Comparative efficacy of non-pharmacological interventions on emergency delirium in postoperative children: a systematic review and Bayesian network meta-analysis

Abstract

Background:

Emergence delirium (ED) is a distressing complication in pediatric patients following general anesthesia, often resulting in self-injury, extended hospitalization, and emotional distress for families. Although non-pharmacological interventions are being increasingly utilized, their effectiveness remains uncertain.

Methods:

We conducted a systematic review and Bayesian network meta-analysis. PubMed, EMBASE, Cochrane CENTRAL, and Web of Science databases were searched from inception to 17 November 2024, with an updated search performed on 24 July 2025. Randomized controlled trials (RCTs) evaluating non-pharmacological interventions in children (<18 years) undergoing general anesthesia were included. Studies involving pharmacological co-interventions were excluded. Two reviewers independently extracted data and assessed the risk of bias using the Cochrane Risk of Bias 2.0 (RoB 2.0) tool. Bayesian random-effects models were applied, and effectiveness of interventions was ranked using surface under the cumulative ranking curve (SUCRA) values.

Results:

A total of 56 RCTs involving 6183 pediatric participants were included. Among 37 non-pharmacological strategies, virtual reality (VR) and ice popsicles were the most effective in reducing ED [VR: risk ratios (RR) = 0.082, 95% credible interval (CrI) 0.010–0.311; ice popsicle: RR = 0.157, 95% CrI 0.057–0.388]. VR also ranked highest for postoperative pain reduction (standardized mean difference = −2.68, 95% CrI −3.47 to −1.88). For anxiety outcomes, parental active participation in anesthesia induction (PAPIA) was most effective in reducing children’s perioperative anxiety (SUCRA = 86.1%), while parental educational programs (EP) ranked highest for reducing caregiver anxiety (SUCRA = 78.3%). No intervention significantly improved compliance on the Induction Compliance Checklist.

Conclusions:

This Bayesian network meta-analysis provides low-to-moderate certainty evidence suggesting that VR and simple oral cold stimulation (ice popsicles) may reduce ED and early postoperative pain in children. PAPIA may modestly reduce children’s perioperative anxiety, while EP appears most effective in alleviating parental anxiety. However, no intervention demonstrated significant improvement in induction compliance.

Introduction

Emergence delirium (ED) is a transient yet distressing neuropsychiatric condition observed in pediatric patients during recovery from general anesthesia. Clinically, it is characterized by cognitive disorientation, restlessness, physical agitation, hallucinations, and emotional instability^[1]. This syndrome is associated with various acute complications, including postoperative agitation, accidental self-harm, prolonged hospitalization, post-discharge behavioral disturbances, emotional burden on patients and caregivers, and increased healthcare resource utilization^[2,3]. A meta-analysis of 16 studies and 9598 children, the global pooled prevalence of ED is approximately 19.2%^[4].

The etiology of ED is complex and multifactorial, involving the interplay between immature central nervous system development, rapid emergence from volatile anesthetics (particularly sevoflurane), unmitigated perioperative pain, and preoperative anxiety^[5,6]. Although pharmacological strategies – primarily dexmedetomidine^[7,8], midazolam^[9], propofol^[10], magnesium sulfate^[11], and nalbuphine^[12] – aim to reduce the incidence and severity of ED by modulating sedation, anxiety, or nociception, their effectiveness is constrained by variability in pediatric pharmacodynamics, unclear optimal dosing, and potential adverse effects including respiratory depression, hypotension, bradycardia, and delayed recovery^[13–16].

There is growing interest in non-pharmacological interventions as they offer developmentally appropriate and safer alternatives with the potential to reduce ED, pain, and anxiety, thereby improving perioperative outcomes^[17–19]. These include cognitive-behavioral strategies (e.g., preoperative education, video modeling, and breathing training), sensory modulation techniques [e.g., music therapy and virtual reality (VR)], emotional support (e.g., parental presence), and physiological stimulation (e.g., acupuncture and transcutaneous electrical stimulation). These approaches are believed to enhance perioperative adaptation by promoting emotional regulation, autonomic stability, and cognitive control. For instance, video-based interventions have been shown to significantly reduce both ED severity and preoperative anxiety, particularly in children under 7 years old^[17]. Similarly, acupuncture therapy has demonstrated a trend toward reduced lower ED incidence^[20]. Moreover, preoperative breathing training delivered via video learning has been shown to reduce both the occurrence and intensity of ED in preschool children by enhancing cognitive regulation and autonomic balance^[21].

Although previous systematic reviews and meta-analyses support the efficacy of non-pharmacological strategies in pediatric perioperative care, they have largely relied on pairwise comparisons, limiting their ability to compare multiple interventions concurrently^[22–25]. This restricts our ability to determine the most effective strategy across diverse intervention types. Network meta-analysis (NMA) addresses this limitation by incorporating both direct and indirect evidence to enable simultaneous comparison and ranking of multiple interventions using both direct and indirect evidence^[26].

In the context of Enhanced Recovery After Surgery and growing emphasis on child-centered perioperative care, identifying optimal non-pharmacological interventions has both clinical and strategic value. Therefore, the present study aimed to compare the relative effectiveness of various non-pharmacological interventions in reducing ED, pain, and perioperative anxiety in children undergoing general anesthesia through a Bayesian NMA. This approach aims to generate high-level comparative evidence to guide clinical decision-making and optimize perioperative care in pediatric surgical populations. In line with current reporting standards for studies involving artificial intelligence, we followed the Transparency In The reporting of Artificial INtelligence (TITAN) 2025 guideline^[27].

Methods

This systematic review and NMA were conducted and reported in accordance with the PRISMA-NMA guidelines, following the reporting standards established by Hutton^[28]. The protocol was registered in the PROSPERO database (CRD42023452957). The methodological quality of this systematic review was assessed and reported in line with the AMSTAR (A Measurement Tool to Assess Systematic Reviews) Guidelines^[29].

Data sources and search strategy

We conducted a comprehensive search of PubMed, EMBASE, the Cochrane Central Register of Controlled Trials, and Web of Science from inception to 17 November 2024, with an updated search performed on 24 July 2025. Search terms included a combination of Medical Subject Headings (MeSH) and free-text keywords. Key MeSH terms included “Pediatrics,” “Child, Preschool,” “Child,” and “Emergence Delirium.” The Cochrane highly sensitive search strategy was used to identify randomized controlled trials (RCTs)^[30]. To maximize sensitivity, we intentionally omitted an intervention-specific search block. In line with Cochrane Handbook guidance (Chapter 4), the electronic strategy combined only population terms (e.g., child, pediatric) and outcome terms [“emergence delirium,” “emergence agitation,” Pediatric Anesthesia Emergence Delirium (PAED)] together with the Cochrane Highly Sensitive RCT filter and omitted specific non-pharmacological keywords (e.g., “virtual reality,” “music therapy,” “acupuncture,” and “parental presence”). Non-pharmacological interventions in this field are described inconsistently and are sometimes labeled with proprietary or non-standard terminology; adding numerous such terms may reduce recall and introduce selection bias. All records retrieved by this broad strategy (including pharmacological studies) were then screened in duplicate at title/abstract and full-text levels to identify eligible non-pharmacological RCTs. Complete search strategies for PubMed, EMBASE, Web of Science, and the Cochrane Central Register of Controlled Trials are provided in Supplemental Digital Content Appendix A, available at: http://links.lww.com/JS9/G647. These strategies were adapted to meet the indexing and syntax requirements of each database. No restrictions were imposed based on the publication year.

HIGHLIGHTS

• Virtual reality (VR) games or a simple ice popsicle may help calm children after surgery, but the data are not yet strong enough to make a firm clinical rule.

• VR is the most effective intervention for reducing postoperative pain in children.

• Parental active participation in anesthesia induction intervention is the most effective intervention for reducing children’s perioperative anxiety.

• Educational program administration effectively reduces parental anxiety.

• No intervention demonstrated a significant improvement in induction compliance.

Inclusion and exclusion criteria

Eligibility screening was conducted using the PICOS framework (Population, Intervention, Comparison, Outcome, and Study Design), and only full-text articles were considered.

Population

Studies were included if they enrolled pediatric patients (<18 years of age) undergoing general anesthesia for surgical procedures.

Interventions and comparisons

Eligible interventions included non-pharmacological strategies aimed at preventing or managing ED, such as maternal voice, monochromatic blue light, mouth breathing training, VR, the ADVANCE strategy, parental presence, video distraction, ice popsicles, and their combinations (e.g., video distraction combined with parental presence). Comparators included standard perioperative care, pharmacological agents (e.g., dexmedetomidine or oral midazolam), or alternative non-pharmacological interventions. To maintain consistency in network modeling, studies that combined non-pharmacological interventions with pharmacological treatments were excluded. Additionally, variations in intensity or delivery modality within the same intervention category (e.g., different types of video content) were considered a single node in the NMA. Detailed definitions and classifications of all intervention nodes are provided in Supplemental Digital Content Appendix D, available at: http://links.lww.com/JS9/G647.

Outcome

The primary outcomes were the severity and incidence of ED in pediatric patients receiving non-pharmacological interventions following general anesthesia. Severity was assessed as a continuous outcome using the PAED scale, a validated tool for quantifying symptom intensity. Incidence was treated as a binary outcome, defined as the proportion of children who met established PAED thresholds or clinical criteria during postanesthesia recovery. To standardize the measurement timing of PAED outcomes across studies, we explicitly defined our primary analytic approach as using the maximum PAED score recorded within the first 30 minutes following emergence from anesthesia. Specifically, when studies reported multiple PAED assessments at different time intervals (e.g., 5, 10, 15, and 30 minutes post-emergence), the highest recorded PAED value within this 30-minute window was extracted for analysis. For studies providing only a single PAED assessment time point (always ≤30 min), that single measurement was utilized directly. This strategy ensures that our analyses consistently capture the peak clinical severity of ED, aligning with previous methodological recommendations and existing literature standards^[31].

Secondary outcomes included preoperative anxiety (measured using the modified Yale Preoperative Anxiety Scale, m-YPAS), postoperative pain (assessed using the Face, Legs, Activity, Cry, and Consolability scale FLACC), induction compliance (evaluated using the Induction Compliance Checklist, ICC), and parental anxiety (assessed using the State-Trait Anxiety Inventory, STAI). These outcomes enabled a comprehensive assessment of both pediatric and caregiver experiences during the perioperative period and supported evaluation of the broader psychosocial effects of the interventions.

Study design

Eligible study designs were RCTs only. Non-randomized studies (e.g., prospective or retrospective cohort, case–control, before–after, quality-improvement projects without a clearly defined concurrent control group) and quasi-randomized trials were excluded to minimize selection bias and confounding and to ensure methodological rigor. We restricted inclusion to RCTs to maximize internal validity in estimating the comparative effectiveness of non-pharmacological interventions. Risk of bias for all included trials was assessed using the Cochrane RoB 2 tool (see Supplemental Digital Content Appendix E, available at: http://links.lww.com/JS9/G647).

Data selection and extraction

Data were independently extracted by two reviewers (W.Y.Z. and Z.C.) using a predesigned structured extraction form. Extracted information included study characteristics, baseline demographics, details of the intervention and control groups, and all prespecified primary and secondary outcomes. All outcomes were recorded at the post-intervention time point.

When necessary, missing numerical data were estimated or converted from reported statistics such as sample size, median, interquartile range, standard error, t-statistic, or P-value. For studies presenting outcome data in graphical format, numerical values were extracted using Origin software. Study authors were contacted via email to obtain missing information or clarify inconsistencies. Discrepancies in data extraction were resolved by consensus or, when needed, through arbitration by a third reviewer (C.H.).

To ensure the reliability of the graphical data extraction, two reviewers independently digitized the same figures, and the extracted values were cross-validated for consistency. For discrepancies exceeding a predefined threshold (≥10% relative difference), the average of the two measurements was used after discussion.

We extracted the surgical context for each study – surgical region; specific procedure; approach (anesthetic technique and airway management; MIS vs non-MIS when reported); operative complexity (ASA class; elective vs emergency; day-case vs inpatient); and setting (see Supplemental Digital Content Appendix J, available at: http://links.lww.com/JS9/G647). To address transitivity, we prespecified a sensitivity analysis restricting the network to low-complexity ENT/ophthalmology day-case surgery (ASA I–II), excluding MIRPE and cardiac catheterization.

Risk of bias assessment

Two reviewers (L.C.X. and L.S.N.) independently assessed the risk of bias for each included study using the Cochrane Risk of Bias 2.0 (RoB 2) tool^[32]. This tool evaluates five domains: (1) the randomization process, (2) deviations from intended interventions, (3) missing outcome data, (4) outcome measurement, and (5) selection of the reported result. Each domain was rated as “low risk,” “some concerns,” or “high risk” according to the criteria outlined in the Cochrane Handbook (version 6.1.0). Given the nature of non-pharmacological interventions, blinding of participants and personnel is generally not feasible. As a result, many studies were rated as having “some concerns” in the domain of deviations from intended interventions how the lack of blinding may have influenced adherence or outcome assessment. This finding is consistent with prior evaluations of behavioral and psychological interventions, in which blinding is inherently limited^[33]. Disagreements between reviewers were resolved through discussion, with arbitration by a third reviewer (C.H.) when necessary.

Statistical analysis

Bayesian NMAs were conducted in R (version 4.1.2; RStudio, Boston, MA, USA) using gemtc (version 0.8-2) and rjags (version 4–10) packages interfaced with JAGS. For continuous outcomes, standardized mean differences (SMDs) or mean differences (MDs) with 95% credible intervals (CrIs) were calculated, whereas for binary outcomes, risk ratios (RRs) with 95% CrIs were reported.

A random-effects model was applied to account for clinical heterogeneity across studies, including differences in country, type of surgery, intervention duration, and patient age. Four Markov chains were run for each outcome, with 50 000 iterations per chain and the first 20 000 discarded as burn-in. Convergence was assessed using trace plots and the Gelman–Rubin–Brooks diagnostic tool.

Relative rankings of interventions were estimated using the surface under the cumulative ranking curve (SUCRA), calculated in R. Model consistency was assessed by comparing the deviance information criterion (DIC) between the consistency and inconsistency models, with a DIC difference less than 5 considered acceptable. Heterogeneity was quantified using the I² statistic. Given that clinical heterogeneity across trials is unavoidable, we prespecified several strategies to minimize potential confounding. A Bayesian random-effects model was used to account for between-study variability in patient age, surgical type, intervention duration, and perioperative context. In addition, we conducted multiple prespecified sensitivity analyses – including restriction to low-complexity ENT/ophthalmology day-case surgery, age-stratified analyses, and outcome-timing analyses – to evaluate whether clinical heterogeneity materially influenced the pooled estimates. These procedures supported the robustness of the main findings despite inherent between-study variability.

To explore developmental stage effects, we pre-specified three age strata – < 5 years (toddler/preschool), 5–8 years (early school–age), and ≥8 years (pre-adolescent) – and performed separate NMAs within each subgroup for both incidence and severity of ED. Treatment effects in each band were estimated as log RRs (for ED incidence) or SMDs (for severity of ED) with 95% CrIs; heterogeneity was quantified by I², and consistency assessed by DIC comparisons (ΔDIC < 5 deemed acceptable).

In parallel, we conducted a Bayesian network meta-regression treating age as a continuous covariate. Raw ages were centered at the sample mean (5.504 years) and scaled by the SD (2.768 years), then entered with a shared slope across all comparisons. Effect modification by age on severity of ED SMDs was evaluated by comparing the DIC of the meta-regression model versus the base model (no covariate).

Network plots were generated in Stata (version 18.1; StataCorp, College Station, TX, USA) using the mvmeta package to illustrate treatment comparisons and network geometry. Comparison-adjusted funnel plots and Egger’s test were also conducted in Stata to assess small-study effects and potential publication bias. All the R and Stata codes used for analysis are provided in Supplemental Digital Content Appendix I, available at: http://links.lww.com/JS9/G647.

Results

Identification of relevant studies

Following deduplication (1925/3992), 2067 records were screened, and 1920 were excluded at title/abstract review. We sought 147 full texts, of which 49 were not retrievable. Of the 98 assessed, 49 were excluded with reasons (Supplemental Digital Content Appendix B, available at: http://links.lww.com/JS9/G647; PRISMA Figure 1). A search update on 24 July 2025 identified 765 additional records and contributed seven further RCTs. In total, 56 RCTs (6183 participants) evaluating 33 non-pharmacological interventions were included.

Figure 1.

PRISMA flow diagram depicting the search and selection process of eligible studies included in the network meta-analysis.

Characteristics of included studies

Fifty-six RCTs published between 2007 and 2025 were conducted across 17 countries spanning Asia, Europe, North America, Africa, and Oceania (Table 1; full references in Supplemental Digital Content Appendix C, available at: http://links.lww.com/JS9/G647). Most trials enrolled day-case ENT/ophthalmology cohorts (e.g., adenotonsillectomy, tympanostomy, and strabismus/cataract), with a minority in urology/orthopedics, hernia/orchiopexy, dental/maxillofacial surgery, and higher-complexity settings such as cardiac catheterization and MIRPE. Study sizes typically ranged from ~40 to ~360 participants; mean or median ages clustered around 4–8 years (range from toddlers to early adolescents), and boys comprised approximately half of the participants, where reported. Definitions of ED/agitation varied, most commonly the PAED scale (thresholds ≥10, ≥ 12), with additional use of Watcha, Aono, EA, ED-I, NUDESC, CAP-D, and related instruments. In design terms, 48 trials were two-arm, 7 were three-arm, and 1 was four-arm. Interventions fell into five broad categories: parental engagement or voice [e.g. PPIA/parental active participation in anesthesia induction (PAPIA), parental presence during induction or emergence, maternal voice]; digital distraction and preparation (VR, virtual operating-room tours or education, tablet/video cartoons, web/app-based education, and large-screen systems); acupuncture or acupoint stimulation (including capsicum plasters); sensory or environmental modulation (blindfolding, earplugs/noise masking, light modulation/shields, active warming, and electric-car transport); and brief comfort or feeding protocols (ice-pops/lollipops, clear liquids, or carbohydrate–electrolyte drinks). Outcomes most frequently assessed were preoperative anxiety (m-YPAS/m-YPAS-SF), emergence delirium/agitation (PAED, Watcha, Aono/EA, ED-I, NUDESC, and CAP-D), ICC, pain (FLACC, CHEOPS, and FPS-R), postoperative behavior (PHBQ/PHBQ-AS), and caregiver anxiety (STAI). Further granularity by surgical context and outcome measurement is provided in Supplemental Digital Content Appendix J, available at: http://links.lww.com/JS9/G647.

Table 1.

Key characteristics of all included studies.

No	Author	Year	Country	Sample size (Male)		Age (Mean ± SD)/Median(SD)/Median[IQR]/Median (range)		Surgery type	Definition of ED	Intervention		Outcomes
				Intervention 1	Intervention 2	Intervention 1	Intervention 2			Intervention 1	Intervention 2
1	Sarah Samnakay	2024	Australia	98 (60)	90 (52)	8.83 ± 2.8	8.80 ± 2.9	Elective/emergency surgery	CAP-D ≥ 9	3D VR	2D Video	m-YPAS-SF, ICC, and CAP-D
2	Cansu Çiftci	2024	Turkey	35 (16)	35 (26)	4.1 ± 1.3	3.8 ± 1.4	Elective surgery	/	Active Warming Group	Control	FLACC and WS
3	Yue Wang	2024	China	80 (39)	80 (32)	4.6 ± 1.1	4.5 ± 1.0	Elective surgery (Strabismus)	PAED ≥ 10	Parents Accompany	Video distraction (tablet)	m-YPAS-SF, PAED, and PHBQ
4	H. Volkan Acar	2012	Turkey	25 (15)	25 (10)	7.2 ± 2.5	7.4 ± 1.8	Elective surgery (ENT)	PAED ≥ 10	Acupuncture(Capsicum plasters)	Control	m-YPAS, PAED, and CHEOPS
5	YUAN-CHI LIN	2009	USA	30 (11)	30 (11)	2.0 ± 1.3	2.4 ± 1.6	Elective surgery (ENT)	EA≥2	Control	Acupuncture	CHEOPS and EA
6	Kristen M. Bailey	2015	Canada	44 (26)	49 (28)	5.0 (2.2)	5.0(2.3)	Elective surgery	/	PPIA	PPIA Preparation	m-YPAS, STAI, FLACC, FPS-R, PAED, and ICC
7	Hui-Hong Liang	2023	China	50 (22)	50 (20)	4.21 ± 1.66	4.38 ± 1.30	Oral surgery (lingual frenectomy/mucocele excision)	Cole≥3	Control	Ice popsicle	FLACC and A Cole 5-Point Scale
8	Cong Wang	2021	China	34 (21)		4.74 ± 1.21		Elective surgery (ENT)	PAED ≥ 10	Mother’s voice		PAED and FLACC
	Cong Wang	2021	China	34 (18)		4.74 ± 1.21		Elective surgery (ENT)	PAED ≥ 10	Stranger voice		PAED and FLACC
	Cong Wang	2021	China	34 (23)		4.50 ± 1.18		Elective surgery (ENT)	PAED ≥ 10	Control		PAED and FLACC
9	Yan-Yan Yang	2020	China	64 (32)	63 (31)	5.1 ± 2.0	4.9 ± 2.1	Elective surgery (bilateral ophthalmic)	WS ≥ 3	Mother’s voice	Control	WS and FLACC m-YPAS
10	Yajun Li	2023	China	29 (11)	29 (17)	8.90 ± 3.71	8.17 ± 3.63	Elective minor surgery (orthopedic)	PAE ≥ 10	Acupuncture	Control	FPS-R and PAED
11	Soumily Bandyopadhyay	2024	Indian	69 (48)	69 (40)	5.3 ± 1.2	5.3 ± 1.3	Elective surgery (eye)	WS > 2	Video distraction + parental presence	Parental Presence Alone	M-YPAS, WS, and CFS
12	Enrico Muzzi	2021	Italy	26 (12)		4.5 [3.6–6.0]		Elective surgery (ENT)	PAED ≥ 10	Auditory stimulation with classical music		PAED, FLACC, FPS-R, and CHEOPS
	Enrico Muzzi	2021	Italy	25 (13)		5.5 [4.1–8.3]		Elective surgery (ENT)	PAED ≥ 10	Auditory stimulation with rhythmic heartbeat noise		PAED, FLACC, FPS-R, and CHEOPS
	Enrico Muzzi	2021	Italy	25 (14)		4.8 [3.7–6.2]		Elective surgery (ENT)	PAED ≥ 10	Ambient noise insulation with masking earplugs		PAED, FLACC, FPS-R, and CHEOPS
	Enrico Muzzi	2021	Italy	28 (17)	5.2 [3.6–7.0]	Elective surgery (ENT)	PAED ≥ 10	Control	PAED, FLACC, FPS-R, and CHEOPS
13	Teresa Franco Castanys	2023	Spain	61 (49)	64 (46)	6.7 ± 2.09	7.0 ± 2.19	Ambulatory surgery (majority circumcision)	PAED ≥ 10	Virtual tour of OR	Control	m-YPAS, STAI, ICC, and PAED
14	Robin Eijlers	2019	Netherlands	97 (56)	73 (34)	7.5 [5.6–10.7]	9.0 [6.4–10.7]	Elective day surgery (maxillofacial/dental/ENT)	PAED ≥ 10	Control	VRE	m-YPAS, FPS-r, FLACC, PAED, and STAI
15	Toshiyuki Hijikata	2015	JAPAN	60 (42)	60 (41)	3.9 (2.5–5.3)	4.3 (3.0–5.6)	Elective surgery (ENT/strabismus)	PAED > 10/Aono ≥ 3	Acupuncture	Control	Aono’s Scale, PAED, and CHEOPS
16	Hong Chen	2024	China	30 (20)		5.5 [5–6]		Elective surgery (ENT)	PAED ≥ 12	Control		PSAS, FLACC, and PAED
	Hong Chen	2024	China	30 (14)		5 [5–6]		Elective surgery (ENT)	PAED ≥ 12	Tablet cartoon distraction		PSAS, FLACC, and PAED
	Hong Chen	2024	China	30 (14)		6 [5–6]		Elective surgery (ENT)	PAED ≥ 12	VR cartoon distraction		PSAS, FLACC, and PAED
17	Gregor Massoth	2024	Germany	26 (12)	25 (11)	1.58 ± 1.22	1.85 ± 1.19	Elective cardiac catheterization	ED I ≥ 9	Mother’s voice	Control	ED I score, m-YPAS, and CHIPPS
18	Xinyu Tang	2023	China	40 (17)	40 (23)	5.7 ± 1.21	5.68 ± 1.58	Elective surgery (ENT)	PAED ≥ 10	Cartoon video	Control	m-YPAS, PAED, ICC, and FLACC
19	Liting Ji	2015	China	51 (28)	51 (23)	6.59 ± 2.49	6.02 ± 2.19	Elective surgery (ENT/Strabismus)	PAED ≥ 10	DrawMD APP aided education	Control	APAIS, m-YPAS, PAED, and STAI
20	Pei-Fang Dong	2022	China	76 (40)		4.55 ± 1.44		Elective surgery (eye)	NU-DES ≥ 2	Control		NU-DESC, m-YPAS, and FPS-R
	Pei-Fang Dong	2022	China	76 (38)		4.85 ± 1.46		Elective surgery (eye)	NU-DES ≥ 2	30-Min Blindfolded		NU-DESC, m-YPAS, and FPS-R
	Pei-Fang Dong	2022	China	72 (45)		4.68 ± 1.46		Elective surgery (eye)	NU-DES ≥ 2	60-Min Blindfolded		NU-DESC, m-YPAS, and FPS-R
21	S. J. KIM	2010	Korea	23 (8)	23 (10)	6.50 ± 3.50	6.83 ± 3.50	Cardiac catheterization	EA > 2	Control	Mother’s voice	EA, m-YPAS, and STAI
22	Omar Soliman	2022	Egypt	30 (30)	30 (30)	6.37 ± 1.4	6.20 ± 1.3	Hypospadias repair surgeries	PAED > 12	Mother’s voice	Control	m-YPAS-SF, PAED, and FLACC
23	Xiaofei Mo	2023	China	20 (15)	20 (17)	3.9 ± 1.4	4.2 ± 1.5	Elective surgery (ophthalmic/urological)	PAED ≥ 10	Lollipops	Control	PAED and FLACC
24	S. Byun	2018	South Korea	33 (14)	33 (17)	5.8 (3–8)	5.7 (3–8)	Elective surgery (ENT/eye)	PAED > 12/WS > 2	Mother’s voice	Stranger Voice	PAED, WS, and FLACC
25	Michelle A. Fortier	2015	USA	38 (24)	44 (22)	4.3 ± 1.8	4.4 ± 1.7	Elective out-patient surgery	/	Webtips	Control	m-YPAS, STAI, PAED, and Nurse-rated pain numeric rating scale
26	Tarek I Ismail	2022	Egypt	120 (58)	120 (57)	5.58 ± 1.53	5.65 ± 1.56	Elective orthopedic surgery	/	Parent active participation(PAPIA)	PPIA	m-YPAS, VFAS FPS-R, and FLACC PAED
27	Constance N. Burke	2009	USA	45 (27)	43 (22)	4.5 ± 1.7	4.8 ± 1.7	Elective MRI	EA ≥ 3	Parent present during emergence in PACU (PP)	Reunion after arousal (PA)	m-YPAS,EA, PHBQ, and PAED
28	Nobuhito Nakamura	2018	JAPAN	50 (37)	50 (37)	3.8 [2.3–5.0]	4.3 [2.6–6.2]	Elective surgery (hernia/orchiopexy)	PAED ≥ 10/Aono ≥ 3	Acupuncture	Control	Aono, PAED, CHEOPS, and m-YPAS
29	Jing Yao	2022	China	30 (20)	30 (18)	4.3 ± 1.1	4.6 ± 1.2	Elective surgery (ENT)	/	Control	PPIA	ICC,PAED,m-YPAS-SF, COMFORT-B, and STAI-T/S
30	Bruno Pastene	2022	France	56 (31)	58 (34)	5 [4–6]	5 [4–7]	Elective surgery (ENT/Eye)	PAED ≥ 10	Control	Transport by electric car	m-YPAS-SF, ICC, PAED, FLACC, and STAI
31	Li-Nan Zhang	2022	China	77 (45)	77 (50)	5.0 (2.0)	5.0 (2.0)	Elective surgery (ENT)	PAED ≥ 10	Video-guided mouth-breathing training	Control	m-YPAS and PAED
32	Yan Pan	2019	China	44 (27)	56 (41)	3.52 ± 0.21	3.77 ± 0.19	Elective surgery (eye)	/	Blindfolded	Control	m-YPAS, PAED, and FLACC
33	Christine S. MARTIN	2020	USA	24 (10)	25 (18)	2.54 ± 1.7	2.52 ± 1.4	Myringotomy tube placement	PAED ≥ 10	Acupuncture	Control Group	PAED
34	Adam C. Adler	2021	USA	51 (27)	51 (27)	4.2 ± 1.4	4.7 ± 1.3	Elective surgery (ENT)	PAED ≥ 12	Control	Monochromatic Blue Light	PAED
35	Yuexi Jin	2021	China	50 (27)	50 (26)	4.72 ± 1.03	4.65 ± 1.81	Elective surgery (Strabismus)	WS ≥ 2	Control	Self-Produced Animation Introduction	STAI and m-YPAS WS
36	Mingyuan Liu	2022	China	54 (29)	55 (24)	/	/	Elective Surgery (Eye)	/	Control	Tablet game	m-YPAS, ICC PAED, and PHBQ
37	Jung-Hee Ryu	2018	South Korea	39 (21)	41 (29)	6 [5–8]	6 [5–7]	Elective surgery	/	Control group	VR tour of OR	PAED, m-YPAS, and PHBQ-AS
38	Fatih Yucedag	2024	Turkey	40 (28)	40 (22)	7.3 ± 2	7.5 ± 1.9	Elective surgery (ENT)	PAED ≥ 10	Child’s preference parental presence	Parents’ preference parental presence	PAED, m-YPAS, and STAI-S/T
39	Nicola G. Clausen	2020	Denmark	30 (25)	30 (26)	4.4 ± 0.3	4.4 ± 0.2	Elective minor surgery (orchidopexy, circumcision)	/	Tablet game	Control	FLACC, PAED, and m-YPAS
40	Doa’a Abdullah Dwairej	2019	Jordan	64 (27)	64 (31)	6.61 ± 1.72	6.50 ± 1.86	Elective minor surgery (ENT, dental)	/	Tablet game	Control	m-YPAS, ICC, and PAED
41	Adriana Carb	2024	Spain	120 (100)	121 (86)	6.7 ± 2.9	6.2 ± 2.8	Elective minor surgery (ENT, orthopedic)	/	VREP (5-min virtual-reality educational video)	Control	m-YPAS, ICC, PAED, STAI, and FPS-R
42	Samuel Rodriguez	2019	USA	27 (14)	25 (14)	6.7 ± 1.7	6.8 ± 2.2	Elective surgery (urology, ophthalmology, orthopedics)	PAED ≥ 10	Tablet	BERT (large screen)	m-YPAS, ICC PAED, and STAI
43	Hyuckgoo Kim	2015	South Korea	34 (15)		5.5 ± 1.0		Elective minor surgery (ENT, eye)	PAED ≥ 10	Video-cartoon distraction		m-YPAS, ICC STAI, CHEOPS, PAED, and PHBQ
	Hyuckgoo Kim	2015	South Korea	33 (12)		5.3 ± 1.4		Elective minor surgery (ENT, eye)	PAED ≥ 10	Parental presence		m-YPAS, ICC STAI, CHEOPS, PAED, and PHBQ
	Hyuckgoo Kim	2015	South Korea	37 (18)		5.0 ± 1.3		Elective minor surgery (ENT, eye)	PAED ≥ 10	Video-cartoon distraction + parental presence		m-YPAS, ICC STAI, CHEOPS, PAED, and PHBQ
44	Yuki Hashimoto	2020	Japan	29 (17)	29 (18)	5 [4–6]	5 [5–6]	Elective minor surgery (ENT, eye)	PAED > 9	VR cartoon distraction	Tablet distraction	m-YPAS, PAED, and EA
45	Yijie Wu	2022	China	48 (42)	48 (43)	8.0 [5.0–9.75]	7.0 [6.0–9.0]	Elective minor surgery (Penoplasty, ENT)	/	Control	Peri-operative VR video	m-YPAS-SF, ICC, PAED, FLACC, and STAI
46	Y. Lin	2018	China	89 (47)	90 (51)	5.0 (3.0–7.0)	4.8 (3.0–7.0)	Elective cataract surgery	PAED ≥ 10	Blindfolded	Control	PAED, m-YPAS, and FPSR
47	Sevtap Hekimoglu Sahin	2022	Turkey	35 (17)		6.9 ± 1.4		Elective minor surgery (orchiopexy, hypospadias repair, and inguinal hernia)	PAED ≥ 12	Silence – earplugs		m-YPAS and PAED
	Sevtap Hekimoglu Sahin	2022	Turkey	35 (16)		6.6 ± 1.5		Elective minor surgery (orchiopexy, hypospadias repair, and inguinal hernia)	PAED ≥ 12	Control		m-YPAS and PAED
	Sevtap Hekimoglu Sahin	2022	Turkey	35 (16)		6.7 ± 1.6		Elective minor surgery (orchiopexy, hypospadias repair, and inguinal hernia)	PAED ≥ 12	Music – children’s songs via CD player		m-YPAS and PAED
48	Qiaosheng Zhong	2018	China	23 (12)	23 (14)	4.60 ± 1.03	4.52 ± 0.99	Elective surgery (ENT)	Aono > 2/PAED ≥ 10	PVOR	Control	Aono, PAED, and OPS
49	Zeev N. Kain	2007	USA	99 (60)		5.4 ± 2		Elective surgery (ENT, urology, and plastics)	Three-point emergence scale = 3	Control		m-YPAS and STAI
	Zeev N. Kain	2007	USA	94 (69)		5.5 ± 2		Elective surgery (ENT, urology, and plastics)	Three-point emergence scale = 3	Parental presence		m-YPAS and STAI
	Zeev N. Kain	2007	USA	96 (62)		5.6 ± 2		Elective surgery (ENT, urology, and plastics)	Three-point emergence scale = 3	ADVANCE behavioral program		m-YPAS and STAI
50	Andi Ade Wijaya Ramlan	2025	Indonesia	67 (53)	70 (52)	4.47 ± 3.08	4.74 ± 3.59	Elective surgery (digestive, urology, ENT, eye, plastic surgery, dental, and orthopedic)	PAED ≥ 10	Carbohydrate-electrolyte fluid (CE)	Control	m-YPAS-SF, PAED, and FLACC
51	Rui Zhang	2025	China	35 (16)	35 (22)	6.0 [5.0–7.0]	5.0 [4.0–6.0]	Elective surgery (bilateral strabismus)	Aono≥ 3/PAED ≥ 16	Light-transmitting eye shields(LT)	Control	Aono, PAED, m-YPAS, and FLACC
52	Mawia Bataineh	2025	Jordan	42 (16)	42 (16)	5.48 ± 1.17	5.10 ± 1.14	Elective surgery (ENT)	PAED ≥ 10	Animated preoperative preparation video	Control	m-YPAS, PAED, and FLACC
53	Jaewon Huh	2024	Korea	27 (22)	27 (23)	4.7 ± 1.0	4.6 ± 0.8	Minimally Invasive Repair of Pectus Excavatum (MIRPE)	PAED ≥ 12/Watcha ≥ 3	Clear liquid (3 mL/kg water/2 h preoperatively)	Control	m-YPAS, PAED FLACC, and Watcha
54	Erin Brown	2025	Australia	50 (29)	50 (28)	6.64 (3–11)	6.38 (3–11)	Elective surgery (dental, ENT, ophthalmology, ortho, pediatric, plastic, radiology)	/	Take 5 video (Animated video teaching parental distraction and anxiety-coping)	Control	m-YPAS-SF, ICC, CAP-D, and FLACC
55	Zeyang Wang	2025	China	120 (59)		8.4 (5–12)		Elective surgery (ENT)	PAED ≥ 12	Maternal voice orientation		PAED and FLACC
	Zeyang Wang	2025	China	120 (57)		8.7 (5–12)		Elective surgery (ENT)	PAED ≥ 12	Maternal voice awakening		PAED and FLACC
	Zeyang Wang	2025	China	120 (62)		8.5 (5–12)		Elective surgery (ENT)	PAED ≥ 12	Control		PAED and FLACC
56	Wen Chen	2025	China	69 (30)	69 (38)	5 [4–6]	5 [4.5–7]	Elective surgery (ENT)	PAED ≥ 12	Watched cartoons	Control	PAED,Watcha, and FLACC

3D VR, three-dimensional virtual-reality video; ADVANCE, behavioral preparation program; Aono, Aono four-point Emergence Agitation Scale; AR, augmented reality headset; BERT, Bedside Entertainment and Relaxation Theater; CAP-D, Cornell Assessment of Pediatric Delirium; CFS, Child Fear Scale; CHEOPS, Children’s Hospital of Eastern Ontario Pain Scale; CHIPPS, Children and Infants Post-operative Pain Scale; COMFORT-B, COMFORT-Behavioral scale; EA, four-point Emergence Agitation scale; ED-I, Emergence Delirium Index; FLACC, Face-Legs-Activity-Cry-Consolability pain scale; FPS-R, ICC, Induction Compliance Checklist; m-YPAS, Modified Yale Pre-operative Anxiety Scale; m-YPAS-SF, m-YPAS Short Form; NU-DESC, Nursing Delirium Screening Scale; OPS, Objective Pain Score; PAED, Pediatric Anesthesia Emergence Delirium scale; PPIA, Parental Presence during Induction of Anesthesia; PAPIA, Parent Active Participation during Induction; PP/PA, Parent Present/Parent Absent at emergence; PSAS, Parent Separation Anxiety Scale; PVOR, Pre-operative Visit to Operating Room; STAI, State-Trait Anxiety Inventory; VR, virtual reality; VREP, Virtual-Reality Education Program; VFAS, Visual Facial Anxiety Scale; VRE, Virtual-Reality operating-theater environment Exposure; Wong-Baker Faces Pain Rating Scale; WS, Watcha four-point agitation score.

Risk of bias of included studies

Across 56 trials, 21 (37.5%) were judged low risk overall, and 35 (62.5%) had some concerns; no study was rated high risk overall. At the domain level, the randomization process was low risk in 53/56 (94.6%); deviations from intended interventions raised some concerns in 33/56 (58.9%); missing outcome data was low risk in 55/56 (98.2%); measurement of outcomes had some concerns in 31/56 (55.4%); and selection of the reported result was low risk in 54/56 (96.4%). Thus, concerns were concentrated in domains related to behavioral-intervention delivery and outcome measurement, whereas randomization, data completeness, and selective reporting were generally well controlled (full study-level assessments in Supplemental Digital Content Appendix E, available at: http://links.lww.com/JS9/G647; domain summaries in Supplemental Digital Content Appendix F, available at: http://links.lww.com/JS9/G647).

Network meta-analysis

Severity of ED (PAED scores)

Forty-six studies comparing 32 non-pharmacological interventions were synthesized (Fig. 2A). In the league table (Fig. 2C), four interventions demonstrated statistically significant reductions in PAED versus routine care: Pre-operative visit to the operating room (PVOR; SMD = −1.63, 95 % CrI −3.25 to −0.01), Maternal voice orientation (MVO; –1.48, −2.59 to −0.37), VR (–1.42, −2.37 to −0.49), and Mother’s voice (MV; –0.71, −1.40 to −0.03). Other pairwise contrasts were imprecise, with CrIs crossing zero. Cumulative ranking curves (Fig. 2B) yielded the following SUCRA values for the top five interventions: PVOR 72.6%, MVO 70.9%, and VR 64.1%. These rankings position PVOR and MVO as the most probable best treatments for reducing PAED, closely followed by VR, consistent with their favorable relative effects. Supplemental Digital Content Appendix G, available at: http://links.lww.com/JS9/G647 lists full SUCRA results.

Figure 2.

Network geometry, cumulative rankings, and relative effects for PAED scores. (A) Network plot of non-pharmacological interventions. Node size is proportional to the total sample size; edge width reflects the number of direct comparisons. (B) Cumulative ranking curves (rank 1 = best). The area under each curve corresponds to SUCRA (higher SUCRA = better ranking). (C) League table of pairwise relative effects on PAED: values are standardized mean differences (SMD) with 95% credible intervals (CrI). Read from row to column; negative SMD indicates lower PAED (favors the row intervention). Estimates in bold are statistically significant (CrI excludes 0). Lower PAED indicates less emergence delirium. Notes: Abbreviations are defined in Appendix D.

Incidence of ED (Binary categorical variable)

Forty-one studies comparing 33 non-pharmacological interventions were synthesized (Fig. 3A). According to the league table (Fig. 3C), several interventions were associated with a significant reduction in the risk of ED compared to routine care. Specifically, VR (RR = 0.082, 95% CrI 0.010–0.311), Ice_Pop (0.157, 0.057–0.388), and PI (0.160, 0.034–0.556) were associated with statistically significant reductions. The estimates for other interventions exhibited wide CrIs overlapping 1, suggesting considerable uncertainty. Cumulative ranking curves (Fig. 3B) revealed the following top five SUCRA values: VR (94.8%), Ice_Pop (88.0%), and PI (85.2%), indicating these interventions are most likely to rank highest in reducing ED incidence. These findings further support their potential clinical utility and align with the observed relative effects. Full SUCRA results are available in Supplemental Digital Content Appendix G, available at: http://links.lww.com/JS9/G647.

Figure 3.

Network geometry, cumulative rankings, and relative effects for emergence delirium (incidence). (A) Network plot of non-pharmacological interventions. Node size is proportional to total sample size; edge width reflects the number of direct comparisons. (B) Cumulative ranking curves (rank 1 = lowest ED risk). The area under each curve corresponds to SUCRA (higher SUCRA = better). (C) League table of pairwise relative effects on ED incidence: odds ratios (OR) with 95% credible intervals (CrI). Read from row to column; OR < 1 favors the row intervention (lower odds of ED), and OR > 1 favors the column intervention. Estimates in bold are statistically significant (CrI excludes 1). Notes: Abbreviations are defined in Appendix D.

Postoperative pain (continuous outcome, SMD)

Twenty-six studies comparing 22 non-pharmacological interventions were synthesized (Fig. 4A). In the league table (Fig. 4C), several interventions showed statistically significant reductions in postoperative pain compared to routine care. Specifically, VR (SMD = −2.68, 95% CrI −3.47 to −1.88), TVG (–1.92, −2.96 to −0.87), and Music (–1.31, −2.03 to −0.59) yielded the most pronounced effects. Other interventions demonstrated effect estimates with wide CrIs that crossed zero, indicating imprecision or lack of significance. Cumulative ranking curves (Fig. 4B) revealed the following top five SUCRA values: VR (99.3%), TVG (93.0%), and Music (85.3%), suggesting these interventions are most likely to be among the best for reducing postoperative pain. These results are consistent with their favorable SMD estimates. Full SUCRA rankings are provided in Supplemental Digital Content Appendix G, available at: http://links.lww.com/JS9/G647.

Figure 4.

Network geometry, cumulative rankings, and relative effects for postoperative pain. A) Network plot of non-pharmacological interventions. Node size is proportional to total sample size; edge width reflects the number of direct comparisons. (B) Cumulative ranking curves (rank 1 = lowest pain). The area under each curve corresponds to SUCRA (higher SUCRA = better). (C) League table of pairwise relative effects on pain: standardized mean differences (SMD) with 95% credible intervals (CrI). Read from row to column; negative SMD indicates lower pain (favors the row intervention). Estimates in bold are statistically significant (CrI excludes 0). Lower pain represents a better outcome (e.g., FLACC/VAS). Notes: Abbreviations are defined in Appendix D.

Preoperative anxiety (m-YPAS scores)

Twenty-three studies comparing 15 non-pharmacological interventions were synthesized (Fig. 5A). According to the league table (Fig. 5C), several interventions were associated with a significant reduction in preoperative anxiety, as measured by m-YPAS, compared to routine care. Specifically, PAPIA (SMD = −2.44, 95% CrI −4.69 to −0.22), TVG (–1.95, −3.40 to −0.52), and TVD (–1.37, −2.62 to −0.14) demonstrated statistically significant effects. The directions of effect for other interventions were favorable but showed imprecise estimates with CrIs overlapping zero. Cumulative ranking curves (Fig. 5B) revealed the following top five SUCRA values: PAPIA (86.1%), TVG (79.2%), and VR_PPIA (72.8%), indicating these interventions are most likely to be effective in reducing preoperative anxiety. These rankings are consistent with their observed relative effects. Full SUCRA rankings are provided in Supplemental Digital Content Appendix G, available at: http://links.lww.com/JS9/G647.

Figure 5.

Network geometry, cumulative rankings, and relative effects for children’s preoperative anxiety (m-YPAS). (A) Network plot of non-pharmacological interventions. Node size is proportional to total sample size; edge width reflects the number of direct comparisons. (B) Cumulative ranking curves (rank 1 = lowest anxiety). The area under each curve corresponds to SUCRA (higher SUCRA = better). (C) League table of pairwise relative effects on m-YPAS: standardized mean differences (SMD) with 95% credible intervals (CrI). Read from row to column; negative SMD indicates lower anxiety (favors the row intervention). Estimates in bold are statistically significant (CrI excludes 0). Lower m-YPAS indicates less preoperative anxiety. Notes: Abbreviations are defined in Appendix D.

Caregiver anxiety (STAI scores)

Eighteen studies comparing 16 non-pharmacological interventions were synthesized (Fig. 6A). The league table (Fig. 6C) indicated several interventions were associated with statistically significant reductions in caregiver anxiety compared to routine care. Specifically, ADVANCE (SMD = −0.37, 95% CrI −0.69 to −0.07), educational program (EP; 0.44, 0.15–0.72), and EP vs. ERC (–0.60, −1.15 to −0.07) showed statistically significant improvements in caregiver-reported STAI scores. Other comparisons yielded imprecise estimates with CrIs crossing zero. Cumulative ranking curves (Fig. 6B) identified the following top five interventions based on SUCRA values: EP (78.3%), Lollipop (76.7%), and ADVANCE (71.5%). These results suggest that distraction techniques and structured educational or engagement strategies are likely effective in alleviating caregiver anxiety. Full SUCRA rankings are provided in Supplemental Digital Content Appendix G, available at: http://links.lww.com/JS9/G647.

Figure 6.

Network geometry, cumulative rankings, and relative effects for parents’ anxiety (STAI). (A) Network plot of non-pharmacological interventions. Node size is proportional to total sample size; edge width reflects the number of direct comparisons. (B) Cumulative ranking curves (rank 1 = lowest anxiety). The area under each curve corresponds to SUCRA (higher SUCRA = better). (C) League table of pairwise relative effects on STAI scores: standardized mean differences (SMD) with 95% credible intervals (CrI). Read from row to column; negative SMD indicates lower anxiety (favors the row intervention). Estimates in bold are statistically significant (CrI excludes 0). Notes: Abbreviations are defined in Appendix D. Lower STAI indicates less parental anxiety.

Induction compliance checklist

Eleven studies comparing nine non-pharmacological interventions were synthesized (Fig. 7A). According to the league table (Fig. 7C), although many comparisons yielded imprecise estimates with wide CrIs crossing zero, several interventions demonstrated promising improvements in induction compliance compared to routine care. Notably, TVD (MD = −1.82, 95% CrI −4.98–0.81), VREP (–1.25, −3.58−1.05), and BERT (–1.55, −6.27−2.57) showed favorable directions of effect, although not reaching statistical significance. Cumulative ranking curves (Fig. 7B) yielded the following top five SUCRA values: TVD (78.7%), BERT (68.7%), and VREP (66.6%), suggesting that these interventions are most likely to improve compliance during anesthetic induction. Full SUCRA results are presented in Supplemental Digital Content Appendix G, available at: http://links.lww.com/JS9/G647.

Figure 7.

Network geometry, cumulative rankings, and relative effects for induction compliance (ICC scale). (A) Network plot of non-pharmacological interventions. Node size is proportional to total sample size; edge width reflects the number of direct comparisons. (B) Cumulative ranking curves (rank 1 = lowest compliance). The area under each curve corresponds to SUCRA (higher SUCRA = better). (C) League table of pairwise relative effects on ICC scores: mean differences (MD) with 95% credible intervals (CrI). Read from row to column; negative MD indicates lower compliance (favors the row intervention). Estimates in bold are statistically significant (CrI excludes 0). Notes: Abbreviations are defined in Appendix D. Lower ICC indicates less induction compliance.

Sensitivity analyses

Restricted low-complexity ENT/ophthalmology set

In sensitivity analyses restricting the network to low-complexity ENT/ophthalmology day-case surgeries (excluding MIRPE and cardiac catheterization), pooled effects and rankings were stable. For PAED score, SUCRA changes were small (median ΔSUCRA +1.1 pp; IQR +1.0 to +1.2; maximum +5.0 pp for MV, rank −4). For ED incidence, SUCRA shifts were minimal (median ΔSUCRA +0.5 pp; IQR +0.4 to +0.7; maximum 2.6 pp for MV, rank 0). See Appendix K (Table K1A–K1B; Figure K1A–K1B).

Age-stratified NMA

ED incidence

When the NMA was stratified by developmental stage, non-pharmacological interventions showed differential rankings, although CrIs were wide in the youngest and oldest groups. In children younger than 5 years, ice popsicle therapy occupied the top rank for reducing the risk of ED (RR 0.21, 95% CrI 0.01–2.22), but no head-to-head comparison achieved statistical significance. Among mid-childhood patients (5–8 years), VR immersion consistently demonstrated the greatest effect (RR 0.52, 95% CrI 0.19–1.44) and outperformed all other comparators in the league table. In the oldest subgroup (>8 years), MVO was highest ranked (RR 0.95, 95% CrI 0.20–4.51), again without significant pairwise differences. These age-specific league tables and the corresponding network plots are presented in Appendix L (Supplemental Digital Content Appendix L, available at: http://links.lww.com/JS9/G647).

ED severity

A parallel age-stratified NMA of ED severity, as quantified by PAED score, revealed a similar pattern. In the under-5 cohort, PVOR ranked first (MD −0.55, 95% CrI −3.55−2.41), though no intervention comparisons reached significance. For children aged 5–8 years, VR again provided the greatest mean reduction in PAED score (MD −0.44, 95% CrI −2.88−2.00), whereas MVO was highest ranked in those older than 8 years (MD −0.23, 95% CrI −2.14−1.67), with similarly nonsignificant CrIs. The full severity league tables and network geometries are likewise available in Appendix L (Supplemental Digital Content Appendix L, available at: http://links.lww.com/JS9/G647).

Outcome measurement timing subgroup analyses

We conducted subgroup analyses to evaluate whether variations in PAED outcome measurement timing influenced our findings. For severity of ED, MVO was most effective at early assessment (≤5 min post-emergence), while PVOR ranked highest at intermediate timings (10–15 min), although pairwise comparisons were not statistically significant. VR consistently ranked highest at delayed assessment (≥30 min). Regarding incidence of ED, MVO was optimal at early assessment, PI ranked highest at intermediate timings, and VR was again top ranked at delayed assessment, though differences remained nonsignificant. Detailed subgroup network diagrams and league tables are provided in Supplemental Digital Content Appendix M, available at: http://links.lww.com/JS9/G647. These results demonstrate stable intervention rankings across various PAED measurement timings, confirming that timing differences did not substantively alter our conclusions.

Consistency and publication bias assessment

Model consistency was assessed by comparing the DIC between consistency and inconsistency models. A difference in DIC of less than 5 across all closed loops was interpreted as indicating good consistency. Comparison-adjusted funnel plots (Supplemental Digital Content Appendix H, available at: http://links.lww.com/JS9/G647) showed no substantial asymmetry, indicating no significant small-study or publication bias.

Discussion

We conducted a comprehensive literature search and performed a NMA of 56 studies to evaluate the comparative effectiveness of various non-pharmacological interventions on ED, postoperative pain, ICC scores, and anxiety outcomes in pediatric surgical patients. VR and ice pops emerged as the most effective strategies for reducing ED incidence, with VR also demonstrating superior efficacy in postoperative pain relief. However, none of the assessed non-pharmacological interventions demonstrated a statistically significant reduction in ICC scores. PAPIA was identified as most effective intervention for reducing anxiety in children, while EP led the ranking for reducing caregiver anxiety.

Our NMA identified VR as the most effective intervention for reducing the severity of ED, while ice popsicle intervention was the most effective in reducing its incidence among pediatric patients. VR functions as a sensory distraction technique, immersing children in interactive environments that divert attention away from the surgical setting and alleviate anxiety levels^[34]. Neuroimaging studies have shown that VR activates brain regions involved in emotional regulation and pain perception, such as the prefrontal and anterior cingulate cortices, promoting a calmer postoperative state^[35,36]. Chen et al reported that immersive VR were associated with a significant reduction in the incidence and severity of ED compared to conventional video distraction methods^[37]. In contrast, ice popsicles induce cold-induced analgesia by activating TRPM8 receptors in the oral mucosa^[38]. These receptors transmit sensory signals to the spinal dorsal horn, where cold stimulation slows the nerve conduction velocity and inhibits pain transmission^[39]. This combined physiological and emotional effect – integrating analgesia with a familiar and pleasurable experience – helps reduce agitation and distress in pediatric patients. Our findings align with previous systematic reviews and RCTs supporting the use of sensory-based interventions to mitigate ED^[17,40]. Despite these promising results, the limited number of available trials warrants cautious interpretation. Nevertheless, our analysis reinforces the clinical value of VR and ice popsicles as adjunctive, low-risk strategies to enhance recovery and comfort in pediatric surgical populations.

Sensitivity analyses reinforced the robustness of our primary findings. When restricting the network to low-complexity ENT/ophthalmology day-case surgeries, effect estimates and SUCRA rankings remained stable across outcomes. This suggests that variations in surgical complexity did not meaningfully bias the comparative efficacy of non-pharmacological interventions. Furthermore, age-stratified analyses revealed developmental differences in intervention effectiveness. Ice popsicles appeared most effective for children under 5 years, VR was most beneficial for those aged 5–8 years, and MVO ranked highest among patients older than 8 years. Notably, subgroup analyses from previous systematic reviews have demonstrated that video-based interventions significantly reduce PAED scores among children under 7 years of age (P = 0.001), supporting the notion that younger children may be more responsive to sensory-based techniques^[17]. These trends, although not statistically significant in our analysis, suggest that developmental stage may influence the response to non-pharmacological strategies and warrant further investigation into age-personalized perioperative care. Finally, subgroup analyses based on the timing of PAED measurement showed consistent intervention rankings regardless of early, intermediate, or delayed assessments, confirming that timing heterogeneity did not substantially alter our conclusions.

In our study, non-pharmacological interventions had no significant impact on ICC scores – a finding consistent with earlier meta-analyses, including those assessing electric car transfers^[41]. However, a trial by Adriana et al reported improvements with a VR-based EP^[42], reflecting inconsistencies that may stem from methodological heterogeneity. These inconsistencies may stem from methodological heterogeneity across studies, including variations in participant characteristics, surgical procedures, and the timing and structure of the interventions. Although our results were not statistically significant, existing evidence suggests that perioperative non-pharmacological interventions may still influence induction behavior. Maintaining effective induction compliance has been shown to reduce preoperative anxiety, minimize ED, and enhance postoperative recovery trajectories. Notably, our NMA identified VR as the most effective intervention for reducing perioperative pain. This finding is supported by several high-quality RCTs demonstrating the analgesic effects of VR in pediatric patients undergoing painful procedures such as burn wound care^[43] and intravenous catheter placement^[44]. Postoperative pain in pediatric patients is primarily attributable to inflammatory responses, tissue injury, and procedural trauma. VR alleviates pain by diverting attention through immersive sensory input, modulating pain-related brain activity, and reducing anxiety-induced pain amplification^[36]. These mechanisms underscore the potential of VR as an adjunctive tool for perioperative pediatric pain management.

Despite advancements in surgical techniques and perioperative care, pediatric patients still experience adverse postoperative outcomes, contributing to higher healthcare costs and reduced quality of life for both children and their families. Previous meta-analyses have shown that non-pharmacological interventions can significantly reduce postoperative anxiety in both pediatric patients and their parents^[45,46]. Our findings highlight EP as the single most effective method to quell preoperative caregiver anxiety. EP works not through medication but through knowledge and support – clarifying each step of the surgical journey, equipping parents with coping skills, and offering a reassuring framework for dialog. This comprehensive psychosocial package alleviates the unknowns and insecurities that so often fuel parental distress^[47,48]. In contrast, PAPIA emerged as the most effective intervention for alleviating postoperative anxiety in children. This finding aligns with previous evidence, including a meta-analysis by Chen et al and highlights the psychological benefits of parental involvement during anesthesia induction^[49]. The PAPIA approach likely promotes emotional regulation through secure attachment, familiar sensory cues, and parental behavioral modeling^[50]. Nevertheless, our study did not detect a significant effect of non-pharmacological interventions on maintaining ICC scores. This may be attributed to methodological heterogeneity across studies, including variations in population characteristics, timing of interventions, and delivery protocols. Collectively, our findings suggest that interventions such as VR, ice pops, EP, and PAPIA may deliver multidimensional benefits across key perioperative outcomes, offering a safe and effective adjunct to standard pediatric surgical care.

Implementation feasibility in real-world setting

Although our analysis suggests potential benefits of immersive VR and ice-popsicle interventions, the practical adoption of these strategies requires careful consideration: VR access and maintenance – Commercial head-mounted displays can cost US $200–600 per unit and require regular disinfection, software updates, and staff training. In resource-constrained hospitals, lower-cost options (e.g., cardboard viewers with smartphones) or group-based tablet distraction could serve as pragmatic substitutes until dedicated VR hardware becomes affordable. Patient selection – VR may be unsuitable for children with severe visual impairment, epilepsy, or pronounced motion sensitivity; brief test sessions before induction are advisable. Infection-control workflow – Single-patient disposable facial interfaces or UV-C disinfection cabinets are recommended to mitigate cross-infection risks, especially in high-turnover day-surgery units. Ice-popsicle administration – Cold oral stimulation should be avoided in children at high risk of postoperative nausea/vomiting or with dietary restrictions (e.g., diabetics). Clear-liquid, sugar-free formulations, and timing ≥30 min after extubation can reduce aspiration and choking risks. Staffing and cultural factors – Successful implementation depends on nursing time, parental cooperation, and local dietary customs; formative implementation studies should accompany future efficacy trials.

Strengths and limitations

This study has several notable strengths. First, it represents the most comprehensive NMA to date evaluating non-pharmacological interventions for ED and related perioperative outcomes in pediatric populations. By synthesizing evidence from 56 trials across diverse geographic regions and clinical settings, this analysis offers robust comparative data on a wide range of interventions. Second, the use of a Bayesian framework enabled the integration of both direct and indirect evidence, improving the precision of effect estimates. Third, the application of SUCRA rankings provided a clinically meaningful hierarchy of interventions, facilitating evidence-based decision-making in perioperative care.

However, this study also has limitations. Many included trials had small sample sizes and moderate methodological quality, potentially compromising the reliability of certain effect estimates. In addition, there was considerable clinical heterogeneity across studies in terms of surgical procedures, intervention modalities, timing, and delivery. Clinical heterogeneity and potential confounding across trials are unavoidable in behavioral and perioperative research. To mitigate these concerns, we conducted multiple prespecified sensitivity analyses – including restriction to low-complexity ENT/ophthalmology surgery, age-stratified analyses, and timing-based analyses – which consistently demonstrated stable rankings and effect estimates, supporting the robustness of the conclusions.

To address concerns regarding within-node heterogeneity (e.g., VR, acupuncture), we conducted exploratory pairwise subgroup analyses (Appendix M). For VR interventions, we distinguished between immersive and interactive VR (e.g., head-mounted displays with environmental interaction) and non-immersive formats (e.g., passive video playback on monitors). For acupuncture, we compared transcutaneous electrical acupoint stimulation with manual needle acupuncture. However, due to the limited number of studies per subgroup, formal stratified network analyses were not feasible. Importantly, these subgroup analyses did not reveal statistically significant differences in ED incidence or severity, suggesting minimal impact on the robustness of node definitions or overall findings.

Additionally, some interventions such as lollipop and ice popsicle were evaluated in only a single study, precluding meaningful subgroup analysis. Although Bayesian models enable indirect comparison of multiple interventions, estimates for sparsely studied nodes may be imprecise, with wide CrIs. These findings should therefore be interpreted with caution.

Furthermore, all included studies were published in English, potentially introducing language bias. Most outcomes were measured using subjective, self-reported instruments, increasing the risk of measurement bias. Moreover, due to the nature of non-pharmacological interventions, participant and personnel blinding was often not feasible, contributing to a relatively high proportion of studies rated as having “some concerns” in the RoB 2 assessment.

Future large-scale randomized trials using standardized, objective outcome measures – and clearly reporting intervention content, timing, and dosage – are warranted to strengthen the current evidence base and allow for more granular modeling of treatment heterogeneity in NMA frameworks.

Conclusions

In this comprehensive Bayesian NMA of 56 studies, no single non-pharmacological intervention demonstrated definitive superiority across all perioperative outcomes in pediatric patients. VR showed the greatest potential for reducing the severity of ED and appeared to have the highest probability of alleviating postoperative pain. Ice popsicle interventions may be associated with a lower incidence of ED. For anxiety-related outcomes, parental presence during induction of anesthesia (PAPIA) was most effective in reducing children’s anxiety, while the EP for parents emerged as the top-ranked intervention for lowering caregiver anxiety. However, none of the included interventions showed a statistically significant improvement in ICC scores. Only 37.7% of the included studies were rated as having low risk of bias, and substantial clinical and methodological heterogeneity – such as variation in intervention techniques, delivery formats, and timing of outcome measurement – limits confidence in the comparative rankings. As such, these findings should be considered hypothesis-generating rather than definitive clinical guidance. Future research should focus on adequately powered, rigorously designed, and transparently reported RCTs. Harmonization of intervention definitions, standardization of outcome measurement timing, and incorporation of longer-term follow-up are needed to confirm or refute the apparent benefits of VR, ice popsicles, PAPIA, and other non-pharmacological strategies in pediatric perioperative care.

Ethical approval

This manuscript is a systematic review and network meta-analysis using publicly available literature.

Therefore, no ethical approval was required.

Consent

Not applicable. No patients or individual data were involved in this systematic review and meta-analysis.

Sources of funding

This study was supported by the Program for Excellent Research and Innovation Team of Higher Education Institutions of Anhui Province (2023AH010081).

Author contributions

C.L. and S.L.: Study conceptualization, methodology development, systematic literature review, data analysis, manuscript drafting, revision, and final approval. Y.W. and J.Z.: Data extraction, quality assessment, manuscript review, and approval. C.Z. and M.Z.: Statistical methodology, data interpretation, manuscript review, revision, and approval. H.C.: Supervision, critical revision of manuscript content, provided clinical insights, and approval. Y.Z. (Corresponding authors): Project oversight, critical manuscript review and revisions, and final approval for publication.

Conflicts of interest disclosure

The authors affirm that no actual or potential conflicts of interest exist in relation to this study, as it was conducted independently and without any commercial or financial affiliations.

Research registration unique identifying number (UIN)

Registered at PROSPERO: CRD42023452957.

Guarantor

Chengxiang Liu and Hong Chen.

Provenance and peer review

This paper was not commissioned; externally peer-reviewed.

Data availability statement

The data supporting this study are available from the corresponding author upon reasonable request.

Presentation

Preliminary findings of this study were not presented at any conference or meeting.