The Music for Autism (M4A) trial evaluates the neurobehavioral outcomes of a music therapy (MT) intervention, compared to a matched play therapy (PT) intervention, on social communication skills, brain connectivity and structural brain changes. In a crossover randomised controlled trial (RCT), 80 children with autism across all levels of functioning, aged 6-12 years, undergo a baseline assessment, which includes measurements of social communication, participation, functional connectivity and brain structure. Participants are then randomly allocated to a sequence of interventions (MT-PT or PT-MT) and assessments are taken before and after each intervention period. Both interventions will target common goals and follow the same structure, while at the same time allowing for flexibility in the therapists' approach. It is hypothesized that 12 weeks of intervention through MT, compared to PT, will improve social communication skills, participation, and other relevant mental health outcomes in children with autism spectrum disorder (ASD), as well as regulate resting-state functional over and under-connectivity and increase grey and white matter volume in specified regions. The investigators also expect changes in functional brain connectivity to correlate with behavioural outcome measures, specifically with improved social communication skills.

M4A will combine biomedical research with clinical outcome research to investigate if 12 weeks of intervention through MT, compared to PT, improve social communication skills, participation, family quality of life, receptive vocabulary, adaptive behavior, and symptom severity in children with ASD, and if this is accompanied by a change in resting state functional connectivity (rsFC) as well as grey and white matter volume change. Additional mental health outcomes include not only core areas of impairment, but also associated problems such as chronic stress, which is a significant issue for people with ASD; impedes learning; and can be reduced through music, both as an outcome in itself and as a mediator for other health outcomes.

M4A will also investigate if clinical improvement is correlated with an increase of rsFC between auditory and striatal/fronto-motor regions as well as a decrease of rsFC between auditory and visual regions in MT compared to PT. Possible changes in grey/white matter volume will be measured by voxel-based morphometry (VBM) in a whole-brain scan before and after the interventions.

Sample size and power: This study will be powered for an effect size of d=0.34. With a two-sided significance level of 5%, a sample of n=70 will be required to detect the effect with 80% power. Attrition is expected at <10%; the study will therefore recruit at least 80 participants. More specifically, the investigators expect to find a mean difference of 4.84 (SD=14.24), corresponding to an effect size d=0.34, on the primary outcome. The investigators expect the scores to be correlated within participants by r≥0.50. Sample size to achieve 80% test power was calculated in R.

Treatment fidelity: All sessions will be recorded on video to help ensure and assess treatment fidelity. Fidelity will be rated by 2 raters, who are trained on a manual for assessment of treatment fidelity, on 4 different dimensions: (1.) Program adherence (number of sessions completed; number and types of activities covered, from the therapist's weekly reports); (2.) Process fidelity (delivery of the theoretical concepts of the intervention); (3.) Content fidelity (establishment of a therapeutic relationship between the participant and the therapist, measured using quality of delivery and participant responsiveness as well as the theoretical principles underlying the interventions); (4.) Programme differentiation between MT and PT ("Music was central to this activity").

Statistical analysis of behavioural outcomes will compare change from before to after each intervention within each participant. The intention-to-treat principle will be followed as applicable in a crossover trial: Participants will be analysed in the group to which they were randomised, regardless of whether they actually received the full allocated intervention. The main analysis will include all participants with valid data for both intervention periods; in addition, multiple imputation of missing outcomes will be used as a sensitivity analysis. Tests will use a two-sided 5% significance level. The two main secondary outcomes, participation and quality of life, will be Bonferroni corrected; the remaining secondary outcomes will be exploratory. Analysis software will be R.

Brain connectivity of frontotemporal regions, measured as rsFC from 6 seeds, will be used as the main neuroscientific outcome. The time-series for each of the seeds will be used to generate individual participant-level maps using whole-brain general linear models at baseline and after the interventions. First-level maps will then be entered into the second-level analyses. For comparison after the intervention, the investigators will use ANCOVA with post-intervention rsFC as a dependent variable, and intervention, baseline rsFC, age, and intelligence quotient (IQ) as covariates. Z-scores of parameter estimates will be used to measure connectivity strength. Results will be reported with a 5% significance level adjusted for multiplicity by family-wise error rate. Z-statistics for each participant from the post-intervention rsFC maps will be used in a linear regression model to evaluate correlation between rsFC and behaviour change. Changes in grey and white matter volume will be assessed in a whole-brain scan using VBM, derived from the anatomical T1 image, acquired at the beginning of each fMRI scan. ROIs include the 6 seeds above as well as other areas identified in our previous review (cerebellum, superior temporal sulcus, temporo-parietal area). The investigators will use SPM12 on Matlab for standard preprocessing and analysis of VBM, and CONN for denoising and rsFC analysis.

Inclusion Criteria:

• Meet diagnostic criteria as specified in Diagnostic and Statistical Manual of Mental Disorders, Fifth edition (DSM-V) and officially diagnosed with autism spectrum disorder (ASD) by a licensed clinical professional using standardised diagnostic tools (Autism Diagnostic Observation Schedule (ADOS), Autism Diagnostic Interview Revised (ADI-R)).

Exclusion Criteria:

• Recent or current music therapy
• Metallic or electronic implants

De-identified clinical and neuroimaging data will be made accessible for re-use by other researchers via platforms such as ENIGMA. We also plan to publish the primary outcome IPD on a public repository such as NSD (www.nsd.no)

Permanent, from the date of publication.

De-identified clinical data will be stored in a publicly available repository (Open Science Foundation, https://osf.io/).

To ensure blinding, the primary outcome measure will be completed by special educators who are familiar with the child (rather than their parents). Success of blinding (for all outcomes) will be verified at the last follow-up.