Deep dive #7 · 9 November 2026

Autism

For health professionals General practice

Autism used to be a diagnosis for young boys. Now the fastest-growing group being diagnosed is adults, many of them women who spent decades masking. Autism is a lifelong difference in how a person communicates and experiences the world, with real strengths, real needs and some very preventable harms. This is what the evidence says across the lifespan.

At a glance

Prevalence: 1 in 31 US 8-year-olds in 2022 (CDC ADDM, 2025); 61.8 million people worldwide, 1 in 127 (GBD 2021); 290,900 Australians, 1.1% (ABS 2022).
Adults: recorded UK diagnoses rose 787% from 1998 to 2018, fastest in adults and females (Russell 2022). In Australia, prevalence is 3.1% under 25 but 0.3% at 25 and over, a gap that likely reflects under-recognition in older adults.
Sex: about 3:1 male to female in high-quality studies, not 4:1. Girls who meet criteria are less likely to be diagnosed (Loomes 2017).
Genetics: heritability about 80% (Bai 2019, 2 million people in 5 countries).
Vaccines: no link. MMR HR 0.93 in 657,461 Danish children (Hviid 2019). The 1998 Wakefield paper was retracted and called fraud.
Health: ADHD in 28% and anxiety in 20% (Lai 2019); premature death OR 2.56 (Hirvikoski 2016).
Support: no medicine treats core autism. Supports target participation, wellbeing and co-occurring conditions.

Biggest take-home: in adults, ask about masking and exhaustion before you settle on "just anxiety".

Language: many autistic people prefer identity-first language ("autistic person"), and we use it here. Some prefer person-first ("person with autism"). Ask, and use what the person prefers. For patients: Autism, Autism in adults and Autism in children.

Definition

DSM-5-TR

A. Social communication and interaction: persistent differences in all 3: social-emotional reciprocity; non-verbal communication; developing, keeping and understanding relationships.
B. Restricted, repetitive patterns: at least 2 of 4: repetitive movements, object use or speech; insistence on sameness or routines; intense, focused interests; sensory hyper- or hyporeactivity.
C. Onset: early developmental period, though features may not show until social demands exceed capacity, or may be masked by learned strategies.
D–E: clinically significant impairment; not better explained by intellectual disability or global developmental delay.
Support levels: 1 (requires support), 2 (substantial), 3 (very substantial), rated separately for each domain. Specifiers: with or without intellectual or language impairment, and associated conditions.

ICD-11

One category (6A02) replaces ICD-10's childhood autism, atypical autism and Asperger syndrome.
Less prescriptive: no fixed symptom count, examples instead, and no operationalised severity. Notes that compensation can hide features, especially with high intelligence.

History

1910s, Eugen Bleuler: coined "autistic" for withdrawal in schizophrenia.
1925, Grunya Sukhareva (Soviet child psychiatrist): described 6 boys with what we'd now call autism, two decades before Kanner. Sula Wolff brought the paper to English readers in 1996.
1943, Leo Kanner: "Autistic disturbances of affective contact": 11 children, 8 boys and 3 girls.
1944, Hans Asperger: published his thesis on "autistic psychopathy". Herwig Czech's 2018 archival study found he publicly legitimised race hygiene and cooperated with Vienna's child "euthanasia" programme.
1981, Lorna Wing: "Asperger's syndrome: a clinical account" placed it with autism in a wider group: the root of today's spectrum.
1998: Wakefield's Lancet paper (12 children) launched the MMR scare. The same year "neurodiversity" reached print (sociologist Judy Singer, journalist Harvey Blume), growing out of online autistic communities.
2010: the Lancet retracted the paper (February); Wakefield lost his UK licence (May). The BMJ called it "an elaborate fraud" in 2011.
2013–2023: DSM-5 merges the subtypes; Australia publishes a national assessment guideline (2018) and its second edition (2023).

Pathophysiology

A highly heritable neurodevelopmental difference. Mechanisms are partly mapped; there is no single cause.

Heritability: median 80.8% across Denmark, Finland, Sweden, Israel and Western Australia; maternal effects 0.4–1.6% (Bai 2019).
Common variants: a GWAS of 18,381 autistic people and 27,969 controls found 5 genome-wide loci, implicating neuronal function and corticogenesis (Grove 2019). Each variant has a tiny effect.
Environment, proven: prenatal valproate: 4.42% autism vs 1.53% background (adjusted HR 2.9; Christensen 2013).
Environment, ruled out: MMR (HR 0.93; Hviid 2019). Prenatal paracetamol: HR 1.05 overall but 0.98 in sibling comparisons (Ahlqvist 2024, 2.48 million children): familial confounding, not causation.
Outcome depends on context: the Supporting Autistic Children Guideline frames barriers to participation as starting in the environment, not the child.

Autism: from genes to support needs, and two claimed causes the evidence rules out (simplified; Bai 2019, Hviid 2019, Ahlqvist 2024). Download this diagram

Epidemiology

Global

GBD 2021: 61.8 million autistic people, age-standardised 788 per 100,000; males 1,065, females 508 per 100,000. Among the top 10 causes of non-fatal health burden under 20.
US (ADDM, 2022 data): 1 in 31 8-year-olds, up from 1 in 36 in 2020; boys more than 3 times as likely; 39.6% also had intellectual disability. Children born in 2018 were 1.7 times as likely to be identified by age 4 as those born in 2014.
Why the rise: in UK primary care data the biggest increases were in adults, females and people without intellectual disability, and the authors concluded "augmented recognition" most likely explains them (Russell 2022).

Australia

ABS 2022: 290,900 autistic people (1.1%), up 41.8% from 2018. Males 1.6%, females 0.7%. 73.0% reported a profound or severe disability.
Work: 42.2% of working-age autistic people employed; unemployment 18.2%, more than double that of people with disability overall (7.5%).

Causes and risk factors

Family history and genetics: the dominant factor (heritability about 80%).
Sex: diagnosed about 3 times more often in males; part of the gap is diagnostic bias (Loomes 2017).
Prenatal valproate: about 3-fold risk (Christensen 2013).
Not risk factors: vaccines or prenatal paracetamol once family factors are accounted for.

Preventable factors

Autism itself isn't something to prevent. This section is about avoidable harms: drug exposures, missed diagnoses and the secondary health effects of unmet needs.

Valproate in pregnancy: avoid where possible, with specialist epilepsy or psychiatric advice and contraception planning.
Periconceptional folic acid: recommended for neural tube defects anyway. In Norway, users had autistic disorder 0.10% vs 0.21% (adjusted OR 0.61); the authors note it can't establish causality (Surén 2013).
Missed and late diagnosis: especially in girls, women and adults who mask.
Secondary harms: untreated anxiety and depression, unemployment and poor access to health care drive much of the excess morbidity and mortality.

Diagnosis essentials

Australian National Guideline (Autism CRC; 2nd edition, 2023)

66 consensus recommendations, NHMRC-approved, for children and adults.
Two pathways: a Lead Practitioner Diagnostic Evaluation (renamed from "single clinician" in 2023; one clinician leads, drawing on others' input) when the picture is clear, or a Consensus Team Diagnostic Evaluation when it isn't.
Holistic: diagnosis sits alongside an assessment of strengths, needs and functioning. Funding eligibility is a separate question from diagnosis.
Language: the update moved to neurodiversity-affirming terminology.

In general practice

Children: act on parent concern and regression; refer for developmental assessment; hearing test.
Adults: clues include lifelong social exhaustion, sensory overwhelm, rigid routines, burnout after life changes, recurrent "treatment-resistant" anxiety or depression, and an autistic child or relative.
Masking: camouflaging combines masking and compensation; motivations are fitting in and connection; consequences include exhaustion and threats to self-perception (Hull 2017, 92 adults). A normal-seeming consultation doesn't exclude autism.
Screen for co-occurring conditions: ADHD 28%, anxiety 20%, sleep-wake disorders 13%, depression 11%, OCD 9% (Lai 2019, pooled clinical and population studies).
Differentials: social anxiety, ADHD, trauma, personality disorders, psychosis, intellectual disability and language disorders. Several can coexist.

Current treatment

"Treatment" here means support for participation, wellbeing and co-occurring conditions, not changing who someone is.

Children

Naturalistic developmental behavioural interventions (NDBI): social communication g=0.36 and autism characteristics g=0.44 on non-caregiver measures. With high-detection-bias studies excluded, only one effect stayed significant (g=0.30). Adverse events were poorly monitored (Project AIM, 252 studies, 13,304 children, 2023).
Behavioural interventions (including ABA-based): g=0.58 for challenging behaviour in RCTs, but mostly caregiver-reported; effects did not survive restriction to blinded measures.
Parent-mediated (PACT): reduced autism symptom severity across about 6 years of follow-up (effect size 0.55); no language difference (Pickles 2016).

Medicines

No medicine treats core autism. Intranasal oxytocin: no difference in social withdrawal (difference −0.2; 290 children; Sikich 2021).
Severe irritability: risperidone response 69% vs 12% on placebo; irritability fell 57% vs 14% (RUPP 2002, n=101). Weight gain and appetite are the main costs; review regularly.
Co-occurring conditions: treat ADHD, anxiety, depression, sleep problems and epilepsy on their own evidence, starting low with medicines.

Adults

Post-diagnosis support, peer connection, workplace and study adjustments, and sensory-friendly care (longer or first appointments, written plans, low-light rooms).
Mental health: treat co-occurring conditions; adapt communication in therapy.
NDIS, in one line: access is based on functional need; Thriving Kids (from October 2026) supports children 8 and under with low to moderate needs without a diagnosis, and NDIS access changes take effect on 1 January 2028.

Emerging treatments and trials

Pre-emptive support in infancy: iBASIS-VIPP in 104 Australian infants (9–14 months) with early signs: 6.7% vs 20.5% met autism criteria at 3 (OR 0.18; NNT 7.2), with better caregiver responsiveness and language (Whitehouse 2021).
Leucovorin: after September 2025 claims, the FDA's March 2026 label update covers only cerebral folate deficiency from FOLR1 gene defects (about 1 in a million), not autism. The largest autism RCT was retracted.
Paracetamol: the 2025 US claims sit against sibling-controlled data showing no association (Ahlqvist 2024).
Policy: Australia's first National Autism Strategy 2025–31 (January 2025) includes a National Roadmap to Improve the Health and Mental Health of Autistic People.

Theoretical frontiers

Hypotheses, not recommendations.

Causes: theories

Hypothesis: thousands of common variants add up. For: heritability 80%, 5 GWAS loci (Grove 2019). Against: each locus explains very little; samples mostly European. Test: larger, ancestrally diverse GWAS combined with sequencing.
Hypothesis: the sex ratio is partly ascertainment. For: 3.25:1 when populations are screened vs 4.56:1 for existing diagnoses (Loomes 2017). Against: a male excess persists. Test: population screening with instruments validated for camouflaging.
Hypothesis: prevalence growth is recognition, not incidence. For: biggest UK rises in adults and females (Russell 2022). Against: a small true rise can't be excluded. Test: repeated population surveys with identical methods.
Hypothesis: prenatal folate pathways matter. For: adjusted OR 0.61 (Surén 2013). Against: observational, residual confounding. Test: sibling comparisons and Mendelian randomisation.
Hypothesis: prenatal paracetamol is causal. For: small unadjusted associations (HR 1.05). Against: sibling HR 0.98, no dose response (Ahlqvist 2024). Test: triangulation across sibling, negative-control and genetic designs.

Prevention: theoretical options

Hypothesis: pre-emptive support improves development. For: iBASIS-VIPP (OR 0.18). Against: 104 infants, one-tailed tests; "fewer diagnoses" is not an outcome autistic people prioritise. Test: a larger multisite RCT with quality-of-life and wellbeing outcomes.
Hypothesis: avoiding valproate cuts preventable cases. For: adjusted HR 2.9. Against: seizure control matters too. Test: registry monitoring after prescribing restrictions.
Hypothesis: earlier diagnosis in girls prevents mental ill-health. For: camouflaging brings exhaustion (Hull 2017). Against: qualitative evidence. Test: a cohort comparing early vs late diagnosis on mental health.
Hypothesis: employment support prevents secondary harm. For: 18.2% unemployment (ABS 2022). Against: no Australian RCT found. Test: an RCT of supported employment for autistic adults.

Treatment: theoretical options

Hypothesis: NDBIs change long-term outcomes. For: blinded social communication effects (Project AIM). Against: one effect survives strict bias limits. Test: active-comparator RCTs with outcomes chosen with autistic people.
Hypothesis: leucovorin helps a folate-antibody subgroup. For: small studies cited by the FDA. Against: the largest RCT was retracted. Test: a new multisite placebo-controlled RCT.
Hypothesis: oxytocin improves social function. For: early small trials. Against: SOARS-B null (Sikich 2021). Test: none justified without new mechanism data.
Hypothesis: reducing camouflaging demands improves wellbeing. For: Hull 2017. Against: no intervention trial. Test: a workplace-adjustment RCT with burnout outcomes.
Hypothesis: better-tolerated options for severe irritability exist. For: risperidone works but causes weight gain. Against: few head-to-head trials. Test: comparative effectiveness trials with metabolic outcomes.

Gaps and open questions

Adult services: recorded diagnoses are rising fastest in adults (Russell 2022). How should adult assessment and post-diagnosis support be organised?
Ageing: we found few data on autistic people in later life.
Mortality: why is premature death more than doubled, and what prevents it?
Equity: we found few data on First Nations autistic Australians or culturally diverse communities.
Intervention harms: adverse events are rarely monitored in trials (Project AIM).

Prognosis and quality of life

Mortality: 2.60% of autistic people died during follow-up vs 0.91% of matched controls (OR 2.56), across almost all causes (Hirvikoski 2016, Sweden, 27,122 people).
Work: 18.2% unemployment in Australia (ABS 2022).
Lifespan: autism is lifelong; skills, support needs and wellbeing change with environment, support and co-occurring conditions.
Early support: PACT symptom benefits persisted about 6 years after the trial.

Weird trivia

Sukhareva first: a Soviet woman psychiatrist described autism in 1925, 18 years before Kanner.
Bleuler's word: "autistic" first described schizophrenia.
Kanner's 11 included 3 girls.
Temple Grandin's hug machine was inspired by cattle calming in a squeeze chute on her aunt's ranch.
Autistic Pride Day (18 June, since 2005) was set on the birthday of the founding group's youngest member.
"Neurodiversity" is often credited to one person, but a 2024 Autism letter argues nobody knows who coined it.

Guidelines

Key trials

RUPP 2002 (NEJM): risperidone for severe irritability, 69% vs 12% response.
PACT 2016 (Lancet): parent-mediated therapy, long-term symptom reduction.
Hviid 2019 (Ann Intern Med): MMR and autism, HR 0.93.
Sikich 2021 (NEJM): oxytocin, null.
Whitehouse 2021 (JAMA Pediatr): pre-emptive iBASIS-VIPP, OR 0.18.
Project AIM 2023 (BMJ): 252 intervention studies, modest effects, poor harm reporting.
Ahlqvist 2024 (JAMA): prenatal paracetamol, null in sibling analysis.

Sources

Information only. Follow the national guideline, local referral pathways and product information. Medicines mentioned treat co-occurring conditions or severe irritability, not autism itself.

Updated 9 November 2026. General education for health professionals; not a substitute for local protocols.

maladywise.com/deep-dives/autism/