Response

In mental health science the identification of biomarkers aims to improve our understanding of mechanisms and support diagnosis, prognosis and personalised treatment approaches. In anxiety, assays may target genetic, neural, pharmacological, electrophysiological, autonomic, behavioural, cognitive or self-reported markers of anxiety disorders. ^{Reference Penninx, Pine, Holmes and Reif1} A promising marker for anxiety should demonstrate several forms of validity: face validity – the convergent association with anxiety symptoms; construct validity (specifically ethiopathogenic validity), which accurately represents underlying mechanisms that cause and maintain anxiety disorder; and predictive validity, which forecasts clinical outcomes, including treatment response or risk of relapse. Additionally, markers should show reliability, sensitivity and specificity, consistently distinguishing individuals with anxiety from those with other conditions and healthy individuals. A useful marker should be measurable, reproducible and applicable across different populations and clinical settings. It is an exacting standard, but Zhang et al’s findings ^{Reference Zhang2} suggest eye-movement and pupillometry indices may prove useful markers in anxiety, extending their findings in depression and schizophrenia.

In Zhang et al’s study, patients with a primary diagnosis of either anxiety or depression (versus healthy controls) completed oculomotor measures of eye movements and pupil size during fixation stability, free-viewing and smooth pursuit. Extracted (principal) component markers performed quite well: differentiating patients with anxiety and depression from healthy controls; correlating with anxiety symptom severity; and mapping to mechanisms described in contemporary neuropsychological models of anxiety (hypervigilance, hyper-arousal, cortico-limbic dysfunction ^{Reference Eysenck, Moser, Derakshan, Hepsomali and Allen3} ). As expected, markers were less able to distinguish anxiety from depression (versus controls), though increased visual exploration (hypervigilance) showed some differentiation of anxiety and depressive illness.

These oculomotor tasks have several benefits. They are non-invasive, comparatively short and engaging for participants (though drop-out was not reported). They provide indices of behavioural, cognitive and autonomic characteristics of anxiety, and eye-tracking tasks can be more reliable than button-press reaction-time measures of cognitive function. Consequently they may provide accessible and well-tolerated measures for young and older individuals with anxiety. In Zhang et al, participants were comparatively young (16–45 years, mean age 25 years), and we need to evaluate tasks and markers in older adults – particularly given the persistence and prevalence of anxiety across the lifespan. Diagnoses were confirmed by psychiatrists, though Hamilton Anxiety Rating Scale (HAM-A) and Hamilton Depression Rating Scale (HAM-D) scores suggested mild levels of anxiety and mild–moderate depression. There was some evidence that markers may covary with anxiety and depressive symptom severity, though further assessment in individuals with moderate/severe anxiety is required. The free-viewing task incorporated social and non-social stimuli, and may usefully be extended to incorporate standardised emotional (negative/threat versus positive/hedonic) stimuli to provide markers that better differentiate anxiety from depression, and may differentiate anxiety subtypes. In Zhang et al’s study the majority of patients with anxiety presented with generalised anxiety disorder – consequently, evaluating markers in all anxiety disorder subtypes, and assessing associations with other symptom scales (e.g. panic, social anxiety) will confirm the sensitivity and specificity of oculomotor markers.

The healthy control group had no personal nor first-degree relative history of anxiety, though levels of HAM-A and HAM-D were not examined in controls. Therefore it remains unclear whether oculomotor markers are sensitive to variations in anxiety and depression in currently healthy individuals and/or whether they predict future risk. In that regard, studies in obsessive–compulsive disorder (OCD) have identified neurocognitive markers of risk in unaffected first-degree relatives of individuals with OCD versus healthy controls without affected relatives. ^{Reference Chamberlain, Fineberg, Menzies, Blackwell, Bullmore and Robbins4}

Zhang et al consider their findings in a dual-dimensional framework in which some markers (e.g. hyper-scanning) may reflect trait-like (risk) markers, with others (e.g. pupillary response) sensitive to current symptom severity/state anxiety. Future studies may examine whether hypervigilance markers are observed in unaffected first-degree relatives (at risk), and whether pupillary metrics better track changes in symptoms over time, and treatment response. Markers are discussed with reference to environmental monitoring and suboptimal information sampling in anxiety, consistent with oculomotor and pupillometric studies of attention, learning, effortful processing and arousal in anxiety. ^{Reference Browning, Behrens, Jocham, O’Reilly and Bishop5} However, mapping markers to mechanisms requires the careful control and manipulation of psychological task parameters – not doing so could promote markers with face validity over ethiopathogenic mechanistic validity. To that end, studies may also assess the sensitivity of markers in healthy human experimental medicine models of anxiety (e.g. pharmacological challenge, threat expectancy manipulations and social stress ^{Reference Graham, Mathew, Marsden, Smith, Smerdon and Hall6–Reference Huneke, Rowlatt, Hyde, McEwan, Maryan and Baldwin8} ) and in back-translated tasks in animal models of anxiety that can dissociate mechanisms and that are used in pharmacological treatment development. ^{Reference Grillon, Robinson, Cornwell and Ernst9}

In sum, we should continue to seek sets of markers that together provide face, construct and predictive validity – sensitive to risk, severity, mechanism and treatment response. Zhang et al’s eye-movement and pupillary metrics may have potential as behavioural markers of physiological, cognitive and emotion processing in anxiety as we look for tractable and scalable markers for use in research and practice.