The key finding: adult-onset ADHD is not simply childhood ADHD persisting

The most consequential finding in the field is that much adult ADHD does not appear to be a straightforward continuation of childhood ADHD — though the current picture is more nuanced than the earliest landmark finding suggested.

A landmark longitudinal study that followed a large birth cohort into adulthood found that the majority of adults who met criteria for ADHD in their twenties had not had ADHD as children — and, strikingly, that most children with ADHD had not continued to meet full criteria as adults (Moffitt et al. 2015). Only a small fraction of childhood ADHD persisted as adult ADHD.

That finding sparked a debate that has since refined the picture. A closer look at the same phenomenon — including repeated comprehensive assessments of a comparison group from ages 10 to 25 — found that most apparent adult-onset ADHD is adolescent-onset (ages 12–16), not de novo adult-onset, and that in many cases the symptoms reflect a missed childhood presentation, a comorbid disorder, or the cognitive effects of substance use rather than a genuinely distinct adult-onset syndrome Asherson and Agnew-Blais (2019). Retrospective recall of childhood ADHD symptoms is itself unreliable — accuracy only about 55% (Breda et al. 2020) — which means some reported “adult onset” is really forgotten childhood onset.

So the current, more careful conclusion is a mix: some adult ADHD is genuinely adolescent-onset; some reflects missed or misrecalled childhood symptoms; some is explained by substance use or other comorbidity. A clean “distinct adult-onset syndrome, appearing out of nowhere” is now the minority reading, not the consensus.

An honest caveat: this “mix” conclusion rests on a small US comparison cohort (the MTA local normative comparison group, n = 239) rather than a population-grade birth cohort, and two of the key studies share a research lineage, so the evidence is a partly self-referential cluster rather than independent replication. The question is therefore best treated as open, not settled — a point the rest of this article’s hedging reflects.

The clinical implication is still direct: an adult who develops ADHD-like symptoms in adulthood is not necessarily reliving an unbroken childhood condition. They may be experiencing an acquired or adolescent-onset process — one that deserves investigation of its cause, not a shrug toward “lifelong wiring.” But the degree to which that process is truly independent of childhood neurodevelopmental risk, rather than a missed childhood presentation, remains unresolved.

Three architectures of the overlap

Our research frames the ADHD-chronic-fatigue relationship through three possible architectures:

Architecture A — shared vulnerability. A common upstream factor — genetic, immune, or developmental — predisposes a person to both ADHD and chronic fatigue independently. The co-occurrence is a selection artifact: the same vulnerability produces both. The genetic architecture of the neurodevelopmental conditions supports this: ADHD and autism share a large fraction of their genetic influences Cross-Disorder Group of the Psychiatric Genomics Consortium (2013).

Architecture B — secondary cascade. The disease process — neuroinflammation, dopaminergic depletion, autonomic dysfunction — produces ADHD-like neuropsychiatric phenotypes in people who would not otherwise have developed them. The ADHD arises because of the illness, not alongside it.

Architecture C — bidirectional amplification. ADHD and the illness each worsen the other — sleep disruption, energy depletion, and stress compound in a loop.

The architectures are not mutually exclusive, and the temporal data matter: when ADHD is documented before fatigue onset — as in the studies where childhood ADHD predicts later chronic fatigue Quadt et al. (2024) — Architecture A or C is supported. Architecture B (acquired ADHD from the illness) applies to a subset of patients who develop new-onset inattention after the illness begins — though the population-level evidence cautions that much apparent “new onset” may be unmasking of pre-existing subclinical neurodivergence rather than true acquisition (see the honest limits below).

The mechanism: disrupted interoception and neuroinflammation

Two proposed routes to acquired ADHD-like features:

Interoceptive disruption. The brain maintains a balance between its predictions about the body and the sensory evidence the body sends back. Acquired features are thought to involve corrupted lower-level signals — brainstem and thalamocortical pathways failing to deliver accurate bodily information. The brain compensates by tightening its priors, producing an outward picture of inattention and difficulty regulating arousal. The computational locus differs from developmental ADHD, and so should the response to treatment.

Neuroinflammation and microglial activation. Sustained neuroinflammation degrading prefrontal, mesolimbic, and thalamocortical circuits can be sufficient to produce ADHD-like features in people whose pre-illness neurodevelopment was intact — the registered speculation examined in Part 2. If this cascade is the proximate cause, then interventions targeting neuroinflammation should partially reverse these secondary features — a testable and not-yet-tested prediction.

The test: context-dependence

Because the two routes differ in stability, they can be told apart empirically:

If acquired ADHD-like inattention shows the same trait stability and context-independence as developmental ADHD — persisting unchanged and not correlating with markers of neuroinflammation — the acquired model is not supported.

In other words: developmental inattention is always there. Acquired inattention should fluctuate with the underlying process (inflammation, metabolic state) and should improve when that process is treated.

The honest limits

• The acquired model is explicitly speculative — a registered hypothesis with low confidence — and a simpler explanation (inflammation directly causing both the symptoms and the underlying process) may account for the findings without any interoceptive framework.
• None of the theoretical components has been directly validated in this context.
• The distinction between acquired and developmental is not yet usable at the bedside: no test currently separates the two in an individual patient.
• Even if some features are acquired and reversible, this does not mean “ADHD can be cured.” It means a subset of the symptom burden in some people may respond to treating its cause.
• The finding that adult ADHD is often adolescent-onset or reflects missed childhood symptoms, rather than a clearly distinct adult-onset syndrome Asherson and Agnew-Blais (2019), establishes a difference between onset groups but does not by itself prove that group is reversible — only that it did not begin in childhood.
• The largest population-grade test found no infection-driven effect. A 20-year national cohort found no independent effect of COVID-19 on ADHD diagnosis or treatment rates (Shkalim Zemer et al. 2024), and the post-COVID rise in ADHD medication is consistent with catch-up diagnosis rather than new onset (Gimbach et al. 2024). This is the strongest counterweight to the “acquired from the illness” architecture (Architecture B): at population scale, infection does not appear to create new ADHD. The honest reading is that most post-infectious attention difficulty reflects unmasking of pre-existing (often subclinical) neurodivergence or symptom overlap, rather than de novo acquisition — though a narrow acquired subset cannot be excluded by the population data.

The clinical consequence — and the harm of getting it wrong

The reason this distinction matters clinically:

• If acquired features are dismissed as “just ADHD”, a treatable underlying process (inflammation, a metabolic deficit, disrupted sleep) goes unaddressed.
• If developmental features are treated as acquired and reversible, patients are offered treatments that will not work, and may blame themselves when they don’t.
• The reverse mislabeling also occurs. A “late-game fader” pattern — post-exertional cognitive fog that mimics the ADHD-inattentive type — can be misread as ADHD and trigger evaluation for an attention disorder, when the underlying issue is energy-driven (ATP-limited) rather than motivation-driven, and worsens with exertion rather than persisting constantly. Both directions of the error send the patient down the wrong treatment path.

The honest position is that acquired ADHD-like features should be supported, not pathologized — and the underlying modifiable drivers (sleep, inflammation, iron, energy) should be addressed regardless of the label, because they affect overall wellbeing even when the core developmental wiring is unchanged.

What to take away

Some ADHD-like features — inattention, impulsivity, executive difficulty — may be acquired after development rather than built in from it. The finding that adult ADHD is often not a straightforward continuation of childhood ADHD Asherson and Agnew-Blais (2019) supports this possibility. If such features are acquired, they should be context-dependent and potentially reversible with treatment of the underlying cause: neuroinflammation, metabolic disturbance, or sleep disruption.

The encouraging part: this means a meaningful fraction of the daily symptom burden in some people may be addressable without claiming to “fix” ADHD itself.

The honest part: this is a low-confidence hypothesis, and the decisive test — showing that acquired inattention fluctuates with underlying inflammation while developmental inattention does not — has not been run.

This concludes the four-part series on the biology of ADHD. Part 1 examined the prefrontal-energy model and the multi-pathway treatment hypothesis, Part 2 the immune and neuroinflammatory strand, Part 3 the dopamine-Nrf2-NLRP3 axis, and Part 4 the distinction between acquired and developmental features. See the series landing page.

This article reflects research hypotheses with explicit, low confidence — not established clinical fact. Discuss any medical decision with a qualified clinician.

The mechanism: a self-amplifying loop

The axis runs through three molecules:

1. Dopamine and Nrf2. Dopamine D1-receptor signaling enhances Nrf2 activity — via protein-kinase-A-mediated phosphorylation and nuclear import. Nrf2 is the master transcription factor of the cell’s antioxidant defenses, switching on over two hundred protective genes. When dopamine tone is low — as it is in ADHD — Nrf2-mediated antioxidant defenses are compromised, and the cell becomes more vulnerable to oxidative stress. A related route runs through dopamine-quinones: chronic low-grade dopamine oxidation produces reactive dopamine-quinone metabolites that deplete glutathione, further impairing Nrf2-driven antioxidant defense and disinhibiting the NLRP3 inflammasome in microglia and immune cells Sies (2017).

2. Nrf2 and NLRP3. Nrf2 normally suppresses the NLRP3 inflammasome, the immune system’s central inflammatory switch, via induction of HO-1 and NQO1. Reduced Nrf2 activity removes that brake. The NLRP3 inflammasome then drives elevated IL-1-beta and IL-18.

3. NLRP3 and dopamine — closing the loop. The elevated inflammation further impairs dopamine synthesis in two ways: it activates the IDO/kynurenine pathway, reducing the availability of BH4 for the enzyme that makes dopamine; and it induces oxidative stress that damages dopaminergic terminals. A further, more direct route runs through the kynurenine metabolite kynurenic acid, which at low concentrations reduces striatal dopamine release — connecting kynurenine overactivation straight to the dopamine deficit rather than only through the BH4 bottleneck Cysique et al. (2023). Lower dopamine → lower Nrf2 → more NLRP3 → more inflammation → even lower dopamine.

The result is a self-amplifying loop: a pre-existing low-dopamine state (as in ADHD) and a bout of inflammation each make the other worse, and the loop tends to persist.

The cross-disease bridge

The hypothesis that matters for ADHD is the lowered threshold for post-infectious fatigue. The mechanistic prediction is specific:

Individuals with ADHD have lower baseline Nrf2 activity and higher basal NLRP3 priming. This lowers the threshold for developing chronic fatigue after an infection. The elevated ADHD-to-chronic-fatigue comorbidity Sáez-Francàs et al. (2012) is a consequence of this pre-existing dopamine-Nrf2-NLRP3 dysregulation.

In plain terms: a person with ADHD starts with the loop already tilted toward inflammation. When an infection adds an inflammatory hit, the loop tips further, and the person is more likely to cross the fatigue threshold than someone without that pre-existing tilt.

This is the same BH4 bottleneck our project has explored across conditions — the cofactor that inflammation diverts away from dopamine synthesis and toward a different branch of its metabolism. The loop is one way of describing why that single bottleneck matters so widely.

Where a mechanism already has a dedicated article

Two dedicated articles carry the established groundwork this part builds on:

• The Inverted-U Is Not One Thing — the Nrf2-mediated hormesis mechanism, the Keap1-Nrf2-ARE pathway, and why Nrf2 drugs (LDN, sulforaphane, quercetin) work only in a narrow dose window. This part does not repeat that content; it extends it to the cross-disease ADHD axis.
• One Cofactor, Six Conditions, One Bottleneck — the BH4 cofactor that inflammation diverts away from dopamine synthesis.

The Nrf2/NLRP3 axis described here is our reading of how those established mechanisms behave in an ADHD-affected, infection-stressed brain. The underlying pharmacology is established; the ADHD-specific cross-disease claim is a registered speculation.

A genetic strand: ADHD’s own architecture, and shared mitochondrial modifiers

Before the cross-disease reading, ADHD has its own well-established genetic architecture, independent of any connection to chronic fatigue. It is among the most heritable of psychiatric conditions.

ADHD’s own genetics. The largest genome-wide study of ADHD — nearly forty thousand cases — identified 27 genome-wide significant risk loci, up from 12 in earlier studies, implicating genes expressed in the brain and in cognitive domains (Demontis et al. 2023). Family and twin studies place its twin-based heritability high — around seventy to eighty percent (Faraone and Larsson 2019). Twin studies confirm that genetic influences on ADHD and autism are substantially shared and are among the strongest genetic correlations in psychiatry (Polderman et al. 2014). Genome-wide studies find a genetic correlation around 0.35–0.41 between ADHD and autism Grove et al. (2019), and when five major psychiatric disorders are analyzed together, ADHD and autism cluster as a single neurodevelopmental group, genetically distinct from the psychotic and mood group (Cross-Disorder Group of the Psychiatric Genomics Consortium 2013). Twin studies confirm this overlap is present from childhood and persists into adulthood Polderman et al. (2014). This is why ADHD and autism co-occur far beyond chance — a meta-analysis found a pooled ADHD prevalence of 28% in autism (Lai et al. 2019), and a register study of nearly two million births found an autism diagnosis elevated the odds of ADHD roughly 22-fold (Ghirardi et al. 2018).

This standalone genetic picture matters for the series because it grounds the “what we know” side: the metabolic hypotheses in Parts 1–3 are added onto a well-replicated genetic and neurobiological foundation, not invented in its place.

One genuinely open question from our paper deserves a flag here: the genetic correlation between ME/CFS itself and ADHD has not been computed. If it turned out to be substantial, it would support the dopamine signal in the chronic-fatigue genome and raise whether childhood ADHD is a direct genetic risk factor for post-infectious fatigue — rather than only a metabolic one. This is an open question, not a finding.

The testable predictions

Because this hypothesis is mechanistic, it makes specific, falsifiable predictions:

1. Chronic-fatigue patients with ADHD comorbidity show lower Nrf2 nuclear translocation in their peripheral blood mononuclear cells than chronic-fatigue patients without ADHD.
2. NLRP3 inflammasome markers (IL-1-beta, caspase-1 activity) are elevated in the comorbid group relative to chronic fatigue alone.
3. Nrf2-activating interventions (sulforaphane, dimethyl fumarate) improve both ADHD symptoms and fatigue in the comorbid group.
4. Nrf2 promoter polymorphisms predict post-infectious fatigue severity in ADHD cohorts.
5. Stimulant exposure provokes inflammation in the comorbid group: our paper’s dopamine-quinone reading predicts that stimulants (which raise dopamine and accelerate its oxidation to quinones) would increase IL-1-beta via NLRP3 in blood-cell cultures from ADHD-comorbid patients, and that this effect would be blocked by Nrf2 agonists or NLRP3 inhibitors. This also sharpens the warning that stimulants are not a free pass in this group — a prediction about a mechanism that could contribute to post-exertional worsening.

Each prediction is testable with existing assays and existing cohorts. None has been run.

The honest limits

This is a speculative hypothesis, with explicit low confidence — a registered speculation, not an established finding.

• The D1-receptor–Nrf2 link is established in cellular models but not in human patients.
• Nrf2 pathway activity has been measured in ADHD patients only once, in a 2026 study of 60 adults with ADHD versus 60 controls, which found serum Nrf2 and HO-1 protein significantly reduced in ADHD and negatively correlated with symptom severity (Gürbüzer, Ozkaya, and Mercantepe 2026). This is consistent with — but does not establish — the axis proposed here: it measured circulating serum protein, not Nrf2 nuclear translocation or NQO1 target-gene expression, and it is adult ADHD only, a single study.
• No study has measured Nrf2 nuclear translocation or NQO1 expression in ADHD cells.
• No prospective ADHD-to-fatigue longitudinal data exist that test the inflammation-mediation chain directly.
• Comorbidity estimates are confounded by diagnostic overlap in symptom reporting.
• Nrf2/NLRP3 crosstalk is well characterized in redox biology, but the cross-disease claim — that ADHD’s pre-existing state specifically lowers the fatigue threshold — is an inference, not a measurement.

The loop is mechanistically plausible and each link has support in isolation, but the chain as a whole is untested.

What to take away

The dopamine-Nrf2-NLRP3 axis offers a self-amplifying loop explanation for a striking pattern: people with ADHD get chronic fatigue at roughly twice the rate, and the elevated risk is mediated by inflammation. It connects three things — dopamine, antioxidant defense, and the inflammatory switch — that are usually studied separately.

Unlike Part 1, this strand does not yet point to a tested intervention. Nrf2-activating agents such as sulforaphane exist and are mechanistically plausible, but no trial has tested them against the ADHD-fatigue axis specifically. Its value is explanatory and predictive: it names a specific, measurable loop and a set of testable predictions.

The encouraging part: the predictions are cheap and testable with existing assays. If they hold, they give the ADHD-to-fatigue association a concrete biological mechanism. If they fail, they narrow the search.

This is Part 3 of a four-part series on the biology of ADHD. Part 1 covers the prefrontal-energy model and the multi-pathway treatment hypothesis. Part 2 covers the immune and neuroinflammatory strand. Part 4 asks when ADHD-like features are acquired and reversible. See the series landing page.

This article reflects a registered research hypothesis with low confidence, not established clinical fact.

The short answer

The whole argument fits in one paragraph and one table. The detailed proof — separating what is established science from what is a proposed model — follows afterward.

The low-dose benefit of LDN depends on partial blockade of its receptor: the cell must still sense some residual signal to mount its compensatory anti-inflammatory response. Whether a receptor is partially or fully blocked is set by the drug concentration relative to that receptor’s sensitivity — the ratio , dose divided by the receptor’s dissociation constant. Raise the dose and rises; the fraction of blocked receptors climbs toward 100%. At 1 mg, the microglial receptor TLR4 sits in the partial-blockade zone and the benefit mechanism runs. At 4 mg, the same receptor is blocked almost completely, the residual signal is gone, and the 1 mg mechanism is switched off — not amplified. Meanwhile, 4 mg engages a different target (TRPM3) that 1 mg never reaches. A null at 4 mg therefore tested the 4 mg mechanism. It did not test the 1 mg mechanism — that mechanism was not running during the 4 mg trial.

The numbers make this concrete. The occupancy column uses the standard Hill equation (established pharmacology); the benefit column uses the simple model proposed in Part B of this article (a hypothesis). The parameter choices are illustrative and unmeasured: mg, the Hill coefficient , and the steepness exponent (the general model in Part B leaves and free):

Dose		(blocked)	(residual)	benefit, rel. to peak
0.7 mg	0.7	0.33	0.67	1.00 — the peak
1 mg	1.0	0.50	0.50	0.84
3 mg	3.0	0.90	0.10	0.06
4 mg	4.0	0.94	0.06	0.02

Read the table from the bottom up. A null at 4 mg speaks only to the bottom row. It says nothing about the first two rows, because those rows describe a state — partial blockade — that a 4 mg dose never produces. And 3 mg is yet another experiment: the low-dose window has closed there too, but different mechanisms engage there — the endorphin plateau and the opening TRPM3 window. Each dose is a different experiment, and the failure of one convicts none of the others.

One clarification: switching a benefit off is not the same as causing harm. At 4 mg the low-dose mechanism simply does not run — the patient is at their untreated baseline for that pathway, not below it. The 4 mg experience is therefore a null, not a worsening — and a null is information only about the state that was actually tested.

The inference rests on an unstated assumption

The argument “if 4 mg doesn’t work, then 1 mg won’t work” is only valid if one condition holds:

The superset hypothesis: the effects of a lower dose are a subset of the effects of a higher dose.

If that is so, then testing the highest dose tests every dose. A failure at the high dose is a failure at all of them. The inference is clean, simple, and saves multiple trials.

The problem: this hypothesis is only true for a drug that hits a single target with a monotonic dose-response curve. LDN is neither.

The chemistry: why a higher dose is not a bigger dose

The superset assumption treats a dose as a volume knob — more drug, more of the same effect. But LDN is a hormetic, multi-target drug, and that changes everything.

What is established science, and what is a proposed model

This is the point where I must separate two very different things: the established science that the superset-failure argument rests on, and a mathematical model I propose to make that argument precise. The former is published, replicated, and citable. The latter is a hypothesis — my own construction — and I will label it as such.

Part A — The established science

Part B — A proposed model (hypothesis, not established science)

I now propose a specific mathematical form. This is my construction — a working hypothesis to make the superset-failure argument precise — and it is explicitly NOT established science. It is not in the literature, it has not been validated, and its parameters are not measured. I present it for what it is: a scaffold.

Throughout Part B, “arm” means one of two opposing influences on the curve that push in different directions as the dose rises. One influence pushes the benefit up; the other, which only becomes dominant later, pulls it back down. It is the fight between these two that produces the inverted-U. The two arms are named below. Likewise, is the maximum benefit the curve would reach if the downward pull never set in — a ceiling, not the value actually reached.

Why failure at one dose does not propagate to others

The inference conflates two very different claims:

1. “This drug is not active in this patient” — a verdict about the drug.
2. “This target is not the rate-limiting pathway in this patient” — a verdict about a target.

A failure at 4 mg yields at most the second: the target combination engaged in the 4 mg band — TRPM3 restoration, orexin disinhibition, and TLR4 in its over-blocked state — is not the pathway limiting symptoms in this patient. And even that verdict is provisional, for the reason developed below: if the 4 mg band’s targets were never actually engaged at a clinical dose, the result says nothing about any mechanism. It certainly says nothing about the TLR4/Nrf2 hormetic state or the endorphin rebound, which engage only at 0.5–3 mg — because those states never occurred during a 4 mg trial.

A simple analogy: a drug blocks two receptors, A and B, with 100-fold higher affinity for A. At a low dose it engages only A — and only partially. At a high dose it engages A completely, and B as well. Now suppose the benefit requires partial engagement of A (the hormetic case). A failed high-dose trial then proves nothing about the low dose: the beneficial state — partial engagement of A — never occurred during the high-dose trial; A was fully blocked, and B’s engagement is beside the point. This is subtype 4a in Part 2 (concentration-dependent target selection), combined with the inverted-U: target selection is why 4 mg reaches targets that 1 mg cannot, and the inverted-U is why 4 mg loses the benefit that 1 mg had.

There is an even sharper version of this for LDN specifically. The TLR4 antagonism at clinical LDN doses is itself uncertain: (+)-naltrexone is a weak TLR4 antagonist, and an optimized derivative required a ~6,200× potency gain to reach nanomolar TLR4 antagonism (Gao et al. 2025). So a 4 mg null may mean something deeper than “this target isn’t rate-limiting” — it may mean the target was never actually engaged at any dose tested. In that case the 4 mg result says nothing about any mechanism, and the 1 mg question is entirely open.

The case where the inference holds

The superset inference is valid in one precise case. If a drug hits a single target, with a single affinity and a monotonic dose-response curve (more dose = more occupancy = more effect, no inversion), then a high-dose failure suffices, and lower doses are redundant. This is a general pharmacology principle — it is the assumption behind standard dose-finding for many single-target, monotonic agents — not a claim specific to LDN or drawn from the paper’s LDN analysis.

LDN is not in this class. Its dose-response curve is not monotonic — benefit rises and then falls within the clinical range, with target optima that do not overlap. For this kind of molecule, the superset hypothesis is not a reasonable shortcut: it is the very hypothesis the dose trial is meant to test, not presuppose.

What this means for a dose trial

The practical corollary is simple and, for this blog, familiar: dose finding is diagnostic (Part 3). Where a benefit appears and disappears on the dose axis reveals which mechanism is limiting. A patient who benefits only at 3–4.5 mg likely has a TRPM3 channelopathy or orexin-sensitive deficit (the paper notes these are not separable by dose alone). A patient who benefits only at 0.5–1.5 mg has TLR4-driven neuroinflammation. A failure at a single dose places the patient in none of these patterns — it only records the failure of that one dose.

For a “no benefit at any dose” to be a solid verdict, you have to have tested the relevant windows — the paper’s own guidance (ch28) is that “non-response cannot be concluded unless the low-dose window has also been tested.” A single dose point cannot produce that verdict. Part 3 makes the same point: “Non-response at 0.5 mg may mean the mechanism requires a higher dose (Pattern 2).” The reverse holds too: no benefit at 4 mg can mean the mechanism requires a lower dose.

The asymmetry no one disputes

There is a sign that the superset hypothesis is fragile, and it is internal to the reasoning itself. No one defends the reverse direction: if 1 mg works, no one concludes that 4 mg will work. On the contrary, it is well known that a higher dose can cancel a benefit present at a low dose — precisely the paradox documented in Part 1.

If you accept that “low dose works” does not predict “high dose works” (because the high dose switches off low-dose mechanisms), then the symmetry is broken: doses are not interchangeable in one direction, and there is no reason to treat them as interchangeable in the other. The superset hypothesis is only coherent if both directions are interchangeable — and they are not.

Certainty estimate

This post draws on the LDN dose-response framework developed in (Loth 2026), which synthesizes the Nrf2 hormesis corpus (Calabrese and Kozumbo 2021), the microglial M1→M2 dose-dependence literature (Kučić et al. 2021), the TRPM3 ME/CFS NK-cell findings (Cabanas et al. 2018, 2021), and the endorphin-rebound pharmacology (Younger, Parkitny, and McLain 2014). The core claim — that a higher dose is not a superset of the effects of lower doses for a non-monotonic multi-target drug — follows directly from the four mechanisms and the diagnostic pattern in the series. The evidence ceiling is explicit: no within-range LDN dose-response trial exists in any condition, and the single-dose trials that do exist are null (Due Bruun et al. 2024; Nielsen, Vaegter, and Due Bruun 2026; Ologunowa et al. 2025). Certainty on the framework itself is low (~0.30); the logical point about dose non-propagation does not depend on the framework being correct.

LDN series: Part 1: Why More Isn’t Better · Part 2: The Inverted-U Is Not One Thing · Part 3: Your LDN Dose Is a Diagnosis · Part 4: The Pharmacopoeia

The idea: inflammation degrading attention circuits

The classic picture of ADHD is developmental: the brain’s dopaminergic and prefrontal circuits were built with a particular configuration from childhood. But an ADHD-like presentation does not require a developmental cause. A later inflammatory process can degrade the very same circuits, producing the same symptoms through a different route.

The mechanism proposed is a cascade:

1. Inflammation raises its markers. Elevated pro-inflammatory cytokines — notably IL-6 and TNF-alpha — are documented independently in both ADHD and chronic fatigue Dunn et al. (2019).
2. Microglia activate. Activated microglia are the brain’s resident immune cells, and their activation is documented in ADHD (Yokokura et al. 2021) and in chronic fatigue (Nakatomi et al. 2014).
3. Dopamine is depleted. Inflammation diverts the cofactor BH4 away from dopamine synthesis toward a different branch of its metabolism, and drives oxidative stress that damages dopaminergic terminals. The result is a lower dopamine supply in the same circuits that primary ADHD affects.
4. Prefrontal function degrades. The result is an ADHD-like clinical phenotype — inattention, impulsivity, executive difficulty — in a person whose pre-illness neurodevelopment was intact.

The key claim: sustained neuroinflammation can be sufficient to produce ADHD-like features in people who did not have ADHD before. That claim is a registered research speculation, not an established finding.

The evidence

Four findings sharpen the case:

1. Neuroinflammation is documented in ADHD. Substantial evidence supports neuroinflammation in ADHD pathophysiology, including elevated pro-inflammatory cytokines in children with ADHD and microglial activation in post-mortem and imaging studies (Dunn et al. 2019). Positron-emission tomography has demonstrated reduced dopamine D1-receptor availability co-localized with microglial activation in ADHD — the two phenomena in the same brain (Yokokura et al. 2021).

2. Neuroinflammation is documented in chronic fatigue — but contested. One study reported substantial elevation in microglial-activation markers across six brain regions (Nakatomi et al. 2014), but a subsequent study using the same tracer found no such elevation (Raijmakers et al. 2021). The neuroinflammation foundation for the same-root model is therefore contested, not established; every downstream claim inherits this uncertainty.

3. The shared inflammatory signal. ADHD traits at age 9 predict doubled chronic-fatigue risk at age 18, and this association is mediated by the inflammatory marker IL-6 (Quadt et al. 2024). This is the single most direct piece of evidence linking the inflammatory state of ADHD to later fatigue — but it establishes a statistical mediation, not a proven mechanism.

4. Convergent pharmacology. Both ADHD and chronic fatigue respond to the same dopamine and noradrenaline reuptake inhibitors Eckey et al. (2025). This is consistent with a shared dopaminergic deficit in both, whether developmental or acquired.

The orexin-dopamine bridge

A further hypothesis in our paper links the inflammatory strand to ADHD’s dopamine deficit through orexin — the wakefulness-promoting peptide. Orexin neurons project to the ventral tegmental area, where they regulate dopamine firing. Deficiency in this system reduces prefrontal dopamine (brain fog) and mesolimbic reward (anhedonia) (Sakurai et al. 1998).

The relevance to ADHD is specific: the same 8.1% lower cerebral glucose metabolism seen in ADHD may reflect a shared hypothalamic orexin deficit, and our paper’s hypothesis proposes that low orexin tone links to low dopamine metabolite and worse attention scores. Direct human evidence now exists: a study of drug-naive children with ADHD found decreased serum orexin A levels compared to controls (Baykal et al. 2019). Because inflammation can suppress orexin neuron function (the pathway by which chronic low-grade neuroinflammation is proposed to drive fatigue), an inflammatory hit that suppresses orexin would further deplete the already-low prefrontal dopamine of an ADHD-affected brain — closing a second route from inflammation to ADHD-like features.

This is a registered speculation with low confidence — the full orexin-inflammation-dopamine-ADHD chain is an inference across literatures, and the direct human orexin-ADHD finding is serum-based and single-study. It is included here for completeness as a distinct mechanism our paper develops, not as an established finding.

The kynurenine-dopamine bridge

A second, more specific route from inflammation to dopamine depletion runs through the kynurenine pathway — the tryptophan-metabolism branch that inflammation (via IDO) pushes toward neuroactive metabolites. Two of its products are directly relevant to dopamine:

• Kynurenic acid is neuroprotective at the NMDA receptor but, at low concentrations, it also reduces striatal dopamine release — an animal finding that connects kynurenine overactivation directly to the dopamine deficit implicated in ADHD (Rassoulpour et al. 2005).
• The pathway’s overactivation in the brain consumes NAD+ — the same cofactor mitochondria need for ATP — raising the possibility that kynurenine load and mitochondrial dysfunction compound each other, with the prefrontal cortex (the most energetically expensive region) showing the first deficit.

In humans recovering from COVID, kynurenine-pathway activation correlates with objective cognitive impairment (Cysique et al. 2023), and mitochondrial-complex gene suppression accompanies COVID cognitive decline (Xu et al. 2025). A 2026 translational viewpoint in Brain, Behavior, and Immunity brings these together explicitly, proposing a shared neuroimmune framework linking Long Covid and ADHD through convergent mechanisms — frontal-striatal-hippocampal dysfunction, catecholamine neuroimmune dysregulation, kynurenine overactivation, and mitochondrial bioenergetic defects (Spanoghe et al. 2026). It also notes that some Long Covid patients report partial benefit from ADHD-targeted medications (methylphenidate, guanfacine, low-dose lithium, dexamfetamine), with pre-existing autonomic dysregulation and post-exertional malaise as limiting factors Fesharaki-Zadeh, Lowe, and Arnsten (2023).

This is a registered speculation: the kynurenic-acid-lowers-dopamine finding is animal data, and the unified framework is a viewpoint without primary data of its own. But it sharpens the mechanism in this article’s cascade — inflammation does not only divert BH4 away from dopamine synthesis; its kynurenine arm can also act directly on dopamine release. The off-label pharmacotherapy reports are clinical observation, not a treatment recommendation.

The same-root hypothesis

Taken together, these lines point to a proposal from our research — the same-root hypothesis: ADHD and chronic fatigue (along with other neuroinflammatory conditions) may share a common root cause — chronic inflammation driving mitochondrial dysfunction and energy failure — with the specific diagnosis determined by which tissues are affected. The causal chain is: inflammation → cytokine-mediated metabolic suppression → mitochondrial ATP deficit → brain energy failure (an ADHD phenotype when compartmentalized) or systemic energy failure (a fatigue phenotype when generalized).

Inflammation is the proposed primary driver; mitochondrial dysfunction is the mechanistic intermediate; symptom expression is determined by which tissues cross their energy threshold first.

This is a registered speculation with low confidence — a hypothesis to test, not an established finding. It is not the claim that all ADHD is inflammatory; it is the claim that a subset of ADHD-like features may be acquired and immune-driven.

Where a mechanism already has a dedicated article

Several adjacent mechanisms have dedicated articles in this blog:

• Why People with ADHD Get ME/CFS at Twice the Rate — the epidemiological bridge between ADHD and post-infectious fatigue.
• Your ADHD Is an Energy Problem — the energy-production argument this strand connects to.
• One Cofactor, Six Conditions, One Bottleneck — the BH4 cofactor that inflammation diverts away from dopamine synthesis.

The honest limits

This framework is explicitly speculative, with low certainty.

• The core neuroinflammation premise in chronic fatigue is contested — one study found it, a replication did not Raijmakers et al. (2021).
• No study has measured the same inflammatory and microglial panel simultaneously across ADHD, chronic fatigue, and healthy controls.
• The same-root hypothesis rests on cross-sectional and retrospective designs; no prospective study has tracked ADHD-like symptom emergence after an inflammatory trigger in a pre-illness-characterized cohort.
• The framework covers a subset of ADHD-like features, not the entire condition. Most ADHD is developmental and stable, not acquired.
• Even where inflammation is present, the causal direction is unresolved: inflammation could drive dopamine depletion, or chronic dopaminergic dysregulation could secondarily stress the immune system.

Convergence does not raise certainty above the weakest link. Until a multi-disease, prospective study exists, this is a research framework, not a finding.

The decisive experiment

The field needs a prospective test: track a cohort after an inflammatory trigger (e.g., a viral infection) and measure whether new-onset inattention and executive difficulty emerge alongside markers of neuroinflammation, in people with no prior ADHD history. The framework predicts that the severity of acquired ADHD-like features should track inflammatory markers and should improve when inflammation is treated.

If acquired inattention follows inflammation and reverses with anti-neuroinflammatory treatment, the same-root model is supported. If no new-onset ADHD-like features appear even in the presence of neuroinflammation, the model fails for that population.

What to take away

The immune and neuroinflammatory model does not claim that ADHD is an immune disease. It claims that a subset of ADHD-like features may be driven by inflammation and microglial activation — an immune process that can deplete dopamine and degrade the same prefrontal circuits that primary ADHD affects through development.

The encouraging part: if some ADHD-like features are immune-driven, they may be acquired and reversible — addressable by treating the inflammation rather than accepted as permanent wiring.

The honest part: this is a research framework, not a clinical recommendation. The neuroinflammation premise is contested, no screening is indicated, and no immunomodulatory treatment for ADHD-like features is established. The decisive prospective study has not been done.

This is Part 2 of a four-part series on the biology of ADHD. Part 1 covers the prefrontal-energy model and the multi-pathway treatment hypothesis. Part 3 explores the dopamine-Nrf2-NLRP3 axis. Part 4 asks when ADHD-like features are acquired and reversible. See the series landing page.

This article reflects research hypotheses from our documentation project. The immune and neuroinflammatory strand of ADHD is an active area of investigation with explicit, low confidence — not established clinical fact. Discuss any medical decision with a qualified clinician.

What ADHD is, on its own

Before the energy model, it helps to state the established facts about ADHD that do not depend on any connection to chronic fatigue. ADHD is one of the most common neurodevelopmental conditions — a systematic review places its worldwide childhood prevalence near five percent (Polanczyk et al. 2007) — and while only a minority of children continue to meet full diagnostic criteria as adults, a larger fraction retain symptoms below that threshold (Part 4 develops this). Its core features are inattention, hyperactivity, and impulsivity, and its classic neurobiology centers on dopamine and noradrenaline: the classic hypothesis is that ADHD involves reduced dopaminergic signaling in circuits that subserve attention and reward, a view supported by decades of pharmacology and by functional imaging showing altered dopamine transporter and receptor availability (Volkow et al. 2011). It is highly heritable and genetically overlapping with autism (Part 3 develops this). Its standard treatments — stimulants and the non-stimulant atomoxetine — act on the catecholamine system, and network meta-analysis confirms their efficacy in reducing ADHD symptoms Mwesigwa et al. (2024).

The energy model in this part is best read as an addition to this foundation: it asks whether a prefrontal fuel shortfall can explain a core slice of the ADHD experience — not whether the dopamine hypothesis is wrong.

The model: a prefrontal cortex running on a thin fuel supply

The prefrontal cortex (PFC) governs executive functions: working memory, impulse inhibition, emotional regulation, attentional control, planning, and the ability to pause and choose a response rather than react. It is the most metabolically expensive region in the human brain. Under ATP scarcity, it is the first brain system to degrade.

The reason is specific. Working memory depends on persistent firing — neurons sustaining activity without external input — which requires the continuous operation of sodium-potassium pumps that burn ATP (Constantinidis et al. 2018). Functional neuroimaging consistently shows that cognitive tasks engaging the PFC produce the largest metabolic signal in the brain. This is not a design flaw; it is the price of higher cognition.

The consequence: any condition that reduces the brain’s ATP delivery or production will impair PFC function before any other brain function — and executive function before memory, motor skills, or vital functions. The brain does not simply crash under a fuel shortfall. It economizes, deprioritizing the most energy-expensive operations to protect the functions it needs to survive. Those energy-expensive operations are precisely the ones we recognize as core to ADHD:

• Sustained attention — holding focus on a task
• Impulse inhibition — pausing before acting
• Working memory — keeping information in mind while using it
• Task-switching — shifting between activities

The evidence: five independent lines point the same direction

If the prefrontal-energy model were an isolated idea, it would be easy to dismiss. It is not isolated. Five independent lines of evidence — drawn from different methods and different disease literatures — converge on the same geometric fact.

Line 1 — Cerebrovascular demand failure. The PFC is the region most sensitive to drops in blood flow because its metabolic rate is highest. In the chronic-fatigue literature, most patients with normal resting cerebral blood flow show abnormal blood-flow reduction during orthostatic challenge — roughly nine in ten in the largest quantitative tilt studies — preferentially starving the PFC when upright and producing postural cognitive impairment: brain fog that worsens when standing (Campen et al. 2020). ADHD itself shows shared prefrontal hypoperfusion (Berthier et al. 2025). Even a modest perfusion drop during upright posture preferentially starves the PFC.

Line 2 — Brain glucose hypometabolism. Position-emission tomography documents regional brain glucose hypometabolism in chronic fatigue Van Der Gucht et al. (2017). Directly in ADHD, the landmark Zametkin study showed 8.1% lower global cerebral glucose metabolism in ADHD adults, with the largest reductions in the superior prefrontal cortex (Zametkin et al. 1990). The ADHD brain is a PFC-hypometabolic brain.

Line 3 — Cerebral creatine depletion. The phosphocreatine shuttle is the brain’s rapid ATP buffer — essential for the high-burst demands of working memory and executive function. Magnetic-resonance spectroscopy documents brain creatine deficiency in chronic fatigue, with significant depletion in the dorsolateral prefrontal cortex and pregenual anterior cingulate (Godlewska et al. 2024). Honest caveat: this MRS creatine-depletion evidence comes from the chronic-fatigue literature, not from a direct study of ADHD brains — no ADHD-specific MRS study has established creatine depletion, and in the ADHD MRS literature creatine is usually used as a reference metabolite rather than measured as a depletion signal. The extrapolation from chronic fatigue to ADHD is a hypothesis, not a finding. If the PFC cannot buffer ATP demand spikes, this directly predicts impaired sustained attention (the PFC runs out of ATP mid-task) and impaired working memory (insufficient phosphocreatine for persistent firing) — but that prediction awaits an ADHD-specific test.

Line 4 — Dopaminergic-noradrenergic deficit. Prefrontal D1 dopamine and alpha-2A noradrenergic receptors operate on an inverted-U dose-response curve: too little catecholamine tone impairs executive function through insufficient receptor activation, while too much impairs it through noise amplification Cools and D’Esposito (2011). The chronic-fatigue literature documents reduced cerebrospinal-fluid catecholamines (Walitt et al. 2024), and the striking stimulant response for brain fog — reported by the large majority of surveyed patients (Eckey et al. 2025) — is consistent with left-arm (suboptimal) PFC catecholamine tone, where low ATP availability compounds low transmitter availability to doubly impair executive circuits.

A further dopamine account in primary ADHD distinguishes incentive-salience instability from tonic depletion. On one influential reading, primary ADHD is characteristically a fluctuation of dopaminergic tone rather than a chronic depletion — producing intermittent hyperfocus alongside inattention (Verma et al. 2016). If that distinction holds, a chronic-fatigue-associated ADHD-like presentation should look different: a more uniform tonic suppression of effort, which may respond differently to stimulants. This prediction has not been directly tested, and the series registers it as a speculation rather than a finding.

Line 5 — Epidemiological gradient. Childhood ADHD is present in nearly 30% of adult chronic-fatigue patients (Sáez-Francàs et al. 2012). ADHD traits at age 9 predict doubled chronic-fatigue risk at age 18, mediated by the inflammatory marker IL-6 (Quadt et al. 2024). This is not merely comorbidity — it is a dose-response gradient consistent with a shared biological substrate: patients who start with lower PFC metabolic reserve are the first to cross the clinical threshold when an additional metabolic hit further reduces brain energy availability.

The synthesis: one mechanism, five paths to it

All five lines converge on a single fact: the PFC is the brain region with the highest ATP demand per gram of tissue. PFC function therefore degrades first under energy scarcity, regardless of whether the scarcity arises from reduced cerebral perfusion (Line 1), impaired glucose metabolism (Line 2), depleted creatine buffering (Line 3), inadequate catecholamine tone (Line 4), or an inherited lower metabolic baseline (Line 5).

Each line is an independent path to the same endpoint — PFC energy failure — and a given patient may have one, several, or all of these pathologies superimposed. The ADHD-like executive dysfunction that appears in chronic fatigue is not a separate comorbidity requiring a separate diagnosis. It is the necessary and predictable consequence of any condition that reduces the brain’s ATP budget below what the PFC needs.

What this could mean for patients: the modifiable part

If part of the symptom burden is a fuel shortfall in the PFC, then part of it may be addressable — not by changing the ADHD brain, but by improving its energy supply. Four targets have real evidence:

1. Creatine — rebuilding the rapid ATP buffer. Creatine supplementation partially restored the depleted phosphocreatine pool in the prefrontal regions where it was most depleted (Godlewska et al. 2024). If the PFC cannot buffer ATP demand spikes, replenishing the buffer is a direct substrate intervention.

2. BH4 cofactor support. Tetrahydrobiopterin (BH4) is the essential helper molecule for the enzymes that make dopamine and noradrenaline. If low BH4 is a real bottleneck — a theme our project has explored across six conditions, and the subject of a dedicated article (see below) — then supporting its production or recycling could relieve the monoamine deficit downstream.

3. Stimulants — temporary compensation, not a cure. Stimulants improve brain fog in the large majority of surveyed chronic-fatigue patients (Eckey et al. 2025), and both ADHD and chronic fatigue respond to the same dopaminergic and noradrenergic reuptake inhibitors (Blockmans et al. 2006). But stimulants temporarily raise catecholamine tone; they do not fix the underlying energy deficit. They compensate for PFC energy failure pharmacologically rather than repairing it.

4. Non-stimulant catecholamine modulators. ADHD’s non-stimulant treatments act on the same prefrontal catecholamine system through a different route, and our paper notes they may be particularly relevant where stimulants carry risk. Guanfacine — an alpha-2A agonist that improves working memory in ADHD — acts postsynaptically in the prefrontal cortex without the same dopaminergic-boosting profile, and the atomoxetine-plus-guanfacine combination used clinically in ADHD is generally well tolerated (Cortese et al. 2018). For a patient whose catecholamine tone is suboptimal, these offer a mechanism to raise prefrontal signaling without the oxidation-and-depletion loop that dopamine-raising stimulants risk.

Safety caveat. In chronic fatigue, stimulants are not a free pass: they can enable activity beyond the body’s true capacity and precipitate post-exertional malaise — the same energy-envelope logic the existing stimulants article develops. They are compensation, not repair, and the dopamine-quinone mechanism in Part 3 predicts they may also provoke inflammation in ADHD-comorbid patients. None of this is a treatment recommendation; discuss any intervention with a qualified clinician.

Where a mechanism already has a dedicated article

Several of these mechanisms have dedicated articles in this blog. This series is the comprehensive overview; the dedicated articles go deeper:

• Your ADHD Is an Energy Problem — the core energy-production argument for ADHD.
• One Cofactor, Six Conditions, One Bottleneck — the BH4 cofactor bottleneck.
• Why Stimulants Help Brain Fog but Not PEM — why stimulants temporarily compensate cognition but do not touch post-exertional malaise.
• Why People with ADHD Get ME/CFS at Twice the Rate — the epidemiological gradient.
• The Crash After Hyperfocus: Is ADHD Already Experiencing Micro-PEM? — our paper’s hypothesis that ADHD hyperfocus is a focal form of the same energy-envelope depletion that underlies post-exertional malaise, triggered by cognitive rather than physical demand.

The honest limits: what the model does not claim

It would be a disservice to patients to oversell this. The model is promising but far from proven, and it has sharp limits.

The model is a convergence reading, not a single replicated finding. Each line has independent evidence, but no study has measured all five in the same ADHD patients at once. The core PFC-specific claim rests on reading across disease literatures.

The cause could be downstream. The observed hypometabolism could equally arise from reduced blood flow, chronic inflammation, or physical deconditioning — none of which requires a glucose-transport or creatine defect. If the real problem is delivery or inflammation rather than the energy machinery itself, substrate-repletion approaches will help less.

Not every ADHD symptom is energy-driven. The model is most confident about inattention, impulse control, working memory, and the exhaustion of sustained effort. It does not claim that the developmental wiring of an ADHD brain is simply a fuel problem that can be fixed. Some ADHD is developmental; that wiring is not reversible by creatine.

Certainty is mostly low. This is a registered speculation. Our project assigns it explicit low confidence — in the range of a hypothesis to test, not a settled finding. It is presented as a testable model, not established clinical fact.

The decisive experiment

There is one test that would separate the model from the alternatives: measuring PFC energy status directly and asking whether executive-function severity tracks it. The framework predicts that executive-function severity should correlate with objective measures of cerebral energy status — the MRS phosphocreatine-to-ATP ratio, FDG-PET PFC glucose uptake, and blood flow during orthostatic challenge — and that treatments improving cerebral energy delivery should improve executive function in proportion to the degree of the baseline PFC energy deficit.

If the correlation holds, the energy model is supported. If executive severity is independent of these markers, the real problem is elsewhere.

An open question: the shape of the cognitive deficit

The prefrontal-energy framework makes a specific prediction about which part of attention fails. A double-blind trial in adolescents with ADHD compared a high-dose L-theanine-caffeine combination against methylphenidate and placebo on a selective-attention task: both the combination and methylphenidate reduced false alarms and sped up the brain’s neural evaluation of targets, but only methylphenidate shortened behavioral reaction time (Nawarathna et al. 2026). The pattern suggests the dopaminergic and non-dopaminergic routes may contribute differently to selection (picking the right target) and speed (deploying the response).

This trial is small (n = 21) and has not yet been independently replicated (Nawarathna et al. 2026). The distinction it points to between a spared selection route and a failing speed route therefore rests on a single small dataset from one research group; an independent replication has not been published.

Two readings point opposite ways, and both are testable:

• If the non-dopaminergic selection route is comparatively spared while the dopaminergic speed route fails, chronic fatigue should show a slow-but-accurate pattern — preserved error rates with selectively slowed responses.
• If the PFC’s ATP-expensive inhibitory networks fail first under energy scarcity (the framework’s own prediction), the pattern should instead be slow-and-inaccurate — increased false alarms alongside slowed responses.

Which pattern chronic-fatigue patients show is an open empirical question. The series registers it as such, and as a registered low-confidence speculation rather than a finding.

The multi-pathway treatment hypothesis

The prefrontal-energy model is one mechanism. It is not the only one. Parts 2–4 of this series examine two further mechanisms (an immune-neuroinflammatory strand, and a dopamine-Nrf2-NLRP3 axis) and one distinction (acquired vs. developmental features). A patient can carry several of these at once — an energy deficit and iron deficiency and an inflammatory strand, for example — each contributing a different slice of the daily burden.

This raises the central treatment question: can several pathways be treated at once, to remove or dampen symptoms and give the patient a more normal life?

What to take away

The prefrontal-energy model reframes a core piece of the ADHD experience — inattention, impulse-control difficulty, working-memory failure, the exhaustion of sustained effort — as the brain’s metabolic reality rather than a behavioral choice. That reframe alone matters: it validates experiences that are often dismissed.

The encouraging part is that some of the fuel variables are modifiable: creatine, iron, BH4 cofactor status, the non-stimulant catecholamine modulators, and (as temporary compensation) stimulants. If even a subset of people with ADHD carries a fixable prefrontal energy deficit, targeted interventions could relieve a meaningful fraction of the daily symptom burden — without pretending to “fix” ADHD itself.

The honest part is that this is a hypothesis, not a settled finding. It points to specific, cheap, and testable experiments, and it should be tested before it is treated as fact.

This is Part 1 of a four-part series on the biology of ADHD, drawn from our research. Part 1 covers the prefrontal-energy model and the multi-pathway treatment hypothesis. Part 2 examines the immune and neuroinflammatory strand. Part 3 explores the dopamine-Nrf2-NLRP3 axis. Part 4 asks when ADHD-like features are acquired and reversible. See the series landing page.

This article summarizes research from our ME/CFS documentation project, where these cross-disease energy models were developed. The ADHD-specific framing here is our reading of the literature; it reflects hypotheses with explicit, often low, confidence — not established clinical fact.

What we know, what our research adds

Every article in this series distinguishes two things clearly:

1. What we know — established scientific findings, cited to the peer-reviewed literature. This includes ADHD’s own well-established science — its core dopamine neurobiology (Volkow et al. 2011), its genetic architecture and overlap with autism Lai et al. (2019), and its standard treatments — alongside the cross-disease findings (the prefrontal hypometabolism in ADHD (Zametkin et al. 1990), the inverted-U catecholamine pharmacology (Arnsten 2011), and the debated adult-onset trajectory Asherson and Agnew-Blais (2019)).
2. What our research adds — the hypotheses our documentation project developed by reading across disease boundaries, and which are not yet established: the prefrontal-energy convergence model, the multi-pathway treatment hypothesis, the acquired-reversible framing, and the dopamine-Nrf2-NLRP3 cross-disease axis.

The distinction matters because the two are not the same confidence. “What we know” is evidence; “what our research adds” is testable hypothesis. The series keeps them separate so you are never misled into treating a hypothesis as a finding.

Each article follows the same shape:

1. What it is — the mechanism, in ordinary words.
2. The evidence — what supports it, and how strong that evidence is.
3. The honest limits — what it does not claim.
4. The decisive experiment — how the idea could be confirmed or refuted.

Part 1: ADHD as a Prefrontal Energy Disorder — and the Multi-Pathway Treatment Hypothesis

The series opens with the most developed strand: a prefrontal cortex running on a thin fuel supply. The prefrontal cortex is the most energy-expensive region in the human brain, so it is the first to degrade under any energy scarcity. Five independent lines — blood-flow failure, glucose hypometabolism, creatine depletion, catecholamine deficit, and an epidemiological gradient — converge on this single point. Part 1 also carries the series’ synthesis: the multi-pathway treatment hypothesis, and what treating several mechanisms at once could mean.

Read the article:

1. ADHD as a Prefrontal Energy Disorder: Five Lines of Evidence, One Mechanism — what ADHD is on its own, the cerebrovascular, glucose, creatine, and catecholamine evidence, the epidemiological gradient, the honest limits, the decisive experiment, the slow-but-accurate open question, and the multi-pathway treatment hypothesis.

Part 2: The Immune and Neuroinflammatory Strand of ADHD

Not all ADHD may have the same cause. A growing line of evidence links ADHD-like features to chronic neuroinflammation and microglial activation — an immune-driven process that can deplete dopamine and degrade the same prefrontal circuits that primary ADHD affects through development. If some ADHD-like symptoms are driven by inflammation, they may be acquired, reversible, and identifiable.

Read the article:

1. The Immune and Neuroinflammatory Strand of ADHD: When Inflammation Shapes Attention — the microglial evidence, the dopaminergic depletion route, the shared pro-inflammatory signal, the orexin-dopamine bridge, the kynurenine-dopamine bridge, the same-root hypothesis, and the honest limits.

Part 3: The Dopamine-Nrf2-NLRP3 Axis

A third strand links dopamine biology to the cell’s antioxidant-defense system and to the immune system’s central inflammatory switch. Low dopamine tone impairs Nrf2-driven antioxidant defenses, which in turn removes a brake on NLRP3-inflammasome activation — and the resulting inflammation further depletes dopamine, closing a self-amplifying loop. This cross-disease axis may explain why people with ADHD carry a lower threshold for post-infectious fatigue.

Read the article:

1. The Dopamine-Nrf2-NLRP3 Axis: An Inflammatory Loop Behind ADHD and Fatigue — the dopamine-Nrf2 link, the Nrf2-NLRP3 brake, the feedback to dopamine synthesis, the BH4 bottleneck, ADHD’s own genetic architecture and its overlap with autism, the shared mitochondrial-genetic strand (haplogroup U, cybrid evidence), and the testable predictions.

Part 4: Acquired vs. Developmental — When ADHD-Like Features Might Be Reversible

The clinically consequential question: are ADHD-like features necessarily present from development, or can some be acquired — and therefore reversible? Much adult ADHD is not a straightforward continuation of childhood ADHD — some is adolescent-onset, some reflects missed or misrecalled childhood symptoms, some is explained by substance use or comorbidity — and acquired inattention after an immune trigger can be reversible when its cause is treated. Telling the two apart determines whether a symptom might respond to treating its cause or is stable wiring.

Read the article:

1. Acquired vs. Developmental: When ADHD-Like Features Might Be Reversible — the adult-onset trajectory, the Architecture A/B/C framings, the interoceptive and neuroinflammatory routes, and the clinical consequences of getting the distinction wrong.

A note on honesty

Every article in this series distinguishes carefully what is confirmed by evidence, what is a promising hypothesis, and what is an individual patient’s experience. Most of the content here is at the hypothesis end — the certainties are explicit, often low, and nothing is presented as established clinical fact or a treatment recommendation.

The underlying science is drawn from our documentation project, which develops these cross-disease energy and immune models in detail. The ADHD-specific framing is our reading of the literature, written to stand alone for anyone curious about the biology.

Discuss any medical decision with a qualified clinician.

The distinction

There are two ways autism-like features can be present:

• Developmental (baseline). The brain’s circuits were built this way from early development. This is the classic picture — stable, trait-like, present from childhood. Parts 1–3 describe developmental mechanisms.
• Acquired (secondary). The underlying wiring developed normally, but a later process — neuroinflammation, an immune insult, a metabolic disturbance — degrades the same circuits, producing the same symptoms through a different route.

The key claim of this part: acquired autism-like features should be context-dependent and potentially reversible, whereas developmental autism is stable and trait-like. Our research registers this as a formal speculation — not an established finding.

The mechanism: disrupted interoception

The proposed route to acquired autism-like features runs through the interoceptive hierarchy — the brain’s system for monitoring and predicting the body’s internal state.

Normally, the brain maintains a balance between its predictions about the body and the sensory evidence the body sends back. Two failure modes produce similar outward symptoms but from opposite ends:

• Developmental ASD is thought to involve overprecise priors — the brain’s predictions are too rigid, suppressing incoming sensory updates. The result: stable, trait-like perceptual rigidity.
• Acquired features are thought to involve corrupted lower-level signals — the brainstem and thalamocortical pathways fail to deliver accurate bodily information. The brain compensates by increasing the precision of its priors to suppress the noise.

The phenomenological result can look identical: sensory overwhelm, difficulty reading internal states, apparent alexithymia (an inability to name one’s emotions — not from absent emotion, but from an inability to link it to a specific bodily state). But the computational locus differs — and so should the response to treatment.

The test: context-dependence

Because the two routes differ in stability, they can be told apart empirically:

If acquired autism-like sensory rigidity shows the same trait stability and context-independence as developmental autism — persisting unchanged and not correlating with markers of neuroinflammation — the acquired model is not supported.

In other words: developmental rigidity is always there. Acquired rigidity should fluctuate with the underlying process (inflammation, metabolic state) and should improve when that process is treated.

A second mechanism: neuroinflammation and microglial activation

A related route is chronic neuroinflammation — sustained microglial and astroglial activation degrading prefrontal, mesolimbic, and thalamocortical circuits. The proposal is that sustained neuroinflammation can be sufficient to produce autism-like and ADHD-like phenotypes in people whose pre-illness neurodevelopment was intact — a registered research speculation.

If this cascade is the proximate cause, then interventions targeting neuroinflammation should partially reverse these secondary phenotypes — a testable and not-yet-tested prediction.

A third mechanism: disrupted sleep and clearance

A further acquired route operates through sleep. Neurodivergent people have high rates of sleep disruption. If sleep — the brain’s primary window for repair and for clearing metabolic waste (the glymphatic system) — is chronically degraded, this compounds any underlying energy or inflammatory disturbance and can worsen cognitive and sensory symptoms over time.

This is the most speculative strand, with the lowest confidence, and it involves a long chain of untested links. It is included for completeness, not because it is well supported.

The honest limits

• The acquired model is explicitly speculative — a registered hypothesis with low confidence — and a simpler explanation (inflammation directly causing both the symptoms and the underlying process) may account for the findings without any interoceptive framework.
• None of the theoretical components has been directly validated in this context.
• The distinction between acquired and developmental is not yet usable at the bedside: no test currently separates the two in an individual patient.
• Even if some features are acquired and reversible, this does not mean “autism can be cured.” It means a subset of the symptom burden in some people may respond to treating its cause.

The clinical consequence — and the harm of getting it wrong

The reason this distinction matters clinically:

• If acquired features are dismissed as “just autism”, a treatable underlying process (inflammation, a metabolic deficit, disrupted sleep) goes unaddressed.
• If developmental features are treated as acquired and reversible, patients are offered treatments that will not work, and may blame themselves when they don’t.

The honest position is that acquired autism-like features should be supported, not pathologized — and the underlying modifiable drivers (sleep, inflammation, iron, energy) should be addressed regardless of the label, because they affect overall wellbeing even when the core developmental wiring is unchanged.

What to take away

Some autism-like features — sensory rigidity, withdrawal, apparent alexithymia — may be acquired after development rather than built in from it. If they are, they should be context-dependent and potentially reversible with treatment of the underlying cause: neuroinflammation, metabolic disturbance, or sleep disruption.

The encouraging part: this means a meaningful fraction of the daily symptom burden in some people may be addressable without claiming to “fix” autism itself.

The honest part: this is a low-confidence hypothesis, and the decisive test — showing that acquired rigidity fluctuates with underlying inflammation while developmental rigidity does not — has not been run.

This concludes the four-part series on the biology of autism. Part 1 examined the brain-energy model and the multi-pathway treatment hypothesis, Part 2 the immune-mediated subset, Part 3 the cerebellar-glutamate connection, and Part 4 the distinction between acquired and developmental features.

This article reflects research hypotheses with explicit, low confidence — not established clinical fact. Discuss any medical decision with a qualified clinician.

The cerebellum is more than motor control

For decades the cerebellum was understood as a motor-coordination structure. A now-extensive literature places it in cognition, affect, and sensory processing as well — the “cerebellar cognitive-affective syndrome.”

The relevance to autism is direct: cerebellar Purkinje cell loss is among the most replicated neuropathological findings in autism. Purkinje cells are the cerebellum’s output neurons, and their vulnerability appears central to the condition.

The shared features: autism and the glutamatergic signal

Our research identified four features shared between autism and ME/CFS — the chronic-fatigue condition our documentation project centers on — suggesting a common glutamatergic substrate (Maccallini 2026):

1. Glutamatergic synapse genetic enrichment in GWAS — the genetic risk signal concentrates in excitatory synaptic genes.
2. Purkinje cell vulnerability to metabolic and inflammatory stress — the same cells that are lost in autism are sensitive to both fuel shortage and immune activation.
3. Sensory hypersensitivity — heightened auditory and tactile sensitivity.
4. Autonomic dysfunction — dysregulation of the involuntary nervous system.

The genetic bridge is the key move. Genome-wide association data shows enrichment in cerebellar and glutamatergic neuronal tissue — independent of the peripheral immune and metabolic signals. This is a wiring-level genetic signal, present from development.

The hypothesis: a shared developmental vulnerability

The proposal, in formal terms (a registered speculation in our research) (Maccallini 2026):

Cerebellar glutamatergic dysfunction is a shared developmental vulnerability. If the cerebellum’s excitatory circuits are built differently — or more vulnerable — from early development, this could explain the disproportionate overlap between autism and other conditions, the sensory hypersensitivity, and the autonomic symptoms.

The mechanism is not that the cerebellum “causes” autism wholesale. It is that a difference in glutamatergic wiring — present from development — produces a specific, measurable pattern of symptoms, chief among them sensory sensitivity.

The testable prediction

This hypothesis is falsifiable, and the prediction is specific:

A cerebellar cell-type polygenic risk score (derived from the genetic enrichment data) will correlate with Sensory Profile questionnaire scores — especially auditory and tactile sensitivity — in autism cohorts, and will be elevated in people with co-occurring autistic traits.

If true, sensory sensitivity would be traceable to a specific, genetically measurable wiring difference — not a vague “overall sensitivity,” but a cerebellar glutamatergic signal that can be quantified in an individual’s genome.

The honest limits

This is a speculative hypothesis, with explicit low confidence — a registered speculation, not an established finding.

• The Purkinje-cell finding is replicated neuropathology, but the link to the genetic signal is an inference.
• The cerebellar and glutamatergic enrichment comes from gene-set analysis, which is sensitive to analytic choices and can be biased toward long, conserved genes.
• The glutamatergic signal is one contributing reading of otherwise-disconnected phenomena — it does not resolve them alone.
• Sensory sensitivity is multiply determined; even if the cerebellar prediction holds, it is one of several contributing mechanisms.

The prediction is testable now, with existing genetic data and existing questionnaires. It has not yet been run.

What to take away

The cerebellar-glutamate model offers a wiring-level explanation for a core autism symptom — sensory sensitivity — that is neither a fuel problem nor an immune problem. It is a developmental difference in how excitatory circuits are built.

Unlike Parts 1 and 2, this strand has no direct treatment implication yet. It is not “modifiable” in the same sense as iron or BH4, and it is not an antibody target. Its value is explanatory and diagnostic: it points to a specific, genetically measurable signature of sensory sensitivity.

The encouraging part: the prediction is cheap and testable with existing data. If it holds, it gives autism’s sensory experience a concrete biological anchor. If it fails, it narrows the search.

This is Part 3 of a four-part series on the biology of autism. Part 1 covers the brain-energy model and the multi-pathway treatment hypothesis. Part 2 covers the immune-mediated subset. Part 4 asks when autism-like features are acquired and reversible.

This article reflects a registered research hypothesis with low confidence, not established clinical fact.

The prenatal route: maternal antibodies

During pregnancy, maternal antibodies cross the placenta. In most cases this is protective. But a body of research suggests that in a subset of autism, maternal antibodies may target fetal brain proteins.

The striking number: approximately 23% of mothers of autistic children carry anti-fetal-brain autoantibodies targeting specific neuronal proteins Meltzer and Van de Water (2017). These antibodies are associated with increased autism risk.

The honest caveats are substantial: - The causal pathway — from maternal antibody to the child’s diagnosis — has not been proven. - The “attributable fraction” depends on assumptions about antibody pathogenicity that remain untested in prospective cohorts. - This defines a distinct immune-associated subgroup, but it is not yet clinically actionable.

Critically, there is no current clinical application: no prenatal screening for anti-fetal-brain antibodies is recommended, and immunomodulation during pregnancy for this indication is untested and potentially harmful.

The postnatal route: infection-triggered antibodies

The prenatal pattern has a postnatal equivalent. Infection can trigger anti-neuronal antibodies that produce autism-like symptoms in previously neurotypical children — structurally analogous to PANDAS/PANS (Pediatric Acute-onset Neuropsychiatric Syndrome), differing only in the developmental window and the circuits targeted Whiteley et al. (2021).

This is not fringe medicine. It is the same disease logic that underlies PANDAS/PANS: infection → immune response → anti-neuronal antibodies → circuit-specific neuropsychiatric phenotype Kirvan et al. (2006). The leading mechanism is molecular mimicry — antibodies formed against a bacterium also recognizing neuronal proteins in the brain.

The evidence for an immune subset within autism

Two findings sharpen the case that an immune-mediated subset exists within the autism label:

1. Anti-neuronal antibodies are elevated in a fraction of autism — regardless of regression status. About 20% of ASD patients carry elevated anti-neuronal antibodies, and this does not neatly align with the regression phenotype (Aslan et al. 2021). The immune subset does not respect the syndromic boundary between regressive and classic autism.

2. Immunotherapy response depends on antibody stratification, not the diagnosis. The clearest proof comes from a pair of IVIG (intravenous immunoglobulin) studies in autism. In one, immunotherapy produced a dramatic response in antibody-stratified patients; in the other, the same treatment was null in unselected autism Plioplys (1998). When recruitment uses the diagnosis alone, the trial is diluted by patients who are not immune-mediated. When it uses antibody profiling, the intervention works in the subset it targets.

This contrast is the single most instructive finding in the field: the therapy is not “effective” or “ineffective” for autism — it is effective for the immune-mediated subset.

The unifying framework: a neuroimmune encephalopathy spectrum

Taken together, these lines point to a broader proposal from our research — the Neuroimmune Encephalopathy Spectrum (NES) — a single disease class defined by an immune trigger, anti-neuronal antibodies targeting specific circuits, and a clinical phenotype determined by which circuits are hit, not by the trigger. It is a registered research speculation, not an established finding.

Under this framework: - Basal-ganglia-directed antibodies → OCD, tics, behavioral regression (PANDAS/PANS) - Brainstem/thalamic antibodies → fatigue, autonomic instability - Diffuse cortical antibodies → psychosis, cognitive collapse (autoimmune encephalitis) - Prenatal maternal antibody transfer → an autism-like developmental phenotype

The practical implication is radical: diagnostic labels (PANDAS, ME/CFS, autism) may be surrogates for circuit identity. The same biology produces all three depending on which circuits the antibodies target. If true, the right question is not “does this patient have autism?” but “which circuits, and are they antibody-targeted?”

The honest limits

This framework is explicitly speculative, with low certainty.

• No study has run the same anti-neuronal antibody panel simultaneously across PANDAS, autism, ME/CFS, and healthy controls.
• The single PANDAS RCT was underpowered and equivocal (Williams et al. 2016).
• The IVIG-in-autism evidence is a single specialist-center series, not a multi-site trial.
• No prospective study has screened children presenting with “regressive autism” for PANS/PANDAS criteria.
• The framework covers a subset within each diagnostic category, not the entire condition. Most autism is not immune-mediated.

Convergence does not raise certainty above the weakest link. Until a multi-disease antibody-profiling study exists, this is a research framework, not a finding.

The decisive experiment

The field needs one study: run the same comprehensive anti-neuronal antibody panel (basal-ganglia, D1/D2 receptor, CaMKII, lysoganglioside, NMDAR, and others) across PANDAS/PANS, regressive autism, ME/CFS, and healthy controls simultaneously, and test two predictions:

1. Antibody profiles cluster by circuit target, not by diagnostic label.
2. Immunotherapy response is predicted by antibody profile, not by diagnosis.

If these hold, the immune-mediated subset of autism becomes identifiable and treatable — regardless of which syndromic label it carries. If they fail, the syndrome-based approach was correct after all.

What to take away

The immune-mediated model does not claim that autism is an immune disease. It claims that a subset of people under the autism label carry a potentially identifiable, potentially treatable immune mechanism — whether from maternal antibodies or an infection-triggered response.

The encouraging part: the Connery/Plioplys contrast is proof-of-concept that selecting the right subset can turn a null treatment into a dramatic response.

The honest part: this is a research framework, not a clinical recommendation. No screening, no prenatal intervention, and no immunotherapy is indicated for autism based on the current evidence. The decisive multi-disease antibody study has not been done.

This is Part 2 of a four-part series on the biology of autism. Part 1 covers the brain-energy model and the multi-pathway treatment hypothesis. Part 3 explores the cerebellar-glutamate connection. Part 4 asks when autism-like features are acquired and reversible.

This article reflects research hypotheses from our documentation project. The immune-mediated autism subset is an active area of investigation with explicit, low confidence — not established clinical fact. Discuss any medical decision with a qualified clinician.

The model: a brain running on a thin fuel supply

In 2026, a review proposed a comprehensive model of autism as a brain energy disorder (Blagojevic-Stokic et al. 2026). The claim is specific:

• The brain’s fuel delivery depends on astrocytes — support cells that convert glucose into lactate and shuttle it to neurons.
• In the model, this shuttle is impaired: reduced glucose uptake, glycogen storage failure, and disrupted astrocyte-to-neuron lactate transfer.
• The result is suboptimal ATP availability — the brain literally does not have enough fuel.

The model’s key move is the “energy-saving adaptation.” Faced with a shortfall, the brain does not simply crash. It economizes: it deprioritizes the most energy-expensive operations to protect the functions it needs to survive. Those energy-expensive operations are precisely the ones we recognize as core to social life:

• Social communication — reading faces, tone, intent
• Cognitive flexibility — switching tasks, adapting to change
• Sensory integration — filtering noise, light, and multiple inputs

Under this model, sensory overload is not a personality quirk. Filtering out irrelevant stimuli requires inhibitory brain networks that burn large amounts of ATP. When ATP is scarce, the brain cannot afford to run those filters. The overwhelm is a direct biological consequence of a fuel budget, not a choice.

The evidence: this is not only in the head

If the brain-energy model were an isolated idea, it would be easy to dismiss. It is not isolated. Three independent lines of evidence point the same direction.

1. Mitochondrial dysfunction is systemic and baseline. A 204-study systematic review and meta-analysis found elevated lactate, pyruvate, and alanine in autism, along with creatine kinase and an elevated lactate-to-pyruvate ratio — the same electron-transport-chain markers seen in energy-metabolism disorders (Frye et al. 2024). Crucially, these are present at baseline, not only after stress. This is a constitutive feature of the body, not an artifact of brain demand.

2. A specific cofactor is consistently low. Tetrahydrobiopterin (BH4) is the essential helper molecule for the enzymes that make dopamine, serotonin, and nitric oxide. A systematic review found consistently lower BH4 in biological samples from autistic individuals (Colpani Filho et al. 2025). Low BH4 means the brain cannot make its key neurotransmitters and cannot regulate its own blood flow — one deficit, multiple downstream failures.

3. A connective-tissue connection. Autistic people are far more likely to have joint hypermobility and Ehlers-Danlos syndromes — a meta-analysis found 7.4× higher odds of EDS (Baeza-Velasco et al. 2025). This matters because hypermobility is linked to dysautonomia and reduced cerebral blood flow, which compounds any energy shortfall at the level of fuel delivery.

What this could mean for patients: the modifiable part

This is where the model becomes encouraging rather than merely explanatory. If part of the symptom burden is a fuel shortfall, then part of it may be addressable — not by changing the autistic brain, but by improving its energy supply. Three targets have real evidence:

1. Ketogenic diet — feeding the brain alternative fuel. Ketones (fat-derived molecules the brain can burn) bypass the damaged glucose/astrocyte step and enter the mitochondrion directly. Modified ketogenic diets improved behavior in children with autism (Lee et al. 2018), and a case report documented metabolic changes on brain scans (Żarnowska et al. 2018). This is proof-of-concept that alternative fuel can help some patients — not evidence that it helps all.

2. BH4 support. If low BH4 is a real bottleneck, interventions that support its production or recycling — such as folinic acid (which fuels the recycling enzyme) or vitamin C (which protects BH4 from oxidation) — could relieve the monoamine and vascular deficits downstream. This is mechanistically grounded but not yet tested in autism-specific trials.

3. Iron. Iron is required for both dopamine synthesis and mitochondrial complex I/II function. For ASD sleep phenotypes specifically, a ferritin threshold below 50 ng/mL is recommended for supplementation (DelRosso, Estrada Chaverri, and Ceballos Fuentes 2026); below that level, iron deficiency compounds both neurotransmitter and energy production. Correcting iron is cheap, safe, and reversible.

These are research hypotheses about modifiable factors, not clinical recommendations. Discuss any intervention with a qualified clinician.

The honest limits: what the model does not claim

It would be a disservice to patients to oversell this. The model is promising but far from proven, and it has sharp limits.

Which symptoms are energy-driven is unspecified. The model is most confident about sensory overload, cognitive rigidity, and the fatigue of socializing. It does not claim that core social cognition — the capacity to read and respond to social cues — is simply a fuel problem that can be fixed. Some autism is developmental wiring; that wiring is not reversible by a diet.

The cause could be something else. The observed brain hypometabolism could equally arise from reduced blood flow, chronic inflammation, or physical deconditioning — none of which requires a glucose-transport defect. If the real problem is delivery or inflammation rather than transport, ketones and glucose-bypass approaches will not help.

The direct evidence is thin. The core model rests on a single 2026 review, not a replicated finding. There is no direct measurement of the glucose transporter (GLUT1) in autistic brain tissue, and no study has measured brain metabolism while controlling for activity level, sleep, and inflammation together.

Baseline, not acquired. The energy systems in an autistic brain were disrupted during development — the circuits were built around the deficit. This matters because a brain built around a constraint may respond differently to an intervention than a brain that developed with full capacity and only later lost it. The model is developmental, so any intervention tested in adults must account for wiring that formed under energy constraint, not around it.

The decisive experiment

There is one test that would separate the model from the alternatives: measuring the CSF-to-plasma glucose ratio. If a glucose-transport bottleneck truly exists, spinal-fluid glucose should be low relative to blood glucose — the same signature used to diagnose GLUT1 deficiency. This is a standard clinical test, already used for GLUT1 deficiency, and has never been run in autism populations to test this specific question.

If the ratio is abnormal, it would define a biologically distinct subgroup with a clear treatment rationale. If it is normal, the transport-defect version of the model is likely wrong, and the real problem is delivery or inflammation instead.

The multi-pathway treatment hypothesis

The brain-energy model is one mechanism. It is not the only one. Parts 2–4 of this series examine two further mechanisms (an immune-mediated subset, and a wiring difference) and one distinction (acquired vs. developmental features). A patient can carry several of these at once — an energy deficit and iron deficiency and disrupted sleep, for example — each contributing a different slice of the daily burden.

This raises the central treatment question: can several pathways be treated at once, to remove or dampen symptoms and give the patient a more normal life?

What to take away

The brain-energy model reframes a core piece of the autism experience — sensory overload, cognitive rigidity, the exhaustion of socializing — as the brain’s metabolic reality rather than a behavioral choice. That reframe alone matters: it validates experiences that are often dismissed.

The encouraging part is that some of the fuel variables are modifiable: ketone-based fuel, BH4 cofactor status, and iron. If even a subset of autistic people carries a fixable energy deficit, targeted interventions could relieve a meaningful fraction of the daily symptom burden — without pretending to “fix” autism itself.

The honest part is that this is a hypothesis, not a settled finding. It points to specific, cheap, and testable experiments, and it should be tested before it is treated as fact.

This is Part 1 of a four-part series on the biology of autism, drawn from our research. Part 1 covers the brain-energy model and the multi-pathway treatment hypothesis. Part 2 examines the immune-mediated autism subset. Part 3 explores the cerebellar-glutamate connection. Part 4 asks when autism-like features are acquired and reversible.

This article summarizes research from our ME/CFS documentation project, where these cross-disease energy models were developed. The autism-specific framing here is our reading of the literature; it reflects hypotheses with explicit, often low, confidence — not established clinical fact.

The measurable finding

The anchor is a controlled radionuclide study. It measured gastric emptying — the time it takes the stomach to clear a meal — in 32 people with chronic fatigue syndrome, using a standardised scan that tracks a small amount of radioactive marker in both a liquid and a solid meal (Burnet and Chatterton 2004).

The result: 23 of 32 (72%) had delayed liquid emptying, and 12 of 32 (38%) had delayed solid emptying. The delay correlated significantly with the mean symptom score (p < 0.001) — the sicker the patient, the slower the stomach.

This is direct ME/CFS evidence that delayed gastric emptying is a real, measurable feature of the illness, not merely a subjective complaint. It can be assessed with a standard gastric emptying test, and it may be a target for treatment rather than something a patient must simply endure.

The honest caveat. The finding rests on a single 2004 cohort of 32 people. The study predates the modern diagnostic criteria for ME/CFS (the 2015 IOM and 2011 ICC criteria), and the finding has not been replicated in a modern-criteria cohort. In our evidence register this is a clinical finding with a certainty of 0.60 — solid, but anchored on one old study. The study did not report the severity of its participants, so we do not know whether the delay is stronger in severe or very-severe patients.

One counterpoint: a single adolescent case report found normal gastric emptying and normal myoelectrical activity in a young person with CFS (Corrado et al. 1998). Delayed emptying is therefore not universal in the illness. The case is one person, pediatric, and carries low evidentiary weight — it does not overturn the larger cohort finding, but it shows the delay is not present in every case.

What it feels like: the anxiety of a stomach you cannot trust

The numbers above are a scan. The experience is different.

When the stomach empties slowly, eating stops being a neutral act. It becomes a risk. You eat, and within minutes you feel full, bloated, or nauseous. You learn to fear the meal. You learn to fear the moment you might vomit in front of other people. You stop eating with others, because you cannot predict what will happen. You start to weigh every bite.

Over time, people tend to change their eating in the same direction. They eat smaller amounts, more often. They cut out the foods that sit heaviest — the fatty ones, the high-fibre ones. And many go to liquids: smoothies, broths, nutritional drinks, soups. Liquids empty faster than solids, so they feel safer. Eating only liquids is not a whim. It is a rational adaptation to a stomach that genuinely clears solids slowly.

This matters for two reasons. First, the anxiety is real and it has a cost. Fear of eating can narrow the diet to the point of malnutrition and weight loss, and it can tip into a disordered relationship with food that needs its own support — not just a diet sheet. A patient who is afraid to eat is a patient who needs a clinician, not more willpower.

Second, the adaptation is not wrong. The shift to small, frequent, low-fat, low-fibre meals and to liquid or semi-liquid nutrition is exactly the first-line, lowest-risk strategy that the evidence supports (see the treatment section below). The patient’s instinct and the evidence point the same way. The gap is that the patient usually arrives at it alone, by trial and error, without the safety net of a clinician watching weight and nutrition.

What the stomach is doing

The clinical picture of gastroparesis in ME/CFS is a recognisable set of symptoms:

• Feeling full after small amounts of food
• Persistent nausea
• Vomiting, especially of undigested food
• Abdominal bloating and discomfort
• Unpredictable blood sugar fluctuations

The mechanism is thought to be autonomic. The vagus nerve controls the motility of the stomach — the waves of contraction that grind food and push it onward. The leading reading is that autonomic dysfunction reduces vagal drive to the stomach in ME/CFS, so it empties slowly; that reading is a hypothesis, not yet proven in ME/CFS. Whatever the cause, the impaired emptying produces the digestive symptoms and the nutritional challenges above.

This is the functional reading: the stomach’s motor program is under-driven by the autonomic nervous system. It is the simplest explanation, and it fits the autonomic dysfunction that ME/CFS is known for. But it is not the only explanation, and it is not the one the evidence has settled on. We return to the alternatives after looking at the limits of the evidence.

There is a second consequence that matters specifically in ME/CFS. Erratic emptying is likely to mean erratic absorption — glucose and nutrients reach the blood in unpredictable bursts rather than a steady stream, because the rate at which the stomach empties shapes the postprandial glucose response (Camilleri 2026). This is a reasonable inference from the physiology, not yet demonstrated in an ME/CFS cohort. For a person already managing a tight energy envelope and orthostatic intolerance, that instability would add a metabolic load on top of the gastric discomfort. It is one more reason the slow stomach is not just a local nuisance but part of the wider energy picture.

The honest limits

Before the mechanism question, the limits. The evidence that ME/CFS patients have delayed gastric emptying is real but thin, and it has specific weaknesses:

• One old cohort. The direct evidence rests largely on the single 2004 cohort of 32 people, which predates modern diagnostic criteria (Burnet and Chatterton 2004).
• One null case. The one direct counter-example is a single adolescent with normal emptying (Corrado et al. 1998).
• No modern replication. No cohort using the current IOM/ICC criteria has replicated the finding.
• Non-specific alternatives. Several explanations other than a distinct ME/CFS gastroparesis lesion can produce the same symptom pattern: medication side-effects (many drugs used in ME/CFS slow gastric emptying), functional dyspepsia with abnormal stomach accommodation and visceral hypersensitivity rather than true emptying delay (Debourdeau et al. 2024), deconditioning or the general effects of severe illness, and subjective reporting bias that inflates symptom-based prevalence.
• Discounted cross-disease support. The mechanistic support from diabetic and idiopathic gastroparesis histology (Grover et al. 2011) and animal models (Wang et al. 2009) comes from other populations and does not by itself establish that ME/CFS shares that pathology.

The practical consequence: before acting on a gastroparesis diagnosis in ME/CFS, a clinician should recognise the evidence is a single old cohort and should exclude the common non-specific causes — drug effects, dyspepsia — with objective testing, rather than assuming a distinct ME/CFS-specific emptying defect. This is a risk assessment, not a positive claim, and the caution applies across all severities.

The mechanism question: three candidates, none confirmed

This is the genuinely open part of the story. We do not yet know what is structurally or functionally wrong with the stomach in ME/CFS. Three candidates are on the table, and they point to different treatments.

Candidate 1 — structural loss of the stomach’s pacemaker cells and enteric nerves. In diabetic and idiopathic gastroparesis, human gastric tissue shows a loss of the interstitial cells of Cajal (the pacemaker cells of the stomach), a reduction in enteric nerve fibres, and altered resident macrophages (Grover et al. 2011). An animal model of diabetic gastroparesis confirms that loss of these cells and nerves is accompanied by slowed emptying (Wang et al. 2009). Whether ME/CFS shares this structural degeneration is unknown — no ME/CFS gastric histology has been published. In our register, the extrapolation of this finding to ME/CFS is an open question with a certainty of 0.40. If it is true, then prokinetic drugs that stimulate surviving nerves have a built-in limit, and treatment would need to target the prevention or replacement of the cell loss instead.

Candidate 2 — autoimmune blockade of the ganglionic acetylcholine receptor. A case of autoimmune gastrointestinal dysmotility after SARS-CoV-2 infection presented with intractable nausea, early satiety, delayed gastric emptying, and antibodies against the ganglionic acetylcholine receptor — and it improved substantially with immunotherapy (Montalvo et al. 2022). By analogy, a subset of ME/CFS patients with gastroparesis could have an autoimmune ganglionic (vagal/enteric) blockade rather than structural nerve loss — a mechanism that would be potentially reversible with immunotherapy rather than irreversible. This is a speculation for ME/CFS with a certainty of 0.30: no ME/CFS cohort has been screened for these antibodies, and the only directly analogous evidence is a single post-viral (Long COVID) case report.

Candidate 3 — functional vagal failure alone. The stomach’s motor program is under-driven by the autonomic nervous system, with no structural cell loss and no autoantibody. This is the simplest reading and the one the clinical picture fits, but it is a diagnosis of exclusion: it holds only if candidates 1 and 2 are ruled out. In our register it is a clinical reading rather than an independently evidenced finding, so we give it no separate certainty value.

These three are not mutually exclusive, and the distinction matters clinically. A structural loss limits what a prokinetic can do. An autoimmune blockade is potentially reversible. A functional failure is the one most likely to respond to prokinetic drugs. The honest position is that we do not yet know which one — or which combination — applies to a given patient, and no test currently separates them in ME/CFS.

Each candidate has a falsifiable prediction. A gastric antral biopsy from ME/CFS patients with documented gastroparesis showing normal pacemaker-cell and enteric-nerve density would falsify the structural-loss reading; evidence of cell or nerve loss would support it. Measuring ganglionic acetylcholine-receptor antibodies in a ME/CFS gastroparesis cohort — if a meaningful subset is antibody-positive and their emptying improves after immunotherapy, the autoimmune-blockade claim is supported; if antibody levels do not track emptying delay or immunotherapy response, it is falsified. Candidate 3 is testable too: if emptying delay tracks non-invasive vagal-tone markers (such as heart-rate-variability indices) and improves with autonomic therapies that raise vagal drive, the functional-failure reading is supported; if emptying stays slow while those markers normalise, it is weakened. The emptying delay itself — the shared feature behind all three — can be measured with a standard gastric emptying scan; telling the candidates apart would need the biopsy and antibody testing above, alongside non-invasive autonomic measures, and none of that has been done in an ME/CFS cohort.

A cross-disease mirror

Gastroparesis is not unique to ME/CFS. It is well-characterised in other settings, and the parallels are informative:

• Diabetes. Gastroparesis is a well-characterised complication of diabetes mellitus, where the structural loss of pacemaker cells and enteric nerves is documented in human tissue (Grover et al. 2011) and in animal models (Wang et al. 2009).
• Post-viral (Long COVID). Autoimmune gastrointestinal dysmotility has been reported after SARS-CoV-2 infection, presenting with intractable nausea, early satiety, delayed gastric emptying, and antibodies against ganglionic acetylcholine receptors, with improvement after immunotherapy (Montalvo et al. 2022).
• POTS and hypermobility. In the autonomic/hypermobility overlap — POTS and hypermobile Ehlers–Danlos syndrome — GI dysmotility and gastroparesis are common and can require non-oral nutritional support in severe cases Aziz et al. (2025).

These parallels support gastroparesis as a plausible manifestation of the autonomic and autoimmune dysfunction implicated in ME/CFS, and as a source of nutritional risk in severe illness. They also give clinicians a management playbook — the POTS/hypermobility framework is the most transferable — while the cross-population gap is flagged: what is established in those populations is not yet proven in ME/CFS.

What to do about it: a safe, ordered pathway

The treatment story has two rungs, and the order matters. The lowest-risk step comes first.

First-line: dietary and nutritional management. For moderate-to-severe gastroparesis, dietary modification and feeding strategy are the first-line, lowest-risk interventions: small frequent meals, low-fat and low-fibre meals to reduce the mechanical load on the stomach, and liquid or semi-liquid nutrient options when solid emptying is slow (Gupta and Lee 2016). Not all liquids are equal: thin broths and soups are low in calories, so a shift to liquids needs guidance on energy and protein targets, not just a switch of texture. In the autonomic/POTS population, GI dysmotility with feeding intolerance is a recognised driver of the need for non-oral (enteral or parenteral) nutritional support in the most affected patients (Tseng et al. 2019). These measures do not reverse the underlying autonomic dysfunction, but they are essential to maintain nutrition and prevent avoidable weight loss in severe and very-severe ME/CFS. They apply across severity; the enteral escalation is a very-severe, feeding-failure measure. This first-line rung is also where a medication review belongs — including, where relevant, a GLP-1 receptor agonist (see below).

This is the rung the patient’s own instinct usually reaches first — the shift to small, frequent, liquid meals described above. The evidence confirms the instinct and adds the safety net: a clinician watching weight, nutrition, and the point at which oral feeding is no longer enough.

Second-line: prokinetics, safety-managed. Prokinetic drugs are the main pharmacological option for gastroparesis and are evidence-based in the general gastroparesis population (Ingrosso et al. 2023). They carry serious risks that are especially relevant to severe and very-severe ME/CFS patients, and none is validated in ME/CFS; nor is any herbal or over-the-counter “motility” supplement, which can interact with cardiac and other medications — do not self-prescribe supplements for this. The prescription drugs considered are:

• Metoclopramide is the only FDA-approved drug for gastroparesis, but it carries a black-box warning for use beyond 12 weeks because of tardive dyskinesia, which may be irreversible; it also causes drowsiness, restlessness, and hyperprolactinaemia (Shakhatreh et al. 2019). In pregnancy it has reassuring first-trimester data on major congenital malformations (Sun et al. 2021), but it still needs specialist review in pregnancy or lactation.
• Domperidone can prolong the cardiac QT interval — a baseline ECG/QT assessment and QTc monitoring are prudent before and during use (Ingrosso et al. 2023); in Europe it is restricted to short-term use on these cardiac grounds (EMA, 2014) (Yüksel and Tuǧlular 2019).
• Erythromycin (as a motilin agonist) loses effectiveness with prolonged use through receptor desensitisation (Ingrosso et al. 2023).
• Highly selective 5-HT4 agonists (for example prucalopride) show efficacy with no excess pooled cardiovascular/QT signal in trials (Patel et al. 2024), but they are not approved for gastroparesis in all jurisdictions.

One class to check, not a prokinetic option: GLP-1 receptor agonists are contraindicated in established gastroparesis. They slow gastric emptying (Nauck et al. 2011) and are a recognised cause of medication-induced gastroparesis (Camilleri and Jencks 2025); they also suppress appetite and can worsen weight loss and nausea — all of which are already problematic in severe ME/CFS. If a patient on a GLP-1 has slow emptying, it is worth asking the prescriber whether the delay began after the drug started; if it clearly did, a clinician may consider a supervised trial hold or discontinuation, which in some cases resolves the picture before any prokinetic is considered. That is a clinical judgement, not established guidance, and the timing question is for the prescriber, not something to decide alone. Do not stop a prescribed GLP-1 on your own. Their use in a patient with gastroparesis requires explicit review against this risk.

Two hard limits apply. First, no prokinetic has been tested in a controlled ME/CFS trial, so dosing and safety must be extrapolated from other populations. Second, prokinetics do not treat the underlying autonomic dysfunction; they are best combined with the dietary measures above. For a severe or very-severe patient, the cautious approach is an evidence-based trial with the lowest-risk agent, close monitoring for extrapyramidal, cardiac, and autonomic effects, and explicit time-limited use. A full drug-interaction review against the individual patient’s medication list is required before use; if that review, or a pregnancy or lactation check, is positive, do not initiate a prokinetic without specialist involvement.

A watch-and-escalate ladder. In this article, “severe” has a concrete operational meaning: unintentional weight loss over weeks, an inability to hold down food or fluids, or reliance on liquids alone to keep weight stable — and in ME/CFS terms this often coincides with being bed-bound or largely unable to manage oral intake. These are our working thresholds, not published triage criteria. A patient already living on liquids and still losing weight is not approaching the escalation point; they are already at the point where a review of non-oral feeding belongs on the table. These are the signals that move a case up the ladder, from the dietary measures above to liquid or semi-liquid nutrition and then, if weight and nutrition keep falling despite that, to non-oral feeding under medical supervision. These are thresholds a clinician should watch and respond to — not lines the patient is expected to draw alone. Three cautions follow. First, escalation past diet — non-oral feeding, and in a severe patient starting any prokinetic — is usually a specialist (gastroenterology or nutrition-team) decision. Second, some signals warrant urgent attention rather than a slow step up: weight loss continuing week over week, an inability to keep fluids down for roughly a day, or faintness or dizziness that is new or worse and tied to fluids not staying down, rather than your usual orthostatic symptoms. Third, seek urgent care rather than waiting on any of the above for vomiting blood or suspected aspiration. In practice, laboratory checks for electrolyte and micronutrient balance support the decision at each step.

This is not a recommendation to take any of these drugs. Medication decisions require a qualified clinician who can weigh the individual patient’s severity, comorbidities, and current medication list.

How to bring this up with your clinician

If this article matches what you feel, the finding is testable. You can ask your clinician whether a gastric emptying study is worth considering given how thin the ME/CFS evidence is, and mention that you would like the result read against a standard protocol. The study can confirm whether emptying is delayed and rule out a normal stomach, but it does not by itself identify the mechanism — it is one useful test, not the whole answer. And if you are losing weight, eating only liquids, or vomiting regularly, say so plainly: those are the signals that make an emptying study and a nutritional assessment worth doing, not optional extras. It is the clinician’s role to decide whether or how to act on those signals, not yours alone. If an emptying study is not immediately available, the safest first-line steps here — dietary measures and a medication review — are still the right place to start, so you are not blocked by waiting for the test.

What to take away

Gastroparesis in ME/CFS is real and testable. Delayed gastric emptying is a documented, measurable feature of the illness — not a vague complaint — and it can be assessed with a standard gastric emptying test (Burnet and Chatterton 2004).

The mechanism is genuinely unresolved. Three candidates are on the table — structural loss of the stomach’s pacemaker cells and enteric nerves (open question, 0.40), autoimmune ganglionic blockade (speculation, 0.30), and functional vagal failure alone — and no test currently separates them in ME/CFS. This uncertainty is honest, and it is not a barrier to care.

There is a safe treatment ladder. Dietary and nutritional management first — small, frequent, low-fat, low-fibre meals and liquid options, with enteral support reserved for feeding failure. Prokinetics second, safety-managed, time-limited, and only after a clinician has weighed the individual risks.

The single most actionable conclusion for a severe case: start with the lowest-risk measures — diet and a medication review — regardless of the scan, and treat the measurable problem objectively rather than assuming a specific lesion until it is demonstrated. The emptying study is the confirmatory test to obtain where available, not a precondition for beginning the safest first-line steps.

This article reflects research with explicit, calibrated certainty — a documented clinical finding (0.60), an open mechanism question (0.40), and a registered speculation (0.30) — not established clinical fact. Discuss any medical decision, including any change to diet or medication, with a qualified clinician.

For the broader GI-dysmotility and SIBO framing, see GI Dysmotility Article 1 and the companion treatment article.

For the comprehensive, fully-cited picture of how gastroparesis is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

What we know, what our research adds

Every article in this series distinguishes two things clearly:

1. What we know — established scientific findings, cited to the peer-reviewed literature (the mitochondrial evidence in autism (Frye et al. 2024), the BH4 finding (Colpani Filho et al. 2025), the maternal-antibody evidence (Braunschweig et al. 2013), and the cerebellar glutamatergic genetic signal (Maccallini 2026)).
2. What our research adds — the hypotheses our documentation project developed by reading across disease boundaries, and which are not yet established: the brain-energy convergence model, the multi-pathway treatment hypothesis, the neuroimmune-encephalopathy-spectrum framework, and the acquired-vs-developmental distinction.

The distinction matters because the two are not the same confidence. “What we know” is evidence; “what our research adds” is testable hypothesis. The series keeps them separate so you are never misled into treating a hypothesis as a finding.

Each article follows the same shape:

1. What it is — the mechanism, in ordinary words.
2. The evidence — what supports it, and how strong that evidence is.
3. The honest limits — what it does not claim.
4. The decisive experiment — how the idea could be confirmed or refuted.

Part 1: Autism as a Brain Energy Disorder — and the Multi-Pathway Treatment Hypothesis

The series opens with the most developed strand: a brain running on a thin fuel supply. When the brain cannot afford its most energy-expensive operations — filtering noise, switching tasks, reading faces — it economizes, and those economized functions are exactly the ones we recognize as core to autism. Part 1 also carries the series’ synthesis: the multi-pathway treatment hypothesis, and what treating several mechanisms at once could mean.

Read the article:

1. Autism as a Brain Energy Disorder: What the Science Says — the astrocyte/GLUT1 model, the mitochondrial and BH4 evidence, the modifiable targets, the honest limits, the CSF:glucose decisive experiment, and the multi-pathway treatment hypothesis.

Part 2: The Immune-Mediated Subset of Autism

Not all autism may have the same cause. A subset may be driven by the immune system — maternal antibodies targeting fetal brain proteins during development, or infection-triggered anti-neuronal responses after birth. The key finding: immunotherapy works in antibody-stratified patients but not in unselected ones. The subset matters because it is potentially identifiable and treatable.

Read the article:

1. The Immune-Mediated Subset of Autism: When the Immune System Shapes the Brain — the maternal-antibody route, the postnatal route, the IVIG stratification evidence, the neuroimmune-encephalopathy-spectrum framework, and the decisive multi-disease antibody study.

Part 3: The Cerebellar-Glutamate Connection

A third strand is neither a fuel problem nor an immune problem: a wiring-level difference in the cerebellum and in glutamatergic circuits — the brain’s excitatory pathways. Cerebellar Purkinje cell loss is among the most replicated findings in autism, and a genetic signal links it to sensory sensitivity. This strand is explanatory and diagnostic, not yet a treatment target.

Read the article:

1. The Cerebellar-Glutamate Connection in Autism: A Wiring-Level Hypothesis — the Purkinje-cell evidence, the glutamatergic genetic signal, the shared features, the testable sensory-sensitivity prediction, and the honest limits.

Part 4: Acquired vs. Developmental — When Autism-Like Features Might Be Reversible

The clinically consequential question: are autism-like features necessarily present from development, or can some be acquired — and therefore reversible? Sensory rigidity, withdrawal, and apparent alexithymia can arise after development rather than from it. Telling the two apart determines whether a symptom might respond to treating its cause or is stable wiring.

Read the article:

1. Acquired vs. Developmental: When Autism-Like Features Might Be Reversible — the interoceptive mechanism, the neuroinflammation route, the context-dependence test, and the clinical consequences of getting the distinction wrong.

A note on honesty

Every article in this series distinguishes carefully what is confirmed by evidence, what is a promising hypothesis, and what is an individual patient’s experience. Most of the content here is at the hypothesis end — the certainties are explicit, often low, and nothing is presented as established clinical fact or a treatment recommendation.

The underlying science is drawn from our documentation project, which develops these cross-disease energy and immune models in detail. The autism-specific framing is our reading of the literature, written to stand alone for anyone curious about the biology.

Discuss any medical decision with a qualified clinician.

The feedback loops that connect the conditions

The eight co-occurring conditions covered in this series are not independent. Every pair has at least one documented — or mechanistically plausible — feedback loop connecting them. The most important loops, the ones that make the machine run, are these:

The amplification ratchet: why longer illness responds less

The primary document develops a specific and sobering hypothesis: each cycle of activation — a mast-cell flare, a POTS episode, a PEM crash — may leave behind a little more structural change. Extra nerve endings (sprouting). More mast cells (proliferation). Lower activation thresholds (receptor sensitisation). Degraded extracellular matrix (MMP-mediated). Receptor internalisation (GPCR autoantibodies).

The amplification ratchet is the idea that the system, once pushed past a tipping point, does not spontaneously return to baseline. Each event ratchets the system one notch further from the healthy state. The mast-cell-ECM bistable model formalises this: two stable states exist — healthy (firm matrix, quiet mast cells) and degraded (soft matrix, hyper-reactive mast cells). Once the system crosses the tipping point from healthy to degraded, returning requires sustained, multi-target intervention — and for some patients, the structural damage may be past the point of any reversibility (Loth 2026).

This has a testable — and falsifiable — prediction (model-derived and unvalidated, so framed as a hypothesis): treatment response should drop measurably with each additional 5 years of illness duration, and this effect should exceed the effect of baseline blood markers. The “5 years” granularity is a modelling convention rather than a measured threshold. If treatment response is identical regardless of illness duration, the ratchet hypothesis loses support.

The clinical implication, if the ratchet is real, is not that long-ill patients should not be treated — it is that treatment expectations should be calibrated to illness duration, and that the time to treat is as early as possible, before the structural changes accumulate.

The clinical imperative: multi-target treatment

The practical consequence of the feedback loops and the amplification ratchet is that single-target treatment is unlikely to be sufficient in established disease. This is the lesson that runs through every article in this series: treating the mast cells without stabilising the POTS, treating the POTS without addressing the gut, treating the gut without calming the mast cells — each can produce a partial response, but none, alone, is likely to break the system.

The ODE models in the primary document predict that a minimum of 4–6 simultaneous drug targets is needed for structural controllability of the interconnected system. This is not polypharmacy for its own sake. It is the mathematical consequence of a system where every node receives input from multiple others — you cannot control a network by controlling one node. One caveat the source files carry and this capstone must not drop: the “4–6” count inherits the unvalidated status of the network topology it is computed on — “structural controllability” is a property of the assumed model, not a dose-finding result, and if the assumed couplings are wrong the number changes.

These two statements reconcile rather than contradict: the model describes the endpoint — a maintenance regimen with several active targets — and the structured trial describes how to reach it. You do not start all targets at once. You add them one at a time, so that when you do reach a simultaneous multi-target regimen, you know what each drug contributes and can drop any that does nothing. The sequence is: trial A → if partial response, maintain A and trial B → if partial response, maintain A+B and trial C → until the symptom burden is acceptable or the drug burden becomes unacceptable.

A necessary safety note before anyone aims at several targets at once. Polypharmacy is where drug interactions concentrate, and the pieces flagged individually across this series need to be assembled before a multi-drug regimen is built: sedation stacking (e.g., a gabapentinoid plus amitriptyline plus ketotifen), orthostatic stacking (e.g., midodrine/fludrocortisone plus an anticholinergic TCA), QT/cardiac stacking (e.g., amitriptyline plus erythromycin plus cimetidine), and CYP3A4 collisions (e.g., cimetidine with ivabradine). Because the series tells patients a combination may be needed, it is obligatory that the interaction and dose review — ideally with one clinician holding the full medication list — happen before combining drugs, not after. A target count is a model estimate, not a license to stack drugs without a consolidated interaction check.

This is slow. It requires patience. It means accepting partial wins — the flushing is better but the fatigue is not, the POTS is controlled but the pain persists — and documenting them honestly, rather than concluding “nothing works” because no single drug fixed everything.

What this series has argued, in one place

1. The conditions that travel with ME/CFS are most often later-arriving amplifiers, not the originating cause. MCAS is not best read as the hidden cause of ME/CFS — in the large two-cohort study, the great majority of affected patients were detected for MCAS only after the illness began, with MCA prevalence rising over the disease course rather than preceding onset (Rohrhofer et al. 2025). Two caveats from the MCAS article carry here: “detected after” is not necessarily “began after” (people are screened after ME/CFS brings them into specialist care, and onset-ordering is retrospective), so this weakens rather than kills a causal reading; and whether MCAS is a genuine amplifier or in part a bystander that co-exists is not yet settled. POTS is likewise best read as a downstream consequence of low blood volume, endothelial dysfunction, and connective-tissue laxity rather than the originating cause. hEDS is best read as a permissive substrate that makes other amplifiers louder. The core energy-immune defect, if it is upstream, is not yet independently demonstrated — the specialist articles flag the causal arrows as not drawn.

2. But calling something an “amplifier” does not mean it is not worth treating. An amplifier that worsens the illness is worth treating. The distinction is between treating the cause — which would cure — and treating the amplifier — which reduces the total illness burden. Most treatment in ME/CFS is amplifier management. That is not a failure. It is accurate medicine for a disease where the cause is not yet understood and the amplifiers are what patients experience every day.

3. The feedback loops mean that treating one amplifier sometimes treats several. Stabilising mast cells may reduce POTS symptoms (less histamine = less vasodilation). Treating SIBO may reduce brain fog (less histamine = less H3-receptor suppression). Addressing the gut may reduce the MCAS burden (more butyrate = more mast-cell stabilisation). The loops are bidirectional, and a treatment that targets one node can have downstream effects on others — which is why the structured-trial architecture uses one drug at a time.

4. The amplification ratchet means that early treatment may matter more than aggressive treatment. If each cycle of activation leaves behind structural change, then preventing the cycles — through pacing, through early amplifier management, through radical energy conservation in the first years of illness — may be more effective than aggressive pharmacological intervention after decades of accumulated damage. This is a hypothesis, not a clinical fact. But it is a hypothesis that, if correct, has different implications than the current “wait and see” approach that characterises most ME/CFS care.

5. A treatment failure in one amplifier says nothing against the others. Failed MCAS treatment does not mean the POTS is not real. Failed POTS treatment does not mean the SFN is psychosomatic. The conditions are mechanically coupled but pharmacologically separable — a drug that works for one mechanism can fail while the overall system remains genuinely pathological. Distinguishing “this drug did not help” from “this condition is not real” is the single most important honest correction this series can offer.

The bottom line

The eight co-occurring conditions covered in this series — MCAS, POTS, hEDS, SFN, fibromyalgia-pain, GI dysmotility/SIBO, chronic infection, and autoimmunity — are not a coincidence. They are a system. The feedback loops that connect them — vascular-inflammation, nerve-mast-cell, gut-brain-histamine, infection-mast-cell-MMP-9, antibody-amplifier — mean that treating one condition in isolation is unlikely to succeed, and that the clinical art is in sequencing trials across the system, accepting partial wins, and not mistaking a single drug failure for a verdict on the illness (Loth 2026).

The amplification ratchet hypothesis — that longer illness is harder to treat because each cycle leaves behind structural change — is both sobering and actionable. Sobering, because it means some damage may be irreversible. Actionable, because it means the time to treat is now, not later — and that aggressive pacing and early amplifier management may prevent the ratchet from advancing further.

The series closes where it opened: the distinction between a co-occurring condition on its own and the same condition on top of ME/CFS. The conditions themselves are real. But on top of ME/CFS, they are amplifiers — and treating them is amplifier management, not curative medicine. Accepting that distinction is not giving up. It is accurate targeting in a complex system, with the honest knowledge that the tools we have are imperfect, the evidence is incomplete, and the patient — not the theory — is the final arbiter of what works.

For the comprehensive, fully-cited picture of how the Septad-plus-fibromyalgia conditions interact, the feedback-loop models, and the amplification ratchet, see (Loth 2026).

First, a plain warning

Everything below describes hospital-based interventions with significant risks: immunoadsorption requires central venous access and carries infection, thrombosis, and hypotension risks. IVIG causes aseptic meningitis, thromboembolism, and anaphylaxis. Rituximab carries a risk of progressive multifocal leukoencephalopathy. Daratumumab causes infusion reactions and immunosuppression. These are treatments for the severely affected, discussed with an immunologist — not self-directed and not first-line.

The treatment evidence, from most to least mature

Expected results

• The best-case treatment response is partial. Even in the positive open-label series, “responder” meant clinically significant improvement — not recovery, not return to work, not normalisation of PEM. The treatments reduce the autoimmune contribution; they do not cure ME/CFS.
• IVIG response may take 3–6 months. The combination of antibody dilution and immunomodulatory Fc-glycan effects is gradual. A single cycle is not an adequate trial, but a null result at 6 months of adequate dosing is sufficient to stop.
• Immunoadsorption works biochemically — IgG drops by 80% — but this does not guarantee clinical improvement. The Anft 2025 discordance (AAbs gone, symptoms unchanged) is the cautionary data point. IA tells you something about the antibody; the clinical response tells you something about whether the antibody was pathogenic.
• Daratumumab responds with an 8–9 month latency, if the signal is real. This is the longest latency of any immune-targeted treatment, and it means a 3–6 month trial may be falsely negative. A 12-month trial is the minimum to judge the daratumumab signal — and such a trial has not been done.
• Treatment decisions in 2026 are made under genuine uncertainty. The IA-PACS-CFS result has been reported preliminarily (a May 2026 conference report suggested no significant benefit over sham) and the EXTINCT controlled result is still awaited (Rücker 2026). The glycoprofile hypothesis is mechanistically elegant but clinically untested. The honest approach is a time-limited, endpoint-defined trial — and the discipline to accept the answer it gives.

An example structured trial

This is an example to give your clinician something to work from — not a self-prescription.

The pre-requisite is documented GPCR autoantibodies on a functional assay (bioassay, not commercial ELISA), a post-infectious onset, and a POTS-dominant autonomic profile. If these are not met, the pre-test probability of a dominantly autoimmune mechanism is low and these treatments are unlikely to be appropriate.

A practical caveat about the pre-requisite: the functional GPCR bioassay is mostly available in research settings, not routine clinical labs — so for most patients the antibody gate cannot actually be met before a trial can even be discussed. Where a functional assay is unobtainable, the practical pre-requisite collapses to the clinical picture alone: clear post-infectious onset plus a POTS-dominant autonomic profile with documented orthostatic symptoms. That is a weaker gate than the antibody test, and it should be read as such — it lowers confidence that the patient is antibody-positive but does not make the treatment inappropriate to raise with a specialist.

If the profile is right, start with the least invasive option: IVIG 0.4 g/kg daily for 5 days, then 0.4 g/kg every 3–4 weeks. Pick a single target symptom — the autonomic symptom that limits you most (standing time, presyncope frequency, heart-rate variability on orthostatic challenge). Reassess at 6 months. Stop if no improvement — a 6-month null IVIG trial is sufficient evidence that immunoglobulin therapy is not the lever. Be aware that IVIG is expensive (often several thousand dollars per infusion cycle) and, being off-label in ME/CFS, is frequently subject to prior authorisation and often denied — so access, not just clinical fit, is a real barrier that should be discussed up front.

If IVIG produces partial improvement but AAbs are still elevated, and the patient is severely affected (housebound or bedbound), discuss immunoadsorption with a specialist centre. A 5-session IA course (10–14 days) should reduce IgG by ~80%. Reassess at 1 month post-treatment. If symptoms have not improved despite adequate IgG reduction, the “antibodies are the dominant pathology” reading is unsupported — this could mean the antibodies were not pathogenic, that the damage they caused is irreversible, or that they were not the main driver; the data as this series discusses do not cleanly separate those. Either way the actionable clinical conclusion is the same: stop pursuing antibody removal and shift to downstream-amplifier management.

Only escalate to daratumumab (where available and under specialist supervision) if IVIG and IA have both failed in the AAb-positive post-infectious subgroup — and only within a controlled research protocol, given the early evidence stage and the on-target NK-cell depletion. Daratumumab is an expensive oncology biologic with no approved ME/CFS indication; in practice it is effectively confined to specialist/research centres that can obtain it, so for most patients it is not a realistically reachable option.

What it means if the treatment does not help

One: a failed immunoadsorption trial is informative. If IgG was reduced by 80% and symptoms did not improve, either the antibodies were not pathogenic (Fc glycoprofile mismatch), the pathology they caused is irreversible (receptor internalisation), or the antibody rebound was too fast. In the first case, further antibody-targeted treatment is unlikely to help. In the second and third, it might — but with a different strategy (maintenance with daratumumab, not pulse with IA). Two: a failed IVIG trial says IVIG does not work for you — nothing more. It does not rule out an autoimmune mechanism. IVIG is immunomodulatory, not antibody-depleting, and the dose, brand, and infusion rate all affect tolerability and possibly efficacy. But if IVIG at standard doses over 6 months produces no benefit, continuing indefinitely is neither evidence-based nor risk-free. Three: the RituxME null was a trial failure, not a hypothesis failure. Rituximab spares plasma cells. The plasma-cell factory model — supported by the daratumumab signal and the 8–9 month latency — is still on the table. The lesson of RituxME is not “autoimmunity is not the mechanism.” It is “do not deplete B-cells and expect to deplete antibodies.” Four: a sequence of failures — immunoadsorption removes the antibodies but symptoms persist, IVIG produces no benefit, daratumumab is unavailable — is genuine evidence that autoimmunity is not the dominant pathology, or that it was dominant once but the damage it caused is now self-sustaining. Either way, it is useful information. Stop pursuing antibody-targeted treatments and focus on the downstream amplifiers (POTS, MCAS, SFN) that the autoimmunity may have triggered.

The falsifiable predictions

• If the glycoprofile predicts IA response — G0F-dominant (pro-inflammatory) profiles respond; sialylated-dominant profiles do not — the Fc-mechanism model is clinically actionable and should be added to pre-treatment testing (Loth 2026).
• If daratumumab responders show a decline in pathogenic antibody titre over 6–9 months with clinical improvement lagging, the plasma-cell-factory model is supported and the latency is a real biological phenomenon, not a placebo effect.
• If IA-PACS-CFS and EXTINCT (Hannover, NCT05954325, n=63) both return null in antibody-positive subgroups, the autoimmune hypothesis “takes a blow comparable to the rituximab-to-RituxME collapse” — the primary document’s own framing.
• If passive transfer of ME/CFS IgG reproduces autonomic dysfunction in mice — the experiment has been done in fibromyalgia and Long COVID but never in ME/CFS — the autoimmune mechanism moves from association to causation.

What you can actually do

• Get the right test, but read it critically. Functional GPCR autoantibody testing (bioassay, not ELISA) is the most meaningful — but it is available only in research settings. If you are offered commercial CellTrend ELISA testing, know that the 2022 replication study found 100% of healthy controls α1-AAb positive, and the test’s clinical utility is disputed. A positive result is ambiguous. A negative result on a high-quality functional assay is more informative than a positive result on a non-specific ELISA.
• The most honest predictor of treatment response is not the antibody titre — it is whether you have a clear post-infectious onset and a POTS-dominant autonomic profile. The β2-AAb signal correlates with autonomic severity; the post-infectious history places you in the subgroup where molecular mimicry is the plausible trigger. If your ME/CFS was gradual, non-post-infectious, and pain-dominant rather than POTS-dominant, the pre-test probability of an autoimmune mechanism being the dominant driver is lower.
• Immunoadsorption is a specialist decision for the severely affected with documented GPCR autoantibodies and a POTS-dominant profile. It is not a first-line treatment and not a generic recommendation. It requires a centre with IA experience and a prescriber who understands the evidence limitations — including the pending IA-PACS-CFS result and the Anft 2025 discordance. It is also among the most expensive and least available interventions described in this series (specialist centre, central venous access, a ~5-session course), so realistic access should be assumed to be a barrier in most systems.
• IVIG is a higher-risk, lower-evidence option than its patient-community reputation suggests. The PANDAS IVIG null is a cautionary parallel. The Eckey 2025 survey data are encouraging but uncontrolled. A 6-month IVIG trial with defined endpoints and a stopping rule is a rational conversation with an immunologist if GPCR autoantibodies are documented and less invasive options have failed — but indefinite IVIG without evidence of benefit is not good medicine. Standard pre-IVIG safety checks — serum IgA level (to exclude IgA deficiency, where anaphylaxis risk is higher), renal function, and a thromboembolic-risk assessment — are part of a responsible workup and should be discussed with the prescriber.
• Accept that the field is unsettled. The daratumumab signal is real but uncontrolled. The IA-PACS-CFS result is now being weighed in its preliminary form (suggesting no clear benefit over sham) and the EXTINCT result is still pending (Rücker 2026). The glycoprofile hypothesis is mechanistically elegant but clinically untested. A treatment decision made in 2026 is made under genuine uncertainty — and the honest clinical approach is to acknowledge that, make a time-limited trial with a stopping rule, and accept the answer the trial gives, whichever direction it points.

The bottom line

Autoimmune-targeted treatments for ME/CFS exist — immunoadsorption, IVIG, rituximab, daratumumab — but the evidence is uneven, the open-label signals exceed the controlled results in a pattern that RituxME made impossible to ignore, and the measurement tools that would identify who will respond (functional bioassays, Fc glycoprofiles) are not clinically available. The daratumumab signal — 60% marked improvement in an uncontrolled pilot — is the most encouraging recent development, but the RituxME precedent demands controlled replication before it changes practice [Fluge et al. (2025)](Loth 2026).

The honest clinical approach is a time-limited, endpoint-defined trial in the antibody-positive, post-infectious, POTS-dominant subgroup — and the discipline to stop if the trial fails and shift attention to the downstream amplifiers that the autoimmunity, if present, had activated.

Next in this series: “Chained together” — the synthesis capstone. How the amplifiers become one disease, and what that means for treatment.

For the comprehensive, fully-cited picture of how autoimmunity is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

First, a plain warning

This is an explanation, not self-medication advice, and I am not a doctor. Autoantibody testing for ME/CFS is not standardised — CellTrend ELISAs, the most commonly used commercial assay, have been shown to produce positive results in 98–100% of healthy controls in a replication study, calling their diagnostic value into question. Immunoadsorption, IVIG, rituximab, and daratumumab are hospital-based interventions with significant risks including infection, anaphylaxis, thromboembolism, and — in the case of rituximab — progressive multifocal leukoencephalopathy. These are treatments for the severely affected, discussed with an immunologist, not self-directed.

The short version, if you only read one part

• GPCR autoantibodies target G-protein-coupled receptors — β1/β2-adrenergic (blood vessels, heart rate), M1/M3/M4 muscarinic (gut, pupils, cognition), and α1-adrenergic (vascular tone). They are found at widely varying rates (from ~29% to nearly everyone) depending on the study, the assay, and the threshold [Loebel et al. (2016)](Bynke et al. 2020).
• The fundamental measurement problem is unresolved. The most common commercial assay (CellTrend ELISA) produced positive results in 100% of healthy controls in one independent replication — meaning the test cannot distinguish patient from control. The highest-quality assay (REAP proteome-wide screen, 7,542 antigen interactions) found zero significant autoantibody signals in ME/CFS — a complete null (Germain et al. 2025).
• Antibody titre is a poor predictor of pathogenicity. Two people with the same β2-AAb level can have completely different disease severity because what matters is not the quantity of antibody but its Fc glycoprofile — the sugar molecules attached to the antibody’s tail that determine whether it activates complement, engages Fc receptors, or sits inert. An agalactosylated (G0F) IgG is pro-inflammatory; a sialylated IgG is anti-inflammatory. Same titre, opposite effects (Loth 2026).
• Passive-transfer evidence from fibromyalgia and Long COVID shows that injecting patient IgG into mice reproduces pain and small-fibre neuropathy — strong evidence that the antibodies are sufficient to reproduce a phenotype (i.e., can be pathogenic, not merely correlative) in those conditions [Goebel et al. (2021)](Mignolet et al. 2026). No ME/CFS-specific passive-transfer study exists yet.
• Autoimmunity is usually post-infectious, not spontaneous. GPCR autoantibodies are generated during infection through molecular mimicry (viral proteins resemble self-receptors) or bystander activation (the immune response to the virus spills over to self-antigens). The antibodies persist after the infection clears — and whether they continue to cause damage depends on the Fc glycoprofile and the host’s ability to clear them.

What are GPCR autoantibodies, for real?

G-protein-coupled receptors (GPCRs) are the largest family of cell-surface receptors in the human body. They sit on the outer membrane of cells and transmit external signals — hormones, neurotransmitters, light, odours — to the inside of the cell. Beta-adrenergic receptors (β1, β2, β3) respond to adrenaline and noradrenaline, controlling heart rate, vascular tone, and bronchial dilation. Muscarinic acetylcholine receptors (M1–M5) respond to acetylcholine, controlling gut motility, pupil constriction, salivation, and — in the brain — cognition and alertness.

Autoantibodies against GPCRs can do three things:

1. Agonistic activation — the antibody mimics the natural ligand, turning the receptor on. An agonistic β2-AAb causes the same effect as adrenaline: increased heart rate, vasodilation. If sustained, receptor desensitisation and internalisation follow — the cell removes the receptor from the surface to protect itself, and the patient becomes functionally deficient in that signalling pathway even though the antibody is still present.

2. Antagonistic blockade — the antibody blocks the natural ligand from binding, turning the receptor off. An antagonistic M3-AAb inhibits gut motility and salivation, mimicking the effects of anticholinergic drugs.

3. Receptor internalisation without activation — the antibody cross-links receptors and triggers β-arrestin-mediated endocytosis, removing the receptor from the surface without activating it. This is functionally equivalent to blockade, but the receptor is physically removed rather than just occupied. This mechanism, demonstrated for NMDAR antibodies in autoimmune encephalitis and extended to GPCRs in the primary document, explains why some autoantibodies cause functional deficits without any measurable agonistic or antagonistic activity (Kim et al. 2026).

The clinical effect depends on which receptors are targeted, in which tissues, with which functional effect, and in which Fc-glycoprofile state. This combinatorial complexity — receptor type × tissue location × functional mode × glycoprofile — is why a single “GPCR autoantibody panel” does not predict symptoms, treatment response, or prognosis.

What the wide prevalence range actually means

The reported prevalence for GPCR autoantibodies in ME/CFS spans a very wide range depending on assay — from about 29% on a functional bioassay to the point where a commercial ELISA reports almost everyone as positive, including healthy controls. This is not biological variability — it is measurement chaos.

• Loebel 2016 (n=268 ME/CFS, n=108 controls) used a bioassay measuring functional receptor activation: 29.5% had ≥1 elevated GPCR AAb. This is the most conservative and most functionally meaningful estimate.
• Bynke 2020 (Sweden, two cohorts, n=24 plasma + n=24 CSF each) used the CellTrend ELISA platform and found significantly elevated M3/M4 autoantibody levels in ME patients versus controls (β1/β2 in one cohort), but reported no single-cutoff prevalence percentage.
• Vernino 2022 (POTS, n=116) tested the commercial CellTrend ELISA: 98.3% of patients AND 100% of controls were α1-AAb positive. The test could not distinguish patient from control — zero diagnostic value.
• Germain 2025 (n=172 participants) ran the highest-resolution platform available — REAP, screening 7,542 antibody–antigen interactions across 6,183 exoproteome proteins — and found zero significant autoantibody signals: no q-value fell below 0.68 (i.e., nothing came close to the significance threshold). Complete null.

The pattern is systematic: the more functional and specific the assay, the lower the prevalence. The less specific the assay, the higher the prevalence — until, at the limit, 100% of healthy controls test positive and the assay collapses. The implication, uncomfortable but honest, is that many of the reported GPCR autoantibody associations may be assay artefacts rather than disease-specific signals (Hall et al. 2022).

The one signal that surfaces most often across the studies is β2-adrenergic autoantibodies — though with an honest caveat. Azcue 2026 found β2-AAb elevated in ME/CFS vs post-COVID and healthy controls (F=3.15, p=0.046), and β2-AAb levels correlated with autonomic symptom severity (r=0.45, p=0.001). Both figures come from a single group and a single assay, and the p=0.046 group difference is borderline — so while β2 is the most frequently reported GPCR target in ME/CFS, it is not independently replicated, and the REAP null above means even this signal does not escape the measurement controversy (Azcue et al. 2026).

The Fc glycoprofile: why the same titre means different things

All IgG antibodies have a conserved glycosylation site at asparagine-297 in the Fc region. The sugar structure attached there — the glycoprofile — determines effector function:

• Agalactosylated (G0F) IgG — no galactose residues. This glycoform binds Fcγ receptors with high affinity, activates complement, and drives antibody-dependent cellular cytotoxicity. It is the pro-inflammatory glycoform, elevated in active rheumatoid arthritis and acute viral infection. If a patient’s β2-AAb is predominantly G0F, it is likely pathogenic.
• Sialylated IgG — terminal sialic acid residues. This glycoform engages DC-SIGN on regulatory macrophages, inducing an anti-inflammatory response. It is the anti-inflammatory glycoform, elevated during pregnancy (when autoimmune diseases often remit) and after successful IVIG therapy. If a patient’s β2-AAb is predominantly sialylated, it may be inert or even protective.
• Bisected (G0FB) IgG — intermediate; elevated in steady-state antibody responses.

The same ELISA titre of β2-AAb — let’s say 15 U/ml — could be predominantly G0F (pathogenic, driving autonomic dysfunction) or predominantly sialylated (inert, a remnant of a past infection that the immune system has already damped). Current commercial testing does not measure the glycoprofile. This means that a positive GPCR autoantibody test, without glycoprofile information, is ambiguous — it tells you the antibody is there, but not whether it is doing anything (Loth 2026).

The post-infectious trigger: molecular mimicry and bystander activation

GPCR autoantibodies are not generated spontaneously. They arise during infection through two mechanisms:

Molecular mimicry. Viral proteins share sequence homology with self-receptors. EBV EBNA-1 shares epitopes with adrenergic and muscarinic receptors. SARS-CoV-2 spike protein shares sequence similarity with β2-AR extracellular loops. The antibody response to the viral protein cross-reacts with the self-receptor — the immune system, targeting the virus, accidentally targets the host.

Bystander activation. During a vigorous antiviral immune response, B-cells that recognise self-antigens — normally kept quiescent by tolerance mechanisms — become activated by the cytokine milieu and begin producing autoantibodies. The autoantibodies were not part of the antiviral response; they were collateral damage from the immune activation.

Both mechanisms explain why GPCR autoantibodies appear after infection and can persist for months to years — and why their pathogenicity depends on whether the post-infectious immune environment shifts toward a pro-inflammatory (G0F) or anti-inflammatory (sialylated) glycoprofile.

Autoimmunity by itself vs ME/CFS with autoimmunity

Classical autoimmunity on its own (no ME/CFS). Lupus, rheumatoid arthritis, myasthenia gravis, autoimmune autonomic ganglionopathy — these are diseases where the autoantibody is the dominant pathology. Removing it (immunoadsorption, rituximab, IVIG) treats the disease. The antibody titre correlates with disease activity. The pathology is well characterised.

ME/CFS with autoantibodies (the situation this series is about). When GPCR autoantibodies sit on top of ME/CFS, three things are different:

• The autoantibodies are one contributor. Removing them may improve autonomic symptoms without fixing the energy failure, the immune dysregulation, or the cerebral hypoperfusion. The best-case treatment outcome is partial — the dysautonomia improves, the PEM does not (Loth 2026).
• Antibody titre does not equal disease severity. Two patients with the same β2-AAb level can have completely different clinical pictures because the Fc glycoprofile, the receptor compartment (plasma membrane vs internalised), the tissue distribution, and the host’s compensatory mechanisms all modulate the antibody’s effect.
• A failed antibody-removal trial says nothing against ME/CFS. If immunoadsorption removes the autoantibodies and symptoms do not improve, the autoantibodies were not the dominant pathology — they were a marker of immune activation, not a driver of the disease. This is the most honest reading of the Anft 2025 result (below) and the one that should shape expectations for any autoantibody-targeted treatment.

The bottom line

Autoimmunity — specifically GPCR autoantibodies — is the most provocative and most contested topic in the Septad. The β2-adrenergic signal is the most frequently reported GPCR finding (but rests on a single group’s borderline p=0.046 result, not independent replication); the commercial assays that report positivity in nearly everyone are not. The Fc glycoprofile determines whether an antibody is pathogenic, and no one is measuring it. The passive-transfer evidence from fibromyalgia and Long COVID provides strong evidence that patient antibodies can reproduce a phenotype in animals — but no ME/CFS-specific transfer study exists (Goebel et al. 2021).

The honest read: GPCR autoantibodies are real in a subset, their pathogenicity depends on factors not captured by current testing, and the treatments that remove them (covered in the companion article) have shown strong signals in uncontrolled studies and sobering nulls in controlled ones. The field is at the point where the mechanism is plausible but the therapeutic implications are unproven — and the distinction matters enormously for patients making treatment decisions.

Next in this mini-series: the treatments — immunoadsorption, IVIG, rituximab, daratumumab, and why the controlled-trial evidence does not match the open-label signals [see the companion article].

For the comprehensive, fully-cited picture of how autoimmunity is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

First, a plain warning

Everything below is a conversation to have with a clinician, not self-medication. Valacyclovir and valganciclovir require renal monitoring; valganciclovir is a teratogen and causes bone-marrow suppression; famciclovir requires dose adjustment in renal impairment. Cimetidine is a CYP450 inhibitor with extensive drug-drug interactions. Mast-cell stabilisers are not antivirals — they intercept the downstream immune activation, not the viral trigger. None of these drugs should be started, stopped, or changed without a prescriber.

The antiviral landscape

The cimetidine story: an H2 antihistamine that modulates immunity

Cimetidine is an H2-receptor antagonist — a heartburn drug. But unlike famotidine (the other H2 blocker commonly used in MCAS), cimetidine has additional immunological effects: it inhibits suppressor T-cell activity, enhances natural-killer-cell function, and — in a 1986 case series — was associated with improvement in a post-EBV fatigue phenotype characterised by POTS, histamine issues, and post-infectious onset (Goldstein 1986).

The honest limit. This is a 40-year-old case series with no controlled trial. The mechanism — H2 blockade on T-suppressor cells → enhanced cellular immunity → improved viral clearance — is plausible but unproven. For reference, the usual H2-blockade dose is 400 mg twice daily, taken with or without food; a trial for this phenotype is typically 4–8 weeks with a defined reassessment. Cimetidine is a potent CYP450 inhibitor (CYP1A2, 2C9, 2D6, 3A4), which means it raises blood levels of beta-blockers, many antidepressants, warfarin, antiarrhythmics, and other POTS/ME/CFS medications. In a patient already on several of these, cimetidine is a drug-interaction risk, not a benign add-on. Famotidine is the safer H2 blocker for routine MCAS treatment; cimetidine is a specific — and speculative — immunomodulatory trial in the post-infectious phenotype, and it requires a full medication review by the prescriber before use.

The mast-cell stabiliser intercept: downstream of the virus

If EBV dUTPase activates mast cells to release MMP-9, and MMP-9 degrades the blood-brain barrier, then mast-cell stabilisers (cromolyn, ketotifen) and MMP-9 inhibitors (low-dose doxycycline) intercept the pathway downstream of the virus. This is not antiviral treatment — it does not reduce viral protein production or clear the infection. It blocks the host response to the viral protein.

This matters because it bypasses the ALR problem. Mast-cell stabilisers do not need the virus to be in lytic replication — they work on the mast cell that has been activated by the viral protein, regardless of what stage the virus is in. The treatment of MCAS — H1 and H2 antihistamines, mast-cell stabilisers, and avoidance of non-specific triggers — is covered in the MCAS articles. The chronic-infection-specific point is that in the EBV-positive, MCAS-positive subset, mast-cell stabilisers may reduce MMP-9-mediated BBB disruption even if antivirals fail, because they target the host side of the virus-host interaction, not the virus side (Loth 2026).

Expected results

• The best-case antiviral response is partial. Even in the positive valganciclovir trial, “responder” meant clinically significant improvement in fatigue and cognition — not recovery, not return to work, not normalisation of PEM. Set expectations accordingly.
• A negative valacyclovir trial is expected, not anomalous. Valacyclovir targets lytic DNA replication; ALR stops before that step. A drug that fails to hit its target because its target is not engaged is a mechanism failure, not a patient failure.
• The valganciclovir signal may or may not replicate. Until an independent replication trial exists — and none is currently registered — the EVOLVE result is a single positive study in a field of nulls. Treat it as a promising signal, not an established treatment.
• Cimetidine, if it helps, helps the post-infectious POTS-MCAS phenotype. If your ME/CFS was gradual-onset, not post-infectious, or if you lack POTS and histamine symptoms, the cimetidine phenotype is a poor match, and famotidine is the safer H2 blocker.
• Mast-cell stabilisers address the host response, not the viral trigger. If EBV is driving your MCAS, stabilising mast cells may reduce symptoms without affecting the underlying viral activity. You will still have high antibody titres. The symptoms improve because the host response to the virus is damped, not because the virus is gone.

An example structured trial

This is an example to give your clinician something to work from — not a self-prescription.

Confirm that the clinical picture is post-infectious (documented EBV mono, HHV-6 roseola, or tick-borne infection preceding ME/CFS onset). Get meaningful serology — EA-D IgG, high/rising VCA IgG, and HHV-6 PCR from whole blood (VCA IgM is a marker of recent/primary infection, not reactivation). If serology suggests reactivation, start valacyclovir 1 g three times daily. Reassess at 6 months. Stop if no improvement — continuing an antiviral past 6 months without benefit is not evidence-based.

If valacyclovir fails and HHV-6 is documented (PCR positive, or ciHHV-6 known), discuss valganciclovir with a specialist. This is a toxic drug — bone-marrow suppression, teratogenicity — requiring regular CBC monitoring. A 3–6 month trial at 900 mg twice daily (taken with food, dose-adjusted for renal function) with pre-defined stopping rules and monthly CBC is the responsible approach. Stop if neutropenia develops or if there is no improvement at 6 months.

If both antivirals fail — or if ALR is suspected (high EA-D without detectable viral load) — shift to the mast-cell intercept: H1 + H2 antihistamines plus a mast-cell stabiliser (cromolyn or ketotifen), targeting the EBV→mast cell→MMP-9 pathway downstream of the virus. Reassess at 12 weeks. A response supports the host-response model; a null result supports neither the antiviral nor the mast-cell model as the dominant mechanism, and the honest move is to stop pursuing the chronic-infection pathway.

The falsifiable predictions

• If mast-cell stabilisers reduce plasma MMP-9 in EBV-reactive ME/CFS patients, but valacyclovir does not, the EBV→mast cell→MMP-9 pathway is host-mediated, not viral-replication-mediated, and the mast-cell intercept is the rational treatment target regardless of antiviral response (Loth 2026).
• If the threat-signal model is correct, antiviral response should be stronger in patients with <3 years illness duration (where the viral-load component w_V·V still dominates the threat signal) than in patients with >10 years (where host factors dominate) — and the difference should exceed the baseline antibody titre difference.
• If EA-D IgG and HHV-6 DNA predict valganciclovir response but VCA IgG and EBNA IgG (markers of past infection, not reactivation) do not, serology triage is clinically actionable and should guide antiviral prescribing rather than the current “try and see” approach.
• If cimetidine + valacyclovir produces a higher responder rate than valacyclovir alone in the post-infectious POTS-MCAS phenotype, the immunomodulatory component is real and the historical cimetidine signal deserves a controlled trial — but this prediction requires a controlled study, not individual experience.

What it means if the treatment does not help

One: a failed valacyclovir trial says valacyclovir does not work for you — nothing more. It does not say EBV is inactive. It does not say ALR is not happening. It says a drug that blocks lytic DNA replication, at the dose you took, for the duration you took it, did not produce a measurable benefit. That is useful pharmacology. It is not a verdict on the viral hypothesis. Two: a failed valganciclovir trial narrows the possibilities more. Valganciclovir covers the poly-herpesvirus spectrum. If a 6-month trial at an adequate dose produces no benefit, the probability that an antiviral — any currently available antiviral — will help you drops substantially. That is useful information, and the honest next step is to shift to mast-cell stabilisation (intercepting the downstream immune activation) and away from further antiviral trials. Three: the threat-signal model from the primary document’s ODE framework predicts that antivirals work only when the viral-load component dominates the threat signal — early in the disease. In long-established ME/CFS, the threat signal is maintained by host factors (immune dysregulation, autoantibodies, metabolic dysfunction) and a ~20% reduction in viral protein production from antivirals is insufficient to shift the system. This model is theoretical, but it offers an honest reading of why antivirals might work for some recently-infected patients and fail for most long-ill patients (Loth 2026).

What you can actually do

• Get meaningful serology, not just IgG. Standard EBV panels (VCA IgG, EBNA IgG) tell you that you were exposed — which >90% of adults are. Meaningful serology for reactivation includes EA-D (early antigen, diffuse pattern), high/rising VCA IgG (and VCA IgM only to tell you the infection was recent/primary rather than a reactivation), dUTPase antibodies (research only, not clinically available yet), and — where accessible — EBV quantitative PCR. HHV-6 PCR from whole blood distinguishes active replication from integrated virus.
• If serology is suggestive and the clinical picture is post-infectious, a time-limited valacyclovir trial (1 g TID, 6 months) is a reasonable conversation to have with a clinician. It is well tolerated, and a negative trial at least closes one door. Do not stay on valacyclovir indefinitely — if there is no change at 6 months, stop.
• Valganciclovir is a specialist decision. The toxicity (bone marrow suppression, teratogenicity), cost (typically several thousand dollars per month, often requiring insurance approval for an off-label ME/CFS indication, and frequently denied), and monitoring burden mean it should be managed by a clinician experienced with the drug, with regular CBC monitoring. A 3–6 month trial with pre-defined stopping rules is the responsible approach.
• Treat the MCAS if it is present. Mast-cell stabilisers intercept the EBV→MMP-9 pathway downstream of the virus and may reduce symptoms regardless of whether the antivirals help. See the MCAS treatment articles.
• Accept that current antivirals may not be the right drugs for the ALR problem. Drugs that target immediate-early gene expression or dUTPase signalling directly do not exist. The antivirals we have were designed for fully lytic herpesvirus infections (shingles, CMV retinitis, genital herpes). Using them for ALR in ME/CFS is a mismatch between mechanism and pharmacology — and a negative trial is not a failure of the viral hypothesis; it is a failure of the available drugs to address the specific viral state that is active.

The bottom line

Chronic infection is the most hypothesis-heavy treatment area in the Septad. The mechanisms are strong — EBV→mast cell→MMP-9, HHV-6 miR-aU14→mitochondrial fragmentation, poly-herpesvirus dUTPase co-activation — but the drugs that target these mechanisms are weak (valacyclovir misses ALR) or toxic (valganciclovir suppresses bone marrow). The mast-cell stabiliser intercept — blocking the host response downstream of the virus — is the one intervention that bypasses the ALR problem and is low-risk in the MCAS-positive subset (Loth 2026).

A failed antiviral trial is expected pharmacology, not a personal failure, and not a disproof of the viral-reactivation hypothesis. The honest clinical approach is a time-limited, endpoint-defined trial with a stopping rule — and if it fails, to shift to mast-cell stabilisation and accept that current antivirals are not the right tools for the job.

Next in this series: Autoimmunity — antibodies that talk to the nervous system, GPCR autoantibodies, and why removing them doesn’t always fix the problem.

For the comprehensive, fully-cited picture of how chronic infection is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

First, a plain warning

This is an explanation, not self-medication advice, and I am not a doctor. Antiviral medications (valacyclovir, valganciclovir) require a prescriber and renal monitoring. Tick-borne infections require specialist evaluation — untreated Lyme can progress to neurological and cardiac complications. If you have a new bull’s-eye rash, facial palsy, or joint swelling after a tick bite, seek urgent medical attention.

The short version, if you only read one part

• A large share of ME/CFS cases follow an acute infection (commonly cited around 70%, though the exact proportion is debated and retrospective), and the Big Four are EBV/mononucleosis, HHV-6 (roseola → neurotropic, integrates into chromosomes), tick-borne Borrelia (Lyme), and SARS-CoV-2 — each of which is documented to precede a subset of ME/CFS and post-infectious fatigue cases.
• The central puzzle is abortive lytic replication (ALR). The virus enters the lytic cycle — expressing immunostimulatory proteins — but stops before producing virions. There is no measurable viral load in blood, but the viral proteins (dUTPases, immediate-early genes) are still being made, and the immune system still sees them. This explains why antibody titres are high but PCR is negative (Cox et al. 2022).
• EBV directly activates mast cells. Recombinant EBV protein added to human mast cells increased MMP-9 release nearly sixfold (2,464 vs 433 pg/ml) — and MMP-9 degrades the blood-brain barrier, allowing peripheral inflammatory mediators to access the brain (Chinnappan et al. 2026).
• Poly-herpesvirus co-reactivation is the rule, not the exception. 72.5% of ME/CFS patients had elevated antibodies to multiple herpesvirus dUTPases vs 31% of controls. EBV, HHV-6, CMV, and VZV can reactivate together — the immune system is fighting a multi-front viral war with depleted resources (Palomo et al. 2026).
• Chronic infection is the most hypothesis-heavy topic in the Septad. It is mechanistically central but harder to treat plainly — antivirals that block DNA replication in fully lytic virus may be ineffective against ALR, and the evidence for them in ME/CFS is weak and unreplicated.

The EBV connection: mechanism, not serology

Epstein-Barr virus infects >90% of the world’s population. In most people, it establishes latency in B-cells and remains quiescent for life, suppressed by T-cell surveillance. In ME/CFS, several lines of evidence point to EBV reactivation — but “reactivation” does not mean the same thing as acute mononucleosis.

HHV-6: the neurotropic integrator

Human herpesvirus 6 (HHV-6) infects nearly all children by age 2 (causing roseola) and establishes latency in T-cells, monocytes, and — critically — in the central nervous system. HHV-6 is the only human herpesvirus that can integrate into the host chromosome (ciHHV-6), creating a permanent genetic reservoir that standard antivirals cannot clear.

In ME/CFS, several findings point to HHV-6 involvement:

• miR-aU14 meets mitochondria. HHV-6 encodes a microRNA, miR-aU14, that inhibits the host’s miR-30 family. miR-30 normally suppresses p53 and DRP1 — removing that inhibition activates p53-dependent DRP1 translocation to mitochondria, causing mitochondrial fragmentation. The virus does not infect mitochondria directly. It sends a microRNA that tells the cell to break its own mitochondria (Schreiner et al. 2020).
• Post-mortem neuroinvasion. HHV-6 miR-aU14 was found in the choroid plexus, hippocampus, amygdala, and dorsal root ganglia of ME/CFS patients at autopsy — and was absent in 24 controls. The sample was n=3, so this is a finding, not a settled fact. But it places a viral nucleic acid in the exact brain regions implicated in ME/CFS cognitive, autonomic, and sensory symptoms.
• Prevalence. Latent or persistent HHV-6 has been reported in around 47% of ME/CFS patients vs roughly 10% of controls in some studies (single-study figures; ranges vary by assay and definition).

Tick-borne disease: Lyme, Bartonella, Babesia

Lyme disease (Borrelia burgdorferi) causes acute infection treated with 2–4 weeks of antibiotics. In 10–20% of cases, symptoms persist for months to years — post-treatment Lyme disease syndrome (PTLDS). PTLDS is reported to overlap heavily with ME/CFS on the core symptom set (fatigue, cognitive dysfunction, musculoskeletal pain, sleep disturbance), making the two conditions nearly clinically indistinguishable in the post-infectious phase.

The mechanism debate mirrors the EBV debate: is there persistent Borrelia (antibiotic-refractory reservoirs, biofilm-protected spirochetes) or is PTLDS a post-infectious immune dysregulation that outlasts the pathogen? The evidence for persistent infection — culture, PCR, xenodiagnosis — exists but is inconsistent and technically challenging. The evidence for post-infectious immune dysregulation — elevated cytokines, T-cell exhaustion, autoantibody generation — is robust and mirrors ME/CFS (Nawrocki et al. 2025).

Bartonella and Babesia are tick-borne co-infections that are less studied than Lyme. Bartonella DNA has been detected in ~26% and Babesia in ~24% of ME/CFS cohorts — but these studies lacked healthy controls, so the baseline community prevalence (which is not zero) is unknown. The honest position: tick-borne co-infections are plausible contributors in the exposed individual; the evidence for them as a general ME/CFS mechanism is weak.

Poly-herpesvirus co-reactivation

The immune system does not fight one herpesvirus at a time. The 72.5% poly-herpesvirus seroreactivity rate (vs 31% in controls) suggests that when immune surveillance weakens — and T-cell exhaustion, NK dysfunction, and CD8+ senescence are all documented in ME/CFS — multiple latent herpesviruses reactivate simultaneously (Palomo et al. 2026).

This creates a treatment problem. Valacyclovir inhibits EBV DNA polymerase (lytic replication) but has no effect on HHV-6. Valganciclovir covers HHV-6 and CMV but is more toxic. No single antiviral covers all four (EBV, HHV-6, CMV, VZV) without additive toxicity. And none of them block ALR, because the early viral proteins are produced before the DNA-polymerase step that the drugs target.

Chronic infection by itself vs ME/CFS with chronic infection

Active chronic infection on its own (no ME/CFS). Active EBV replication (PCR-positive, detectable viral load), active Lyme (culture or PCR-confirmed), active HHV-6 (viremia in transplant patients) — these are infectious diseases treated with antivirals or antibiotics. If the pathogen is identified and a drug exists that targets it, treatment is straightforward: give the drug, monitor the pathogen, confirm clearance.

ME/CFS with chronic immune stimulation (the situation this series is about). When the infection history is in the past and the current problem is ALR, immune dysregulation, or post-infectious autoimmunity, the treatment logic changes:

• Antivirals that work against lytic replication may not work against ALR. The viral proteins driving the immune response are produced before the drug’s target step. A negative antiviral trial does not prove the virus is irrelevant — it proves the drug, at that dose, did not interrupt the specific replication stage that is active.
• Eradication may be impossible. Integrated HHV-6 cannot be removed. Latent EBV in memory B-cells is lifelong. The treatment goal shifts from “clear the pathogen” to “suppress the immune activation the pathogen is causing.”
• The host immune defect may be the rate-limiting factor, not the viral load. If T-cells are exhausted and NK cells are dysfunctional, giving an antiviral without addressing the immune defect is like giving a firefighter a smaller hose while the water pressure is dropping. The fire is not the pathogen; the fire is the immune response to the pathogen, and the response is failing because the immune system is depleted.

The bottom line

Chronic infection is the most mechanistically central and therapeutically frustrating of the Septad conditions. The evidence that herpesviruses and tick-borne pathogens are involved in ME/CFS is strong — abortive lytic replication, poly-herpesvirus co-reactivation, EBV→mast cell→MMP-9, and HHV-6 mitochondrial fragmentation are not fringe hypotheses. But the evidence that targeting them with current antivirals improves ME/CFS is weak and unreplicated — and the distinction between ALR (which antivirals do not block) and lytic replication (which they do) explains why (Loth 2026).

The treatment companion article covers the antivirals (valacyclovir, valganciclovir, famciclovir), the immunomodulators (cimetidine), the mast-cell stabilisers that intercept the EBV→MMP-9 pathway downstream of the virus, and the honest “what if the antivirals don’t work” discussion.

Next in this mini-series: the antivirals, the evidence, and what a negative trial does and does not tell you [see the companion article].

For the comprehensive, fully-cited picture of how chronic infection is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

First, a plain warning

Everything below is a conversation to have with a clinician, not self-medication. SIBO antibiotics and prokinetics are prescription medications. Elimination diets — especially prolonged low-FODMAP or low-histamine — risk malnutrition and should be supervised by a dietitian. Elemental diet requires medical supervision. DAO supplements are unregulated.

The treatment ladder

Expected results

• Meal spacing works quickly or not at all. If 4–5 hour gaps between meals reduce bloating within 2 weeks, the MMC is functional and motility is the lever. If it does nothing, the MMC may be too impaired for meal spacing alone to restore it.
• The low-histamine diet is a 3–4 week diagnostic probe. If symptoms improve, continue with structured reintroduction. If nothing changes, stop the restriction — do not drift into long-term malnutrition.
• Rifaximin works within the 14-day course or not at all. If bloating, distension, and bowel habits normalise, SIBO was present. If nothing changes — and the breath test was positive — one honest possibility is a false-positive breath test (~30–40% rate).
• Prokinetics prevent recurrence, not acute SIBO. If SIBO recurs within 3 months of antibiotics, the MMC is not being maintained — add a prokinetic. If SIBO does not recur without prokinetics, the MMC recovered on its own.
• DAO is a diagnostic probe, not a long-term fix. If DAO supplementation improves post-meal symptoms, HIT is a likely contributor and dietary management is the strategy — though response can also reflect placebo. DAO supplementation is suggestive of the mechanism; it does not replace dietary histamine avoidance.

An example structured trial

This is an example to give your clinician something to work from — not a self-prescription.

Start with meal spacing: 4–5 hours between meals, no snacking. Reassess bloating and early satiety at 2 weeks. If improved, continue — you have a motility-dependent component.

Add the low-histamine diet for 4 weeks. Reassess at 4 weeks. If post-meal flushing, tachycardia, and brain fog improve by >50%, histamine is a contributor. If nothing changes, stop the diet — histamine is not the dominant mechanism.

If bloating, early satiety, and alternating bowel habits persist despite meal spacing, get a lactulose breath test. If hydrogen or methane is elevated, treat with rifaximin 550 mg TID ± neomycin 500 mg BID for 14 days. Reassess at 4 weeks post-treatment. If SIBO symptoms recur within 3 months, add a prokinetic (erythromycin 50 mg at bedtime). Reassess at 12 weeks. Stop the prokinetic if SIBO has not recurred at 6 months — the MMC may have recovered.

If all of the above fails and SIBO is confirmed on repeat testing, consider a 14-day elemental diet under medical supervision.

Only add one intervention at a time. Adding everything at once — meal spacing + low-histamine + rifaximin + prokinetic simultaneously — makes it impossible to know what helped.

What it means if the treatment does not help

One: SIBO treatment may fail because the breath test was a false positive. Breath tests have a ~30–40% false-positive rate, and the 2024 consensus statement’s “remains unproven” conclusion is a genuine epistemic limitation. If rifaximin and herbal antimicrobials produce no benefit, one honest possibility is that SIBO was never the problem — the symptoms had a different mechanism (visceral hypersensitivity, MCAS, rapid transit).

Two: the low-histamine diet may fail because histamine is not the problem. If 4 weeks of strict dietary histamine reduction produces no change, and DAO supplementation also fails, histamine is unlikely to be a major contributor — stop dietary restriction and look to other mechanisms.

Three: prokinetics may fail because the MMC is structurally damaged, not just slowed. If ICCs are destroyed by autoantibodies, not just functionally suppressed, a drug that stimulates residual ICC function may have limited effect. The distinction matters for expectations — prokinetics support residual function; they do not regenerate lost pacemaker cells.

Four: the gut may not be the dominant amplifier. If you have treated SIBO, eliminated histamine, restored butyrate, and taken prokinetics — and the fatigue, brain fog, and PEM are unchanged — the gut was a passenger, not a driver, in your specific case. That is honest and useful information. Stop pursuing gut-targeted treatments and move to the next amplifier.

Five: elemental diet may fail because the underlying cause was never the bacteria. Elemental diet removes all dietary substrate for bacterial fermentation — if 14 days of exclusive elemental diet produces no improvement, the symptoms were not bacterial-overgrowth-driven. They may be visceral hypersensitivity, MCAS, or rapid transit. That is a finding — the most aggressive SIBO treatment available has been tested and has failed, and the probability that SIBO is the dominant mechanism drops substantially.

The falsifiable predictions

• If mast-cell stabilisers (cromolyn, ketotifen) improve GI dysmotility independent of diet changes, mast-cell mediators are a direct motility brake, and stabilisers are upstream of prokinetics in the treatment algorithm.
• If butyrate supplementation reduces mast-cell degranulation markers (urinary methylhistamine, prostaglandin D₂ metabolite) in SIBO-positive ME/CFS, the butyrate-deficiency→mast-cell-disinhibition pathway is clinically relevant and the microbiome-mast-cell axis is a targetable mechanism.
• If prokinetics prevent SIBO recurrence in patients with normal ICC function (anti-CdtB/anti-vinculin negative) but fail in patients with positive anti-ICC antibodies, the ICC-damage mechanism is clinically actionable and antibody testing should guide prokinetic expectations.

What you can actually do

• Start with meal spacing. Four to five hours between meals, no snacking. This costs nothing, requires no prescription, and directly supports MMC function. If it helps, you have identified a motility-dependent symptom component.
• Trial a low-histamine diet for 4 weeks with a clear stop-date. If symptoms improve, histamine is relevant. If nothing changes, stop the restriction.
• Get a lactulose breath test, but read it critically. A positive test suggests SIBO; a negative test does not rule it out. Treat the clinical picture, not a borderline number.
• If SIBO is treated, add a prokinetic to prevent recurrence. Antibiotics clear the overgrowth; only restored MMC function prevents it from coming back.
• Consider DAO before meals as a diagnostic probe. Consistent DAO-related improvement suggests HIT; needing antihistamines as well suggests an additional endogenous mast-cell (MCAS) component; neither helping suggests histamine is not a major contributor. Treat each of these as a suggestive signal, not a proof — response can also reflect placebo or unrelated gut improvement.
• Accept that the gut is one amplifier among several. Treating it may reduce the total illness burden without fixing the energy crisis. That is not failure — it is accurate targeting in a multi-amplifier system.

The bottom line

GI dysmotility treatment in ME/CFS runs from meal spacing and low-histamine elimination through SIBO antibiotics and elemental diet to DAO, butyrate, and prokinetics. Diet is the first lever — the intervention the patient controls directly — and the pharmacology is supportive, not central (Loth 2026).

A failed trial of any one intervention tells you something specific — which mechanism is not dominant in your case — and that is useful information, not a personal failure. The structured-trial architecture (one intervention at a time, named endpoint, named stop-rule) is how you distinguish signal from noise in a multi-mechanism system.

For the comprehensive, fully-cited picture of how GI dysmotility and the gut microbiome are weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

First, a plain warning

This is an explanation, not self-medication advice, and I am not a doctor. Rapidly progressive weight loss, inability to keep down liquids, severe abdominal pain, or blood in the stool require urgent evaluation. Elimination diets — especially prolonged low-FODMAP or low-histamine — risk malnutrition and should be supervised by a dietitian.

The short version, if you only read one part

• GI symptoms affect 70–90% of ME/CFS patients — nausea, bloating, early fullness, alternating diarrhoea and constipation, food intolerances — making the gut the most consistently involved organ system outside the brain.
• Three mechanisms converge. (1) Autonomic neuropathy damages the interstitial cells of Cajal and the vagus, slowing gut motility — food sits, bacteria grow (SIBO). (2) Mast cells in the gut mucosa degranulate, releasing histamine, tryptase, and prostaglandins that further slow motility and increase permeability. (3) Sulfate-reducing bacteria produce hydrogen sulfide (H₂S), which directly poisons mitochondrial Complex IV — the same target as cyanide — draining cellular ATP (Nicholls et al. 2013).
• SIBO is real in ME/CFS but its diagnosis is contested. Breath tests have high false-positive rates, and a 2024 international consensus statement concluded the SIBO hypothesis “remains unproven” (Kashyap et al. 2024). Treat the clinical picture, not the breath-test number.
• Treating the gut rarely fixes the energy crisis, but ignoring the gut makes the energy crisis worse. The gut is an amplifier — histamine overload, nutrient malabsorption, and H₂S toxicity all drain a system that is already in deficit.

Three mechanisms

The DAO–histamine connection: why food makes you flare

Diamine oxidase (DAO) is the intestinal enzyme that degrades dietary histamine before it can enter the bloodstream. DAO is produced by gut epithelial cells; its activity is reduced by mucosal inflammation, SIBO, and genetic polymorphisms in the AOC1 gene (present in ~15–20% of the population). When DAO activity is low, high-histamine foods (aged cheeses, fermented foods, cured meats, wine, tomatoes, spinach) deliver a histamine load that the gut cannot clear.

This matters in ME/CFS because the distinction between histamine intolerance (HIT — impaired DAO-mediated degradation of dietary histamine, with normal mast-cell function) and MCAS (excessive endogenous histamine production from hyperactive mast cells) changes the treatment. HIT responds to DAO supplementation and a low-histamine diet; MCAS requires mast-cell stabilisers and antihistamines. The two can coexist, and the primary document proposes a diagnostic probe: supplement DAO before meals. If DAO alone reduces post-meal symptoms, HIT is the likely contributor; if antihistamines are needed in addition, an endogenous mast-cell (MCAS) component is likely; if neither DAO nor antihistamines help, histamine is probably not the dominant driver (Loth 2026). These probe responses are suggestive signals rather than proofs — placebo, diet changes, and non-histamine mechanisms can all confound an individual trial.

The order matters: gut problems are usually downstream, not the cause

GI dysmotility in ME/CFS is almost always downstream of the systemic illness, not its cause. The ICC damage is post-infectious (antibodies generated during the initial gastroenteritis persist and continue to damage the pacemaker cells). The vagal denervation is post-viral (COVID, EBV, or other neurotropic infection damaging the vagus directly). The mast-cell–gut loop is a consequence of systemic MCAS, not a primary gut disorder. The hydrogen sulfide problem is driven by the bacterial overgrowth that follows the motility failure.

But — as with every amplifier in this series — “downstream” does not mean “does not matter.” Once the gut loop is running, it adds a histamine load, a nutrient-malabsorption burden, and an H₂S energy drain on top of an already-depleted system. Treating the gut does not fix the core energy/immune defect, but not treating the gut means the system drains faster than it can recover.

Gut problems by themselves vs ME/CFS with gut problems

IBS/SIBO on their own (no ME/CFS). IBS is a disorder of gut-brain interaction — visceral hypersensitivity, altered motility, and microbiome shifts produce pain, bloating, and altered bowel habits. SIBO is a defined overgrowth treatable with antibiotics. Treatment is gut-focused: diet, prokinetics, antibiotics, and the symptoms are confined to the GI tract. Fatigue, if present, is secondary to malnutrition and resolves with gut treatment.

ME/CFS with gut problems (the situation this series is about). When GI dysmotility sits on top of ME/CFS, three things are different:

• The gut is one amplifier. The histamine overload, the H₂S energy drain, and the nutrient malabsorption all worsen the systemic energy deficit — but fixing the gut does not fix the PEM or the immune dysregulation (Loth 2026).
• Diet is the first lever — but it is a lever, not a cure. A low-histamine diet reduces the histamine load but does not stop the mast cells from degranulating. Meal spacing supports the MMC but does not regenerate lost ICCs. The dietary interventions are partial, supportive, and worth doing — and they are not curative.
• A failed gut treatment says nothing against your ME/CFS. Because the gut is one amplifier among several, a SIBO treatment that fails to improve fatigue does not mean the SIBO was not real — it means the SIBO was not the dominant drain on your energy budget, and the honest next step is to look elsewhere.

The bottom line

The gut is the most consistently involved organ system in ME/CFS outside the brain, and it is the one system where the patient has direct agency through diet. Three mechanisms — ICC/vagal motility failure, mast-cell–driven permeability and dysmotility, and hydrogen sulfide mitochondrial poisoning — are not mutually exclusive, and they converge on the same endpoint: a gut that steals energy rather than providing it (Loth 2026).

Next in this mini-series: how to actually treat the gut — meal spacing, low-histamine elimination, SIBO antibiotics and herbals, elemental diet, DAO, butyrate, prokinetics, and the honest “what if nothing helps” discussion [see the companion article].

For the comprehensive, fully-cited picture of how GI dysmotility and the gut microbiome are weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).

First, a plain warning

Everything below is a conversation to have with a clinician, not self-medication. LDN is a compounded medication requiring a prescription. COX-2 inhibitors have cardiovascular and renal risks. Gabapentinoids cause physiological dependence. NMDA antagonists (ketamine) are controlled substances requiring supervised administration. Work with a pain specialist who understands ME/CFS.

The treatment architecture

An example structured trial

This is an example to give your clinician something to work from — not a self-prescription.

Start with LDN 0.5–1.5 mg daily (morning, to avoid insomnia — see above), titrate to 4.5 mg over 4–8 weeks. Reassess the single target symptom — resting burning pain, tender-point sensitivity, or post-exertional pain amplification — at 12 weeks. If no benefit at 4.5 mg, stop — doses above 4.5 mg lose the TLR4-selective effect.

If LDN produces partial benefit, add celecoxib 200 mg daily for 2 weeks as a diagnostic probe. If burning pain and heat sensitivity improve within 2 weeks, a COX-2/PGE₂-mediated component is suggested (not confirmed — placebo or a general analgesic effect can look identical) — but prolonged COX-2 inhibition carries cardiovascular risk, so use it to inform the mechanism and consider intermittent rather than continuous dosing. If celecoxib has no effect at 2 weeks, stop — the COX-2 pathway is probably not a dominant amplifier.

If LDN + celecoxib have both been trialled, add an SNRI (duloxetine 30 mg, titrate to 60 mg over 4 weeks) or pregabalin (25–50 mg at bedtime, titrate to the maximum tolerated dose). Reassess at 8 weeks. Stop if sedation or activation prevents reaching an effective dose — the therapeutic window in ME/CFS is narrower than in pure fibromyalgia, and a null result at the maximum tolerated dose is a valid trial.

Add PEA 300–600 mg twice daily as a low-risk adjunct at any point in the sequence. Its broad-spectrum mechanism (mast-cell stabilisation, microglial modulation, peripheral-sensitisation reduction) makes it complementary to LDN and SNRIs rather than redundant. Reassess at 8 weeks.

Only add one intervention at a time. The goal is to find the combination that reduces pain enough to improve function without adding sedation or cognitive slowing that worsens the core illness. The structured-trial architecture is the only way to know what each drug contributes.

Expected results

• LDN takes weeks, not days. The anti-microglial effect builds gradually. Titrate to 4.5 mg over 4–8 weeks; reassess at 12 weeks. Vivid dreams and initial insomnia are common and usually resolve. If no benefit at 4.5 mg after 12 weeks, stop — LDN has a sharp dose-response curve, and doses above 4.5 mg lose the TLR4-selective effect and enter standard opioid-antagonist territory.
• SNRIs and gabapentinoids have a therapeutic lag and a side-effect cost. The effective dose for pain is often higher than the starting dose. A drug stopped at 2 weeks because of sedation was never given a fair trial — but a drug that causes intolerable sedation at the lowest dose is a genuine failure and should be stopped.
• COX-2 inhibitors work quickly (days) or not at all. If celecoxib reduces burning pain within a week, a COX-2/PGE₂-mediated component is suggested — though not proven, since a non-COX-2 analgesic or placebo effect can look identical. If it does nothing after 2 weeks, the loop is probably not the dominant contributor, and prolonged COX-2 inhibition carries cardiovascular risk — stop.
• PEA is slow and subtle. Expect gradual improvement over 4–8 weeks, not dramatic relief. The benefit, if it comes, is often described as “less sensitised” rather than “less pain” — a reduction in the amplification rather than the signal.
• Partial response is the most common outcome. Pain reduced from 8/10 to 5/10 on LDN + PEA, with SNRIs adding a further point but causing unacceptable side effects — that is a real-world treatment result, not a failure.

What it means if the treatment does not help

One: failed LDN does not disprove central sensitisation. LDN blocks microglial TLR4 — one pathway among several. If the dominant mechanism is deficient descending inhibition (low CSF norepinephrine), an SNRI may help where LDN did not. Two: sedation may prevent reaching an effective gabapentinoid dose. The maximum tolerated dose in ME/CFS may be below the analgesic dose. This is a pharmacological access problem, not evidence that the pain is not central. Three: the PEM-driven re-sensitisation loop may overwhelm the treatment. If every minor exertion triggers an ASIC3/P2X4/TLR4 re-sensitisation cascade, then even effective medications are fighting a moving target. The pain management cannot succeed if the energy envelope is not respected — pacing is not an alternative to medication; it is the condition under which medication has a chance to work. Four: a sequence of well-run trials (LDN, SNRI, gabapentinoid, PEA, COX-2 inhibitor) with no benefit is genuine evidence that the available pharmacology is not sufficient. That is a limitation of current medicine, not a verdict on the pain being “psychological.” It means the focus shifts to non-pharmacological strategies (pacing, graded sensory exposure, cognitive strategies) and radical energy conservation — not because the pain is not real, but because the drugs are not adequate.

The falsifiable predictions

• If LDN reduces pain in ME/CFS more in the PEM-amplified pain subgroup (pain worsens 12–48h post-exertion) than in the stable-pain subgroup, the TLR4-driven PEM re-sensitisation is clinically relevant (Loth 2026).
• If celecoxib reduces heat hyperalgesia on QST but does not improve VO₂ peak or PEM threshold, the COX-2/PGE₂/TRPV1 loop is a pain amplifier, not an energy-failure mechanism — consistent with the nociplastic-amplifier model.
• If TRPA1 antagonists reduce post-exertional pain more than baseline pain, the oxidative-stress-to-TRPA1 pathway is a PEM-specific pain mechanism.
• If fibromyalgia-only patients (no ME/CFS) show normal exercise muscle perfusion on NIRS while ME/CFS+fibromyalgia patients show impaired perfusion, the vascular-TRPV1 mechanism distinguishes the ME/CFS-overlap pain from pure fibromyalgia pain.

What you can actually do

• Start with LDN. Lowest side-effect burden, mechanistically relevant to microglial-driven central sensitisation, and evidence-supported in fibromyalgia. Titrate slowly, give it 12 weeks, and stop if no benefit at 4.5 mg.
• Use COX-2 inhibitors as a diagnostic probe, not a maintenance strategy. If celecoxib 200 mg daily for 2 weeks reduces burning pain, a COX-2/PGE₂-mediated component is suggested (not proven — a placebo or general analgesic effect can look the same), which tells you something about the pain mechanism even if you do not stay on the drug long-term.
• Add PEA. Low risk, multi-modal mechanism, meta-analytically effective across pain types. It is not a replacement for LDN or SNRIs — it is an adjunct that may shift the sensitivity without adding side effects.
• For SNRIs and gabapentinoids, start low, go slow, and name a stop-rule. The goal is to find a dose that reduces pain without worsening brain fog. If no dose in the therapeutic range achieves that, the drug has failed — document it and move on.
• Respect the energy envelope. Pain that worsens after crashes is a PEM signal. Medications that improve pain cannot compensate for an energy budget that is continuously overdrawn. Pacing is the foundation on which pain management is built — not an alternative to it.

The bottom line

Fibromyalgia-pattern pain in ME/CFS is treated with a multi-modal approach: LDN for microglial TLR4, SNRIs and gabapentinoids for central amplification (with sedation caveats), COX-2 inhibitors for the PGE₂/TRPV1 loop, PEA for broad-spectrum sensitisation reduction, and NMDA antagonists for severe wind-up. The PEM-driven re-sensitisation — absent in pure fibromyalgia — means the energy envelope and the pain envelope are coupled, and pacing is the foundation on which all pharmacological pain management rests (Loth 2026).

A failed trial of any one drug does not mean the pain is not real. It means that drug, at that dose, in that patient, was not sufficient — and the honest next step is to try the next mechanism or to accept that current pharmacology has limits and shift the focus to function within them.

Next in this series: The gut that misbehaves — GI dysmotility, SIBO, and why the stomach is the first lever ME/CFS patients can pull.

For the comprehensive, fully-cited picture of how fibromyalgia-pattern pain is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).