Binge-eating disorder (BED; American Psychiatric Association, 2013) is the most prevalent eating disorder in adults, with a mean lifetime prevalence of up to 2.8% (Galmiche et al., 2019). BED is defined by recurrent episodes of binge eating involving eating unambiguously large amounts of food in a discrete period of time, accompanied by a sense of loss of control over eating. Regular compensatory behaviors to prevent weight gain are absent contrary to bulimia nervosa. BED co-occurs with mental disorders, including major depressive disorder, and somatic disorders such as Type 2 diabetes mellitus and essential hypertension (Udo and Grilo, 2019). Cognitive-behavioral therapy is considered as first-line treatment for BED in evidence-based international clinical guidelines (Hilbert et al., 2017). Meta-analyzes showed large-sized reductions in binge-eating episodes when compared to inactive control conditions (Hilbert et al., 2019). However, only 46–52% of patients remained abstinent from binge eating in the long term (Hilbert et al., 2020), indicating the need for further treatment optimization.

Recent evidence suggested that aberrant brain activity patterns are implicated in the development and maintenance of BED (Aviram-Friedman et al., 2018; Donnelly et al., 2018). Specifically, a systematic review on studies using electroencephalography (EEG) revealed increased fronto-central beta activity in individuals with BED and overweight or obesity in resting state and during food cue presentation relative to healthy, normal-weight controls (Blume et al., 2019). These consistent findings initiated the development of brain-directed treatment options which directly target these neural deviations in BED (Imperatori et al., 2018; Blume et al., 2022). EEG-neurofeedback (NF) is such a brain-directed treatment approach that seeks to enable individuals to gain voluntary control over their brain functions by providing them real-time feedback on their brain activity (Yucha and Montgomery, 2008), for example via visual stimuli. Building on operant learning, participants are reinforced if they regulate the signal in the desired direction (Sherlin et al., 2011). EEG-NF has shown potential in treating adults with BED (Blume et al., 2022), overweight and obesity (Fattahi et al., 2017; Leong et al., 2018), and subthreshold eating disturbances (Schmidt and Martin, 2015, 2016, 2020). Specifically, reductions in binge-eating episodes and eating disorder psychopathology were found at post-treatment and at 6-month follow-up of 12 sessions of food-specific EEG-NF, targeting fronto-central high beta activity, or functional near-infrared spectroscopy-based NF, targeting prefrontal cortex upregulation, in a recent randomized-controlled trial for patients with BED (Hilbert et al., 2023). Another randomized-controlled pilot study consistently demonstrated decreases of binge-eating episodes after 10 sessions of food-specific EEG-NF, addressing the reduction of fronto-central high beta (23–28 Hz) and theta activity, and at 3-month follow-up in patients with BED (Blume et al., 2022). These symptom improvements were accompanied by significant physiological reductions of high beta activity during resting state and food-cue presentation at posttreatment compared to pretreatment.

Despite these promising findings, the underlying mechanisms of EEG-NF are poorly understood. There is an ongoing debate about the factors which affect NF outcomes, including participants’ use of mental strategies (Thibault et al., 2015; Muñoz-Moldes and Cleeremans, 2020). Among the few available studies on the effects of participants’ individual strategy use on regulation success, Hasslinger et al. (2020) examined the nature and role of mental strategies in response to slow cortical potentials NF in n = 30 children and adolescents (9–17 years) with attention-deficit/hyperactivity disorder (ADHD) over 25 sessions. Based on semi-structured interviews conducted after each fifth session, the children’s mental strategies were grouped into four categories – cognitive, emotional, physiological, and unspecified regulation – using an inductive-deductive qualitative approach. Only patients using strategies from the cognitive (such as focusing on the task) or emotional regulation category (reaching a kind of meditative state), showed success in targeted up- and downregulation of the NF signal (i.e., physiological changes of the EEG activity) and subjective (i.e., self-reported) improvement of behavioral symptoms, while no effects were found for other strategies (Hasslinger et al., 2020).

Among healthy adults (n = 20), those who did not use any mental strategy after the first and last of 10-session sensorimotor rhythm NF examined written strategy reports (Kober et al., 2013) showed better learning slopes (i.e., greater correlation between session number and targeted sensorimotor rhythm upregulation) than those applying visual, auditory, cheering, relaxation, or concentration strategies based on self-report (Kober et al., 2013). A further study delivering 8 gamma enhancement EEG-NF sessions to n = 8 healthy women showed that concentration on the task was subjectively perceived as the most successful strategy based on written strategy reports (Khodakarami and Firoozabadi, 2020), while associations between strategy use and EEG-NF regulation success were not examined. A single session of EEG-NF on short-term memory training via alpha band upregulation in n = 32 healthy students demonstrated that participants who reported to use positive strategies (e.g., entertainment, love, family) showed stronger increases in alpha band activity compared to those using neutral (e.g., calculation, work, number) or negative (i.e., quarrel, anger, accident) strategies (Nan et al., 2012). Likewise, strategies applied during a single sensorimotor rhythm upregulation NF session differed between n = 62 individuals with a positive versus negative NF learning slope, suggesting a link between strategies and NF regulation (Autenrieth et al., 2020).

It is crucial that none of these studies investigated the use of mental strategies to downregulate high beta activity, employed food stimuli, or was concerned with treating BED. Thus, it remains to date unclear which strategies are applied by patients with BED over the course of multiple sessions of food-specific NF, and whether strategy use influences brain-based and subjective regulation success. In this context, the present study sought to explore the use of strategies in adults with BED during 12 sessions of NF in which individually salient food pictures were presented (for detail, see Hilbert et al., 2023). Based on a systematic qualitative approach, session-wise reported mental strategies were categorized, hypothesizing to identify cognitive, emotional, physiological, and unspecified strategies, in line with previous studies (Kober et al., 2013; Hasslinger et al., 2020). We aimed at determining the use of mental strategies and its relationship with objective (i.e., high beta activity during regulation beneath the baseline) and subjective (i.e., self-reported) regulation success in the EEG-NF arm, which was considered separately from the real-time functional near-infrared spectroscopy (rtfNIRS) group due to the pertinent differences in the targeted processes and because the parent study (Hilbert et al., 2023) revealed differential responses in the EEG-NF versus rtfNIRS-NF group. We expected that NF strategies would differ in modulating objective EEG-based and subjective regulation success (Autenrieth et al., 2020; Hasslinger et al., 2020).

This study was part of the single-center, assessor-blinded, randomized-controlled, three-armed feasibility study of rtfNIRS-NF for BED (NIRSBED; Hilbert et al., 2023; DRKS00014752, www.drks.de). The total sample of N = 78 individuals was randomized to 12 60-min sessions of rtfNIRS-NF, delayed rtfNIRS-NF, or high-beta EEG-NF, delivered over 8 weeks. In agreement with the Declaration of Helsinki, this project was approved by the local Ethics Committee of the University of Leipzig (474-ek). Details were reported elsewhere (Hilbert et al., 2023).

This was the first study to investigate the use of mental strategies over 12 sessions of food-specific EEG- and rtfNIRS-NF in a clinical sample of patients with BED. Our study provided evidence for the diversity of mental strategies used to self-regulate brain activity, as reflected in eight strategy categories applied during NF treatment delivered in different imaging modalities (i.e., EEG-NF, rtfNIRS-NF). Patients mostly used “Imagination,” followed by “Behavior,” and “Self-Talk” strategies. Regarding brain-based success in the EEG-NF arm, the use of “Imagination,” “Behavior” and “Thought” strategies was associated with higher EEG-based regulation success in the NF regulation and transfer tasks, defined as the intended decrease in high beta relative to baseline activity. “Emotion” strategies were, in contrast, linked to lower EEG-based regulation success, and the use of “Concentration” strategies uniquely predicted lower subjective success, without an association with brain-based regulation success.

Our study showed that patients were able to identify a wide range of potential strategies in the absence of explicit instructions, using guided discovery only. We decided to consider patients from all randomization arms, benefitting from a larger database to derive a classification system with high validity and reliability (Erlingsson and Brysiewicz, 2013), also considering its application to future studies. The eight strategy categories that were extracted based on the written strategy reports by patients across all NF arms were comparable to categories found during previous qualitative analyzes of mental strategies reported in EEG-NF studies in healthy and patient samples (Kober et al., 2013, 2017; Hasslinger et al., 2020), including imaginative, behavioral, and emotional strategies. Considerable variability in strategy use between patients and sessions (see Supplementary Figure S1) suggested the continuous use of trial-and-error learning during the NF treatment.

Regarding the content of the mental strategies, patients across NF arms commonly applied task-oriented mental strategies encompassing the depicted food stimuli, such as imagining the food in a disgusting context in the most frequently used “Imagination” strategies. Altered food-related processing patterns specific to BED (Schmidt et al., 2016; Stojek et al., 2018; Chami et al., 2019) and BED with concurrent obesity (Paslakis et al., 2017) possibly fostered the use of disorder-specific food-related strategies in our sample with BED and overweight or obesity even in the absence of disorder-relevant stimuli (i.e., when only bars were presented on the screen during EEG-NF regulation trials). Previous clinical studies similarly reported the use of disorder-specific strategies, such as strategies targeting arousal regulation in ADHD with pertinent deficits in attention and impulsivity (Hasslinger et al., 2020); kinesthetic strategies in Parkinson’s disease with pertinent motion deficits (Bichsel et al., 2021); or food-related strategies in overweight and obesity (Spetter et al., 2017; Kohl et al., 2019). Accordingly, a study in healthy participants interpreted the lack of differences in the mental strategies used for various types of feedback signals (i.e., bars vs. worm avatars) in terms of the lacking match with participants’ preferences (Autenrieth et al., 2020). Importantly though, patient populations, as opposed to healthy participants, likely have a higher intrinsic motivation to perform better during NF treatment (Haugg et al., 2021) due to their goal to alleviate disease-specific symptoms (Blume et al., 2020). Indeed, the high burden of suffering of patients with BED relative to healthy participants may have increased their motivation to perform well during NF. This increased motivation may in turn have stimulated patients’ efforts to select a strategy targeting their pertinent BED symptoms. Alternatively, the “Imagination” strategies used by our patients may benefit their urges to eat binge foods through cognitive reappraisal, as binge foods were mostly imagined in ways that stressed their undesirability. This conjecture was supported by the demonstrated link between similar imaginative strategies encompassing reappraisal of the depicted foods and improved regulation of the urge to ingest craved foods in healthy participants (Giuliani et al., 2013). Regarding the use of mental strategies between groups, there are differences in the strategies used not only between EEG-NF and the rtfNIRS-NF group, but also between rtfNIRS-NF and delayed rtfNIRS-NF (see Supplementary Table S2). Thus, it is unlikely that neurocognitive mechanisms are the only factor with an influence on strategy use. Future research assessing the association between the target process and the emergence of mental strategies remains therefore imperative. Although studies highlighted the possible advantages of combining different modalities (Li et al., 2022), future research on the combination of these methods in eating disorders remains yet outstanding.

Turning to brain-based regulation success in the EEG-NF arm, high beta activity at its core has been identified to correlate with high performance and cognitive processing, but also rumination, overthinking, and worry (Demos, 2005). Regarding eating disorders, high beta activity was previously implicated as a neuronal marker for increased awareness of food cues in individuals with BED and with obesity (Blume et al., 2019; Hiluy et al., 2021). Increased fronto-central beta activity relative to individuals with normal weight correlated with eating disorder psychopathology in individuals with BED, while individuals with obesity also showed increased beta activity in the resting state. The consistently reported increased beta activity in fronto-temporal regions aligned with functional magnetic resonance imaging (fMRI)-based reports of alterations in the prefrontal cortex in BED and obesity (Donnelly et al., 2018; Mele et al., 2020). Most recently, studies on fNIRS in patients with BED and obesity suggest prefrontal hypoactivation in response to food stimuli when compared to normal weight controls (Rösch et al., 2021; Veit et al., 2021). Although these findings indicated dysfunctionalities in a brain network dedicated towards food-cue awareness and attentional bias, especially regarding the balance between inhibition and disinhibition, the exact interplay between EEG-based findings and the underlying brain activity assessed via fMRI or fNIRS has not yet been evaluated (Blume et al., 2019; Hiluy et al., 2021). Exploratory studies on the interplay of beta activity and neuronal activity linked beta activity to gamma-aminobutyric acid (GABA)-mediated inhibitory processes (Jensen et al., 2005; Yamawaki et al., 2008; Hall et al., 2011; Baumgarten et al., 2016). Most recently, a study on coordinated looking and reaching in non-human primates showed beta frequency to positively correlate with neural mechanisms responsible for inhibiting natural behavior such as coordinated movement (Hagan and Pesaran, 2022). Regarding the connection with BED and disordered eating, the modulation of GABA action has been mentioned as a potential target for future studies (Mele et al., 2020), but overall, research on the neural response to food stimuli is rare (Chae and Lee, 2023).

Regarding the association between strategy use and brain-based regulation success in the EEG-NF arm, “Emotion” and “Self-Talk” strategies were related to lower EEG-based regulation success. Verbal strategies, such as “Self-Talk” strategies, likely recruit substantial cognitive resources by requiring a level of awareness (Bastian et al., 2017) that hinders the integration of attentional and introspective processes (Alderson-Day and Fernyhough, 2015). Thus, “Emotion” and “Self-Talk” strategies may be regarded as explicit mental strategies which possibly induce cognitive overload (Kober et al., 2013) and thereby prevent the integration of various introspective processes necessary for NF success (Bagdasaryan & Le Van Quyen, 2013). Improved NF learning, defined as a positive correlation between session number and higher NF regulation success, was in contrast speculatively linked to a high level of automated and therefore subconscious control (Wood et al., 2014). Indeed, “Thought” strategies, subsuming cognitive attempts for free mind wandering which likely indicate a more automatic regulation, and “Behavior” strategies which presumably require little mental effort, emerged as positive predictors for greater EEG-based NF regulation success in healthy adults (Hardman et al., 1997; Kober et al., 2017). In summary, the search for an explicit mental strategy may induce cognitive overload (Khodakarami and Firoozabadi, 2020) and thereby hamper NF regulation success, whereas the absence of conscious strategy access potentially promoted internalization (Strehl, 2014).

Finally, recording subjective success via self-reports after each session may itself have kept patients motivated to reach their pre-defined goal of minimizing bar sizes during the 12 EEG-NF sessions (Abele and Spurk, 2009). The reporting of freely chosen strategies after each of the 12 NF sessions is unique to the present study, whereas previous studies only asked patients to report their strategies after the first and last of ten EEG-NF sessions (Kober et al., 2013). This thorough assessment of the mental strategies used possibly fostered patients’ continuous reflection on their own strategy use and thereby enhanced their motivation in the present study. Nevertheless, subjective success did not translate to changes in presumed underlying brain activity patterns in the EEG-NF arm, as the strategies that were most successful in terms of larger EEG-NF regulation success were not subjectively perceived as the most successful. This present finding stands in contrast to a previous study, showing that patients with epilepsy who successfully controlled their slow cortical potentials were able to sense their success after 30 of 35 EEG-NF sessions with an 8-week practice phase in between (Kotchoubey et al., 2002). Thus, more treatment sessions may be required before patients are able to correctly estimate the success of certain strategies. In light of limited evidence, the differences in objective, i.e., brain-based, and subjective regulation success and their association with the use of mental strategies should be explored in future studies.

To our knowledge, no previous study examined the mental strategies used by patients with BED to self-regulate food-specific brain activity in NF treatment. We uniquely developed a new and fine-grained classification system via qualitative content analysis, based on written strategy reports provided after each session of EEG- or rtfNIRS-NF. Written self-report is the most common and feasible method to record mental strategies after NF (Nan et al., 2012; Kober et al., 2013, 2017). Assessment via interview is also possible (Hasslinger et al., 2020), however, the written strategy reports limited a potential interviewer bias on the participant (Davelaar et al., 2018) which likely affected previous coding systems (Hasslinger et al., 2020), including the induction of a good-subject bias (Nichols and Maner, 2008). Importantly, every single strategy report was considered carefully, and categories only emerged after several steps of analysis and consensus within the research team. Previous strategy classification approaches, on the contrary, were based on the experimenter’s subjectively determined categories (Nan et al., 2012; Kober et al., 2013), or made use of existing classification systems developed in healthy samples (Kober et al., 2017; Autenrieth et al., 2020). In addition, instead of solely relying on consensus between the raters (Hasslinger et al., 2020), raters in the present study independently assigned strategies to the preformed categories before reaching a consensus. High inter-rater reliability supported the validity of our new classification system, making it well-suited to categorize the strategies used by patients with BED for food-specific NF self-regulation. A further strength of this study was its analytical approach, ensuring robustness in small samples (Luke, 2017) via the use of linear mixed models, while comprehensively including all random effects in accordance with recommendations (Barr et al., 2013) in order to account for non-independent data (Brauer and Curtin, 2018).

A possible limitation was that we could not trace back strategies to single trials, as all strategies used during one session were reported at the end of the NF session. Consequently, the percentage of time a strategy was used during one session could not be determined, resulting in a potential over- or underestimation of the strategies’ relevance. It remains unknown whether patients used a combination of strategies rather than a single strategy to modulate brain activity during each trial. This assessment of strategies via self-report after the end of a NF session may have introduced order effects in the form of primacy or recency effects (i.e., patients may have remembered their first or last mental strategies; Murdock, 1962) and errors in reporting. In addition, patients may not have remembered all strategies they used during the 12 NF trials. While strategy assessment via interview would also have been subject to both of these effects and despite the previously discussed pertinent advantages of recording subjective success via self-report, we cannot rule out that other strategies may have been spontaneously verbalized if strategies were assessed through an interview (e.g., Hasslinger et al., 2020). In addition, qualitative data analysis is inherently subject to experimenter’s subjective decisions, although the consensus discussions and the assessment of inter-rater reliability aimed to increase generalizability of the present findings. In this context, it deserves mention that the current classification of patients’ mental strategies was based on the full sample of patients who underwent the NF procedure, irrespective of the processes targeted through each specific imaging modality. Importantly though, the emergence of strategies may depend on the underlying process even if the provided feedback was similar across the processes (through individually appetitive food cues in our study). Future studies are therefore needed to disentangle the degree to which the emergence of strategies is affected by the processes targeted through the imaging modality, even if feedback is given through a similar brain-computer interface. Another limitation was the implicit data analytic assumption that the association between strategy use and brain-based or subjective regulation success was the same across participant. In fact, previous research emphasized that strategies which were successful on the group level may not be successful on the participant level and vice versa (Barth et al., 2016). Regarding the sample, we cannot rule out for that patients’ comorbidities and subsequently, their medication, had an influence on the selection and use of mental strategies. However, since BED is an affliction with a high potential for comorbid somatic conditions (Udo and Grilo, 2019), this was taken into account during patient selection: The use of medication such as antidiabetics, which might have an influence on eating behavior, weight and/or executive function and therefore, on the use of mental strategies was only permissible after a minimum duration and if the dosage had been stable for several months (see Supplementary material of Hilbert et al., 2023), to minimize effects on the use of mental strategies. Future explorative analyzes should consider the potential influence of patients’ characteristics, in particular the role of concurrent weight status and sex on mental strategy use and on the link between subjective and brain-based success. Lastly, the current study focused on patients randomized to EEG-NF, with likely differential links between strategy use and EEG- versus rtfNIRS-based regulation success reported elsewhere (Rösch et al. (2023) “Mechanisms underlying fNIRS-neurofeedback over the prefrontal cortex for participants with binge-eating disorder”). Future studies may harness the pertinent advantages of both imagining modalities in combined applications to provide more comprehensive information on the functional activity of the brain through the concurrent assessment of neuronal electricity activity (EEG-NF) and metabolic response (fNIRS-NF; Li et al., 2022).

In summary, our study extended previous research on the use of mental strategies throughout NF treatment by firstly deriving a classification system for mental strategies used in response to individually appetitive food cues in adult patients with BED. The classification system provided valuable information about efficient strategies targeting BED-specific processing of individually appetitive food stimuli. Based on our findings, future clinical NF studies could examine whether explicit instructions of imaginative strategies, which were linked to better EEG regulation success, were likewise associated with favorable clinical outcomes (Herwig et al., 2019). Strategies which were linked to greater brain-based regulation success were more frequently used than the strategies that prove herein less successful, in line with the presumed operant conditioning principle underlying NF learning. Successful strategies were likely internalized due to the continuous reinforcement during the NF sessions, and then shown in an automatic manner not requiring conscious attention. Importantly though, the differential link between the mental strategies used and objective and subjective success warrants replication in future research, which should ideally also assess whether brain-based and subjective success translate to clinical symptom improvements.

The raw data supporting the conclusions of this article will be made available upon request to the corresponding author.

The studies involving humans were approved by Ethics Committee of the University of Leipzig (474-ek). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

AH and RS: conceptualization, methodology, software, and project administration. SR and RS: validation and investigation. JW and SR: formal analysis and writing – original draft preparation. AH: resources, supervision, and funding acquisition. JW, RS, and SR: data curation. AH, JW, RS, and SR: writing – review and editing. SR: visualization. All authors contributed to the article and approved the submitted version.