Introduction

The electroencephalogram (EEG) shows promising clinical utility in being non-invasive, easy to record and cost-effective compared to other neuroimaging methods (Hajcak et al., 2019). Event-related potentials (ERPs) on the EEG are indexes of task-related brain activity time-locked to stimuli or other events and potential biomarkers of mental disorders (Hajcak et al., 2019). Several ERPs have been associated with diagnostic categories and severity of diagnosis-specific symptoms (Luck & Kappenman, 2011). However, for the emotional disorders, one of the main causes of suffering worldwide, ERP findings remain inconsistent (González-Robles et al., 2018; Watson et al., 2022; Widiger & Oltmanns, 2017). While differences between ERP studies in paradigm design and preprocessing and analysis methods limit comparison of some results, discrepancies also stem from issues inherent in categorical taxonomies and which are especially abundant in the emotional disorders (Michelini et al., 2021). Recent evidence indicate that at least some ERPs are more closely related to transdiagnostic measures of psychopathology than to diagnostic categories (Donaldson et al., 2020; Macedo et al., 2021; Pasion & Barbosa, 2019; Riesel et al., 2022). Given this, a comprehensive investigation of the associations between classic ERPs, of which some were discovered in the 1960’s, and transdiagnostic measures of psychopathology is called for (Latzman & DeYoung, 2020; Polich, 2020). In this study, we aim to do so by assessing a mixed sample for correlations between ERPs and transdiagnostic measures of psychopathology while accounting for the effects of medication and psychotherapy treatment.
Biomarkers in psychiatry could greatly improve clinical practice in providing objective measures of psychopathology and treatment outcome (Singh & Rose, 2009). The discovery of such markers requires a sound psychopathology framework from which to derive biobehavioral targets to examine (Latzman & DeYoung, 2020; Michelini et al., 2021). ERP studies have traditionally evaluated differences in ERP measures such as peak or average amplitude or latency between diagnostic groups based on categorical taxonomies ICD and DSM (DSM-IV-TR., 2000; WHO, 2004). These taxonomies posit that mental disorders are discrete entities with specific symptoms and clear-cut boundaries between healthy and ill and between diagnoses (Clark et al., 2017). However, clinical reality shows that comorbidity among the emotional disorders is in the range of 40 to 80% and that symptom profiles of patients with the same diagnosis varies greatly (González-Robles et al., 2018). This suggests that categorical taxonomies do not capture the true nature of psychopathology (Clark et al., 2017; González-Robles et al., 2018). There being no straightforward way to account for comorbidity in case-control designs, most ERP studies merely report concurrent diagnoses and rely on the primary diagnosis as a sufficient classification of the sample (Petrolini & Vicente, 2022; Zald & Lahey, 2017). To see why this can be problematic, consider the error-related negativity (ERN) which is robustly enhanced (increased amplitude) in some anxiety disorders such as obsessive-compulsive disorder (OCD) compared to healthy comparison subjects (Macedo et al., 2021). Somewhat inconsistent results indicate that ERN is attenuated (decreased amplitude) in depression (Klawohn et al., 2020). It is clear to see how a study examining ERN in depression while not accounting for anxiety-related comorbidity might end up with null results. Heterogeneity within disorders and arbitrary boundaries between healthy and ill pose similar loss-of-information problems in studies based on categorical taxonomies (Michelini et al., 2021). In fact, with the notable exception of ERN in OCD, decades of research has revealed no robust deviations in ERPs or other EEG measures in any of the emotional disorders as defined in the categorical taxonomies, e.g., depression or major depressive disorder (MDD) (de Aguiar Neto & Rosa, 2019), generalized anxiety disorder (GAD) (Maron & Nutt, 2022), panic disorder (PD) (Howe et al., 2014) and social anxiety disorder (SAD) (Al-Ezzi et al., 2020).
Alternative frameworks of psychopathology transcends arbitrary diagnostic boundaries in considering transdiagnostic symptoms which are shared among disorders as the basic building blocks of mental disorders (Clark et al., 2017). The Hierarchical Taxonomy of Psychopathology (HiTOP) is an empirical and data-driven attempt to describe the full range of psychopathology (Kotov et al., 2017). In the HiTOP, transdiagnostic symptoms and maladaptive traits at the lowest level of the hierarchy are clustered based on shared features into subfactors roughly corresponding to categorical diagnoses, which in turn are joined into higher-level spectra such as the Internalizing and Thought disorder spectra. Accordingly, the emotional disorders share core symptoms and traits but are further up in the hierarchy allocated to the Fearand Distress subfactors, the latter containing depression and GAD. The HiTOP comes with several advantages for neuroimaging research (Conway et al., 2022; Corr & Mobbs, 2023; Kotov et al., 2022; Latzman & DeYoung, 2020; Michelini et al., 2021; Perkins et al., 2019). By design, the HiTOP deals with comorbidity, symptom heterogeneity within disorders and arbitrary boundaries between healthy and ill. In contrast to categorical taxonomies, the HiTOP encourages studies of mixed samples characterized at different levels of the hierarchy capturing the full range of psychopathology (Conway et al., 2022; Latzman & DeYoung, 2020). In other words, subjects included in a study based on the HiTOP do not need to fulfill some diagnostic criteria or score above some threshold, but are fully characterized in terms of homogeneous dimensional constructs. A given ERP measure can thereby be investigated in terms of being a marker of a transdiagnostic symptom or trait, of a subfactor or of a whole spectrum. Relying on the HiTOP is also advantageous when selecting biobehavioral targets with which associations to ERPs are sought. For transdiagnostic measures, rather than using sub scales of rating scales developed in categorical setting, measures consistent with the HiTOP would directly place results in the context of a comprehensive empirical model of mental disorders (Perkins et al., 2019).
Ample evidence support that biological measures align more closely to transdiagnostic constructs than to diagnostic categories (Kotov et al., 2020; Waszczuk et al., 2020; Watson et al., 2022). The Research Domain Criteria (RDoC) was launched to encourage research into such biobehavioral constructs cutting across diagnostic boundaries (Cuthbert & Insel, 2010). In line with this, several ERPs have recently been recast as markers of transdiagnostic psychopathology. ERN and its counter-part, the correct-related negativity (CRN), previously solely associated with OCD, are now conceived as markers of specific transdiagnostic measures of anxiety and negative affect in the Internalizing spectrum (Macedo et al., 2021; Pasion & Barbosa, 2019; Riesel et al., 2022). Mismatch negativity (MMN) and some other ERP components in auditory oddball paradigms do not seem to be uniquely related to a chronic diagnosis of schizophrenia but to symptoms shared by a range of psychotic disorders (Donaldson et al., 2020; Parker et al., 2021). These developments being very recent, only the ERN, CRN, and to a lesser extent the late-positive potential, have thus far been cast in the light of transdiagnostic psychopathology in the emotional disorders (Granros, 2021). A comprehensive investigation of the associations between other classic ERPs and transdiagnostic markers of psychopathology in the emotional disorders is lacking. Conversely, further validation of the HiTOP with biological measures is called for (Perkins et al., 2019).
The aim of the present study was to examine the associations between a set of transdiagnostic measures of psychopathology and a range of ERPs elicited by thee classic paradigms (the Eriksen Flanker, the auditory Attended Oddball and the auditory Unattended Oddball) (Luck & Kappenman, 2011). Measures of transdiagnostic psychopathology were assessed with validated self-report measures covering symptoms and traits consistent with the HiTOP Internalizing spectra. We included 50 patients with emotional disorders undergoing 14 weeks of UP transdiagnostic group cognitive behavioral psychotherapy and 37 healthy comparison subjects (HC) matched in age and sex (Barlow et al., 2017; Reinholt et al., 2021). Patients were assessed with EEG and self-report questionnaires three times: before, 10 weeks into, and within one week after treatment. The majority of HCs were assessed once but some a second time after at least two months in order to account for normal variation in the models.
To evaluate the associations between ERPs and measures of transdiagnostic psychopathology, we conducted robust mass univariate linear regression based on single-trial ERP analysis as implemented in the EEGLAB toolbox LIMO EEG (Delorme & Makeig, 2004; C. R. Pernet et al., 2011). LIMO EEG is based on statistical parametric mapping (SPM), as in the analysis of fMRI data, and provides a complete workflow from preprocessed EEG data to the evaluation of single-trial subject-level ERPs at group level with a range of robust statistical measures (Kiebel & Friston, 2004; C. R. Pernet et al., 2021). Employing a hierarchical generalized linear model (GLM) approach, the method makes redundant several choices required in traditional ERP methods known to inflate false positives and influence group level statistics (Feuerriegel & Bode, 2022; Luck & Gaspelin, 2017). Instead of requiring the á priori selection of channel and time window regions of interest, as well as methods for peak or average amplitude extraction, LIMO EEG models the subject-level single-trial GLM across all channels and time points concurrently. False positives are controlled through bootstrap methods and threshold-free cluster enhancement (TFCE) (Maris & Oostenveld, 2007; Mensen & Khatami, 2013; C. R. Pernet, 2015). Consequently, the investigate scope is vastly expanded without loss of statistical power and can reveal effects at other channels and time periods than what is traditionally investigated (Fields & Kuperberg, 2020). Recognizing current issues in the preprocessing of ERP data, we relied on a novel cleaning pipeline based on an empirical evaluation of other well-established pipelines, the Reduction of Electroencephalographic Artifacts (RELAX) Bailey et al. (2022); Bailey et al. (2023). Given evidence that robust single-trial methods allows for less aggressive cleaning of ERP data, thereby preserving more brain activity, we applied a less strict than default cleaning of artifacts and noise (Alday & van Paridon, 2021; Delorme, 2022).
Establishing associations between measures of transdiagnostic psychopathology and ERPs, many of which are related to specific neural functioning, would be an important step toward biomarkers in psychiatry and would increase our understanding of the neural basis of mental disorders (Hajcak et al., 2019; Lavoie et al., 2019).
Given the exploratory nature of the study, we refrain from making specific hypotheses. However, as found in two recent studies, we expected the ERN to be related to one or more measures in the Internalizing spectrum (Macedo et al., 2021; Riesel et al., 2022).