Research Round-Up: March 2026 - From Molecular Mechanisms to the Continuing Wait for Assessment
This past fortnight has brought several significant updates in autism research. We have seen new findings on a molecular pathway that might link different genetic variants, a potential biomarker that could help identify autistic children earlier, and a large meta-analysis looking at mathematical ability. Meanwhile, on the clinical side, the NHS assessment crisis continues to impact families as the government’s National Autism Strategy enters its final year.
My goal is to share this science in a way that is practical and relevant for parents, professionals, and autistic adults trying to navigate a complex system.
A Shared Molecular Pathway Across Forms of Autism
A recent collaboration between the Hebrew University of Jerusalem and Yale, published in Molecular Psychiatry (Bhabi et al., 2026), has identified a chain reaction inside nerve cells that appears to go wrong across multiple genetic forms of autism.
They found that a gas called nitric oxide, which is produced naturally in the brain, can chemically tag a protective protein called TSC2 when present in excess. This tagging destroys TSC2. Because TSC2 normally keeps a cellular growth pathway called mTOR in check, its loss leads to overactive mTOR, which then disrupts normal neuronal growth and connectivity.
The researchers demonstrated this mechanism in two different genetic mouse models of autism (Shank3 and Cntnap2). More importantly, they validated it in clinical samples from children with ASD, including those with identified SHANK3 mutations and those with no specific genetic cause found yet. Nitric oxide inhibitors successfully reversed the effect in the laboratory.
You can read my previous post exploring this pathway in more depth here: A Shared Druggable Switch: How Nitric Oxide, TSC2, and mTOR Connect Multiple Forms of Autism.
What does this mean for families? This research is still at the laboratory stage and does not offer an immediate treatment. However, it points toward a shared biological mechanism common across different genetic subtypes. If these findings are replicated in further studies, this could eventually lead to therapies that benefit a much broader group of autistic individuals rather than just those with specific gene mutations.
Reference: Bhabi, S. et al. (2026). Nitric oxide-mediated S-nitrosylation of TSC2 drives mTOR dysregulation across Shank3 and Cntnap2 models of autism spectrum disorder. Molecular Psychiatry. Nature Publishing Group.
Vasopressin: A Replicating Biomarker
Finding reliable biological markers for autism has been a long-standing challenge. This is why a new publication from Karen Parker’s group at Stanford in Autism Research (2026) is worth noting.
Oztan, Kong, Jiang, Crain and their colleagues report that cerebrospinal fluid (CSF) levels of vasopressin, a neuropeptide involved in social behaviour, are significantly lower in autistic children compared to non-autistic peers. They also found lower vasopressin levels in newborns who were later diagnosed with autism, which suggests it might have some predictive value from birth. These findings have now been replicated across multiple independent groups.
Low vasopressin specifically correlates with differences in social communication, but not with repetitive behaviours. This makes sense biologically and raises the possibility that vasopressin levels could eventually help us identify children who might benefit most from targeted interventions.
A word of caution. Checking vasopressin in CSF requires a lumbar puncture, which is too invasive for routine screening. For this to become a useful clinical tool, we still need to find out if it can be reliably measured in blood or through other less invasive methods. But as a piece of the biological puzzle, it is a very robust finding.
A Closer Look: Does this move the clinical needle?
The short answer is: this paper doesn’t change the clinical picture, and it was never really going to, given the design. But it’s worth being precise about why.
What the paper actually adds: The Parker group has been building the CSF vasopressin story for nearly a decade now. This new paper attempts two things: first, to replicate the CSF AVP–social symptom correlation in a methodologically different sample (postmortem, older individuals, LC-MS/MS rather than ELISA, ADI-R rather than ADOS, informant-reported rather than clinician-observed); and second, to show that CSF AVP actually reflects hypothalamic AVP gene expression, which blood AVP does not.
The replication argument is the stronger claim conceptually. They emphasise that the association held across age, collection time, quantification platform, clinical instrument, rater type, and symptom domain. That’s a reasonable point about convergent validity. If you keep getting the same directional finding despite changing multiple methodological parameters, that’s worth noting.
Why it doesn’t move the clinical needle: The problem is that convergent validity across methods doesn’t compensate for a fundamental issue with this study: statistical power. With an N=6 in Study 1, you have essentially no ability to distinguish a true moderate effect from noise. The reported correlation of -0.85 is almost certainly inflated. A correlation that large, if real, would make CSF AVP one of the strongest biomarkers in all of psychiatry. That seems implausible for a complex, heterogeneous condition like autism.
More importantly, from a clinical biomarker perspective, you need to demonstrate several things this paper doesn’t and can’t address: sensitivity and specificity in a diagnostically relevant population, test-retest reliability, performance against existing clinical tools, and — critically — whether measuring CSF AVP changes any clinical decision you’d actually make. The paper is framed around eventual clinical utility, but as the authors themselves acknowledge, CSF surveillance requires invasive sampling procedures that will be difficult to integrate into routine clinical care. Frankly, no parent is going to consent to a lumbar puncture to confirm an autism diagnosis that can be made behaviourally.
The treatment-selection argument: The more interesting claim is the treatment-stratification one — that CSF AVP could identify a subgroup who would respond to vasopressin replacement therapy. This is where the story gets more biologically coherent. But again, the therapeutic evidence base is itself preliminary — a small pilot trial — so you’d be using an unvalidated biomarker to select patients for an unvalidated treatment.
Bottom line: This is a mechanistically interesting paper from a group doing careful translational work under genuinely difficult constraints — postmortem neuropathological specimens with matched clinical data are extraordinarily rare. But as a contribution to clinical biomarker development, it’s a proof-of-concept stepping stone, not a clinical advance. The sample sizes mean the specific parameter estimates should be treated as provisional at best. For anyone in clinical practice, nothing here changes how you’d assess or manage a child referred for possible autism.
Reference: Oztan, O., Kong, J., Jiang, S., Crain, C. et al. (2026). CSF vasopressin concentration as a biomarker of social symptoms in ASD. Autism Research. Wiley.
Mathematical Ability in Autistic Populations
I often see the impact of stereotypes in my clinic. An international team recently published a systematic review and meta-analysis of 66 studies in Nature Human Behaviour (2026) examining mathematical ability in autistic populations.
Contrary to the common stereotype of autistic people as natural mathematicians, the study found that on average, autistic individuals scored lower on standardised maths assessments than their non-autistic peers. However, they also showed far greater variability. The spread of scores was much wider, meaning some autistic individuals excelled while others struggled significantly.
Autism is a condition defined by wide individual differences. Applying any group stereotype to an individual child is always unhelpful. I frequently see children denied extra educational support because school staff assume they must be good at maths, and I see others whose genuine talents go unsupported because the focus is entirely on their difficulties.
For parents and educators: This is a reminder to assess the child in front of you. If your child is struggling with maths, they need the right support. Neither their diagnosis nor stereotypes should get in the way of individualised assessment and planning, as emphasized by NICE guidelines.
Reference: Systematic review and meta-analysis of the proficiency and variability of mathematical ability in populations with autism spectrum disorder (2026). Nature Human Behaviour.
Girls Are Still Being Diagnosed Too Late
A landmark study published in the BMJ in February 2026 analysed data from about three million Swedish children over forty years. The study showed that while boys are diagnosed with autism at a ratio of roughly 3 to 1 compared to girls before the age of ten, that ratio narrows to almost 1 to 1 by age twenty.
Autistic girls are not rarer than we thought; they are simply being identified much later. The reasons are well-documented, from diagnostic criteria historically focused on male presentations to the fact that many girls learn to mask or hide their autistic traits at great personal cost.
As a clinician and a parent, this finding confirms what many of us have been saying for years. Late diagnosis ultimately means years of unnecessary struggle with mental health and a lack of access to adjustments that could make a meaningful difference.
For families: Trust your instincts. If you feel something is not quite right, pursue an assessment. Being told that your daughter makes eye contact or has friends should not automatically rule out an autism diagnosis. Autistic girls often manage these things, though frequently at an exhausting personal cost.
Reference: BMJ (2026). Sex differences in autism diagnosis: Swedish population study. British Medical Journal.
The NHS Assessment Crisis
As the government’s 2021-2026 National Autism Strategy enters its final year, NHS autism assessment wait times remain far above the NICE-recommended thirteen weeks. Some children’s services are reporting waits of over thirty months. The statutory guidance update promised in 2024 has still not been published, and it appears unlikely that the strategy’s targets will be met by 2026.
There is some practical news: from April 2026, Clinical Partners will resume NHS Right to Choose autism assessments. The Right to Choose pathway in England allows patients to request an assessment by a qualified provider other than their local NHS service, which remains the best route to a faster assessment for many families. However, wait times here are also increasing due to high demand.
What can families do?
- Ask your GP about Right to Choose. You have a legal right under the NHS Constitution to choose your provider for a first outpatient appointment. Your GP should discuss these options with you.
- Document everything. Keep a record of when you raised concerns and what the waiting time estimate was.
- Contact the National Autistic Society’s helpline (0808 800 4104) for local guidance, as provision varies significantly by area.
References: - NHS Digital. Autism statistics. digital.nhs.uk. - National Autistic Society. Autism assessment waiting times. autism.org.uk. - NICE (2017). Autism spectrum disorder in under 19s: recognition, referral and diagnosis. Clinical guideline CG128/CG170.
What I Am Watching Next
The Autism Research journal has recently published several new papers on autism mechanisms, including work by Magnuson and colleagues (2026), which I will be reviewing soon. I am also following the Lancet’s recent review on paracetamol exposure and autism risk, evaluating the evidence to provide clearer guidance for families.
I will be back in a fortnight with the next update. Please do get in touch if there is a specific study you would like me to cover.
Reference for upcoming review: Magnuson, J. E. et al. (2026). Autism Research. Wiley.
Dr Odet Aszkenasy is a Consultant Developmental Paediatrician and parent. She writes fortnightly research summaries for Practical Autism Research.