Spectrum Connect reviews published research on interventions parents are exploring for their autistic children — so you can see where the evidence actually stands. No agenda, no selling, no cherry-picking. Just the studies, our method, and what it means for you.
Bumetanide (Oral)Not supported — no clear benefit found
Key Takeaways
The largest, most careful trials found no benefit — two large randomized trials (211 children, ages 2–17) compared bumetanide to a placebo for about six months. Neither found it improved core autism traits, and the drug's maker stopped developing it for autism after these results.
Real side effects, with no offsetting benefit — bumetanide is a diuretic: it causes frequent urination and can lower blood potassium, which is why the trials monitored blood tests. That's a real burden to weigh when the trials found no clear benefit.
A “responder subgroup” idea is unproven — some researchers are re-analyzing the trial data to see if a small subgroup responded. That's an early, after-the-fact idea, not evidence the drug works, even for some children.
If your child's doctor has raised bumetanide, this research summary — not a recommendation — is a starting point for that conversation, including about the monitoring it requires.
What this means for you
Bumetanide is a diuretic — a “water pill” normally used to remove extra fluid from the body — tried for autism because of a theory about how brain cells handle a salt called chloride. Two large, carefully run trials (one in children 2–6, one in ages 7–17, 211 children in the main phase) compared it against a placebo over about six months. Neither found a benefit on the main measure of autism traits, and the company developing the drug stopped its autism program afterward.
Earlier, smaller studies had looked more promising, but they were smaller and less rigorous, and the larger trials didn't confirm them. Because bumetanide is a water pill, it makes children urinate much more often and can lower blood potassium — expected, manageable effects, but a real burden when the trials found no clear benefit on the other side.
Two large, careful trials agree bumetanide doesn't help core autism traits — and it comes with a real, monitored side-effect burden.
Who was studied. Autistic children and adolescents ages 2–17, given oral bumetanide at varying doses for about six months, with blood tests to monitor for low potassium. Generalizability beyond this population is not established.
Where the studies landed
Not supported — no clear benefit found
Tap a band to see what those studies actually said.
Points against it helping2
The two definitive Phase III trials (SIGN-1 and SIGN-2, n=211) found no significant difference from placebo on the primary autism-severity measure — and a 2024 meta-analysis incorporating those trials confirms no benefit once the strongest evidence is weighted properly.
Mixed / superseded5
Earlier, smaller Phase 2 trials and a positive 2021 review all pre-date the Phase III results and were not confirmed by them — the field moved on once the larger trials reported. A separate 2024 review reported mixed conclusions. A 2026 machine-learning reanalysis raises a possible “responder subgroup” hypothesis, but it's exploratory, not confirmatory evidence.
Tap any tile to read that study
Each tile is one study. The ringed tiles are systematic reviews or meta-analyses that pool multiple trials — the stronger kind, when they're current.
See the research behind thisSearch strategy, PRISMA flow & evidence strength — 7 studies
01
Where we looked
This run was a scoping search only — done via general web search, not the reproducible Boolean search of record and not the PubMed/Epistemonikos API layer we use on a fully conformant run. That means we can't publish reproducible per-database counts or a formal PRISMA flow for this run. Below is the search string a full conformant pass would run against PubMed, the Cochrane Library, and Epistemonikos — we haven't executed it against the database APIs yet.
("autism spectrum disorder"[MeSH] OR "ASD"[tiab] OR autis*[tiab]) AND ("bumetanide"[MeSH] OR "bumetanide"[tiab]) AND ("randomized controlled trial"[pt] OR "systematic review"[pt] OR "meta-analysis"[pt])Run on PubMed →
02
What we did with what we found
—web-search scoping (counts not reproducible this run)
n/aduplicates, off-topic, and superseded records set aside
Two large Phase III trials (n=211) found no significant difference between bumetanide and placebo on the primary autism-severity measure at 26 weeks. The drug's sponsor discontinued development for autism after these results.
How sure
Moderate
Repetitive / routinized behavior
Results here are mixed and low-certainty — not proof of benefit, and not confirmed by the strongest trials. A post-hoc reanalysis has raised a possible responder subgroup, but that's a hypothesis, not evidence.
How sure
Low
Tolerability / safety
Frequent urination and lowered blood potassium are expected with this diuretic and require blood-test monitoring. These are real, manageable side effects — but a real burden when weighed against no confirmed benefit.
How sure
Moderate
Ray Kawai · Protocol v4.5BCAT · Open record · Gate D pending
Spectrum Connect is not a medical provider, and nothing here is medical advice. This page shows where the research stands and how we got there. It is not a recommendation, and it is not a substitute for your child’s doctor or care team. What you do with it is yours to decide, together with them.
Test run — not for publication · Awaiting independent sign-off · not medical advice
Think we got something wrong?
We publish the whole record so it can be checked — and that only counts if we act on what you find. If a number looks wrong, a study is missing or has been retracted, or we’ve read a finding in a way the evidence doesn’t support, tell us.
You don’t need a research background to file one. “This doesn’t match what our doctor told us” is a useful report. Every one reaches a person: we reply within seven days, and within thirty we have either corrected the page or told you when we will. Substantive reports send the affected steps back through the protocol and need fresh sign-off before anything here changes.
The full record for bumetanide, open for anyone who wants to check our work.
Who does each step
A research agent does the mechanical and drafting work. A person checks it. An independent expert signs it before anything is published. code automatic · agent AI draft a human verifies · human a named person decides.
01
Define humanquestion + outcomes
Does bumetanide help autistic children? Pre-specified subgroups: age 2–6 vs 7–17, and the claimed EEG/biomarker “responder” subgroup (flagged post-hoc). Protocol v4.5 + Amendment v4.6 Rev C, Track A (pharmacological, contested).
02
Register humanPROSPERO + OSF
R1 prospective public registration not filed this run — a blocking conformance item. Logged as a deviation.
03
Search codedatabases
Web-search scoping only this run, not the Boolean search of record or the PubMed/Epistemonikos API layer. Per-database counts are not reproducible, so no publishable PRISMA flow. Logged as the deviation of record.
3.5
Intake checks codestanding + retraction
Citations resolve to real, non-retracted records via the search index. The pivotal Fuentes 2023 trial carries a published October 2023 correction revising its conflict-of-interest statement — an administrative correction, not a retraction; distinguished and passed. 0 DOIs were independently dereferenced this run — the fetch layer was blocked by bot-detection.
04
Screen agenthumantwo reviewers
Scoping search only; dual independent screening not performed this run — single AI screener.
Gate A
At least one solid review available? Yes — eligible systematic reviews exist (not all Critically Low). Route: overview of reviews (Track A).
05
Appraise agenthumanAMSTAR 2 / RoB 2
RoB 2 for the pivotal Phase III trials (Low–Some concerns, provisional; industry-funded, noted). AMSTAR 2 for the reviews (Low–Moderate, provisional). The 2026 machine-learning reanalysis is explicitly not a systematic review — exploratory, hypothesis-generating only, and cannot be a verdict basis. Dual independent human rating not performed — single AI appraiser.
06
Map overlap agentcodeshared trials
The reviews draw on the same trials (SIGN-1/2, BAMBI, Zhang, and others). Whole-review overlap matrix (CCA) not computable this run — included-study lists not retrieved, Epistemonikos matrices unreachable. Unverifiable cells marked, not asserted.
Gate B
Overlap resolved? Not fully — high overlap expected but not quantified this run. Flagged to resolve at the primary-study level.
07
Synthesize agenthumanper outcome
The two definitive Phase III trials found no benefit on core autism severity; a 2024 meta-analysis incorporating them agrees. Earlier positive Phase 2 trials and reviews pre-date the Phase III results and were not confirmed by them.
Gate C
Strongest reviews agree? The apparent controversy dissolves on two grounds: the positive meta-analyses pre-date the negative Phase III trials and are superseded, and the earlier Phase 2 signal did not replicate in Phase 3. Among credible, up-to-date reviews there is no unresolved pro-benefit disagreement — reported honestly, including the active, exploratory responder-subgroup direction, without manufacturing a live controversy.
08
Rate certainty agenthumanGRADE
Core autism severity: Moderate (no-benefit direction) — two Phase III RCTs could support High, but single-sponsor funding and heterogeneity argue for Moderate; either way the direction is no-benefit. Repetitive behavior: Low, unclear/mixed direction — the post-hoc responder signal must not, on its own, move this to a benefit-flavored label. Two human raters would settle both exact levels.
Gate D
Independent sign-off — required before any publishing.Cannot pass: independent reviewer bench not staffed (≥2 required), and process is non-conformant — registration, PRESS review, live per-database counts, screener calibration, and dual rating at appraisal and certainty all pending. Correctly blocked pre-publication.
09
Set readout codedecision table
Core autism severity {no_benefit, Moderate, no harm signal, no unresolved disagreement} → row 5, “Not supported: studied, no benefit found.” Repetitive behavior separately rated {unclear, Low} would mechanically map to row 6 (“Promising but contested”) — deliberately not surfaced as a second top-level verdict here, since the run's own review flags that render as a defect on low-certainty, unclear-direction data (see below).
10
Translate agenthumanplain language
Written for a tired parent. No dose or monitoring schedule given as advice — that's a decision for your child's prescribing doctor.
11
Publish codeopen record
Not yet published as a fully-conformant run — staging draft, pending Gate D.
Gate E
Living surveillance. Highest-value trigger: a prospective RCT that pre-specifies the EEG/biomarker responder hypothesis (not post-hoc), or any European regulatory-review outcome. Note the sponsor has discontinued its program, so new confirmatory trials may be sparse.
The people accountable
RK
Lead synthesizer · steps 1, 4, 5, 7, 8, 10
Ray Kawai
BCAT (IBCCES) · Registered Behavior Technician · BS Cell Biology, UC Davis — this run's dual independent human rating not yet performed (single AI appraiser)
Active
+
Independent clinical sign-off · recruiting / Gate D
Open role — recruiting
A conflict-free developmental pediatrician, child psychiatrist, or clinical psychologist (PhD/PsyD) who did not produce the synthesis.
Unfilled
Why the empty slots are shown. We do not display experts we do not have. Roles still open are shown as open.
Points against it helpingTwo Phase III RCTs · double-blind · n=211
Bumetanide oral solution for the treatment of children and adolescents with autism spectrum disorder: results from two randomized phase III studies (SIGN-1, SIGN-2)
Fuentes J, Parellada M, Georgoula C, et al. · Autism Res 2023;16(10):2021-2034 · DOI 10.1002/aur.3005 · PMID 37794745
What it looked at
Two Phase III, double-blind, placebo-controlled trials (n=211 in the double-blind phase) — one in children 2–6, one in ages 7–17 — measuring core autism severity (CARS2) at 26 weeks.
What it found
No superior effect of bumetanide over placebo on the primary CARS2 outcome in either trial. The sponsor discontinued development of bumetanide for autism after these results.
Quality — our provisional read
Provisional RoB 2: Low to Some concerns. Industry-funded (Servier/Neurochlore) — noted, not disqualifying. Carries a published October 2023 correction to its conflict-of-interest statement (administrative, not a retraction).
Mixed — lower-quality reviewSystematic review + meta-analysis · pooled
Can bumetanide be a miraculous medicine for autism spectrum disorder: meta-analysis evidence from RCTs
Xiao HL, Zhu H, Jing JQ, et al. · Res Autism Spectr Disord 2024;114:102363
What it looked at
A meta-analysis of RCTs of bumetanide in autism.
What it found
Reported findings that don't resolve cleanly with the pivotal Phase III trials — read alongside the stronger, more current evidence rather than as a standalone verdict basis.
Superseded by later trialsSystematic review + meta-analysis · pooled
Treatment effect of bumetanide in children with ASD: a systematic review and meta-analysis
Wang T, Shan L, Miao C, Xu Z, Jia F · Front Psychiatry 2021;12:751575
What it looked at
Earlier trials of bumetanide in autistic children, pre-dating the Phase III results.
What it found
Reported a positive effect — but its search predates the negative Phase III trials, so this finding is superseded by more current, more rigorous evidence.
Quality — our provisional read
Superseded — search date predates the pivotal Phase III trials.
Exploratory — not a verdict basisPost-hoc machine-learning reanalysis
Treating autism with bumetanide: identification of responders using the Q-Finder machine-learning algorithm
Transl Psychiatry 2026 · DOI 10.1038/s41398-026-03848-3
What it looked at
A machine-learning re-examination of the already-null Phase III trial data, looking for a subgroup that might have responded.
What it found
Suggests a possible responder subgroup — an after-the-fact, data-driven hypothesis, not confirmatory evidence. It would need a fresh trial designed to test it directly.
Quality — our provisional read
Not a systematic review — exploratory, hypothesis-generating reanalysis of already-negative data. Cannot be a verdict basis on its own.