Sulforaphane & Autism: What RCTs and the 2025 Meta-Analyses Show

Sulforaphane is not an established autism treatment. The human evidence is clinically interesting but mixed: some randomized trials report improvements on selected behavioral measures, others miss their primary outcome or show no significant between-group benefit, and two 2025 meta-analyses reached different pooled conclusions.

That disagreement is important rather than inconvenient. Participant age, outcome scale, treatment duration, formulation, dose, assessor and whether sulforaphane was used alone or alongside medication can all change the result.

Sulforaphane autism trials at a glance

The 2014 trial: why sulforaphane became an autism research topic

The influential 2014 randomized trial enrolled 44 males aged 13–27 years with moderate to severe autism. Participants received broccoli-sprout-derived sulforaphane or placebo for 18 weeks followed by a four-week washout.^[1]

Several behavioral measures improved during treatment, and scores moved back toward pretreatment values after withdrawal. The withdrawal pattern made the result biologically intriguing and helped drive subsequent clinical research.

But the study was small, male-only and restricted to adolescents and young adults. It did not establish efficacy across young children, females, all autism severity levels or every sulforaphane formulation.

The 2020 adjunctive trial: positive, but alongside risperidone

In 60 children aged 4–12 years, sulforaphane was tested as an adjunct to risperidone rather than as a stand-alone treatment. Irritability and hyperactivity/noncompliance improved more than placebo over 10 weeks, while several other ABC-C domains did not significantly differ.^[2]

Why this matters: a positive adjunctive result cannot be rewritten as proof that sulforaphane alone produces the same effect.

The 2021 pediatric trial: primary outcome null, secondary signals positive

Zimmerman and colleagues randomized 57 children aged 3–12 years to sulforaphane or placebo. In the initial 15-week double-blind phase, the primary overall clinical outcome did not significantly differ between groups.^[3]

A caregiver-rated secondary scale and several biological markers showed positive signals. This is exactly why primary and secondary endpoints must remain separate: a study can be biologically interesting without being a positive primary efficacy trial.

The larger multicenter study: caregiver and clinician ratings disagreed

The multicenter trial enrolled 108 children and measured both caregiver- and clinician-rated outcomes. Caregiver-rated scales did not significantly differ between sulforaphane and placebo, while clinician-rated scores improved and a responder subset emerged.^[4]

This rater discrepancy is not a minor detail. It weakens any simple claim that the trial was either clearly positive or clearly negative.

The 2023 trial: a genuinely null counterweight

A prospective randomized double-blind study in young children found no significant between-group improvement on ABC, SRS-2 or ADOS-2 outcomes.^[5]

The sample was small and dropout was substantial, but the result matters because it demonstrates that positive findings have not reproduced consistently across age groups and study designs.

What do the two 2025 meta-analyses add?

The two contemporary meta-analyses synthesized overlapping sulforaphane trials but did not produce identical headline conclusions.

Guo 2025 found no significant pooled improvement in ABC total score. Irritability and hyperactivity showed signals in analyses of interventions lasting more than 10 and up to 18 weeks, while other domains were less consistent.^[6]

Wang 2025, a systematic review and meta-analysis restricted to six randomized trials, reported small pooled improvements in several symptom domains, including total symptoms, aberrant behavior, hyperactivity, social interaction, social communication and restricted/repetitive behaviors. Several other domains were not significantly different, heterogeneity was substantial for some analyses, and common adverse-event rates did not significantly differ from control.^[7]

Editorial interpretation: neither synthesis should be cherry-picked as the final answer. Their disagreement is a signal that pooled estimates depend on outcome definition, scale, assessor, duration and analytic choices. The evidence supports continued research into selected symptom domains; it does not establish a broad or standard autism-treatment effect.

Why do the studies disagree?

• Different ages. Trials range from preschool-aged children to young adults.
• Different outcome scales. ABC, SRS, ADOS and clinician impression scales measure overlapping but not identical constructs.
• Caregiver vs clinician ratings. The largest trial produced different conclusions depending on rater.
• Adjunctive vs monotherapy designs. One positive trial added sulforaphane to risperidone.
• Different formulations and delivery. Broccoli-sprout extracts and branded glucoraphanin/myrosinase systems do not produce identical exposure.
• Duration and analysis. Both duration and the way outcomes are pooled can affect the apparent signal.

What doses have been studied?

Autism trials commonly report dose in micromoles of delivered sulforaphane and often adjust dose by body weight. That is very different from copying a milligram number from a retail broccoli-extract label.

For example, the 2014 trial used weight-adjusted broccoli-sprout-derived sulforaphane, while the 2020 adjunctive trial used 50 or 100 µmol/day depending on body weight. Because precursor and active-sulforaphane systems differ in conversion, retail formula weight should not be treated as equivalent to trial dose.

See our sulforaphane dosage guide and bioavailability guide for dose-unit and delivery differences.

What about Avmacol?

Avmacol, a glucoraphanin + active-myrosinase formulation, has been used in autism research and has one of the stronger product-specific research footprints in the category. The larger multicenter trial produced mixed caregiver-versus-clinician results rather than a uniformly positive outcome.^[4]

Product-specific trial participation is scientifically stronger than borrowing evidence from unrelated formulations. It is still not an approved treatment claim.

Is sulforaphane safe for children with autism?

The trials generally report manageable short-term adverse effects, but that does not create a blanket pediatric wellness recommendation. Participants in clinical trials were screened, monitored and treated under study protocols.

Children with seizure disorders, gastrointestinal disease, significant comorbidities or concurrent medications need individualized review. Long-term routine use of concentrated sulforaphane products has not been established for every pediatric population.

See our broader sulforaphane safety guide.

Bottom line

The best current description is “mixed but research-worthy.” Individual RCTs include positive, mixed and null findings. The two 2025 meta-analyses also disagree in their headline pooled conclusions: one is more conservative for ABC total score, while the other reports small pooled benefits across several domains.

That uncertainty is exactly why sulforaphane should not be presented as established autism therapy or used to replace behavioral, educational or medical care.