The Efficacy and Safety of Acupuncture for the Treatment of Children with Autism Spectrum Disorder: A Systematic Review and Meta-Analysis

Objectives. We aimed to summarize and critically evaluate the available evidence regarding the efficacy and safety of acupuncture for children with autism spectrum disorder (ASD). Methods. We searched 13 databases for studies published up to December 2016. Randomized controlled trials (RCTs) evaluating the efficacy of acupuncture for children with ASD were included. Outcome measures were the overall scores on scales evaluating the core symptoms of ASD and the scores for each symptom, such as social communication ability and skills, stereotypies, language ability, and cognitive function. Effect sizes were presented as mean differences (MD). Results. Twenty-seven RCTs with 1736 participants were included. Acupuncture complementary to behavioral and educational intervention significantly decreased the overall scores on the Childhood Autism Rating Scale (CARS) (MD −8.10, 95% CI −12.80 to −3.40) and the Autism Behavior Checklist (MD −8.92, 95% CI −11.29 to −6.54); however, it was unclear which of the ASD symptoms improved. Acupuncture as a monotherapy also reduced the overall CARS score. The reported adverse events were acceptable. Conclusions. This review suggests that acupuncture may be effective and safe for pediatric ASD. However, it is not conclusive due to the heterogeneity of the acupuncture treatment methods used in the studies.


Introduction
Autism spectrum disorder (ASD) is a neurodevelopmental disorder with an etiology that remains incompletely understood. The core symptoms of ASD include persistent deficits in social communication and interaction and restricted, repetitive patterns of behavior, interests, or activities [1]. The prevalence of ASD as reported in various studies ranges from as low as 1 in 500 to as high as 1 in 50 [2,3].
Many types of treatments are available for ASD, but none have yet been developed that effectively treat the core symptoms. Conventional treatments for ASD include pharmacological therapy and behavioral and educational interventions (BEI). However, risperidone, a commonly used medication for the treatment of maladaptive behaviors in ASD, has adverse effects such as weight gain, fatigue, drowsiness, and tremors [4,5]. Furthermore, the majorities of high quality BEI require 20 to 40 hours of treatment per week, and take a long time to show benefits [6][7][8].
The National Center for Complementary and Integrative Health describe "complementary" as a therapy used alongside conventional medicine, and "alternative" as a therapy used

Methods
This systematic review and meta-analysis was conducted according to the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions [22]. Reporting was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [23]. The protocol of this review was registered in the International Prospective Register of Systematic Reviews, PROSPERO (CRD42017054544).

Search Strategy.
Two authors (B. L. and J. L.) performed a comprehensive search of the following 13 electronic databases from their inception dates to 31 December 2016: six English databases (Cochrane Library, MEDLINE via EBSCO, EMBASE via Elsevier, AMED via EBSCO, CINAHL via EBSCO, and PsycARTICLES via ProQuest), two Chinese databases (Chinese National Knowledge Infrastructure (CNKI) and Wanfang Data), two Japanese databases (Japanese Institutional Repositories Online (JAIRO) and Citation Information by NII (CiNii)), and three Korean databases (Oriental Medicine Advanced Searching Integrated System (OASIS), Korean Traditional Knowledge Portal (KTKP), and KoreaMed). We also searched the reference lists of the relevant papers to identify additional trials. There was no restriction on language. In addition to the studies published in the journal, we also included grey literature, such as conference proceedings and degree theses.
The following search terms were used in MEDLINE: (autis * OR pervasive developmental disorder * OR childhood disintegrative disorder OR asperger * OR autism spectrum disorder OR child development disorders, pervasive) AND (acup * OR needl * OR trigger point OR body acupuncture OR scalp acupuncture OR tongue acupuncture OR auricular acupuncture OR pharmacopuncture OR bee OR acupuncture OR acupuncture therapy OR acupuncture points OR acupuncture, ear OR electroacupuncture). The details of search strategies used in all databases are presented in the Supplementary Material (Supplement 1).

Types of Studies.
We included randomized controlled trials (RCTs) that evaluated the efficacy of acupuncture for children with ASD, including those using a quasi-random method such as alternate allocation or allocation by birth date. We included both parallel and crossover studies.

Types of Participants.
We included patients with ASD under the age of 18 years, regardless of gender or race, diagnosed by standard criteria such as the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Diseases (ICD). We accepted diagnoses by assessment tools such as the Autism Diagnostic Observation Scale (ADOS), Autism Diagnostic Interview-Revised (ADI-R), Childhood Autism Rating Scale (CARS), Chinese Classification of Mental Disorder (CCMD), and other validated tools. Studies on ASD were included even if they did not refer to the diagnostic criteria.

Types of Interventions.
We included studies on acupuncture involving the insertion of needles into traditional acupoints or into nonmeridian points as experimental interventions, regardless of the acupuncture treatment method. Studies that did not involve skin penetration, such as those using acupressure, were excluded. Control interventions included pharmacological interventions, BEI, and CAM, including combinations of two or more therapy types. We also included sham acupuncture, which refers to a needle placed in an area close to but not into acupoints, as a control intervention. Studies in which the other treatments were applied to both experimental and control groups in the same manner were also included. We excluded trials that compared only different forms of acupuncture.

Types of Outcome
Measures. The outcome measures were as follows: (1) the overall score on scales evaluating the core features of ASD, including CARS, the Autism Behavior Checklist (ABC 1 ), the Aberrant Behavior Checklist (ABC 2 ), Evidence-Based Complementary and Alternative Medicine 3 the Autism Treatment Evaluation Checklist (ATEC), and the Ritvo-Freeman Real Life Rating Scale (RFRLRS); (2) social interaction skills, communication ability, or stereotypy, which are essential symptoms of ASD; and (3) language ability or cognitive function, which are associated features of ASD.

Study Selection and Data Extraction.
After eliminating duplicate studies, two authors (B. L. and J. H. C.) independently screened the titles and abstracts of the searched studies to identify those that potentially met the inclusion criteria. Thereafter, the texts of the remaining articles were obtained and screened for eligibility by two authors (J. H. C. and H. K. S.).
One author (B. L.) extracted the data from the selected studies, and a different author (S. H. C.) reviewed the data. Discrepancies were resolved through discussion with a third author (G. T. C.). The extracted information included study design, country, demographic characteristics of the participants, diagnostic criteria, details of the experimental and control interventions, outcomes, adverse events, and information for assessing the risk of bias. We contacted the primary authors of the included studies via email if additional information was needed.

Quality Assessment.
Two authors (J. H. C. and H. K. S.) independently evaluated the methodological quality of the included studies, using the Cochrane Collaboration's tool [22]. The following characteristics were assessed: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessments, incomplete outcome data, selective reporting, and other bias. Each domain was evaluated and categorized into three groups: "low risk," "unclear," or "high risk." If the relevant information was not mentioned in the article, we evaluated the domain as "unclear." Disagreements between the review authors over the risk of bias were resolved by discussion, with the participation of another author (J. L.).

Statistical Analysis.
For studies using the same type of intervention, comparator, and outcome measure, quantitative synthesis was conducted by meta-analysis using the Review Manager software, version 5.3 (Cochrane, London, UK). Descriptive analysis was conducted when the number of reported studies was only one, or when it was considered that heterogeneity was too high for the results to be synthesized. For the meta-analyses, risk ratios (RR) and 95% CIs were calculated for dichotomous data, and mean differences (MD) and 95% CIs were used for continuous data. We used prepost differences or end-point scores as outcome measures, presented for each included study. If among the included studies there were some measuring the change in a value, and others measuring the final value, of the same outcome measures, we planned to synthesize the data by calculating the final value as the initial value plus the change, if possible. We examined heterogeneity among the studies using the Higgins 2 test. We considered 2 ≥ 50% to be indicative of substantial heterogeneity, and 2 ≥ 75% to be indicative of serious heterogeneity. In the meta-analyses, a random effects model was used when the heterogeneity was significant, while a fixed effects model was used when the heterogeneity was not significant or the number of studies included in a metaanalysis was very small, in which case estimates of interstudy variance have poor accuracy [24].

Study Description
3.1.1. Literature Search. We identified 1763 records through a database search, and 10 additional records from the reference lists of relevant papers. After removing duplicates, 1519 records remained. A review of the titles and abstracts excluded 1404 records. After assessing the full text of the remaining 115 articles, we finally included 27 articles  in the systematic review and 17 articles [27, 29, 32-36, 38-43, 45-47, 49] in the meta-analysis (Figure 1). Table 1 describes the characteristics of the included studies. The most common diagnostic reference was the DSM. Five studies [25,[40][41][42][43] received institutional review board approval before the study was conducted. Consent forms were obtained from research participants in 22 studies [25-30, 32, 34-43, 46-49, 51]. Only 4 studies [40][41][42][43] registered the protocol before the trial.

Study Characteristics.
Experimental interventions involved manual acupuncture in 19 trials [25-28, 31-34, 36, 39, 42-48, 50, 51], electroacupuncture in 6 trials [30,35,37,38,41,49], and both manual acupuncture and electroacupuncture in 2 trials [29,40]. The most frequently used acupoints were GV20 [25,26,37,38,40,49] and EX-HN3 [26,29,37,38,40,41], used in six trials each. The depth of insertion ranged from 0.3 cm to 1.5 cun, and the heterogeneity was high, depending on the location of the acupoints. (Cun, a traditional Chinese unit of length, corresponds to 3.0303030 cm [52].) The most commonly used needle retention time was 30 minutes; the most frequent number of treatment sessions was 120; the most commonly used frequency and duration of acupuncture treatment were 5 times/week and 8 weeks. There was no follow-up study after treatment in any of the studies. Details of the acupuncture treatment methods used can be found in Table 2.

Quality Assessment.
Of the 27 trials included, only 3 [25,41,46] used sealed or opaque envelopes and were rated as having "low risk" of bias in allocation concealment. A further 6 studies [30,39,[48][49][50][51] were rated as "high risk" because they did not use appropriate concealment methods, and the remaining 18 studies [ Studies included in the meta-analysis (n = 17) were evaluated as "unclear" due to lack of information. Nine studies [25,26,29,36,[40][41][42][43]46] used adequate methods of random sequence generation, using a randomization table or computer-generated randomization number. Because it is difficult to blind participants and therapists to acupuncture manipulation procedures, all studies were evaluated as "high risk" in the performance bias category. Only five trials [29,[40][41][42][43] reported the blinding of outcome assessment, and the remaining studies [25-28, 30-39, 44-51] were evaluated as "unclear" due to lack of information. Twenty studies [25-27, 29, 30, 33-40, 42, 45-50] had "low risk" of attrition bias, but five [31,32,41,44,51] were "unclear" because they did not provide statistical methods to deal with missing values. In two studies [28,43], there were also dropouts, but they were rated "high risk" because of statistical methods using per-protocol analysis. Twenty-four studies [25-43, 45, 46, 49-51] were assessed as "low risk" in reporting bias, while two [47,48] reported only effective rates without presenting raw data, and one [44] did not report all scheduled results. We contacted the primary authors of the studies [47,48] to obtain the raw data via email, but did not receive the data. Except for the eight studies [25,26,[30][31][32][33][34]44] with possible baseline imbalances among participants in the experimental and control groups, the remaining studies [27-29, 35-43, 45-51] were evaluated as "low risk" in the other bias domain. Methodological assessments for each included study are presented in Figure 2.      One study [29] found that the total ABC 2 scores were improved in the acupuncture plus BEI group compared with the BEI group ( < 0.01), but in another study [40] the differences were not significant ( = 0.205). One study [29] did not present the standard deviation, thus quantitative synthesis could not be performed. In one study [40], which evaluated the overall RFRLRS score, no significant difference was found between the two groups ( = 0.752).
Of the two studies [38,49] that assessed social interaction skills using ABC 1 , Zhao et al. [49] found significant improvement in the acupuncture group compared with the controls ( = 0.019), while no such effect was found by Wang et al. [38] ( > 0.05). We could not do a metaanalysis because of considerable baseline variations between the studies. Wong [40] reported that there were no significant differences between the two groups when social interaction was assessed by a subscale of the ADOS ( = 0.987) or RFRLRS ( = 0.268).
There was no difference between the two groups in one study [40], in which subscales of the ABC 2 and ADOS were used to assess communication skills ( = 0.183, = 0.604; resp.).
(4) Stereotypy. There was no difference between the two groups in one study [40] in which a subscale of the ABC 2 was used to assess stereotypy ( = 0.484).
Two studies [38,49] evaluated language ability using a subscale of the ABC 1 , one of which [49] showed a significant improvement in the acupuncture plus BEI group compared with the controls ( < 0.05), while the other [38] did not ( > 0.05). We could not conduct a meta-analysis due to the heterogeneity of the baseline values. Zeng et al. [47] found a significant improvement in language ability in the acupuncture group compared with the controls, when assessed with TER calculated using the Psychoeducational Profile (PEP) ( < 0.05), while no significant difference was found in other studies evaluated with the Arabic language test [25], Peabody Picture Vocabulary Test [38], Reynell Developmental Language Scales (RDLS) [43], and RFRLRS [40].    [40,43] showed a comparative benefit of using acupuncture in conjunction with BEI (MD 3.01, 95% CI 0.99 to 5.04, 2 = 91%) (Figure 4). Two studies [31,47] reported greater improvements in cognitive function, evaluated by the PEP-revised and TER of the PEP, respectively, in the acupuncture plus BEI group than in the control group ( = 0.05, = 0.02; resp.).

Complementary Therapy to Pharmacotherapy.
One study [50] compared acupuncture plus pharmacotherapy to pharmacotherapy alone, and descriptive analysis was performed. The study showed that the addition of acupuncture to risperidone improved abnormal behaviors, including stereotypy assessed by the TER, compared with risperidone alone ( < 0.05).

Complementary Therapy to BEI Combined with Music
Therapy. Two studies [30,48] compared acupuncture plus a combination therapy including both BEI and music therapy, which is a form of CAM, to the combination therapy. Only descriptive analysis was performed, because of the difference in the outcome measures evaluated in the two studies.
(1) Overall Scores on Scales Evaluating the Core Features of ASD. Li and Liu [30] reported that the total CARS and ABC 1 scores were lower when acupuncture was added to the combination therapy (both < 0.01). Zeng et al. [48] reported that the TER calculated using the CARS was higher in the acupuncture group ( = 0.008).
(2) Social Interaction Skills. One study [48] reported that social interaction skills were not significantly different between the two groups, when assessed by subscales of the PEP3 ( > 0.05).
(3) Communication Ability. Zeng et al. [48] reported that communication abilities were not significantly different between the two groups, when assessed by subscales of the PEP3 ( > 0.05).
(4) Language Ability. Both studies measured language ability, using either a subscale of the PEP3 [48] or the Gesell Developmental Schedules [30], and both showed a significant improvement in the acupuncture group compared to the control group ( < 0.05, < 0.01; resp.).
(5) Cognitive Function. One study [48] reported that cognitive function was higher in the acupuncture group compared with the control group, when assessed by subscales of the PEP3 ( = 0.0005).
(1) Overall Scores on Scales Evaluating the Core Features of ASD. Two studies [34,46] found greater improvement in the overall CARS score after treatment in the acupuncture group than in the BEI group (MD −1.54, 95% CI −2.61 to −0.46, 2 = 0%) ( Figure 5).
The ABC 1 [34], and the TER calculated with the CARS [46], also showed a significant improvement in the acupuncture group compared to the control group (both < 0.05).
(2) Social Interaction Skills. Yang [46] measured social interaction skills using a subscale of the CARS but found no significant difference between the two groups ( > 0.05).
(3) Communication Ability. Yang [46] evaluated communication skills using subscales of the CARS. There were no significant differences in nonverbal communication abilities between the two groups ( > 0.05), but verbal communication abilities were significantly different in favour of the acupuncture group ( < 0.05).
(4) Language Ability. Gao [28] reported that language ability was significantly improved in the acupuncture group compared with the BEI group ( < 0.05), when measured by a subscale of the PEP.
(5) Cognitive Function. In one study [28], there was no difference between interventions when cognitive function was evaluated by a subscale of the PEP ( > 0.05).

Alternative Therapy to BEI Combined with Music
Therapy. Two studies [27,39] compared acupuncture with a combination therapy consisting of both BEI and music therapy. Both studies showed improved total CARS scores in the acupuncture group compared with the control group, after treatment (MD −6.86, 95% CI −7.96 to −5.75, 2 = 0%) ( Figure 5).

Acupuncture versus Sham
Acupuncture. Two studies [41,42] compared acupuncture plus BEI with sham acupuncture plus BEI.
(3) Communication Ability. One study [41] reported that there was no difference between the two groups when assessing communication ability using subscales of the ABC 2 ( = 0.849).

Other Effects of Acupuncture in ASD.
In addition to the above-mentioned outcome measures, acupuncture has other potential effects on ASD. First, acupuncture might improve social adaptation. Li and Liu [30] found that participants who received acupuncture showed improved ability for social adaptation, as measured by the Gesell Developmental Schedules, compared to those who did not receive acupuncture ( < 0.01). Wang et al. [37] found that the social adaptive quotient and TER calculated were higher when acupuncture was added to BEI than for BEI alone (both < 0.05). Second, acupuncture might improve sleeping habits. In a study that evaluated the TER using the Children's Sleep Habits Questionnaire [26], the rate for the acupuncture plus BEI group was higher than that of the BEI-alone group ( = 0.001).

Safety.
Six RCTs with 306 participants [35,[40][41][42][43]49] provided data on adverse events associated with acupuncture (Table 3). Four trials [35,42,43,49] reported that no adverse events occurred during the interventions. Wong and Chen [41] reported that some participants experienced minor superficial bleeding, crying, or irritability during acupuncture. Another study [40] reported one case of a worsening sleep pattern and one case of increased hyperactivity and ritualistic behavior.

Discussion
In the present study, we conducted a systematic review of 27 RCTs with 1736 participants, to evaluate the efficacy of acupuncture for children with ASD. The outcome assessment of our review was summarized in two aspects: improvement of overall symptoms of ASD (evaluated with the overall scores on scales such as CARS, ABC 1 , ABC 2 , ATEC, and RFRLRS, which are specific outcome measures for ASD) and improvement in each of the major symptom categories of ASD (including social interaction skills, communication ability, stereotypy, language ability, and cognitive function).
When acupuncture was added to BEI, our pooled results indicated that reductions in the overall scores on the CARS, ABC 1 , and ATEC, and TERs based on the overall scores on the CARS and ABC 1 , were significantly higher compared with the BEI-alone group. Acupuncture was not demonstrably effective in reducing the overall ABC 2 and RFRLRS scores as an adjunctive therapy, but the results were unreliable because only a very small number of studies evaluated this. As a complementary therapy for BEI, acupuncture therapy provided inconsistent results in improving the individual symptoms of ASD. This may have been a result of evaluating each symptom using nonspecific scales for ASD, such as the SPT or WeeFIM. There were many studies that evaluated the total scores on ASD-specific outcome measures, but the scores on the subscales that evaluated each individual ASD   symptom were rarely presented. Therefore, the number of studies evaluating each major symptom of ASD was insufficient for assessing statistical validity. The effect of acupuncture as a complementary therapy alongside pharmacotherapy, or as an adjunctive therapy to music therapy combined with BEI, could not be assessed because there were only one or two studies available. When acupuncture was used as an alternative therapy for BEI, or for BEI plus music therapy, the total CARS scores of the acupuncture group showed greater improvement than those of the control groups. There was only one study evaluating the efficacy of acupuncture for each of the ASD symptoms, such as social interaction skills, communication ability, language ability, and cognitive function. Therefore, the results concerning the efficacy of acupuncture as a monotherapy for each symptom of ASD should be interpreted with caution.
Our review found that acupuncture had no significantly different effect on any outcome compared to sham acupuncture. In addition, while the results indicate that acupuncture might improve social adaptability and sleeping habits, these results cannot be confirmed because of the limited number of studies.
According to TCM's holistic approach, stimulating the acupoints related to various organs or viscera of the human body through meridians corrects the disharmony and dysregulation of organ systems. In TCM, the pathogenesis of ASD is considered to be related to disharmony of the five viscera, six bowels, and meridians, mainly dysregulation of the heart, liver, spleen, kidney, and brain [15]. Dysregulation and disharmony of the organs associated with ASD may have been corrected through acupuncture treatment, and this may have resulted in an improvement in overall scores on ASDspecific outcome measures, such as the CARS and ABC 1 . On a scientific basis, the pathogenesis of ASD is incompletely understood. Acupuncture can modulate neurotrophic factors [19] and neurotransmitter systems such as serotonin [53], GABA [17], and glutamate [16], which are considered strong candidates for roles in ASD [18], and this could lead to improvement in the symptoms of ASD. Further investigations should be conducted to clarify the specific mechanisms of acupuncture in treating ASD, and to examine which core symptoms of ASD can be improved by acupuncture.
We did not find strong evidence to show that acupuncture is associated with any serious adverse events. However, only six RCTs (22.22%) reported acupuncture safety data. Therefore, we could not draw a firm conclusion about the safety of acupuncture for ASD.
Two systematic reviews regarding acupuncture for ASD [20,21] have been published during the last few years. Compared with the previous systematic reviews [20,21], which included 10 and 11 articles, respectively, our review included many additional published RCTs. In other words, by including 27 RCTs and 1736 participants, this is the most comprehensive and recent review of acupuncture for patients with pediatric ASD. Unlike previous studies, which referred to the treatment used in the control group as "conventional treatment," we analyzed the interventions used in the controls by classifying them as pharmacological interventions, BEI, and CAM. We could confirm that acupuncture as a complementary therapy improved the total CARS and ABC 1 scores, which are tools for measuring the key symptoms of ASD, but could not specifically determine the symptoms that improved. The total CARS scores before and after treatment were reported in ten studies [27, 30, 32-36, 39, 45, 46], but there was only one study [46] that showed the scores on subscales evaluating each symptom. Meanwhile, we did not find much improvement in the low methodological quality of the included studies. Furthermore, owing to the heterogeneity of the acupuncture methods used in each study, we could not draw any other conclusions.

4.1.
Limitations. This review has several limitations, and each should be considered seriously. First, we found that significant heterogeneity existed for several of the outcomes. This may have been caused by the heterogeneity of the recruited participants with ASD, as ASD comprises a range of symptoms and associated problems. Moreover, unlike other clinical trials, acupuncture clinical trials involve various factors that cause heterogeneity, such as selection of acupoints, frequency and duration of treatments, depths of insertion, and needle retention times. Although the treatments used in the control groups were classified in this review as pharmacological interventions, BEI, or CAM, the variety of BEI treatments may have caused heterogeneity. Second, the low methodological quality of some studies remains a challenge. Many studies showed performance bias, because acupuncture therapy is difficult to blind. The low methodological quality of some studies can lead to overestimation of the effects of acupuncture on ASD. Additionally, although we searched the literature regardless of language, most of the included studies were conducted in China (74.07%), which suggests a potential for location bias and limits the generality of this review. We intended to estimate publication bias with a funnel plot, but none of the outcome measures were used in more than 10 studies. Accordingly, we could not assess publication bias in this review.

Implications for Future Research.
After the publication of the DSM-5 in 2013 [1], a single diagnosis of "ASD" has replaced previous subtypes. ASD is a highly heterogeneous disorder that involves various symptoms. Therefore, it is necessary to identify the subgroups of ASD children that can benefit from acupuncture. To resolve this in detail, subscales as well as total scores on the indicators evaluating ASD, such as the CARS, ABC 1 , and ABC 2 , should be measured and presented to distinguish the symptoms that might be improved by acupuncture. Only four of the studies included in this review [25,36,42,43] used CARS scores as diagnostic criteria to specify the severity of the symptoms. In future studies, the severity of ASD should be specified in participants via appropriate inclusion criteria.
Since there were variations in the types of acupuncture used, and a lack of standardization in the selection of acupoints, future research must standardize and generalize acupuncture treatment for ASD. In the studies included in this review, GV20 and EX-HN3 were frequently used. The most frequent needle retention time was 30 minutes, the number of treatment sessions was 120, the frequency of treatment was 5 times/week, and the duration of treatment was 8 weeks. To accumulate evidence, we should select specific acupuncture treatment methods based on existing studies. Additionally, the included studies did not conduct follow-ups after the treatment ended. We suggest long-term clinical trials with longer follow-up periods that include serial measurements of outcomes to determine the effects of acupuncture over time.
To improve the methodological quality, a strict protocol should be planned prior to future trials. Moreover, we expect rigorous, well-designed clinical trials in full compliance with the Standards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) recommendations [54]. In this review, only six studies reported adverse events related to acupuncture, and all events were mild. To draw clear conclusions about the safety of acupuncture for pediatric ASD, future studies should measure and report all adverse events using guidelines such as the CONSORT for Harms Data Recommendations [55]. Moreover, parents of children with ASD experience considerable levels of stress [56]; therefore, we suggest that the quality of life of parents of children with ASD should be considered as an additional important outcome measure in future clinical trials.

Conclusion
The results of this systematic review indicated that acupuncture may improve the overall symptoms of ASD. The reported adverse events associated with acupuncture were mild and acceptable. However, due to heterogeneities among the trials and the low methodological quality of some of the studies, additional rigorous, well-designed RCTs with larger sample sizes should be performed to confirm these results.