A proof-of-concept multicenter, cross-sectional, prospective case-control cohort study was conducted in 22 FM patients who met the 2010 ACR diagnostic criteria for FM (40) and 30 ME/CFS who fulfilled the 2011 ICC definition for ME/CFS (41). Subjects were recruited by clinicians from two outpatient referral centres (Hospital VIAMED Santa Angela de la Cruz, Seville, Spain and Vall d’Hebron Hospital, Barcelona, Spain) from February 2018 to March 2019. The sample comprised 26 age- and sex-matched sedentary healthy volunteers, who were recruited from each local community by word of mouth. Ten FM patients, 12 with ME/CFS and 13 sedentary healthy volunteers were recruited from the hospital in Sevilla and the rest from the local hospital in Barcelona so as to avoid sampling bias in the selection of the target population. The participants were Caucasian, from the same geographical area, and had a sedentary lifestyle with a similar Mediterranean dietary pattern at the time of inclusion.

The exclusion criteria for participation in the study included current or past diagnosis of autoimmune conditions (such as coeliac disease), food allergies and/or intolerances, haematological conditions, cardiovascular diseases, metabolic and endocrine disturbances (thyroid-related conditions), infectious diseases, neuropsychiatric disorders (psychosis/major depression). Individuals were also excluded if they had morbid obesity, were smokers, were pregnant and/or breast-feeding, or had history of substance misuse or any underlying symptoms that might influence the clinicians’ ability to distinguish between FM and ME/CFS diagnoses.

All study participants were informed of the research procedures and signed a written informed consent form prior to enrolment, in accordance with the 2013 Declaration of Helsinki. This study was approved by the local Research Ethics Committees (reference number: GutME-0634; on January 17, 2018). After providing consent, all participants underwent a clinical examination. Finally, data were analysed in an irreversibly anonymized fashion. A detailed summary of the participants’ demographic and clinical characteristics is shown in Table 1 .

Participants were asked to complete validated self-reported questionnaires under the supervision of two trained investigators, who ensured compliance. These questionnaires included the Fatigue Impact Scale (FIS-40), the Composite Autonomic Symptom Score (COMPASS-31), the Pittsburgh Sleep Quality Index (PSQI), and the Short Form 36-item (SF-36) health survey, which were used to compile data on participants’ demographic characteristics and current health status.

Fatigue was assessed by the Fatigue Impact Scale (FIS-40), a self-administered 40-item questionnaire which includes three subscales (scored from 0 to 4) reflecting the perceived feeling of fatigue in physical (10 items), cognitive (10 items), and psychosocial functions (20 items). The sum of the three scales yields a global score ranging from 0 to 160. Higher scores indicate more functional limitations resulting from fatigue; scores above 120 points are taken to indicate severe fatigue, while scores of 120 points or less are taken to reflect mild/moderate fatigue (42).

The frequency and severity of autonomic symptoms were evaluated by using the validated self-administered 31-item Composite Autonomic Symptom Score (COMPASS-31), comprising six main domains: orthostatic intolerance (4 items), vasomotor (3 items), secretomotor (4 items), gastrointestinal (12 items), bladder (3 items), and pupillomotor systems (5 items). The overall COMPASS-31 score ranges from 0 to 100, with higher scores indicating worse autonomic symptoms (43).

Sleep disturbances were assessed by the standardized, self-administered 19-item Pittsburgh Sleep Quality Index (PSQI) questionnaire, which comprises seven components of sleep quality assessed over a one-month interval (scored from 0 to 3): subjective sleep quality, latency, sleep duration, habitual sleep efficiency, sleep perturbations, use of sleeping medication, and daytime dysfunction. The global PSQI score can range from 0 to 21 points, with score above 5 representing poorer subjective sleep quality (44).

Participants’ general physical and mental health was assessed using the short form 36-item (SF-36) questionnaire, a 36-item self-report health survey conducted over a 4-week period. The SF-36 comprises eight health domains, focusing on limitations in physical activities due to health issues, limitations in social activities due to physical or/and emotional problems, limitations in everyday activities due to health and/or psychological problems, bodily pain, general mental health (including psychological distress and well-being), vitality and overall health perceptions. The eight domains were weighted and summarized in physical component summary (PCS) scores and mental component summary (MCS) scores ranging from 0 to 100. Higher scores indicate better health-related physical and mental quality of life (45).

Fasting blood samples were collected from each participant directly in K₂EDTA tubes (Vacutainer, BD Biosciences, Madrid, Spain) by venipuncture upon confirmation of the diagnosis. Plasma samples were obtained by centrifugation at 2,000 x g for 15 minutes at 4°C within 1 hour of the blood collection; then they were collected immediately and frozen in aliquots at –80°C until further analysis.

Plasma ZO-1 (Catalog #: MBS706368, MyBioSource, San Diego, CA), LPS (Catalog #: MBS266722, MyBioSource, San Diego, CA), sCD14 (Catalog #: RK01060, Abyntek Biopharma, Vizcaya, Spain), and IL-1β (Catalog #: MBS2510385, MyBioSource, San Diego, CA) were measured using commercially available ELISA kits according to the manufacturers’ instructions. Circulating human anti-β-lactoglobulin antibody levels (IgG isotypes) were assayed using a validated home-made ELISA protocol as detailed below. All the plasma samples were measured in blind duplicates for each biomarker. For all protocols, absorbance at 450 nm with 570 nm correction was measured in a microplate reader, and corrected absorbance was interpolated in each standard curve to determine the concentrations (Sigma Plot).

Circulating anti-β-lactoglobulin antibody levels (IgG isotypes) were tested using a standard direct ELISA protocol (46). Briefly, microtitre 96-well polystyrene plates (Nunc, Roskilde, Denmark) were coated with 1 μg/μL β-lactoglobulin from cow’s milk as target protein (Cat # L7880; Sigma Aldrich, Madrid, Spain) diluted in 50 mM Na₂CO₃/NaHCO₃ buffer (pH 9.6) at 4°C overnight. After blocking with 1X phosphate-buffered saline containing 0.05% Tween-20 (PBS-T) and 3% BSA for 90 minutes at room temperature, the pre-coated plates were incubated with the human plasma samples diluted in blocking buffer (dilution from 1:200 to 1:2000) for two hours at room temperature. Plates were washed out three times with PBS-T for five minutes and then each well was incubated with an HRP-conjugated mouse anti-human IgG1 Fc secondary antibody at dilution 1:5000 (Cat # A-10648; Thermo Fisher Scientific, MA, USA) for three hours at room temperature. Between incubations, plates were washed three times with PBS-T. Plates were revealed using TMB substrate (Thermo Fisher Scientific, MA, USA) for 30 minutes at room temperature in the dark and then 50 μl of stop solution (2M sulphuric acid) was added to each well. The absorbance (O.D.) was read immediately at 450 nm with 570 nm correction using a microplate reader (Varioskan Flash, Thermo Electron Corporation, NH, USA). The results were given as antibody levels per ng/ml on a standard curve with anti-human beta-LGB chimera monoclonal antibody (clone 25I2; CABT-L2429; Creative Diagnostics, NY, USA) considering the dilution of each sample. The standard curve was prepared using serial dilutions from 8000 ng/ml to 31.25 ng/ml and included five standard concentrations within the indicated range. Based on the signal-to-noise ratio of assay, the limit of detection was 1.96 ng/ml. In each ELISA assay, endogenous negative and positive controls of human plasma samples were included. Negative controls were tested from healthy donors who had not ingested dairy products for at least one year. Positive controls were assayed from patients recently diagnosed with celiac disease, before the start of treatment, and who habitually ingested dairy products.

All the data obtained were checked for normality with the Kolmogorov-Smirnov and Shapiro-Wilk tests. Normally distributed data were presented as means ± standard error of the mean (SEM). Statistical analysis of non-parametric data in the three study groups was performed using an analysis of non-parametric data evaluated by the Kruskal-Wallis test and compared by Dunn’s multiple comparison test. Scale and subscale scores were examined using the non-parametric Mann-Whitney U test (two-group comparisons). Fisher’s exact test was used to compare the frequency of the reported categorical variables between the groups. Box plots and ROC curves were generated using GraphPad Prism software. The area under the curve (AUC) was calculated so as to compare the overall diagnostic accuracy of gut biomarkers for predicting FM and ME/CFS. Cut-off values with the highest accuracy were selected as the diagnostic cut-off points. Correlations were analysed using Spearman’s test with the R package. Statistical significance was set at P < 0.05 (two-tailed). Only adjusted P-values presenting significant differences are shown. Statistical analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism version 9.5.1 for Windows (GraphPad software, Boston, MA, USA).

Seventy-eight participants, comprising 22 FM patients, 30 ME/CFS patients and 26 healthy-matched controls were included in the study. Table 1 shows the demographics and clinical characteristics of the study population. No significant differences were observed for age, gender, BMI, hemodynamic variables, marital status, illness duration, family background, and concomitant medication between the study cohorts from each site. In this group, most patients with FM (n = 12) and ME/CFS (n = 25) had an illness duration of more than 120 months (10 years), which was self-reported as long-lasting fatigue and chronic pain (data not shown). More than half of the patients reported frequent medication use, including analgesics, non-steroidal anti-inflammatories, anticonvulsants, anxiolytics and antidepressants, but none of the controls were taking medication.

The clinical assessment based on self-reported outcome measures showed that patients with FM and ME/CFS had significantly higher scores on the FIS-40, COMPASS-31 and PSQI questionnaires than controls (all P < 0.0001), whereas healthy controls had significantly higher SF-36 scores than the patients’ groups (all P < 0.0001).

The measurement of intestinal biomarkers proposed in this study revealed that FM patients had higher presence of increased gut permeability and microbial translocation than ME/CFS patients and matched controls. As shown in Figure 1 , plasma levels of intestinal barrier function biomarkers and inflammation were as follows: IgG anti-β-LGB antibodies (A), ZO-1 (B), LPS (C), sCD14 (D), and IL-1β (E) in individuals with FM, ME/CFS, and healthy controls. FM patients had significantly higher levels of IgG anti-β-LGB antibodies, ZO-1, LPS and sCD14 than ME/CFS and controls (P < 0.001 for all). In ME/CFS patients, plasma levels of anti-β-LGB, ZO-1, LPS, and sCD14 were significantly higher than in controls (P < 0.01), but lower than in FM cases (P < 0.001). There were no significant differences in IL-1β levels in patients with FM and ME/CFS and healthy controls.

Correlations between the proposed biomarkers of intestinal barrier function and inflammation and self-reported outcome measure scores in the study cohorts are displayed in Figure 2 . Briefly, in the FM cohort, ZO-1 was significantly correlated with anti-β-LGB antibodies (r = 0.91; P < 0.001), LPS (r = 0.83; P < 0.001) and sCD14 (r = 0.65; P < 0.01), and with physical and mental health component scores on the SF-36 questionnaire (r = 0.51 and r = -0.51; both P < 0.05) respectively. In this cohort, anti-β-LGB was strongly correlated with ZO-1 (r = 0.91; P < 0.001), LPS (r = 0.91; P < 0.001), and sCD14 (r = 0.86; P < 0.001) and also with physical health component scores on the SF-36 questionnaire (r = 0.43; P < 0.05). In contrast, IL-1β negatively correlated with overall COMPASS-31 scores (r = -0.45; P < 0.05) ( Figure 2A ).

Analysis of the ME/CFS cohort showed significant positive correlations between ZO-1 and anti-β-LGB (r = 0.82; P < 0.001), LPS (r = 0.91; P < 0.001) and sCD14 (r = 0.61; P < 0.001), and COMPASS-31 scores (r = 0.45; P < 0.05), whereas anti-β-LGB was positively correlated with ZO-1 (r = 0.82; P < 0.001), LPS (r = 0.89; P < 0.001) and sCD14 (r = 0.65; P < 0.001) ( Figure 2B ).

In healthy controls, ZO-1 was positively and significantly correlated with sCD14 (r = 0.42; P < 0.05) and opposed with physical healthy component scores on the SF-36 questionnaire (r = -0.46; P < 0.05); while anti-β-LGB was negatively correlated with physical and mental health component scores on the SF-36 questionnaires (r = -0.46 and r = -0.48; both P < 0.05), respectively ( Figure 2C ). Multipanel scatter dot plots for statistically significant correlations between intestinal barrier function biomarkers and self-reported outcome measures in FM, ME/CFS and healthy controls are depicted ( Supplementary Figs S1-S3 ).

Analyses of the diagnostic power of each circulating gut function biomarker with regard to predictively distinguishing FM and ME/CFS from healthy controls are displayed in Figures 3 and 4 respectively. As shown in the ROC curve analysis for FM, compared to the reference, anti-β-LGB (AUC = 1.00; 95% CI: 1.00-1.00; P < 0.0001), ZO-1 (AUC = 0.980; 95% CI: 0.94-1.00; P < 0.0001), LPS (AUC = 0.996; 95% CI: 0.98-1.00; P < 0.0001), and sCD14 (AUC = 0.949; 95% CI: 0.88-1.00; P < 0.0001) were able to distinguish between FM patients and healthy controls as demonstrated by the AUC values using a univariate model ( Figures 3A-E ).

ROC curve analysis for ME/CFS showed that compared to the reference, anti-β-LGB (AUC = 0.943; 95% CI: 0.86-1.00; P < 0.0001), ZO-1 (AUC = 0.934; 95% CI: 0.86-1.00; P < 0.0001), LPS (AUC = 0.806; 95% CI: 0.68-0.93; P < 0.0001), and sCD14 (AUC = 0.794; 95% CI: 0.68-0.92; P < 0.0001) were able to distinguish between ME/CFS patients and healthy controls as demonstrated by the AUC values using a univariate model ( Figures 4A-E ).

Also, an ROC curve analysis to predictively distinguish patients with FM and ME/CFS is shown in Figure 5 . It showed that compared to the reference, anti-beta-LGB (AUC = 0.991; 95% CI: 0.97 to 1.00; P < 0.0001), ZO-1 (AUC = 0.882; 95% CI: 0.78 to 0.97; P < 0.0001), LPS (AUC = 0.953; 95% CI: 0.89 to 1.00; P < 0.0001), and sCD14 (AUC = 0.800; 95% CI: 0.67 to 0.92; P = 0.0002) were able to distinguish between patients with FM and ME/CFS as demonstrated by the AUC values using a univariate model ( Figures 5A-E ).