, with regard to acute pain VAS scores, it has been estimated that 9 to 14 mm on a 0 to 100 mm VAS is the minimum clinically significant difference (Kelly 1998; Kelly 2001). Thus, main results from the meta‐analyses regarding preoperative anxiety when melatonin was compared with placebo (12 mm for primary meta‐analysis and 12 mm for the sensitivity analysis) could be considered clinically relevant.

Whether the anxiolytic effect of melatonin can be applied to all surgical patients remains unclear, as many factors influence the risk of preoperative anxiety. Among these are age, sex, type of surgery, type of anaesthesia, and cultural and religious differences (Caumo 2001a; Domar 1989; Kindler 2000; Lovering 2006). Younger age and female sex have been shown to be independent risk factors for preoperative anxiety (Caumo 2001a; Domar 1989; Kindler 2000). This may influence the external validity of our results, as two of the studies in this review included only patients older than 60 years (Capuzzo 2006; Ismail 2009), and five of the studies in this review included only women (Caumo 2007; Caumo 2009; Khezri 2013; Naguib 1999; Naguib 2000).

Conflicting opinions can be found in the literature regarding preoperative anxiety and type of surgery. Caumo 2001a showed that medium or major surgery (classified according to blood loss, degree of pain, invasiveness, degree of monitoring required, and length of stay in hospital due to the surgical procedure) leads to higher preoperative anxiety. In contrast, Domar 1989 showed no difference regarding type of surgery and preoperative anxiety. The type of anaesthesia used ‐ regional versus general ‐ can also influence anxiety levels in different directions (Haugen 2009; Mitchell 2008; Mitchell 2010; Mitchell 2012). As far as general anaesthesia is concerned, many patients fear waking up during surgery or not waking up after surgery (Mitchell 2010; Ramsay 1972).

Furthermore, for the most part, general anaesthesia is used for major surgery, which in itself may influence the risk of anxiety (Caumo 2001a). As far as regional anaesthesia is concerned, patients experience the anxiety of being awake during the procedure, involving all the noises, lights, and pain associated with this (Mitchell 2008). The studies included in this review vary from minor to major surgery, performed with general, regional, or topical anaesthesia. We conducted three subgroup analyses exploring effects of anaesthetic modality, age of participants, and dose of melatonin on heterogeneity for our primary analysis: preoperative anxiety melatonin versus placebo. When anaesthetic modality was assessed, statistical heterogeneity presented as an I² value was close to equal to our primary analysis in the general anaesthesia group (I² = 51% in subgroup and I² = 49% in primary analysis). The effect estimate was close to our primary analysis as well (‐12.25 in subgroup analysis and ‐11.69 in primary analysis). In the other group (topical, local, or spinal anaesthesia), statistical heterogeneity totally disappeared (I² = 0%), but the effect estimate was close to our primary analysis (‐10.97). Testing for subgroup differences indicated no statistically significant subgroup effect (P = 0.52). However, there were far more studies in one of the subgroups, which is why the analysis might not be able to detect subgroup differences. It appears that anaesthetic modality does not explain the heterogeneity in our primary analysis.

In our subgroup analysis exploring the effect of participant age, it appears that melatonin has a lesser effect in an older population (> 60 years). Statistical heterogeneity disappeared in the group > 60 years of age (I² = 0). However, this subgroup included only three studies, so this conclusion cannot be made with certainty. These subgroup differences did not reach statistical significance; however, because one subgroup contained only three studies, the analysis might not be able to detect subgroup differences.

When the effect of the dose of melatonin administered was explored, statistical heterogeneity was still present in both the ≥ 6 mg group and the < 6 mg group (I² = 57% and I² = 22%, respectively). Effect estimates in both subgroups were close to our primary analysis (‐12.28 in ≥ 6 mg group and ‐10.98 in < 6 mg group). Testing of subgroup differences did not reach statistical significance. The dose of melatonin did not explain heterogeneity in our primary analysis.

Cultural and religious differences have been shown to influence the actual perception of anxiety (Lovering 2006). Sixteen of the 25 studies were carried out in Middle Eastern countries (Saudi Arabia, Turkey, and Iran), one in Italy, one in Romania, two in Brazil, four in India, one in Singapore, and one in Nepal. This could lead to an imbalance and could influence external validity as some cultures are over‐represented and others are under‐represented.

In 21 of the 27 included studies, the method used to measure preoperative anxiety was the VAS (in one study, a numerical rating scale (NRS) (Capuzzo 2006)), and only three studies used the State‐Trait Anxiety Inventory (STAI). The VAS and the STAI as anxiety‐measuring techniques were used to measure preoperative and postoperative anxiety and have been validated in a surgical population (Kindler 2000). To date, the gold standard for anxiety evaluation is the STAI, but its architecture of 20 to 40 multiple choice questions for anxiety alone limits its use as a bedside instrument. In contrast, the VAS allows patients to easily indicate their degree of preoperative or postoperative anxiety by simply marking a point on a horizontal line. The simple VAS method is very easily applied for both doctor and patient and has proved a useful and valid measure of preoperative anxiety (Kindler 2000; Millar 1995).

Of the 27 studies in our review, nine studies administered melatonin sublingually, and 18 administered it orally as tablets. With sublingual administration (comparable to intravenous administration), first‐pass metabolism is bypassed, and this leads to variation in bioavailability compared to oral administration (Brzezinski 1997). Due to heterogeneity of the method of administration used by included studies, additional studies are required to perform relevant subgroup analyses.

Of the 15 studies assessing postoperative anxiety, only five studies measured anxiety six hours postoperatively, whereas the remaining studies mainly assessed anxiety in the immediate postoperative period. Hence, more studies are warranted to clearly determine effects of melatonin on postoperative anxiety in the postoperative period.

The half‐life of melatonin was examined in a systematic review (Harpsoe 2015). This review included 22 studies that explored the pharmacokinetics of melatonin and concluded that the half‐life of melatonin was approximately 45 minutes when melatonin was administered orally or intravenously. Heizmann 1983 examined the pharmacokinetics of midazolam in six healthy males. These investigators found that the half‐life was 2.3 hours when given intravenously and was almost the same when given orally. This might have an effect on postoperative anxiety in that melatonin, benzodiazepines, and placebo were administered preoperatively. We found no difference in postoperative anxiety when melatonin was compared with benzodiazepines, whereas we found a small difference when melatonin was compared with placebo; however, this difference might not be clinically relevant. In future studies, melatonin could be administered in the immediate postoperative period to better determine the effect of melatonin on postoperative anxiety.

Quality of the evidence

Almost half of the included studies had low risk of selection bias, and at least 70% had low risk of attrition, performance, and detection bias. Most of the included studies had unclear risk of reporting bias; however, some had high risk because the protocol and the manuscript were not identical.

We performed sensitivity analyses for all primary and secondary outcomes when we excluded studies with overall high risk of bias to test the robustness of the estimated effect. We did not find that inclusion of studies with high risk of bias altered our conclusions, except for postoperative anxiety, when melatonin was compared with placebo. When studies with overall high risk of bias were excluded from this analysis, the effect was lost.

The estimate of effect for the primary outcome (preoperative anxiety) was judged as having evidence of moderate certainty, based on GRADE assessment, for the comparison of melatonin versus placebo and melatonin versus benzodiazepines. Evidence was downgraded by one level for our primary outcome preoperative anxiety for the comparison of melatonin versus placebo due to substantial heterogeneity and overall high risk of bias. However, sensitivity analysis from which all studies with overall high risk of bias were excluded showed a similar result as our main meta‐analysis. Therefore, we chose to downgrade the certainty of evidence by only one level, because we concluded that inclusion of studies with high risk of bias did not alter conclusions. Evidence was downgraded by one level for the comparison of melatonin versus benzodiazepines for preoperative anxiety due to substantial heterogeneity and overall high risk of bias. The sensitivity analysis exploring whether studies with high risk of bias would alter the conclusions showed a similar result as the main analysis for this comparison. For this reason, we decided to downgrade by only one level. The estimate of effect for the secondary outcome (postoperative anxiety) was judged as having evidence of low certainty for melatonin versus placebo and for melatonin versus midazolam. This was due to large heterogeneity, small numbers of participants, and overall high risk of bias.

When exploring heterogeneity in our review, we found I² of 49% for our main analysis (Figure 4). We considered this to be moderate and not a substantial issue; hence, we performed only random‐effects model meta‐analyses. In the process of analysing the data, we also performed a sensitivity analysis (Table 4) by excluding studies that reported only median (interquartile range (IQR) or range) for VAS data on preoperative anxiety. The I² value for this sensitivity analysis was 34%. The statistical heterogeneity seen in our primary analysis is suspected to be due to clinical diversity. Studies varied in study design, population, anaesthesia, and type of surgery. We performed subgroup analysis by which we examined the effects of anaesthetic modality, dose of melatonin, and age of participants (Table 5); however, none of the subgroup analyses reached statistical significance upon testing for subgroup differences. When the different subgroups are examined, it appears that age of participants might explain some of the statistical heterogeneity found in our primary analysis. It appears that older age is an effect modifier; however, some unexplained heterogeneity is still present in our primary analysis, which we have not been able to explain through subgroup analysis.

Potential biases in the review process

To obtain additional information, we contacted the authors of 20 of the included studies. Four authors answered sufficiently (Capuzzo 2006; Dianatkhah 2015; Jain 2019; Marzban 2016), one author answered insufficiently (Ionescu 2008), and information regarding the remaining 15 studies was not obtained as study authors did not reply, despite repeated attempts. Khanna 2019 did not provide any contact information; hence, we were unable to contact study authors. This might have introduced a potential source of bias, in that some of these studies were excluded from meta‐analysis or sensitivity analysis and therefore if included, could have altered the results.

Marzban 2016 was only single‐blinded, which creates possible bias, but the remaining studies were double‐blinded.

A single review author (BKM) performed data extraction in duplicate. Even though data extraction was done in duplicate, this could present a potential source of bias due to the possibility of duplicating the single review author's biases.

Ionescu 2008 used a short version of the STAI. The six‐item STAI has been validated previously (Marteau 1992); however, we were unable to retrieve a conversion key for this questionnaire and therefore decided not to include this study in the meta‐analysis because we were unable to pool results across the two scales (STAI and six‐item STAI).

Several studies did not report on adverse events; therefore it is not possible to conclude with certainty, from the data on adverse effects collected in this review, that melatonin is better tolerated than benzodiazepines. However, several studies reported that benzodiazepines caused impairment of psychomotor and cognitive functions, whereas melatonin, for the most part, did not, or did so to a lesser extent than benzodiazepines. So, it appears that melatonin is tolerated better than benzodiazepines.

When studies presented data only as graphs or figures, we read values directly from the graphs. This could have introduced minor errors because it is not particularly precise compared to other software methods. This could have introduced uncertainty to the exact results.

Agreements and disagreements with other studies or reviews

Two other systematic reviews have investigated the effect of melatonin as an anxiolytic in the perioperative period (Andersen 2014a; Yousaf 2010). Yousaf 2010 identified 10 studies investigating perioperative anxiety. These studies are also included in our review, together with 17 others (Abbasivash 2019; Dianatkhah 2015; Hoseini 2015; Jain 2019; Javaherforooshzadeh 2018; Khanna 2019; Khare 2018; Khezri 2013; Khezri 2013b; Khezri 2016; Marzban 2016; Norouzi 2019; Patel 2015; Pokharel 2014; Seet 2015; Torun 2019; Turkistani 2007). The study by Turkistani et al. was not included due to lack of a pre‐intervention anxiety score (exclusion criterion). The rest of the studies were not included because they were published after the search date. We chose not to believe that lack of this specific assessment (pre‐intervention anxiety score) should be considered a potential confounder due to the randomized design of all included studies; hence, this was not an exclusion criterion in our review. Andersen 2014a identified 14 studies investigating perioperative anxiety. Twelve of the studies included in that review are also included in our review (Acil 2004; Capuzzo 2006; Caumo 2007; Caumo 2009; Ionescu 2008; Ismail 2009; Khezri 2013; Mowafi 2008; Naguib 1999; Naguib 2000; Naguib 2006; Turkistani 2007); two studies examined anxiety in children and are not included in our review.

The findings of the reviews mentioned above ‐ Yousaf 2010 and Andersen 2014a ‐ and the findings of ours agree that melatonin premedication is effective in ameliorating perioperative anxiety. However, in Yousaf 2010, no quantitative analyses were undertaken. This was mainly explained by the fact that retrieved data were presented in a graphical fashion or as median and range. Furthermore, the review authors found the heterogeneity of the studies too extensive to synthesize the data quantitatively (Yousaf 2010). Andersen 2014a performed meta‐analysis but explained that the analysis was very heterogeneous, which was partially resolved by the exclusion of studies in which median (IQR or range) was converted to mean (SD). Andersen 2014a included Acil 2004 in their meta‐analysis; we chose not to include this one because the study did not report an SD.

Due to the review authors' assessment of heterogeneity (Yousaf 2010), they concluded that future studies should focus on investigating effects on more varied surgical populations and the optimal dosing regimen.

Figure 1

Study flow diagram.

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 4

Forest plot of comparison: 1 Melatonin versus placebo, outcome: 1.1 Preoperative anxiety (VAS) (mm) with subgroup 1.1.1 Final VAS scores and subgroup 1.1.2 Change VAS scores.

Figure 5

Funnel plot of comparison: 1 Melatonin versus placebo, outcome: 1.1 Preoperative anxiety (VAS) [mm].

Figure 6

Forest plot of comparison: 2 Melatonin versus benzodiazepine ‐ preoperative anxiety, outcome: 2.1 Preoperative anxiety (VAS) [mm].

Analysis 1.1

Comparison 1: Melatonin versus placebo, Outcome 1: Preoperative anxiety (VAS)

Analysis 1.2

Comparison 1: Melatonin versus placebo, Outcome 2: Postoperative anxiety (VAS) [mm]

Analysis 1.3

Comparison 1: Melatonin versus placebo, Outcome 3: Postoperative anxiety (STAI)

Analysis 2.1

Comparison 2: Melatonin versus benzodiazepine, Outcome 1: Preoperative anxiety (VAS) [mm]

Analysis 2.2

Comparison 2: Melatonin versus benzodiazepine, Outcome 2: Postoperative anxiety (VAS) [mm]

Summary of findings 1. Summary of findings

Outcomes	*Illustrative comparative risks (95% CI)**		No. of participants (studies)	Quality of the evidence (GRADE)	Comments
Melatonin compared with placebo
Patient or population: patents undergoing elective surgery Setting: hospital Intervention: melatonin Comparison: placebo
	Assumed risk	Corresponding risk
	Placebo	Melatonin
Preoperative anxiety (VAS) VAS (0 to 100 mm) measured approximately 50 to 120 minutes after premedication 0: no anxiety 100: maximum anxiety possible	Mean VAS total ranged across control groups from 22.7 to 66.5, and mean change in VAS ranged across control groups from 4 to ‐22	Mean VAS in intervention groups was 11.69 lower (13.80 lower to 9.59 lower) Lower score indicated less preoperative anxiety compared to placebo	1264 (18 studies)	⊕⊕⊕⊖ Moderate^a	Melatonin most likely decreases preoperative anxiety compared with placebo
Preoperative anxiety (STAI) STAI (20 to 80) measured approximately 120 minutes after premedication 20: no anxiety 80: maximum anxiety possible	Mean STAI in control group measured just before entrance to the operating room was 39.73	Mean STAI in intervention group measured just before entrance to the operating room was 41.18	44 (1 study)	⊕⊖⊖⊖ Very low^b	Because only 1 study examined preoperative anxiety using an STAI, no meta‐analysis was performed
Preoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured approximately 90 minutes after premedication 6: no anxiety 24: maximum anxiety possible	Mean STAI in control group measured at patient arrival to the operating room was 13.5	Mean STAI in intervention group measured at patient arrival to the operating room was 11.6	36 (1 study)	⊕⊕⊖⊖ Low^c	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
Immediate postoperative anxiety (VAS) VAS (0 to 100 mm) measured after surgery, in recovery, or at discharge from recovery room 0: no anxiety 100: maximum anxiety possible	Mean VAS total ranged across control groups from0 to 48, and mean change in VAS ranged across control groups from ‐4.7 to ‐6.5	Mean VAS in intervention groups was 5.04 lower (9.52 lower to 0.55 lower) Lower score indicated less postoperative anxiety compared to placebo	524 (7 studies)	⊕⊕⊝⊝ Low^d	Melatonin may have an effect on postoperative anxiety compared with placebo; however, this effect was below the minimum clinical effect
Delayed postoperative anxiety (STAI) STAI (20 to 80) measured 6 hours after surgery 20: no anxiety 80: maximum anxiety possible	Mean STAI ranged across control groups from 42.2 to 42.5	Mean STAI in intervention groups was 5.31 lower (8.78 lower to 1.84 lower) Lower score indicated less postoperative anxiety compared to placebo	73 (2 studies)	⊕⊕⊝⊝ Low^e	Melatonin may have an effect on postoperative anxiety compared with placebo; however, this effect was below the minimum clinical effect
Postoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured 1 hour and 6 hours after surgery 6: no anxiety 24: maximum anxiety possible	Mean STAI value in control group 1 hour after surgery was 11 Mean STAI value in control group 6 hours after surgery was 11.6	Mean STAI value in melatonin group 1 hour after surgery was 8 Mean STAI value in melatonin group 6 hours after surgery was 7.9	36 (1 study)	⊕⊕⊝⊝ Low^f	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; STAI: State‐Trait Anxiety Inventory; VAS: visual analogue scale.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
^aThe certainty of evidence was downgraded by one level due to unclear overall risk of bias and the presence of substantial heterogeneity. We chose not to downgrade by two levels because sensitivity analysis excluding all studies with high risk of bias showed a similar effect estimate; we therefore concluded that high risk of bias in the included studies did not affect conclusions. ^bWe chose to downgrade the evidence by three levels due to imprecision and high risk of bias: only one study with 44 participants examined preoperative anxiety using STAI; this study also had overall high risk of bias. ^cWe chose to downgrade the evidence by two levels due to imprecision: only one study with 36 participants examined preoperative anxiety using a six‐item STAI. ^dThe certainty of evidence was downgraded by two levels due to large heterogeneity of the studies (I² = 89%) and overall high risk of bias. Several of the included studies had overall high risk of bias, making the overall risk of bias for the outcome high. When all studies with high risk of bias were excluded from the sensitivity analysis, the effect of the intervention was lost, which is why we suspect that inclusion of studies with overall high risk of bias may alter conclusions. ^eThe certainty of evidence was downgraded by two levels due to the small numbers of participants. ^fThe certainty of evidence was downgraded by two levels due to the small numbers of participants.

Summary of findings 1. Summary of findings

Summary of findings 2. Summary of findings

Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
Melatonin compared with benzodiazepine
Patient or population: patients undergoing elective surgery Setting: hospital Intervention: melatonin Comparison: benzodiazepine (midazolam, alprazolam)
	Assumed risk	Corresponding risk
	Benzodiazepine	Melatonin
Preoperative anxiety (VAS) VAS (0 to 100 mm) measured approximately 90 minutes after premedication 0: no anxiety 100: maximum anxiety possible	Mean VAS total ranged across control groups from 3.6 to 16.7, and mean change in VAS ranged across control groups from ‐7.7 to ‐50	Mean VAS in intervention groups was 0.78 higher (2.02 lower to 3.58 higher) A higher score indicated greater preoperative anxiety compared to benzodiazepine		409 (7 studies)	⊕⊕⊕⊝ Moderate^a	Melatonin most likely has little or no effect on preoperative anxiety compared with benzodiazepines
Preoperative anxiety (STAI) STAI (20 to 80) 20: no anxiety 80: maximum anxiety possible	No studies available	No studies available	‐	‐	‐	‐
Preoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured approximately 90 minutes after premedication 6: no anxiety 24: maximum anxiety possible	Mean STAI in benzodiazepine group measured at patient arrival to the operating room was 10.5	Mean STAI in melatonin group measured at patient arrival to the operating room was 11.6		35 (1 study)	⊕⊕⊖⊖ Low^b	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
Immediate postoperative anxiety (VAS) VAS (0 to 100 mm) measured approximately 90 minutes after surgery or in recovery room 0: no anxiety 100: maximum anxiety possible	Mean VAS in control group was 7.4 and mean change in VAS ranged across control groups from ‐5.3 to ‐6.4	Mean VAS in intervention groups was 2.12 lower (4.61 lower to 0.36 higher) Lower score indicated less postoperative anxiety compared to benzodiazepine		176 (3 studies)	⊕⊕⊝⊝ Low^c	Melatonin had little or no effect on postoperative anxiety compared with benzodiazepines
Postoperative anxiety (6‐item STAI) STAI 6‐item (6 to 24) measured 1 hour and 6 hours after surgery 6: no anxiety 24: maximum anxiety possible	Mean STAI value in benzodiazepine group 1 hour after surgery was 10.4 Mean STAI value in benzodiazepine group 6 hours after surgery was 9.3	Mean STAI value in melatonin group 1 hour after surgery was 8 Mean STAI value in melatonin group 6 hours after surgery was 7.9		35 (1 study)	⊕⊕⊖⊖ Low^d	Because only 1 study examined preoperative anxiety using a 6‐item STAI, no meta‐analysis was performed
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; STAI: State‐Trait Anxiety Inventory; VAS: visual analogue scale.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
^aThe certainty of evidence was downgraded by one level due to high overall risk of bias for the outcome and substantial heterogeneity. We decided not to downgrade the evidence by two levels because sensitivity analysis excluding all studies with overall high risk of bias showed a similar effect estimate; we therefore concluded that high risk of bias in the included studies did not alter conclusions. ^bWe chose to downgrade the evidence by two levels because only one study with 35 participants examined preoperative anxiety using a six‐item STAI. ^cThe certainty of evidence was downgraded by two levels due to the small numbers of participants. One study had overall high risk of bias, but sensitivity analysis excluding this study showed a similar effect, which is why we chose not to downgrade by another level. ^dThe certainty of evidence was downgraded by two levels due to the small numbers of participants.

Summary of findings 2. Summary of findings

Table 1. Harms reported in primary study reports

Author, year	Comparison	Harms
Abbasivash 2019	Melatonin, placebo	No harms reported
Acil 2004	Melatonin, midazolam, placebo	"The melatonin group showed increased levels of sedation 90 min after premedication with respect to placebo... This group showed decreased levels of sedation with respect to midazolam..." (page 555) "Furthermore, in the preoperative period, impairment in psychomotor performance was more significant in the midazolam group. In the Trail Making A and B test....the melatonin and midazolam groups exhibited a significantly poorer performance compared with placebo. However, in the Word Fluency test, the midazolam group showed a significant impairment...whereas there was no difference between the scores of the melatonin and placebo groups... The placebo group showed better postoperative performance on the Word Fluency test. Amnesia was only significant in the midazolam group..." (page 556) No harms reported
Capuzzo 2006	Melatonin, placebo	No harms reported
Caumo 2007	Melatonin, placebo	No harms reported
Caumo 2009	Melatonin, clonidine, placebo	No harms reported
Dianatkhah 2015	Melatonin, oxazepam	"A smaller proportion of the participants experience delirium in the melatonin group (n=4, 0.06%) than in the oxazepam group (n=9, 0.12%), but this difference was not statistically significant (P value = 0.187)" (page 125) No harms reported
Hoseini 2015	Melatonin, clonidine, gabapentin, placebo	No harms reported; however, the frequency of vomiting and the severity of nausea were measured, and no differences between groups were observed (Table 3) (page 123)
Ionescu 2008	Melatonin, midazolam, placebo	"Amnesia scores, assessed as the number of remembered pictures, were significantly better (the score of the remembered pictures was greater) in the melatonin group in comparison to the midazolam group at every evaluation time, whereas there were no significant difference between the melatonin and placebo groups" (page 11) "No side effects of melatonin were noted" (page 10)
Ismail 2009	Melatonin, placebo	"Contraray to the control group, MAP decreased significantly after melatonin premedication. No incidence of hypotension or bradycardia requiring intervention was reported in groups...One patient in the melatonin group complained of dizziness, and another patient in control group suffered nausea" (page 1148)
Jain 2019	Melatonin, placebo	"In our study, there were no untoward incidences of bradycardia, cardiac arrhythmias, respiratory depression, nausea, hypotension, anaphylaxis, and drug interactions, in any of the groups" (page 20)
Javaherforooshzadeh 2018	Melatonin, gabapentin, placebo	"In this study, a single dose of gabapentin was used, thus, patients did not report any side effects. Ismail et al. found that MAP was significantly reduced after melatonin premedication, although it was described that this difference, at some points, was unimportant between the groups and was consistent with our results" (page 5)
Khanna 2019	Melatonin, pregabalin, alprazolam	"In or study we found that patients in group M were more sedated as compared to group P or group A at all intervals, and the difference at all intervals was statistically significant, whereas sedation score in patients of group P and group A was comparable at all intervals, and the difference at all intervals was statistically insignificant" (page 70) "Side effects like headache, dizziness were comparable in all groups..." (page 70)
Khare 2018	Melatonin, alprazolam, placebo	"Our results showed that both melatonin and alprazolam caused significant sedation in patients as compared to placebo. Among Group M and Group A, melatonin caused less sedation than alprazolam" (page 661) "In our study, alprazolam caused change in orientation score in patients when compared to melatonin and placebo groups...There was a decline in cognitive function in Group A as compared to Group P, whereas the cognitive function was enhanced or maintained in Group M..." (pages 661‐662) No harms reported
Khezri 2013	Melatonin, placebo	"No patient developed hypoxia, hypotension, or bradycardia. Only one patient in the melatonin group complained of a mild headache" (page 322)
Khezri 2013b	Melatonin, gabapentin, placebo	"Significant differences were observed between sedation scores during RBB placement in gabapentin and placebo groups. The difference in sedation scores during RBB placement in melatonin versus gabapentin and placebo was insignificant" (page 584) "No patient developed hypoxia, hypotension, bradycardia, excessive drowsiness (or sleepiness), nausea, and vomiting during surgery. One patient in the melatonin group complained of mild headache, and one in the gabapentin group of severe dizziness while staying in the ward" (page 584)
Khezri 2016	Melatonin, melatonin, placebo	"As shown in Table 3, apart of headache, no significant differences were found in the three groups in terms of other intraoperative and postoperative side effects including pruritus, nausea, vomiting, and respiratory depression. The incidence of headache in group M₆ was significantly higher than other groups" (page 966) "All newborns in our study were free of any adverse effect" (page 967)
Marzban 2016	Melatonin, gabapentin, placebo (midazolam)	Views sedation No harms reported
Mowafi 2008	Melatonin, placebo	"Melatonin premedication reduced MAP compared to control group... No incidence of hypotension or bradycardia requiring intervention was reported in either group" (page 1424) One patient complained of dizziness and two in the melatonin group had excessive sleepiness" (page 1424)
Naguib 1999	Melatonin, midazolam, placebo	"Patients who received midazolam and melatonin showed increased levels of sedation at 60 and 90 min... Furthermore, patients in the midazolam group showed significantly (P<0.05) higher levels of sedation compared with the melatonin group at 30 and 60 min after premedication..," (page 877) "However, in the preoperative period only patients in the midazolam group experienced significant impairment of psychomotor skills. After operation, patients who received midazolam or melatonin had increased levels of sedation at 30 min and impairment of performance of the DSST... Amnesia was notable only in the midazolam group for one preoperative event" (pages 878‐879) "No side effects were noted" (page 878)
Naguib 2000	Melatonin, midazolam, placebo	"...patients who received premedication with 0.05, 0.1 or 0.2 mg/kg sublingual midazolam or melatonin had a significant decrease in anxiety levels (Figure 1) and increase levels of sedation preoperatively... After operation, patients who received 0.2 mg/kg midazolam premedication had increased levels of sedation at 90 min compared with the 0.05 and 0.1 mg/kg melatonin groups" (pages 877‐878) "However, in the preoperative period, only patients in the three midazolam groups experiences significant impairment in psychomotor skills... In addition, patients in the three midazolam groups had impairment of performance on the DSST at 15, 30, 60, and 90 minutes postoperatively... Amnesia was notable only with the 0.2 mg/kg midazolam group for two preoperative events" (pages 477‐478) "No side‐effects were noted" (page 477)
Naguib 2006	Melatonin, placebo	"Here, oral premedication with 0.2 mg/kg melatonin approximately 50 min before induction of anaesthesia significantly reduced preoperative anxiety and increased sedation without impairment of orientation..." (page 1450) No harms reported
Norouzi 2019	Melatonin, placebo	"In addition, no significant difference was found in orientation between both before melatonin administration and in recovery (P>0.05), while it was statistically significant before anesthesia induction (P=0.44) and lower in the melatonin group before induction... In addition, there was no significant difference in sedation between the two groups..." (pages 64‐65) "The results of this double‐blinded clinical trial showed that MAP was lower in the melatonin group..." (page 65) No harms reported
Patel 2015	Melatonin, midazolam, placebo	"This showed that psychomotor and cognitive functions were not affected in melatonin group patients whereas they were significantly affected in midazolam group patients... This showed that midazolam produced the maximum derangement in both psychomotor and cognitive functions after premedication and before surgery" (pages 39‐40) "The intergroup comparison of sedation scores showed that midazolam produced the highest degree of sedation when compared to melatonin and placebo. Melatonin also showed sedative properties when compared with placebo" (page 41) No harms reported
Pokharel 2014	Melatonin, alprazolam, melatonin + alprazolam, placebo	"In our patients, alprazolam produced more sedation scores than placebo at 60 min after premedication, but the difference was not statistically significant. However, our patients who received alprazolam got sedated half an our earlier than placebo... We too found that the melatonin administration was associated with earlier onset of sleep than placebo" (pages 3‐4) "More number of patients in groups receiving the combination drugs and alprazolam (9 each) did not recognize the picture shown at 60 min after premedication...Amnesia for two events was notable in maximum number of patients in the group receiving the combination of alprazolam and melatonin. However the difference was statistically significant only between groups receiving combination drugs (5 (26%)) and placebo (0) for only one event" (page 3) "There was no statistical difference between the groups in the number of people reporting occurrence of nausea, vomiting, dizziness, headache, or restlessness (Table 1)" (page 3)
Seet 2015	Melatonin, placebo	No harms reported
Torun 2019	Melatonin, midazolam, placebo	"Although sedation levels were considerably higher in the melatonin group than in the placebo group at 25, 30, and 35 minutes, during this increase, patient RSS scores did not exceed 3 and did not affect cognitive or psychomotor functions. No side effects were encountered" (page 6)
Turkistani 2007	Melatonin, melatonin, no premedication (placebo)	No harms reported
DSST: Digit Symbol Substitution Test. MAP: mean arterial pressure. RBB: retrobulbar block. RSS: Ramsey Sedation Scale.

Table 1. Harms reported in primary study reports

Table 2. Sensitivity analysis ‐ primary and secondary outcomes ‐ exclusion of studies with an overall high risk of bias

Outcomes	Statistical method	Studies	Participants	Effect estimate (I²)
Preoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	13	936	‐11.20 (‐13.87 to ‐8.53) (54%)
Final VAS scores	MD (IV, Random, 95% CI)	10	778	‐10.49 (‐13.97 to ‐7.00) (65%)
Change VAS scores	MD (IV, Random, 95% CI)	3	158	‐12.59 (‐16.23 to ‐8.95) (0%)
Postoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	3	236	‐0.79 (‐3.67 to 2.09) (0%)
Final VAS scores	MD (IV, Random, 95% CI)	1	138	0.00 (‐4.94 to 4.94) (‐)
Change VAS scores	MD (IV, Random, 95% CI)	2	98	‐1.20 (‐4.75 to 2.35) (0%)
Preoperative anxiety VAS [mm] ‐ melatonin vs benzodiazepine ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	5	315	0.85 (‐3.01 to 4.72) (66%)
Final VAS scores	MD (IV, Random, 95% CI)	2	133	‐0.95 (‐7.97 to 6.07) (55%)
Change VAS scores	MD (IV, Random, 95% CI)	3	182	2.49 (‐3.68 to 8.66) (79%)
Postoperative anxiety VAS [mm] ‐ melatonin vs benzodiazepine ‐ excluding studies with an overall high risk of bias	MD (IV, Random, 95% CI)	2	122	‐2.02 (‐5.82 to 1.78) (0%)
CI: confidence interval. IV: inverse variance. MD: mean difference. SD: standard deviation. VAS: visual analogue scale.

Table 2. Sensitivity analysis ‐ primary and secondary outcomes ‐ exclusion of studies with an overall high risk of bias

Table 3. Primary and secondary outcomes as reported in the primary study reports

Author, year	Preoperative VAS	Preoperative STAI	Preoperative anxiety HAM‐A	Preoperative BAI	Postoperative VAS	Postoperative STAI	Postoperative HAM‐A	Postoperative BAI
Abbasivash 2019	￬ (90 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Acil 2004	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	￬ (90 min postop) compared to placebo ￬ (90 min postop) compared to midazolam	NM	NM	NM
Capuzzo 2006	→ (90 min after premed) compared to placebo	NM	NM	NM	→ (in recovery room) compared to placebo	NM	NM	NM
Caumo 2007	NM	NM	NM	NM	NM	￬ (6 h postop) compared to placebo	NM	NM
Caumo 2009	NM	NM	NM	NM	NM	￬ (6 h postop) compared to placebo → (6 h postop) compared to clonidine	NM	NM
Dianatkhah 2015	NM	NM	→ (before surgery) compared to oxazepam	NM	NM	NM	￬ (after surgery) compared to oxazepam	NM
Hoseini 2015	NM	→ (120 min after premed) compared to placebo → (120 min after premed) compared to clonidine → (120 min after premed) compared to gabapentin	NM	NM	NM	NM	NM	NM
Ionescu 2008	NM	→ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	￬ (1,6 and 24 h postop) compared to placebo) ￬ (1 h and 24 h postop) compared to midazolam → (6 h postop) compared to midazolam	NM	NM
Ismail 2009	￬ (90 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Jain 2019	￬ (120 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Javaherforooshzadeh 2018	￬ (85 min after premed) compared to placebo → (85 min after premed) compared to gabapentin	NM	NM	NM	￬ (1 h after arrival to recovery room) compared to placebo ￬ (6 h after arrival to recovery room) compared to placebo → (6 h after arrival to recovery room) compared to gabapentin	NM	NM	NM
Khanna 2019	NM	NM	NM	→ (60 min after premed) compared to pregabalin → (60 min after premed) compared to alprazolam	NM	NM	NM	→ (1, 2, 6, 12 hours after surgery) compared to pregabalin → (1, 2, 6, 12 hours after surgery) compared to alprazolam
Khare 2018	￬ (120 min after premed) compared to placebo → (120 min after premed) compared to alprazolam	NM	NM	NM	NM	NM	NM	NM
Khezri 2013	￬ (60 min after premed) compared to placebo	NM	NM	NM	￬ (before discharge from recovery room) compared to placebo	NM	NM	NM
Khezri 2013b	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to gabapentin	NM	NM	NM	￬ (postoperative before discharge) compared to placebo → (postoperative before discharge) compared to gabapentin	NM	NM	NM
Khezri 2016	￬ (20 min after premed) compared to placebo	NM	NM	NM	→ (in recovery room) compared to placebo	NM	NM	NM
Marzban 2016	→ (90 min after premed) compared to placebo/midazolam → (90 min after premed) compared to gabapentin	NM	NM	NM	→ (in recovery room) compared to placebo/midazolam → (in recovery room) compared to gabapentin)	NM	NM	NM
Mowafi 2008	￬ (90 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Naguib 1999	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	→ (90 min postop) compared to placebo → (90 min postop) compared to midazolam	NM	NM	NM
Naguib 2000	￬ (90 min after premed) compared to placebo → (90 min after premed) compared to midazolam	NM	NM	NM	→ (90 min postop) compared to placebo → (90 min postop) compared to midazolam	NM	NM	NM
Naguib 2006	￬ (50 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Norouzi 2019	￬ (50 min after premed) compared to placebo	NM	NM	NM	￬ (in recovery room) compared to placebo	NM	NM	NM
Patel 2015	￬ (60 to 90 min after premed) compared to placebo → (60 to 90 min after premed) compared to midazolam	NM	NM	NM	NM	NM	NM	NM
Pokharel 2014	→ (60 to 90 min after premed) compared to placebo → (60 to 90 min after premed) compared to alprazolam	NM	NM	NM	NM	NM	NM	NM
Seet 2015	→ (30 to 60 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
Torun 2019	￬ (60 min after premed) compared to placebo → (60 min after premed) compared to midazolam	NM	NM	NM	NM	NM	NM	NM
Turkistani 2007	￬ (approximately 100 min after premed) compared to placebo	NM	NM	NM	NM	NM	NM	NM
→: no difference between groups. ￬: lower, difference compared to placebo or midazolam. BAI: Beck Anxiety Inventory. HAM‐A: Hamilton Anxiety Rating Scale. NM: not measured. STAI: State Trait Anxiety Inventory. VAS: visual analogue scale.

Table 3. Primary and secondary outcomes as reported in the primary study reports

Table 4. Sensitivity analysis ‐ primary and secondary outcomes

Outcome	Statistical method	Studies	Participants	Effect estimate (I²)
Preoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies not reporting outcome in mean (SD)	MD (IV, Random, 95% CI)	10	621	‐11.90 (‐14.24 to ‐9.55) (34%)
Final VAS scores	MD (IV, Random, 95% CI)	7	463	‐11.34 (‐14.62 to ‐8.06) (55%)
Change VAS scores	MD (IV, Random, 95% CI)	3	158	‐12.59 (‐16.23 to ‐8.95) (0%)
Postoperative anxiety VAS [mm] ‐ melatonin vs placebo ‐ excluding studies not reporting outcome in mean (SD) or reporting SD values of zero	MD (IV, Random, 95% CI)	4	246	‐4.31 (‐7.18 to ‐1.44) (39%)
Final VAS scores	MD (IV, Random, 95% CI)	2	148	‐6.09 (‐8.74 to ‐3.44) (0%)
Change VAS scores	MD (IV, Random, 95% CI)	2	98	‐1.20 (‐4.75 to 2.35) (0%)
Preoperative anxiety VAS [mm] ‐ melatonin vs benzodiazepine ‐ excluding studies not reporting outcome in mean (SD) and an additional study due to lack of blinding	MD (IV, Random, 95% CI)	5	315	0.91 (‐3.02 to 4.38) (67%)
Final VAS scores	MD (IV, Random, 95% CI)	2	133	‐0.95 (‐7.97 to 6.07) (55%)
Change VAS scores	MD (IV, Random, 95% CI)	3	182	2.61 (‐3.68 to 8.90) (80%)
CI: confidence interval. IV: inverse variance. MD: mean difference. SD: standard deviation. VAS: visual analogue scale.

Table 4. Sensitivity analysis ‐ primary and secondary outcomes

Table 5. Subgroup analysis ‐ preoperative anxiety ‐ melatonin vs placebo

Outcome	Statistical method	Studies	Participants	Effect estimate (I²)	Test for subgroup differences (P)
Anaesthetic modality	MD (IV, Random, 95% CI)	17	1136	‐12.13 (‐14.00 to ‐10.26) (31%)	0.52
General anaesthesia	MD (IV, Random, 95% CI)	11	796	‐12.25 (‐14.85 to ‐9.64) (51%)
Spinal, regional, or topical anaesthesia	MD (IV, Random, 95% CI)	6	340	‐10.97 (‐13.91 to ‐8.02) (0%)
Age of participants	MD (IV, Random, 95% CI)	17	1184	‐11.78 (‐13.99 to ‐9.85) (50%)	0.16
Age > 60 years	MD (IV, Random, 95% CI)	3	258	‐8.04 (‐13.58 to ‐2.50) (0%)
Age ≤ 60 years	MD (IV, Random, 95% CI)	14	946	‐12.36 (‐14.62 to ‐10.09) (50%)
Dose of melatonin	MD (IV, Random, 95% CI)	17	1216	‐11.71 (‐13.91 to ‐9.50) (52%)	0.54
Melatonin dose ≥ 6 mg	MD (IV, Random, 95% CI)	10	735	‐12.28 (‐15.21 to ‐9.35) (57%)
Melatonin dose < 6 mg	MD (IV, Random, 95% CI)	7	481	‐10.98 (‐13.88 to ‐8.09) (22%)
CI: confidence interval. IV: inverse variance. MD: mean difference. ST: standard deviation. VAS: visual analogue scale.

Table 5. Subgroup analysis ‐ preoperative anxiety ‐ melatonin vs placebo

Comparison 1. Melatonin versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Preoperative anxiety (VAS) Show forest plot	18	1264	Mean Difference (IV, Random, 95% CI)	‐11.69 [‐13.80, ‐9.59]

1.1.1 Final VAS scores	14	1066	Mean Difference (IV, Random, 95% CI)	‐11.58 [‐14.08, ‐9.08]
1.1.2 Change VAS scores	4	198	Mean Difference (IV, Random, 95% CI)	‐11.96 [‐15.48, ‐8.45]
1.2 Postoperative anxiety (VAS) [mm] Show forest plot	7	524	Mean Difference (IV, Random, 95% CI)	‐5.04 [‐9.52, ‐0.55]

1.2.1 Final VAS scores	5	426	Mean Difference (IV, Random, 95% CI)	‐6.04 [‐11.69, ‐0.40]
1.2.2 Change VAS scores	2	98	Mean Difference (IV, Random, 95% CI)	‐1.20 [‐4.75, 2.35]
1.3 Postoperative anxiety (STAI) Show forest plot	2	73	Mean Difference (IV, Random, 95% CI)	‐5.31 [‐8.78, ‐1.84]

Comparison 1. Melatonin versus placebo

Comparison 2. Melatonin versus benzodiazepine

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Preoperative anxiety (VAS) [mm] Show forest plot	7	409	Mean Difference (IV, Random, 95% CI)	0.78 [‐2.02, 3.58]

2.1.1 Final VAS scores	3	187	Mean Difference (IV, Random, 95% CI)	0.68 [‐2.54, 3.91]
2.1.2 Change VAS scores	4	222	Mean Difference (IV, Random, 95% CI)	2.31 [‐3.39, 8.01]
2.2 Postoperative anxiety (VAS) [mm] Show forest plot	3	176	Mean Difference (IV, Random, 95% CI)	‐2.12 [‐4.61, 0.36]

2.2.1 Final VAS scores	1	54	Mean Difference (IV, Random, 95% CI)	‐2.20 [‐5.48, 1.08]
2.2.2 Change VAS scores	2	122	Mean Difference (IV, Random, 95% CI)	‐2.02 [‐5.82, 1.78]

Comparison 2. Melatonin versus benzodiazepine