Depression is one of the best-studied indications for both neurofeedback and biofeedback. Heart-rate-variability biofeedback keeps the highest evidence level in unipolar depression; EEG neurofeedback (frontal alpha asymmetry) and fMRI neurofeedback are rated efficacious in adults, one level below, once each technique is judged on its own trials. For adolescents the level is lower and rests partly on clinical judgement rather than on trials; for children it is the lowest level, because no controlled study has been done in that age group at all. None of this replaces psychotherapy or medication: it is one further option, to discuss with a clinician.
AAPB5Efficacious and specificNeuroLogic4Efficaciouslower vs AAPB
Level 4 (AAPB level assigned jointly to biofeedback and neurofeedback, and pooling EEG with fMRI protocols.) For EEG amplitude training alone the randomised evidence is Choi 2011 (n = 24, alpha-asymmetry NF at F3/F4 vs psychotherapy placebo: only the NF group increased right frontal alpha and asymmetry scores and improved on HAM-D and BDI-II; 6 of 12 NF participants vs none of the placebo group). Controlled positives without a randomised placebo add to it: Lee 2019 (treatment-resistant MDD, NF augmentation n = 12 vs medication only n = 12, superior on all outcomes, 58 % response, 50 % remission) and Wang 2019 (MDD with GAD, alpha asymmetry n = 24 / high-beta down-training n = 23 / no treatment n = 23: both NF arms reduced BDI-II and BAI-II, controls unchanged), plus open-label series (Cheon 2016: 55 % remission, 75 % response; Peeters 2014). Meta-analyses pooling EEG and fMRI are positive (Xia 2024, 22 studies, symptoms g = −0.600; Fernández-Alvarez 2022, MDD between-group g = 1.050, significant only in RCTs). Level 5 is not assigned to the EEG row on its own: no independent replication of randomised superiority over a bona fide comparator, and in Wang 2019 the asymmetry arm improved without achieving the trained EEG change. No clinician-delivered EEG amplitude trial in diagnosed patients was published in 2022-2026.
Level 1 (the AAPB does not rate neurofeedback for depression before adolescence; its Level 5 pools ages and rests on adult trials). No controlled trial of EEG amplitude training has been run in children with a diagnosed depressive disorder. The nearest evidence is adolescent (Hashemian & Sadjadi 2015, null between arms) or adult; paediatric samples appear only as comorbid subgroups inside ADHD and anxiety studies, where depressive symptoms are a secondary measure. Level 1 records the absence of a paediatric evidence base. It is not a demonstration that the method fails in children — nothing has been tested at a clinical dose, so nothing can be claimed either way.
(frontal F3 protocol, not detailed) — adolescent major depressive disorder
Ages
Adolescent (12-17)
AAPB5Efficacious and specificNeuroLogic3Probably efficaciouslower vs AAPB
Level 3 (AAPB level assigned jointly to biofeedback and neurofeedback; the AAPB's Level 5 pools ages and asks that adolescent findings be read separately from adult ones.) This level is a NeuroLogic judgement that goes beyond what the controlled adolescent trials demonstrate, and it is flagged as such deliberately. The only EEG trial is Hashemian & Sadjadi 2015: n = 28 adolescents with DSM-5 MDD, all on fluoxetine 20 mg, randomised to NF (n = 14) or placebo NF fed back from another person's EEG (n = 14), twenty 30-minute sessions over 10-12 weeks at F3; both groups improved on HAM-D with no statistical difference between groups (within-group d 0.74 for real NF vs 0.49 for placebo), and the protocol is not described. Melnikov 2021, applying the same La Vaque criteria used throughout this tool, places every EEG protocol in depression at level 2. The remaining adolescent work is single-cohort fMRI amygdala-hippocampus training (Quevedo 2019; Ahrweiler 2022, n = 34 depressed) with no clinical treatment endpoint. The 3 rests on the adult amplitude base at 4, on the demonstration that the regulation target is engageable at this age, and on the publisher's clinical appraisal — not on an adolescent trial that carries it. A reader who wants the level the published adolescent trials support on their own should read 2; the reasoning for the difference is set out in the perspective on this page.
AAPB5Efficacious and specificNeuroLogic4Efficaciouslower vs AAPB
Level 4 (AAPB level assigned jointly to biofeedback and neurofeedback; the neurofeedback half of the Level 5 rests substantially on fMRI.) Young 2017: double-blind RCT, n = 36 unmedicated MDD, two sessions of left-amygdala vs parietal-control rt-fMRI NF — symptoms fell by more than 50 % with 32 % remission, vs 8 % and 6 % in controls. Compère 2023: double-blind RCT, n = 35 (amygdala 16 / parietal 19) before 10 CBT sessions — lower depressive symptomatology and higher remission at 6 months and 1 year. Tsuchiyagaito 2023: double-blind sham-controlled RCT, n = 39, PCC-rTPJ connectivity NF — brooding reduced in the active but not the sham group. Misaki 2025, pooled analysis of the amygdala trials, n = 95: symptom reduction in the active (d = −0.704) but not the control group. Held at 4 rather than 5 because the positive RCTs come from one institute, the independent trials were null between arms (Mehler 2018, n = 32, 43 % responding and 38 % remitting overall with no group difference; Jaeckle 2023, n = 43 randomised, BDI-II −46 % in both arms) and the pooled RCT estimate is non-significant (González Méndez 2022, 4 RCTs, SMD −0.32, 95 % CI −0.73 to 0.10, low certainty).
LORETA and variants (swLORETA, source-space z-score)
Ages
Adult (18+)
Techniques
LORETA, qEEG-guided (incl. z-score)
AAPB0Not ratedNeuroLogic2Possibly efficacious
Level 2. Wu 2024: non-randomised controlled trial, n = 48 MDD with anxiety symptoms, ten one-hour swLORETA z-score sessions vs treatment as usual — lower BDI-II and BAI at post-test than controls, fewer EEG abnormalities and reduced beta current source density in prefrontal, cingulate and amygdala regions. Huang 2026: same design and group, n = 61 — reduced theta and beta z-coherence between prefrontal regions and amygdala, symptoms fell in the NF group, but the control group also improved on depression. Faridi 2024: three-arm randomised trial, n = 30 men on methadone maintenance, 20 sessions of LORETA z-score NF vs attention-bias modification vs methadone alone — larger BDI and BAI reduction with NF, sustained at 1 month, but ten per arm and blinding not described. Two non-randomised studies from one group plus one small randomised trial in a comorbid population: Level 2.
Infra-low frequency (ILF) / infra-slow (ISF)
Ages
Adult (18+)
Techniques
Infra-low frequency (ILF)
AAPB0Not ratedNeuroLogic2Possibly efficacious
Level 2. Perez 2025 (ISAD): double-blind, sham-controlled three-arm pilot RCT, n = 60 adult women with internalising disorders, six sessions of source-localised infraslow NF — clinically important HADS improvements across all arms, no active-vs-sham difference, and no differential modulation in the targeted 0.01-0.1 Hz band. One adequately controlled study in a mixed anxiety/depression sample, with a session count well below clinical practice. Equal-to-sham is not disqualifying under the framework used here, but there is no treatment-as-usual or active-comparator evidence in depression.
Level 2. Lin 2026: multicentre, assessor-blind RCT, n = 50 with mild-to-moderate post-stroke depression, fNIRS-guided NF plus art therapy integrated into CBT vs CBT alone for 12 weeks — HAMD-17 change −11.4 ± 2.5 vs −8.6 ± 2.5 (d = 1.1), remission 80 % vs 40 %, larger IL-6 and TNF-α reductions. The best-controlled neurofeedback trial in the 2022-2026 window, but the neurofeedback is bundled with art therapy so its own contribution cannot be isolated; the sample is post-stroke rather than primary MDD; no replication. Level 2 pending an independent trial that separates the components.
EEG-fMRI-pattern (EFP) neurofeedback (reward system / amygdala)
Ages
Adult (18+)
Techniques
EEG-fMRI pattern (EFP)
AAPB0Not ratedNeuroLogic2Possibly efficacious
Level 2. Amital 2025: single-arm multicentre pilot of reward-system EEG-fMRI-pattern neurofeedback (Prism device) in MDD with anhedonia — 49 screened, 34 completers (77 %), ten sessions; HDRS-17 fell by eight points on average (95 % CI −10.5 to −5.41), 78.47 % clinical improvement, 32.25 % remission, SHAPS-C −6.3 points. An adequately sized open study with well-identified outcomes but no control arm and 23 % attrition. The method is EEG-acquired but trains an fMRI-derived model, so it is shown as its own row. The foundational amygdala-EFP literature (Keynan 2019) was developed for stress and PTSD rather than depression.
Biofeedback
Heart rate variability biofeedback — unipolar depression (MDD, including depression after cardiac surgery)
Ages
Adult (18+)
Techniques
HRV — resonance-frequency breathing
AAPB5Efficacious and specificNeuroLogic5Efficacious and specificsame vs AAPB
Level 5 maintained (AAPB level assigned jointly to biofeedback and neurofeedback; on the biofeedback side every counted study is HRV.) Caldwell & Steffen 2018: RCT, 21 female students meeting MDD criteria randomised to five HRV biofeedback sessions plus psychotherapy or psychotherapy alone (CBT or ACT) — greater decreases in depressive symptoms and greater HRV improvement at 6-week follow-up; the psychotherapy-only arm improved without HRV change. Patron 2013: RCT, n = 26 with MDD after cardiac surgery, five 45-minute HRVB sessions plus usual care vs usual care — superior on RSA and CES-D, improvement correlated with RSA increase. Two independent settings with a bona fide or usual-care comparator, backed by a consistent quasi-experimental base (Karavidas 2007, n = 11 open label, about 50 % symptom reduction; Lin 2019, n = 48 MDD with insomnia, superior to standard care and maintained at 1 month; Bunthumporn 2012, n = 100, superior to usual care) and by Fernández-Alvarez 2022 (MDD between-group g = 1.050). The 2022-2026 window adds a positive waitlist pilot (Schumann 2022, n = 30) and one null active-comparator trial (Gao 2026, n = 59, vs relaxation training: both arms improved, no group difference).
Heart rate variability biofeedback — depressive symptoms secondary to a medical or psychiatric condition
Ages
Adult (18+)
Techniques
HRV — resonance-frequency breathing
AAPB5Efficacious and specificNeuroLogic4Efficaciouslower vs AAPB
Level 4 (AAPB level assigned jointly to biofeedback and neurofeedback, and pooled across unipolar and secondary depression; this row disaggregates depressive symptoms in medical or psychiatric comorbidity.) Donnelly 2023: meta-analysis of 9 trials, n = 428, HRVB vs standard treatment in comorbid adults — depressive symptoms g = 0.478 (95 % CI 0.212 to 0.743, I² = 42.9 %). Vann-Adibe 2026: 18 remote HRVB studies, n = 1352 — depression g = −0.41 (k = 10, I² = 72.6 %), robust to excluding non-randomised and high-risk studies. Yen 2023: RCT, n = 61 methamphetamine users, HRVB plus usual care vs usual care — greater reduction in depressive symptoms and better sleep at end of intervention and follow-up. Fernández-Alvarez 2022: depressive symptoms secondary to other conditions, between-group g = 0.303. Randomised superiority over usual care in independent settings meets Level 4; Level 5 is not assigned because comparators are usual care or waitlist, outcomes are self-report, and the perinatal trial was null on depression (Chen 2024, n = 86 post-caesarean).
, modality unspecified or multimodal — depressive symptoms in non-MDD samples
Ages
Adult (18+)
Techniques
Other biofeedback
AAPB0Not ratedNeuroLogic3Probably efficacious
Level 3. Wang 2024: 4-week RCT, n = 104 college students with anxiety or depression symptoms, biofeedback vs waitlist — depression improved (p = 0.001) along with anxiety, insomnia and stress. Cho 2024: RCT, 78 community adults with PHQ-9 ≥ 10 or PDSS ≥ 9 randomised to three sessions of VR-based or therapist-delivered biofeedback — MADRS fell 70.0 % and PHQ-9 64.1 % in the VR arm, no significant difference between arms, no untreated control. Birling 2026: network meta-analysis of 182 RCTs, n = 16 835, in cancer — biofeedback ranked first for depressive symptoms vs usual care (SMD −1.18), on few small trials. Chen 2025: waitlist pilot in Parkinson's disease, n = 19, mixed results. Waitlist-controlled and comparative studies in non-MDD samples without a named signal: Level 3.
By age
Adolescent (12-17)
In adolescents, neurofeedback for depression has been tested in only one small trial, in which young people receiving real neurofeedback and those receiving a look-alike version improved to the same degree, all while on an antidepressant. The other work consists of single fMRI sessions in the laboratory. Whether neurofeedback helps depressed adolescents therefore cannot be said from the trials alone; the level shown here is higher than those trials support, and reflects the clinical judgement set out in Brendan's perspective. In children no study exists at all, which is why that age group carries the lowest level.
NeuroLogic Level 3 (AAPB 5 pooled) — a publisher's judgement above the level the adolescent trials support on their own, which is 2. Hashemian & Sadjadi 2015 (n = 28, fluoxetine, 20 sessions at F3, NF = placebo NF on HAM-D); Quevedo fMRI cohort (single session, no treatment endpoint). No study below age 12.
Adult (18+)
In adults, heart-rate-variability biofeedback outperformed psychotherapy alone and usual care in two independent trials, in a handful of sessions with home practice: it is the best-demonstrated approach here. Alpha-asymmetry EEG neurofeedback has one good randomised trial and several positive controlled studies, but no independent randomised replication yet. fMRI neurofeedback shows striking results in one research centre, less clear results elsewhere, and remains out of reach of community practice.
AAPB Level 5 assigned jointly to neurofeedback and biofeedback, pooling alpha asymmetry with fMRI protocols and all HRV work. NeuroLogic disaggregates: EEG amplitude NF adult 4 (Choi 2011 plus controlled positives, no independent randomised replication), adolescent 3 (a NeuroLogic judgement above what the trials carry — Hashemian 2015 is null between arms and supports 2; see the rationale on that row), child 1 (no controlled paediatric trial exists); fMRI NF 4 (Young 2017, Compère 2023, Tsuchiyagaito 2023 from one institute; Mehler 2018 and Jaeckle 2023 null); HRV biofeedback 5 in unipolar MDD (Caldwell & Steffen 2018, Patron 2013) and 4 for depressive symptoms secondary to a medical or psychiatric condition (Donnelly 2023, Vann-Adibe 2026).
Protocols
Frontal alpha asymmetry (F3/F4 referenced to Cz, Rosenfeld protocol), alpha/theta or beta at F3/Pz; amygdala fMRI neurofeedback in research settings; resonance-frequency HRV; 10-30 neurofeedback sessions, 5-10 supervised HRV sessions with daily home practice.
Limits
33 studies in the AAPB base, many of them small; no clinician-delivered EEG amplitude trial in diagnosed patients since 2022; the positive fMRI trials come from one institute and independent trials are null between arms; fMRI neurofeedback remains a research tool; one adolescent trial only and no evidence below age 12; no EMG biofeedback study in a depressed sample despite the AAPB abstract.
Study base
33 peer-reviewed publications in the AAPB base (2023); 2022-2026 base: 36 publications indexed in the archive (26 neurofeedback, 8 HRV biofeedback, 2 other biofeedback).
Brendan's perspective
We sit at 4 where the AAPB has 5, and the reason is disaggregation rather than disagreement. Their level 5 pools EEG with fMRI and with the whole HRV literature; judged on its own trials, HRV biofeedback keeps the 5 in unipolar depression (Caldwell & Steffen 2018, Patron 2013) while EEG amplitude training holds 4 on Choi 2011 plus controlled positives, with no independent randomised replication. The adolescent row is the one I have moved, and I want to be straight about why. The published adolescent trial evidence supports a 2, not a 3: Hashemian 2015 is the only EEG study, null between arms, protocol undescribed, and Melnikov 2021 applies the same criteria we use and lands every EEG protocol in depression at 2. I have put it at 3 anyway, as a publisher's judgement rather than a reading of the trials. My reasons: the adult amplitude base is solid at 4, the adolescent fMRI work (Quevedo 2019, Ahrweiler 2022) shows the regulation target is engageable at this age even though it carries no clinical endpoint, and adolescents in my own practice respond at least as well as adults to protocols read off the qEEG. Someone who wants the number the adolescent trials carry on their own should read 2, and I would not argue with them. Children sit at 1 for the plainest of reasons: nobody has run a controlled trial of amplitude training in a depressed child, so there is nothing to support, and a level 1 says exactly that and not that the method fails. In clinic I start from frontal alpha asymmetry at F3/F4, knowing the FAA base is real but contested and its test-retest reliability questioned, and I add resonance-frequency HRV early, because an unregulated autonomic system is noise in the learning loop. What the client should be told: this is an addition to psychotherapy or medication, not a replacement — and no clinician-delivered EEG amplitude trial in diagnosed patients has been published since 2022.
Choi et al. (2011) Is alpha wave neurofeedback effective with randomized clinical trials in depression? A pilot study doi:10.1159/000322290
Young et al. (2014) Real-time fMRI neurofeedback training of amygdala activity in patients with major depressive disorder doi:10.1371/journal.pone.0088785
Young et al. (2017) Randomized clinical trial of real-time fMRI amygdala neurofeedback for major depressive disorder: Effects on symptoms and autobiographical memory recall doi:10.1176/appi.ajp.2017.16060637
Caldwell & Steffen (2018) Adding HRV biofeedback to psychotherapy increases heart rate variability and improves the treatment of major depressive disorder doi:10.1016/j.ijpsycho.2018.01.001
Patron et al. (2013) Biofeedback assisted control of respiratory sinus arrhythmia as a bio-behavioral intervention for depressive symptoms in patients after cardiac surgery: A preliminary study doi:10.1007/s10484-012-9202-5
Compère et al. (2023) Enhanced efficacy of CBT following augmentation with amygdala rtfMRI neurofeedback in depression doi:10.1016/j.jad.2023.07.063
Xia et al. (2024) Uncovering the power of neurofeedback: a meta-analysis of its effectiveness in treating major depressive disorders doi:10.1093/cercor/bhae252
Fernández-Alvarez et al. (2022) Efficacy of bio- and neurofeedback for depression: a meta-analysis doi:10.1017/S0033291721004396
Donnelly et al. (2023) A meta-analysis investigating the outcomes and correlation between heart rate variability biofeedback training on depressive symptoms and heart rate variability outcomes versus standard treatment in comorbid adult populations doi:10.23750/abm.v94i4.14305
Gao et al. (2026) Baseline Depressive Symptoms and Heart Rate Variability Indices Predict HRV Biofeedback Outcomes in Young Adults with Depression doi:10.1007/s10484-026-09785-7
Lin et al. (2026) Functional near-infrared spectroscopy-guided neurofeedback combined with art therapy and cognitive behavioral therapy for post-stroke depression: A randomized controlled trial doi:10.1016/j.jad.2026.121494
Melnikov (2021) The current evidence levels for biofeedback and neurofeedback interventions in treating depression: A narrative review doi:10.1155/2021/8878857
