Normalizing uncertain threat: What change in anxiety-potentiated startle can tell us about anxiety treatment

Identifying biomarkers and candidate mechanisms underlying psychological disorders represents an intriguing path toward improving treatments. The promise of this line of research is twofold. First, it can facilitate a more targeted approach to treatment by enabling interventions to specifically target the mechanisms responsible for changes in psychological dysfunction, thereby leading to more potent and personalized interventions. In addition, the presence or absence of change in a biomarker or mechanism during treatment may serve as a prognostic indicator of treatment response or maintenance of gains, with the potential to be a more sensitive indicator of clinical change than alternative measures and to inform clinical decision-making (e.g., whether to continue or alter treatment approach). Importantly, the evidence that must be accumulated to establish a potential biomarker as a prognostic indicator or targetable mechanism of change is substantial, and successful translation of biological and other basic research into improved interventions has been somewhat limited (1). Nonetheless, the research by Hoge et al., (2) in this issue of Biological Psychiatry investigates many of the initial questions required for evaluating a candidate mechanism of change (illustrated in Figure 1), specifically with regard to anxiety- and fear-potentiated startle (APS, FPS). Results also raise critical questions that need to be addressed to move this line of research forward, particularly related to linking biological constructs and clinical outcomes.

Figure 1.

Illustration of the evidence needed to establish a potential biomarker as a plausibly targetable mechanism of change, and the extent to which such evidence was established in Hoge et al. (2) with regard to anxiety-potentiated startle.

One important strength of this study by Hoge et al., (2) is the examination of not just whether APS and FPS change during treatment, but also whether such measures meaningfully distinguish anxiety patients from non-anxious controls before and after intervention. In doing so, this work adds to the literature on the extent to which elevated APS and FPS characterize anxiety disorders while also narrowing in on the component of the startle response that is the most meaningful indicator (and potential mechanism) of change, which in this case appears to be startle under uncertain threat (i.e., APS). Moreover, the examination of post-treatment APS responses in comparison to non-anxious controls showed that treatment “normalized” this elevated startle response in patients with anxiety disorders. Such a finding indicates such findings are clinical meaningful and rules out the possibility that reductions in APS in the treated groups were merely the result of retesting effects.

A second strength of the study design is the examination of two active treatments – escitalopram and mindfulness-based stress reduction (MBSR) – with different purported mechanisms of action. Establishing that an anxiety-related process is more effectively targeted by one treatment than another can help strengthen the evidence for causality of the candidate mechanism (e.g., if greater mechanism change in turn leads to greater symptom change), further intervention development by illustrating how to best intervene upon the mechanism, and lead to a more personalized approach to treatment. Results in Hoge et al., (2) were ultimately inconclusive regarding differential treatment effects: while there were no significant differences between MBSR and escitalopram in the magnitude of APS change during treatment when directly compared, only the escitalopram group showed significantly greater APS change than the untreated healthy controls. Nonetheless such findings are important in establishing that two distinct treatments both impact APS. Were escitalopram to show a direct advantage in APS improvement over MBSR or other treatment in future research (e.g., for certain subgroups of anxiety patients), it could be used to specifically target those with particularly elevated APS, consistent with a precision medicine approach to treatment.

The possible benefits of targeting a candidate mechanism in such a personalized manner, however, depend on changes in the variable of interest being associated with changes in clinical outcomes, and this is where the promise of APS as a mechanism becomes more tenuous. Small correlations between changes in APS and changes in the State-Trait Anxiety Inventory (STAI; r = .22) and Center for Epidemiologic Studies Depression scale (CES-D; r = .18), alongside non-significant relationships with the Penn State Worry Questionnaire, the Beck Anxiety Inventory (BAI), and the Clinical Global Impressions scale, indicate that improvements in APS and subjective clinical measures are largely distinct. From an experimental therapeutics standpoint (3), such weak correspondence with symptom improvement raises questions about the extent to which APS is a meaningful target for intervention. Subjective distress and impairment are what ultimately bring patients in for treatment, and a focus on biological measures without consideration for their link to subjective experiences can be an impediment to progress in intervention research (4). In this regard, Hoge et al., (2) should be commended for not just examining correspondence of APS changes with symptom improvement, but also subjective fear and anxiety during the threat task, which significantly improved in both treatments groups relative to controls despite only a trend for differences at baseline. Unfortunately, the authors do not report whether improvements in subjective APS, were associated with symptom improvement, missing an opportunity to better understand subjective distress under uncertain threat as potential mechanism of change.

Two caveats are worth noting regarding the minimal correspondence between APS and symptom change. First, the subjective measures used in this study to examine symptom change were primarily broad assessments of anxiety and depression, and thus may have largely captured general distress and negative affect more distally related to APS and the specific psychological processes underlying patient’s primary symptoms. Elevated APS has been found to predict greater fear symptoms but not general distress or anhedonia-apprehension symptoms (5), which could partly explain the weak or nonsignificant associations with measures like the STAI, CES-D, and BAI. Examining APS change in relation to self-report measures that assess a more targeted set of symptoms or psychological constructs, particularly those related to uncertain threat, may increase the likelihood of successfully linking biological and subjective changes. In fact, the National Institute of Mental Health’s Research Domain Criteria (RDoC) matrix (6) delineates several constructs assessed by self-report (e.g., anxiety sensitivity, fear of negative evaluation, behavioral inhibition) under the Potential Threat (“Anxiety”) classification, which is also how APS is classified. A construct like anxiety sensitivity, which is implicated across anxiety disorders and has also been linked to APS (7) could be an ideal candidate for future research attempting to link reductions in APS to subjective measures. Examining such measures could help to establish APS’s clinical relevance more clearly.

A second caveat is that despite the weak relationship seen between change in APS and clinical outcomes across the whole sample, there may be individual differences in the strength of such relationships. Examining potential moderators or subgroups of patients for whom APS change is more closely linked to clinical improvement is an important step toward precision medicine in this domain, particularly given the heterogeneity in diagnostic groups examined by Hoge et al., (2). For instance, there is some evidence that elevated APS is more characteristic of fear-based disorders (e.g., social anxiety) than distress-based disorders (e.g., generalized anxiety) (8), thus improvements in APS may be more closely tied to treatment response among those with predominantly fear-based presentations.

When considering these relationships between biological and symptom change, it is also worth considering which treatment-related changes are ultimately the most important. Self-reported symptom measures are not without their flaws, and Hoge et al., (2) appropriately point out they are more prone to demand characteristics than biological measures. Thus, it is certainly plausible that APS is just as meaningful of an indicator of long-term outcomes in functioning and remission of diagnosis than subjective assessments, if not more so. For instance, startle reactivity in a different threat-related paradigm has been found to predict subsequent development of anxiety disorders beyond the effects of neuroticism and depression (9). Longer-term assessment of clinical outcomes, as well as functioning, well-being, and quality of life in relation to APS within or outside the context of treatment would be valuable to better understanding the clinical relevance of the construct. Similarly, more frequent assessments during treatment would enable investigation of the time course of APS and its relation to subsequent symptom during treatment, which would be the ultimate test of its clinical importance should it emerge as a predictor or mediator of treatment-related changes.

The above discussion focuses on the way treatment-related changes in a potential biomarker can be evaluated as candidate mechanism of change or prognostic indicator of treatment efficacy, as this would be the most exciting possible implication of this work for improving clinical outcomes. However, it is appropriate to view the work of Hoge et al., (2) with a less ambitious goal in mind: to better understand the biological correlates of clinical anxiety. In this way, Hoge et al., (2) have added to the body of research on APS as a distinguishing feature of anxiety by illustrating not just a difference in APS between clinical and non-anxious groups, but also by establishing that APS normalizes in the clinical groups after two distinct treatments. From that perspective, this study is a successful step toward better understanding the interaction between responses to uncertain threat and anxiety treatment, building on existing evidence on the role of APS in anxiety while opening further questions about its links to clinical change.