In arthritis the modality matters. For knee osteoarthritis, relearning how to walk with real-time feedback on knee loading now has randomised evidence, including against sham retraining, with pain reduction lasting a year. Traditional EMG biofeedback added to strengthening exercise has repeatedly done no better than exercise alone, and the older rheumatoid arthritis trials of EMG and thermal biofeedback remain small. Neurofeedback has one unreplicated trial.
Updated :
What the research shows
EEG neurofeedback — standard amplitude training
EEG neurofeedback — alpha training as an add-on to rheumatoid arthritis treatment
Ages
Adult (18+)
Techniques
Alpha training
AAPB0Not ratedNeuroLogic2Possibly efficacious
Level 2. Grekhov 2018 (cited in the AAPB base, but the AAPB's rating covers biofeedback): alpha-enhancement neurofeedback, 12-14 daily sessions, added to medication and physiotherapy in rheumatoid arthritis (n = 60) vs medication and physiotherapy alone (n = 30) — reduced joint-syndrome pain on OMERACT III criteria and higher subjective control; the few non-responders were mainly stage III disease. One adequately sized but unblinded trial with a treatment-as-usual control, published in a regional journal, without replication; no 2022-2026 amplitude-neurofeedback study in arthritis passed screening.
Level 1. Mathew 2022: source-localised infraslow-fluctuation neurofeedback (pregenual and dorsal anterior cingulate, primary somatosensory cortices), nine 30-minute sessions, double-blind sham-controlled feasibility trial in chronic painful knee osteoarthritis, n = 22 — 91 % retention, no serious adverse events; pain severity and interference changed similarly in the active and sham arms, physical function numerically favoured the active arm (6.2 vs 1.6 points). Secondary analyses (Mathew 2025) report changes in EEG activity and effective connectivity in the targeted regions. A feasibility trial not powered for efficacy: equal-to-sham does not disqualify the method, but no adequately powered study exists.
Biofeedback
Real-time gait-retraining biofeedback (knee loading and kinematics) — knee osteoarthritis
Ages
Adult (18+)
Techniques
Other biofeedback
AAPB3Probably efficaciousNeuroLogic4Efficacioushigher vs AAPB
Level 4 (the AAPB Level 3 is a pooled rating across all biofeedback modalities for arthritis). Real-time gait-retraining biofeedback for knee osteoarthritis now has randomised superiority in independent settings. Segal 2015 (AAPB): physiotherapist-directed gait training with visual feedback vs usual care, n = 56 adults aged 60 or over — better mobility disability and pain at 3 months, no significant difference at 6 and 12 months. Uhlrich 2025: personalised foot-progression-angle retraining vs sham retraining (the participant's natural foot angle as target), n = 68, participants and analysts masked — greater 1-year reductions in medial knee pain (between-group difference −1.2, 95 % CI −1.9 to −0.5) and peak knee adduction moment, with less MRI-estimated cartilage change. Wan 2026: n = 50, three arms — all groups improved pain and function, and only gait-pattern feedback produced a lasting reduction in the first peak knee adduction moment (−7.6 %). Rynne 2022 meta-analysis: gait retraining vs control, knee adduction moment SMD −1.10 and WOMAC SMD −0.86 (six studies, high heterogeneity). Level 5 would need sham superiority replicated in a second setting; Uhlrich 2025 is so far the only sham-controlled trial.
EMG and thermal biofeedback — rheumatoid arthritis and osteoarthritis (adult)
Ages
Adult (18+)
Techniques
EMG biofeedback, Thermal biofeedback
AAPB3Probably efficaciousNeuroLogic3Probably efficacioussame vs AAPB
Level 3 maintained (the AAPB Level 3 is a pooled rating across all biofeedback modalities). Rheumatoid arthritis: Flor 1983, EMG biofeedback vs pseudotherapy and conventional medical treatment (n = 8 per arm, 12 sessions) — greater reductions in pain, negative cognitions and physiological measures; Achterberg 1981, thermal biofeedback with relaxation, n = 24 — less muscle tension and pain in both training directions, and greater physical-function change than physiotherapy in the second phase; Astin 2002 meta-analysis of psychological interventions in RA (five biofeedback trials), small post-intervention effects (pain 0.22, functional disability 0.27). Knee osteoarthritis: EMG biofeedback added to exercise has repeatedly shown no significant difference from exercise alone — Yilmaz 2010 (n = 40), Durmus 2007 (n = 50), Mahnik 2026 (n = 93, three arms, 15 sessions, assessor-blind, no between-group difference at any point to 6 months), Chang 2025 (n = 14) — and the French 2024 meta-analysis found no clinically significant adjunctive effect; Raeissadat 2018 (n = 46) favoured biofeedback on VAS pain only. After arthroplasty, Sklempe Kokic 2022 (n = 131) and Kokic 2023 (n = 90) found no additional benefit. Preece 2026, cognitive muscular therapy (EMG-guided muscle retraining within a behavioural programme) vs usual care, feasibility RCT, n = 82: WOMAC fell 17.1 points vs 2.8. No HRV biofeedback trial in arthritis was identified.
EMG and thermal biofeedback — juvenile idiopathic arthritis (child and adolescent)
Ages
Child (under 12) · Adolescent (12-17)
Techniques
EMG biofeedback, Thermal biofeedback
AAPB3Probably efficaciousNeuroLogic3Probably efficacioussame vs AAPB
Level 3 (the AAPB Level 3 is a pooled rating; the AAPB's paediatric evidence concerns juvenile rheumatoid arthritis). Eid 2016: EMG biofeedback-guided isometric exercise added to conventional physiotherapy vs physiotherapy alone, children aged 8-13, n = 36, 12 weeks — significantly greater improvement in pain, quadriceps strength and functional ability at 12 weeks. Lavigne 1992: relaxation, EMG and thermal biofeedback package vs delayed treatment, n = 8, with blinded physician and physiotherapist reports — reduced self- and mother-rated pain, maintained at 6 months. One adequately sized randomised trial and one very small one from different groups; no 2022-2026 paediatric publication passed screening. Child and adolescent bands are pooled because neither trial allows a split.
By age
Child (under 12)
In children the only evidence is two juvenile arthritis trials in which EMG or thermal biofeedback was added to physiotherapy or relaxation, both positive but one very small; gait retraining and neurofeedback have not been studied.
Adolescent (12-17)
In adolescents there is no trial of its own: the level rests on the juvenile arthritis trials, which included children up to age 13, and on the adult evidence.
In short
Clinical reading
AAPB Level 3 (biofeedback, all modalities pooled). NeuroLogic separates the rows: gait-retraining biofeedback for knee osteoarthritis Level 4 (Segal 2015; Uhlrich 2025 sham-controlled; Wan 2026; Rynne 2022 meta-analysis); EMG and thermal biofeedback Level 3 in adults (Flor 1983, Achterberg 1981; repeated null adjunctive trials in knee osteoarthritis, Mahnik 2026) and in juvenile arthritis (Eid 2016, Lavigne 1992); alpha neurofeedback Level 2 (Grekhov 2018, single trial); infraslow neurofeedback Level 1 (Mathew 2022, feasibility).
Protocols
Feedback-guided gait retraining (foot progression angle, knee adduction moment, gait pattern), 6-12 laboratory sessions with feedback fading; surface EMG on the quadriceps and peri-articular groups during exercise; thermal biofeedback and assisted relaxation in rheumatoid arthritis.
Limits
Gait retraining requires an instrumented laboratory or a validated wearable and has one sham-controlled trial; EMG biofeedback added to exercise has not outperformed exercise alone in knee osteoarthritis or after arthroplasty; the rheumatoid arthritis trials are old and small; paediatric evidence is limited to two juvenile arthritis trials; neurofeedback is unreplicated.
Study base
Older base of small RCTs in rheumatoid arthritis and knee osteoarthritis; the 2022-2026 window adds two gait-retraining RCTs and several null adjunctive EMG trials. 2022-2026 base: 11 publications indexed in the archive (10 biofeedback, 1 neurofeedback).
Brendan's perspective
The AAPB gives arthritis one pooled level 3 across all biofeedback modalities, and that hides the interesting part. Feedback-guided gait retraining for knee osteoarthritis is a different intervention from EMG-assisted quadriceps exercise, and it is the one that earns level 4: Uhlrich 2025 beat sham retraining on knee pain a year out, with Segal 2015 and Wan 2026 pointing the same way. EMG biofeedback added to exercise has now failed to add anything several times over (Mahnik 2026), so I would not present it as training worth adding. The neurofeedback rows are thin and rated accordingly: Grekhov 2018 is a single unblinded trial, Mathew 2022 a feasibility study that did not separate from sham. Arthritis is not a condition I see often. When it comes alongside something else, I work on pain and arousal — qEEG-based alpha/SMR amplitude training with HRV biofeedback — and I say plainly that this addresses how someone carries the pain, not the joint itself.
Flor et al. (1983) Efficacy of EMG biofeedback, pseudotherapy, and conventional medical treatment for chronic rheumatic back pain doi:10.1016/0304-3959(83)90124-0
Achterberg, McGraw & Lawlis (1981) Rheumatoid arthritis: a study of relaxation and temperature biofeedback training as an adjunctive therapy doi:10.1007/BF00998870
Eid, Aly & El-Shamy (2016) Effect of electromyographic biofeedback training on pain, quadriceps muscle strength, and functional ability in juvenile rheumatoid arthritis doi:10.1097/PHM.0000000000000524
Segal et al. (2015) Intensive gait training for older adults with symptomatic knee osteoarthritis doi:10.1097/PHM.0000000000000264
Uhlrich et al. (2025) Personalised gait retraining for medial compartment knee osteoarthritis: a randomised controlled trial doi:10.1016/S2665-9913(25)00151-1
Wan et al. (2026) Six-week biofeedback gait retraining programme for people with knee osteoarthritis: A randomised controlled trial doi:10.1016/j.clinbiomech.2026.106776
Rynne et al. (2022) Effectiveness of gait retraining interventions in individuals with hip or knee osteoarthritis: A systematic review and meta-analysis doi:10.1016/j.gaitpost.2022.04.013
Mahnik et al. (2026) Electromyographic biofeedback or neuromuscular electrical stimulation added to isometric quadriceps exercise does not improve pain and functional outcomes in knee osteoarthritis: a randomized controlled trial doi:10.1097/MRR.0000000000000689
Preece et al. (2026) Cognitive muscular therapy™ for knee osteoarthritis: A feasibility randomised controlled trial doi:10.1016/j.ocarto.2026.100796
