Acupuncture for children with autism spectrum disorder: A systematic review and meta-analysis

Abstract

Background:

The aim of this study was to assess the efficiency and safety of acupuncture in core symptomatic improvement of children with autism spectrum disorder (ASD).

Methods:

We searched the following databases: Cochrane Library, PubMed, Embase, Medline, China National Knowledge Infrastructure (CNKI), Wanfang, Chinese Science and Technology Periodical (VIP) and Chinese Biological Medicine (CBM), from 1 January 2012 to 25 September 2022. The Autism Behavior Checklist (ABC), Childhood Autism Rating Scale (CARS), and Autism Treatment Evaluation Checklist (ATEC) were adopted as outcome indicators. Three reviewers independently assessed the risk of bias (ROB) and the Grading of Recommendations Assessment, Development, and Evaluation (GRADE)assessment. Utilizing Review Manager (RevMan) 5.3 and Stata 12.0, data were analyzed.

Results:

A total of 38 trials were included, and 2862 participants participated in qualitative synthesis and meta-analysis. Only 1 trial was assessed as having a low ROB, and 37 trials were assessed as having an overall high ROB. The quality of evidence for most indicators were considered very low by the GRADE criteria. The results showed that acupuncture groups might have a higher clinical effective rate than nonacupuncture groups (relative risk [RR] = 1.33,95% confidence interval [CI] = 1.25–1.41; heterogeneity: x²=18.15, P = .64, I² = 0%). Regarding changes in ABC scores, the acupuncture groups might exhibit greater decrease than nonacupuncture groups (MMD = −6.06, 95%CI = −7.25 to −4.87, P < .00001; heterogeneity: x² =73.37, P = .03, I² = 77%). In terms of changes in CARS score, acupuncture group may benefit more than nonacupuncture group (MMD = −3.93, 95%CI = 4.90 to −2.95, P < .00001; heterogeneity: x²=234.47, P < .00001, I² = 90%). Additionally, in terms of ATEC score, acupuncture groups showed more benefit than nonacupuncture groups (MMD = −10.24, 95%CI = −13.09 to −7.38, P < .00001; heterogeneity: x²=45.74, P = .04, I² = 85%). Both subgroup analysis and sensitivity analysis are existing heterogeneity. Only 1 RCT study involved adverse events with mild symptoms that did not interfere with treatment and evaluation.

Conclusion:

Children with ASD may benefit from acupuncture because of its effectiveness and safety. Nevertheless, given the low quality of the evidence for the assessed outcomes and the high ROB of analyzed trials, the results should be regarded with caution.

1. Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental disorder that begins early (usually before 3 years old) and is characterized by trouble in social communication, abnormal repetitive behaviors, narrow interests and atypical sensory sensitivity.^[1] On account of large-scale prevalence and surveillance research, it was estimated that approximately 1% to 2% of general population has ASD, although these studies are widely variable by country, age at an assessment, and other criteria.^[2–4] If children with ASD do not take effective intervention measures timely, it will have serious implications on their self-care ability and quality of life and prove a heavy burden indeed to the family on occasion. It is necessary to find practical methods to intervene.

Improving autistic children physical and mental function require multidisciplinary treatment,^[5] such as drug treatment including mental stimulant, atypical antipsychotics, antidepressants and alpha 2 adrenergic agonists. Although drug treatment comes into effect quickly, the side effect is bigger, which may result in children with malnutrition and skeletal dyplasia. It is also easy to relapse once the medication has been stopped.^[6,7] The use of behavioral interventions, such as applied behavior analysis (ABA),^[8] treatment and education of autistic and related communication handicapped children (TEACCH),^[9] music therapy (MT), and others, has been increasing. These interventions have certain effects on improving the symptoms and social emotions of children with ASD, but their long-term effects are still unknown. On balance, there is a lack of effective interventions to improve the core symptoms of children with ASD.

With the development of traditional Chinese medicine research in recent years, traditional Chinese medicine therapy represented by acupuncture has gradually entered the field of rehabilitation treatment for children with ASD. Some studies have shown that acupuncture on the prefrontal cortex (PFC) of autistic rats can improve the learning and memory ability, which may be closely related to the up-regulation of CX43, CX32, and CX36 gene expression levels in the frontal cortex.^[10] In addition, some studies have shown that acupuncture on the Zusanli (ST36) acupoint can reduce the expression of TXNIP and inhibition of NLRP3 inflammasome activation alleviates behavioral disorders associated with PFC in autistic rats.^[11,12] The safety and effectiveness of acupuncture for children with ASD has been implemented in a number of randomized controlled trials (RCTs). Compared to other conventional treatments, acupuncture is broadly accepted for its validity, accessibility, and low cost. Further analysis and discussion are needed due to the different methodological quality and the inconsistent results in various studies, especially in China.

Therefore, according to literature analysis, effective rate (significant effect rate + valid effect rate = effective rate) based on the results of scales, Autism Behavior Checklist (ABC), Childhood Autism Rating Scale (CARS) and Autism Treatment Evaluation Checklist (ATEC) were adopted as relatively reliable measures to assess the efficacy of acupuncture for children with ASD. Moreover, few meta-analyses were found recently. In order attain stronger evidence for clinical practice, this meta-analysis evaluated the effectiveness and safety of acupuncture for the treatment of children with ASD by contrasting its differences from other treatment interventions.

2. Methods

2.1. Registration and protocol

The International Prospective Register of Systematic Reviews (PROSPERO) has recorded the protocols for this meta-analysis and systematic review under registration number CRD42022362582.

2.2. Literature search strategy

We searched through 8 databases of literature on children with ASD to compare the effectiveness and safety of acupuncture with other treatment interventions. Without regard to language, 8 databases were examined between January 1, 2012 and September 25, 2022. Eight databases were available, 4 of which are in English (Cochrane Library, PubMed, Embase, and Medline) and 4 of which were in Chinese (China National Knowledge Infrastructure [CNKI] Database, China Biomedical Database [CBM], China Science and Technology Journals [VIP] Database, and Wanfang Database).

The search criteria used to find records in the English database included “autism spectrum disorder,” “autism,” “electroacupuncture,” and “acupuncture,” (Supplementary Appendix S1, http://links.lww.com/MD/I534). The Chinese database also contains “zibizheng,” “guduzheng,” “zhenjiu,” “dianzhen” and “zhenci” (Supplementary Appendix S2, http://links.lww.com/MD/I535). Before assessing the articles in accordance with the exclusion and inclusion criteria, we searched for publications in the aforementioned databases, read titles and abstracts in the preliminary screening process, and the complete texts in the last screening process.

2.3. Medical record report types

This systematic review includes all RCTs, regardless of language or publication status, that looked at the validity and security of acupuncture for children with ASD.

2.4. Participant types

It would be all unrestricted variables for children with ASD, regardless of their gender, race, ethnicity, duration of therapy, income level, or whether they were inpatients or outpatients. For children with ASD, we excluded individuals with a defined disease or condition, such as mental retardation, schizophrenia, childhood disintegrating disorder, Asperger syndrome, attention deficit and hyperactivity disorder (ADHD), Rett syndrome, and other disorders that are easily confused with ASD. The study was selected according to the following requirements: Age: 0 to 12 years; ASD was diagnosed according to DSM-5 criteria or by radiological examination; ASD diagnosis was evaluated according to standard diagnostic criteria or validity, ABC, CARS and ATEC scales; The probable causal relationship between ASD and ASD symptoms and their mutual relationship in clinical management.

2.5. Intervention types

There was also acupuncture or acupuncture-related combination therapy. RCTs of acupuncture combined with traditional Chinese medicine or moxibustion therapy was not included. Acupuncture procedures not frequently used in clinical practice, such as conventional acupoint injection or acupoint thread-embedding, ear acupuncture, and intradermal acupuncture were also eliminated. Instead, we selected procedures that are extensively utilized in clinical practice and Chinese acupuncture texts. Control interventions will be sham acupuncture and other rehabilitation programs.

2.6. Comparison types

The comparison includes common rehabilitation therapies, such as ABA, TEACCH, and MT. The combination of traditional Chinese medicine (such as herbal medicine and moxibustion) was not within the scope.

2.7. Outcome assessments

The effective rate, ABC, CARS, and ATEC scores were used as outcome measures to evaluate the validity of acupuncture on the improvement of ASD core symptoms. Therefore, any remaining results besides the 4 assessment criteria were excluded. Nevertheless, some trials included the scales described above, their statistical methods only included subscale score analysis, without summary scores, and did not include corresponding outcome indicators analysis. Besides, although some studies used other scales besides the above scales simultaneously, they were not included in the analysis because other scales were not commonly used.

2.8. Data extraction

Three researchers (TTL, SPL, and YCC) independently finished the full procedure of gathering and extracting data in accordance with susceptibility criteria. The researchers discussed divergent viewpoints until they were all in agreement. JY and LL worked as advisors. Finally, 38 trials were eligible for inclusion. The extracted information is consistently saved in Excel format. Title, author, year of publication, sample size (male/female), methods of intervention, methods of comparison, outcome indicators, and outcome consequences were primary categories of information that were retrieved.

2.9. Risk of bias assessment

Using the Cochrane Risk of Bias (ROB) tool, 3 researchers (TTL, YCC, and DXW) independently evaluated the ROB in the included trials, contending that the evidence was only valuable if the study accurate methodology was disclosed. JY resolved the discrepancy between the opinions of the 2 researchers. Seven distinct areas make up the Cochrane ROB tool, and each is rated as having a low, unclear, or high ROB based on assessment criteria. When more than half of the specific domains were evaluated as low ROB, it indicated that the trial had low ROB. However, when more than half of the specific domains were evaluated as high ROB or the ROB was unclear, it indicated that the trial had high ROB.^[13,14]

2.10. GRADE assessment

Three researchers (TTL, SPL, and YXZ) scored the Recommendation Assessment, Development, and Evaluation (GRADE) to determine the level of evidence for each outcome. JY coordinated when the opinions of the 3 researchers differed. To assess the degree of certainty of all pooled outcomes, the quality of the evidence for each outcome is presented in the form of GRADE criteria. According to 3 upgrading criteria (including large effect, plausible confounding, and dose response gradients) and 5 degrading criteria (including ROB, inconsistency, indirectness, imprecision, and disclosure bias), the GRADE criteria assessed the strength of the evidence for each outcome. Finally, there are 4 levels, respectively evaluated as high, moderate, low, or very low,^[15] to evaluate the quality of evidence.

2.11. Synthesis and evaluation of data

Review Manager (RevMan) 5.3 and Stata 12.0 were used to analyze the data. Response rate, mean, and standard deviation (SD) statistics were chosen because of the dichotomous and continuous nature of the data from the included trials. The included trials’ heterogeneity was examined using the x² test (α = 0.1), and the I² statistical approach was used to quantify the degree of heterogeneity. When P > .1 and I² < 50%, homogeneity between trials was taken into consideration, and the data used a fixed effects model (FEM) to evaluate. When P < .1 and I² > 50%, heterogeneity between trials should be concerned and the data used a random effects model (REM) to evaluate.^[14,16] For the purpose of addressing significant clinical heterogeneity, group analysis and sensitivity analysis were utilized. In order to evaluate for any potential publication bias, Egger test and funnel plots were used, and the results were summarized in the form of a forest plot.

3. Results

3.1. Search results

A total of 1048 studies in all were identified; 237 of them came from the Cochrane Library, 38 from PubMed, 42 from Medline, 116 from Embase, 82 from CBM, 154 from CNKI, 85 from VIP, and 294 from the Wanfang Database. In total, 38 studies were included in the meta-analysis after being found eligible following many layers of screening. Figure 1 shows the selection procedure for the literature.

Figure 1.

Flow chart of literature search and screening.

3.2. Characteristics of the included trials

Thirty-eight^[17–54] trials in China, were carried out with 3352 participants: 1467 were distributed into experimental groups (acupuncture groups), and 1395 were distributed into control groups (nonacupuncture groups). For each RCT, baseline data were gathered prior to the start of the study. Treatment lasted for 3 to 10 months. Of the 38 studies, 35 trials^{17–49,51,52} used manual acupuncture (MA) and only 3 trials^[50,53,54] used electroacupuncture (EA). Among them, 7 trials^{18,24,29,33,34,46,49} used Jin 3-needle acupuncture method (J3N), which is adopted by more researchers, and the other fragmented studies remained were method of syndrome differentiation and acupoint selection or different schools of acupuncture. The main assessment tools were overall effective rate, ABC and CARS; Twenty-two trials^{17,18,21–23,25,29,31–33,35,37,39,41,43,45,47,50–54} assessed effectiveness using effective rate, Eighteen trials^{17,18,20,22–24,28,35,36,40,41,43,45–48,51–53} assessed effectiveness using ABC, and 25 trials^{17,18,20,22–24,26,30,32–40,43,45,46,48,50–53} assessed effectiveness using CARS. Eight trials^{[18,19,26,30,35,42,43,52]} assessed effectiveness using ATEC, which was regarded as the secondary evaluation outcome. Table 1 presents the characteristics of the analyzed trials.

Table 1.

Characteristics of analyzed trials.

Authors	Yr	Age/yrs		Sex (male/female)		Disease course		Methods of intervention	Comparison	Treatment duration	Main outcomes
		O	C	O	C	O	C
Zhao et al[53]	2015	NA		NA		NA		EA + RT	RT(TEACCH)	3 mo	ABC,RR
Zeng et al[28]	2016	3.62 ± 1.36	3.60 ± 1.52	(36/24)	(16/9)	NA		A(J3N) + RT	RT(ABA,CT,TEACCH,SE)	6 mo	CARS,PEP-3,RR
Liao et al[36]	2016	9.24 ± 1.64	9.51 ± 1.50	NA		3.28 ± 1.61	3.26 ± 1.31	A + RT	RT(ABA,SIT)	6 mo	CARS,RR
Hu et al[43]	2016	4.8 ± 1.79	4.7 ± 1.87	(28/3)	(23/2)	NA		A + RT	RT(ABA,MT,TEACCH,SIT)	10 mo	PEP-3
Qiu et al[24]	2017	5.2 ± 1.5	5.4 ± 1.7	(14/11)	(16/9)	NA		A + RT	RT(ST,FT)	3 mo	CARS,PEP-3,RR
Jiang et al[38]	2017	8.34 ± 1.22	8.26 ± 1.18	(21/14)	(22/13)	4.44 ± 0.35	4.21 ± 0.17	A + RT	RT(SE,BCR)	3 mo	CARS,RR
Li et al[40]	2017	4.8 ± 0.53	4.46 ± 0.57	(28/17)	(30/15)	NA		A + RT	RT(MT,TEACCH)	3 mo	ABC,Gesell,RR
Zeng and Ouyang[50]	2017	5.21 ± 1.39	4.98 ± 1.24	(18/9)	(19/8)	NA		A + RT	RT(SE)	4 mo	ABC,CARS,RR
Tu and Li[16]	2018	5.17 ± 2.44	5.86 ± 2.52	(29/2)	(31/3)	NA		A + RT	RT(ABA,FTT,SIT)	3 mo	ABC,CARS
Zhang[32]	2018	6.75 ± 2.48	6.67 ± 2.43	(30/20)	(32/18)	NA		A(J3N) + RT	RT(MT,SE,ABA)	6 mo	CARS,RR
Zhou et al[35]	2018	4.89 ± 1.33	5.09 ± 1.31	(37/8)	(35/10)	2.20 ± 1.15	2.13 ± 1.18	A + RT	RT(ABA,EACCH,SIT)	NA	ABC,CARS,ATEC,PEP-3,SF-36
Ba et al[37]	2018	5 ± 1	5 ± 1	(18/7)	(16/9)	1.11 ± 0.54	1.16 ± 0.61	A + RT	RT(ABA,SIT)	3 mo	CARS
He,Wang et al[21]	2019	2.8 ± 0.66	2.8 ± 0.71	(31/9)	(23/6)	NA		A + RT	RT(CT,FT,MT)	6 mo	ABC,CARS,S-S
He,Liu et al[22]	2019	2.8 ± 0.69	2.8 ± 0.56	(30/7)	(23/3)	NA		A + RT	RT(CT,FT,MT)	6 mo	ABC,CARS,Gesell,S-S
Gao et al[29]	2019	NA	(22/8)	(25/5)	NA	A + RT	RT(BRC,ABA)	3 mo	CARS,ATEC
Wang et al[46]	2019	3.81 ± 0.64	3.85 ± 0.57	(29/16)	(25/12)	NA		A + RT	RT	4 mo	ABC, Oral function Assessment
Zhou et al[48]	2019	3.21 ± 1.12	3.31 ± 1.14	(14/16)	(13/17)	NA		A(J3N) + RT	RT(TEACCH,ABA,SIT)	6 mo	CARS,PEP-3
Dang et al[17]	2020	4.74 ± 0.75	4.69 ± 0.77	(32/11)	(34/9)	1.60 ± 0.29	1.57 ± 0.25	A(J3N) + RT	RT(MT,BRC,TEACCH,SIT)	6 mo	ABC,ATEC,PEP-3,WPPSI
Zou et al[49]	2020	NA		(40/15)	(39/16)	NA		A + RT	RT(ABA,MT,TEACCH,SIT)	3 mo	CARS,EEG,BI
Wang et al[47]	2020	4 ± 1	4 ± 1	(22/13)	(20/15)	2.06 ± 0.82	1.94 ± 0.83	A + RT	RT(SE)	5 mo	ABC,CARS,DQ,BI
Zhao et al[20]	2021	2.8 ± 1.3	2.5 ± 1.5	(27/9)	(25/11)	NA	A + RT	RT(SE,BRC,CT)	3 mo	ALC,RR
Huang et al[23]	2021	NA	NA	NA	A(J3N) + RT	RT(TEACCH,ABA)	3 mo	ABC,CARS,SRS,CSHQ
Feng et al[25]	2021	6.79 ± 2.16	7.15 ± 2.28	27/16	25/14	NA	A + RT	RT(TEACCH,ABA,SIT,SE)	3 mo	CARS,ATEC
Yu et al[27]	2021	5.12 ± 2.42	5.46 ± 2.01	29/24	30/23	2.38 ± 1.30	2.45 ± 1.65	A + RT	RT(TEACCH)	3 mo	ABC,ATEC,PEP-3
Lai et al[30]	2021	NA		(24/6)	(25/5)	NA		A + RT	RT(SIT, CT, rTMS)	6 mo	WPPSI,RR
Zhang et al[34]	2021	8 ± 2	8 ± 2	(18/10)	(16/11)	5.11 ± 1.50	5.07 ± 1.54	A + RT	RT(SIT)	6 mo	ABC,CARS,ATEC,RR
Yu[41]	2021	6.7 ± 1.1	3.8 ± 1.2	(16/12)	(17/11)	3.3 ± 0.4	3.1 ± 0.5	A + RT	RT(BRC,TEACCH)	3 mo	ATEC,
Ren et al[45]	2021	4.13 ± 1.24	4.61 ± 1.33	(30/24)	(31/23)	NA		A(J3N) + RT	RT(TEACCH)	4 mo	ABC,CARS,Gesell,S-S
Lin et al[52]	2021	6.21 ± 2.09	6.14 ± 2.15	(14/11)	(12/13)	NA		EA + RT	RT(ABA,SE,OT,TEACCH)	4 mo	ABC,CARS
Dang et al[18]	2022	2.82 ± 0.56	2.81 ± 0.57	(24/6)	(23/7)	NA		A + RT	RT(ABA,CT,TEACCH,SE)	NA	ATEC,PEP-3,S-S
Zhang et al[19]	2022	4.11 ± 0.8	4.18 ± 0.87	26/21	28/21	2.46 ± 1.23	2.21 ± 1.24	A + RT	RT(TEACCH,BRC,SIT)	6 mo	ABC,CARS,PEP-3
Huang et al[33]	2022	3.35 ± 1.39	3.26 ± 1.51	20/13	19/13	2.42 ± 1.25	.39 ± 1.41	A + RT	RT(CT,SIT)	3 mo	Gesell,PEP-3
Tang[31]	2022	8 ± 1	8 ± 1	(42/8)	(41/9)	1.90 ± 0.20	1.85 ± 0.25	A + RT	RT(SIT,TEACCH,MT)	6 mo	CARS,PEP-3,Gesell,PedsQLTM4.0,RR
Hu et al[26]	2022	7 ± 3	7 ± 4	(45/39)	(47/37)	NA		A(J3N) + RT	RT(ABA)	NA	CARS,PEP-3
Liu et al[39]	2022	4.52 ± 0.45	4.70 ± 0.59	(24/20)	(25/19)	NA		A + RT	RT(SIT,SE)	3 mo	ABC,CARS,PEP-3,ABC,WPPSI
Zhou[42]	2022	8.02 ± 2.52	8.89 ± 2.78	(20/15)	(21/14)	6.11 ± 1.03	6.56 ± 1.32	EA + RT	RT(TEACCH,SIT)	3 mo	ABC,CARS,ATEC
Wang et al[44]	2022	5 ± 0	5 ± 0	(28/15)	(27/16)	2.01 ± 0.35	1.99 ± 0.33	A + RT	RT(ABA,MT,TEACCH,SIT)	6 mo	ABC,CARS,SM,BI
Lin et al[51]	2022	4.6 ± 0.7	4.7 ± 0.7	(21/9)	(23/7)	1.8 ± 0.4	1.8 ± 0.4	A + RT	RT(TEACCH)	3 mo	ABC,CARS,ATEC,RR

A = acupuncture, ABA = Applied Behavior Analysis, ABC = Autism Behavior Checklist, ALC = Assessment form of language and communication effect in children with autism, ATEC = Autism Treatment Evaluation Checklist, BI = Biochemical indexes, BRC = Behave remedy conditioning, CARS = childhood autism rating scale, CE = conductive education, CSHQ = Children’s Sleep Habits Questionnaire, CT = Cognitive training, DQ = development quotient, EA = electroacupuncture, Gesell = Gesell Developmental Schedules, FT = family training, J3N = Jin’s 3 needling, MT = music therapy, pedsQLTM4.0 = pedsQLTM multidimensional fatigue scale 4.0, PEP-3 = Psycho-educational Profile-3, RR = response rate, RT = rehabilitation therapy, S-S = sign- significant relation, SE = special education, SF-36 = SF-36 health scale, SIT = sensory integration training, SRS = Social Responsiveness Scale, TEACCH = treatment and education of autistic and related communication handicapped children, WPPSI = Wechsler preschool and primary scale of intelligence.

3.3. ROB assessment

Only 1 trial^[52] out of the 38 was evaluated to have a low ROB, while the other 37 were evaluated to have a high ROB. Figure 2 displays the ROB for the experimental population.

Figure 2.

(A) Risk of bias graph. (B) Risk of bias summary.

3.4. GRADE assessment

Based on GRADE criteria, the quality of evidence for effective rate was considered low; the quality of evidence for ABC, CARS, and ATEC scores were considered very low due to inadequate allocation concealing, a lack of participants and personnel blinding (although this is challenging to accomplish in the design and implementation of experiments), ambiguous selective reporting, uncertain blinding of outcome assessments and significant heterogeneity (Table 2).

Table 2.

GRADE assessment of effective rate, ABC, CARS, and ATEC in acupuncture on children with ASD.

Certainty assessment							No. of patients		Effect		Certainty	Importance
No. of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Acupuncture	Nonacupuncture	Relative (95% CI)	Absolute (95% CI)
Effective rate
22	Randomized trials	Very serious*	Not serious†	Not serious	Not serious	None	708/815 (86.9%)	499/760 (65.7%)	RR 1.33 (1.25–1.41)	217 more per 1000 (from 164 more to 269 more)	⨁⨁◯◯ Low
ABC score
18	Randomized trials	Very serious*	Serious‡	Not serious	Not serious	None	701	672	-	MD 6.06 lower (7.25 lower to 4.87 lower)	⨁◯◯◯ Very low
CARS score
25	Randomized trials	Very serious*	Very serious§	Not serious	Not serious	None	694	673	-	MD 3.88 lower (4.95 lower to 2.8 lower)	⨁◯◯◯ Very low
ATEC score
8	Randomized trials	Very serious*	Serious∥	Not serious	Not serious	None	282	274	-	MD 10.24 lower (13.09 lower to 7.38 lower)	⨁◯◯◯ Very low

ABC = autism behavior checklist, ATEC = autism treatment evaluation checklist, CARS = childhood autism rating scale.

^*Inadequate allocation concealing, a lack of participants and personnel blinding, ambiguous selective reporting, uncertain blinding of outcome assessments.

^†x² = 18.15, P = .64, I² = 0%.

^‡x² = 73.37, P = .03, I² = 77%.

^§x² = 234.47, P < .00001, I² = 90%.

^∥x² = 45.74, P = .04, I² = 85%.

3.5. Primary outcomes

3.5.1. Effective rate.

Twenty-two trials^{[17,18,21–23,25,29,31–33,35,37,39,41,43,45,47,50–54]} reported the efficacy of acupuncture versus no acupuncture in treating ASD. Using FEM, the acupuncture groups may have a higher clinical response rate than the nonacupuncture groups (relative risk [RR] = 1.33, 95%CI = 1.25–1.41) with low heterogeneity between 2 groups (x² =18.15, P = .64, I² = 0%). The effective rate of acupuncture groups and nonacupuncture groups was statistically significant (Z = 9.72, P < .00001; Fig. 3).

Figure 3.

Forest plot of effective rate.

3.5.2. ABC score.

Eighteen trials^{[17,18,20,22–24,28,35,36,40,41,43,45–48,51–53]} using the ABC scale reported that acupuncture could reduce the total scores of ABC compared with nonacupuncture. The heterogeneity of ABC score (x² =73.37, P = .03, I² = 77%) was high in the included trials. Using REM, and MMD values was −6.06 (95%CI: −7.25 to −4.87), the analysis indicated a statistically significant difference existed in ABC scores between the acupuncture groups and the nonacupuncture groups (Z = 9.96, p0.00001; Fig. 4).

Figure 4.

Forest plot of ABC scores. ABC = autism behavior checklist.

According to the subgroup analysis, 2 subgroups indicated that the ABC scores of the acupuncture groups were not statistically different from those of the nonacupuncture groups (published before 2017:MMD = −6.11, 95%CI = −14.85 to 2.64, P = .17; published in 2020: MMD = −8.69, 95% CI = −17.78 to 0.40, P = .06); The other 4 subgroups in ABC scores of the acupuncture groups compared with the nonacupuncture groups showed statistically significant (published in 2018: MMD = −4.99, 95%CI = −8.61 to −1.36, P = .007; published in 2019: MMD = −7.09, 95% CI = −9.49 to −4.86, P < .00001; published in 2021: MMD = −6.03, 95% CI = −8.12 to −3.94, P < .00001; published in 2022: MMD = −5.37, 95% CI = −8.00 to −2.74, P < .00001). Five subgroups showed high heterogeneity of ABC scores published before 2017: x² =19.89, P < .00001, I² = 95%; published in 2018: x² =8.67, P = .003, I² = 88%; published in 2020: x² =9.95, P = .002, I² = 90%; published in 2021: x² =15.79, P = .003, I² = 75%; published in 2022: x² =8.88, P = .003, I² = 66%), only 3 trials published in 2019 showed no heterogeneity (x² =0.34, P = .84, I² = 0%; Fig. 4).

3.5.3. CARS score.

Twenty-five trials^{17,18,20,22–24,26,30,32–40,43,45,46,48,50–53} using CARS have reported that acupuncture can reduce the total score of CARS when comparing with nonacupuncture. Using REM, the CARS score (x² =234.47, P < .00001, I² = 90%) of the included trials was highly heterogeneous and MMD value was −3.93 (95%CI: −4.90 to −2.95). A statistically significant difference was observed between the acupuncture groups and the nonacupuncture groups in CARS score (Z = 7.91, P < .00001; Fig. 5).

Figure 5.

(A) Forest plot of CARS scores-based on published year. (B) Forest plot of CARS scores-based on average age. CARS = childhood autism rating scale.

3.5.3.1. Subgroup A.

According to the analysis of the subgroup analysis based on time of publication, the CARS scores of the acupuncture groups when comparing to nonacupuncture groups decreased statistically significantly in all 6 subgroups (published prior 2017:MMD = −5.88, 95%CI = −10.25 to −1.51, P = .008; published in 2018: MMD = −2.71, 95%CI = −3.54 to −1.87, P < .00001; published in 2019: MMD = −2.37, 95% CI = −3.56 to −1.18, P < .00001; published in 2020: MMD = −4.68, 95% CI = −6.57 to −2.79, P < .00001; published in 2021: MMD = −3.09, 95% CI = −5.37 to −0.81, P = .008; published in 2022: MMD = −4.89, 95%CI = −5.93 to −3.85, P < .00001). Four subgroups showed high heterogeneity of CARS scores published before 2017: x² =43.40, P < .00001, I² = 95%; published in 2020: x² =2.37, P = .12, I² = 58%; published in 2021: x² =50.88, P < .00001, I² = 92%; published in 2022: x² =22.42, P = .001, I² = 73%), only 4 trials published in 2018 (x² =2.84, P = .42, I² = 0%) and 3 trials in 2019 showed no heterogeneity (x² = 0.05, P = .97, I² = 0%; Fig. 5A).

3.5.3.2. Subgroup B.

All 3 categories demonstrated statistically significant differences in CARS scores between acupuncture groups and nonacupuncture groups in subgroup analysis based on average age of inclusion (the average age < 6 years: MMD = −3.28, 95%CI = −4.10 to −2.46, P < .00001; the average age ≥ 6 years: MMD = −5.37, 95%CI = −6.71 to −4.02, P < .00001; not mention: MMD = −1.40, 95%CI = −1.40 to −0.46, P < .00001). Two subgroups showed high heterogeneity in ABC scores between acupuncture groups and nonacupuncture groups (the average age < 6 years: x² =30.90, P = .002, I² = 61%; the average age ≥ 6 years: x² =73.72, P < .00001, I² = 89%; not mention: x² =1.20, P = .27, I² = 17%; Fig. 5B).

3.6. Secondary outcomes

Eight trials^{18,19,26,30,35,42,43,52} using ATEC assessment reported that acupuncture could reduce the total score of ATEC when comparing with nonacupuncture. The heterogeneity of ATEC scores (x² =45.74, P = .04, I² = 85%) was high in the included trials. Using REM, MMD value was −10.24 (95%CI: −13.09 to −7.38). The analysis indicated a statistically significant difference existed in the ATEC scores between acupuncture groups and nonacupuncture groups (Z = 7.04, P < .00001; Fig. 6).

Figure 6.

Forest plot of ATEC scores. ATEC = autism treatment evaluation checklist.

According to the publication year of study, the ATEC scores of the acupuncture groups were statistically significant compared with the nonacupuncture groups (published before 2020: MMD = −10.10, 95%CI = −14.94 to −5.27, P < .00001; published in 2021: MMD = −7.63, 95% CI = −10.58 to −4.68, P < .00001; published in 2022: MMD = −12.86, 95%CI = −15.33 to −10.39, P < .00001). Two subgroups showed heterogeneity in ATEC scores published before 2020: x² =0.29, P = .59, I² = 0%; published in 2021: x² =3.67, P = .16, I² = 46%), and 3 trials published in 2022 (x² =45.74, P = .04, I² = 69%) showed no higher heterogeneity (Fig. 6).

3.7. Sensitivity analysis

According to the sensitivity analysis of the exclusion approach, the heterogeneity of trials published in 2021 was mainly generated from Lin^[53] and that of trials published in 2022 was mainly generated from Zhou^[43] in CARS with the year of publication as the subgroup analysis condition. After removal of these 2 trials, the heterogeneity of 2 subgroups decreased (published in 2021: x² =3.49, P = .32, I² = 14%; published in 2022: x² =5.10, P = .4, I² = 2%), but showed no obvious decrease in overall heterogeneity (x² =152.71, P < .00001, I² = 86%). The main source of heterogeneity of ATEC scale was Zhou,^[43] and the overall heterogeneity was reduced after removing this experiment (x² =9.75, P = .14, I² = 38%). After further analysis of these 2 trials,^[43,53] we found that both trials involved EA intervention in ASD, which may have resulted in a different result compared with manual acupuncture. The sensitivity analysis of ABC scale was subgroup analysis according to acupuncture manipulation, age, course of disease, publication bias and other factors, and the improvement was not apparent.

3.8. Publication bias

The effective rate of 22 trials (P = .95), ABC score of 18 trials (P = .33) and CARS score of 25 trials (P = .093) were used as the outcome indicators, and funnel plots were made to analyze publication bias. The results of the Egger test indicated that there was no discernible publication bias. Eight trials’ ATEC scores, however, manifested an obvious publication bias (P = .009; Supplementary Appendix S3, http://links.lww.com/MD/I536).

3.9. Safety of acupuncture

Only 1 of the 38 RCTs included adverse events; the others did not. Wang^[48] reported 2 cases of adverse reactions during treatment in acupuncture group, 1 case of ecchymosis at the acupuncture site, and 1 case of itching after acupuncture. The symptoms are mild, the treatment is timely, and there is no great impact on study.

4. Discussion

4.1. Main finding summary

For the aim of comparing the validity and security of acupuncture to nonacupuncture, this meta-analysis evaluated 38 RCTs with 3352 participants. A comparison of effective rate among 1575 subjects enrolled in 22 RCTs showed that acupuncture improved autistic symptom function before and after treatment. Comparison of ABC scores among 1373 subjects enrolled in 18 RCTs showed that acupuncture improved autistic symptom functioning before and after treatment. The CARS scores of 2041 participants in 25 RCTs were compared before and after treatment to see if acupuncture improved autistic symptom function. A subgroup analysis revealed that the differences in the average age of the included treatments, the type of acupuncture therapy, and the study timing of publication were the main causes of heterogeneity. A comparison of CARS scores before and after treatment in 2041 subjects enrolled in 25 RCTs showed that acupuncture improved autistic symptom. Subgroup analysis indicated the existed differences in average age of the included treatments, the type of acupuncture therapy, and the publication time of the study were the main sources of heterogeneity. After sensitivity analysis, Lin^[53] and Zhou^[43] were found to be the main reasons for the heterogeneity of the publication time in 2021 and 2022. The mean CARS scores showed that the onset age of ASD was less than 6 years old and timely intervention treatment would improve the course of ASD more than 6 years old.

In addition, acupuncture improved the functioning of autistic symptoms, as demonstrated by a comparison of ATEC scores before and after treatment in 556 participants in 8 RCTs. A subgroup analysis revealed that the timing of the study publication was a factor in the source of heterogeneity. The study of Lin^[53] and Zhou^[43] was the primary contributor to the heterogeneity, according to the sensitivity analysis of the exclusion approach. ATEC showed publication bias, which may be related to consistent source trials and the small number of trials. Although some outcome indicators were analyzed with high heterogeneity, their clinical significance cannot be ruled out because I² is not an absolute measure of heterogeneity.^[55]

One RCT out of the 38 included an adverse event, while the others did not. Mild local subcutaneous hematoma and itching were 2 adverse effects mentioned in RCTs. However, because they are minor, the symptoms do not affect treatment or evaluation.

4.2. Present evidence applicability

Our meta-analysis indicated that acupuncture combined with rehabilitation training is superior to rehabilitation training alone. Moreover, when paired with related scale analyses, the ABC, CARS, and ATEC scores in the groups receiving both rehabilitation training and acupuncture was lower than those in the groups receiving only rehabilitation training. Acupuncture combined with rehabilitative exercises may increase effectiveness even further. The safety of acupuncture intervention in ASD is further demonstrated by the fact that there were not many negative side effects throughout therapy. The effectiveness and safety of acupuncture over the long-term and its capacity to decrease the recurrence rate are difficult to evaluate due to the inconsistent use of acupuncture location and duration in previous research and the dearth of long-term and follow-up efficacy data.

4.3. Limitations and deficiency

The conclusions of the review should be regarded cautiously due to a number of limitations. Firstly, the majority of the trials that were considered had high overall ROB levels. Only 1 of the 38 trials was determined to have a low ROB (2.6%), while the remaining 37 were determined to have a high risk (97.4%) based on the risk assessment. Only 1 RCT was performed for double blindness, 37 RCTs were not performed for double blindness. In general, 97.37% trials have not specified the assignment and 39.47% of the method of random sequence generation have not been specified. In only 2 of the 38 trials had the same trained physician or researcher assess the outcomes, while 10 results had a professional physician, therapist, or specialist evaluation team measure the outcomes. Overall, 97.37% did not describe the blinding of the outcome assessor. Disease course was not reported in 22 trials, and age or sex was not reported in 6 trials; as a result, the quality of the trials is low.

Secondly, the quality of evidence for effective rate was considered low; the quality of evidence for ABC, CARS, and ATEC scores were considered very low due to inadequate allocation concealing, a lack of participants and personnel blinding, ambiguous selective reporting, uncertain blinding of outcome assessments and substantial heterogeneity. Due to the specific nature of acupuncture treatment, it is difficult to blind participants and personnel, which influences the entire quality of evidence when assessed using GRADE criteria.

Thirdly, a significant portion of the RCTs included were published in Chinese, and the trials were all completed in China, despite the fact that we did not exclude publications published in other languages during the search process. As a result, source bias and a lack of adequate representation in this study were inadequate.

Finally, clinical trials with larger sample sizes are needed to validate these findings, as some of the studies included in this meta-analysis had limited sample sizes. To further increase the accuracy of evaluation, it is also considerable to use more efficient and advanced methods or assessment tools, such as the Autism Diagnostic Observation Schedule scale^[56,57] and functional near-infrared spectroscopy.^[58]

4.4. Consensus and divergence from other literature

A preceding meta-analysis^[59] evaluated the effectiveness of scalp acupuncture for children with ASD and described a promising research direction for further exploration on the potential value of acupuncture. Regarding the primary outcomes, the characteristics of the literature that was included, and the evaluation indicators employed for children with ASD, this study differs from previous systematic reviews. Previous studies have not been considered to include more literature analysis, and this paper conducted a subgroup analysis of relevant literature to further classify each part of the acupuncture test. A variety of scale indicators to assess the improvement of core symptoms of autism were incorporated into the analysis, which can provide different indicators for reference analysis. Moreover, whereas 26 of the 38 RCTs analyzed in this evaluation were published after 2019, none of the RCTs analyzed in the prior systematic analysis were published after 2018. There is a wide range of characteristics among the trials that were analyzed.

5. Conclusion

Based on the results of response rate, ABC, CARS, and ATEC scores, acupuncture could be inferred to be safe and efficient in enhancing fundamental ASD functionality. While acupuncture may be used as a complementary therapy for children with ASD, it is also important to take into account the issue of whether children will cooperate and accept EA. In the short term, the use of acupuncture in combination to different rehabilitation therapies including ABA, TEACCH, and MT may increase efficacy even more. Because of the high ROB in the included trials and the low quality of the data supporting the evaluation conclusions, these results should, nevertheless, be treated with caution. Future clinical trials should be designed from multi-centers, using high-quality randomized, double-blind, controlled test method with larger sample sizes in order to assess acupuncture validity. Meanwhile, a more standardized and united principle should be established in acupuncture operation processes.

Author contributions

Conceptualization: Li Li, Jin Yu.

Data curation: Shaoping Lin, Yuecai Chen, Yunxuan Zhao, Dexin Wang, Li Li.

Funding acquisition: Jin Yu.

Investigation: Shaoping Lin.

Methodology: Tingting Lun, Dexin Wang, Li Li.

Project administration: Jin Yu.

Validation: Yunxuan Zhao.

Visualization: Yunxuan Zhao.

Writing – original draft: Tingting Lun, Shaoping Lin.

Writing – review & editing: Tingting Lun, Shaoping Lin, Yuecai Chen, Jin Yu.