SMR neurofeedback is one of the most-studied historical applications, with reduced seizure frequency in drug-resistant patients; slow cortical potential training was replicated by an independent group with effects persisting ten years. It never replaces antiepileptic medication.
Updated :
What the research shows
EEG neurofeedback — standard amplitude training
(SMR and slow cortical potentials) — seizure reduction
AAPB4EfficaciousNeuroLogic4Efficacioussame vs AAPB
Level 4 maintained (AAPB level printed jointly for SMR-based and SCP-based protocols; ages pooled, see control-quality note). Two meta-analyses in medically refractory epilepsy: Sterman 2000 (24 studies, 243 patients, 82 % with a seizure reduction of at least 50 %) and Tan 2009 (10 studies, 79 % of SMR-trained patients with a statistically significant reduction in seizure frequency). One randomised controlled trial (Lantz & Sterman 1988, n = 24, refractory focal seizures) found significant reductions from baseline in the neurofeedback arm compared with non-contingent feedback and no treatment. SCP training was replicated by an independent group: Kotchoubey 2001 (controlled outcome study, n = 64; about two thirds of patients who learn to control their SCPs reduce their seizures), with persistence up to ten years (Strehl 2014, n = 23). Technique note: SMR carries the randomised evidence; SCP carries the independent replication and the long-term follow-up, and Fumuro 2025 (n = 12, 35 supervised sessions, uncontrolled) again ties seizure reduction to acquired SCP control. Not Level 5: the only sham-controlled randomised trial with a seizure outcome (Morales-Quezada 2019, adolescents already controlled on medication) showed no seizure-frequency effect, so superiority over a credible sham in two independent settings is not established. No clinical-fidelity randomised seizure trial has appeared since 2022.
SMR-based neurofeedback — nonseizure manifestations of epilepsy (cognition, quality of life)
AAPB3Probably efficaciousNeuroLogic3Probably efficacioussame vs AAPB
Level 3 maintained (ages pooled). Morales-Quezada 2019 (n = 44, aged 10-18, focal epilepsy controlled by medication; randomised, double-blind, sham-controlled, exploratory): the SMR arm improved significantly on letter-category reaction time but not on the other attention-switching measures; all three arms (SMR, SCP, sham) improved on seizure-related quality of life; no effect on seizure frequency. Cheng 2024 (n = 36 children with idiopathic epilepsy syndromes and comorbid ADHD, retrospective, EEG biofeedback vs no training, protocol not stated in the abstract): improvement in forward and backward digit span only, no change in the other cognitive domains tested, no seizure outcome. One exploratory randomised trial and one retrospective controlled comparison, from different groups, with partly convergent cognitive gains: probably efficacious, not efficacious.
AAPB1Not empirically supportedNeuroLogic1Not empirically supportedsame vs AAPB
Level 1 maintained. LORETA z-score neurofeedback in epilepsy rests on two case series from the same group: Frey & Koberda 2015 (n = 6, default-mode-network training in drug-refractory patients, notable seizure reduction in 5 of 6) and Koberda & Frey 2015 (n = 10, some improvement in every patient). AAPB states that controlled studies have not yet been done; nothing in the 2022-2026 window changes this. Consistent with the Level 1 given to LORETA for ADHD.
AAPB1Not empirically supportedNeuroLogic1Not empirically supportedsame vs AAPB
Level 1 maintained. Walker & Kozlowski 2005: 10 consecutive patients trained to normalise qEEG power and coherence, 9 of 10 seizure-free — a case series. Middlebrooks 2016, a conference abstract only: two-channel bivariate versus four-channel multivariate coherence training in children with autism and seizures, at least 14 sessions, interictal spiking reduced by 56 % and 81 % respectively, with no clinical seizure outcome. AAPB is explicit that no controlled study exists, and an unpublished abstract cannot carry a level; no 2022-2026 record.
Level 1. AAPB did not rate infra-low frequency training. The only published record is a single case: Schmidt 2023, an eight-year-old with Dravet syndrome trained for 2.5 years with no significant medication change, reporting reduced seizure frequency and severity, improved sleep and reversal of neurodevelopmental decline. Glaubig 2026 (narrative review of paediatric slow-oscillation neurofeedback) classes ILF and ISF as still emerging and requiring further validation, and cites an earlier ISF paediatric epilepsy case series (Legarda 2011, to verify). Case-level evidence only; the row makes the absence of controlled evidence visible where ILF is actively marketed for paediatric epilepsy.
Level 2, pending verification. Peripheral biofeedback was not rated by AAPB. The rationale for electrodermal biofeedback is Nagai's work in drug-resistant epilepsy — a controlled study (Nagai 2004) and a later pilot randomised trial (Nagai 2019) reporting reduced seizure frequency after sympathetic-arousal training; neither is in the sweep or AAPB, so no figures are quoted and both references are marked to verify. Schach 2022 (n = 30, quasi-randomised to arousal, sham or relaxation EDA biofeedback, one 30-minute session under continuous EEG) states that there is evidence that EDA biofeedback can reduce seizure frequency, but is itself a mechanistic study: only arousal training produced a prolonged robustness-enhancing effect on functional networks, other network effects were non-specific, the sham group increased EDA more than the arousal group, and there was no effect on attentional-executive function or mood and no seizure outcome. Spurgeon 2026 (118 lifestyle-intervention studies, biofeedback k = 14) gives no biofeedback-specific responder rate. Level 3 becomes defensible if the Nagai trials verify as randomised controlled seizure-outcome studies.
In short
Clinical reading
AAPB Levels 4 (SMR and SCP, seizure reduction), 3 (SMR, nonseizure manifestations) and 1 (connectivity and LORETA z-score) all confirmed by NeuroLogic, ages pooled for want of age-band data. The 4 rests on Sterman 2000 and Tan 2009, Lantz & Sterman 1988 (RCT) and the SCP replication of Kotchoubey 2001 / Strehl 2014; no 5, since superiority over a credible sham on seizure frequency is not shown. New rows: ILF Level 1 (one case), electrodermal biofeedback Level 2 pending verification of the Nagai trials.
Protocols
Sensorimotor rhythm uptraining (12-15 Hz) at central sites (C3, C4 or Cz), often with simultaneous theta suppression, or slow cortical potential training (about 35 sessions); long series, alongside medical treatment.
Limits
The literature is old and methodologically uneven; no clinical-fidelity randomised seizure trial since 2022; the only sham-controlled trial (adolescents already controlled on medication) showed no seizure effect; no analysis by age band; connectivity, LORETA z-score and ILF protocols have no controlled base; no trial measured medication reduction. Intracranial (iEEG) hippocampal neurofeedback pilots (Matsuhashi 2026; Koizumi 2023) target memory preservation before resective surgery, are invasive and research-only, and are not rated here.
Study base
Mature but old: 1970s-2000s case series, two meta-analyses (Sterman 2000; Tan 2009), one small RCT and a 10-year SCP cohort. 2022-2026 base: 39 records screened, 3 kept on triage, 2 publications indexed in the archive; one clinical seizure study only (Fumuro 2025, SCP, n = 12, uncontrolled).
Brendan's perspective
Level 4, same as the AAPB, and this is one I would not move in either direction. The base is unusually good for our field: Sterman 2000 and Tan 2009 in medically refractory epilepsy, a randomised trial with a non-contingent-feedback arm (Lantz & Sterman 1988) and, rare in neurofeedback, an independent replication with ten-year persistence on the slow-cortical-potential side (Kotchoubey 2001; Strehl 2014). It is also unusually old: no clinical-fidelity randomised seizure trial since 2022, and Morales-Quezada 2019 was run in adolescents already controlled on medication, which is a floor effect rather than a refutation. In clinic this is SMR uptraining at central sites with theta suppression, or SCP, long series, always alongside the neurologist and never a reason to touch medication. Look at the level 1 rows: LORETA z-score and connectivity rest on case series from their own originators, and ILF on a single case (Schmidt 2023) — worth knowing when you see ILF marketed for paediatric epilepsy. And no trial has measured medication reduction. One thing from my own practice that I have never seen properly studied, offered as the anecdote it is: with epileptic clients — particularly those who feel a seizure coming — it is worth adding skin conductance biofeedback. The task is deliberately simple: learn to bring down a single physiological measure, and with it the arousal level, on demand. In some of these individuals I have watched that skill applied at the onset of an aura appear to abort what would otherwise have become a seizure. I have no controlled data for that, no idea how often it holds, and I am aware of every way a clinician can fool himself about seizure counts. I still teach it, because the skill costs the person very little and the ceiling on it looks high.
Lantz & Sterman (1988) Neuropsychological assessment of subjects with uncontrolled epilepsy: Effects of EEG feedback training doi:10.1111/j.1528-1157.1988.tb04414.x
Sterman et al. (1974) Biofeedback training of the sensorimotor electroencephalogram rhythm in man: Effects on epilepsy doi:10.1111/j.1528-1157.1974.tb04016.x
Lubar & Bahler (1976) Behavioral management of epileptic seizures following EEG biofeedback training of the sensorimotor rhythm doi:10.1007/BF00998692
Kotchoubey et al. (2001) Modification of slow cortical potentials in patients with refractory epilepsy: A controlled outcome study doi:10.1046/j.1528-1157.2001.22200.x
Strehl et al. (2014) Sustained reduction of seizures in patients with intractable epilepsy after self-regulation training of slow cortical potentials — 10 years after doi:10.3389/fnhum.2014.00604
Morales-Quezada et al. (2019) Neurofeedback impacts cognition and quality of life in pediatric focal epilepsy: An exploratory randomized double-blinded sham-controlled trial doi:10.1016/j.yebeh.2019.106570
Fumuro et al. (2025) Self-regulation of slow cortical potential and seizure suppression by scalp electroencephalography: Early prediction of therapeutic efficacy doi:10.1016/j.clinph.2024.11.018
Matsuhashi et al. (2026) Optimization of memory neurofeedback system utilizing intracranial electroencephalogram of the hippocampus doi:10.1016/j.clinph.2025.2111490
Koizumi et al. (2023) Intracranial Neurofeedback Modulating Neural Activity in the Mesial Temporal Lobe During Memory Encoding: A Pilot Study doi:10.1007/s10484-023-09595-1
