Solving ME/CFS · disease model · Evidence atlas
A hypothesis about the disease mechanism (pathophysiology) of ME/CFS: a self-sustaining loop across immune, hormonal, vascular and nervous systems. It covers what a crash is under this model, and what the model implies for treatment, diagnostics and research. The evidence behind each link is in the evidence atlas.
ME behaves like a self-sustaining loop that a virus started and that no longer needs the virus. Common triggers are EBV, COVID and other viral infections. In the version below, each link is something measured in patients. What isn't proven is that the links join up this way.
| Link | What happens |
|---|---|
| 1 · Immune priming | A big infection leaves the immune system primed. In interferon-treated hepatitis C patients, a third still had fatigue 6 months after treatment ended, predicted by a bigger early immune surge. Severe acute illness predicts post-viral ME. |
| 2 · Brain and hormone control | DecodeME genetics points at brain and pituitary. One uncontrolled study found ADH undetectable after fluid deprivation in most patients, though copeptin (a steadier ADH marker) was normal in the 13 tested. The stress-hormone axis is mildly blunted. Small nerve fibres that tighten veins are lost in many patients. |
| 3 · Low volume, small heart | Less plasma and red cell volume, smaller heart chambers, too little blood returning to the heart (preload failure). Brain blood flow falls on standing. |
| 4 · Vessels and capillaries | Artery lining doesn't dilate properly. Muscle capillaries are fewer, and bed rest doesn't explain it (Charlton 2026); thickened capillary walls have so far only been shown in a conference poster. Pericytes (the cells that wrap capillaries and control flow) may be damaged. VEGF results are mixed. |
| 5 · Tissue under-supply | Effort or standing asks for more blood flow than this system can deliver. Oxygen extraction in muscle is poor. Muscle sodium is raised. |
| 6 · Damage and immune reaction | A day after exertion: mitochondrial function drops, fibres are damaged, immune cells move in, amyloid deposits appear. Over the next 24 h, blood proteins show immune signalling changes and stressed metabolism. Complement and platelets are activated in long COVID. |
| 7 · Sickness response | Immune signals reach the brain (vagus nerve, blood–brain barrier) and switch on the sickness programme: malaise, exhaustion, pain, the feeling of being poisoned. Interferon studies show it runs partly through dopamine circuits in the basal ganglia. |
| 8 · Arousal and sleep | The sympathetic system runs hot to hold blood pressure up. That wrecks sleep, and poor sleep worsens autonomic control and immune balance, which feeds back to link 2. |
In this model a crash is links 5→6→7 running in sequence. Effort, standing or even sensory load asks for blood flow the vessels can't deliver. Muscle and brain are briefly under-supplied and take damage. The damage triggers an immune reaction that builds over a day or two. That explains the 12–72 hour delay, which ordinary running-out-of-fuel can't. The immune signals then drive the brain's sickness response, the same machinery that makes severe flu feel like dying. Each crash adds arousal and poor sleep (link 8), which lowers the threshold for the next one. That fits the shrinking crash threshold many patients describe.
Two things make this more than a story. Muscle biopsies a day after exercise show the damage directly, in long COVID (Appelman 2024). And one large proteomics study reported the biggest blood-protein changes in the recovery window rather than during the effort (Germain 2025), though our person-by-person re-analysis of that data found no single protein that clearly changed. Two things are still unknown: why the sickness response stays switched on, and how much of the vessel damage is cause versus consequence.
A loop can be broken at any link, and pushing on several weak links at once may matter more than finding the "root". The candidates below are tagged with the links they act on.
Ordered by suggested start. Try fits the model and the risk is low. With doctor needs a prescriber or test. Not now means the risk outweighs the likely gain for a severe, weight-losing patient.
| Candidate | Links | Status |
|---|---|---|
| Diagnostic bloods: copeptin and osmolality, standing noradrenaline, paired crash vs baseline bloods | 1, 2, 6 | With doctor |
| Home soft-shell oxygen chamber | 4, 5, 6 | Try |
| Pyridostigmine | 3, 5 | With doctor |
| Red / near-infrared light | 5, 6 | Try |
| Ear vagus nerve stimulation (tVNS) | 7, 8 | Try |
| CoQ10 + NADH | 5 | Try |
| Low-dose aripiprazole | 7 | With doctor |
| BPC-157 (oral) | 4, 6 | Possible, with care |
| Low-dose rapamycin | 1, 6 | After RCT |
| GLP-1 drugs / retatrutide | 1, 4, 7 | Not now |
Why: these show which links are worst in a given person. Copeptin with paired blood and urine osmolality tests the ADH story directly. An EBV/CMV PCR is worth adding to the crash set. Standing noradrenaline shows whether POTS is the high-adrenaline type. A home phlebotomist can draw blood in bed. Taking one set on a stable day and another 24–48 h into a crash would show what moves during a crash. Freeze spare plasma both times.
Crash-vs-baseline panel:
Mechanism: breathing oxygen under pressure dissolves extra oxygen in plasma, which can reach tissue that red cells under-supply. That works on links 4–5 directly. At full clinical doses (2 ATA, 100% O₂) there's also evidence of new vessel growth and a rise in circulating vessel-repair cells. That matters if capillary loss is part of the picture.
Dose: Rental home chambers typically run up to 1.5 ATA with a 93–98% concentrator. What the person breathes depends on the mask: from ~50% oxygen through a loose mask to ~90% through a sealed mask with a reservoir bag. That's roughly 35–65% of the dissolved oxygen in the clinical trials. The best ME/CFS data at full dose (Charité, 40 sessions, n=30) showed modest average gains, with a third improving meaningfully and no one crashing from it. Nobody has tested this lower dose. It may simply work less, or it may sit under a threshold.
Get the most from it:
Risks: ear and sinus pain on pressurising, confinement, compressor noise, the effort of getting in and out. Fire safety with oxygen: no electronics, heat packs or synthetic fabrics inside.
Mechanism: boosts the nerve signal that tightens veins and helps blood return to the heart (link 3, preload failure). Evidence: one positive single-dose RCT in ME/CFS (Systrom); a longer long COVID trial was flat overall. In practice: the best-matched prescription option when preload failure or POTS is prominent. Side effects are gut cramps and urgency.
Mechanism: light at 600–850 nm is absorbed by mitochondria (cytochrome c oxidase). This releases nitric oxide, which dilates local vessels, and raises ATP in cells (link 5). Near-infrared reaches a few cm into tissue. Evidence: no ME/CFS trials; small positive long COVID and fibromyalgia studies. In practice: done lying down, very low risk. Goggles for near-infrared, and watch for light sensitivity.
Mechanism: stimulating the vagus nerve through the ear calms sympathetic arousal and dampens inflammatory signalling through the "cholinergic anti-inflammatory pathway" (links 7–8). Evidence: small POTS and long COVID studies, mixed. In practice: £200–500 device, in bed, very low risk. Can be timed for before sleep.
Mechanism: two cofactors of the mitochondrial energy chain (link 5). Evidence: two small RCTs in ME/CFS showed less fatigue (sponsor-linked); CoQ10 alone was negative. In practice: over the counter, cheap, low risk.
Mechanism: at 0.25–2 mg it nudges dopamine signalling and may calm microglia. That targets the basal-ganglia dopamine dip in the sickness response (link 7). Evidence: 74% of 101 patients reported improvement, retrospective with no control. In practice: off-label prescription. It can worsen insomnia or cause restlessness, so start after sleep is steadier.
Mechanism, mostly from rat and cell studies:
Caveat: almost all of this comes from one research group in animals, and human data is a handful of small reports. So the mechanism map is plausible, but the human effect is unknown. The product is unregulated: use the oral form and get the batch independently tested.
Mechanism: mTOR inhibition boosts cellular clean-up (autophagy) and reins in over-active immune cells (links 1, 6). Evidence: 72% of ~40 reported improvement, open-label; Mount Sinai RCT due late 2026. In practice: specialist prescription; worth revisiting when that RCT reports.
Mechanism: GLP-1 receptors sit on immune cells, vessel lining and brain. Scheibenbogen's rationale is anti-inflammatory effects, better endothelial function and capillary perfusion, and calmer neuroinflammation. That fits links 1, 4 and 7, so it can't be ruled out as helpful. Why not now:
Charité's study is in overweight patients only. Would change if: weight is stable and a controlled ME result appears. If tried then, it should be low-dose licensed semaglutide through a doctor, not grey-market retatrutide.
If the model is right, these are the experiments that would test it fastest.