Summary: An Integrated Neurological Model
The evidence from the NIH deep phenotyping study and decades of prior research supports an integrated model of neurological dysfunction in ME/CFS (Walitt et al. 2024). An initiating trigger such as infection or other stressor disrupts central nervous system homeostasis. Microglial activation persists beyond acute illness, producing chronic low-grade neuroinflammation. Catecholamine and tryptophan pathway abnormalities develop, affecting dopamine, norepinephrine, and serotonin signaling (neurotransmitter dysregulation). The temporal-parietal junction and related regions fail to accurately process effort-related information (integrative brain dysfunction). Parasympathetic withdrawal and sympathetic dysregulation produce cardiovascular and multi-organ effects (autonomic dysfunction). Reduced cerebral blood flow limits brain metabolic capacity (cerebrovascular compromise). Finally, fatigue, cognitive dysfunction, orthostatic intolerance, and other symptoms emerge from these converging abnormalities as the clinical manifestations.
This model explains why ME/CFS patients experience fatigue fundamentally different from normal tiredness: the brain’s basic mechanisms for perceiving, estimating, and responding to effort are dysfunctional. Treatment approaches targeting these specific neurological abnormalities may prove more effective than those addressing peripheral fatigue or deconditioning.
Stimulants (amphetamines, methylphenidate, modafinil) are generally contraindicated in ME/CFS despite their effectiveness in other fatigue conditions. While they may temporarily mask fatigue by artificially boosting alertness and motivation, they do not address the underlying energy deficit and may enable activity levels that exceed the patient’s true physiological capacity. This can precipitate post-exertional malaise (PEM) and potentially cause permanent deterioration (see Warning amphetamine permanent risk). The neurological model presented here explains why: stimulants affect perceived effort and motivation (downstream of the TPJ dysfunction) without correcting the fundamental mismatch between the brain’s effort calculations and actual metabolic capacity. Patients may feel capable of activity that their bodies cannot sustain, leading to crashes. This differs fundamentally from stimulant use in conditions like ADHD or narcolepsy, where the underlying metabolic machinery is intact.
A further concern specific to ME/CFS is a metabolic paradox created by stimulant-induced appetite suppression. Catecholaminergic stimulants (including methylphenidate and amphetamines) suppress hunger signals through dopamine and norepinephrine pathways (Volkow, Wang, and Baler 2012). In a patient with already-impaired energy metabolism, this suppression of compensatory hunger—the body’s attempt to increase substrate availability in the face of cellular energy deficit—removes a critical homeostatic signal. The patient feels less hungry not because their energy needs are met, but because the medication has silenced the hunger circuit. The result is that patients may further reduce caloric intake at precisely the time when metabolic support is most needed. Clinicians using stimulants in selected ME/CFS patients (see Chapter Symptom-Based Management) should actively monitor body weight and encourage structured meal schedules independent of appetite perception.
Certainty: 0.55. Basement membrane thickening in ME/CFS capillaries (Wüst et al. 2024 (Wüst et al. 2024)) impairs both vascular perfusion AND glymphatic clearance—the brain’s waste removal system that operates during sleep through perivascular spaces surrounded by basement membranes. Slow breathing at 6 breaths/min enhances arterial pulsation, which is a key driver of glymphatic flow (Section Neuronal Protein Tracing Reveals Physiological Brain Clearance Architecture). Combined with elevated head sleeping position (15–30°), this intervention may enhance both cerebral perfusion and glymphatic clearance simultaneously.
Evidence base: Bragee et al. (2020) documented that 78% of ME/CFS patients meet IIH (idiopathic intracranial hypertension) criteria, suggesting elevated intracranial pressure that may impair glymphatic drainage (Bragée et al. 2020). Capillary basement membrane thickening (Wüst et al. 2024 (Wüst et al. 2024)) physically restricts glymphatic flow through perivascular spaces. Slow breathing improves POTS symptoms (Ryabkova et al. 2024 (Ryabkova et al. 2024)) via enhanced vagal tone, suggesting central autonomic benefits.
Proposed mechanism: Slow breathing (6 breaths/min, ~10-second cycle: 4-second inhale, 6-second exhale) increases respiratory sinus arrhythmia and arterial pulsatility amplitude. This enhanced pulsation drives CSF flow through perivascular spaces, mechanically overcoming basement membrane restriction. Elevated head position uses gravity to assist venous drainage, reducing intracranial pressure and facilitating glymphatic efflux.
Testable predictions:
- ME/CFS patients practicing 6 breaths/min breathing for 15 minutes before sleep will show improved next-morning cognitive performance (processing speed, working memory) compared to normal breathing
- DTI-ALPS glymphatic index will improve after 4 weeks of nightly slow-breathing + elevated-head protocol
- Patients with confirmed capillary basement membrane thickening (via muscle biopsy) will show greater response than those without structural BM changes
Treatment implications: Slow-paced breathing is zero-cost, low-risk, and addresses both perfusion and clearance simultaneously. Elevated head sleeping (wedge pillow or bed elevation 15–30°) requires no equipment beyond pillows or adjustable bed. This protocol may serve as an adjunct to sleep hygiene and other interventions for unrefreshing sleep and brain fog.
Certainty: 0.35. Bragee et al. (2020) documented that 45% of ME/CFS patients have Chiari malformation type I versus 0.5–1% in the general population (Bragée et al. 2020)—a 45- to 90-fold increase that cannot be explained by congenital prevalence alone. This suggests acquired tonsillar descent: post-infectious dural inflammation and thickening increase intracranial pressure and reduce CSF compliance, pushing cerebellar tonsils inferiorly without requiring structural skull abnormality.
Supporting evidence: Ramirez-Paesano et al. (2023) described CCI (craniocervical instability) as a common finding in ME/CFS, with dural thickening contributing to mechanical compression of neural structures Ramirez-Paesano et al. (2023). The Chiari prevalence in ME/CFS exceeds congenital prevalence by orders of magnitude, strongly favoring acquired over congenital etiology.
Proposed mechanism: Post-infectious inflammation triggers dural mast cell activation and fibroblast proliferation, producing dural thickening and reduced compliance. CSF pressure transmission is impaired, creating pressure gradients that drive tonsillar descent. Unlike congenital Chiari, this acquired form may be reversible with anti-inflammatory treatment or dural decongestion.
Falsifiable predictions:
- ME/CFS patients with acquired Chiari will show dural thickening on cervical MRI (dural diameter >3 mm) compared to congenital Chiari controls
- Anti-inflammatory treatment (e.g., LDN, mast cell stabilizers) will reduce dural thickness measured by serial MRI and improve CSF flow metrics
- Cervical decompression surgery will show differential response: acquired Chiari patients will improve more than congenital controls, suggesting reversibility
Treatment implications: Unlike congenital Chiari which primarily requires surgical decompression, acquired Chiari may respond to anti-inflammatory approaches targeting dural inflammation (mast cell stabilizers, LDN) and reducing intracranial pressure (slow breathing, elevated head sleep position). Surgical intervention should be considered only after inflammatory optimization fails.
Certainty: 0.55. Glymphatic flow occurs through perivascular spaces surrounded by capillary basement membranes (BM). Wüst et al. (2024) documented BM thickening and collagen IV deposition in ME/CFS skeletal muscle capillaries (Wüst et al. 2024). If this extends to cerebral capillaries, BM hypertrophy physically restricts glymphatic flow, producing waste accumulation. Impaired clearance of inflammatory mediators (IL-1\(\beta\), TNF-\(\alpha\)) from perivascular spaces then sustains local inflammation that promotes further BM thickening—a bidirectional impairment loop.
Evidence chain: Glymphatic dysfunction is documented in ME/CFS via reduced DTI-ALPS index (Chaganti 2025 in Long COVID, relevant to ME/CFS overlap) (Chaganti, Talekar, and Brew 2025). Capillary BM thickening is documented in muscle (Wüst 2024 (Wüst et al. 2024)) and likely extends systemically. BM-Glymphatic structural coupling is established in neuroanatomy: perivascular spaces are bounded by basement membranes; BM integrity determines glymphatic hydraulic conductivity.
Proposed mechanism: Step 1: Post-infectious inflammation triggers BM thickening via fibroblast collagen IV deposition Step 2: Thickened BM reduces hydraulic conductivity of perivascular spaces, impairing glymphatic clearance Step 3: Reduced clearance allows inflammatory mediators and metabolic waste to accumulate Step 4: Local inflammation sustains fibroblast activation and collagen IV production, further thickening BM Step 5: Loop reinforces unless interrupted at either structural (BM) or clearance (glymphatic) node
Testable predictions:
- Cerebral capillary basement membrane thickness (measurable via post-mortem tissue or advanced imaging) will correlate inversely with DTI-ALPS glymphatic index
- Patients with more severe BM thickening will show greater symptom burden (brain fog, unrefreshing sleep) independent of other variables
- Interventions that reduce glymphatic load (slow breathing, sleep optimization) will slow BM thickening progression in longitudinal studies
Treatment implications: Interventions should target both arms: (1) reduce BM thickening via anti-inflammatory and antifibrotic approaches, (2) enhance glymphatic clearance despite BM restriction via arterial pulsation enhancement (slow breathing), sleep optimization, and CSF pressure management.
Certainty: 0.40. Cervical instability contributes to ME/CFS symptoms in some patients (Bragee 2020 (Bragée et al. 2020), Ramirez-Paesano 2023 Ramirez-Paesano et al. (2023)). Beyond structural CCI from ligamentous laxity, muscle fatigue may produce functional instability: sustained neck muscle exhaustion reduces tone, impairing proprioceptive input from cervical mechanoreceptors, which creates cervicothoracic instability mimicking CCI without structural laxity.
Proposed mechanism: Neck muscles (deep cervical flexors, multifidi) provide proprioceptive feedback about head position and cervical stability. When these muscles fatigue—which occurs rapidly in ME/CFS due to mitochondrial dysfunction and impaired ATP synthesis—proprioceptive signal quality degrades. The brain receives ambiguous information about cervical stability, reduces confidence in head movements, and may compensate with pathological muscle tension patterns. This functional instability produces symptoms indistinguishable from structural CCI (headache, dizziness, nausea, brain fog) despite normal structural imaging.
Falsifiable predictions:
- ME/CFS patients without structural CCI (normal cervical MRI) will show proprioceptive deficits on cervical joint position sense testing correlating with muscle fatigue scores
- Cervical proprioceptive training (targeted exercises for deep cervical flexors) will improve symptoms even in patients without structural instability
- EMG-measured cervical muscle endurance time will predict symptom severity during head-dependent activities (reading, computer use) independently of structural findings
Clinical implications: Differentiating functional from structural CCI guides treatment: structural CCI requires orthopedic stabilization (collar, PT, potentially surgery), while functional instability responds to proprioceptive training and graded neck muscle endurance exercise without requiring invasive intervention.
Asymmetry-driven functional CCI: Skeletal asymmetry below the cervical spine (pelvic obliquity, rotoscoliosis, leg length discrepancy) imposes sustained compensatory demand on cervical stabilizers to maintain horizontal gaze — the same neck muscles whose fatigue drives functional CCI. In this cascade, asymmetry is the upstream mechanical source: pelvic obliquity → lumbar/thoracic countercurves → cervical tilt → chronic subclinical neck muscle overwork → premature fatigue → proprioceptive degradation → functional CCI (Skeletal Asymmetry as a Primary Mechanical Trigger of the ME/CFS Cascade, Pathway 4). This identifies a potentially reversible driver of functional CCI — postural correction (shoe lifts, orthotics) could reduce the compensatory neck load and delay the onset of proprioceptive fatigue — and provides a testable mechanism linking skeletal asymmetry to the cervicothoracic instability phenotype common in ME/CFS.
Mechanical-autonomic extension: The structural/functional distinction may extend beyond the craniocervical junction to the thoracolumbar sympathetic chain, where vertebral rotation or osteophytic encroachment could produce analogous mechanical dysautonomia (Thoracolumbar Sympathetic Chain Irritation from Skeletal Asymmetry).
Certainty: 0.40. CCI may compress superior cervical ganglion (SCG) and stellate ganglion (cervicothoracic ganglion), producing sympathetic overactivity that mimics or exacerbates POTS. The SCG provides sympathetic innervation to the head (pupil dilation, sweating), while the stellate ganglion innervates the upper extremities and thoracic viscera. Compression at the craniocervical junction could produce dysautonomia that responds to ganglion-specific interventions.
Evidence base: Bragee et al. (2020) documented 45% Chiari prevalence in ME/CFS (Bragée et al. 2020), with high overlap between CCI symptoms and autonomic dysfunction. Ramirez-Paesano et al. (2023) described CCI as a treatable cause of ME/CFS symptoms Ramirez-Paesano et al. (2023). Anatomical studies confirm that SCG and stellate ganglion lie in close proximity to the craniocervical junction; tonsillar descent in CCI could mechanically compress these structures.
Proposed mechanism: Tonsillar descent and dural thickening in CCI reduce space in the craniocervical junction. The SCG and stellate ganglion, positioned laterally to the vertebral column at C2–C3 and C7–T1 respectively, may experience intermittent compression during head movements or sustained poor posture. Compression produces aberrant sympathetic signaling: intermittent compression may cause irritative firing (excessive sympathetic output), while chronic compression may cause conduction block followed by compensatory hypersensitivity.
Testable predictions:
- ME/CFS patients with CCI+ symptoms will show abnormal sympathetic testing (excessive standing norepinephrine, inappropriate heart rate response) disproportionate to other autonomic measures
- Stellate ganglion block (diagnostic/therapeutic injection) will transiently improve POTS symptoms in CCI+ patients
- Upright cervical MRI will show reduced ganglion space correlating with sympathetic overactivity markers
Treatment implications: Stellate ganglion block could serve both diagnostic (improvement confirms sympathetic contribution) and therapeutic (providing symptom relief) purposes. Cervical stabilization (collar, physical therapy, potentially surgery) addresses the root mechanical cause rather than just modulating sympathetic output.
Certainty: 0.10. The sympathetic ganglion compression documented at the craniocervical junction (Cervical Sympathetic Ganglion Compression in CCI) may not be the only mechanical-autonomic interface. The paravertebral sympathetic chain extends from T1 to L2 along the anterolateral vertebral bodies — the same region affected by rotoscoliosis and vertebral rotation. Schulte et al. (2010) documented sympathetic chain dysfunction in 12/31 patients after anterior surgical approach to the thoracic spine for scoliosis correction (Schulte et al. 2010) — a proof of principle that mechanical disruption can alter autonomic output, but a surgical lesion qualitatively different from the chronic, intermittent, spontaneous compression hypothesised here (see Skeletal Asymmetry Hypothesis: Evidence Base is Near-Zero for this distinction). Idiopathic rotoscoliosis or osteophytic encroachment could spontaneously compress or irritate these ganglia, producing multi-organ dysautonomia without requiring CCI. The anatomical distribution predicts specific symptom patterns: T5–T9 compression → splanchnic nerve dysfunction → gastroparesis, hepatobiliary dyskinesia; T10–L1 → renal blood flow dysregulation; T1–T4 → cardiopulmonary symptoms. This mechanism would explain dysautonomia in ME/CFS patients without cervical hypermobility or Chiari — a population currently unexplained by the CCI framework. Origin: Gerlier 2026-07-21, personal communication.
Testable prediction: EOS full-spine radiography in n=200 ME/CFS patients will identify a subgroup with rotoscoliosis >10° or vertebral rotation >grade 2, and this subgroup will have higher COMPASS-31 scores than spinally aligned ME/CFS patients, independent of Beighton score and CCI status.
Consequence: If confirmed, this would expand the structural-autonomic ME/CFS phenotype beyond CCI to include thoracic/lumbar mechanical drivers, identifying patients who may benefit from postural correction and targeted sympathetic chain interventions rather than cervical surgery.
Certainty: 0.30. Autism spectrum disorder (ASD) and ME/CFS share immune dysregulation, mitochondrial dysfunction, and sensory hypersensitivities. The astrocyte-derived protein S100B — elevated in both ASD and ME/CFS — activates microglial RAGE, triggering a calcineurin-dependent NFAT (nuclear factor of activated T-cells) transcriptional program that sustains neuroinflammation (Guloksuz et al. 2014) (Donato 2009). This microglial NFAT loop may represent a shared mechanistic node between neurodevelopmental (ASD) and acquired neuroinflammatory (ME/CFS) conditions.
Mechanism. Step 1: Stressed or damaged astrocytes release S100B into the extracellular space. Step 2: S100B engages RAGE on microglia, activating phospholipase C-gamma and increasing intracellular Ca2+ . Step 3: Elevated Ca2+ activates calcineurin, a Ca2+/calmodulin-dependent phosphatase. Step 4: Calcineurin dephosphorylates NFAT, exposing its nuclear localization signal. Step 5: Nuclear NFAT drives transcription of pro-inflammatory cytokines (IL-6, TNF-alpha), COX-2, and inducible nitric oxide synthase (iNOS). Step 6: Sustained NFAT activity prevents microglial return to the surveillant state, maintaining chronic low-grade neuroinflammation. Step 7: Elevated iNOS-derived NO impairs mitochondrial respiration in neighbouring neurons, contributing to the neuronal energy deficit documented in ME/CFS (Chapter Energy Metabolism and Mitochondrial Function).
ASD-ME/CFS overlap. Both conditions show elevated S100B in serum and CSF, microglial activation on TSPO-PET (where available), and evidence of calcineurin pathway dysregulation. The shared pathway suggests that neuroinflammatory processes in ME/CFS and ASD may converge on a common RAGE-calcineurin-NFAT axis, even though the triggering events differ (neurodevelopmental in ASD, post-infectious in ME/CFS). If confirmed, this would identify S100B as a shared biomarker and calcineurin/NFAT as a shared therapeutic target across diagnostic boundaries.
Falsifiable predictions. (1) ME/CFS CSF will show elevated S100B correlating with microglial activation markers (sTREM2, YKL-40). (2) NFAT nuclear translocation (immunofluorescence) will be increased in post-mortem ME/CFS brain microglia compared to controls. (3) S100B stimulation of ME/CFS-derived microglia (iPSC-derived) will produce exaggerated NFAT-dependent cytokine release compared to control microglia. (4) Calcineurin inhibitors (tacrolimus, cyclosporine A) will reduce NFAT translocation and cytokine production in ME/CFS microglial cultures.
Limitations. S100B has not been specifically measured in ME/CFS CSF — evidence is from serum and from analogies to other neuroinflammatory conditions. The RAGE-calcineurin-NFAT pathway in human microglia is incompletely characterized; most evidence comes from rodent models. Calcineurin inhibitors have significant systemic toxicities (nephrotoxicity, hypertension) that limit clinical translation. The ASD-ME/CFS connection is inferred from parallel elevations in S100B, not from co-morbidity studies or family studies.
Certainty: 0.35. ADHD is overrepresented in ME/CFS populations, and stimulants are contraindicated in ME/CFS (Section Summary: An Integrated Neurological Model, Warning Stimulant Contraindication and Metabolic Paradox), suggesting a shared dopaminergic dysfunction that paradoxically diverges in treatment response. A proposed mechanistic bridge: chronic low-grade dopamine oxidation produces dopamine-quinones that deplete glutathione and impair NRF2-mediated antioxidant defence, disinhibiting the NLRP3 inflammasome in microglia and immune cells (Segura-Aguilar 2018) (Sies 2017).
Mechanism. Dopamine is normally stored in synaptic vesicles. Excessive cytosolic dopamine — from impaired vesicular packaging (VMAT2 dysfunction, now demonstrated in long COVID by PET: Observation Striatal Dopaminergic Terminal Loss Quantified by VMAT2 PET in Long COVID), reduced dopamine transporter (DAT) expression, or mitochondrial ROS impairing vesicular integrity — undergoes auto-oxidation to dopamine-quinones. These quinones: (1) covalently modify cysteine residues on KEAP1, preventing NRF2 nuclear translocation and reducing antioxidant response element (ARE) transcription; (2) deplete glutathione (GSH) directly through conjugation; (3) activate NLRP3 via mitochondrial ROS and lysosomal damage pathways. Unrestrained NLRP3 drives IL-1beta production, which amplifies neuroinflammation and impairs dopamine neuron function, creating a self-reinforcing dopaminergic-inflammatory loop.
ME/CFS context. If this loop operates in ME/CFS, it would explain: (a) the ADHD overrepresentation (shared dopaminergic vulnerability), (b) contraindication of stimulants (dopamine-elevating drugs would accelerate quinone production and NLRP3 activation), and (c) the cognitive dysfunction pattern (impaired prefrontal dopamine signalling from chronic neuroinflammation). The model predicts that ME/CFS patients with ADHD comorbidity would show higher inflammatory markers, worse cognitive function, and greater sensitivity to exertional triggers than those without ADHD.
Falsifiable predictions. (1) ME/CFS patients with ADHD comorbidity will show elevated urinary dopamine metabolites and dopamine-quinone adducts (5-S-cysteinyl-dopamine) compared to ME/CFS without ADHD and healthy controls. (2) NRF2 target gene expression (NQO1, HO-1, GCLC) will be reduced in PBMCs from ME/CFS+ADHD vs ME/CFS-only patients. (3) Stimulant exposure (methylphenidate, amphetamine) in ME/CFS PBMC cultures will increase IL-1beta production via NLRP3, and this effect will be blocked by NRF2 agonists (sulforaphane) or NLRP3 inhibitors (MCC950). (4) NRF2 agonist treatment (sulforaphane, dimethyl fumarate) will improve cognitive function preferentially in ME/CFS+ADHD patients compared to ME/CFS-only in a pilot trial.
Limitations. Dopamine-quinone formation has not been measured in any ME/CFS sample. The ADHD-ME/CFS co-occurrence rate is not precisely known (estimates vary from 5% to 25% depending on assessment method). Stimulant-induced IL-1beta increases in ME/CFS are untested. The NRF2-NLRP3 connection is well-established in other neuroinflammatory contexts (Parkinson’s, multiple sclerosis) but not in ME/CFS. New VMAT2 PET evidence in long COVID (Observation Striatal Dopaminergic Terminal Loss Quantified by VMAT2 PET in Long COVID) provides partial empirical support for the VMAT2 dysfunction premise, though generalisation to non-COVID ME/CFS remains unconfirmed.
Certainty: 0.40. Blitshteyn (2025) has proposed that the dorsolateral inferior medulla — containing nucleus tractus solitarius (NTS), rostral ventrolateral medulla (RVLM), and dorsal motor nucleus of the vagus (DMV) — represents a shared CNS localization for autonomic dysfunction across POTS, ME/CFS, and Long COVID (Blitshteyn 2025) (Wagoner et al. 2019). This region integrates baroreflex afferents, controls sympathetic outflow, and receives humoral signals via the adjacent area postrema (a circumventricular organ with a leaky BBB). Chronic neuroinflammation at this site could produce the autonomic signatures common across all three conditions: sympathetic predominance, baroreflex impairment, and vagal withdrawal.
Evidence. (1) TSPO-PET studies in Long COVID show neuroinflammation in medullary regions, though not specifically resolved to dorsolateral medulla (VanElzakker et al. 2024) (Moen and Iwasaki 2025). (2) Wagoner (2019) documented brainstem inflammation in pediatric autoimmune neuropsychiatric disorders, providing precedent for medullary neuroinflammation as a pathogenic mechanism. (3) Post-mortem ME/CFS studies show brainstem glial activation, though sampling has not focused on dorsolateral medulla. (4) Area postrema lacks a functional BBB, allowing circulating autoantibodies and inflammatory mediators direct access to NTS neurons — providing a mechanistic route from systemic autoimmunity to brainstem dysfunction.
Falsifiable predictions. (1) ME/CFS+POTS patients show higher TSPO-PET signal in dorsolateral medulla than healthy controls and POTS-only patients. (2) TSPO-PET signal in this region correlates with COMPASS-31 orthostatic subscale (r > 0.4) and baroreflex sensitivity (r < -0.4). (3) CSF markers of neuroinflammation (NfL, YKL-40) correlate with medullary TSPO-PET signal (r > 0.5). (4) GPCR autoantibody-positive patients have ≥20% higher medullary TSPO-PET signal than autoantibody-negative patients, linking the autoantibody and neuroinflammation pillars.
Falsified if (a) TSPO-PET signal in dorsolateral medulla does not differ between ME/CFS+POTS and controls (p ≥ 0.05 by group comparison), or (b) TSPO-PET signal does not correlate with any autonomic measure (p ≥ 0.05 for all), or (c) GPCR autoantibody status does not stratify TSPO-PET signal.
Limitations. No TSPO-PET study has specifically resolved the dorsolateral medulla in ME/CFS — current evidence is from whole-brainstem analyses. Wagoner (2019) is from pediatric autoimmune literature, not directly generalizable to ME/CFS. Area postrema–NTS autoantibody access route is anatomically plausible but experimentally untested. TSPO-PET cannot distinguish microglial activation from astrocytic reactivity or endothelial TSPO expression.
Certainty: 0.35. Circulating α1-adrenergic and M2/M4 muscarinic autoantibodies may cross the area postrema (circumventricular organ with leaky BBB) and bind their cognate receptors on NTS baroreflex neurons (Blitshteyn, Doherty, and Steinman 2026) (Sunami et al. 2024) (Gunning et al. 2019). Chronic low-level agonism triggers GPCR internalization via β-arrestin-mediated endocytosis, reducing the density of functional adrenergic and muscarinic receptors on second-order baroreflex neurons. This shifts the baroreflex operating point rightward — the brainstem reads mean arterial pressure as lower than actual because fewer receptors transduce the afferent signal — resulting in inappropriate compensatory sympathetic activation despite normal or elevated MAP, exactly the pattern seen in hyperadrenergic POTS.
Mechanism. Step 1: GPCR autoantibodies in periphery access the CNS via area postrema fenestrated capillaries. Step 2: Autoantibodies bind α1-AR and M2/M4 receptors on NTS neurons, mimicking or blocking endogenous ligand binding. Step 3: Chronic engagement biases receptor signaling toward β-arrestin recruitment rather than G protein activation, triggering clathrin-mediated internalization. Step 4: Internalized receptors are sequestered in endosomes and either recycled slowly or degraded, reducing surface receptor density. Step 5: With fewer functional receptors, the NTS generates a blunted baroreflex afferent signal for any given MAP — the brain perceives lower pressure than actual. Step 6: The RVLM receives an inappropriately strong sympathoexcitatory signal, driving tachycardia and vasoconstriction despite adequate MAP. Step 7: Over weeks to months, chronic receptor internalization produces sustained baroreflex resetting, distinguishing POTS from acute hypovolemic tachycardia.
Evidence. (1) α1-AR and M2 autoantibodies are documented in POTS and ME/CFS (Fedorowski et al. 2017) (Gunning et al. 2019). (2) GPCR autoantibodies from POTS patients show functional activity in cell-based assays (CHO cell cAMP/Ca2+ flux) (Fedorowski et al. 2017) (Kharraziha et al. 2020). (3) Area postrema is adjacent to NTS and lacks BBB — autoantibodies have direct anatomical access. (4) GPCR internalization via β-arrestin in response to chronic agonism is a well-established pharmacological phenomenon.
Falsifiable predictions. (1) In ME/CFS+POTS patients, CSF GPCR autoantibody titers (α1, M2, M4) correlate with baroreflex sensitivity deficit (r > 0.4). (2) Baroreflex operating point (MAP at half-max HR) is shifted rightward in GPCR autoantibody-positive vs negative patients and healthy controls. (3) GPCR autoantibody-positive patients show reduced NTS fMRI BOLD response to baroreflex activation (Valsalva phase IV vs phase II). (4) Immunoadsorption that reduces GPCR autoantibody titers by ≥70% partially restores baroreflex sensitivity within 2–4 weeks.
Falsified if (a) no correlation between CSF GPCR autoantibody titer and baroreflex sensitivity (r ≤ 0.2, p ≥ 0.05), or (b) baroreflex operating point does not differ between autoantibody-positive and negative patients, or (c) autoantibody removal does not improve baroreflex function.
Limitations. NTS GPCR internalization by autoantibodies is inferred from standard pharmacology, not demonstrated in human NTS tissue. Area postrema autoantibody access has not been experimentally confirmed. No study has measured CSF GPCR autoantibodies alongside baroreflex testing in the same patients. Autoantibody internalization time course (weeks to months) is speculative.
Certainty: 0.30. If the dorsolateral medulla harbors persistent antigen — viral remnants, neo-epitopes from oxidative damage, or misfolded proteins — meningeal B cell aggregates or perivascular cuffs could form, creating local autoantibody production that sustains a feed-forward neuroimmune loop (Blitshteyn 2025) (Moen and Iwasaki 2025) (VanElzakker et al. 2024). The NTS/RVLM/DMV region is adjacent to the fourth ventricle choroid plexus, a known immune surveillance site. Chronic neuroinflammation could recruit B cells via CXCL13/CXCR5, establish tertiary lymphoid structures, and drive intrathecal GPCR autoantibody synthesis independent of peripheral B cell activity.
Evidence. (1) B cell aggregates and tertiary lymphoid structures occur in other neuroinflammatory diseases (MS meningeal B cell follicles, NMDA receptor encephalitis perivascular cuffs). (2) CXCL13 is elevated in CSF during neuroinflammation and predicts intrathecal B cell recruitment. (3) GPCR autoantibodies are found in CSF as well as serum in ME/CFS (Bynke et al. 2020), consistent with at least some CNS production.
Predictions. (a) CSF GPCR autoantibodies differ from paired serum in titer ratio — CSF/serum index > 1.5 for ≥1 GPCR target — indicating intrathecal synthesis rather than passive diffusion. (b) CSF oligoclonal bands specific to GPCR targets are present in ≥20% of ME/CFS+POTS patients. (c) TSPO-PET signal in dorsolateral medulla correlates with CSF autoantibody titer (r > 0.5) but not with serum titer. (d) Post-mortem medulla from ME/CFS+POTS patients shows CD138+ plasma cells and CD20+ B cells in perivascular cuffs around NTS.
Falsified if (a) CSF GPCR autoantibody index (CSF/serum ratio relative to albumin ratio) does not exceed 1.0 in any patient, or (b) no oligoclonal bands are found in CSF, or (c) TSPO-PET does not correlate with CSF autoantibody measures.
Limitations. CXCL13 has not been measured in ME/CFS CSF. No post-mortem medulla immunohistochemistry exists for ME/CFS — the entire hypothesis rests on analogies to MS and autoimmune encephalitis. Intrathecal autoantibody production, if present, would not distinguish cause from consequence (reactive B cell infiltration vs primary CNS autoimmune process).
1 GLP-1 Receptor Agonists: Central Nervous System Research Directions
The genetic pathway enrichment findings (Chapter Genetic and Epigenetic Factors, Section Convergent Brain-Enriched Genetic Architecture in Fibromyalgia and ME/CFS) and the orexin-GLP-1 reciprocal dysregulation hypothesis Ruhrländer, Schieffer, and Schieffer (2026) together motivate CNS-directed research on GLP-1 receptor agonists in ME/CFS. All considerations below are research-stage; no clinical data exist.
No study has measured whether ME/CFS patients have normal endogenous GLP-1 secretion, glucose-dependent insulinotropic polypeptide (GIP), or related incretin responses to a standardised meal. This is the most basic unresolved question. If GLP-1 secretion is blunted — due to dysbiosis, vagal dysfunction, or L-cell mitochondrial impairment — then GLP-1 RAs would function as hormone replacement. If secretion is normal, the genetic pathway enrichment reflects receptor-level or downstream pathway involvement, and GLP-1 RAs act supraphysiologically. A simple meal tolerance test with serial GLP-1, GIP, glucose, insulin, and C-peptide measurements (0, 15, 30, 60, 90, 120 min) would answer this question.
Prediction: Blunted GLP-1 AUC in ME/CFS vs BMI-matched controls during mixed meal tolerance test. GLP-1 AUC correlates negatively with fatigue severity and positively with vagal tone (HRV high-frequency power). Null result: no difference, falsifying the deficiency hypothesis.
Status: Standard endocrine assay; zero ME/CFS data. Estimated cost: USD 200–300 per patient.
Certainty: 0.25. (0.20→0.25: convergence with Central NE Deficiency — GLP-1 incretin resistance and catecholamine deficiency both predict cerebral energy homeostasis failure via complementary pathways from different physiological systems.)
GLP-1 receptors on neurons and glia regulate cerebral glucose uptake, mitochondrial biogenesis, and insulin sensitivity. If ME/CFS involves central GLP-1 resistance — analogous to the well-documented central insulin resistance in Alzheimer’s disease (“Type 3 diabetes”) — neurons would be metabolically starved despite normal or elevated blood glucose. This could explain the reduced cerebral glucose metabolism observed in some ME/CFS PET studies and the subjective experience of “brain fog” as “brain not getting enough fuel.” GLP-1 RAs at supraphysiological doses might overcome resistance, improving neuronal energy availability independent of peripheral effects (Athauda et al. 2026).
Testable prediction: Acute IV GLP-1 (7-36) infusion in ME/CFS should show blunted cerebral FDG uptake response compared to healthy controls (FDG-PET). Chronic GLP-1 RA treatment should increase resting cerebral glucose metabolism in frontal cortex.
Status: Entirely inferential. No central GLP-1 sensitivity measurements exist in any disease.
Certainty: 0.25. (0.20→0.25: convergence with State-Dependent Endothelial Dysfunction — glycocalyx shedding in T2D microvascular literature and state-dependent ED in ME/CFS physiology point to same state-dependent endothelial failure from complementary domains.)
Endothelial dysfunction in ME/CFS (Chapter Cardiovascular Dysfunction) includes glycocalyx degradation, impaired flow-mediated dilation, and microvascular permeability. GLP-1 receptors on endothelial cells activate eNOS, increase NO bioavailability, and preserve glycocalyx integrity. By stabilising the glycocalyx, GLP-1 RAs could reduce exercise-induced microvascular leakage and immune cell extravasation into muscle tissue — mechanically interrupting the cascade that leads to PEM. Exercise-induced glycocalyx shedding is a normal physiological response that resolves within hours; pathological persistence occurs in endothelial dysfunction and parallels the 24–48h delayed worsening characteristic of PEM.
Testable prediction: ME/CFS patients should show elevated serum glycocalyx degradation products (syndecan-1, hyaluronan) at 48h post-CPET compared to sedentary controls. A 4-week course of liraglutide should reduce post-exercise glycocalyx shedding and correlate with reduced PEM severity.
Status: No glycocalyx study in ME/CFS. GLP-1 RA glycocalyx protection established in T2D (meta-analytic).
Certainty: 0.35. (0.30→0.35: convergence — GLP-1 RA NK enhancement from T2D pharmacology and clinical evidence of corticosteroid NK suppression both indicate NK function is pharmacologically modifiable, via complementary evidence from different domains.)
NK cell dysfunction is one of the most replicated immune findings in ME/CFS (Chapter Immune System Dysfunction). GLP-1 receptors are expressed on NK cells and their activation enhances cytotoxicity in T2D (Deng, Chen, and Shi 2025). No study has examined GLP-1 RA effects on ME/CFS NK cells. Ex vivo exposure of ME/CFS PBMCs to GLP-1 RA (liraglutide, semaglutide) would provide the first human ME/CFS-specific mechanistic data with zero patient drug exposure.
Testable prediction: ME/CFS PBMC NK cells exposed to GLP-1 RA (1–100 nM) should show dose-dependent increase in K562 target cell killing vs vehicle. Effect size correlates with baseline NK dysfunction severity. Null: no effect, narrowing the immune pathways potentially responsible for any clinical benefit.
Status: Feasible with small blood volumes; ex vivo design eliminates tolerability concerns; zero ME/CFS data.
2 Research Priorities and Cross-Disease Bridges
The major barrier to GLP-1 RA neuroprotection is blood-brain barrier penetration. Most GLP-1 RAs reach CNS primarily through circumventricular organs and vagal afferents — limited and variable access. Intranasal delivery bypasses the BBB and delivers GLP-1 directly to the brain via olfactory and trigeminal pathways, achieving high CNS concentrations with minimal systemic exposure. This could target neuroinflammation, cerebral glucose metabolism, and orexin-system modulation without GI adverse effects, weight loss, or cost of systemic GLP-1 RAs. Intranasal insulin is under investigation in Alzheimer’s disease. Intranasal GLP-1 has shown CNS effects in healthy volunteers (reduced appetite, altered fMRI brain activity). No intranasal GLP-1 study in any disease has progressed past Phase 1.
Status: Technically feasible but unvalidated. Formulation stability and safety in ME/CFS entirely unknown. Falsifiability: intranasal GLP-1 achieves ≤2× CSF-to-plasma ratio vs subcutaneous route → CNS-targeted delivery advantage is negligible, falsifying the route-specific hypothesis.
If central GLP-1 resistance is present in ME/CFS, measuring the acute cerebral metabolic response to IV GLP-1 (7-36) infusion by FDG-PET could provide a direct CNS GLP-1 sensitivity readout. Healthy brains show increased FDG uptake in prefrontal cortex and hypothalamus after GLP-1 infusion. Blunted FDG response would indicate central GLP-1 resistance, suggesting need for higher doses or CNS-penetrant compounds. The cost (approximately USD 3,000–5,000) and radiation exposure limit scalability to proof-of-concept studies.
Status: GLP-1 FDG-PET response documented in small healthy/T2D studies (n=10–20). No ME/CFS infusion study exists. CNS resistance not established. Falsifiability: acute IV GLP-1 at the highest safe dose produces no significant increase in prefrontal FDG uptake vs saline in ME/CFS → falsifies central GLP-1 sensitivity as a measurable phenomenon.
A single biomarker will never capture the full GLP-1 pathway. An integrative score combining four cheap, accessible measures could classify patients by pathway integrity: (1) postprandial GLP-1 AUC (meal tolerance test), (2) fasting DPP-4 activity (serum fluorogenic assay), (3) HRV high-frequency power (vagal tone, 5-min recording), and (4) faecal butyrate producers (qPCR for Faecalibacterium prausnitzii, Roseburia). Each component independently reflects GLP-1 pathway function (secretion, degradation, neural sensitivity, nutritional input). Combined into a weighted score (e.g., PCA-derived), it would identify patients with “multi-level GLP-1 pathway failure” vs isolated defects. No integrative GLP-1 score exists in any disease.
Status: Components individually validated; integrative approach novel; validation requires treatment outcome data. Falsifiability: score predicts GLP-1 RA treatment response with AUC ≤0.55 → individual components mis-classify individually and in combination, score offers no predictive value over chance.
Gardner 2026 reports protective ME/CFS alleles overlap with T2D protection. This creates a testable prediction: GLP-1 RA effects in ME/CFS may occur through the same pathways that protect against T2D, which are distinct from glucose lowering — specifically, anti-inflammatory (NF-kB, CRP), endothelial (eNOS, glycocalyx), and mitochondrial (PGC-1α, biogenesis) effects (Ren et al. 2025) (Athauda et al. 2026). If true, GLP-1 RAs could benefit normal-weight, normoglycaemic ME/CFS patients as much as pre-diabetic ones, and baseline HbA1c would not predict symptom response. This would substantially expand the eligible population for any future trial.
Testable prediction: In a GLP-1 RA ME/CFS trial, baseline HbA1c will NOT predict symptom response. CRP reduction and symptom improvement will correlate, but neither will correlate with HbA1c change. Falsified if only pre-diabetic ME/CFS patients respond.
Status: Glucose-independence of GLP-1 RA anti-inflammatory effects established in T2D. Protective allele overlap new. No normoglycaemic ME/CFS data.
Standardised mixed meal tolerance test with serial GLP-1, GIP, PYY, and glucose measurements (0, 15, 30, 60, 90, 120 min) could stratify patients into low/normal/high GLP-1 responders. Patients with low GLP-1 AUC are predicted to benefit most from GLP-1 RAs (replacement rationale). Those with normal/high GLP-1 AUC may benefit more from DPP-4 inhibitors (which raise already-adequate GLP-1) or may be GLP-1 RA non-responders. This biomarker would enrich a trial for predicted responders, reducing sample size requirements.
Testable prediction: In a GLP-1 RA trial, baseline GLP-1 AUC tertile significantly predicts treatment response. Lowest tertile shows ≥2× the effect size of highest tertile for PEM improvement. Falsified if response is uniform across tertiles.
Status: Standard endocrine assay; GLP-1 AUC variability 2–3× between healthy individuals; zero ME/CFS data.
Serum DPP-4 activity can be measured by a simple fluorogenic substrate assay (approximately USD 20/sample). Elevated DPP-4 predicts rapid GLP-1 degradation. Patients with high DPP-4 activity may benefit more from DPP-4 inhibitors (blocking degradation) than from GLP-1 RAs (which are DPP-4-resistant but still rely on endogenous GLP-1 production for vagal afferent effects). Patients with low DPP-4 activity may be GLP-1 analogue-responsive regardless. This biomarker stratifies the choice between two drug classes — DPP-4 inhibitors (cheap, oral, well-tolerated) vs GLP-1 RAs (injectable, costly, less well-tolerated) — based on individual biology, not assumption.
Testable prediction: In a crossover trial of sitagliptin vs liraglutide, baseline DPP-4 activity predicts the better-tolerated and more-effective drug at an individual level. High DPP-4 → sitagliptin response; low DPP-4 → liraglutide response.
Status: DPP-4 activity varies 3× between individuals, genetically regulated, modifiable by inflammation. Zero ME/CFS data. DPP-4→treatment selection untested in any disease.
GLP-1 exerts many CNS effects through vagal afferents (NTS → hypothalamus pathway), not direct BBB penetration. Vagal tone is reduced in ME/CFS (HRV evidence), which would blunt the sensitivity of the vagal GLP-1 signalling pathway even if GLP-1 levels are normal. Auricular transcutaneous vagal nerve stimulation (tVNS) activates vagal afferents directly, potentially restoring the sensitivity of the GLP-1→vagal→CNS axis. This would enhance the effect of endogenous GLP-1 without changing GLP-1 levels. tVNS is non-invasive, safe, and self-administered at home.
Testable prediction: 20 min daily auricular tVNS × 8 weeks increases postprandial vagal HRV response (high-frequency power) and subjective energy (VAS) after a standardised meal vs sham stimulation. Null: tVNS improves HRV but does not alter GLP-1-related outcomes, confirming vagal dysfunction but suggesting vagal and GLP-1 pathologies are dissociable.
Status: Vagal GLP-1 pathway well-characterised in preclinical models. tVNS RCTs show benefit in pain, inflammation. No tVNS × GLP-1 combination study exists.
Endogenous GLP-1 is released from intestinal L-cells in response to nutrients. L-cells express mechanosensors (PIEZO2, TRPA1) and chemosensors (FFAR2/3 for SCFAs). The gut → vagus → NTS → hypothalamus circuit is the major physiological route for GLP-1 signalling, independent of BBB penetration. In ME/CFS, dysbiosis (reduced SCFAs, altered bile acids), slow gut motility, and vagal dysfunction may collectively suppress endogenous GLP-1 release, creating a vicious cycle: low GLP-1 → reduced vagal tone → gut dysmotility → further GLP-1 suppression. This cycle could be interrupted by prokinetics, prebiotics, or vagal nerve stimulation — entirely without pharmacological GLP-1 RAs. The orexin-GLP-1 reciprocal regulatory cycle described by Ruhrländer 2026 is positioned within this broader gut-brain axis Ruhrländer, Schieffer, and Schieffer (2026).
Testable prediction: ME/CFS patients show blunted postprandial GLP-1 response to a mixed meal vs controls. Baseline GLP-1 correlates with α-diversity and SCFA (butyrate) levels. Vagal tone (HRV high-frequency power) predicts GLP-1 response amplitude. Restoring SCFA levels (prebiotics, butyrate) increases GLP-1 AUC.
Status: Each link individually plausible. Integrated cycle entirely hypothetical. ME/CFS GLP-1 secretion dynamics unmeasured.
GIP receptors are also expressed in the brain, and GIP-GLP-1 co-agonism (tirzepatide) shows synergistic metabolic effects. However, Forny Germano 2024 found tirzepatide ineffective in Alzheimer’s mice — the same negative result as semaglutide — suggesting that dual agonism alone is insufficient without CNS penetrance (Forny Germano et al. 2024). The practical question: does combined incretin agonism cover a broader range of PrecisionLife-identified ME/CFS genes than GLP-1 alone? This is answerable computationally by intersecting the ME/CFS risk gene set with GLP-1R, GIPR, and GCGR target pathways. If GIPR/GCGR targets add no significant coverage beyond GLP-1R alone, dual agonism offers no genetic advantage.
Testable prediction: GIPR and GCGR target pathways do NOT add significant additional ME/CFS gene coverage beyond GLP-1R targets. Falsified if dual agonism significantly expands pathway coverage.
Status: Computational analysis feasible now without any patient data. Vear 2025 Nature Metabolism reviews dual/triple agonists. No ME/CFS-specific compound data.
PrecisionLife is developing combinatorial genetic classifiers to predict GLP-1 RA response. An intermediate validation step before any ME/CFS patient exposure: retrodict GLP-1 RA efficacy using the ME/CFS-derived classifier in existing T2D/obesity GLP-1 RA trials (EXSCEL, REWIND, LEADER), where genotype and outcome data exist. If the ME/CFS-derived classifier predicts GLP-1 RA response in non-ME/CFS populations, it cross-validates the genetic pathway enrichment framework without requiring an ME/CFS trial. This could raise the certainty of the PrecisionLife findings from 0.45 (conference presentation) to ~0.60 (independent validation).
Testable prediction: The classifier predicts HbA1c response in existing T2D GLP-1 RA trial cohorts. Falsified if no cross-cohort prediction is achieved.
Status: Requires access to existing GLP-1 RA trial genotype data. No such analysis has been reported. Fast, cheap derisking step.
No pharmacological intervention has been shown to prevent or reduce PEM. Since PEM involves delayed (24–72h) worsening after exertion, a substudy embedded in a trial would use standardised CPET provocation at baseline and end-of-treatment, with serial symptom and biomarker measurement at 0, 24, 48, 72h. The hypothesis: GLP-1 RA pre-treatment reduces the AUC of PEM symptoms and post-exertional inflammatory markers (IL-6, CRP) compared to placebo. This directly tests whether GLP-1 RAs modify the defining symptom of ME/CFS.
Testable prediction: GLP-1 RA × 8 weeks reduces CPET PEM symptom AUC ≥30% vs placebo, with parallel reduction in post-CPET IL-6 spike at 24h. Null: no effect on PEM timeline despite improved baseline symptoms, suggesting GLP-1 RA mechanism is independent of PEM generation.
Status: PEM provocation is ME/CFS-specific methodology. No prior GLP-1 RA PEM provocation study. Unclear whether single-dose or chronic dosing needed.
Qi et al. 2024 (Nature Metabolism) demonstrated that Desulfovibrio-produced H2S suppresses GLP-1 secretion from L cells by inhibiting mitochondrial function (Qi et al. 2024). Bismuth subsalicylate chelates luminal H2S, rescuing GLP-1 production in mice. ME/CFS patients have altered gut microbiota including potential increases in H2S producers. BSS is an OTC, cheap, well-tolerated, oral medication with >50 years of safety data. This provides a mechanism to increase endogenous GLP-1 without injected drugs.
Testable prediction: BSS 524 mg QID × 2 weeks increases postprandial GLP-1 AUC by ≥50% in ME/CFS vs placebo. GLP-1 increase correlates with reduction in breath H2S. Null: no GLP-1 change, suggesting H2S-mediated GLP-1 suppression is not relevant in ME/CFS.
Status: Compelling mouse model; OTC human safety established; zero human GLP-1 data with BSS in any condition.
Palmitoylethanolamide (PEA) activates PPAR-α, which upregulates GLP-1 receptor expression in the brain. By increasing GLP-1R density, PEA could enhance the response to both endogenous and exogenous GLP-1 without increasing GLP-1 levels — a “sensitivity enhancer” approach distinct from secretagogues or receptor agonists. PEA is OTC, safe, with positive RCT data in fibromyalgia and chronic pain. PEA also reduces mast cell activation (relevant to ME/CFS-MCAS). No study has examined PEA × GLP-1 axis interaction.
Testable prediction: PEA 600mg BID × 8 weeks increases CSF GLP-1R expression (surrogate) and enhances GLP-1 response to a mixed meal (higher peak, longer duration) in ME/CFS vs placebo.
Status: PEA safety established; GLP-1R upregulation mechanism is preclinical; combination effect entirely unexamined.
The orexin↔︎GLP-1 reciprocal disruption described by Ruhrländer Ruhrländer, Schieffer, and Schieffer (2026) suggests that neither pathway alone can be re-normalised without addressing its partner. A triple-intervention protocol simultaneously targeting GLP-1 (dietary/protein-first), glymphatic clearance (lateral sleep positioning, sleep hygiene), and orexin stabilisation (morning light exposure, circadian meal timing) could break the self-reinforcing cycle. Each intervention alone may be insufficient; combination may produce synergy.
Testable prediction: A 4-week triple protocol (morning protein-first meal + lateral sleep + 30 min morning bright light) improves DSQ-PEM by ≥40% vs a single intervention (protein-first alone). Triple synergy indicated by disproportionate improvement vs additive individual effects.
Status: Entirely hypothetical. No triple circuit study exists. Adherence burden may be high. Each component is zero-cost and zero-risk.
Severe/very-severe patients cannot access injectable GLP-1 RAs (cost, tolerance, refrigeration, travel). Even oral drugs may be problematic (swallowing difficulty, GI sensitivity). An adapted approach: (a) sublingual berberine (bypasses GI, avoids nausea), (b) liquid whey protein for GLP-1 stimulation (easier than solids), (c) chronobiotic meal timing within the patient’s achievable schedule, (d) transdermal magnesium (supports GLP-1 secretion via calcium-dependent pathways). Focus entirely on endogenous GLP-1 enhancement — no exogenous drug. This addresses the population most in need and least served.
Testable prediction: Severe-ME/CFS-adapted protocol (sublingual berberine 200mg + liquid whey protein 20g, both TID) is tolerable (≥80% completion) and shows ≥20% improvement in Bell Disability Scale at 12 weeks.
Status: Severe ME/CFS adaptation is needed regardless. GLP-1 relevance in severe subgroup unknown. Safety profile acceptable for all components individually.
Two human disease models offer potential mirrors of GLP-1 dysregulation relevant to ME/CFS: (1) Post-gastric-bypass reactive hypoglycaemia, driven by GLP-1 hypersecretion from remodelled gut anatomy — producing autonomic symptoms (palpitations, sweating, tremor, fatigue) that partially overlap ME/CFS. If ME/CFS involves GLP-1 deficiency (rather than excess), the symptom profile should be the inverse. (2) Anorexia nervosa, involving chronic caloric restriction, reduced GLP-1 secretion, fatigue, and bradycardia. Unlike AN, where refeeding restores GLP-1, ME/CFS GLP-1 deficiency (if present) may be calorie-independent (L-cell dysfunction, vagal dysregulation, DPP-4 elevation). Comparing meal-challenge GLP-1 dynamics across post-bypass, AN, and ME/CFS would test whether GLP-1 pathophysiology is relevant to ME/CFS symptom generation and whether the deficiency is primary or secondary to low intake.
Testable prediction: ME/CFS patients have lower GLP-1 AUC than BMI-matched controls and comparable GLP-1 to AN patients — but unlike AN, GLP-1 does NOT normalise after 4 weeks of caloric normalisation. Post-bypass patients do not report ME/CFS-like PEM.
Status: Post-bypass and AN GLP-1 data are well-characterised. ME/CFS GLP-1 data entirely absent. Cross-disease comparison is novel and testable with standard meal challenges.
The existing pathophysiological DAG (Chapter Integrative Models and Multi-System Pathophysiology) lacks a GLP-1 node. Several quantitative extensions are warranted: (1) Endogenous GLP-1 dynamics node — add [Endogenous GLP-1] with input edges from gut SCFA levels, meal protein content, circadian phase, and DPP-4 activity; output edges to orexin stability, cerebral glucose uptake, and systemic inflammation. The GLP-1 node should be modelled with a circadian forcing function (peak AM, trough PM) and a nutrient-input response function. (2) DPP-4/CD26 degradation kinetics — modelled as Michaelis-Menten: d[GLP-1]/dt = secretion_rate(t, nutrients) − (Vmax_DPP4 × [GLP-1])/(Km + [GLP-1]). If DPP-4 activity is elevated in ME/CFS, the Vmax term increases, producing GLP-1 deficiency non-linearly: a 2× DPP-4 increase requires a 3–4× secretion increase to maintain normal GLP-1 levels. (3) GLP-1 RA PKPD model with subgroup-specific parameters — extend existing T2D PKPD models by adding ME/CFS-specific parameters (reduced vagal tone shifts EC50 for CNS effects; reduced cerebral glucose metabolism alters dose-response for cognitive endpoints; baseline inflammation modifies anti-inflammatory dose-response). These extensions would predict that inflammatory-subgroup patients require lower doses for anti-inflammatory effects but higher doses for CNS effects — informing trial design.
Status: GLP-1 dynamics well-quantified in endocrinology. ME/CFS-specific parameters unmeasured. DPP-4 kinetics established (Km ~5 µM for GLP-1). GLP-1 RA PKPD models exist for T2D. All parameterisation requires ME/CFS-specific data. Falsifiability: GLP-1-augmented DAG does NOT improve PEM severity prediction (ΔR² ≤ 0.10) over the current model without GLP-1 node → GLP-1 pathway adds negligible explanatory power.
Measuring GLP-1 requires venipuncture, precise timing (GLP-1 half-life ~2 min with DPP-4), and expensive ELISA kits. Urinary GLP-1 9-36 amide (the primary DPP-4 cleavage product) could provide a non-invasive, integrated measure of GLP-1 secretion over hours. If validated against serum GLP-1 AUC, this would enable repeated home-based measurement — before/after meals, PEM episodes, drug administration — without clinical visits. No validated human urinary GLP-1 assay exists. Technical challenges are significant (peptide degradation, concentration variability). If feasible (r ≥ 0.6 correlation with serum GLP-1 AUC), transformative for at-home GLP-1 monitoring in n-of-1 trials.
Status: Speculative. GLP-1 fragments excreted in urine in rodents. No validated human urinary GLP-1 assay. Technical feasibility uncertain. Falsifiability: urinary GLP-1 9-36/creatinine correlates with serum GLP-1 AUC at r < 0.3 despite optimised collection protocol → the approach is infeasible as a clinical monitoring tool.
3 Retinal and Corneal Imaging as CNS Windows
The retina is embryologically part of the central nervous system and the only CNS tissue accessible by non-invasive, high-resolution imaging. Recent Long COVID studies have demonstrated that retinal and corneal imaging can detect neuroinflammation, small fiber pathology, and protein aggregation — findings that may be relevant to ME/CFS where direct CNS tissue access is severely limited.
(Origin: literature synthesis.) (Certainty: 0.25.) Three independent lines of evidence support the anterior visual pathway as a potential biomarker window into CNS pathology:
Retinal Aβ pathology from viral exposure. Miller et al. (2025, Science Advances) demonstrated that SARS-CoV-2 induces Alzheimer’s disease-related amyloid-β (Aβ) pathology in ex vivo human retinal explants and iPSC-derived retinal organoids (Miller et al. 2025). The Aβ accumulation was reversible with an NRP1 inhibitor (not clinically available). While this was an ex vivo study (not in living patients), it establishes that SARS-CoV-2 exposure can directly trigger amyloid pathology in CNS tissue — and that the retina is a visible readout. Baraniuk found amyloid products in ME/CFS as early as 2010, and Pretorius has documented fibrin amyloid microclots, suggesting amyloid pathology may be convergent across compartments.
Retinal microcirculation correlates with fatigue. Schlick et al. (2022, n=40) used optical coherence tomography angiography (OCT-A) to demonstrate that retinal microcirculation impairment correlates with chronic fatigue severity in post-COVID syndrome (Schlick et al. 2022). Reduced retinal vessel density and increased foveal avascular zone area — markers of microvascular dysfunction — distinguished fatigued from non-fatigued patients, independent of systemic inflammatory markers.
Corneal nerve fiber loss as neuroinflammation marker. Cañadas et al. (2023, n=50) used corneal confocal microscopy to demonstrate reduced corneal small nerve fiber density and increased immune cell (Langerhans cell) infiltration in Long COVID patients two years post-infection Cañadas et al. (2023). Corneal nerve fiber loss is a validated marker of small fiber neuropathy, and the presence of immune cells suggests ongoing local neuroinflammation at a CNS-adjacent site. Notably, Azcue et al. (2025) have already demonstrated increased corneal small fiber tortuosity in ME/CFS using the same technique (Azcue et al. 2025), providing direct evidence of corneal-level pathology in ME/CFS itself — though retinal OCT-A has not yet been studied. This corneal-nerve arm is further corroborated by the symptom-linked post-COVID ocular syndrome described below (Post-COVID Ocular Syndrome Directly Confirms the Dysautonomia–Neuropathy–Immune Triad in an Accessible Tissue); the retinal OCT-A arm, however, remains untested in ME/CFS, so the overall speculation certainty is unchanged.
Relevance to ME/CFS: Small fiber neuropathy is documented in approximately 30% of ME/CFS patients. Retinal and corneal imaging could provide: (a) a non-invasive window into CNS neuroinflammation (retinal microglia are CNS microglia), (b) an objective correlate of the brain fog and cognitive dysfunction reported by patients, and (c) a longitudinal biomarker that can be repeated without radiation (unlike PET) or lumbar puncture. Corneal confocal microscopy has already been performed in ME/CFS and shows increased small fiber tortuosity (Azcue et al. 2025), but retinal OCT-A has not yet been studied.
Falsifiable prediction: A cross-sectional study of 100 ME/CFS patients versus matched controls would show: (a) reduced retinal vessel density on OCT-A correlating with cognitive dysfunction severity, (b) reduced corneal nerve fiber density on confocal microscopy correlating with SFN symptoms, and (c) retinal OCT markers of neuroinflammation (increased retinal nerve fiber layer thickness in acute-phase patients, thinning in chronic patients) that track with disease duration. Falsified if retinal and corneal parameters are indistinguishable from controls in patients with documented cognitive dysfunction and SFN.
Consequence: If validated (and this remains entirely unvalidated for ME/CFS — cert 0.25), a non-invasive eye scan could provide information currently only obtainable through invasive procedures — making repeated CNS biomarker measurements feasible in clinical trials. This is a research-tool proposal, not a clinical replacement for established diagnostics.
(Origin: literature synthesis.) Miller’s finding that SARS-CoV-2 induces retinal Aβ (Miller et al. 2025), combined with VanElzakker’s 4-fold elevated p-tau217 in Long COVID (VanElzakker et al. 2024), and large epidemiological evidence of increased dementia risk in post-COVID populations, collectively raise the question of whether post-infectious syndromes accelerate neurodegenerative processes. However, there is an alternative hypothesis: Aβ may be an antimicrobial peptide produced as a protective response to infection, and its accumulation in post-infectious syndromes may represent a successful defense mechanism rather than early Alzheimer’s pathology. Increased new-onset dementia in COVID-19 patients over 50 is driven primarily by respiratory and vascular problems, not by the Alzheimer’s-associated amyloid/tau profile. The retinal Aβ paradigm could potentially distinguish between these two models — protective antimicrobial response vs pathological neurodegeneration — by tracking whether Aβ accumulation correlates with neurodegeneration markers (retinal nerve fiber layer thinning) or with infection resolution markers. No such study has been done.
Consequence: If retinal Aβ accumulation in post-infectious syndromes is protective rather than pathological, anti-amyloid therapies (which are being explored for Alzheimer’s) would be contraindicated — they would remove a defense mechanism. Distinguishing between these models is essential before any anti-amyloid intervention is considered for ME/CFS or Long COVID.
(Evidence type: observational study — prospective cross-sectional.) (Certainty: 0.75.) Moustardas et al. (2026, Nature Communications) examined 100 non-hospitalised patients with persistent ocular symptoms 3 months to 3 years after mild COVID-19 against 32 mild-COVID controls without ocular symptoms (Moustardas et al. 2026). One in three affected patients was on full or part-time sick leave. The defining clinical observation is that standard ophthalmic examination was normal — abnormalities emerged only on specialised testing, exactly the “normal routine workup, pathology on specialised assay” pattern that characterises ME/CFS itself.
Specialised testing revealed a convergent triad: (a) peripheral neuropathy — corneal neurodegeneration and chronic dendritic/T-cell activation on corneal confocal microscopy, extending the small-fiber findings of Cañadas et al. Cañadas et al. (2023) and Azcue et al. (Azcue et al. 2025) from research biomarker to symptom-linked pathology; (b) dysautonomia — weakened autonomic pupillary reflexes (pupils admitting excess light, mechanistically explaining photophobia) and adult-onset strabismus attributed to oculomotor nerve involvement; and (c) immune dysregulation — a tear-fluid proteomic signature of CD4+ T-cell dysregulation that matched protein patterns previously reported in blood and tissue in severe and fatal COVID-19. Two diagnostic models achieved 77% (instrument-only) to 91% (instrument + tear proteomics) accuracy.
This finding is confirmatory of the causal cascade developed in Chapter Causal Hierarchy: Root Causes, Amplifiers, and Consequences: a post-viral insult propagating through autonomic, small-fiber, and T-cell-mediated immune channels to produce organ-specific dysfunction. The eye is simply a tissue where all three channels are simultaneously and non-invasively measurable. It is not evidence that post-COVID ocular syndrome is identical to ME/CFS — the Moustardas cohort was not screened against ME/CFS diagnostic criteria (see Limitation). The convergent argument assembling this finding with the SFN model and T-cell readout is synthesised in The Eye as a Convergent Readout of the Dysautonomia–Neuropathy–Immune Triad.
Consequence: Patients whose eye symptoms after a mild infection are dismissed because routine eye exams look normal now have an objective explanation and a testable diagnostic pathway; for ME/CFS research, the same corneal-confocal and tear-proteomic tools offer an accessible window onto the dysautonomia and small fiber neuropathy the disease is thought to share.
The Moustardas et al. cohort (Moustardas et al. 2026) comprised post-COVID patients selected for ocular symptoms, not patients meeting ME/CFS diagnostic criteria. The mechanistic overlap (dysautonomia, small fiber neuropathy, T-cell dysregulation) is compelling, but the study does not establish that these patients have ME/CFS, nor that ME/CFS patients as a group exhibit this ocular syndrome. Severity applicability to the ME/CFS spectrum is unknown. Direct evidence would require applying the same multimodal ocular protocol to a criteria-defined ME/CFS cohort. The finding is single-study and not yet independently replicated. A literature search found no studies that examined and failed to find post-COVID ocular abnormality, so publication bias toward positive findings cannot be excluded; dry-eye and convergence complaints also have mundane confounders (increased screen time during and after illness) that a symptom-selected cohort cannot separate from neuropathic causes.
Consequence: This result should be read as strong support for a shared mechanism, not as proof that ME/CFS patients have this eye syndrome — the honest current status is “highly suggestive, awaiting a direct ME/CFS study.”
Two additional open questions arise from this evidence but remain speculative. First, whether corneal Langerhans cell density tracks CNS microglial activation — shared myeloid origin does not establish functional equivalence, and no study has measured corneal LC density and TSPO-PET signal simultaneously in any disease. Second, whether the observed tortuosity-dominant CCM pattern in ME/CFS reflects systemic SFN, trigeminal-specific neuropathy, or dry eye secondary to autonomic dysfunction — concurrent IENFD and tear-film assessment could distinguish these, but the data do not exist. Both questions are discussed in detail in Chapter Cardiovascular Dysfunction.