Effects of Fentanyl on Emergence Agitation in Children under Sevoflurane Anesthesia: Meta-Analysis of Randomized Controlled Trials

Background and Objectives The goal of this meta-analysis study was to assess the effects of fentanyl on emergence agitation (EA) under sevoflurane anesthesia in children. Subjects and Methods We searched electronic databases (PubMed, Embase, Web of Science and the Cochrane Central Register of Controlled Trials) for articles published until December 2014. Randomized controlled trials (RCTs) that assessed the effects of fentanyl and placebo on EA under sevoflurane anesthesia in children that the outcome were the incidence of EA, postoperative pain, emergence time or adverse effects were included in this meta-analysis. Results A total of 16 studies, including 1362 patients (737 patients for the fentanyl group and 625 for the placebo group), were evaluated in final analysis. We found that administration of fentanyl decreased the incidences of EA (RR = 0.37, 95% CI 0.27~0.49, P<0.00001) and postoperative pain (RR = 0.59, 95% CI 0.41~0.85, P = 0.004) but increased the incidence of postoperative nausea and vomiting (PONV) (RR = 2.23, 95% CI 1.33~3.77, P = 0.003). The extubation time (WMD = 0.71 min, 95% CI 0.12~1.3, P = 0.02), emergence time (WMD = 4.90 min, 95% CI 2.49~7.30, P<0.0001), and time in the postanesthesia care unit (PACU) (WMD = 2.65 min, 95% CI 0.76~4.53, P = 0.006) were slightly increased. There were no significant differences in the time to discharge of day patients (WMD = 3.72 min, 95% CI -2.80~10.24, P = 0.26). Conclusion Our meta-analysis suggests that fentanyl decreases the incidence of EA under sevoflurane anesthesia in children and postoperative pain, but has a higher incidence of PONV. Considering the inherent limitations of the included studies, more RCTs with extensive follow-up should be performed to validate our findings in the future.


Introduction
Literature search strategy A comprehensive literature search was performed in December 2014. We searched electronic databases, including PubMed, Embase, Web of Science, and the Cochrane Central Register of Controlled Trial. The key search terms were as follows: sevofluran Ã , emergence agitation/ excit Ã / delirium/ confusion, (postoperative/ postanesthetic) (agitation/ confusion / behavioral change Ã ), children/ infant, and fentanyl. The manual searching of the references of the retrieved studies were used to extend the search. Only English articles were considered. When necessary, we contacted the authors for additional unpublished data.

Inclusion and exclusion criteria
RCTs comparing fentanyl with placebo (normal saline) administered perioperatively to reduce EA incidence in pediatric patients (aged 1-14 years) with sevoflurane anesthesia were included in this systematic review. We excluded letters to the editor, editorials, case reports, reviews, and animal studies.

Data extraction and outcome measurements
Two independent authors extracted and summarized data from eligible trials. Disagreements were resolved by discussion with other authors. We extracted the following data from each eligible trial: first author, publication year, patient ages, type of surgery, number of patients, sedative premedication, dose, timing, and route of administration of fentanyl/placebo, sevoflurane anesthesia protocol, perioperative analgesia, the EA incidence, postoperative pain, emergence time, extubation time, time in postanesthesia care unit (PACU), time to discharge and adverse events.
The primary outcome is the incidence of emergence agitation (EA). EA incidence was defined as the incidence of participants with postoperative behavioural disturbance during emergence from anesthesia, which was measured by the authors of included studies. The secondary outcomes examined in this study included pain incidence in PACU, extubation time, emergence time, time in the PACU, the time to discharge of day patients and adverse events, such as the incidence of PONV, respiratory adverse events and haemodynamic changes requiring intervention. Pain incidence in PACU was defined by the authors of the studies using the Objective Pain Scale (OPS), Children's and Infant's Postoperative Pain Scale (CHIPPS) or fourpoint Verbal Rating Scale. Extubation time was defined as the time interval from anesthetic discontinuation to extubation. Emergence time was measured as the time between discontinuation of anesthesia and spontaneous eye opening. Time in the PACU was defined as the time interval from anesthetic discontinuation to discharge from the PACU. The time to discharge of day patients was defined as the time between anesthetic discontinuation and discharge from the hospital of day patients. The incidence of PONV was assessed by evaluating nausea and vomiting behaviors from the entrance of patients into the PACU to 24 h after surgery.

Quality assessment and statistical analysis
We examined the quality of studies included in the meta-analysis using the Cochrane Collaboration's tool for assessing risk of bias [16]. The domains included a random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting.
Meta-analyses were conducted using Review Manager 5.3 (Cochrane Collaboration, Oxford, UK) and Stata software, version 12.0 (Stata Corporation, College Station, TX). Relative risks (RRs) and weighted mean differences (WMDs) were used to compare dichotomous and continuous variables, respectively, both with corresponding 95% confidence intervals (CIs). A confidence interval for an RR of <1 indicated that the incidence of the test target in the fentanyl group was lower than that in the placebo group. If studies presented continuous data as median and range values, the means and standard deviations were transformed as described by Hozo et al. [17] Statistical heterogeneity was assessed using the chi-square test with a significance of P<0.10 [18]. Heterogeneity was quantified with the I 2 statistic [19]. If P>0.10 and I 2 <50%, fixed effects analysis was conducted to calculate the pooled OR; otherwise, a random effects model was used [16]. We conducted subgroup analyses to investigate possible causes of heterogeneity. Sensitivity analyses were performed by removing each trial individually to evaluate the quality and consistency of the results. To evaluate whether potential publication bias might have affected statistical results, we applied funnel plots, Begg's test and Egger's test. All statistical tests were 2-sided.
We conducted subgroup analysis separately because confounding factors, such as premedication, type of surgery, preschool-aged children and pain, may have affected the incidence of EA (Table 3). Subgroup analysis of the timing of administration revealed that the use of fentanyl both before and at the end of surgery resulted in a preventive effect against EA (RR = 0.39, 95% CI 0.28~0.54, P<0.00001, I 2 = 59%; RR = 0.26, 95% CI 0.15~0.47, P<0.00001, I 2 = 0%). Analysis of 12 intravenous fentanyl trials showed that this intervention was effective (RR = 0.35, 95% CI 0.24~0.50, P<0.00001, I 2 = 53%). Three intranasal studies of this drug had RR of 0.30(95% CI 0.12~0.72, P = 0.008, I 2 = 50%), and one oral study had an RR of 0.54 (95% CI 0.35~0.83, P = 0.005). The effect of midazolam is still an ongoing debate. The meta-analysis by Zhang et al found midazolam had a significant effect on preventing EA [57] while the metaanalysis by Dahmiani et al found midazolam to be ineffective for the prevention of EA. On the contrary, midazolam might even trigger EA [5]. To eliminate the effects of this drug, we performed subgroup analysis, including 9 trials without premedication and 4 studies with midazolam premedication, and showed the prevention of EA in the fentanyl group (RR = 0.34, 95% CI 0.23~0.50, P<0.00001, I 2 = 55%; and RR = 0.34, 95% CI 0.13~0.90, P = 0.03, I 2 = 63%, respectively). Ear, nose and throat (ENT) procedures were reported to be independent risk factors for EA [58]. The protocols of 8 studies included ENT procedure for children. The pooled RR of ENT procedure studies was 0.45 (95% CI 0.32~0.64, P<0.0001, I 2 = 33%). When pooled analysis was limited to studies of patients who underwent minor urologic or inguinal surgery and received an appropriate regional block with enough local anesthetics, we found that the pooled  RR was 0.34 (95% CI 0.21~0.57, P<0.0001, I 2 = 0%). Seven trials evaluated EA in preschool children younger than 7 years of age. We found that the pooled RR was 0.33 (95% CI 0.21~0.52, P<0.00001) but that the I 2 remained high at 59%.
Adverse events. The assessment of 9 studies [4,12,43,44,50,52,53,55,56]together showed that PONV occurred in 103 of 454 patients in the fentanyl group and 42 of 388 patients in the placebo group. The pooled results showed that fentanyl significantly increased the PONV incidence in the children under sevoflurane anesthesia (RR = 2.23, 95% CI 1.33~3.77, P = 0.003, I 2 = 42%) (Fig 8). One study [44] reported that a participant experienced suspicious laryngospasm, and 4(6%) patients had airway obstruction in the fentanyl group. Another study [53] showed that the risk of drug-related respiratory adverse events was higher for patients receiving oral transmucosal fentanyl citrate (OTFC) than for other patients; however, most of the adverse events were mild. No study reported hemodynamic events requiring intervention in any arm.

Methodological qualities of included studies and potential sources of bias
The methodological qualities of the included trials were showed in Table 4. No study was found to beat a high risk of bias for any of the criteria considered. The blinding of participants and personnel, the blinding of the outcome assessment, the presence of incomplete outcome   data, and selective reporting were determined to be at a low risk of bias in all included studies. Random sequence generation was unclear in five trials [45,47,49,51,53], and allocation concealment was unclear in 14 studies [4,12,43,44,46,[48][49][50][51][52][53][54][55].
A funnel plot of the included studies that reported the incidence of EA showed potential publication bias (Begg's test, P = 0.022, Egger's test, P = 0.023) (Fig 9). Considering the effect of the missing trials, we conducted a trim-and-fill analysis and the analysis showed "no trimming performed; data unchanged".

Discussion
This systematic review and meta-analysis of 16 RCTs, including 1362 patients, indicates that fentanyl significantly reduces the incidence of EA under sevoflurane anesthesia in children and decreases postoperative pain but it increases the incidence of PONV. The extubation time, emergence time, and time in the PACU were slightly prolonged. We found no significant difference in the time to discharge of day patients.
Several previous meta-analyses indicate that fentanyl can reduce the incidence of EA under sevoflurane anesthesia in children [59], whereas the meta-analysis by Dahmiani et al state that intravenous. fentanyl failed to prevent EA [5]. In our subgroup analysis, we found that both intravenous and intranasal fentanyl showed to be effective. The reasons for these conflicting results may be due to inclusion of only two studies in the meta-analysis by Dahmiani et al.
Fentanyl, a short-acting opioid analgesic, is used to reduce the incidence of pain. Some investigators have argued that pain experienced during impaired consciousness in children results in severe EA [13,58,60]. Our findings also showed that fentanyl decreased the incidence of pain in children in the PACU (RR = 0.59, 95%CI 95%CI 0.41~0.85, P = 0.004) and reduced the incidence of EA (RR = 0.37, 95% CI 0.27~0.49, P<0.00001) in children under sevoflurane anesthesia. However, it was still difficult to fully identify EA or pain-induced behavioral disorders in the children evaluated in the present study. Locatelli et al suggested that the splitting of PAED scale into ED1 and ED2 scores might help to separate ED from pain [61]. In addition, previous studies have reported a frequent incidence of EA in patients who have received sevoflurane for genitourinary surgery with an adequate caudal block and for non painful interventions, such as magnetic resonance imaging [7,54]. Following restriction of the studies achieving a high level of pain relief during surgery by regional nerve block, the preventative effect of fentanyl remained significant (RR = 0.34, 95% CI 0.21~0.57, P<0.00001, I 2 = 0%). Thus, it is hard to establish an explicit relationship between pain and EA, and pain may not be the only factor affecting the occurrence of EA in children. Fentanyl is effective for EA in a rather unspecific way. Whatever the reason for EA might be pain, delirium, agitation for other reasons such as parental separation, hunger, thirst etc, fentanyl provides analgesia and sedation and hence disrupts agitation and crying. It resolves the problem even without knowing the exact underlying cause, especially in those situations where there might be an overlap between pain and delirium. Some studies have demonstrated that rapid awakening is one of the factors contributing to EA [62] because of the low blood-gas solubility and rapid emergence characteristics of sevoflurane. In the current study, the children administered fentanyl were found to have a slightly prolonged extubation time (WMD = 0.71 min, 95% CI 0.12~1.3, P = 0.02, I 2 = 0%), emergence time (WMD = 4.90 min, 95% CI 2.49~7.30, P<0.0001, I 2 = 95%) and time in the PACU (WMD = 2.65 min, 95% CI 0.76~4.53, P = 0.006) and a lower incidence of EA. Some authors have found that the incidence of EA is not reduced by delayed emergence from sevoflurane anesthesia in children [63]. Therefore, it is still difficult to confirm that fentanyl reduces the incidence of EA by preventing rapid emergence from sevoflurane anesthesia.
The incidence of PONV was significantly higher in the fentanyl group than the placebo group (RR = 2.23, 95% CI 1.33~3.77, P = 0.003, I 2 = 42%). However, a lack of postoperative follow-up after more than 24 hours may have been a limiting factor in the interpretation of these study results. Other adverse events were reported in two studies; however, we did not find any serious adverse events in any of the included trials. Additional adverse events were infrequent in most studies mentioning ' no adverse events' and in those not addressing them at all. Thus, we were notable to ascertain safety.
Between-study heterogeneity was significant for some of the continuous variables but was not significant for the dichotomous outcomes. Different surgery types, children's ages, premedication, timing and the route of administration were described in the included studies. These differences may have resulted in the significant between-study heterogeneity. The effect of heterogeneity may have been reduced by using the random effects model, but not abolished.
Some limitations need to be considered for the present study. The main limitation is that the incidence of EA may have been greatly influenced by the uses different scales with different cut-off values to define the presence of EA and some of the scales are not validated [64]. Because small children cannot verbalize pain, anxiety, thirst or hunger, it is difficult to interpret their behaviors [65]. Although some studies used a reliable pain scale and the PAED scale to decrease errors associated with pain, a clear differentiation between EA and agitation because of pain could not be guaranteed. Future systematic reviews should explore different EA assessment tools separately when a sufficient amount of data is available. In addition, the follow-up time was generally short; therefore, any impacts on the long-term outcome of EA remain to be validated. Furthermore, we restricted the study selection to the English language and unpublished studies were not included in this meta-analysis adding a language bias and publication bias. Some studies reported that the exclusion of non-English studies may result in more conservative estimates of treatment effects, because studies with positive results were more likely to be published and more likely to be published in English [66]. Nevertheless, we searched for studies with multiple strategies, included and evaluated the methodological qualities of the studies with strict criteria, and minimized heterogeneity with subgroup analysis. Therefore, we provide the up-to-date information on this topic.

Conclusions
In conclusion, this systematic review and meta-analysis indicates that fentanyl may be associated with a decreased incidence of EA in children under sevoflurane anesthesia in addition to reduced postoperative pain, but has a higher incidence of PONV. However, considering the inherent limitations of the included studies, more RCTs with extensive follow-up should be performed to validate our findings in the future.