Open questions — fibromyalgia¶
Last curated: 2026-08-28. The research frontier: concrete, answerable questions the assembled literature exposes, each with the evidence showing why it is open and what would close it. Stable IDs (OQ-n); closed questions move to the Closed section with the closing citation. Supporting PMIDs/NCTs were verified during the sessions that wrote them (CONVENTIONS §1).
How to read priority¶
Tier 1 = answering it would change practice or reorder the field; a study design is conceivable today. Tier 2 = important, but blocked on tools, cohorts, or a Tier-1 answer. Tiers are curator judgment — argue with them.
Dots not yet connected¶
Cross-domain junctions where two bodies of evidence both exist in this knowledge base but no study has ever joined them. Synthesis opportunities — often cheaper than new data collection.
| # | Dot A | Dot B | The missing junction | Powers |
|---|---|---|---|---|
| D1 | 26 GWAS loci from 2.5M-person biobank analysis (PMID 42521817) | ACR-criteria-defined clinical FM cohorts | No replication bridges biobank diagnosis labels and criteria-defined FM — nobody knows if they are the same phenotype | OQ-4 |
| D2 | Brain-exclusive heritability enrichment in the FM GWAS (PMID 42521817) | Skeletal-muscle tissue specificity in the chronic-widespread-pain GWAS (PMID 33926923) | The two largest genetic studies point at different organs; no reconciliation analysis exists | OQ-4 |
| D3 | IgG passive-transfer causation in mice (PMID 34196305) | The therapeutic registry: zero randomized IgG-lowering trials (confirmed absence; only a 7-patient retrospective series, PMID 31702528) | The causal claim has never met an interventional test | OQ-3 |
| D4 | Sleep as the strongest modifiable prospective predictor (dose-dependent RR 3.43, PMID 22081440) | Prevention trials: none | A sleep-treatment prevention trial in high-risk populations has never been run | OQ-6 |
| D5 | A drug approved on a sleep mechanism (sublingual cyclobenzaprine 2025, PMID 40627411) | The pain-first incumbents (duloxetine, pregabalin) | No head-to-head trial tests sleep-first vs pain-first strategy | OQ-6 |
| D6 | Diagnostic classifiers with high internal accuracy — fMRI signature 92%/94% (PMID 27583567), metabolomics, spectroscopy | Locked-model, multi-site external validation | Never attempted for any FM classifier — not attempted-and-failed | OQ-2, OQ-7 |
| D7 | Microbiome causation in germ-free mice + open-label FMT relief (PMID 40280127) | A strictly-controlled human microbiome null (zero taxa survive FDR; PMID 42550534) | No blinded FMT trial; no design reconciles the causal chain with the null | OQ-8 |
| D8 | Placebo mechanics precisely quantified (30.8% ≥30%-responders, 45% of drug effect; PMID 23137770, 21429668) | FM trial design | No trial-design innovation (or deliberate clinical harnessing) has been tested against these numbers | OQ-9 |
| D9 | Six live diagnostic criteria sets (PMID 27916278, 30453109) | Outcome validation | No criteria set has ever been validated against prognosis or treatment response — only against other criteria | OQ-1 |
| D10 | Diagnosed and criteria-positive populations barely overlap (κ≈0.3; PMID 27281286, 31777779) | Claims-based genetics, cost estimates, and trial cohorts | Every label-based dataset inherits an unknown phenotype mixture; nobody has quantified the downstream distortion | OQ-1 |
| D11 | Concomitant FM inflates inflammatory-arthritis disease-activity scores (DAS28 +1.24; PMID 29788461) | Treat-to-target escalation trials | No trial tests FM screening before biologic escalation | OQ-14 |
| D12 | Patient-priority outcomes — fatigue, sleep, fibrofog, fluctuation (PMID 18640807) | Regulatory pain-centric endpoints (PMID 21953205) | Nobody has re-scored the trial evidence under patient-weighted endpoints to see if treatment rankings reorder | OQ-5, OQ-16 |
When a junction study appears, connect the dot here and sharpen or close the powered question.
Tier 1 — highest-value directions¶
OQ-1. What population does the FM evidence base actually describe?¶
The measurement crisis under everything: physician-diagnosed and criteria-positive populations are roughly equal in size but only ~26–35% overlapping (κ≈0.3) — ~3/4 of people carrying the diagnosis fail criteria and ~3/4 of criteria-positive people are undiagnosed. Every claims cohort, cost estimate, biobank GWAS, and trial inherits an unknown mixture, and no criteria set has ever been validated against outcomes rather than against other criteria. A criteria-vs-label cohort with long-term outcomes would recalibrate the entire literature. (PMID 27281286, 31777779, 27916278, 32206792) → diagnostic criteria, epidemiology
OQ-2. Is any objective finding specific to fibromyalgia rather than generic to chronic pain?¶
The central-sensitization findings that define the field — augmented fMRI responses, glial PET signal, insular glutamate — largely lack chronic-pain control groups, and where comparisons exist the same signals appear in vulvodynia, low back pain, and sickle cell disease. A multi-condition, locked-protocol comparison would either give FM a biological boundary or dissolve it into "chronic pain in general." (PMID 26864780, 23578957, 25582579, 39615812) → central pathophysiology, biomarkers
OQ-3. Is fibromyalgia (in part) an antibody-mediated disease?¶
Patient IgG reproduces pain hypersensitivity and skin denervation in mice — but every transfer paper shares consortium authors, antibody-binding signal dissociates from pathogenicity in COVID sera, and no randomized IgG-lowering trial exists. Independent replication plus one adequately powered immunotherapy RCT would settle the biggest mechanistic claim of the decade. (PMID 34196305, 40408228, 31702528) → autoimmunity
OQ-4. Do the new genetics replicate in criteria-defined FM — and which tissue do they indict?¶
The 2.5M-person GWAS (26 loci, HTT coding variant, sex-invariant architecture) used heterogeneous biobank labels; nothing shows the loci transfer to ACR-criteria FM, its brain-exclusive heritability contradicts the muscle signal in the CWP GWAS, and no registered trial targets any implicated pathway (e.g., GPR52). (PMID 42521817, 33926923) → genetics, omics
OQ-5. Can treatment response be predicted — for anyone, with anything?¶
Every Cochrane review concludes "a minority benefit substantially"; EULAR's 2017 research agenda demanded response prediction; a decade later no validated baseline predictor exists for any drug or non-drug therapy. Given NNTs of 5–11, prospective responder identification is the single highest-leverage clinical question. (PMID 29489029, 27684492, 27377815) → pharmacologic therapy
OQ-6. Is sleep the causal lever?¶
Sleep problems predict incident chronic widespread pain dose-dependently across four independent prospective cohorts; the newest FDA approval acts on a sleep mechanism; yet no prevention trial has ever targeted sleep in high-risk people, and no head-to-head compares sleep-first vs pain-first pharmacology. (PMID 22081440, 24574238, 40627411) → sleep, fatigue & cognition, pharmacologic therapy
OQ-7. Can any FM diagnostic classifier survive external validation?¶
fMRI signature (92%/94% vs healthy controls), vibrational spectroscopy, metabolomics, microbiome ML — all report high internal accuracy; none has faced locked-model multi-site validation with chronic-pain comparators; and free symptom data (fatigue alone, 86%) currently beats the best metabolite model. The circularity problem (symptom-defined gold standard) needs an explicit design answer. (PMID 27583567, 38411371, 26089700) → biomarkers
OQ-8. Does the gut microbiome cause fibromyalgia pain — in humans, under blinding?¶
Composition differences + mouse FMT causation + open-label human relief vs a strictly controlled null where zero taxa survive correction: the line is directly contested, in a condition with an 18.6% placebo 50%-response rate. A blinded FMT/defined-consortium trial and a harmonized multi-cohort composition study would decide it. (PMID 31219947, 40280127, 42550534, 23137770) → omics
OQ-9. Can trial design tame the placebo/nocebo problem it has quantified?¶
Placebo accounts for ~45% of drug effect, 30.8% of placebo patients reach the 30% responder bar, and 10.9% drop out of placebo arms from adverse events — yet FM trials keep the same parallel-group templates. Design innovation (run-in strategies, context-controlled arms, deliberate clinical harnessing) is testable now. (PMID 21429668, 23137770, 22120916) → outcomes & measurement
Tier 2 — by theme¶
Nosology & epidemiology¶
- OQ-10. Taxon or dimension? No modern latent-structure/taxometric analysis of WPI/SSS item data at population scale exists; the dimensional case rests on regression smoothness. (PMID 23424058, 8774169) → history & nosology
- OQ-11. True incidence under modern criteria — claims-based incidence exceeds plausible criteria-based incidence ~10-fold; no population incidence study under 2016 criteria exists. (PMID 16755239, 10555910) → epidemiology
- OQ-12. Is the sex ratio partly recognition bias? Criteria choice moves F:M from 13.7:1 to 2.3:1; genetic architecture is sex-invariant; male under-detection studies are sparse. (PMID 25323744, 30212526, 42521817, 33771468) → diagnostic criteria, patient experience
- OQ-13. Is the suicide excess (SMR ~3.4) FM-specific, psychiatric-mediated, or ascertainment artifact — and is it modifiable? Adjusted analyses inconsistently lose significance; the two mortality metas disagree in ways that track cohort definition (label vs criteria). No mediation or intervention study. (PMID 37429737, 37929630, 34867495) → epidemiology
- OQ-14. Does screening for concomitant FM before biologic escalation improve inflammatory-arthritis care? Score inflation quantified; management trial missing. (PMID 29788461) → comorbidities
- OQ-15. Post-COVID and post-whiplash FM — selection-biased estimates (post-COVID ~30%) and a 20-year-unresolved trauma contradiction (21.6% vs 1/153) both need controlled prospective designs. (PMID 34426540, 9082932, 16652434) → genetics & risk factors
Mechanism¶
- OQ-16. Do patient-weighted endpoints reorder treatment effectiveness? Patients rank fatigue/sleep/cognition/fluctuation above the pain-centric regulatory endpoints; nobody has re-scored the evidence base accordingly. (PMID 18640807, 21953205, 38333024) → outcomes & measurement
- OQ-17. Small-fiber pathology: cause, consequence, or comorbid subgroup? IgG transfer puts denervation downstream; silent-nociceptor sensitization suggests upstream drive; ~half of series find no severity tracking; no longitudinal ordering study and no treatment-response stratification. (PMID 34196305, 24243538, 30314675) → peripheral pathophysiology
- OQ-18. Why do peripheral blocks reduce evoked hyperalgesia but not clinical pain beyond placebo? Replicated dissociation; expectancy explains a fifth of clinical-pain variance. (PMID 19540671, 24193993) → peripheral pathophysiology
- OQ-19. Are QST abnormalities trait or state? Thresholds shift with any treatment and with baseline anxiety — awkward for a fixed-lesion model and for QST-based trial enrichment. (PMID 39225326) → central pathophysiology
- OQ-20. Does the insula show augmentation or (as newly reported) reduced activation? A direct directional conflict in the imaging literature, unreconciled. (PMID 12115241, 38294039) → central pathophysiology
Treatment¶
- OQ-21. Combination pharmacotherapy at scale — routine in practice, tested only in one n=41 crossover; Cochrane found the literature un-meta-analyzable. (PMID 26982602, 29457627) → pharmacologic therapy
- OQ-22. Why did low-dose naltrexone shrink from pilot promise to powered null? Dose, small-study bias, or endpoint choice — a definitive dose-ranging trial would close it. (PMID 23359310, 38258677, 40540205) → pharmacologic therapy
- OQ-23. Does ANY therapy work beyond 6 months? No reliable long-term efficacy/safety evidence for any FM drug; real-world 1-year adherence ~33%. (PMID 37160297, 22792005) → pharmacologic therapy
- OQ-24. Can the guideline contradictions be reconciled — and will the 2025 approval force it? The contradictions are methodological, not evidential (panel composition predicts recommendations); NICE's umbrella approach drew formal international objection; no consensus process has followed. (PMID 28034828, 34712885) → guidelines, registry
- OQ-25. Exercise works on average — but dose, type-matching, and adherence machinery are unsolved, and effect sizes decay without maintenance. (PMID 28636204) → non-pharmacologic therapy
Patient-centered research¶
- OQ-26. Does earlier diagnosis improve outcomes, or is the post-diagnosis pathway the real target? Diagnosis brings relief then frequent abandonment ("hardly helpful"); no trial tests structured post-diagnosis care. (PMID 20420681, 18041660) → patient experience
- OQ-27. Can clinician-attitude interventions reduce stigma-driven harm? FM ranks last of 38 conditions in physician disease-prestige surveys; attitude problems are quantified; no intervention trial exists. (PMID 28319909, 39093781, 40310228) → patient experience
- OQ-28. Are online FM communities net-protective or net-harmful? Both effects documented observationally; no design separates them. (PMID 30066603, 33355406) → patient experience
- OQ-29. Why is there no FM patient-partnered research infrastructure (registry, priority-setting partnership) of the kind rarer diseases have — and what would it change? (PMID 17349056, 34295021) → patient experience
Closed¶
(None yet. When a question closes, move it here with the closing citation and date.)