Skip to content

Minerbi A, Gonzalez E, Brereton NJB, Anjarkouchian A, Dewar K, Fitzcharles MA, Chevalier S, Shir Y. Altered microbiome composition in individuals with fibromyalgia. Pain. 2019;160:2589-2602. PMID 31219947

One-paragraph summary

Faecal microbiomes of 77 women with fibromyalgia (FM) and 79 control participants were compared using 16S rRNA gene amplification and whole-genome shotgun sequencing. Differential abundance analysis identified significant differences in several bacterial taxa between patients and unrelated controls. Variance in microbiome composition was explained by FM-related variables more than by any other innate or environmental variable measured, and taxon abundances correlated with clinical indices of FM severity. Consistent with the observed changes in butyrate-metabolising species, targeted serum metabolite analysis confirmed differences in circulating butyrate and propionate concentrations. Machine-learning classification using microbiome composition alone separated patients from controls with a receiver-operating-characteristic area under the curve of 87.8%. The authors describe this as the first demonstration of gut microbiome alteration in a non-visceral pain condition.

Key findings

  • Multiple bacterial taxa differ in relative abundance between women with FM and controls; the differences are not attributable to the other innate or environmental variables measured.
  • FM-related variables explained more compositional variance than any other measured covariate — an unusual result, since diet, BMI and age normally dominate microbiome variance.
  • Taxon abundances correlated with clinical severity, not merely with case status, which is the more demanding and more interesting claim.
  • Serum butyrate and propionate concentrations differed between groups, providing a functional metabolite readout congruent with the taxonomic changes in butyrate-metabolising species.
  • Microbiome-only machine-learning classification reached ROC AUC 87.8%, immediately raising the possibility of a diagnostic aid in a condition with no laboratory test.

Limitations

  • Women only, single centre. Generalisability to men with FM — a substantial and under-studied group — is untested.
  • Healthy-control comparator. Patients were not compared against other chronic-pain or rheumatic conditions, so the finding cannot distinguish an FM signature from a generic chronic-pain, deconditioning or medication signature.
  • Cross-sectional. Direction of causation is unresolved by this design; altered activity, sleep, diet quality, gut motility and analgesic and antidepressant exposure could all reshape the microbiome downstream of FM.
  • Medication exposure. FM populations carry heavy exposure to antidepressants, gabapentinoids and opioids, all of which alter gut flora; the design cannot fully partition this.
  • AUC 87.8% is an internal-validation figure on a modest sample with a very high-dimensional feature space — the regime in which classification performance is systematically optimistic. No independent cohort was tested.
  • The reference standard is symptom-based FM criteria, so the classifier is being scored against a questionnaire (see the circularity discussion in biomarkers).

Why it matters

This paper opened the gut–brain line in fibromyalgia and did so with a testable, falsifiable structure: a composition claim, a severity correlation, and a metabolite bridge. The same group then systematically closed the obvious objections — showing that dietary intake does not explain the signature (PMID 35328942), and extending the metabolite bridge to serum secondary bile acids, with α-muricholic acid depletion correlating with pain intensity and fatigue (PMID 35587528). The programme culminated in a causal experiment: faecal transplant from FM patients, but not controls, into germ-free mice induced pain and parallel molecular phenotypes, and replacing the FM microbiota alleviated it (PMID 40280127).

Its significance is therefore double-edged and both edges matter. It is the origin of the only line in FM research that has reached a genuine causal demonstration in an animal model with a rescue arm. It is also the origin of a classification claim (AUC 87.8%) that has since been directly contradicted: a strictly controlled study of paired oral and faecal samples, excluding comorbidity and standardising diet, medication, hormonal phase and recent infection, found no significant differences in α-diversity or β-diversity in either compartment and no genus surviving FDR correction (Celik 2026, PMID 42550534). The tension between a robust causal signal in mice and an absent compositional signal under strict clinical control is the central unresolved question in FM microbiome research, and the most plausible reconciliation — that the effect is real but confined to a phenotypic or comorbidity-defined subgroup — is exactly the stratification hypothesis the field has never tested.

Cited by wiki pages

  • omics and emerging science
  • biomarkers