Raising Each Other
Menu
What actually works, and what I wish I had done first
Learn · where to start, in order

What actually works, and what I wish I had done first

Our first year, honestly

My son was born in March 2023. By the time he was two, he had almost no words: one, on a good day, and often none.

What I reached for first was the biology. I read about folate and antibodies and the gut. We did a course of nystatin. We started low-dose folinic acid, on prescription, and we still hold it. I never did the antibody test that everyone in the forums talks about. I spent my nights reading papers instead of sleeping, and I built a whole app to keep track of it all.

I don’t regret the reading. But here is what nobody sat me down and said, and what I would tell myself two years ago:

The things with the best evidence behind them were the least exciting ones, and most of them were free. They were not a supplement or a test. They were a communication device, the way I play with my own child, and finding out whether he was in pain.

This page is the one I needed then. It is ordered, first thing first, and every claim says how strong the evidence for it actually is.

In three sentences

If your child is newly diagnosed and you have limited time, money and energy, the evidence points at a short list: give your child a way to communicate right now, learn a way of playing and talking that you can do every day, and hunt down pain, constipation, poor sleep and anxiety.

None of it is dramatic and none of it is sold in a bottle.

Everything else (including the biology I write about elsewhere on this site) sits after these, not instead of them.

THIS FIRST1A way to talk todayAAC: pictures, a board, a tablet that speaksmoderate, consistent2Play and talk, taught to youparent-mediated programme (PACT)best long-term3Pain, sleep, constipation, anxietywhat gets seen as “behaviour”strong that it's common4A seizure conversation on recordepilepsy in about one in tenstrong (prevalence)5Then a programmeESDM, JASPER, PRT, if you can get onemoderate for languageEverything else comes after these, not instead of them.
Figure 1. The order. Ranked by how much evidence stands behind each step, not by how loudly it is sold.

1. Give them a way to talk today. Do not wait for speech

If your child has few or no words, the single most useful thing you can do this month is start AAC: augmentative and alternative communication. Pictures, a board, or an app on a tablet that speaks when you press a picture.

Almost every parent is told, usually kindly, some version of: if you give him a device, he will stop trying to talk.

This is not true, and we have known it is not true for a long time.

In a randomised trial of 61 minimally verbal autistic children, the group that began with a speech-generating device did better on spontaneous spoken words, on new words and on comments than the group working on spoken words alone. Not equal: better. A review of 11 separate studies found not one case where speech went down.

WHAT PARENTS ARE TOLD“Give him a device and he’llstop trying to talk.”WHAT THE TRIAL FOUND61 children, randomisedStarted WITH a deviceSpoken words onlyThe device-first group did better:more spontaneous words, more new words, more comments
Figure 2. It isn’t true, and hasn’t been for a long time. The figure shows the direction of the finding, not its size. We don’t print numbers we haven’t checked in the original paper.

Think of it like a child learning to walk who is given a push-along walker. The walker does not teach the legs to stop trying. It gets the child across the room today, and the walking comes anyway.

Evidence: moderate, and consistent. This is one of the few places in autism where the research and the folk wisdom point in opposite directions, and the research is clear.

What this means on Monday: ask for a speech and language assessment, and ask specifically about AAC. Don’t accept “let’s wait and see if speech comes first.”

2. The therapy that works through you

The intervention with the strongest long-term evidence in this entire field is not a clinic. It is a way of playing and communicating with your own child, taught to the parent.

PACT, a parent-mediated programme, is the only early autism intervention with a randomised trial followed up six years later that still showed a difference. Autism severity was reduced (effect size 0.55), and the proportion of children in the severe category was 46% in the PACT group against 63% in usual care.

Six years. No waiting list. No monthly fee.

Evidence: the best long-term evidence available for any early intervention.

What this means on Monday: ask your service about parent-mediated or parent-coaching programmes. If none exists near you, the underlying ideas are learnable: following your child’s lead, leaving space for them to start the interaction, responding to what they are already interested in rather than redirecting them.

3. Find the pain

This is the most neglected item on the list, and often the fastest visible change.

Autistic children have far more gut trouble than other children: general GI symptoms are about four and a half times more common, constipation nearly four times, abdominal pain two and a half times. Nearly half have GI symptoms at any given time.

Here is the part that matters. A child who cannot tell you it hurts will show you instead: as irritability, as self-injury, as sleep that falls apart, as “behaviour” that everyone starts trying to manage behaviourally. The official paediatric guidance says this plainly: investigate and treat gut problems in an autistic child exactly as you would in any other child.

Sleep is the same story. Roughly one in eight has a diagnosed sleep-wake problem and far more have trouble in practice. Behavioural approaches come first and do work: about 18 minutes faster to sleep, about 24 minutes more of it. Melatonin comes second, and the important detail almost nobody mentions: the effect is a curve, not a line. Around 5.7 mg is where it works best. More is not better; more can be worse.

And anxiety at 20%, ADHD at 28%. Both are treatable, both making everything else look worse when they are missed.

Evidence: strong that these are common; moderate that treating them helps overall.

What this means on Monday: if something has got worse and nobody can say why, check for constipation, toothache, ear infection and reflux before assuming it is autism.

4. Put a seizure conversation on the record

About one in ten autistic people develops epilepsy: around 7% in childhood, rising to roughly 19% in adulthood. It is not urgent today, and it is worth your paediatrician knowing you know.

What this means on Monday: ask what a seizure would look like in your child, and what you should do if you see one. Then stop worrying about it.

5. Then a programme, if you can get one

Early intensive and naturalistic developmental interventions (ESDM, JASPER, PRT and the rest) do help, mostly with communication. They are worth having. They are also, honestly, oversold.

The Cochrane review rated the certainty of the evidence low to very low. A careful re-analysis found that reported effects were inflated by how outcomes were measured. Most of the social-communication measures were taken by people who knew which group the child was in, and many measured almost exactly the thing that had just been taught rather than whether it transferred to real life. No branded model has been shown to beat any other.

Evidence: moderate for language, low for everything else.

What this means on Monday: if you can access one, take it. Choose it on whether your child is content in the room and whether the goals are your child’s goals, not on the brand name or the number of hours.

On ABA specifically

Many autistic adults object to ABA, and the objections deserve to be taken seriously rather than waved away: suppressing stimming that was doing no harm, the historical goal of making a child “indistinguishable from peers”, compliance trained without the child’s assent, and autistic people being left out of setting the goals.

The honest position on harm is that it is under-studied, not disproven. Trials in this field almost never record adverse events at all. Modern practice has moved towards assent-based and naturalistic methods, unevenly.

You do not need to pick a side to make a good decision. Ask: is my child happy in that room, and are these goals for my child’s benefit or for other people’s comfort?

What the guidelines actually say

Both major guideline bodies land in the same place. The first-line recommendation for core autism is a play-based social-communication intervention involving parents and carers.

NICE also publishes a list of things not to use for core autism, which is worth knowing before someone offers you one: antipsychotics, antidepressants, anticonvulsants, gluten-free/casein-free exclusion diets, secretin, chelation, hyperbaric oxygen, auditory integration training, neurofeedback.

About “recovery”

Somewhere between 3% and 25% of autistic children eventually no longer meet the diagnostic criteria. That range is enormous because the studies differ, and the children who do are systematically the ones who started with milder difficulties, stronger cognitive scores and an earlier diagnosis.

Even among them, brain imaging still shows language being processed differently, and rates of ADHD and phobias stay high.

No intervention has been shown to reliably produce this, which means nobody can sell it to you. Anyone who implies they can is telling you something about themselves.

Your child’s life can be good either way. That is not a consolation prize; it is the actual goal.

What to ask at the next appointment

  • Can we start AAC now, rather than waiting to see if speech comes?
  • Is there a parent-coaching or parent-mediated programme I can join?
  • Can we rule out constipation, reflux, dental pain and ear infection as a cause of the behaviour that changed?
  • What would a seizure look like in my child, and what should I do?
  • Which of the things I am doing would you stop, if it were your child?

How to read the evidence grades on this site

Established: replicated, and accepted by the bodies that write guidelines. Plausible mechanism, unproven: the biology makes sense and the trials are small, few, or not yet done. Interesting. Not a promise. Marketing: the loudest claims, usually made by whoever sells the test, the supplement or the clinic visit.

We say which one applies, every time. If we cannot source a number, we leave it out.

Sources

Verified 13 September 2026. Not medical advice. I am a parent, not a doctor. Nothing here is a dose or a prescription, and your child’s paediatrician knows your child.