Sleep-Glymphatic-Neuroinflammation Pathway Study

1 Background and Rationale

The glymphatic system—the brain’s waste clearance mechanism that is maximally active during sleep—may play a critical role in neuroinflammation control. Poor sleep in ME/CFS could impair glymphatic clearance of neuroinflammatory mediators, perpetuating microglial activation and progressive symptom deterioration. Conversely, treating underlying sleep disorders might interrupt this pathway and slow disease progression. This study tests whether interventions targeting sleep disorders in ME/CFS patients with documented sleep pathology produce measurable reductions in neuroinflammation and slowing of PEM progression.

2 Study Design

2.1 Design Overview

Randomized, prospective study of ME/CFS patients with documented sleep disorders, comparing immediate treatment to delayed (6-month wait-list control) treatment.

2.2 Participants

  • n=60 ME/CFS patients (ages 18–60) with documented sleep disorders
  • Sleep disorder documentation via home sleep apnea testing: OSA (AHI \(\geq\) 5 events/hour) or other sleep pathology (insomnia with documented sleep fragmentation, periodic breathing, etc.)
  • Mild to moderate severity (Bell scale 40–75)
  • No prior treatment with CPAP, BiPAP, or formal sleep disorder management
  • Willing to be randomized to immediate vs. delayed treatment

2.3 Randomization

1:1 randomization to:

  • Immediate treatment (n=30): Sleep disorder management initiated at baseline
  • Delayed treatment (n=30): Sleep management begins at 6-month mark (wait-list control) Stratification by sleep disorder type (OSA vs. other).

3 Interventions

3.1 Immediate Treatment Arm

OSA patients (CPAP/BiPAP):

  • Polysomnography or split-night study to determine therapeutic pressure

  • CPAP or BiPAP initiation with titration protocol

  • Mask fitting and desensitization

  • Monthly adherence monitoring for first 3 months, then quarterly

  • Target: \(\geq\) 4 hours per night use Insomnia or fragmentation:

  • Sleep restriction therapy with gradual sleep window expansion

  • Cognitive-behavioral therapy for insomnia (CBT-I) protocol

  • Pharmacological support: Trazodone or mirtazapine as needed

  • Weekly coaching sessions for first 4 weeks

3.2 Delayed Treatment Arm

Standard sleep hygiene education only for first 6 months; treatment as above begins at month 6.

4 Measures

4.1 Sleep Assessment (baseline, 6 months, 12 months)

  • Home sleep apnea testing or portable sleep monitoring
  • Sleep diary (7-day baseline, 7-day at each timepoint)
  • Actigraphy (7-day continuous at each timepoint)
  • Pittsburgh Sleep Quality Index (PSQI)

4.2 Neuroinflammation Markers (baseline, 6 months, 12 months)

  • CSF sampling (optional lumbar puncture subset, n=20 per arm): TNF-\(\alpha\), IL-6, IL-1\(\beta\), MCP-1, neopterin
  • Serum inflammatory markers: High-sensitivity CRP, IL-6, TNF-\(\alpha\), IL-1\(\beta\)
  • TSPO-PET imaging (subset, n=15 per arm): microglial activation assessment

4.3 Clinical Outcome Measures (baseline, 6 months, 12 months)

  • PEM crash diary (continuous): Frequency, severity, recovery duration
  • Bell Disability Scale
  • Cognitive function (Montreal Cognitive Assessment)
  • Autonomic function (NASA Lean Test, HRV)
  • DSQ-PEM

5 Outcomes

5.1 Primary Outcomes

  • Change in inflammatory marker trajectory (serum IL-6) from baseline to 12 months: Immediate vs. delayed arm
  • PEM threshold stability: Rate of change in PEM severity over 12 months in treated vs. untreated groups

5.2 Secondary Outcomes

  • Individual inflammatory markers (TNF-\(\alpha\), IL-1\(\beta\), CRP)
  • CSF markers in subset (if available)
  • TSPO-PET signal change in imaging subset
  • Bell Disability Scale change baseline to 12 months
  • Cognitive function improvement
  • PEM crash frequency change

5.3 Tertiary Outcomes

  • Sleep quality improvement and correlation with inflammatory markers
  • CPAP/BiPAP adherence as moderator of treatment effect
  • Slow-wave sleep percentage (from polysomnography in subset) and correlation with inflammatory markers

6 Analysis Plan

  • Primary analysis: Linear mixed-effects model comparing serum IL-6 trajectory over 12 months between arms
  • Secondary outcomes: ANCOVA for Bell scale and other clinical outcomes at 12 months, adjusted for baseline values
  • Mediation analysis: Does change in sleep quality mediate the relationship between treatment arm and inflammation?
  • Adherence analysis: Among immediate arm, does CPAP adherence predict inflammatory marker improvements?

7 Sample Size Justification

With n=30 per arm and 3 measurement timepoints:

  • 80% power to detect 30% difference in IL-6 slope between arms (assuming SD=0.8 in log IL-6) at \(\alpha\)=0.05
  • Based on preliminary data: treated sleep apnea showing 25–35% reduction in inflammatory markers in healthy controls

8 Expected Outcomes and Implications

If sleep treatment reduces neuroinflammation and slows PEM progression:

  • Establishes sleep as critical disease-modifying factor

  • Validates screening and treating sleep disorders as standard ME/CFS care

  • Identifies glymphatic function as therapeutic target

  • May support clinical guidelines for aggressive sleep disorder management If sleep treatment shows minimal effect:

  • Suggests neuroinflammation maintenance is not primarily glymphatic-dependent

  • Indicates other mechanisms of microglial activation are primary

  • Redirects focus toward direct microglial targeting (LDN, other agents)

ImportantHypothesis: Metabolic-Immune Crosstalk in LDN Response

LDA-induced metabolic changes (glucose intolerance, insulin resistance) amplify neuroinflammation and create a therapeutic ceiling that limits cognitive benefit over time. Concurrent metformin administration in patients developing prediabetes will preserve LDN efficacy and enhance cognitive outcomes compared to LDN monotherapy. HbA1c improvements correlate with neuroinflammatory marker reduction (MCMC Research 2024).

NoteOpen Question: Could LDN vs Metformin Ever Be a Phenotype-Guided First-Line Choice?

A proposed stratification framework suggests choosing between repurposed drugs by dominant mechanism: LDN for a neuroimmune/TRPM3-dominant phenotype, metformin for an immunometabolic phenotype (insulin resistance, AMPK/mTOR dysregulation). The current evidence base does not support this as a triage rule, and the two drugs are not equally evidenced: metformin has been tested in established disease and failed to show symptom benefit — one completed RCT was negative ((Lim et al. 2026)) and a second adaptive trial stopped its metformin arm for futility ((Reis et al. 2026)) — whereas LDN has never been tested in a completed placebo-controlled RCT for this endpoint at all. “Neither can be selected by validated biomarker” is therefore not a symmetric null: it reflects one drug that has been tested and not shown benefit, and one that remains untested.

The strongest case for metformin is not that it failed, but that the failed trials were unselected: Lim2026 and Reis2026 enrolled established-disease patients without stratifying by insulin resistance, the exact design that can mask a true subgroup effect. Metformin’s one positive RCT-level signal — COVID-OUT, acute-phase initiation, 41% relative reduction in Long COVID incidence (6·3% vs 10·4% by day 300, (Bramante et al. 2023)) — arose in a defined metabolic/timing context, and post-infectious insulin resistance is documented in the same trigger population ((Sanhueza et al. 2026)). A metabolically defined subgroup could therefore benefit even though unselected trials did not detect it. Against this, the mechanism-specific endpoints (REE, NETosis) have never been measured in ME/CFS, and no negative trial has been re-analysed by HOMA-IR subgroup — so the subgroup hypothesis is unverified, not refuted. (Absence of a measured response marker is also not proof that no responsive subgroup exists; the interaction may simply be unmeasured.)

On the LDN side, TRPM3 NK-cell dysfunction is a reproducible abnormality restored in vitro by naltrexone (E. Sasso et al. 2026) (E. M. Sasso et al. 2025), but restoration of channel function has never been shown to predict clinical response; metformin’s immunometabolic reach (Th17/Treg balance, B-cell biology, flares in SLE/RA trials) is likewise real but modest and inconsistent (Ursini et al. 2018) (Sun et al. 2020) (Gharib et al. 2021). Neither mechanism has demonstrated clinical-response prediction, so for a given patient the choice between them is currently empirical rather than biomarker-guided. The two drugs could in principle act on different branches of the same process (Metabolic-Immune Crosstalk in LDN Response), but no synergy mechanism or co-administration evidence exists.

Falsifiable prediction: Two designs could test the phenotype hypothesis. The cheaper first step is an enrichment trial — enrol only insulin-resistant (HOMA-IR-high) established ME/CFS patients and randomize to metformin vs placebo, with REE and NET/insulin endpoints alongside symptoms. Only if that shows signal is the fuller TRPM3 × HOMA-IR stratified randomization against LDN warranted (an interaction test needing several-fold the n of a main-effect trial). The joint prediction is that metformin benefits the HOMA-IR-high stratum and LDN the TRPM3-abnormal/insulin-sensitive stratum, with a significant phenotype × drug interaction. Indeterminate if subgroup signal appears on mechanism endpoints (REE, NETosis) without symptom improvement — that would establish metabolic modulation without clinical benefit, and would require a separate clinical-endpoint trial rather than declaring the hypothesis falsified.

Consequence: This is a research question, not a treatment protocol — no dose, monitoring schedule, or patient-selection rule is supplied here. No validated marker selects either drug today; for a patient who presents with post-viral fatigue and insulin resistance now, neither metformin nor LDN has demonstrated established-disease symptom benefit, and the choice is empirical — metformin’s evidence is negative in unselected trials (with an untested subgroup hypothesis), and LDN’s is mechanistic and untested in a completed RCT. The question defines what a future decision aid would need — validated response predictors — and warns against importing metformin into established ME/CFS on the strength of its acute-phase prevention signal. (Severity applicability: unknown — no severity-stratified data; this is a research question, not clinical guidance. Origin: brainstorm/assessment.)

References

Bramante, Carolyn T, John B Buse, David M Liebovitz, Jacinda M Nicklas, Michael A Puskarich, Ken Cohen, et al. 2023. “Outpatient Treatment of COVID-19 and Incidence of Post-COVID-19 Condition over 10 Months (COVID-OUT): Randomised, Quadruple-Blind, Placebo-Controlled Trial.” The Lancet Infectious Diseases 23 (10): 1119–29. https://doi.org/10.1016/S1473-3099(23)00299-2.
Gharib, Mahmoud, Walaa Elbaz, Ebtissam Darweesh, Nagwa Ali Sabri, and May Ahmed Shawki. 2021. “Efficacy and Safety of Metformin Use in Rheumatoid Arthritis: A Randomized Controlled Study.” Frontiers in Pharmacology 12: 726490. https://doi.org/10.3389/fphar.2021.726490.
Lim, So Yun, Jacob Lee, Euijin Chang, Ji-Soo Kwon, Choi Young Jang, Yubin Seo, Jin Ju Park, et al. 2026. “Neither Metformin nor Ursodeoxycholic Acid Effectively Treats Postacute Sequelae of COVID-19: A Randomized Clinical Trial.” Annals of Internal Medicine 179 (4): 478–85. https://doi.org/10.7326/ANNALS-25-04883.
MCMC Research. 2024. “Neurometabolic Modelling of PEM and Central Sensitisation.” 2024. https://www.mcmc-research.com/post/neurometabolic-modelling-of-pem-and-central-sensitisation.
Reis, Gilmar, Eduardo Augusto dos Santos Moreira Silva, Daniela Carla Medeiros Silva, Lehana Thabane, Thiago Santiago Ferreira, Luiza Lanna França Reis, et al. 2026. “The Effect of Fluvoxamine and Metformin for Fatigue in Patients with Long COVID: An Adaptive Randomized Trial.” Annals of Internal Medicine. https://doi.org/10.7326/ANNALS-25-03959.
Sanhueza, Sergio, Camilo Cabrera, Romina Quiroga, et al. 2026. “De Novo COVID-19-Associated Insulin Resistance Drives Dysregulated Neutrophil Extracellular Trap Formation (NETosis) Four Months After Infection.” Frontiers in Immunology 17: 1787799. https://doi.org/10.3389/fimmu.2026.1787799.
Sasso, Etianne Martini, Natalie Eaton-Fitch, Peter Smith, Katsuhiko Muraki, and Sonya Marshall-Gradisnik. 2025. “Low-Dose Naltrexone Restored TRPM3 Ion Channel Function in Natural Killer Cells from Long COVID Patients.” Frontiers in Molecular Biosciences 12: 1582967. https://doi.org/10.3389/fmolb.2025.1582967.
Sasso, Etianne, Peter Smith, Sonya Marshall-Gradisnik, et al. 2026. “Multi-Site Validation of TRPM3 Ion Channel Dysfunction in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” Frontiers in Medicine, January. https://doi.org/10.3389/fmed.2025.1703924.
Sun, Fangfang, Huijing Wang, Zhe Liu, Shikai Geng, Haiting Wang, Xiaodong Wang, Ting Li, Laurence Morel, Weiguo Wan, and Liangjing Lu. 2020. “Safety and Efficacy of Metformin in Systemic Lupus Erythematosus: A Multicentre, Randomised, Double-Blind, Placebo-Controlled Trial.” The Lancet Rheumatology 2 (4): e210–16. https://doi.org/10.1016/S2665-9913(20)30004-7.
Ursini, Francesco, Emilio Russo, Gianluca Pellino, Salvatore D’Angelo, Agostino Chiaravalloti, Giovambattista De Sarro, Roberto Manfredini, and Roberto De Giorgio. 2018. “Metformin and Autoimmunity: A ‘New Deal’ of an Old Drug.” Frontiers in Immunology 9: 1236. https://doi.org/10.3389/fimmu.2018.01236.