When Therapy Helps Some Autistic Adults but Not Others: What 7,175 NHS Records Reveal

Draft for review only — not for publication or scheduling

When psychological therapy does not help, it can be tempting to conclude that you have failed, or that therapy itself is pointless. For autistic adults, that conclusion may be especially painful when the therapeutic approach has not been adapted to your sensory, communication, functional or cultural needs.

A 2026 study offers a more nuanced picture. Rather than describing one average response, researchers examined how depression and anxiety symptoms changed across 7,175 autistic adults who accessed routine psychological therapies in English primary-care mental-health services.

The findings were not a simple success story or a simple failure story. Some people improved rapidly. Some improved more gradually. Most remained broadly stable. A small number experienced worsening symptoms during the period studied.

This matters because a single average outcome can conceal the very different journeys taking place underneath.

What did the study examine?

Pender, El Baou, O'Nions and colleagues (2026) analysed routinely collected NHS data from autistic adults who accessed NHS Talking Therapies for Anxiety and Depression — the service formerly known as IAPT — between 2012 and 2019.

Participants had an autism diagnosis recorded in their hospital or mental-health records, met clinical thresholds for depression or anxiety at the start of treatment, received at least three assessment or treatment sessions, and had been discharged from the service.

Depression was measured with the PHQ-9 and anxiety with the GAD-7, both collected at every appointment. The first eight sessions were modelled.

The researchers used a statistical technique called growth mixture modelling. In straightforward terms, this looks for different patterns — or trajectories — of symptom change over time.

They identified five depression trajectories and seven anxiety trajectories.

The five depression trajectories

• Moderately severe, not improving — 39.0% of the sample

• Moderate, limited improvement — 36.7%

• Moderately severe, improving gradually into the mild range — 18.6%

• Severe, improving rapidly into the minimal range — 4.3%

• Moderate, deteriorating into the moderately severe range — 1.5%

The seven anxiety trajectories

• Severe, not improving — 33.1%

• Moderate, limited improvement — 20.5%

• Severe, improving gradually into the mild range — 18.4%

• Moderate, improving into the mild range — 13.9%

• Moderate to mild, minimal improvement — 6.7%

• Severe, improving rapidly into the minimal range — 4.9%

• Moderate, deteriorating into the severe range — 2.6%

What the proportions actually tell us

It would be comforting to lead with the rapid improvers. Honesty requires leading with the largest groups.

For depression, the two non-improving classes together accounted for roughly three-quarters of the sample. For anxiety, three classes showing no or minimal improvement accounted for around sixty per cent. Rapid improvement was the experience of about one person in twenty in each model.

This is consistent with earlier work showing that autistic adults are less likely than non-autistic adults to improve or recover in these services.

It is worth being equally precise about the other end. Deterioration was uncommon: 1.5% in the depression model and 2.6% in the anxiety model. If you are approaching therapy with dread, the evidence does not suggest that getting worse is the likely outcome. The more common outcome, unfortunately, is that not much changes.

The improvement that did occur was substantial where it happened. Almost everyone in the rapidly improving groups showed change beyond measurement error, and most crossed out of the clinical range altogether.

The important message is this: there was no single autistic therapy experience.

A vital caution: association is not causation

This was an observational routine-care study, not a randomised controlled trial. It cannot prove that therapy caused improvement, stability or deterioration.

People whose symptoms worsened may have been experiencing additional stressors, unsuitable environments, physical health difficulties, social exclusion, burnout or barriers to accessing the right level of care. The dataset cannot tell us that therapy itself caused the worsening.

Likewise, when symptoms improved during therapy, we cannot assume that therapy alone produced the change. Natural recovery, medication, practical support, relationships, housing, employment or other factors may also have contributed.

The study shows patterns of symptom change during routine psychological therapy. It does not establish a direct cause-and-effect relationship.

The study can tell us	The study cannot tell us
Autistic adults followed several different symptom trajectories	Exactly why each person improved or deteriorated
Some people improved rapidly or gradually	That therapy directly caused those changes
Some people experienced little change or worsening	That an individual person will follow a predicted trajectory
Daily-living difficulties were associated with poorer outcomes in some severe-symptom groups	That functional difficulties are a personal defect
Ethnically minoritised participants were more likely to experience worsening anxiety	The precise cultural, structural or clinical mechanisms involved

What the study cannot tell us

Several limitations deserve to be stated plainly, because they change how much weight any individual should place on these findings.

The measures were not designed for autistic people. The PHQ-9, GAD-7 and the functioning scale used have not been validated in autistic populations. The authors note that current measures do not capture all aspects of autistic experience, and that autistic burnout — increasingly understood as distinct from depression, and as a response to poorly adapted environments — may not be distinguishable from depression on a PHQ-9.

“Recovery” may not mean what it appears to mean. The authors are explicit that the meaning of improvement for autistic people should not be assumed to be the same as for non-autistic people.

The statistical groupings are uncertain at the individual level. The models’ entropy scores were 0.56 and 0.61, which indicates limited confidence about which group any particular person belongs to. The trajectories describe patterns in a population. They are not a diagnostic category and they are not a prediction about you.

The sample excludes those who left early. Only people who received at least three sessions and were discharged were included. Anyone who dropped out sooner — plausibly including some of those for whom the therapy was least accessible — is not in this dataset. Neither are those who were never referred.

Undiagnosed autistic people are absent. Inclusion depended on a formal autism diagnosis recorded in NHS records. Because autism remains widely underdiagnosed, particularly in women and in ethnically minoritised communities, many autistic people accessing these services would not have been counted.

The findings concern autistic adults who accessed NHS psychological therapy in England. They should not automatically be generalised to people without a formal diagnosis, to children, or to every autistic person.

The hidden clinical context: daily living is not background detail

One of the most clinically important findings concerned functioning in daily living, measured using the Work and Social Adjustment Scale.

Among people with severe or moderately severe depression or anxiety before treatment, greater difficulties with everyday functioning were associated with a lower likelihood of following an improving trajectory. These difficulties covered four areas:

• Managing tasks at home

• Participating in private leisure

• Taking part in social leisure

• Maintaining close relationships

Employment was assessed separately, and was excluded from the main analysis because not everyone was in work. When it was added back in a sensitivity analysis, the pattern held.

This should not be interpreted as “people with more difficulties are less motivated”. That would place the burden in the wrong location.

A person may struggle with daily living because their environment is inaccessible, unpredictable, sensory-intensive or socially hostile. They may be spending enormous cognitive energy on executive functioning, emotional regulation and camouflaging before they even arrive at a therapy appointment.

The authors themselves reach for this reading. They suggest that difficulties in home, leisure and relationships probably reflect poorly adapted and unsuitable environments, and that these difficulties then perpetuate the health disparities autistic people already face.

In this context, asking someone to complete extensive homework, attend appointments at inconvenient times, tolerate overwhelming surroundings or describe emotions using abstract language may increase the load on an already overloaded operating system.

A therapy plan that addresses anxiety while ignoring the conditions generating exhaustion may be trying to mop the floor while the tap remains open.

Social leisure, masking and the “Masking Debt”

The study found particularly important associations between social leisure difficulties and anxiety outcomes. Lower difficulty with social leisure at the outset was associated with a greater likelihood of both rapid and gradual improvement from severe anxiety.

Social leisure has a specific definition in the scale used. It refers to activities done with other people: parties, bars, clubs, outings, visits, dating and entertaining at home.

For some autistic adults, social participation is not simply a question of confidence. It may involve:

• Monitoring facial expressions and tone of voice

• Rehearsing conversational responses

• Suppressing natural movements or communication styles

• Managing noise, light, touch and unpredictable changes

• Recovering for hours or days afterwards

This sustained camouflaging can create what we might call Masking Debt: the accumulated physiological and psychological cost of repeatedly performing acceptability at the expense of authenticity.

Research has linked camouflaging with psychological distress (Cook et al., 2021). Autistic burnout may then emerge when the nervous system can no longer maintain the same level of compensation, often involving profound exhaustion, reduced tolerance of sensory input and a loss of previously available skills (Raymaker et al., 2020).

The study authors raise exactly this possibility. They suggest that people reporting higher difficulty with social leisure — the group least likely to benefit from anxiety treatment — may have been affected by autistic burnout, and that support for social participation, camouflaging and burnout may be useful before or alongside psychological therapy.

This is an important shift. We are not merely asking, “How can we reduce your anxiety?” We are also asking, “What is your anxiety responding to, and what would make your life more inhabitable?”

Cultural responsiveness is essential

The researchers also found that participants identifying as belonging to the global majority — that is, an ethnically minoritised group in England — were more likely to follow a worsening anxiety trajectory than participants identifying as White. The association was substantial in size, though the confidence intervals were wide.

This finding must be handled carefully. It does not suggest that ethnicity causes poorer outcomes, and it does not explain what happened in any individual case.

The authors are candid about the limitations. Ethnicity had to be collapsed into a single binary category because the numbers in individual groups were too small to analyse. Ethnically minoritised participants were underrepresented in the cohort, and they were more likely to have missing data at the end of treatment. The authors describe this compression as regrettable and call for finer-grained work on the intersection of autism, race and ethnicity.

Nevertheless, it is an important signal of possible inequity, and one the authors think warrants targeted attention and cultural adaptation of therapy, not only autism-specific adaptation.

An autistic person may experience overlapping forms of misunderstanding or marginalisation related to autism, race, ethnicity, language, religion, gender, migration history or socioeconomic circumstances. A culturally responsive therapist should remain curious about these intersecting realities rather than assuming that a standardised protocol is culturally neutral.

This may involve:

• Asking how the person understands their difficulties

• Exploring experiences of racism, exclusion or institutional mistrust

• Considering family, community and cultural expectations

• Checking whether examples and therapeutic language feel relevant

• Offering interpreters or communication support where appropriate

• Recognising that “recovery” may mean autonomy and authenticity, not simply appearing more socially typical

What personalised, neurodiversity-affirming therapy may involve

Guidance and research discussed by the authors support adaptations such as more concrete structure, written or visual information, environmental adjustments and communication accommodations (Loizou et al., 2024; National Institute for Health and Care Excellence, 2012).

It is worth noting that this guidance was already in place in England throughout the period the study covers. The dataset does not record whether any of it was actually applied. We do not know how many of these 7,175 people received a therapy that had been adapted at all.

In practice, adaptation might include:

Agreeing the structure of sessions in advance. A clear agenda can reduce uncertainty and cognitive load.

Allowing additional processing time. Silence is not necessarily disengagement. You may need time to translate an internal experience into words.

Using written summaries. A brief recap after the session can support memory and reduce the pressure to retain information while emotionally activated.

Adapting emotional language. Feelings may be explored through body sensations, patterns, images, scales or concrete examples rather than relying only on abstract labels.

Assessing daily-life barriers. Therapy should consider sleep, food, sensory environments, work, relationships, transport and domestic demands.

Discussing camouflaging and burnout without pathologising autistic identity. The aim is not to make you look less autistic. It is to reduce the cost of surviving environments that may not accommodate you.

At Life Changes 4 Good Consulting, we use the Name–Normalise–Redirect framework, developed by Dr Danielle Baillieu, to help translate insight into a compassionate next step:

Name: “I am overwhelmed by the demands of this environment.”

Normalise: “My nervous system is responding to sustained sensory and social load; this is not evidence of laziness or moral failure.”

Redirect: “I will identify one accommodation or support that reduces the load today.”

This framework does not replace assessment or treatment. It offers a micro-step between shame and action.

The existential question and the practical response

Existential therapy asks questions such as: Who am I when I am no longer organising my life entirely around other people’s expectations? What makes life meaningful, safe and worth inhabiting?

These questions can feel like the vertigo of freedom when masking has shaped your identity for years. Yet therapy also needs a practical “how”.

Cognitive Behavioural Therapy may help identify patterns between situations, thoughts, emotions and behaviour. Acceptance and Commitment Therapy can support values-based action without requiring you to eliminate every uncomfortable feeling first. Both approaches work best when adapted to the person’s communication style, sensory profile and actual capacity.

The goal is not intensity. It is consistency.

Not a dramatic transformation overnight. A therapeutic process that respects your nervous system, notices what is and is not working, and adjusts accordingly.

If therapy does not seem to be helping

A lack of early improvement should not be treated as a verdict.

The researchers found that trajectories could be distinguished with greater confidence at the third session than at the start. But they are unusually direct about how this should and should not be used. A lack of improvement by the third session, they write, should not provoke pessimism in either the person receiving therapy or the clinician, and should not be read as an indicator of poor overall outcome. Some people simply need more processing time.

What they propose instead is that the third session is a good moment for collaborative review — a point at which the person can say how the adaptations are or are not working, and at which care can be stepped up, augmented or adapted further.

Such a review might ask:

• Does the therapy model fit my needs?

• Are the communication methods accessible?

• Is the pace too fast?

• Is burnout being mistaken for depression?

• Are practical barriers being addressed?

• Do I feel safe enough to be honest?

• Do cultural and identity-related experiences have space in the work?

You are allowed to say, “This approach is not reaching me.” That is clinical information, not defiance.

If you are in immediate distress

If you or someone else is at immediate risk of serious harm, call 999 or go to A&E.

For urgent NHS mental-health support in England, call NHS 111 and select the mental health option (option 2).

You can contact the Samaritans on 116 123 at any time.

Parents or carers supporting a young person can contact the YoungMinds Parents Helpline on 0808 802 5544.

For information about neurodiversity and therapy, you can explore our neurodiversity resources and our therapies, techniques and research articles. You may also find our article on neurodivergent burnout helpful.

Compassionate final note

The findings from these 7,175 NHS records do not say that autistic adults are destined to benefit, or destined not to benefit, from therapy.

They say something more human and more useful: people respond differently, and the context matters.

They also say something uncomfortable that services need to hear. For most of the autistic adults in this study, routine therapy as it was delivered did not produce meaningful change. That is a finding about provision, not about the people receiving it.

If therapy has not helped you yet, this is not proof that you are broken. It may be an invitation to examine the fit, the adaptations, the pace, the environment and the support surrounding the therapy.

Your needs are not inconvenient data points. They are part of the clinical picture.

Intensity is not the same as progress. Consistency, collaboration and compassionate adaptation may be where meaningful change begins.

References

Cook, J., Hull, L., Crane, L., & Mandy, W. (2021). Camouflaging in autism: A systematic review. Clinical Psychology Review, 89, 102080. https://doi.org/10.1016/j.cpr.2021.102080

Loizou, S., Pemovska, T., Stefanidou, T., Foye, U., Cooper, R., Kular, A., Greenburgh, A., Baldwin, H., Griffiths, J., Saunders, K. R. K., Barnett, P., Minchin, M., Brady, G., Ahmed, N., Parker, J., Chipp, B., Olive, R. R., Jackson, R., Timmerman, A., ... Johnson, S. (2024). Approaches to improving mental healthcare for autistic people: Systematic review. BJPsych Open, 10(4), e128. https://doi.org/10.1192/bjo.2024.707

National Institute for Health and Care Excellence. (2012). Autism spectrum disorder in adults: Diagnosis and management (NICE Guideline CG142; last updated 2021). https://www.nice.org.uk/guidance/cg142

Pender, R., El Baou, C., O'Nions, E., Spector, A., Buckman, J. E. J., Richards, M., Pilling, S., John, A., Stott, J., Saunders, R., Crane, L., & Mandy, W. (2026). Symptom change in depression and anxiety during psychological therapy for autistic adults. Nature Mental Health, 4, 279–287. https://doi.org/10.1038/s44220-025-00567-4

Raymaker, D. M., Teo, A. R., Steckler, N. A., Lentz, B., Scharer, M., Delos Santos, A., Kapp, S. K., Hunter, M., Joyce, A., & Nicolaidis, C. (2020). “Having all of your internal resources exhausted beyond measure and being left with no clean-up crew”: Defining autistic burnout. Autism in Adulthood, 2(2), 132–143. https://doi.org/10.1089/aut.2019.0079

Website: Lifechanges4good.com — Email: dr.danielle@Lifechanges4good.comEmail:dr.danielle@Lifechanges4good.com