In Gut-Autism Data, Constipation Explained More Than Diagnosis

What the research found

A gut microbiome study measures the bacterial DNA in one sample of stool. Sequencing establishes which bacterial groups are present, and in what proportion to one another, at a single moment in a single child. The procedure records nothing about the brain or about behaviour, and nothing about which came first.

Kiana West and colleagues at Second Genome, a California biotechnology company, working with researchers at Stanford University and Oregon State University, gathered thirteen such datasets covering ten cohorts and 690 children between two and eighteen years old, drawn from the United States, Canada, China, Italy and India. Rather than pool the conclusions other authors had already published, they retrieved the raw sequencing files and reprocessed every one through the same pipeline. The analysis appeared in Scientific Reports in 2022.

Going back to raw data was necessary because the published literature does not agree with itself. Surveys of the autistic gut have run for more than a decade, and the bacteria they nominate change from paper to paper. Alongside the sequences, the team assembled 52 pieces of information that the original studies had recorded about the children, and tested every pairwise combination, 580 tests in all. Age, sex and bowel function emerged as the variables that mattered.

Some bacterial signals did hold across cohorts. Two sequence variants of Ruminiclostridium_E siraeum were less abundant in autistic children in twelve of the thirteen datasets, and Barnesiella intestinihominis in eleven. The order Burkholderiales was more abundant in autistic children in nine of ten, at an effect size the authors themselves describe as small, a log2 fold change of 0.59.

Between 40 and 70 percent of autistic children have gastrointestinal problems, and stool consistency ranks among the strongest known influences on the bacteria a sample contains. When the team removed the children with bowel dysfunction and repeated the comparison, many of the autistic-versus-neurotypical differences disappeared, which the authors take as a sign that those differences had followed constipation and diarrhoea more closely than the diagnosis. Comparing autistic children with constipation against autistic children with diarrhoea separated far more bacterial groups than comparing autistic children against neurotypical ones.

Nineteen bacterial groups survived into the normal-bowel subset at statistical significance, most of them at small effect sizes. The four largest rested on two to four datasets each, and two were species so rare that they appeared in 7 percent and 3 percent of the neurotypical children and in none of the autistic ones. In the two datasets that reported age, sex and bowel function together, adjusting for all three changed the answer outright: of four strains significant in one dataset before adjustment, one survived it, and seventeen new ones appeared.

What this means for you and your child

Nothing in this work was a trial. No child received a probiotic, a diet or a transplant, and no outcome was measured. A design of this kind can register association and no more, which leaves the direction of every association it finds entirely open.

The authors list diet and medication among the confounders their own analysis could not address, because too few of the original studies had recorded them, and they note that extreme food selectivity is common in autistic children. They cite work published in Cell in 2021 by Yap and colleagues reporting that autism-related dietary preferences mediate the association between autism and the gut microbiome. A narrow diet, adopted for sensory reasons, would be expected to produce a distinctive microbiome as its consequence.

The composition of the cohorts constrains the claims further. Boys made up between 70 and 89.5 percent of every dataset, and when the autistic-versus-neurotypical comparison was run separately within each sex, no bacterial group appeared on both lists. A microbiome finding drawn from this literature is, in practice, a finding about boys.

What the paper supplies to a parent is a set of questions. Its authors state plainly that simple case-control comparisons are unsuitable for this field, and that studies should at minimum score samples on the Bristol Stool Scale, a standard seven-point description of stool form, and report the result. Only two of the ten published cohorts had recorded bowel function in both the autistic and the neurotypical group.

Nine of the ten cohorts contained fewer than 100 children and five contained fewer than 50, a restriction on statistical power the authors identify as severe. Pooling was designed to get past exactly that, and it did narrow the list of credible bacteria considerably. The result is a short list of candidates for further study, offered by the authors as a starting point for identifying patient subgroups.

Every product marketed to parents on the strength of a gut bacterium draws on this literature. Two questions do most of the work in assessing one: how many independent cohorts showed that bacterium moving in the same direction, and whether anyone recorded whether the children were constipated. For the bacteria West and colleagues could examine, the first answer ranged from twelve datasets down to two, and across most of the field the second answer is no.