Skip to content

Clinical trials landscape in fibromyalgia

TL;DR — Fibromyalgia (FM) drug development ended a fifteen-year drought in August 2025 with FDA approval of sublingual cyclobenzaprine (TNX-102 SL, Tonmya), on the back of two positive Phase 3 trials, RELIEF (NCT04172831) and RESILIENT (NCT05273749), the latter showing a −1.8 vs −1.2 point difference in weekly average daily pain at 14 weeks (Lederman 2026, PMID 40627411). The effect size is small and comparable to duloxetine and pregabalin. The second regulatory-track programme is now an enriched-enrolment randomised-withdrawal trial of esreboxetine (AXS-14, FORWARD, NCT07398417, n = 620) whose primary endpoint is time to loss of therapeutic response — an explicit design response to FM's defining trial pathology: a pooled 18.6% rate of 50% pain reduction on placebo across 18 registration trials, with placebo accounting for roughly 45% of the response seen in active-drug arms (Häuser 2012, PMID 23137770; Häuser 2011, PMID 21429668). The most striking recent result is not pharmacological at all: self-guided digital acceptance and commitment therapy delivered by smartphone produced a 71% vs 22% PGIC responder rate in PROSPER-FM (NCT05243511; Gendreau 2024, PMID 38991582). As of this session's ClinicalTrials.gov queries there are 189 interventional FM trials in recruiting, not-yet-recruiting or active-not-recruiting status, and 43 registered studies using the term "nociplastic pain". The graveyard is instructive: sodium oxybate produced two positive Phase 3 trials plus a positive long-term extension and never reached market, and no NGF-antibody trial in FM exists at all.

How to read trial stages on this page

  • Drug trials carry FDA-defined phases, reported by ClinicalTrials.gov: phase 1 (safety and dose-finding, tens of participants), phase 2 (preliminary efficacy), phase 3 (confirmatory, endpoint-powered), phase 4 (post-marketing); "early phase 1" denotes exploratory testing before formal phase 1. On this page the registration-track programmes are explicit phase 3 (TNX-102 SL's RELIEF/RALLY/RESILIENT; FORWARD, whose registry entry states Phase 3), while most academic drug studies sit at phase 2 or carry no phase.
  • Behavioural, exercise and neuromodulation trials are mostly phase-less: ClinicalTrials.gov assigns "NA" to interventional studies outside the drug-phase framework, which covers the bulk of the 189 active FM trials (§1). A behavioural trial run to regulatory standard can still borrow drug-phase vocabulary — PROSPER-FM, an industry-sponsored digital therapeutic study, was published as "a phase 3, multicentre, randomised controlled trial" (PMID 38991582). The phase field encodes regulatory pathway, not evidential weight.
  • Enriched-enrolment randomised withdrawal (EERW) — the FORWARD design, discussed in §7.2 — changes what both the efficacy and the safety numbers mean. Efficacy: the double-blind phase answers "does benefit persist in people already responding?", not "does the drug work in unselected patients?". Safety: everyone randomised has, by construction, already tolerated the drug through the open-label phase, so adverse-event and discontinuation rates from the randomised phase systematically understate first-exposure tolerability and are not comparable with parallel-group trial safety tables. The open-label phase carries the honest tolerability signal; in the two pregabalin EERW trials, 54% of the 1,492 who started titration never reached randomisation (PMID 27684492, §7.2).
  • Safety numbers below are reported as: any treatment-emergent adverse event (TEAE) rate per arm, discontinuations due to AE, serious AEs (SAE), and each trial's signature AE — because in a condition with only patient-reported endpoints, a distinctive side effect is also an unblinding mechanism (§7.3).

1. How this page was built

All trial data below were retrieved from ClinicalTrials.gov during the session dated in the frontmatter. Query counts, all interventional unless stated:

Query Total returned
Condition "fibromyalgia", recruiting / not-yet-recruiting / active-not-recruiting 189
Condition "nociplastic pain" (all study types, all statuses) 43
Condition "fibromyalgia" + intervention exercise (recruiting/active) 48
Condition "fibromyalgia" + intervention digital therapeutic / smartphone app / ACT 36
Condition "fibromyalgia" + intervention transcranial magnetic stimulation 30
Condition "fibromyalgia" + intervention TNX-102 / cyclobenzaprine 12
Condition "fibromyalgia" + intervention psilocybin / MDMA / ketamine / psychedelic 12
Condition "fibromyalgia" + intervention naltrexone 8
Condition "fibromyalgia" + intervention immunoglobulin / rituximab / IVIG 3
Condition "fibromyalgia" + intervention tanezumab / fasinumab / nerve growth factor 3 (none NGF-directed)
Condition "fibromyalgia" + intervention sodium-channel blocker / lacosamide / suzetrigine / NaV1.7 1

The distribution is itself the finding. Of 189 active trials, the overwhelming majority are small, single-centre, non-pharmacological studies of exercise, manual therapy, electrical stimulation and app-delivered behavioural interventions, typically with 20–100 participants and no blinding of participants. Industry-sponsored, multi-site, placebo-controlled drug trials number in the low single digits.

2. The 2025 approval: TNX-102 SL

Sublingual cyclobenzaprine was approved by the FDA in August 2025 — the first new pharmacologic treatment approved for FM in more than fifteen years (Abd-Elsayed 2026, PMID 42479282). Its rationale is pharmacokinetic rather than novel-target: bedtime sublingual dosing produces peak plasma concentrations 4–5 hours after administration, aligning exposure with the middle of the sleep period, and the proposed mechanism is 5-HT2A and α1-adrenergic antagonism affecting sleep architecture and central pain processing rather than peripheral muscle relaxation.

NCT Name Design N Status Readout
NCT01903265 Phase 2b/3 DB RCT, TNX-102 SL 2.8 mg vs placebo 205 Completed —
NCT02015234 F202 12-month open-label extension 158 Completed —
NCT02436096 AFFIRM Phase 3 DB RCT, 2.8 mg vs placebo 519 Completed —
NCT02589275 — 3-month open-label extension 375 Completed —
NCT02829814 — Phase 3 DB RCT, 2.8 mg vs placebo 51 Terminated Stopped after 51 of planned enrolment
NCT04172831 RELIEF Phase 3 DB RCT, 2.8→5.6 mg vs placebo, 14 wk 503 Completed Pain −1.9 vs −1.5, p = 0.01; PGIC not significant; FIQR, PROMIS sleep/fatigue, daily sleep quality improved (PMID 37165930)
NCT04508621 RALLY Phase 3 DB RCT, 5.6 mg vs placebo, 14 wk 514 Completed Missed primary endpoint: pain −1.6 (95% CI −1.8 to −1.3) vs −1.3 (−1.5 to −1.1), p = 0.115 — registry-posted results only, no journal publication located in PubMed this session (ClinicalTrials.gov results for NCT04508621, accessed 2026-08-28); halted at interim analysis partly for high AE-related discontinuations in both arms during peak-COVID recruitment (PMID 40627411)
NCT05273749 RESILIENT Phase 3 DB RCT, 2.8 mg ×2 wk → 5.6 mg ×12 wk vs placebo 457 Completed Pain −1.8 (SE 0.12) vs −1.2 (SE 0.12), p < 0.001; all six secondary endpoints p ≤ 0.001 (PMID 40627411)

Two features of this programme are worth extracting because they generalise.

The near-miss and the replication. The early trials ran at 2.8 mg only (the Phase 2b/3 study and AFFIRM); the programme then moved to a 5.6 mg target dose reached after a two-week 2.8 mg run-in — the identical schedule in RELIEF and RESILIENT (PMIDs: 37165930, 40627411; RALLY was dosed at 5.6 mg throughout per its registration). On that schedule, RELIEF produced a marginal result: it hit its primary pain endpoint at p = 0.01 but missed PGIC, the endpoint closest to whether patients think they are better. RESILIENT then reproduced the design and cleared the primary endpoint and every one of six secondary endpoints at p ≤ 0.001. A 14-week trial that separates on diary pain but not on global impression is exactly the ambiguous result FM trials repeatedly produce — and the programme needed a second, cleaner replication to resolve it.

The tolerability trade. In RESILIENT the most common adverse events were local: oral hypoesthesia 23.4% vs 0.4% placebo, abnormal taste 11.3% vs 0.9%, oral paraesthesia 6.9% vs 0.9%. Systemic events were uncommon (somnolence 3.0% vs 1.3%). This is a deliberate design choice — shifting adverse effects from systemic to local — but it also functions as a partial unblinding mechanism: roughly a quarter of active-arm participants experience a distinctive oral sensation that placebo participants do not. In a condition with an 18.6% placebo 50%-response rate, any cue that reveals allocation matters. The sponsor's counter-evidence: a post hoc analysis in RESILIENT found the treatment effect similar in active-arm patients with an oral sensory AE (LSMD 0.72 NRS units vs placebo, p = 0.003, n = 71) and without one (LSMD 0.62, p < 0.001, n = 160), with no difference between the two subgroups (p = 0.70) — consistent with the effect not being driven by AE-mediated unblinding, though a post hoc subgroup split cannot fully exclude it (PMID 40627411).

Safety readouts across the programme (phase 3 trials with published or registry-posted safety data):

Trial Any TEAE Discontinued for AE Serious AEs Signature (oral) AEs, active vs placebo Source
RELIEF 59.7% vs 46.3% Not reported in paper 2/248 vs 5/255 (registry) Oral hypoesthesia 17.3% vs 0.4%; oral paraesthesia 5.6% vs 0.4%; abnormal taste 4.4% vs 0.4% PMID 37165930; ClinicalTrials.gov results for NCT04172831, accessed 2026-08-28
RALLY Registry reports only AEs ≥5% frequency Not posted 6/256 (2.3%) vs 0/258 — none in a pattern suggesting drug causation (single events: pneumonias, pancreatitis, fracture, breast carcinoma) Oral hypoesthesia 28.1% (72/256) vs 0.8% (2/258); abnormal taste 10.2%; oral discomfort 9.0%; tongue discomfort 6.3% ClinicalTrials.gov results for NCT04508621, accessed 2026-08-28 (no journal publication)
RESILIENT 58.9% vs 36.7% 6.1% vs 3.5% Serious TEAE 0.9% vs 1.3%; severe 1.3% each arm; no deaths Oral hypoesthesia 23.4% vs 0.4%; abnormal taste 11.3% vs 0.9%; oral paraesthesia 6.9% vs 0.9%; any oral-cavity TEAE 42.9% vs 10.2%, 63.8% resolving in <60 min PMID 40627411

Systemic tolerability is the programme's selling point: somnolence 3.0% vs 1.3% and weight-gain TEAEs 0.4% in both arms in RESILIENT (PMID 40627411) — against the dizziness/somnolence/weight-gain profile of the gabapentinoid comparator class (§7.4).

3. Active and recent drug trials

NCT Name / agent Design N Status Notes
NCT07398417 FORWARD — AXS-14 (esreboxetine), Axsome Phase 3; 12-wk open-label, then 12-wk double-blind randomised withdrawal in responders 620 Recruiting (51 sites; status and Phase 3 label re-verified 2026-08-28) Primary endpoint: time from randomisation to loss of therapeutic response; started 2026-01, primary completion 2028-03
NCT05900466 Metformin Phase 2 RCT vs placebo 72 Recruiting Mechanistic proof-of-concept, Univ. of Utah
NCT06946940 Cannabis oil 5% THC / 5% CBD Phase 2, sham-controlled, "biosignature of response" 150 Recruiting Rambam; explicitly seeks a response biomarker
NCT07498023 FRIDINHA — full-spectrum cannabis extract, dose-ranging Phase 1/2 open-label multicentre 36 Active, not recruiting —
NCT07783295 Full-spectrum cannabis 5:1 CBD:THC + CB2/TRPV1 polymorphisms Phase 2 80 Not yet recruiting Pharmacogenetic stratification attempt
NCT07230171 Esketamine + pregabalin/venlafaxine vs conventional Multicentre 92 Recruiting Beijing Tiantan
NCT07186751 Pregabalin ± venlafaxine Multicentre 750 Recruiting Largest active FM drug trial by enrolment
NCT07382921 Pregabalin ± tizanidine PROBE (open-label, blinded endpoint) 164 Not yet recruiting —
NCT07753018 Curcumin + piperine vs placebo Phase 2 80 Not yet recruiting —
NCT07293767 Silymarin vs diosmin/hesperidin vs no intervention — 90 Recruiting Third arm is no-treatment, not placebo
NCT07585045 Oral magnesium glycinate vs vitamins RCT, FIQR outcome 86 Recruiting In patients not responding to standard treatment

Low-dose naltrexone

Eight registered FM naltrexone studies exist. The definitive one is negative.

NCT Design N Result
NCT04270877 FINAL — single-centre, randomised, double-blind, placebo-controlled, 12 wk, naltrexone 6 mg/day, women 99 Between-group difference in pain intensity −0.34 (95% CI −0.95 to 0.27), p = 0.27, Cohen's d 0.23 — not superior to placebo (PMID 38258677)
NCT04270877 (secondary) QST and functional outcomes 92 analysed Only CPM differed between groups; authors judged it a chance finding uncorrelated with clinical improvement (PMID 40214857)
NCT02806440 Crossover, double-blind, placebo-controlled 58 Completed
NCT04739995 INNOVA — 12-month RCT with cost-utility and neurobiological outcomes 99 Status unknown
NCT04502251 LDN + tDCS factorial vs placebo/sham 92 Completed
NCT00568555 Stanford pilot 53 Completed
NCT02107014 LDN immune monitoring, open-label 9 Completed
NCT00855972 Juvenile primary FM 0 Withdrawn

FINAL is the cautionary tale of the last decade: a treatment adopted widely on the basis of small crossover pilots, then tested properly and found no better than placebo, with 84% (41/49) of LDN and 86% (43/50) of placebo participants reporting adverse events — a near-identical nocebo profile. The rest of the safety ledger is equally symmetric: discontinuations due to AE 4/49 (8%) on LDN versus 3/50 (6%) on placebo, one serious adverse event (in the placebo group), and no deaths (Bruun 2024, PMID 38258677). LDN's reputation as an exceptionally benign drug survived its efficacy trial; what did not survive is the claim that the benefit patients feel on it exceeds what an identical-appearing tablet produces.

Psychedelics and dissociatives

NCT Agent Design N Status Readout
NCT05128162 Psilocybin 15 mg then 25 mg + psychotherapy Phase 2a open-label pilot 17 registered Completed 5 participants reported; no serious adverse events, transient BP/HR elevation, 4/5 transient headache; Cohen's d −2.1 pain severity, −1.8 pain interference, −2.5 sleep disturbance at 1 month; recruitment stopped early over generalisability concerns and changed FDA psychedelic guidance (PMID 40171515)
NCT06368492 Psilocybin vs hypnosis script Multicentre 35 Recruiting Maastricht
NCT05068791 Psilocybin vs dextromethorphan Early Phase 1 30 Recruiting UAB
NCT05548075 Psilocybin, EEG biomarkers Observational 20 Unknown Imperial College London
NCT06066853 MDMA-assisted therapy, fMRI hyperscanning Early Phase 1 20 Recruiting Spaulding
NCT04938713 Ketamine vs esketamine Crossover 50 Unknown —
NCT04436250 S-ketamine low vs high dose Exploratory 23 Terminated —

The psilocybin pilot illustrates the interpretive trap — and also what a pilot is actually for: safety was its prespecified primary outcome, and that endpoint it answered cleanly (no serious adverse events; transient dosing-day blood pressure/heart rate elevations normalising by end of treatment; transient headache in 4 of 5 participants; PMID 40171515). Effect sizes of d = −2.1 in an open-label study of five people, in the condition with the field's largest documented placebo response, carry essentially no efficacy information. The authors say so themselves, describing the result as establishing a basis for randomised trials. The three ongoing randomised or comparator-controlled psilocybin studies are the ones that will matter.

Immune-directed

Only three registered studies matched immunoglobulin-type interventions, and none is a randomised IVIG trial in FM:

  • NCT03058224 — IGN-ES001 vs acetaminophen in chronic widespread pain with or without FM, IgNova GmbH, n = 230, 23 sites, completed 2017. No publication located.
  • NCT04158752 — galcanezumab (anti-CGRP) for trigeminal/glossopharyngeal pain associated with small-fibre neuropathy/FM, Phase 4, n = 26, completed.
  • NCT04381793 — "Recovery Factors" supplement in CFS and FM, n = 43, completed.

The published human IVIG experience in FM is a retrospective series of 7 patients with both FM and small-fibre neuropathy, reporting fewer symptoms and improved skin-biopsy nerve fibre density after 6 months (Metyas 2020, PMID 31702528). Immune treatment responsiveness has been reported in seropositive (TS-HDS or FGFR-3 antibody) cryptogenic small-fibre neuropathy — 8 patients improving on IVIG with a 42% reduction in pain scores (p = 0.02) — in a population the authors note is "often misdiagnosed as other conditions including fibromyalgia" (Zeidman 2021, PMID 34019003). Given the passive-transfer evidence for pronociceptive IgG in FM (autoimmunity and inflammation), the absence of a randomised IgG-lowering trial is the single most conspicuous gap in the FM pipeline.

4. Neuromodulation

Thirty registered FM trials involve transcranial magnetic stimulation. The field is characterised by many small single-centre studies and almost no multi-site replication.

NCT Modality Design N Status
NCT03658694 rTMS of precentral gyrus, conventional vs sham then conventional vs patterned International multicentre, adaptive, 6 sites 80 Completed
NCT01942538 Maintenance rTMS × 6 months vs sham in 3-week responders 2 sites 78 Completed
NCT05963321 Cerebellar tDCS RCT, 2 sites 92 Active, not recruiting
NCT06009159 Transcutaneous vagus nerve stimulation RCT 60 Recruiting (MGH)
NCT06912334 Transcutaneous VNS Double-blind sham-controlled 120 Recruiting (Athens)
NCT04777500 Auricular transcutaneous VNS — 60 Not yet recruiting (MGH)
NCT07260864 rTMS left vs right motor cortex with fMRI Randomised 20 Recruiting
NCT06746922 Intermittent theta-burst stimulation vs sham RCT 30 Recruiting
NCT05395494 D-cycloserine + iTBS vs placebo Phase 2 47 Terminated
NCT07423377 Vestibulocortical stimulation RCT 56 Recruiting (Mount Sinai)
NCT07239427 tDCS within an active pain-coping programme RCT 68 Recruiting
NCT07323199 HIIT ± tDCS RCT 40 Not yet recruiting
NCT07525063 TMS, pain and reward circuits — 30 Enrolling by invitation (Duke)

The maintenance-rTMS design (NCT01942538) is the methodologically most interesting: it randomises only participants who responded to an initial 3-week course, which is a neuromodulation analogue of enriched enrolment and partially controls for the very large non-specific effect of repeated clinic attendance with a device.

5. Digital and behavioural interventions

This is where FM's largest recent effect size was produced.

NCT Name Design N Status Readout
NCT05243511 PROSPER-FM Phase 3, 25 US community sites, 12-wk self-guided smartphone ACT vs active control (daily symptom tracking + educational materials), blind-to-hypothesis allocation 275 (257 women, 18 men) Completed PGIC responders 71% (99/140) vs 22% (30/135); difference in proportions 48.4% (95% CI 37.9–58.9), p < 0.0001; no device-related safety events (PMID 38991582)
NCT05011162 REACT-FM Real-world smartphone ACT 130 Completed —
NCT05149222 REACT-FM-EXT Extension 50 Completed —
NCT05433337 PROSPER-FM-EXT Extension 60 Completed —
NCT05962437 Spanish STANZA 3-arm RCT: digital ACT vs symptom-tracking app vs treatment as usual, with cost-utility and physiological outcomes 360 Recruiting —
NCT05573685 Click Therapeutics DiNaMo Randomised, single-blind basket across RA, diabetic neuropathy, FM, IBS 124 Completed —
NCT06988761 Mayo Clinic – Lin Health behavioural health RCT 218 Completed —
NCT06208514 Brief Pain Exposure Therapy for nociplastic pain RCT across FM, SLE, chronic pelvic pain, CLBP 125 Recruiting Univ. of Michigan
NCT06567886 At-home morning bright light + sleep stabilisation for chronic nociplastic pain RCT 390 Recruiting Largest active behavioural FM trial
NCT07465991 Digital CBT for insomnia in musculoskeletal disease Longitudinal + embedded RCT 142 Recruiting Oxford
NCT05525598 Internet treatment for severe functional somatic disorders (incl. FM) RCT 166 Recruiting Aarhus

On safety, PROSPER-FM reported no device-related safety events in either arm (PMID 38991582) — which is the expected result for a self-guided app and, read against §7.3, the deeper point: a digital behavioural intervention has no signature adverse event to unblind participants, no discontinuation-for-AE channel to distort imputation, and no risk–benefit trade for a regulator to weigh. The entire tolerability axis on which every drug programme on this page is judged (and on which sodium oxybate died, §8.1) is simply absent.

PROSPER-FM's design choices deserve study by anyone running an FM trial. It used an active comparator rather than waitlist; it used a blind-to-hypothesis approach, telling participants both arms were potentially effective therapies under evaluation; and it reported a responder proportion rather than a mean change. The 22% control response rate is almost exactly the pooled placebo 50%-pain-response rate from drug trials, which is a useful cross-modality consistency check. What it cannot rule out is that a 12-week interactive app produces more engagement, expectancy and behaviour change than a tracking app regardless of the ACT content — the classic attention-control problem in behavioural trials.

6. Exercise and implementation

Forty-eight active or recruiting FM trials involve exercise. Almost all are efficacy studies of a specific modality in a single centre with 20–90 participants — a striking allocation of effort given that exercise already has the strongest evidence base in FM (non-pharmacologic therapy) and the open question is implementation, not efficacy.

The few trials addressing delivery rather than modality:

  • NCT06880653 — dose–response of physical activity on arthritis-attributable outcomes with an adaptive intervention (DREAM), n = 285, across arthritis diagnoses including FM.
  • NCT06948500 — immersive virtual-reality therapeutic exercise with altered visual feedback, n = 80.
  • NCT07304167 — video-based exergame vs home exercise, n = 20.
  • NCT06672419 — AI-guided exercise with computer vision plus pain neuroscience education, tele-rehabilitation, n = 50, 2 sites.

7. Trial-design pathologies specific to fibromyalgia

7.1 Placebo and nocebo response

The quantitative anchor is a systematic review and meta-analysis restricted to randomised, double-blind, placebo-controlled parallel-group trials of the four drugs submitted for FM approval — duloxetine, milnacipran, pregabalin and sodium oxybate. Across 18 studies and 3,546 patients on placebo:

  • pooled estimate of 50% pain reduction on placebo: 18.6% (95% CI 17.4–19.9)
  • pooled estimate of drop-out due to adverse events on placebo (nocebo): 10.9% (95% CI 9.9–11.9) (Häuser 2012, PMID 23137770)

A broader analysis across 30 trials with 3,846 patients on placebo found 30% pain reduction in 30.8% and 50% reduction in 18.8% of placebo recipients, with risk ratios of true drug versus placebo of only 1.38 (30% response) and 1.57 (50% response) (Häuser 2011, PMID 22120916).

The most consequential figure comes from a comparison of FM against painful diabetic peripheral neuropathy across 72 FM trials (9,827 patients) and 70 DPN trials (10,297 patients): the pooled weighted mean difference in the FM placebo arms was 7.69 (95% CI 6.10–9.29) on a 0–100 pain scale versus 13.96 (11.93–15.99) in FM active-drug arms — placebo accounted for 45% of the response in the drug groups in FM. Placebo response correlated with the active-drug effect size (r = 0.69 in FM) and with year of study initiation and baseline pain, but not with age, sex or race (Häuser 2011, PMID 21429668).

Three design consequences follow:

  1. Correlated placebo and active response means the placebo arm is not a nuisance to be minimised independently; trials with high placebo response also have high active response, so the difference — the estimand — is squeezed from both sides.
  2. Baseline pain predicts placebo response, which interacts badly with entry criteria that require a minimum pain score: regression to the mean is built into the design.
  3. Nocebo drop-out near 11% on placebo means adverse-event-driven attrition is not evidence of drug toxicity, and imputation choices for those drop-outs materially change results.

7.2 Enrichment and randomised withdrawal

The standard response to a large placebo response is enriched enrolment randomised withdrawal (EERW): treat everyone open-label, then randomise only responders to continue or switch to placebo. FORWARD (NCT07398417) does exactly this — 12 weeks open-label esreboxetine, then randomisation of responders 1:1 to drug or placebo for 12 weeks, with primary endpoint time to loss of therapeutic response.

The design's problems are documented. A systematic review of 28 EERW chronic-pain trials (13,662 patients in the open/single-blind phase, 7,937 in the double-blind phase) found no consistent primary endpoint: 18 used pain intensity on VAS or 11-point NRS, one used a 4-point NRS, one used loss of therapeutic response and eight used time to loss of therapeutic response — with definitions of that outcome differing considerably between trials. Only 2 of 8 trials using time-to-LTR reported the proportion achieving ≥50% pain relief, versus 14 of 18 using NRS or VAS. The authors recommend NRS as primary endpoint with conservative imputation and time-to-LTR as secondary (Kopsky 2022, PMID 35210848).

Broader methodological critique of chronic-pain trials adds: a consistent bias favouring active over placebo in small, short trials; inflation of results by last-observation-carried-forward imputation when adverse-event withdrawal is high; and a skewed benefit distribution that makes mean pain change misleading relative to responder analysis in which withdrawal counts as non-response (Moore 2013, PMID 23794643).

FM already has its own EERW precedent, in pregabalin. The Cochrane review of pregabalin for FM pain includes two EERW trials: 1,492 participants began open-label dose titration, only 687 were randomised — the majority never reached the double-blind phase — and of those randomised, 40% maintained therapeutic response on pregabalin versus 20% on placebo (NNT 5). Normalised back to the population that started titration, the NNT rises to 12, and about 10% of the initial population achieved maintained response — almost exactly the net-benefit figure the classic parallel-group pregabalin trials produce (Derry 2016, PMID 27684492). That convergence is the most useful single fact for reading FORWARD: enrichment changes which patients are counted, not how many are ultimately helped.

Applied to FM specifically: EERW answers "does the drug maintain benefit in people already benefiting?" That is a legitimate regulatory question and a poor guide to first prescription, because the open-label phase's response is itself substantially placebo. It also launders the safety table: everyone in an EERW double-blind phase has pre-tolerated the drug, so its AE and discontinuation rates cannot be compared with parallel-group safety data (see the stage primer above).

7.3 Subjective endpoints and the absence of an objective anchor

Every primary endpoint in the table above is patient-reported: daily diary NRS pain, PGIC, FIQR, PROMIS. There is no validated objective outcome and no biological entry criterion (biomarkers), which produces four compounding problems:

  • Unblinding via side effects. Distinctive adverse events (oral hypoesthesia in 23.4% on TNX-102 SL; somnolence and dizziness with sodium oxybate; dry mouth with SNRIs) partially reveal allocation, and the outcome is the participant's own rating.
  • No mechanistic falsification. A trial that fails cannot distinguish "the mechanism is wrong" from "the endpoint is too noisy" from "the population was too heterogeneous."
  • Heterogeneous population by construction. Criteria-defined FM almost certainly contains biologically distinct subgroups; an unstratified trial dilutes a real subgroup effect toward null.
  • Comparability failure. Different trials use different responder thresholds and different instruments, which is why cross-trial synthesis in FM is so weak (outcomes and measurement).

7.4 The safety baseline: what the approved drugs' numbers look like

Every new agent's tolerability is implicitly judged against the three FDA-approved drugs, so their pooled safety numbers belong on this page. From the Cochrane syntheses of the registration-era trials:

Drug class Any-AE burden Signature AEs (vs placebo) Discontinuation due to AE Serious AEs
Pregabalin 300–600 mg (5 classic-design RCTs, 3,283 participants, plus 2 EERW trials) 70–90% of participants in all arms, active and placebo Dizziness NNH 3.7; somnolence NNH 7.4; weight gain NNH 18; peripheral oedema NNH 19 (all doses combined) ~10 percentage points higher than placebo No difference vs placebo (very low quality evidence) (Derry 2016, PMID 27684492)
SNRIs — duloxetine and milnacipran (18 RCTs, 7,903 participants) — Nausea, insomnia, somnolence (NNTH calculated per event) 19% (794/4,166) vs 10% (292/2,863), NNTH 14 (95% CI 10–25) No difference vs placebo (RD −0.00, 95% CI −0.01 to 0.00) (Welsch 2018, PMID 29489029)

Three readings. First, the 70–90% any-AE rate in placebo arms of pregabalin trials is the same phenomenon as FINAL's 86% placebo AE rate (§3) and Häuser's 10.9% placebo nocebo drop-out (§7.1): in FM, background symptom load makes "adverse event" a low-specificity signal, and only the difference and the signature events carry information. Second, the SNRI class doubles AE-driven dropout versus placebo — the Cochrane authors' overall judgement was that on average the potential benefits of duloxetine and milnacipran were outweighed by their potential harms (PMID 29489029) — which is the comparator backdrop for TNX-102 SL's 6.1% vs 3.5% discontinuation figures (§2) and for esreboxetine, whose FORWARD open-label phase will generate the comparable first-exposure numbers. Third, serious AEs do not differ from placebo for any approved FM drug; FM drug safety is a tolerability problem, not a catastrophic-risk problem — sodium oxybate (§8.1) being the exception that proves the rule.

8. The completed-trial graveyard

8.1 Sodium oxybate: two positive Phase 3 RCTs and a positive extension, no approval

Study Design N Result
Russell 2011 (PMID 21397402) Phase 3, 14-wk, DB RCT, SXB 4.5 g and 6 g/night vs placebo 548 ≥30% pain reduction in 54.2% (4.5 g) and 58.5% (6 g) vs 35.2% placebo, both p < 0.001; fatigue, sleep disturbance and FIQ all significantly improved
Spaeth 2012 (PMID 22294641) Phase 3, international, 108 centres in 8 countries 573 ≥30% pain reduction 42.0% (4.5 g) and 51.4% (6 g) vs 26.8% placebo; Jenkins sleep score improved 20–25% vs 0.5%
Spaeth 2013 (PMID 24286114), NCT00423605 38-week open-label extension, 130 sites, 7 countries, Jazz Pharmaceuticals 560 57.0% completed; 23.0% discontinued for adverse events; 68.8% achieved ≥30% pain reduction and 69.7% ≥30% FIQ reduction at endpoint

The safety readouts, from the same papers: in Russell 2011 the most common adverse events — headache, nausea, dizziness, vomiting, diarrhoea, anxiety and sinusitis — each occurred at at least twice the placebo incidence (PMID 21397402); in the international trial the AEs with ≥5% incidence and at least twice placebo were nausea, dizziness, vomiting, insomnia, anxiety, somnolence, fatigue, muscle spasms and peripheral oedema (PMID 22294641). In the 38-week extension (n = 560, dose-titratable 4.5–9 g/night), 23.0% discontinued for adverse events, serious AEs occurred in 3.6%, and the most frequent AEs (≥5%) included nausea, headache, dizziness, vomiting, anxiety, insomnia and somnolence (Spaeth 2013, PMID 24286114). Note what the profile is: systemic, CNS-and-GI, dose-carried — the exact opposite of the local-oral trade TNX-102 SL later made (§2), on a molecule PubMed itself indexes under "Central Nervous System Depressants" (MeSH classification of PMID 24286114).

Those responder differences (roughly 15–25 percentage points over placebo) are larger than anything the currently approved FM drugs achieved. Sodium oxybate was formally submitted for approval in FM — it is one of the four agents included in the meta-analysis of "drugs applying for approval for fibromyalgia syndrome treatment" (Häuser 2012, PMID 23137770) — and no FM approval followed: the 2026 review of the field states that the August 2025 approval of TNX-102 SL was the first new FM approval in more than fifteen years (PMID 42479282), which places the last FM approval before 2010 and excludes sodium oxybate.

The reason for non-approval appears to be a risk–benefit story rather than an efficacy story: sodium oxybate is the sodium salt of γ-hydroxybutyrate (GHB), an endogenous GABA-derived compound (Russell 2011, PMID 21397402). The rest of the regulatory account — GHB's controlled-substance scheduling, abuse and CNS-depression liability, the narcolepsy restricted-distribution programme, and the specific record of advisory committee vote, complete response letter, and the agency's stated reasoning — could not be verified from PubMed or ClinicalTrials.gov in this session and is marked [unverified] here; it should be sourced from FDA documents in a later pass.

A separate low-dose GHB investigator-initiated trial was negative: 25 women randomised to 25 mg/kg oral GHB at bedtime versus placebo for 15 weeks, all participants also receiving operant behavioural pain treatment, showed no between-group differences in pain intensity (p = 0.61), depressive mood (p = 0.16), physical impairment (p = 0.25) or sleep quality (p = 0.44), with improvement in both groups attributed to the behavioural therapy (Reuter 2017, PMID 27807735).

What it taught. First, that efficacy is not sufficient in FM: a condition regarded as non-life-threatening, affecting a large population, with substantial psychiatric comorbidity, sets a high safety bar for a controlled substance. Second, that the sleep hypothesis has genuine pharmacological support — sodium oxybate's benefits ran through sleep quality, and the eventually-approved TNX-102 SL is explicitly a sleep-targeting agent with a bedtime pharmacokinetic profile. The mechanism survived; the molecule did not. A systematic review of 47 RCTs and 11,094 participants addressing sleep in FM found that pregabalin and sodium oxybate moderately improved sleep (with uncertainty around the evidence) while amitriptyline, milnacipran and duloxetine showed no significant sleep benefit, and that the largest sleep effect came from CBT for insomnia (SMD −0.63, 95% CI −0.98 to −0.27) rather than from any drug (Pathak 2025, PMID 40084994).

8.2 Sodium channels and the NGF antibodies that never came

The only registered sodium-channel-blocker trial in FM is NCT00401830 — a UCB Pharma parallel, randomised, double-blind, placebo-controlled multicentre proof-of-concept trial of lacosamide 400 mg/day, n = 159, 25 sites, completed 2008. No further FM sodium-channel programme followed, and the current generation of selective NaV1.8/NaV1.7 analgesics has not entered FM trials.

No NGF-antibody trial in fibromyalgia exists. A ClinicalTrials.gov search for tanezumab, fasinumab or nerve-growth-factor interventions in FM returned three studies, none of which involves an NGF-directed agent. This is a negative worth stating precisely, because the FM literature sometimes implies an NGF-antibody failure that did not happen: the anti-NGF programmes were developed in osteoarthritis and chronic low back pain, encountered rapidly progressive osteoarthritis as a class safety signal, and were never taken into FM.

8.3 Terminated trials

Terminations cluster in the small-academic and dose-finding categories: TNX-102 SL NCT02829814 (stopped at 51 participants), D-cycloserine plus iTBS NCT05395494 (n = 47), S-ketamine NCT04436250 (n = 23), cyclobenzaprine ER augmentation NCT01041495 (n = 37), rTMS in Gulf War veterans NCT01608321 (n = 17), and juvenile FM naltrexone NCT00855972 (withdrawn, n = 0). Underpowered single-centre trials that terminate contribute nothing to evidence and consume the same participant goodwill as trials that finish.

9. What the graveyard taught

  1. Effect size is not the binding constraint; risk–benefit is. Sodium oxybate produced the largest responder differences in FM history and was not approved.
  2. Sleep is a validated target. Both the largest failed programme and the newest approval act on sleep, and both showed sleep-quality improvement alongside pain reduction.
  3. Placebo response is the tax on every FM trial, and it is correlated with active response, so it cannot be engineered away by better blinding alone.
  4. Enrichment shifts, rather than solves, the problem. It buys statistical power at the cost of answering a narrower question, with unstandardised endpoints (PMID 35210848).
  5. Non-pharmacological interventions currently produce the largest reported responder differences (48.4 percentage points in PROSPER-FM) — although with an active-control rather than placebo comparator, and without participant blinding.
  6. The pipeline does not match the biology. The strongest mechanistic evidence in FM is for pronociceptive IgG and for a causal gut microbiome contribution, and neither has a randomised, blinded, controlled trial in the registry. Nor does the pipeline yet contain anything nominated by FM's own genetics: the first FM GWAS prioritised the druggable orphan GPCR GPR52 alongside HTT (Kerrebijn 2026, PMID 42521817), and no registered FM trial targets it. Every agent in §2 and §3 is a repurposed drug developed for another indication.

Open questions

  • Will FORWARD's randomised-withdrawal design (NCT07398417) produce a result that generalises to first-line prescribing, given that the open-label lead-in response is itself substantially placebo (PMID 23137770) and that time-to-loss-of-response is defined inconsistently across EERW trials (PMID 35210848)?
  • Why was sodium oxybate not approved despite ≥30% responder rates of 42–58.5% versus 26.8–35.2% on placebo across two Phase 3 trials (PMIDs: 21397402, 22294641)? The efficacy record is public; the regulatory reasoning is not verified here.
  • Does an IgG-lowering intervention (IVIG, plasma exchange, or a B-cell-directed agent) reduce FM pain? Passive-transfer evidence is strong (PMID 34196305) and the registry contains no randomised trial; the only human data are a 7-patient retrospective series (PMID 31702528).
  • Is faecal microbiota transplantation effective in FM under blinded, sham-controlled conditions? The existing human trial is open-label (PMID 40280127) in the condition with an 18.6% placebo 50%-response rate.
  • Do the ongoing randomised psilocybin trials (NCT06368492, NCT05068791) replicate the open-label pilot's large effect sizes (PMID 40171515), or does the effect collapse against an active comparator as low-dose naltrexone did (PMID 38258677)?
  • Can any FM trial demonstrate a biologically stratified treatment effect? NCT06946940 (cannabis response biosignature) and NCT07783295 (CB2/TRPV1 polymorphisms) are the only registered attempts identified, and both are small.
  • Is the 48.4-percentage-point PGIC difference in PROSPER-FM (PMID 38991582) attributable to ACT content or to differential engagement with an interactive versus a passive app? The active-control design constrains but does not resolve this.
  • Given 48 active exercise trials and an already-established evidence base, what is the marginal value of another single-centre exercise-modality comparison versus one adequately powered implementation trial?

References

  1. Abd-Elsayed A, Knezic A, Asfour J, Dewan P, Hasoon J, Reilly M. Sublingual cyclobenzaprine for fibromyalgia: pharmacokinetic rationale, clinical evidence, and place in therapy. Curr Pain Headache Rep. 2026;30:1. PMID 42479282
  2. Aday JS, McAfee J, Conroy DA, et al. Preliminary safety and effectiveness of psilocybin-assisted therapy in adults with fibromyalgia: an open-label pilot clinical trial. Front Pain Res. 2025;6:1527783. PMID 40171515
  3. Bruun KD, Christensen R, Amris K, et al. Naltrexone 6 mg once daily versus placebo in women with fibromyalgia: a randomised, double-blind, placebo-controlled trial. Lancet Rheumatol. 2024;6:e31-e39. PMID 38258677
  4. Bruun KD, Christensen R, Amris K, et al. Effect of naltrexone on spinal and supraspinal pain mechanisms and functional capacity in women with fibromyalgia: exploratory outcomes from the randomized placebo-controlled FINAL trial. CNS Drugs. 2025;39:685-692. PMID 40214857
  5. Cai W, Haddad M, Haddad R, et al. The gut microbiota promotes pain in fibromyalgia. Neuron. 2025;113:2161-2175.e13. PMID 40280127
  6. Derry S, Cording M, Wiffen PJ, Law S, Phillips T, Moore RA. Pregabalin for pain in fibromyalgia in adults. Cochrane Database Syst Rev. 2016;9:CD011790. PMID 27684492
  7. Gendreau RM, McCracken LM, Williams DA, et al. Self-guided digital behavioural therapy versus active control for fibromyalgia (PROSPER-FM): a phase 3, multicentre, randomised controlled trial. Lancet. 2024;404:364-374. PMID 38991582
  8. Goebel A, Krock E, Gentry C, et al. Passive transfer of fibromyalgia symptoms from patients to mice. J Clin Invest. 2021;131:e144201. PMID 34196305
  9. Häuser W, Bartram-Wunn E, Bartram C, Reinecke H, Tölle T. Systematic review: placebo response in drug trials of fibromyalgia syndrome and painful peripheral diabetic neuropathy — magnitude and patient-related predictors. Pain. 2011;152:1709-1717. PMID 21429668
  10. Häuser W, Bartram-Wunn E, Bartram C, Tölle TR. [Placebo responders in randomized controlled drug trials of fibromyalgia syndrome: systematic review and meta-analysis]. Schmerz. 2011;25:619-31. PMID 22120916
  11. Häuser W, Sarzi-Puttini P, Tölle TR, Wolfe F. Placebo and nocebo responses in randomised controlled trials of drugs applying for approval for fibromyalgia syndrome treatment: systematic review and meta-analysis. Clin Exp Rheumatol. 2012;30(6 Suppl 74):78-87. PMID 23137770
  12. Kerrebijn I, et al. The genetic architecture of fibromyalgia across 2.5 million individuals. Nat Med. 2026;32:3060-3070. PMID 42521817
  13. Kopsky DJ, Szadek KM, Schober P, Vrancken AFJE, Steegers MAH. Study design characteristics and endpoints for enriched enrollment randomized withdrawal trials for chronic pain patients: a systematic review. J Pain Res. 2022;15:479-496. PMID 35210848
  14. Lederman S, Arnold LM, Vaughn B, Kelley M, Sullivan GM. Efficacy and safety of sublingual cyclobenzaprine for the treatment of fibromyalgia: results from a randomized, double-blind, placebo-controlled trial. Arthritis Care Res. 2023;75:2359-2368. PMID 37165930
  15. Lederman S, Arnold LM, Vaughn B, Engels JM, Kelley M, Sullivan GM. Pain relief by targeting nonrestorative sleep in fibromyalgia: a phase 3 randomized trial of bedtime sublingual cyclobenzaprine. Pain Med. 2026;27:86-94. PMID 40627411
  16. Metyas S, Chen C, Quismorio A, Abdo N, Kamel K. Improvement of nerve fiber density in fibromyalgia patients treated with IVIg. Curr Rheumatol Rev. 2020;16:280-284. PMID 31702528
  17. Moore RA, Derry S, Wiffen PJ. Challenges in design and interpretation of chronic pain trials. Br J Anaesth. 2013;111:38-45. PMID 23794643
  18. Pathak A, Kelleher EM, Brennan I, et al. Treatments for enhancing sleep quality in fibromyalgia: a systematic review and meta-analysis. Rheumatology (Oxford). 2025;64:4495-4516. PMID 40084994
  19. Reuter E, Tafelski S, Thieme K, et al. [Treatment of fibromyalgia syndrome with gamma-hydroxybutyrate: a randomized controlled study]. Schmerz. 2017;31:149-158. PMID 27807735
  20. Russell IJ, Holman AJ, Swick TJ, Alvarez-Horine S, Wang GY, Guinta D. Sodium oxybate reduces pain, fatigue, and sleep disturbance and improves functionality in fibromyalgia: results from a 14-week, randomized, double-blind, placebo-controlled study. Pain. 2011;152:1007-1017. PMID 21397402
  21. Spaeth M, Bennett RM, Benson BA, Wang YG, Lai C, Choy EH. Sodium oxybate therapy provides multidimensional improvement in fibromyalgia: results of an international phase 3 trial. Ann Rheum Dis. 2012;71:935-42. PMID 22294641
  22. Spaeth M, Alegre C, Perrot S, et al. Long-term tolerability and maintenance of therapeutic response to sodium oxybate in an open-label extension study in patients with fibromyalgia. Arthritis Res Ther. 2013;15:R185. PMID 24286114
  23. Welsch P, Üçeyler N, Klose P, Walitt B, Häuser W. Serotonin and noradrenaline reuptake inhibitors (SNRIs) for fibromyalgia. Cochrane Database Syst Rev. 2018;2:CD010292. PMID 29489029
  24. Zeidman LA, Kubicki K. Clinical features and treatment response in immune-mediated small fiber neuropathy with trisulfated heparin disaccharide or fibroblast growth factor receptor 3 antibodies. J Clin Neuromuscul Dis. 2021;22:192-199. PMID 34019003

Registry results cited

  • ClinicalTrials.gov posted results for NCT04508621 (RALLY) — primary endpoint and adverse-event module, accessed 2026-08-28. No journal publication of RALLY was located in PubMed this session.
  • ClinicalTrials.gov posted results for NCT04172831 (RELIEF) — serious-adverse-event module (supplementing PMID 37165930), accessed 2026-08-28.

Trials cited (ClinicalTrials.gov, retrieved 2026-08-28)

NCT00401830 · NCT00423605 · NCT00568555 · NCT00855972 · NCT01041495 · NCT01608321 · NCT01903265 · NCT01942538 · NCT02015234 · NCT02107014 · NCT02436096 · NCT02589275 · NCT02806440 · NCT02829814 · NCT03058224 · NCT03658694 · NCT04158752 · NCT04172831 · NCT04270877 · NCT04381793 · NCT04436250 · NCT04502251 · NCT04508621 · NCT04739995 · NCT04777500 · NCT04938713 · NCT05011162 · NCT05068791 · NCT05128162 · NCT05149222 · NCT05243511 · NCT05273749 · NCT05395494 · NCT05433337 · NCT05525598 · NCT05548075 · NCT05573685 · NCT05900466 · NCT05962437 · NCT05963321 · NCT06009159 · NCT06066853 · NCT06208514 · NCT06368492 · NCT06567886 · NCT06672419 · NCT06746922 · NCT06880653 · NCT06912334 · NCT06946940 · NCT06948500 · NCT06988761 · NCT07186751 · NCT07230171 · NCT07239427 · NCT07260864 · NCT07293767 · NCT07304167 · NCT07323199 · NCT07382921 · NCT07398417 · NCT07423377 · NCT07465991 · NCT07498023 · NCT07525063 · NCT07585045 · NCT07753018 · NCT07783295