Bumetanide for Autism: Why a Promising Drug Failed in Trials

The short answer

Bumetanide is a diuretic (water tablet) that was investigated for autism because of a theory about the brain chemical GABA. Early small trials looked promising. But two large, well-designed Phase 3 trials involving 422 children found no significant benefit over placebo on their main or key secondary outcomes. The sponsor discontinued development for autism.

A 2026 reanalysis proposed possible symptom-profile “responder” groups. It did not find a subgroup that reproduced a benefit on the Phase 3 trials' main outcome, and it has not produced a validated test for choosing children. The overall clinical conclusion has not changed.

This guide explains what the theory was, why the early results were encouraging, what went wrong in the larger trials, and what it means for families.

What is bumetanide?

Bumetanide is a prescription diuretic — a drug that increases urine output by affecting how the kidneys handle salt and water. It has been used in mainstream medicine for decades to treat fluid retention (oedema) in conditions like heart failure and kidney disease.

It is not a new or experimental drug. What was new was the idea that it might help with autism — not through its diuretic effect, but through a different mechanism involving brain chemistry.

The theory: GABA and chloride imbalance

The rationale for testing bumetanide in autism centres on a neurotransmitter called GABA (gamma-aminobutyric acid). In the mature brain, GABA is the main inhibitory neurotransmitter — it calms neural activity. But in early brain development, GABA actually has an excitatory (stimulating) effect. This is because of how chloride ions are balanced inside and outside neurons.

Two proteins control this balance:

• NKCC1 — pumps chloride into neurons (making GABA excitatory)
• KCC2 — pumps chloride out (making GABA inhibitory)

During normal brain development, KCC2 gradually takes over from NKCC1, and GABA switches from excitatory to inhibitory. The hypothesis was that in some autistic individuals, this switch may be delayed or incomplete — meaning GABA remains partly excitatory when it should be inhibitory. This could contribute to the excitation/inhibition imbalance often discussed in autism neuroscience.

Bumetanide blocks NKCC1. By reducing chloride inside neurons, it could — in theory — restore GABA's normal inhibitory function and correct that imbalance.

It is a biologically plausible theory. But biological plausibility does not guarantee clinical effectiveness.

What the early trials showed

Pilot studies (2010–2015)

Several small, early studies reported encouraging results:

• A 2012 randomised controlled trial (60 children over 3 months) found significant reductions in CARS scores (autism severity) compared to placebo
• Improvements were also seen on the Clinical Global Impressions scale
• A 2015 Chinese study combining bumetanide with behavioural therapy reported improvements in both groups
• Brain imaging studies suggested bumetanide might affect how autistic children process faces — normalising activity in the amygdala and fusiform face area

These results were genuinely promising and generated significant research interest. The mechanism made sense, the early numbers were positive, and the side effects seemed manageable.

Phase 2 trials

A larger Phase 2 trial (88 participants, ages 2–18) tested different doses. Results were mixed — some improvements were observed, but there was no clear dose-response relationship (higher doses did not consistently work better). The most common side effect was mild hypokalaemia (low potassium), which was manageable with supplementation. Higher doses caused more urination and higher dropout rates.

What happened in the Phase 3 trials

This is where the story changes.

Two large, multicentre, randomised, double-blind, placebo-controlled Phase 3 trials were conducted:

• SIGN 1 — children and adolescents aged 7–17
• SIGN 2 — children aged 2–6

Each trial randomised 211 children—107 to bumetanide and 104 to placebo—and tested oral solution over 26 weeks. The primary outcome was change on the CARS-2 autism-severity scale.

Both trials were stopped early. In SIGN 1, the average CARS-2 difference was −0.45 points (95% confidence interval −1.64 to 0.74; p=.455). In SIGN 2, it was +0.35 points (−1.04 to 1.75; p=.617). Neither result showed a treatment benefit. The Social Responsiveness Scale, Clinical Global Impression, Vineland Adaptive Behavior Scale and quality-of-life outcomes also did not show consistent benefit.

The sponsor — Servier — discontinued development of bumetanide for autism.

The full results were published in Autism Research in October 2023. This single publication reports the two Phase 3 studies; it is not a third separate trial.

Side effects in the Phase 3 trials

The expected diuretic burden was common. Low potassium occurred in 19.6% of children taking bumetanide versus 3.8% taking placebo in SIGN 1, and 16.8% versus 1.9% in SIGN 2. Thirst and increased urination were also more common with bumetanide. More children taking bumetanide stopped because of adverse events or needed potassium supplements.

One five-year-old died after recurrent cardiac arrests during and after open-label exposure. The investigator judged the death unrelated to bumetanide, but the paper states that a causal role could not be excluded. This does not show that bumetanide caused the death; it is an important part of an honest safety account.

Why did the early trials look positive but the large ones didn't?

This is a common pattern in medicine, and it is worth understanding:

1. Small studies can overestimate effects. With only 60 participants, random variation can produce results that look significant but don't replicate in larger samples.
2. Placebo effects are strong in autism trials. Parents and clinicians who believe a treatment is working can unconsciously rate improvements more generously. This is especially true with subjective outcome measures like behaviour rating scales.
3. Open-label bias. Some early studies were not blinded — participants knew they were receiving the drug. This can inflate apparent benefit.
4. Publication bias. Positive small studies get published and cited. Negative small studies often don't. This creates a misleadingly positive evidence base before the definitive trial is run.
5. Biological plausibility ≠ clinical effectiveness. The GABA/chloride hypothesis may be partially correct, but that does not mean a drug that targets one part of the pathway will produce a measurable clinical effect in a complex, heterogeneous condition like autism.

This is exactly why large, well-controlled trials exist — and why families should wait for them before investing hope (or money) in treatments supported only by small studies.

Is anyone still researching bumetanide for autism?

The 2026 “responder subgroup” paper

Researchers used a machine-learning method to look again at the two negative Phase 3 datasets. They tested thousands of possible combinations of baseline symptoms to see whether a smaller group might have responded.

Three symptom-profile rules produced a repeatable signal on the parent-rated SRS-2 secondary outcome across the two existing datasets. But no subgroup reproduced a benefit on CARS-2, the Phase 3 primary outcome. The analysis also excluded some children who stopped because of adverse events, withdrew or had missing main-outcome data, and the rules were developed and checked using the same two trials rather than a new prospective study.

The paper's authors describe the profiles as candidates for further validation. They are not a blood test, brain scan or questionnaire that a clinician can use to predict benefit today. A properly pre-registered trial that enrols the proposed subgroup would be needed before using these rules to select children or justify prescribing.

Other small studies

A 2025 Swedish waitlist study involved only 15 children and had no blinded placebo group. Nine completed treatment; four parents reported clear improvement, but the available adaptive-behaviour and communication comparisons were not statistically significant. Two children stopped after behavioural deterioration or hyperactivity, and four more discontinued during longer follow-up. This cannot overturn the Phase 3 evidence.

The broader question of whether the excitation/inhibition imbalance theory holds up in autism is still actively researched — but bumetanide as the drug to address it has, for now, been set aside by the pharmaceutical industry.

It is possible that future drugs targeting the same pathway with better specificity or fewer side effects could be developed. But that is speculative.

What UK guidance says

NICE says autistic people should not be offered medication for the core features of autism. Bumetanide is licensed in the UK as a loop diuretic for oedema caused by heart, kidney or liver disease—not for autism.

Any autism use would be off-label and contrary to the current negative efficacy evidence. When bumetanide is used for a licensed medical reason, UK product information requires clinical monitoring of electrolytes such as potassium and sodium, hydration, blood pressure and kidney function. That monitoring need is a safety warning, not an autism-treatment protocol.

What this means for families

If you have come across bumetanide whilst researching autism treatments, here is the honest picture:

• The theory behind it was biologically plausible but remains unproven
• Early small trials looked promising
• Large, well-designed trials found no benefit over placebo for the core features of autism
• The drug's developer has stopped pursuing it for autism
• It has real side effects (low potassium, excessive urination) that require monitoring
• The 2026 subgroup paper does not provide a validated way to identify a child who will benefit
• Current evidence does not support bumetanide as an effective treatment for the core features of autism

This is a case where the research process worked as it should. A hypothesis was tested, early signals were investigated, and when the definitive trials came back negative, development was discontinued. That is not a failure of science — it is science functioning properly.

Bumetanide is also the textbook example of a wider pattern in autism research: encouraging Phase 2, disappointing Phase 3. If you've seen this story repeat across other "promising" autism treatments, our guide on why so many autism trials fail explains the structural reasons — and gives you a 5-question checklist for reading the next breakthrough headline.

Questions parents should ask

1. If a clinician suggests bumetanide for my child, what evidence are they relying on? The Phase 3 trials were negative. Any recommendation should acknowledge this.
2. Are the side effects worth the risk for an unproven benefit? Hypokalaemia requires blood monitoring. Increased urination can be disruptive and distressing, particularly for young children.
3. Could the time and effort be better directed? Ask what communication, health, sensory, educational or daily-living support best matches the child's identified need.

The bottom line

Bumetanide for autism is a cautionary example of how promising early results can fail to translate into proven treatments. The biological theory was plausible and the early data were encouraging. But when tested properly at scale, the drug did not deliver measurable benefit for the core features of autism.

This does not mean the research was wasted. Understanding why bumetanide failed will inform future approaches to targeting the excitation/inhibition balance in autism. But for families today, the picture is clear: current evidence does not support bumetanide as an effective treatment for the core features of autism.

If you encounter it being promoted online or offered by a clinician, the Phase 3 evidence should be the starting point for any conversation.

---

Frequently asked questions

Is bumetanide still being studied for autism?

Researchers continue to analyse possible subgroups and biomarkers, but the main pharmaceutical development programme was discontinued after negative Phase 3 results. No subgroup or biomarker is validated for selecting children, and the GABA/chloride hypothesis is still being explored through research.

Can I get bumetanide prescribed for my autistic child?

Bumetanide is a prescription diuretic available in the UK, but it is not licensed for autism. Any use for this purpose would be entirely off-label and not supported by current evidence. Given the negative Phase 3 trials, most clinicians would not recommend it.

What are the side effects?

The main concerns are low potassium, thirst, increased urination and dehydration. Bumetanide can also affect other electrolytes, blood pressure and kidney function. These are predictable effects of a diuretic and require medical supervision and blood monitoring when the medicine is used for a licensed medical reason.

Can a test show whether my child is a bumetanide responder?

No validated clinical test can do this. The 2026 symptom-profile analysis and earlier EEG, cytokine and brain-imaging studies are exploratory. They have not been confirmed in a prospective trial that shows a test can predict benefit over placebo.

Why did the small trials show benefit but the large ones didn't?

This is common in medicine. Small trials are more susceptible to placebo effects, random variation, and bias. Large, well-controlled trials provide more reliable evidence. The Phase 3 trials enrolled hundreds of children and found no benefit for the core features of autism — this is the most trustworthy evidence available.

Does this mean the GABA theory of autism is wrong?

Not necessarily. The GABA/chloride imbalance hypothesis may have some validity in certain individuals. But targeting it with bumetanide specifically did not produce a clinical benefit in the broad autism population tested. Future research may explore more targeted approaches.

---

Sources and further reading

• Fuentes et al. (2023) — Bumetanide oral solution for the treatment of children and adolescents with autism spectrum disorder: results from two randomised Phase III studies. Autism Research. PubMed: 37794745
• Rabiei et al. (2026) — retrospective machine-learning subgroup analysis of the two Phase 3 trials. Translational Psychiatry.
• Lemonnier E et al. (2012) — A randomised controlled trial of bumetanide in the treatment of autism in children. Translational Psychiatry.
• Lemonnier E et al. (2017) — Effects of bumetanide on neurobehavioral function in children and adolescents with ASD. Translational Psychiatry.
• Fernell et al. (2025) — 15-child waitlist study of bumetanide. Acta Paediatrica.
• NICE QS51 — medication should not be used for the core features of autism
• NHS — bumetanide uses, cautions and monitoring information
• UK Summary of Product Characteristics — bumetanide oral solution

---

Disclaimer: SENDPath provides information for families navigating SEND in Kent and beyond. We are not clinicians. Nothing on this page constitutes medical advice. Always consult a qualified healthcare professional before making treatment decisions for your child.