Neurosurgical treatment can be offered only in a minority of patients with drug resistant epilepsy (DRE) (around 20%) [1], and is effective in about 50% (ranging from 36 to 93% according to the localization and the etiology of the epilepsy) of cases [2]. Therefore, a large population is either not eligible or presents a pathology that is insufficiently improved by neurosurgery.

In this population, other palliative therapeutic strategies through neuromodulation (vagal nerve stimulation, VNS; deep brain stimulation, DBS; responsive neurostimulation, RNS; or transcranial direct current stimulation, tDCS) can be proposed to reduce seizure frequency and comorbidities linked to epilepsy [3], [4], [5]. Most of these techniques are invasive procedures with intrinsic surgical risk and are not accessible to all patients.

Behavioral therapies (cognitive behavioral therapy, mindfulness, self-management, psycho-educational intervention and depression intervention) are possible adjunctive treatments for epilepsy. These methods have few side effects, minimal cost and may be effective options for example during pregnancy. The current literature suggests that these approaches may lead to a reduction in seizure frequency, improvement of quality of life and psychiatric comorbidities such as anxiety and depression in epileptic patients [6], [7]. The International League Against Epilepsy Psychology Task Force highly recommends incorporating them into holistic epilepsy care [8]. Furthermore, behavioral therapies give the patient an active tool to increase self-efficacy in seizure management. Some epileptic patients report to adopt certain cognitive, emotional, and behavioral strategies to reduce seizure intensity and even to prevent seizure occurrence after such psychological intervention [9], [10], [11].

Electroencephalographic neurofeedback (EEG-NFB) is a biofeedback procedure measuring an electrophysiological activity using a brain-computer interface (BCI) in order to extract a neurophysiological signal of interest that is presented in real-time to the participant. As such, it may be considered to be a unique therapy involving both neural and behavioral modulation at the same time [12]. This procedure is usually employed to train a specific brain activity. The information from the EEG recording is usually displayed on a computer screen in a visual way through BCI in real time (note that other sensory modalities could be used). A closed loop of information is therefore formed with the patient's brain activity and the feedback that is displayed on the screen. This feedback loop typically consists in decomposing and processing the EEG signal and consequently presenting it in real time. The biomarker of interest may be based on classic spectral analysis, with broadband power spectral density as a feature to be employed as a feedback, or, with a more complex approach, transformation of the extracted EEG signal into one specific frequency band of interest serving as a biomarker. Such a signal may also be used for coherence or connectivity analysis or their combination [13]. The selected feature, or biomarker, is then presented to participants through the feedback displayed on the screen, enabling them to modify their neural activity to achieve the desired neural state. The objective of the task is determined a priori and the patient has immediate feedback regarding the accomplishment of the objective. The self-control of the electroencephalographic signals in real time is achieved volitionally through learning, by trial and error trial [14]. For example, patients can learn through reward by successfully moving an object in space like a spaceship in a game or increasing the value of a thermometer. Through this process of trial and error, patients develop strategies to modify the EEG feature (Fig. 1).

In clinical settings, the training aims to facilitate self-regulation of the putative neural substrates underlying a specific pathology [13]. In contrast to other neuromodulation methods such as DBS, VNS or tDCS, NFB requires an active engagement of patients who must constantly apply and adapt their strategies to alter brain activity in the intended procedure.

Thus, NFB has been used to modulate functional networks in the treatment of various neurological and psychiatric pathologies [15], [16], [17], [18]. Biofeedback and NFB have been proposed as alternative adjunctive therapies for drug-resistant epilepsy over the past fifty years [19], [20], [21], [22], [23].

Two independent meta-analyses regarding the use of NFB in the treatment of epilepsy provided evidence of its efficacy in the reduction of seizure frequency (Tan et al., 2009) (Nigro, 2019). However, there are strong methodological limitations in the studies reviewed in these meta-analyses. Indeed, most summarized studies were uncontrolled, non-blinded and non-randomized.

Additionally, up to 60% pf patients with DRE frequently suffer from cognitive and psychiatric comorbidities [24], [25]. A recent review reported the efficacy of NFB training in major depressive disorder [26]. Furthermore, research studies have shown promising efficacy of NFB in emotional regulation [27], [28].

A better understanding of the mechanisms through which NFB operates, the progress in neuroimaging and BCI, and a more widespread network-based approach to phenotype pathological brain activities have led to a renewal of interest in this method [13], [18], [29]. Identifying the factors influencing learning and self-regulation of brain activity, as well as a better targeting of neurophysiological variables reflecting pathological changes in functional brain networks, might help to develop rigorous protocols and optimize the potential effects of NFB trainings.

In this narrative review we aim to:

present an overview of the physiological rationale of NFB and biofeedback as neuromodulation techniques and the different protocols used in epilepsy;

discuss specific and non-specific factors underlying NFB efficacy in terms of seizure reduction and improvement in comorbidities;

discuss the limitations of available studies and how to improve future protocols.