Unproven treatments and how to spot them

Autistic children have been offered treatments aimed at removing autism since autism was first described, and the market for them has never gone away. The modern "biomedical" movement grew through the 1990s, built on the idea that autism is caused by something in the body that can be cleared out, from mercury to gut bacteria to a missing vitamin. Its flagship project, Defeat Autism Now!, arose in 1995 from collaboration among members of the Autism Research Institute, which advocated a specific protocol for diagnosis and treatment.¹

Secretin was the movement's first treatment to be tested at scale. In 1998 this gut hormone was proposed as a treatment after parents and doctors reported sudden changes in speech, and it was given intravenously to thousands of children. A Cochrane review of 16 studies found no evidence that single- or multiple-dose intravenous secretin is effective, and concluded that it "should not be recommended or administered as a treatment" for autism.² A separate 2011 review of seven randomized trials reached the same verdict.³

One of these treatments has killed a child. A five-year-old autistic boy died in a physician's office during his third chelation treatment; the office had given him edetate disodium instead of edetate calcium disodium, which stripped the calcium from his blood and stopped his heart.⁴ Chlorine dioxide, sold to parents as "Miracle Mineral Solution" and given as drops or enemas, has been marketed as a cure-all for autism and much else; United States poison centers recorded 53 exposures to the bleach product between 2000 and March 2020, and 13 of those people were hospitalized.⁵ Hyperbaric oxygen chambers, where a child breathes oxygen under pressure, are also sold for autism. A Cochrane review in 2016 found only one trial, in 60 children, which showed no improvement, and side effects were more common with the treatment.⁶

Chelation means giving a drug that grabs heavy metals so the body can pass them out. It is real medicine for genuine metal poisoning. For autism it rests on the claim that mercury from vaccines causes autism, a claim the evidence does not support (Contested causes). That death shows how the treatment can go wrong even in a doctor's office: the office used a form of the drug that has not been recommended for children since 1991, and the result was fatal.⁴ The authors of the case report wrote that "the use of chelation therapy in autistic children has not been validated and can have tragic consequences."⁴

Chlorine dioxide, the industrial disinfectant at the center of the "Miracle Mineral Solution" brand, has been sold to parents as a cure, sometimes as a drink and sometimes as an enema. Of the 53 poison-center exposures to chlorine dioxide products sold as Miracle Mineral Solution, recorded in the United States between 2000 and March 2020, the most common effects were vomiting, nausea, abdominal pain and diarrhea, and about a quarter of the people exposed needed hospital care.⁵ An adult who drank it for months, believing it would protect him from COVID-19, died after it perforated his intestine.⁷

Diets are among the most heavily marketed treatments, and the gluten-free, casein-free diet is the best known. A meta-analysis of six randomized trials covering 143 children found no effect on core autism symptoms, with certainty rated low to very low; the same review found the diet might cause gastrointestinal side effects (a risk ratio of 2.33, on an interval that crosses no effect).⁸ A double-blind challenge trial gave 14 children on the diet snacks containing gluten, casein, both or neither; none of the challenges produced a significant change, and the authors concluded the results do not support general use of the diet.⁹ An umbrella review that pooled 15 meta-analyses and 118 studies found small improvements in overall symptoms, communication, social interaction and hyperactivity, with vitamin D the largest effect, but none in stereotyped behavior, irritability, lethargy or inappropriate speech.¹⁰

The gut microbiome is the newer pitch: the idea that changing a child's gut bacteria changes autism itself. As a group signature, a distinctive autistic microbiome is not established; the differences that studies report are confounded by diet, antibiotic use and other conditions.¹¹

Fecal microbiota transplant is a different case from chelation and bleach. It has real early trials in autistic people. An early trial was open-label and uncontrolled: 18 children aged 7 to 16 with significant gut problems had antibiotics, a bowel cleanse and then transplants, and their gut and behavioral symptoms improved, with a follow-up eight weeks after treatment ended.¹² A two-year follow-up of the same 18 children reported that the gut gains held, but there was still no control group.¹³ The first double-blind, placebo-controlled trial, in 103 children, found no significant difference between the transplant and placebo groups on its main outcome; one secondary measure of socialization favored the transplant group, and there were no serious side effects, though mild ones were reported somewhat more often in the transplant group.¹⁴

A 2023 meta-analysis claimed large benefits from fecal transplants and was quoted widely. It was retracted on 16 July 2026, after the publisher found that three of the studies it pooled were not randomized trials at all.¹⁵

Stem-cell clinics sell infusions of a child's own or donated cells as a way to "reset" the immune system or repair the brain. What has actually been tested is narrow. A phase II trial of cord-blood infusion in 180 children in the United States found no improvement on its primary measure of social communication; a subgroup analysis suggested a possible gain in children without an intellectual disability.¹⁶ A smaller single-center, double-blind, placebo-controlled trial, in 34 children aged 3 to 8 with autism and a related immune problem, gave infusions of donated cord-blood cells and found a small improvement on a parent-rated scale over 13 weeks, with no major side effects; the confidence interval only just excluded no effect.¹⁷ An open-label randomized trial in 54 children aged 3 to 7 in Vietnam gave bone-marrow cells into the spine along with education, and compared the result with education alone; it reported improvements over 12 months, but with no placebo group, so expectation was not controlled.¹⁸ A 2026 review says the stem-cell literature is mostly preclinical (animal and cell models), with a small clinical literature it calls early-stage and a need for larger, more rigorous trials.¹⁹ A 2025 review of medical fraud counts unproven stem-cell injections among the "medical grifts" that cause harm and waste.²⁰ No trial has been replicated by an independent group as of October 2026.

Brain stimulation has drawn a large volume of recent trials; a single 2026 meta-analysis pooled 53 randomized trials.²¹ Three techniques are studied: repetitive transcranial magnetic stimulation (rTMS), which uses a magnetic coil held to the head; transcranial direct current stimulation (tDCS); and transcranial alternating current stimulation (tACS). The trials below are checked against a "sham" comparison, where the device is set up but delivers no active stimulation; most of the wider literature is not.

The largest trial randomized 200 children aged 4 to 10 at three hospitals in China to real or sham stimulation; 167 of them were boys and all had an IQ of at least 50. Parent ratings of social responsiveness improved more in the treatment group at one month, a large effect. The researchers hold a patent on an rTMS device.²² A second sham-controlled trial, in 67 autistic people aged 6 to 30 whose average age was 10, 88 percent of them male and all with cognitive or adaptive delays, reported that 55 percent responded to the real treatment against 29 percent to sham on a clinician-rated measure at week 12, with no follow-up after the treatment stopped; its senior author is a paid consultant to a brain-stimulation company and a co-author works for it.²³ A third, in 47 children, 39 of them boys, found improvement on the study's rating scale at week one and week four after a course of 40-hertz tACS.²⁴

Against those single trials, a meta-analysis of 53 randomized trials found pooled effects that were large on some rating scales and small on others, but rated the certainty of the evidence low to very low, because 85 percent of the trials were open-label; it concluded that rTMS "cannot yet be recommended as routine clinical treatment" for autism.²¹ Clinics already sell brain stimulation for autism off-label, although no device is cleared for autism as of October 2026. And the adult evidence is thin: no published sham-controlled trial has reported core-symptom outcomes in adults as of October 2026, the small adult studies that exist measure brain activity rather than symptoms,²⁵ and a sham-controlled trial in people aged 14 to 40, due to have finished in 2024, had published no results as of October 2026.²⁶

Leucovorin (also called folinic acid, a form of folate) became the most visible autism treatment of 2025. On 22 September 2025 the US Department of Health and Human Services named it "a promising treatment for symptoms of autism", pointing to an analysis of 23 publications from 2009 to 2024 on leucovorin for cerebral folate deficiency, where 85 percent were reported to see some benefit; it describes the evidence for leucovorin in the genetic forms of that condition as case reports and case series.²⁷

The largest leucovorin trial, published in 2024, was a randomized, double-blind trial in 80 children at a hospital in India that reported the supplement improved autism symptoms.²⁸ It was retracted on 29 January 2026, after a statistical review was unable to reproduce its results from the data it provided.²⁸ A 2026 meta-analysis pooling five randomized trials found no clear overall benefit, with a possible effect on language and communication; it has not been peer reviewed.²⁹ On 10 March 2026 the Food and Drug Administration approved an expanded use of leucovorin for cerebral folate deficiency in patients with a confirmed change in the FOLR1 gene, a condition far rarer than autism, and not for autism itself.³⁰ The American College of Medical Toxicology published a 2026 position statement that the current evidence does not support routine use of leucovorin for autism.³¹

The episode shows how quickly demand can move ahead of evidence: in the two weeks after the announcement, internet searches for leucovorin were 379 percent higher than expected, about 1.2 million extra searches, though on a wide interval (95 percent confidence interval 79 to 801 percent).³² The claims themselves, and the evidence on vaccines and acetaminophen that they sit beside, are checked in Contested causes.

Several features recur across the treatments sold for autism.

A promise to remove autism itself is the clearest sign, with the child described as sick or damaged. Chlorine dioxide was sold to parents as a "cure-all", and brain-stimulation clinics advertise treatment for "treatment-resistant autism".^5,33

Official backing that arrives ahead of the evidence is another. The leucovorin announcement of September 2025 rested on small studies, and the largest leucovorin trial was retracted four months after the announcement.^27,28

A device or drug cleared for something else, sold for autism anyway, is a third. Brain-stimulation clinics lean on the fact that their equipment is "FDA-cleared", while no such device is cleared for autism as of October 2026.³⁴

A case built on testimonials rather than trials, with the trials that do exist often small or unblinded, is a fourth. In the rTMS literature, 51.9 percent of trials did not report safety data.²¹

Silence about harm is the last. The clinics that sell these treatments rarely say that chelation has killed a child, that bleach has perforated a bowel, or that stimulation carries its own risks.^4,7

Autistic writers describe these treatments as a warning. Eric Garcia calls the field "a slew of cottage industries offering supposed 'cures'", and notes that they rest on a logic he rejects: "it is better to gamble with children's health by giving them pseudoscientific treatments than to let them be autistic."^c1 He names chlorine dioxide, hyperbaric chambers and chelation among them, and writes that there is evidence they can be harmful.^c1

The market side is described by Meghan Ashburn, Jules Edwards and Morénike Giwa Onaiwu, who write that "autism is a Billions (with a capital B) dollar market" and that "the billions of dollars in the autism market don't actually go to autistic people."^c2 For a parent, they write, the trap is set early: "from the moment we type 'What is autism?' in our search engine, we are being preyed upon", as the same deficit framing that makes a child look broken "makes us vulnerable to companies who offer the hope of improvement or the chance of a cure."^c2

Steve Silberman's history of the era records the biomedical movement's own paperwork, where a flagship parent report admitted that "no well-controlled outcome studies have yet been performed", and closes with a researcher's advice to parents: if a practitioner promises a cure, "run in the other direction."^c3