Intracranial Neuromodulation for Pain

Evidence status. No intracranial target is FDA-approved for chronic pain in the United States. Epidural motor cortex stimulation carries a weak European guideline recommendation on very low quality evidence and is performed off-label with hardware labeled for other indications; deep brain stimulation for pain failed its two prospective multicenter trials, was never approved, and remains off-label and investigational. Most evidence is observational, heterogeneous, and vulnerable to selection and reporting bias.

1.Pain has lateral and medial systems

The lateral system links spinothalamic input with VPL/VPM, S1, and S2 and carries location, intensity, and sensory quality. The medial system recruits intralaminar and medial thalamus, insula, anterior cingulate cortex (ACC), prefrontal cortex, amygdala, and striatum. It shapes salience, unpleasantness, attention, action, and learning. Neither is an isolated channel; both are embedded in descending control and state-dependent networks.

2.Why the relay model failed

Classical DBS paired sensory thalamus for neuropathic or deafferentation pain with PAG/PVG for nociceptive pain. Early series often required a temporary trial and accepted permanent implantation when stimulation produced substantial pain reduction. Some patients achieved durable benefit, but the two prospective multicenter trials that could have settled the question did not. Both were industry-sponsored: the first enrolled 196 patients with the Model 3380 lead, the second 50 patients with the Model 3387, and success was prespecified as at least half of patients reporting at least 50% pain relief at one year. Neither trial met that threshold on protocol analysis, and the second closed early for slow enrollment, high attrition, and low efficacy. That result, reported by Coffey in 2001, is why deep brain stimulation has never been approved for pain by the FDA. Heterogeneous diagnoses, tolerance, loss of benefit after internalization, unblinding by paresthesia, variable targets, and outcomes centered on pain intensity all contributed.

The lesson is not that relay targets never work. It is that a relay-based biomarker (a warm or tingling field covering the painful body region) is an incomplete surrogate for durable, meaningful analgesia.

3.Motor cortex stimulation

Epidural motor cortex stimulation (MCS) places a paddle or strip over the precentral gyrus contralateral to pain. Benefit is thought to emerge through corticothalamic, corticocortical, and descending modulatory networks rather than direct stimulation of the painful representation. The best-known phenotypes are central post-stroke pain and trigeminal neuropathic pain; evidence for spinal cord injury pain, plexus injury, phantom pain, and CRPS is less consistent.

Preoperative motor mapping, neuronavigation, and sometimes a noninvasive rTMS response help frame the hypothesis, but none is a validated stand-alone predictor. A temporary externalized trial is used by some programs. The European Academy of Neurology guideline offers a weak recommendation for MCS in chronic refractory neuropathic pain, but on very low quality of evidence, the lowest grade in the scheme and below the mostly low-quality evidence behind its weak recommendations for spinal cord stimulation. A weak recommendation on very low evidence is permission to proceed in a selected patient at an experienced center, not an indication. That same guideline grades DBS for neuropathic pain inconclusive; the two intracranial modalities are not in the same evidentiary position.

4.Sensory thalamus: VPL and VPM

The sensory thalamic target is selected somatotopically: VPM for face and VPL for body, with the painful territory mapped by recording and stimulation when an awake workflow is used. Paresthesia that covers the painful area can support localization. Know the neighbors, because each one gives a different unwanted effect. Laterally the internal capsule gives motor contraction and is the hard stop. Medial to VPL lies VPM, and medial to VPM the centromedian and parafascicular nuclei, so medial spread from a VPM target does not simply fail: it changes the experiment, recruiting the medial system instead of the lateral one. Posteriorly lies pulvinar, anteriorly the motor thalamus (Vim), and inferiorly the thalamic fasciculus and zona incerta, with the medial lemniscus entering from below.

Peripheral neuropathic pain has generally performed better than central pain in pooled DBS literature. Long-term attrition remains the defining caution. A target that explains the body map still may not explain the disease network.

One programming caution matters when adapting movement-disorder experience. Historical sensory-thalamic pain protocols often use lower frequencies than movement-disorder DBS, and some patients experience worsening pain at higher settings; parameter-response relationships are not universal. A trainee arriving from tremor programming who opens at habitual Vim settings can make the patient worse and then conclude the target failed. Reassess stimulation, hardware, and the pain diagnosis when symptoms worsen, and take the specific frequency and pulse-width windows from a current primary source rather than from memory.

5.PAG/PVG and descending analgesia

Periaqueductal and periventricular gray stimulation was developed for nociceptive pain and engages endogenous descending analgesic systems. Warmth or analgesia can occur; ocular phenomena, vertigo, nausea, autonomic effects, and mood change can signal spread. The phenotype-target rule is historical rather than absolute, and modern practice sometimes combines PAG/PVG and thalamic leads during a trial.

The mechanism carries its own failure mode. Stimulation-produced analgesia from this region is at least partly opioidergic (naloxone reversal was reported in humans in the 1970s, although later studies did not consistently reproduce it), and tolerance is a recognized late cause of lost benefit here. Explain that early benefit may fade; comparative prediction of durability from an opioidergic versus non-opioidergic target has not been established.

This remains an expert-center, off-label strategy. It should not displace treatment of a reversible nociceptive generator, cancer-pain procedures matched to prognosis, or established spinal/peripheral neuromodulation.

6.ACC and ventral striatum: changing suffering

ACC DBS developed from the observation that cingulotomy can reduce the affective burden of pain. Patients may describe that pain remains present but is less intrusive or distressing. That makes quality of life, activity, mood, and pain interference essential outcomes. A small randomized pilot combining bilateral ACC with sensory thalamic leads reported improved EQ-5D utility during the ACC-on period compared with baseline, without a significant change in pain intensity. That pattern is the dissociation the affective hypothesis predicts, and it is why intensity alone cannot adjudicate this target, but the study did not show superiority over ACC-off stimulation. It also reported two epileptic seizures among eight patients, one of them intraoperative and severe enough to postpone the operation, alongside transient motor and attentional disturbance and one persistent gait disturbance. Longer-term series support feasibility, but efficacy is not established, and the seizure liability of stimulation in cingulate territory belongs in the consent conversation rather than in a footnote.

Ventral capsule/ventral striatum and related reward-motivation targets extend the same network logic. They are best understood as research targets for cognitive-affective dimensions, not as proven replacements for thalamus, MCS, SCS, or treatment of the pain generator.

7.Biomarker-guided and closed-loop stimulation

The critique in Part I, that a paresthesia field is an incomplete surrogate, invites an obvious reply: find a physiological readout of the pain state itself and stimulate against it. Intracranial recording from orbitofrontal and anterior cingulate cortex in a first-in-human series of four patients showed that sustained spectral power changes track self-reported chronic pain and are dissociable from the transient signatures of acute experimental pain, providing an early, small-sample biomarker result rather than a validated clinical pain meter. A follow-on trial selected personalized targets within cortico-striatal-thalamocortical circuits during a staged inpatient trial, implanted five responders with sensing-capable devices, and ran a double-blind sham-controlled crossover, reporting that personalized closed-loop stimulation was superior to sham, with benefit lasting up to 3.5 years in some participants. That second result is available only as a preprint and should be read as such until a peer-reviewed version appears.

Read it as five patients under an investigational device exemption, not as a therapy. The sample is tiny, the targets differ between subjects, and personalization makes conventional multicenter replication hard to design. The result supports further controlled study, not routine adoption. Other small sham-controlled intracranial studies, including a 2025 posterior-superior insula trial in ten highly selected participants, also report preliminary signals; no target has established broadly generalizable efficacy.

8.The workup is a mechanism conference

Confirm a stable diagnosis and exhaust disease-specific, pharmacologic, psychological, rehabilitation, interventional, and less invasive neuromodulation options. Map pain distribution and evoked phenomena; separate continuous burning pain from paroxysms; document sensory loss, allodynia, motor deficit, autonomic change, sleep, mood, substance use, and functional goals. Review litigation or compensation issues without treating them as a moral judgment.

Intracranial therapy is inappropriate when expectations require cure, the phenotype remains unstable, untreated psychiatric disease prevents safe participation, substance use is uncontrolled, or longitudinal programming and measurement are not feasible. Depression and anxiety are common comorbidities and should be treated, not used as automatic exclusions.

9.Define success before the incision

A reasonable protocol records pain intensity and interference, Brief Pain Inventory or McGill dimensions, EQ-5D or SF-36, analgesic dose, sleep, activity, and one or two concrete patient goals. Prespecify the trial duration, blinded or sham epochs when feasible, rescue rules, and the threshold for internalization. A 50% intensity reduction is often used, but a smaller change with a large functional gain may be more meaningful, and the reverse may be true.