EEG-based Neurofeedback to Improve Emotion Regulation in Major Depressive Disorder: A Randomized Clinical Trial

The goal of this clinical trial is to evaluate whether EEG-based neurofeedback targeting the emotion regulation network through swLORETA can improve emotional regulation and reduce symptoms in adults with Major Depressive Disorder (MDD) who have not responded sufficiently to first-line treatments. The main questions it aims to answer are: * Does EEG-neurofeedback improve emotional self-regulation and reduce clinical symptoms in patients with MDD with or without anxiety symptoms? * Are changes in EEG resting-state activity and stress biomarkers (e.g., cortisol) associated with clinical improvement? Researchers will compare an active neurofeedback group, a sham (placebo) neurofeedback group, and a treatment-as-usual control group to see if real-time EEG-neurofeedback leads to greater improvement in mood, emotional regulation, and neurophysiological indicators than placebo or no additional intervention. Participants will: * Receive 10 sessions of either real or sham EEG-neurofeedback (or no sessions in the control group) over 5 weeks. * Complete clinical, psychological, and neurophysiological assessments before (week 0) and after the intervention (week 6). * Provide repeated saliva samples to assess stress-related biomarkers at week 0 and week 6. * Continue their standard pharmacological treatment throughout the study.

This is a single-blind, randomized, placebo-controlled clinical trial designed to evaluate the efficacy of EEG-based neurofeedback (EEG-NF) as a complementary intervention for individuals diagnosed with Major Depressive Disorder (MDD) who have shown an insufficient response to first-line pharmacological treatments. Seventy-two adult participants aged 18 to 65 with a confirmed diagnosis of MDD and a minimum threshold score on a validated depression severity scale will be recruited through psychiatric services. Participants will be randomized into three groups: an active neurofeedback group, a sham neurofeedback group (receiving non-contingent feedback), and a treatment-as-usual control group. The intervention is based on swLORETA Z-score neurofeedback using the NeuroGuide® platform, targeting key regions within the fronto-limbic circuit, including the anterior cingulate cortex, medial prefrontal cortex, amygdala, insula, posterior cingulate cortex, precuneus, habenula, and nucleus accumbens. These areas are implicated in emotion regulation and have shown altered activity and connectivity in individuals with MDD. The EEG will be acquired using a 24-channel EEG system (eego™, ANT Neuro) with electrodes placed according to the 10-20 international system, and real-time feedback will be provided via interactive gamified visual interfaces. Participants in the active neurofeedback group will undergo 10 training sessions over five weeks (two sessions per week), each consisting of a baseline resting EEG, five blocks of training based on operant conditioning (reinforcing reductions in abnormal Z-scores), a transfer task to assess generalization, and a post-training EEG resting-state with eyes closed. The sham group will follow an identical structure but receive feedback from another participant's EEG, ensuring the participant receives non-contingent, placebo-controlled feedback. EEG data will be preprocessed in real time, and feedback will be calculated based on frequency-specific parameters (delta, theta, alpha, beta, high beta) across power, coherence, and phase metrics. Participants will continue taking their usual medication during the trial. In addition to EEG, salivary cortisol will be measured as a neuroendocrine biomarker of stress regulation, collecting saliva with Salivette® kits at awakening, +30 minutes, +60 minutes, 10:00h, and 21:00h, on two consecutive days at baseline and post-treatment. The data will be used to calculate the cortisol awakening response (CAR) and diurnal slope following established guidelines. EEG recordings at baseline and after the intervention will be analyzed to assess changes in power spectra, connectivity, and network-level dynamics. In parallel, machine learning classifiers will be trained on vectorized EEG features to identify neurophysiological signatures of response to the intervention, using 80/20 train-test splits with 50 repeated iterations to ensure generalizability. Recursive Feature Elimination (RFE) will be employed to rank and select the most informative features, which will also be examined as potential predictors of clinical response. The study complies with the CRED-nf checklist for neurofeedback research, incorporates rigorous sham control and randomization procedures, and adheres to ethical standards per the Declaration of Helsinki and European regulations for low-risk, non-invasive clinical trials. EEG acquisition systems and analysis software (NeuroGuide and eego™) are certified medical devices under FDA (510(k): K041263) and CE Class IIa regulations. Participants are blinded to their group allocation, and informed consent is obtained before enrollment. All data will be pseudo-anonymized and stored securely on REDCap (Research Electronic Data Capture), ensuring traceability and compliance with data protection legislation.