Central Sensitization and Pain Amplification
Chronic widespread pain affects the majority of ME/CFS patients and does not fully correlate with peripheral tissue damage. Central sensitization—amplification of pain signals within the spinal cord and brain—provides a mechanism for pain that persists independent of peripheral nociceptive input. This section develops a model of central sensitization that treats pain as an active feedback process rather than a passive readout.
1 Spinal Cord Sensitization Dynamics
Central sensitization is mediated by activity-dependent changes in dorsal horn neuron excitability, driven by NMDA receptor (NMDAR) activation, glial cell activation, and neuropeptide release. The model tracks the sensitization state \(\mathcal{S}\) of dorsal horn circuits:
\[ \frac{d \mathcal{S}}{d t} = k_{\text{wind}} \cdot P_{\text{noci}}(t) \cdot \frac{\mathcal{S}}{K_{\text{NMDA}} + \mathcal{S}} + k_{\text{glia}} \cdot \mu_{1,\text{spinal}} - k_{\text{resolve}} \cdot \mathcal{S} \cdot \frac{[\text{endorphin}]}{K_{\text{endorphin}} + [\text{endorphin}]} \tag{1}\]
where \(P_{\text{noci}}(t)\) is the peripheral nociceptive input (from tissue inflammation, small fiber neuropathy, or mechanical stimulation), \(k_{\text{wind}}\) is the wind-up rate (NMDAR-dependent potentiation), \(\mu_{1,\text{spinal}}\) is spinal glial activation (analogous to the CNS microglial model, Equation microglia), \(k_{\text{resolve}}\) is the resolution rate (modulated by endogenous opioids), and the \(\mathcal{S}\)-dependent term in the wind-up ensures positive feedback: existing sensitization amplifies further nociceptive input.
2 Pain as a Feedback Process
The sensitization state modifies the symptom generation function for pain (Equation symptom mapping):
\[ \text{Pain}(t) = (1 + \alpha_{\text{sens}} \cdot \mathcal{S}) \cdot [w_4 \cdot \mathbf{C}_{\text{pro}}(t) + w_5 \cdot \mu_1(t) + w_6 \cdot [\text{ROS}](t)] \tag{2}\]
where \(\alpha_{\text{sens}}\) is the sensitization gain. Critically, pain itself feeds back into the system: pain increases sympathetic tone (\(S\) in the ANS model, Equation ans balance), disrupts sleep (elevating \(\theta_{\text{on}}\) in the sleep model), and consumes attentional resources (increasing effective cognitive energy demand). This bidirectional coupling—invisible without the mathematical model—predicts that pain reduction interventions should improve not only pain but also autonomic function, sleep quality, and cognitive capacity, because they break a feedback amplification loop. The model quantifies this: for typical ME/CFS parameters, a 50% reduction in \(\mathcal{S}\) is predicted to improve sleep efficiency by 10–15% and reduce resting sympathetic tone by 8–12%, mediated entirely through the pain \(->\) ANS \(->\) sleep pathway rather than direct effects on sleep or autonomic circuits.
Low-dose naltrexone operates in this model through two mechanisms: immune modulation (Response to Immune Interventions) and upregulation of endogenous endorphin production (increasing \([\text{endorphin}]\) in Equation central sensitization), accelerating sensitization resolution. The model predicts that LDN’s analgesic effect should precede its anti-inflammatory effect by the difference in timescales: endorphin upregulation (days) versus cytokine network remodeling (weeks).
3 Small Fiber Neuropathy Coupling
Small fiber neuropathy (SFN)—documented in a subset of ME/CFS patients—reduces peripheral nerve fiber density, paradoxically both reducing normal sensory input and increasing aberrant nociceptive signaling. The model represents SFN as a slow variable modifying the nociceptive input function:
\[ \frac{d \mathcal{F}}{d t} = -k_{\text{degen}} \cdot ([\text{ROS}] + \alpha_{\text{auto}} \cdot [\text{Ab}]) + k_{\text{regen}} \cdot \frac{[\text{NGF}] \cdot [\text{ATP}]}{(K_{\text{NGF}} + [\text{NGF}])(K_{\text{ATP,regen}} + [\text{ATP}])} \tag{3}\]
where \(\mathcal{F}\) is small fiber density (normalizing between 0 and 1), \(k_{\text{degen}}\) is the degeneration rate (driven by oxidative stress and autoantibodies), \(k_{\text{regen}}\) is the regeneration rate (dependent on nerve growth factor and ATP availability), and the nociceptive input increases as fiber density decreases below a threshold: \(P_{\text{noci}} = P_0 \cdot (1 + \beta_{\text{SFN}} \cdot (1 - \mathcal{F}/\mathcal{F}_0)^+)\), where \(( \cdot )^+\) denotes the positive part. The model predicts that SFN is both a consequence and an amplifier of ME/CFS: ROS and autoantibodies (from the disease) cause fiber loss, and fiber loss increases pain signaling, which increases sympathetic tone and energy demand, worsening the underlying disease. This feedback loop explains why SFN prevalence increases with ME/CFS disease duration.
(Origin: brainstorm.) (Certainty: 0.35.) The state variable \(\mathcal{F}\) in Equation sfn dynamics is abstract — it has never been anchored to a repeatable measurement in ME/CFS, forcing the model to rely on infrequent skin-punch biopsy for intraepidermal nerve fiber density. Corneal confocal microscopy provides a non-invasive, radiation-free, repeatable measurement of corneal nerve fiber density (CNFD), already shown to be reduced (and nerve tortuosity increased) in ME/CFS (Azcue et al. 2025) and in post-COVID cohorts with ocular symptoms (Moustardas et al. 2026) Cañadas et al. (2023). Treating normalised CNFD as a direct observable of \(\mathcal{F}\) makes the SFN ODE empirically identifiable from serial eye scans rather than serial biopsies.
Falsifiable prediction: In a longitudinal ME/CFS cohort, serial CNFD (corneal confocal) will track the model trajectory of \(\mathcal{F}\): declining during symptom worsening (high ROS/autoantibody terms) and recovering during remission (adequate NGF/ATP terms), with CNFD change rate proportional to \(d \mathcal{F} \\/ d t\). A joint fit of the ODE to paired corneal-confocal and skin-biopsy IENFD data will yield consistent \(k_{\text{degen}}\) and \(k_{\text{regen}}\) estimates across both tissues (within a fixed scaling constant). Falsified if corneal and skin fiber-density trajectories diverge systematically, or if CNFD is static despite documented clinical fluctuation in SFN symptoms.
Consequence: If confirmed, an eye scan taken repeatedly at low cost and zero risk could substitute for repeated skin biopsies as the way to track nerve damage and recovery over time — turning an abstract model term into something a clinical trial can actually measure.
Whether corneal nerve tortuosity and fiber density reflect distinct pathological processes or the same process at different stages is unknown — longitudinal CCM data in ME/CFS, which do not exist, would be needed to determine if these variables track independently.
Is central sensitization in ME/CFS reversible if the driving inputs (inflammation, nociception) are removed, or does prolonged sensitization produce structural changes (synaptic remodeling, glial priming) that persist independently? The model parameterizes this as the ratio \(k_{\text{resolve}} / k_{\text{wind}}\): if resolution is fast relative to wind-up, sensitization reverses when inputs diminish; if resolution is slow (due to structural changes), sensitization persists as a “pain memory.” This distinction has direct treatment implications: reversible sensitization responds to anti-inflammatory and analgesic therapy, while structurally maintained sensitization may require neuromodulatory interventions (NMDAR antagonists, transcranial stimulation) targeting the dorsal horn directly.
4 Nerve Sheath Inflammation Dynamics
The preceding SFN model (Equation sfn dynamics) tracks distal small fiber density. A complementary model is needed for pain arising from the nerve sheath itself, where Schwann cell activation, endoneurial hypoxia, perineurial mast cell degranulation, and autoimmune attack converge (see Kynurenine–NMDA Link to Central Sensitization for the biological mechanisms). The model introduces a nerve sheath inflammation state \(\mathcal{N}_s\) representing the aggregate inflammatory burden within the endoneurial and perineurial compartments:
\[ \frac{d \mathcal{N}_s}{d t} = underbrace(k_{\text{schw}} \cdot [\text{ROS}] \cdot \frac{[\text{CGRP}]}{K_{\text{CGRP}} + [\text{CGRP}]}, \text{Schwann cell TRPA1 activation}) + underbrace(k_{\text{peri}} \cdot H_{\text{mc}}, \text{perineurial mast cell input}) + underbrace(k_{\text{isch}} \cdot (1 - \frac{O_{2,\text{endo}}}{O_{2,\text{endo}}^0})^+, \text{endoneurial ischemia}) - k_{\text{res},N} \cdot \mathcal{N}_s \tag{4}\]
where \(k_{\text{schw}}\) is the rate of Schwann cell-mediated neurogenic inflammation (driven by the combination of ROS and CGRP, reflecting the TRPA1 pathway), \(H_{\text{mc}}\) is the mast cell histamine release rate from the mast cell model (Mast Cell Mediators and Histaminergic Symptom Generation), \(O_{2,\text{endo}}/O_{2,\text{endo}}^0\) is the fractional endoneurial oxygen delivery (reduced by endothelial dysfunction), and \(k_{\text{res},N}\) is the spontaneous resolution rate.
The endoneurial oxygen term couples to the endothelial dysfunction model. In ME/CFS, vasa nervorum share the microvascular pathology of the systemic endothelium:
\[ O_{2,\text{endo}} = O_{2,\text{endo}}^0 \cdot \frac{[\text{NO}]}{K_{\text{NO,endo}} + [\text{NO}]} \cdot \frac{\text{MAP}}{K_{\text{MAP,endo}} + \text{MAP}} \tag{5}\]
where NO availability (from the BH₄ model, Tetrahydrobiopterin Competition Model) and mean arterial pressure (from the ANS model, Autonomic Nervous System Models) jointly determine endoneurial perfusion. BH₄ depletion reduces NO, constricting vasa nervorum; orthostatic hypotension further compromises nerve perfusion during upright posture.
The nerve sheath inflammation state modifies nociceptive input to the central sensitization model:
\[ P_{\text{noci,total}} = P_0 \cdot (1 + \beta_{\text{SFN}} \cdot (1 - \frac{\mathcal{F}}{\mathcal{F}_0})^+) + \gamma_{\text{sheath}} \cdot \mathcal{N}_s \tag{6}\]
where \(\gamma_{\text{sheath}}\) converts nerve sheath inflammation into nociceptive input units. The total nociceptive input \(P_{\text{noci,total}}\) replaces \(P_{\text{noci}}\) in Equation central sensitization, feeding into the central sensitization wind-up term. The model predicts that nerve sheath pain and SFN pain have different temporal profiles: SFN pain worsens slowly over months to years (governed by slow fiber degeneration \(\mathcal{F}\)), while nerve sheath pain fluctuates on the timescale of hours to days (governed by the faster dynamics of mast cell activation, posture-dependent ischemia, and ROS levels). This prediction is clinically testable through pain diary analysis correlated with posture, activity, and mast cell markers.
5 Periarticular and Muscular Pain Model
Joint and muscle pain in ME/CFS arise from tissue-specific nociceptive generators that are then amplified by the central sensitization model (see Convergent Nerve Sheath Vulnerability in ME/CFS for biological mechanisms). The model introduces two tissue-compartment nociceptive sources that feed into the total nociceptive input.
5.1 Periarticular Nociceptive Input
Periarticular pain is driven primarily by local mast cell degranulation in the joint capsule and synovium, where mast cell density is 10–50\(\\times\) higher than in subcutaneous tissue:
\[ P_{\text{peri}}(t) = \rho_{\text{mc,joint}} \cdot H_{\text{mc}}(t) \cdot (1 + \alpha_{\text{NGF}} \cdot [\text{NGF}]) \tag{7}\]
where \(\rho_{\text{mc,joint}}\) is the joint-to-systemic mast cell density ratio (reflecting the anatomical concentration), \(H_{\text{mc}}(t)\) is the systemic mast cell activation level, and the NGF term represents nerve growth factor-mediated upregulation of TRPV1 on periarticular nociceptors. In hEDS patients, an additional mechanical term applies:
\[ P_{\text{peri,hEDS}} = P_{\text{peri}} + k_{\text{lax}} \cdot (1 - \text{stability} / \text{stability}_0) \cdot dot(q) \tag{8}\]
where \(dot(q)\) represents joint movement rate and the stability ratio captures hypermobility-related microtrauma at fascial attachment sites (coupling to the hEDS model in Chapter Integrated Multi-System Models, Extended Subsystem Couplings).
5.2 Muscular Metabolic Nociception
Deep muscle pain arises from metabolic nociceptor activation driven by the interaction of low pH, elevated lactate, and extracellular ATP—the “metabolic danger triad” sensed by ASIC3 channels:
\[ P_{\text{musc}}(t) = k_{\text{ASIC}} \cdot underbrace([\text{H}^+]_m \cdot [\text{Lac}]_m \cdot [\text{ATP}]_{\text{ext,} m}, \text{ASIC3 combinatorial activation}) \cdot (1 + \alpha_{\text{isch}} \cdot (1 - \frac{dot(V) O_{2,\text{musc}}}{dot(V) O_{2,\text{musc}}^0})^+) \tag{9}\]
where \([\text{H}^+]_m\), \([\text{Lac}]_m\), and \([\text{ATP}]_{\text{ext,} m}\) are intramuscular proton, lactate, and extracellular ATP concentrations (from the energy metabolism model, Chapter Energy Metabolism Models), \(k_{\text{ASIC}}\) is the ASIC3 activation coefficient, and the ischemia term represents microvascular dysfunction reducing muscle oxygen delivery \(dot(V) O_{2,\text{musc}}\) below baseline \(dot(V) O_{2,\text{musc}}^0\). The multiplicative ASIC3 activation term captures the experimentally established combinatorial gating: ASIC3 responds to the simultaneous presence of all three signals, not their individual levels.
Post-exertional upregulation of ASIC3 gene expression (persisting 48 h) is modeled as a slow modulation:
\[ \frac{d k_{\text{ASIC}}}{d t} = k_{\text{up}} \cdot E(t) - k_{\text{down}} \cdot (k_{\text{ASIC}} - k_{\text{ASIC}}^0) \tag{10}\]
where \(E(t)\) is exertion intensity, \(k_{\text{up}}\) is the upregulation rate, and \(k_{\text{down}}\) governs return to baseline \(k_{\text{ASIC}}^0\). In ME/CFS, \(k_{\text{down}}\) is reduced (slower return to baseline), consistent with the 48 h persistence observed in gene expression studies.
5.3 Total Nociceptive Input with Tissue Compartments
The complete nociceptive input to the central sensitization model becomes:
\[ P_{\text{noci,total}}(t) = underbrace(P_0 \cdot (1 + \beta_{\text{SFN}} \cdot (1 - \frac{\mathcal{F}}{\mathcal{F}_0})^+), \text{SFN (distal)}) + underbrace(\gamma_{\text{sheath}} \cdot \mathcal{N}_s, \text{nerve sheath}) + underbrace(P_{\text{peri}}(t), \text{periarticular}) + underbrace(P_{\text{musc}}(t), \text{muscular}) \tag{11}\]
This four-compartment nociceptive model replaces the simpler \(P_{\text{noci}}\) in Equation central sensitization. Each compartment has distinct temporal dynamics and therapeutic targets:
- SFN compartment: Slow (months–years); modifiable by neuroprotection (alpha-lipoic acid, IVIG for autoimmune SFN).
- Nerve sheath compartment: Intermediate (hours–days); modifiable by mast cell stabilization, antioxidants, and postural management.
- Periarticular compartment: Fast–intermediate (minutes–hours); modifiable by mast cell stabilizers and anti-NGF approaches. In hEDS: joint stabilization reduces mechanical input.
- Muscular compartment: Fast during exertion, slow post-exertional decay (48 h); modifiable primarily by pacing (reducing \(E(t)\)) and metabolic support (CoQ10, NADH to improve \(dot(V) O_{2,\text{musc}}\)).
The multi-compartment model generates a clinically actionable prediction: patient-specific pain phenotyping (by temporal dynamics, anatomical distribution, and biomarker correlation) should identify which compartments dominate, enabling targeted treatment. A patient whose pain fluctuates with posture and mast cell episodes (nerve sheath + periarticular dominant) requires a different strategy than a patient whose pain tracks exertion with 48 h decay (muscular dominant) or a patient with slowly progressive burning pain (SFN dominant). The central sensitization state \(\mathcal{S}\) acts as a gain multiplier on all compartments—reducing \(\mathcal{S}\) (via LDN, NMDA antagonists) benefits all pain phenotypes, while compartment-specific interventions provide targeted improvement.
Does the dominant pain compartment shift predictably as ME/CFS progresses? Early-stage patients may be muscular-dominant (metabolic stress without yet established SFN or sensitization), while late-stage patients may be SFN-dominant (accumulated fiber loss) with high central sensitization gain. If so, pain treatment strategies should evolve with disease stage—a prediction testable through longitudinal QST, skin biopsy, and pain phenotyping in cohort studies.