Skip to content

Neuromodulation and experimental therapy

TL;DR — Deep-brain stimulation (DBS), repetitive transcranial magnetic stimulation (rTMS), psilocybin and other biological strategies remain experimental in AN. Evidence is dominated by feasibility studies and highly selected small samples; sham control, expectancy, concurrent care and durability are major uncertainties (Gallop 2022, PMID 35179712). The whole DBS literature amounts to 36 patients with patient-level data (Shaffer 2023, PMID 36724522); a 2026 rTMS meta-analysis pooled 195 active and 132 sham treatments but found no significant BMI benefit (Ewis 2026, PMID 42226264); the psilocybin evidence is a 10-participant open-label phase 1 study (Peck 2023, PMID 37488291). Invasive interventions require especially strong consent, capacity and risk governance.

Modality map

Modality Proposed target Evidence stage Central limitation
rTMS Dorsolateral prefrontal, parietal and extrastriate body-area targets Several small sham-controlled trials plus case series; 195 active and 132 sham treatments pooled across AN and BN studies (Ewis 2026, PMID 42226264) No significant pooled BMI effect in AN; targets and outcomes differ
DBS Subcallosal cingulate, nucleus accumbens, ventral anterior limb of internal capsule 11 studies, 36 patients with patient-level data (Shaffer 2023, PMID 36724522) Open-label throughout; no blinded on/off periods anywhere in the literature
Psilocybin-assisted therapy Cognitive flexibility/experiential change One 10-participant open-label phase 1 (Peck 2023, PMID 37488291) No control arm; psychological support co-administered; safety-only endpoints
Ketamine Rapid affective/cognitive effects Case-level/early investigation Core-AN efficacy and medical safety unknown
Neurofeedback Self-regulation of circuit activity Pilot Target engagement and clinical mediation

rTMS

The field's own early commentary framed rTMS as a candidate treatment before controlled data existed (Bainbridge 2014, PMID 24269202 — a two-page letter, not a trial). An open case series of five women with enduring AN receiving ~20 sessions of neuronavigated high-frequency rTMS to the left dorsolateral prefrontal cortex reported improved eating-disorder and affective symptoms with stable BMI at post-treatment, further gains at 6 months (3/5 "recovered" on the EDE-Q), but waning psychopathology effects and some weight loss by 12 months (McClelland 2016, PMID 26537308).

The TIARA study randomized 34 community-dwelling adults with severe and enduring AN (illness duration ≥3 years, at least one previous completed treatment) 1:1 to 20 sessions of real or sham MRI-guided high-frequency left-DLPFC rTMS over 4 weeks, in addition to treatment as usual. rTMS was safe and well tolerated. Between-group effect sizes from baseline to 4-month follow-up were small for BMI (d=0.2, 95% CI −0.49 to 0.90) and for eating-disorder symptoms (d=0.1, 95% CI −0.60 to 0.79), medium for quality of life, and moderate-to-large for mood (d=0.61–1.0), all nominally favouring rTMS (McClelland 2018, PMID 30012789).

The evidence has widened without becoming definitive. A double-blind proof-of-concept trial randomized 20 patients to active or sham extrastriate-body-area theta-burst stimulation plus body-perception training, with a further 20 treatment-as-usual comparators. Body Shape Questionnaire improvement was faster after active stimulation and persisted to six months, but only 10 patients received active stimulation (Boehme 2026, PMID 42386712). A 2026 systematic review pooled 13 AN/BN studies (195 active and 132 sham treatments): rTMS did not significantly improve BMI in AN (mean difference 0.19, 95% CI −0.50 to 0.88), while depression improved modestly over sham (SMD −0.41, 95% CI −0.81 to −0.02) and no serious adverse events were reported (Ewis 2026, PMID 42226264). The controlled evidence therefore remains small and heterogeneous, with no established durable remission effect.

DBS

The DBS literature is small, open-label and surgically consequential. A phase 1 pilot implanted six patients with chronic treatment-refractory AN at the subcallosal cingulate; one serious adverse event (a seizure during programming) occurred, alongside panic attack, nausea, air embolus and pain, and three of six exceeded their historical baseline BMI at 9 months (Lipsman 2013, PMID 23473846). The extended open-label trial enrolled 16 patients (mean age 34 years, SD 8; mean illness duration 18 years, SD 6); seven of 16 (44%) had serious adverse events, mostly attributable to the underlying illness, and two requested device removal or deactivation (Lipsman 2017, PMID 28238701). A separate randomized study allocated eight patients to subcallosal cingulate or nucleus accumbens stimulation by comorbidity profile; mean BMI at 6 months did not differ from the pre-operative mean (p=0.84), although it did rise against a constructed BMI reference value (p=0.02), five of eight gained ≥10% on that reference, quality of life improved (p=0.03), and three had cutaneous complications (Villalba Martínez 2020, PMID 32580399).

A contemporaneous commentary framed the 2017 trial as a genuine attempt to change the brain to treat AN while noting how early the field remained (McAdams 2017, PMID 28238702). Pooling patient-level data from 11 studies and 36 patients (mean age 38.07 years, mean BMI at surgery 12.58 kg/m²), a network meta-analysis ranked the subcallosal cingulate as the best-supported target for BMI change at 6 and 9–12 months (P-scores 0.94 and 0.98) — while stating that blinded on/off periods are necessary to confirm it (Shaffer 2023, PMID 36724522). That caveat is the whole point: as of September 2026 no DBS study in anorexia nervosa has included a blinded stimulation-off period, so every BMI change in this literature is confounded with intensive concurrent care, regression to the mean and natural history. Long-term follow-up has been reported as a research letter without a control condition (De Vloo 2021, PMID 33687970). Ethical analysis must address irreversible procedural risks, device dependence, capacity under illness, rescue care and the possibility that desperation amplifies therapeutic misconception.

Psychedelic-assisted therapy

The 2023 phase 1 study administered a single 25-mg dose of synthetic psilocybin with psychological support to 10 adult women meeting DSM-5 criteria for AN or partial-remission AN (mean BMI 19.7 kg/m², SD 3.7), at a single academic research institute (NCT04661514). Its aims were safety, tolerability and feasibility. No clinically significant changes occurred in ECG, vital signs or suicidality; two participants developed asymptomatic hypoglycaemia that resolved within 24 hours; all adverse events were mild and transient (Peck 2023, PMID 37488291). The accompanying Nature Medicine commentary situates the result within the wider psychedelic-therapy debate but is itself a two-page comment, not data (Majić 2023, PMID 37488290). This is proof that carefully governed administration can be studied; it is not evidence of superiority to established care, and with n=10, no control and no efficacy endpoint it cannot be. Key design variables are psychological support, dose, expectancy, cardiovascular/metabolic eligibility, suicidality monitoring and durability. Current recruiting work is catalogued in clinical trials (NCT07169747; live ClinicalTrials.gov API, 2026-09-02).

Around 20–30% of people with an eating disorder fail to respond to the best available treatments and develop persistent, treatment-refractory illness — the population from which every DBS and most rTMS samples are drawn, and the reason these interventions are pursued at all despite their evidence base (Bryson 2024, PMID 38503683).

Transcranial direct current stimulation

tDCS is cheaper, more portable and lower-risk than either rTMS or DBS, and its AN evidence base is correspondingly earlier. A single-centre double-blind placebo-controlled trial in Poland is delivering 30 25-minute sessions at 2 mA to the DLPFC (F3 anode / F4 cathode) twice daily over three weeks alongside treatment as usual in female inpatients aged 13–25 with BMI ≤ 17.5 kg/m², with EAT-26 change as the primary outcome (NCT05814458; registry status active, not recruiting, estimated enrolment 40) (Rząd 2024, PMID 38757134 — protocol).

Its published interim analysis covers 20 adolescent females (10 active, 10 sham). Eating-disorder symptoms fell significantly in the active group at end of treatment (p = 0.003) and at follow-up (p = 0.02) with no significant change in the sham group; BMI rose more in the active group (13.78% vs 7.31%) but the difference was not statistically significant (p = 0.10); adverse effects were mild and transient (Rząd 2025, PMID 40725729). Two cautions apply and the authors state the second: a within-group significant change alongside a within-group null in the comparator is not a between-group difference, and this is an interim analysis published to justify continuing the trial, not a result. An independent pilot targeted a different site entirely — the left inferior parietal lobe rather than DLPFC — in 20 participants randomized to 10 sessions of high-definition tDCS or sham, with feasibility and acceptability as the primary aims (Phillipou 2019, PMID 31433523). Two active trials, two different targets, no shared outcome measure.

Cognitive remediation therapy: mechanism-targeted and repeatedly null

CRT is the clearest test in AN of the "model → target → clinical outcome" chain, because it was explicitly built to correct the set-shifting and central-coherence inefficiencies described in neurobiology and cognition. The results do not support the chain.

Synthesis Corpus Result
Hagan 2020, PMID 32408265 7 RCTs plus one quasi-RCT; quality fair to good CRT was not associated with improved central coherence versus control at end of treatment (g = 0.25, 95% CI −0.35 to 0.85, k = 3); set-shifting results were mixed because measures differed across studies; no advantage on eating-disorder-related or other psychological outcomes. CRT may reduce dropout, on insufficient data
Alserihi 2024, PMID 39540007 6 studies, 4 meta-analysed; GRADE-appraised (PROSPERO CRD42023411784) Cognitive flexibility: MD −0.21 (95% CI −0.44 to 0.02) across 3 trials and 413 participants; EDE-Q: MD −0.25 (−0.76 to 0.27); eating-disorder quality of life: MD −0.19 (−0.41 to 0.03), all with I² = 0%. Conclusion: efficacy "remains inconclusive"

Two independent meta-analyses, five years apart, find CRT does not reliably move the very neuropsychological variables it was designed to move, and does not move eating-disorder outcomes. Read alongside the SWAN secondary analysis, in which participants with poorer baseline central coherence improved faster on standard therapy (Keegan 2022, PMID 35715854), the case for treating set-shifting and coherence as therapeutic targets in AN is weaker now than when CRT was introduced. Canadian guidance nevertheless does not list cognitive remediation among its GRADE-graded recommendations, positive or negative (Couturier 2020, PMID 32021688); NICE NG69 recommendation text read on 2026-09-02 does not name it either — see guidelines.

Exposure-based and psychedelic approaches beyond psilocybin

Virtual-reality body exposure. Thirty-five patients with AN (16 experimental, 19 control) received treatment as usual, with the experimental group additionally receiving five sessions of embodiment-enhanced VR body exposure. Eating-disorder symptoms fell in both groups, but fear of gaining weight and body-image disturbance were significantly lower in the experimental group both immediately after the intervention and at three months (Porras-Garcia 2021, PMID 33578767). The design is unblinded with an inactive add-on comparator, so attention and expectancy are not controlled; the specific targeting of fear of weight gain — the criterion at the centre of the diagnosis — is nonetheless a departure from weight-focused endpoints.

MDMA-assisted therapy has no AN-specific trial. The only randomized evidence is a pre-specified exploratory analysis inside a placebo-controlled pivotal trial of MDMA-assisted therapy for severe PTSD in 90 adults, of whom 15% had EAT-26 scores in the clinical range at baseline despite the absence of purging or low weight. Among 82 completers, total EAT-26 fell significantly more after MDMA-assisted therapy than placebo (p = .03), with larger effects in women scoring ≥11 (p = .0012) and ≥20 (p = .0478) (Brewerton 2022, PMID 35272210). This is eating-disorder symptomatology in a PTSD population, not AN, and it is a secondary analysis.

Ketamine likewise has no controlled AN trial. An open-label study gave five adults weight-recovered from AN but with persistent eating-disorder cognition a therapeutic ketogenic diet followed by six ketamine infusions, with follow-up to 6 months. All completed without significant adverse effects, and the group improved on clinical impairment (p = 0.008), EDE-Q global (p = 0.006) and eating (p = 0.005), shape (p = 0.016) and weight (p = 0.032) concerns; one participant relapsed at four months (Calabrese 2022, PMID 35997954). Five participants, no control, two co-administered interventions — the design cannot attribute the change to either component.

Evidence thresholds

Gate Minimum evidence before routine use
Feasibility Recruitment, retention and protocol delivery
Safety Prespecified adverse events and long enough follow-up
Target engagement Demonstrated change in proposed mechanism
Efficacy Credible comparator, blinded assessment, core outcomes
Durability Relapse, function and quality of life beyond acute novelty
Generalizability Broader severity, sex/gender, age and comorbidity

Open questions

  • Can any modality outperform credible sham plus specialist care on durable remission? Sham-controlled trials remain small and heterogeneous, and the 2026 meta-analysis found no significant BMI effect (McClelland 2018, PMID 30012789; Boehme 2026, PMID 42386712; Ewis 2026, PMID 42226264).
  • Which neural target mediates benefit rather than merely changing scan measures? Target ranking currently rests on 36 unblinded patients pooled across 11 DBS studies (Shaffer 2023, PMID 36724522).
  • Would a blinded stimulation-on/off crossover be ethically and practically deliverable in severe enduring AN, and what would it cost in trial retention (Shaffer 2023, PMID 36724522; Lipsman 2017, PMID 28238701)?
  • How should capacity and voluntariness be protected in invasive trials for severe enduring illness?
  • Which cortical target should non-invasive stimulation use? Active AN trials are stimulating the DLPFC and the inferior parietal lobe on different rationales, with no shared outcome measure (Rząd 2025, PMID 40725729; Phillipou 2019, PMID 31433523).
  • If cognitive remediation reliably fails to improve the cognitive inefficiencies it targets, is the set-shifting/central-coherence model wrong, the intervention weak, or the measures insensitive (Hagan 2020, PMID 32408265; Alserihi 2024, PMID 39540007; Keegan 2022, PMID 35715854)?
  • Does targeting fear of weight gain directly, as virtual-reality body exposure does, outperform weight-focused treatment on relapse — and does the effect survive an attention-matched control (Porras-Garcia 2021, PMID 33578767)?
  • Do the psychedelic and dissociative signals in AN come from the drug, the accompanying psychological support, or the population? Every AN-relevant dataset is either open-label, uncontrolled, or drawn from a non-AN sample (Peck 2023, PMID 37488291; Brewerton 2022, PMID 35272210; Calabrese 2022, PMID 35997954).

References

  1. Gallop L, et al. Neuromodulation and eating disorders. Curr Psychiatry Rep. 2022. PMID 35179712.
  2. Bainbridge K, et al. rTMS as a treatment for anorexia nervosa. Brain Stimul. 2014;7:149-150. Letter. PMID 24269202.
  3. De Vloo P, et al. Long-term follow-up of deep brain stimulation for anorexia nervosa. J Neurol Neurosurg Psychiatry. 2021;92:1135-1136. Research letter. PMID 33687970.
  4. Majić T, et al. Psilocybin for the treatment of anorexia nervosa. Nat Med. 2023;29:1906-1907. Commentary on Peck 2023. PMID 37488290.
  5. McAdams CJ, et al. Changing the brain to treat anorexia nervosa. Lancet Psychiatry. 2017;4:262-263. Commentary on Lipsman 2017. PMID 28238702.
  6. Bryson C, et al. Established and emerging treatments for eating disorders. Trends Mol Med. 2024. PMID 38503683.
  7. Peck SK, et al. Psilocybin therapy for females with anorexia nervosa: a phase 1, open-label feasibility study. Nat Med. 2023;29:1947-1953. PMID 37488291.
  8. McClelland J, et al. Randomised controlled feasibility trial of real versus sham rTMS in adults with severe and enduring anorexia nervosa: the TIARA study. BMJ Open. 2018;8:e021531. PMID 30012789.
  9. McClelland J, et al. Repetitive transcranial magnetic stimulation treatment in enduring anorexia nervosa: a case series. Eur Eat Disord Rev. 2016;24:157-163. PMID 26537308.
  10. Lipsman N, et al. Subcallosal cingulate deep brain stimulation for treatment-refractory anorexia nervosa: a phase 1 pilot trial. Lancet. 2013;381:1361-1370. PMID 23473846.
  11. Lipsman N, et al. Deep brain stimulation of the subcallosal cingulate for treatment-refractory anorexia nervosa: 1 year follow-up of an open-label trial. Lancet Psychiatry. 2017;4:285-294. PMID 28238701.
  12. Villalba Martínez G, et al. A randomized trial of deep brain stimulation to the subcallosal cingulate and nucleus accumbens in patients with treatment-refractory, chronic, and severe anorexia nervosa: initial results at 6 months of follow up. J Clin Med. 2020;9:1946. PMID 32580399.
  13. Shaffer A, et al. Efficacy of deep brain stimulation for the treatment of anorexia nervosa: a systematic review and network meta-analysis of patient-level data. Neurosurg Focus. 2023;54:E5. PMID 36724522.
  14. Rząd Z, et al. The efficacy of transcranial direct current stimulation in the treatment of anorexia nervosa: a randomized double-blind clinical trial. Front Psychiatry. 2024;15:1284675. Protocol. PMID 38757134. (NCT05814458)
  15. Rząd Z, et al. Efficacy of transcranial direct current stimulation in the treatment of anorexia nervosa — interim results from an ongoing, double-blind, randomized, placebo-controlled clinical trial. J Clin Med. 2025;14:5063. Interim analysis. PMID 40725729.
  16. Phillipou A, et al. High-definition transcranial direct current stimulation in anorexia nervosa: a pilot study. Int J Eat Disord. 2019;52:1274-1280. PMID 31433523.
  17. Hagan KE, et al. A preliminary systematic review and meta-analysis of randomized-controlled trials of cognitive remediation therapy for anorexia nervosa. Eat Behav. 2020;37:101391. PMID 32408265.
  18. Alserihi AR, et al. Effect of cognitive remediation therapy in anorexia nervosa: a systematic review and meta-analysis. Front Psychiatry. 2024;15:1484457. PMID 39540007.
  19. Porras-Garcia B, et al. AN-VR-BE: a randomized controlled trial for reducing fear of gaining weight and other eating disorder symptoms in anorexia nervosa through virtual reality-based body exposure. J Clin Med. 2021;10:682. PMID 33578767.
  20. Brewerton TD, et al. MDMA-assisted therapy significantly reduces eating disorder symptoms in a randomized placebo-controlled trial of adults with severe PTSD. J Psychiatr Res. 2022;149:128-135. PMID 35272210.
  21. Calabrese L, et al. Ketogenic diet and ketamine infusion treatment to target chronic persistent eating disorder psychopathology in anorexia nervosa: a pilot study. Eat Weight Disord. 2022;27:3751-3757. PMID 35997954.
  22. Keegan E, et al. An exploratory examination of executive functioning as an outcome, moderator, and predictor in outpatient treatment for adults with anorexia nervosa. J Eat Disord. 2022;10:83. PMID 35715854.
  23. Couturier J, et al. Canadian practice guidelines for the treatment of children and adolescents with eating disorders. J Eat Disord. 2020;8:4. PMID 32021688.
  24. Boehme R, et al. Theta burst stimulation of extrastriate body area for body perception in anorexia nervosa: a randomized controlled trial. Transl Psychiatry. 2026;16. PMID 42386712.
  25. Ewis DK, et al. Efficacy and safety of repetitive transcranial magnetic stimulation in anorexia and bulimia nervosa: a systematic review and meta-analysis. Ann Gen Psychiatry. 2026;25. PMID 42226264.