Part 4: The Pharmacopoeia — Every Dose Range for Every Medication, and What It Means
Anyone who has been through the trial-and-error cycle enough times knows the pattern by heart: start a drug, notice something, titrate up, lose the something, titrate down, get the something back. The doctor records “responder” or “non-responder” and moves on — a binary that throws away most of the information. Where on the dose axis a response appeared, and where it disappeared, is the diagnostic signal. The rest is lossy compression.
The three previous posts established why: the eighteen drugs with inverted-U dose-response curves fall into four mechanistically distinct categories, and the dose at which benefit appears, peaks, and inverts can reveal which mechanism is rate-limiting. This post is the patient-facing companion to the consolidated reference table now in the primary paper (ch32, sec-14). What follows is a walk through every drug, every clinically relevant dose band, and what each band may mean — benefit, loss of benefit, side effects, and silence all carry information.
Two caveats before we begin. First, every cross-drug inference here is untested. The cluster-correlation claim — that Nrf2 drugs’ windows should track each other but not catecholamine drugs’ windows — is the core falsifiable hypothesis, and it awaits the HIP-B trial. These are empirical bets, not established fact. Second, the framework may be wrong. The dose at which a drug works, and the dose at which it stops working, remains information about biology regardless of whether the interpretive layer on top of it holds up. The dose-response curve is yours. The theory around it is provisional.
1 Category 1: Nrf2-mediated hormesis
These drugs trigger anti-inflammatory and antioxidant gene expression through the Keap1-Nrf2-ARE pathway. The benefit is the cell’s own adaptive response, not the drug’s direct action. Too little drug fails to fire the stress signal; too much extinguishes it or overwhelms the cell’s compensatory machinery. The therapeutic window is therefore narrow by design.
1.1 LDN (Low-Dose Naltrexone)
LDN hits four targets, each at a different plasma concentration: TLR4/Nrf2 hormetic priming (peaks at 0.5–1.5 mg), opioid compensatory upregulation (plateaus by 1.5–3.0 mg), TRPM3 ion channel restoration (likely 3.0–4.5 mg), and orexin disinhibition (tracking the TLR4 curve). The right dose depends entirely on which mechanism is dominant — see Part 1 for the full four-mechanism cascade.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.25–0.5 mg | Micro-dose probe | If it works: High sensitivity; a narrow window across all Nrf2-cluster drugs is likely. If it doesn’t: The triggering threshold may not be reached — try 1.0–1.5 mg. Or baseline oxidative tone is low and a larger perturbation is needed to fire the sensor. Side effects: Flare at micro-dose points to homeostatic overshoot (see Part 3): the system’s buffering capacity is too low to contain the adaptive response. Try 0.1 mg sublingual drops; if the flare persists, dose-finding is premature. |
| 0.5–1.5 mg | TLR4/Nrf2 hormetic priming | If it works: TLR4-driven neuroinflammation is rate-limiting and the microglia are Nrf2-responsive. The width of this window (benefit lost at 1.0 mg vs. persisting to 1.5 mg) estimates Nrf2 reserve. If it doesn’t: Benefit lost below 1.5 mg means narrow Nrf2 reserve. No benefit at all suggests TLR4 is not the rate-limiting input on the microglia. Side effects: Vivid dreams, initial sleep disruption — opioid receptor engagement detectable even at micro-doses. Usually transient (1–2 weeks). Persistent beyond 3 weeks: try 0.25 mg morning + 0.25 mg evening. |
| 1.5–3.0 mg | Transition zone: TLR4/Nrf2 fading, opioid compensation rising | If it works: Maintained benefit suggests opioid upregulation is contributing or Nrf2 reserve is broad enough to sustain priming. If it doesn’t: Loss at 1.5–2.0 mg is the most common pattern — a pure TLR4/Nrf2-dominant response. Side effects: “Wired” feeling suggests orexin disinhibition engaging. If tolerable, continue. |
| 3.0–4.5 mg | TRPM3 restoration, opioid plateau, orexin disinhibition. TLR4/Nrf2 extinguished | If it works: Benefit appearing here for the first time points to TRPM3 channelopathy or orexin deficiency. Benefit maintained from lower doses means opioid compensation is the dominant driver. If it doesn’t: No benefit here (and 0.5–1.5 mg also failed) means none of LDN’s four mechanisms is rate-limiting — the differential narrows to D2-pathway (LDA), autoantibodies, B-cell pathology, MCAS, or structural causes. Side effects: Excessive wakefulness; when coupled with loss of anti-inflammatory benefit, the net effect is neutral-to-negative. |
| 4.5–9.0 mg | All mechanisms past optima; approaching mu-opioid antagonism | If it works: No mechanism has an optimum above 4.5 mg — any benefit was already achieved at 4.5. If it doesn’t: Expected — all four mechanisms are past their dose optima. The drug hasn’t failed; the dose is too high. Side effects: Irritability, dysphoria — mu antagonism engaging. Do not escalate further. |
| ≥50 mg | Full mu-opioid antagonist | If it works: No therapeutic rationale in ME/CFS. If it doesn’t: All LDN-specific mechanisms extinguished. Side effects: Anhedonia, worsened pain — stop, return to the effective low dose. |
What it implies for other drugs: Benefit in the 0.5–1.5 mg range is consistent with response to other Nrf2 activators (sulforaphane, low-dose melatonin, low-dose lithium, quercetin). Benefit only at 3.0–4.5 mg suggests Nrf2 activators are unlikely to help and TRPM3-directed interventions (microdose lithium for PIP₂ stabilization) or orexin-directed agents deserve consideration.
1.2 Low-Dose Lithium (0.3–5 mg element lithium)
Lithium works through GSK-3β inhibition, which derepresses Nrf2. At higher concentrations, IMPase inhibition enters the picture, creating a biphasic two-target dynamic — the same logic as rapamycin’s mTORC1/C2 but with a different enzyme pair.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.3–1.0 mg | GSK-3β partial inhibition, mild Nrf2/ARE activation | If it works: Broad Nrf2 reserve — even minimal GSK-3β inhibition suffices. Consistent with broad windows for the rest of the cluster. If it doesn’t: Threshold not reached. Does not mean GSK-3β is irrelevant. Side effects: Essentially none; below the side-effect threshold for any known lithium mechanism. |
| 1.0–5.0 mg | Dose-dependent GSK-3β inhibition, Nrf2 activation; IMPase inhibition begins | If it works: GSK-3β is rate-limiting for Nrf2 activation. Benefit lost above this range suggests IMPase sensitivity limits the therapeutic ceiling. If it doesn’t: No benefit at any dose suggests GSK-3β is not the rate-limiting Nrf2 repressor, or IMPase inhibition is offsetting the benefit through PIP₂ depletion — impossible to separate without PI measurement. Side effects: Mild cognitive dulling (rare) — lithium should be stopped, as GSK-3β is not then a viable target. Thirst or polyuria at these doses: discontinue. |
| 5–15 mg | IMPase inhibition dominates | If it works: Benefit only here means the mechanism is not GSK-3β/Nrf2 — possibly IP₃ modulation or PIP₂ dynamics. If it doesn’t: Loss from a lower effective dose means IMPase inhibition is counteracting the Nrf2 benefit. Side effects: Cognitive dulling, emotional blunting, possible thyroid effects with prolonged use. |
| ≥150 mg | Psychiatric doses | If it works: Not applicable to ME/CFS hormesis. If it doesn’t: Expected. Side effects: Full lithium toxicity profile (tremor, nephrotoxicity, hypothyroidism). Do not use. |
What it implies for other drugs: A lithium window narrower than the LDN window implicates IMPase sensitivity rather than Nrf2 reserve as the limiting factor. A lithium window broader than the LDN window suggests TLR4 tone limits LDN specifically, not Nrf2 reserve globally.
1.3 Melatonin (0.3–5 mg)
Melatonin has two separable effects: a chronobiotic (circadian) signal that is dose-independent, and Nrf2 activation via MT1/MT2 receptors that is hormetic. Interpreting the dose-response curve requires distinguishing which of these is producing the benefit.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.1–0.3 mg | Ultra-low-dose circadian probe | If it works: Split-dose resolution of grogginess at higher doses — CYP1A2 slow-metabolizer pattern. If it doesn’t: Chronobiotic threshold may not be reached. Side effects: Essentially none; below the side-effect threshold. |
| 0.3–1.0 mg | Physiological: circadian phase-advance + mild Nrf2 activation via MT1/MT2 | If it works: Sleep-onset benefit only → circadian mechanism (no Nrf2 inference). Anti-inflammatory benefit (reduced malaise, clearer cognition) → Nrf2 pathway contributing, consistent with response to other cluster drugs. If it doesn’t: Circadian-only benefit means Nrf2 is not contributing — cluster inference does not apply. Side effects: Morning grogginess → slow CYP1A2 metabolism. Split dose or drop to 0.1–0.3 mg. |
| 1.0–3.0 mg | Supraphysiological: MT1/MT2 + direct antioxidant effects | If it works: Benefit exceeding what 0.3–1.0 mg achieved → direct antioxidant effects contributing, or rapid metabolizer. If it doesn’t: Benefit declining from 1 mg → MT1/MT2 desensitization; supraphysiological signal lacks circadian specificity. Side effects: None beyond dose-dependent sedation. |
| 3.0–5.0 mg | High-dose: MT1/MT2 desensitized, Nrf2 component extinguished | If it works: Only chronobiotic effect on sleep onset may persist — Nrf2-mediated anti-inflammatory benefit likely lost. If it doesn’t: Expected partial loss — the hormetic anti-inflammatory component is absent. Side effects: Sedation, no therapeutic gain over lower doses. |
| >5 mg | Sedation only | If it works: Placebo at best through the hormetic framework. If it doesn’t: Expected. Side effects: Dose-dependent sedation, no mechanistic rationale. |
1.4 Sulforaphane (broccoli sprout extract, glucoraphanin-standardized)
Sulforaphane directly modifies Keap1 cysteine residues to release Nrf2 — it is the purest Nrf2 probe in the cluster.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 10–30 mg glucoraphanin | Keap1 cysteine modification, Nrf2 release | If it works: Exceptionally sensitive Nrf2 sensor, ample Nrf2 reserve. Narrow window here → Nrf2 cluster likely the bottleneck. If it doesn’t: Threshold not reached; escalate. Side effects: GI discomfort → glucosinolate metabolism issue; persistence → gut microbiome lacks myrosinase-producing bacteria. |
| 30–60 mg glucoraphanin | Dose-dependent Keap1 suppression | If it works: Moderate Nrf2 degradation rate; GSK-3β tone within normal range. If it doesn’t: Benefit lost by 60 mg → narrow Nrf2 reserve, all cluster drugs should show narrow correlated windows. Side effects: GI symptoms as above. |
| 60–100 mg glucoraphanin | Near-complete Keap1 suppression | If it works: High Nrf2 degradation rates (Keap1 overexpression or GSK-3β hyperactivity); requires near-complete Keap1 suppression. If it doesn’t: Loss is a saturation plateau (Keap1 fully modified), not true hormetic inversion. Side effects: GI symptoms as above. |
| No dose-response | Nrf2 reserve flat | If it works: Maximally activated or inactivated beyond rescue. If it doesn’t: Cluster drugs unlikely to help. Side effects: — |
1.5 NAC (N-Acetylcysteine)
NAC’s dose-response is dominated by a biochemical quirk: at 600 mg, a thiol-radical burst can exhaust glutathione before net synthesis kicks in, producing transient paradoxical worsening in roughly 20–30% of patients. This is a pharmacokinetic event, not a drug failure.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 600 mg | Thiol-radical burst, glutathione synthesis initiation | If it works: Benefit sustained → paradoxical worsening absent; low oxidative tone. If it doesn’t: Transient paradoxical worsening expected in 20–30% — a pharmacokinetic event, not a drug failure. Side effects: GI upset, transient worsening (oxidative burst exhausts glutathione before synthesis). |
| 1,200 mg | Glutathione synthesis, redox-sensitive Keap1 Nrf2 activation | If it works: Glutathione synthesis pathway intact; Nrf2 activation through redox-sensitive Keap1 functional. If it doesn’t: Nrf2 activation rate-limited downstream; Keap1 modifiers (sulforaphane) or GSK-3β inhibitors (lithium) bypassing glutathione may still work. Side effects: GI upset. |
| No benefit at any dose | Glutathione synthesis or Nrf2 activation rate-limited downstream | If it works: — If it doesn’t: Keap1 modifiers (sulforaphane) or GSK-3β inhibitors (lithium) may still bypass the block. Side effects: GI upset. |
1.6 Quercetin
Quercetin straddles two categories — COMT inhibition (catecholamine cluster) and Nrf2 activation (Nrf2 cluster) — and the dose range reveals which mechanism is driving the benefit.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 250–500 mg | COMT inhibition (catecholamine-cluster) + Nrf2 activation via PI3K/Akt (Nrf2-cluster) | If it works: Cognitive benefit → catecholamine-cluster signal. Anti-inflammatory benefit → Nrf2-cluster signal. If it doesn’t: Either mechanism may be insufficiently engaged — escalate to isolate. Side effects: Headache → assess COMT genotype: Val/Val may need more; Met/Met may overshoot at any dose. |
| 500–1,000 mg | Both mechanisms engaged at moderate intensity | If it works: Broad benefit → both COMT and Nrf2 contributing. If it doesn’t: Loss here → narrow window spanning both mechanisms. Side effects: Monitor for noradrenergic signs (tachycardia, hypertension). |
| 1,000–2,000 mg | Both mechanisms at peak → risk of overshoot | If it works: Sustained benefit → broad COMT and Nrf2 reserve. If it doesn’t: COMT inhibition overshoot → noradrenergic excess. Catechol moiety auto-oxidation → pro-oxidant stress. Side effects: Hypertension, tachycardia, headache, paradoxical inflammation. |
2 Category 2: Partial-agonist inverted-U
This is not hormesis — no compensatory upregulation, no stress signal, no adaptive response. Aripiprazole has roughly 25% intrinsic activity at the D2 receptor compared to dopamine’s 100%. At low occupancy, it supplements a deficient endogenous signal (net agonism). At higher occupancy, it displaces the more effective natural ligand (net antagonism). The curve is identical in any population with dopamine deficit, not ME/CFS-specific, but the occupancy at which the inversion occurs reveals D2 receptor reserve.
2.1 LDA (Low-Dose Aripiprazole, 0.2–2 mg)
LDA acts at three levels, all below 2 mg: microglial D2/D3 (≤1 mg, raises activation threshold, suppresses cytokine release), mesocorticolimbic D2/D3 (≤2 mg, restores cognitive and motivational dopamine tone), and 5-HT1A autoreceptors (≤2 mg, stabilizes serotonergic output). All three are extinguished above 2 mg as receptor engagements shift to a standard antipsychotic profile (5-HT2A, H1, α1).
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.2–0.5 mg | Microglial D2/D3 partial agonism | If it works: Highly sensitive microglial D2 receptors. Microglial D2/D3 activation likely rate-limiting. If it doesn’t: Dosing problem, not a D2 problem — escalate, don’t conclude D2 non-involvement. Side effects: Akathisia/restlessness → paradoxical: presynaptic autoreceptor engagement, transient dopamine dip before postsynaptic agonism takes over. |
| 0.5–1.0 mg | Microglial D2/D3 sustained, mesocorticolimbic D2/D3 begins | If it works: Microglial-driven inflammation is dominant; cognitive benefit emerging. If it doesn’t: Benefit lost by 1.0 mg → narrow D2 receptor reserve. Side effects: Restlessness, activation. |
| 1.0–2.0 mg | Mesocorticolimbic D2/D3 + 5-HT1A autoreceptors | If it works: Broad D2 reserve. Dopamine system unlikely rate-limiting. Serotonergic stabilization contributing. If it doesn’t: Moderate D2 reserve — benefit window is dose-constrained. Side effects: Sedation, weight gain beginning to appear. |
| >2 mg | Standard antipsychotic profile (5-HT2A, H1, α1) dominates; D2 partial agonism extinguished | If it works: No therapeutic rationale. If it doesn’t: Expected — all low-dose mechanisms extinguished. Side effects: Akathisia, sedation, weight gain, metabolic effects. LDA window closed. |
Load-bearing clinical logic: LDA non-response followed by LDN response means microglial involvement via TLR4, not D2. LDN non-response followed by LDA response means the reverse. Both patterns confirm microglial involvement through different receptors — the two drugs are mechanistically complementary.
3 Category 3: Catecholamine inverted-U at prefrontal cortex
The inverted-U is the native operating curve of D1 and α2A circuits in the prefrontal cortex — established neuroscience, across species and methods. These drugs push the system along its existing curve. The inversion point reveals where baseline catecholamine tone sits. Drugs that increase tone (modafinil, duloxetine) push from the left arm toward the peak; drugs that decrease tone (beta-blockers, guanfacine) push from the right arm toward the peak. Someone who benefits from both a low-dose elevator and a low-dose reducer has a paradoxically broad peak — a wide range of tone is tolerated.
3.1 Modafinil
Modafinil blocks DAT and NET, raising synaptic dopamine and norepinephrine.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 12.5–25 mg | Micro-dose DAT/NET blockade probe | If it works: Extremely high sensitivity to dopamine/norepinephrine elevation. Consistent with narrow catecholamine-cluster windows. If it doesn’t: Threshold not reached; baseline dopamine may be very high. Side effects: Essentially none. |
| 25–50 mg | DAT/NET blockade, catecholamine elevation | If it works: Low-normal baseline dopamine; narrow catecholamine-cluster windows expected. Benefit in this range → low inversion point. If it doesn’t: No benefit → not a dopamine-deficit problem or baseline too high. Side effects: Headache, anxiety (sympathetic activation). |
| 50–100 mg | Moderate DAT/NET blockade | If it works: Moderate baseline dopamine; standard therapeutic range. If it doesn’t: Benefit lost by 100 mg → narrow reserve, confirmatory of low inversion point. Side effects: Headache, anxiety, tachycardia. |
| 100–200 mg | Full DAT/NET blockade | If it works: Higher baseline dopamine, broader windows. If it doesn’t: — Side effects: Sympathetic overactivation, insomnia. |
| 200–400 mg | Supra-therapeutic DAT/NET blockade | If it works: Sustained benefit → high baseline dopamine, very broad windows. Rare. If it doesn’t: No benefit at any dose → guanfacine (α2A, distinct from DAT) may work where modafinil doesn’t. Side effects: Anxiety, insomnia, cardiovascular strain. Not recommended as first-line. |
Key cross-drug test: LDN × modafinil inversion point correlation tests whether hormetic reserve is a genuine systems trait spanning completely distinct receptors (TLR4 vs. DAT/NET) and cell types (microglia vs. presynaptic terminals).
3.2 Duloxetine
Duloxetine has two dose-response profiles folded into one drug: SERT inhibition (serotonergic, monotonic saturation — no inverted-U) and NET inhibition (noradrenergic, catecholamine inverted-U).
| Dose | Mechanism | Interpretation |
|---|---|---|
| 10–20 mg | SERT inhibition (serotonergic only) | If it works: Pain modulation, sleep improvement. Serotonergic-dominant pattern. If it doesn’t: NET threshold not reached — no catecholamine inference. Side effects: SNRI discontinuation syndrome if stopped abruptly; taper ≥3 weeks. |
| 20–40 mg | SERT saturation + NET inhibition engaging | If it works: Both serotonergic and noradrenergic benefits. Energy and motivation appearing → NET component engaged. If it doesn’t: Energy not emerging → NET not yet reached or baseline NE already at peak. Side effects: GI, activation, insomnia. |
| 40–60 mg | Full SERT + NET at peak therapeutic range | If it works: Broad serotonergic and catecholamine reserve. If it doesn’t: Energy/motivation peaking here means moderate catecholamine reserve. Side effects: Increased side effects; sweat, BP elevation. |
| 60–90 mg | SERT plateau, NET overshoot risk | If it works: Pain benefit may persist (SERT on plateau) while energy benefit inverts (NET on right arm) — mechanisms successfully dissociated. If it doesn’t: Energy benefit lost → NET on right arm of inverted-U while SERT remains on plateau. The drug hasn’t failed; the NET component is past optimum. Side effects: Dose-dependent sympathetic activation. |
| 90–120 mg | Both SERT and NET supra-therapeutic | If it works: No additional benefit over 60 mg. If it doesn’t: Expected loss of noradrenergic benefit. Side effects: Maximum side-effect burden; SNRI discontinuation syndrome risk increases. |
3.3 Beta-Blockers
Unlike modafinil and duloxetine, which increase catecholamine tone, beta-blockers reduce it — pushing from the right arm toward the peak.
| Dose | Mechanism | Interpretation |
|---|---|---|
| Propranolol 5–10 mg or Atenolol 12.5–25 mg | Peripheral β-AR blockade | If it works: Orthostatic tachycardia controlled without cognitive dulling → peripheral sympathetic excess with intact central tone. If it doesn’t: Peripheral β-AR contribution may be minimal — not an autonomic problem. Side effects: Minimal at these doses. |
| Propranolol 10–20 mg | Peripheral + central β-AR modulation | If it works: Cognitive improvement with orthostatic control → central β-AR modulation moving tone toward the peak; baseline tone was excessive. If it doesn’t: Cognitive dulling at any dose → baseline tone is low, not high; beta-blockade pushes it further left. Side effects: Fatigue, brain fog, bradycardia. |
| Propranolol >20 mg | Non-selective central β-AR blockade | If it works: Sustained orthostatic control without cognitive cost → broad catecholamine reserve. If it doesn’t: Expected cognitive dulling. Combining with guanfacine or clonidine is counterproductive — the system needs more catecholamine tone, not less. Side effects: Cognitive dulling, fatigue, depression, bradycardia. |
3.4 Guanfacine
Guanfacine’s inverted-U is built into the ratio of postsynaptic (therapeutic) to presynaptic (sedating) α2A agonism.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.5–1.0 mg | Postsynaptic α2A dominates | If it works: Improved working memory, focus, emotional regulation. Postsynaptic effects dominate at therapeutic doses. If it doesn’t: Sedation without cognitive improvement → presynaptic α2A agonism dominant even at low levels; guanfacine is probably the wrong drug. Side effects: Sedation, dry mouth, hypotension. |
| 1.0–2.0 mg | Postsynaptic α2A + emerging presynaptic effects | If it works: Sustained cognitive benefit → broad α2A reserve. Correlated windows with modafinil support catecholamine-cluster coherence (circuit overlap with modafinil’s D1-mediated effect). If it doesn’t: Cognitive dulling without therapeutic gain → presynaptic inhibition has overtaken postsynaptic benefit. Side effects: Sedation, bradycardia, hypotension. |
| >2 mg | Presynaptic α2A sedation dominates | If it works: No additional therapeutic gain expected. If it doesn’t: Expected — all therapeutic α2A effects past optimum. Side effects: Pronounced sedation, bradycardia, hypotension. |
3.5 Gabapentinoids
Gabapentinoids are not a clean catecholamine-cluster probe — they reduce glutamate release broadly, not just noradrenergic tone.
| Dose | Mechanism | Interpretation |
|---|---|---|
| Gabapentin 100–300 mg or Pregabalin 25–75 mg | α2δ-mediated sensory gating | If it works: Sensory gating improvement (α2δ-mediated) rather than optimal PFC catecholamine tone — limits cross-drug inferential value. If it doesn’t: Threshold not reached; escalate carefully. Side effects: Cognitive dulling at any dose is expected and dose-dependent. |
| Gabapentin 300–900 mg or Pregabalin 75–150 mg | Broad glutamate reduction | If it works: Broader benefit may involve noradrenergic tone reduction; catecholamine-cluster inference is confounded by glutamatergic effects. If it doesn’t: — Side effects: Dose-dependent sedation, cognitive dulling. |
| Gabapentin >900 mg or Pregabalin >150 mg | Sedation dominates | If it works: No additional therapeutic gain expected. If it doesn’t: Sedation overwrites any diagnostic signal. Side effects: Pronounced sedation, cognitive impairment. Not recommended for diagnostic dose-finding. |
4 Category 4: Concentration-dependent target selection and biphasic biochemistry
These drugs share the inverted-U curve shape with the previous categories but arise from mechanisms specific to each compound: differential binding affinity across targets (4a), pharmacokinetic thresholds involving CNS penetration (4b), or biphasic metabolism (4c). They carry no cross-drug predictive value within this category, though some offer pathway-specific inferences.
4.1 Rapamycin (sirolimus) — Category 4a
Rapamycin inhibits mTORC1 with roughly 10-fold higher affinity than mTORC2. At intermittent low doses (0.5–2 mg/week), mTORC1 is partially blocked — autophagy is restored, mitophagy improves, and senescent-cell SASP is suppressed. At higher weekly doses (3–6 mg) or daily dosing, mTORC2 inhibition kicks in, producing insulin resistance and immunosuppression. The therapeutic window reflects the FKBP12-mTOR binding affinity ratio, not a host-tissue hormetic response, which is why weekly intermittent dosing preserves selectivity — mTORC2 has a longer off-rate and recovers during the drug-free interval.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.5–2 mg/week (intermittent) | mTORC1-selective inhibition | If it works: mTORC1-driven autophagy failure is rate-limiting. Autophagy restored, mitophagy improved, SASP suppressed. If it doesn’t (mTORC1 suspected): Autophagy machinery downstream of mTORC1 may be broken — the gate opens, but the pathway behind it is inoperable. Side effects: Minimal at intermittent low doses. |
| 3–6 mg/week or daily dosing | mTORC2 inhibition engaging | If it works: Benefit from lower dose lost → mTORC2 selectivity breached; insulin resistance and immunosuppression offset autophagy gains. If it doesn’t: Expected loss from mTORC1 window. Side effects: Mouth ulcers, lipid/glucose elevations — mTORC2 inhibition has breached the therapeutic window. |
| Daily dosing, standard doses | Full mTORC1 + mTORC2 inhibition | If it works: No therapeutic rationale (immunosuppression is the primary effect). If it doesn’t: Expected. Side effects: Insulin resistance, immunosuppression, impaired wound healing. |
Rapamycin’s window width says nothing about LDN or modafinil windows. It may, however, predict response to other mTORC1-selective interventions: intermittent fasting, metformin, or spermidine that induce autophagy through mTORC1 without engaging mTORC2 at all.
4.2 Corticosteroids — Category 4a
Corticosteroids are not a single drug with one dose-response curve — they are two drugs separated by a concentration threshold. Physiological replacement provides the anti-inflammatory glucocorticoid signal a hypofunctional HPA axis fails to produce endogenously. Supraphysiological dosing suppresses the HPA axis itself, and the taper produces rebound inflammation often worse than baseline.
| Dose | Mechanism | Interpretation |
|---|---|---|
| Hydrocortisone 5–10 mg/day or Prednisone 2.5–5 mg/day | Physiological glucocorticoid replacement | If it works: Dramatic response → trigger ACTH/cortisol workup for adrenal insufficiency. The HPA axis is hypofunctional. If it doesn’t: Physiological replacement insufficient — underlying inflammation is not GR-mediated or the HPA axis is intact. Side effects: Minimal at replacement doses. Taper will be difficult because the underlying HPA hypofunction is still present. |
| Prednisone 5–15 mg/day | Supraphysiological: HPA suppression threshold | If it works: Glucocorticoid resistance or non-GR-mediated inflammation — mechanistically informative but clinically dangerous; therapeutic margin narrows as HPA suppression deepens. If it doesn’t: May need higher dose — but risk-benefit calculus shifts rapidly. Side effects: HPA axis suppression begins. Taper will be difficult because the axis has been suppressed and needs time to recover. |
| Prednisone >15 mg/day | Full HPA suppression + systemic glucocorticoid effects | If it works: Benefit validates a glucocorticoid-responsive mechanism, but at high cost. If it doesn’t: Expected — side-effect burden limits utility. Side effects: HPA suppression, osteoporosis risk, immunosuppression, weight gain, glucose dysregulation. Rebound inflammation on taper often worse than baseline. |
4.3 DORAs (Dual Orexin Receptor Antagonists) — Category 4b
DORAs sit on a pharmacokinetic cliff. Partial orexin blockade improves sleep by quieting the wakefulness drive. Near-complete blockade can overshoot into sleep paralysis and sleep-onset hallucinations — a narcolepsy-mimic that degrades subjective sleep quality even as total sleep time increases.
| Dose | Mechanism | Interpretation |
|---|---|---|
| Daridorexant 25 mg or Suvorexant 10 mg | Partial orexin blockade | If it works: Sleep improved by quieting the wakefulness drive without overshoot. Appropriate orexin tone reduction. If it doesn’t: Insufficient blockade; partial antagonism not reaching therapeutic threshold. Or orexin is not the rate-limiting wake-drive signal. Side effects: Minimal at these doses. |
| Daridorexant 50 mg or Suvorexant 20 mg | Near-complete orexin blockade | If it works: High orexin tone requiring near-complete blockade for sleep onset. If it doesn’t: Overshoot into sleep paralysis and sleep-onset hallucinations — a narcolepsy-mimic degrading subjective sleep quality. Side effects: Sleep paralysis, sleep-onset hallucinations, next-day somnolence. Daridorexant (t½ = 8h) preferred over suvorexant (t½ = 12h) to minimize next-day carryover. |
4.4 H1 Antihistamines — Category 4b
The clinical question is whether the antihistamine is benefiting because of peripheral H1 blockade (MCAS symptom control) or because of central H1 blockade (sedation, anxiolysis). Fexofenadine answers this: it is a P-glycoprotein substrate actively effluxed from the CNS, with near-zero brain penetration at any dose.
| Dose / Agent | Mechanism | Interpretation |
|---|---|---|
| Fexofenadine 60–180 mg | Peripheral H1 blockade only (no CNS penetration) | If it works: Symptoms controlled → target is peripheral (MCAS). Confirms peripheral histamine is the driver. If it doesn’t: Peripheral H1 is either not the cause, or dose is insufficient. Side effects: Minimal; no sedation (no CNS penetration). |
| Cetirizine 5–10 mg | Peripheral + central H1 blockade | If it works and fexofenadine didn’t: Central effects are contributing — perceived benefit may not be through peripheral H1 at all. Anxiolysis/sedation may be the actual mechanism. If it doesn’t: H1 is not the rate-limiting pathway. Side effects: Sedation (CNS-penetrant). |
4.5 Allopregnanolone — Category 4c
Allopregnanolone is biphasic at the GABA-A receptor because it likely binds two sites with opposing concentration-response curves. The dose-response is highly individual, depending on GABA-A subunit composition, hormonal state, and stress history.
| Concentration / Context | Mechanism | Interpretation |
|---|---|---|
| ~1–5 nM (luteal-phase equivalent) | GABA-A site 1: paradoxical anxiogenesis | If it causes anxiety: Paradoxically anxiogenic at low concentrations — a known phenomenon in PMDD. GABA-A subunit composition favors excitatory response at low allopregnanolone. If it doesn’t cause anxiety: GABA-A receptor subtype profile favors inhibition even at low concentrations. Side effects: Anxiety, agitation, irritability. |
| >10 nM (pregnancy-level) | GABA-A site 2: anxiolysis and sedation | If it works: Anxiolytic and sedating at higher concentrations. GABA-A receptor profile supports inhibition at these levels. If it doesn’t: GABA-A receptor insensitivity or competing excitatory tone. Side effects: Sedation, tolerance with prolonged use. |
4.6 Ketotifen — Category 4b
Ketotifen carries two independent therapeutic actions: mast-cell stabilization (non-H1) and H1 blockade.
| Dose | Mechanism | Interpretation |
|---|---|---|
| 0.25–1.0 mg | Mast-cell stabilization dominates; CNS penetration minimal | If it works: MCAS benefit without sedation → confirms mast-cell stabilization mechanism. Dose-dependent and sedation-limited response confirms mast-cell involvement. If it doesn’t: Sedation even at 0.25 mg → high CNS penetration sensitivity, argues for peripheral-only interventions. Side effects: Minimal; sedation at 0.25 mg is itself diagnostic — points to high CNS penetration sensitivity. |
| 1.0–2.0 mg | Mast-cell stabilization + CNS H1 blockade | If it works: Additional MCAS benefit at cost of sedation — acceptable trade-off if mast-cell stabilization dose-dependent. If it doesn’t: CNS H1 blockade produces clinically significant sedation in most people, offsetting mast-cell gains. Side effects: Clinically significant sedation in most people. |
| 2.0–4.0 mg | CNS H1 blockade dominates; sedation cost | If it works: No additional benefit over 1.0–2.0 mg. If it doesn’t: Sedation cost exceeds the mast-cell stabilization gain — net benefit inverts. Side effects: Pronounced sedation; gains from mast-cell stabilization overwritten. |
5 Reading the pattern across drugs
The diagnostic yield is not in any single drug’s dose-response curve. It emerges from comparing across all the drugs a person has tried.
Nrf2-cluster drugs all showing narrow windows — benefit lost at low doses across LDN, sulforaphane, lithium, melatonin — points to depleted Nrf2 transcriptional reserve. There is a significant Nrf2-dependent inflammatory component, but the adaptive machinery is exhausted. The strategy shifts toward Nrf2-sparing interventions: reduce allostatic load, avoid drugs depending on Nrf2 for benefit, consider sulforaphane as preconditioning before other cluster drugs. The mirror — broad windows throughout — means Nrf2 reserve is ample, and Nrf2 is probably not the bottleneck.
The same logic applies to the catecholamine cluster. Narrow windows everywhere imply prefrontal catecholamine dynamic range is compressed, with baseline tone sitting close to the inverted-U peak; small perturbations in either direction cause trouble. Dosing must be precise for all catecholamine-active drugs, and the patient is simultaneously sensitive to drugs that raise tone and drugs that lower it. The opposite constellation — high doses needed across the cluster, no inversions at clinical doses — points to very low baseline tone, monotonic-increasing benefit, and doses that would invert for others being well-tolerated.
A narrow rapamycin window alongside broad Nrf2 and catecholamine windows is a drug-specific signal: mTORC1/C2 selectivity is tight, but hormetic reserve in other systems is intact. mTORC1-driven autophagy may still be a valid target through precisely dosed rapamycin or mTORC1-selective alternatives (intermittent fasting, metformin). No dose-response relationship in any cluster — all drugs producing flat or monotonic curves — is an informative null: the framework has identified no accessible target, and investigation should turn to non-pharmacological or structural causes.
Mixed patterns localize the lesion. Narrow Nrf2 with broad catecholamine windows implicates the immune-inflammatory system as the bottleneck while cognitive reserve is preserved — the expected pattern for TLR4-dominant neuroinflammation. Anti-inflammatory drugs get the careful titration; cognitive drugs are dosed more freely. The reverse — broad Nrf2 with narrow catecholamine — points to catecholamine-dominant pathophysiology (prominent orthostatic intolerance, low supine norepinephrine, COMT Val/Val), reversing the dosing priorities. Every cluster narrow, all drugs inverting at very low doses, suggests global adaptive compromise: cells at the edge of decompensation, any pharmacological perturbation risking overshoot. The strategy is ultra-low, ultra-slow — stabilize, don’t aim for improvement. The narrowness is the severity indicator.
6 Certainty
| Claim | Certainty |
|---|---|
| The four-category mechanistic decomposition | Moderate (Category 1: Nrf2 hormesis in toxicology, not clinical dosing); High (Category 2: receptor-occupancy property); High (Category 3: canonical neuroscience); Variable (Category 4: subtype-dependent) |
| Dose-range interpretations are mechanistically accurate | Low-to-moderate — grounded in each drug’s target engagement, but clinical relevance in ME/CFS is unestablished |
| Cross-drug cluster correlation predictions | Low — the core falsifiable hypothesis, never tested |
| “No benefit at any dose” rules out a drug’s mechanisms as rate-limiting | Moderate — valid only when “no benefit” comes from systematic dose-finding, not a single dose |
| Cluster assignment predicts subtype and treatment priority | Low — coherent, untested; awaiting HIP-B |
| Global narrowness indicates severe adaptive compromise | Low-to-moderate — consistent with the framework but indistinguishable from coincident drug-specific sensitivities without correlation data |
This post draws on the hormesis framework developed in Loth 2026 (https://yannickloth.github.io/health-me-cfs/). The four-category mechanistic decomposition (Part 2) and five LDN dose-response patterns (Part 3) provide the conceptual architecture. Specific dose ranges and interpretations are synthesized from the 17-drug hormetic-inversion framework, the HIP-B trial specification, the mechanistic clustering analysis, and clinical observation — see the primary paper for full mechanistic cascades.
Series: Part 1: Why More Isn’t Better · Part 2: The Inverted-U Is Not One Thing · Part 3: Your LDN Dose Is a Diagnostic