Probiotics and autistic children: what the trials actually found
If you have read that probiotics help autistic children, you have read a summary of a literature that does not say that. The trials exist, several are well run, and the honest picture they paint is narrower and more interesting than the headlines.
Here is what has actually been tested.
The most cited trial was negative
In 2019 a team in Taiwan ran a randomised, double-blind, placebo-controlled trial of Lactobacillus plantarum PS128 in 80 autistic boys aged 7 to 15, over four weeks. It is the study that put this probiotic on the map.
Its three pre-specified primary outcomes were the Aberrant Behavior Checklist, the Social Responsiveness Scale and the Child Behavior Checklist. None of them differed between the probiotic and the placebo. Clinical global impression scored 3.64 against 3.66, a p value of 0.94.
One finding survived: in an age-stratified subgroup of the younger children, aged 7 to 12, a secondary measure of opposition and defiance improved against placebo. The authors themselves noted a high placebo effect in the trial. (Liu et al., Nutrients, 2019, PMID 30979038.)
That is a real result worth knowing about. It is not the same claim as “probiotics improve autism symptoms”.
The largest and longest trial
An Italian team randomised 85 autistic preschoolers, mean age 4.2, to a multi-strain probiotic or placebo for six months, the longest such trial run. The primary outcome was the calibrated severity score of the ADOS, the standard observational assessment.
No difference between groups. An exploratory subgroup of the 55 children without gastrointestinal symptoms showed a larger fall in social-affect severity on the probiotic, and the 30 children with gastrointestinal symptoms showed gains in gut symptoms and adaptive functioning. (Santocchi et al., Frontiers in Psychiatry, 2020, PMID 33101079.)
The faecal transplant story, and its correction
The most dramatic results in this field came from an open-label study at Arizona State University: 18 autistic children aged 7 to 16 given a course of antibiotics, a bowel cleanse and then a microbiota transfer. Gastrointestinal symptoms fell 82 per cent and autism severity scores fell 22 per cent. At the two-year follow-up, autism severity was 47 per cent below baseline and 44 per cent of the children scored below the diagnostic cut-off. (Kang et al., Microbiome, 2017, PMID 28122648; Scientific Reports, 2019, PMID 30967657.)
Those are extraordinary numbers, and the authors are careful about them. The study had no placebo group, so as they write, every assessment is subject to placebo effect. Twelve of the eighteen families changed medications, diets or supplements during the follow-up period.
In 2024 a Chinese team ran the blinded version: 103 children, randomised, double-blind, placebo-controlled oral faecal transplant. The primary outcome, the Social Responsiveness Scale, showed a difference of 1.49 points in favour of the transplant with a confidence interval crossing zero and a p value of 0.362. One secondary measure out of roughly twenty, a socialisation domain, reached significance. (Wan et al., Clinical and Translational Medicine, 2024, PMID 39187939.)
Better design, smaller effect. That pattern repeats through this whole literature.
The trial that separated the two questions
A 2025 Spanish trial gave L. plantarum and L. brevis to 42 autistic children and 38 children with ADHD, aged 5 to 16, for twelve weeks. On intention to treat there were no significant between-group differences on any parent-reported or neuropsychological outcome.
Its companion paper analysed the same children's gut bacteria and found the microbiome had genuinely shifted in the autistic children who took the probiotic: richness and diversity both rose. (Rojo-Marticella et al., Research on Child and Adolescent Psychopathology, 2025, PMID 39798036; Novau-Ferré et al., Research in Developmental Disabilities, 2025, PMID 40184961.)
The bacteria moved. The symptoms did not. Those are two different things and this pair of papers is the cleanest demonstration of it anywhere.
What the pooled analyses say
They disagree, and the disagreement is informative.
Reviews restricted to parallel-group randomised trials find little or nothing. One 2025 analysis of 15 trials found a pooled effect on autism symptoms of −0.05 with a confidence interval crossing zero and no heterogeneity at all. Another found no significant improvement in autism severity, gut symptoms or comorbid psychopathology. Reviews that also include open-label studies report moderate benefits.
A careful 2024 review of 12 randomised trials in 630 children under 20 found a small effect on core symptoms of −0.21, and stated plainly that the data do not provide high-quality evidence, with high risk of bias in 9 of the 12 trials. (Kotowska et al., Clinical Nutrition ESPEN, 2024, PMID 39173907.)
Where it is more solid
Gastrointestinal symptoms. A meta-analysis of 6 randomised trials in 302 children found a meaningful improvement in a gut symptom index while finding no difference on behaviour, social responsiveness, development or global impression. (Zeng et al., Italian Journal of Pediatrics, 2024, PMID 38902804.)
Constipation, diarrhoea and abdominal discomfort are common and genuinely distressing in autistic children, and they are the outcomes probiotics move most consistently.
What to take from it
Probiotics change gut bacteria. They help gut symptoms. Across every trial that pre-specified a core symptom outcome and tested it properly between groups, they have not changed core autism symptoms or ADHD symptoms.
The positive results in this field are overwhelmingly within-group comparisons, subgroups identified after the fact, or one outcome among twenty. That is not a reason to dismiss the research. It is a reason to read the next headline slowly.