Introduction

Neuropathic pain is defined as pain caused by a lesion or disease of the somatosensory nervous system.1,2 Peripheral neuropathic pain (PNP) includes postherpetic neuralgia (PHN), painful polyneuropathy, peripheral nerve injury pain, trigeminal neuralgia and painful radiculopathy in the International Classification of Diseases, 11th Revision (ICD-11) framework.3 Population estimates vary because studies use different ascertainment tools, but pain with neuropathic characteristics is commonly reported in approximately 7–10% of the general population.2,4 This epidemiological burden is clinically important because the diagnostic label does not identify a single mechanism or a single treatment-responsive subgroup.

Current international guidance combines structured diagnosis with staged treatment. The 2023 joint guideline from the European Academy of Neurology, European Pain Federation and Neuropathic Pain Special Interest Group of the International Association for the Study of Pain (NeuPSIG) recommends validated screening tools, neuroanatomically plausible symptom mapping and targeted confirmatory testing to establish neuropathic pain.5 The 2025 NeuPSIG treatment update recommends tricyclic antidepressants, serotonin-noradrenaline reuptake inhibitors and alpha-2-delta ligands as first-line therapies; topical agents as second-line options; and botulinum toxin type A, repetitive transcranial magnetic stimulation (rTMS) and opioids as weak third-line options.6 Earlier NeuPSIG syntheses and contemporary evidence reviews likewise show modest average pharmacological benefits and persistent unmet need.7,8 Although pooled estimates suggested that approximately one additional patient would benefit for every four patients treated and that treatment-related adverse events were uncommon, the 2025 update gave rTMS only a weak third-line recommendation because the evidence was of low certainty.6 These broad algorithms provide a rational baseline but do not identify which PNP phenotypes are most likely to benefit, supporting phenotype-stratified selection rather than uniform escalation.

The clinical relevance of this framing is practical. A patient with PHN and allodynia, a patient with painful diabetic polyneuropathy and sensory loss, and a patient with radiculopathy after nerve-root compression may all be described as having PNP. Yet their peripheral generators, psychological burden, treatment constraints and likely stimulation response may differ. Reviews of chronic pain epidemiology and associated factors show that chronic pain rarely travels alone; it commonly interacts with sleep, mood, disability and social participation.9 Neuropathic pain reviews therefore increasingly treat mechanism, phenotype and function as inseparable parts of clinical interpretation.10

The review also distinguishes rTMS from the broader field of brain stimulation. Non-invasive stimulation includes transcranial magnetic stimulation (TMS), rTMS, theta-burst stimulation and transcranial direct-current stimulation (tDCS), each with different physical principles and biological assumptions.11 Reviews of non-invasive techniques also place TMS beside vagus-nerve and direct-current approaches, which helps prevent overgeneralization across modalities.12 Therapeutic TMS reviews further emphasize the need to separate experimental physiology from clinical prescription.13 Annual-review and clinical-neuroscience articles describe how stimulation waveform, coil geometry, cortical state and network context influence the induced electric field and downstream response.14 These principles help explain why the same nominal frequency can produce different effects across patients and studies.

This revised narrative review therefore makes a deliberately narrower and more original claim: rTMS should be evaluated as a phenotype-stratified cortical analgesia strategy, not as a uniform intervention for all PNP. The organizing framework links peripheral aetiology, central amplification, cortical target selection and multidomain response. This structure preserves the positive signal for primary motor cortex (M1) stimulation while explaining why evidence remains weak for several common peripheral neuropathic phenotypes.

The central challenge for a narrative review is therefore not to repeat that rTMS is promising. That claim is already familiar. The more useful contribution is to explain why the same stimulation protocol may appear persuasive in one peripheral neuropathic phenotype and inconclusive in another. We therefore organize the evidence around a translational chain: peripheral lesion phenotype, central pain-modulation network, stimulation target, clinical response and maintenance strategy. This structure allows the review to reach a distinct conclusion while remaining conservative about the evidence.

Literature Search Approach

To support this narrative synthesis, we searched PubMed/MEDLINE, Embase, Web of Science Core Collection and the Cochrane Library from database inception through 30 June 2026. Search concepts combined terms for PNP and its major aetiologies (including PHN, diabetic neuropathy (DNP), radiculopathy and peripheral nerve injury) with rTMS, motor cortex stimulation, dorsolateral prefrontal cortex (DLPFC), neuromodulation, phenotype, protocol, maintenance and clinical outcomes. We prioritized randomized and sham-controlled trials, systematic reviews and meta-analyses, consensus or guideline documents, and clinically relevant mechanistic studies. Because the purpose was a narrative synthesis rather than a systematic review, study selection was interpretive; no Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-based screening, de novo meta-analysis or formal study-level risk-of-bias assessment was performed.

Conceptual Framework and Evidence Boundaries

This narrative review is organized around a clinical and mechanistic problem rather than a formal systematic-review methodology. PNP is not a single treatment entity. It is a family of disorders in which peripheral nerve injury, sensory loss, ectopic activity and inflammatory signaling interact with spinal and supraspinal gain control.15 Broad clinical reviews emphasize the same point: pain distribution, aetiology, sensory signs and comorbidity should shape treatment interpretation.16 For this reason, the evidence below is read through phenotype, target and response domain rather than through a generic rTMS-versus-sham question.

The first boundary is diagnostic. The review uses PNP to mean pain linked to a lesion or disease of the peripheral somatosensory system, including PHN, DNP, radiculopathy and traumatic or postsurgical nerve injury. The diagnostic logic follows modern grading principles and current international assessment guidance, which distinguish symptoms alone from neuroanatomically plausible pain with sensory signs and confirmatory tests.17 European guidance on neuropathic pain assessment also supports using clinical examination and disease context before assigning a neuropathic label.5 This distinction matters because non-neuropathic pain can improve after procedures through expectation or nonspecific care.

The second boundary is therapeutic. rTMS is interpreted as a neuromodulatory adjunct, not as a replacement for established pharmacological and rehabilitation care. Neuropathic pain pharmacotherapy has a long evidence base, including older evidence-based recommendations and later comparative reviews.18 Gabapentin evidence illustrates both the importance and limits of conventional drug treatment in chronic neuropathic pain.19 Contemporary recommendations still support gabapentinoids, antidepressants and topical options for selected patients, while acknowledging limited average effect sizes and tolerability problems.7 French multidisciplinary recommendations similarly frame non-pharmacological approaches as part of a staged care model rather than as stand-alone cures.20

The third boundary is mechanistic. Peripheral input is necessary for many syndromes, but persistent neuropathic pain is maintained by central amplification and altered inhibitory control. Baron et al emphasized how peripheral drive sustains central sensitization.21 Human nerve-injury studies show that biological and behavioural markers can track pain after peripheral injury.22 Microglial adenosine triphosphate (ATP) signaling, transient receptor potential (TRP) channels and endocannabinoid pathways add molecular context, although most of these targets are not directly tested by clinical rTMS trials.23 TRP-channel reviews are particularly useful for separating peripheral nociceptor sensitization from cortical network hypotheses.24

The fourth boundary is evidence strength. Some references in this review support direct clinical efficacy, such as randomized trials and meta-analyses of rTMS. Other references support background mechanisms, sham design, stimulation physics, patient selection or comparison with invasive neuromodulation. This separation prevents a common error in narrative reviews: using a mechanism paper as if it proved clinical efficacy. It also allows the field to retain useful mechanistic breadth without overstating the readiness of rTMS for routine care. Accordingly, randomized trials and meta-analyses are used to judge clinical efficacy, whereas mechanistic studies are used only to assess biological plausibility and generate testable hypotheses.

Mechanistic Rationale: Treating a Peripheral Pain Disorder Through Cortical Networks

The rationale for rTMS in PNP rests on the coupling between peripheral input and central gain. Peripheral lesions can generate ectopic activity, inflammatory signaling and altered afferent traffic, but persistent pain is maintained partly through spinal sensitization, thalamo-cortical reweighting and affective salience networks.2,3,17 A cortical intervention is therefore plausible when the therapeutic goal is not to repair the peripheral nerve directly, but to reduce amplification and strengthen descending inhibitory control.

TMS was introduced as a non-invasive method for activating human motor cortex,25 and modern neurophysiological guidance describes how pulse intensity, coil geometry and stimulation frequency shape cortical output.26 Therapeutic rTMS for neuropathic pain most often targets M1 contralateral to the painful region. The updated European recommendations rated high-frequency M1 rTMS for neuropathic pain as having definite efficacy, although the recommendation aggregates heterogeneous neuropathic conditions.27

Mechanistic work suggests that M1 stimulation does not act only at the cortical point under the coil. Motor cortex stimulation can engage thalamic, anterior cingulate, orbitofrontal and periaqueductal grey (PAG) circuits that influence both sensory-discriminative and affective dimensions of pain.28 Human pharmacological experiments further support an opioidergic component: naloxone reduced M1 rTMS-induced analgesia in healthy volunteers, DLPFC rTMS analgesia also showed opioid sensitivity, and later work showed region- and dose-dependent involvement of endogenous opioid peptides during M1 and DLPFC stimulation.29–31 These data support a network model of analgesia, while also warning that mechanistic evidence from healthy-volunteer pain models cannot be directly equated with durable benefit in chronic peripheral neuropathy.

Several mechanistic literatures support, but do not prove, the use of cortical stimulation in PNP. Imaging work in chronic pain has linked persistent symptoms to reorganization of sensory and affective networks.32 A broader theory of chronic pain emphasizes that the brain gradually learns pain-related predictions, salience and avoidance patterns.33 Cognitive and emotional control systems can further shape pain intensity and distress, which is relevant when DLPFC stimulation is considered as an adjunctive target.34 Insular-cortex stimulation studies add another network-level example, because the insula links sensory, salience and autonomic dimensions of pain.35

DLPFC stimulation has direct experimental relevance even though it is not yet a standard peripheral-neuropathic-pain protocol. In a capsaicin pain model, DLPFC rTMS modulated experimentally induced pain, supporting the idea that executive-control nodes can influence pain processing.36 This evidence should be treated as mechanistic support, not as proof that DLPFC stimulation is clinically effective for every peripheral neuropathic phenotype.

Motor cortex stimulation has a distinct history in refractory pain. Early invasive motor cortex stimulation studies showed that central and peripheral deafferentation pain could respond to cortical stimulation in selected patients.37 Later mechanistic reviews proposed that M1 stimulation modulates thalamic, cingulate, insular and brainstem circuits rather than acting only through the hand or face representation under the coil.28 Brain opioid receptor density may also predict the efficacy of motor cortex stimulation, supporting a biologically plausible link between cortical stimulation and endogenous analgesia.38

Physiological studies add another layer. High-frequency rTMS can modify cortical excitability, while theta-burst protocols demonstrate that patterned stimulation can induce durable changes in human motor cortex.39 The direction and magnitude of plasticity depend on baseline excitability, synaptic inhibition and receptor systems. Gamma-aminobutyric acid type B (GABAB)-mediated intracortical inhibition is one candidate mechanism through which cortical state may shape response variability.40 These findings justify measuring neurophysiological state in future pain trials rather than treating all patients as interchangeable stimulation targets.

Targeting accuracy is also part of mechanism. Figure-eight and H-coil designs generate different electric-field distributions.41 Deep TMS methods were developed to reach less superficial networks, but broader fields can complicate mechanistic interpretation.42 Navigated TMS can improve reproducibility by anchoring the coil to individual anatomy rather than to scalp landmarks alone.43Figure 1 integrates this multiscale account. At the peripheral level, voltage-gated sodium channel 1.7 (NaV1.7) and transient receptor potential vanilloid 1 (TRPV1)-mediated ion channel sensitization, along with P2X purinergic receptor 4 (P2X4)-dependent microglia signalling in the dorsal horn, collectively drive spinal sensitization and spinothalamic pain transmission. Meanwhile, rTMS targeting the M1 is hypothesized to engage thalamic, prefrontal, insular regions, as well as the descending pain modulatory pathway formed by the PAG and rostral ventromedial medulla (RVM). The DLPFC, a core node of cognitive and affective pain regulation, is also annotated as a cortical component in the network. These molecular labels are mechanistic anchors, not validated treatment-selection biomarkers.

Figure 1 Multiscale mechanism map for rTMS analgesia in PNP. Peripheral injury is represented from ion-channel sensitization NaV1.7 / TRPV1 to dorsal-horn microglia signalling (P2X4) and spinothalamic transmission. M1 rTMS is positioned within cortical (M1 /DLPFC/ insula) and descending periaqueductal grey-rostral ventromedial medulla (PAG-RVM) pathways. Molecular annotations identify biologically plausible nodes, not validated treatment-selection biomarkers.

Phenotype-Stratified Mechanistic Interpretation

PHN and painful DNP should not be treated as different labels for the same rTMS target. PHN commonly presents with a painful, dermatomally organized region and dynamic allodynia after herpes zoster.44 This phenotype is compatible with continued peripheral drive superimposed on dorsal-horn and supraspinal gain control.21 A reproducible painful territory may make contralateral M1 selection more tractable, which is consistent with the positive high-frequency M1 trials in PHN.45 This is a phenotype-linked explanation, not proof that one central remodeling pattern determines response.

Painful DNP has a different biological starting point. Its length-dependent axonal injury coexists with metabolic, vascular and inflammatory influences.46 Sensory loss, autonomic features and foot-risk factors add clinical heterogeneity that a single cortical hotspot cannot capture.47 Diffuse distal deafferentation may therefore weaken a one-to-one mapping between pain territory and an M1 target. The short 10-Hz M1 trial provides a useful signal.48 However, the pathology-stratified synthesis did not establish a reliable DNP-specific effect.49

The proposed phenotype-stratified framework is hypothesis-generating: it links peripheral lesion pattern, sensory phenotype and comorbidity to central gain and targetability, and then to the probability that a trial-defined rTMS regimen will produce a multidomain response. Figure 1 maps the biological layers of this chain, and Table 1 translates it into phenotype-linked mechanisms, direct clinical evidence and trial-informed options. Neither the figure nor the table should be interpreted as a validated patient-level prediction rule.

Table 1 Phenotype-Stratified Clinical Translation Matrix for rTMS in PNP

Clinical Evidence Across PNP

The clinical evidence has expanded substantially since early single-session studies. A 2022 systematic review of 38 randomized controlled studies across neuropathic pain conditions found that active rTMS reduced pain more than sham stimulation, with an overall effect size of −0.66 (95% CI, −0.87 to −0.46).58 More recent peripheral neuropathy-focused evidence is more informative for this review. A 2026 meta-analysis assessed with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework and restricted to PNP included nine randomized sham-controlled trials with 435 patients and reported improved pain intensity after rTMS (SMD, −1.11; 95% CI, −1.72 to −0.49).55 The same meta-analysis found no statistically significant increase in common adverse events, but highlighted heterogeneity and limited long-term data.

A separate 2026 systematic review stratified painful peripheral neuropathies by pathology and stimulation modality. It included 18 trials, with 13 trials and 569 participants in the quantitative synthesis.49 This analysis is especially important because it separated PHN from DNP, radiculopathy, brachial plexus injury and post-cancer-treatment neuropathy. Motor cortex rTMS produced modest pain reductions in PHN from immediate post-treatment assessment (MD, −1.96; 95% CI, −2.96 to −0.97) to 3-month follow-up (MD, −1.23; 95% CI, −2.34 to −0.14), whereas evidence for other peripheral neuropathies was non-significant and low or very low certainty.49

Across these studies, confidence in clinical efficacy is constrained by small phenotype-specific samples, variable sham credibility and blinding, inconsistent outcome definitions, heterogeneous targets and stimulation schedules, and short or incomplete follow-up. These limitations increase risk of bias and reduce the clinical interpretability of pooled effects, particularly when diverse neuropathic aetiologies are combined. We therefore treat the pooled estimates as evidence of a group-level analgesic signal rather than proof of a universal or validated phenotype-specific protocol.49,55

The largest individual multicentre trial is therefore best read as a target-specific test rather than a final clinical algorithm. In the 2021 Brain trial, 149 treated patients with chronic PNP received active M1, active DLPFC or sham rTMS. The protocol used 10 Hz stimulation, 3000 pulses per session and 15 sessions over 22 weeks. M1 rTMS reduced average pain intensity relative to sham, whereas DLPFC rTMS did not.57 This result supports M1 as the principal analgesic target, while leaving open the possibility that DLPFC stimulation may be more relevant to mood, sleep or pain appraisal in selected contexts.

Older clinical studies remain useful because they show how the field developed before the largest multicentre trial. Lefaucheur et al reported that pain relief from precentral rTMS depended on pain origin and stimulation site.59 Early work also suggested that precentral stimulation could produce transient analgesia in neurogenic pain.60 A later review summarized chronic-pain rTMS studies and highlighted the same problems that persist today: small samples, heterogeneous protocols and variable follow-up.61

Some peripheral or deafferentation syndromes are informative even when they are not the core target of this review. Parietal rTMS has been reported to transiently improve phantom limb pain-like symptoms, suggesting that non-M1 cortical targets can influence body representation and pain perception.62 Motor cortex rTMS has been shown to restore defective intracortical inhibition in chronic neuropathic pain, linking analgesia to measurable cortical physiology.63 The duration of relief after motor cortex rTMS has also been examined in interventional neurophysiology studies, underscoring the need to separate immediate response from sustained treatment benefit.64

Dose accumulation also deserves attention. Daily-session studies reported longer-lasting antalgic effects than single-session designs in selected chronic pain cohorts.65 This observation does not settle the optimal schedule, but it supports the clinical logic of induction plus maintenance rather than one-off stimulation.

Condition-specific evidence should be read with clinical context. PHN has a well-described epidemiology and treatment burden in older adults.44 Herpes zoster reviews similarly emphasize age, immune status and persistent allodynia as important determinants of long-term disability.66 For this phenotype, rTMS trials and meta-analyses provide the clearest direct signal, but the surrounding clinical literature explains why sleep, mood and daily function should be measured alongside pain intensity.

Painful DNP requires a different interpretation. Consensus recommendations describe diagnosis, severity grading and management as part of a broader diabetes-care pathway.67 Updated reviews of DNP biomarkers and therapies show that metabolic, inflammatory, vascular and nerve-regeneration mechanisms can coexist.46 A systematic review of non-invasive neuromodulation for painful DNP found potential benefit but also emphasized limited trial size and protocol heterogeneity.68

The DNP literature also reminds reviewers not to over-interpret short-term pain scores. tDCS trials in diabetic polyneuropathy have reported improvements in quality of life and physical fitness, suggesting that neuromodulation studies may detect functional effects even when analgesic effects are modest.69 Clinical reviews of DNP emphasize sensory loss, autonomic features, ulcer risk and comorbidity, none of which can be captured by a single visual analogue scale.47 This is why future rTMS studies in DNP should include sensory phenotype and foot-risk status.

Other peripheral neuropathic syndromes are less mature as rTMS indications. Malignancy-related neuropathic pain has been tested in a randomized clinical trial, but cancer pain involves disease burden, treatment effects and systemic factors that may not generalize to PHN or DNP.70 Bladder pain syndrome and interstitial cystitis studies suggest that rTMS can be explored in chronic pelvic pain, but the diagnostic overlap with visceral and centralized pain mechanisms is substantial.71 Orofacial neuropathic pain meta-analysis provides another adjacent evidence stream, useful for target and frequency hypotheses but not a substitute for peripheral limb neuropathy trials.72Figure 2 translates the evidence hierarchy into a phenotype-sensitive map. Its forest-style panel reproduces published pooled estimates rather than presenting a new meta-analysis, while the remaining panels isolate the factors that still limit cross-trial comparison and clinical translation.

Figure 2 Evidence map for rTMS in PNP. (a–d) Identify the evidence hierarchy, published pooled estimates, translation barriers and minimum reporting variables. (a) Summarizes the relative evidence signal by phenotype. (b) Displays published pooled estimates cited in the manuscript: overall PNP SMD, −1.11 (95% CI, −1.72 to −0.49),55 PHN immediate MD, −1.96 (95% CI, −2.96 to −0.97), and PHN 3-month MD, −1.23 (95% CI, −2.34 to −0.14).49 (c and d) Identify translation barriers and minimum reporting variables. These cited estimates are not a new pooled analysis. Panel letters are retained to identify the four subfigures.

Condition-Specific Evidence

PHN is the peripheral neuropathic phenotype with the most consistent rTMS signal. The condition is common in older adults after herpes zoster and can remain refractory despite conventional treatment.44 A 2024 meta-analysis focused on PHN concluded that rTMS reduced pain compared with control, but the included evidence was still limited by small trials and protocol differences.50 The primary trials point in the same direction. A 40-patient sham-controlled trial found greater visual analogue scale reduction after 10 sessions of 10 Hz M1 rTMS than after sham stimulation.45 A 60-patient randomized trial comparing 5 Hz, 10 Hz and sham stimulation reported that both active frequencies reduced pain, with 10 Hz producing greater pain relief than 5 Hz over follow-up.73

Newer PHN studies add nuance rather than a simple positive conclusion. In a 2023 sham-controlled trial, M1 stimulation produced stronger and more durable analgesia than DLPFC stimulation, while both targets improved sleep quality.51 A 2026 single-centre trial in patients with PHN and comorbid depression undergoing interventional neuromodulation reported a lower rate of poor prognosis at 3 months with adjunctive 10 Hz rTMS than with sham stimulation (27.4% versus 42.7%; odds ratio, 0.51; 95% CI, 0.27–0.97).74 These findings suggest that PHN may be the best near-term indication for pragmatic rTMS trials, especially when pain, sleep and mood outcomes are measured together.

Painful DNP is clinically important but less secure as an rTMS indication. Diagnostic and management standards emphasize that DNP is heterogeneous, length-dependent and influenced by metabolic control, sensory loss and foot-risk factors.75,76 The rTMS evidence remains small. A 2022 randomized trial enrolled 22 patients with diabetic PNP, of whom 20 completed the study, and reported short-term reductions in numeric pain ratings after five sessions of 10 Hz M1 stimulation.48 However, the 2026 peripheral neuropathy meta-analysis found non-significant evidence for DNP when stratified by condition.49 The defensible conclusion is therefore that DNP is a priority phenotype for trials, not yet a mature routine indication.

Radiculopathy and traumatic or postsurgical nerve injury occupy a middle ground. A two-centre sham-controlled comparative study in lumbosacral radiculopathy randomized 35 patients and found active 10 Hz rTMS superior to tDCS and sham for pain intensity after three daily sessions.52 By contrast, a 2024 randomized sham-controlled parallel trial of navigation-guided rTMS in upper-limb neuropathic pain randomized 30 patients and did not show significant pain relief on the primary outcome, although pain-related disability improved and no serious adverse events were observed.53 An H-coil crossover study in 17 patients reported statistically significant effects on pain intensity and symptoms of anxiety and depression, but the small sample and coil-specific stimulation field make generalization difficult.54

The mixed findings across phenotypes are not failures of the field; they are the signal that PNP should not be treated as one biological entity. Prediction work based on the multicentre trial dataset has begun to test whether clinical variables can identify responders to motor cortex rTMS.56 That direction is likely more useful than simply increasing sample size without stratifying aetiology, sensory phenotype, psychological comorbidity and stimulation exposure.

Safety, Tolerability and Patient Selection

rTMS is non-invasive, but it is not risk-free. The 2021 expert safety guidelines update the 2009 consensus guidance and emphasize protocol limits, screening, staff training, seizure preparedness, device-specific considerations and special contexts such as neuronavigation and combined stimulation. Within established safety parameters, the risk of seizure is considered low, but individual risk assessment remains mandatory, especially in patients with epilepsy risk factors, intracranial metal, implanted devices or medications that may lower seizure threshold.77,78

Patient selection should begin with diagnostic certainty rather than treatment enthusiasm. The grading system for neuropathic pain separates possible, probable and definite neuropathic pain using history, neuroanatomically plausible distribution, sensory signs and confirmatory tests where appropriate.17 For a patient with poorly defined pain, rTMS may add procedural complexity without addressing the mechanism. For a patient with confirmed PNP, persistent disability and limited tolerance of systemic treatments, a time-limited rTMS trial can be justified if outcomes are specified before treatment.

Outcome measurement should include more than pain intensity. Pain diaries and visual analogue or numeric scales are necessary, but patients and reviewers also need information on function, sleep, mood, medication use, global impression of change and adverse events. This is especially important because some trials show improvement in disability or sleep even when the primary pain outcome is neutral.51,53 A responder definition should be set prospectively, using a clinically meaningful pain reduction together with functional or patient-global benefit.

Safety interpretation should be historical as well as contemporary. Early safety recommendations for rTMS defined seizure-risk screening, stimulation limits and staff preparedness.79 Later consensus safety guidance refined these recommendations for modern devices and clinical contexts.77 A systematic review of theta-burst safety is relevant because newer pain protocols may borrow patterned stimulation approaches from other fields.80

The safety discussion also depends on sham design. Sham TMS can generate scalp sensation and sound without delivering the same intracerebral field, but some sham systems still induce measurable electric fields or motor-evoked potentials.81 This matters in pain trials because expectation, discomfort and blinding credibility can influence reported analgesia. Pain-specific TMS guidelines therefore recommend credible sham conditions, predefined outcomes and careful adverse-event reporting.82

Patient selection should also consider conventional treatment history. Reviews of painful neuropathy and guideline overviews describe the common sequence of diagnosis, first-line pharmacotherapy, adverse-effect management and escalation when pain remains disabling.83 European Federation of Neurological Societies (EFNS) pharmacological guidance similarly supports structured medication trials before moving to more specialized interventions.84 Overviews of neuropathic-pain guidelines show that disagreement often reflects different evidence thresholds rather than entirely different clinical principles.85

Protocol and Reporting Implications

The best-supported protocol family for future trials remains high-frequency stimulation of M1 contralateral to the painful region. The largest multicentre trial used repeated 10 Hz sessions across a 22-week treatment pathway.57 Its design is useful because it moved beyond a single exposure and asked whether maintenance treatment could sustain benefit.

PHN trials illustrate why schedule details should be interpreted phenotype by phenotype. One study tested a ten-session 10 Hz M1 course.45 A separate frequency-comparison study evaluated 5 Hz and 10 Hz stimulation and found the higher-frequency schedule more favourable in that context.73

Earlier therapeutic rTMS studies used different maintenance and exposure patterns.86 Additional trials varied pulse number, treatment duration and follow-up windows.87 Radiculopathy evidence came from a shorter comparative course rather than a long maintenance protocol.52 This heterogeneity prevents a single recommended prescription, so protocol tables should be treated as evidence maps rather than clinical guidelines.

Neuronavigation and individualized targeting are attractive because pain somatotopy, cortical reorganization and coil placement may influence response. Pain-specific TMS research guidelines recommend rigorous sham control, credible blinding, standardized outcomes and careful reporting of coil position, motor threshold, pulse dose and concurrent treatments.82 For translational research and clinical-development studies, these same items should be documented as quality-control variables. For research, they are essential to make trials comparable.

Comparisons with other neuromodulation approaches should be conservative. tDCS is simpler and cheaper to deliver, but direct evidence in PNP is inconsistent and, in the radiculopathy trial, rTMS was superior to tDCS.52,88 Spinal cord and peripheral nerve stimulation have established roles in selected chronic pain syndromes, but they involve implantation, device management and procedure-related risks.89 The clinical niche for rTMS is therefore best described as a reversible, non-invasive bridge between pharmacotherapy and invasive neuromodulation, not as a replacement for either.

Protocol reporting should include the stimulation technology itself. Basic International Federation of Clinical Neurophysiology (IFCN) procedures for non-invasive electrical and magnetic stimulation establish terminology for motor threshold, coil orientation and induced responses.26 Earlier IFCN guidance also described foundational procedures for electrical and magnetic stimulation of the brain, spinal cord and roots.90 Formal rTMS therapeutic guidelines extended these principles into clinical indications, safety levels and protocol reporting.91 These details are not cosmetic. Without them, a reader cannot determine whether two studies delivered comparable cortical exposure.

A useful protocol table should therefore separate target localization, dose and schedule. Rapid-rate TMS studies established that frequency and train structure can shape motor-cortex response.92 Determinants of non-invasive stimulation plasticity include baseline cortical state and individual variability.93 Reviews of therapeutic TMS emphasize that these sources of variability are one reason why a positive group effect does not automatically translate into a fixed clinical prescription.94

Comparison with other neuromodulation approaches can help define the clinical niche of rTMS. Neurostimulation consensus recommendations support spinal cord and peripheral nerve stimulation for selected refractory neuropathic pain syndromes.54 EFNS neurostimulation guidance also placed rTMS beside other stimulation approaches for neuropathic pain, while recognizing differences in evidence strength and invasiveness.95 Mechanistic work on spinal cord stimulation addresses which neural elements may be activated and how dorsal-column stimulation may suppress pain.96 Reviews of spinal cord stimulation mechanisms further show why invasive neuromodulation cannot be treated as the same therapeutic category as non-invasive rTMS.97

Peripheral nerve stimulation and deep brain stimulation mark two additional boundaries. Long-term peripheral nerve stimulation outcomes in reflex sympathetic dystrophy show that targeted peripheral neuromodulation can benefit selected patients but requires implantation and follow-up.98 Deep brain stimulation principles illustrate the opposite end of the invasiveness spectrum, with strong mechanistic appeal but substantial surgical complexity.99 Modern reviews of deep brain stimulation for chronic pain support its role only in highly selected refractory cases.100

These comparisons define rTMS as a reversible cortical probe and potential adjunctive therapy. It is more focal and physiologically direct than tDCS, less invasive than spinal cord or peripheral nerve stimulation, and far less invasive than deep brain stimulation. However, that position also creates a burden of proof. Trials must show not only that pain decreases after stimulation, but also that the response is reproducible, durable and clinically worth the repeated treatment visits.

For phenotype-specific clinical translation, protocol details should be described as trial-informed starting points, not as universal prescriptions. In PHN, contralateral M1 stimulation at 10 Hz, 1500 pulses per session for 10 sessions has direct sham-controlled support.45 In painful DNP, 10-Hz M1 stimulation for five sessions has been tested, but this evidence is not sufficient for a routine indication-specific protocol.48 In mixed PNP, the largest multicentre study used 10 Hz, 3000 pulses per session and 15 sessions over 22 weeks.57 These values should not be interpreted as a validated patient-level treatment algorithm.

Maintenance is the least settled part of the pathway. One continuous trial enrolled induction responders into four weekly M1 sessions, but it did not establish a universally effective maintenance schedule.87 Practical continuation decisions should therefore require a predefined multidomain response assessment, including pain, function, sleep, medication burden and adverse events. Baseline phenotype and early change can be used as stratification variables, but clinical-variable response models remain preliminary rather than validated selection tools.56Table 2 specifies the minimum variables that should be recorded to make these decisions transparent and comparable across studies. Figure 3 presents the review’s central proposition as a proposed research and clinical-development framework, not an established clinical pathway; baseline phenotype and symptom burden guide target localization and dose selection, followed by induction and early multidomain monitoring, with iterative feedback used to generate hypotheses for future validation.

Table 2 Minimum Reporting and Clinical-Translation Checklist for rTMS Studies in PNP

Figure 3 Proposed adaptive precision rTMS framework for PNP. This schematic presents a hypothesis-generating research and clinical-development framework rather than an established or validated clinical pathway. Baseline phenotyping includes pain aetiology, pain distribution, sensory phenotype, sleep, mood, function, medication tolerance, and neuropathic pain grading or diagnostic certainty. Target and dose selection are individualized according to cortical target, coil type and laterality, neuronavigation or motor-hotspot localization, stimulation frequency, intensity, pulses per session, and total number of sessions. During the induction phase, high-frequency rTMS is typically delivered over M1 contralateral to the painful body region, with the DLPFC considered as an optional modulatory target. Early multidomain response monitoring evaluates pain intensity, physical function, walking or daily activity, sleep quality, mood, patient global impression, adverse events, and medication use. Based on clinical response and tolerability, patients are stratified as responders, partial responders, or non-responders/intolerant patients, guiding maintenance stimulation, retargeting or dose adjustment, or transition to alternative care. Solid arrows indicate the proposed primary clockwise sequence, whereas dashed arrows indicate feedback loops for adaptive care. (Bold text identifies workflow modules and key category labels; it does not denote evidence strength or a validated recommendation.).

Limitations of the Evidence Base

This narrative, non-systematic review has several limitations that should shape its claims. First, trial samples remain small for condition-specific inference, even though pooled analyses are now larger. Second, stimulation schedules, pulse doses, coil types, sham methods and maintenance protocols differ widely. Third, follow-up is often short, and long-term durability beyond a few months remains uncertain. Fourth, broad neuropathic pain meta-analyses can obscure important differences between PHN, DNP, radiculopathy and traumatic nerve injury.49,55

Because study selection was not conducted as a systematic review, we did not undertake duplicate screening, a de novo meta-analysis or a formal study-level risk-of-bias assessment. The synthesis may therefore be affected by selection bias, and its conclusions should be read as a structured, hypothesis-generating interpretation of the literature rather than a comprehensive estimate of treatment effectiveness. Clinical efficacy judgments were based on randomized trials and quantitative syntheses; mechanistic evidence was used only to support biological plausibility.

Publication bias and expectation effects also matter in procedural pain trials. The 2015 pharmacotherapy review found evidence that publication bias can overstate treatment effects in neuropathic pain trials.7 Although this observation comes from drug studies, it reinforces the need for credible sham stimulation, patient and assessor blinding, registration, predefined responder analyses and reporting of negative outcomes in rTMS research.

One limitation of the broader evidence base is that related chronic-pain studies do not all describe the same biology. Functional imaging meta-analyses show distributed brain responses to pain, but these responses do not specify which patients will respond to M1 rTMS.101 Primary somatosensory cortex reorganization after spinal cord injury demonstrates central plasticity after nervous-system damage, but central injury is not equivalent to peripheral neuropathy.102 These studies are valuable for mechanism, not for direct efficacy claims.

A second limitation is that prediction science remains immature. Conditioned pain modulation has predicted chronic post-operative pain in earlier work, suggesting that endogenous inhibition can identify vulnerability.103 Whether similar measures predict rTMS response in PNP is still uncertain. Biomarker reviews in chronic pain show promise, but the field has not yet converged on a clinically usable rTMS-response signature.32

Future Directions

The next generation of studies should be phenotype-specific. PHN is ready for larger pragmatic trials that test induction plus maintenance schedules and multidomain outcomes. DNP trials should stratify by neuropathy severity, glycaemic context, sensory loss and comorbid depression. Radiculopathy and nerve injury trials should separate compressive, traumatic and postsurgical mechanisms rather than pooling them under a broad peripheral neuropathic label.

Mechanistic substudies should be built into clinical trials. Quantitative sensory testing, conditioned pain modulation, cortical excitability, electroencephalography, functional imaging or biomarker panels may identify patients whose pain is more centrally amplified and therefore more likely to respond to cortical stimulation. The goal is not to make rTMS more technologically elaborate for its own sake, but to prevent repeated underpowered trials in biologically mixed populations.

Implementation studies are also needed. If rTMS is to move from specialized centres into pain rehabilitation pathways, trials must report treatment burden, adherence, cost, access, patient preference and compatibility with medications, psychological care and physical rehabilitation. Without these data, a statistically significant short-term pain reduction will remain difficult to translate into a reimbursable clinical service.

Future studies should connect mechanistic measurement to clinically meaningful endpoints. Allodynia and hyperalgesia have distinct clinical manifestations and mechanisms, and both should be reported rather than hidden under a single pain-intensity score.104 Reviews of treatment advances in neuropathic pain also suggest that mechanism-based phenotyping will be needed to combine drugs, rehabilitation and neuromodulation rationally.105 Endocannabinoid and TRP-channel literatures further show that peripheral molecular targets may influence which patients remain peripherally driven and which become more centrally amplified.106

The most useful trials will therefore be neither purely pragmatic nor purely mechanistic. They should be pragmatic enough to measure disability, sleep, medication burden and treatment attendance. They should also be mechanistic enough to report sensory phenotype, cortical excitability, target localization and maintenance exposure. This combined design would make it possible to ask whether rTMS is an analgesic procedure, a network-modulation therapy or a selection tool for more invasive neuromodulation.

Conclusion

rTMS is no longer merely an experimental curiosity for PNP. The strongest current evidence supports high-frequency M1 stimulation, particularly in PHN, with a generally favourable safety profile when consensus precautions are followed. However, the evidence does not justify a uniform clinical algorithm for all PNP. Accordingly, future clinical trials and translational development should evaluate rTMS within an aetiology- and phenotype-stratified, mechanism-informed framework for selected patients, using transparent responder definitions and maintenance strategies. This proposal does not establish a clinical algorithm; it defines a testable research agenda for matching peripheral lesion phenotypes to central pain-modulation targets.

Abbreviations

ATP, Adenosine Triphosphate; CI, Confidence Interval; DLPFC, Dorsolateral Prefrontal Cortex; DNP, Diabetic Neuropathy; EFNS, European Federation of Neurological Societies; GABAB, Gamma-Aminobutyric Acid Type B; GRADE, Grading of Recommendations Assessment, Development and Evaluation; ICD-11, International Classification of Diseases, 11th Revision; IFCN, International Federation of Clinical Neurophysiology; M1, Primary Motor Cortex; MD, Mean Difference; NaV1.7, Voltage-Gated Sodium Channel 1.7; NeuPSIG, Neuropathic Pain Special Interest Group; P2X4, P2X Purinergic Receptor 4; PAG, Periaqueductal Grey; PAG-RVM, Periaqueductal Grey-Rostral Ventromedial Medulla Pathway; PHN, Postherpetic Neuralgia; PNP, Peripheral Neuropathic Pain; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; rTMS, Repetitive Transcranial Magnetic Stimulation; RVM, Rostral Ventromedial Medulla; SMD, Standardized Mean Difference; tDCS, Transcranial Direct-Current Stimulation; TMS, Transcranial Magnetic Stimulation; TRP, Transient Receptor Potential; TRPV1, Transient Receptor Potential Vanilloid 1.

Data Sharing Statement

No new datasets were generated or analysed for this narrative review.

Ethics Approval and Informed Consent

Not applicable because this narrative review did not involve human participants, human data or animal experiments.

Author Contributions

Jie Zhuang and Dorota Duhova contributed equally to this work, and Lifeng Qian and Lei Fang contributed equally to this work. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Funding

This work was supported by the Zhejiang Provincial Rehabilitation Medicine Association Scientific Research Fund (No. ZKKY2024010) and the Research Project of Shanghai Second Rehabilitation Hospital (No. Y2025-07).

Disclosure

The authors report no conflicts of interest in this work.

References

1. Jensen T, Baron R, Haanpää M, et al. A new definition of neuropathic pain. Pain. 2011;152(10):2204–17. doi:10.1016/j.pain.2011.06.017

2. Colloca L, Ludman T, Bouhassira D, et al. Neuropathic pain. Nat Rev Dis Primers. 2017;3(1):17002. doi:10.1038/nrdp.2017.2

3. Scholz J, Finnerup N, Attal N, et al. The IASP classification of chronic pain for ICD-11: chronic neuropathic pain. Pain. 2019;160(1):53–59. doi:10.1097/j.pain.0000000000001365

4. van Hecke O, Austin S, Khan R, Smith B, Torrance N. Neuropathic pain in the general population: a systematic review of epidemiological studies. Pain. 2014;155(4):654–662. doi:10.1016/j.pain.2013.11.013

5. Truini A, Aleksovska K, Anderson CC, et al. Joint European Academy of Neurology-European Pain federation-neuropathic pain special interest group of the international association for the study of pain guidelines on neuropathic pain assessment. Eur J Neurol. 2023;30(8):2177–2196. doi:10.1111/ene.15831

6. Soliman N, Moisset X, Ferraro M, et al. Pharmacotherapy and non-invasive neuromodulation for neuropathic pain: a systematic review and meta-analysis. Lancet Neurol. 2025;24(5):413–428. doi:10.1016/S1474-4422(25)00068-7

7. Finnerup N, Attal N, Haroutounian S, et al. Pharmacotherapy for neuropathic pain in adults: a systematic review and meta-analysis. Lancet Neurol. 2015;14(2):162–173. doi:10.1016/S1474-4422(14)70251-0

8. Moisset X. Neuropathic pain: evidence based recommendations. Presse Med. 2024;53(2):104232. doi:10.1016/j.lpm.2024.104232

9. Mills SE, Nicolson KP, Smith BH. Chronic pain: a review of its epidemiology and associated factors in population-based studies. Br J Anaesth. 2019;123(2):e273–e283. doi:10.1016/j.bja.2019.03.023

10. Cohen SP, Mao J. Neuropathic pain: mechanisms and their clinical implications. BMJ. 2014;348(feb05 6):f7656. doi:10.1136/bmj.f7656

11. Wagner T, Valero-Cabre A, Pascual-Leone A. Noninvasive human brain stimulation. Annu Rev Biomed Eng. 2007;9(1):527–565. doi:10.1146/annurev.bioeng.9.061206.133100

12. George MS,