Effects of sevoflurane versus other general anaesthesia on emergence agitation in children

Abstract

Background

Sevoflurane is an inhaled volatile anaesthetic that is widely used in paediatric anaesthetic practice. Since its introduction, postoperative behavioural disturbance known as emergence agitation (EA) or emergence delirium (ED) has been recognized as a problem that may occur during recovery from sevoflurane anaesthesia. For the purpose of this systematic review, EA has been used to describe this clinical entity. A child with EA may be restless, may cause self‐injury or may disrupt the dressing, surgical site or indwelling devices, leading to the potential for parents to be dissatisfied with their child's anaesthetic. To prevent such outcomes, the child may require pharmacological or physical restraint. Sevoflurane may be a major contributing factor in the development of EA. Therefore, an evidence‐based understanding of the risk/benefit profile regarding sevoflurane compared with other general anaesthetic agents and adjuncts would facilitate its rational and optimal use.

Objectives

To compare sevoflurane with other general anaesthetic (GA) agents, with or without pharmacological or non‐pharmacological adjuncts, with regard to risk of EA in children during emergence from anaesthesia. The primary outcome was risk of EA; secondary outcome was agitation score.

Search methods

We searched the following databases from the date of inception to 19 January 2013: CENTRAL, Ovid MEDLINE, Ovid EMBASE, the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCOhost), Evidence‐Based Medicine Reviews (EBMR) and the Web of Science, as well as the reference lists of other relevant articles and online trial registers.

Selection criteria

We included all randomized (or quasi‐randomized) controlled trials investigating children < 18 years of age presenting for general anaesthesia with or without surgical intervention. We included any study in which a sevoflurane anaesthetic was compared with any other GA, and any study in which researchers investigated adjuncts (pharmacological or non‐pharmacological) to sevoflurane anaesthesia compared with no adjunct or placebo.

Data collection and analysis

Two review authors independently searched the databases, decided on inclusion eligibility of publications, ascertained study quality and extracted data. They then resolved differences between their results by discussion. Data were entered into RevMan 5.2 for analyses and presentation. Comparisons of the risk of EA were presented as risk ratios (RRs) with 95% confidence intervals (CIs). Sevoflurane is treated as the control anaesthesia in this review. Sensitivity analyses were performed as appropriate, to exclude studies with a high risk of bias and to investigate heterogeneity.

Main results

We included 158 studies involving 14,045 children. Interventions to prevent EA fell into two broad groups. First, alternative GA compared with sevoflurane anaesthesia (69 studies), and second, use of an adjunct with sevoflurane anaesthesia versus sevoflurane without an adjunct (100 studies). The overall risk of bias in included studies was low. The overall Grades of Recommendation, Assessment, Development and Evaluation Working Group (GRADE) assessment of the quality of the evidence was moderate to high. A wide range of EA scales were used, as were different levels of cutoff, to determine the presence or absence of EA. Some studies involved children receiving potentially inadequate or no analgesia intraoperatively during painful procedures.

Halothane (RR 0.51, 95% CI 0.41 to 0.63, 3534 participants, high quality of evidence) and propofol anaesthesia were associated with a lower risk of EA than sevoflurane anaesthesia. Propofol was effective when used throughout anaesthesia (RR 0.35, 95% CI 0.25 to 0.51, 1098 participants, high quality of evidence) and when used only during the maintenance phase of anaesthesia after sevoflurane induction (RR 0.59, 95% CI 0.46 to 0.76, 738 participants, high quality of evidence). No clear evidence was found of an effect on risk of EA of desflurane (RR 1.46, 95% CI 0.92 to 2.31, 408 participants, moderate quality of evidence) or isoflurane (RR 0.76, 95% CI 0.46 to 1.23, 379 participants, moderate quality of evidence) versus sevoflurane.

Compared with no adjunct, effective adjuncts for reducing the risk of EA during sevoflurane anaesthesia included dexmedetomidine (RR 0.37, 95% CI 0.29 to 0.47, 851 participants, high quality of evidence), clonidine (RR 0.45, 95% CI 0.31 to 0.66, 739 participants, high quality of evidence), opioids, in particular fentanyl (RR 0.37, 95% CI 0.27 to 0.50, 1247 participants, high quality of evidence) and a bolus of propofol (RR 0.58, 95% CI 0.38 to 0.89, 394 participants, moderate quality of evidence), ketamine (RR 0.30, 95% CI 0.13 to 0.69, 231 participants, moderate quality of evidence) or midazolam (RR 0.57, 95% CI 0.41 to 0.81, 116 participants, moderate quality of evidence) at the end of anaesthesia. Midazolam oral premedication (RR 0.81, 95% CI 0.59 to 1.12, 370 participants, moderate quality of evidence) and parental presence at emergence (RR 0.91, 95% CI 0.51 to 1.60, 180 participants, moderate quality of evidence) did not reduce the risk of EA.

One or more factors designated as high risk of bias were noted in less than 10% of the included studies. Sensitivity analyses of these studies showed no clinically relevant changes in the risk of EA. Heterogeneity was significant with respect to these comparisons: halothane; clonidine; fentanyl; midazolam premedication; propofol 1 mg/kg bolus at end; and ketamine 0.25 mg/kg bolus at end of anaesthesia. With investigation of heterogeneity, the only clinically relevant changes to findings were seen in the context of potential pain, namely, the setting of adenoidectomy/adenotonsillectomy (propofol bolus; midazolam premedication) and the absence of a regional block (clonidine).

Authors' conclusions

Propofol, halothane, alpha‐2 agonists (dexmedetomidine, clonidine), opioids (e.g. fentanyl) and ketamine reduce the risk of EA compared with sevoflurane anaesthesia, whereas no clear evidence shows an effect for desflurane, isoflurane, midazolam premedication and parental presence at emergence. Therefore anaesthetists can consider several effective strategies to reduce the risk of EA in their clinical practice. Future studies should ensure adequate analgesia in the control group, for which pain may be a contributing or confounding factor in the diagnosis of EA. Regardless of the EA scale used, it would be helpful for study authors to report the risk of EA, so that this might be included in future meta‐analyses. Researchers should also consider combining effective interventions as a multi‐modal approach to further reduce the risk of EA.

Plain language summary

Agitation in children after sevoflurane anaesthesia

Review question

We reviewed the evidence looking at how often children wake up agitated after a sevoflurane general anaesthetic compared with other general anaesthetics. We also reviewed evidence looking at the effects of other treatments (e.g. a medication given during the anaesthetic, the presence of a parent when a child wakes up) on how often children wake up agitated after receiving a sevoflurane anaesthetic.

Background

Sevoflurane is a commonly used anaesthetic gas for children because it can be breathed in by face mask and works very quickly in getting children off to sleep. Sevoflurane is given continuously during an operation to keep the child asleep, and it is turned off when it is time for the child to wake up. It is very common for children, especially preschool children, to wake up restless, agitated, delirious or thrashing around after receiving a sevoflurane anaesthetic. We call this "emergence agitation." It can occur even when no pain is present and usually resolves within 30 minutes of waking up. Children with emergence agitation may injure themselves, bump the operation wound and pull out drips or wound drains. Emergence agitation can be distressing for parents and caregivers. We wanted to discover whether the rate of emergence agitation is lowered when different anaesthetics are used. We also wanted to know whether treatments can be given to reduce the rate of emergence agitation when sevoflurane is used.

Study characteristics

The evidence is current to January 2013. We found a total of 158 studies involving 14,045 children. A total of 69 studies compared a sevoflurane anaesthetic with a different anaesthetic, and 100 studies looked at treatments to reduce the rate of emergence agitation with a sevoflurane anaesthetic. Most of these treatments were medications that were compared with a dummy treatment (placebo) or with no medication. We reran the search in April 2014 and will address identified studies of interest when we update the review.

Key results

The medications propofol, halothane, alpha‐2 agonists (dexmedetomidine, clonidine), opioids (e.g. fentanyl) and ketamine reduce the rate of emergence agitation, whereas no clear evidence of an effect was found for the anaesthetic gases desflurane and isoflurane, the premedication midazolam and parental presence when a child wakes up from anaesthesia.

Quality of the evidence

Overall the evidence is of moderate to high quality. Researchers should consider combining effective interventions to see whether the risk of EA can be reduced further.

Summary of findings

Summary of findings for the main comparison. Other GA/adjunct versus sevoflurane anaesthesia for reducing risk of emergence agitation.

Other GA/adjunct versus sevoflurane anaesthesia for reducing risk of emergence agitation
Patient or population: patients with emergence agitation Settings: hospital setting (postanaesthesia care unit) Intervention: other GA/adjunct versus sevoflurane anaesthesia
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Other GA/adjunct versus sevoflurane anaesthesia
Emergence agitation—Propofol induction and maintenance	Study populationa		RR 0.35 (0.25 to 0.51)	1098 (14 studies)	⊕⊕⊕⊕ highb
	314 per 1000	110 per 1000 (78 to 160)
	Moderatea
	368 per 1000	129 per 1000 (92 to 188)
Emergence agitation—Dexmedetomidine	Study populationa		RR 0.37 (0.29 to 0.47)	851 (12 studies)	⊕⊕⊕⊕ highb
	434 per 1000	160 per 1000 (126 to 204)
	Moderatea
	368 per 1000	136 per 1000 (107 to 173)
Emergence agitation—Fentanyl	Study populationa		RR 0.37 (0.27 to 0.5)	1247 (15 studies)	⊕⊕⊕⊕ high
	442 per 1000	164 per 1000 (119 to 221)
	Moderatea
	368 per 1000	136 per 1000 (99 to 184)
Emergence agitation—Clonidine	Study populationa		RR 0.45 (0.31 to 0.66)	739 (9 studies)	⊕⊕⊕⊕ high
	436 per 1000	196 per 1000 (135 to 288)
	Moderatea
	368 per 1000	166 per 1000 (114 to 243)
Emergence agitation—Halothane	Study populationa		RR 0.51 (0.41 to 0.63)	3534 (34 studies)	⊕⊕⊕⊕ highc
	294 per 1000	150 per 1000 (121 to 186)
	Moderatea
	368 per 1000	188 per 1000 (151 to 232)
Emergence agitation—Isoflurane	Study populationa		RR 0.76 (0.46 to 1.23)	379 (6 studies)	⊕⊕⊕⊝ moderated
	328 per 1000	249 per 1000 (151 to 403)
	Moderatea
	368 per 1000	280 per 1000 (169 to 453)
Emergence agitation—Desflurane	Study populationa		RR 1.46 (0.92 to 2.31)	408 (6 studies)	⊕⊕⊕⊝ moderated
	176 per 1000	258 per 1000 (162 to 408)
	Moderatea
	368 per 1000	537 per 1000 (339 to 850)
*The basis for the assumed risk (e.g. 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; RR: Risk ratio.
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.

^aWe provided 1 typical risk value for the outcome of emergence agitation and assigned this as "moderate risk." We used the mean control group risk of emergence agitation for all control group participants across all comparisons for this purpose.
 ^bLarge effect with precise result of appreciable benefit.
 ^cSensitivity analyses of studies assessed as having high risk of bias did not change the results. Large effect with precise result of appreciable benefit.
 ^dDowngraded for possible imprecision as the result of wide confidence intervals.

Background

Sevoflurane is an inhaled volatile anaesthetic that is widely used in paediatric anaesthetic practice. First described in 1975 (Wallin 1975), its use commenced in Japan in 1992 and became widespread in 1995 (Holzki 1999). Sevoflurane is a non‐pungent, insoluble agent that facilitates smooth, rapid induction and emergence (Holzki 1999; Johr 2002). However, since its introduction, postoperative behavioural disturbance, observed predominantly in the paediatric population (Cole 2002; Voepel‐Lewis 2003), has become an important clinical issue. Some studies suggest that sevoflurane is associated with the highest risk of behavioural disturbance of all current general anaesthetics, while others show conflicting results (Abbotts 2006; Foesel 2001; Johr 2002; Vlajkovic 2006).

These behavioural changes have been described in the literature using a variety of descriptive terms, such as emergence agitation (EA), emergence delirium (ED) and postanaesthetic excitation. No consensus has been reached regarding a definition (Cole 2002; Vlajkovic 2006); however the condition has been described as a mental disturbance during recovery from general anaesthesia that may consist of hallucinations, delusions and confusion manifested by moaning, restlessness, involuntary physical activity and thrashing about in the bed (Sikich 2004). Emergence delirium appears to represent a subset of EA, as not all agitated children are truly delirious (Bajwa 2010; Malarbi 2011). For the purpose of this systematic review, the term 'emergence agitation' will be used to encompass this clinical entity. Until recently, no reliable and validated scale has been available to measure EA. Concern has been expressed as to the reliability of research results and the ease of comparing studies; this probably played a part in the development in 2004 of the Pediatric Anesthesia Emergence Delirium (PAED) scale (Sikich 2004). Although the PAED scale is now the most frequently used scale in research studies, at least one investigator has described a PAED scale modification (Locatelli 2013), and some study authors are now reporting EA by using two scales simultaneously—typically PAED plus one other scale (Patel 2010; Li 2011; Na 2013; Salik 2011). In addition, some trial authors are using different PAED scores as cutoffs for the presence of EA (Characteristics of included studies). This suggests that no one scale currently fulfils all relevant requirements in determination of EA (Bajwa 2010). 
 
 The potential adverse effects of EA are mostly short lived (Veyckemans 2001). It would be unusual for children with EA to be discharged from the postanaesthetic care unit (PACU), as a restless child may cause self‐injury, the dressing or surgical site may be disrupted and indwelling devices have the potential to become dislodged. To prevent such outcomes, the child may require pharmacological or physical restraint. Pharmacological management provides the disadvantage of exposing the child to medications such as opioids and sedatives. These drugs themselves could have adverse effects, and their administration could delay discharge from the PACU or hospital. The psychological and long‐term consequences of EA are largely unknown, but it has been suggested that maladaptive behaviours, for example, withdrawal, sleeping and eating problems, may be associated with EA (Holzki 1999; Kain 2004). Extra care is required to manage a patient with EA, and this might strain already limited nursing resources. Caregivers are at risk of injury when managing these children and may feel dissatisfied with the quality of available anaesthetic care. Parents who witness EA may become concerned regarding future anaesthetic experiences for their child (Houck 2005). Additional costs and potential delays to discharge may be significant.
 
 The exact aetiology of EA remains unclear; however research to date suggests numerous predisposing factors (Voepel‐Lewis 2003). Anaesthetic factors may include rapid emergence and the intrinsic characteristics of the anaesthetic. The newer volatile anaesthetic agents, such as sevoflurane, allow faster emergence, which potentially results in early manifestation of acute pain and anxiety (Wells 1999). Some authors have suggested that sevoflurane exerts a stimulating or even neurotoxic effect on the central nervous system (Constant 1999; Johr 2002; Vlajkovic 2006). Patient‐related factors include age, preoperative anxiety and the temperament of the child. The risk of EA is highest in preschoolers, potentially because of psychological immaturity in this age group (Voepel‐Lewis 2003). Surgery‐related factors include pain and type of surgery. Pain may increase the risk of EA (Lynch 1998), and the behaviour of a child in pain may mimic EA. Otorhinolaryngology and ophthalmological procedures carry an increased risk of EA; however this phenomenon has been observed even after non‐painful imaging procedures (Voepel‐Lewis 2003).

Sevoflurane may be a major contributor to the development of EA. Therefore, an evidence‐based understanding of the risk/benefit profile regarding sevoflurane compared with other general anaesthetic (GA) agents and adjuncts would facilitate its rational and optimal use.

Description of the condition

As stated above, although the condition described by the authors of included studies included terms such as ED, EA and postoperative excitation, we have used the term EA throughout this review to describe postoperative behavioural disturbance during emergence from anaesthesia.

Description of the intervention

Alternative anaesthetic agent to sevoflurane or adjunct with sevoflurane anaesthesia. Sevoflurane is treated as the control anaesthesia in this review.

How the intervention might work

Interventions may delay emergence, allowing for washout of sevoflurane before emergence occurs or modifying emergence effects on the brain in some other way yet to be determined.

Why it is important to do this review

Approximately one‐third of children experience EA after sevoflurane anaesthesia if a preventative intervention is not used (Bajwa 2010).

Objectives

To compare sevoflurane with other general anaesthetic agents, with or without pharmacological or non‐pharmacological adjuncts, with regard to risk of EA in children during emergence from anaesthesia. The primary outcome was risk of EA; the secondary outcome was agitation score.

Methods

Criteria for considering studies for this review

Types of studies

We included all randomized and quasi‐randomized, published and unpublished controlled clinical studies.

Types of participants

We included children younger than 18 years of age presenting for general anaesthesia with or without surgical intervention.

Types of interventions

We have included any sevoflurane anaesthetic with or without nitrous oxide compared with any other general anaesthetic.

Types of general anaesthetics included were other volatile anaesthetics, for example, isoflurane, desflurane and halothane; and any other general anaesthetics, for example, propofol or ketamine.

Pharmacological adjuncts such as fentanyl or other opioids, propofol, midazolam, ketamine, dexmedetomidine or clonidine or non‐pharmacological adjuncts such as parental presence.

Types of outcome measures

Primary outcomes

Emergence agitation (EA) was defined as the number of participants with postoperative behavioural disturbance, as measured by the authors of included studies.

Secondary outcomes

• Postoperative behavioural disturbance or agitation scores such as PAED (Sikich 2004).

• Number of participants with EA during the postoperative period after leaving the PACU, as measured by the authors of included studies.

Search methods for identification of studies

Electronic searches

We searched the Cochrane Central Register of Controlled Trials (CENTRAL, 2013, Issue 1; see Appendix 1) and the following electronic medicine, nursing, psychology and medical databases: MEDLINE (Ovid SP, 1966 to January 2013; see Appendix 2), EMBASE (Ovid SP, 1980 to January 2013; see Appendix 3), the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCOhost, 1982 to January 2013; see Appendix 4), Evidence‐Based Medicine Reviews (EBMR) and the Web of Science (1954 to January 2013; see Appendix 5).

We searched MEDLINE using medical subject headings (MeSH) and text words. We combined this search with the Cochrane highly sensitive search strategy, phases one and two, as provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2005).

We adapted the other databases as appropriate.

We searched for ongoing clinical trials and unpublished studies via Internet searches on the following sites.

• http://www.controlled‐trials.com.

• http://www.update‐software.com.

• http://clinicalstudyresults.org.

• http://centrewatch.com,

We reran the search in April 2014 and will address identified studies of interest when we update the review.

Searching other resources

Trials were also identified by:

• manual searching: Relevant conference proceedings abstracts will be searched;

• snowballing: Reference lists of relevant articles will be checked; and

• contacts: Relevant trial authors will be contacted by e‐mail to identify additional studies.

We did not apply language or publication restrictions.

Data collection and analysis

Selection of studies

Review authors identified titles and abstracts of studies during the initial search. Potentially relevant studies were retrieved in full‐text version and were evaluated for inclusion by two review authors working independently.

Data extraction and management

Two review authors independently extracted data from relevant studies using a standardized data collection form. We resolved disagreements by discussion. If additional information was required, we contacted the authors of the relevant study.

Two review authors independently assessed the following.

• Randomization.

• Allocation concealment, which was graded according to blinding of participants and personnel according to the standard scheme of The Cochrane Collaboration.

• Blinding of outcome assessments.

• Losses to follow‐up and treatment of withdrawals.

• Selective reporting.

Sensitivity analysis was performed to assess major differences with regard to methodological quality.
 Excluded studies are listed in the Characteristics of excluded studies table, and reasons for exclusion are given.

Measures of treatment effect

Sevoflurane is treated as the control anaesthesia in this review. We have presented the results from dichotomous data as risk ratios (RRs) with 95% confidence intervals (CIs); categorical scales of EA were reported in the text. We performed, as applicable, meta‐analyses on subgroups of trials that used similar outcome reporting strategies, for example, all those reporting number of children with postoperative EA in each group.

Unit of analysis issues

We identified no unit of analysis issues.

Dealing with missing data

We contacted authors of studies to ask for missing data; this was reported in the notes section of the Characteristics of included studies table. For example in one study, EA was reported only in the intervention group, and study authors provided control group data for this outcome upon request (Le Berre 2001). Included studies with > 15% withdrawals were assessed as having high risk of attrition bias.

Assessment of heterogeneity

We applied the I² statistic (Higgins 2002) to test for heterogeneity among the studies. We used the random‐effects model to limit the effects of heterogeneity between trials, but when substantial inconsistency (I² > 40%) was found, we explored the reasons for this.

Assessment of reporting biases

Outcomes stated in the methods section and not reported by trial authors were noted and stated in the Characteristics of included studies table or the Main results section. We sought publication bias by using funnel plots.

Data synthesis

We used the Review Manager software of The Cochrane Collaboration (RevMan 5.2) to perform quantitative analysis. The different scales and definitions of EA were managed by using measures of EA as defined by the authors of included studies. We calculated dichotomous data by using risk ratios (RRs) and 95% confidence intervals (CIs). When pooled analyses were not possible, we reported the trial results of individual studies separately. Other data were presented as reported in the original trial. Summary of findings tables were prepared by using the GRADE process and software.

Subgroup analysis and investigation of heterogeneity

Originally, subgroup analyses were preplanned to be undertaken to compare different age groups, different types of surgery—emergency or elective, premedication adjuncts, intraoperative adjuncts, nitrous oxide or air, intravenous fluids or no fluids and surgical versus non‐surgical procedures such as magnetic resonance imaging (MRI), when sufficient numbers of studies were identified. It became clear during performance of this review that these planned analyses were not feasible, so no subgroup analyses were performed. We structured our review to include different interventions for EA, presented as subgroups.

In exploring heterogeneity when three or more trials were found and I² > 40%, we found potentially inadequate analgesia in the control group; multiple routes of adjunct administration such as intravenous (IV) or caudal (e.g. for clonidine as an adjunct) and adenotonsillectomy/adenoidectomy (as painful procedures known to have a high risk of EA) versus other procedures were potential candidates for subgroup analysis in future updates.

Sensitivity analysis

Sensitivity analyses were performed in those studies in which one or more factors indicated a high risk of bias, when compared with all studies.

Results

Description of studies

See the Characteristics of included studies and Characteristics of excluded studies table (below).

Results of the search

Our search yielded 158 included studies investigating 14,045 children and excluded 30 trials (Figure 1). We reran the search in April 2014 and found a further 120 citations with 29 studies of interest. We will address identified studies of interest when we update the review.

1.

Study flow diagram. We reran the search in April 2014. We found a further 29 studies of interest, which we will assess for eligibility when we update the review.*

Included studies

Interventions to prevent EA fell into two broad groups. First, alternative GA compared with sevoflurane anaesthesia (69 studies, of which 65 reported risk of EA), and second, use of an adjunct with sevoflurane anaesthesia versus sevoflurane without an adjunct (100 studies, of which 75 reported risk of EA). Eleven studies investigated both an alternative GA and an adjunct.

All studies included preschool‐aged children, and most studies (57%) limited inclusion to children eight years of age or younger. Only 12 studies (7.5%) included children beyond the age of 12 years. A total of 25 studies included infants younger than one year of age. Nearly all studies excluded American Society of Anesthesiology (ASA) status classification III, IV and V children and those with a history of chronic illness, developmental delay or preexisting behavioural problems.

Among studies in which a sedative premedication was not the intervention, 56 involved administration of a sedative premedication to all children (mostly midazolam), and one study allowed sedative premedication at the discretion of the attending anaesthetist.

Three main categories of procedures were noted in the included studies: children having non‐painful investigations (such as MRI) under general anaesthesia; those undergoing surgery with effective (as defined by study authors) regional analgesia; and children having surgery with or without intraoperative analgesia, in which pain is likely to be a contributing factor to emergence agitation. Some studies involved children receiving potentially inadequate or no analgesia intraoperatively during painful procedures.

Postoperative behavioural disturbance during emergence from anaesthesia was most commonly termed emergence agitation, although the terms emergence delirium and excitation were also used. Numerous definitions of EA during emergence from anaesthesia were based on a wide range of scales. Furthermore, different researchers used a range of cutoffs to determine the presence or absence of EA. Most but not all studies reported the risk of EA, and this measure was the primary outcome for this review. Most studies used a three‐ or five‐point categorical scale to define a cutoff for EA. More recent studies have tended to report the PAED scale as an outcome for EA. At least three studies have reported on both of these outcomes.

Our secondary outcome—"number of participants with EA in the postoperative period after leaving the PACU, as measured by the authors of included studies"—was not reported in any of the included studies.

Although we had planned to pool continuous data when the same EA scale was used, in practice we found no scales suitable for this form of meta‐analysis.

Further details are given in the Characteristics of included studies table.

Excluded studies

Of the 30 excluded studies, 11 (37%) failed to report EA as an outcome (Akinci 2008; Ariffin 1997; Delvi 2007; El‐Hennawy 2009; Funk 2000; Greenspun 1995; Ingelmo 2007; Isik 2006b; Piat 1994; Sarner 1995; Wagner 2003).

Two studies (Hung 2005; Ibrahim 2001) (7%) included adult participants.

Twelve studies (Almenrader 2007; Chen 2010; Choi 2011; Ertugrul 2006; Kawaai 2008; Kawaraguchi 2002; Malmgren 2004; Mckay 2011; Ozer 2003; Shaban 2008; Steinmetz 2007; Uysal 2011) (40%) failed to compare EA versus a control group.

Three studies (Cole 2002; Kain 1999; Mizrak 2011) (10%) did not investigate sevoflurane, and in one study (Kain 2007), it was unclear whether sevoflurane was used.

One article (Mayer 2006) was retracted after publication.

Further details are given in the Characteristics of excluded studies table.

Risk of bias in included studies

See Figure 2 and Figure 3.

2.

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

3.

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

Allocation

One study was quasi‐randomized (Murray 2002). All but three (Erden 2012; Mikawa 2002; Shin 2001) of the remaining 157 included studies reported that the study was randomized, but only 79 studies (50%) reported the method of random sequence generation applied. Allocation concealment was adequately reported in only 30 trials (19%).

Blinding

Given that a vast majority of children were preschoolers, with the exception of parental presence studies, all children were effectively blinded to the intervention of interest. Most trials (133 studies) reported blinding of the outcome assessors. The seven studies in which it was clear that the outcome assessor was not blinded were assessed as having high risk of bias (Aguilera 2003; Johannesson 1995; Michalek‐Sauberer 1998; Milic 2010; Sury 1996; Tripi 2004; Villani 1998).

Incomplete outcome data

One hundred two studies (65%) reported no withdrawals of study participants after randomization. Only three studies (Chandler 2012; Keaney 2004; Kulka 2001b) had > 15% withdrawals and were assessed as having high risk of attrition bias.

Selective reporting

All included studies reported either risk of EA or an EA score, except for three studies (Erden 2012; Jacob 2012; Salik 2011), which failed to report actual data but provided a P value for differences between groups.

Other potential sources of bias

Some evidence of asymmetry can be seen in both funnel plots (Figure 4 and Figure 5). Trials may be missing from the small negative trials corner (i.e. in favour of sevoflurane and not in favour of the other drugs), which suggests possible publication bias.

4.

Funnel plot of comparison: 1 Any other GA versus sevoflurane anaesthesia, outcome: 1.1 Emergence agitation.

5.

Funnel plot of comparison: 2 Adjunct versus placebo/No adjunct during sevoflurane anaesthesia, outcome: 2.1 Emergence agitation.

Effects of interventions

See: Table 1

Comparison 1. Any other GA versus sevoflurane anaesthesia
 Studies investigating risk of EA when comparing alternative GAs with sevoflurane are summarized in Analysis 1.1. Results of different individual agents compared with sevoflurane are reported below.

1.1. Analysis.

Comparison 1 Any other GA versus sevoflurane anaesthesia, Outcome 1 Emergence agitation.

• Sevoflurane versus halothane

All studies (34 trials) for this comparison reported risk of EA and show that halothane has a lower risk of EA when compared with sevoflurane (RR 0.51, 95% CI 0.41 to 0.63).

• Sevoflurane versus isoflurane

All six studies (Bortone 2006; Le Berre 2001; Meyer 2007; Singh 2009; Singh 2012; Valley 1999) investigating this comparison reported risk of EA. No difference in EA was found.

• Sevoflurane versus desflurane

Six studies (Cohen 2002; Demirbilek 2004; Singh 2012; Uzun 2003; Valley 2003; Welborn 1996) investigating this comparison reported risk of EA. No difference in EA was found. One study (Ahishakiye 2011) used a five‐point scale and reported no difference in agitation scores.

• Sevoflurane versus propofol induction and maintenance (including total intravenous anaesthesia [TIVA])

Fourteen studies (Abdel‐Halim 2002; Auerswald 2006; Chandler 2012; El‐Sada 2005; Guard 1998; Gurkan 1999; Hanna 2004; Liao 2010; Lopez Gil 1999; Nakayama 2007; Ou 2011; Ozer Kocak 2001; Picard 2000; Seo 2011) investigated risk of EA for this comparison and found that propofol anaesthesia reduced risk of EA (RR 0.35, 95% CI 0.25 to 0.51). It should be noted that in one of these studies (Ou 2011), the TIVA arm included ketamine 1.5 mg/kg at induction of anaesthesia, and that another study (Hanna 2004) compared TIVA with sevoflurane anaesthesia with 1 mg/kg of propofol at the end of anaesthesia, yet showed TIVA propofol to be superior in reducing EA.

One additional study reported lower mean PAED scores in the propofol group (Kol 2008).

• Sevoflurane versus propofol maintenance (after sevoflurane induction)

Eight studies (Auerswald 2006; Bryan 2009; Cohen 2003; Cohen 2004; Konig 2009; Kubo 2001; Pieters 2010; Uezono 2000) reported risk of EA and found that the propofol group had a lower risk of EA (RR 0.59, 95% CI 0.46 to 0.76). Two further studies (Ahishakiye 2011; Jacob 2012) found lower EA scores in the propofol group. One study (Ahishakiye 2011) used a five‐point scale, and another small study reported that PAED scores indicated a greater frequency of EA in the sevoflurane group without reporting the risk (Jacob 2012).

• Sevoflurane versus ketamine anaesthesia

One small study (El‐Sada 2005) included only 20 participants and showed no difference in risk of EA.

• Sevoflurane versus midazolam anaesthesia

One study (Milic 2010) found lower risk of EA, although it should be noted that the outcome assessor was not blinded.

• Sevoflurane versus other GA combinations

Two further comparisons were reported in one small study (Welborn 1996).

Comparison 2. Adjuncts to sevoflurane anaesthesia

Studies investigating the effectiveness of an adjunct for EA while sevoflurane anaesthesia was administered are summarized in Analysis 2.1. Results of different individual adjuncts compared with no adjunct or placebo are reported below.

2.1. Analysis.

Comparison 2 Adjunct versus placebo/No adjunct during sevoflurane anaesthesia, Outcome 1 Emergence agitation.

• Propofol bolus

At induction

One study (Viitanen 1999a) investigated propofol 3 mg/kg induction and showed no effect on risk of EA, whilst another study (Bal 2006) investigated 2 to 2.5 mg/kg propofol induction with no difference in EA scores between groups. Two studies (Bal 2006; Tsai 2008) investigated 1 mg/kg of propofol just after sevoflurane induction. The former showed no difference in scores for EA whilst the latter showed no effect on risk of EA.

At the end of anaesthesia

Five studies (Abu‐Shahwan 2008; Aouad 2007; Bakhamees 2009; Kim 2011; Lee 2010b) investigating the effect of 1 mg/kg of propofol administered at the end of anaesthesia showed decreased risk of EA (RR 0.58, 95% CI 0.38 to 0.89).

Two further studies (Chiba 2003; Sun 2010) reported only EA scores. Sun 2010 gave propofol 1 mg/kg and found lower PAED scores, whilst Chiba 2003 compared propofol 1 mg/kg versus 2 mg/kg vs intralipid placebo and found no difference in mean EA scores across the three groups.

In view of I² > 40%, we investigated heterogeneity and found that when we removed the trial with an MRI setting, I² decreased from 62% to 49%, and EA was still significantly reduced (RR 0.66, 95% CI 0.46 to 0.93). When only the two studies of adenotonsillectomy were analysed, the reduction in EA was no longer evident (RR 0.74, 95% CI 0.42 to 1.32) (I² = 52%).

• Thiopentone

One study (Aguilera 2003) reported no difference in risk of EA following thiopentone induction, and another study (Hegazy 2009) administered 2 to 3 mg/kg thiopentone after sevoflurane induction and found reduced risk of EA and lower PAED scores in children undergoing MRI scans with mean duration of less than 20 minutes.

• Clonidine

Nine studies (Bock 2002; De Kort 2007; Ghai 2010; Ghosh 2011; Kulka 2001b; Lankinen 2006; Malviya 2006; Tazeroualti 2007; Tesoro 2005) investigated risk of EA and showed an overall reduction in risk of EA (RR 0.45, 95% CI 0.31 to 0.66) (I² = 61%).

Bergendahl 2004 compared midazolam and clonidine premedication and showed reduced EA scores in children younger than five years of age receiving clonidine. Mikawa 2002 investigated oral clonidine premedication 4 mcg/kg and found that it was more effective when compared with clonidine 2 mcg/kg, midazolam 0.5 mg/kg, diazepam 0.4 mg/kg or placebo, without influencing discharge readiness.

In view of I² > 40%, we investigated heterogeneity with respect to studies with regional block analgesia versus systemic analgesia and route of administration (IV vs caudal). I² for these analyses remained unaffected, as did the effectiveness of this intervention. When clonidine (all routes) was used in conjunction with a regional block (seven studies), it was effective in reducing EA (RR 0.37, 95% CI 0.23 to 0.59) (I² = 55%), whereas in the two systemic analgesia studies (one of which was an adenoidectomy study), this effect was no longer evident (RR 0.74, 95% CI 0.49 to 1.12) (I² = 19%).

• Dexmedotomidine

Twelve studies (Asaad 2011; Erdil 2009; Guler 2005; Ibacache 2004; Isik 2006a; Lili 2012; Meng 2012; Ozcengiz 2011; Patel 2010; Saadawy 2009; Sato 2010; Shukry 2005) investigating this intervention found a large overall reduction in risk of EA, with I² = 0 (RR 0.37, 95% CI 0.29 to 0.47). An additional four studies reported lower EA scores for this intervention (Al‐Zaben 2010; Anand 2011; Mizrak 2013; Salik 2011).

• Ketamine

Oral premedication

Two studies (Abdelmawgoud 2012; Khattab 2009) have shown this to be an effective intervention with an overall reduction in risk of EA.

Ketamine bolus after induction

One study showed no reduction in risk of EA compared with placebo (Analysis 2.1.9), although study authors reported lower risk of EA (Tsai 2008).

Ketamine 0.25 mg/kg bolus at end of anaesthesia

Three studies (Abu‐Shahwan 2007; Dalens 2006; Lee 2010a) show an overall reduction in risk of EA (RR 0.30, 95% CI 0.13 to 0.69).

In view of I² > 40%, we investigated heterogeneity with respect to potentially inadequate analgesia for painful surgery versus any other setting such as MRI, or painful surgery with analgesia. When the study in which no analgesia was reported to be administered to the control group undergoing adenotonsillectomy was removed (Lee 2010a), I² decreased from 44% to 0%, yielding RRs of 0.30 (95% CI 0.13 to 0.69) and 0.43 (95% CI 0.22 to 0.81), respectively.

• Fentanyl

Fifteen included studies showed an overall decrease in risk of EA (RR 0.37, 95% CI 0.27 to 0.50) (I² = 54%). One further study investigating intranasal fentanyl 1 mcg/kg (Rampersad 2010) reported no difference in risk of EA (but did not present the risk data) or mean EA scores (data were reported).

In view of I² > 40%, we investigated heterogeneity with respect to potentially inadequate analgesia for painful surgery versus any other setting such as MRI, or painful surgery with analgesia or route of administration.

When the three studies of adenotonsillectomy/adenoidectomy (Bakhamees 2009; Demirbilek 2004; Erdil 2009) in which limited (rectal paracetamol only) or no analgesia was reported to be administered to the control group were removed, I² remained high at 61% but still showed fentanyl to be an effective means of decreasing EA (RR 0.34, 95% CI 0.23 to 0.49).

Analysis of IV fentanyl versus non‐IV fentanyl (intranasal, transmucosal) studies showed that this intervention was still effective in the 12 IV fentanyl studies (RR 0.35, 95% CI 0.24 to 0.51) (I² = 55%) (Asaad 2011; Bakhamees 2009; Cravero 2003; Demirbilek 2004; Erdil 2009; Im 2004; Inomata 2010; Isik 2005; Jung 2010; Li 2011; Makharita 2009; Tazeroualti 2007). The three non‐IV studies (Binstock 2004; Finkel 2001; Galinkin 2000) had an RR of 0.42 (95% CI 0.22 to 0.77) (I² = 59%).

• Other opioids (including tramadol)

Five studies (Dong 2010; Gouda 2004; Na 2013; Ozturk 2009; Shen 2012) show an overall reduction in risk of EA when remifentanil is administered (RR 0.50, 95% CI 0.30 to 0.85).

Two studies (De Kort 2007; Li 2011) show no overall reduction in risk of EA when sufentanil is administered.

One study (Kim 2009) administered alfentanil and reduced the risk of EA.

One study (Dalens 2006) found nalbuphine to be effective in reducing the risk of EA.

One study (Abdelmawgoud 2012) administered dextromethorphan and found it to be effective in reducing the risk of EA.

One study (Murray 2002) found that oxycodone premedication was ineffective in reducing the risk of EA.

Two studies investigated tramadol. One showed reduced risk of EA (Fan 2000), and the other (Sun 2009) found lower PAED scores.

• Midazolam

Oral midazolam premedication

Seven studies (Jiang 2010; Ko 2001; Kubo 2001; Lapin 1999; Ozcengiz 2011; Viitanen 1999b; Viitanen 1999c) showed no overall reduction in risk of EA (RR 0.81, 95% CI 0.59 to 1.12) (I² = 50%).

Two further studies (Arai 2005; Kazak 2010) found no difference in EA scores.

One study (Breschan 2007) compared 0.5 mg/kg versus 1 mg/kg oral premedication and found no difference in the risk of EA.

In view of I² > 40%, we investigated heterogeneity with respect to midazolam dose; studies with regional block analgesia versus systemic or no analgesia and adenoidectomy versus other procedures. The only difference was found when the two adenoidectomy studies were removed, resulting in I² = 4% and an overall reduction in EA (RR 0.68, 95% CI 0.53 to 0.87).

Flumazenil reversal of midazolam premedication

This comparison showed no difference in the risk of EA in one study (Araki 2005).

Midazolam IV at induction of anaesthesia

One study (Byon 2012) showed no difference in PAED scores.

Midazolam IV at the end of anaesthesia

Two studies (Kim 2011; Kulka 2001a) reduced the risk of EA.

One study (Bae 2010) showed a reduction in EA scores.

• Non‐steroidal anti‐inflammatory drugs (NSAIDs)

Three studies were suitable for meta‐analysis. Two investigated ketorolac (Davis 1999; Im 2004), and one, ibuprofen premedication (Shin 2001); these studies showed a decrease in risk of EA (RR 0.43, 95% CI 0.26 to 0.73). One further study investigating ketorolac (Rampersad 2010) reported no difference in risk of EA (but did not present the risk data) nor mean EA scores (data were reported), and another, investigating diclofenac, showed lower PAED scores (Sun 2009).

• Regional blocks (local anaesthetic)

Caudal

Two studies (Aouad 2005; Araki 2005) showed reduced risk of EA.

Subtenon

One study (Seo 2011) showed reduced risk of EA but provided no analgesia to the control group.

Peritonsillar bupivacaine

One small study (Guler 2002) showed no difference in risk of EA.

• 5‐HT₃ antagonists

Three studies investigated this intervention—two with ondansetron (Erden 2012; Hosten 2011) and one with tropisetron (Lankinen 2006)—with no overall effect in EA. Erden 2012 reported no difference in agitation scores but did not report the risk of EA, so this study does not appear in Analysis 2.1.25.

• Hydroxyzine

One study (Koner 2011), which added this intervention to a midazolam premedication, showed reduced risk of EA.

• Melatonin

One study (Ozcengiz 2011) showed no reduction in risk of EA with melatonin compared with placebo (Analysis 2.1.28), although study authors reported lower risk of EA.

• Magnesium sulphate

One study (Yucel 2012) showed no effect on risk of EA or PAED scores, whilst Apan 2010 also showed no difference in agitation scores.

• Gradual cessation of sevoflurane

One study (Oh 2005) showed no difference in risk of EA when compared with rapid cessation of sevoflurane.

• Sevoflurane concentration

One study (Liao 2011) showed no difference in PAED scores with lower concentrations of sevoflurane titrated to BIS (Bispectral Index).

• Nitrous oxide washout

One study (Shibata 2005) showed lower agitation scores when inhaled nitrous oxide concentration was maintained at the end of surgery until BIS had reached 80 to wash out sevoflurane.

• Diazepam

Diazepam 0.25 mg/kg added to midazolam 0.25 mg/kg for premedication showed decreased EA scores compared with midazolam 0.5 mg/kg alone and placebo (Arai 2005).

• Sufentanil + clonidine (multi‐modal)

One study (De Kort 2007) investigated a multi‐modal approach to EA and found that sufentanil and clonidine in combination (four of 20 participants experienced EA) decreased the risk of EA further than when either drug was used alone (eight of 20 with EA) or against placebo (13 of 19 with EA).

• Parental presence on emergence

Two studies (Burke 2009; Tripi 2004) showed no difference in the risk of EA.

• Acupuncture

One study (Lin 2009) reported lower EA scores with acupuncture after induction of anaesthesia for children undergoing bilateral myringotomy and tympanostomy tube insertion.

• Parental presence at induction of anaesthesia (PPIA)

PPIA showed lower EA scores when all participants were premedicated with midazolam (Arai 2007). Another study (Kazak 2010) using a lower dose of midazolam premedication (0.25 mg/kg) in the PPIA group compared with the non‐PPIA group, which received 0.5 mg/kg midazolam, showed no effect on EA scores.

• Airway management

One study (Lee 2011) showed lower risk of EA when a laryngeal mask airway (LMA) was removed deep compared with an endotracheal tube (ETT) removed awake. No difference between ETT deep versus LMA deep extubation was reported. However no LMA awake group was included; therefore no comparison was performed with ETT awake versus LMA awake or LMA deep versus LMA awake.

Discussion

Strengths of the review

This review is the most comprehensive to date examining the effects of interventions used to prevent EA associated with sevoflurane anaesthesia in children.

Limitations

The main limitation with a review of this type is outcome definition. In this case, we designated EA as the universal term for our outcome of interest, irrespective of the term used by study authors, provided their description reflected this outcome. A degree of variability in scales and definitions was noted, along with varying cutoffs for what might be considered EA by different authors. For example, investigators using the PAED scale sometimes regarded > 10 as the cutoff for EA, whilst others used ≥ 16 as the cutoff for EA. As with a hypotension review (Cyna 2006), we used the definition of EA provided by study authors to pool these data in our meta‐analyses.

In several studies, analgesia was stated to be an effective means of reducing EA when the effect of pain on the child's behaviour was a potential confounder in determining this outcome. Some authorities express the opinion that undertreated pain contributes to EA (Rosen 2013). This was a reason for using potentially inadequate analgesia in our sensitivity analyses, where it was shown to be an important factor in generating heterogeneity between trials.

We note that several studies are awaiting assessment and acknowledge that there will be a lag time in assessing and incorporating these studies in future reviews. However, it appears unlikely that these studies will impact our key findings.

Summary of main results

As can be seen from the Results section and from the meta‐analyses, we have found a large number of well‐designed trials of adequate size with low risk of bias, generating outcomes of interest that are likely to be relevant to clinical practice.

Our key findings were that propofol, halothane, alpha‐2 agonists (dexmedetomidine, clonidine), opioids (e.g. fentanyl) and ketamine reduce the risk of EA (Main results; Summary of findings table 1). No clear evidence showed an effect of desflurane, isoflurane, midazolam premedication and parental presence at emergence in reducing the risk of EA.

Overall completeness and applicability of evidence

This review is very likely to represent research findings to date and to be applicable to clinical practice. In view of the funnel plot findings detailed in the Results section, we suggest some caution about the magnitude of the findings in favour of other GAs/adjuncts, but not the direction of effect.

Sensitivity analyses of adenotonsillectomy/adenoidectomy studies seem to show that this setting is more refractory to reductions in EA with propofol bolus, midazolam premedication and clonidine. In contrast, fentanyl and ketamine remain effective, irrespective of this context.

Quality of the evidence

This review shows good evidence of benefit or lack of benefit for several interventions, as stated above. Eleven studies had one or more factors designated as causing high risk of bias, but sensitivity analyses in the nine studies in which this was possible (Chandler 2012; Johannesson 1995; Keaney 2004; Kulka 2001b; Michalek‐Sauberer 1998; Murray 2002; Sury 1996; Tripi 2004; Villani 1998) showed no difference in findings. The other two were single studies for which sensitivity analyses were impossible (Aguilera 2003; Milic 2010).

We have noted some significant heterogeneity, using a random‐effects model, with some of the comparisons, namely, halothane, clonidine, fentanyl, midazolam premedication, propofol 1 mg/kg bolus at end and ketamine 0.25 mg/kg bolus at end. Sensitivity analysis showed minimal changes in overall findings.

Potential biases in the review process

See "Limitations" at the beginning of the Discussion.

Agreements and disagreements with other studies or reviews

Our findings are consistent with those described in the halothane versus sevoflurane review by Kuratani 2008, and we have included all of the 23 studies in their review plus an additional 11 studies investigating EA. Dahmani 2010 reviewed pharmacological interventions for sevoflurane and desflurane anaesthesia, such as propofol, midazolam, ketamine, fentanyl, alpha‐2 agonists and perioperative analgesia, for preventing EA; the only indication of disagreement with the current review was a statement that IV fentanyl was ineffective in preventing EA. This discrepancy appears to be due to inclusion of fewer studies in this previous review (only two IV fentanyl studies vs 12 in the current review). One meta‐analysis of randomized controlled trials compared propofol anaesthesia (induction and maintenance, or maintenance after sevoflurane induction) versus sevoflurane for reducing EA in children (Sun 2008). Although we included more studies and separated the analyses for propofol maintenance after sevoflurane induction from those for propofol for induction and maintenance, our findings were consistent.

Authors' conclusions

Implications for practice.

Propofol, halothane, alpha‐2 agonists (dexmedetomidine, clonidine), opioids (e.g. fentanyl) and ketamine have been shown to reduce the risk of EA when compared with sevoflurane anaesthesia. No clear evidence shows an effect in reducing the risk of EA with desflurane, isoflurane, midazolam premedication and parental presence at emergence.

Implications for research.

Data on the effects of halothane versus sevoflurane on EA were adequate, and further research is unlikely to generate useful additional findings with respect to EA risk in children. Further investigation of dexmedetomidine and use of propofol‐based anaesthesia, as described in the included studies, is also unlikely to generate further useful data, as these interventions have been shown to be clearly effective. It could be useful to apply a multi‐modal approach by using two or more of the effective interventions reported in this review.

Future studies should ensure adequate analgesia in the control group, for which pain may be a confounding factor for the diagnosis of EA.

Regardless of the EA scale used, it would be helpful for study authors to report the risk of EA, so that this might be included in future meta‐analyses.

Our secondary outcome, "number of participants with EA in the postoperative period after leaving the PACU, as measured by the authors of included studies," was not reported in any of the included studies and is probably of limited clinical relevance. Therefore it will be excluded as an outcome in future reviews.

In view of the likely different mechanisms in new‐onset maladaptive behaviours (e.g. sleep or eating disturbance, poor concentration) in the postoperative period after discharge from hospital, we have excluded and will be excluding maladaptive behaviours after the immediate emergence period in future updates of this review.

Appendices

Appendix 1. Search strategy for CENTRAL

#1 sevofluran*
 #2 MeSH descriptor Propofol explode all trees
 #3 MeSH descriptor Halothane explode all trees
 #4 MeSH descriptor Isoflurane explode all trees
 #5 MeSH descriptor Anesthetics, General explode all trees
 #6 MeSH descriptor Enflurane explode all trees
 #7 MeSH descriptor Anesthetics explode all trees
 #8 MeSH descriptor Receptors, GABA‐A explode all trees
 #9 MeSH descriptor Anesthesia, General explode all trees
 #10 MeSH descriptor Anesthetics, Intravenous explode all trees
 #11 MeSH descriptor Nitrous Oxide explode all trees
 #12 MeSH descriptor Thiopental explode all trees
 #13 MeSH descriptor Analgesics, Opioid explode all trees
 #14 Propofol or Halothane or Isoflurane or an?esth* or Enflurane or "Nitrous Oxide" or Thiopental or (Opioid near analg*)
 #15 (#2 OR #3 OR #4 OR #5 OR #6 OR #7 OR # OR #9 OR #10 OR #11 OR #12 OR #13 OR #14)
 #16 MeSH descriptor Psychomotor Agitation explode all trees
 #17 MeSH descriptor Postoperative Complications explode all trees
 #18 MeSH descriptor Anesthesia Recovery Period explode all trees
 #19 MeSH descriptor Confusion explode all trees
 #20 MeSH descriptor Delirium explode all trees
 #21 (emergence near (agitation or excit* or delirium or confusion))
 #22 ((postoperative or postan?esthetic) near (agitation or confusion or behavio?ral change*))
 #23 (#16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22)
 #24 (#1 AND #15 AND #23)
 #25 Child* or Adolescent or Infant*
 #26 adult*
 #27 (#26 AND NOT ( #25 AND #26 ))
 #28 (#24 AND NOT #27)

Appendix 2. Search strategy for MEDLINE (Ovid SP)

1. sevofluran*.mp. 
 2. exp Propofol/ or exp Halothane/ or exp Isoflurane/ or exp Anesthetics, General/ or exp Enflurane/ or exp Anesthetics/ or exp Receptors, GABA‐A/ or exp Anesthesia, General/ or exp Anesthetics, Intravenous/ or exp Nitrous Oxide/ or exp Thiopental/ or exp Analgesics, Opioid/ 
 3. exp Psychomotor Agitation/ or exp Postoperative Complications/ or exp Anesthesia Recovery Period/ or exp confusion/ or exp delirium/ or (emergence adj3 (agitation or excit$ or delirium or confusion)).mp. or ((postoperative or postan?esthetic) adj3 (agitation or confusion or behavio?ral change$)).mp.
 4. exp Child/ or exp Adolescent/ or exp Infant/ or child*.mp. 
 5. exp Adult/ or adult*.mp. 
 6. 5 not (4 and 5)
 7. (1 and 2 and 3) not 6
 8. ((randomized controlled trial or controlled clinical trial).pt. or randomized.ab. or placebo.ab. or drug therapy.fs. or randomly.ab. or trial.ab. or groups.ab.) not (animals not (humans and animals)).sh. 
 9. 8 and 7

Appendix 3. Search strategy for EMBASE (Ovid SP)

1. exp sevoflurane/ or sevofluran*.mp.
 2. exp propofol/ or exp halothane/ or exp isoflurane/ or exp anesthetic agent/ or exp enflurane/ or exp 4 aminobutyric acid a receptor/ or exp general anaesthesia/ or exp intravenous anesthetic agent/ or exp nitrous oxide/ or exp thiopental/ or exp narcotic analgesic agent/
 3. exp restlessness/ or exp postoperative complication/ or exp anesthetic recovery/ or exp confusion/ or exp delirium/ or (emergence adj3 (agitation or excit$ or delirium or confusion)).mp. or ((postoperative or postan?esthetic) adj3 (agitation or confusion or behavio?ral change$)).mp.
 4. exp child/ or exp adolescent/ or exp infant/ or child*.mp.
 5. exp adult/ or adult*.mp.
 6. 5 not (4 and 5)
 7. (1 and 2 and 3) not 6
 8. (randomized‐controlled‐trial/ or randomization/ or controlled‐study/ or multicenter‐study/ or phase‐3‐clinical‐trial/ or phase‐4‐clinical‐trial/ or double‐blind‐procedure/ or single‐blind‐procedure/ or (random* or cross?over* or factorial* or placebo* or volunteer* or ((singl* or doubl* or trebl* or tripl*) adj3 (blind* or mask*))).ti,ab.) not (animal* not (human* and animal*)).sh.
 9. 8 and 7

Appendix 4. Search strategy for CINAHL (EBSCOhost)

S1 ("Propofol") or (MM "Propofol") or ("Halothane") or (MM "Halothane") or ("Isoflurane") or (MM "Isoflurane") or (MH "Anesthetics, General") or "Enflurane" or (MH "Anesthetics") or (MH "Anesthetics, General") or (MH "Anesthetics, Intravenous") or (MH "GABA Agents") or (MH "Nitrous Oxide") or ("Thiopental") or (MM "Thiopental") or (MH "Analgesics, Opioid") 
 S2 (MH "Psychomotor Agitation+") or (MH "Postoperative Complications") or (MH "Anesthesia Recovery") or (MH "Confusion+") or (MM "Delirium") or TX emergence and TX ( agitation or excit* or delirium or confusion ) or (TX ( postoperative or postan?esthetic ) and TX ( agitation or confusion or behavio?ral change*))
 S3 Child* or Adolescent or Infant*
 S4 adult*
 S5 S4 not (S3 and S4)
 (S1 and S2 and sevofluran*) not S5

Appendix 5. Search strategy for ISI Web of Science

#1 TS=sevofluran*
 #2 TS=(Propofol or Halothane or Isoflurane or an?esth* or Enflurane or GABA‐A or Nitrous Oxide or Thiopental or Opioid*)
 #3 TS=Psychomotor Agitation or TS=Anesthesia Recovery Period or TS=(confusion or delirium) or TS=(emergence SAME (agitation or excit* or delirium or confusion)) or TS=((postoperative or postanesthetic or postanaesthetic) SAME(agitation or confusion or behavioral change*))
 #4 #3 AND #2 AND #1
 #5 TS=(Child* or Adolescent or Infant*)
 #6 TS=adult*
 #7 #4 not (#6 not (#5 and #6))

Data and analyses

Comparison 1. Any other GA versus sevoflurane anaesthesia.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Emergence agitation	65		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.1 Halothane	34	3534	Risk Ratio (M‐H, Random, 95% CI)	0.51 [0.41, 0.63]
1.2 Isoflurane	6	379	Risk Ratio (M‐H, Random, 95% CI)	0.76 [0.46, 1.23]
1.3 Desflurane	6	408	Risk Ratio (M‐H, Random, 95% CI)	1.46 [0.92, 2.31]
1.4 Propofol induction and maintenance	14	1098	Risk Ratio (M‐H, Random, 95% CI)	0.35 [0.25, 0.51]
1.5 Propofol maintenance after sevoflurane induction	8	738	Risk Ratio (M‐H, Random, 95% CI)	0.59 [0.46, 0.76]
1.6 Ketamine anaesthesia	1	20	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.22, 2.52]
1.7 Halothane induction + desflurane maintenance	1	40	Risk Ratio (M‐H, Random, 95% CI)	11.00 [1.56, 77.40]
1.8 Halothane induction + sevoflurane maintenance	1	40	Risk Ratio (M‐H, Random, 95% CI)	3.0 [0.34, 26.45]
1.9 Midazolam anaesthesia	1	140	Risk Ratio (M‐H, Random, 95% CI)	0.02 [0.00, 0.39]

Comparison 2. Adjunct versus placebo/No adjunct during sevoflurane anaesthesia.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Emergence agitation	75		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.1 Propofol induction	1	52	Risk Ratio (M‐H, Random, 95% CI)	0.5 [0.20, 1.26]
1.2 Propofol 1 mg/kg just after induction	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.91 [0.50, 1.64]
1.3 Propofol 1 mg/kg bolus at end	5	394	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.38, 0.89]
1.4 Thiopentone induction	1	100	Risk Ratio (M‐H, Random, 95% CI)	0.6 [0.29, 1.24]
1.5 Thiopentone bolus just after induction	1	50	Risk Ratio (M‐H, Random, 95% CI)	0.06 [0.00, 0.97]
1.6 Clonidine	9	739	Risk Ratio (M‐H, Random, 95% CI)	0.45 [0.31, 0.66]
1.7 Dexmedetomidine	12	851	Risk Ratio (M‐H, Random, 95% CI)	0.37 [0.29, 0.47]
1.8 Ketamine oral premedication	2	169	Risk Ratio (M‐H, Random, 95% CI)	0.20 [0.09, 0.44]
1.9 Ketamine 0.25 mg/kg IV just after induction	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.31, 1.30]
1.10 Ketamine 0.25 mg/kg IV bolus at end	3	231	Risk Ratio (M‐H, Random, 95% CI)	0.30 [0.13, 0.69]
1.11 Fentanyl	15	1247	Risk Ratio (M‐H, Random, 95% CI)	0.37 [0.27, 0.50]
1.12 Alfentanil	1	100	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.51, 0.98]
1.13 Remifentanil	5	284	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.30, 0.85]
1.14 Sufentanil	2	105	Risk Ratio (M‐H, Random, 95% CI)	0.34 [0.10, 1.19]
1.15 Sufentanil + clonidine	1	39	Risk Ratio (M‐H, Random, 95% CI)	0.29 [0.12, 0.74]
1.16 Nalbuphine	1	57	Risk Ratio (M‐H, Random, 95% CI)	0.10 [0.01, 0.71]
1.17 Oxycodone	1	64	Risk Ratio (M‐H, Random, 95% CI)	1.21 [0.73, 2.02]
1.18 Tramadol	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.36 [0.14, 0.95]
1.19 Dextromethorphan	1	77	Risk Ratio (M‐H, Random, 95% CI)	0.23 [0.11, 0.45]
1.20 NSAID	3	189	Risk Ratio (M‐H, Random, 95% CI)	0.43 [0.26, 0.73]
1.21 Midazolam premedication	7	370	Risk Ratio (M‐H, Random, 95% CI)	0.81 [0.59, 1.12]
1.22 High dose vs usual dose midazolam premedication	1	115	Risk Ratio (M‐H, Random, 95% CI)	1.16 [0.74, 1.84]
1.23 Flumazenil reversal of midazolam premedication	1	30	Risk Ratio (M‐H, Random, 95% CI)	1.0 [0.76, 1.32]
1.24 Midazolam IV bolus	2	116	Risk Ratio (M‐H, Random, 95% CI)	0.57 [0.41, 0.81]
1.25 5‐HT3 antagonists	2	117	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.34, 1.51]
1.26 Hydroxyzine	1	84	Risk Ratio (M‐H, Random, 95% CI)	0.13 [0.03, 0.51]
1.27 Melatonin	1	50	Risk Ratio (M‐H, Random, 95% CI)	0.25 [0.06, 1.06]
1.28 Magnesium sulphate	1	52	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.33, 2.21]
1.29 Caudal block	2	74	Risk Ratio (M‐H, Random, 95% CI)	0.22 [0.03, 1.77]
1.30 Subtenon block	1	250	Risk Ratio (M‐H, Random, 95% CI)	0.38 [0.21, 0.69]
1.31 Peritonsillar bupivacaine	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.16, 1.21]
1.32 Gradual sevoflurane cessation	1	85	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.61, 1.98]
1.33 Parental presence on emergence	2	180	Risk Ratio (M‐H, Random, 95% CI)	0.91 [0.51, 1.60]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Abdel‐Halim 2002.

Methods	Randomized controlled trial
Participants	2‐10 years, ASA I‐II, bone marrow aspiration, intrathecal aspiration of CSF Exclusion criteria: none stated Recruitment: 60 children (20 in each of 3 groups) Location: Turkey
Interventions	Halothane group: halothane induction and maintenance in oxygen Propofol group: propofol induction and maintenance Control group: sevoflurane induction and maintenance in oxygen
Outcomes	EA defined as "crying or doing abnormal excitatory movements on emergence" Other outcomes: somnolence, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded to maintenance technique but not induction technique (but would not be viable to blind child between IV and inhalational induction), anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Abdelmawgoud 2012.

Methods	Randomized controlled trial
Participants	4‐10 years, ASA I, adenotonsillectomy Exclusion criteria: history of cardiovascular, pulmonary or neurological disease, chronic cough, bronchial asthma, coagulation defects, allergy to study drugs or recent URTI in past 2 weeks Recruitment: 120 children randomly assigned, 116 analysed (38 control group, 39 dextromethorphan group, 39 ketamine group) Location: Egypt
Interventions	Dextromethorphan group: dextromethorphan 1 mg/kg oral premedication 60 minutes before surgery Ketamine group: ketamine 5 mg/kg oral premedication 60 minutes before surgery Control group: placebo oral premedication 60 minutes before surgery All participants: parental presence at induction if separation from parents "unsuccessful" (separation score of 3 or 4), sevoflurane 8% induction with 50% nitrous oxide, IV insertion, intubation, rectal paracetamol 15 mg/kg, maintenance with sevoflurane 3%‐4% and 50% nitrous oxide, gauze soaked with lidocaine 2% applied to tonsillar bed by surgeon
Outcomes	EA defined as score of 4 or 5 on the following 5‐point scale: 1 = obtunded with no response to stimuli 2 = asleep but responsive to movement and stimuli 3 = awake and appropriately responsive 4 = crying and difficult to console 5 = wild thrashing behaviour that requires restraint Other outcomes: separation scores before induction of anaesthesia, cooperation score at induction, duration of emergence, time in PACU, vital signs and side effects in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"closed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	anaesthetist and child blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 patients excluded because of bleeding in the surgical region after extubation: 2 from control group, 1 from ketamine group, 1 from dextromethorphan group
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Abu‐Shahwan 2007.

Methods	Randomized controlled trial
Participants	4‐7 years, ASA I‐II, dental repair Exclusion criteria: behavioural problems, developmental delay Recruitment: 85 children randomly assigned, 80 analysed (42 in intervention group, 38 in control group) Location: Canada
Interventions	Ketamine group: ketamine 0.25 mg/kg IV 10 minutes before end of surgery Control group: saline placebo IV 10 minutes before end of surgery All participants: premedication with acetaminophen 30 mg/kg and midazolam 0.5 mg/kg, sevoflurane with nitrous oxide for induction and maintenance, mivacurium 0.25 mg/kg and ketorolac 1 mg/kg IV before intubation
Outcomes	EA defined as PAED score ≥ 16/20 at any time during first 30 minutes in PACU Other outcomes: CHEOPS 10‐point scale for pain, adverse anaesthetic or surgical outcomes, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"patients were randomized into a placebo‐controlled, double‐blinded study," block randomization in blocks of 10 participants, method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	anaesthetist and child blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	5 children excluded because of intraoperative opioid administration (no details of which groups they were from)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Abu‐Shahwan 2008.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I‐II, MRI examination Exclusion criteria: mental retardation, need for sedative premedication Recruitment: 84 children randomly assigned, 83 analysed (42 intervention group, 41 control group) Location: Canada
Interventions	Propofol group: 1 mg/kg propofol (maximum 30 mg) at completion of diagnostic procedure Control group: inhalation induction with sevoflurane/nitrous oxide and sevoflurane/nitrous oxide maintenance. Saline instead of propofol at completion of diagnostic procedure
Outcomes	EA defined as PAED score ≥ 16/20 at any time during first 30 minutes in PACU Other outcomes: time in PACU, adverse anaesthetic or surgical outcome
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"each patient was allocated to one of the two groups using a concealed random number generator"
Allocation concealment (selection bias)	Unclear risk	"each patient was allocated to one of the two groups using a concealed random number generator"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	anaesthetist and child blinded (anaesthesia assistant prepared and administered study drug)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 of the children from the control group excluded because of propofol administration during induction
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Aguilera 2003.

Methods	Randomized controlled trial
Participants	2‐14 years, ASA I‐II, ENT surgery Exclusion criteria: history of sleep apnoea, parental preference for a particular type of induction, not correctly premedicated, contraindication to study drug Recruitment: 110 children randomly assigned, 100 analysed (50 in each group) Location: United Kingdom
Interventions	Intervention group: thiopentone IV induction followed by mivacurium 0.2 mg/kg Control group: inhalational induction with sevoflurane and nitrous oxide All participants: midazolam 0.5 mg/kg premedication, sevoflurane/nitrous oxide maintenance, morphine 0.1 mg/kg IV given intraoperatively when indicated
Outcomes	"Emergence anxiety" defined as 1 or 2 on the following 4‐point scale (measured 30 minutes after arrival to recovery room): 1 = crying or distressed, uncooperative 2 = anxious but cooperative 3 = awake and calm 4 = asleep Other outcomes: anxiety on arrival to anaesthetic room and on induction of anaesthesia (same scale as above) and late behavioural changes assessed through questionnaire to parents after 1 week, induction time, recovery time
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	anaesthetist not blinded (not viable), not viable to blind child between IV and inhalational induction; however midazolam premedicated at the time of induction
Blinding of outcome assessment (detection bias) All outcomes	High risk	outcome assessor not blinded to anaesthesia technique for emergence assessment
Incomplete outcome data (attrition bias) All outcomes	Low risk	10 children excluded because of protocol violations or non‐evaluability of study parameters: errors in giving premedication to 4 children, 3 children refused the induction method they had been allocated (2 from IV group, 1 from inhalational group), 1 child had "dysphoric reaction during the recovery period precluding satisfactory evaluation of the degree of anxiety and the recovery time," 2 children allocated to IV group had to be anaesthetized by inhalation route because of difficulty locating suitable veins
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ahishakiye 2011.

Methods	Randomized controlled trial
Participants	2‐7 years, elective tonsillectomy and adenoidectomy Recruitment: 266 analysed (93 propofol group, 85 desflurane group, 88 control group) Location: Belgium
Interventions	Propofol group: propofol maintenance Desflurane group: desflurane maintenance Control group: sevoflurane maintenance All participants: sevoflurane induction, standardized analgesic treatment
Outcomes	EA assessed using a 5‐point scale (scale not described) and reported mean (SD) scores Other outcomes: time to extubation, time in PACU, sedation scores, parental interview 24 hours postoperatively regarding agitation and combative behaviour at home
Notes	Abstract only
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	report the number who "completed the study" but not the number randomly assigned
Selective reporting (reporting bias)	Low risk	reported EA scores

Al‐Zaben 2010.

Methods	Randomized controlled trial
Participants	1‐12 years, ASA I, hypospadias repair Exclusion criteria: history of allergy to dexmedetomidine, renal or hepatic dysfunction, cardiac disease, behavioural disturbances or mental retardation, need for sedative or analgesic premedication Recruitment: 48 children (24 in each group) Location: Jordan
Interventions	Dexmedetomidine loading dose 1 mcg/kg IV over 10 minutes after induction of anaesthesia, then infusion at 0.7 mcg/kg/h until end of surgery Control group: saline placebo All participants: sevoflurane induction with 50% nitrous oxide in oxygen, fentanyl 2 mcg/kg and atracurium 0.5 mg/kg IV, tracheal intubation, maintenance with 2% sevoflurane (adjusted to maintain BIS 40‐60) with 60% nitrous oxide. Increases in BP or HR > 20% baseline treated with IV fentanyl 1 mcg/kg, decreases in BP or HR < 20% baseline treated with fluid boluses, ephedrine, atropine as necessary
Outcomes	Behaviour scores measured in PACU up to 120 minutes after arrival, using scale as described by Watcha: 0 = child asleep 1 = calm 2 = crying but can be consoled 3 = crying and cannot be consoled 4 = agitated and thrashing around Report mean (SD) behaviour scores, do not report incidence of EA Other outcomes: additional intraoperative fentanyl requirements, sevoflurane requirement, HR and BP intraoperatively and in PACU, time to extubation, pain scores, time to first analgesic requirement, number of participants requiring analgesia (morphine) in PACU, morphine dose in PACU, adverse events
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization table"
Allocation concealment (selection bias)	Unclear risk	"computer‐generated randomization table used to assign each patient," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child binded, syringe containing saline or dexmedetomidine prepared "by an anaesthesiologist not involved in anaesthetic management of the patients"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported postoperative behaviour scores

Anand 2011.

Methods	Randomized controlled trial
Participants	6 months‐6 years, ASA I‐II, lower abdominal surgeries Exclusion criteria: known allergy to study drugs, suspected coagulopathy, infection at site of caudal block, history of developmental delay, neurological diseases and skeletal deformity Recruitment: 60 children (30 in each group) Location: India
Interventions	Intervention group: caudal 0.25% ropivacaine 1 mL/kg + dexmedetomidine 2 mcg/kg Control group: caudal 0.25% ropivacaine 1 mL/kg + 0.5 mL normal saline All participants: oral midazolam premedication 0.5 mg/kg 1 hour before induction with 8% sevoflurane and 50% nitrous oxide in oxygen, laryngeal mask airway, maintenance with 3% sevoflurane and 50% nitrous oxide with sevoflurane concentration adjusted to maintain haemodynamic changes within 30% of baseline
Outcomes	Emergence behaviour assessed using the following 4‐point scale and reported mean (SD) emergence behaviour scores: 1 = calm 2 = not calm but could be easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited, disoriented. Incidence of EA not reported Other outcomes: time to emergence, sedation scores, pain scores, heart rate, blood pressure, analgesic duration
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"simple randomized sampling was done by lottery method"
Allocation concealment (selection bias)	Unclear risk	"simple randomized sampling was done by lottery method," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child and anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported emergence behaviour scores

Aono 1997.

Methods	Randomized controlled trial
Participants	Healthy boys aged 3‐10 years undergoing circumcision (aged 3‐5 years) or inguinal hernia repair (aged 6‐10 years) Exclusion criteria: agitation or crying before or during anaesthesia induction, > 10% increase in preoperative stable value of heart rate or systolic blood pressure after skin incision, participants reporting pain during recovery period Recruitment: 116 boys randomly assigned, 112 analysed (56 in each group) Location: Japan
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: diazepam 0.2 mg/kg premedication, maintenance with 1 MAC of anaesthetic agent, caudal block 0.25% bupivacaine
Outcomes	EA defined as grade 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited, disoriented Other outcomes: time to extubation, time to emergence
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	dice throw
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded to agent used
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 exclusions from control group: 2 for signs of incomplete pain relief, 1 for agitation before induction. 1 exclusion from intervention group for signs of incomplete pain relief
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Aouad 2005.

Methods	Randomized controlled trial
Participants	2‐6 years, inguinal hernia repair Exclusion criteria: developmental delay, neurological disease, indication of RSI, contraindication for caudal analgesia, failed caudal block, preop agitation Recruitment: 48 children randomly assigned, 44 analysed (22 in each group) Location: Lebanon
Interventions	Intervention group: caudal block 1 mL/kg 0.25% bupivacaine Control group: 1 mcg/kg fentanyl IV before skin incision with further 1 mcg/kg fentanyl if required All participants: sevoflurane induction and maintenance
Outcomes	EA defined as score of 3 or 4 on following 4‐point scale: 1 = calm 2 = not calm but could be easily consoled 3 = moderately agitated or restless and not easily calmed 4 = combative, excited or disoriented, thrashing around Other outcomes: number of children requiring analgesia in recovery
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	random numbers table
Allocation concealment (selection bias)	Unclear risk	"randomized," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 exclusions: 2 in fentanyl group due to preop agitation, 2 in caudal group due to inadequate caudal block
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Aouad 2007.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, strabismus surgery Exclusion criteria: mental disease, neurological disease, treatment with sedatives, full stomach, indication for rapid sequence induction Recruitment: 80 randomly assigned, 77 analysed (36 control, 41 intervention) Location: Lebanon
Interventions	Intervention group: 1 mg/kg propofol at completion of surgery Control group: saline bolus at completion of surgery All participants: midazolam 0.5 mg/kg oral premedication, inhalational induction with sevoflurane, 1 mg/kg IV lidocaine, 15 mg/kg IV paracetamol, 1 mg/kg IV dexamethasone (maximum 16 mg), sevoflurane maintenance with 60% nitrous oxide
Outcomes	2 scales used to assess EA: PAED scores measured (means ± SD reported) but EA defined as grade 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited, disoriented Highest scores from time of LMA arrival until calm were reported. Different assessors for each scale Other outcomes: time to LMA removal, time to emergence, discharge time from PACU, parental satisfaction
Notes	PAED scale scores were correlated with dichotomous outcomes on the 4‐point scale, with study authors concluding that a threshold PAED scale score > 10 was the best discriminator between presence and absence of EA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	random numbers generated by a computer
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 participants excluded from the saline group because of incomplete data collection
Selective reporting (reporting bias)	Low risk	reported incidence of EA and PAED scores

Apan 2010.

Methods	Randomized controlled trial
Participants	3‐16 years; ASA I; adenoidectomy, tonsillectomy or both with or without myringotomy and pressure equalization tube placement Exclusion criteria: preexisting renal disease, cardiovascular dysfunction, neurological or psychological disorders or any conditions that may influence co‐operation, respiratory disease, known allergy to study drugs, receipt of any analgesic or sedatives within 2 days before surgery, parent refusal Recruitment: 118 randomly assigned, 110 analysed (55 in each group) Location: Turkey
Interventions	Intervention: magnesium sulphate 30 mg/kg IV infusion over last 10 minutes of sevoflurane anaesthesia Control: normal saline infusion of same volume (20 mL) over last 10 minutes of sevoflurane anaesthesia All participants: EMLA cream and awake IV access, midazolam 0.1 mg/kg IV premedication, propofol 2‐2.5 mg/kg IV induction, vecuronium 0.1 mg/kg, fentanyl 1 mcg/kg, then intubation, maintenance with sevoflurane at 1 MAC with nitrous oxide in oxygen (35%), reversal with neostigmine and atropine
Outcomes	EA assessed using Aono's 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = moderately agitated or restless 4 = combative, excited, disoriented Other outcomes: pain/discomfort scores, time to extubation, time to eye opening, time to emergence, time to full Aldrete score, time to interaction, time to discharge from PACU, magnesium levels preinfusion and post infusion, laryngospasm, nausea and vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomization procedure was performed using a computer‐generated number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded, "independent observer assessed the agitation level"
Incomplete outcome data (attrition bias) All outcomes	Low risk	8 withdrawals: 4 in study group and 3 in control group due to "unexpected reasons or parent refusal" and 1 in study group due to "inappropriate data collection" (presumably this last one was actually from the control group, as investigators report 55 patients from each group completing the study)
Selective reporting (reporting bias)	Low risk	reported agitation scores

Arai 2005.

Methods	Randomized controlled trial
Participants	1‐7 years, ASA I‐II, adenotonsillectomy Exclusion criteria: long‐term therapy with hepatic enzyme‐inducing drugs and dysfunction of the central nervous or cardiovascular system, refusal to take the whole dose of premedication Recruitment: 42 children (14 in each of 3 groups) Location: Japan
Interventions	Mi group: midazolam 0.5 mg/kg oral premedication Mi + Di group: midazolam 0.25 mg/kg and diazepam 0.25 mg/kg oral premedication Control group: no premedication All participants: study premedications given 45 minutes before induction, sevoflurane induction and maintenance in 100% oxygen, tracheal intubation, local anaesthetic infiltration of the tonsillar fossae
Outcomes	Emergence behaviour assessed on the following 5‐point scale: 1 = obtunded with no response to stimulation 2 = asleep but responsive to movement or stimulation 3 = awake and responsive 4 = inconsolable crying 5 = thrashing behaviour requiring restraint Reported median (range) scores for emergence behaviour, not incidence of EA Other outcome: sedation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	computer‐generated random numbers
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	children blinded between 2 active premeds but investigators did not describe using a placebo premed for the control group, anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported emergence scores

Arai 2007.

Methods	Randomized controlled trial
Participants	1‐3 years, ASA I, minor plastic surgery Exclusion criteria: history of chronic illness, prematurity, developmental delay, previous surgery, parental insistence on a particular study group Recruitment: 60 children randomly assigned, 58 analysed (19 midazolam group, 20 PPIA group, 20 midazolam + PPIA group) Location: Japan
Interventions	Midazolam group: midazolam 0.5 mg/kg PO premedication 40 minutes before induction PPIA group: parental presence (mother) at induction of anaesthesia Midazolam + PPIA group: both of the above All participants: sevoflurane induction in 100% oxygen, fentanyl 1 mcg/kg and vecuronium 0.1 mg/kg to facilitate tracheal intubation, sevoflurane maintenance 1.5%‐2.5% in 60% oxygen and fentanyl 4 mcg/kg
Outcomes	Emergence behaviour assessed on the following 5‐point scale: 1 = obtunded with no response to stimulation 2 = asleep but responsive to movement or stimulation 3 = awake and responsive 4 = inconsolable crying 5 = thrashing behaviour requiring restraint Reported median (range) scores for emergence behaviour, not incidence of agitation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	computer‐generated random numbers
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	not viable to blind child and anaesthetist or to parental presence
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 children excluded as a result of refusal to take oral midazolam (1 from midazolam group and 1 from midazolam + PPIA group)
Selective reporting (reporting bias)	Low risk	reported emergence behaviour scores

Araki 2005.

Methods	Randomized controlled trial
Participants	1‐8 years, ASA I‐II, elective herniorrhaphy Exclusion criteria: allergy to drugs involved, previous adverse anaesthesia experience, mental retardation, long‐term medication that could react with midazolam Recruitment: 60 children recruited (15 in each of 4 groups) Location: Japan
Interventions	Group 1: no caudal and IV saline placebo at the end of surgery Group 2: no caudal and IV flumazenil 0.02 mg/kg at the end of surgery Group 3: caudal with 1 mL/kg 0.25% bupivacaine and IV saline placebo at the end of surgery Group 4: caudal with 1 mL/kg 0.25% bupivacaine and IV flumazenil 0.02 mg/kg at the end of surgery All children received 0.5 mg/kg midazolam premedication, induced with 5% sevoflurane with 66% nitrous oxide in oxygen, maintained with 3% sevoflurane with 66% nitrous oxide in oxygen
Outcomes	EA defined as a score of 2 or 3 on the following scale: 1 = asleep, calm or mildly agitated but easily consolable 2 = moderately agitated or restless but inconsolable 3 = hysterical, crying inconsolably, thrashing Other outcomes: effect of oral midazolam on entering the operating room
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned," "computer‐generated random number"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Asaad 2011.

Methods	Randomized controlled trial
Participants	5‐10 years, ASA I, elective surgery of expected duration 30‐60 minutes (e.g. inguinal hernia repair, hydrocoele, circumcision) under general anaesthesia with caudal block Exclusion criteria: chronic or acute intake of any sedative or analgesic drug, known adverse effects of study drugs, failure of caudal block (defined as an increase in heart rate or mean arterial blood pressure > 10% of preincision value) Recruitment: 90 children randomly assigned, 88 analysed (28 in fentanyl group, 30 in each of the other 2 groups) Location: Egypt
Interventions	Fentanyl group: 1 mcg/kg IV fentanyl (in 10 mL over 10 minutes) Dexmedetomidine group: 0.15 mcg/kg IV dexmedetomidine (in 10 mL over 10 minutes) Control group: 10 mL saline placebo IV over 10 minutes All participants: sevoflurane induction and maintenance with 50% nitrous oxide in oxygen, spontaneous ventilation, intubation without relaxant, caudal block with 0.5 mL 0.25% bupivacaine
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = moderately agitated or restless 4 = combative, excited, disoriented Other outcomes: pain scores (CHIPPS), modified Aldrete scores, haemodynamics values, time to eye opening, time to discharge from PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned by a concealed envelope"
Allocation concealment (selection bias)	Low risk	"concealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child and anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 withdrawals from fentanyl group due to meeting definition of failed caudal block
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Auerswald 2006.

Methods	Randomized controlled trial
Participants	1‐5 years, ASA I‐III, adenoidectomy Exclusion criteria: none described Recruitment: 103 children in 4 groups (24 group 1, 27 group 2, 25 group 3, 27 group 4) Location: Germany
Interventions	Group 1: sevoflurane induction in oxygen, sevoflurane maintenance in 60% nitrous oxide with oxygen Group 2: sevoflurane induction in oxygen, sevoflurane maintenance in air/oxygen with alfentanil IV 20 mcg/kg, then 10‐mcg/kg boluses as required (maximum 100 mcg/kg/h) Group 3: sevoflurane induction in oxygen, propofol + alfentanil maintenance Group 4: TIVA with propofol + alfentanil All participants: premedication with 0.5 mg/kg midazolam 30 minutes preoperatively and EMLA cream/paracetamol 40 mg/kg 60 minutes preoperatively
Outcomes	EA defined as score of 0 on the following scale: 0 = very agitated, not consoled by nurse 1 = light agitation, easily consoled 2 = not agitated, sleeping Other outcomes: effect of premedication at induction, pain on injection with propofol, intubating conditions, postoperative vomiting, pain scores
Notes	English abstract, full paper in German
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomized by drawing lots"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	anaesthetist not blinded, children not blinded between IV induction group and inhalational induction groups
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether PACU nurse performing agitation assessment was blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	12% of participants randomly assigned to group 4 (TIVA) were reassigned to group 3 protocol (sevo induction followed by IV maintenance) as uncooperative for IV insertion
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Bae 2010.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I, strabismus surgery Exclusion criteria: none described Recruitment: 60 children (15 in each of 4 groups) Location: Korea
Interventions	Group I (control): normal saline 2 mL IV at cessation of sevoflurane Group II: midazolam 0.025 mg/kg IV at cessation of sevoflurane Group III: midazolam 0.05 mg/kg IV at cessation of sevoflurane Group IV: normal saline 2 mL IV at cessation of sevoflurane + treatment with IV midazolam if agitation score of 4 or 5 occurred All participants: no premedication; induction with IV thiopental sodium, vecuronium and atropine; intubation; maintenance with sevoflurane 2%‐3% and 50% nitrous oxide; ketorolac 0.8 mg/kg; reversal with glycopyrrolate and pyridostigmine; parental presence in PACU
Outcomes	EA assessed on the following scale: 1 = asleep 2 = calm 3 = crying but can be consoled 4 = crying and cannot be consoled 5 = agitated and thrashing around Reported mean (SD) agitation scores, not incidence of EA Other outcomes: preinduction agitation scores, time to extubation, duration of emergence agitation, time to discharge from PACU, effect of IV midazolam in PACU to treat emergence agitation in group IV participants who had an emergence agitation score of 4 or 5
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly placed in 4 groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	"randomly placed in 4 groups," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	children blinded, not stated whether treating anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported agitation scores

Bakhamees 2009.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I, adenotonsillectomy ± insertion of grommets Exclusion criteria: history of previous anaesthesia, bad hospital experience, known allergy to study drugs, diagnosed psychological or behavioural problems, participants who required propofol or fentanyl at any stage of anaesthesia other than planned Recruitment: 120 children (40 in each group) Location: Kingdom of Saudi Arabia
Interventions	Group F: fentanyl 1.5 mcg/kg before start of surgery and 0.1 mL/kg saline at the end of surgery Group FP: fentanyl 1.5 mcg/kg before start of surgery and propofol 1 mg/kg (maximum 30 mg) at the end of surgery Group C (control): 0.1 mL/kg saline at end of procedure All participants: midazolam 0.5 mg/kg oral premedication, sevoflurane induction and maintenance with 50% nitrous oxide, rectal paracetamol 40 mg/kg
Outcomes	EA was measured on a 10‐point scale: 1 = calm and sleepy 2‐4 = mild agitation 5‐7 = medium agitation 8‐10 = severe agitation Reported incidence of "any agitation" (used in forest plot) and incidence of agitation score of 5 or higher for longer than 5 minutes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Bal 2006.

Methods	Randomized controlled trial
Participants	2‐10 years, ASA I‐II, elective adenotonsillectomy and adenoidectomy Exclusion criteria: cardiac, pulmonary, hepatic, renal diseases and known psychiatric disorder Recruitment: 120 children recruited (40 in each group) Location: Turkey
Interventions	Group P: propofol induction Group SP: sevoflurane induction followed by propofol 1 mg/kg Group S: sevoflurane induction All participants: no premedication, EMLA cream, parents not present at induction, induction as per group allocation, vecuronium 0.1 mg/kg, intubation, fentanyl 1 mcg/kg, sevoflurane 2%‐4% maintenance with 50% nitrous oxide, metamizole 15‐20 mg/kg, metoclopramide 0.25 mg/kg
Outcomes	"Anxiety" at 30 minutes after arrival in PACU was measured on a 4‐point scale: 1 = agitated, crying, uncooperative 2 = anxious but cooperative 3 = awake and calm 4 = asleep Reported median score and mean score ± SD Other outcomes: anxiety on arrival to anaesthetic room or at induction of anaesthesia and postoperative behavioural disturbances in the following week
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly allocated using random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	anaesthetist and child not binded to IV versus inhalational induction (not feasible)
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported "anxiety" scores

Bebawy 2005.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, inguinal hernia repair Exclusion criteria: contraindication to NSAIDs; history of sleep apnoea, developmental delay or psychological disorders Recruitment: 80 children (40 in each group) Location: Egypt
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: premedication with oral midazolam 0.5 mg/kg and rectal diclofenac 2 mg/kg, nitrous oxide during induction and maintenance, IV fentanyl 1 mcg/kg, local infiltration with ropivacaine 3 mg/kg
Outcomes	EA defined as 3 or 4 on the following scale: 1 = awake, calm, cooperative 2 = crying, requires consoling 3 = irritable, restless, screaming, inconsolable 4 = combative, disoriented EA measured on admission to PACU, then every 5 minutes Other outcomes: modified Yale Preoperative Anxiety Scale, a separation scale, an induction scale, objective pain scale (0‐10 inch), emergence and recovery times
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned," computer‐generated table
Allocation concealment (selection bias)	Unclear risk	"randomly assigned," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded (not feasible)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Bergendahl 2004.

Methods	Randomized controlled trial
Participants	1‐11 years, ASA I, adenoidectomy and/or tonsillectomy Exclusion criteria: none stated Recruitment: 104 children randomly assigned, 100 analysed (48 clonidine group, 52 midazolam group) Location: Sweden
Interventions	Intervention group: rectal clonidine premedication 5 mcg/kg 30‐60 minutes preinduction Control group: rectal midazolam premedication 300 mcg/kg 30‐60 minutes preinduction All participants: rectal atropine 40 mcg/kg included with the study drug premedication, sevoflurane induction in oxygen, IV fentanyl 2 mcg/kg, intubation, maintenance in sevoflurane 2%‐4% with nitrous oxide 65%, rectal paracetamol 20 mg/kg, awake extubation
Outcomes	Emergence delirium defined as a "confusion" score ≥ 3 on the following scale: 1 = calm 2 = could be easily calmed 3 = could not be easily calmed 4 = combative, excited, disoriented "Confusion" scores assessed on arrival in PACU and every 30 minutes for first 2 postoperative hours and reported sum of scores Other outcomes: Objective Pain Scale (OPS) scores for first 2 hours, number of rescue doses of fentanyl (for OPS ≥ 5), sedation, shivering, vomiting, parental assessment of analgesia, sedation and sleeping pattern over 24 hours
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomization was performed according to a computer‐generated random list"
Allocation concealment (selection bias)	Low risk	"both patients and investigators were blinded with respect to the randomized treatment and the study code was opened once the entire study was concluded"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"both patients and investigators were blinded with respect to the randomized treatment"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"investigators were blinded with respect to the randomized treatment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 exclusions due to refusal to accept rectal premedication, major unexpected haemorrhage, severe bronchospasm after intubation necessitating cancellation of surgery and refusal to accept mask induction
Selective reporting (reporting bias)	Unclear risk	reported sum of "confusion" scores, but not incidence of delirium

Beskow 1999.

Methods	Randomized controlled trial
Participants	8 months‐18 years, ASA I‐II, minor surgery Exclusion criteria: infection, emergence surgery requiring rapid sequence induction, mental retardation, need for preoperative opioids, history of marked agitation after anaesthesia Recruitment: 62 children (31 in each group) Location: Sweden
Interventions	Intervention group: halothane maintenance Control group: sevoflurane maintenance All participants: rectal premedication with midazolam 0.4 mg/kg and atropine 0.02 mg/kg, local anaesthetic ointment before IV cannulation, parental presence at induction, IV induction with propofol 3 mg/kg (4 children in halothane group and 6 children in sevoflurane group had inhalational inductions caused by cannulation difficulties or fear of needles), maintenance study agent with 65% nitrous oxide, caudal block or local infiltration with bupivacaine 0.5‐1 mL/kg 0.25%, when appropriate
Outcomes	EA defined as a score greater than 5 on a 10‐point visual analogue scale (VAS) with VAS scores assessed upon arrival in PACU and at 15, 30, 45, 60, 75, 90 and 120 minutes after arrival Other outcomes: number of children with pain (pain assessed with VAS), vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	"randomly assigned," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"assessment was done by a specially trained nurse anaesthetist who was unaware of the agent given"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Binstock 2004.

Methods	Randomized controlled trial
Participants	2‐10 years, ASA I‐II, day case surgery Exclusion criteria: none stated Recruitment: 129 children randomly assigned, 125 analysed (23 group 1, 27 group 2, 25 group 3, 26 group 4, 24 group 5) Location: USA
Interventions	5 groups to study the effects of 2 different doses of oral transmucosal fentanyl (OTFC) premedication with or without ondansetron: Group 1: OTFC normal dose (10‐15 mcg/kg) premedication + ondansetron 0.1 mg/kg IV Group 2: OTFC normal dose (10‐15 mcg/kg) premedication + placebo Group 3: placebo + ondansetron 0.1 mg/kg IV Group 4: placebo + placebo Group 5: OTFC low dose (100 mcg fixed dose) premedication + placebo All participants: sevoflurane with nitrous oxide for induction and maintenance, caudal blocks with 0.125% bupivacaine for surgery below umbilicus
Outcomes	Postoperative "anxiety/agitation" defined as not "calm" but "clinging or apprehensive" or "crying" in PACU. Time point used in forest plot is upon arrival in PACU Other outcomes: number of participants requiring analgesia in PACU, caregiver and parental satisfaction, PONV, respiratory events, time to emergence, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"all study drugs and placebos were prepared by the OR pharmacy and labelled with only the patients' name and hospital numbers"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"patients were monitored after anesthesia by an independent blinded observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 withdrawals reported for the following reasons: attending anaesthetists did not approve of the child's enrolment after study of drug administration anaesthetic gas was changed intraoperatively surgery was cancelled after study drug was opened discretion of anaesthetist was applied in the interest of participant's health
Selective reporting (reporting bias)	Low risk	reported incidence of "anxiety/agitation"

Bock 2002.

Methods	Randomized controlled trial
Participants	3‐8 years, ASA I‐II, minor surgery with caudal block Exclusion criteria: known endocrine disease, family history of malignant hyperthermia, aortic stenosis, signs of infection, preoperative agitation Recruitment: 80 children randomly assigned, 72 analysed (18 in each of 4 groups) Location: Germany
Interventions	Group I: clonidine 1 mcg/kg added to caudal + saline placebo IV Group II: clonidine 3 mcg/kg added to caudal + saline placebo IV Group III: no clonidine added to caudal + clonidine 3 mcg/kg IV Group IV: no clonidine added to caudal + saline placebo IV All participants: oral midazolam 0.4 mg/kg premedication, sevoflurane with 70% nitrous oxide induction and maintenance, atropine 0.1 mg/kg and atracurium 0.3 mg/kg IV, intubation, caudal block with 0.175% bupivacaine 1 mL/kg, caudal block judged to be effective if end‐tidal sevoflurane concentration could be reduced to 1.0% or lower without movement or increase in heart rate/blood pressure (> 15% of presurgical level) with surgical stimulus, parental presence at emergence
Outcomes	EA defined as total score ≥ 3 at any time point for items 3‐5 of the Pain/Discomfort Scale (as described by Hanallah et al. Anesthesiology 1987;66:832‐4) Other outcomes: nausea and vomiting, hypotension, bradycardia, time to PACU discharge
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomized into four groups according to a computer‐generated code"
Allocation concealment (selection bias)	Unclear risk	"randomized into four groups," method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "all of the drug combinations in the study were prepared before induction of anesthesia, leaving the observer unaware of group assignment," but not stated whether treating anaesthetist was blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"one blinded observer... recorded all data"
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	2 participants withdrawn because of preoperative agitation, and 6 participants withdrawn because the end‐tidal concentration of sevoflurane could not be reduced to below the predetermined threshold
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Bortone 2006.

Methods	Randomized controlled trial
Participants	1‐6 years, ASA I‐II, elective subumbilical surgery suitable for regional block Exclusion criteria: chronic illness, developmental delay, parents not able to collaborate during induction, premedication, ineffective regional block Recruitment: 110 children randomly assigned (54 sevoflurane group, 56 isoflurane group) Location: Italy
Interventions	Intervention group: isoflurane maintenance at 1‐1.5 MAC Control group: sevoflurane maintenance at 1‐1.5 MAC All participants: un‐premedicated, induction with sevoflurane up to 8% in air/oxygen 50:50 mix with parental presence, LMA with sevo 3%‐5% maintenance until regional block (penile, caudal ilioinguinal) performed with 0.25% bupivacaine, rectal paracetamol 20 mg/kg, randomization to study maintenance agent once block deemed effective
Outcomes	EA defined as 3 or 4 on the following scale: 1 = calm child: no kind of intervention 2 = consolable child: requires only physical contact with parents 3 = agitated child: a screaming and crying child 4 = aggressive child: must be physically restrained to avoid harm
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"children were randomly allocated using a computer‐generated list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"children, their parents as well as the researcher who gathered outcome data, were blinded to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals after randomization
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Breschan 2007.

Methods	Randomized controlled trial
Participants	6 months‐5 years, scheduled minor surgery Exclusion criteria: regional block not working Recruitment: 115 children (57 children in intervention group, 58 children in control group) Location: Austria
Interventions	Intervention group: rectal midazolam 1 mg/kg 10‐15 minutes before surgery Control group: rectal midazolam 0.5 mg/kg 10‐15 minutes before surgery All participants: general anaesthesia induced with propofol or sevoflurane and maintained with 1.5% sevoflurane, appropriate regional block (caudal, penile or local infiltration)
Outcomes	EA defined as a score of 3 measured on following 3 point scale: 1 = calm and cooperative 2 = mildly anxious and agitated but consolable 3 = severely agitated, totally out of control and inconsolable Other outcomes: sedation scores, time to awakening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomisation list," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor (recovery room nurse) blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Bryan 2009.

Methods	Randomized controlled trial
Participants	18 months‐7 years, ASA I‐II, MRI brain Exclusion criteria: inpatients, tracheostomy, family history of malignant hyperpyrexia, allergy to propofol, diagnostic or therapeutic procedures in addition to MRI Recruitment: 201 children randomly assigned, 200 analysed (101 in GAS group, 99 in GAP group) Location: USA
Interventions	GAP group: propofol maintenance (i.e. following sevoflurane induction, a propofol bolus of 2 mg/kg given, then 200 mcg/kg/min infusion maintained) without LMA GAS (control) group: sevoflurane 2% maintenance with LMA All participants: no premedication, sevoflurane induction, maintenance agent increased if movement occurred and decreased if apnoea occurred
Outcomes	EA defined as PAED score > 10 Also reported median PAED scores with interquartile ranges Other outcomes: pausing of MRI scan, respiratory complications, timeliness of care (including time in PACU)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"A block‐randomized plan stratified by age was used to assign the patients to the two treatment groups" "In addition, block randomization with block size 8 was used to ensure that the allocation ratio of the number of subjects in each treatment was 1:1 after every 8th randomized participant in each age‐treatment stratum" method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor (research co‐ordinator) blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	One participant was excluded from the final analysis (not stated which group) because a radiologist added sequences of the spine after scanning the brain, although stated that "the analysis population was the intent‐to‐treat population"
Selective reporting (reporting bias)	Low risk	reported incidence of EA and complete PAED score data

Burke 2009.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I‐III, neurologically intact children, elective MRI, with sevoflurane induction and isoflurane maintenance Exclusion criteria: children requiring preoperative sedation Recruitment: 88 children randomly assigned (45 in parent present group, 43 in parent absent group) Location: USA
Interventions	Parent present (intervention) group: parent reunited with child before emergence in PACU Parent absent (control) group: parent reunited per routine practice after arousal of patient in PACU at discretion of PACU nurse
Outcomes	EA scored on the following 4‐point scale: None = quiet, calm, cooperative Mild = restless/crying, quiets with reassurance Moderate = agitated, restless, non‐purposeful, difficult to control Severe = agitated, thrashing, kicking, non‐purposeful, incoherent, inconsolable Children scoring moderate or severe also scored on PAED scale Other outcomes: duration of agitation, parent and nurse survey, modified Post‐Hospitalization Behaviour Questionaire on day 1 and day 7 (parents telephoned)
Notes	Sevoflurane only used at induction and isoflurane used for maintenance in all participants
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	child and anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor (trained observer) blinded as to purpose of the study
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals after randomization
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Byon 2012.

Methods	Randomized controlled trial
Participants	4‐12 years, ASA I‐II, strabismus surgery Exclusion criteria: history of motion sickness, previous PONV, gastrointestinal tract disorders or administration of antiemetic in the 24 hours before surgery Recruitment: 446 children randomly assigned, 405 analysed (203 in midazolam group, 202 in control group), 41 withdrawals described in risk of bias table Location: Republic of Korea
Interventions	Intervention group: midazolam 0.1 mg/kg IV before induction of anaesthesia Control group: no midazolam All participants: no premedication, ramosetron 6 mcg/kg IV before induction of anaesthesia, induced with thiopental 6 mg/kg and atropine 0.02 mg/kg, sevoflurane 8% after loss of consciousness, LMA insertion, maintenance with sevoflurane 2%‐2.5% with nitrous oxide 50%
Outcomes	EA (secondary outcome) assessed with PAED scale on arrival to PACU and after 30 minutes in PACU. Report mean (SD) PAED scores at these 2 time points. Do not report incidence of EA Other outcomes: incidence of nausea, retching or vomiting (primary outcome); number of participants requiring postoperative analgesics
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double blind"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	"41 patients were excluded because they were lost to follow‐up" (22 from midazolam group, 19 from control group)
Selective reporting (reporting bias)	Low risk	reported PAED scores

Calderon 1996.

Methods	Randomized controlled trial
Participants	3 months‐10 years, ASA I‐II, day case infraumbilical or ENT procedures < 3 hours' surgical time Exclusion criteria: pneumonia in the previous year, anticipated difficult intubation, CVS or CNS disease, congenital cardiac or respiratory anomalies, renal or hepatic insufficiency, family history of muscle disorders (malignant hyperthermia), general anaesthesia in past 2 weeks Recruitment: 39 participants (20 in control group, 19 in intervention group) Location: Spain
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: midazolam premedication 0.2 mg/kg intranasally, nitrous oxide during induction and maintenance, caudal or ilioinguinal block for infraumbilical procedures, fentanyl 2 mcg/kg IV for ENT procedures, postoperative analgesia with paracetamol 10 mg/kg
Outcomes	Reported emergence agitation rate but no specific scale or definition used Other outcomes: time until awakening, recovery (Aldrete scale) scores, adverse effects during recovery period
Notes	English abstract, full paper in Spanish
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	report incidence of EA

Chandler 2012.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, elective strabismus repair Exclusion criteria: not stated Recruitment: 112 participants randomly assigned, 94 analysed (47 in each group, 18 exclusions described in risk of bias table) Location: Canada
Interventions	Intervention group: TIVA propofol and remifentanil Control group: sevoflurane and nitrous oxide induction and maintenance All participants: oral premedication of paracetamol 20 mg/kg and ibuprofen 10 mg/kg, LMA
Outcomes	Emergence delirium defined as PAED score ≥ 10 (PAED scores recorded 5‐minutely from LMA removal) Other outcomes: induction behaviour (PACBIS), pain scores (FLACC scale), time to LMA removal, time in PACU
Notes	Abstract only (conference proceedings) Full paper published in Pediatric Anesthesia 2013;23(4):309‐15 (outside 19 January 2013 search date)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"masked investigator assessed ED"
Incomplete outcome data (attrition bias) All outcomes	High risk	18 exclusions (1 failed IV, 17 protocol deviations)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Chiba 2003.

Methods	Randomized controlled trial
Participants	1‐7 years, short (< 1 hour) general anaesthesia (urology, ENT, paediatric surgery, ophthalmology, orthopaedics, plastics) Exclusion criteria: none stated Recruitment: 90 children (30 in each of 3 groups) Location: Japan
Interventions	Group P1: propofol 1 mg/kg 5 minutes before end of operation Group P2: propofol 2 mg/kg 5 minutes before end of operation Control group: intralipid 0.2 mL/kg 5 minutes before end of operation All participants: no premedication, sevoflurane 7% induction, sevoflurane maintenance 2.5%‐3.5%, sevoflurane ceased 5 minutes before end of operation
Outcomes	EA assessed with new scoring system, which investigators devised (scoring range ‐4 to 10), and report mean scores for each group Other outcomes: time to extubation, time to emergence
Notes	English abstract, full paper in Japanese
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double‐blinded"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"double‐blinded"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	report EA scores

Chiu 1999.

Methods	Randomized controlled trial
Participants	1‐10 years, ASA I‐II, < 25 kg, elective urology surgery < 1 hour Exclusion criteria: none stated Recruitment: 40 children (20 in each group) Location: Malaysia
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance
Outcomes	Restlessness and agitation evaluated using the 3 subjective components of the Objective Pain Scale. If the child was crying inconsolably and was thrashing and hysterical, he was reported to be agitated Other outcomes: emergence time, PONV
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"....recorded by a second investigator blinded to the agents used"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Cohen 2002.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I‐II, adenoidectomy ± myringotomy tubes Exclusion criteria: sleep apnoea, developmental delay, psychological disorders Recruitment: 100 children (50 in each group) Location: USA
Interventions	Intervention group: desflurane maintenance 4%‐6% Control group: sevoflurane maintenance 1.5%‐2% All participants: no premedication, sevoflurane induction in nitrous oxide/oxygen 70:30 mix, mivacurium for ETT, nitrous oxide during maintenance, fentanyl 2.5 mcg/kg, ondansetron 0.1 mg/kg
Outcomes	EA measured on following 3‐point scale with score of 3 defined as severe agitation: 1 = calm 2 = agitated but consolable 3 = severely agitated, inconsolable Other outcomes: pain scores (Objective Pain Scale), time until discharge (Steward recovery score); incidence of agitation, pain, vomiting at home 24 hours after surgery
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	report incidence of EA

Cohen 2003.

Methods	Randomized controlled trial
Participants	2‐36 months, ASA I‐II, elective ambulatory surgery Exclusion criteria: < 40 weeks postconceptional age, signs of developmental delay, procedures expected to last < 30 minutes Recruitment: 53 children (26 sevoflurane group, 27 propofol group) Location: USA
Interventions	Intervention group: propofol maintenance (200 mcg/kg/h) Control group: sevoflurane maintenance All participants: sevoflurane induction, 60% nitrous oxide during induction and maintenance; IV fentanyl 2 mcg/kg or caudal block with 0.25% bupivacaine according to surgical procedure
Outcomes	EA defined as 3 on the following 3‐point scale: 1 = calm 2 = agitated but consolable 3 = severely agitated, inconsolable Other outcomes: time to emergence, recovery and discharge from PACU, agitation and pain at home 24 hours postoperatively
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"an observer blinded to anesthetic technique recorded degree of agitation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Cohen 2004.

Methods	Randomized controlled trial
Participants	< 3 years (2‐33 months), ASA I‐II, ambulatory surgical procedures Exclusion criteria: none stated Recruitment: 56 children (28 in each group) Location: USA
Interventions	Intervention group: propofol maintenance 200 mcg/kg/min Control group: sevoflurane 1.5%‐2.5% maintenance All participants: sevoflurane 8% induction with nitrous oxide 60%, mivacurium to facilitate intubation, nitrous oxide 60% during maintenance, caudal block 1 mL 0.25% bupivacaine for infraumbilical procedures, fentanyl 1 mcg/kg IV for other procedures
Outcomes	"Recovery assessments" performed and reported, along with number of participants requiring additional opioids or midazolam to control pain or agitation during recovery Other outcomes: ionized calcium (primary outcome) and magnesium concentrations, haemodynamics, recovery times, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"assessment was performed by investigator blinded to the anaesthetic technique"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported an agitation outcome

Cravero 2000a.

Methods	Randomized controlled trial
Participants	6 months‐10 years, ASA I‐II, bilateral pressure equalization tube insertion with same surgeon Exclusion criteria: none stated Recruitment: 43 children (22 intervention group, 21 control group) Location: USA
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: nitrous oxide throughout anaesthesia, rectal paracetamol 25 mg/kg
Outcomes	EA defined as "the child thrashing and requiring restraint for > 3 min" Other outcomes: nausea and vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	computer‐generated blocks of 8 participants
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded observer....recorded the level of agitation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Cravero 2000b.

Methods	Randomized controlled trial
Participants	6 months‐10 years, ASA I‐II, MRI scans Exclusion criteria: emotional disorders, cognitive delays, neurological conditions affecting communication Recruitment: 32 children (17 intervention, 15 control) Location: USA
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: no premedication, nitrous oxide throughout anaesthesia
Outcomes	EA assessed with the following 5‐point scale: 1 = obtunded with no response to stimulation 2 = asleep but responsive to movement or stimulation 3 = awake and responsive 4 = crying 5 = thrashing behaviour that requires restraint High threshold EA definition = level 5 for > 3 minutes Low threshold EA definition = level 4 for > 3 minutes High threshold definition used for this review Other outcomes: time to discharge from PACU and secondary recovery unit, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated random assignment in blocks of five"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"a blinded observer ...recorded the level of agitation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Cravero 2003.

Methods	Randomized controlled trial
Participants	18 months‐10 years, ASA I‐II, MRI scans Exclusion criteria: midazolam premedication, neurological condition limiting communication, defined psychological/emotional disorder, cognitive delay Recruitment: 32 participants (16 intervention, 16 control) Location: USA
Interventions	Intervention group: fentanyl 1 mcg/kg IV 10 minutes before end of anaesthesia Control group: saline IV 10 minutes before end of anaesthesia All participants: sevoflurane induction with nitrous oxide, sevoflurane maintenance in 100% oxygen
Outcomes	EA defined as a score ≥ 4 for ≥ 5 minutes despite all calming efforts by the child's parents/guardians and nursing personnel using the following 5‐point scale: 1 = obtunded with no response to stimulation 2 = asleep but responsive to movement or stimulation 3 = awake and responsive 4 = crying 5 = thrashing behaviour that requires restraint Other outcomes: duration of anaesthesia, duration of agitation, itching or vomiting, time to hospital discharge
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number assignment"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether treating anaesthetist was blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"one trained observer, blinded to patient group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Dalens 2006.

Methods	Randomized placebo‐controlled trial
Participants	6 months‐8 years, MRI scans Exclusion criteria: haemodynamic/respiratory unstable participants, severe mental retardation/physical disabilities, sedative/anticonvulsant treatment, emergency MRI scans, extremely distressed/uncooperative children, parental refusal Recruitment: 90 participants (28 control, 33 group K, 29 group N) Location: Canada
Interventions	Group K: ketamine 0.25 mg/kg IV at end of procedure Group N: nalbuphine 0.1 mg/kg IV at end of procedure Control group: saline at end of procedure All participants: sevoflurane induction in 100% oxygen, sevoflurane maintenance in air/oxygen, 10 mL/kg Ringer's solution followed by maintenance, LMA
Outcomes	EA defined as a score ≥ 4 for ≥ 5 minutes despite all calming efforts by the child's parents/guardians and nursing personnel using the following 5‐point scale: 1 = obtunded with no response to stimulation 2 = asleep but responsive to movement or stimulation 3 = awake and responsive 4 = crying 5 = thrashing behaviour that requires restraint
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random list was established by the department of Epidemiology and Human Statistics of our institution"
Allocation concealment (selection bias)	Low risk	"The vials used for the study were prepared by the pharmacy of our institution. All vials were identical and contained the same volume of transparent and unidentifiable fluid. The patient's anesthesiologist and the nursing staff were unaware of the drug administered"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	as above
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Davis 1999.

Methods	Randomized controlled trial
Participants	1‐5 years, ASA I‐II, ambulant surgery for bilateral myringotomy Exclusion criteria: none stated Recruitment: 200 participants (50 in each of 4 groups) Location: USA
Interventions	Group 1: halothane + IV ketorolac 1 mg/kg Group 2: halothane + IV saline placebo Group 3: sevoflurane + IV ketorolac 1 mg/kg Group 4: sevoflurane + IV saline placebo All participants: intranasal midazolam 0.2 mg/kg premedication, induction and maintenance with study anaesthetic agent with nitrous oxide
Outcomes	EA defined as a score of 2 or 3 on the following 3‐point scale: 1 = asleep, calm, or mildly agitated but easily consolable 2 = moderately agitated or restless but inconsolable 3 = hysterical, crying inconsolably, or thrashing Other outcomes: vomiting, number of participants requiring rescue analgesia in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number code"
Allocation concealment (selection bias)	Low risk	"injections were prepared by the hospital pharmacist and placed in specially labelled syringes. Patients, physicians, and the research nurse were blinded to the syringes contents"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	as above
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

De Kort 2007.

Methods	Randomized controlled trial
Participants	2‐8 years, ASA I‐II, circumcision, orchidopexy or inguinal hernia repair Exclusion criteria: insufficient block suspected by haemodynamic variables Recruitment: 80 children randomly assigned, 79 analysed (19 placebo group, 20 in each of the other 3 groups) Location: Belgium
Interventions	Group 1: saline placebo Group 2: sufentanil 0.1 mg/kg Group 3: clonidine 2 mcg/kg Group 4: sufentanil 0.1 mg/kg + clonidine 2 mcg/kg All participants: sevoflurane induction and maintenance, caudal or penile block, then randomly assigned to treatments above
Outcomes	EA assessed on the following 5‐point scale: 1 = obtunded with no response to stimulation 2 = asleep but responsive to movement/stimulation 3 = awake and responsive 4 = crying 5 = thrashing behaviour that requires restraint Incidence of score ≥ 4 used as incidence of EA for this review
Notes	We have assumed that study authors reported clonidine dosage of 2 mg/kg as a typographical error, and that the actual dosage was 2 mcg/kg Abstract only
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"allocated randomly in a double‐blind fashion," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"blinded solution of sufentanil and/or clonidine"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"blinded solution of sufentanil and/or clonidine"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 participant from placebo group discontinued from the study because of inadequate penile block
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Demirbilek 2004.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I, adenotonsillectomy ± grommets Exclusion criteria: "none of the patients had a history of sleep apnoea, developmental delay or psychological disorders" Recruitment: 120 children (30 in each of 4 groups) Location: Turkey
Interventions	Group 1: sevoflurane + saline Group 2: sevoflurane + fentanyl 2.5 mcg/kg during induction Group 3: desflurane + saline Group 4: desflurane + fentanyl 2.5 mcg/kg during induction All children: oral midazolam 0.5 mg/kg premedication, thiopental induction, atracurium to facilitate endotracheal intubation, maintenance with study anaesthetic agent, rectal acetaminophen 30 mg/kg
Outcomes	EA defined as a score of 3 on the following 3‐point scale: 1 = asleep/calm 2 = agitated but consolable 3 = severely agitated/inconsolable Other outcomes: pain scores, vomiting, emergence time
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, treating anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded anesthesiologist also assessed the degree of agitation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Dong 2010.

Methods	Randomized controlled trial
Participants	3‐7 years, ASA I‐II, adenotonsillectomy Exclusion criteria: psychological or emotional disorders, abnormal cognitive development, developmental delay, known history of allergy to drugs in the protocol Recruitment: 60 participants (30 in each group) Location: China
Interventions	Intervention group: remifentanil infusion 1 mcg/kg/min intraoperatively Control group: no remifentanil All participants: IM atropine 0.01 mg/kg 30 minutes preoperatively, IV induction fentanyl 3 mcg/kg and propofol 2.5 mg/kg, intubation after vecuronium 0.1 mg/kg, maintenance with sevoflurane 1.5%‐2.5% with 50% nitrous oxide, sevoflurane concentration adjusted to maintain stable haemodynamics
Outcomes	Emergence agitation defined as PAED scale score ≥ 10 PAED score assessed 5‐minutely for first 30 minutes in PACU Other outcomes: time to eye opening, time to extubation, haemodynamics, SpO2
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"concealed random number generator"
Allocation concealment (selection bias)	Unclear risk	"concealed random number generator"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"scored by two nurses who were blinded to the anaesthetic used"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

El‐Sada 2005.

Methods	Randomized controlled trial
Participants	1‐7 years, ASA I‐II, surgery for congenital esotropia, paralytic or myogenic strabismus Exclusion criteria: disturbed neuromuscular transmission, treatment with aminoglycosides Recruitment: 30 participants (10 in each group) Location: Egypt
Interventions	Propofol group : propofol TIVA (2.5 mg/kg induction, then 10 mg/kg/h) S‐ketamine group: IV induction and maintenance with S‐ketamine (0.5 mg/kg, then 1 mg/kg/h) Control group: sevoflurane induction and maintenance, nitrous oxide All participants premedicated with 0.2 mg/kg midazolam and 0.02 mg/kg atropine IM
Outcomes	EA defined as "crying, difficult to console, or wild thrashing behaviour that require restraint" Other outcomes: PONV, respiratory depression
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"randomized to one of the three groups using sealed envelope assignment, concealing the randomization from the attending anaesthesia staff until directly before induction of anaesthesia"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique but blinded to maintenance, attending anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor in PACU blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Epstein 1995.

Methods	Randomized controlled trial
Participants	9 months‐16 years, ASA I‐II, elective inpatient otorhinolaryngological or orthopaedic procedures Exclusion criteria: none stated Recruitment: 40 participants (20 in each group) Location: USA
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: 2:1 nitrous oxide/oxygen mixture, vecuronium 0.1 mg/kg, intubation, maintained at 1.5 MAC for first 20 minutes then 0.75 MAC until end of surgery, fentanyl 1 mcg/kg 15 minutes before end, reversed with atropine/neostigmine, extubated awake
Outcomes	Reported "excitement" on emergence Other outcomes: haemodynamics, induction and emergence times, coughing, breath‐holding
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor of emergence was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	report incidence of "excitement" on emergence

Erden 2012.

Methods	Randomized controlled trial
Participants	2‐7 years, infraumbilical surgery with caudal block Exclusion criteria: none stated Recruitment: 40 participants (20 in each group) Location: Turkey
Interventions	Intervention group: ondansetron 0.1 mg/kg IV Control group: placebo All participants: oral midazolam 0.5 mg/kg premedication, sevoflurane anaesthesia, caudal block
Outcomes	EA assessed with PAED scale at 10, 20, 30, 60, 120, 180 minutes. Reported comparison of agitation scores between groups (P value) but not the scores Other outcomes: induction quality, parental separation, sedation and pain scores
Notes	Abstract only (conference proceeding)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"assigned to two groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Unclear risk	reported comparison of agitation scores between groups (P value) but not scores or incidence of EA

Erdil 2009.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I, adenoidectomy ± myringotomy/insertion of tubes Exclusion criteria: sleep apnoea, developmental delay, psychological disorders Recruitment: 90 children (30 in each of 3 groups) Location: Turkey
Interventions	Group F: fentanyl 2.5 mcg/kg IV after tracheal intubation Group D: dexmedetomidine 0.5 mcg/kg IV after tracheal intubation Control group: saline All participants: sevoflurane induction with 50% nitrous oxide, atracurium 0.6 mg/kg for tracheal intubation, dexamethasone 0.5 mg/kg for pain and vomiting, sevoflurane maintenance in oxygen
Outcomes	EA assessed on following 5‐point scale with severe agitation defined as ≥ 4: 1 = sleeping 2 = awake, calm 3 = irritable, crying 4 = inconsolable, crying 5 = severe restlessness, disorientation, thrashing around Other outcomes: haemodynamics, pain, PONV, emergence times
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	study drugs and saline were "prepared by a team member not involved in data recording." All given as 10 mL volume via syringe labelled with child's study number
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ersin 2005.

Methods	Randomized controlled trial
Participants	Children with intellectual disability (mean age 11 years), dental treatment under GA Exclusion criteria: none stated Recruitment: 86 participants (44 intervention, 42 control) Location: Turkey
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: midazolam premedication 0.5‐0.75 mg/kg oral or PR, nitrous oxide, atracurium 0.5 mg/kg and atropine 10‐20 mcg/kg IV, diclofenac 2‐2.5 mg/kg PR, local infiltration with lignocaine for extractions
Outcomes	EA defined as VAS > 5 (0‐10 100‐mm scale) Agitation distinguished from pain by the fact that the child was uncooperative, screaming, could not be consoled and tried to leave his/her bed Other outcomes: pain (VAS), PONV, bleeding, laryngospasm, bronchospasm
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly selected by use of a computer‐generated table to receive either..."
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"specially trained recovery nurse, who was unaware of the anesthetic method used"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Fan 2000.

Methods	Randomized controlled trial
Participants	1‐8 years, ASA I, inguinal surgery Exclusion criteria: signs of infection, emergency surgery, mental retardation, GA within past 3 months, contraindication to use of tramadol Recruitment: 40 children (20 in each group) Location: Taiwan
Interventions	Intervention group: tramadol 1 mg/kg IV when suturing Control group: no tramadol All participants: EMLA 1 hour before surgery, premedication with 0.01 mg/kg of atropine orally, thiamylal 4 mg/kg IV induction, sevoflurane maintenance in oxygen, lignocaine infiltration applied before skin incision
Outcomes	EA measured on a visual analague scale (VAS). Reported number of children with VAS > 5 and mean VAS scores Other outcomes: pain, coughing, nausea and vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	child blinded, not stated whether treating anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"emergence agitation was recorded....by a blinded anesthesiologist in the PACU"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Finkel 2001.

Methods	Randomized controlled trial
Participants	6 months‐5 years, ASA I‐II, bilateral myringotomy Exclusion criteria: none stated Recruitment: 150 participants (51 fentanyl 1 mcg/kg, 50 fentanyl 2 mcg/kg, 49 control group) Location: USA
Interventions	Intervention groups: intranasal fentanyl 1 mcg/kg or 2 mcg/kg Control group: intranasal saline All participants: sevoflurane with 60% nitrous oxide induction and maintenance, PR paracetamol 40 mg/kg
Outcomes	EA defined as 3 or more on the following 4‐point scale: 1 = calm 2 = crying but can be consoled 3 = crying and cannot be consoled 4 = agitated and thrashing around Other outcomes: vomiting, pain (Objective Pain Scale)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"the study drug was prepared in a 1‐mL syringe for each patient by a member of the research team who was not involved in the care or observation of the patient being studied"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Galinkin 2000.

Methods	Randomized controlled trial
Participants	9 months‐6 years, weight < 25 kg, ASA I‐II, bilateral myringotomy Exclusion criteria: allergy to fentanyl or midazolam, history suggestive of increased risk of malignant hyperthermia, risk of airway obstruction, neurodevelopmental delay, cardiopulmonary disease, active history of untreated gastro‐oesophageal reflux Recruitment: 265 children randomly assigned, 260 analysed (64, 69, 66 and 61 in each group) Location: USA
Interventions	4 groups: Sevoflurane + nasal fentanyl 2 mcg/kg Sevoflurane + nasal saline placebo Halothane + nasal fentanyl 2 mcg/kg Halothane + nasal saline placebo All participants: oral premedication with acetaminophen 10 mg/kg and midazolam 0.5 mg/kg, study inhalation agent with nitrous oxide for induction and maintenance
Outcomes	Severe EA defined as a score of 4 on the following 4‐point scale: 1 = calm 2 = not calm but consolable 3 = not easily calmed, moderate agitation, restlessness 4 = combative, excited, disoriented, thrashing Other outcomes: pain (CHEOPS), vomiting, adverse respiratory events
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"anesthesiologist, research observer, patient, parents and PACU nurse were all blinded to the intranasal solution administered, which was drawn up by a third party with no further involvement in the study" "research observer, patient, parents and PACU nurse were all blinded to the inhalation agent used"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	5 withdrawals due to protocol violations (unclear to which groups participants were assigned)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ghai 2010.

Methods	Randomized controlled trial
Participants	1‐6 years, ASA I‐II, cataract surgery Exclusion criteria: history of allergy to midazolam, clonidine or bupivacaine; seizure disorder; developmental delay; any neurological disease potentially associated with symptoms of agitation; family history of malignant hyperthermia; or ineffective sub‐Tenon block during surgery (defined as rise in heart rate or blood pressure of 20% above baseline or requirement of > 1.5% sevoflurane) Recruitment: 124 children randomly assigned, 120 analysed (39 group C1, 41 group C2, 40 control group) Location: India
Interventions	Group C1 = clonidine 1 mcg/kg IV Group C2 = clonidine 2 mcg/kg IV Control group: normal saline IV All participants: oral midazolam premedication 0.5 mg/kg, sevoflurane 8% induction in 40:60 mix of oxygen/nitrous oxide, maintenance with sevoflurane 1%‐1.5% with 33% oxygen in nitrous oxide, spontaneous ventilation with LMA, sub‐Tenon block by surgeon 0.08‐0.1 mL/kg 0.5% bupivacaine after proparacaine eyedrops, IV paracetamol 10 mg/kg
Outcomes	EA defined as Pain/Discomfort Score (PDS) ≥ 3 using items 3‐5 only. PDS measured at 15‐minute intervals for 90 minutes. Severe agitation defined as PDS > 6 Other outcomes: parental separation scores, induction scores, time to discharge, incidence of cardiovascular and respiratory depression, nausea and vomiting, intraoperative heart and systolic blood pressure
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated random number"
Allocation concealment (selection bias)	Low risk	"kept in opaque sealed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child, anaesthetist, surgeon all blinded: "the envelope was opened by an independent anaesthetist not involved in the study, who prepared the study drugs"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 exclusions: 1 excluded from group C1 as subtenon block could not be administered because of compromised sclera, 2 excluded from group C2 because of incomplete postoperative data, 1 excluded from control group because of incomplete postoperative data
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ghosh 2011.

Methods	Randomized controlled trial
Participants	1‐5 years, ASA I‐II, elective urogenital and lower limb surgery Exclusion criteria: history of allergic reaction to local anaesthetics, aspirin ingestion in the preceding 4 weeks, history of bleeding disorder, preexisting neurological and spinal disorder and infection at the local site Recruitment: 90 (30 in each of 3 groups) Location: India
Interventions	Clonidine 1 mcg/kg in local anaesthetic caudal block Clonidine 0.75 mcg/kg in local anaesthetic caudal block Control group: no clonidine in local anaesthetic caudal block All participants: no premedication, sevoflurane 4%‐5% induction with 50% nitrous oxide, intubation after rocuronium 1 mg/kg, maintenance with sevoflurane 1% with oxygen/nitrous oxide in 1:2 ratio, no opioids given, caudal block 0.75 mL/kg of 0.25% bupivacaine, reversal with neostigmine/atropine at end of surgery
Outcomes	EA defined as Pain/Discomfort Score (PDS) ≥ 3 using items 3‐5 only Other outcomes: hypotension, bradycardia, respiratory compromise
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"A resident doctor who did not take part in the care of the children prepared the study drugs. The study investigators, as well as the children and their parents were unaware of the contents of the study drug"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"The same investigator, who was blind to the randomisation process, observed all of the patients in PACU"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Gouda 2004.

Methods	Randomized controlled trial
Participants	3‐8 years, ASA I‐II, elective myringotomy Exclusion criteria: any neurological condition, medication that might affect anaesthesia or recovery Recruitment: 40 participants (20 in each group) Location: Egypt
Interventions	Remifentanil 0.25 mcg/kg IV at conclusion of surgical procedure Control group: saline placebo 10 mL IV at conclusion of surgical procedure All participants: sevoflurane with nitrous oxide induction and maintenance, PR acetaminophen 15 mg/kg
Outcomes	EA defined as a score ≥ 4 for ≥ 5 minutes on the following 5‐point scale, despite all calming efforts by the child's parents or guardians and nursing personnel: 1 = no response to stimuli 2 = responsive to stimuli but asleep 3 = awake and responsive 4 = crying and difficult to console 5 = wild thrashing, behaviour requiring restraint Other outcomes: time in PACU, nausea/vomiting, 4‐point pain scale (1 = none, 2 = mild, 3 = moderate, 4 = severe)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomised into 2 groups", method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist was blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"nurse blinded to patient group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Guard 1998.

Methods	Randomized controlled trial
Participants	2‐8 years, ASA I‐II, elective urological procedures as outpatients Exclusion criteria: none stated Recruitment: 50 children (25 in each group) Location: Canada
Interventions	Intervention group: propofol induction and maintenance Control group: sevoflurane induction and maintenance All participants: no premedication, 70% nitrous oxide before IV access then continued throughout, rocuronium and reversal, lumbar or caudal epidural block ± catheter insertion
Outcomes	"Excitation" during emergence reported Other outcomes: haemodynamics, time in PACU, number of participants requiring analgesia in PACU, number of participants with delayed discharge from hospital, adverse anaesthetic or surgical outcomes, caregiver and parental satisfaction
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random number tables"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	children blinded (all received 70% nitrous oxide initially), anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"research nurse who was blinded to the anaesthetic assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of "excitation" during emergence

Guler 2002.

Methods	Randomized controlled trial
Participants	Inclusion criteria: 3‐6 years, ASA I, adenotonsillectomy Exclusion criteria: none stated Recruitment: 40 children (20 in each group) Location: Turkey
Interventions	Intervention group: peritonsillar infiltration of 0.25% bupivacaine 0.5 mL/kg beforesurgical procedure Control group: peritonsillar infiltration of 0.9% saline 0.5 mL/kg before surgical procedure All participants: premedication (medication not stated), sevoflurane 8% induction with 60% nitrous oxide, sevoflurane 2%‐3% maintenance, rocuronium 0.6 mg/kg, intubation, atropine and neostigmine reversal
Outcomes	EA defined as score of 3 or 4 on following 4‐point scale: 1 = calm 2 = crying but can be consoled 3 = crying, cannot be consoled 4 = agitated and thrashing Other outcomes: pain scores, analgesic requirement, vomiting, haemodyamics
Notes	English abstract, full paper in Turkish
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist or surgeon blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Guler 2005.

Methods	Randomized controlled trial
Participants	3‐7 years, ASA I, adenotonsillectomy Exclusion criteria: midazolam premedication due to anxiety Recruitment: 60 participants (30 in each group) Location: Turkey
Interventions	Dexmedetomidine 0.5 mcg/kg IV 5 minutes before end of surgery Control group: saline placebo IV 5 minutes before end of surgery All participants: oral acetaminophen 15 mg/kg premedication, sevoflurane with nitrous oxide induction and maintenance, tracheal intubation facilitated by vecuronium 0.1 mg/kg IV, reversal with 0.05 mg/kg neostigmine and 0.02 mg/kg atropine IV
Outcomes	Severe EA defined as score ≥ 4 on the following 5‐point scale: 1 = sleeping 2 = awake, calm 3 = irritable, crying 4 = inconsolable crying 5 = severe restlessness, disorientation, thrashing around Other outcomes: vomiting, pain (5‐point scale: 1 = no pain, asleep; 2 = mild pain but without distress; 3 = stating that the throat was very painful but not distressed or moderately tearful and consolable; 4 = considerable distress; 5 = screaming and struggling violently)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated randomization program"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "anesthesiologist was blinded to the drug," both given as 5‐mL volume
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded research observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Gurkan 1999.

Methods	Randomized controlled trial
Participants	3‐15 years, ASA I, strabismus surgery Exclusion criteria: allergy to any of the drugs to be used, symptomatic medical illness Recruitment: 40 children (20 in each group) Location: Turkey
Interventions	Intervention group: propofol induction and maintenance Control group: sevoflurane induction and maintenance All participants: nitrous oxide 66% throughout surgery, vecuronium 0.1 mg/kg for intubation, neostigmine and atropine for reversal, acetaminophen 10 mg/kg oral or rectal in PACU for pain
Outcomes	Reported incidence of EA but did not give a definition of EA Other outcomes: incidence of oculocardiac reflex, PONV
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned using a table of random numbers"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded nurse"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Hallen 2001.

Methods	Randomized controlled trial
Participants	3‐8 years, ASA I, elective myringotomy Exclusion criteria: none stated Recruitment: 60 participants (30 in each group) Location: Sweden
Interventions	Halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: premedication with rectal midazolam 0.3 mg/kg and paracetamol 25 mg/kg, nitrous oxide throughout
Outcomes	Postoperative excitation defined as 2 or higher on the following scale: 0 = quiet, calm 1 = crying, not restless 2 = crying and restless 3 = crying, restless, uncontrollable Other outcomes: PONV, pain (4‐point scale: 1 = none, 2 = mild, 3 = moderate, 4 = severe)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random numbers table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor (experienced nurse) blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of postoperative excitation

Hanna 2004.

Methods	Randomized controlled trial
Participants	4‐10 years, ASA I‐II, adjustable strabismus surgery Exclusion criteria: history of neurological disease, mental retardation, muscle disease or previous history of pseudocholinesterase deficiency Recruitment: 30 participants (15 in each group) Location: Egypt
Interventions	Intervention group: propofol induction and maintenance for placement of adjustable sutures, allowed to briefly emerge, then propofol 1 mg/kg for adjustment of sutures Control group: sevoflurane induction and maintenance for placement of adjustable sutures, allowed to briefly emerge, then propofol 1 mg/kg for adjustment of sutures All participants: 50% nitrous oxide in oxygen, mivacurium bolus then infusion, remifentanil 0.5 mcg/kg/min infusion
Outcomes	EA defined as "combative, excited and disoriented behaviour that requires transient physical restraint." Assessment performed after adjustment of sutures Other outcomes: extubation time, PONV
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly divided into two equal groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"independent observer, blinded to the nature of the anaesthetic used"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Hegazy 2009.

Methods	Randomized controlled trial
Participants	2‐8 years, ASA I‐II, elective cranial MRI Exclusion criteria: diagnostic imaging for trauma, those for whom any of the agents investigated were contraindicated, upper respiratory infection, known gastro‐oesophageal disease, craniofacial malformation, intracranial bleeding, hypertension, full stomach Recruitment: 50 children (25 in each group) Location: Egypt
Interventions	Intervention group: thiopental 2‐3 mg/kg IV given after sevoflurane induction Control group: no thiopental All participants: no premedication, sevoflurane 8% induction, IV access, LMA insertion, sevoflurane 1.5%‐2% maintenance in 100% oxygen
Outcomes	EA defined as PAED score ≥ 16 at any time in the first 15 minutes after emergence. PAED scores assessed upon emergence and every 5 minutes for first 15 minutes and mean (SD) peak PAED scores reported Other outcomes: recovery times, heart rate, blood pressure, SpO2, nausea, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"patients were blinded to group assignment," anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"staff collecting data were blinded to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Hosten 2011.

Methods	Randomized controlled trial
Participants	1‐6 years, ASA I, minor surgery below umbilicus (inguinal hernia repair, undescended testis, hypospadias, circumcision) Exclusion criteria: neurological or psychiatric disease, attention deficit disorder, emergency procedure, medical contraindication to performance of caudal block or a previous anaesthesia experience, caudal block judged inadequate (heart rate and/or blood pressure increase > 10% baseline on skin incision, mCHEOPS pain score > 4 in recovery room) Recruitment: 70 children randomly assigned, 67 analysed (34 ondansetron, 33 control) Location: Turkey
Interventions	Ondansetron 0.1 mg/kg IV up to maximum of 4 mg Control group: placebo All participants: premedication with rectal midazolam 0.5 mg/kg 30 minutes previously, sevoflurane 8% induction with 50% nitrous oxide in oxygen, LMA insertion, randomization to ondansetron or placebo, caudal block 0.8 mL/kg 0.25% levobupivacaine with skin incision 15 minutes after block completed, maintenance with sevoflurane 1%‐2% (corrected for stable intraoperative haemodynamics) in 60:40 nitrous oxide/oxygen mixture, awake LMA removal
Outcomes	EA defined as score > 5 on 10‐point scale (TPS) ranging from 1 (calm or asleep) to 10 (worst possible and inconsolable agitation). TPS assessed every 10 minutes for first 30 minutes in recovery room Other outcomes: mYPAS preoperative anxiety scores, mCHEOPS pain scores in recovery, time to eye opening, time to discharge readiness, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomisation was performed immediately before anaesthetic induction by drawing prepared numbers from closed envelopes"
Allocation concealment (selection bias)	Low risk	"closed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"independent observer blinded to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 exclusions: 1 from each group for inadequate caudal block intraoperatively (received fentanyl following skin incision), and 1 exclusion from placebo group for mCHEOPS pain score > 4 in PACU
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Hsieh 1999.

Methods	Randomized controlled trial
Participants	3 months‐1 year, 5‐11 kg, ASA I‐II, day case urology Exclusion criteria: history of cardiac, respiratory, hepatic, renal, neurological or neuromuscular disease Recruitment: 40 participants (20 in each group) Location: Taiwan
Interventions	Halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: nitrous oxide, atropine 0.01 mg/kg and atracurium 0.5 mg/kg IV after induction, ilioinguinal and iliohypogastric blocks before surgery
Outcomes	Postoperative "restlessness and agitation" reported without a formal definition Other outcomes: cough, laryngospasm, breath‐holding and vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random number tables"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"observer who was blinded"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ibacache 2004.

Methods	Randomized controlled trial
Participants	1‐10 years, ASA I, subumbilical surgery Exclusion criteria: chronic or acute intake of sedative or analgesic drug, adverse effect to any study drug, previous anaesthesia, failure of caudal block Recruitment: 90 participants (30 in each of 3 groups) Location: Chile
Interventions	Dexmedetomidine 0.15 mcg/kg IV Dexmedetomidine 0.30 mcg/kg IV Control group: saline placebo All participants: sevoflurane with nitrous oxide induction and maintenance, caudal block 0.5‐1 mL/kg 0.25% bupivacaine, study drug given after induction
Outcomes	EA defined as 3 or higher on the following 4‐point scale: 1 = calm 2 = not calm but easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited, disoriented Other outcomes: haemodynamics, emergence and recovery times, pain scores (CHIPPS)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random numbers generated by a computer"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "administration of anesthesia and study drugs....were made by two investigators blinded as to the study drugs"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"evaluated by a blinded nurse"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Im 2004.

Methods	Randomized controlled trial
Participants	3‐10 years, ASA I‐II, tonsillectomy and adenoidectomy Exclusion criteria: developmental delay, epilepsy, obstructive sleep apnoea, previous history of emergence agitation Recruitment: 60 children (15 in each of 4 groups) Location: Korea
Interventions	Group 1: nil Group 2: ketorolac 0.5 mg/kg Group 3: fentanyl 1 mcg/kg Group 4: ketorolac 0.5 mg/kg + fentanyl 1 mcg/kg All participants: thiopentone 5 mg/kg IV induction, rocuronium 0.6 mg/kg, atropine 0.02 mg/kg, sevoflurane maintenance in oxygen/nitrous oxide, study drug given IV after intubation
Outcomes	EA defined as a score of 3 on following scale: 0 = asleep 1 = calm, crying, can be consoled 2 = cannot be consoled 3 = hysterical, agitated, thrashing around Other outcomes: severe pain on objective pain/discomfort scale, faces scale
Notes	English abstract, full paper in Korean
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double blinded design"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"a blinded observer evaluated each patient"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Inomata 2010.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, elective minor surface surgery Exclusion criteria: airway malformation, clinical evidence of a difficult airway, asthma, any sign of upper respiratory infection, participants taking CNS depressants, CNS disorders including spinal cord dysfunction, developmental delay, autism, participants who received rescue doses of fentanyl after surgery (i.e. complained of pain after surgery or tried to remove dressing after surgery) Recruitment: 150 children randomly assigned, 139 analysed (45 group F1, 48 group F2, 46 control group) Location: Japan
Interventions	Group F1: fentanyl 1 mcg/kg bolus IV, then infusion 0.5 mcg/kg/h Group F2: fentanyl 2 mcg/kg bolus IV, then infusion 1 mcg/kg/h Control group: normal saline All participants: no premedication, sevoflurane 5% induction in oxygen, intubation after bolus of study drug, sevoflurane 1.5%‐2.5% maintenance (titrated to maintain haemodynamics within 20% baseline) in oxygen, local anaesthetic (ropivacaine) field block, awake extubation
Outcomes	PAED scores assessed continuously for 15 minutes after emergence, with the score that lasted the longest considered the agitation score. Reported mean (SD) PAED scores. Also reported frequency of emergence agitation (PAED score > 10) for the 2 intervention groups Other outcomes: intubating conditions, time to extubation, time to recovery, haemodynamics
Notes	Successful correspondence with study author to obtain frequency of agitation (PAED score > 10) in the control group (unpublished data)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly allocated to one of three groups using computer‐generated numbers"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, attending anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"assessed by an anaesthesiologist blinded to the treatment group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	11 exclusions: 4 control group participants, 5 group F1 participants and 2 group F2 participants excluded after administration of fentanyl rescue doses
Selective reporting (reporting bias)	Low risk	reported PAED scores and incidence of EA

Isik 2005.

Methods	Randomized placebo‐controlled trial
Participants	18 months‐10 years, ASA I‐II, cranial MRI Exclusion criteria: none stated Recruitment: 51 children (17 in each of 3 groups) Location: Turkey
Interventions	Group Fi: fentanyl 1 mcg/kg IV at induction of anaesthesia (2 mL saline 10 minutes before end anaesthesia) Group Fe: fentanyl 1 mcg/kg IV 10 minutes before end anaesthesia (2 mL saline at induction of anaesthesia) Group P: placebo control group (2 mL saline at induction and 10 minutes before end of anaesthesia) All participants: no premedication, sevoflurane induction and maintenance with 50% nitrous oxide, IV insertion, LMA
Outcomes	EA defined as a score of 4 or 5 on a 5‐point agitation scale (in Turkish). EA assessed every 5 minutes for first 25 minutes Other outcomes: recovery times, side effects
Notes	English abstract, full paper inTurkish
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly divided into three groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double‐blind"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Isik 2006a.

Methods	Randomized controlled trial
Participants	18 months‐10 years, cranial MRI, no premedication Exclusion criteria: psychological disorder, cognitive delay, chronic or acute sedative drug intake, neurological condition limiting communication Recruitment: 42 participants (21 in each group) Location: Turkey
Interventions	Group D: dexmedetomidine 1 mcg/kg IV Group P: saline placebo All participants: sevoflurane with 50% nitrous oxide induction and maintenance, insertion of IV catheter then study drug given over 2 minutes Location: Turkey
Outcomes	EA assessed on the following 5‐point scale and delirium defined as a score ≥ 4 for ≥ 5 minutes despite calming efforts of parents: 1 = sleeping 2 = awake/calm 3 = irritable, crying 4 = inconsolable crying 5 = severe restlessness, disorientation Other outcomes: haemodynamics, emergence and recovery times, PONV, bronchospasm
Notes	Reported incidence of EA and incidence of delirium. EA incidence used in forest plot of this review
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomization list," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"administration of anesthesia and study drugs....were made by investigators blinded to the study drugs"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Jacob 2012.

Methods	Randomized controlled trial
Participants	2‐7 years, MRI Exclusion criteria: none stated Recruitment: 21 children (10 in sevoflurane group, 11 in propofol group) Location: USA
Interventions	Intervention group: propofol maintenance with supplemental oxygen Control group: sevoflurane maintenance with LMA All participants: sevoflurane induction
Outcomes	EA assessed with PAED scale but no report of EA incidence or PAED scores (stated that "PAED score revealed that ED was more frequent in children in Group S") Other outcomes: in vivo proton MR spectroscopy (1H‐MRS) to detect cerebral metabolites
Notes	Abstract only (conference proceeding)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	not stated
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Unclear risk	EA incidence or PAED scores not reported

Jiang 2010.

Methods	Randomized controlled trial
Participants	Children undergoing cataract surgery; ASA I‐II; normal development; normal cardiovascular, respiratory, liver, kidney and neurological symptoms; no sedatives within 24 hours of surgery; mean (SD) ages: 2.4 (1.3) years in midazolam group and 2.7 (1.7) years in control group Recruitment: 48 children (24 in each group) Location: China
Interventions	Oral midazolam premedication 0.5 mg/kg 30 minutes before operation Control group: oral placebo 30 minutes before operation All participants: sevoflurane 8% induction with nitrous oxide 60%, spontaneous ventilation via LMA, maintenance with sevoflurane 2%‐3% with nitrous oxide 50%
Outcomes	EA defined as a score of 4 or higher on the following 5‐point scale: 1 = asleep 2 = awake and calm 3 = agitated and crying 4 = inconsolable 5 = delirious and inconsolable Other outcomes: emergence time, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	computer‐generated randomization numbers
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	not stated
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	report incidence of EA

Johannesson 1995.

Methods	Randomized controlled trial
Participants	Children, ASA I, ENT surgery, median (range) age: group H 4.2 (1.1‐7.3) years, group S 4.0 (2.1‐7.5) years Exclusion criteria: none stated Recruitment: 40 participants (18 in halothane group, 22 in sevoflurane group) Location: Sweden
Interventions	Group H: halothane induction and maintenance Group S: sevoflurane induction and maintenance All participants: rectal midazolam and atropine premedication, nitrous oxide, intubation, paracetamol 10 mg/kg
Outcomes	Immediate behaviour during emergence from anaesthesia was evaluated according to the following scale: calm; cooperative; crying unhappily; or being turbulent. Those who were crying unhappily and were evaluated as turbulent were classified as excited Other outcomes: haemodynamics, PONV
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	"open randomized study," not stated whether outcome assessor was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Jung 2010.

Methods	Multi‐centre randomized controlled trial
Participants	3‐10 years, ASA I, surgery for strabismus or entropion Exclusion criteria: previous anaesthesia; cardiovascular, pulmonary or neurological disease; current upper respiratory tract infection Recruitment: 95 children "studied," 93 analysed (25 group 1, 23 group 2, 24 group 3, 21 group 4) Location: Korea
Interventions	Group 1: ketamine 1.5 mg/kg IV induction + fentanyl 1.5 mcg/kg IV after induction Group 2: ketamine 1.5 mg/kg IV induction + no fentanyl Group 3: thiopental sodium 5 mg/kg IV induction + fentanyl 1.5 mcg/kg IV after induction Group 4: thiopental sodium 5 mg/kg IV induction + no fentanyl All participants: no premedication, rocuronium 0.4 mg/kg, intubation, maintenance with 2%‐3% sevoflurane (adjusted to maintain stable haemodynamics and adequate anaesthesia) with 50% nitrous oxide in oxygen, ketorolac 0.5 mg/kg, ondansetron 0.1 mg/kg, pyridostigmine/glycopyrrolate reversal, awake extubation
Outcomes	EA defined as score of 3 on following 3‐point scale: 1 = asleep, calm or mildly agitated 2 = moderately agitated or restless but consolable 3 = hysterical, crying inconsolable, or thrashing Other outcomes: preoperative anxiety, time to extubation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned to one of four groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"the state of emergence was evaluated by an anesthesiologist who was blind to the anesthetic technique"
Incomplete outcome data (attrition bias) All outcomes	Low risk	95 children "studied" but outcome data reported for only 93; reasons for exclusion of 2 participants not stated
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kain 2005.

Methods	Randomized controlled trial
Participants	3‐10 years, ASA I‐II, outpatient surgery Exclusion criteria: chronic illness, prematurity, developmental delay Recruitment: 102 participants (52 control group, 50 halothane group) Location: USA
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: induction with study inhalation agent with nitrous oxide, procedure‐specific analgesia protocols
Outcomes	"Emergence behaviour was rated based on a scale developed and validated by Keegan et al" (no other explanation) and reported incidence of ED Other outcomes: postoperative maladaptive behavioural changes (PHBQ, primary outcome), sleep, recovery inventory, pain (Paediatric Pain Measure for Parents)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child and parents blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all research assistants who gathered outcome data were blinded to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kataria 1996.

Methods	Randomized controlled trial
Participants	0‐18 years, ASA I‐II, surgery > 1 hour Exclusion criteria: significant cardiac or respiratory abnormalities; renal, central nervous system or hepatic disease; history of unusual response to halogenated anaesthetics; had received a general anaesthesia in previous 2 weeks; taking medication known to affect MAC, hepatic or renal function Recruitment: 431 participants randomly assigned, 428 analysed (214 in each group) Location: 12 institutions from 6 nations (1 from Austria, 1 from Belgium, 1 from United Kingdom, 2 from Switzerland and 6 from USA)
Interventions	Halothane induction and maintenance Control group: sevoflurane induction and maintenance Preoperative medication (midazolam or fentanyl) was standardized at each centre, study agent was administered with 60%‐70% nitrous oxide, muscle relaxants were used when needed, opioids (fentanyl) were permitted but standardized at each site for both groups, morphine was given postoperatively when needed for pain
Outcomes	Reported incidence of "excitement" on emergence (no definition given) Other outcomes: emergence and recovery times, plasma fluoride levels, adverse effects
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"open‐label," "trained, independent observers"
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 children in the halothane groups were discontinued from the study because of the appearance of cardiac dysrhythmia (supraventricular ectopies and ventricular bigeminy). Administration of halothane was discontinued, and the anaesthetic was changed to isoflurane in both cases, with resolution of dysrhythmia
Selective reporting (reporting bias)	Low risk	reported incidence of "excitement"

Kazak 2010.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I, routine procedures of less than 90 minutes Exclusion criteria: use of sedatives or hypnotics within the previous month, use of theophylline or hepatic enzyme‐inducing drugs, presence of severe central nervous system dysfunction, increased intracranial pressure, malformation of cardiovascular system, hypertonus, hyperthyroidism Recruitment: 60 children (20 in each of 3 groups) Location: Turkey
Interventions	Group M: midazolam oral premedication 0.5 mg/kg orally Group MP: midazolam oral premedication 0.25 mg/kg + parental presence at induction Group P: placebo oral premedication + parental presence at induction All participants: sevoflurane 7% induction in oxygen, rocuronium 0.15 mg/kg, LMA insertion, sevoflurane 1.5%‐2.5% maintenance with 50% nitrous oxide in oxygen, paracetamol 15‐20 mg/kg rectally before surgery started, reversal of muscle relaxant, awake LMA removal
Outcomes	Postanaesthetic recovery was evaluated using FLACC score, Anxiety Scale score, visual analogue scale (VAS) pain score and Observer Pain Scale (OPS) score every 10 minutes Other outcomes: sedation scores preoperatively and postoperatively, Anxiety Scale scores preoperatively, parental satisfaction, parental interviews at 1 and 14 days postoperatively for changes in child's behaviour
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated into three groups by sealed envelope method," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded (placebo premedication), not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"physician in the postanaesthesia care unit blinded to the study groups"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported postoperative behavioural scores

Keaney 2004.

Methods	Randomized controlled trial
Participants	Children, ASA I‐II, day cases, mean (SD) age in months: 39.9 (31.3) sevoflurane, 47.7 (35.3) halothane Exclusion criteria: previous GA, central nervous system or psychological disorders Recruitment: 197 children randomly assigned, 120 analysed (63 sevoflurane group, 57 halothane group) Location: Ireland
Interventions	Halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: nitrous oxide, analgesia (PR diclofenac 1 mg/kg and local anaesthesia by regional technique or local infiltration) was administered as appropriate and normally practised
Outcomes	"Distress on emergence" measured as a score of 3 or higher on following 4‐point scale: 1 = calm and cooperative 2 = mildly anxious but consolable 3 = tearful and inconsolable 4 = totally out of control Other outcomes: Post‐Hospital Behavioural Questionnaire (PHBQ) for behavioural disturbance
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"the emergence score was documented by the recovery room nursing staff, who were blinded to the anesthetic agent"
Incomplete outcome data (attrition bias) All outcomes	High risk	77 withdrawals from study (i.e. 39% of those randomly assigned): 22 excluded because of breach of protocol (not mentioned how many from each group); 55 excluded because of failure to return questionnaire (halothane group n = 25, sevoflurane group n = 30)
Selective reporting (reporting bias)	Low risk	reported emergence scores

Khattab 2009.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, elective dental surgery Exclusion criteria: developmental problems, cerebral palsy, Down syndrome, inborn errors of metabolism, history of epileptic fits, body weight less than 10 kg or greater than 26 kg, known allergy to any medications used, procedures lasting less than 30 minutes or longer than 2 hours Recruitment: 95 children randomly assigned, 46 analysed in each group, 3 exclusions described in risk of bias table Location: Qatar
Interventions	Group M: oral midazolam 0.5 mg/kg premedication Group KM: oral midazolam 0.5 mg/kg + ketamine 2 mg/kg premedication All participants: premedication with study drugs mixed with ibuprofen suspension 10 mg/kg, sevoflurane 8% induction in 100% oxygen, atropine 0.02 mg/kg, fentanyl 1 mcg/kg, cisatracurium 0.15 mg/kg, intubation then maintenance with sevoflurane 1.5%‐2% in 35:65 oxygen/nitrous oxide mix, IV paracetamol 15 mg/kg, neostigmine/atropine reversal, awake extubation
Outcomes	EA assessed with the following 5‐point scale: 1 = obtunded with no response to stimuli 2 = asleep but responsive to movement and stimuli 3 = awake and appropriately responsive 4 = crying and difficult to console 5 = wild thrashing behaviour that requires restraint. EA defined as an agitation score ≥ 4 for longer than 5 minutes despite all calming efforts by the parents Other outcomes: time to emergence, postoperative fentanyl consumption, time to hospital discharge
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned...via computer‐generated random numbers"
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"study drugs were prepared by a trained nurse (who was not involved in any other part of the study) into identical 5 ml‐syringes," "the anesthetist... was blind to the premedication"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all scores and observations were recorded by the anesthesia technician, who was unaware of the type of premedication"
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 exclusions for procedure duration exceeding 2 hours (2 from group M and 1 from group KM)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kim 2009.

Methods	Randomized controlled trial
Participants	3‐10 years, ASA I‐II, adenotonsillectomy Exclusion criteria: airway disease, sleep apnoea, developmental delay, psychological disorder Recruitment: 105 children randomly assigned, 100 analysed (32 group A10, 34 group A20, 34 control group) Location: Korea
Interventions	Group A10: alfentanil 10 mcg/kg Group A20: alfentanil 20 mcg/kg Control group: saline All participants: sevoflurane 8% with nitrous oxide induction and maintenance, ketorolac 1 mg/kg, dexamethasone 0.1 mg/kg, study drug IV after induction
Outcomes	"Severe agitation" defined as 3 or 4 on 4‐point Aono's scale: 1 = calm 2 = not calm but could be easily consoled 3 = moderately agitated 4 = combative, excited or disoriented, thrashing around PAED scale also assessed, with mean and range PAED scores reported Other outcomes: emergence and recovery times, pain scores (CHIPPS), number of participants requiring analgesia in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated using a sealed envelope system," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "an independent researcher prepared the study syringe for each patient"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"the anaesthesiologist who recorded the result was blinded to the patient group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	5 participants excluded: 1 from control group because of breath‐holding, 3 from group A10 (1 because of postoperative bleeding, 1 as the result of systemic epinephrine absorption, 1 as the result of breath‐holding), 1 from Group A20 because of postoperative bleeding
Selective reporting (reporting bias)	Low risk	reported incidence of EA and PAED scores

Kim 2011.

Methods	Randomized controlled trial
Participants	1‐13 years, ASA I‐II, strabismus surgery Exclusion criteria: refusal by parents, neurological disease, developmental delay, history of previous surgery, ASA III‐IV, airway disease Recruitment: 101 children (35 group M, 31 group P, 35 group S) Location: Korea
Interventions	Group M: midazolam 0.05 mg/kg IV at end of anaesthesia Group P: propofol 1 mg/kg IV at end of anaesthesia Group S: saline IV at end of anaesthesia (control group) All participants: atropine 0.01 mg/kg IM premedication, IV induction thiopentone 5 mg/kg, rocuronium 0.6 mg/kg, intubation, maintenance with sevoflurane 2%‐3% with 50% nitrous oxide in oxygen, IV paracetamol 10 mg/kg, awake extubation, parental presence in PACU; 5 minutes before end of surgery, sevoflurane and nitrous oxide were discontinued and study drugs were given
Outcomes	EA defined as score of 3 or 4 over 10 minutes during PACU stay using the following scale: 1 = calm 2 = not calm but could be easily calmed 3 = not easily calmed, moderately agitated or restless 4 = excited or disoriented Other outcomes: time to emergence, number of participants needing fentanyl to treat severe agitation (score of 4), laryngospasm, nausea/vomiting, desaturation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned by means of random numbers generated by a computer"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double‐blind"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"in the PACU, a nurse who was blinded to the patient's study group recorded the degree of emergence agitation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ko 2001.

Methods	Initially randomized controlled trial but no longer randomized after 1 month (all children in intervention group)
Participants	12‐110 months, ASA I‐II, elective outpatient surgery Recruitment: during randomized phase of trial, 43 children randomly assigned (22 control, 21 midazolam). Total of 88 children in study, including non‐randomized phase (22 control, 66 midazolam). Location: Taiwan
Interventions	Midazolam 0.2 mg/kg oral premedication at least 10 minutes before induction Control group: placebo premedication All participants: sevoflurane with nitrous oxide induction and maintenance, atropine 0.01 mg/kg before intubation
Outcomes	EA defined as > 3 on a 10‐point scale (0 = none, 5 = moderate, 10 = worst) If score ≥ 5, IV meperidine 0.5 mg/kg was given. If 10 minutes after M1, agitation still persisted with a score ≥ 5, a second dose of meperidine (0.5 mg/kg) was given. After another 10 minutes, if agitation continued to persist, IV propofol 1 mg/kg was given Other outcomes: time in PACU, parental and PACU nurse satisfaction
Notes	No longer RCT after 1 month; as the authors believed the midazolam group was doing so much better, they decided to give everyone midazolam. Correspondence with study authors revealed that the number of participants in the midazolam group was 21 and the incidence of EA was 11 cases before randomization ceased Entered data for randomized phase only in forest plot
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	study changed from double‐blind (initially both anaesthetist and outcome assessor were blinded, as was major nursing caregiver) to single‐blind part‐way through the study (only outcome assessors remained blinded)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded throughout study
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals for randomized portion of the study (randomized part of study was stopped early because of high incidence of emergence agitation in control group)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kobylarz 2001.

Methods	Randomized controlled trial
Participants	Inclusion criteria: 2 weeks‐14 years, ASA I‐II, various surgeries except cardiothoracic and neurosurgical Recruitment: 102 children (38 halothane, 64 sevoflurane) Location: Poland
Interventions	Intervention group: halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: analgesia or regional anaesthesia technique, subgroups received midazolam/atropine premedication versus no premedication
Outcomes	EA defined as the need for constant care of the emerging child with 1‐to‐1 nursing
Notes	English abstract, full paper in Polish
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	not stated
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kol 2008.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, cranial MRI for diagnosis of convulsion, epilepsy and headache Exclusion criteria: upper respiratory infection, known gastro‐oesophageal disease, craniofacial malformation, intracranial bleeding, hypertension, full stomach Recruitment: 88 children randomly assigned, 79 analysed (37 propofol, 42 sevoflurane) Location: Turkey
Interventions	Intervention group: propofol induction and maintenance Control group: sevoflurane induction and maintenance All participants: parental presence on induction, no premedication, EMLA cream applied, IV cannula inserted awake, randomization after insertion of IV cannula
Outcomes	EA assessed with PAED scale and reported mean (SD) PAED scores Other outcomes: haemodynamics, MRI interruption, emergence times, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random allocations in a ratio of 1:1 in balanced blocks of 8"
Allocation concealment (selection bias)	Low risk	"sequentially numbered, opaque, sealed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	7 in the propofol group and 2 in the sevoflurane group were excluded from data analysis because of protocol violations resulting from lack of recording of clinical data
Selective reporting (reporting bias)	Low risk	reported PAED scores

Koner 2011.

Methods	Randomized controlled trial
Participants	1‐7 years, ASA I‐II, infraumbilical day case surgery Exclusion criteria: allergy to study drugs, psychological or emotional disorders, developmental delay, contraindication to caudal block, caudal block failure (heart rate increase at skin incision) Recruitment: 95 children randomly assigned, 84 analysed (42 in each group) Location: Turkey
Interventions	Group MH: midazolam 0.5 mg/kg and hydroxyzine 1 mg/kg oral premedication 30 minutes before Group M: midazolam 0.5 mg/kg oral premedication 30 minutes before All participants: no parental presence at induction, induction with sevoflurane 7% with nitrous oxide 60% in oxygen, maintenance sevoflurane 2% with 60% nitrous oxide, IV alfentanil 10 mcg/kg, LMA, caudal block 0.25% bupivacaine (0.8 mL/kg up to 16 mL), parental presence in PACU
Outcomes	PAED scale score assessed every 5 minutes for first 30 minutes in PACU EA defined as PAED score ≥ 16 at any time during the 30‐minute follow‐up period Median PAED scores for each group also reported Other outcomes: parental separation score, induction quality, sedation quality (before induction), Ramsay sedation score at 30 minutes in PACU, pain scores in PACU (CHIPPS)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomisation and allocation....was performed using computerised numbers"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	7 withdrawals from group M (5 did not take premedication, 2 delayed transfer to operating room) and 4 from group MH (3 did not take premedication, 1 had caudal block failure)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Konig 2009.

Methods	Randomized controlled trial
Participants	2‐12 years, ambulatory dental surgery under general anaesthesia Exclusion criteria: psychiatric, behavioural or developmental history; medical contraindication to any of the study drugs; history of malignant hyperpyrexia; ASA III or higher Recruitment: 184 children randomly assigned, 179 analysed (91 in sevoflurane group, 88 in propofol group) Location: USA
Interventions	Intervention group: propofol maintenance Control group: sevoflurane maintenance All children: 20 mg/kg paracetamol preoperatively, midazolam 0.5 mg/kg at the discretion of the anaesthetist, sevoflurane induction, nitrous oxide during maintenance, boluses of fentanyl if clinically indicated by heart rate/blood pressure criteria
Outcomes	EA measured on PAED scale every 5 minutes after waking for 30 minutes. Highest score during this period for each of the 5 behaviours was used to compute the final PAED score Reported median PAED scores with interquartile ranges and number of participants with PAED score > 10. Number of participants with PAED score > 10 used for this review Other outcomes: PONV, number of nursing interventions in the recovery room, time to discharge readiness, parental satisfaction
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned by a computer‐generated randomization table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"trained observers were blinded to the anesthetic techniques and intraoperative drugs"
Incomplete outcome data (attrition bias) All outcomes	Low risk	5 participants were excluded from the data analysis (1 from sevoflurane group and 4 from propofol group). One exclusion was the result of equipment failure, and 4 exclusions were due to airway concerns on the part of the responsible anaesthesiologist, leading to the decision not to extubate in a deep plane of anaesthesia
Selective reporting (reporting bias)	Low risk	reported incidence of EA and PAED scores

Kubo 2001.

Methods	Randomized controlled trial
Participants	3 months‐6 years, ASA I‐II, hernia day surgery Exclusion criteria: allergies Recruitment: 110 children randomly assigned, 104 analysed (28 group S, 23 group P, 29 group MS, 24 group MP), 6 withdrawals described in risk of bias table Location: Japan
Interventions	Group S: sevoflurane 2% maintenance with 70% nitrous oxide until end of surgery Group P: propofol maintenance (2.5 mg/kg, then 20 mg/kg/h for 10 minutes, then 10 mg/kg/h, then 5 mg/kg/h when closing wound, ceased at end of surgery) Group MS: as for group S and oral premedication with midazolam 0.5 mg/kg and famotidine 1 mg/kg 30 minutes preinduction Group MP: as for group P and oral premedication with midazolam 0.5 mg/kg and famotidine 1 mg/kg 30 minutes preinduction All participants: sevoflurane induction with 70% nitrous oxide, regional block 0.5 mg/kg bupivacaine 0.25%
Outcomes	EA assessed on the following 3‐point scale: 1 = none—child is lying quietly with no crying 2 = mild—occasional movement and/or crying; no need for restraint 3 = marked—thrashing and/or needs restraint and/or constant crying Other outcomes: recovery scores (Steward scores), pain scores, PONV, time to oral intake, time to discharge home
Notes	English abstract, full paper in Japanese
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	6 excluded because they required "top‐up" medication for light anaesthesia (1 from group S, 5 from group P)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kulka 2001a.

Methods	Randomized controlled trial
Participants	2‐7 years , ASA I‐II, penile operations Exclusion criteria: mental retardation, neurological disease, insufficient pain therapy with penile block in previous operations or in PACU, behavioural difficulties Recruitment: 46 children randomly assigned, 40 analysed (20 in each group) Location: Germany
Interventions	Intervention group: midazolam 0.1 mg/kg IV before termination of anaesthesia Control group: saline placebo All participants: sevoflurane induction and maintenance in 2:1 nitrous oxide in oxygen, penile block, paracetamol 15 mg/kg PR
Outcomes	Agitaton defined as non‐purposeful movement that requires intervention from staff. Severe agitation defined as requirement for midazolam in PACU Other outcomes: PONV, pain scores
Notes	English abstract, full paper in German
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	opaque envelopes
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double‐blind placebo controlled"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"double‐blind"
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 exclusions from each group due to insufficient penile block
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Kulka 2001b.

Methods	Randomized controlled trial
Participants	Aged 2‐7 years, ASA I‐II, undergoing circumcision Exclusion criteria: mental retardation, history of agitation, any other neurological disease, inappropriate pain therapy Recruitment: 49 children randomly assigned, 40 analysed (20 in each group) Location: Germany
Interventions	Intervention group: clonidine 2 mcg/kg IV 5 minutes after start of surgery Control group: saline placebo All participants: midazolam premedication 0.5 mg/kg orally 30 minutes before induction, sevoflurane with nitrous oxide induction and maintenance penile block (0.1 mL/kg 0.5% bupivacaine), paracetamol 15 mg/kg suppository
Outcomes	EA defined as "unpurposeful movement requiring restraint." If midazolam required, episode was classed as severe Other outcomes: PONV, pain and discomfort scores
Notes	Correspondence with study author confirmed that 20 participants were analysed in each group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double blind," "the anesthesiologist was blinded to the drug"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"evaluation was performed by a second investigator, who was blinded to the patient's treatment schedule"
Incomplete outcome data (attrition bias) All outcomes	High risk	9 participants excluded from analysis (6 from clonidine group and 3 from placebo group). Reasons for exclusion were increase in heart rate and blood pressure, indicating ineffective regional penile block (5 clonidine group, 3 placebo group); and reported pain in the area that had been operated on (1 from clonidine group)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Lankinen 2006.

Methods	Randomized controlled trial
Participants	1‐7 years old, ASA I‐II, outpatient adenoidectomy Exclusion criteria: allergy to study medications, asthma, cardiovascular disease Recruitment: 75 children (26 control group, 25 tropisetron group, 24 clonidine group) Location: Finland
Interventions	Tropisetron 0.1 mg/kg IV Clonidine 1.5 mcg/kg IV Control group: saline placebo All participants: sevoflurane induction and maintenance in 100% oxygen, alfentanil 20 mcg/kg IV, rectal diclofenac 1 mg/kg, study drug given after induction
Outcomes	EA defined as a score of 3 or higher on the Modified Pain/Discomfort scale Other outcomes: emergence and discharge times, number of participants requiring analgesia, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number assignment"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"the study was double‐blind and placebo‐controlled"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all observers, as well as the children and their parents, were unaware of the contents of the study drug"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Lapin 1999.

Methods	Randomized controlled trial
Participants	6 months‐6 years, ASA I‐II, myringotomy and tube insertion Exclusion criteria: none stated Recruitment: 104 children randomly assigned, 100 analysed (28 in sevomidaz group, 24 in each of the other 3 groups) Location: USA
Interventions	Halo group: halothane Halomidaz group: oral midazolam 0.5 mg/kg + halothane Sevo group: sevoflurane Sevomidaz group: oral midazolam 0.5 mg/kg + sevoflurane All participants: induction and maintenance with study inhalational agent with nitrous oxide, rectal paracetamol 15‐30 mg/kg
Outcomes	EA defined as PACU nurse answering "yes" to the question: "Did this patient become uncontrollable to the point that you had to restrain the patient and you could not perform any other functions?" Other outcomes: recovery and discharge home times
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"the recovery room nurses and discharging nurses were blinded to the anesthetic technique"
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 participants excluded because of problems with premedication
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Le Berre 2001.

Methods	Randomized controlled trial
Participants	2 months‐6 years, ASA I‐II, non‐ambulatory elective surgical procedures below umbilicus, orthopaedic surgery of lower limb and surgery requiring completion time of at least 90 minutes Exclusion criteria: none stated Recruitment: 40 children (20 in each group) Location: France
Interventions	Intervention group: isoflurane maintenance Control group: sevoflurane maintenance All participants: rectal midazolam 0.2 mg/kg 60 minutes before induction, sevoflurane induction with 60% nitrous oxide, caudal block for analgesia
Outcomes	EA defined as "disoriented behaviour characterised by inconsolable agitation" Other outcomes: recovery and discharge times, postoperative vomiting
Notes	Correspondence with study author to clarify how many participants in isoflurane group experienced agitation. Study author replied that no participants who received isoflurane showed agitation
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"a single‐blinded observer, who was not informed about the halogenated agent the patient had received"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported EA incidence

Lee 2010a.

Methods	Randomized controlled trial
Participants	2‐14 years, ASA I‐II, adenotonsillectomy or adenoidectomy Exclusion criteria: cognitive or developmental disorders Recruitment: 93 children randomly assigned, 90 analysed (30 in each of 3 groups) Location: Republic of Korea
Interventions	Group K0.25: ketamine 0.25 mg/kg IV 10 minutes before end of surgery Group K0.5: ketamine 0.5 mg/kg IV 10 minutes before end of surgery Control group: saline 10 minutes before end of surgery All participants: atropine 0.01 mg/kg IM premedication 30 minutes before, thiopentone 5 mg/kg IV induction, rocuronium 0.6 mg/kg, intubation, sevoflurane 1.5%‐3% with 50% nitrous oxide maintenance, pyridostgimine 0.2 mg/kg glycopyrrolate 0.008 mg/kg reversal, awake extubation
Outcomes	EA assessed on arrival to PACU, then 5‐minutely on the following 4‐point scale: 1 = asleep 2 = awake but calm 3 = agitated but consolable 4 = severely agitated and difficult to console Highest score during PACU stay recorded. Study authors have not defined EA, but we have used incidence of a score of 4 in the forest plot of this review Other outcomes: extubation time, pain scores (modified CHEOPS), PONV, laryngospasm
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"allocated randomly to one of three groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"the experimental drugs were administered by physicians not participating in the anesthesia"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"emergence agitation was observed by nurses who were blinded to the patients' groups"
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 withdrawals (1 from each group) for laryngospasm after extubation
Selective reporting (reporting bias)	Low risk	reported EA scores

Lee 2010b.

Methods	Randomized controlled trial
Participants	3‐8 years, ASA I, adenotonsillectomy without myringotomy Exclusion criteria: mental or neurological disease, use of sedative medication Recruitment: 90 children randomly assigned, 88 analysed (44 in each group) Location: Republic of Korea
Interventions	Propofol group: propofol 1 mg/kg at end of surgery Control group: saline 0.1 mL/kg at end of surgery All participants: thiopental sodium 1 mg/kg IV before entering theatre, induction with thiopental sodium 5 mg/kg IV and atracurium 0.5 mg/kg, intubation, maintenance with sevoflurane 2%‐2.5% with 50% nitrous oxide, ketorolac 1 mg/kg IV, awake extubation, parental presence in PACU
Outcomes	EA assessed with 2 scales simultaneously (Aono's 4‐point scale and PAED scale) at 5, 15 and 30 minutes after emergence. EA defined as score of 3 or 4 on the 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = moderately agitated or restless and not easily calmed 4 = combative, excited or disoriented, thrashing around Time point of 5 minutes used in forest plot PAED scale used to assess severity of EA and reported mean (SEM) PAED scores at 3 time points Other outcomes: pain scores (CHIPP scale), Ramsay sedation scores after emergence, emergence time, time to discharge from PACU, incidence of PONV
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated randomization"
Allocation concealment (selection bias)	Low risk	"sealed envelopes which were opened on the day of the scheduled operation"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 exclusion from propofol group due to intraoperative ST depression on EKG, 1 exclusion from saline group due to bleeding in the surgical region after extubation
Selective reporting (reporting bias)	Low risk	reported incidence of EA and PAED scores

Lee 2011.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I‐II, elective subumbilical surgery (inguinal herniorrhaphy, hydrocoelectomy or orchidopexy) Exclusion criteria: known history of a neurological or psychological disorder, history of previous anaesthesia, presence of sleep apnoea, participants in whom intubation or LMA insertion failed on the first attempt Recruitment: 168 children (56 in each of 3 groups) Location: Republic of Korea
Interventions	Interventions: ETT removed awake, ETT removed deep, LMA removed deep All participants: no premedication, PPIA, sevoflurane induction 8% in nitrous oxide/oxygen (3/1) mixture, IV access, glycopyrrolate 0.004 mg/kg, insertion of airway device as per randomization, fentanyl 2 mcg/kg IV, ondansetron 0.1 mg/kg, maintenance with sevoflurane in air/oxygen mixture to maintain HR and BP within 20% of preinduction values and BIS of 40‐60, PPEA
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited or disoriented Other outcomes: behaviour on induction (3‐point scale), duration of anaesthesia (induction to eye opening), length of stay in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated randomization table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, treating anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"emergence agitation was evaluated by one anaesthesiologist, who was unaware of each patient's group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Lerman 1996.

Methods	Randomized controlled trial
Participants	1‐12 years, ASA I‐II, elective ambulatory surgery Exclusion criteria: none stated Recruitment: 375 children randomly assigned, 374 analysed (124 halothane, 250 sevoflurane) Location: USA and Canada, multi‐centre (5 centres) phase 3 trial for sevoflurane
Interventions	Halothane induction and maintenance Control group: sevoflurane induction and maintenance All participants: un‐premedicated, 60% nitrous oxide, caudal block or local anaesthetic wound infiltration, fentanyl given according to haemodynamic/respiratory parameters but not within 30 minutes of the conclusion of surgery
Outcomes	EA defined as involuntary movement of 1 or more extremities during recovery from anaesthesia Other outcomes: modified Aldrete score, Objective Pain Discomfort score, induction and emergence characteristics, child or caregiver satisfaction
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned to receive either sevoflurane or halothane in a 2:1 ratio using a computer‐generated randomization schedule"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"recorded by a nurse, who was blinded to the treatment administered"
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 child from halothane group excluded from analysis because of severe airway obstruction and hypoxia caused by a mucous plug
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Li 2011.

Methods	Randomized controlled trial
Participants	3‐11 years, ASA I‐II, adenotonsillectomy Exclusion criteria: history of airway disease, sleep apnoea, developmental delay, psychological disorders, those crying on arrival to the operating theatre Recruitment: 105 children randomly assigned,100 analysed (32 sufentanil, 34 fentanyl, 34 control) Location: China
Interventions	Sufentanil 0.2 mcg/kg IV Fentanyl 2 mcg/kg IV Control group: saline placebo All participants: sevoflurane 8% induction with 60% nitrous oxide, study drug given 1 minute after loss of the eyelash reflex, intubation, sevoflurane maintenance 1.5%‐2.5%, tramadol 2 mg/kg, dexamethasone 0.1 mg/kg
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale (Aono's scale): 1 = calm 2 = not calm but could be easily consoled 3 = moderately agitated or restless and not easily calmed 4 = combative, excited or disoriented, thrashing around Also measured and reported median (range) PAED scores Other outcomes: time to extubation, time in PACU, pain scores (CHIPPS)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated using a sealed envelope system," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "an independent researcher prepared the syringe for each patient and the anesthesiologist who recorded the result was blinded to the patient group"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"an independent researcher prepared the syringe for each patient and the anesthesiologist who recorded the result was blinded to the patient group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	5 withdrawals after randomization (3 from sufentanil group—1 postoperative bleeding, 1 systemic epinephrine absorption, 1 breath‐holding; 1 from fentanyl group—postoperative bleeding; 1 from control group—breath‐holding)
Selective reporting (reporting bias)	Low risk	reported incidence of EA and PAED scores

Liao 2010.

Methods	Randomized controlled trial
Participants	1‐4 years, ASA I‐II, undergoing rigid bronchoscopy for removal of tracheal or bronchial foreign body Exclusion criteria: known allergy to any anaesthetic agent, family history of malignant hyperthermia, coagulopathy, legal guardian's refusal to sign consent Recruitment: 64 children (32 in each group) Location: China
Interventions	Group TIVA: propofol target controlled‐infusion (TCI) with remifentanil 0.05‐1 mcg/kg/min Group VIMA: sevoflurane induction 8% and maintenance 2.5%‐3.5% All participants: no premedication, BIS 40‐60, lignocaine 4% spray to airway, dexamethasone 0.1 mg/kg
Outcomes	"Excitement after procedure" was recorded Other outcomes: haemodynamics throughout procedure, induction and emergence times, intubating conditions, breath‐holding, cough, laryngospasm, desaturation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"allocated randomly," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique but blinded to maintenance, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Liao 2011.

Methods	Randomized controlled trial
Participants	3‐12 years, ASA I‐II, elective urological outpatient surgery Exclusion criteria: history of premature delivery; reported developmental delay; deafness; significant cardiovascular, respiratory, neurological disease; receiving medication known to affect the central nervous system Recruitment: 160 children (54 in SP group, 52 in BIS group, 54 in AAI group) Location: Taiwan
Interventions	BIS (bispectral index) group: sevoflurane anaesthesia controlled by BIS 40‐60 AAI (A‐line autoregressive index) group: sevoflurane anaesthesia controlled by AAI 15‐30 SP (standard practice) group: sevoflurane anaesthesia controlled by clinical parameters All participants: no premedication, PPIA, sevoflurane induction with 50% nitrous oxide, LMA insertion, sevoflurane maintenance (concentration control determined by group allocation) with spontaneous ventilation with 50% oxygen in air, fentanyl 1 mcg/kg 5 minutes before incision
Outcomes	PAED scores, assessed every 5 minutes after awakening for 30 minutes Other outcomes: recovery time (primary outcome), incidence of PONV, parental satisfaction, sevoflurane consumption
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"allocated randomly by computer‐generated randomization table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded (monitors on all participants), trained investigators (first and second anaesthesiologists) blinded to anaesthetic techniques and grouping of participants
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"third investigator, who was also unaware of the grouping of the patient, was responsible for the assessment of the patient during the emergence"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported PAED scores

Lili 2012.

Methods	Randomized controlled trial
Participants	3‐7 years, ASA I‐II, vitreoretinal surgery Exclusion criteria: asthma, cardiovascular disease, abnormal renal or hepatic function, any known sensitivity to study medication, previous sedative or analgesic medication use, history of past anaesthesia Recruitment: 60 children (30 in each group) Location: China
Interventions	Dexmedetomidine 0.5 mcg/kg IV Control group: saline placebo All participants: no premedication, sevoflurane 8% induction in oxygen, IV insertion, atropine 0.01 mg/kg IV, propofol 2 mg/kg, remifentanil 0.5 mcg/kg, cistracurium 0.15 mg/kg to facilitate intubation, study drug given over a 10‐minute period after induction of anaesthesia, maintenance of anaesthesia with sevoflurane 1%‐2% and remifentanil 0.2 mcg/kg/min in oxygen with BIS maintained at 40‐60
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = calm but could be easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited, or disoriented Other outcomes: pain (CHIPPS scale), extubation time, emergence time, intraoperative blood pressure and heart rate, intraocular pressure, airway events and incidence/severity of coughing after extubation, oxygen desaturation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"allocated randomly (by computer‐generated random numbers)"
Allocation concealment (selection bias)	Low risk	"study drugs were prepared by the hospital pharmacy in identical containers marked with consecutive numbers"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"patients and investigators were unaware which was dexmedetomidine or normal saline (double‐blind)"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Lin 2009.

Methods	Randomized controlled trial
Participants	1‐6 years, ASA I‐II, bilateral myringotomy and tube placement as a first procedure Exclusion criteria: neurological disease, developmental delay, analgesia or sedatives within 36 hours before surgery Recruitment: 60 children (30 in each group) Location: USA
Interventions	Intervention group: acupuncture at 2 points: LI‐4(he‐gu) and HT‐7(shen‐men) points. Manual stimulation for 10 seconds and needles kept in situ for 10 minutes Control group: no acupuncture All participants: no premedication, sevoflurane induction with 70% nitrous oxide in oxygen followed by addition of sevoflurane
Outcomes	EA measured on the following 4‐point scale and reported median (range) scores: 1 = asleep/calm 2 = mildly agitated but easily consolable 3 = moderately agitated or restless and inconsolable 4 = hysterical, crying inconsolably or thrashing Other outcomes: pain scores (CHEOPS), number of participants requiring analgesia in PACU, vomiting, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	blinded observer in PACU evaluating pain and agitation
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported EA scores

Lopez Gil 1999.

Methods	Randomized controlled trial
Participants	6‐144 months, ASA I‐II, surgery below umbilicus Exclusion criteria: risk of aspiration, requiring high‐pressure ventilation Recruitment: 120 participants (60 in each group) Location: Spain
Interventions	Propofol induction (3 mg/kg) and maintenance (5 mg/kg/h) Control group: sevoflurane induction and maintenance All participants: premedication with oral midazolam 0.5 mg/kg, IV access before induction, nitrous oxide during induction and maintenance, fentanyl bolus 3 mcg/kg followed by an infusion 1 mcg/kg/h. If surgery to last longer than 30 minutes, atracurium 0.5 mg/kg and pressure‐controlled ventilation. Postoperative analgesia with metimazole 40 mg/kg
Outcomes	Reported number of participants who were "agitated" on emergence. No definition of EA given Other outcomes: time to LMA insertion/removal, haemodynamics, oxygen saturation, emergence times, number of participants requiring analgesia
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	child not blinded to induction technique, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	outcome assessor of EA blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Makharita 2009.

Methods	Randomized controlled trial
Participants	3‐8 years, ASA I‐II, bilateral myringotomy and tube placement Exclusion criteria: known allergy to study medicines; bronchial asthma, cardiovascular disease, psychological disorder or cognitive delay or any neurological condition that would limit a parent's ability to communicate with, or understand, nursing personnel; treatment with sedatives or anticonvulsants; extremely distressed and uncooperative children; parental refusal Recruitment: 80 children (40 in each group) EA Location: Egypt
Interventions	Fentanyl 1 mcg/kg IV 10 minutes before end of anaesthesia Control group: saline placebo All participants: rectal acetaminophen 40 mg/kg 90‐120 minutes before induction of anaesthesia, parental presence at induction, sevoflurane induction and maintenance
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but easily calmed 3 = moderately agitated or restless 4 = excited or disoriented Other outcomes: recovery time, time to hospital discharge, adverse events
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number assignment"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double blind," "the patients randomly received the prepared syringe (either IV saline or fentanyl) in a blinded fashion, the researcher did not know the solution nature"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"a research observer who was blinded to the treatment groups"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Malviya 2006.

Methods	Randomized controlled trial
Participants	2‐10 years, healthy children, otological or pulse dye laser procedures Exclusion criteria: cardiovascular disease or neurological impairment, sensitivity to clonidine or ketorolac or requirement for premedication with midazolam Recruitment: 133 children randomly assigned, 120 analysed (59 clonidine group, 61 placebo group)
Interventions	Clonidine 2 mcg/kg IV Control group: saline placebo All participants: oral acetaminophen preoperatively, sevoflurane with nitrous oxide induction, maintenance with isoflurane, IV placement after induction and then random assignment to receive clonidine or placebo, ketorolac given for analgesia Location: USA
Outcomes	EA assessed on the following 4‐point scale: 0 = quiet, calm 1 = mildly agitated but consolable 2 = moderately agitated, non‐purposeful and inconsolable 4 = severely agitated PACU nurse was asked to subjectively qualify whether the agitation was indicative of delirium Other outcomes: blood pressure, emergence time, time in PACU; parent telephone interview regarding sleepiness, agitation, aggression, anxiety, fear or nightmares in first 24 hours after surgery
Notes	Sevoflurane for induction only, maintenance with isoflurane
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"table of random numbers"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "all care providers were blinded to group assignment"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all care providers were blinded to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	13 cases were excluded (unclear from which groups they came)—6 required premedication; 2 had medical exclusions; in 3 cases, IV cannulation was not obtained; 1 procedure was more extensive than planned and 1 was cancelled (Note: 5 participants in the clonidine group and 6 participants in the placebo group refused oral acetaminophen preoperatively but were included in the final analysis)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Meena 2009.

Methods	Randomized controlled trial
Participants	6 months‐3 years, ASA I‐II undergoing cleft lip/palate repair Exclusion criteria: other associated congenital anomalies, major systemic illness, previous history of agitation following anaesthesia, hypersensitivity to anaesthetic drugs Recruitment: 60 children (30 in each group) Location: India
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: oral midazolam 0.5 mg/kg premedication 1 hour before the procedure, induction and maintenance with study agent and 60% nitrous oxide, fentanyl 1 mcg/kg IV, rocuronium and reversal
Outcomes	Reported incidence of postoperative "excitement/restlessness" Other outcomes: haemodynamics, recovery times and scores, number of participants requiring analgesia in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated random number code"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"observed in the post‐operative period by an independent observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Meng 2012.

Methods	Randomized controlled trial
Participants	5‐14 years, ASA I‐II, tonsillectomy Exclusion criteria: known history of allergy to study drugs, history of previous surgery; participants receiving other analgesics such as non‐steroidal anti‐inflammatory drugs; participants receiving α‐methyldopa, clonidine or a beta‐blocker; psychiatric disorder, neurological or neuromuscular disorders; those who had used drug, alcohol or tobacco 2 weeks before the study began Recruitment: 120 children (40 in each of 3 groups) Location: China
Interventions	Dexmedetomidine 0.5 mcg/kg IV loading dose, then infusion 0.2 mcg/kg/h Dexmedetomidine 1 mcg/kg IV loading dose, then infusion 0.4 mcg/kg/h Control group: lactated Ringer's placebo All participants: midazolam 40 mcg/kg IV premedication, IV induction with propofol 1.2‐2 mg/kg, sufentanil 0.4 mcg/kg, cisatracurium 0.15 mg/kg, intubation, maintenance with sevoflurane 1.5%‐2.5% and remifentanil infusion, BIS maintained 45‐55, tropisetron 0.1 mg/kg, neostigmine and atropine reversal
Outcomes	EA defined as a score of 3 or 4 on the following scale: 1 = awake and calm, cooperative 2 = crying, requiring consoling 3 = irritable, restless, screaming, inconsolable 4 = combative, disoriented, thrashing. EA score recorded at 5‐minute intervals for first 30 minutes, then at 10‐minute intervals until 60 minutes. Data in forest plot represent time point 5 minutes after extubation Other outcomes: heart rate, blood pressure, extubation time, time in PACU, pain scores (VAS), sedation scores
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomized using a computer‐generated random numbers table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double blind," "all study drugs were prepared by an anesthesiologist, who was blinded to the details of the study"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all postoperative observations and scores were performed by the same anesthesiologist, who was blinded to the group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Meyer 2007.

Methods	Randomized controlled trial
Participants	1‐6 years, ASA I‐II, minor surgery below umbilicus Exclusion criteria: neurological disease, psychiatric disease and attention deficit disorder Recruitment: 63 children randomly assigned, 59 analysed (30 sevoflurane group, 29 isoflurane group) Location: USA
Interventions	Isoflurane maintenance Control group: sevoflurane maintenance All participants: oral midazolam 0.4 mg/kg premedication 30 minutes preoperatively; IV induction with thiopentone, alfentanil and mivacurium; caudal block; rectal paracetamol; awake extubation. If IV access failed after 2 attempts, inhalational induction performed (2 children in isoflurane group and 4 in sevoflurane group)
Outcomes	EA defined as > 5 on a 10‐point scale and ED as > 7 on a 10‐point scale where 1 = calm/asleep and 10 = worst possible/inconsolable Other outcomes: postoperative anxiety, time in PACU, number of participants requiring analgesia, caregiver and parental satisfaction, PONV
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"according to randomization," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"independent observer blinded for the group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 participants excluded (not specified which groups)—2 because of insufficient caudal anaesthesia, 1 because of fentanyl given on induction and 1 as the result of additional surgical intervention on the face
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Michalek‐Sauberer 1998.

Methods	Randomized controlled trial
Participants	2‐16 years, ASA I‐III, elective surgical procedure > 1 hour duration Exclusion criteria: none stated Recruitment: 42 children (21 in each group) Location: Austria
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: midazolam premedication (1 mg/kg rectal or 0.01 mg/kg IV for older children) 15 minutes before induction, induction and maintenance with study agent with nitrous oxide, vecuronium used to facilitate intubation at the discretion of anaesthetist, analgesia provided in IV increments of fentanyl 2‐3 mcg/kg or by a regional anaesthetic technique
Outcomes	Reported incidence of mild, moderate and severe excitement on emergence (moderate/severe excitement on emergence used in forest plot in this review) Other outcomes: haemodynamics, emergence times, modified Aldrete scores, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"a sealed envelope was opened to allocate patients randomly," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	"patients were observed by an independent, non‐blinded observer throughout the study period"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Mikawa 2002.

Methods	Randomized controlled trial
Participants	2‐11 years, undergoing minor (urological, ophthalmological, otological and orthopaedic) surgery Exclusion criteria: none stated Recruitment: 175 children (35 in each of 5 groups) Location: Japan
Interventions	Group CLO‐2: oral clonidine 2 mcg/kg (up to 150 mcg) Group CLO‐4: oral clonidine 4 mcg/kg (up to 150 mcg) Group MID: 0.5 mg/kg (up to 10 mg) Group DIA: 0.4 mg/kg (up to 10 mg) Control group: placebo All participants: premedication with study drug 30‐60 minutes before anaesthesia, sevoflurane with nitrous oxide induction and maintenance, rectal diclofenac 12.5 mg or 25 mg given immediately after induction of anaesthesia
Outcomes	EA assessed on following 4‐point scale: 0 = asleep 1 = calm 2 = agitated but consolable 3 = severely agitated and inconsolable
Notes	Attempted to contact study authors regarding randomization, no reply
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The word "randomized" was not reported, but prospective study with equal numbers in each group, ethics approval and informed participant consent, so randomization assumed
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"we blindly recorded agitation score"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals reported
Selective reporting (reporting bias)	Low risk	reported agitation scores

Milic 2010.

Methods	Randomized controlled trial
Participants	All children scheduled for elective cleft lip and palate surgical repair (3 months‐10 years) Exclusion criteria: contraindications for surgery, infection, anaemia, asthma, severe heart defects, mental retardation Recruitment: 140 children randomly assigned (76 midazolam anaesthesia group, 64 sevoflurane anaesthesia group) Location: Croatia
Interventions	Midazolam anaesthesia group: IV induction with midazolam 0.05 mg/kg, fentanyl 0.005 mg/kg, vecuronium 0.1 mg/kg, then maintenance with IV boluses of midazolam (0.025 mg/kg every 45 minutes), fentanyl (0.001 mg/kg depending on heart rate) and vecuronium (0.01 mg/kg every 40 minutes, or if increase in ETCO2), reversal with atropine and neostigmine Sevoflurane anaesthesia group: sevoflurane induction 5%‐8% and sevoflurane maintenance 0.8%‐1% in oxygen/air mixture, IV fentanyl 0.005 mg/kg, then additional boluses of 0.001 mg/kg depending on heart rate All participants: oral midazolam premedication, sevoflurane 5%‐8% "preinduction" for IV cannulation, then sevoflurane ceased and randomly assigned to either group, 2% lidocaine infiltration by surgeon
Outcomes	EA assessed over 2 hours in PACU on following scale: 0 = child sleeps peacefully 1 = agitation Study authors stated: "only cases in which agitation persisted after the administration of analgesics were included in the analysis" (agitation that disappeared after analgesia was considered pain‐related) Other outcomes: intraoperative complications (difficult intubation, ventricular asystole, bronchospasm), PONV, perioperative oxygen saturation, heart rate, haemoglobin and haematocrit
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	"it was not possible to achieve postoperative blinding," "the anesthesiologists carried out the anesthesia and collected the data"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Mizrak 2013.

Methods	Randomized controlled trial
Participants	5‐15 years, ASA I‐II, adenotonsillectomy Exclusion criteria: hypertension, psychiatric disorders, drug allergy, cardiovascular and clotting disorders, peptic ulcers, those who could not use visual analogue scales, adenoidectomy alone Recruitment: 60 children (30 in each group) Location: Turkey
Interventions	Intervention group: dexmedetomidine IV 0.5 mcg/kg over 10 minutes after induction of anaesthesia Control group: saline placebo All participants: midazolam 0.05 mg/kg IM premedication 45 minutes before induction, sevoflurane induction and maintenance with 50% nitrous oxide, rocuronium 0.5 mg/kg, intubation
Outcomes	EA assessed on the following 5‐point scale with median (standard error of the mean) scores reported at the 15th minute postoperatively: 1 = sleeping 2 = awake and calm 3 = irritable and crying 4 = inconsolable crying 5 = severe restlessness and disorientation, purposelessly wanting to get out of bed, wanting to stand on the bed, shouting, crying or mumbling aloud Other outcomes: haemodynamics, intraoperative blood loss, operation time, recovery time, pain scores (VAS), analgesic requirement, sedation scores, Hb, INR, APTT
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double blind," "syringes prepared by another anesthetist"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"double blind"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported agitation scores

Moore 2003.

Methods	Randomized controlled trial
Participants	3‐12 years, day case general or ENT surgery Exclusion criteria: history of allergic or other serious adverse experiences with anaesthesia, severe cardiovascular/metabolic or CNS disease, anticipated airway difficulties, if regimen was expected to include succinylcholine Recruitment: 347 children randomly assigned, 322 analysed (163 control group, 159 intervention group) Location: United Kingdom
Interventions	Intervention group: halothane maintenance (after propofol induction) Control group: sevoflurane maintenance (after sevoflurane induction) All participants: no premedication, nitrous oxide, rectal diclofenac 12.5‐25 mg, paracetamol if unsuitable for diclofenac, IV opioids/local infiltration/regional block depending on procedure
Outcomes	Reported incidence of "recovery mental state" according to the following 3 categories: alert and awake drowsy agitated or distressed Other outcomes: PONV, adverse events during induction and recovery, recovery and discharge times
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random number sequence"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded to maintenance agent, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"recovery room nurse was asked to judge the patients mental state in recovery," not stated whether they were blinded to group assignment
Incomplete outcome data (attrition bias) All outcomes	Low risk	25 withdrawals: 15 for protocol violation, 5 for cancellation of operation, 5 for withdrawal of consent
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Murray 2002.

Methods	Quasi‐randomized controlled trial
Participants	< 7 years old, myringotomy tube placement Exclusion criteria: none stated Recruitment: 130 children (33 in halothane/oxycodone and halothane groups, 32 in sevoflurane/oxycodone and sevoflurane groups) Location: USA
Interventions	Halothane/oxycodone group: halothane anaesthesia with oxycodone 0.1 mg/kg preinduction Halothane group: halothane anaesthesia with no oxycodone Sevoflurane/oxycodone group: sevoflurane anaesthesia with oxycodone 0.1 mg/kg preinduction Sevoflurane group: sevoflurane anaesthesia with no oxycodone
Outcomes	EA defined as score of 4 or 5 on following 5‐point scale: 1 = sleeping 2 = awake/calm 3 = irritable 4 = inconsolable 5 = agitated, combative, restless Other outcomes: number of children with EA for > 10 minutes, PONV, discharge times
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	"assigned to receive either no premedication or oxycodone," "then alternately assigned to either a sevoflurane or halothane inhalation anaesthetic"
Allocation concealment (selection bias)	High risk	"alternately assigned"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"nurse scoring the patient's behaviour was blinded to the preoperative medication and the inhalation anaesthetic"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Na 2013.

Methods	Randomized controlled trial
Participants	2‐6 years, ASA I‐II, adenotonsillectomy Exclusion criteria: sleep apnoea, developmental delay, psychological disorder, any neurological disorder Recruitment: 84 children (42 in each group) Location: Republic of Korea
Interventions	Intervention group: maintenance with sevoflurane 1% and remifentanil (1 mcg/kg bolus at induction, then infusion 0.5 mcg/kg/min until intubated, then 0.25 mcg/kg/min) Control group: maintenance with sevoflurane 2%‐3% without remifentanil All participants: no premedication, IV induction with thiopental 5 mg/kg and rocuronium 0.6 mg/kg, ketorolac 1 mg/kg, dexamethasone 0.15 mg/kg, nitrous oxide 60% during maintenance, peritonsillar 2% lignocaine injection
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but easily calmed 3 = moderately agitated or restless 4 = combative, excited, disoriented Also assessed EA with PAED scale and reported median (IQR) PAED scores and incidence of PAED scores ≥ 10 (also used as a second definition of EA by authors) Other outcomes: duration of EA, time to emergence, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated, randomised group allocation"
Allocation concealment (selection bias)	Low risk	"sealed‐envelope method"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"in the PACU, one anaesthesiologist blinded to the patient group evaluated EA"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Naito 1991.

Methods	Randomized controlled trial
Participants	1‐7 years, pulse dye laser for port wine stain Exclusion criteria: none stated Recruitment: 30 children (15 in each group) Location: Japan
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: induction and maintenance with study agent and nitrous oxide
Outcomes	Incidence of postoperative restlessness and agitation was reported. No further definition given Other outcomes: emergence and recovery times, vomiting, complications
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"allocated randomly," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"assessed by the nurse, who was not informed which anaesthetic had been used"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Nakayama 2007.

Methods	Randomized controlled trial
Participants	2‐11 years, ASA I‐II, otorhinolaryngological procedures Exclusion criteria: known allergy to study drugs, sleep apnoea, developmental delay, psychological disorders Recruitment: 186 children randomly assigned, 176 analysed (89 sevoflurane, 87 propofol) Location: Japan
Interventions	Control group: sevoflurane induction and maintenance Intervention group: propofol induction and maintenance
Outcomes	EA defined as a score of 3 or 4 on following 4‐point scale: 1 = calm, 2 = not calm but could easily be consoled 3 = not easily calmed, moderately agitated or restless 4 = combative, excited or disoriented Other outcomes: emergence time, time in PACU, haemodynamics, PONV, laryngospasm, preschool versus school‐aged incidence of EA
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"during the recovery period, an independent anesthesiologist, who was blinded to the anaesthetic used, recorded all the observations and measurements"
Incomplete outcome data (attrition bias) All outcomes	Low risk	10 children excluded because of insufficient analgesia (6 propofol, 4 sevoflurane)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Oh 2005.

Methods	Randomized controlled trial
Participants	1‐7 years, ASA I‐II, elective urological surgery Exclusion criteria: history of sleep apnoea, developmental delay, psychological disorder, any neurological disorder Recruitment: 85 children (43 in group G, 42 in group I) Location: Republic of Korea
Interventions	Group G: gradual cessation of sevoflurane at end of surgery (0.1% per minute) Group I (control): immediate cessation of sevoflurane at end of surgery All participants: no premedication; IV induction with thiopental, atropine, rocuronium; sevoflurane maintenance 2%‐2.5% with 50% nitrous oxide, pyridostigmine and glycopyrrolate reversal
Outcomes	EA defined as grade 3 or 4 on following 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = moderately agitated or restless 4 = combative, excited or disoriented Other outcomes: time to BIS 70/80/90, time to first movement, time to extubation, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated numbers"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"an observer blinded to the patient group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ou 2011.

Methods	Randomized controlled trial
Participants	2‐11 years, ASA I‐II, strabismus surgery Exclusion criteria: none stated Recruitment: 60 children (30 in each group) Location: China
Interventions	Intervention group (TIVA): IV induction ketamine 1.5 mg/kg, midazolam 0.1 mg/kg, propofol 1 mg/kg, vecuronium 0.1 mg/kg, remifentanil 2 mcg/kg, intubation, maintenance with propofol 3‐6 mg/kg/h with remifentanil 0.1‐0.125 mcg/kg/min Control group: sevoflurane induction, midazolam 0.1 mg/kg IV, vecuronium 0.1 mg/kg IV, intubation, sevoflurane maintenance 2%‐3%
Outcomes	Reported incidence of EA but no definition for EA Other outcomes: time to extubation, time to emergence, incidence of oculocardiac reflex, PONV
Notes	English abstract, full paper in Chinese No ketamine in control (sevoflurane) group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly divided," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ozcengiz 2011.

Methods	Randomized controlled trial
Participants	3‐9 years, ASA I‐II, oesophageal dilatation procedures Exclusion criteria: lack of consent; known adverse reactions to dexmedetomidine, midazolam or melatonin; mental retardation, developmental delay, neurological or psychiatric illness that may be associated with agitation Recruitment: 100 children (25 in each group) Location: Turkey
Interventions	Group MD: midazolam 0.5 mg/kg oral premedication Group D: dexmedetomidine 2.5 mcg/kg oral premedication Group ML: melatonin 0.1 mg/kg oral premedication Group P: placebo oral premedication All participants: premedication (according to randomization) given with paracetamol to be easily drinkable 40‐45 minutes before induction, sevoflurane 8% induction and 2%‐4% maintenance with 50% nitrous oxide, IV vecuronium for intubation, no supplemental analgesic, atropine and neostigmine reversal
Outcomes	EA defined as score of 3 or 4 on the following scale: 1 = awake and calm, cooperative 2 = crying, requiring consoling 3 = irritable/restless, screaming, inconsolable 4 = combative, disoriented, thrashing EA scale assessed on arrival to PACU and at 5, 10, 15, 30 and 60 minutes after arrival to PACU Other outcomes: heart rate, blood pressure, oxygen saturation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomization was performed using a table random method"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"for the purpose of creating double‐blind conditions, neither the researcher who attended clinical applications and observations nor the parents were informed as to which drug was administered"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all patients were evaluated by the same researcher who was unaware which drug was administered"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ozer Kocak 2001.

Methods	Randomized controlled trial
Participants	Children aged 6.3 ± 1.6 years, ASA I‐II, adenoidectomy with bilateral myringotomy and insertion of tubes and/or tonsillectomy Exclusion criteria: none stated Recruitment: 30 children (15 in each group) Location: Turkey
Interventions	Group I: TIVA propofol (3 mg/kg/h) and remifentanil (0.5 mcg/kg/min) with 50% oxygen in air/oxygen mix Group II: sevoflurane 2.5%‐3% maintenance with 50% nitrous oxide in oxygen All participants: IV cannula inserted awake after EMLA cream, IV midazolam premedication 0.03‐0.05 mg/kg, IV induction with remifentanil 1 mcg/kg and propofol 2 mg/kg, cisatracurium, intubation, rectal paracetamol, neostigmine and atropine reversal
Outcomes	EA assessed on the following 3‐point scale: 1 = asleep or calm 2 = mildly agitated, crying but consolable 3 = hysterical, crying inconsolably
Notes	Inconsistencies in data between abstract and results section of full paper. Emailed study author to clarify with no response. In the forest plot of this review, used incidence of EA reported in the abstract
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"nurse, blinded to the anaesthetic technique, assessed the quality of emergence and postoperative agitation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Ozturk 2009.

Methods	Randomized controlled trial
Participants	2‐6 years, elective fibreoptic bronchoscopy Exclusion criteria: haemoptysis, previous fibreoptic bronchoscopy Recruitment: 50 children (25 in each group) Location: Turkey
Interventions	Remifentanil 1 mcg/kg bolus over 2 minutes, then infusion 0.15 mcg/kg/min Control group: no remifentanil All participants: premedication with oral midazolam 0.5 mg/kg and nebulized 4% lignocaine 4 mg/kg, sevoflurane induction and maintenance in 100% oxygen, IV access after induction, then given IV atropine 10 mcg/kg
Outcomes	EA defined as a score ≥ 4 on the following scale: 1 = sleeping 2 = awake, calm and cooperative 3 = crying, requiring consoling 4 = irritable/restless, screaming, inconsolable 5 = combative, disoriented, thrashing Measured at emergence, 5, 10, 15 minutes Other outcomes: thoracic wall rigidity during bolus, coughing, emergence and recovery times
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized by sealed envelope," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"persons scoring the children for agitation and cough were unaware of group allocation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Patel 2010.

Methods	Randomized controlled trial
Participants	2‐10 years, ASA II‐III, elective tonsillectomy and adenoidectomy Exclusion criteria: known allergy to alpha‐2 agonists, developmental delay, cardiac and craniofacial abnormalities, anxiety disorder, chronic disabilities or pain syndrome and use of psychotherapeutic medications, beta‐blockers, digoxin, cimetidine, alpha‐2 agonists, anticonvulsants or psychotropic medications Recruitment: 137 children randomly assigned, 122 analysed (61 in each group) Location: USA
Interventions	Intervention group: dexmedetomidine IV 2 mcg/kg over 10 minutes (as soon as IV access obtained) followed by 0.7 mcg/kg/h until 5 minutes before end of surgery Control group: fentanyl IV 1 mcg/kg as soon as IV access obtained All participants: no premedication, sevoflurane 8% induction with 60% nitrous oxide in oxygen, IV access, rocuronium 0.6 mg/kg to facilitate intubation, maintenance with 1 MAC sevoflurane and 60% nitrous oxide as long as BIS remained below 60 (if BIS reached 60 or higher, sevoflurane increased to reduce BIS to below 60), IV dexamethasone 0.5 mg/kg (maximum 10 mg), rectal acetaminophen 30‐40 mg/kg (maximum 1000 mg) before start of surgery, fentanyl 0.5‐1 mcg/kg for increase in heart rate or systolic blood pressure to above 30% baseline, glycopyrrolate and neostigmine reversal
Outcomes	EA evaluated on 2 scales (5‐point scale described by Cole, PAED scale) at the same time intervals (arrival in PACU, every 5 minutes for first 15 minutes, then every 15 minutes for next 2 hours). "Severe EA" was defined as a score of 4 or 5 on the 5‐point scale. Incidence of "severe EA" on arrival to PACU used in the forest plot of this review. Also reported percentage of participants with PAED score 10 or above Other outcomes: intraoperative heart rate and blood pressure, time to emergence, time to extubation, Objective Pain Scale (OPS) score, morphine rescue in PACU (incidence, dosage), desaturation episodes below SpO2 95%
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"random number table was used to assign subjects"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "anesthesiologists and data collectors in operating room (OR) were not blinded"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"observers in the postanesthesia care unit (PACU) were blinded to treatment groups"
Incomplete outcome data (attrition bias) All outcomes	Low risk	15 withdrawals for the following reasons: surgery was cancelled for 2 participants, 1 refused to participate after enrolling, 1 had an intraoperative complication, 11 participants who completed the study had deviation from study protocol or incomplete data
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Picard 2000.

Methods	Randomized controlled trial
Participants	3‐10 years old, ASA I‐II, elective tonsillectomy Exclusion criteria: children screaming before arrival to operating room Recruitment: 50 children randomly assigned, 46 analysed (24 control group, 22 propofol group) Location: Switzerland
Interventions	Propofol induction and maintenance Control group: sevoflurane induction and maintenance All participants: nitrous oxide 60% during maintenance, alfentanil 20 mcg/kg and atracurium 0.5 mg/kg to facilitate intubation, rectal paracetamol 20 mg/kg and ibuprofen 10 mg/kg, local infiltration with bupivacaine 2 mg/kg
Outcomes	EA defined as score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm 3 = moderate agitation, restless 4 = combative, excited, disoriented Other outcomes: emergence time, time in PACU, PONV, laryngospasm, bronchospasm
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"a random numbers table was used to assign children"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique but blinded to maintenance, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"all observations and measurements were recorded by an independent anaesthesiologist, who was blinded to the anaesthetic given"
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 withdrawals: 2 excluded because of postoperative bleeding and 2 because of missing data
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Pieters 2010.

Methods	Randomized controlled trial
Participants	3‐7 years, ASA I‐II, elective tonsillectomy and adenoidectomy Exclusion criteria: presence of a genetic syndrome, allergy to study medications, behavioural disorders, use of psychiatric medications, non‐English speaking families, participants requiring sedative medication before going to the operating room Recruitment: 42 children randomly assigned, 38 analysed (19 in each group) Location: USA
Interventions	Intervention group: propofol maintenance 150‐300 mcg/kg/h Control group: sevoflurane maintenance 1.5%‐4% All participants: sevoflurane induction with nitrous oxide, fentanyl 2 mcg/kg IV, then intubation, mechanical ventilation with air/oxygen mix, maintenance agent titrated to clinical effect, ondansetron 0.1 mg/kg, dexamethasone 0.2 mg/kg, awake extubation
Outcomes	PAED scores assessed 1 minute after awake extubation, then every 10 minutes, for a total of 30 minutes. Emergence delirium defined as PAED score ≥ 16. Dichotomous data in forest plot are for PAED score ≥ 16 at any time in PACU. Also reported median (range) and maximum PAED scores Other outcomes: pain scores (CHEOPS), extubation time, time in PACU, fentanyl dose in PACU, Aldrete score post extubation, nursing and parental satisfaction, PONV, hospital length of stay, anaesthetic complications, subsequent emergency room admissions
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated numbers in blocks of four"
Allocation concealment (selection bias)	Low risk	"sequentially numbered opaque envelopes opened immediately prior to entering OR"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"postoperatively agitation and pain were measured by blinded study personnel"
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 excluded from sevoflurane group after recruitment because of administration of preoperative midazolam; 2 excluded from propofol group—1 because of administration of diphenhydramine in the PACU after a cutaneous reaction to an antibiotic, and another because of loss of the IV cannula in the PACU
Selective reporting (reporting bias)	Low risk	reported incidence of emergence delirium

Rampersad 2010.

Methods	Randomized controlled trial
Participants	1‐5 years, ASA I‐II, bilateral myringotomy procedures Exclusion criteria: any procedure in addition to the bilateral myringotomy procedure, chronic medical issues (ASA III or higher), developmental delay making agitation and pain scoring harder, any allergy or contraindication to study medications, receiving opioid medications or NSAIDs on a long‐term basis Recruitment: 229 randomly assigned, 228 analysed (75 group AF, 76 group AK, 77 group A) Location: USA
Interventions	Group AF: intranasal fentanyl 1 mcg/kg after induction but before start of surgery Group AK: intramuscular ketorolac 1 mg/kg after induction but before start of surgery Group A (control): no fentanyl or ketorolac All participants: oral midazolam 0.5 mg/kg premedication, sevoflurane and nitrous oxide for induction and maintenance by mask, rectal acetaminophen 40 mg/kg after induction but before start of surgery
Outcomes	EA defined as score ≥ 3 on the following 4‐point scale: 1 = calm 2 = crying but can be consoled 3 = crying, cannot be consoled 4 = agitated and thrashing around Study authors state that no statistically significant differences were noted between the 3 groups for the incidence of EA at any time point during recovery, but they do not report the incidence of EA in each of the 3 groups (therefore no forest plot data for this study) Other outcomes: pain scores (CHEOPS)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"standard randomization was carried out by the pharmacy," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sequentially numbered envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"observer remained blinded as to group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 participant withdrawn from study before the start of surgery, as treating surgeon determined that an additional procedure was required
Selective reporting (reporting bias)	Low risk	reported EA scores

Rieger 1996.

Methods	Randomized controlled trial
Participants	2‐10 years, ASA I‐II, adenoidotomy, otomicroscopy and myringotomy Exclusion criteria: fever > 38 or purulent nasal secretions (all children had a mild upper respiratory infection) Recruitment: 41 children (19 halothane group, 22 sevoflurane group) Location: Germany
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: rectal premedication with midazolam 0.5 mg/kg and atropine 0.02 mg/kg, inhalational induction and maintenance with study agent, endotracheal intubation following suxamethonium 1.5 mg/kg
Outcomes	Reported incidence of "excitation" on emergence Other outcomes measured: pain scores, time in PACU, modified Aldrete score, complications and side effects
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned, by a computer‐generated list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"recovery was characterized and evaluated by an independent observer," not stated whether observers were blinded to anaesthetic agent
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Saadawy 2009.

Methods	Randomized controlled trial
Participants	1‐6 years, male, ASA I, having unilateral hernia/orchidopexy Exclusion criteria: contraindication to caudal, such as bleeding diathesis, preexisting neurological/spinal surgery or abnormality of sacrum, local anaesthetic allergy or conduction problems Recruitment: 60 children (30 in each group) Locations: Egypt and Saudi Arabia
Interventions	Intervention group: caudal block with bupivacaine + dexmedetomidine 1 mcg/kg Control group: caudal block with bupivacaine All participants: propofol 3‐4 mg/kg induction, LMA insertion, sevoflurane with nitrous oxide maintenance
Outcomes	EA defined as score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but easily calmed 3 = not easily calmed, moderately agitated or restless 4 = combative, excited or disoriented Other outcomes: haemodynamics, sedation scores, postoperative analgesia requirement, side effects such as vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "the study drugs were randomly prepared by an anesthetist, who was not involved in the study, in an unlabelled syringe, and handed to the anesthesiologist, who was blind to the identity of the drug"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"observer‐blinded study"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Salik 2011.

Methods	Randomized controlled trial
Participants	6 months‐13 years, ASA I‐III, neurointerventional radiology procedures Exclusion criteria: none stated Recruitment: 28 participants (14 in each group) Location: USA
Interventions	Dexmedetomidine 1 mg/kg loading dose over 10 minutes, followed by infusion 0.4‐0.7 mg/kg/h beginning 1 hour before end of procedure and continuing for 1 hour of recovery in the PACU/PICU Control group: normal saline placebo All participants: sevoflurane and nitrous oxide induction, maintenance with sevoflurane in oxygen/air mix with remifentanil IV infusion, IV hydromorphone before extubation
Outcomes	EA assessed using 5‐point scale (Cole) and PAED scale. Do not report incidence of EA or agitation/PAED scores, so no data suitable for forest plots. "Statistical significance was seen at the P < 0.05 level for the Cole scale between 10 and 30 minutes, the period of greatest EA" Other outcomes: heart rate, blood pressure, excessive sedation
Notes	Abstract only (conference publication)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double‐blind"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"double‐blind"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Unclear risk	do not report actual incidence of EA nor agitation scores

Sato 2010.

Methods	Randomized controlled trial
Participants	1‐9 years, ASA I‐II, same‐day or overnight stay surgery Exclusion criteria: mental retardation, neurological or heart disease, uncontrollable asthma, any type of acute illness Recruitment: 81 children (39 intervention group, 42 control group) Location: Japan
Interventions	Intervention group: IV dexmedetomidine 0.3 mcg/kg after induction in 5 mL saline over 10 minutes Control group: saline placebo All participants: no premedication, parental presence during sevoflurane 8% induction in oxygen, IV access, LMA insertion, sevoflurane 2%‐5% maintenance in air/oxygen mix to provide stable heart rate, blood pressure and spontaneous ventilation, rectal acetaminophen 40 mg/kg or diclofenac 1 mg/kg, ropivacaine infiltration (except for laser irradiation or myringotomy with tube insertion), deep LMA removal, parental presence in PACU
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = calm 2 = not calm but could be easily calmed 3 = not easily calmed, moderately agitated, or restless 4 = combative, excited, or disoriented Other outcomes: heart rate, blood pressure, pain (CHIPPS), parental satisfaction, adverse events, modified Aldrete score, time to PACU discharge
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned...using a randomization list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "attending anesthesiologists, surgeons and nurses were blinded to the treatment assignment"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"observer blinded to the patient assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Seo 2011.

Methods	Randomized controlled trial
Participants	4‐10 years, ASA I‐II, extraocular muscle surgery for strabismus Exclusion criteria: intense preanaesthetic anxiety or crying Recruitment: 260 randomly assigned, 250 analysed (60 group SS, 60 group SL, 65 group BS, 65 group BL), 10 withdrawals described in risk of bias table Location: Republic of Korea
Interventions	Group SS: sevoflurane, subtenon saline injection Group SL: sevoflurane, subtenon lignocaine injection Group BS: propofol/remifentanil, subtenon saline injection Group BL: propofol/remifentanil, subtenon lignocaine injection All participants: no oral premedication, IV thiopental sodium 2‐3 mg/kg sedation in the preanaesthetic waiting room just before parental separation, IV induction with propofol 2‐2.5 mg/kg, rocuronium 0.8 mg/kg, intubation, adjustment of sevoflurane or propofol/remifentanil to maintain heart rate and blood pressure within 20% baseline during maintenance, nitrous oxide during maintenance
Outcomes	EA defined as score of 4 or higher on the following 5‐point scale (Cole): 1 = asleep 2 = awake and calm 3 = irritable or consolable crying 4 = inconsolable crying 5 = severe restlessness EA assessed on arrival to PACU and at 10‐minute intervals for first 30 minutes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐derived randomization list"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"two well trained [observers], who were blinded to which group the patient was assigned"
Incomplete outcome data (attrition bias) All outcomes	Low risk	10 withdrawals because of incomplete data collection (5 from group SS, 5 from group SL)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Shen 2012.

Methods	Randomized controlled trial
Participants	2‐5 years, ASA I‐II, elective electronic cochlear implantation Exclusion criteria: current upper respiratory tract infection, asthma, mental disease, congenital disease Recruitment: 50 children (25 in each group) Location: China
Interventions	Remifentanil 0.02‐0.05 mcg/kg/min (titrated to maintain adequate spontaneous ventilation) Control group: saline All participants: no premedication, sevoflurane 8% induction in oxygen, fentanyl 2 mcg/kg with propofol 1 mg/kg to facilitate intubation, ondansetron 0.15 mg/kg, dexamethasone 0.2 mg/kg, sevoflurane adjusted to MAC of 1.3 for maintenance of anaesthesia, extubation at sevoflurane MAC 1.3 in control group whereas sevoflurane reduced to 1.0 MAC 10 minutes before extubation in remifentanil group
Outcomes	EA defined as PAED score > 10 Other outcomes: respiratory rate, end‐tidal CO2, incidence of use of oral airway after deep extubation, time to emergence, time in PACU, PONV, coughing
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"anesthesia assistant blinded to the group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Shibata 2005.

Methods	Randomized controlled trial
Participants	Children, ASA I, minor otolaryngological surgery, mean (SD) ages: 6 (2) and 7 (2) years in 2 groups Exclusion criteria: none stated Recruitment: 20 children (10 in each group) Location: Japan
Interventions	Nitrous oxide washout group: sevoflurane discontinued at completion of surgery but nitrous oxide continued until BIS reached 80 Control group: sevoflurane and nitrous oxide discontinued at completion of surgery All participants: no premedication, sevoflurane and nitrous oxide induction and maintenance, local anaesthesia on the surgical field
Outcomes	EA measured on a 3‐point scale: 1 = asleep or calm 2 = mildly agitated, crying but consolable, restless 3 = hysterical, crying inconsolably, thrashing
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded nurse observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported EA scores

Shin 2001.

Methods	Randomized controlled trial
Participants	4‐7 years, tonsillectomy Exclusion criteria: none stated Recruitment: 29 children (16 ibuprofen group, 13 placebo group) Location: Republic of Korea
Interventions	Intervention group: oral ibuprofen 5 mg/kg 60 minutes preoperatively Control group: oral placebo All participants: oral chloral hydrate 50 mg/kg premedication 60 minutes preoperatively, sevoflurane with nitrous oxide induction and maintenance, vecuronium 0.1 mg/kg, fentanyl 1 mcg/kg after intubation
Outcomes	Incidence of EA reported, with EA defined as "frenzy, tumultuous, very worried or upset and showing this in their behaviour, movement or voice" (Korean translation) Other outcomes: pain scores (VAS 1‐5), Aldrete scores, vomiting, postoperative bleeding
Notes	English abstract, full paper in Korean
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"assigned into two groups," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	blinded outcome assessor
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Shukry 2005.

Methods	Randomized controlled trial
Participants	1‐10 years old, ASA I‐II, elective outpatient surgical procedure Exclusion criteria: known adverse reaction to dexmedetomidine, mental retardation, developmental delay, neurological/psychiatric illness Recruitment: 50 children randomly assigned, 46 analysed (23 in each group)
Interventions	Dexmedetomidine infusion 0.2 mcg/kg/h (commenced 5 minutes after securing airway and ceased 15 minutes after arrival to PACU) Control group: saline placebo infusion All participants: sevoflurane induction and maintenance in oxygen to achieve BIS of 40‐60, intraoperative fentanyl at blinded anaesthetist's discretion, parental presence on emergence Location: USA
Outcomes	ED defined as a score of 3 or 4 on the following 4‐point scale for > 3 minutes: 0 = asleep 1 = calm 2 = crying, consolable 3 = crying, unconsolable 4 = agitated, thrashing Other outcomes: haemodynamics, time to extubation, time in PACU, pain scores
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomized by a computer‐generated program"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "blinded anaesthesia team"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	4 withdrawals: 1 due to breach of protocol, 1 due to IV dislodgement, 1 due to history of severe ED, 1 due to history of seizures
Selective reporting (reporting bias)	Low risk	reported incidence of ED

Singh 2009.

Methods	Randomized controlled trial
Participants	1‐12 years, ASA I‐II, elective surgery for spinal dysraphism Exclusion criteria: previous spinal surgery; cardiac, respiratory, renal, hepatic disease; hydrocephalus, Chiari malformation, history of seizures Recruitment: 80 children (40 in each group) Location: India
Interventions	Isoflurane maintenance Control group: sevoflurane maintenance All participants: sevoflurane with nitrous oxide and maintenance, rocuronium 1 mg/kg, fentanyl 2 mcg/kg IV with further 1 mcg/kg repeated every hour and last dose about 15 minutes before end of surgery
Outcomes	EA was evaluated by using 3 points of the Objective Pain Scale. "If the child was crying inconsolably, thrashing, and was hysterical, he or she was reported to be 'agitated'" Ohter outcomes: haemodynamics, emergence times, modified Aldrete scores, nausea/vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomized according to computer‐generated sequence of numbers"
Allocation concealment (selection bias)	Low risk	"sealed envelope technique"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Singh 2012.

Methods	Randomized controlled trial
Participants	4 months‐7 years, ASA I‐II, elective subumbilical surgery Exclusion criteria: active airway disease, sleep apnoea, developmental delay, psychological disorder, neurological disorder, cardiovascular abnormality or requirement of postoperative ventilation Recruitment: 75 children (25 in each of 3 groups) Location: India
Interventions	Group I: isoflurane maintenance Group D: desflurane maintenance Group S (control group): sevoflurane maintenance All participants: no premedication, sevoflurane 8% induction in 100% oxygen, IV access, fentanyl 2 mcg/kg IV, atracurium 500 mcg/kg IV, dexamethasone 0.5 mg/kg IV (maximum 8 mg), maintenance with study drug (1.0‐1.2 MAC) with nitrous oxide, caudal block 0.2% bupivacaine and rectal paracetamol 30 mg/kg before surgical incision. If heart rate/blood pressure rose by 20%, then caudal block considered ineffective and excluded from study
Outcomes	EA defined as PAED score ≥ 16 Other outcomes: time to extubation, time to emergence, time in PACU, time to readiness for discharge home, pain scores (FLACC), adverse events
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"double‐blinded"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"double‐blinded"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Sun 2009.

Methods	Randomized controlled trial
Participants	4‐7 years, ASA I‐II, adenotonsillectomy Exclusion criteria: sensitive to aspirin Recruitment: 45 children (15 in each of 3 groups) Location: China
Interventions	Intervention groups: Group 1: diclofenac 12.5 mg rectal immediately after intubation Group 2: diclofenac 12.5 mg rectal at end of surgery Control group: no diclofenac All participants: oral midazolam 0.5 mg/kg premedication, IV induction with midazolam 0.1 mg/kg, atropine, ketamine 2 mg/kg, remifentanil 0.5 mcg/kg, vecuronium 100 mcg/kg, intubation, sevoflurane maintenance (1 MAC) with nitrous oxide 50%
Outcomes	EA assessed with PAED scale at 10, 20, 30 minutes after arrival to PACU and report median (range) PAED scores at each time point. Do not report incidence of EA Other outcomes: extubation time, time in PACU, pain scores, modified Aldrete scores
Notes	English abstract, full paper in Chinese
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly divided," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, not stated whether anaesthetist blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported PAED scores

Sun 2010.

Methods	Randomized controlled trial
Participants	Children, ASA I‐II, adenotonsillectomy, mean (SD) ages: 5.8 (1.5), 6.1 (1.5), 6.5 (1.6) years in 3 groups Exclusion criteria: none stated Recruitment: 90 children (30 in each of 3 groups) Location: China
Interventions	Intervention groups: Tramadol group: tramadol 1 mg/kg IV 30 minutes before end of operation Tramadol + propofol group: tramadol 1 mg/kg IV 30 minutes before end of operation and propofol 1 mg/kg at end of operation Control group: saline 0.1 mL/kg IV 30 minutes before end of operation All participants: oral midazolam premedication 0.5 mg/kg (maximum 15 mg) 20‐30 minutes preoperatively, sevoflurane 7% induction with 60% nitrous oxide, IV access, vecuronium 0.1 mg/kg, intubation, sevoflurane 1.5%‐2% maintenance with nitrous oxide 50%
Outcomes	EA assessed with PAED scale every 10 minutes after arrival to PACU and report median (range) PAED scores. Do not report incidence of EA Other outcomes: time to extubation, time in PACU, modified Aldrete scores, pain scores (CHIPPS), PONV
Notes	English abstract, full paper in Chinese
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly divided," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	not stated whether outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported PAED scores

Sury 1996.

Methods	Randomized controlled trial
Participants	6 months‐6 years, ASA I‐II, day case general surgical, urological or orthopaedic Exclusion criteria: history of epilepsy Recruitment: 40 children (20 in each group) Location: United Kingdom
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: no sedative or anxiolytic premedication, induction and maintenance with study agent and nitrous oxide, analgesia appropriate to surgery soon after induction (local anaesthetic block and/or diclofenac suppository)
Outcomes	Incidence of "excitement" on emergence reported Other outcomes: emergence and recovery times, complications
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	"the observer had been present during induction and was therefore not blinded to the choice of vapour"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Tang 2009.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I‐II, cleft lip and/or palate surgery Exclusion criteria: weight deviating by more than 20% from average, obstructive sleep apnoea, psychological disorders, any congenital problem, difficult airway, developmental delay Recruitment: 54 children (26 intervention group, 28 control group) Location: China
Interventions	Intervention group: "multimodal analgesia" with fentanyl 0.5 mcg/kg 10 minutes before end of surgery and rectal paracetamol (2‐4 years given 120 mg, 5‐7 years given 325 mg) just after intubation Control group: no fentanyl 10 minutes before end and no paracetamol All participants: no premedication, sevoflurane 5% induction with 50% nitrous oxide, IV access, fentanyl 2 mcg/kg IV, atracurium 0.6 mg/kg, intubation, local anaesthesia (bupivacaine + lidocaine) to surgical site by surgeon at beginning of surgery, sevoflurane 2%‐2.5% maintenance with 50% nitrous oxide, ondansetron 0.1 mg/kg
Outcomes	EA defined as a score of 4 or 5 on the following scale: 1 = slow, not responding to stimulus 2 = drowsy but responds 3 = awake, calm, cooperative 4 = crying, difficult to console 5 = agitated, needing restraint Other outcomes: sedation, Objective Pain Scale, time to readiness for PACU discharge, adverse effects
Notes	English abstract, full paper in Chinese
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly allocated," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"an observer who was blinded"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Tazeroualti 2007.

Methods	Randomiszed controlled trial
Participants	1‐6 years, ASA I‐II, circumcision Exclusion criteria: family history of malignant hyperthermia, mental retardation, any neurological disease with potential for agitation Recruitment: 68 children randomly assigned, 60 analysed (20 midazolam group, 21 clonidine 2 mcg/kg, 19 clonidine 4 mcg/kg) Location: Belgium
Interventions	Clonidine 2 mcg/kg oral premedication Clonidine 4 mcg/kg oral premedication Control group: midazolam 0.5 mg/kg oral premedication All participants: sevoflurane with nitrous oxide for induction and maintenance, penile block (bupivacaine 0.5% 0.3 mL/kg), rectal paracetamol 30 mg/kg
Outcomes	Modified Objective Pain Scale used to measure agitation: 3 of 5 items used to assess agitation, with total score ≥ 3 defined as agitation Other outcomes: emergence time
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"...drew the sealed envelope," i.e. drawing lots
Allocation concealment (selection bias)	Low risk	"an anaesthesiologist not involved in the clinical protocol prepared the randomization envelopes," "sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "the anaesthetist who administered the premedication was excluded from the perioperative management and postoperative assessment of the child'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"independent observer"
Incomplete outcome data (attrition bias) All outcomes	Low risk	8 children excluded (unclear to which groups assigned): 7 for ineffective penile block at the time of incision and 1 because of a study protocol violation
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Tesoro 2005.

Methods	Randomized controlled trial
Participants	ASA I‐II, day surgery cases, mean (SD) ages: 3.2 (1.4) years (no age limit set) Exclusion criteria: history of seizures or mental illness Recruitment: 171 randomly assigned, 169 analysed (91 clonidine group, 78 control group) Location: Italy
Interventions	Intervention group: clonidine 2 mcg/kg IV just before start of surgery Control group: saline placebo All participants: oral midazolam 0.5 mg/kg premedication, sevoflurane induction in air, LMA, regional/central block at beginning or local infiltration at end with 2.5 mg/kg ropivacaine, rectal paracetamol 30 mg/kg at end of surgery
Outcomes	EA defined as "nonpurposeful movements that could be calmed and did not require restraint" for ≥ 5 minutes. Severe EA defined as "nonpurposeful movements that could not be calmed and required restraint" for ≥ 5 minutes Other outcomes: emergence time, number of participants requiring analgesia in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned to a treatment group," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, "the anesthesiologist caring for the child was blinded to the patients' group assignment"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"a second anesthesiologist also blinded to the treatment group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 withdrawals: 1 inadequate caudal block, 1 participant with seizures
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Tripi 2004.

Methods	Randomized controlled trial
Participants	18 months‐9 years, ASA I‐II, inguinal or penile surgery Exclusion criteria: none stated Recruitment: 100 randomly assigned, 92 analysed (46 in each group) Location: USA
Interventions	Intervention group: parental presence at emergence Control group: no parental presence at emergence All participants: oral midazolam premedication 0.5‐0.75 mg/kg mixed with acetaminophen, sevoflurane and nitrous oxide induction, maintenance with isoflurane and nitrous oxide via face mask or LMA, caudal block 1 mL/kg 0.125% bupivacaine, no opioids or muscle relaxants
Outcomes	Emergence behaviours recorded using 2 measures: the Operating Room Behaviour Rating Scale and the Operating Room Cooperation Rating Scale (1 = total co‐operation, 7 = total lack of co‐operation). Severe emergence distress defined as co‐operation score of 5 or greater Other outcomes: parental survey of child's temperament, parent anxiety, parent satisfaction, postoperative negative behavioural changes 1 and 4 weeks following surgery
Notes	Sevoflurane used only at induction and isoflurane used for maintenance for all participants
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomized," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	child and anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	"one research assistant was responsible for collecting all data, however she was not blinded to the presence or absence of a parent during emergence"
Incomplete outcome data (attrition bias) All outcomes	Low risk	8 withdrawals because of deviations from the protocol (4 from each group): 7 received IV fentanyl, 1 received sevoflurane maintenance
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Tsai 2008.

Methods	Randomized controlled trial
Participants	12‐110 months, ASA I‐II, elective outpatient surgeries Exclusion criteria: none stated Recruitment: 60 children (20 in each of 3 groups) Location: Taiwan
Interventions	Propofol 1 mg/kg just after induction Ketamine 0.25 mg/kg just after induction Control group: saline placebo All participants: midazolam 0.2 mg/kg premedication, sevoflurane induction followed by IV insertion and administration of study drug, sevoflurane maintenance
Outcomes	EA defined as score > 3 when measured on a 10‐point visual analogue scale (VAS) Other outcomes: severity of EA (VAS scores), duration of PACU stay, parent and nurse satisfaction
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"the i.v. agents and the i.v. set were covered to ensure that all personnel were blinded to the treatment"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"PACU nurse blinded"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Uezono 2000.

Methods	Randomized 2‐period cross‐over study
Participants	1‐5 years, ASA I‐II, eye examination for retinoblastoma Exclusion criteria: history of neurological disorder, series of radiation therapy Recruitment: 18 participants, 16 of whom received each of the 2 types of anaesthesia on separate occasions (8 received sevoflurane maintenance first and 8 received propofol maintenance first) Location: Japan
Interventions	Propofol maintenance (0.1‐0.4 mcg/kg/min after a 2‐mg/kg bolus) Control group: sevoflurane maintenance All participants: oral midazolam 0.5 mg/kg as premedication, sevoflurane induction in air/oxygen, rectal paracetamol 3 mg/kg, intubation facilitated by vecuronium 0.1 mg/kg
Outcomes	EA defined as "inconsolable crying, combative behavior, or thrashing" Other outcomes: time in PACU, PONV, parental satisfaction, number of participants with pain
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, treating anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"blinded anesthesiologist"
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 withdrawals due to not having repeat procedure within 6 months of the first
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Uzun 2003.

Methods	Randomized controlled trial
Participants	4‐12 years, ASA I‐II, tonsillectomy and/or adenoidectomy, adenoidectomy and/or tube insertion, tonsillectomy and/or tube insertion Exclusion criteria: long‐term sedative medication, central nervous system disease, history of allergy to drugs Recruitment: 50 children (25 in each group) Location: Turkey
Interventions	Desflurane 6%‐7% maintenance Control group: sevoflurane 2%‐2.5% maintenance All participants: oral midazolam 0.5 mg/kg premedication, propofol 2‐2.5 mg/kg IV induction, alfentanil 10 mcg/kg, cisatracurium 0.1 mg/kg, rectal paracetamol 20 mg/kg, dexamethasone 150 mcg/kg, maintenance with study volatile and 60% nitrous oxide in oxygen
Outcomes	EA defined as a score of 3 on following 3‐point scale: 1 = asleep or calm 2 = mildly agitated, crying but consolable, restless 3 = hysterical, crying inconsolably, thrashing Other outcomes: emergence time, recovery time, perioperative heart rate and blood pressure
Notes	English abstract, full paper in Turkish
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Valley 1999.

Methods	Randomized controlled trial
Participants	4 months‐14 years, ASA I‐II, elective surgery below the umbilicus Exclusion criteria: history of airway disease Recruitment: 40 children (20 in each group) Location: USA
Interventions	Isoflurane maintenance Control group: sevoflurane maintenance All participants: sevoflurane induction or propofol IV induction, maintenance with study agent and nitrous oxide, ETT placement facilitated by muscle relaxants, regional or local anaesthesia depending on procedure
Outcomes	Emergence delirium defined as "disoriented behaviour characterized by inconsolable agitation" Other outcomes: vomiting, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"random assignment," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"research nurse, blinded to the assigned group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of emergence delirium

Valley 2003.

Methods	Randomized controlled trial
Participants	6 months‐13 years, ASA I‐II, elective surgery below the umbilicus Exclusion criteria: history of airway disease Recruitment: 48 children (24 in each group) Location: USA
Interventions	Desflurane maintenance Control group: sevoflurane maintenance All participants: premedication with oral midazolam 0.5 mg/kg at anaesthetist's discretion, sevoflurane or propofol induction at anaesthetist's discretion, maintenance with study agent with nitrous oxide, regional or local block at end of procedure
Outcomes	EA defined as "disoriented behavior characterized by inconsolable agitation" Other outcomes: number of participants needing analgesia in PACU, vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"randomly assigned," method of random sequence generation not stated
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"research nurse, blinded to the assigned group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Viitanen 1999a.

Methods	Randomized controlled trial
Participants	1‐3 years, ASA I‐II, ambulatory adenoidectomy with or without myringotomy Exclusion criteria: none stated Recruitment: 52 children (26 in each group) Location: Finland
Interventions	Propofol induction 3 mg/kg IV Control group: sevoflurane induction All participants: no sedating premedication, EMLA cream, IV atropine 10 mcg/kg and IV alfentanil 10 mcg/kg before induction, sevoflurane and nitrous oxide maintenance, mivacurium to facilitate intubation, rectal paracetamol 20 mg/kg
Outcomes	Emergence delirium defined as total score > 3 on the pain/discomfort scale described by Hannallah et al (Hanallah 1994) Other outcomes: intubating conditions, recovery times, PONV, laryngospasm, number of participants requiring pain relief
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned, by computer‐based random numbers listing"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child not blinded to induction technique; however all children underwent IV cannulation before induction, inducing anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"recovery was assessed by a specially trained nurse, who was unaware of the induction method"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of emergence delirium

Viitanen 1999b.

Methods	Randomized controlled trial
Participants	1‐3 years, ASA I‐II, ambulatory adenoidectomy Exclusion criteria: allergic to drugs used, taking medication that would interact with midazolam (e.g antibiotics, antiepileptics, sedatives) Recruitment: 60 children (30 in each group) Location: Finland
Interventions	Midazolam 0.5 mg/kg oral premedication 30 minutes before induction of anaesthesia Control group: placebo All participants: IV cannula inserted before induction, atropine 0.01 mg/kg and alfentanil 10 mcg/kg IV, sevoflurane and nitrous oxide induction and maintenance, rectal acetaminophen 20 mg/kg after intubation
Outcomes	"Postanesthetic excitement" defined as total score > 3 using the 3 objective components (crying, movement and agitation) of the pain/discomfort scale reported by Hannallah et al (Hanallah 1994) Other outcomes: number of children requiring analgesia in PACU, time to emergence, vomiting, laryngospasm
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"all children were randomized according to a computer‐generated random numbers program in a double‐blinded fashion"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"all observers, as well as the children and their parents, were unaware of the contents of the premedicant"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals for EA outcome
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Viitanen 1999c.

Methods	Randomized controlled trial
Participants	1‐3 years, ASA I‐II, ambulatory adenoidectomy Exclusion criteria: allergy to drugs used, recent or long‐term medication that could interact with midazolam Recruitment: 60 children (30 in each group) Location: Finland
Interventions	Midazolam 0.5 mg/kg oral premedication 30 minutes before induction Control group: placebo All participants: IV cannula inserted before induction after EMLA cream, atropine 10 mcg/kg and alfentanil 10 mcg/kg given IV, lignocaine 10 mg IV with propofol induction of 3 mg/kg with increments of 0.5 mg/kg to achieve acceptance of face mask, mivacurium 0.2 mg/kg to facilitate tracheal intubation, maintenance with sevoflurane and 70% nitrous oxide, rectal acetaminophen 20 mg/kg after intubation
Outcomes	"Arousal distress" defined as total score > 3 on a pain/discomfort scale based on that of Hannallah et al (Hanallah 1994). Total numbers for incidence of "arousal distress" used in forest plot Other outcomes: time to emergence, time to discharge, vomiting, airway difficulty, number needing postoperative pain relief in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated random numbers listing"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"all observers, as well as the children and their parents, were unaware of the contents of the oral premedicant"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	as above
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Viitanen 2000.

Methods	Randomized controlled trial
Participants	1‐3 years, ASA I‐II, adenoidectomy with or without myringotomy Exclusion criteria: none stated Recruitment: 40 children (20 in each group) Location: Finland
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: no sedating premedication, induction and maintenance with study agent and 70% nitrous oxide, tracheal intubation, then rectal diclofenac 12.5 mg
Outcomes	"Post‐anaesthetic excitement" defined as total score > 3 using a modified pain/discomfort scale based on that used by Hannallah et al (Hanallah 1994) Other outcomes: laryngospasm, PONV, number of participants requiring analgesia postoperatively
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly allocated by use of a computer‐generated table"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"recovery of all children was evaluated by the same trained recovery nurse, who was blinded to the anaesthetic method used"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Villani 1998.

Methods	Randomized controlled trial
Participants	3‐12 years, ASA I‐II, elective urological, abdominal and orthopaedic surgery Exclusion criteria: ASA III or greater; pulmonary, heart, CNS, neuromuscular or renal disease; obese participants (> 30% of ideal body weight) Recruitment: 64 children (32 in each group) Location: Spain
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: oral flunitrazepam 0.05 mg/kg premedication, induction and maintenance with study agent
Outcomes	Reported "agitation during emergence" without a further definition Other outcomes: hypotension during induction, laryngospasm, vomiting, renal function tests (urea, creatinine)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"using sealed envelopes, patients were randomly selected," method of random sequence generation not stated
Allocation concealment (selection bias)	Low risk	"sealed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	"all observations and measurements were made and recorded by a trained observer, who was not blind to the choice of vapour"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Walker 1997.

Methods	Randomized controlled trial
Participants	2‐13 years, ASA I‐II, day case surgery Exclusion criteria: history of significant cardiac, respiratory, renal, hepatic or musculoskeletal abnormalities; participants receiving drugs that potentially affect renal or hepatic function, recent general anaesthesia, expected duration of surgery > 3 hours Recruitment: 52 children randomly assigned, 50 analysed (26 halothane group, 24 sevoflurane group) Location: Australia
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: premedication with oral midazolam 0.5 mg/kg (maximum 15 mg), oral paracetamol preinduction or rectal paracetamol post induction, induction and maintenance with study agent and nitrous oxide, regional block when appropriate with bupivacaine 2.5 mg/kg
Outcomes	Incidence of "excitement" on emergence reported Other outcomes: induction and emergence times; objective pain/discomfort scores; haemodynamics, respiratory events, shivering, nausea and vomiting on emergence; blood and urine biochemistry
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization schedule"
Allocation concealment (selection bias)	Low risk	"the drug allocation was contained in a sealed envelope"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"emergence events were noted and graded in severity by the nurse observer," not stated whether they were blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 withdrawals from sevoflurane group: 1 because of problems with vaporizer and 1 as the result of cancelled surgery
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Welborn 1996.

Methods	Randomized controlled trial
Participants	1‐7 years, ASA I‐II, adenoidectomy with bilateral myringotomy and insertion of tubes Exclusion criteria: none stated Recruitment: 80 children (20 in each of 4 groups) Location: USA
Interventions	Group 1: sevoflurane induction and maintenance Group 2: halothane induction and sevoflurane maintenance Group 3: halothane induction and maintenance Group 4: halothane induction and desflurane maintenance All participants: oral midazolam 0.5 mg/kg premedication, 60% nitrous oxide, no opioids given in theatre
Outcomes	"Postoperative agitation" defined as child was crying inconsolably, thrashing and hysterical Other outcomes: number of participants requiring analgesia in PACU, postoperative vomiting
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomly assigned via a computer‐generated random numbers table"
Allocation concealment (selection bias)	Low risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"a single‐blinded, independent observer evaluated each patient"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Weldon 2004.

Methods	Randomized controlled trial
Participants	12 months‐6 years, ASA I, inguinal hernia repair Exclusion criteria: emergency procedures, medical contraindication to placement of a block, mental retardation, developmental delay, attention‐deficit/hyperactivity disorder, psychiatric illness, history of paradoxical excitation with sedatives or previous episode of emergence delirium Recruitment: 80 children randomly assigned, 68 analysed (34 in each group) Location: USA
Interventions	Halothane anaesthesia Control group: sevoflurane anaesthesia All participants: oral premedication with midazolam 0.5 mg/kg mixed with ibuprofen 10 mg/kg, induction and maintenance with study agent and nitrous oxide, caudal block (1 mL/kg of a mixture of equal volumes of 0.25% bupivacaine and 1% lignocaine), caudal block judged inadequate if the child's heart rate increased by > 20% within 60 seconds of skin incision
Outcomes	EA defined as 3 or 4 on the following 4‐point scale (assessed on arrival to PACU and at 5‐minute intervals): 1 = awake and calm 2 = crying, requiring consoling 3 = irritable/restless, screaming, inconsolable 4 = combative, disoriented, thrashing Severe EA defined as a score that remained > 3 for 5 minutes after arrival of a parent Data in forest plot are from the 5‐minute time point Other outcomes: Yale Preoperative Anxiety Scale (YPAS) score, separation scale, induction scale, parental satisfaction scores
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization program"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Low risk	child blinded, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"trained observer blinded to the inhaled anesthetic group"
Incomplete outcome data (attrition bias) All outcomes	Low risk	12 children (6 from each group) excluded because of heart rate response at skin incision (suspected ineffective block)
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Yucel 2012.

Methods	Randomized controlled trial
Participants	3‐7 years, ASA I‐II, adenotonsillectomy with or without myringotomy Exclusion criteria: known allergy to study drugs, sensitivity or contraindication to opioids or local anaesthetic, renal or liver impairment, acute pharyngeal infection, history of asthma, clotting disorder Recruitment: 52 children (26 in each group) Location: Turkey
Interventions	Magnesium sulphate 15 mg/kg IV (commenced 2 minutes after intubation in 20 mL saline over 20 minutes) Control group: saline placebo All participants: midazolam 0.5 mg/kg oral premedication and rectal paracetamol 30‐40 mg/kg 1 hour before induction, parental presence at induction if agitated despite premedication, sevoflurane 8% induction with nitrous oxide 50%, IV insertion, cisatracurium 0.2 mg/kg IV, alfentanil 20 mg/kg IV, intubation, maintenance with sevoflurane 1.5%‐2.5% with nitrous oxide 60%, nesostigmine and atropine reversal
Outcomes	EA defined as PAED score ≥ 16 Other outcomes: preinfusion and postinfusion magnesium levels, time to extubation, time to eye opening, pain scores (Oucher Visual Analog Pain Scale), adverse events, haemodynamics
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization program"
Allocation concealment (selection bias)	Low risk	numbered, identical syringes
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"neither the treatment assignment or the contents of the syringe were known to anaesthesia staff"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"research observer who was blinded to the treatment allocation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	no withdrawals
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Zand 2011.

Methods	Randomized controlled trial
Participants	2‐7 years, ASA I‐II, anaesthesia for short (less than 0.5 hour) outpatient subumbilical surgery, compatible with peripheral nerve block Exclusion criteria: history of chronic illness or developmental delay, mental retardation, attention‐deficit/hyperactivity disorder, psychiatric illness or paradoxical excitation with sedatives Recruitment: 169 randomly assigned, 167 analysed (44 group S, 40 group SM, 41 group H, 42 group HM) Location: Iran
Interventions	Group S: sevoflurane with parental presence at induction and no premedication Group SM: sevoflurane with oral midazolam 0.5 mg/kg premedication Group H: halothane with parental presence at induction and no premedication Group HM: halothane with oral midazolam 0.5 mg/kg premedication All participants: induction and maintenance with study volatile and 60% nitrous oxide via face mask, rectal diclofenac 2 mg/kg, peripheral field block by surgeon with 0.25% bupivacaine
Outcomes	EA defined as a score of 3 or 4 on the following 4‐point scale: 1 = awake and calm, cooperative 2 = crying, requiring consoling 3 = irritable/restless, screaming, inconsolable 4 = combative, disoriented, thrashing Other outcomes: emergence time, time in PACU
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer‐generated randomization program"
Allocation concealment (selection bias)	Unclear risk	method of allocation concealment not stated
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	child not blinded to parental presence or premedication, anaesthetist not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"one recovery room nurse blinded to the group assignment"
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 participant each from group H and group SM was excluded because of ventricular arrhythmia and regurgitation of gastric content, respectively
Selective reporting (reporting bias)	Low risk	reported incidence of EA

Abbreviations used in tables:

AAI = A‐line autoregressive index.

APTT = activated partial thromboplastin time.

ASA = American Society of Anesthesiologists.

BIS = Bispectral Index.

BP = blood pressure.

CHEOPS = Children's Hospital Eastern Ontario Pain Scale.

CHIPPS = Children's and Infants’ Postoperative Pain Scale.

CNS = central nervous system.

CSF = cerebrospinal fluid.

CVS = cardiovascular system.

EA = emergence agitation.

ED = emergence delirium.

EKG = electrocardiogram.

ENT = ear, nose and throat.

ETCO₂ = end‐tidal carbon dioxide.

ETT = endotracheal tube.

FLACC = face, legs, activity, cry, consolability.

Hb = haemoglobin.

HR = heart rate.

INR = international normalized ratio.

IQR = interquartile range.

IV = intravenous.

LMA = laryngeal mask airway.

MAC = minimum alveolar concentration.

mCHEOPS = modified Children's Hospital Eastern Ontario Pain Scale.

mYPAS = modified Yale Preoperative Anxiety Scale.

MRI = magnetic resonance imaging.

NSAID = non‐steroidal anti‐inflammatory drug.

OPS = Objective Pain Scale.

OTFC = oral transmucosal fentanyl citrate.

OR = operating room.

PACBIS = Perioperative Adult Child Behavioral Interaction Scale.

PAED = Pediatric Anesthesia Emergence Delirium.

PACU = postanaesthesia care unit.

PICU = paediatric intensive care unit.

PDS = pain/discomfort score.

PHBQ = Post‐Hospital Behavioural Questionnaire.

PONV = postoperative nausea and vomiting.

PPIA = parental presence at induction of anaesthesia.

PPEA = parental presence at emergence from anaesthesia.

PR = per rectum.

RCT = randomized controlled trial.

RSI = rapid sequence induction.

SD = standard deviation.

SEM = standard error of the mean.

SpO₂ = peripheral capillary oxygen saturation.

TCI = target‐controlled infusion.

TPS = ten‐point scale.

TIVA = total intravenous anaesthesia.

URTI = upper respiratory tract infection.

VAS = visual analogue scale.

YPAS = Yale Preoperative Anxiety Scale.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Akinci 2008	EA not studied
Almenrader 2007	No control group
Ariffin 1997	EA not studied
Chen 2010	No control group (comparison of midazolam, propofol and ketamine)
Choi 2011	No control group (comparison of remifentanil and nitrous oxide during maintenance)
Cole 2002	Study drugs halothane and isoflurane, not sevoflurane
Delvi 2007	EA not studied
El‐Hennawy 2009	EA not studied. Aim of study to measure analgesic effects of clonidine and dexmedetomidine in caudal blocks with the FLACC (face, legs, activity, cry, consolability) pain scale used for this purpose
Ertugrul 2006	No control group (comparison of ketamine, meperidine and tramadol)
Funk 2000	EA not studied and researchers have not controlled for sevoflurane
Greenspun 1995	EA not studied
Hung 2005	Includes adult participants
Ibrahim 2001	Adult participants only
Ingelmo 2007	EA not studied
Isik 2006b	EA not studied
Kain 1999	Sevoflurane not used (halothane induction with isoflurane maintenance)
Kain 2007	Unclear whether all participants were maintained with sevoflurane. Attempted to contact study author for clarification but no reply
Kawaai 2008	No control group
Kawaraguchi 2002	No control group
Malmgren 2004	No control group
Mayer 2006	Article retracted in 2011 by editor‐in‐chief of Anesthesia & Analgesia for "unethical conduct of research," as IRB approval could not be confirmed
Mckay 2011	No standardized sevoflurane control group (switched mid‐study from halothane maintenance to sevoflurane maintenance when halothane became unavailable)
Mizrak 2011	No sevoflurane anaesthesia
Ozer 2003	No control group (comparison of tramadol and meperidine)
Piat 1994	EA not studied
Sarner 1995	EA not studied
Shaban 2008	No control group
Steinmetz 2007	Fentanyl dose not controlled between the 2 comparison groups
Uysal 2011	No control group
Wagner 2003	All participants received isoflurane maintenance after sevoflurane induction with no reported risk of EA

Abbreviations used:

EA = emergence agitation.

IRB = Institutional Review Board.

Characteristics of studies awaiting assessment [ordered by study ID]

Abdelhalim 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Abdulatif 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Ali 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Aydin 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Celebi 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Chandler 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Chen 2013a.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Chen 2013b.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Cheng 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Elgebaly 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Gai 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Gupta 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Hasanein 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Hasani 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

He 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Khalifa 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Kim 2013a.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Kim 2013b.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Kim 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Kodra 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Li 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Locatelli 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Oofuvong 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Park 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Pedersen 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Rashad 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Rizk 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Sethi 2013.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Sheta 2014.

Methods
Participants
Interventions
Outcomes
Notes	Not yet assessed

Differences between protocol and review

• The title of the published protocol (Stephens 2008) has been edited to include updated terminology for our primary outcome—emergence agitation (EA).

• Minor rewording of Objective revised. "To compare sevoflurane with other general anaesthetic agents, with or without pharmacological or non‐pharmacological adjuncts, with regard to risk of EA in children during emergence from anaesthesia." Change in terminology of our primary outcome to risk of EA.

• A new secondary outcome of agitation score has been added because EA scores have been developed and used since the writing of the original protocol for this review. Protocol Methods updated to make future reviews more relevant to clinical practice. We removed subgroup analyses for different age groups; different types of surgery—emergency or elective surgery; other premedication adjuncts, especially other preemptive analgesics; use of different opioids; use of other intraoperative analgesic adjuncts; and surgical versus less invasive techniques, when sufficient numbers of studies are identified. These were replaced with more relevantly worded subgroups.

• Added "agitation scores such as PAED" as a secondary outcome in light of this measure of EA used in more recent studies.

• Removal of "Number of participants with other, new‐onset maladaptive behaviours (e.g. sleep or eating disturbance, poor concentration) during the postoperative period, as measured by the authors of included studies" as a secondary outcome, as multiple factors other than sevoflurane will be influencing this outcome, which are quite separate from those influencing our primary outcome of EA.

• Sensitivity and subgroup analyses have been revised so that future versions of this review will incorporate those most likely to influence clinical practice.

• 'Types of interventions' has been modified and clarified further since publication of the Protocol to include "any sevoflurane anaesthetic with or without nitrous oxide compared with any other general anaesthetic. The types of general anaesthetics included were other volatile anaesthetics, for example, isoflurane, desflurane and halothane; and any other general anaesthetics, for example, propofol or ketamine. Pharmacological adjuncts such as use of fentanyl or other opioids, propofol, midazolam, ketamine, dexmedetomidine or clonidine or non‐pharmacological adjuncts such as parental presence."

• The published protocol stated that primary outcomes were "The number of participants with postoperative behavioural disturbance as measured by the authors of included studies. Emergence delirium will be defined as stated above: a mental disturbance during recovery from general anaesthesia consisting of hallucinations, delusions and confusion manifested by moaning, restlessness, involuntary physical activity and thrashing about in the bed. The review now states, "Emergence agitation (EA) was defined as the number of participants with postoperative behavioural disturbance as measured and/or reported by the authors of included studies.This updated primary outcome reflects changes in nomenclature and understanding of EA since the time the protocol was written.

• Similarly, we have updated both secondary outcomes. The published protocol stated, "Secondary outcomes, Number of participants with other, new‐onset maladaptive behaviours (e.g. sleep or eating disturbance, poor concentration) during the postoperative period, as measured by the authors of included studies. Number of participants with emergence delirium during the postoperative period after leaving the PACU, as measured by the authors of included studies." Our review now states the secondary outcomes as "Postoperative behavioural disturbance or agitation scores such as PAED (Sikich 2004) and Number of participants with EA during the postoperative period after leaving the PACU, as measured by the authors of included studies."

• New authors were added to the review team since the protocol was published in April 2008: David Costi joined in 2008, very soon after the protocol was published; Samira Ahmed and James Ellwood joined in 2010; Cheryl Chooi and Laura Burgoyne joined in 2012.

Contributions of authors

David Costi (DC), Allan M Cyna (AMC), Samira Ahmed (SA), Kate Stephens (KS), Penny Strickland (PS), James Ellwood (JE), Jessica N Larsson (JNL), Cheryl Chooi (CC), Laura L Burgoyne (LLB), Philippa Middleton (PM)

Conceiving of the review: AMC, KS, PS.
 Co‐ordinating the review: DC, AMC.
 Undertaking manual searches: DC, KS, PS, JNL.
 Screening search results: DC, KS, PS, JNL.
 Organizing retrieval of papers: DC, KS, PS, JNL.
 Screening retrieved papers against inclusion criteria: DC, SA, KS, PS, JE, JNL, CC, LLB, AMC.
 Appraising quality of papers: DC, PM, SA, KS, PS, JE, JNL, CC, LLB, AMC.
 Abstracting data from papers: DC, SA, KS, PS, JE, JNL, CC, LLB, AMC.
 Writing to authors of papers to ask for additional information: DC, JE.
 Obtaining and screening data on unpublished studies: N/A.
 Managing data for the review: DC, PM, AMC.
 Entering data into Review Manager (RevMan 5.2): DC, SA.
 Analysing RevMan statistical data: PM, DC, AMC.
 Performing other statistical analyses not using RevMan: N/A.
 Performing double entry of data: (data entered by person one): N/A.
 Interpreting data: DC, AMC, PM.
 Making statistical inferences: PM.
 Writing the review: DC, AMC, PM, KS, PS.
 Securing funding for the review: N/A.
 Performing previous work that was the foundation of the present study: N/A.
 Serving as guarantor for the review (one author): DC.
 Taking responsibility for reading and checking the review before submission: AMC, DC.

Sources of support

Internal sources

• Child Youth and Women's Health Service (CYWHS), Australia.

External sources

• No sources of support supplied

Declarations of interest

David Costi: received a SPANZA research grant (The Society for Paediatric Anaesthesia in New Zealand and Australia) for a randomized controlled study of EA in 2010.

Allan M Cyna: received travel and accommodation expenses for speaking at the New Zealand Society of Hypnosis ASM in 2013.

Samira Ahmed: none known.

Kate Stephens: none known.

Penny Strickland: none known.

James Ellwood: none known.

Jessica N Larsson: none known.

Cheryl Chooi: none known.

Laura L Burgoyne: none known.

Philippa Middleton: none known.

New