Refining Biomarker Evaluation in ASD

Abstract

This commentary reflects on reasonable biomarker expectations in ASD by addressing three key questions: What is a biomarker? What is required for a biomarker in ASD? How can biomarkers be useful in ASD? In addressing these queries, a path forward emerges based on clear definition of the objective for any given ASD biomarker and evaluation of each biomarkers relative to current best practices.

At present, there are no biomarkers to effectively guide clinical practice or clinical research in autism spectrum disorder (ASD). There is, however, great interest in and extensive research on the development of biomarkers for ASD. Controversy surrounds the viability of extant biomarkers for ASD; some of the most well-studied biomarkers do not display all potentially desirable properties of biomarkers, leading to questions regarding their utility (Key and Corbett, 2020). This commentary reflects on reasonable biomarker expectations in ASD by addressing three key questions: What is a biomarker? What is required for a biomarker in ASD? How can biomarkers be useful in ASD? In addressing these queries, a path forward emerges based on clear definition of the objective for any given ASD biomarker and evaluation of each biomarker relative to current best practices.

What is a biomarker?

The FDA-NIH BEST (Biomarkers, EndpointS, and other Tools) Resource (Group, 2016-) operationalizes a biomarker as “a defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or biological responses to an exposure or intervention, including therapeutic interventions… A biomarker is not an assessment of how an individual feels functions, or survives.” Seven specific categories of biomarkers are defined: susceptibility/risk, diagnostic, monitoring, prognostic, predictive, pharmacodynamic/response, and safety. Implicit in this framework is the notion that a biomarker may be effective for one purpose yet have limited utility for another purpose. For example, a marker of risk for ASD (i.e., susceptibility/risk) may provide no useful information for evaluating whether a pharmacologic treatment has engaged its desired target (i.e., pharmacodynamic/response). Also implied by this structure is that distinct types of data derived from accordingly designed studies are required to assess the viability of a biomarker for a specific domain. Based on this framework, a biomarker for ASD might be operationalized as a measurement of a biological process relevant to a particular clinical objective (or, as described by the FDA, its context of use), with evaluation of its fitness necessarily being specific to that given context of use.

What is required for a biomarker of ASD?

Following from the above, there could be multiple biomarkers for ASD, corresponding to multiple contexts of use. For example, an eye-tracking assay of visual attention in 6-month-old infants might represent an effective susceptibility/risk marker for ASD. This same biomarker, applied in school-aged children, might represent an effective index of response to behavioral treatment (i.e., pharmacodynamic/response). However, these distinct contexts of use are independent and would need to be validated by distinct studies and data types; beyond required standard psychometric properties, such as test-retest reliability, the viability of an ASD biomarker in one context of use provides limited information regarding its appropriateness for a different context. A challenge for biomarker evaluation in ASD has been a tendency to concurrently evaluate the fitness of a biomarker for any context of use based on its viability across multiple contexts of use. For example, potential biomarkers for ASD have been evaluated critically because available data were not unilaterally supportive of concurrent effectiveness in indexing clinical status, reliably measuring development over time, stratifying individuals into clinically meaningful subgroups, and demonstrating association with specific symptom domains (Vettori et al., 2019). While it is desirable that a single biomarker might meet all these criteria, meeting only subsets of these criteria are required for effectiveness in a given context of use. In ASD, a common instantiation of this problematic approach to evaluating a biomarker is to expect a biomarker to uniformly apply to all individuals with ASD and not apply to individuals who do not have autism; though this would be required for a diagnostic biomarker, it may not be needed for other contexts of use; indeed, it is likely that biomarkers useful for ASD may also be relevant to other neurodevelopmental conditions. By setting unattainable standards, the field risks foregoing significant practical benefits of more modestly successful biomarkers in pursuit of potentially unrealistic cross-purpose markers.

How can biomarkers be useful in ASD?

At present, nearly all clinical practice and research in ASD and other psychiatric conditions relies on phenotypic information collected via subjective clinical measures that are resource consumptive and prone to bias (Jones et al., 2017). Estimators of clinical status that are biologically based may offer significant advantages. Because most treatments for ASD enact change via the central nervous system, direct measurements of this activity may be more likely to effectively or efficiently detect change at this level (i.e., target engagement or early efficacy), prior to potential detection via behavioral methods. Progress from clinical to biomarker based evaluation may also support practical goals, such as economy and scalability (McPartland, 2016). In short, a biomarker that performs similar functions to the current status quo, i.e., clinical ratings and caregiver reports, could yield significant practical benefits.

Looking forward

ASD biomarker development will benefit from thoughtful and realistic evaluation of biomarkers operationalized within specific contexts of use, using data collected to be appropriate for that purpose. Given limited replicability among biomarkers findings in ASD, the field should continue to rigorously interrogate well-studied markers (McPartland et al., 2020) while simultaneously evaluating the viability of novel indices (Rossion, 2020). Indeed, for a condition as heterogeneous and complex as ASD, it may be beneficial to target common brain systems using complementary assays. As evident in this series of commentaries (McCracken et al., 2021), increased communication and collaboration with regulatory agencies can facilitate progress through optimization of biomarker study design and data interpretation in alignment with regulatory procedures. The nascency of biomarker development in ASD is a key reason for its great potential to make great progress in the short term.