Medical treatment of autism spectrum disorder in children: Current evidence, controversies, and clinical challenges

Abstract

BACKGROUND

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental condition associated with debilitating comorbidities [e.g., aggression, irritability, gastrointestinal (GI) issues]. Medical management primarily targets these symptoms, as no drug is Food and Drug Administration-approved for core social-communication deficits.

AIM

To synthesize the efficacy and safety of five major pharmacological classes and evaluate the emerging evidence for biomarker-driven (precision medicine) interventions in pediatric ASD.

METHODS

Following PRISMA guidelines, we systematically reviewed randomized controlled trials (RCTs) for five classes: Atypical antipsychotics, stimulants, selective serotonin reuptake inhibitors, metabolic/nutritional, and microbiota-gut-brain axis agents. Quantitative meta-analysis for antipsychotics (n = 5 RCTs pooled) used the random-effects model, reporting I² to quantify heterogeneity.

RESULTS

Atypical antipsychotics are the only drugs with robust, established efficacy for severe irritability: Pooled analysis for risperidone (n = 3 RCTs) showed a significant mean difference of approximately -11.0 on Aberrant Behavior Checklist-Irritability subscale (I² approximately 72%). Risperidone carries a greater metabolic burden (e.g., weight gain) than aripiprazole. Stimulants and selective serotonin reuptake inhibitors, respectively. Emerging therapies demonstrate targeted potential: Microbiota transfer therapy significantly improved GI and behavioral symptoms in cohorts with GI disease. Similarly, the efficacy of High-dose folinic acid was concentrated in the subgroup with folate receptor-α autoantibodies.

CONCLUSION

The management of ASD demands a shift to a precision medicine model, as the efficacy of interventions is highly variable and concentrated in specific patient subgroups. Future research must prioritize the validation of biological biomarkers (metabolic, genetic, neurophysiological) to reliably predict treatment response, guiding the selection of targeted therapies, and addressing current evidence gaps.

Key Words: Autism spectrum disorder; Pharmacological; Pediatric; Risperidone; Aripiprazole; Fecal microbiota transplantation; Microbiota-gut-brain axis; Precision medicine; Methylcobalamin; Folinic acid

Core Tip: The core finding of this systematic review is that the medical management of autism spectrum disorder is primarily focused on controlling debilitating comorbid symptoms (e.g., aggression, gastrointestinal distress) rather than improving core social deficits, for which no medication is approved. The field is rapidly transitioning from a broad, empirical treatment approach to one of precision medicine. This shift is driven by evidence that highly targeted interventions, such as microbiota transfer therapy for children with gastrointestinal abnormalities.

INTRODUCTION

Autism spectrum disorder (ASD) is a complex and heterogeneous neurodevelopmental condition characterized by persistent deficits in social communication and interaction, alongside restricted and repetitive patterns of behavior, interests, or activities[1]. The global prevalence of ASD has risen considerably over recent decades, with current estimates indicating that approximately one in every 36 children is affected, according to the Centers for Disease Control and Prevention. This growing prevalence underscores the escalating public health and socioeconomic impact of ASD on families, healthcare systems, and society at large[2]. While behavioral and educational interventions remain the cornerstone for addressing the core social and communication deficits of ASD, they often prove insufficient for managing the wide range of associated medical and psychiatric comorbidities. These comorbidities - including irritability, aggression, severe anxiety, sleep disturbances, seizures, and persistent gastrointestinal (GI) abnormalities - are the features that most substantially impair the child’s daily functioning and quality of life[3]. Consequently, the primary therapeutic goal of medical interventions in ASD is to complement behavioral approaches by targeting these specific, debilitating comorbid symptoms and their underlying biological manifestations, thereby enhancing adaptive function and responsiveness to educational programming.

The medical management of ASD remains complex and controversial. Pharmacologic agents like risperidone and aripiprazole [the only Food and Drug Administration (FDA)-approved medications for ASD-associated irritability] show effectiveness but are limited by metabolic and neurological side effects[4]. Other established drug classes, including stimulants, antidepressants, and anticonvulsants, vary widely in their effectiveness and tolerability. Concurrently, a broad range of biomedical and metabolic interventions - such as folinic acid, methylcobalamin, and gut-brain axis-modulating therapies - have been investigated[5]. Additionally, novel and experimental methods, including vasopressin receptor antagonists and Bumetanide, are currently under investigation, often raising ethical and methodological concerns due to their early stage of development[6].

Despite the proliferation of studies across these diverse classes, the available evidence is fragmented, inconsistent, and often of variable methodological quality, ranging from randomized controlled trials (RCTs) to small-scale or uncontrolled studies. This heterogeneity has contributed to significant confusion among clinicians and necessitates a comprehensive and critical synthesis of current evidence to inform rational, evidence-based clinical decision-making.

Therefore, this systematic review aims to address this knowledge gap with the following primary and secondary objectives. The primary objectives were to evaluate the efficacy and safety of established pharmacological agents [antipsychotics, stimulants, and selective serotonin reuptake inhibitors (SSRIs)] for managing ASD comorbidities in children and to critically synthesize the evidence for emerging biomedical and targeted interventions (metabolic/nutritional, microbiota-gut-brain axis, and neurophysiological) and their potential role in a precision medicine model. The secondary objectives were to clarify ongoing controversies regarding the efficacy, safety, and ethical justification of both established and experimental therapies and to identify key clinical and research challenges and guide future research priorities in this rapidly evolving field.

MATERIALS AND METHODS

Protocol registration

This systematic review was prospectively registered in the International Prospective Register of Systematic Reviews under the title “Medical Treatment of Autism Spectrum Disorder in Children: Current Evidence, Controversies, and Clinical Challenges” (Registration ID: CRD420251175162). The review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines.

Eligibility criteria

Population: We included children and adolescents (≤ 18 years) diagnosed with ASD using the Diagnostic and Statistical Manual of Mental Disorders or the International Classification of Diseases criteria. Studies involving adults only or non-human subjects were excluded (detailed criteria are summarized in Supplementary Table 1).

Interventions: Any medical or pharmacological intervention for ASD was eligible, including pharmacologic agents, nutritional/dietary supplements, metabolic support therapies, and gut-brain axis modulation strategies. We excluded purely behavioral, educational, or psychotherapeutic interventions without a medical component.

Comparators: Eligible comparators included placebo, standard care, or other pharmacological/biomedical interventions.

Study designs: We included RCTs, quasi-experimental studies, and meta-analyses. Observational studies (cohort studies) were included for safety/tolerability and for emerging interventions for which RCTs are lacking. Case reports, narrative reviews, and editorials were excluded.

Language and date restrictions: Only studies published in English were included, with no restriction on publication date.

Information sources and search strategy

A comprehensive literature search was conducted across the following electronic databases: PubMed, MEDLINE, EMBASE, Scopus, PsycINFO, Cochrane Central Register of Controlled Trials, and Web of Science. Search terms combined controlled vocabulary and free-text terms related to ASD and medical or pharmacological treatments, including specific terms such as “antipsychotics”, “methylcobalamin”, “probiotics”, “immunotherapy”, and “stem cell therapy”. Additional studies were identified through manual citation tracking and clinical trial registries.

Study selection

Two reviewers independently screened titles, abstracts, and full-text articles using predefined eligibility criteria. Discrepancies were resolved by consensus or consultation with a third reviewer. The screening process followed the PRISMA flowchart.

Data extraction

Data extraction was performed independently by two reviewers using a standardized data collection form to capture study characteristics, participant details, intervention and comparator types, outcome measures, and main findings.

Risk of bias and quality assessment

The risk of bias was evaluated using the Cochrane Risk of Bias 2.0 tool for RCTs and the Newcastle-Ottawa Scale for observational studies. The overall certainty of evidence will be assessed using the GRADE approach across five domains: Risk of bias, inconsistency, indirectness, imprecision, and publication bias.

Data synthesis

While all full-text RCTs were included for qualitative review, studies were excluded from the quantitative meta-analysis if they did not report sufficient statistical data [e.g., mean and SD, or an adequate sample size (n) to calculate an effect size (standardized mean difference (SMD) or mean difference (MD)].

For all quantitative syntheses, continuous outcomes were pooled using the SMD, and dichotomous outcomes were pooled using the relative risk (RR), both with 95% confidence intervals (CIs). All meta-analyses were performed using the DerSimonian-Laird random-effects model to account for the anticipated clinical and methodological heterogeneity inherent in ASD trials. Statistical heterogeneity was quantified by the I² statistic (I² > 50% indicating substantial heterogeneity).

Publication bias assessment

Publication bias was assessed using visual inspection of funnel plots for outcomes with at least 10 included studies. When appropriate (i.e., a sufficient number of studies), Egger’s regression test was used to statistically examine funnel plot asymmetry.

Outcomes

The primary outcome is the efficacy and safety of medical and pharmacological interventions in children with ASD. Secondary outcomes include improvements in behavioral, cognitive, social, and adaptive functioning, as well as adverse events and tolerability.

Ethics and dissemination

As this review is based solely on published data, ethical approval is not required. The findings will be disseminated through publication in a peer-reviewed journal and presentations at relevant academic conferences.

RESULTS

We studied different pharmacological and biological interventions for children with autism. Figure 1 shows the PRISMA flow chart of the study. We classified the interventions into the following categories.

Figure 1  The PRISMA flow chart of the study.

Antipsychotics

Antipsychotic medications, acting across dopaminergic and serotonergic pathways, are widely used to manage behavioral and psychiatric symptoms in ASD. Risperidone and aripiprazole remain the only FDA-approved treatments for irritability in children with ASD, while several other agents - including quetiapine, ziprasidone, olanzapine, haloperidol, pimozide, paliperidone, clozapine, and lurasidone - have been explored with limited or mixed evidence[7]. Robust RCTs and subsequent meta-analyses support the efficacy and safety of the FDA-approved agents. Supplementary Tables 2-5 summarize the key characteristics and findings of the primary RCTs for these two agents.

FDA-approved antipsychotics

Risperidone: Risperidone, a dopamine D₂ and serotonin 5-hydroxytryptamine receptor type 2A (5-HT₂A) antagonist, has been approved by FDA since 2006 for irritability in autistic children aged 5-16 years[8]. It is highly effective for aggression, self-injury, and severe tantrums[9], with strong support from the pivotal Research Units on Pediatric Psychopharmacology (RUPP) trial and multiple RCTs[9-13]. Table 1 summarizes the different features of using risperidone for children with ASD. The RUPP study demonstrated a 56.9% reduction in Aberrant Behavior Checklist-Irritability subscale (ABC-I) vs 14.1% with placebo (P < 0.001) and a 69% responder rate (vs 12% placebo)[9], with sustained benefit at 6-21 months[13]. Meta-analyses confirm significant reductions in irritability, hyperactivity, and stereotypy[14]. Studies from Oman and Egypt replicated improvements on Clinical Global Impression-Improvement (CGI-I), Childhood Autism Rating Scale (CARS), and Autism Treatment Evaluation Checklist (ATEC)[15,16], and an Iranian RCT showed CARS improvement that relapsed after discontinuation unless adjunctive therapy was added[17]. Improvements in core ASD features are limited, though long-term follow-up showed modest gains in social withdrawal and Vineland social skills[13,18].

Table 1 Risperidone for pediatric autism spectrum disorder.

Feature	Key findings
Indication	FDA-approved for irritability (aggression, self-injury, tantrums) associated with autistic disorder in ages 5-16 years
Efficacy	Strong evidence for reducing irritability, aggression, self-injury, and tantrums. Moderate evidence for reducing hyperactivity and stereotypy. Limited/inconsistent evidence for improving core social-communication symptoms
Key adverse effects	Weight gain: Significant, often rapid, linked to increased appetite; risk increases with duration and possibly younger age. Metabolic changes: Increased risk of insulin resistance, metabolic syndrome; changes in glucose, lipids, leptin, and adiponectin observed. Sedation/somnolence: Common, especially initially, often mild-moderate and transient. Hyperprolactinemia: Frequent, dose-dependent. Enuresis: Increased risk observed in long-term use. Tremor: Increased risk, though EPS is generally low
Cognitive effects	Appears to have no detrimental effect; some studies suggest potential minor improvements in attention/recognition memory in testable children
Pharmacokinetics/TDM	Sum trough concentration (risperidone + 9-hydroxyrisperidone-risperidone) correlates with efficacy and side effects (weight gain, sedation, prolactin). Proposed TDM target range: 3.5-7.0 μg/L to balance efficacy and weight gain. Simulation studies suggest TDM can improve achievement of the target range and potentially reduce side effects

FDA: Food and Drug Administration; EPS: Extrapyramidal symptoms; TDM: Therapeutic drug monitoring.

Adverse effects include rapid and significant weight gain (e.g., +2.7 kg in 8 weeks; +5.4 kg in 24 weeks) with metabolic abnormalities such as increased insulin, homeostasis model assessment of insulin resistance, alanine aminotransferase, leptin, and metabolic syndrome[19-22]. Sedation is common early in treatment. Hyperprolactinemia is frequent and dose-dependent, documented in Thai, Omani, and Dutch cohorts[15,23,24]. Extrapyramidal symptoms (EPS) rates are low (approximately 8%-9%), though tremor risk is increased[13,14,25]. Cognitive function is not impaired and may improve slightly. Therapeutic drug monitoring (TDM) studies show that higher active-moiety trough levels correlate with better efficacy but more weight gain and hyperprolactinemia, supporting a proposed therapeutic range of 3.5-7.0 μg/L[24,26]. Pharmacogenetic data remain inconclusive, though brain-derived neurotrophic factor (BDNF) Val66Met may predict insulin resistance[27,28]. Table 2 summarises comparison of risperidone effects vs placebo in children with ASD based on extracted data analysis.

Table 2 Risperidone vs placebo in pediatric autism spectrum disorder (based on extracted data analysis).

Outcome measure	No. studies (pooled)	Result (pooled estimate)	95% confidence interval	I2 heterogeneity	P value	Notes
Efficacy (ABC change)		Mean difference (random-effects)				Negative mean difference favors risperidone
ABC Irritability	3	-11.08	(-14.39 to -7.78)	72% (substantial)	< 0.0001	Robust effect, but high heterogeneity
ABC Stereotypy	3	-3.91	(-6.37 to -1.44)	55% (moderate/substantial)	0.002
ABC Hyperactivity	3	-7.97	(-12.26 to -3.69)	81% (high)	0.0003
ABC Social Withdrawal	3	-0.55	(-0.85 to -0.24)	12% (low)	0.0005	Small but consistent effect
ABC Inappropriate Speech	3	-2.27	(-3.85 to -0.70)	41% (moderate)	0.005
Safety (adverse events)		Risk ratio (random-effects)				Risk ratio > 1 indicates higher risk with risperidone
Increased appetite	3	2.45	(1.29-4.65)	30% (moderate)	0.006	Common risk
Somnolence	2	4.14	(1.81-9.47)	0% (low)	< 0.001	Consistent, common risk
Drooling	2	4.70	(1.52-14.54)	65% (substantial)	0.005	Significant, but variable risk
Tremor	2	8.22	(1.56-49.82)	22% (low)	0.01	Significantly increased risk
Fatigue	3	2.18	(0.70-6.85)	78% (high)	0.18	Trend towards higher risk, high variability
Constipation	2	1.31	(0.10-16.38)	0% (low)	0.84	No significant difference
Weight gain (continuous)		Mean difference (kg) (random-effects)				Positive mean difference indicates more weight gain with Risperidone
Mean weight gain diff (kg)	3	1.97	(1.52-2.41)	0% (low)	< 0.0001	Significant and consistent weight gain over approximately 8 weeks

Pooled estimates are calculated using the random-effects model to account for clinical and methodological heterogeneity. The I² statistic explicitly quantifies heterogeneity. High heterogeneity (I² > 50%) suggests that the actual effect size varies significantly across studies, cautioning against a single, universal interpretation. These results highlight the trade-offs associated with risperidone: Its apparent efficacy in addressing challenging behaviors comes at the cost of significant, but variable, adverse effects, particularly metabolic ones (low I² for weight gain), which necessitate careful patient selection and monitoring. Therapeutic drug monitoring may offer a path to optimize this balance. ABC: Aberrant Behavior Checklist.

Aripiprazole: Aripiprazole - a dopamine D₂/5-HT₁A partial agonist and serotonin 5-hydroxytryptamine receptor type 2A antagonist - was FDA-approved in 2009 for irritability in ASD (ages 6-17)[29]. Seven trials were reviewed; three (Owen 2009, Marcus 2009, Ichikawa 2017) were pooled in a meta-analysis[30-32], with four additional studies contributing to qualitative synthesis[33-36]. Aripiprazole significantly reduces irritability, with a pooled MD of -5.71 on ABC-I (95%CI: -8.38 to -3.04; P < 0.0001)[30-32], with improvements seen by week 1-2. CGI-I response rates reached 52% vs 14% with placebo (Marcus 2009)[31] (Figure 2A). Benefits extend to hyperactivity and stereotypy, with sustained effects up to 52 weeks in open-label follow-up[33,36]. Effects on core ASD symptoms remain limited; drug concentrations do not predict effectiveness[35].

Figure 2 Forest plot. A: Pooled meta-analysis of aripiprazole efficacy on Aberrant Behavior Checklist-Irritability (8 weeks). Forest plot showing the mean difference in change from baseline in the Aberrant Behavior Checklist-Irritability subscale across randomized controlled trials evaluating aripiprazole in children and adolescents with autism spectrum disorder. Negative values indicate greater improvement in irritability compared with placebo. Squares represent study-level effect sizes with 95% confidence intervals, scaled by inverse-variance weighting. The diamond represents the pooled random effects estimate. A vertical line at mean difference = 0 indicates no treatment effect. All included studies demonstrated greater symptom reduction with aripiprazole than placebo; B: Pooled risk ratio for adverse events associated with aripiprazole in autism spectrum disorder. Forest plot summarizing the pooled risk ratios of adverse events reported in clinical trials of aripiprazole for irritability associated with autism spectrum disorder. Effect sizes are displayed on a logarithmic scale. A vertical reference line at risk ratios = 1.0 denotes no difference in adverse event risk between aripiprazole and placebo. Squares represent individual study risk ratios with 95% confidence intervals, proportional to study weight. The pooled random-effects estimate is shown as a diamond. Values to the right of the reference line indicate an increased likelihood of adverse events with aripiprazole. ABC: Aberrant Behavior Checklist; RRs: Risk ratios.

Aripiprazole exhibits linear pharmacokinetics and a prolonged elimination half-life (> 75 hours), enabling once-daily dosing. It is extensively metabolized by cytochrome P450 2D6 and cytochrome P450 3A4 to its active metabolite, dehydro-aripiprazole, which contributes to the overall clinical effect[37]. Aripiprazole is generally well tolerated. Pooled analysis shows increased somnolence (RR = 2.89), fatigue (RR = 4.17), increased appetite (RR = 2.88), and modest weight gain (+0.98 kg in 8 weeks)[30-32] (Figure 2B). Long-term data report +4.6 kg weight gain by 52 weeks[33]. Unlike risperidone, aripiprazole lowers prolactin. EPS (akathisia, tremor) occurs more often than placebo but remains manageable[30-32]. Corrected QT interval effects are minimal. Higher trough concentrations predict metabolic adverse effects - higher body mass index z-scores and hemoglobin A1c - highlighting a potential role for TDM in safety monitoring[35]. Table 3 provides a direct comparison, demonstrating that the two agents have comparable efficacy for irritability but distinct safety profiles. Risperidone is associated with higher risks of weight gain and hyperprolactinemia, whereas aripiprazole is associated with a higher risk of EPS (akathisia, tremor) but a more favorable metabolic and endocrine profile.

Table 3 Comparison between risperidone vs aripiprazole in pediatric autism spectrum disorder.

Feature	Risperidone	Aripiprazole
FDA approval	Approved in 2006 for irritability (aggression, self-injury, tantrums) in children with autistic disorder aged 5-16 years	Approved in 2009 for the same indication in children and adolescents aged 6-17 years
Mechanism of action	Dopamine D2 and serotonin 5-HT2A receptor antagonist	Dopamine D2 and serotonin 5-HT1A partial agonist, 5-HT2A antagonist
Core clinical target	Severe behavioral symptoms - irritability, aggression, self-injury, tantrums	Same behavioral symptoms; sometimes preferred for milder irritability or when metabolic risk is a concern
Efficacy (ABC-I)	Robust reduction (approximately 50%-60%) vs placebo in multiple RCTs (RUPP 2002, Shea 2004). Effects sustained up to 21 months	Similar magnitude of improvement (approximately 40%-60%) vs placebo; benefits evident within 1-2 weeks and maintained up to 52 weeks
Effects on other ABC subscales	Significant improvements in hyperactivity, stereotypy, and inappropriate speech; modest gains in social withdrawal	Moderate improvement in hyperactivity and stereotypy; inconsistent effects on social withdrawal
Effect on core ASD symptoms	Minimal or inconsistent improvement in social and communication domains; benefits, mainly indirect via behavior control	Similar - limited impact on core ASD features, though better adaptive engagement may follow behavioral improvement
Weight gain	Significant, often rapid (mean +2.7 to +5.4 kg in 8-24 weeks); linked to increased appetite and metabolic changes	Milder (mean +1.3 kg in 8-12 weeks); generally, not associated with metabolic syndrome
Metabolic effects	Marked rise in insulin, glucose, HOMA-IR, leptin, and fall in adiponectin; increased risk of metabolic syndrome	Minimal changes in glucose or lipid parameters; low metabolic liability overall
Prolactin	Increases prolactin (dose-dependent); may cause galactorrhea, gynecomastia	Usually reduces or normalizes prolactin due to partial D2 agonism
Sedation/somnolence	Common (40%-60%); often transient but dose-related	Mild to moderate; usually transient; less sedation than risperidone
EPS	Low-moderate risk; tremor approximately 8%-9%, dose-related	Low risk overall; akathisia slightly more frequent than with risperidone
Cognitive effects	No cognitive decline; some studies show mild improvement in attention/recognition	Neutral cognitive profile; no significant impairment reported
Pharmacokinetics	Metabolized by CYP2D6 → active metabolite 9-hydroxyrisperidone-risperidone; TDM useful (target sum trough 3.5-7 μg/L)	Metabolized by CYP2D6 and CYP3A4 → active dehydro-aripiprazole; TDM not routinely required
Pharmacogenetics	CYP2D6 poor metabolizers have higher active moiety levels; BDNF Val66Met linked to insulin resistance	CYP2D6 poor metabolizers show higher exposure; limited evidence of clinical impact from DRD2/HTR variants
Duration of benefit	Sustained efficacy with continued use; relapse on discontinuation	Sustained up to 1 year; relapse possible on abrupt discontinuation
Overall clinical impression	Highest efficacy for irritability and aggression, but greater metabolic and endocrine burden	Comparable efficacy, better metabolic profile, slightly higher akathisia risk; favorable long-term tolerability

This table provides a critical head-to-head comparison of the two Food and Drug Administration-approved atypical antipsychotics for pediatric autism spectrum disorder: Risperidone and aripiprazole. Both drugs are highly effective, demonstrating robust, comparable efficacy in reducing severe, challenging behaviors such as irritability, aggression, and tantrums (the primary target of Food and Drug Administration approval), with effects maintained over time. Their contrasting tolerability profiles primarily drive the choice between them. Risperidone is associated with a greater metabolic and endocrine burden, specifically significant weight gain and marked increases in prolactin (leading to potential side effects like gynecomastia). Aripiprazole, while showing comparable efficacy, has a significantly better metabolic profile (less weight gain, minimal change in glucose/Lipids) and tends to reduce prolactin levels. The trade-off is that aripiprazole carries a slightly higher risk of akathisia (a type of extrapyramidal symptom or extrapyramidal symptoms). Clinically, risperidone is often chosen for patients requiring the highest reduction in severe aggression. At the same time, aripiprazole is preferred for patients with pre-existing metabolic risk factors or those with milder symptoms, where the priority is to minimize metabolic side effects.

FDA: Food and Drug Administration; 5-HT₂A: Serotonin 5-hydroxytryptamine receptor type 2A; 5-HT₁A: Serotonin 5-hydroxytryptamine receptor type 1A; ABC-I: Aberrant Behavior Checklist-Irritability subscale; RCTs: Randomized controlled trials; RUPP: Research Units on Pediatric Psychopharmacology (a study group); ABC: Aberrant Behavior Checklist; ASD: Autism spectrum disorder; HOMA-IR: Homeostasis model assessment of insulin resistance; EPS: Extrapyramidal symptoms (motor side effects like tremor or dystonia); CYP2D6: Cytochrome P450 2D6 (a liver enzyme involved in drug metabolism); TDM: Therapeutic drug monitoring (measuring drug levels in blood); CYP3A4: Cytochrome P450 3A4 (a liver enzyme involved in drug metabolism); BDNF: Brain-derived neurotrophic factor; Val66Met: Valine to Methionine substitution at position 66 (a polymorphism in the brain-derived neurotrophic factor gene); DRD2: Dopamine D2 receptor; HTR: 5-hydroxytryptamine receptor (serotonin receptor).

Non-approved antipsychotics

Haloperidol has shown short-term benefits for irritability, stereotypy, and social withdrawal, with stronger effects in older children or those with higher IQ. However, its use is limited by sedation and EPS risk[38,39]. Evidence for pimozide is mixed: A small open-label study showed broad behavioral improvement with minimal side effects[40], whereas a double-blind trial found benefit mainly for sleep and excretion disorders, without significant behavioral change[41].

Paliperidone - risperidone’s active metabolite - shows preliminary benefit for irritability, aggression, and hyperactivity, particularly in individuals who previously responded to risperidone, with a tolerability profile similar to but possibly slightly milder than risperidone[42,43]. Clozapine is reserved for severe, treatment-resistant aggression or self-injury, with case series showing notable improvement but significant risks including agranulocytosis, seizures, and metabolic complications, requiring intensive monitoring[44,45]. Lurasidone has shown modest improvement in irritability and global functioning in early studies, with a relatively favorable metabolic profile; however, evidence remains limited. Therefore, larger RCTs are needed[46].

Cannabis-based medicine intervention

The use of cannabis-based medicines (CBMs), particularly those rich in cannabidiol (CBD) and containing low amounts of Δ9-tetrahydrocannabinol (THC), has grown based on the role of the endocannabinoid system in modulating neurotransmission and behavior. This has prompted significant parental interest and “real-world” use for managing ASD-related symptoms. The therapeutic goal is primarily to manage associated behavioral problems such as irritability, aggression, anxiety, and sleep disturbances, with some reports suggesting potential benefits for core symptoms[47]. Supplementary Table 6 shows the various studies concerned with CBM intervention in patients with ASD.

Observational and retrospective evidence

Three of the provided studies were observational or retrospective, reflecting “real-world” use and parent-reported outcomes. These studies, while susceptible to placebo and expectancy bias, consistently reported positive results. Bar-Lev Schleider et al[48] in 2019 analyzed data from 188 ASD patients (2015-2017) treated with cannabis oil (30% CBD and 1.5% THC; a 20:1 ratio). After 6 months, of the 93 patients assessed, 30.1% reported “significant improvement” and 53.7% reported “moderate improvement”[48]. Barchel et al[49] in 2018 reported on the prospective follow-up of 53 children (median age 11) using oral CBD. Parents reported high rates of improvement in comorbid symptoms, including self-injury/rage attacks (67.6% improved), hyperactivity (68.4% improved), and sleep problems (71.4% improved). Anxiety improved in 47.1%[49]. Mazza et al[50] in 2024 conducted a retrospective observational study on 30 children (5-18 years) with moderate to severe ASD using a 33:1 CBD: THC extract. The study reported “significant improvements” in communicative skills, attention, and eye contact, and a reduction in aggression and irritability, as noted by both clinicians and parents.

Only one study in the provided set was a randomized, placebo-controlled trial. Silva et al[51] in 2024 conducted a 12-week, double-blind, parallel-group RCT in 60 children aged 5-11 years, comparing a CBD-rich cannabis extract to a placebo. This trial reported positive results. The treatment group showed statistically significant improvements compared to placebo in social interaction (P = 0.0002), anxiety (P = 0.016), and psychomotor agitation (P = 0.003). Concentration also improved in cases with mild ASD (P = 0.01).

The safety profile was relatively consistent across both the observational studies and the RCT. Bar-Lev Schleider et al[48] in 2019 reported restlessness (6.6%) as the most common side effect. Barchel et al[49] in 2018 reported mild adverse effects, mostly somnolence and a change in appetite. Mazza et al[50] in 2024 noted “minimal untoward effects”. The Silva et al[51] trial in 2024 reported few adverse effects, with only 9.7% of the treatment group (3 children) experiencing adverse events, including dizziness, insomnia, colic, and weight gain. The need for more rigorous RCTs is recognized in the literature. Sannar et al[52] in 2025 published a protocol paper detailing the methods for an upcoming 27-week, randomized, placebo-controlled crossover trial. This future study will evaluate CBD for irritability (ABC-I) and aggression.

Therefore, the evidence for CBMs in pediatric ASD consists of highly positive but low-quality observational data and one positive, higher-quality RCT. The open-label and retrospective studies suggest high rates of parent-reported improvement in behavior, anxiety, and sleep. This is supported by the single available RCT (Silva et al[51]), which found that a CBD-rich extract was superior to a placebo in improving social interaction, anxiety, and agitation. Across all studies, CBMs appear to be well-tolerated, with the most common adverse effects being mild somnolence, changes in appetite, and restlessness. More RCTs are needed to confirm these findings.

Anti-depressants: SSRIs are hypothesized to modulate the serotonergic system to treat repetitive behaviors (obsessive-compulsive-like symptoms) and associated anxiety in ASD[53]. However, SSRIs are generally not recommended for core ASD symptoms due to inconsistent efficacy and the risk of behavioral activation.

Sertraline: Sertraline, an SSRI, showed early promise in open-label trials[54]. However, recent placebo-controlled RCTs primarily involving young children (ages 2-6 years) have yielded negative results on primary outcomes. A pooled meta-analysis of two RCTs (AlOlaby et al[55] in 2017 and AlOlaby et al[56] in 2020) showed no statistically significant difference between sertraline and placebo on the primary outcome (CGI-I responder rate). Similarly, pooled analysis of the change in ABC-I scores showed no significant benefit over placebo. A key finding was a significant gene-by-treatment interaction: Children with the high-activity solute carrier family 6 member 4 long/Long genotype showed significant improvement with sertraline compared to placebo (P = 0.04), while S-carriers experienced worse outcomes[55]. Cytochrome P450 family 2 subfamily C member 19 metabolizer status was also a significant predictor of adverse drug reactions. Higher-dose sertraline was associated with a significantly higher rate of adverse drug reactions compared to placebo, including hyperactivity, insomnia, stereotypy, and decreased appetite[57]. Low doses were better tolerated[56]. Supplementary Table 7 summarizes the various studies concerned with sertraline in ASD.

Fluoxetine: Fluoxetine, a SSRI, has been investigated for the treatment of core ASD symptoms - particularly repetitive behaviors - given its established efficacy in obsessive-compulsive disorder. Evidence derives from early open-label studies, mechanistic neuroimaging research, and three placebo-controlled RCTs. A pooled meta-analysis was performed on the two largest parallel-group trials (Herscu et al[58] in 2020 and Reddihough et al[59] in 2019) (Supplementary Table 8).

Initial non-placebo studies provided early support. Fatemi et al[60] in 1998 reported improvement in lethargy and trends toward benefit in irritability and stereotypy in a small retrospective cohort (n = 7). DeLong et al[61] in 2002, in a large open-label trial (n = 129), observed a 69% favorable response rate, strongly associated with a family history of affective disorders or exceptional intellectual achievement.

In contrast, larger placebo-controlled trials yielded largely negative results. Herscu et al[58] in 2020 found no significant difference between fluoxetine and placebo on the Children’s Yale-Brown Obsessive Compulsive Scale-Pervasive Developmental Disorder version total score or CGI-I responder rates (36% vs 41%) in 158 children with ASD. Reddihough et al[59] in 2019 reported a small but non-robust benefit favoring fluoxetine in the primary adjusted analysis (MD = -2.01, P = 0.03), which lost significance after prespecified additional adjustments (MD = -1.17, P = 0.21), leading the authors to conclude null findings. Pooling change-from-baseline data from both trials (n = 304) yielded a non-significant overall effect (MD = -0.18, 95%CI: -1.33 to 0.97; P = 0.76), indicating no superiority of fluoxetine over placebo for reducing repetitive behaviors.

A smaller 16-week crossover study by Hollander et al[62] in 2005 reported benefit with low-dose liquid fluoxetine (mean 9.9 mg/day) on the Children’s Yale-Brown Obsessive Compulsive Scale compulsion subscale (P < 0.03), though not on overall autism severity. Differences from larger trials may relate to its crossover design, smaller sample, or focus on a subscale rather than the total score. Supporting biological plausibility, Chantiluke et al[63] in 2015 demonstrated via functional magnetic resonance imaging (fMRI) that acute fluoxetine normalized medial prefrontal cortex activation during cognitive flexibility tasks in boys with ASD.

Adverse events across trials were characterized primarily by behavioral activation rather than sedation. Herscu et al[58] in 2020 reported similar rates of activation symptoms in fluoxetine and placebo groups (42% vs 45%), with insomnia and GI symptoms also common. Reddihough et al[59] in 2019 noted high dropout rates, particularly in the fluoxetine group (41% vs 30%), often due to irritability and agitation. Hollander et al[62] in 2005 found no significant differences in treatment-emergent adverse effects between fluoxetine and placebo at low doses.

Other antidepressants: Evidence for other antidepressants is largely unsupportive or based on low-quality studies. An extensive, 12-year placebo-controlled study of Citalopram found it “scarcely effective” and associated with significant adverse effects[64]. Escitalopram is poorly supported, with reported side effects such as hyperactivity and aggression leading to early treatment termination[65]. Fluvoxamine has yielded conflicting results, with early positive reports contradicted by subsequent studies showing no clinical benefits and significant adverse events like agitation and hyperactivity[66-68]. Paroxetine is the least studied SSRI and carries the most significant risk of suicide among SSRIs in this population[69,70].

Regarding SNRIs, mirtazapine showed positive results in an open-label examination for aggression and anxiety, though not core symptoms[71]. Venlafaxine appeared “quite promising” in case reports and a retrospective review[72,73]; however, a randomized, double-blind study found that despite some improvement in children treated with venlafaxine, there were no significant differences between the venlafaxine and placebo groups[74]. Finally, trazodone is primarily supported by evidence for treating sleep disorders in ASD[75], with only case reports suggesting efficacy for aggression[76].

Mood stabilizers: Mood disorders are common in children and adolescents with ASD, and approximately 50% receive medication to manage comorbid behavioral or mood symptoms. Antiepileptic drugs, such as valproic acid, are frequently used as mood stabilizers. This dual use is relevant as seizures co-occur in 10% to 30% of young patients with ASD. Lithium is another classic mood stabilizer considered for children with ASD, particularly those presenting with symptoms of a mood disorder like mania, elevated mood, or paranoia, which may or may not accompany irritability. While atypical antipsychotics are approved for irritability, antiepileptic drugs and lithium represent the core “mood stabilizer” class used for mood lability[77].

Sodium valproate: Valproate (divalproex sodium) is an antiepileptic drug also used as a mood stabilizer. Its investigation into ASD is driven by the high comorbidity of epilepsy (10%-30%) and mood lability in this population, as well as its known efficacy in treating impulsive aggression in other disorders. We reviewed three double-blind, placebo-controlled trials. The 12-week study by Hollander et al[78] in 2010 randomized 27 children and adolescents (mean age 9.46) to receive either divalproex sodium or a placebo. The study found a significant improvement in irritability, with the divalproex group showing notably greater gains on the ABC-I subscale (P = 0.048). Additionally, the responder rate (CGI-I for irritability) was significantly higher in the divalproex group (62.5%) compared to the placebo group (9%) (P = 0.008). Another 8-week pilot RCT by Hollander et al[79] in 2006 (n = 13) focused on repetitive behaviors, revealing that divalproex sodium was significantly more effective than placebo in reducing these symptoms, as measured by the Children’s Yale-Brown Obsessive-Compulsive Scale (P = 0.037), with a large effect size. Furthermore, a large 12-week RCT by Aliyev[80] in 2018 (n = 100) compared valproate sodium to placebo for overall severity, showing a significant improvement in global symptoms. The responder rate on the Clinical Global Impression (CGI) Scale was 80% in the valproate group, compared to 12% in the placebo group. The case report by Carta et al[81] in 2024 supports the use of intravenous (IV) valproate in emergency settings. An 11-year-old boy with ASD and attention-deficit/hyperactivity disorder (ADHD) experiencing a psychotic episode with severe aggression, who was unresponsive to first-line therapies, was successfully and safely treated with IV-valproate. Regarding safety, the included studies report that valproate was generally well-tolerated. Hollander et al[78] in 2010 noted three discontinuations, two in the active group and one in the placebo, due to lack of efficacy or increased irritability. Hollander et al[79] in 2006 did not report specific adverse events but implied good tolerability. Aliyev[80] in 2018 made the strong claim that “Side effects were not observed”. Carta et al[81] in 2024 reported that IV-valproate led to “safe and prompt clinical success” with a lower risk of adverse events than standard high-dose medications.

The evidence for other mood stabilizers in pediatric ASD is limited and conflicting. Lamotrigine has been poorly studied; a single 8-week randomized, double-blind, placebo-controlled study in 28 children found no significant differences between the lamotrigine and placebo groups[82]. Levetiracetam has shown “conflicting results”. While an early open-label study suggested benefit in a small subgroup[83], the only randomized double-blind study (n = 20) found it ineffective for aggression and mood instability. It noted it “seemed instead to increase aggressive behaviors possibly”[84]. Oxcarbazepine lacks any double-blind, placebo-controlled trials. A retrospective study (n = 30) reported improvements in aggression and sleep, but “important adverse effects” occurred in approximately a quarter of patients, including worsening of irritability[85]. In contrast, Topiramate showed positive results in a double-blind, placebo-controlled study (n = 40), which found that adding topiramate to risperidone reduced irritability, stereotyped behaviors, and hyperactivity more than risperidone alone[86]. Lithium has been studied retrospectively for comorbid mood symptoms, with one review (n = 30) reporting improvement in elevated mood, mania, and paranoia, although nearly half the sample experienced side effects[87]. Supplementary Table 9 shows the different studies concerned with using valproate (divalproex) in children with ASD.

Psychostimulants: A high rate of comorbidity exists between ASD and ADHD, with studies indicating that 40%-70% of children with ASD also present with significant ADHD symptoms. Given that psychostimulant drugs are the primary treatment for ADHD, their use has been extensively studied in ASD to manage these overlapping symptoms of hyperactivity and impulsivity. However, tolerability is a major clinical challenge, as patients with ASD often cannot tolerate doses equivalent to those used for ADHD alone and may experience adverse effects such as increased irritability or stereotyped behaviors[88,89].

Methylphenidate: The evidence for methylphenidate (MPH) in ASD is derived from several short-term, placebo-controlled, crossover trials. We reviewed five placebo-controlled, crossover RCTs involving a total of 146 children (aged 5-13 years). The treatment duration in these trials was typically very short, lasting only 1 week per dose. These studies[90-94] showed that MPH provided a statistically significant benefit for ADHD-like symptoms but not for core ASD symptoms. Hyperactivity was the most consistently improved symptom. There was a significant and clinically relevant benefit for hyperactivity as rated by teachers (SMD = -0.78, 95%CI: -1.13 to -0.43; P < 0.001). This confirms findings from the primary trials: Quintana et al[90] in 1995 reported “modest but statistically significant improvement” in hyperactivity, Handen et al[91] in 2000 found 8 of 13 children responded on the Conners hyperactivity index, and Posey et al[92] in 2007 found significant improvement, noting hyperactivity and impulsivity improved most. Regarding inattention, the studies also showed a significant, though not clinically relevant, benefit for teacher-rated inattention (MD = -2.72 on the Swanson, Nolan, and Pelham Rating Scale, 4^th edition; 95%CI: -5.37 to -0.06; P = 0.04). Pearson et al[93] in 2013 also noted significant declines in inattentive behavior, as reported by parents.

On the other hand, the core ASD symptoms, stereotypical behaviors, and social communication did not show significant or conclusive improvement. Handen et al[91] in 2000 found “no changes... on the CARS”, and Posey et al[92] in 2007 found “no significant effects on... stereotyped and repetitive behavior”. However, a secondary analysis by Jahromi et al[94] in 2009 of the RUPP trial data (n = 33) reported that MPH did have a “significant positive effect... on children’s use of joint attention initiations, response to bids for joint attention, self-regulation, and regulated affective state”. Pearson et al[93] in 2013 also noted parent-reported improvements in social skills.

Notably, a primary challenge with MPH in the ASD population is tolerability. Handen et al[91] in 2000 noted that this group seems “particularly susceptible to adverse side effects”, including social withdrawal and irritability, especially at higher doses. However, other trials reported a more benign profile. Quintana et al[90] in 1995 found “no significant side effects including worsening stereotypic movements”, and Pearson et al[93] in 2013 concluded that MPH was “well-tolerated” with side effects “similar to those seen in typically developing children with ADHD”. Based on the included data, the only adverse effect found to be significantly more likely with MPH was reduced appetite (risk ratio = 8.28, 95%CI: 2.57-26.73; P < 0.001). Supplementary Table 10 shows the different studies and RCTs concerned with using MPH in children with ASD.

Atomoxetine: Atomoxetine has been investigated as a non-stimulant pharmacologic option for managing behavioral symptoms in children with ASD, particularly when comorbid ADHD features are present. We reviewed four studies[95-98] and included two of them[95,96] that evaluated atomoxetine monotherapy vs placebo for ADHD symptoms in children with ASD in a pooled meta-analysis. The first study by Harfterkamp et al[95] in 2012 (n = 97) was an 8-week parallel trial that found atomoxetine (1.2 mg/kg/day) resulted in a significant improvement in ADHD symptoms compared to placebo. The primary outcome, as measured by the ADHD Rating Scale score, showed a significant reduction (MD = -6.7, P < 0.001). The second study by Arnold et al[96] in 2006 (n = 16) was a 6-week crossover pilot trial that found atomoxetine to be significantly superior to placebo on the primary outcome, the Aberrant Behavior Checklist (ABC) Hyperactivity subscale (P = 0.043, effect size d = 0.90). Combining the data from these two trials (total n = 113) confirms the efficacy of atomoxetine for hyperactivity. The pooled SMD was -0.68 (95%CI: -1.06 to -0.31). This result is statistically significant (P = 0.0004). The pooled analysis shows a moderate-to-large effect size, indicating that atomoxetine is effective in reducing hyperactivity symptoms in children with ASD.

Regarding the core symptoms of autism, the study by Eslamzadeh et al[97] in 2018 evaluated atomoxetine (0.5-1.2 mg/kg/day) as an add-on to risperidone over 8 weeks (n = 44). This trial found that atomoxetine augmentation provided significant improvement compared to placebo augmentation on the CGI-Severity and CGI-Improvement scales (P ≤ 0.05). It also showed significant improvement in the CARS total score and 7 CARS subscales, including relationship to people, emotional response, nonverbal communication, and activity level (all P ≤ 0.05). This suggests atomoxetine may have benefits for core ASD symptoms, but only when used as an adjunct to risperidone.

Across all trials, atomoxetine was reported as generally well-tolerated, with an adverse effect profile like that seen in children with ADHD only. Harfterkamp et al[95] in 2012 reported adverse events in 81.3% of the atomoxetine group vs 65.3% of the placebo group (not a significant difference). The most common adverse events were nausea, decreased appetite, fatigue, and early-morning awakening. Arnold et al[96] in 2006 reported adverse events as “otherwise tolerable” and noted no tendency to cause stereotypy. However, one subject was rehospitalized for recurrent violence while on atomoxetine. Eslamzadeh et al[97] in 2018 (adjunct trial) reported the most common adverse events as mood change, irritability, and GI disturbance, noting that they “tend to subside”. The open-label study by Zeiner et al[98] in 2011 also found the drug “well tolerated”, with nausea and headache as the most common adverse events.

From our results, we found that both MPH and atomoxetine have moderate, evidence-based efficacy for ADHD symptoms in children with ASD; MPH often yields larger/earlier benefit but may be less well tolerated in a subset, while atomoxetine is a reasonable, generally well-tolerated non-stimulant alternative - choice should be individualized based on prior response, side-effect vulnerability, and clinician/family preference. Table 4 compares the use of both drugs in children with ASD. Table 5 shows the clinical algorithm for selecting MPH vs atomoxetine in children with ASD and ADHD symptoms. Supplementary Table 11 shows the different studies and RCTs concerned with using atomoxetine in children with ASD.

Table 4 Comparison of methylphenidate vs atomoxetine in children with autism spectrum disorder and attention-deficit/hyperactivity disorder symptoms.

Feature	Methylphenidate	Atomoxetine
Evidence base	5 placebo-controlled crossover RCTs (total n = 146)	2 placebo-controlled RCTs included in pooled analysis (total n = 113)
Effect on hyperactivity	Significant improvement (teacher-rated SMD approximately -0.78, P < 0.001)	Significant improvement (pooled SMD approximately -0.68, P = 0.0004)
Effect on inattention	Small but significant reduction (P = 0.04); not clinically large	Significant reduction on ADHD-RS in parallel trial (MD: -6.7, P < 0.001)
Effect on core ASD symptoms	No primary benefit; secondary signals for joint attention/self-regulation	No primary benefit; possible improvement only when combined with risperidone
Tolerability profile	Higher risk of irritability, emotional lability, decrease appetite (RR: 8.28)	Generally well tolerated; nausea, decrease appetite, fatigue most common
Risk of behavioral activation	Higher (especially at higher doses)	Lower
Trial duration	Very short: 1 week per dose	6-8 weeks
Clinical role	Often first-line if tolerated	Alternative when stimulants are not tolerated or contraindicated

This table compares the two most common pharmacological treatments for attention-deficit/hyperactivity disorder symptoms, primarily hyperactivity and inattention, when they co-occur in children with autism spectrum disorder: Methylphenidate, a stimulant, and atomoxetine, a non-stimulant. Both drugs demonstrate significant and comparable efficacy in reducing hyperactivity, which is the primary behavioral target. Methylphenidate, while often chosen as the first-line agent for attention-deficit/hyperactivity disorder in the general population, has been shown to have a higher risk of behavioral activation (increased irritability, emotional lability) in children with autism spectrum disorder compared to neurotypical peers, potentially leading to discontinuation. Atomoxetine, a selective norepinephrine reuptake inhibitor, offers a generally well-tolerated alternative with a lower risk of behavioral side effects, though side effects like nausea and decreased appetite are still common. Crucially, neither drug shows a primary benefit on core autism spectrum disorder symptoms (social or communication deficits). The choice between them balances potent efficacy (methylphenidate) against a better tolerability profile (atomoxetine). Atomoxetine is often preferred when methylphenidate is not tolerated or contraindicated, or as a co-treatment when combined with atypical antipsychotics like risperidone. RCTs: Randomized controlled trials; SMD: Standardized mean difference (a measure of effect size); ADHD-RS: Attention-Deficit/Hyperactivity Disorder Rating Scale; MD: Mean difference; ASD: Autism spectrum disorder; RR: Relative risk (ratio of probability of an event occurring in an exposed group vs unexposed group).

Table 5 Clinical algorithm for selecting methylphenidate vs atomoxetine in children with autism spectrum disorder and attention-deficit/hyperactivity disorder symptoms.

First-line medication choice
MPH is recommended as initial pharmacotherapy in most children with ASD and comorbid ADHD symptoms, provided that no major tolerability concerns are present. MPH demonstrates the largest pooled effect size for hyperactivity reduction (SMD: -0.78) and has rapid onset of action (within days)
ATX is recommended as first-line therapy when:
The child has a history of stimulant-induced irritability, behavioral activation, or emotional dysregulation
Comorbid anxiety, tics, or sleep disturbance is present
Parents prefer a non-stimulant medication
Cardiac risk factors delay or preclude stimulant use
Stepwise treatment approach
Initiate MPH at low dose and titrate gradually based on response and tolerability
Reassess after 2-4 weeks. If inadequate response or intolerable adverse effects occur, switch to ATX
If ATX is started first and response remains suboptimal after 6-8 weeks of optimized dosing, switch to MPH
Combination therapy (MPH + ATX) may be considered only in specialist care after monotherapy failure, with clear target symptoms and close monitoring
Clinical considerations
MPH is associated with a higher risk of appetite suppression and irritability but provides faster and stronger symptom reduction
ATX offers more stable behavioral control, is better tolerated in emotionally reactive children, and may improve global ASD severity when combined with risperidone
Both agents require monitoring of appetite, sleep, heart rate, blood pressure, and behavioral changes at baseline and follow-up visits

MPH: Methylphenidate; ASD: Autism spectrum disorder; ADHD: Attention-deficit/hyperactivity disorder; SMD: Standardized mean difference; ATX: Atomoxetine.

Alpha-2 adrenergic receptor agonists: Alpha-2 adrenergic receptor agonists, such as clonidine, modulate the noradrenergic system. They have been investigated in ASD to manage associated behavioral symptoms. Their use is primarily aimed at reducing hyperarousal behaviors, including hyperactivity, inattention, impulsivity, and sleep disturbances, which are common in children with ASD and often limit the effectiveness of other interventions[99].

Clonidine: The evidence for clonidine’s efficacy is limited to two small, short-term, placebo-controlled crossover trials from 1992, which yielded mixed-to-modest results. A study by Fankhauser et al[100] in 1992 in 9 autistic males found that transdermal clonidine (approximately 0.005 mg/kg/day) was significantly superior to placebo in improving social relationships, affectual responses, sensory responses, and CGI. Conversely, a study by Jaselskis et al[101] in 1992 in 8 hyperactive boys with autism found only a “modest” effect; while parent (Conners) and teacher (ABC) ratings for irritability and hyperactivity improved, clinician ratings (CGI) and other teacher ratings were not significantly different from placebo. This trial evidence is supplemented by open-label data, such as a retrospective study by Ming et al[102] in 2008 on 19 children, which reported that clonidine was effective in reducing sleep initiation latency and night awakenings, with less improvement in hyperactivity, mood instability, and aggressiveness. Across these studies, the most common adverse effects were sedation, drowsiness, and fatigue, which were generally described as tolerable and most prominent during the first two weeks of treatment. Supplementary Table 12 summarizes the included studies concerned with using clonidine in children with ASD.

Guanfacine: Guanfacine, an alpha-2 adrenergic receptor agonist, has been studied as an alternative to psychostimulants for managing hyperactivity in children with developmental disabilities. A large, 8-week, multisite RCT by Scahill et al[103] in 2015 (n = 62) demonstrated that extended-release guanfacine (GEXR) was safe and effective for children with ASD and ADHD symptoms. GEXR treatment resulted in a 43.6% decline on the ABC-Hyperactivity (ABC-H) subscale, compared to a 13.2% decrease in the placebo group (effect size = 1.67). Furthermore, 50% of the guanfacine group were responders (rated “much improved” or “very much improved” on the CGI-I), compared to only 9.4% of the placebo group. A small (n = 11) double-blind, crossover trial by Handen et al[104] in 2008 also found significant benefits on the ABC-H (45% were responders) and global improvement ratings.

Secondary analyses of the Scahill et al[103] in 2015 trial data, reported by Politte et al[105] in 2018, revealed additional benefits. GEXR was found to be effective in reducing oppositional behavior (44% decline vs 12% for placebo; P = 0.004) and, more modestly, repetitive behaviors (24% decline vs < 1% for placebo; P = 0.01) as measured by the Children’s Yale-Brown Obsessive Compulsive Scale-ASD. No significant improvements were seen for anxiety or sleep habits. Across these studies, guanfacine was generally well-tolerated. The most common adverse events were drowsiness, fatigue, and decreased appetite. Supplementary Table 13 describes the different studies concerned with using guanfacine in children with ASD.

N-methyl-D-aspartate-receptor antagonists: Memantine, an N-methyl-D-aspartate glutamate receptor antagonist approved for Alzheimer’s disease, has been investigated in ASD based on theories of excessive glutamatergic activity[106]. Trials have examined its potential to modulate learning and behavior in children with ASD. Early evidence for memantine monotherapy comes from open-label trials, which have produced mixed but promising results. A preliminary 8-week, open-label pilot study by Owley et al[107] in 2006 in 14 children (ages 3-12) with pervasive developmental disorders found significant improvements in memory function and on several subscales of the ABC, including hyperactivity, lethargy, and irritability. However, the study reported no significant improvements in language or nonverbal IQ, and no subject was deemed “much-improved” or “very much improved” on the CGI-I scale. A much larger open-label observational study by Chez et al[108] in 2007 (n = 151) reported more positive clinician-derived outcomes. This study found “significant improvements” in language function, social behavior, and self-stimulatory behaviors (though improvement was less in the latter).

The most substantial evidence for memantine comes from a trial evaluating its use as an add-on to existing antipsychotic treatment. A 10-week, randomized, double-blind, placebo-controlled trial by Ghaleiha et al[109]in 2013 assessed memantine (up to 20 mg/day) as an adjunctive treatment to a stable dose of risperidone (up to 3 mg/day) in children with autism. The study found a significant benefit for the combination therapy. The memantine + risperidone group showed significantly greater reductions in ABC-Community (ABC-C) subscale scores for irritability, stereotypic behavior, and hyperactivity compared to the placebo + risperidone group. Memantine was generally well-tolerated in the studies provided. In the adjunctive RCT, Ghaleiha et al[109]in 2013 found no significant difference in the frequency of side effects between the memantine and placebo groups. In the large open-label trial (Chez et al[108] in 2007), chronic use appeared to have “no serious side effects”. However, agitation led to discontinuation in approximately 10% of patients. Supplementary Table 14 shows the different studies of using memantine in children with ASD.

D-cycloserine: D-cycloserine (DCS) is a partial agonist of the N-methyl-D-aspartate glutamate receptor. Its investigation into ASD stems from its known ability to enhance extinction learning and its potential to improve negative symptoms (like social withdrawal) in other neuropsychiatric disorders. The drug is highly dose-dependent, with low, intermittent doses (e.g., 50 mg) being considered most effective for enhancing learning. The primary hypothesis in ASD is that DCS does not treat symptoms directly but rather potentiates or enhances the effects of behavioral interventions, such as social skills training (SST)[110].

Early pilot data from Posey et al[111] in 2004 in 14 subjects with autistic disorder suggested a potential benefit, showing significant improvement on the ABC Social Withdrawal subscale and the CGI scale. This preliminary positive signal led to a more rigorous, larger trial. The definitive study by Minshawi et al[112] in 2016 was a 10-week, double-blind, placebo-controlled trial in children with ASD (n = 67 randomized) designed to test if 50 mg of DCS administered weekly prior to group SST enhanced outcomes compared to placebo plus SST. This trial was negative in the short term. At the end of the 10-week treatment period (and at the 1-week post-treatment follow-up), there was no statistically significant difference between the DCS and placebo groups on the primary outcome, the Social Responsiveness Scale (SRS) total score (P = 0.45). A Cochrane systematic review (Aye et al[113] in 2021) that analyzed this single trial confirmed these negative short-term findings, reporting “low certainty evidence of little to no difference” for social interaction, communication, repetitive behaviors, or global improvement at the end of active treatment. However, a subsequent long-term, blinded follow-up analysis of the same trial participants by Wink et al[114] in 2017 revealed a “sleeper effect”. At the 22-week follow-up (11 weeks after treatment ended), the DCS group showed significantly greater maintenance of the skills learned during SST than the placebo group. This was demonstrated by a significantly lower (better) SRS total raw score in the DCS group vs the placebo group at 22 weeks (P = 0.042). This suggests DCS did not enhance the acquisition of social skills but may have enhanced the consolidation and durability of those skills.

Across all studies, low-dose (50 mg) intermittent DCS was found to be safe and well-tolerated. Posey et al[111] reported that it was “well tolerated” at most doses used. Minshawi et al[112] in 2016 and Wink et al[114] in 2017 both reported that DCS was well-tolerated. The most frequently reported adverse event was irritability, which was reported with equal frequency in both the DCS and placebo groups. The Aye et al[113] review confirmed no difference in serious adverse events or dropout rates between the groups. Supplementary Table 15 shows the different studies of using DCS in children with ASD.

Gamma-aminobutyric acid type B receptor agonists: Baclofen, a gamma-aminobutyric acid (GABA) type B receptor agonist, and its active enantiomer, arbaclofen (STX209), have been investigated in ASD based on the hypothesis that they may correct an imbalance of excitatory to inhibitory neurotransmission. This rationale is supported by computational drug-repositioning studies, such as Gao et al[115] in 2021, which identified baclofen as a high-potential candidate drug for ASD based on its interaction with network-specific core genes.

The evidence for arbaclofen as a standalone treatment is mixed. An 8-week, open-label trial by Erickson et al[116] in 2014 in 32 children (ages 5-17) provided initial positive signals. This study reported significant improvements on the primary endpoint, the ABC-I subscale, as well as on the Lethargy/Social Withdrawal subscale, the SRS, and the Children’s Yale-Brown Obsessive Compulsive Scale-Pervasive Developmental Disorder version[116]. However, a subsequent large, 8-week, randomized, placebo-controlled, phase 2 trial by Veenstra-VanderWeele et al[117] in 2017 in 150 participants (ages 5-21) did not meet its primary endpoint. The study found no significant difference between arbaclofen and placebo on the ABC Social Withdrawal/Lethargy subscale. Despite the negative primary outcome, the 2017 RCT noted potential signals in secondary analyses, including improvement on the clinician-rated CGI-Severity scale and, in a post-hoc analysis, the Vineland Adaptive Behavior Scales socialization domain.

A separate study evaluated baclofen not as a monotherapy, but as an add-on treatment for children already taking risperidone. A 10-week, randomized, double-blind, placebo-controlled trial by Mahdavinasab et al[118] in 2019 randomized 64 children (ages 3-12) to receive either baclofen or placebo in addition to their existing risperidone treatment. This trial found that the combined treatment (baclofen + risperidone) exerted a significantly greater effect on hyperactive symptoms (as measured by the ABC-H subscale; P < 0.001) than risperidone alone. However, the study found no significant adjunctive benefit for other ABC subscales, such as irritability or lethargy.

The safety profile of baclofen/arbaclofen was generally consistent across studies, with agitation and sedation being the most common adverse events. The open-label Erickson et al[116] in 2014 trial reported agitation and irritability as the most common adverse events, though they often resolved without dose changes. The large Veenstra-VanderWeele et al[117] in 2017 RCT reported affect’s lability (11%) and sedation (9%) as the most common adverse events. The adjunctive Mahdavinasab et al[118] in 2019 trial reported that the combination was safe and efficacious, implying good tolerability. Supplementary Table 16 shows the different studies of using Baclofen in children with ASD.

Glutamatergic modulators: N-acetylcysteine (NAC) is a glutamatergic modulator and an antioxidant precursor under investigation for ASD due to its potential to correct glutamatergic pathway imbalances and address chronic redox imbalance. It is often studied as an adjunctive therapy to reduce irritability, hyperactivity, and repetitive behaviors, and potentially enhance social responsiveness[119]. The most comprehensive evidence comes from a meta-analysis by Lee et al[119] in 2021, which pooled data from five RCTs (4 of which were included in the analysis). This analysis, covering 8-12 weeks of NAC supplementation, found a significant improvement in the ABC total score (MD = 1.31). They also found significant improvements in the ABC-H subscale (MD = 4.80) and the ABC-I subscale (MD = 4.07), as well as in social awareness, as measured by the SRS (MD = 1.34). However, they did not find a significant difference in the pooled results for the Repetitive Behavior Scale.

Other individual RCTs show conflicting results, suggesting that the efficacy of NAC may be context dependent. When used as monotherapy, Hardan et al[120] in 2012 conducted a pilot RCT of NAC monotherapy (n = 33, 12 weeks) and found a significant improvement on the ABC-I subscale (P < 0.001) compared to placebo. Conversely, Wink et al[121] in 2016 (n = 31, 12 weeks) conducted a pilot RCT targeting social impairment. They found no significant difference between NAC and placebo on the primary outcome, the CGI-I anchored to social impairment (P > 0.69). Similarly, the large RCT by Dean et al[122] in 2017 (n = 102, 6 months) found no differences between NAC and placebo on any primary or secondary outcome measures, including the SRS and Repetitive Behavior Scale.

However, when used as adjunctive therapy (n = 33, 12 weeks), the evidence becomes more substantial. Two separate 8-10-week RCTs (Moghimi-Sarani[123] in 2013, n = 40; Nikoo et al[124] in 2015, n = 40) both found that NAC + risperidone was significantly superior to placebo + risperidone in reducing ABC-I scores. Nikoo et al[124] in 2015 also found a significant benefit for the ABC-H/Noncompliance subscale. However, Moghimi-Sarani[123] in 2013 noted that NAC did not change the core symptoms of autism.

In addition, open-label, case series, and retrospective data generally support the positive findings for behavioral symptoms. A retrospective study by Nalbant and Erden[125] in 2023 (n = 37) found that 8 weeks of NAC (400-600 mg/day) reduced ABC scores for irritability, stereotypy, and hyperactivity, and improved CARS scores for social withdrawal and communication. A case series by Çelebi et al[126] in 2017 (n = 8 completers) suggested that NAC (1200-2700 mg/day) as an adjunct to risperidone significantly improved stereotypic behaviors (P = 0.025) and CGI-I scores (P = 0.006). Dean et al[127] in 2019, in a qualitative follow-up to their negative RCT, found that despite the null quantitative result, parents in the NAC group qualitatively reported improved calmness, decreased aggression, and greater improvements in verbal communication.

NAC was consistently reported as safe and well-tolerated across all study designs, including the meta-analysis by Lee et al[119] in 2021. The RCT by Hardan et al[120] in 2012 and the pilot trial by Wink et al[121] in 2016 both reported that NAC was “well tolerated”. Ghanizadeh and Moghimi-Sarani[123] in 2013 reported the most common adverse events as constipation, increased appetite, fatigue, nervousness, and drowsiness, and concluded that NAC was “generally well-tolerated”. The Dean et al[122] in 2017 RCT found no significant difference in the number or severity of adverse events between NAC and placebo. Supplementary Table 17 shows the different studies of using NAC in children with ASD.

Anti-inflammatory medications

Some studies suggest anti-inflammatory medications may help manage specific autism symptoms by reducing inflammation in the brain, potentially improving behaviors like irritability, hyperactivity, and stereotyped actions. Medications explored include corticosteroids (such as prednisolone), anti-inflammatory drugs (such as ibuprofen), and other options, including celecoxib, minocycline, and omega-3 fatty acids. However, these findings are preliminary, and more large-scale research is needed to confirm effectiveness and safety, especially in the long term[128].

Prednisolone: The use of corticosteroids, such as prednisolone, in ASD is predicated on the hypothesis that neuroinflammation or autoimmune processes contribute to the pathophysiology of the disorder, particularly in the subgroup of children with regressive autism (R-ASD) who experience a loss of previously acquired skills[129]. This approach draws parallels to the treatment of Landau-Kleffner syndrome, a condition characterized by regression and epileptic aphasia, which responds to steroid therapy. Prednisolone has shown potential benefits in some studies for certain children with ASD, particularly those with developmental regression, by improving language skills and reducing irritability and hyperactivity[130].

Two RCTs have investigated prednisolone, each focusing on a different primary outcome. A 12-week, single-blind, placebo-controlled RCT by Malek et al[131] in 2020 enrolled 37 children with regressive ASD. Participants received prednisolone (1 mg/kg/day) or placebo as an adjunct to risperidone. The prednisolone group showed significantly greater improvement than the placebo group on the CARS (P < 0.001) and four ABC-C subscales: Irritability (P = 0.026), hyperactivity (P = 0.039), lethargy (P < 0.001), and stereotypy (P = 0.026). Inflammatory biomarkers were also significantly decreased. No significant difference was found for the inappropriate speech subscale. A 24-week, double-blind, placebo-controlled RCT by Brito et al[129] in 2021 enrolled 38 male children (ages 3-7) with ASD to evaluate the effect of prednisolone (initial dose 1 mg/kg/day) on language skills. The study found that prednisolone increased the global language development score specifically in children younger than 5 years who had developmental regression (P = 0.0057). Vocal acts also improved significantly in this subgroup (P = 0.004). However, the benefit was less evident in the overall sample, leading the authors to conclude that the benefit was “more evident” in younger children with regression. Retrospective data support the potential for corticosteroids to improve neurophysiological and clinical markers in regressive autism. Duffy et al[132] in 2014 reviewed 20 steroid-treated children with regressive ASD (STAR group) vs 24 untreated controls. They found that steroid treatment was associated with a significant increase in the frequency modulated auditory evoked response (a measure of language cortex function) and significant improvements in language and behavior scores, which were retained one year after treatment.

While generally considered safe in these short-to-medium term trials, corticosteroid use is associated with known metabolic and immune risks. Brito et al[129] in 2021 reported “mild” side effects, including hyperglycemia (5 patients), hypertension (2 patients), and varicella infection (2 patients), but stated these did not contraindicate use. Malek et al[131] in 2020 detected no significant adverse events during their 12-week trial. A case report by Figueiredo[133] in 2021 highlighted a potential neuropsychiatric adverse effect, describing the recurrence of motor tics in a 7-year-old boy with ASD following acute prednisolone treatment. Supplementary Table 18 shows the different studies of using prednisolone in children with ASD.

Pregnenolone: Pregnenolone is an endogenous neurosteroid that modulates GABA neurotransmission. It has been investigated in ASD based on the hypothesis that it may correct excitatory/inhibitory imbalances. While research is ongoing regarding its potential to treat irritability and sensory issues, clinical data are currently limited to adolescents and young adults, with limited safety data for younger children[134].

The most substantial evidence comes from a study evaluating pregnenolone as an adjunctive therapy to antipsychotics in adolescents. Ayatollahi et al[135] in 2020 conducted a double-blind, placebo-controlled RCT in 64 adolescents (ages 11-17 years) with ASD. Participants received either pregnenolone or a placebo in addition to risperidone. The study found that the pregnenolone group exhibited significantly greater improvement than the placebo group on several subscales of the ABC-C: Irritability (P = 0.025), stereotypy (P = 0.029), and hyperactivity (P = 0.004). However, there were no significant differences between groups for lethargy or inappropriate speech. Fung et al[136] in 2014 conducted a pilot, open-label, 12-week trial in 12 adults (mean age 22.5 years) with ASD. This study reported a statistically significant improvement in ABC-I (P = 0.028), ABC-Lethargy (P = 0.046), and sensory sensitivity (Short Sensory Profile, P = 0.009). While positive, the open-label nature and adult population limit the generalizability to children. A recent methodological report highlights the difficulty in interpreting subjective symptom improvement in pregnenolone trials. McGrath et al[137] in 2025 analyzed a single-blind, 2-week placebo lead-in phase for a trial in adolescents and young adults (ages 14-25). They found that placebo administration alone resulted in a 30.2% decrease in irritability symptoms (P < 0.001). This high placebo response underscores the necessity of the rigorous double-blind design used by Ayatollahi et al[135] in 2020 to confirm actual drug effects.

In the adolescent and adult populations studied, pregnenolone appeared well-tolerated. Ayatollahi et al[135] in 2020 found no significant difference in the frequency or severity of adverse effects between the pregnenolone and placebo groups in adolescents. Fung et al[136] in 2014 reported no severe side effects in adults. Reported adverse events included single episodes of tiredness, diarrhea, and depressive affect, though causality was not definitive. It should be noted that there is currently no reliable information regarding the safety of pregnenolone in younger children or its long-term safety in adolescents. Supplementary Table 19 shows the different studies of using Pregnenolone in children with ASD.

Celecoxib: Celecoxib is a selective cyclooxygenase-2 inhibitor. Its investigation as a treatment for ASD is based on the hypothesis that ASD pathophysiology involves the activation of the inflammatory response system. By inhibiting cyclooxygenase-2 and reducing neuroinflammation, celecoxib may improve behavioral symptoms associated with the disorder[138]. The primary evidence comes from a 10-week, randomized, double-blind, placebo-controlled trial by Asadabadi et al[139] in 2013. This study enrolled 40 children with autism and evaluated the efficacy of celecoxib as an adjunctive therapy to risperidone, where they were randomized to receive either risperidone + celecoxib (up to 300 mg/day) or risperidone + placebo. The study found that the combination of celecoxib and risperidone was significantly superior to risperidone alone on several key measures. By week 10, the celecoxib group showed significantly greater improvement in irritability (P < 0.001), lethargy/social withdrawal (P < 0.001), and stereotypic behavior (P < 0.001). A significantly higher proportion of patients in the celecoxib group achieved a “complete response” compared to the placebo group (55% vs 20%; P = 0.022). However, no significant difference was observed between groups for hyperactivity/noncompliance (P = 0.202) or inappropriate speech (P = 0.802). In this 10-week trial, celecoxib appeared to be safe when added to risperidone. The frequency of side effects was similar between the two groups, suggesting that adding celecoxib did not significantly increase the adverse event burden compared with risperidone monotherapy[139].

Hormonal therapy in children with autism: Hormonal therapy for autism, specifically the use of oxytocin and vasopressin, is a subject of ongoing research with mixed results; studies show potential benefits for social and behavioral issues in some children, but there is no established medical treatment, and it is not widely used[140]. Other hormones, such as growth hormone and gonadotropin-releasing hormone agonists, have been explored for specific conditions associated with autism. However, their use is limited to specific genetic syndromes or challenging behaviors like severe aggression[141]. It is crucial to understand that hormonal therapy is not a cure and its effectiveness varies, with more research needed to determine which individuals might benefit and to understand the long-term effects.

Oxytocin: Oxytocin is a neuropeptide involved in social bonding, trust, and empathy. Research suggests that some individuals with ASD may have lower oxytocin levels or altered oxytocin receptor function. Intranasal oxytocin has been investigated as a treatment to target core social deficits, including social withdrawal, emotion recognition, and repetitive behaviors. The efficacy of chronic intranasal oxytocin treatment (ranging from 4 to 24 weeks) remains mixed[142-145].

Parker et al[142] in 2017 conducted a 4-week, parallel-group RCT (n = 32). They found that oxytocin (24 IU twice daily) significantly enhanced social abilities on the SRS compared to placebo, particularly in children with low pretreatment oxytocin levels. Yatawara et al[143] in 2016 conducted a 5-week crossover RCT (n = 31) and reported significant improvements in caregiver-rated social responsiveness with oxytocin (24 IU/day) compared to placebo. Sikich et al[144] in 2021 conducted the largest trial to date (n = 290) in a 24-week parallel-group RCT. They found no significant difference between oxytocin and placebo on the primary outcome, the ABC-modified Social Withdrawal subscale (P = 0.61), or on secondary social function measures. Dadds et al[145] in 2014 conducted a short-term (4-day) crossover trial (n = 38) and found no significant improvement in emotion recognition or social interaction skills.

Studies examining the acute effects (single dose) suggest that oxytocin modulates neural circuits, though clinical benefits vary. Guastella et al[146] in 2010 found that a single dose of oxytocin improved performance on the “reading the mind in the eyes” task (emotion recognition) in 16 male youths with ASD. Gordon et al[147] in 2013 (fMRI, n = 17) showed that oxytocin increased brain activity in regions associated with social processing (striatum, medial prefrontal cortex) during social judgments. Mayer et al[148] in 2021 (fMRI, n = 25) found that oxytocin increased amygdala responsiveness to physical pain stimuli but did not otherwise substantially modulate empathy-related neural activation. Across all studies, intranasal oxytocin was well tolerated. Sikich et al[144] in 2021 reported that the incidence and severity of adverse events were similar between oxytocin and placebo groups over 24 weeks. Yatawara et al[143] in 2016 noted common adverse events included thirst, urination, and constipation, but overall safety was good. The effects of oxytocin are modest and inconsistent. Children with lower baseline oxytocin levels may respond better to the treatment, but more research is needed. Supplementary Table 20 shows the different studies of using oxytocin in children with ASD.

Vasopressin agonists/antagonists: Arginine vasopressin (AVP) is a neuropeptide critical for regulating social behavior in mammals. Research in children with ASD has identified dysregulation in the vasopressin system. Studies have shown that mean cerebrospinal fluid AVP concentrations are significantly lower in children with autism compared to controls and are associated with greater social symptom severity[149]. Similarly, plasma levels of AVP are lower in mothers of autistic children, though findings in children’s plasma levels have been mixed[150]. Blood AVP concentrations have also been identified as a potential biomarker for theory of mind ability in children with ASD[151]. These findings have led to two opposing therapeutic strategies: (1) Agonism: Administering intranasal vasopressin to correct a potential deficiency; and (2) Antagonism: Blocking the vasopressin 1a (V1a) receptor (e.g., with balovaptan) to modulate social signaling pathways.

Vasopressin agonists (intranasal vasopressin): Evidence from a phase 2 RCT supports the use of intranasal AVP in children. Parker et al[152] in 2019 conducted a 4-week, double-blind, placebo-controlled RCT in 30 children (ages 6-12.9 years). The study found that intranasal AVP treatment significantly enhanced social abilities compared to placebo, as measured by the primary outcome, the SRS, 2^nd Edition (SRS-2) total score (P = 0.0052, effect size d = 1.40). AVP also diminished anxiety symptoms and some repetitive behaviors. An fMRI study by Zink et al[153] in 2011 showed that intranasal vasopressin use by healthy males modulates activity in the left temporoparietal junction, a key node of the theory of mind network, suggesting a neurobiological mechanism for its prosocial effects. In the Parker et al[152] in 2019 trial, intranasal AVP was well tolerated with minimal side effects. There were no dropouts and no differences in adverse event rates compared to placebo. No changes were observed in vital signs, electrocardiography, or clinical chemistry.

Vasopressin antagonists (balovaptan/RG7713): In contrast to the agonist strategy, trials of vasopressin antagonists have mainly yielded negative results in both children and adults, despite initial promise. Hollander et al[154] in 2022 (aV1ation study): A large, 24-week phase 2 RCT in children and adolescents (ages 5-17) found no statistically significant difference between balovaptan (V1a antagonist) and placebo on the primary endpoint, the Vineland-II two-domain composite score of socialization and communication (P = 0.91). A phase 3 RCT in adults by Jacob et al[155] in 2022 (n = 322) was terminated early for futility. The study found no improvement in the Vineland-II two-domain composite score with balovaptan vs placebo. An earlier phase 2 trial in adult men (Bolognani et al[156] in 2019, VANILLA study) had shown “clinically meaningful improvements” on the Vineland-II composite score for specific doses (4 mg or 10 mg), despite missing the primary SRS-2 endpoint. Similarly, a single-dose proof-of-mechanism study (Umbricht et al[157] in 2017) using the antagonist RG7713 showed some improvement in eye-tracking but reduced ability to detect certain emotions. Balovaptan was generally well-tolerated in both pediatric and adult trials, with adverse event rates similar to placebo. No new safety concerns emerged[154,156].

Consequently, the evidence currently favors vasopressin agonism (intranasal AVP) over antagonism for treating social deficits in pediatric ASD. A rigorous RCT demonstrated that intranasal AVP significantly improved social responsiveness and reduced anxiety in children with ASD. Conversely, despite early interest, large-scale trials of the V1a antagonist balovaptan failed to show efficacy in improving social communication in both children and adults. The evidence currently favors vasopressin agonism (intranasal AVP) over antagonism for treating social deficits in pediatric ASD. A rigorous RCT demonstrated that intranasal AVP significantly improved social responsiveness and reduced anxiety in children with ASD. Conversely, despite early interest, large-scale trials of the V1a antagonist balovaptan failed to show efficacy in improving social communication in both children and adults. Supplementary Table 21 shows the different studies of using vasopressin agonists and antagonists in children with ASD.

Immunotherapy: Emerging evidence indicates that a subset of children with ASD may exhibit various forms of immune dysregulation - including chronic inflammation, altered cytokine signaling, and the presence of autoantibodies - which has fueled growing interest in immunomodulatory treatment strategies. Maternal anti-fetal brain autoantibodies have been identified as a potential risk factor for ASD, further supporting an immune-related pathway in a subgroup of affected children[158]. Although immunotherapy is not an established or approved treatment for ASD, several experimental interventions such as IV immunoglobulin (IVIG), low-dose interleukin-2 (Ld IL-2), M2 macrophage-derived secretome, stem cell-based approaches, and other targeted immunomodulators have been evaluated in preliminary studies and early-phase clinical trials[159]. These therapies aim to correct specific immune abnormalities and, in selected cases, have been associated with improvements in behavioral or developmental outcomes. However, the evidence remains inconsistent, study populations are small, and observed benefits appear restricted to biologically defined subgroups[160]. Overall, immunotherapy for ASD remains an exploratory field, and rigorous, large-scale trials are necessary to determine whether any children may meaningfully benefit from immune-targeted interventions.

IVIG: IVIG is being explored to potentially neutralize the brain autoantibodies and modulate immune function, thereby improving behavioral and cognitive symptoms. The evidence for IVIG efficacy comes from a mix of open-label studies, case series, and a few controlled trials. Rossignol and Frye[161] in 2021 conducted a systematic review and meta-analysis of 27 publications. They found that IVIG treatment was significantly associated with improvements in total aberrant behavior, irritability, hyperactivity, and social withdrawal, often with large effect sizes. Boris et al[162] in 2009 reported on 26 autistic children treated with monthly IVIG for 6 months. They observed substantial decreases in hyperactivity, inappropriate speech, irritability, lethargy, and stereotypy. However, regression to pre-IVIG status occurred in 22 of the 26 children within 2-4 months of discontinuation. Connery et al[163] in 2018 presented a case series of 31 children screened for autoimmune encephalopathy. They reported statistically significant improvements on the SRS and the ABC, with immune biomarkers (such as anti-dopamine D2 L antibodies) predicting response. Мальцев ДВ and Свтушенко[164] in 2016 studied 78 children with ASD and folate cycle genetic deficiencies treated with high-dose IVIG. They reported “complete elimination” of the ASD phenotype in 21 patients and “marked improvement” in 33 others, along with improvements in pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections, epilepsy, and GI syndromes.

Niederhofer et al[165] in 2003 conducted a double-blind, placebo-controlled crossover study in 12 children. They found statistically significant improvements on the ABC subscales (irritability, hyperactivity, inadequate eye contact, inappropriate speech) for the immunoglobulin group compared to placebo (P < 0.05). However, clinician ratings showed no significant differences. Plioplys[166] in 1998 conducted an open-label trial with 10 children. Only one child (10%) showed “very significant improvement”, while four showed mild improvement and five had no response. The authors cautioned that the mild improvements could be due to placebo effects and recommended caution given the high costs and low response rates.

Melamed et al[167] in 2018 found significant improvements in communication and social interaction in a pilot study of 14 children selected for specific immune dysfunction markers. Jyonouchi et al[168] in 2011 found that ASD children with ASD requiring IVIG for specific antibody deficiencies had distinct cytokine profiles (lower pro-inflammatory cytokines), suggesting a specific neuroimmune subgroup. Adverse events were generally reported as limited but not uncommon. Rossignol and Frye[161] in 2021 listed headaches, vomiting, worsening behaviors, anxiety, fever, nausea, fatigue, and rash as adverse events. Connery et al[163] in 2018 noted that adverse effects were common (62%) but mainly limited to the infusion period, with only 6% discontinuing treatment. Niederhofer et al[165] in 2003 reported significantly higher rates of drowsiness and decreased activity in the immunoglobulin group compared to placebo. Supplementary Table 22 summarizes studies on the use of IVIG in children with ASD.

Emerging and other Immunotherapies

Beyond IVIG and corticosteroids, researchers are exploring targeted immunotherapies to address specific immune dysregulations observed in ASD, such as imbalances in T-cell subpopulations (e.g., Th1/Treg ratio) and neuroinflammation. These therapies are currently experimental and mainly in the early phases of investigation (Table 6).

Table 6 Emerging immunotherapies in autism spectrum disorder.

Intervention	Ref.	Population	Key findings	Status/safety
Low-dose IL-2	Case reports/series (Chen et al[159] in 2025, Li et al[253] in 2024)	Small N (children)	Positive: Improvements in speech, social interaction, sleep. Mechanism: Corrected Th1/Treg immune imbalance	Experimental. No adverse events reported in these cases
M2 macrophage secretome	Clinical trial (Shevela et al[171], 2024)	n = 71 (3-13 years)	Positive: Reduced language impairment and autistic-like behavior via intranasal delivery	Experimental. Reported as safe and well-tolerated
GcMAF	In vitro study (Siniscalco et al[172], 2014)	Macrophage cells	Mechanistic: Normalized endocannabinoid gene expression in cells	Unproven/risky. Unlicensed; no clinical evidence provided
Allergy immunotherapy	Case report (Sood et al[174], 2016)	n = 1 (8 years)	Feasibility: Behavioral strategies allowed successful treatment of comorbid allergies	Standard therapy for atopy (not ASD core symptoms)

Table 6 presents highly experimental and emerging immunotherapies for autism spectrum disorder, based on the hypothesis that a subset of autism spectrum disorder is driven by immune dysregulation or neuroinflammation. The current evidence base is extremely low certainty, relying primarily on case reports/series and preclinical studies rather than large-scale randomized controlled trials. Low-dose interleukin-2 and the M2 macrophage secretome show POSITIVE initial signals in small human trials, demonstrating improvements in symptoms like speech, social interaction, and language impairment. The mechanism is hypothesized to involve correcting the imbalance between Th1 and Treg cells. These interventions are highly experimental but have been reported as safe in the limited cases studied. Gc protein-derived macrophage activating factor is listed primarily for its mechanistic interest (in vitro), but it is generally considered unproven and risky due to a lack of clinical evidence and issues surrounding its licensing and production. Allergy immunotherapy is relevant for treating comorbid atopy (allergies), not core autism spectrum disorder symptoms, though managing comorbidities can indirectly improve behavior. Overall, these interventions target specific biological pathways (e.g., T-cell differentiation, macrophage function) that are hypothesized to be dysfunctional in autism spectrum disorder, but they remain far from clinical recommendation and require extensive, rigorous double-blind randomized controlled trial validation. IL-2: Interleukin-2 (a type of cytokine); M2 macrophage: A type of macrophage (immune cell) associated with tissue repair and anti-inflammation; GcMAF: Gc protein-derived macrophage activating factor; Th1: T helper type 1 cell (an immune cell); Treg: Regulatory T cell (an immune cell that suppresses immune responses); In vitro: Experiment performed in a test tube or culture dish (outside a living organism); ASD: Autism spectrum disorder.

Ld IL-2: Individuals with ASD often exhibit immune dysfunction, specifically a deficiency in regulatory T cells and an elevated Th1/Treg ratio. Low-dose IL-2 (Ld IL-2) is hypothesized to stimulate Treg expansion, thereby restoring immune balance and reducing neuroinflammation. Chen et al[159] in 2025 presented case reports of four children (average age 6) with ASD and immune abnormalities treated with Ld IL-2. Following treatment, assessments (CARS, ABC, ATEC) revealed substantial behavioral improvements, particularly in speech/Language and sleep quality. Immune testing showed a marked decrease in the proportion of cytotoxic T cells and a correction of the cytotoxic T cells/Treg ratio. Li et al[169] reported on an 8-year-old child treated with subcutaneous Ld IL-2. After six courses, the child showed improvements in speech, social initiative, and adaptability. These behavioral gains persisted for three months post-treatment and coincided with a reduced Th1/Treg ratio. A study in BTBR mice (an autism model) by Li et al[170] in 2025 supported these clinical observations, finding that Ld IL-2 increased Treg levels, reduced central nervous system inflammation, and ameliorated core autistic behaviors. The behavioral improvements were diminished upon Treg cell depletion, confirming the mechanism of action.

M2 macrophage secretome: M2 macrophages release immunoregulatory and neurotrophic factors that may combat neuroinflammation and aid neural regeneration. Shevela et al[171] in 2024 conducted a clinical trial (n = 71) in children (ages 3-13) with severe language impairment and autistic-like behavior. Daily intranasal inhalations of M2 macrophage-conditioned medium for 30 days were well tolerated. Two-thirds of the children showed clear clinical improvement in language impairments, autistic-like behavior, and ADHD symptoms. The effects appeared within a month and persisted for 6 months during follow-up.

Gc protein-derived macrophage activating factor: Gc protein-derived macrophage activating factor (GcMAF) is an unlicensed product. Regulatory bodies note there is no scientific evidence to support its claims for autism, and it poses health risks. Evidence is limited to in vitro studies. Siniscalco et al[172] in 2014 investigated the effects of GcMAF on blood monocyte-derived macrophages from autistic patients. The study found that GcMAF treatment normalized dysregulated endocannabinoid gene expression and downregulated macrophage overactivation in cell culture. However, this study did not assess clinical efficacy or safety in humans.

Allergy immunotherapy: Comorbidity between ASD and atopy (allergies/asthma) is common, but children with ASD may struggle with the medical procedures required for treatment[173]. Sood et al[174] presented a case report of an 8-year-old boy with ASD demonstrating that subcutaneous allergy immunotherapy can be successfully administered using behavioral modification strategies (visual schedules, positive reinforcement). This intervention treated the comorbid allergic condition rather than core ASD symptoms.

Metabolic interventions

Methylcobalamin and folinic acid (methylation support): Research indicates that children with ASD often exhibit metabolic abnormalities, specifically in the folate-dependent one-carbon metabolism and transsulfuration pathways. Studies have consistently found lower brain levels of methylcobalamin (active B12) in autistic subjects across the lifespan, as well as plasma deficits in methionine, S-adenosylmethionine, and reduced glutathione, alongside elevated homocysteine and oxidized glutathione[175]. These findings suggest impaired methylation capacity and chronic oxidative stress, providing a rationale for supplementation with methylcobalamin and folinic acid.

For efficacy of methylcobalamin (methyl B12) as a monotherapy, Hendren et al[176] in 2016 conducted an 8-week, randomized, double-blind, placebo-controlled trial (n = 57) of subcutaneous methyl B12 (75 μg/kg every 3 days). The primary outcome, the clinician-rated CGI-I score, was significantly better in the methyl B12 group (2.4) than in the placebo group (3.1) (P = 0.005). This clinical improvement correlated with increased plasma methionine and cellular methylation capacity. However, no significant improvements were observed in parent-rated measures such as the ABC or the SRS. Meanwhile, Bertoglio et al[177] in 2010 conducted a 12-week crossover RCT (n = 30) using a similar subcutaneous dose (64.5 μg/kg). This study found no statistically significant difference between methyl B12 and placebo on overall behavioral measures or glutathione status. However, a subgroup of responders (30%) showed clinically significant improvement and increased glutathione levels.

However, open-label and case reports reported a positive response to high-dose methylcarbamide monotherapy. Čorejová et al[178] in 2022 studied high-dose oral methylcobalamin syrup (500 μg daily) in 25 patients for 200 days. They reported gradual improvements in social, cognitive, and behavioral domains, which correlated with improved glutathione redox status. In addition, a case report by Čorejová et al[179] in 2015 described the cessation of nocturnal enuresis in an 18-year-old male with autism following methylcobalamin treatment, which returned when treatment stopped. While generally considered safe, Geier and Geier[180] in 2010 raised a concern about the potential toxicity of cobalt. Their cohort study found that subcutaneous methylcobalamin injections significantly increased plasma and urinary cobalt levels, often exceeding occupational exposure limits.

High-dose folinic acid monotherapy: Folinic acid (leucovorin) can increase the cerebral level of folate by using a different mechanism that avoids folate receptor-α, as shown in Figure 3. Fewer RCT studies examine the efficacy of high-dose folinic acid monotherapy in the management of children with ASD. Frye et al[181] in 2018 conducted a 12-week, double-blind, placebo-controlled RCT (n = 48) of high-dose folinic acid (2 mg/kg/day, max 50 mg) in children with ASD and language impairment. They found that folinic acid treatment resulted in a significant improvement in verbal communication (as measured by the Clinical Evaluation of Language Fundamentals-4^th Edition) compared with placebo. The response was strongest in children who were positive for folate receptor-α autoantibodies (FRAA), suggesting a targeted mechanism for cerebral folate deficiency.

Figure 3 Folate transport and metabolism pathway targeted by folinic acid in autism spectrum disorder. This figure illustrates the biological mechanism underlying cerebral folate deficiency in a subset of children with autism spectrum disorder. It provides the mechanistic rationale for targeted treatment with high-dose folinic acid (leucovorin). Under normal conditions, active folate (5-methyltetrahydrofolate) is transported across the blood-brain barrier by the folate receptor alpha. In children with cerebral folate deficiency, circulating folate receptor alpha autoantibodies bind and block the folate receptor alpha receptor, impairing folate transport into the central nervous system despite normal peripheral serum folate levels. This blockade reduces cerebral folate availability and disrupts downstream metabolism in the folate cycle. Folinic acid, a reduced folate derivative, bypasses the folate receptor alpha-dependent transport mechanism because it can enter the brain via alternative, non-folate receptor alpha transporters. Once inside the central nervous system, folinic acid is converted to 5-methyltetrahydrofolate, restoring folate cycle function and improving neurological and behavioral outcomes in biomarker-positive children. This figure highlights the translational precision-medicine link, showing that folate receptor alpha autoantibodies are predictive biomarkers of clinical responsiveness to folinic acid therapy. 5-MTHF: 5-methyltetrahydrofolate.

Several studies analyzed data from an open-label trial of combined therapy (75 μg/kg methyl B12 injections + 400 μg oral folinic acid twice daily). James et al[182] in 2009 (n = 40) demonstrated that 3 months of combined treatment significantly increased cysteine and glutathione concentrations and normalized the glutathione redox ratio (P < 0.008). Frye et al[183] in 2013 (n = 37 from the same cohort) found significant improvements in Vineland Adaptive Behavior Scale scores (average improvement of 7.7 months) over the treatment period. Greater metabolic improvement (glutathione redox status) correlated with greater gains in expressive communication and daily living skills. Delhey et al[184] in 2018 and Vargason et al[185] in 2018 re-analyzed this data, confirming that the combined B12/folinic acid treatment significantly shifted metabolic profiles toward those of neurotypical controls, more effectively than high-dose folinic acid alone for metabolic markers. Supplementary Table 23 summarizes the studies concerned with using methylcobalamin and folinic acid in children with ASD.

L-carnitine (mitochondrial support): L-carnitine is an essential nutrient that transports long-chain fatty acids into the mitochondria for energy production (β-oxidation). Research indicates that children with ASD may have lower blood levels of acyl-carnitines, suggesting potential mitochondrial dysfunction and abnormal fatty acid metabolism[186]. Furthermore, defects in the carnitine biosynthetic pathway (e.g., mutations in the TMLHE gene) have been linked to an increased risk of autism. Supplementation with L-carnitine or its derivative, acetyl-L-carnitine, is investigated to correct these metabolic deficits and improve behavioral symptoms.

Evidence from placebo-controlled trials supports the efficacy of L-carnitine, particularly as an adjunctive therapy. Shakibaei Jelvani[187] in 2023 conducted an 8-week RCT (n = 50) and found that adding L-carnitine to risperidone significantly reduced the ABC total score and the lethargy/social isolation subscale compared to placebo. Nasiri et al[188] in 2024 conducted a 10-week RCT (n = 68) of L-carnitine (150 mg/day) plus risperidone. They reported significantly greater reductions in irritability (P = 0.033) and hyperactivity (P < 0.001) in the L-carnitine group compared to the placebo group. Although distinct from carnitine, a study on L-carnosine (800 mg/day) by Hajizadeh-Zaker et al[189] in 2018 (n = 70) found significant improvement in hyperactivity/noncompliance (P = 0.044) but not irritability when added to risperidone.

Goin-Kochel et al[190] in 2019 conducted an open-label pilot study of high-dose carnitine (up to 400 mg/kg/day) in 10 males with ASD. They observed significant improvements over time in hyperactivity (ABC), social communication, and social approach behaviors. However, after statistical correction for multiple comparisons, these changes were not significant, likely due to the small sample size. Two studies in valproate-induced autism rat models (Zahedi et al[191] in 2023; Zahedi et al[192] in 2024) demonstrated that acetyl-L-carnitine alleviated autism-like behaviors (social interaction, repetitive behaviors). Mechanisms included reducing oxidative stress, improving mitochondrial function (biogenesis, membrane potential), and reducing neuroinflammation by modulating gut microbiota and short-chain fatty acids.

L-carnitine was generally well-tolerated, but high doses are associated with specific side effects. Goin-Kochel et al[190] in 2019 reported that heavy body odor (“fishy” smell) and diarrhea were the most common side effects at high doses (up to 400 mg/kg/day), leading some participants to discontinue or fail to reach target doses. The adjunctive trials using lower doses (e.g., 150 mg/day) reported good safety profiles. Nasiri et al[188] in 2024 found no significant difference in adverse effects compared to placebo. Supplementary Table 24 summarizes the studies concerned with using L-carnitine in children with ASD.

Coenzyme Q10: Coenzyme Q10 (CoQ10) and its reduced active form, ubiquinol, are lipid-soluble antioxidants and critical components of the mitochondrial electron transport chain. Its use in ASD is based on the hypothesis that oxidative stress and mitochondrial dysfunction are key pathophysiological mechanisms in the disorder[193]. A large meta-analysis of 87 studies confirmed that children with ASD exhibit significant oxidative stress marker aberrations, including lower levels of antioxidants like glutathione and higher levels of oxidative markers, supporting the rationale for antioxidant therapy[194]. Furthermore, genome-wide association studies have linked serum CoQ10 levels to genetic loci associated with neuronal diseases, including autism[195].

Evidence for CoQ10/ubiquinol efficacy comes primarily from open-label and retrospective studies, which report generally positive findings. Mousavinejad et al[196] in 2018 studied 90 children with ASD treated with CoQ10 (30-60 mg/day) for 100 days. They found that supplementation improved oxidative stress markers (reducing malondialdehyde) and was associated with improvements in sleep disorders (P = 0.005) and GI problems (P = 0.004). Gvozdjáková et al[197] in 2014 reported on 24 children treated with ubiquinol (100 mg/day) for 3 months. They noted improvements in communication (12%-21%), sleep (34%), and food rejection (17%) in children whose plasma CoQ10 levels reached a therapeutic threshold (> 2.5 μmol/L). However, an “expression of concern” was published regarding the Gvozdjáková et al[197] in 2014 study due to a lack of ethical approval, registration, and conflict-of-interest disclosure (funding from a ubiquinol manufacturer), thereby limiting the reliability of these findings[198].

In addition, Legido et al[199] in 2018 conducted an open-label pilot trial (n = 11) of a “MitoCocktail” containing CoQ10, carnitine, and alpha-lipoic acid. They reported significant improvements in mitochondrial function (respiratory chain complex I/IV ratio) and behavioral scores on the ABC-Lethargy (P < 0.01) and ABC-Inappropriate Speech (P < 0.02) subscales. These gains waned 3 months after treatment cessation. Cucinotta et al[200] in 2022 performed a retrospective chart review of 59 patients with neurodevelopmental disorders (including 36 with ASD) treated with ubiquinol, vitamin E, and B-complex. A positive clinical outcome (CGI-I ≤ 3) was recorded in 76.27% of patients, with improvements in cognition (44%), adaptive functioning (44%), and social motivation (32%).

CoQ10 and ubiquinol appear to be safe and well-tolerated in the pediatric ASD population. Cucinotta et al[200] in 2022 reported mild side effects in 30.5% of patients, with the most frequent being increased hyperactivity (15.3%) - no significant adverse events led to treatment discontinuation. Mousavinejad et al[196] in 2018 and Legido et al[199] in 2018 did not report any serious adverse events. Supplementary Table 25 summarizes studies on the use of CoQ10 (ubiquinol) in children with ASD.

Sulforaphane: Sulforaphane (SFN), an isothiocyanate derived from broccoli sprouts, targets multiple metabolic and immunologic pathways relevant to ASD. It is known to upregulate antioxidant capacity (via nuclear factor erythroid 2-related factor 2 activation), improve glutathione synthesis, reduce mitochondrial dysfunction, and lower neuroinflammation. Given these broad physiological effects, SFN has been investigated as a treatment for both core and associated symptoms of ASD[201].

Evidence from placebo-controlled trials is generally positive, though the magnitude of effect varies by population and outcome measure. Singh et al[202] in 2014 conducted the first major RCT in 44 young men (ages 13-27) with moderate-to-severe ASD. SFN treatment (50-150 μmol daily) for 18 weeks resulted in substantial improvements compared to placebo. The SFN group showed a 34% decline in ABC scores (P < 0.001) and a 17% decline in SRS scores (P = 0.017). CGI-I was also significantly greater (P < 0.015). Notably, symptoms returned toward pretreatment levels after discontinuation[202]. Ou et al[203] in 2024 conducted a larger multicenter RCT in 108 children (ages 4-12) in China. Results were mixed: Caregiver ratings showed no significant change, but clinician-rated scales showed significant improvement in the SFN group. One-third of participants had at least a 30% decrease in scores. Effects were greater in children older than 10 years. Zimmerman et al[204] in 2021 studied 57 children (ages 3-12) over 36 weeks (15 weeks double-blind). While the primary outcome was not statistically significant, secondary caregiver ratings on the ABC improved significantly at 15 weeks (Cohen’s d = -0.96). Clinical improvements correlated with positive changes in urinary biomarkers (glutathione, mitochondrial respiration, inflammation). Magner et al[205] in 2023 conducted a smaller RCT (n = 40) in younger children (3-7 years). While ABC and SRS-2 scores improved in both groups, the difference was not statistically significant, leading the authors to find “no significant clinical improvement” in this specific age cohort.

Momtazmanesh et al[206] evaluated SFN as an adjunct to risperidone in 60 children (ages 4-12). The combination (risperidone + SFN) was significantly superior to risperidone + placebo in reducing irritability (P = 0.001) and hyperactivity/noncompliance (P = 0.015). No additional benefit was found for lethargy, stereotypy, or inappropriate speech. Lynch et al[207] in 2017 followed up with participants from the original Singh et al[202] in 2014. They reported a case series of 16 families, many of whom continued SFN supplements. Caregivers articulated sustained positive effects over the ensuing 3 years, supporting the long-term utility of the intervention.

SFN was consistently reported as safe and well-tolerated across all studies. Ou et al[203] in 2024 and Momtazmanesh et al[206] in 2020 found no significant difference in adverse events between SFN and placebo groups. Zimmerman et al[204] in 2021 reported rare side effects, including insomnia, irritability, and intolerance to taste/smell (likely due to the broccoli sprout origin), but no serious adverse events. Supplementary Table 26 summarizes the studies concerned with using SFN in children with ASD.

Diuretics

The use of diuretics - particularly bumetanide, a loop diuretic - has gained research interest as a potential neurobiological intervention for ASD.

Bumetanide: Bumetanide, a loop diuretic, is investigated in ASD based on the “excitatory/inhibitory imbalance” hypothesis. Unlike traditional diuretic use for fluid or blood pressure regulation, the therapeutic rationale in ASD is based on restoring inhibitory neurotransmission, specifically by modulating the brain’s GABAergic system. In immature neurons and potentially in ASD, intracellular chloride levels are elevated due to high expression of the (Na⁺-K⁺-2Cl^- cotransporter 1) NKCC1 chloride importer. This causes GABA to act paradoxically as an excitatory rather than inhibitory neurotransmitter[208]. Bumetanide inhibits the NKCC1 chloride importer, reducing neuronal chloride and potentially shifting GABA back toward its typical inhibitory function, thereby improving neuronal network stability[209].

The clinical evidence for bumetanide is characterized by a distinct dichotomy: Robust positive findings in phase 2 and smaller RCTs, contrasted by a lack of efficacy in large, multi-center phase 3 trials. Numerous independent trials have reported significant improvements in core ASD symptoms. Lemonnier et al[210] in 2012 and Lemonnier et al[211] in 2017) consistently found significant reductions in CARS (P < 0.004), SRS, and CGI scores with 3 months of treatment. Dai et al[212] in 2021 (n = 120) and Zhang et al[213] in 2020 (n = 83) both demonstrated significant reductions in symptom severity (CARS) and CGI scores compared to controls. Shaker et al[214] in 2024 (n = 80) reported statistically significant decreases in CARS scores after 6 months (P < 0.001). Hajri et al[215] in 2019 and Fernell et al[216] in 2021 provided open-label and waitlist-control evidence supporting improvements in communication and cognitive abilities.

On the other hand, Fuentes et al[217] in 2023 reported on two large, international phase 3 trials (SIGN 1 and 2, total n = 422). Both were terminated early for futility, as there was no significant difference between bumetanide and placebo on the primary endpoint in either children (7-17 years) or young children (2-6 years). Sprengers et al[218] in 2021 (BAMBI trial, n = 92) found no superior effect on the primary outcome (SRS-2), though secondary analysis showed benefit for repetitive behaviors. However, evidence suggests bumetanide may enhance the effects of behavioral interventions. Du et al[219] in 2015 conducted a pilot RCT (n = 60) comparing applied behavior analysis (ABA) alone vs ABA + bumetanide. The combined group showed significantly lower ABC and CGI scores (P < 0.05) than the ABA-only group, suggesting a synergistic effect.

Recent studies using neuroimaging and electrophysiology provide strong biological validation for bumetanide’s proposed mechanism, even where clinical scales vary. Dai et al[212] in 2021 and Zhang et al[213] in 2020 used magnetic resonance spectroscopy (MRS) to confirm that bumetanide treatment restores excitatory/inhibitory balance by significantly decreasing the GABA/glutamate ratio in the insular and visual cortices. Importantly, this physiological change correlated with symptom improvement, directly linking the mechanism to the clinical outcome. Hadjikhani et al[220] in 2018 used fMRI to show that bumetanide normalized amygdala activation during eye contact. In addition, Mollajani et al[221] in 2025 found that bumetanide improved facial emotion recognition accuracy. This was accompanied by increased amplitude and decreased latency of N170 and N250 event-related potentials, indicating faster and more robust neural processing of social stimuli. To predict responders to bumetanide, Juarez-Martinez et al[222] in 2023 used machine learning on electroencephalography (EEG) data to predict which children would respond to bumetanide, achieving up to 92% accuracy, suggesting that efficacy may be limited to a specific biological subgroup. Bumetanide is generally safe but carries specific mechanism-based side effects. The most consistently reported adverse events are hypokalemia (low potassium), polyuria (increased urination), thirst, and dehydration. Hypokalemia occurred in up to 51% of participants in some trials but was effectively managed with oral potassium supplements and monitoring (Hajri et al[215] in 2019 and Sprengers et al[218] in 2021).

However, the use of bumetanide in ASD remains experimental. The magnitude of benefit varies across studies, and long-term efficacy has not been firmly established. Safety concerns primarily include hypokalemia, dehydration, mild metabolic alkalosis, and increased urination, necessitating monitoring of electrolytes and hydration status. Given the small sample sizes and ongoing debate regarding clinical significance, bumetanide is not recommended as a routine therapy. However, it represents a promising avenue for targeted treatment - particularly in biologically defined subgroups of children with ASD who show evidence of chloride channel dysfunction or excitatory-inhibitory imbalance. Supplementary Table 27 summarizes the studies concerned with using bumetanide in children with ASD.

Spironolactone: Spironolactone, a mineralocorticoid receptor antagonist and potassium-sparing diuretic, is being explored for the treatment of ASD due to its pleiotropic effects. Beyond its diuretic action, it possesses potent anti-inflammatory and anti-androgenic properties[223]. The rationale for its use is supported by findings that subsets of children with ASD exhibit neuroinflammation and elevated levels of adrenal androgens (e.g., dehydroepiandrosterone, androsterone), which are known to modulate neurotransmission. Additionally, recent research suggests spironolactone may act as an antagonist to the epidermal growth factor receptor 4 (ErbB4) receptor, a protein implicated in neurodevelopment and often dysregulated in ASD[224].

Currently, the primary efficacy data comes from rodent models of autism, which show promising behavioral and biochemical improvements. Zarate-Lopez et al[225] in 2025 investigated spironolactone in a mouse model of autism induced by valproate. They found that prenatal valproate exposure led to social deficits, repetitive behaviors, and dysregulation of the ErbB4/mammalian target of rapamycin signaling pathway in the striatum and prefrontal cortex. Treatment with spironolactone (50 mg/kg) antagonized ErbB4 and reduced repetitive behaviors, consistent with the downregulation of this aberrant signaling pathway. Mirza et al[226] in 2023 evaluated spironolactone in a rat model of ASD induced by propionic acid. Post-natal administration of spironolactone (25 mg/kg and 50 mg/kg) ameliorated impairments in social behavior, anxiety, and repetitive behavior. Biochemically, it restored levels of BDNF and synapsin II in the brain, which were reduced by propionic acid exposure.

Direct clinical evidence in humans remains sparse and preliminary. Bradstreet et al[227] in 2007 presented a medical hypothesis paper reviewing the rationale for spironolactone use. They briefly described a single case report demonstrating objective clinical improvements in an autistic child following spironolactone administration. They proposed that its dual anti-inflammatory and anti-androgen profile makes it a “desirable intervention”, particularly for the hyperandrogenic subgroup of children. Majewska et al[228] in 2014 provided biomarker support for the anti-androgen rationale. Their study of salivary steroids in prepubertal children found significantly higher levels of androgens (androstenediol, dehydroepiandrosterone, androsterone) in children with ASD compared to controls, suggesting that modulating these hormones (as spironolactone does) could be a valid therapeutic target.

While spironolactone has a well-established safety profile for other indications, specific safety data for its use in pediatric ASD is limited. Bradstreet et al[227] in 2007 noted that spironolactone generally has a “desirable safety profile” but emphasized the need for controlled trials to define risks in this specific population. Known risks in other populations include hyperkalemia and endocrine effects (e.g., gynecomastia) due to its anti-androgenic activity. Supplementary Table 28 summarizes the studies concerned with using spironolactone in children with ASD.

Microbiota-gut-brain axis interventions

The “microbiota-gut-brain axis” is a bidirectional communication pathway implicated in the pathophysiology of ASD. Metagenomic studies consistently reveal gut dysbiosis in children with ASD, characterized by reduced alpha diversity and specific taxonomic shifts, such as lower levels of Bifidobacterium, Prevotella, and Veillonellaceae, and higher levels of Clostridium, Bacteroides, and Lactobacillus compared to neurotypical peers[229]. These microbial alterations correlate with the severity of both GI and behavioral symptoms, providing a rationale for modulation via probiotics (beneficial bacteria), prebiotics (dietary fibers), or synbiotics (combinations of both). Table 7 summarizes the various biotic interventions that modulate the gut-brain axis.

Table 7 The different biotic interventions to modulate the gut-brain axis.

Intervention type	Definition/source	Key mechanisms on gut-brain axis	Examples	Observed effects (general and specific)
Probiotics	Live microorganisms (e.g., bacteria or yeast) that, when administered in adequate amounts, confer a health benefit on the host	Directly modulate the gut microbial composition. They produce short-chain fatty acids like butyrate, which can cross the blood-brain barrier. They also modulate the production of neurotransmitters (e.g., gamma-aminobutyric acid, serotonin) and reduce systemic inflammation	Bacteria: Lactobacillus species (Lactobacillus rhamnosus, Lactobacillus acidophilus), Bifidobacterium species (Bifidobacterium longum, Bifidobacterium infantis)	General: Improved mood, reduced anxiety, improved gut health (e.g., reduced IBS symptoms). Specific: Probiotics with Fructo-oligosaccharide in children with autism spectrum disorder were found to improve autism-related symptoms, increase beneficial bacteria, and reduce the hyper-serotonergic state and dopamine metabolism disorder
Prebiotics	Selectively fermented ingredients (often non-digestible fibers) that nourish and result in specific changes in the composition and/or activity of the beneficial gastrointestinal microbiota	Indirectly influences the brain by serving as food for beneficial gut bacteria, thereby increasing short-chain fatty acid production (especially butyrate). They enhance gut barrier function and indirectly influence the nervous system by reducing inflammation and modulating short-chain fatty acid signaling	Fibers: Fructans (inulin, Fructo-oligosaccharides), Galacto-oligosaccharides	General: Selective growth of beneficial bacteria (Bifidobacteria, Lactobacillus), enhanced mineral absorption, reduced constipation. Specific: Fructo-oligosaccharide used in combination with probiotics in children with autism spectrum disorder was associated with improved symptoms and beneficial gut change
Synbiotics	A product containing both Probiotics (live microorganisms) and Prebiotics (their selectively utilized food substrate) in a combined form	Combines the direct seeding and modulating effects of probiotics with the selective nourishment and short-chain fatty acid-boosting effects of prebiotics, potentially offering a synergistic effect on the gut-brain axis	Commercial blends containing strains such as Bifidobacterium and a fiber source such as Fructo-oligosaccharides or inulin	Effects are similar to or enhanced versions of those observed with single-agent interventions, often formulated to maximize the survival and activity of the probiotic strain

This table provides a clear taxonomy of interventions targeting the microbiota-gut-brain axis, a critical communication pathway between the central nervous system and the gut. These interventions aim to correct microbial dysbiosis, which is hypothesized to contribute to neurodevelopmental conditions like autism spectrum disorder. Probiotics are live microbial supplements that directly modulate the gut's composition and function, producing crucial metabolites like short-chain fatty acids (e.g., butyrate) that can influence the brain and modulate neurotransmitter systems (e.g., serotonin). Prebiotics are non-digestible fibers that indirectly support brain health by selectively feeding beneficial bacteria, thereby boosting short-chain fatty acid production and enhancing the intestinal barrier. Synbiotics combine both components to produce a synergistic effect potentially. In the context of autism spectrum disorder, specific studies on the combination of probiotics and fructo-oligosaccharides have shown promise, not only improving gut health but also reducing autism-related symptoms and correcting key metabolic issues, such as reducing the hyper-serotonergic state and dopamine metabolism disorder. This highlights a targeted strategy to manage autism spectrum disorder-associated symptoms and comorbidities by leveraging the connection between the gut and the brain. IBS: Irritable bowel syndrome.

Probiotics and prebiotics: Two major recent meta-analyses indicate a divergence between physiological/GI improvements and core behavioral changes. Zeng et al[230] in 2024 conducted a meta-analysis of 6 RCTs (n = 302). They found that probiotics significantly decreased the total GI severity index (MD = -0.59, P < 0.05). However, there was no statistical difference between probiotics and placebo in core ASD symptoms measured by the ABC, the SRS, or the CGI. In addition, Rahim et al[231] in 2023 conducted a meta-analysis of 10 studies. They found no significant effect on autism-related behavioral symptoms (SMD = -0.07, P = 0.65). However, they observed significant effects on brain connectivity, specifically decreased frontopolar power in beta and gamma bands, which correlated with changes in tumor necrosis factor-α levels.

In contrast to the pooled meta-analyses, recent large-scale RCTs utilizing specific strains or synbiotics have reported robust behavioral benefits, suggesting strain-specificity is key. Narula Khanna et al[232] in 2025 (RCT, n = 180): This large trial reported a significant reduction in symptom severity in the probiotic group compared to placebo. They found a significant improvement in SRS-2 (47.77% vs 23.33%; P = 0.000) and a significant reduction in ABC-2 scores for social withdrawal (40%), stereotypy (37%), and hyperactivity (34%). Lin et al[233] in 2024 (RCT, n = 60) evaluated the Bacteroides fragilis BF839 supplement in children with ASD. They found significant improvements in ABC total scores, CARS scores, and GI symptoms at week 16 compared with placebo, particularly in children with baseline CARS scores ≥ 30. In addition, Phan et al[234] in 2024 performed a large open-label “precision synbiotic” study (n = 170 completers). They found that supplementation increased gut diversity and improved GI discomfort, language, comprehension, and cognition over 3 months.

However, we found that probiotics appear most effective when combined with other therapies such as oxytocin or ABA. Kong et al[235] in 2020 found that Lactobacillus plantarum PS128, when combined with intranasal oxytocin, produced synergistic improvements in ABC and SRS scores and in gut microbiome networks compared to placebo. Li et al[236] in 2021 and Niu et al[237] in 2019 both demonstrated that combining probiotics with ABA resulted in significantly lower ATEC scores than ABA therapy alone.

Based on many studies, the clinical use of probiotics in ASD reveals that while generic multi-strain mixtures show mixed behavioral results, specific “psychobiotic” strains demonstrate consistent efficacy for core symptoms. Lactobacillus plantarum PS128 has demonstrated specific efficacy in ameliorating oppositional/defiant behaviors and enhancing social behaviors[238], while Bacteroides fragilis BF839 has significantly improved ABC and CARS scores[233]. Additionally, Lactobacillus reuteri is associated with oxytocin-dependent behavioral improvements and directional improvements in social preference[235]. In contrast, Bifidobacterium and multi-strain formulas (e.g., Visbiome/Vivomixx) appear most effective for GI symptoms and quality of life rather than core behavioral symptoms[239]. Effective dosing varies significantly by strain; positive trials have utilized 60 billion CFUs per day for Lactobacillus plantarum PS128[236], 2 million CFUs per day for Bacteroides fragilis BF839[234], and high doses ranging from 450 billion to 900 billion bacteria per day for multi-strain formulations[240,241].

Regarding the duration of therapy, while clinical effects on GI symptoms may be observed as early as 4 weeks[240], significant behavioral changes are more consistently observed in trials lasting 12 weeks to 16 weeks[233,242]. Maintenance treatment up to 6 months is suggested by some studies to be beneficial for sustained effects and gut colonization[243]. Several factors influence efficacy; outcomes are enhanced by younger age, with improvements more pronounced in children aged 7-12 or preschool age than in adolescents, and by the presence of GI symptoms, where the “GI group” shows greater improvements in adaptive functioning[242]. Combination therapy significantly enhances efficacy, such as pairing probiotics with ABA, oxytocin, or dietary interventions[236,244]. The use of synbiotic formulations with prebiotics also enhances colonization[235]. Conversely, factors that decrease efficacy include the vast heterogeneity of ASD microbiomes, which makes “one size fits all” approaches fail, and the lack of strain specificity in generic probiotics lacking psychobiotic properties. Supplementary Table 29 summarizes studies on the use of probiotics/prebiotics in children with ASD.

Fecal microbiota transplantation: Fecal microbiota transplantation (FMT), also referred to as microbiota transfer therapy (MTT) or washed microbiota transplantation, involves transferring gut microbiota from healthy donors to patients[245]. The rationale lies in the “microbiota-gut-brain axis” as shown in Figure 4. Animal models provide strong causal evidence: Transplanting fecal microbiota from children with ASD into mice or honeybees induces ASD-like behaviors, memory impairment, and neuroinflammation in the recipients. Conversely, transplanting healthy microbiota into ASD mouse models alleviates social deficits and restores excitatory/inhibitory balance[246].

Figure 4 The microbiota-gut-brain axis in autism spectrum disorder and the mechanism of microbiota transfer therapy (microbiota transfer therapy/fecal microbiota transplantation). This figure depicts the pathophysiological disturbances of the microbiota-gut-brain axis observed in autism spectrum disorder and illustrates how microbiota transfer therapy (microbiota transfer therapy/fecal microbiota transplantation) may reverse these abnormalities. In autism spectrum disorder, gut dysbiosis - characterized by reduced beneficial bacteria and increased pathobionts - leads to elevated production of harmful metabolites (e.g., p-cresol, indoles) and contributes to increased intestinal permeability (“leaky gut”). This allows microbial metabolites and inflammatory mediators to cross the intestinal barrier, triggering immune activation and the release of pro-inflammatory cytokines. Through both systemic circulation and vagal-nerve signaling, these immune and metabolic disturbances reach the brain, promoting neuroinflammation and exacerbating autism spectrum disorder symptoms. The therapeutic pathway shows how microbiota transfer therapy/fecal microbiota transplantation introduces a healthier and more diverse microbial community into the gut. This intervention increases beneficial bacteria, enhances production of short-chain fatty acids such as butyrate, propionate, and acetate, and helps restore intestinal barrier integrity. Short-chain fatty acids exert anti-inflammatory effects locally and centrally, ultimately reducing neuroinflammation and supporting symptom improvement. The figure highlights how a mechanism-based intervention can target a specific autism spectrum disorder endophenotype (gut dysbiosis/Low short-chain fatty acids state), illustrating the precision-medicine rationale for microbiota-directed therapies. ASD: Autism spectrum disorder; MTT: Microbiota transfer therapy; FMT: Fecal microbiota transplantation; SCFA: Short-chain fatty acids.

Recent RCTs present a mixed but promising picture, with some variability based on the specific protocol and primary outcomes used. A double-blind RCT (DB-RCT) by Wang et al[247] in 2024 demonstrated significant efficacy. Following MTT, the treatment group showed significant reductions in GI symptoms (P < 0.0001) and core ASD symptoms, as measured by CARS (P < 0.0001), SRS (P = 0.0002), and ABC scores (P < 0.0001), compared to controls. This was accompanied by a reduction in urinary 5-hydroxyindoleacetic acid (a serotonin metabolite). However, a large DB-RCT by Wan et al[248] in 2024 (n = 103) found no significant difference between oral MTT and placebo on the primary outcome, the SRS-2 (P = 0.03 within group, but not significant between groups). However, significant benefits were observed in secondary outcomes, specifically the socialization domain of the Vineland-3 adaptive behavior scale.

Open-label studies have consistently reported robust improvements, with some evidence of long-term durability. Kang et al[249] in 2017 reported that an intensive MTT protocol (antibiotics + bowel cleanse + MTT) led to an 80% reduction in GI symptoms and significant improvements in core ASD symptoms in 18 children. A 2-year follow-up of the same cohort (Kang et al[250] in 2019) showed that these improvements were not only maintained but autism-related symptoms improved even further post-treatment. Gut diversity increased and remained elevated, with enrichment of Bifidobacteria and Prevotella. Ye et al[251] in 2022 followed 328 patients for up to 5 years. They observed significant improvements in ABC and CARS scores that persisted for 36 months to 48 months. However, scores tended to return to baseline levels by the 60^th month, suggesting the potential need for maintenance therapy. Pan et al[252] in 2022 found that washed microbiota transplantation significantly improved ABC, CARS, and sleep scores, and noted that repeated courses (additional treatments) led to significantly better outcomes than a single course.

Considering the method of fecal transplanation (route and donor), Li et al[253] in 2024 (n = 98) compared delivery methods and found that upper GI routes (capsules or nasal jejunal tube) resulted in greater reductions in ASD symptoms (ABC, CARS, SRS) compared to lower GI routes (transendoscopic enteral tube), with a lower incidence of adverse events in the capsule group. In addition, He et al[254] in 2026 demonstrated in mice that targeted MTT from donors selected for high Lactobacillus abundance was more effective than non-targeted MTT. Wu et al[255] in 2025 validated a simplified protocol using pediatric donors, which effectively modulated the microbiota (increasing Faecalibacterium) and alleviated symptoms.

MTT appears to be generally safe in this population. The most frequently reported side effects were mild-to-moderate GI issues (diarrhea, abdominal pain, nausea, vomiting) and transient fever. Multiple studies, including the RCT by Wan et al[248] in 2024 (n = 103) and the large cohort by Ye et al[251] in 2022 (n = 328), reported no serious adverse events related to the treatment. Supplementary Table 30 summarizes studies on the use of FMT in children with ASD. To summarize the results of this study review, Table 8 presents the level of evidence for common pharmacological and biomedical interventions in children with ASD.

Table 8 Level of evidence of the common pharmacological and biomedical interventions in pediatric autism spectrum disorder.

Intervention type	Mechanism of action	Evidence level	Recommended dose (pediatric)	Common side effects	Key clinical notes
Antipsychotics
Risperidone	Antagonist of D2 (dopamine) and serotonin 2A receptors	High (FDA-approved for irritability)[1]	Initial: 0.25-0.5 mg/day. Target: 0.5-3.0 mg/day (weight-based)[2]	Weight gain, increased appetite, sedation, hyperprolactinemia, enuresis, tremor[3]	Most effective for irritability/aggression. Requires monitoring of weight, metabolic profile, and prolactin[4]
Aripiprazole	Partial agonist of D2 and serotonin 1A; antagonist of serotonin 2A	High (FDA-approved for irritability)[5]	Initial: 2 mg/day. Target: 5-15 mg/day[6]	Sedation, vomiting, tremor, akathisia, weight gain (less than risperidone)[7]	Metabolic profile generally better than risperidone. Can lower prolactin levels[8]
Psychostimulants
Methylphenidate	Blocks reuptake of norepinephrine and dopamine	Moderate[9]	Start low (e.g., 2.5-5 mg/day) and titrate. Max 2 approximately mg/kg/day or 60 mg/day[10]	Appetite suppression, insomnia, irritability, emotional lability[11]	Effective for ADHD symptoms but less tolerated in ASD than in pure ADHD. “Irritability” is a dose-limiting side effect[12]
Non-stimulants
Atomoxetine	Selective norepinephrine reuptake inhibitor	Moderate[13]	0.5 mg/kg/day to 1.2-1.4 mg/kg/day[14]	Nausea, fatigue, decreased appetite, early morning awakening[15]	Good alternative if stimulants aren’t tolerated. Effects may take weeks to appear[16]
Alpha-2 agonists
Guanfacine	Selective alpha-2A adrenergic receptor agonist	Moderate (RCTs available)[17]	1-4 mg/day (extended release)[18]	Drowsiness, fatigue, hypotension, dry mouth[19]	Effective for hyperactivity and impulsivity. Monitor BP and pulse[20]
Clonidine	Non-selective alpha-2 adrenergic agonist	Low/moderate (small RCTs)[21]	Oral/transdermal. Approximately 0.1-0.2 mg/day (titrated)[22]	Sedation (prominent), dizziness, hypotension, irritability[23]	Useful for hyperarousal and sleep. More sedating than guanfacine[24]
Mood stabilizers
Valproate	Increases GABA; blocks Na+ channels; epigenetic modulation	Low/moderate (mixed RCTs)[25]	Titrated to serum levels (50-100 μg/mL)[26]	Weight gain, GI upset, tremor, hair loss, thrombocytopenia, hepatotoxicity (rare)[27]	May reduce irritability/aggression. Avoid in metabolic disorders (mitochondrial)[28]
CBD	Modulates endocannabinoid system (low affinity CB1/CB2)	Low/emerging (one positive RCT)[29]	Variable. e.g., 20:1 CBD:THC oil or pure CBD[30]	Somnolence, decreased appetite, restlessness, diarrhea[31]	Shows promise for severe behavioral problems. Long-term safety data in kids is limited[32]
SSRI/antidepressants
Fluoxetine	SSRI	Low/negative for core symptoms[33]	Low dose start (e.g., 2.5-5 mg)[34]	Behavioral activation (hyperactivity, agitation), insomnia, GI upset[35]	Not effective for core RRBs in large trials. Use for comorbid anxiety/depression[36]
Sertraline	SSRI	Low/negative for core symptoms[37]	Low dose start (e.g., 25 mg)[38]	Hyperactivity, insomnia, GI disturbance[39]	May benefit a subgroup with specific genetic profile (solute carrier family 6 member 4 long/Long genotype)[40]
Metabolic/dietary
Methyl B12	Cofactor for methionine synthase; supports methylation	Low (small RCTs)[41]	64.5-75 μg/kg subcutaneous injection every 3 days[42]	Generally safe. Hyperactivity/agitation possible. Cobalt accumulation[43]	May improve clinical global impression in some children[44]
Folinic Acid	Reduced folate; supports methylation/DNA synthesis	Low/moderate (one positive RCT)[45]	High dose (up to 2 mg/kg/day; max 50 mg)[46]	Excitement, insomnia, aggression (rare)[47]	Improved verbal communication in FRAA-positive children[48]
L-Carnitine	Transports fatty acids for mitochondrial oxidation	Low (small/pilot studies)[49]	50-100 mg/kg/day (up to 400 mg/kg used in pilots)[50]	Fishy body odor, diarrhea/GI upset[51]	May help a subset with mitochondrial dysfunction. Odor is a limiting side effect[52]
Sulforaphane	Upregulates antioxidant response (Nrf2 pathway)	Low/moderate (mixed RCTs)[53]	Derived from broccoli sprout extract (variable potency)[54]	GI upset, insomnia, smell/taste aversion[55]	showed benefit in young adults; results in children are mixed[56]
Gut-brain axis
Probiotics	Modulate gut microbiota and gut-brain signaling	Low/moderate (mixed)[57]	Strain-dependent (e.g., Lactobacillus plantarum, Lactobacillus reuteri)[58]	GI bloating, gas (usually mild)[59]	Some strains improve GI symptoms and potentially some behavioral symptoms (anxiety)[60]
Fecal Transplant (FMT)	Restores diverse/healthy gut microbiota	Low/emerging (open label/small RCTs)[61]	Oral capsules or rectal enema (experimental protocols)[62]	GI symptoms (diarrhea, pain), fever (transient)[63]	Promising long-term data on GI and core symptoms in small cohorts. Investigational[64]
Experimental
Bumetanide	Diuretic; reduces intracellular chloride (restores GABA inhibition)	Low/controversial (failed phase 3)[65]	0.5-1.0 mg twice daily[66]	Hypokalemia (frequent), diuresis, dehydration[67]	Phase 3 trials terminated for futility. May benefit specific biological subgroups[68]
Oxytocin	Neuropeptide; enhances social cognition/bonding	Low/negative (large RCT negative)[69]	Intranasal spray (various doses, e.g., 24 IU)[70]	Nasal irritation, epistaxis[71]	Large trial showed no benefit over placebo[72]
Immunotherapy (IVIG)	Modulates immune system; neutralizes autoantibodies	Very low (case series/uncontrolled)[73]	Intravenous (hospital based)[74]	Headache, fever, aseptic meningitis, risk of infection/thrombosis[75]	Reserved for rare cases with identifiable autoimmune encephalitis/markers[76]

FDA: Food and Drug Administration; ADHD: Attention-deficit/hyperactivity disorder; ASD: Autism spectrum disorder, RCT: Randomized controlled trial; GABA: Gamma-aminobutyric acid; BP: Blood pressure; GI: Gastrointestinal; CBD: Cannabidiol; THC: Tetrahydrocannabinol; SSRI: Selective serotonin reuptake inhibitor; RRBs: Restricted and repetitive behaviors; FRAA: Folate receptor-α autoantibodies; Nrf2: Nuclear factor erythroid 2-related factor 2; FMT: Fecal microbiota transplantation.

DISCUSSION

This systematic review and meta-analysis of five major therapeutic classes confirm the persistent dichotomy and central challenge in the medical management of ASD. While established pharmacological interventions, particularly FDA-approved antipsychotics, offer effective treatment for severe behavioral comorbidities like irritability and aggression, our comprehensive analysis reveals that safe and efficacious pharmacological options for the core deficits of social communication and restricted, repetitive behaviors remain elusive. This fundamental dichotomy necessitates a shift in focus from generalized symptom management to a highly individualized, precision medicine approach.

Efficacy vs safety: The therapeutic trade-off

The most robust evidence continues to support the use of FDA-approved atypical antipsychotics - risperidone and aripiprazole - for managing severe irritability, aggression, and self-injury. Our pooled analysis aligns with previous literature showing large effect sizes for these indications. However, this efficacy comes at a significant cost[256]. The metabolic burden of these agents, particularly the substantial weight gain associated with risperidone (mean +1.97 kg over approximately 8 weeks in our analysis) and the risks of metabolic syndrome, necessitates a careful risk-benefit calculation. While aripiprazole offers a comparatively favorable metabolic profile and reduced risk of hyperprolactinemia, it carries a higher risk of akathisia[257].

In contrast, psychostimulants (MPH) and non-stimulants (atomoxetine, guanfacine) demonstrate moderate efficacy for comorbid ADHD symptoms. Crucially, our review highlights that children with ASD are less tolerant of stimulants than their neurotypical peers, with higher rates of irritability and social withdrawal. Atomoxetine and guanfacine appear to be viable, better-tolerated alternatives for this population, although their effect sizes for hyperactivity are generally smaller than those of stimulants[258].

Regarding probiotics/prebiotic use, meta-analyses (Zeng et al[230] in 2024; Liu et al[238] in 2022 generally conclude that these supplements show no significant benefit for core ASD symptoms (social interaction, restricted/repetitive behaviors). However, they consistently demonstrate significant improvement in co-morbid GI symptoms (Zeng et al[230] in 2024; Arnold et al[239] in 2019 and secondary behavioral issues like irritability and hyperactivity. Specific probiotic strains, such as Lactobacillus plantarum PS128, have shown efficacy in ameliorating opposition/defiance and improving hyperactivity in younger children. Furthermore, studies suggest these interventions modulate the axis by reducing hyper-serotonergic states, decreasing inflammatory markers (tumor necrosis factor-α), and modifying brain electrical activity (EEG) towards a more typical pattern, sometimes synergizing with other therapies like ABA or oxytocin. Overall, while general probiotics may primarily treat co-morbid GI issues, intense or targeted microbial interventions show potential for a genuine impact on secondary and core ASD symptoms, emphasizing the need for personalized approaches.

FMT, often used in conjunction with antibiotics and bowel cleanse as MTT, for treating ASD. This is one of the most intensive interventions targeting the microbiota-gut-brain axis. The clinical evidence is MIXED, particularly between early open-label findings and recent large, DB-RCTs. Strong positive signal (was observed in open label studies. The original work by Kang et al[249] in 2017 and Kang et al[250] in 2019 demonstrated robust, long-term improvement in both GI symptoms and ASD symptoms, with improvements sustained or further enhanced two years post-treatment. However, we observed mixed DB-RCT outcomes. Later large-scale RCTs show inconsistency. One trial (Wan et al[248] in 2024) was negative on the primary social measure (SRS-2) but positive on a secondary adaptive measure (Vineland-3 socialization), suggesting that the benefit may be subtle or manifest in functional, adaptive domains rather than core symptom severity. Another DB-RCT (Wang et al[247] in 2024) reported broad positive effects on multiple behavioral scales. Evidence also suggests the effect is not permanent; improvements may wane after several years (Ye et al[251] in 2022), and repeated courses may be necessary for maximal benefit (Pan et al[252] in 2022). The route of administration (oral capsules vs tube delivery) may also affect efficacy and the profile of adverse events. Overall, FMT/MTT holds significant therapeutic potential, backed by strong biochemical rationale and large effect sizes in early studies, but further research is critical to standardize the protocol and identify the specific subgroup of children most likely to respond.

Heterogeneity in risperidone meta-analysis

The substantial heterogeneity observed in the risperidone meta-analysis - particularly for ABC-I (I² = 72%) and ABC-H (I² = 81%) - reflects key clinical and methodological differences across the included trials. Several factors likely contributed to this variability. First, participant characteristics differed considerably, including variation in age ranges (from preschool-aged children to adolescents), baseline ASD severity, and the presence of comorbid intellectual disability or ADHD. Younger children and those with more severe baseline irritability often show greater symptom fluctuations, which can amplify between-study variability. Second, behavioral co-interventions differed across trials. Some studies incorporated concurrent structured behavioral programs, while others did not report the intensity or type of psychosocial support. Because risperidone’s efficacy often interacts with behavioral therapy, inconsistent reporting of these co-interventions likely contributed to outcome disparities[15-17].

Third, dose ranges and titration schedules varied widely across studies (e.g., fixed-dose, flexible-dose, and weight-based titration), creating differences in cumulative drug exposure. Fourth, treatment duration ranged from 6 weeks to 8 weeks, which may affect the magnitude of symptom improvement since risperidone’s behavioral benefits and adverse effects accumulate over time. Fifth, geographical and cultural differences - including diet, environmental factors, and parental reporting styles - may have influenced behavioral ratings and side-effect reporting. Finally, differences in outcome measurement (parent-reported vs clinician-rated vs teacher-rated instruments) and baseline rating inflation or deflation contributed to measurement heterogeneity[26-28].

Although subgroup analysis was considered, it was not feasible due to the small number of trials (n = 3 per outcome), limited reporting of stratified data, and insufficient statistical power. As PRISMA guidelines recommend, we therefore interpreted the pooled estimates with appropriate caution. Significantly, the presence of substantial heterogeneity does not undermine the consistent direction of effect, which remained firmly in favor of risperidone across all ABC subscales. Instead, the heterogeneity highlights the clinical diversity of ASD phenotypes and treatment contexts, underscoring the need for more targeted, biomarker-informed approaches to pharmacotherapy.

The “core symptom” gap and failed candidates

A significant finding of this review is the lack of efficacy for multiple drug classes previously hoped to treat core ASD symptoms. SSRIs such as fluoxetine and sertraline, once theorized to reduce repetitive behaviors via serotonergic modulation, failed to show benefit over placebo in large RCTs for this indication[55]. Similarly, the “social neuropeptide” oxytocin, despite early promise, demonstrated no significant benefit in social responsiveness in the largest multicenter trial to date[144]. This pattern of early-phase two promise followed by phase three futility - also seen with the vasopressin antagonist balovaptan[154-157] and the diuretic bumetanide[217-219] - underscores the heterogeneity of ASD and the difficulty of identifying a “one-size-fits-all” biological treatment.

The shift toward precision medicine

A prominent theme emerging across all therapeutic classes reviewed - antipsychotics, stimulants, SSRIs, metabolic/nutritional agents, and microbiota-gut-brain axis therapies - is the marked inter-individual variability in treatment response, consistent with the heterogeneous neurobiological underpinnings of ASD. The conflicting results across many interventions suggest that the biological heterogeneity of the ASD population may mask efficacy. Across interventions, the traditional “one-size-fits-all” pharmacologic model demonstrated limited effectiveness, particularly for core ASD symptoms. Instead, accumulating evidence supports the paradigm shift toward precision medicine, in which identifiable biological, behavioral, or genetic markers inform treatment selection. Our review identified several interventions that appear effective only within specific biological subgroups.

Stimulant and SSRI response also demonstrated biological stratification. Children with ASD and co-occurring ADHD showed favorable responses to MPH and atomoxetine, yet a substantial proportion experienced behavioral activation or emotional lability[90-92]. Notably, in the sertraline trial, response was significantly moderated by solute carrier family 6 member 4 (serotonin transporter) genotype: Children with the high-activity L/L genotype showed improvement, whereas S-allele carriers worsened[55]. This gene-by-treatment interaction represents one of the strongest demonstrations to date that pharmacogenetics can predict medication response in ASD.

Similarly, the utility of methylcobalamin and folinic acid is a classic example of biomarker-driven therapy. While methylcobalamin monotherapy has yielded mixed results overall, the RCT on high-dose folinic acid demonstrated significant benefits on verbal communication. Crucially, this response was strongly predicted by the presence of FRAA in the cerebrospinal fluid or with laboratory evidence of cerebral folate deficiency or methylation disruption. This finding reframes the therapy: These agents are not intended for all children with ASD. However, they are highly effective for the subgroup with confirmed defects in folate transport or methylation capacity[183-185]”. Similarly, emerging evidence suggests that the benefits of methylcobalamin, vitamin D supplementation, or L-carnitine may be most tremendous in children with corresponding metabolic or mitochondrial biomarkers.

While general probiotic supplementation shows inconsistent results for core ASD symptoms, MTT/FMT has demonstrated significant, sustained improvements in both behavioral symptoms and GI function in cohorts with documented microbial dysbiosis and comorbid GI abnormalities. The dramatic improvements reported in MTT trials suggest that these therapies are most effective when administered to the subgroup of children with ASD whose pathology is driven, at least in part, by severe gut microbial and metabolic abnormalities[246-248]. These findings support the growing view that GI abnormalities represent a clinically and biologically distinct ASD subgroup with unique therapeutic responsiveness.

The conflicting trial results for bumetanide highlight the urgent need for biomarkers. Positive results from early, small trials were not replicated in the large phase 3 trials, suggesting the drug is not broadly effective. However, subsequent studies are exploring the use of EEG or neuroimaging MRS to measure the GABA/glutamine/glutamate complex ratio to identify the specific subgroup of children with a true excitatory/inhibitory imbalance where bumetanide can successfully restore excitatory/inhibitory. Patients are most likely to respond to NKCC1 inhibition[212,213,222].

Treatments like IVIG and corticosteroids showed dramatic improvements in select cases of “regressive” or “autoimmune” autism but are essentially ineffective or risky for the broader ASD population[132,163,164]. This evidence strongly supports a shift away from trial-and-error prescribing toward a precision medicine approach, in which biomarkers guide therapeutic selection.

Integrating precision medicine into clinical practice

Together, the evidence across therapeutic classes indicates that ASD pharmacotherapy is most effective when targeted to biologically specific subgroups rather than applied broadly. This shift has significant implications for clinical practice. The future of ASD management should integrate precision medicine into clinical practice by incorporating the following, as shown in Figure 5: Biomarker-guided treatment selection (e.g., FRAA testing before folinic acid; baseline GI assessment before microbiota-directed therapies). Pharmacogenetic profiling, especially for serotonergic agents and drugs metabolized by cytochrome P450 2D6 or cytochrome P450 C19. TDM for antipsychotics, particularly risperidone, to optimize the balance between benefit and metabolic liability. Neurophysiological markers, such as EEG or MRS signatures of excitatory-inhibitory imbalance, to identify potential responders to emerging treatments such as bumetanide. Phenotype-based stratification, such as behavioral profiles, sensory reactivity, and co-occurring psychiatric features. These considerations underscore that the primary barrier to precision medicine in ASD is not a lack of candidate treatments, but rather the absence of validated, clinically deployable biomarkers. Future RCTs should therefore incorporate stratification and enrichment designs to identify responder subgroups and reduce heterogeneity that dilutes treatment effects in conventional trial designs.

Figure 5 The translational precision medicine pipeline for medical management of autism spectrum disorder. This figure illustrates the proposed precision-medicine framework for autism spectrum disorder, demonstrating how biomarker-based stratification can guide individualized medical treatment. The pipeline begins with Patient Stratification, where children with autism spectrum disorder undergo biomarker screening, including folate receptor-α autoantibodies (predicting folinic acid response), gut microbiota composition (predicting response to microbiota transfer therapy/fecal microbiota transplantation), metabolic panels (e.g., carnitine, branched-chain amino acids levels), and relevant genetic polymorphisms (e.g., methylenetetrahydrofolate reductase variants). These biomarkers identify biologically meaningful endophenotypes within the heterogeneous autism spectrum disorder population. The second stage, targeted intervention, maps each endophenotype to a mechanism-driven therapy - for example, folinic acid for cerebral folate deficiency, microbiota transfer therapy for gut dysbiosis/Low short-chain fatty acids states, and atypical antipsychotics for severe irritability/aggression. The third stage, objective outcome measures, emphasizes the use of biological markers such as cerebrospinal fluid folate levels, fecal short-chain fatty acids concentrations, and electroencephalography/functional magnetic resonance imaging signatures to monitor treatment response, moving beyond reliance on subjective scales (e.g., Aberrant Behavior Checklist, Childhood Autism Rating Scale). The pipeline concludes with personalized outcome and future research, showing that responders experience symptom improvement and enhanced quality of life, whereas non-responders re-enter the pipeline for re-evaluation. This model highlights how biomarker-driven stratification can enable mechanism-based, individualized treatments in autism spectrum disorder. ASD: Autism spectrum disorder; FMT: Fecal microbiota transplantation; BCAA: Branched-chain amino acids; MTHFR: Methylenetetrahydrofolate reductase; FRα: Folate receptor alpha; SCFA: Short-chain fatty acids; CSF: Cerebrospinal fluid; EEG: Electroencephalography; ERP: Event-related potential; ABC: Aberrant Behavior Checklist; SRS: Social Responsiveness Scale.

Emerging frontiers: Gut-brain axis and cannabinoids

Interventions targeting the microbiota-gut-brain axis represent a promising frontier. While generic probiotics showed mixed results for behavior, specific “psychobiotic” strains (e.g., Lactobacillus plantarum PS128) and FMT demonstrated significant, and in the case of FMT, long-term improvements in both GI and core ASD symptoms[235,251]. Similarly, CBD has emerged as a potent intervention for severe behavioral disruptions, with one positive RCT and consistently favorable observational data[47]. However, the long-term neurodevelopmental safety of cannabinoids in children remains a critical evidence gap.

Limitations

The quality of the primary studies limits this review. Many trials, particularly those involving nutritional supplements and immunotherapies, suffered from small sample sizes, short durations (< 12 weeks), and open-label designs susceptible to placebo effects. Another major limitation of our quantitative synthesis is the consistent finding of substantial statistical heterogeneity (I² > 50%) across multiple primary outcomes (e.g., ABC-I, ABC-H). This high heterogeneity, inherent to the ASD phenotype and to methodological variability across trials and outcome measures (e.g., parent vs clinician ratings), further complicates cross-study comparisons and underscores the need to use the random-effects model. However, it also suggests that the single pooled estimate may not uniformly apply to all children with ASD. Future research must focus on identifying clinical or biological subgroups (e.g., age, co-occurring GI symptoms, specific genetic markers) that may explain this observed variability in treatment response.

Clinical implications and recommendations for practice

Based on the systematic review and meta-analysis of the current evidence, we propose the following framework to guide clinical decision-making in the pharmacological management of pediatric ASD. These recommendations prioritize the balance between efficacy, safety, and the emerging role of precision medicine.

Hierarchical management of irritability and aggression: Our findings confirm that atypical antipsychotics remain the only pharmacological class with high-certainty evidence for treating severe irritability, aggression, and self-injury. However, the significant metabolic burden associated with these agents necessitates a stratified approach.

Selection of agent: Aripiprazole should be considered a primary option when metabolic syndrome or hyperprolactinemia is a concern, given its more favorable metabolic profile compared to risperidone. Risperidone remains a robust option for acute stabilization but requires rigorous monitoring of weight, glucose, and lipids[34-36].

Optimization: To mitigate risks, TDM for risperidone is advisable where available, targeting active moiety levels between 3.5 μg/L and 7.0 μg/L to maximize efficacy while minimizing weight gain[26]. Alternatives: For cases refractory to antipsychotics or where adverse effects are intolerable, our review supports the consideration of valproate, particularly in children with mood lability or subclinical epileptiform EEG abnormalities[78-81].

Tailored management of comorbid ADHD symptoms: While psychostimulants are highly effective in neurotypical ADHD, our review highlights that children with ASD have a narrower therapeutic window and a higher propensity for adverse effects such as irritability and social withdrawal.

Dosing strategy: A “start low, go slow” approach is essential, initiating treatment at 25%-50% of the typical pediatric dose. Non-stimulant alternatives: Atomoxetine and guanfacine (extended release) have emerged as valuable alternatives[97,105]. While their effect sizes for hyperactivity are more modest than MPH, they offer a superior tolerability profile and are particularly effective for children with co-occurring anxiety or hyperarousal.

Re-evaluating the role of SSRIs: A critical implication of this review is the need to refine the prescribing indications for SSRIs. Large-scale RCTs have consistently failed to demonstrate the efficacy of fluoxetine or sertraline for core repetitive behaviors in ASD[55,58]. Consequently, these agents should not be prescribed solely for the treatment of restricted and repetitive behaviors. Their utility is best reserved for the treatment of distinct, co-occurring anxiety disorders or depression, with careful monitoring for behavioral activation.

The transition to biomarker-guided metabolic therapy: Our analysis suggests that metabolic and nutritional interventions should move from generalized empiricism to a precision medicine model. Methylation and folate support: High-dose folinic acid and methylcobalamin should be prioritized for children with identified metabolic markers, such as FRAA or methylation deficits, where they have demonstrated specific benefits for verbal communication[183]. Mitochondrial support: L-carnitine and CoQ10 appear most beneficial in subsets of children with suspected mitochondrial dysfunction, though larger trials are needed to solidify these indications[186].

Gut-brain axis interventions as adjunctive care: Interventions targeting the gut-brain axis, including probiotics and MTT, show promise primarily for the management of GI comorbidities. Probiotics: Specific strains (e.g., Lactobacillus plantarum PS128) or multi-strain formulations (e.g., Visbiome) may be used as adjunctive therapy to alleviate GI pain and constipation, which can secondarily improve behavioral regulation[240,241]. FMT: While long-term data are promising, MTT remains investigational and should be restricted to clinical trial settings or cases of severe, refractory dysbiosis until standardized safety protocols are established.

Limitation of investigational therapies: Finally, based on current phase 3 futility data, agents such as bumetanide, balovaptan, and intranasal oxytocin cannot be recommended for routine clinical practice at this time[155,157,213,218]. Their use should remain restricted to research settings, particularly studies aimed at identifying specific biological responders (e.g., electrophysiological phenotyping with bumetanide). Table 9 summarizes the decision matrix for some of the common disorders encountered in children with ASD.

Table 9 Summary decision matrix for common associated disorders frequently met in children with autism spectrum disorder.

Primary symptom	1st line intervention	2nd line/alternative	Monitoring key
Irritability/aggression	Aripiprazole or risperidone	Valproate, cannabidiol	Weight, glucose, lipids, prolactin (risperidone)
Hyperactivity	Methylphenidate	Guanfacine, atomoxetine	Irritability, sleep, appetite
Anxiety/depression	CBT (behavioral)	Sertraline or fluoxetine	Behavioral activation (agitation)
Core social deficits	Behavioral therapy	Lactobacillus plantarum PS128 (probiotic)	Nonspecific
Language/speech	Speech therapy	High-dose folinic acid (if FRAA+)	Sleep disturbance (excitement)

CBT: Cognitive behavioral therapy; FRAA: Folate receptor-α autoantibodies.

CONCLUSION

The current medical management of ASD requires a hierarchical approach, with robust behavioral and educational interventions remaining the indispensable foundation for core symptoms. The systematic evidence confirms that, for debilitating and severe behavioral comorbidities like irritability and aggression, atypical antipsychotics (risperidone and aripiprazole) are the sole FDA-approved pharmacological options. While these agents are essential clinical tools, the meta-analysis underscores the need for continuous metabolic and endocrine monitoring. It highlights the challenges posed by high patient heterogeneity and the inability of these agents to address core ASD deficits. Consequently, the field is undergoing a fundamental paradigm shift toward precision medicine, moving beyond generalized symptom management to focus on specific, biologically defined patient subgroups. This review identifies compelling evidence for targeted, lower-risk adjunctive interventions that harness this approach, such as high-dose folinic acid for children with confirmed cerebral folate deficiency, and MTT/FMT for those with comorbid GI abnormalities and gut dysbiosis. To overcome treatment heterogeneity and translate these biological insights into effective clinical practice, future research must be aggressively reoriented: The next generation of RCTs must prioritize the identification and validation of reliable stratification biomarkers (e.g., folate receptor alpha autoantibodies, specific microbial signatures, or neurophysiological markers) to reliably match these novel biological treatments to their most likely responders. In summary, the integration of biological, genetic, metabolic, and microbiota-derived markers into clinical decision-making represents the definitive, transformative direction for ASD therapeutics, promising to usher in an era of personalized, mechanism-driven medical care for the heterogeneous ASD population.