The science

The big picture

Research is converging on a loop:

1. The autonomic nervous system (the automatic system that controls heart rate, blood pressure and blood flow) doesn't work properly.
2. Less blood reaches the brain and muscles, especially when upright or active.
3. Cells run short of oxygen and switch to an inefficient backup energy system. Their energy factories (mitochondria) get damaged.
4. Post-exertional malaise follows as the body struggles to clean up and repair.

Autoantibodies, immune dysfunction and brain inflammation may drive this loop in some patients. Different people may be affected at different points, which could explain why no single treatment works for everyone.

The brain and neuroinflammation

• Imaging meta-analysis (2024): 65 brain imaging studies covering 1,529 patients, pooled to look at neuroinflammation (inflammation in the brain). It found abnormalities in “pivotal network hubs bridging reason and emotion” that disrupt connections with the limbic system. Observational¹
• An earlier review of 55 studies found widespread disruption of the brain network that controls the autonomic nervous system, though findings weren’t consistent across studies. Observational²
• Important nuance: this isn’t the severe brain inflammation seen in encephalitis. The ME Association’s reading of the evidence is that it points to low-level inflammation, possibly from an immune response to the original infection that never fully switched off.³
• White matter study (March 2026, Human Brain Mapping): 67 patients and 67 healthy people were compared using a new diffusion-based imaging measure of neuroinflammation. A later correction by the authors revised how the signal should be interpreted, so treat this as early-stage work. It could become a biomarker (an objective test). Observational⁴
• The “sickness response” model (Komaroff and Dantzer, 2025): infection and inflammation activate brain circuits that produce fatigue, pain, brain fog and sleepiness. This is a normal protective programme (the same one that makes you want to lie down with the flu) that, in ME/CFS and Long COVID, fails to switch off. Hypothesis⁵
• Johns Hopkins (ongoing): testing whether a leaky blood-brain barrier lets immune cells into the brain and drives cognitive problems.⁶

The brain’s waste-clearance system

Every night during sleep, the brain flushes out metabolic waste through what’s called the glymphatic system.

• Griffith University (2026) was among the first to show that this system is impaired in ME/CFS. In 31 patients and 27 healthy people, a lower glymphatic index was linked to worse sleep and poorer concentration. Observational^7,8
• Why it matters: it suggests a direct link between poor sleep and brain fog. Broken sleep means less clean-up time, which means more fog. It’s one reason to check for sleep disorders like sleep apnoea.

Blood flow to the brain

• Reduced blood flow to the brain is common. A 2025 systematic review found it significantly reduced in 12 of the 26 ME/CFS studies it examined, and reduced further in people who also have orthostatic intolerance.⁹ One of those studies found blood flow in the brain’s limbic (emotion and memory) region was about 34% lower than in healthy people. Observational¹⁰
• The brainstem: Griffith University MRI studies have found abnormal signals in the brainstem’s blood-pressure control centre, the midbrain and the hypothalamus, impaired connections within the brainstem, and changes consistent with reduced brainstem myelin. Observational^11,12,13
• Brigham & Women’s Hospital (January 2026): in 143 people with Long COVID and 170 with ME/CFS, blood flow to the brain when upright was reduced in about 9 in 10, and small fibre nerve damage was found in about half to two-thirds. In many, blood flow dropped even without a rise in heart rate or a drop in blood pressure, linked to breathing out too much carbon dioxide (“hypocapnic cerebral hypoperfusion”). Observational¹⁴

• Speculative: a 2026 paper proposed that some severe cases may involve instability where the skull meets the neck, deforming the brainstem. This is a hypothesis only. Hypothesis¹⁵

Muscles and energy: the “two engines”

Stanford whole-body PET scan (presented May 2026, not yet published): in 15 patients and 15 healthy people, researchers expected to see brain inflammation. The striking signal was instead in the muscles (thighs, shoulders), and it matched how fatigued people were. It supports a model where poor blood flow causes a self-reinforcing cycle of damage in muscle energy factories. Treat it as preliminary until the paper appears. Observational^16,17

Here’s the plain-language version of what seems to be going on.

In ME/CFS, the switch to the backup engine trips far too early, for a shower or a conversation rather than a sprint. Several two-day exercise tests have found that the point at which patients tip into anaerobic energy production drops sharply on the second day, even compared with equally unfit controls,^18,19,20,21 though a 2026 study of 58 patients found no day-two drop, so this is still debated,²² and muscle biopsies in Long COVID and ME/CFS show more anaerobic fibre types and damage after exertion that bed rest alone doesn’t explain.^23,24 When the switch trips:

• energy runs short, which causes weak legs and slow thinking
• lactate and acid build up, bringing heavy limbs and malaise
• the mitochondria take damage and get even worse at their job

PEM is that cascade playing out over 24–48 hours as the body tries to clean up and repair. It’s also why pacing works: staying below the switch point avoids the damage.

The immune system and TRPM3 (Griffith University)

Griffith University’s National Centre for Neuroimmunology and Emerging Diseases (NCNED) in Queensland has argued for a decade that ME/CFS is an ion channel disorder.

• A calcium channel called TRPM3 on natural killer (NK) immune cells doesn’t open properly, so calcium can’t get in and the cell can’t do its job.
• TRPM3 is also found in the brain, brainstem, muscles and heart. That could explain how one faulty channel causes symptoms all over the body.
• The drug naltrexone removes a brake on TRPM3. In the lab, it restored the channel’s function. Lab study^25,26

Strengths: five groups of patients, two techniques, two illnesses, all with consistent results. That is unusual in ME/CFS research. Limitations: very small samples (9–10 per group), lab doses far higher than what a tablet achieves, and no controlled clinical results yet. The researchers also hold a patent on a TRPM3-based test.

Griffith is now starting a placebo-controlled trial of LDN in ME/CFS and Long COVID (TreatMELC). See trials and the LDN section.

Genetics: DecodeME

The world’s largest ME/CFS genetics study compared more than 15,000 people with ME/CFS against about 260,000 controls. It found eight genetic signals, mainly in the immune and nervous systems. ME/CFS is polygenic (many genes, each with small effects). The results are still a preprint awaiting peer review. Observational³¹ More on DecodeME.

Blood tests in development

There’s no approved diagnostic test yet, but several are being developed:

• Edinburgh (2025, EMBO Molecular Medicine): using UK Biobank data from 1,455 people with ME and 131,000 controls, researchers found 116 blood measures that differed in both men and women, and showed they weren’t explained by inactivity. Observational³²
• Cornell (2025): a model based on RNA fragments in the blood was 77% accurate on new samples. The researchers say that isn’t high enough for a diagnostic test yet. Observational^33,34
• EpiSwitch (a commercial test, 2025) reported 92% sensitivity and 98% specificity in 47 severely ill patients versus 61 healthy people. Independent experts note the small numbers and that it wasn’t tested against other illnesses, so treat it sceptically until replicated. Observational^35,36

The biggest lesson of 2026

Trials that tested a drug on all patients failed, but in several of them some patients clearly responded, and those responses were buried in the overall “no effect” result. This happened with LDN, immunoadsorption and BC007.³⁷ Researchers are now selecting patients by biomarker so each treatment is tested on the people it’s most likely to help. The clearest example so far: BC007 failed in an unselected trial but improved fatigue and function in a small trial that only enrolled people with specific autoantibodies.³⁸ See 2026 results.