Discussion

The main aim of this multisite RCT was to assess the clinical and cost-effectiveness of the URB parent/carer group intervention versus a psychoeducation parent/carer group LAA for parents/carers of young autistic children aged 3 to 9 years 11 months, who present with functionally impactful RRB.

Families were recruited from clinical services across three research sites, two in the North of England and one in Scotland. Parents/carers were randomised to either URB or LAA and our analyses indicate that both interventions were run by trained facilitators with fidelity to the manuals. A small number of SAEs were noted for participants, none of which were deemed to be related to participation in the study, indicating that there were no identified serious risks associated with either intervention. Rates of and types of ESI were comparable across arms suggesting that any differences detected were not attributable to differences in families’ circumstances or experiences during participation.

All children had previously received a clinical diagnosis of ASD. Autistic traits were assessed at baseline using a play-based observational assessment (the ADOS-2) and a parent/carer questionnaire (the SRS-2). Adaptive functioning was assessed using the VABS-3. The mean scores reported across all three ADOS-2 modules produced comparison scores of six, indicating ‘moderate’ levels of autism spectrum related symptoms which are representative of a clinical sample of autistic children. Given COVID-19 restrictions, only around two-thirds of the sample were able to complete the ADOS-2, which is an in-person assessment, so caution needs to be applied in using these data to describe the representativeness of the sample. SRS-2 scores indicated that the sample demonstrated characteristics that are strongly associated with a diagnosis of autism. The mean baseline ABC score (67.63), derived from the VABS-3, was within the ‘low’ range of adaptive functioning, indicating difficulties with adaptive behaviour and daily living skills that are consistent with a clinical sample of autistic children. For comparison to a wider sample, based on standardisation data, a mean score of 63.3 is reported for autistic children aged 3–8 with a concurrent intellectual disability and a mean score of 76.8 is reported for autistic children aged 3–8 without a concurrent intellectual disability. The range of scores reported for the SRS-2 and the VABS-3 ABC suggests a sample of autistic children with mixed abilities, including children with and without a concurrent intellectual disability.

Recruitment to and participation in the study was significantly impacted by the onset of the COVID-19 pandemic in March 2020. At this point, the study was still recruiting participants, running intervention groups in person, and undertaking all assessments directly with parents/carers and children. Following extensive discussion, a decision was made to continue with the trial, necessitating a number of significant procedural changes, with all consent and assessment visits moved to remote delivery (via telephone) and intervention groups delivered online using free-to-access platforms. The ADOS-2 assessment was no longer able to take place. The move to remote delivery of the study did not seem to have impacted significantly on total recruitment to the trial (though recruitment at the Lothian site was completely closed from March 2020 to the end of the recruitment window). It is apparent however that a number of families were lost to follow-up and the primary end point, with missing data for 72 families at this follow-up point. The primary clinical outcome for the study was the CGI-I scale at 24 weeks. The CGI-I comprised data from a range of sources, which captured the presence and impact of functionally impactful RRB (the vignettes), RRB more generally (the RBQ-2 and Teacher RBQ) as well as adaptive functioning (the VABS-3). An ITT analysis indicated that at the primary end point (24 weeks) there was no evidence of an intervention effect on CGI-I, blind rated by trained assessors. It is important to consider the findings in relation to statistical power. Despite good enrolment to the trial, at the primary end point, due to difficulties contacting families for follow-up, the primary analysis was underpowered with 31.72% missing data (power = 70–75%). CGI-I was also analysed on its original scale and the result was in agreement with the primary analysis (see Appendix 3, Table 26). Thus, there is inconclusive evidence that the URB intervention was more effective than the attentional control (LAA), when considering global improvement in child functioning.

The results of the economic evaluation demonstrate that the URB intervention is unlikely to be considered cost-effective when compared to an active comparator such as LAA. These findings align with a previous study, the original Paediatric Autism Communication Trial (PACT) which found that the PACT intervention (a parent-mediated communication-focused therapy) resulted in a small additional effect at a greater cost.64 However, in keeping with the PACT, both studies demonstrated significant improvement in outcomes that are more proximal to the treatment target of the intervention. Further analysis of the PACT data has highlighted both a significant difference between the intervention and treatment as usual (TAU) and also persistence of the benefit 6 years after the end of the study.13 There are therefore a number of factors to consider around the interpretation of how best to evaluate both the clinical and cost-effectiveness of autism interventions which in turn can help inform future research in this area for autistic children and their families and the wider autism community.

We also considered the findings in relation to the secondary outcome measures. It is important to bear in mind two considerations when reflecting on the findings of the secondary outcomes. First, the study collected less than the expected data at the primary end point. Second, much of the study took place during a period of uncertainties, including for many the absence of their usual daily routines and support. Thus, taking part in the research study and having access to the support of a parent/carer group intervention may have provided participating families with a different experience from their peers.

The most proximal outcomes to the URB intervention are the TBVs, which capture a detailed description of a behaviour and its functional impact, from parent/carer interviews. Pairs of vignettes from different time points are then compared and expertly rated as indicating response/non-response. Our analysis demonstrated that children in the URB group were significantly more likely to be rated as responders at 10 and then at 24 weeks, with the odds of improvement in functionally impactful RRB on average 92% higher in the URB arm than the LAA arm at the primary end point (24 weeks). In addition, our analysis indicates that there was a statistically significant difference on impact on the family at 10 and 24 weeks, with those in the URB arm more likely to be rated as ‘responders’ than families in the LAA arm. These findings indicate that the URB intervention is more effective at supporting parents/carer to understand and reduce the impact of their child’s functionally impactful RRB than the LAA intervention and the maintenance of that effect 6 months post intervention is also clinically meaningful.

Restricted and repetitive behaviour more generally were also captured using two questionnaire measures (RBQ-2 parent and RBQ-2 teacher). A significant difference in parent/carer-reported RRB between the arms was found at 10 weeks follow-up, with lower RRB scores reported for those in the URB arm compared to the LAA arm. However, closer inspection of the data suggests that the difference can be accounted for by an increase in reported sensory motor RRB by parents/carers in the LAA arm, rather than a decrease in scores in the URB arm, and this pattern is not apparent at subsequent time points. Our findings therefore indicate that when considering parent/carer report of RRB more generally, rather than one specific functionally impactful RRB, as captured in the vignettes described above, there is no advantage conferred by participation in the URB intervention. Similar conclusions must be drawn when considering general RRB rated by teachers (using the RBQ-2 Teacher total scores) where we find a significant difference in RRB between the two intervention arms at 52 weeks. Further inspection reveals that this can be accounted for by a significant decrease in scores in the LAA arm at 52 weeks compared to the baseline. There were no significant changes in teacher-rated RRB over time in the URB arm.

One possible interpretation of these self-report findings, in relation to the vignette-based results, is that the URB intervention was successful in having an impact on identifying and specifically moderating functionally impactful RRB. A more general reduction in RRB as a whole was not observed. It would be unwise to conjecture further without more knowledge of all the different RRBs exhibited by the autistic children in our sample, but certainly a general extinguishing of RRB should not be considered either the goal of this intervention, or a sign of success.

As can be seen from the results for the other clinical effectiveness secondary outcomes, there is a general trend towards improvement across a range of both child and parent/carer outcomes, within both intervention arms, but no significant differences between the arms, indicating that participation in either intervention (or professionally run parent groups in general) may be beneficial for parents/carers and families.

Child adaptive behaviour was captured using the VABS-3 at baseline and then again at 24 weeks. For the URB group, significant changes can be seen in the domains of daily living skills, socialisation and the composite score comparing the 24-week outcomes to the baseline. These improvements are also found for children whose parents/carers participated in the LAA group, with the addition of a significant improvement in the communications domain also. No significant differences were found between the arms, indicating that while gains in adaptive functioning over time are apparent, these improvements are not specific to either intervention. It is also important to note that these changes, although statistically significant, are small and do not represent a clinically meaningful change in adaptive function.

Turning now to the parent/carer and family functioning measures, the story is similar. The PSE37 was used to record the presence/absence of behaviours typically exhibited by autistic children and parents/carers’ confidence in managing these behaviours. A significant difference between the arms was found on the PSE at 10 weeks follow-up, with the URB parents/carers reporting higher levels of self-efficacy compared to the baseline and to the LAA group, while the LAA group scores did not differ from the baseline. This difference between the arms is not maintained at 24 or 52 weeks follow-up; instead, within both arms we see significant increases in parent/carers self-efficacy compared to the baseline.

These findings are similar to those found with an APSI. Using the APSI, parenting stress can be seen to significantly reduce in the URB group at all follow-ups; for the LAA group no significant reduction was apparent at 10 weeks, but a significant reduction was apparent by 24 weeks and maintained at 52 weeks. There are no detected differences between the arms at any time point. Across these two measures then, we see more immediate positive impacts on the PSE and stress in the URB group, but over a longer time period both parent/carer groups yield benefits in this domain.

Parental well-being was further assessed using the WEMWBS at 24 and 52 weeks. No significant differences were found comparing the URB and LAA arms at either time point. Within the arms, a significant favourable change in scores was found for both arms at 52 weeks, but not at 24 weeks.

The AFEQ provides an overview of the family experiences of families including an autistic child. It provides scores within four domains [experience of being a parent/carer; family life; child development, understanding and social relationships; child symptoms (feelings and behaviour)] as well as a total score. For the family life and child development domains, significant changes are present at 24 weeks and 52 weeks for the URB arm, whereas a significant change only becomes apparent at 52 weeks for the LAA arm. For child symptoms there are no significant findings within or between arms at any time point.

Taken together, the findings from the secondary measures relating to parental functioning (self-efficacy, stress and well-being) and family functioning suggest that for the families that participated in the study both intervention programmes yielded positive outcomes, with no statistically significant differences between the two approaches. The results highlight that both the URB and the LAA interventions (or professionally delivered parent/carer groups) have the potential to be generally beneficial to families of young autistic children experiencing functionally impactful RRB. Of course, these were the families who had agreed to take part in the research. The continuity of being part of the trial especially during a time of great uncertainty for families and the wider community may have also been relevant.

It is important to reflect on potential explanations for the lack of positive effect of the URB intervention on the primary outcome (CGI-I) at the primary end point. A number of, potentially interrelated, explanations are worthy of consideration, including the suitability of the primary outcome to detect a difference, the suitability of the primary end point (24 weeks), the lack of statistical power to detect an effect (due to missing data) and of course the possibility that our hypothesis that URB would be superior to LAA is incorrect.

Regarding the suitability of the primary outcome and end point, the CGI-I measure was selected as it was in keeping with the priorities of the autism community in that it is a measure that captures global functioning that is relevant to everyday life experiences. In addition, this outcome was used during the pilot study and has been widely used in other clinical trials with autistic children and their families. However, the aim of both interventions was not to reduce the symptoms of autism, and also our understanding on whether and how changes in overall RRBs (including those considered to have a functional impact) might have affected global functioning is a complex judgement especially when the context for many families changed dramatically during the COVID-19 pandemic. Further, published early social communication intervention studies with autistic children often target proximal behaviours such as joint attention or parental synchrony. In such studies, a theoretical developmental framework supposes that gain of required proximal skills would lead to downstream impacts on wider abilities. This theoretical positioning reduces the burden on such trials to demonstrate either longer-term impact or change in measures of broader skills and everyday functioning. In contrast, the study’s primary outcome relied on evidence of change not only in the proximal outcome (target vignette) but across a range of domains (SCQ, VABS and parent/carer and teacher-reported RBQ). At the primary end point we do find a significant difference between the two arms in relation to the TBVs, the most proximal measure to the intervention, in favour of the URB arm. This short-term change in proximal skills corresponds with primary end points in other early autism intervention trials and suggests that 24 weeks may have been too early to detect evidence of broadening real-life impact.

With respect to the measures of health status used within the economic evaluation, the EQ-5D-5L is the recommended instrument by regulatory bodies such as NICE in their guidelines for health technology appraisals to express outcomes in terms of QALYs50 is not appropriate to measure health-related quality-of-life impacts of children aged under 12 years. It was utilised here to measure quality of life for the caregiver of the autistic child. There is an alternative version of the EQ-5D, the EQ-5D-Y that can be used to elicit health-related quality of life of younger children. However, it currently does not have a validated scoring system. Therefore, we used another tool, the CHU-9D, to estimate the impacts on health-related quality of life of younger children. Both the responses to the EQ-5D-5L and CHU-9D were used to estimate QALYs for parents/carers and children, respectively. The utility valuations that underpin the estimation of QALYs for both the EQ-5D and the CHU-9D only reflect health on the day that the measurements were completed. Brazier et al. noted that there are important elements of health that are not fully encompassed within the EQ-5D-5L (or CHU-9D) including well-being and capabilities.65 Some of these impacts of interventions to help understand and respond to RRB may fall within this purview. These include such things as parental reassurance and changes in the ability to plan for the future. Future research could potentially utilise tools such the ICECAP-A approach66 to measure parental QALYs from a capability and well-being approach. The view is supported by a recent review of economic evaluation approaches for children with neurodevelopmental disorders which advocates for greater use in approaches such as the capability approach.67

In addition to utilising well-being and capability approaches, there are also future opportunities to assess benefits from such interventions within the context of a cost–benefit approach. Potential options include the use of contingent valuation methods, which allow participants to value the intervention in monetary terms. The advantage of this approach is it allows respondents to consider all the facets of the intervention as they value it. While there are challenges in using this approach to value health care, a similar approach has been used to value early intervention for those with dementia.68 Exploring differing approaches to valuing interventions for the parents/carer of autistic children could form the basis of future research.

Another consideration regarding the results of the economic evaluation is the time horizon for the study, which in this case was 52 weeks. As this was a parenting intervention it is possible that the timescale of benefits (and cost) could extend into the future as the child grows into an adolescent and an adult. As such, a longer period of follow-up would be welcomed. This would provide an opportunity to consider the perspective of the autistic adult which is an understandable high priority of the autism community.69,70 Further, this would also provide a mechanism to examine any long-term harms of the intervention, something that is rarely done.69 It can be difficult to measure longer-term outcomes within the context of a trial. However, using other study designs such as observational approaches could provide valuable information to inform a cost-effectiveness modelling approach. The modelling approach could adopt a much longer time horizon, such as 20 years, or a lifetime time horizon to truly understand the long-term impacts of such early years' interventions.

While our fidelity assessment indicated that both interventions were successfully delivered with fidelity to their respective manuals, with different therapists involved in each arm of the trial, there is the possibility of inadvertent personalisation and informal discussion of RRB by therapists in the LAA arm. Some personalisation was noted within the LAA arm during fidelity checks, possibly due to small group numbers (i.e. one or two parents/carers in attendance) and therapists’ desire to support families during the COVID-19 pandemic. This would, of course, reduce the potential difference between the content of the two interventions and increase the sample required to detect any effect, in what is already an underpowered trial.

This study has a number of strengths. The URB intervention was developed in partnership with parents/carers in response to an identified underserved clinical need and contributes significantly to the range of interventions available to parents/carers of autistic children by providing the first manualised parent/carers mediated intervention that focuses on functionally impactful RRB. Despite the challenges and restrictions of COVID-19, the clinical and research teams responded rapidly to the unfolding situation and adapted research methods and clinical delivery to enable the trial to continue to deliver the study objectives, with recruitment closing permanently in only one research site (Lothian). One unanticipated benefit of these challenging circumstances was the opportunity to deliver the assessments and interventions via remote means. Our fidelity analysis, which included ratings of sessions delivered in the original face-to-face and remote formats, indicates that both methods of delivery are feasible, can be done with fidelity to the manual and are acceptable to parents/carer. Indeed, informal feedback from parents/carers during COVID-19 suggests that they were appreciative of the efforts of the research and clinical teams to continue with the study during a time when many clinical, educational and support services were closed.

On the other hand, only a minority of families (n = 47) completed their participation in the trial up to the 24-week follow-up assessment entirely prior to the onset of COVID-19 restrictions and therefore less than 25% of those families recruited to the study experienced the trial in its original format. For all other families, a combination of arrangements was in place. Some families had baseline assessment pre COVID-19 and all other procedures (intervention and follow-up assessments) during COVID-19 restrictions, while others had baseline assessment and participated in their intervention pre COVID-19 and then completed the primary end-point follow-ups during COVID-19 restrictions and some had all study procedures during COVID-19 restrictions. It is impossible to determine the precise impact of these unanticipated and fluctuating circumstances on the families who were part of the study, or indeed on the research and clinical team charged with delivering the trial in such uncertain times. Although the study recruited and randomised 227 families, a significant number were lost to follow-up at the 24-week primary end point and it is plausible that COVID-19 contributed substantially to this attrition rate.

Coronavirus-19 may also have impacted in other ways on the outcomes of the study. Many parents/carers reported changes in their child’s well-being as a consequence of the pandemic. For some, these changes may have been difficult, with an increase in worry and anxiety related to the uncertainty about how the pandemic would unfold and disruptions to routine. For others, the changes may have been more positive with the child benefiting from a less demanding lifestyle with the focus on staying at home with family. For many families it may have been a combination of the two at different times.71,72 For all parents/carers the closure of schools and other services meant that they lost their expected everyday support networks and often became sole full-time caregivers of their autistic child. In the context of these difficult uncertain and changing circumstances, the significant and positive changes detected in parental self-efficacy, well-being and stress and the improvements in family functioning in both arms are not to be underestimated.

Statistical considerations

Please see Chapter 2, Statistical Methodology where the COVID-19 implications on statistical analyses are discussed.

Equality, diversity and inclusion and patient and public involvement

Recruitment via clinical teams and data from available baseline characterisation measures indicates that the sample was comprised of autistic children with mixed abilities, including children with and without a concurrent intellectual disability. This is consistent with our inclusion criteria (no restrictions based on ability). The age range for children eligible for the study was widened in response to feedback from clinical teams that the intervention was needed for older children too. In collaboration with PPI advisers, we ensured that participant-facing documents were inclusive and accessible. Parent/carer group sessions were purposefully offered in areas of economic deprivation to support participation of families from these areas. Despite recruiting from a relatively large geographical area, the sample was not representative of the ethnic diversity across the UK. Indeed, one reason given for declining to take part in the study was the lack of availability of translators. The majority of children were from white ethnic backgrounds (85.9%). This is consistent with research indicating that there is a need for more culturally responsive practices for autistic people. Future research should aim to actively explore any potential reasons why families from non-white backgrounds are under-represented and promote recruitment polices that aim to rectify this imbalance and incorporate issues relating to diversity into research questions.73

Our research team comprised of individuals from a range of cultural and ethnic backgrounds, as well as those with lived experience. The research staff involved in the study benefited from training on a range of clinical measures and research and statistical techniques and some have now progressed to clinical training and others to research posts. Opportunities were provided for undergraduate students and trainees to work as interns during the study, gaining valuable skills and knowledge to further their career prospects.

Patient and public involvement

Parent/carer representatives were involved in the initial development of the URB programme (see Introduction), as well as acting as full and active members of the TSC. PPI considerations are also embedded within our dissemination strategy. We have established a working group and are in the process of developing guidelines for how to proactively engage with the autism community in relation to the interpretation and dissemination of the findings of the study. Outputs will be guided by close liaison with PPI advisers to ensure that the materials are accessible and inclusive, and that the language is appropriate. For example, outputs will be sensitivity-read by paid, autistic sensitivity readers. Conference presentations will include contributions from PPI representatives. Working closely with our PPI parent advisers and representatives from the autism community, we are planning a series of dissemination events and publications for the wider autism and more general community, in addition to academic and clinical outputs. For example, a URB newsletter for parents/carers, clinicians and local stakeholders who helped with recruitment will be produced and distributed at all sites. PPI representatives, with support from the research team, will submit an article to the INVOLVE newsletter and we hope to be able to present the study findings at appropriate parent/carer/third sector/professional local, national and international conferences including NAS Annual Conference, Autism Europe, and the International Society for Autism Research (INSAR). We will publish a FAQ on the URB website that addresses questions from the autistic community. We will also host a free, online ‘town hall’ event after publishing results where autistic people, and anyone from the broader autism community, can come and ask questions from a panel of the research team. Reports will be submitted in accessible newsletters such as Your Autism and Your Impact (NAS) and Asperger United. A feedback event will be held which will be available across all sites.

The novel URB intervention evaluated in this trial was developed over 12 years ago and piloted almost 10 years before the publication of this report. In that time there has been an explosion in recognition of the imperative, and understanding of how, to work with autistic people as partners in research (Fletcher-Watson et al. 2019, 2021). Reporting the results of this trial provides an important point at which to engage with the autism community – not only parents/carers and practitioners, who already have played key roles in the creation of URB and its evaluation, but also autistic people. While many decisions are in the past, there is room for autistic influence in interpreting and sharing these findings and determining the appropriate next steps for research and practice. We will be convening a consultant group to partner with us in this work. It is already clear that future research should work with autistic people to examine how we distinguish between RRB that are/are not reasonable targets for intervention, and what alternate strategies for managing anxiety and distress might be most useful for autistic children to adopt.74,75

Limitations

• The study did not recruit to target, and experienced high attrition at follow-up. The study was therefore underpowered at the primary end point to detect a difference between the two arms.
• The study employed a number of retention strategies that were unfortunately not successful. The retention strategies employed included consideration of parent burden in selection of outcome measures, frequent family newsletters reporting on study progress, follow-up assessments where possible being completed by the same RA so a rapport could be built up, flexibility in timings and location of follow-up assessments, reminder texts, e-mails, letters and phone calls.
• Delivery of the trial was significantly impacted by the onset of COVID-19. It is not possible to determine whether the change of delivery methods impacted on the findings reported.
• There is a lack of ethnic diversity amongst the families who participated in the study.
• Although no measure of co-occurring problems was included in the study protocol, the range of scores reported for the SRS-2 and VABS-3 ABC subscale provided indirect markers of the range of difficulties commonly experienced by young autistic children.
• Direct observational assessment of the children was not possible following onset of the COVID-19 pandemic.
• The follow-up of the study may have been too short to detect differences in effectiveness and cost-effectiveness of the interventions.
• An ongoing methodological challenge is that there is no single measure or methods to aggregate health status and QALYs across carers and child.

Measures of health status themselves may not reflect the broader impacts of the interventions. Other outcomes such as those relating to the ICECAP may be worth exploring in future studies.

There is a lack of any qualitative work such as a process evaluation to help explore potential explanations for the limited effectiveness of the URB parent group. Qualitative semistructured interviews may have provided insights into low attendance rates and factors that impacted on recruitment and retention rates.

Strengths

• Families were successfully recruited from a range of clinical services across three research sites.
• Adaptations to delivery of both research assessments and the two parent group-based interventions were possible to enable the study to continue despite restrictions related to the pandemic. However, a great deal of additional work was required by the research team and the therapists as each NHS trust adopted different virtual packages based on local policies/guidance.
• Trained facilitators were able to deliver the programmes with fidelity to the manuals in both in-person and remote formats. A very small number of unrelated SAEs were noted indicating that there were no identified risks associated with either intervention.
• The economic evaluation used rigorous, widely accepted and explicit methods to estimate relative efficiency. This was done in three complementary ways, accompanied by extensive stochastic and deterministic sensitivity analysis which gave rise to broadly consistent conclusions.

Implications for practitioners

• Our data suggest that those parents/carers who took part in the research found both interventions to be helpful, feasible and acceptable, and we demonstrated that both can be delivered with fidelity to the manual in in-person and remote formats.
• Many parents/carers did not take part in the research citing a range of reasons including time commitments, child care, finances, and travel arrangements. Those who did take part during the COVID-19 pandemic found face-to-face and online delivery of the parent group intervention, and telephone assessment appointments acceptable. Providing carefully tailored and flexible opportunities for families to attend appointments for assessments and/or access to interventions may improve efficient use of limited clinical resources without necessarily impacting of quality-of-service delivery.
• The use of online delivery for both parent groups was found to be acceptable and feasible. Online delivery can increase efficiency in a variety of ways for both parents and healthcare professionals.
• Some families who declined to take part indicated that they could not identify any functionally impactful RRBs. The discussion about RRBs and social communication needs are important as part of the diagnostic assessment process and when planning appropriate interventions to meet the needs of young autistic children and their families.

Implications for parents/carers and autistic children

• Participation in the URB programme resulted in a greater likelihood of reduction in engagement in functionally impactful RRB by the child and an improvement in the impact of those functionally impactful RRB on the family.
• Parents/carers who attended URB and LAA reported an increase in self-efficacy, a reduction in parenting stress and improvement in well-being over time indicating that both programmes are helpful to parents/carers of young autistic children.
• Participation in either the URB or LAA parent groups can be mutually beneficial and supportive for parents of autistic children.
• A flexible approach to the timely delivery of both assessments and interventions with opportunities for face-to-face and/or virtual delivery of services was acceptable to parents/carers who took part in this research.

Implications for research

In future research it will be important to determine:

• How can participation in autism research of autistic people from non-white backgrounds be improved? Issues relating to diversity should be built into research questions.
• What are the mechanisms of change in functionally impactful RRB?
• What should be included in a standardised set of outcome measures for autism intervention studies and what should be the optimal primary end point to detect meaningful change?
• How can recruitment and retention of families be improved?
• How can attendance rates at parent group sessions be improved?
• Whether longer longitudinal follow-up, which could be utilised in economic modelling studies over a longer time horizon, could change conclusions on effectiveness and cost-effectiveness.
• What other measures could be used to quantify benefit for economic evaluations in autism intervention studies?