Raising Each Other
Menu
Milk and gluten: what was actually found, and what is being sold
Learn · milk, gluten, and what was actually checked

Milk and gluten: what was actually found, and what is being sold

Our son has been off dairy and gluten from the start. I am not neutral and I won’t pretend to be. But when I sat down to check what really stands behind that decision, it turned out three things are true at once, and usually you only ever hear one of them.

Constipation, reflux, painfour times more common than in peersThe child cannot say itno words for “my stomach hurts”It shows as behaviourscreaming, head, sleep, refusing foodsame picture, two readingsGoes to a behaviour programmethe pain stays, the behaviour returnsGoes to a gastroenterologistoften sorted in a month
Figure 1. The best-evidenced thing on this page is not the diet but this: pain a child cannot report comes out as behaviour. How it is read decides who the child ends up seeing.

In three sentences

Gut trouble in autistic children is real and common. It is the best-evidenced thing in this whole story and the least talked about. The reason given for years as to why milk and gluten harm the brain (so-called “opioid peptides”) was looked for in children’s urine and was not found. And yet the diet still shows a small effect in trials, there is one narrow mechanism with direct evidence behind it, and there is a cost that nobody selling you the diet will mention.

The part that is rarely said out loud

Start with the most certain thing, because it is the thing most often skipped.

A meta-analysis published in Pediatrics in 2014 found that autistic children are more than four times as likely to have general gastrointestinal complaints, around three times as likely to have constipation and diarrhoea, and about twice as likely to have abdominal pain, compared with their peers.

That is not a theory about what causes autism. It is a finding about what daily life is like for autistic children.

And it carries a consequence that changes the ordinary day: a child who cannot say that something hurts shows the pain as behaviour. Screaming, head-banging, refusing food, waking at night, “regression”: any of it can be constipation now in its third week. Read as behaviour, it goes to a behaviour programme. Read as pain, it goes to a gastroenterologist and is often sorted in a month.

Grade: established. This is the only thing on this page that needs no leap of faith at all.

The story that sold the diet

Why would milk and gluten have anything to do with the brain? The explanation given is called the opioid excess theory: casein from milk and gluten from wheat are said to break down in some children into opiate-like peptides (casomorphin, gluteomorphin); those peptides cross the gut wall, reach the brain through the blood, and the child seems “absent, as if drugged”.

The story is seductive because it explains everything at once, which is why it has survived thirty years.

Then somebody went looking for those peptides. A study published in 2008 in Archives of Disease in Childhood has a title that leaves no room for doubt: “Absence of urinary opioid peptides in children with autism.” They measured with sensitive methods and did not find them.

Grade: tested and did not hold. Not “we lack evidence”: it was looked for and was not there.

This matters for one reason that has nothing to do with diet: when somebody explains a mechanism to you with great confidence, the question is not whether it sounds logical, but whether anyone ever went and checked.

THE CLAIMWHAT WAS CHECKEDWHAT CAME BACKThe gut is more oftenin troublemeta-analysis, Pediatrics 2014true: 4× more likelyMilk keeps folate receptorantibodies up12 children off milk, 2008titre falls, narrow groupThe diet helps symptoms8 trials, 297 childrensmall effect, poorly blindedOpioid peptides in urinelooked for, sensitive methodnot found
Figure 2. The same four claims you will hear in one conversation, sorted by what happened when somebody checked them.

So what do the trials themselves show

The interesting part is that the diet did not disappear along with its explanation.

A 2021 meta-analysis (Nutrition Reviews) pooled eight randomised trials and 297 children from Iran, Norway, Poland, Spain, the USA and Denmark. Trials ran from six weeks to a year. What it found:

  • stereotyped behaviours: improvement, SMD −0.41 (95% CI −0.68 to −0.15)
  • cognition: improvement, SMD −0.46 (95% CI −0.91 to −0.01)
  • communication and social difficulties: no difference at all

“SMD” is a way of comparing trials that used different scales; 0.4 is a small-to-moderate effect.

Now the limitations, and they are serious. The trials are small. They measured with different instruments. Most ran under four months. And the one that matters most: only one of the eight stated that the children were blinded, that is, that they did not know whether they were eating ordinary or gluten-free food. In all the others both the parent doing the rating and the child knew what was going on, and a parent who has cooked separately for three months is not a neutral judge. Not because they lie, but because nobody is.

The authors themselves conclude the benefits are “promising” and call for larger, double-blind trials.

Also: the Cochrane review of these same diets was withdrawn in 2019. A withdrawal does not mean the diet was disproved; it means that review no longer stands as a current source.

Grade: weak but real evidence. Not nothing, and not a confirmation either.

One narrow mechanism that did hold up

There is one place where the link between milk and the brain has direct evidence. But it is a far narrower claim than “milk causes autism”, and it is worth getting exactly right.

The folate-binding protein in cow’s milk (FBP) resembles the human folate receptor closely enough that the immune system can confuse the two. Antibodies made against the milk protein then block the door folate uses to get into the brain. We write about this in detail on the leucovorin page.

When Ramaekers and colleagues took milk out for twelve children with cerebral folate deficiency in 2008, the titre of those antibodies fell over the following months.

Notice what that claim is and is not. It is: in children who have those antibodies, milk keeps them up. It is not: milk causes autism, nor that every autistic child has those antibodies.

And one more caveat we have added since we first wrote about folate: the test for those antibodies is not as reliable as it is made to sound. In March 2026 the Child Neurology Society said testing is “not recommended as a basis for clinical decisions”, and the American Academy of Pediatrics said that no such test is FDA-approved.

Grade: direct evidence, small study, narrow application.

The cost nobody mentions

This is the part you will not find on the pages selling you gluten-free flour.

An NIH study followed 75 boys aged 4–8 with autism. Their bones were growing longer but not thickening normally; by seven and eight they were noticeably thinner than their peers’. Among the boys on a dairy-free diet (nine of them), bones were about 20% thinner, against about 10% in the boys with no dietary restriction.

Nine children is few and the authors themselves ask for larger studies. But the direction makes sense and is no surprise: milk is the main source of calcium and vitamin D for small children, autistic children often already eat very narrowly, and they are indoors more and move less.

The conclusion is not “don’t take milk out”. It is: if you take milk out, calcium and vitamin D become your responsibility: a conversation with your paediatrician or a dietitian, not something that sorts itself out.

If you try it, try it as a trial

The biggest harm from this diet is not nutritional. It is this: a family takes out milk and gluten, nothing obvious happens, and the diet stays forever, because nobody ever formally ended it. Years later the child eats narrowly, the family struggles at birthdays and restaurants, and nobody can say whether any of it ever helped.

So if you try:

  1. Change one thing. Milk first or gluten first, not both. Change two things and you cannot know which one worked.
  2. Set an end date before you start. Six to eight weeks is enough to see an effect on behaviour and stools. Write the end date down first.
  3. Decide in advance what you are watching. Two or three concrete things: stools, night waking, number of meltdowns. Not “general impression”: impressions bend towards the effort you put in.
  4. Write it down every day, briefly. Memory after six weeks is not data. One bad week erases three good ones.
  5. Put the food back at the end. This is the step everyone skips and the only one that actually answers the question. If nothing changes when it goes back, the diet was not working, and you have handed your family back half its life.

None of this is our invention. It is what an honest elimination trial looks like, and how gastroenterologists run them.

What to ask your doctor

  • Has my child actually been examined for constipation? Not “do they go”, but examined. Constipation in children can present as diarrhoea.
  • Does the behaviour that worries me have a pattern around meals, nights or stools?
  • If we take milk out, how do we cover calcium and vitamin D, and should anything be measured?
  • Does it make sense to test for coeliac disease before we take gluten out? (The test does not work once a child is already gluten-free.)
  • If the trial shows nothing in eight weeks, do we agree that we put the food back?

Our case

Maxi has been off milk and gluten from the start, alongside folinic acid. Which means we do not know which did what. We started everything at once, like most families, because when it is your child you are not thinking about experimental design.

If I could do it again I would introduce one thing at a time and write it down. Not because the outcome would differ, but because today I would know what to tell another parent who asks.

He sleeps well and always has, so on sleep, this page has nothing to offer from our own experience.

Evidence grades

  • Established: several independent studies agree and the finding has been repeated.
  • Direct evidence, narrow application: a real finding, but in a small group or a narrow condition.
  • Weak evidence: studies exist, but small, short or poorly blinded.
  • Tested and did not hold: the claim was tested and not confirmed.
  • Marketing: somebody sells it and there is no study behind it.

Sources

Verified: 17 September 2026.