Arbaclofen and Autism: What the 2026 Trial Actually Found

The short answer

The fair interpretation is not “it works” or “it was a complete failure”. It is: the confirmatory outcomes were negative, while exploratory findings may help researchers design another study. That is a much weaker position than an available treatment.

What is arbaclofen?

Arbaclofen—also called R-baclofen or STX209—is one half, or enantiomer, of the medicine baclofen. It activates GABA-B receptors, part of a signalling system that usually reduces nerve-cell activity.

Researchers have long studied whether differences in excitatory and inhibitory signalling might contribute to particular autistic difficulties. Earlier open-label and small studies created interest in arbaclofen for social functioning. A 2017 randomised trial of 150 participants did not meet its primary social-withdrawal outcome, but post-hoc analyses suggested a possible signal among participants with fluent speech. The 2026 study was designed around that narrower group.

How did the 2026 trial work?

The AIMS-2-TRIALS study was a 16-week, randomised, double-blind, placebo-controlled Phase 2 trial. It recruited at seven sites in Spain, France and the UK between 2019 and 2022.

• 123 autistic children and teenagers aged 5 to 17 were randomised.
• 122 were included in the main analysis: 59 on arbaclofen and 63 on placebo.
• Participants used fluent or complex verbal language, corresponding to ADOS modules 3 or 4.
• The primary outcome was change on the Vineland-3 socialisation domain.
• A key secondary outcome was clinician-rated autism severity using the CGI-S.

This design matters. It tested a group that earlier analyses had suggested might be more likely to respond, rather than including every autistic child.

What did the main results show?

The primary Vineland socialisation result was an adjusted 1.30-point difference favouring arbaclofen, but the 95% confidence interval ran from −2.6 to 5.1. Because that interval includes no difference and possible disadvantage as well as possible benefit, the result was not statistically significant.

The key clinician-rated CGI-S outcome also showed no meaningful difference: −0.03 points, with a 95% confidence interval from −0.74 to 0.69.

Those are the results the trial was chiefly designed to test. On that basis, it did not establish efficacy.

What about the positive secondary outcomes?

Many parent, teacher and quality-of-life measures moved in the same direction and several reached nominal statistical significance. These included parts of the Social Responsiveness Scale, Autism Impact Measure, Aberrant Behavior Checklist and quality-of-life questionnaires. An exploratory responder definition on the SRS-2 was met by 27% on arbaclofen and 14% on placebo.

That sounds encouraging. Three cautions stop it becoming proof:

1. Many outcomes were tested. The paper says no correction was made for multiple comparisons. The more measures researchers test, the more likely one will appear positive by chance.
2. The primary and key clinician outcome were negative. Secondary findings are better used to form the next hypothesis than to overturn the main answer.
3. Most positive results came from questionnaires. They may reflect real change, but expectation, missing data and rater differences can affect scores. Teacher questionnaires were available for only about two-thirds of participants, partly because of COVID-19 disruption.

The study authors themselves call for larger trials. SENDPath agrees: the results justify research, not a treatment claim.

What did the trial find about side effects?

Sleep-related problems were reported substantially more often with arbaclofen: 34 of 59 participants (57.6%) compared with 22 of 63 (34.9%) on placebo. Six participants taking arbaclofen withdrew, compared with two taking placebo. Three withdrawals were judged possibly due to adverse events—two in the arbaclofen group and one in the placebo group.

One serious adverse event—psychotic symptoms—occurred in the placebo group. That event cannot be attributed to arbaclofen. The trial was not large enough to detect rare harms or settle long-term safety.

Parents sometimes hear that GABA-targeting medicines are “calming”. The sleep findings are a reminder that individual nervous systems can respond in unexpected ways and that a theory does not predict a child's experience.

Who does this evidence apply to?

The trial included 5-to-17-year-olds able to use fluent or complex speech. It does not tell us whether arbaclofen helps nonspeaking children, younger children, autistic adults or people with different intellectual and medical profiles.

This is not a criticism of recruiting a defined group. Narrow studies can answer clearer questions. It does mean headlines such as “arbaclofen improves autism” are far broader than the research.

Is arbaclofen available for autism in the UK?

No marketed arbaclofen autism treatment is available in the UK. The medicine remains investigational. NICE does not recommend medication for the core features of autism in children and young people.

A UK site participating in a trial does not mean the NHS has adopted the treatment. Trial participation has a protocol, eligibility rules, consent, monitoring and oversight that ordinary prescribing does not recreate.

Questions to ask if someone offers a similar treatment

• Is this actually arbaclofen, or ordinary baclofen?
• What diagnosed problem is being treated?
• How is a failed primary outcome being explained?
• What sleep and withdrawal effects will be monitored?
• Is the offer part of a registered, ethically approved trial?

The bottom line

The 2026 arbaclofen study was careful, publicly supported research that answered an important question. Its main answer was negative. The secondary pattern is interesting but uncertain, and sleep problems were more common with treatment.

Arbaclofen may be studied again. For now, it is not a proven autism treatment, it is not an ordinary baclofen prescription, and it is not something a family can safely recreate outside a regulated trial.

See the wider 2026 autism-treatment research guide, or learn how to assess primary outcomes, subgroups and placebo response.

Sources and further reading

• Parellada et al. — AIMS-2-TRIALS arbaclofen Phase 2 study, 2026
• Arbaclofen 2026 trial — free full text
• ClinicalTrials.gov — AIMS-2-TRIALS study record
• Veenstra-VanderWeele et al. — earlier 2017 arbaclofen Phase 2 trial
• NICE CG170 — autism support and management recommendations

Disclaimer: Information only, not medical advice. Evidence and source status were checked on 29 August 2026. Do not start, stop, share or substitute baclofen or another prescription medicine based on an online article.