Sickness Behavior as Overarching Integrative Framework
Sickness behavior—the coordinated constellation of fatigue, social withdrawal, anhedonia, cognitive slowing, hyperalgesia, and appetite suppression induced by immune activation—provides an overarching framework for understanding ME/CFS symptomatology. This section argues that many core ME/CFS symptoms represent a chronic, dysregulated sickness behavior state mediated by cytokines, prostaglandins, and vagal afferents acting on hypothalamic and limbic circuits. The adaptive logic of acute sickness behavior (conserving energy for immune defense) becomes maladaptive when chronically engaged without resolution.
1 Cytokine-to-Brain Signaling Routes
Peripheral cytokines communicate with the brain through three complementary routes (Dantzer et al. 2000) (McCusker and Kelley 2013). The humoral pathway operates via circumventricular organs—notably the organum vasculosum of the lamina terminalis (OVLT) and the area postrema—where an incomplete blood-brain barrier permits direct cytokine diffusion into adjacent hypothalamic tissue. A saturable transport pathway operates through carrier-mediated influx systems at the BBB endothelium for IL-1\(\beta\), IL-6, and TNF-\(\alpha\) (McCusker and Kelley 2013). The neural pathway exploits vagal afferents: cytokines at peripheral infection sites activate cytokine receptors on paraganglia of the vagus nerve, transmitting immune signals to the nucleus tractus solitarius and thence to hypothalamic circuits within minutes (Dantzer et al. 2000) (Huerta, Masters, and Matheny 2025).
At the blood-brain barrier endothelium, circulating cytokines trigger prostaglandin E2 (PGE2) synthesis, which diffuses into the hypothalamic parenchyma to activate EP3 and EP4 receptors, suppressing wake-active orexin neurons and promoting sleep-pressure circuits (Matsumura and Kobayashi 2004) (Marty et al. 2008).
2 Hypothalamic Integration and the Sickness Behavior Program
Peripheral proinflammatory cytokines communicate with the hypothalamus via the neural and humoral pathways described above. Once there, cytokines and PGE2 activate distinct receptor populations that coordinate a unified behavioral program: IL-1\(\beta\) suppresses wake-active neurons in the lateral hypothalamus, IL-6 and TNF-\(\alpha\) activate sleep-promoting circuits in the ventrolateral preoptic area, and PGE2 potentiates these effects via EP3/EP4 signaling. The downstream result is the sickness behavior syndrome—fatigue, anorexia, hyperalgesia, social withdrawal, fever, and sleep dysregulation—all orchestrated by hypothalamic cytokine integration (Dantzer et al. 2008) (Dantzer 2008). This program is adaptive in acute infection (conserving energy for immune defense), but becomes pathological when chronically engaged. (Mechanism: well-established in animal and human models; certainty: High for acute sickness behavior, Medium for chronic persistence in ME/CFS).
3 Persistent Sickness Behavior in ME/CFS
In acute infection, sickness behavior resolves as pathogens are cleared and cytokines return to baseline. In ME/CFS, this resolution fails: persistent immune activation (elevated TGF-\(\beta\), NK hypofunctionality, activated microglia) maintains cytokine drive, while HPA axis hyporesponsiveness removes the glucocorticoid anti-inflammatory brake (Morris and Maier 2013). Post-exertional malaise may represent an acute exacerbation of this state, in which exercise triggers an abnormal cytokine spike that transiently recapitulates full sickness behavior (Dantzer 2008) (Morris and Maier 2013). (Certainty: Medium. Mechanism coherent; direct PGE2 measurement in ME/CFS lacking.)
4 Dedicated Sickness and Torpor Neural Circuits
If the sickness behavior that accompanies ME/CFS is generated by dedicated, recently identified neural circuits rather than by an unspecific diffuse response — a hypothesis with present-tense evidence today only in rodents — then these circuits would offer a circuit-level mechanistic account of how neuroinflammation produces the core symptoms. In rodents, three such circuits have been mapped: a circuit originating in the area postrema and projecting to the brainstem that suppresses eating, drinking, and movement (sickness behavior); a circuit from the organum vasculosum of the lamina terminalis to the ventromedial preoptic area that produces appetite suppression, warmth seeking, and fever; and a circuit centered on the median preoptic nucleus that can induce a torpor-like hypometabolic state (Hrvatin et al. 2020) (Komaroff and Dantzer 2025). Preoptic EP3-receptor neurons may act as a two-way switch routing the organism between protective fever and torpor-like hypometabolic states (Machado et al. 2025). The area-postrema→brainstem circuit, which suppresses movement and appetite, is the arm with the closest phenotypic match to the ME/CFS symptom set; the fever (VMPO) and torpor (MnPO) arms are included as context but are less clearly engaged in the human phenotype (ME/CFS patients do not show a torpor vital-sign signature of hypothermia or bradycardia). Applied cautiously to ME/CFS, engagement of these circuits would generate fatigue, cognitive slowing, and malaise and could contribute to a chronically low-energy state that overlaps the metabolic-safety-mode and CNS energy crisis models (Sections CNS Energy Crisis as Trigger-Capable Root Cause, Harmful Advice: The “Power of Positive Thinking”). An important rival must be engaged: any low-energy state in ME/CFS is also explainable by reduced physical activity (deconditioning), which the paper’s bed-rest rebuttal distinguishes from circuit-driven pathology (Section Sixty-Day Bed Rest Fails to Reproduce the ME/CFS Muscle Phenotype — Deconditioning Is Not the Explanation). Whether the dedicated-circuit account adds explanatory value beyond the already-integrated diffuse neuroinflammation → sickness-behavior framework (present in this chapter) remains an open question; the model changes no immediate treatment prediction (existing anti-neuroinflammatory rationale is unchanged) and its pharmacodiagnostic utility is not yet established. (Certainty: 0.40. The rodent circuits are well established; the application to chronic human ME/CFS is inferential and rests on a single review by the framework’s originator — no independent, ME/CFS-cohort evidence exists. This certainty reflects the review’s synthesis, not independent replication.)
(Translation gap: animal (rodent) → human. The dedicated circuits and preoptic switch were demonstrated in mice/rats; their chronic engagement in ME/CFS is not evidenced in any patient cohort, and the rodent “torpor” phenotype (defended hypothermia, bradycardia) is not reproduced in ME/CFS.)
Falsifiable prediction: The strongest test identifies whether a specified circuit — not “regional neuroinflammation” in general — drives symptoms. A single long-COVID TSPO-PET study found regional (limbic) glial signal correlated with symptoms but no global elevation (Tuomaala et al. 2026); because limbic regions are not the area-postrema/preoptic circuits named here, that finding is indirect and does not by itself confirm this model. The prediction here is that engagement of the area-postrema→brainstem (sickness) circuit specifically tracks symptom severity. This requires a modality able to resolve these millimeter-scale brainstem/hypothalamic nuclei, which current TSPO-PET (~4–6 mm effective resolution) and standard fMRI cannot do reliably — so near-term human testing requires ultra-high-field fMRI or higher-resolution tracers, and is a research (not clinical) objective. Absence of any circuit-specific activation change with symptoms in such a study would falsify the claim that a dedicated circuit (rather than diffuse inflammation or a plasticity-maintained state) mediates the symptoms. An observation that would refute the neuroinflammation-input claim itself: consistent absence of neuroinflammatory signal (TSPO-PET or CSF) in ME/CFS cohorts would remove the model’s stated input, independent of circuit findings.
Consequence: If dedicated sickness/torpor circuits are confirmed as drivers in ME/CFS, this would give researchers a specific brain pathway to study — while the interim clinical point is unchanged: no new test or drug follows today, and routine inflammatory markers are frequently normal in ME/CFS without excluding the diagnosis.
4.1 The EP3R “Stuck Switch” and Circuit-Leveraged Mechanisms
A speculation (not a finding) that a defective fever↔︎torpor toggle could, in principle, contribute to persistent low-energy fatigue in ME/CFS: preoptic EP3-receptor neurons route between fever (hypermetabolic) and torpor-like (hypometabolic) states in rodents (Machado et al. 2025). If that two-way switch could be left in the energy-conserving position after the acute infection — e.g., via EP3R desensitization, biased signaling, or failure of the fever arm to re-engage — it would offer one way to account for persistent low-energy symptoms with normal peripheral inflammatory markers, without requiring ongoing neuroinflammation (Komaroff and Dantzer 2025). Note that this restates persistence in circuit vocabulary rather than identifying a demonstrated lesion: no EP3R desensitization, receptor bias, or switch-latch state has been shown in ME/CFS or even in the rodent model as a persistent (post-washout) state. A simpler rival reading must be stated plainly: the absence of peripheral inflammation in chronically-fatigued patients (Omdal null) may mean the sickness-behavior model simply does not extend to that inflammation-negative subset, whose symptoms may have a different cause (Omdal et al. 2026) — the stuck-switch mechanism is needed only if the phenotype in that subset is to be retained within this framework. (Certainty: 0.25. The EP3R two-way switch is rodent-established; the persistent-switch-failure application to human ME/CFS has zero direct evidence and no demonstrated latch in any species.)
(Translation gap: animal (rodent) → human. No human evidence of EP3R desensitization or switch failure exists. Rodent torpor-with-hypothermia physiology is not reproduced in ME/CFS.)
(Severity applicability: unknown — not severity-stratified; presented as a general candidate without confirmed applicability to any severity subgroup.)
Falsifiable prediction: The discriminating prediction must target a reversible switch, not a permanently stuck one. If a transient PGE2 load can latch the EP3R switch into sustained torpor output after ligand washout (bistability — the property this hypothesis requires), the rodent model would first demonstrate a persistent, acute-challenge-reversible state; absent that, the human question never arises. In humans, the phenotype prediction is a persistently low defended temperature set-point with acute disengageability — i.e., normal physiology returns briefly under a strong thermoregulatory/arousal challenge, then reverts. Because a peripheral immune-cell (e.g., PBMC) EP3R measure cannot be assumed to report preoptic-neuron EP3R state, PBMC reads are not a valid proxy and are excluded from this prediction; a CNS-proximal readout (CSF or post-mortem marker, ultra-high-field imaging of the MnPO) would be required. The claim is falsified if no persistent, reversible latch is demonstrable in the rodent switch under PGE2 washout, or if ME/CFS patients show no maintained low set-point with acute disengageability.
Origins: brainstorm (re-run) — EP3R two-way-switch failure. This is a hypothesis-generation entry, not clinical guidance; no switch-resetting intervention exists or is in trials.
Consequence: If the brain’s energy-saving switch can be temporarily stuck (and reversibly unlatched) rather than permanently broken, it would point toward a possible recovery mechanism — but this remains a target-identification hypothesis with no treatment pathway today, and no safe bedside test follows from it.
An untested hypothesis that the area postrema could be a point of autoimmune and pharmacological convergence in ME/CFS. Two facts about this single brainstem structure are established separately: (a) it is a route by which circulating GPCR autoantibodies can access brainstem autonomic nuclei, and (b) it is one of the densest GLP-1-receptor-expressing regions in the brain (Azcue et al. 2026) (Komaroff and Dantzer 2025). If GPCR autoantibodies in ME/CFS also bound or interfered with GLP-1 receptors in this region, they could blunt GLP-1-receptor-dependent anti-inflammatory signaling and help maintain neuroinflammation and sickness-circuit activation. This requires an explicit, currently-unsupported assumption: the ME/CFS GPCR autoantibodies implicated here (literature-relevant targets include adrenergic and muscarinic receptors) would have to cross-react with or sterically impede GLP-1 receptors — no such cross-family interaction has been demonstrated in any system. Without that premise, the two structures’ co-location does not by itself entail an interaction; a simpler, established mechanism (agonistic autoantibodies acting on their own receptors) already explains part of the autonomic symptom profile. (Certainty: 0.20. The anatomical co-location is real; the interference step is speculative and rests on an untested cross-reactivity assumption; GPCR-autoantibody findings in ME/CFS are themselves contested and single-source here (Azcue et al. 2026).)
(Translation gap: autoantibody occupancy of GLP-1 receptors at the area postrema in ME/CFS has not been demonstrated.)
(Severity applicability: unknown — the underlying autoantibody and area-postrema data are not severity-stratified.)
Falsifiable prediction: The decisive test is direct and cheap: purified patient IgG (autoantibody-positive vs -negative vs control) in a competitive binding assay against labeled GLP-1 on area-postrema membrane preparations. If ME/CFS GPCR autoantibodies do NOT displace GLP-1 binding, the competitive-interference claim is falsified even if the symptoms otherwise fit. Clinical predictions (blunted GLP-1-induced nausea; reduced GLP-1-agonist anti-inflammatory effect in autoantibody-positive patients) are confounded by vagal dysfunction, gastroparesis, and dysautonomia common in ME/CFS, so they only secondarily support the mechanism and cannot stand alone.
Origins: brainstorm (re-run) — area-postrema dual-role gate. This is hypothesis generation, NOT a prescribing recommendation: no GLP-1-receptor-agonist use in ME/CFS is supported by any trial, GPCR-autoantibody testing is research-only, and GLP-1 receptor agonists carry specific risks in (especially severe) ME/CFS — appetite suppression and slowed gastric emptying worsen intake/weight loss; nausea aggravates orthostatic intolerance in comorbid POTS; gastroparesis is a documented concern. Prescribing decisions must not be made on this hypothesis.
Consequence: If a real autoantibody interference at GLP-1 receptors were demonstrated, it would help explain why inflammation and fatigue can persist — Accordingly, today this informs research design only, not patient care.
Hibernating mammals (arctic ground squirrel, Syrian hamster) show a deep, torpor-like hypometabolic state that is reversible — they periodically “arouse” back to euthermia (brief, metabolically costly returns roughly every one to three weeks). The molecular machinery of arousal is a candidate template for studying how a conserved energy-state switch could be re-engaged — a resource relevant to the torpor-like low-energy state proposed in ME/CFS, conditional on that state sharing mechanism with torpor, which is not established (see constraint below) (Komaroff and Dantzer 2025) (Drew 2026). Candidate metabolic-reactivation signals include: mTORC1 reactivation as a conserved cross-species metabolic-reactivation switch (Wu and Storey 2021); an arachidonic-acid → PPARα/TRPV-Ca²⁺ lipid signal implicated in hibernation metabolic regulation (single T2T-genome/multi-omics study, not yet replicated) (Wang et al. 2026); BAT UCP1 thermogenesis as the primary rewarming heat source in hibernators (of limited relevance to ME/CFS, which shows no defended-hypothermia phenotype) (Hunstiger, Johannsen, and Oliver 2023); and non-neuronal CNS sensors (choroid plexus, tanycytes, pars tuberalis) that may initiate arousal independently of classic hypothalamic thermoregulatory neurons (Markussen et al. 2024). The entry signal — central adenosine A1-receptor activation inducing hypothermia/torpor in a non-hibernating rat (Shimaoka et al. 2018) — is listed here as an entry-direction signal (whose antagonism, a shift toward A2A, is a candidate arousal-direction readout); it is not itself an arousal signal, and whether drug-induced hypothermia reproduces natural torpor rather than a nonspecific depressant effect is contested (central A1AR agonism produces hypothermia in essentially every mammal tested without a demonstrated torpor program). (Certainty: 0.40 that the animal arousal machinery is a legitimate, well-characterized basic-science resource; its translation to ME/CFS is far less certain — see constraint below. No human/ME/CFS data link these signals to the ME/CFS low-energy state.)
(Translation gap: animal/model-system → human. All findings are from hibernating rodents, hamsters, dwarf lemur, or in-vitro mammalian cells; the only confirmed primate hibernator, the fat-tailed dwarf lemur, may not share the rodent arousal circuitry (Blanco et al. 2024).)
(Critical constraint — torpor ≠ ME/CFS phenotype: the defining physiological feature of animal torpor is a regulated, defended drop in body temperature with bradycardia; ME/CFS patients are normothermic or mildly low-normal. The arousal machinery therefore evolved to reverse a state ME/CFS patients do not demonstrably occupy, so at most it informs a partial, non-physiologic torpor-like energy deficit — which proportionally weakens the strength of the cross-species analogy and any pharmacological extrapolation. Note also that hibernation biology has been promoted as a drug-discovery resource for stroke/trauma/cardiac surgery for decades with limited clinical translation to date, a track record that tempers expectation.)
A note on overlap with the paper’s existing ISR content: a core arousal/recovery mechanism — translational stalling and mitochondrial fragmentation reversible by ISR inhibition or mTORC1 reactivation, reported in non-hibernating mammalian cells as a “hibernation-like” adaptive pausing response (Jobava et al. 2021) — substantially overlaps the integrated-stress-response (ISR) and WASF3/PERK content already in this book (ch07 Sections Post-Exertional Malaise Timing Reflects Integrated Stress Response Kinetics, Chronic ISR Activation in ME/CFS: Adaptive Initiation, Maladaptive Persistence, PEM as Biphasic ISR Cycle: Adaptive Initiation and Maladaptive Persistence; the ch07 ISR assay content; muscle WASF3/shRNA rescue evidence). The hibernation biology here does not propose a distinct mechanism; it contributes (a) a specific assayable signature (polysome:monosome ratio, stalled-initiation-codon enrichment) and (b) a cross-species precedent that ISR/pausing is a reversible state, not irreversible damage. This speculation should be read as a research-resource angle on the existing ISR framework, not a novel mechanism.
Falsifiable prediction: Because positive pausing/ISR markers alone are not specific (they also accompany generic chronic-inflammation-driven ISR activation and metabolic inactivity), the discriminating test must include a matched-sedentary/deconditioned control arm and a pausing-specific signature. If the APR/ISR engagement is real in ME/CFS, patient-derived PBMCs or iPSC-derived cells should show: (i) a reduced polysome:monosome ratio and enrichment of stalled-initiation-codon mRNAs relative to both healthy AND matched-sedentary controls; and (ii) partial reversibility of this signature on ISR inhibition (ISRIB) or mTORC1 reactivation that exceeds, by a pre-stated margin (e.g., a statistically significant treatment-by-group interaction, not merely a numerically larger mean), the reversible component in the matched-sedentary control. The hibernation-resource claim is falsified as ME/CFS-relevant if no arousal/recovery signal reverses the low-energy signature in non-hibernating disease-relevant cells or a non-hibernating rodent model (noting that a fully validated rodent ME/CFS model does not yet exist). This is a research prediction on patient samples; it does not describe an intervention for any patient today.
Safety caveat (Duffy): Arousal/reactivation itself increases oxidative damage in hibernators (Duffy, Staples, and Tessier 2022), who have co-evolved antioxidant protections that ME/CFS patients likely lack (documented oxidative stress). Any future attempt to force a metabolically-reactivated state could be net-harmful and must be gated by this caveat.
Origins: /integrate-topic hibernation-biology-torpor-arousal (Phase-4 re-run of the komaroff-dantzer cycle, idea 8.1; consolidated here with the adaptive-pausing/ISR angle).
Consequence: Hibernators show that a low-energy state can be reversibly exited, and the same cellular “pausing” signaling overlaps research already underway in ME/CFS cells — so the greatest value here is a cross-species encouragement and an assay (instead of a new mechanism or treatment). But this is early-stage basic science, the waking process damages cells, no human data exist, and nothing here is usable by patients today.
4.2 Constraints on the Torpor/Sickness-Circuit Model
The dedicated sickness/torpor neural-circuit model, applied to chronic human ME/CFS, is a hypothesis subject to strong constraints and to rival explanations that must be engaged honestly (Komaroff and Dantzer 2025). (1) Human-animal translation gap is maximal for torpor: torpor is a small-mammal adaptation; no primate other than the fat-tailed dwarf lemur naturally torpors, and the median preoptic torpor circuit was identified in rodents (Hrvatin et al. 2020) (Machado et al. 2025). There is zero direct human evidence of this circuit’s engagement in ME/CFS, and the animal “torpor” phenotype — a defended drop in body temperature and bradycardia — is not observed in ME/CFS, so the model predicts at most a partial/attenuated, non-physiologic torpor-like state, which proportionally weakens the rodent analogy. (2) The neuroinflammation input may be absent in established disease: two independent lines — no peripheral inflammatory-biomarker association with persistent fatigue (Omdal null) and a TSPO-PET finding of regional but not global glial signal in long COVID — leave open whether the neuroinflammation that would drive these circuits is present in chronic ME/CFS (Omdal et al. 2026) (Tuomaala et al. 2026). A consistent absence of neuroinflammatory signal in ME/CFS cohorts would remove the model’s stated input. (3) Rival persistence mechanisms are under-rendered: a low-energy symptom state is equally explained by reduced physical activity (deconditioning), which the paper’s bed-rest evidence rebuts as the sole cause but which cannot be presumed absent Sixty-Day Bed Rest Fails to Reproduce the ME/CFS Muscle Phenotype — Deconditioning Is Not the Explanation; and experience-dependent neural plasticity in ordinary (non-torpor) sensitized circuits — the chapter’s own documented persistence mechanism — is an existing rival that requires no cross-species torpor extrapolation and that this model does not currently distinguish from a “dedicated-circuit” account. (4) The model may not apply to all patients: the absence of peripheral inflammation in some chronically-fatigued patients (Omdal null) permits the parsimonious reading that the sickness-behavior/circuit model does not extend to that inflammation-negative subset, whose symptoms may have a different cause. (5) Severity confounding: the review presents the circuit model as general, but severe/very-severe ME/CFS may involve structural CNS pathology outside the sickness-behavior framework, and no severity-stratified circuit data exist; all proposed tests in this section are research-only and unsuitable for severe/bedbound patients. (6) Acute-to-chronic extrapolation: sickness behavior was characterized as an acute (~days) response; extending it to explain multi-year persistence is an extrapolation, and the persistence mechanisms (vicious cycles, autoimmunity, dysbiosis, mitochondrial dysfunction) are conceptually independent of the circuit model. (Working certainty: 0.40 for the parent hypothesis; these constraints primarily apply to the torpor-specific and chronic-severe claims.)
Observed would-be falsifiers: the neuroinflammation-input claim is refuted by consistent absence of neuroinflammatory signal in ME/CFS cohorts; the dedicated-circuit claim is refuted by absence of circuit-specific activation change with symptoms under a sufficiently high-resolution imaging protocol. Naming these keeps the model exposed to falsification rather than accommodating every outcome.
Consequence: These limitations mean the “stuck hibernation switch” or “dedicated brain circuit” explanation for chronic ME/CFS is a hypothesis awaiting direct human evidence — not a confirmed mechanism — and is especially unverified for bedbound patients who may have progressed to structural, non-circuit pathology. No new test or treatment follows today.
5 Orexin/Hypocretin Suppression as a Central Mediator of ME/CFS Fatigue and Sleep Dysfunction
The orexin (hypocretin) system provides a mechanistic link between the cytokine-driven sickness behavior described above and several cardinal ME/CFS symptoms: fatigue, unrefreshing sleep, autonomic instability, and the “tired but wired” paradox. Orexin-A and orexin-B are hypothalamic neuropeptides produced by a small population (~70,000 in humans) of neurons in the lateral and posterior hypothalamus. These neurons project widely to cortical, brainstem, and spinal targets and regulate sleep-wake transitions, arousal, autonomic function, and energy homeostasis (Sakurai et al. 1998).
Evidence for orexin suppression in ME/CFS. López-Amador (2025) conducted an integrative review of 27 studies examining the orexin system in ME/CFS and found consistent evidence of reduced orexin-A levels, with variable orexin-B responses that may serve as a subtyping biomarker (López-Amador 2025). The review identified hypothalamic orexinergic dysfunction as a central pathophysiological feature of ME/CFS, supported by concurrent hypocortisolism and autonomic dysregulation. CSF orexin-A levels in ME/CFS appear to fall in an intermediate range (\(\sim\) 250 \(\pm\) 30 pg/mL) — below healthy controls but above the \(<\) 110 pg/mL diagnostic threshold for narcolepsy type 1 (López-Amador 2025) (Rauf et al. 2025).
Mechanism: cytokine-induced orexin suppression. Animal studies have established a precise circuit through which inflammation suppresses orexin neurons. Grossberg et al. (2011) demonstrated that LPS-induced inflammation suppresses perifornical hypothalamic orexin neuron activity, reduces CSF orexin-A, and blocks dark-phase orexin activation (Grossberg et al. 2011). Central orexin-A replacement fully reversed inflammation-induced lethargy, establishing causal direction: orexin suppression drives fatigue rather than merely accompanying it. Importantly, orexin neurons themselves lack direct cytokine receptors; suppression is mediated by lateral hypothalamic neurotensin interneurons rather than the PGE2/melanocortin pathways (Grossberg et al. 2011). Gaykema and Goehler (2009) confirmed this using Fos immunohistochemistry: LPS reduced Fos expression in lateral orexin neurons during exploration (47% → 25%) and dark-phase activity (42% → 9%), with parallel suppression of histaminergic tuberomammillary neurons — suggesting co-shutdown of the orexin–histamine arousal axis (Gaykema and Goehler 2009). Weymann et al. (2014) extended these findings to chemotherapy-induced fatigue, showing the same cytokine → hypothalamic inflammation → orexin suppression → locomotor fatigue pathway, again reversible by central orexin-A rescue (Weymann et al. 2014).
Certainty: 0.50. In ME/CFS, chronic low-grade neuroinflammation — sustained by peripheral cytokine signalling (Section Inflammatory Cytokine-Induced Somnolence and Fatigue) and microglial activation (Section Microglia Activation and Neuroinflammatory Fatigue) — persistently suppresses lateral hypothalamic orexin neurons via neurotensin interneuron-mediated inhibition (Grossberg et al. 2011) (Gaykema and Goehler 2009). The resulting partial orexin deficiency (intermediate between health and narcolepsy) produces:
- Fatigue: Reduced orexin-mediated arousal drive; causal reversal demonstrated by orexin-A rescue in animal fatigue models (Grossberg et al. 2011) (Weymann et al. 2014)
- Unrefreshing sleep: Impaired REM gating produces sleep fragmentation; orexin neuron activity normally suppresses inappropriate REM intrusions during NREM (Ito et al. 2023)
- “Tired but wired” paradox: Partial (not complete) orexin loss may create state-instability where neither wake nor sleep is fully consolidated — the system oscillates between insufficient arousal and insufficient sleep depth
- Autonomic instability: Orexin modulates baroreflex sensitivity, sympathetic tone, and electrolyte balance; dysfunction may contribute to POTS/OI Ruhrländer et al. (2025)
- Circadian desynchronisation: Orexin neurons integrate metabolic, circadian, and immune signals; their dysfunction decouples these systems
Replication status: Not yet replicated as a unified ME/CFS-specific mechanism. Individual components (reduced orexin-A: reviewed across 27 studies (López-Amador 2025); cytokine-orexin suppression: 3 independent animal studies (Grossberg et al. 2011) (Gaykema and Goehler 2009) (Weymann et al. 2014); REM gating by orexin: optogenetic causal proof (Ito et al. 2023)) are well-replicated. No ME/CFS-specific CSF orexin measurement study with simultaneous symptom correlation has been published.
Testable predictions:
- ME/CFS patients with CSF orexin-A \(<\) 200 pg/mL have more severe fatigue and unrefreshing sleep than those with higher levels
- Anti-inflammatory interventions (LDN, tocilizumab) that reduce neuroinflammation should increase CSF orexin-A levels and improve fatigue
- Central orexin replacement (OX2R agonist) should improve fatigue and sleep quality in orexin-low ME/CFS patients
- Orexin-B/orexin-A ratio distinguishes ME/CFS subtypes with different treatment response profiles
Treatment implication: OX2R-selective agonists (danavorexton, TAK-861/oveporexton) are in clinical development for narcolepsy (Rauf et al. 2025). If ME/CFS involves partial orexin deficiency, these agents represent a mechanistically rational therapeutic class — but no ME/CFS trial has been conducted. This is a research hypothesis, not a recommendation.
Certainty: 0.25. Parhizkar et al. (2025) demonstrated in P301S/E4 tauopathy mice that lemborexant — a dual orexin receptor antagonist (DORA) — reduces tau phosphorylation via cAMP/PKA pathway inhibition, preventing neurodegeneration independently of sleep promotion (Parhizkar et al. 2025). The critical dissociation: zolpidem increased sleep duration but provided no neuroprotection, establishing orexin signaling (not sleep duration) as the mechanistic driver. Lucey et al. (2023) provided human validation in an RCT (n=38): suvorexant 20 mg reduced CSF p-tau181/T181 ratio by ~10–15% in cognitively unimpaired adults (Lucey et al. 2023). Together, these studies identify a production-prevention pathway — orexin antagonism reduces tau phosphorylation before aggregation begins — that is mechanistically distinct from the clearance-enhancement pathway (glymphatic removal of already-accumulated waste) and the damage-reversal pathway (HSP70-mediated repair, Chapter Neurological and Neurocognitive Dysfunction).
Relevance to ME/CFS. Three axes of relevance, none yet tested in ME/CFS:
Differential DORA safety: The paper’s existing glymphatic medication warning (Section Glymphatic Dysfunction and Brain Waste Accumulation) lists DORAs among potentially glymphatic-impairing agents via LC-NE pathway effects (Zhu, Yang, and Hashimoto 2025). However, Parhizkar et al. demonstrate a countervailing benefit: even if DORAs mildly suppress NE oscillations, they may simultaneously reduce tau phosphorylation via PKA. This creates an unresolved risk–benefit trade-off specific to DORAs (not Z-drugs, which carry glymphatic impairment without tau protection). The net effect in ME/CFS — where orexin is already suppressed — is unknown.
Orexin tone paradox: ME/CFS involves partial orexin suppression (Section Constraints on, and Rival Readings of, the Torpor/Sickness-Circuit Model), while Parhizkar 2025’s protection requires pharmacological orexin antagonism. If endogenous orexin is already low, the PKA pathway may already be downregulated in ME/CFS — potentially providing endogenous protection against tau hyperphosphorylation. Conversely, the “tired but wired” phenotype suggests incomplete suppression: residual orexin tone during supposed “sleep” periods may drive both sleep fragmentation and pathological tau phosphorylation, in which case DORAs could provide dual benefit. This is entirely unresolved.
Neurodegeneration risk context: The paper’s chronic glymphatic impairment speculation (Section Can Coupling Strength Be Estimated from Reduced-Modality Data?) predicts elevated long-term tau/amyloid accumulation in ME/CFS from impaired clearance. Parhizkar 2025 adds the complementary possibility that tau may be produced faster (via orexin→PKA→phosphorylation) as well as cleared slower (via glymphatic failure) — a double hit on tau homeostasis. This is a cross-disease extrapolation from a tauopathy model; no ME/CFS tau phosphorylation data exist.
Falsifiable prediction: ME/CFS patients should show higher CSF p-tau181/T181 ratios than age-matched controls, with ratio correlating with CSF orexin-A levels and unrefreshing sleep severity.
Limitations: Single preclinical study, male mice only, tauopathy model (not ME/CFS). Human validation (Lucey 2023) is acute single-dose, cognitively unimpaired participants. The PKA pathway has not been studied in ME/CFS. All claims are cross-disease extrapolation.
Certainty: 0.25. This is a hypothetical pathway proposed from mechanism, with no primary evidence that daridorexant — or any dual orexin receptor antagonist (DORA) — reduces \(\beta\)-amyloid (\(A \beta\)) in humans or animals. The existing tau arm (Section Orexin→PKA→Tau Phosphorylation: A Production-Side Mechanism Distinct from Glymphatic Clearance) is supported by real data (suvorexant reduced CSF p-tau181 by ~10–15% in humans (Lucey et al. 2023); lemborexant reduced tau phosphorylation in mice (Parhizkar et al. 2025)), but tau is a different protein from \(A \beta\), and those data are for suvorexant and lemborexant, not daridorexant. The extension from tau to \(A \beta\) is analogy, not evidence; a literature search finds only review-level discussion of the orexin–Alzheimer’s rationale, with no primary daridorexant+\(A \beta\) study.
The hypothetical pathway. Three documented links, none yet combined as a tested drug effect, would jointly predict a daridorexant-driven reduction in \(A \beta\): (1) orexin signaling promotes neuronal activity and activity-dependent \(A \beta\) production — so orexin blockade could reduce \(A \beta\) production upstream; (2) slow-wave sleep drives glymphatic \(A \beta\) clearance (Section Glymphatic Dysfunction and Brain Waste Accumulation) — so daridorexant’s improvement and consolidation of sleep architecture (St Onge, Phillips, and Rowe 2022) could enhance \(A \beta\) clearance. A tau analogy ((Parhizkar et al. 2025)) bears only on a general “orexin→proteostasis” theme, not on \(A \beta\) production directly — tau and \(A \beta\) are generated by different proteolytic pathways, so the tau data do not mechanistically support an \(A \beta\)-production effect. The two arms that are directly relevant — production-side reduction (from orexin blockade) and clearance-side enhancement (from better sleep) — could sum to a net reduction in brain \(A \beta\) burden over chronic use.
Critical competing arm (must not be omitted). The pathway is not unambiguously beneficial. DORAs may impair glymphatic clearance by suppressing norepinephrine-mediated vasomotion (Zhu, Yang, and Hashimoto 2025) (Section Glymphatic Dysfunction and Brain Waste Accumulation), so a daridorexant-driven increase in sleep duration could coexist with reduced per-hour clearance quality. Whether the net effect on \(A \beta\) is reduction, neutral, or even accumulation is unresolved — both directions are mechanistically plausible, exactly as the chapter’s existing DORA risk–benefit analysis already notes for tau.
Relevance to ME/CFS. ME/CFS is documented to have chronic glymphatic-impairment risk (Section Can Coupling Strength Be Estimated from Reduced-Modality Data?), so the long-term \(A \beta\)/tau-accumulation question has potential relevance; but no ME/CFS-specific \(A \beta\) or tau phosphorylation data exist, and daridorexant has not been tested for any \(A \beta\) endpoint in any population. This speculation is therefore cross-disease and entirely mechanistic — it does not alter the sleep-management guidance for daridorexant already given (Chapter Action Plans for Mild to Moderate Cases) and is not a basis for treatment expectation.
Falsifiable prediction: If the pathway holds, chronic DORA use (e.g., daridorexant 25–50 mg nightly) would reduce CSF \(A \beta\) 42, \(A \beta\) 40, or the \(A \beta\) 42/\(A \beta\) 40 ratio relative to baseline over weeks-to-months in cognitively unimpaired adults — a testable outcome that no study has yet run. A null result (no \(A \beta\) change despite improved sleep) would be inconsistent with the net-reduction prediction but would not cleanly adjudicate the production arm, because the competing glymphatic-impairment arm could mask a genuine production-side reduction; it would leave the tau arm (Section Orexin→PKA→Tau Phosphorylation: A Production-Side Mechanism Distinct from Glymphatic Clearance) intact regardless.
Limitations: No primary daridorexant+\(A \beta\) study exists; the pathway is inferred from (a) the orexin→\(A \beta\) production link and (b) sleep→glymphatic clearance, with the DORA tau data ((Parhizkar et al. 2025), (Lucey et al. 2023)) relevant only to a general orexin→proteostasis theme, not to \(A \beta\) production. The glymphatic-impairment arm could oppose any benefit. All claims are cross-disease extrapolation; nothing here is demonstrated.
Consequence: For a non-specialist: daridorexant is a sleep drug already used in ME/CFS. Related sleep drugs in the same class (suvorexant, lemborexant) have real evidence of helping protect the brain by reducing tau, and daridorexant might share that effect — but the idea that it also reduces the Alzheimer’s protein \(A \beta\) is only a mechanistically plausible guess, not a proven effect, and it could even slow waste removal in some patients. Its value for sleep stands on its own evidence; the \(A \beta\) claim should not be relied on.
The two speculations in this section (Orexin→PKA→Tau Phosphorylation: A Production-Side Mechanism Distinct from Glymphatic Clearance and Daridorexant to Beta-Amyloid Reduction: A Hypothetical, Untested Pathway) jointly argue a convergent point that neither states alone: dual orexin receptor antagonists (DORAs) may exert a neuroprotective effect on the brain’s protein-clearance/production axis through two distinct arms of orexin antagonism. The tau arm is evidenced: suvorexant reduced CSF p-tau181 by ~10–15% in humans ((Lucey et al. 2023)) and lemborexant reduced tau phosphorylation and protected hippocampal volume in mice ((Parhizkar et al. 2025)), via orexin→PKA pathway inhibition that is independent of sleep duration. The \(A \beta\) arm is hypothetical: no primary study shows any DORA reducing \(\beta\)-amyloid; the pathway is inferred from orexin’s promotion of activity-dependent \(A \beta\) production and from sleep-dependent glymphatic \(A \beta\) clearance, but tau and \(A \beta\) are generated by different proteolytic pathways, so the tau data do not mechanistically support an \(A \beta\)-production effect.
The convergence does not make the two arms additive. They are constrained by a shared competing effect: DORAs may impair glymphatic clearance by suppressing norepinephrine-mediated vasomotion ((Zhu, Yang, and Hashimoto 2025)), so any neuroprotection from reduced tau/\(A \beta\) production could be offset by reduced waste clearance. For tau, the risk-benefit is genuinely unresolved; for \(A \beta\), the competing arm makes even the direction of the net effect uncertain (reduction, neutral, or accumulation). The honest bottom line across both arms: DORAs plausibly act on orexin-driven protein dyshomeostasis, but the neuroprotective claim is evidenced only for tau, and only for suvorexant/lemborexant, not daridorexant — and the glymphatic-impairment arm could negate either benefit. This is a research hypothesis, not a basis for treatment expectation; daridorexant’s value for sleep stands on its own evidence.
Consequence: For a non-specialist: this class of sleep drug may, as a side benefit, help protect the brain by reducing one Alzheimer’s-linked protein (tau) — that part has real evidence. Whether it also reduces a second one (beta-amyloid) is only a plausible guess, and it could even slow the brain’s waste removal in some patients. So the neuroprotection idea is interesting but not a reason to take the drug; its sleep benefit is the established reason.
5.1 Pharmacological Convergences on the cAMP/PKA Tau Phosphorylation Axis
Certainty: 0.15–0.25 across agents. The Parhizkar et al. (Parhizkar et al. 2025) demonstration that DORAs reduce tau phosphorylation via cAMP/PKA suggests that other drug classes acting on the cAMP/PKA pathway — or on parallel tau kinases — may share this neuroprotective mechanism. None of the agents below have been tested for tau effects in ME/CFS; all claims are cross-disease pharmacological extrapolation.
DORA class comparison. All dual orexin receptor antagonists should share the PKA-mediated tau phosphorylation reduction, but site-specificity may differ. Lucey et al. (2023) showed suvorexant 20 mg reduced CSF p-tau181/T181 by ~10–15% in humans but did not affect phosphorylation at S202 or T217 (Lucey et al. 2023). Parhizkar et al. demonstrated lemborexant reduces phosphorylation at Ser202, Ser409, and Thr205 in mice (Parhizkar et al. 2025). Whether this represents DORA-specific pharmacology (lemborexant blocking OX1R and OX2R with different kinetics than suvorexant), species differences (mouse vs human), or chronic-vs-acute dosing effects is unresolved. Daridorexant — the third approved DORA, already discussed for ME/CFS sleep (Chapter Action Plans for Mild to Moderate Cases) — has not been tested for tau phosphorylation effects. No published study has directly compared tau phosphorylation profiles across DORAs.
Propranolol via β-adrenergic→cAMP→PKA. β-adrenergic receptors signal through Gs → adenylyl cyclase → cAMP → PKA — the same intracellular cascade downstream of orexin receptors. Non-selective β-blockers such as propranolol may therefore partially dampen PKA activity through a receptor population (β-adrenergic) that is mechanistically convergent with orexin signalling at the level of cAMP production. This creates a dual-benefit rationale already partially present in the paper: low-dose bedtime propranolol is discussed for autonomic coupling restoration (Section Post-Exertional Malaise May Involve Inflammation-Induced Routing Disruption of Brain Clearance); the PKA dampening provides an independent — but entirely untested — neuroprotective argument. Key caveat: propranolol does not block orexin receptors, and the relative contribution of β-adrenergic vs orexin signalling to neuronal cAMP/PKA tone is unknown. The Hussain et al. (2025, SfN presentation) experimental TBI cocktail combined an alpha-blocker with a beta-blocker to improve glymphatic clearance, not to modulate tau phosphorylation — the PKA mechanism is distinct.
Prazosin revisited. The paper already includes prazosin among glymphatic-impairing medications (Section Glymphatic Dysfunction and Brain Waste Accumulation), citing Zhu et al. (Zhu, Yang, and Hashimoto 2025) for NE oscillation suppression. However, alpha-1 adrenergic receptors also signal through Gq → phospholipase C → IP3/DAG → calcium — a pathway that also activates PKA indirectly via calcium-sensitive adenylyl cyclase. Iliff’s group (University of Washington) found in a retrospective study that veterans taking prazosin for trauma-related nightmares showed reduced dementia risk, and they are now prospectively testing prazosin’s glymphatic effects in rodents and nonhuman primates. The two mechanisms — NE oscillation suppression (potentially impairing glymphatic clearance) vs PKA pathway dampening (potentially reducing tau phosphorylation) — are in opposition for glymphatic outcomes but may be synergistic for tau outcomes. This tension is unresolved; the retrospective dementia finding suggests the net effect may favour neuroprotection in some populations. No ME/CFS prazosin data exist.
Lithium (GSK3β inhibitor). GSK3β is a major tau kinase distinct from PKA — it phosphorylates tau at Ser396/404 rather than the Ser202/Ser409/Thr205 sites affected by DORAs. Lithium is an established GSK3β inhibitor at therapeutic concentrations (0.6–1.2 mmol/L). Guttuso et al. (2024) conducted an RCT (n=60) of low-dose lithium aspartate (20–25 mg/day, far below bipolar dosing) in Long COVID cognitive impairment and found significant improvement in fatigue and cognitive endpoints, though the mechanism (GSK3β vs neurotrophic vs anti-inflammatory) was not established (Guttuso, Zhu, and Zahra 2024). Relevance to the tau phosphorylation discussion is speculative: if tau phosphorylation contributes to ME/CFS neurocognitive symptoms (unestablished), lithium could provide complementary tau kinase inhibition (GSK3β) to DORAs’ PKA inhibition, targeting different phospho-sites. This is a multi-kinase tau suppression concept entirely untested in any condition.
Valproate (HDAC inhibitor with GSK3β inhibition). Valproate inhibits both histone deacetylases (HDACs) and GSK3β — providing a dual mechanism of epigenetic modulation (potentially relevant to the epigenetic consolidation lock, Chapter Causal Hierarchy: Root Causes, Amplifiers, and Consequences) and tau kinase inhibition. Like lithium, it is an “old, dirty drug” with established human safety data but significant adverse effect burden (tremor, weight gain, hepatic effects, teratogenicity) that limits its candidacy for ME/CFS. The GSK3β inhibitory component at clinically achievable concentrations is modest. No ME/CFS valproate data exist.
Resolving the orexin tone paradox. The Parhizkar et al. temporal finding — active-phase dosing (suppressing high endogenous orexin) more effective than inactive-phase — initially appears to complicate ME/CFS, where orexin is already partially suppressed (~250 pg/mL CSF vs narcolepsy \(<\) 110 pg/mL). However, the resolution is that the ME/CFS problem is not absolute level but circadian misplacement: the “tired but wired” phenotype reflects orexin that is too low during wake (insufficient arousal) and too high during sleep (failed withdrawal, producing alpha-delta intrusion and fragmented NREM). The nocturnal orexin signal — the signal that should be absent during deep sleep but persists in ME/CFS — is the very signal that DORAs suppress. Bedtime DORA administration targets precisely the pathologically elevated nocturnal orexin window without suppressing the already-deficient daytime tone (DORAs have a ~10-hour half-life, clearing by morning). The active-phase dosing in mice (lights-off, when orexin peaks) is mechanistically analogous to bedtime dosing in humans (sleep onset, when orexin should withdraw but fails to in ME/CFS). This reframes the paradox as a dosing alignment question: DORAs are expected to provide dual benefit in ME/CFS — sleep consolidation + PKA-mediated tau protection — because they target the nocturnal orexin surplus rather than the daytime deficiency. The critical caveat is that intermittent or late-night DORA use could still suppress daytime orexin if the drug persists past morning awakening; short-acting DORAs (suvorexant, ~12h half-life at 20 mg) may be preferable to longer-acting formulations for ME/CFS patients specifically to avoid encroaching on the already-deficient daytime orexin window.
5.2 Medications and Glymphatic Clearance: Reconciling the Evidence
Certainty: 0.15–0.30. The paper’s existing glymphatic medication warning (Section Glymphatic Dysfunction and Brain Waste Accumulation) groups Z-drugs and DORAs together as potentially impairing glymphatic clearance via NE oscillation suppression. Three lines of evidence now suggest this grouping may be incorrect — DORAs and Z-drugs have opposing effects on glymphatic function, and several agents may actually enhance clearance:
DORAs: unmeasured glymphatic effects, but evidence of tau production-side benefit via PKA. The Zhu et al. (Zhu, Yang, and Hashimoto 2025) citation in the existing warning infers DORA harm from general adrenergic pharmacology but never directly tested DORAs on glymphatic flow. The mechanistic direction is unclear: (1) orexin withdrawal during sleep is normal physiology — DORAs facilitate this natural withdrawal, which could permit the LC to settle into the infraslow (\(\sim\) 0.05 Hz) oscillatory pattern that Hauglund et al. (Hauglund et al. 2025) identified as the glymphatic pump driver; (2) however, ME/CFS already involves partial orexin suppression (Section Constraints on, and Rival Readings of, the Torpor/Sickness-Circuit Model), and further pharmacological suppression could worsen clearance if orexin tone is already below the optimal range for LC oscillatory dynamics. Neither prediction has been directly tested: no study has measured glymphatic flow before and after any DORA in any species. (3) Lucey et al. (2023) demonstrated suvorexant reduced CSF p-tau181 by ~10–15% in sleeping humans (Lucey et al. 2023) — but this could operate via PKA-mediated reduction of tau phosphorylation (a production-side mechanism, Section Orexin→PKA→Tau Phosphorylation: A Production-Side Mechanism Distinct from Glymphatic Clearance) rather than glymphatic clearance enhancement, or via improved sleep architecture independent of glymphatic pumping. The p-tau reduction is consistent with DORA benefit but does not distinguish between mechanisms. (4) Parhizkar et al. showed lemborexant preserved hippocampal volume by 30–40% in tauopathy mice while zolpidem gave zero protection (Parhizkar et al. 2025) — again, consistent with either production-side PKA inhibition or clearance enhancement. In summary: DORAs have downstream evidence of tau-related neuroprotection but no direct glymphatic measurement. The net glymphatic effect — enhancing (facilitated NE withdrawal) or impairing (further orexin suppression in already-suppressed ME/CFS) — is unknown. The current evidence supports DORAs as mechanistically distinct from Z-drugs in tau outcomes, not in glymphatic clearance per se.
Propranolol: normalising NE oscillations, not suppressing them. Low-dose bedtime propranolol (10–20 mg, already discussed for autonomic coupling in Section Post-Exertional Malaise May Involve Inflammation-Induced Routing Disruption of Brain Clearance) may improve glymphatic clearance by damping excessive nocturnal sympathetic tone — the “brain asleep, heart awake” decoupling pattern that SleepFM identified as the strongest disease predictor (Thapa et al. 2026). In ME/CFS dysautonomia, NE is not simply elevated or suppressed but dysregulated: sympathetic surges during supposed NREM sleep fragment sleep architecture, while LC fails to settle into coherent infraslow oscillation. Propranolol at low dose acts as a normaliser, not a suppressor — reducing pathological sympathetic surges without abolishing the low-amplitude oscillatory signal needed for vasomotion. This is distinct from high-dose beta blockade which would suppress both pathological and physiological NE signalling.
Clonidine: stabilising LC oscillatory amplitude. Low-dose clonidine (0.025–0.05 mg at bedtime, titrated over weeks) reduces LC firing rate but preserves oscillatory competence, unlike Z-drugs which suppress oscillation amplitude by ~50% (Hauglund et al. 2025). By reducing the excessive tonic NE drive that keeps the LC from settling into its oscillatory state, clonidine may paradoxically increase oscillation quality — restoring the infraslow rhythm that drives vasomotion and glymphatic flow. No direct glymphatic data exist for clonidine; this inference is from the LC-NE oscillation physiology established by Hauglund et al.
Trazodone: SWS consolidation without NE suppression. Low-dose trazodone (25–50 mg) increases slow-wave sleep duration and consolidates sleep architecture without suppressing NE oscillations (Zhu, Yang, and Hashimoto 2025). This makes it mechanistically distinct from Z-drugs: it improves the window for glymphatic clearance (more SWS) without impairing the pump (preserved NE oscillations). Trazodone is already discussed in the paper’s treatment chapters for ME/CFS sleep; the glymphatic argument provides an additional mechanistic rationale beyond subjective sleep improvement.
Revised medication classification. The existing glymphatic warning should be read with the following revision, pending direct experimental glymphatic measurements in ME/CFS:
| Agent | NE oscillation effect | Net glymphatic prediction |
|---|---|---|
| Z-drugs (zolpidem) | Suppress amplitude \(\sim\) 50% | Impair |
| DORAs (suvorexant, lemborexant) | May normalise OR may worsen (already-suppressed ME/CFS orexin) | Unknown (production-side tau benefit via PKA established; net glymphatic effect unmeasured) |
| Low-dose propranolol | Normalise (dampen surges, preserve oscillation) | Potential enhance |
| Low-dose clonidine | Stabilise (reduce tonic drive, preserve oscillatory) | Potential enhance |
| Low-dose trazodone | Neutral (consolidates SWS window) | Potential enhance via window extension |
| Alpha-1 blockers (prazosin) | Suppress amplitude | Potential impair |
| Quetiapine (low-dose) | Suppress via alpha-1 blockade | Potential impair |
The critical clinical implication: DORAs demonstrate downstream neuroprotection (tau reduction) independently of glymphatic function, but the mechanism may be production-side (PKA inhibition) rather than clearance-side. Z-drugs lack both glymphatic enhancement and tau neuroprotection. Neither drug class has been directly tested for glymphatic effects in humans. This uncertainty should be communicated clearly to patients considering sleep medication choices: the evidence for DORA benefit is more promising than for Z-drugs, but the mechanism and magnitude of any glymphatic effect remain entirely unknown.
Prescribing framework: when, how often, which events. The above classification is mechanistic; the practical question is when each agent should be taken, whether nightly or as needed, and which clinical events trigger use.
DORAs — nightly, scheduled. Daridorexant 25–50 mg 30 minutes before bed. The 52-week extension study demonstrated no tolerance or withdrawal with either continuous or intermittent use (Kunz et al. 2022), making DORAs the only sleep medication class with explicit intermittent safety data. However, glymphatic clearance is cumulative — each non-restorative night adds metabolic waste that the next night must clear. The rationale for scheduled nightly use is that ME/CFS sleep architecture is persistently impaired (not episodic), and the neuroinflammatory-glympathic vicious cycle operates on a nightly basis, not an occasional one. PRN use — taking a DORA only after a subjectively bad night — is mechanistically counterproductive: by the time the patient notices non-restorative sleep, the waste has already accumulated, and the missed clearance window has closed. Dose timing is critical: the drug must be taken 30 minutes before bedtime to suppress the nocturnal orexin signal that should be withdrawing but fails to in ME/CFS. Taking a DORA at 3 AM after waking unfreshed is too late — the glymphatic window in the first NREM cycle is already lost.
Trazodone — nightly, scheduled. 25–50 mg at bedtime. The rationale for scheduled use is identical: SWS consolidation must happen during sleep, not after it fails. Trazodone’s mechanism (SWS window extension without NE suppression) makes it complementary to DORAs: DORAs normalise the orexin→NE oscillatory signal, trazodone extends the window in which that oscillation drives clearance.
Low-dose propranolol/clonidine — nightly, scheduled, at bedtime. Both agents function as NE oscillation normalisers, and the glymphatic pump operates during each NREM cycle. However, propranolol also has a PRN use case documented in Chapter Integrative and Personalized Treatment Approaches: 10–20 mg before high-risk activities as a pacing enforcement tool (pharmacological heart rate ceiling). This PRN use is for daytime pacing, not nocturnal glymphatic support — the mechanisms are distinct.
After PEM. Sleep architecture fragmentation predictably worsens during post-exertional malaise (Section Pupillometry as Noninvasive LC Functional Readout in ME/CFS), with CNS energy deficits intensifying 24–72 hours post-exertion. This is the event where glymphatic-enhancing agents are most needed — the waste production is highest and the endogenous clearance capacity is lowest. The clinical implication: the night before an anticipated PEM window (e.g., if the patient has a medical appointment that will trigger PEM) is not the time to skip the scheduled DORA/trazodone dose. Conversely, adding an extra agent on the night after a crash has no role — the glymphatic window cannot be recovered retrospectively, and polypharmacy at bedtime increases the risk of accumulated sedation and cognitive impairment. The strategy is consistent nightly scheduled use, with dosing adjustments made pre-emptively, not reactively.
What to avoid. Z-drugs should not be used as “rescue sleep” after a bad night — the evidence now shows they increase sleep duration while suppressing the NE oscillations needed for glymphatic clearance, and provide zero tau neuroprotection. They solve the problem the patient feels (wakefulness) while worsening the problem the patient needs solved (waste clearance). The same applies to PRN quetiapine or prazosin at bedtime for sleep — the alpha-1 blockade component impairs vasomotion regardless of dose timing.
Contraindications and side effects. The glymphatic benefit of these agents must be weighed against their safety profiles, which differ substantially. Trazodone is the lowest-risk entry point; DORAs have the strongest mechanistic rationale but the highest practical barriers.
- Trazodone 25–50 mg. Key contraindications: MAOI co-administration (serotonin syndrome risk). Critical side effects: next-day sedation at doses above 50 mg; priapism (~1/6000, rare but permanent damage if untreated — warn male patients); orthostatic hypotension. ME/CFS-specific concern: already-impaired daytime alertness worsens if dose too high; keep at or below 50 mg.
- DORA (daridorexant) 25–50 mg. Key contraindications: narcolepsy (absolute); severe hepatic impairment (Child-Pugh C); strong CYP3A4 inhibitors. Critical side effects: sleep paralysis (less than 1 percent); complex sleep behaviours (sleep-walking, rare); next-day somnolence (dose-dependent, ~3 percent at 25 mg); headache. ME/CFS-specific concern: borderline CSF orexin levels in some patients raise narcolepsy misdiagnosis concern; $350–$450/month without insurance; Schedule IV controlled substance — many clinicians refuse as policy; no ME/CFS-specific data.
- Propranolol 10–20 mg. Key contraindications: asthma/COPD (beta-blockade triggers bronchospasm); bradycardia (HR below 60); hypotension (SBP below 100). Critical side effects: fatigue worsening, exercise intolerance, cold extremities, vivid nightmares. ME/CFS-specific concern: already documented as harm in treatment chapters; ivabradine preferred if pure HR reduction needed; low-dose bedtime use for glymphatic normalisation is entirely speculative with no published human glymphatic data.
- Clonidine 0.025–0.05 mg. Key contraindications: hypotension (SBP below 100); bradycardia; severe coronary insufficiency. Critical side effects: morning sedation, dry mouth, rebound hypertension on abrupt discontinuation. ME/CFS-specific concern: no adult ME/CFS dosing data exists; orthostatic intolerance may worsen; dangerous if patient is too fatigued to refill on time; entirely speculative for glymphatic use — no published data.
Of these four, only trazodone is a straightforward prescription: generic (~$4/month), 50+ years of safety data, universally accepted off-label use for insomnia. DORAs face the highest refusal risk — Schedule IV status, high cost, no FDA indication for ME/CFS, and the legitimate absence of ME/CFS-specific outcome data. A reasonable clinician would correctly state that evidence for DORAs in ME/CFS does not exist. Propranolol and clonidine for nocturnal NE normalisation are entirely speculative — no published human glymphatic data for either agent — and refusal on evidence-based grounds is appropriate. The paper provides mechanistic rationale to motivate trials, not clinical assertions that override the absence of data.
Falsifiable predictions. (a) DTI-ALPS glymphatic index will improve after 4 weeks of lemborexant vs placebo in ME/CFS patients, while zolpidem will show no change or worsening. (b) CSF p-tau181/T181 ratio will decrease on DORA and remain unchanged or increase on zolpidem, matched for sleep duration improvement. (c) NE oscillation amplitude during NREM sleep will increase on DORA, remain stable on propranolol/clonidine/trazodone, and decrease on zolpidem/prazosin.
Falsifiable prediction. A head-to-head comparison of lemborexant vs placebo in ME/CFS, with CSF p-tau181/T181 ratio as primary endpoint at 4 weeks, would directly answer: (a) whether ME/CFS patients have elevated baseline p-tau vs controls, and (b) whether further orexin suppression meaningfully reduces it. A null result would suggest either that ME/CFS does not involve tau hyperphosphorylation, or that endogenous orexin suppression has already “taken care of it.” A positive result would motivate DORA + GSK3β inhibitor combination studies for multi-kinase tau suppression.
Post-infectious context. The orexin suppression pathway has direct parallels across post-infectious conditions. Narcolepsy type 1 is a confirmed post-infectious orexin disease: H1N1 influenza and the Pandemrix vaccine triggered autoimmune destruction of \(>\) 95% of orexin neurons in genetically susceptible individuals (HLA-DQB1*06:02) (Rauf et al. 2025). ME/CFS may occupy a less severe position on the same spectrum: functional suppression rather than destruction, intermediate orexin levels rather than abolition, and reversibility (in principle) if the inflammatory drive resolves. In Long COVID/PASC, Ruhrländer et al. (2025) proposed orexin disruption as an explanation for POTS, endothelial dysfunction, and metabolic dysregulation Ruhrländer et al. (2025). Heinicke et al. (2025) measured plasma orexin-A in 78 ICU COVID patients and found that levels correlated with survival (Cohen’s \(d=0.4\)) and inversely with length of stay, with neuropeptide perturbations persisting at 2.5-year follow-up (Heinicke et al. 2025).
Certainty: 0.35. (0.30→0.35: SleepFM (Thapa et al. 2026) demonstrates that cross-modal physiological decoupling during sleep predicts disease onset across 130+ conditions, supporting the coupling-integrity mechanisms described in this loop at the general-principle level; this does not validate any specific ME/CFS mechanism) The orexin suppression and glymphatic failure mechanisms described above may form a closed positive feedback loop entirely within the CNS:
Neuroinflammation (activated microglia) → orexin neuron suppression (via PGE2/EP3 signalling at the BBB endothelium and/or neurotensin interneuron-mediated inhibition in the lateral hypothalamus — the two routes may be parallel or context-dependent (Grossberg et al. 2011)) → reduced orexin tone → impaired LC NE oscillatory quality → reduced vasomotion amplitude → impaired glymphatic clearance → metabolite accumulation (tau, amyloid-\(\beta\), adenosine) → microglial activation → more neuroinflammation
Once established, this loop operates independently of peripheral immune activation. Even if peripheral cytokine drive resolves (e.g., viral clearance), the CNS PGE2 source (activated microglia responding to accumulated waste) is self-sustaining. This may explain why ME/CFS persists long after the triggering infection has cleared.
Testable predictions:
- CSF PGE2 levels should correlate inversely with CSF orexin-A in ME/CFS patients
- COX-2 inhibitors (which reduce PGE2) should transiently improve sleep quality (measurable by overnight PSG delta power) in ME/CFS
- EP3-selective antagonists should restore orexin neuron activity and improve glymphatic clearance in animal models of post-infectious fatigue
Treatment implication: If this loop is a primary disease maintenance mechanism, breaking it at the PGE2 node (anti-inflammatory), the orexin node (OX2R agonist), or the glymphatic node (SWS-enhancing interventions) could each be sufficient to interrupt the cycle — consistent with the multi-lock framework’s prediction that some single-node interventions succeed in mild/moderate subtypes (Chapter Causal Hierarchy: Root Causes, Amplifiers, and Consequences). Research-stage hypothesis only, not a clinical recommendation.
Certainty: 0.40. (0.35→0.40: SleepFM (Thapa et al. 2026) demonstrates cross-modal decoupling during sleep predicts disease onset across 130+ conditions, supporting the orexin-mediated coupling disruption mechanism at the general-principle level; no specific ME/CFS mechanism validated) Orexin neurons in the lateral hypothalamus project densely to the locus coeruleus (LC) (Sakurai et al. 1998). During normal sleep, orexin withdrawal permits LC norepinephrine (NE) to settle into the infraslow (\(\sim\) 0.05 Hz) oscillatory pattern that drives cerebral vasomotion and glymphatic clearance (Hauglund et al. 2025). If orexin-A is chronically low (as in ME/CFS, Section Constraints on, and Rival Readings of, the Torpor/Sickness-Circuit Model), the LC never receives a clean on–off signal. Instead of crisp wake-state (high tonic NE) transitioning to sleep-state (infraslow NE oscillations), the LC operates in a dysregulated intermediate state: insufficient tonic drive for full wakefulness (daytime fatigue), yet failing to settle into the coherent infraslow oscillation needed for glymphatic pumping.
This reframes orexin deficiency as simultaneously a sleep quality problem (via disrupted LC oscillatory dynamics) and a wakefulness problem — explaining the “tired but wired” paradox where neither wake nor sleep is fully consolidated. SleepFM’s cross-modal decoupling framework (Thapa et al. 2026) predicts that the EEG–autonomic dissociation produced by orexin deficiency should be detectable as elevated reconstruction error during NREM sleep. No published paper connects orexin deficiency specifically to impaired NE oscillatory quality (as distinct from NE level).
If orexin-A (OX1R-preferring) is selectively reduced while orexin-B (OX2R-preferring) is preserved (López-Amador 2025), the result is: the histamine arousal system (OX2R-driven, tuberomammillary nucleus) remains partially active, while LC modulation (OX1R-driven) is impaired (Sakurai et al. 1998) — producing the combination of residual arousal drive with disrupted sleep architecture.
Testable predictions:
- ME/CFS patients with lower CSF orexin-A should show reduced DTI-ALPS glymphatic index
- ME/CFS patients should show reduced coherence of infraslow pupil oscillations (an LC NE proxy) vs controls
- The orexin-B/orexin-A ratio should correlate with “tired but wired” severity and inversely with objective SWS duration
Certainty: 0.30. (0.25→0.30: SleepFM (Thapa et al. 2026) demonstrates cross-modal decoupling during sleep predicts disease onset across 130+ conditions, supporting the alpha-delta decoupling mechanism at the general-principle level; no specific ME/CFS mechanism validated) The NE infraslow oscillation (\(\sim\) 0.05 Hz) drives vasomotion at a specific frequency (Hauglund et al. 2025). Thalamocortical delta oscillations (0.5–4 Hz) nest within these slower NE oscillations — the two must be phase-locked for optimal glymphatic flow, because the neural slow wave drives blood volume changes that in turn drive CSF pulsation (Fultz et al. 2019).
If CaV3.1 T-type calcium channels are dysfunctional in ME/CFS (Section Sleep Architecture Failure Hypothesis), producing alpha instead of delta oscillations, the thalamocortical oscillations may no longer phase-lock properly with the NE infraslow rhythm. The result is not just impaired sleep quality but specifically impaired resonance between the neural oscillation and the vascular oscillation. Both rhythms may be present but decoupled — producing an incoherent oscillatory pattern that cannot drive coordinated fluid dynamics. This predicts that alpha-delta sleep is worse for glymphatic function than simply reduced delta power, because incoherent oscillations actively disrupt the sequential slow-wave → blood-volume-change → CSF-pulse cascade.
Testable predictions:
- Phase-amplitude coupling between infraslow (\(\sim\) 0.05 Hz) and delta (0.5–4 Hz) EEG oscillations should be reduced in ME/CFS versus controls during NREM sleep
- Alpha-delta sleep phenotype should show worse DTI-ALPS glymphatic scores than low-delta-without-alpha sleep, distinguishing frequency disruption from amplitude disruption
- Interventions that specifically restore delta frequency (not just increase total slow-wave power) should improve glymphatic metrics
Research feasibility: Phase-coupling analysis can be performed computationally on existing ME/CFS polysomnography datasets at minimal cost — no new data collection required.
Certainty: 0.30. (0.25→0.30: SleepFM (Thapa et al. 2026) demonstrates cross-modal decoupling during sleep predicts disease onset across 130+ conditions, supporting the glymphatic-coupling mechanism at the general-principle level; no specific ME/CFS mechanism validated) Physical and cognitive exertion increase neuronal metabolic waste production (extracellular potassium, lactate, adenosine, and potentially tau fragments). In health, the next sleep episode clears this waste via glymphatic flow. In ME/CFS with impaired glymphatic clearance, exercise-generated waste accumulates additively because overnight clearance is insufficient.
This predicts a ratchet effect: each day’s activity adds waste that is only partially cleared overnight. The cumulative effect manifests as progressive cognitive deterioration over days of normal-for-the-patient activity, with recovery requiring multiple days of rest during which even the impaired glymphatic system eventually catches up.
This mechanism explains why cognitive PEM can occur independently of muscular PEM — the brain has its own waste accumulation dynamic independent of muscle damage via the NCX1/AIMM pathway (Chapter Energy Metabolism and Mitochondrial Function).
Testable predictions:
- Cognitive PEM severity should correlate with objective sleep quality (delta power) from the preceding night
- Plasma NfL or p-tau should show transient increases after cognitive exertion in ME/CFS patients but not controls
- Interventions improving SWS quality should improve cognitive PEM recovery (processing speed at 48h post-exertion, greater than 50 percent return to baseline) more than muscular PEM recovery (handgrip endurance at 48h post-exertion, less than 20 percent return to baseline). A null result — equivalent or reversed cognitive-muscular recovery — refutes selective cognitive benefit.
Certainty: 0.30. (0.25→0.30: SleepFM (Thapa et al. 2026) demonstrates cross-modal decoupling during sleep predicts disease onset across 130+ conditions, supporting the dual-route coupling disruption mechanism at the general-principle level; no specific ME/CFS mechanism validated) Orexin neurons project to sympathetic preganglionic neurons and regulate cardiovascular sympathetic tone (Sakurai et al. 1998) Ruhrländer et al. (2025). Reduced orexin-A thus impairs glymphatic clearance via two independent routes (orexin receptor subtype projections — OX1R to LC, OX2R to tuberomammillary nucleus — are well-established pharmacology): (1) the LC NE oscillation route described in The Orexin–Vasomotion–Glymphatic Triad: A Unified Sleep Failure Model, and (2) peripheral vascular sympathetic regulation of arterial vasomotion. Perivascular sympathetic nerves controlling cerebral arterial vasomotion receive descending brainstem control that is modulated by orexinergic input. Orexin deficiency therefore reduces both the central (LC-mediated NE oscillation) and peripheral (sympathetic vasomotor) components of the glymphatic pump simultaneously.
Testable predictions:
- ME/CFS patients with documented OI (POTS/NMH) should show worse DTI-ALPS glymphatic index than ME/CFS patients without OI, independent of sleep quality measures
- The subgroup with both low CSF orexin-A (less than 200 pg/mL) and severe autonomic dysfunction (COMPASS-31 greater than 40) should show DTI-ALPS index less than 0.25 — lower than any subgroup with only one of these features (DTI-ALPS 0.25–0.35) and lower than ME/CFS patients with neither feature (DTI-ALPS greater than 0.30). A null result — no significant DTI-ALPS difference between the orexin-low+OI subgroup and the orexin-normal+no-OI subgroup — refutes the combined predictor claim.
Certainty: 0.20. ME/CFS patients show flattened cortisol diurnal rhythm (Chapter Endocrine and Metabolic Dysfunction). Orexin neurons receive input from the suprachiasmatic nucleus (SCN) and normally show circadian variation (Sakurai et al. 1998). If neuroinflammatory suppression of orexin (via PGE2/EP3 or neurotensin interneuron pathways — see PGE2–EP3 Self-Sustaining Feedback Loop: Orexin Suppression as a CNS Disease Maintenance Mechanism) is tonic (constant) rather than phasic, the orexin rhythm may flatten or shift independently of the cortisol rhythm.
When two master circadian outputs (HPA axis cortisol, hypothalamic orexin) desynchronise relative to each other, the result is internal desynchrony — distinct from simple circadian delay. The patient is not merely shifted but incoherent: alternating periods of paradoxical alertness (one system promoting wakefulness while the other promotes sleep) and sudden crashes (both systems simultaneously at nadir). This may explain the characteristic ME/CFS complaint of unpredictable energy fluctuations that do not follow a recognisable daily pattern.
Testable predictions:
- 24-hour orexin-A profiling (serial CSF sampling; no validated saliva assay currently exists) in ME/CFS should show reduced amplitude and/or phase shift relative to cortisol rhythm
- The degree of orexin–cortisol phase separation should correlate with subjective symptom variability (assessed by hourly symptom logging)