Immune-Modulating Medications
1 Low-Dose Naltrexone (LDN)
Low-dose naltrexone (LDN) has emerged as one of the most commonly used off-label treatments for ME/CFS, despite limited controlled trial data.
1.1 Mechanism of Action
Naltrexone at standard doses (50 mg) blocks opioid receptors to treat addiction. At low doses (1–4.5 mg), the mechanism differs:
- Transient opioid blockade: Brief receptor occupancy may trigger compensatory endorphin upregulation (Younger, Parkitny, and McLain 2014)
- Glial cell modulation: LDN may antagonize Toll-like receptor 4 (TLR4) on microglia, reducing neuroinflammation (Younger, Parkitny, and McLain 2014)
- Immune modulation: Effects on T regulatory cells and cytokine balance reported; in vitro evidence of TRPM3 ion channel restoration in ME/CFS NK cells (Cabanas et al. 2021)
- Endorphin rebound: Overnight blockade may increase morning endorphin levels (Younger, Parkitny, and McLain 2014)
1.2 Dosing Protocols
Typical protocols involve:
- Starting dose: 0.5–1.5 mg at bedtime
- Gradual titration over weeks to months
- Target dose: 3–4.5 mg (individual optimization required)
- Compounding pharmacy often needed for low doses
1.3 Evidence in ME/CFS
Evidence remains preliminary:
- A retrospective study of ME/CFS patients reported symptomatic improvement in a majority of respondents (Olli Polo, Pesonen, and Tuominen 2019)
- No large randomized controlled trials completed in ME/CFS
- A small randomized crossover trial in fibromyalgia showed significant pain reduction (Oscar Polo et al. 2019); larger reviews support anti-inflammatory effects (Younger, Parkitny, and McLain 2014)
- Patient community reports generally favorable
1.4 Side Effects
Generally well-tolerated:
- Vivid dreams (common, usually transient)
- Sleep disturbance initially
- Headache
- Nausea (rare)
While LDN is generally well-tolerated, severe psychiatric reactions including depression and suicidal ideation have been reported in a subset of ME/CFS patients. These reactions appear more common in individuals who exhibit paradoxical responses to other medications.
Risk factors for psychiatric adverse effects:
- History of paradoxical medication reactions
- Pre-existing mood vulnerability
- Concurrent use of other neuroactive medications
Monitoring protocol:
- Screen for mood changes during first 2–4 weeks of treatment
- Use PHQ-2 or similar brief screening at each dose adjustment
- Ensure caregiver/family awareness to observe for behavioral changes
- Have immediate discontinuation plan ready
- Discontinue immediately if depressive symptoms or suicidal ideation emerge
LDN’s reputation as a “harmless” intervention may lead to inadequate monitoring. Patients and prescribers should maintain vigilance for mood changes, particularly in the “paradoxical reactor” phenotype (see Section Medication Sensitivity Phenotypes).
2 Low-Dose Naltrexone (LDN)
While LDN is generally well-tolerated, severe psychiatric reactions including depression and suicidal ideation have been reported in a subset of ME/CFS patients. These reactions appear more common in individuals who exhibit paradoxical responses to other medications.
Risk factors for psychiatric adverse effects:
- History of paradoxical medication reactions
- Pre-existing mood vulnerability
- Concurrent use of other neuroactive medications
Monitoring protocol:
- Screen for mood changes during first 2–4 weeks of treatment
- Use PHQ-2 or similar brief screening at each dose adjustment
- Ensure caregiver/family awareness to observe for behavioral changes
- Have immediate discontinuation plan ready
- Discontinue immediately if depressive symptoms or suicidal ideation emerge
LDN’s reputation as a “harmless” intervention may lead to inadequate monitoring. Patients and prescribers should maintain vigilance for mood changes, particularly in the “paradoxical reactor” phenotype (see Section Medication Sensitivity Phenotypes).
Patient community reports describe synergistic benefits from combining LDN with other interventions. One frequently mentioned combination involves LDN (at bedtime), NAD+ precursors (nicotinamide riboside or NMN, in the morning), and melatonin (at bedtime for circadian regulation). The theoretical rationale combines: (1) LDN’s anti-neuroinflammatory effects, (2) NAD+’s role in mitochondrial energy production and cellular repair, and (3) melatonin’s effects on sleep architecture, circadian rhythm, and its own anti-inflammatory properties. Individual case reports describe dramatic improvements, including return to work after prolonged disability. However, this represents anecdotal evidence only—no controlled trials have evaluated this specific combination, and publication bias strongly favors positive reports. The heterogeneous nature of ME/CFS means that treatments helping some patients may be ineffective or harmful for others. Patients considering such combinations should work with knowledgeable physicians and implement changes sequentially to identify individual responses.
Testable prediction: A controlled crossover trial of LDN + NAD+ precursors + melatonin vs LDN monotherapy shows greater improvement in PEM frequency and cognitive processing speed at 12 weeks. Falsified if the combination produces no benefit over LDN alone. Limitations: Evidence is anecdotal only; no controlled data for this specific combination exist.
The majority of medications in this chapter are used off-label in ME/CFS, with evidence imported from other conditions. Key epistemic boundaries:
- LDN efficacy in ME/CFS rests on retrospective data, patient community reports, and a small fibromyalgia crossover trial—no large RCT has been completed in ME/CFS specifically.
- Antiviral protocols rest on limited controlled evidence: the Lerner 2007 valacyclovir trial (Lerner et al. 2007) was placebo-controlled but small and conducted by a single group; the Montoya 2013 valganciclovir RCT (n=30) (Montoya et al. 2013) showed significant improvement in mental fatigue, fatigue severity, and cognitive function in responders, but the overall effect was driven by a subgroup, and neither trial has been independently replicated.
- Mitochondrial supplements (CoQ10, L-carnitine, D-ribose, NADH) have individually small or single-study evidence bases in ME/CFS; combination protocols are extrapolated from biochemical reasoning rather than clinical testing.
- The “Energy Profile” classification (Categories A/B/C) assigned to each medication is a rational framework that has not been validated by measuring actual metabolic processing costs in ME/CFS patients.
2.1 Mechanism of Action
2.2 Dosing Protocols
2.3 Evidence in ME/CFS
2.4 Side Effects
2.5 What the Response Tells Us About LDN
Low-dose naltrexone targets microglial TLR4 (reducing neuroinflammation), TRPM3 (restoring calcium channel function), and the opioid/endorphin system — making it one of the most broadly informative probes available. Because LDN is energy-neutral, it can be tried even in severe patients.
2.5.1 If LDN works: improved fatigue, brain fog, PEM tolerance, or pain
2.5.1.1 Finding 1 — Neuroinflammation was present
LDN blocks TLR4 receptors on microglia (the brain’s resident immune cells). TLR4 is a danger sensor — when chronically activated by cellular debris, bacterial fragments, or inflammatory signals entering the brain, microglia release cytokines (IL-1β, TNF-α) that produce fatigue, cognitive slowing, and pain sensitivity. This is neuroinflammation — persistent low-grade immune activation in the nervous system.
Why LDN response implicates it: LDN blocked TLR4 → microglia calmed → inflammatory signaling dropped → symptoms improved. Therefore, neuroinflammation was contributing.
Certainty of this inference: Low to Medium. LDN’s TLR4 antagonism is established in vitro (Younger, Parkitny, and McLain 2014), and neuroinflammation is documented in ME/CFS (Hundreds of Blood Biomarkers Distinguish ME/CFS, Independent of Inactivity). But we lack direct evidence that LDN’s clinical effect in ME/CFS is TLR4-mediated, and whether LDN reaches brain microglia at clinical doses (1–4.5 mg) is unknown.
What this finding does NOT tell us: What is causing the neuroinflammation. The microglia could be responding to autoantibodies entering the brain (Oxidative Stress Sensing Polymorphisms as Safe Mode Predisposition), viral components, mitochondrial debris, or vagal stress signals from the body. LDN calms microglia without removing the trigger.
Action: Anti-neuroinflammatory strategies (PEA, luteolin, quercetin) become relevant. Identifying and treating the upstream trigger could produce deeper improvement than LDN alone.
Level of action: Partial root cause. LDN calms the microglia (reducing neuroinflammation) but does not remove whatever is activating them. If the trigger is removed and neuroinflammation resolves, LDN may no longer be needed. If the trigger persists, LDN is providing symptomatic neuroimmune suppression without addressing the driver.
2.5.1.2 Finding 2 — TRPM3 ion channel dysfunction was present
LDN restores TRPM3-mediated calcium flux in NK cells in vitro (Cabanas et al. 2018). TRPM3 is a calcium channel — calcium is the universal cellular “on switch.” When TRPM3 fails, immune cells cannot kill infected cells, neurons cannot release neurotransmitters efficiently, and vascular smooth muscle cannot regulate blood flow. TRPM3 dysfunction is the most replicated ion channel finding in ME/CFS, documented across six independent NK cell studies (Cabanas et al. 2021).
Why LDN response implicates it: LDN has been shown to restore TRPM3 function in ME/CFS NK cells in a dish. If this restoration occurs in a living patient — in neurons, blood vessels, and muscle, not just NK cells — it would improve immune function, neurotransmitter release, and blood flow regulation across multiple systems simultaneously.
Certainty of this inference: Low to Medium. The central link — LDN restores TRPM3 in living humans, producing clinical benefit — has never been demonstrated. All TRPM3 restoration evidence is from isolated NK cells in vitro. We do not know whether LDN achieves the same effect in the tissues that generate symptoms.
What this finding does NOT tell us: Whether TRPM3 dysfunction is the root cause or a downstream consequence. GPCR autoantibodies can deplete PIP2 (a membrane lipid TRPM3 needs to open), causing secondary TRPM3 dysfunction (PIP2 Exhaustion Links GPCR Autoantibodies to Multi-Channel Dysfunction). LDN response does not distinguish primary channelopathy from secondary PIP2 depletion.
Action: TRPM3-targeting strategies (pregnenolone sulfate, PIP2 maintenance) become relevant. TRPM3 function measurement, when clinically available, could predict LDN response and track treatment effect.
Level of action: Potential root cause. TRPM3 channelopathy is classified as a trigger-capable root cause (GPCR Autoantibody Cascade as Trigger-Capable Root Cause). If LDN restores TRPM3 function, it is partially correcting a core disease mechanism — but whether the restoration is complete and sustained is unknown.
2.5.1.3 Finding 3 — Orexin suppression may have been relieved
LDN reduces microglial activation in the hypothalamus → less inflammatory signaling (PGE2, TNF-α) suppressing orexin neurons → more orexin release → improved wakefulness and cognition. Orexin (hypocretin) falls into an intermediate “gray zone” in ME/CFS CSF — lower than healthy but not as low as narcolepsy (Chronic Orexin Suppression as a Central Driver of ME/CFS Fatigue and Sleep-Wake Instability).
Why LDN response implicates it: If microglial inflammation was functionally suppressing orexin neurons, and LDN reduced that inflammation, orexin release would increase, improving the cognitive symptoms that orexin governs (attention, motivation, wakefulness). This is the least-tested LDN mechanism in ME/CFS.
Certainty of this inference: Low. Each link is individually plausible (TLR4→microglia established, inflammation→orexin suppression established in animal models (Grossberg et al. 2011), orexin reduced in ME/CFS CSF). But no study has measured orexin before/after LDN treatment in ME/CFS patients.
What this finding does NOT tell us: Whether orexin neurons are damaged (autoantibody-targeted, requiring neuronal repair) or intact but functionally suppressed (inflammation-driven, reversible by calming microglia). LDN response is consistent with either.
Action: If this mechanism contributes, orexin receptor agonists and strategies protecting orexin neurons from inflammation become relevant.
Level of action: Symptom management. Orexin suppression is a downstream consequence of neuroinflammation, not a root cause. Restoring orexin signaling improves wakefulness and cognition but does not address what caused the suppression in the first place.
2.5.1.4 Finding 4 — Endorphin-mediated pain relief
At standard doses (50 mg), naltrexone blocks opioid receptors. At low doses (1–4.5 mg), brief overnight blockade triggers compensatory endorphin upregulation. Endorphins are natural painkillers and mood regulators.
Why LDN response implicates it: The endorphin rebound effect is well-documented in pain conditions (Younger, Parkitny, and McLain 2014). Pain relief and mood improvement on LDN likely involve this mechanism — but endorphin elevation alone is unlikely to explain PEM reduction, cognitive improvement, or immune modulation.
Certainty of this inference: Medium for pain improvement, Low for other symptoms.
What this finding does NOT tell us: Whether pain is inflammatory, neuropathic, or centralized — all respond to endorphins.
Action: If pain relief is the main benefit, the underlying disease process may still be progressing — this is symptomatic relief, not disease modification. Treating the pain driver directly (neuroinflammation, autoantibodies, mast cell activation) may produce more durable benefit than relying on endorphin compensation alone.
Level of action: Symptom management. Endorphin elevation relieves pain and improves mood but does not address any disease mechanism. The underlying pathology continues unchecked.
2.5.1.5 Finding 5 — The dose-response curve shape identifies which mechanism is rate-limiting
LDN’s four mechanisms engage at different doses (hormetic dose response): TLR4/Nrf2 anti-inflammatory priming peaks at 0.5–1.5 mg, compensatory endorphin upregulation plateaus at 1.5–3.0 mg, TRPM3 restoration may require 3.0–4.5 mg, and orexin disinhibition tracks the TLR4 dose-response. The dose at which a patient responds is therefore itself a diagnostic readout.
Why dose-response implicates mechanism: Benefit at 0.5–1.5 mg that disappears at 3.0–4.5 mg → TLR4/Nrf2 hormetic mechanism dominant — higher doses remove the basal TLR4 tone sustaining the anti-inflammatory priming, so benefit collapses (Calabrese and Kozumbo 2021) (Kučić et al. 2021). Benefit that plateaus at 1.5–3.0 mg and persists at 4.5 mg → endorphin compensation dominant (a hormetic mechanism would have worsened at the high dose). Benefit only at 3.0–4.5 mg → TRPM3 or orexin dominant. A biphasic pattern (benefit at 0.5 mg, loss at 1.5 mg, return at 3.0 mg) → two non-overlapping optima engaged: TLR4 hormetic at low dose plus a second mechanism at higher dose.
Certainty of this inference: Low. The multi-target dose-optimum model is mechanistically grounded but no within-range dose-response trial has been conducted in any condition.
What this finding does NOT tell us: The TRPM3 dose-response is unknown — all in vitro data are single-concentration (Cabanas et al. 2018) — so high-dose-only response cannot be confidently attributed to TRPM3 over orexin. Symptom profile helps: wakefulness/cognition-dominant benefit suggests orexin; multi-system benefit suggests TRPM3.
Action: Titrate slowly across the full range (0.5 → 4.5 mg) and record response at each dose — the titration is a diagnostic procedure, not just dose-finding. If escalation extinguishes benefit, return to the effective low dose: this is not a failed titration but identification of the hormetic window.
Level of action: Varies by identified mechanism — TRPM3 is a trigger-capable root cause (GPCR Autoantibody Cascade as Trigger-Capable Root Cause); TLR4 calming is a partial root cause; endorphin and orexin effects are symptom management.
2.5.1.6 What a positive response does NOT reveal
- Which of the four mechanisms is responsible — at a single fixed dose. LDN targets all four simultaneously, so response at one dose cannot be attributed to any single mechanism. However, the dose-response curve shape discriminates (Finding 5): benefit lost on escalation implicates the TLR4/Nrf2 window; plateau without loss implicates endorphin compensation; high-dose-only benefit implicates TRPM3 or orexin.
- Whether response is pharmacological or placebo. LDN has never been demonstrated superior to placebo in an ME/CFS RCT. Some or all responses may be placebo effects or natural fluctuation.
- The upstream cause. LDN calms microglia and may restore TRPM3, but does not identify what triggered these dysfunctions — infection, autoantibodies, metabolic collapse, or genetic vulnerability.
2.5.2 If LDN does NOT work (adequate dose ≥ 3–4.5 mg, ≥ 8–12 weeks)
2.5.2.1 Interpretation 1 — Neuroinflammation may not be the dominant mechanism
If ME/CFS does not involve significant microglial TLR4-driven neuroinflammation, LDN’s primary target is absent. However, LDN may simply fail to reach brain microglia at clinical doses (individual blood-brain barrier permeability, drug metabolism). Non-response does not prove neuroinflammation absent.
2.5.2.2 Interpretation 2 — TRPM3 dysfunction may be absent or not LDN-responsive
TRPM3 channelopathy may not be universal in ME/CFS. Some patients may have normal TRPM3 function, or dysfunction from mechanisms LDN cannot correct (direct channel-blocking autoantibodies rather than regulatory impairment). LDN may also restore NK cell TRPM3 without affecting neuronal or vascular TRPM3.
2.5.2.3 Interpretation 3 — Dose wrong — possibly too high, not too low
No dose-response studies exist. Some patients may need higher doses (up to 6 mg) or longer treatment. But the hormetic model (hormetic dose response) implies the opposite failure mode is equally likely: if the patient’s therapeutic mechanism is the TLR4/Nrf2 window (0.5–1.5 mg), a trial conducted only at 3.0–4.5 mg sits above the therapeutic window and produces apparent non-response even in a true responder. Non-response cannot be concluded unless the low-dose window has also been tested.
2.5.2.4 Interpretation 4 — Objective benefit without subjective improvement
TRPM3 may improve on testing without the patient feeling better. TRPM3 dysfunction present but not rate-limiting — other mechanisms drive symptoms despite channel restoration.
2.5.2.5 Interpretation 5 — Paradoxical reactor
A subset of ME/CFS patients worsen on LDN (Medication Sensitivity Phenotypes). Their adverse response is itself diagnostic — their ME/CFS involves mechanisms aggravated by opioid modulation or TLR4 blockade.
2.5.2.6 Interpretation 6 — LDN may not work better than placebo in ME/CFS
No large RCT has demonstrated superiority to placebo. If LDN = placebo, all response patterns interpreted here are placebo effects or natural fluctuations. This cannot be excluded without a controlled trial.
2.5.2.7 Key caveat on non-response
Non-response is weaker evidence than response. It makes a mechanism less likely but does not exclude it — the drug may simply not reach its target. This is more than theoretical: (+)-naltrexone is a weak TLR4 antagonist, and an optimized derivative required ~6200× potency gain to reach nanomolar TLR4 antagonism (Gao et al. 2025) — so clinical-dose racemic LDN (4.5–6 mg) may never engage TLR4, making a null result the expected outcome of a mechanism that was never actually tested at dose (Nielsen, Vaegter, and Due Bruun 2026). LDN non-response also does not exclude immune involvement in general: autoantibody, B-cell, mast-cell, and T-cell mechanisms are not directly targeted by LDN.
The fibromyalgia evidence illustrates the non-response reading. The FINAL trial (n=99) showed no significant primary pain difference (Due Bruun et al. 2024), and a 30% responder re-analysis of six secondary non-pain outcomes was null on all of them (Nielsen, Vaegter, and Due Bruun 2026); an independent 12-month RCT and a meta-analysis also found no between-group benefit (Rodríguez-Freire et al. 2026) (Ologunowa et al. 2025). This does not rule out benefit in ME/CFS: fibromyalgia may lack the TRPM3 channelopathy that has been reported in ME/CFS NK cells (Cabanas et al. 2021), so the FM null may not generalize to the mechanism LDN is best-documented to address in ME/CFS. It likewise does not exclude the OGFr/enkephalin-rebound mechanism, which may require longer duration or a fatigue-specific subtype (Zagon and McLaughlin 2018).
2.5.3 How LDN combines with other medications
LDN works + pyridostigmine works: Step 1: LDN → neuroinflammation/TRPM3 present. Step 2: Pyridostigmine increases acetylcholine, improving cerebral blood flow and parasympathetic function. Step 3: Both working = neuroinflammation → autonomic dysfunction compensated. Inflammatory-to-autonomic pathway, consistent with brainstem neuroinflammation disrupting autonomic centers (Trigger-Capable Mechanisms). → Prioritize anti-inflammatory treatment (upstream); add autonomic support for downstream symptoms.
LDN works + valacyclovir does NOT work: Step 1: LDN → neuroinflammation present. Step 2: Valacyclovir non-response → active herpesvirus replication unlikely as main driver. Step 3: Neuroinflammation without active viral replication → post-infectious mechanism: infection triggered persistent immune changes after viral clearance. → Stop antiviral pursuit. Focus on immunomodulation and trigger identification.
LDN works + antihistamines do NOT work: Step 1: LDN → neuroinflammation present. Step 2: Antihistamine non-response → mast cell activation unlikely as primary neuroinflammation driver. Step 3: Neuroinflammation from non-mast-cell sources — autoantibodies, microglial priming, metabolic stress. → Distinguishes MCAS-driven from autoantibody/post-infectious subtypes — treatment differs fundamentally.
LDN does NOT work + cimetidine works: Step 1: LDN non-response → TLR4/TRPM3 pathway less likely. Step 2: Cimetidine uniquely enhances T-cell function and inhibits CYP450. Step 3: Cimetidine-positive, LDN-negative → T-cell component, possibly viral reactivation. Matches VIM phenotype (Domain 6 Evidence Limitations — No ME/CFS RCT Exists). → VIM phenotype identified: cimetidine + amino acids + metabolic support + antivirals if indicated.
LDN does NOT work + pyridostigmine does NOT work + LDA works: Step 1: LDN non-response → neuroinflammation less likely. Step 2: Pyridostigmine non-response → cholinergic autonomic dysfunction less likely. Step 3: LDA response → dopamine signaling impairment. Step 4: Dopamine-responsive, neuroinflammation-non-responsive, autonomic-non-responsive → primary dopaminergic deficit. → Prioritize cognitive/dopaminergic support with caution: dopamine compensates downstream while upstream driver continues.
2.5.4 Limitations
- No large ME/CFS RCT: All inferences rest on known mechanisms, retrospective data, fibromyalgia trials, and community reports — not controlled ME/CFS evidence.
- Mechanism specificity uncertain: LDN targets multiple mechanisms simultaneously; response cannot be attributed to any single one.
- No biomarker confirms target engagement: Cannot measure TLR4 blockade or TRPM3 restoration in individual patients.
- TRPM3 testing not clinically available: Research-grade flow cytometry only.
- Orexin and endorphin mechanisms are the least-tested in ME/CFS: No pre/post LDN treatment data exist.
- Dose-response unknown: Optimal dose may differ by mechanism and by individual.
- Overall inference certainty: Low to Medium. If future RCTs show LDN ≤ placebo, every inference above collapses — responses would be indistinguishable from placebo plus natural fluctuation.
(Origin: medication-differential-analysis)
3 Immunoglobulins (IVIG)
4 Rituximab
5 Other Immunomodulators
The majority of medications in this chapter are used off-label in ME/CFS, with evidence imported from other conditions. Key epistemic boundaries:
- LDN efficacy in ME/CFS rests on retrospective data, patient community reports, and a small fibromyalgia crossover trial—no large RCT has been completed in ME/CFS specifically.
- Antiviral protocols rest on limited controlled evidence: the Lerner 2007 valacyclovir trial (Lerner et al. 2007) was placebo-controlled but small and conducted by a single group; the Montoya 2013 valganciclovir RCT (n=30) (Montoya et al. 2013) showed significant improvement in mental fatigue, fatigue severity, and cognitive function in responders, but the overall effect was driven by a subgroup, and neither trial has been independently replicated.
- Mitochondrial supplements (CoQ10, L-carnitine, D-ribose, NADH) have individually small or single-study evidence bases in ME/CFS; combination protocols are extrapolated from biochemical reasoning rather than clinical testing.
- The “Energy Profile” classification (Categories A/B/C) assigned to each medication is a rational framework that has not been validated by measuring actual metabolic processing costs in ME/CFS patients.