Table of Contents
Companion to: Spikeopathy Pathways: mTOR, Glymphatic Clearance, and the SIRT1/PGC-1α Axis. Read that first — this article assumes you understand the two-system central model (intracellular mTOR/autophagy + extracellular glymphatic clearance) and the mast-cell third arm.
Where this sits. This article individualises a larger model — spikeopathy as systemic clearance and tolerance failure under persistent antigenic drive. The model itself and the full lever map live on the Spikeopathy hub.
Plain English
Your body has different waste-clearance systems. Some work inside cells. Others work between cells and in the spaces around the brain. When spike protein lingers, it can interfere with these systems in different ways depending on the person.
Two people with the same exposure history can end up with very different dominant problems. One person's main issue might be that their brain isn't clearing waste properly at night. Another's might be that their cells can't clean up damaged mitochondria. A third might have immune cells that are more likely to tolerate the spike instead of clearing it.
This article gives you a way to think about which system is most likely dominant in you, based on your symptoms and some basic tests. It's not a diagnosis. It's a working map to help you and your clinician focus on the levers that actually match your pattern.
You don't have to try everything. You can start with the shortlist that fits your dominant subtype and see what moves.
Why this article exists
The pathways article maps the mechanism. It doesn't tell you which mechanism is dominant in you. "Spikeopathy" as a single umbrella hypothesis has the same problem "cancer" did a century ago — correct at the mechanistic level, useless at the bedside. Two people with the same exposure history can sit in the same clinic with entirely different symptom clusters, entirely different biomarker profiles, entirely different drug shortlists. Both correctly diagnosed with spikeopathy. Both needing different treatment.
The community is mechanism-rich and measurement-poor. Dozens of proposed levers, no clean way to decide which levers apply to which patient. This article tries to fix that. Three things make it different from what's out there:
Mechanism-based subtyping. Existing PASC subtyping work (Su 2022, Thaweethai 2023, Klein 2023, Wang 2026) clusters patients empirically — by symptoms, by multi-omics signatures. Useful, but it tells you that subtypes exist, not why. This framework clusters by the underlying clearance/tolerance mechanism the pathways article identifies. Nobody has published a comprehensive mechanism-based subtype framework for PASC. I checked the literature through July 2026. This is the differentiator.
Orderable biomarkers, mapped. Every subtype has a biomarker signature you can actually test, with UK ordering tiers (NHS-routine, NHS-specialist, private, research-only) and reference ranges. Where PASC-specific ranges don't exist — and for most of the markers that matter, they don't — that is said explicitly. Track within-person trends, not population cutoffs.
Falsifiable. The framework makes specific predictions. If they fail, the framework gets revised or abandoned. That section is at the end, not buried.
Evidence protocol (same as the main article). Every claim gets a tag.
[ESTABLISHED]= multiple RCTs or strong human data.[MECHANISTIC]= in vitro or animal, pathway level.[HUMAN-IMAGING]= clinical cohorts with imaging or biomarkers.[HYPOTHESIS]= convergent inference, not yet confirmed at scale. Weigh each claim on its own. Don't bundle them. Full grading system, sourcing standard, and correction policy on the Methodology page. None of the subtypes are validated as clinical diagnoses. This is a framework, not a diagnostic manual.
Piece 1 — The five mechanistic subtypes
Each subtype is a hypothesis about which node of the clearance/tolerance network is the rate-limiting failure in a given person. In reality, subtypes overlap and most people have a dominant subtype plus one or two secondary features. The framework's value is in identifying the dominant failure mode — the node where intervention buys you the most.
The biomarker column is shorthand; full ranges and ordering tiers are in Piece 2. The "first-line lever" column is shorthand; full shortlists are in Piece 3 (pending).
| # | Subtype (dominant failure) | Hallmark symptoms | Mechanism (one-line) | Sharpest biomarker(s) | First-line lever class |
|---|---|---|---|---|---|
| A | Glymphatic-dominant | Brain fog worse than fatigue; cognitive slowing; sleep non-restorative no matter how long; morning headaches; symptom flare after poor sleep | Astrocytic AQP4 mislocalisation + sleep-dependent clearance failure. Spike-driven inflammatory cytokines (IL-6, TNF-α, IL-1β) detach AQP4 from endfeet; CSF-ISF exchange collapses; amyloidogenic spike fragments, tau, α-synuclein back up in parenchyma | DTI-ALPS < 1.6 [HUMAN-IMAGING]; rising NfL (age-adjusted); rising GFAP (age-adjusted) | Sleep optimisation + lateral sleep position + PBM/NIR; glymphatic is the single node where sleep architecture interventions are not optional |
| B | Autophagy / mTOR-dominant | Profound post-exertional malaise; "hit by a bus" after minor exertion; cold extremities; insulin resistance signs; hunger dysregulation | Intracellular autophagy failure. Spike-p53-mTOR survival trap keeps spike-loaded cells alive instead of clearing them; concurrent insulin resistance further activates mTOR; protein aggregates accumulate inside neurons and endothelium | HOMA-IR > 2.0 (or fasting insulin > 8 mIU/mL); hsCRP > 3; optionally TMEM106B genotype (risk allele) | mTOR modulation (rapamycin prescription-only, or gentler AMPK activators: berberine, fasting, spermidine); not strength training until PEM resolves |
| C | Mitochondrial / SIRT1-PGC-1α-dominant | Fatigue dominates over cognition; exercise intolerance with normal cardiac workup; muscle pain; cold intolerance; "battery doesn't recharge" | NAD⁺ depletion + SIRT1 inactivation + PGC-1α suppression → mitochondrial biogenesis collapses → electron transport chain underperforms → lactate accumulates → both clearance systems lose their energy supply | Resting lactate > 2.0 mmol/L; post-exertional lactate > 5 mmol/L; CPET VO₂max below age-predicted; NAD⁺ trend (research-only) | NAD⁺ precursors (NR, NMN, niacin) + ubiquinol + PBM/NIR; the single subtype where mitochondrial support is the primary lever, not adjunctive |
| D | Tolerance / EPO-EPAR-dominant | Tissue hypoxia signs (cold extremities, pallor, slow wound healing); microclot symptoms (chest tightness, breathlessness on minimal exertion); bruising; menorrhagia; poor response to iron supplementation | Spike-driven EPO axis disruption + microclot formation + immune-tolerance trap (RAGE/IL-10) keep tissue oxygen delivery below demand. Cells survive but can't perform. Microclots capillary-level block perfusion | EPO low-normal or inappropriately normal for degree of anaemia; D-dimer elevated; fibrinogen elevated; ferritin low or normal despite anaemia signs; Thioflavin T microclot imaging (research) | Nattokinase/lumbrokinase excluded here — see framework-lev note. Address microclot + EPO axis with clinician; review iron handling, not just iron supplementation |
| E | Mast-cell / neuroimmune-dominant | Flushing; food sensitivities; histamine intolerance; POTS overlap; symptom flares with stress or allergens; skin manifestations; GI symptoms; anxiety/sense-of-doom episodes | Meningeal mast-cell degranulation (tryptase, chymase, histamine, IL-1β, IL-33) → BBB opening → mast-cell–microglia loop. The third upstream arm identified in the pathways article. Sits upstream of both clearance systems (proposed, Theoharides 2026, not proven in human tissue) | Baseline serum tryptase + acute:baseline delta (Valent rule: ≥20% rise + ≥2 ng/mL, drawn 1–4 h post-event); urinary 11β-PGF2α / LTE4 / N-methylhistamine (24h); CBC eosinophils (rough) | Mast-cell stabilisers (luteolin, quercetin OTC; ketotifen, cromolyn Rx; LDN off-label Rx); H1/H2 antihistamine stacking; low-histamine diet |
How to use the table
- Dominant subtype, not exclusive subtype. Most people have one dominant plus one or two secondary features. The framework's value is identifying the dominant failure — the node where intervention buys you the most. Treating secondary subtypes before the dominant one usually produces modest or absent response. Don't do that.
- Symptom pattern > single biomarker. Your symptoms are the first-pass filter. The biomarker confirms or redirects. Don't diagnose yourself off one number.
- Sharpest biomarker isn't always orderable. DTI-ALPS is the single most informative glymphatic test and isn't available on the NHS for this indication. Where the sharpest test is out of reach, the matrix in Piece 2 gives you the best orderable proxy.
- PEM points to B or C. Post-exertional malaise is the single most discriminating symptom. If PEM dominates, you're in autophagy/mTOR (B) or mitochondrial/SIRT1 (C) territory, not glymphatic-dominant (A). The lactate profile and HOMA-IR separate B from C.
What's genuinely new here
Verification check against the published literature through July 2026:
- Su et al. 2022 (
[HUMAN]) identified four symptom-cluster subtypes empirically using multi-omics. Useful taxonomy. Not mechanism-based. - Thaweethai et al. 2023 (
[HUMAN]) found three symptom-cluster subtypes in a large PASC cohort. Same pattern: empirical, not mechanistic. - Klein et al. 2023 (
[HUMAN]) used serum proteomics to define two endotypes. Closer to mechanism, but does not map to clearance pathways. - Wang et al. 2026 (
[HUMAN]) extended the endotype approach with longitudinal proteomics. Still mechanism-agnostic at the subtype level.
Nobody has published comprehensive PASC subtyping by clearance/tolerance mechanism. The framework above is a hypothesis built on convergent mechanism work — it is not validated and should not be read as a clinical diagnosis. The value is in testable predictions, which Piece 4 operationalises.
Piece 2 — Biomarker matrix: ordering tiers mapped to mechanism
Reference ranges caveat first. Validated PASC-specific reference ranges do not exist for: NfL, GFAP, DTI-ALPS, plasma spike, IL-6 (in PASC context), NAD⁺, or urinary mast-cell mediators. For these markers, track within-person trends, not population cutoffs. A number that is "in range" by general-population reference may be deeply abnormal for you. A rising trend is more actionable than any single snapshot.
All ranges below are approximate; always use your lab's reference and interpret in clinical context. Age-adjusted where noted. UK ordering tiers are indicative; access varies by ICB and by individual clinician willingness.
Tier 1 — NHS-routine (ask GP; rarely refused)
| Biomarker | Range (age-adjusted where noted; use lab-specific) | Maps to subtype |
|---|---|---|
| hsCRP | <1 mg/L optimal; <3 acceptable; >3 chronic inflammation; acute >10 | All — esp. B (autophagy/mTOR), E (mast-cell) |
| Ferritin | 13–150 µg/L ♀, 30–400 µg/L ♂ (adult). Interpret with hsCRP: high ferritin + high CRP = inflammation, not iron overload | C (mitochondrial/SIRT1), B (autophagy/mTOR) |
| HbA1c | <39 mmol/mol | B (autophagy/mTOR), C (mitochondrial) |
| D-dimer | <0.5 µg/mL FEU | D (tolerance/EPO-EPOR) — microclot-adjacent |
| Fibrinogen | 2–4 g/L | D (tolerance) |
| EPO | 4–24 mIU/mL | D (tolerance/EPO-EPOR) — direct |
| TSH / free T4 | 0.4–4.0 mIU/L / 10–20 pmol/L | All — non-specific (spike affects thyroid tone) |
| CBC + differential | Lab-specific; look at eosinophil and basophil trends | E (mast-cell) — rough screen only |
Tier 2 — NHS-specialist (referral or sympathetic GP)
| Biomarker | Range / note | Maps to subtype |
|---|---|---|
| Tryptase (serum) | Baseline 1–15 ng/mL (modern consensus includes HαT carriers; traditional upper limit <11.4–11.5 ng/mL). MCAS diagnostic rule (Valent 2019, PMID 31256161; 2023 update): acute rise ≥20% + ≥2 ng/mL above individual baseline, drawn 1–4 h post-event | E (mast-cell) — diagnostic |
| IL-6 | <5–7 pg/mL typical healthy upper limit (assay-dependent). No PASC-specific range. | B (autophagy/mTOR), E (mast-cell) |
| Lactate (resting ± post-exertional) | <2.0 mmol/L resting; >5 mmol/L post-exertion suspicious | C (mitochondrial/SIRT1) — core |
| Homocysteine | <10 µmol/L ideal; <15 acceptable | B (autophagy/mTOR) — methylation |
| CPET (cardiopulmonary exercise test) | VO₂max, anaerobic threshold; compare to age-predicted | C (mitochondrial/SIRT1) — functional |
| HOMA-IR (fasting glucose × fasting insulin / 22.5; often Tier 2 in UK) | <1.0 excellent; <2.0 normal; 2.0–2.9 mild IR; ≥3.0 significant (Gayoso-Diz 2013) | B (autophagy/mTOR) — core |
| Urinalysis ± 24h N-methylhistamine | Specialist-only; often refused | E (mast-cell) |
Tier 3 — Private / self-pay (UK: ~£30–£300 per marker)
| Biomarker | Range / note | Maps to subtype |
|---|---|---|
| NAD⁺ (whole blood) | No validated reference range; track trend only | C (mitochondrial/SIRT1) |
| Organic acids (urine) | Krebs cycle intermediates vs lab reference | C (mitochondrial/SIRT1) |
| 11β-PGF2α / LTE4 (24h urine) | No PASC range. More stable than serum tryptase; 24h collection | E (mast-cell) — urinary adjuncts supportive, not specific (limitations noted in 2025 reviews) |
| Salivary cortisol curve | AM peak / PM trough | D (tolerance) — HPA axis |
Tier 4 — Research-only / trial enrolment (not clinically orderable in UK)
| Biomarker | Range / note | Maps to subtype |
|---|---|---|
| NfL (plasma, Simoa) | Strongly age-dependent. 15–19 y ~2.5 pg/mL; 50–59 y ~7.3; 80–85 y ~19.5 (upper ranges 40+ in healthy elderly). Elevations >2–3× age-adjusted ULN suggest ongoing neuronal injury. (Practical Neurology 2023; Cooper 2023.) No PASC-specific range. | A (glymphatic) — axonal injury |
| GFAP (plasma) | Age-dependent. 15–19 y ~50 pg/mL; 50–59 y ~71; 70+ y 120–150+ upper limit. (Arslan 2025: ULN 38 pg/mL <50 y; 73 pg/mL 50–70 y; 156 pg/mL >70 y.) No PASC range. Abbott i-STAT available for some TBI contexts. | A (glymphatic) — astrocyte injury/activation |
| UCH-L1 | Less age-stratified data; typical healthy <10–360 pg/mL depending on assay. Often paired with GFAP in TBI panels. Elevated in acute injury | A (glymphatic) — acute neuronal injury |
| DTI-ALPS index (MRI) | <1.6 = impaired glymphatic flow. Higher = better clearance. Proprietary post-processing. (He 2025, Frontiers in Psychology.) | A (glymphatic) — direct, imaging |
| Plasma/serum spike (S1, Simoa) | Research assays only. No clinical cutoff. Swank 2023: ~60% of PASC at 2–12 mo, 0 matched controls. Rong 2024: tissue persistence up to 4 y post-infection | All — exposure/reservoir marker |
| Tissue spike (gut/mucosal biopsy IHC) | Research labs only | All — reservoir |
Cross-cutting notes
- Minimum informative panel for the framework: hsCRP + ferritin + tryptase + HOMA-IR + EPO + (CPET or resting + post-exertional lactate). Six tests, all Tier 1–2. Covers the four dominant subtypes (B, C, D, E) with a single panel.
- Adding A (glymphatic) requires Tier 4 — DTI-ALPS is the only direct test, and it is not orderable on the NHS for this indication. Best orderable proxies for A: NfL + GFAP (private Simoa, ~£150–250 each) + symptom pattern.
- Indicative UK costs (London, mid-2026, for budgeting only): NHS Tier 1 ~£0–50; tryptase private ~£45–85; HOMA-IR ~£35–60; NAD⁺ ~£120–200; NfL/GFAP Simoa ~£150–250 each; DTI-ALPS ~£600–900 if you can find a research-friendly radiologist.
- Always interpret in clinical context; always use your lab's reference; none of these markers are diagnostic of spikeopathy. They are mechanism-mapped signals to track alongside symptoms and the lever classes in Piece 3.
Excluded levers (and why)
This framework explicitly excludes nattokinase and lumbrokinase from all subtype shortlists (Piece 3, pending), despite their presence in the main article's Tier 2 compound matrix. Reason, stated plainly in the main article row:
In the context of this article's amyloidogenic-spike + clearance-failure model, these agents are presented only as a microclot/plumbing lever… They are not included in the subtype shortlists in the stratification companion because enzymatic fragmentation of amyloidogenic material sits in tension with the core hypothesis that persistent spike fragments contribute to ongoing pathology.
The in-vitro spike-degradation data (Tanikawa 2022) and fibrinaloid microclot degradation work (Grixti et al.) are real and remain relevant to the Spike Persistence companion for the microclot/persistence audience. They are excluded here because the subtype framework treats persistent/amyloidogenic spike material as something to clear via normal pathways (glymphatic flow, autophagy, immune recognition), not to enzymatically fragment. Cleaving an amyloid-forming protein into smaller pieces generates fragments that can spread further and re-aggregate in tissues (heart, brain, vasculature) — arguably worse than the intact protein if downstream clearance is impaired.
This is a framework-internal consistency choice, not a clinical recommendation against nattokinase in any individual case. Clinician judgement applies, especially for the microclot-dominant (subtype D) picture where the plumbing argument has its own separate logic.
Piece 3 — Per-subtype compound shortlists
How to use. Match your dominant subtype from the symptom clusters (Piece 1) + biomarker signature (Piece 2). Run the shortlist for 6–12 weeks with daily tracking. Most people show one primary node plus partial overlap. Reassess before layering. None of this is medical advice — clinician oversight required, especially for prescription items (rapamycin, ketotifen, cromolyn, LDN, EPO-axis workup) and anything that hits CYP3A4.
| Subtype | Priority shortlist (3–4 levers) | Rationale (mechanistic fit) + evidence tier |
|---|---|---|
| A. Glymphatic-dominant | Sleep optimisation + lateral sleep position; melatonin (0.3–1 mg, 30–60 min before bed); PBM / NIR (810–1070 nm); sulforaphane (broccoli sprouts or standardised glucoraphanin); baicalin | Slow-wave sleep and arterial pulsation drive CSF flow; melatonin shifts sleep architecture toward NREM-dominated cycles. Cytokine reduction (sulforaphane via Nrf2, baicalin via NF-κB) supports AQP4 polarisation. PBM has small human-pilot signal in post-COVID cognitive symptoms. [ESTABLISHED for sleep→glymphatic; MECHANISTIC for cytokine→AQP4; HUMAN-PILOT for PBM] |
| B. Autophagy / mTOR-dominant | Time-restricted eating (8–10 h window); berberine (500 mg, 2–3× daily); spermidine; urolithin A | Targeted mTOR modulation + autophagy/mitophagy induction without the blunt suppression that risks energy-starved neurons (the dual-mTOR trap from the main article). Fasting and berberine both activate AMPK → mTOR inhibition. Spermidine and urolithin A induce autophagy downstream. Berberin caveat: moderate CYP3A4 inhibitor — polypharmacy review mandatory (statins, SSRIs, anticoagulants, many CCBs). [ESTABLISHED for autophagy induction; MECHANISTIC for spikeopathy application] |
| C. Mitochondrial / SIRT1-PGC-1α-dominant | NAD⁺ precursors (NR 300 mg or NMN 500 mg, AM); ubiquinol (reduced CoQ10, 200–400 mg); PBM / NIR; TUDCA (500–1500 mg) | NAD⁺ is the direct SIRT1 substrate — the single most targeted lever for the core node. Without NAD⁺, SIRT1 is inactive and the entire PGC-1α → mitochondrial biogenesis axis collapses. Ubiquinol rescues electron transport directly. PBM upregulates cytochrome c oxidase. TUDCA relieves ER stress that compounds mitochondrial failure. [MECHANISTIC; HUMAN-PILOT for PBM in cognitive symptoms; HUMAN for NAD⁺ precursor effect on NAD⁺ levels but not yet on PASC endpoints] |
| D. Tolerance / EPO-EPOR-dominant | Baicalin (cGAS-STING dampening); sulforaphane (Nrf2); low-histamine diet + trigger avoidance; PBM; clinician-led review of EPO axis, iron handling, and microclot picture | Reduce antigenic drive and innate immune overactivation; support clearance without forcing a tolerance break the system can't sustain. Nattokinase/lumbrokinase excluded here — see Excluded levers. The microclot angle has its own logic at the Spike Persistence companion, but not in this framework's amyloidogenic-spike framing. EPO axis workup is clinician territory — do not self-prescribe iron or androgens. [MECHANISTIC throughout — no human RCTs for baicalin in PASC/PVS] |
| E. Mast-cell / neuroimmune-dominant | Baicalin; luteolin and/or quercetin (OTC flavonoids); strict low-histamine diet + trigger avoidance; consider H1/H2 antihistamine stacking (OTC, clinician-aware); prescription options with clinician: ketotifen, cromolyn, LDN | Mast-cell stabilisation + mediator reduction (tryptase, chymase, histamine, IL-1β, IL-33). BBB integrity preserved upstream of both clearance systems. Quercetin has small acute-COVID RCT signal. Luteolin has prior antiviral-transcription history (Mehla 2011, HIV-1 Tat abrogation) and crosses BBB. Prescription mast-cell stabilisers are well-established for MCAS but untested for PASC/PVS glymphatic or spike-clearance endpoints. [MECHANISTIC; strongest clinical phenotype correlation; NO RCTs for PASC/PVS endpoints] |
General rules across all subtypes
- Start low, go slow. Layering multiple mechanistic levers at once destroys your ability to attribute response.
- Daily symptom log. Energy, brain fog, sleep quality, flare frequency, 1–10 scale. Memory is unreliable; logs aren't.
- Re-test key surrogate biomarkers at 6–12 weeks. hsCRP, HOMA-IR, lactate profile, tryptase (if E) — whichever mapped to your dominant subtype. Single snapshots don't track change; serial measurements do.
- Mandatory polypharmacy review before starting. Berberine, baicalin, and high-dose CBD all carry CYP3A4 interaction risk. Quercetin modifies P-glycoprotein. If your clinician doesn't know what CYP3A4 is, find one who does.
- Nattokinase/lumbrokinase excluded from all five shortlists. See Excluded levers for the framework rationale. The microclot plumbing argument survives at the Spike Persistence companion — it's a context-specific lever, not a subtype-framework one.
- Rapamycin is deliberately absent from B. Prescription-only, blunt mTOR inhibitor, dual-edge risk in spikeopathy (the dual-mTOR model: hits reservoirs but starves energy-neurons). Mentioned in the main article. Not a shortlist item. If a clinician offers it, the conversation is between you and them.
Piece 4 — Honest gaps and falsifiability
Honest gaps
The five categories in Piece 1 are proposed mechanistic subtypes, not validated clinical endotypes. No published cohort has stratified PASC patients by dominant clearance or tolerance mechanism. That gap is the reason this framework exists. It's also the reason the categories rest on convergent mechanism biology rather than on demonstrated patient-level separation.
- Your symptom cluster is inferred from the underlying biology, not derived from formal cluster analysis of patient-reported data. Empirical subtyping work (Su 2022, Thaweethai 2023, Klein 2023, Wang 2026) has been run. Mechanistic subtyping has not.
- All biomarkers in Piece 2 are surrogates. None are specific to spikeopathy. Most lack validated PASC reference ranges — flagged in the table but worth restating: a "normal" NfL, GFAP, DTI-ALPS, plasma spike, IL-6, NAD⁺, or urinary mast-cell mediator result does not exclude any subtype. Track trends, not cutoffs.
- Overlap between subtypes is the rule, not the exception. Pure subtypes are rare. The framework's value is identifying the dominant failure node — not producing a tidy single-label diagnosis.
- The mast-cell / neuroimmune arm (E) has the strongest observable clinical phenotype right now (MCAS/PASC overlap is well-documented; Weinstock 2021, Wechsler 2022, Wang 2021). But direct evidence of spike protein stored inside meningeal mast cells in humans — the centrepiece of the Theoharides 2026 proposal — remains limited. This is stated honestly in the main article's mast-cell honest-gap paragraph.
What this framework is not. A diagnostic manual. A treatment protocol. A claim that the five subtypes are real, distinct, and clinically useful. It's a hypothesis about where the dominant node sits — a map to think with, not a label to attach to yourself.
Falsifiability — trackable predictions
This framework, like the main article's central model, makes specific predictions. If they fail, the framework should be revised or retired. Last updated: 14 July 2026. This section will be updated as data emerge.
Predictions that would strengthen the framework (2026–2027):
- Clinical-grade plasma spike assays (Simoa or equivalent) will detect persistent full-length or S1 spike in a meaningful fraction (≥30–50%) of well-characterised persistent PASC cases, compared with near-zero rates in matched asymptomatic recovered controls. Failure here weakens the reservoir premise that all five subtypes rest on.
- Longitudinal DTI-ALPS studies will demonstrate that interventions targeting sleep architecture and arterial pulsation produce measurable improvements in glymphatic index that correlate with symptom change specifically in glymphatic-dominant (A) patients — and produce smaller or absent responses in non-A subtypes. This is the single cleanest test of subtype-specific lever response.
- HOMA-IR and lactate biomarkers will separate subtype B from subtype C in formal cluster analysis, even where symptoms overlap. If B and C are indistinguishable on these markers across cohorts, they should be merged.
Predictions that would strengthen the framework (2027+):
- Single-cell or tissue-level studies will confirm or refute EPOR-mediated tolerance bias in subtype D patients and mast-cell spike storage in subtype E patients. Both are currently mechanism-inferred, not tissue-confirmed.
- Formal cluster analyses using mechanistic biomarkers (DTI-ALPS, mitochondrial complex activity, tryptase dynamics, HOMA-IR, NfL) will either validate, merge, or replace these five working categories. If the data consistently produce four, six, or seven clusters with different boundaries, the framework is wrong about where the dominant nodes sit.
What would kill the framework:
- A well-powered, multi-centre cohort finds no mechanistic separation between putative subtypes on biomarker panels — symptom clusters exist but map to no underlying mechanism signature. The framework becomes a relabelling exercise with no predictive value.
- Targeted interventions do not move both surrogate biomarkers and symptoms in subtype-specific fashion. If a glymphatic-dominant patient's DTI-ALPS improves on sleep+PBM but symptoms don't, or symptoms improve but DTI-ALPS doesn't, the lever→mechanism→symptom chain is broken at some link and the framework's clinical utility collapses.
- A competing framework (e.g., purely autoimmune, purely dysautonomia, purely microbiome-driven) accounts for the same patient population with fewer assumptions and superior predictive power. The framework should then be retired in favour of the better model, not defended.
Why this matters. A framework that cannot be falsified is a belief system, not a hypothesis. The value of writing the predictions down is that they can be tested, and if they fail, the framework goes. This is the discipline that separates mechanism-based stratification from the endless "try everything" literature.
[Piece 5 (summary table + main-article cross-link) — applied to the main article, not this companion. Article remains draft: true until reviewed and the publish step (flip draft, activate stratification-companion link in nattokinase row) is run.]
Monitoring plus mechanistic support. Not treatment or diagnosis. Work with a clinician who understands the data.