Assessment of outcomes in autism clinical trials over the course of development

The promise of developing more effective pharmacological treatments for autism spectrum disorder (ASD) is tempered by the considerable challenges of designing clinical trials and developing robust outcome measures. ASD involves multiple symptoms within the broad domains of social impairments and restricted, repetitive behaviors. Individual symptoms may or may not be present and will differ depending on the individual’s age, developmental level, and sex. Improved outcomes can represent a reduction in symptoms, such as reduced social withdrawal, or increases in rate of skill acquisition. Well-validated outcome measures for many core symptoms, such as impaired social motivation and sensory sensitivities, are lacking. There is also a need for validated self-report measures that have been co-developed by investigators and autistic individuals to ensure that these measures reflect outcomes that are meaningful to the individuals for which new treatments are being developed. Similarly, there are very few outcome measures that are appropriate for individuals with ASD who are minimally verbal.

One approach to addressing some of these challenges has been to utilize standardized adaptive behavior ratings, such as the Vineland Adaptive Behavior Scales (VABS), that assess a broad range of socialization, communication, and daily living skills and have been normed and developmentally-referenced (Chatham et al., 2018). Improvements in adaptive behavior have strong face validity in terms of quality of life. Minimally clinically important differences have been defined for the VABS (Chatham et al., 2018). However, increases in adaptive behaviors necessitate that the individual has adequate opportunities during the clinical trial to acquire the skills that are being assessed, requiring a trial design of adequate length for such changes to occur. Rates of skill acquisition would be expected to be variable; for example, acquiring a skill might take longer for persons who have comorbid intellectual disability.

Another challenge in utilizing many standardized assessments of adaptive skills is their reliance on caregiver report. Caregivers might not have sufficient opportunity to observe behavioral changes, especially if the behavior occurs at school or with low frequency. Thus, involving teachers in the outcome assessment can be helpful. Furthermore, both caregivers and clinicians are susceptible to the effects of expectancy/placebo. Even in the absence of a treatment, changes in caregiver-reported outcomes have been shown to occur over an 8-week period. Reviewing symptoms with caregivers can direct their attention to the targets of the treatment such that behavioral changes can reflect a response to the assessment process itself (Jones, Carberry, Hamo, & Lord, 2017). The use of eye-tracking and electroencephalography (EEG) measures to assess changes in social attention/processing could provide more objective assessments (Dawson, Bernier, & Ring, 2012). Several event-related potential (ERP) studies of face processing in children and adults with ASD find differences in face processing (Kang et al., 2018), and latency of the N170 ERP component, which is sensitive to face stimuli, is a strong ASD biomarker candidate (J. McPartland, Dawson, Webb, Panagiotides, & Carver, 2004; J. C. McPartland, 2017). While objective, such measures are not necessarily meaningful unless they are correlated with changes in clinical outcomes and quality of life.

Digital phenotyping is also being explored as an objective, quantitative, and scalable assessment method (Dawson & Sapiro, 2019). Computer vision analysis, for example, can quantify a wide range of ASD-relevant behaviors, including patterns of social attention, orienting, and facial expressions. Although recent studies indicate that such approaches might be useful for detecting early ASD symptoms (Dawson & Sapiro, 2019), their utility for assessing outcomes in clinical trials is still under investigation (Ness et al., 2019). A clear advantage of digital behavioral measurement is that it offers better precision and resolution, which might provide greater assay sensitivity.

Finally, most participants in ASD clinical trials are involved in background community-based treatments, both pharmacological and behavioral, and these will introduce variation in clinical trials. A large percentage of individuals with ASD are on psychoactive medications (Goel, Hong, Findling, & Ji, 2018). Excluding all individuals on such medications would result in study samples that do not reflect the general population of persons with ASD. Careful monitoring of the types and amounts of behavioral and pharmacological interventions is necessary. One strategy for reducing variability due to background interventions is to provide all study participants with the same behavioral intervention and randomize participants to receive placebo or drug. This requires additional staff and expertise, however, and this design provides information about add-on effects of the drug and not drug effects, per se.

While much progress in understanding the genetic and neural basis of ASD is being made, it is imperative that similar rates of progress occur in designing better outcome measures for ASD clinical trials. Together, such parallel progress offers hope that we will not only be able to discover new pharmacological treatments but also be able to know more definitively whether they are helpful in improving outcomes for persons with ASD.