What Russia Adds to the Gut-Brain Evidence Base: 96 Children, Six Months, and No Symptom Claim
- A Kazan group followed 96 children with autism spectrum disorder (aged 3-8, mean age 5.38) and 39 neurotypical controls; 91 children were prescribed a rotational or elimination diet for six months and all received a multi-strain probiotic at 1 billion CFU per sachet plus 990 mg of fructooligosaccharides for 1.5 months. Sixty children completed follow-up.
- No autism symptom outcome was measured at any point. ADOS-2 was administered once, to confirm diagnosis at entry, and never repeated. The endpoints were 16S rRNA taxonomic composition and two fecal markers, zonulin and calprotectin.
- The ASD and control groups did not differ significantly on either fecal marker (p>0.05), and no comparison anywhere in the study found a significant difference in alpha diversity. Fecal zonulin fell in the paired before-after comparison (p=0.018, 42 paired measurements); lower fecal calprotectin was seen in the elimination-diet subgroup (p=0.023).
- Effect sizes in the comparisons by feeding type and food selectivity ranged from 0.015 to 0.034, which the authors describe as small. The intervention was not randomised and had no control arm; the authors state that causal conclusions cannot be drawn and that the contributions of the diet and the probiotic cannot be disentangled.
If you want to know how thin the gut-brain evidence base actually is, do not read the marketing. Read a careful study by people who had every incentive to overclaim and declined. A Kazan team published a six-month gut microbiota cohort in children with autism this May, and the most instructive thing about it is the outcome they did not measure.
What was actually measured, and what was not
The design is real clinical work. Ninety-six children aged 3-8 were diagnosed at the Republican Clinical Psychiatric Hospital named after V.M. Bekhterev in Tatarstan, by ICD-10 criteria with ADOS-2 confirmation, against 39 neurotypical controls. Stool went to the Institute of Fundamental Medicine and Biology at Kazan Federal University for 16S rRNA sequencing of the V3-V4 region on Illumina MiSeq, processed through QIIME2 and DADA2 against SILVA 138. Ethics approval came from Kazan State Medical University. Ninety-one children were prescribed a rotational (74) or elimination (17) diet; five needed neither. All received a multi-strain probiotic at 1 billion CFU per sachet with 990 mg of fructooligosaccharides for 1.5 months. Sixty completed the six-month follow-up.
Now the outcomes. ADOS-2 was used to establish who had autism – and then never administered again. There is no behavioural endpoint in this study, no symptom scale, no repeated developmental measure. What was measured was bacterial relative abundance and two fecal markers of intestinal inflammation and permeability: calprotectin (93 at baseline, 57 at follow-up, 38 controls) and zonulin (86, 44, 22). This is not a criticism of the authors. It is the point. A group with a psychiatric hospital, an ADOS-2-trained cohort and six months of contact declined to claim that diet and probiotics changed autism, because their design could not support it.
And the intermediate markers were themselves unkind to the premise. The ASD and control groups showed no statistically significant difference in fecal zonulin or calprotectin at all (p>0.05) – the "leaky gut in autism" starting assumption was not reproduced in this cohort. Alpha diversity differed in no comparison in the paper: not ASD versus control, not breastfed versus formula-fed, not selective versus non-selective eaters, not rotational versus elimination diet, not baseline versus follow-up. Effect sizes in the comparisons by feeding type and food selectivity ran 0.015 to 0.034. The two positive signals are narrow: zonulin declined in a paired within-child comparison (p=0.018) in a non-randomised design with no control arm, so regression to the mean, maturation and the six-month passage of time are all live explanations; and calprotectin was lower in the elimination-diet subgroup (p=0.023), which the abstract renders as "reduced" but the body describes as lower levels among children who were already on elimination diets at enrollment – a cross-sectional contrast, not a within-child change.
Why this matters for the field
The commercial gut-brain story runs on studies exactly like this one, stripped of exactly these caveats. Take the same cohort, quote the zonulin p-value, mention the autism diagnosis in the same sentence, and you have a psychobiotic marketing claim. The authors did not do this. They wrote that causal conclusions cannot be drawn, that the individual contributions of the diet and the probiotic cannot be separated, that the sample was small, that the sex imbalance leaves residual confounding, and that relative abundance in compositional data does not reflect absolute quantity. That is what a contribution to an evidence base looks like: it makes the base more honest, not larger.
There is a second contribution here, and it is specifically regional. The authors repeatedly hit findings that contradict the published literature – a genus increased in their cohort where Chinese data describe a decrease – and attribute the discrepancy both to cohort differences, China versus Russia, and to the nature of eating behaviour itself. They are right to flag it, and clinicians should read it as a warning. Microbiome reference norms are population-specific, built on diet, geography and early feeding patterns. A Russian or CIS clinician handed a foreign-derived "dysbiosis panel" and a supplement recommendation is being sold an inference that this study suggests does not travel. The practical consequence is narrow and worth stating plainly: this work supports stratifying patients by early nutritional history and eating behaviour, and it does not support telling a parent that a probiotic will change their child's autism. If a supplement rep cites gut-brain research at you, the useful question is the one this study answers honestly – what was the outcome measure, and was anybody randomised?
A group with a psychiatric hospital, an ADOS-2-confirmed cohort and six months of contact declined to claim that diet and probiotics changed autism – because their design could not support it, and that refusal is the contribution.
This is a non-randomised cohort with no control arm for the intervention, a small sample, a pronounced sex imbalance, and no ability to separate the effects of the diet from the probiotic. All conclusions are associational, and the relative abundance of compositional metagenomic data does not reflect absolute bacterial quantity.