Noninvasive transcranial brain stimulation in central post-stroke pain: A systematic review

1 Introduction

Chronic pain is one of the most common complaints in adult patients who seek medical care and is associated with the limitation of physical activity, suffering, opioid dependence, and substantial economic impact [1]. Chronic pain syndromes are common in stroke survivors, affecting more than 1 in 10 patients, with the majority occurring in the first month post-stroke [2]. Central post-stroke pain (CPSP) is a neuropathic pain syndrome that results either acutely or in the chronic phase of a cerebrovascular event and it is a result of lesions at any level of the somatosensory tract of the brain, including the medulla, thalamus, and cerebral cortex [3]. Hyperalgesia and allodynia are predominant characteristics of neuropathic pain resulting from lesion or disease of the somatosensory nervous system and are the main targets of the CPSP treatment.

The pathophysiological mechanisms involved in CPSP are complex, multifaceted, and poorly understood. Some theories have been proposed, such as central imbalance, central sensitization, central disinhibition, other thalamic adaptive changes, and local inflammatory responses [4]. In addition, the coexistence of psychological factors, which is associated with the vast number of central mechanisms, makes CPSP treatment a complex challenge [5]. Various studies using neuroimaging, such as positron emission tomography (PET) [6,7] and functional magnetic resonance imaging (fMRI) [8,9], were conducted to investigate post-stroke neural network activation and connectivity changes. Both ipsilateral and contralateral changes were observed, but a strong association between ipsilateral activation of motor cortex and motor recovery was found [10]. By contrast, fMRI overactivity of the contralesional motor cortex in the chronic phase was associated with poor recovery and, for this reason, the interhemispheric inhibition theory, which postulates imbalanced neural excitation and inhibition, was proposed to explain sustained motor and somatosensorial symptoms in post-stroke patients [11,12]. Additionally, limbic structures are associated with the affective part of the pain circuit and can also be affected in CPSP [13]. Kim et al. [14] identified, through lesion network mapping, that lesions causing CPSP are related to a specific brain network associated with metabolic changes, which could be clinically important for neuromodulation therapies. They identified twelve neuromodulation targets, of all of them ipsilateral M1 had the best evidence of efficacy and they extended into S1 and the superior parietal lobe, regions linked with sensation or attentional modulation of sensory perception. They also verified prospectively that transcranial magnetic stimulation (TMS) targeting the sensorimotor cluster of their pain network improved pain scores in seven patients with CPSP, which shows that treatments focusing S1 and M1 at the same time could be better than these regions isolated [14].

The management of CPSP includes pharmacological and non-pharmacological treatment. However, in practice, and in spite of the available options, central pain is difficult to treat, being a source of suffering for patients and a constant challenge for the physician [5].

Noninvasive brain stimulation (NBS) has emerged as a promising approach for managing various pain disorders, including neuropathic pain, visceral pain, and migraine. Among the techniques employed in NBS, TMS stands out as the most commonly utilized, followed by transcranial direct current stimulation (tDCS) [15].

In TMS, magnetic pulses are administered via a coil placed on the scalp. These pulses, through electromagnetic induction, generate currents and electric fields (E-fields) within the cortex. Depending on the intended outcome, TMS can be delivered as single pulses or in repetitive trains. Single pulse TMS is used to investigate brain function, while repetitive TMS (rTMS) is employed to induce lasting alterations in brain activity beyond the stimulation period [16,17].

TMS pulses reliably depolarize neurons when the strength of the magnetic pulses and consequent induced current and E-field exceed the threshold value of cortical neurons. Notably, rTMS applied over the motor cortex induces sustained changes in motor-evoked potentials (MEP) amplitudes that persist post-stimulation. These changes are determined by the frequency of rTMS: high frequency (HF) (>5 Hz, typically 10–20 Hz in clinical practice) activates the targeted neural network, while low frequency (<1 Hz) deactivates it, akin to long-term potentiation (LTP) or long-term depression effects [16,17]. Beyond the frequency employed, the orientation of the coil, the type of coil, the number of sessions, and the intensity influence the stimulation effect [18].

Transcranial electric stimulation techniques involve the application of current to two or more surface electrodes, with one electrode at least placed on the scalp, and they are used mostly at the subthreshold level. There are countless electric stimulation patterns with different frequencies, intensities, and durations; the simplest approach is tDCS, which has the low end of the frequency spectrum. tDCS drives to a polarization of neuronal membranes. A surface anode hyperpolarizes the superficial layers and depolarizes layer 5, by increasing pyramidal tract neuronal firing rate, and vice versa with cathode stimulation. Generally, at rest, anodal stimulation conducts to excitation quantify commonly in MEP increase after tDCS, while cathodal stimulation leads to inhibition [19].

Clinically, tDCS is a portable, low-cost, and well-tolerated non-invasive neuromodulatory technique, which gave rise to hundreds of published human studies to potentially treat several neurologic and psychiatric disorders, even intractable chronic pain. On the other hand, no corporation leads to tDCS development, and studies are so heterogeneous that clinicians might have difficulties evaluating its benefits and therapeutic conditions. Fregni et al. [20] summarized recommendations on tDCS efficacy/safety according to clinical indication in nine disorders, including neuropathic pain, fibromyalgia, migraine, and postoperative acute pain.

Over the years, the literature has demonstrated the efficacy of TMS in chronic pain and, for that reason, in recent recommendations, TMS of the primary motor cortex contralateral to the pain area is indicated as an effective therapeutic option, with level of evidence A for neuropathic pain [21,22]. Severe adverse effects such as seizures are extremely rare with these treatments [23].

In cases of refractory CPSP, NBS therapy such as rTMS or tDCS can be recommended [3,4]. Considering the preliminary evidence of the efficacy and safety of TMS for treating some types of pain, multiple authors recognize the importance of a greater investment in developing TMS procedures in this context [24].

2 Objectives

The aim of this systematic review is to analyze the efficacy of NBS (TMS and tDCS) in the treatment of CPSP.

3 Methods

This systematic review is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guideline 2020 statement. Institutional review board and informed consent were not relevant for this systematic review because it relied on published information.

4 Results

4.1 Study selection

The literature search resulted in 3,943 articles. The automatic removal of duplicates resulted in 2,958 articles, which were manually analyzed by the two authors, obtaining a total of 2,674 articles eligible for this review.

After the first selection, based on titles and abstracts, 25 articles were selected and their full texts were read by the two authors (RSM and NRF). Most of the articles were excluded due to the study design and included other forms of neuropathic pain, besides CPSP. In these cases, when was not possible to extract the data of post-stroke pain separately, the article was excluded.

Only one RCT assessing the effectiveness of tDCS was found; however, due to inappropriately analyzed data, the authors opted not to include it in this paper [28].

A total of five articles were deemed suitable for this review. The selection scheme (flow chart) for this systematic review is shown in Figure 1.

Figure 1

PRISMA flow chart.

4.2 Risk of bias assessment

Figure 2 shows a summary of the risk of bias assessment of the included articles, evaluated with the RoB 2.0 tool. All of the studies showed a low risk of bias concerning missing outcome data, deviation from intended interventions, and selection of the reported result. Overall, the risk of bias due to the measurement of outcome was low. The randomization process was revealed to be a high risk of bias in two of the five articles, due to the non-concealed allocation sequence until participants were enrolled and assigned to interventions. The crossover studies showed low risk when carryover effects were analyzed.

Figure 2

Risk of bias (RoB 2.0) of the included studies.

Regarding the strength of evidence overall classification, the GRADE tool demonstrated a low quality of evidence for rTMS studies (Table 2).

Table 2

Quality of evidence of non-invasive neuromodulation for the treatment of Central Poststroke Pain

Certainty assessment							Summary of findings
No of participants (studies)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence
rTMS (5 RCTs), active group, N = 75, sham group, N = 67	Seriousa	Not serious	Not serious	Seriousb	All plausible residual confounding would suggest spurious effect, while no effect was observed	⨁⨁◯◯ low	rTMS is a promising approach for the treatment of poststroke pain.

⨁⨁◯◯ represents that the classification has obtained in the evaluated categories (risk of bias, inconsistency, indirectness, imprecision). ⨁⨁ represents categories without serious compromise of study quality (indirectness, imprecision), while ◯◯ represents categories that may negatively influence the quality of evidence (risk of bias, inconsistency).

^aThe item risk of bias received a downgrade due to the major of studies did not perform Allocation concealment properly.

^bThe item imprecision received a downgrade because patients available for this analysis do not therefore meet the IOS criterion.

5 Discussion

We performed a systematic review of the efficacy of NBS in the treatment of post-stroke pain, focusing on the two most common techniques: TMS and tDCS. Although a systematic review was published in 2019 [34], only three databases and articles from 2007 to 2018 were included, one out of six was a randomized controlled trial and only one was considered “good/excellent” in quality risk of bias assessment. Since NBS is an emerging therapy, new literature has been published in the last few years, including randomized controlled trials, which motivated the authors of this review to carry out a new systematic review in this field.

Recent evidence has shown that HF rTMS applied in the M1 area contralateral to the painful side is an effective treatment for peripheral neuropathic pain (Level A evidence) [22]. However, this systematic review reveals that there is still a scarcity of RCTs evaluating the effectiveness of NBS in post-stroke pain in the literature, despite an increase in the number of RCTs over the past 3 years. In comparison to the systematic review conducted by Ramger et al. [34] in 2019, this review includes three articles published between 2020 and 2021.

Only the study conducted by De Oliveira et al. [33] did not show a statistically significant reduction in pain intensity. De Oliveira et al. [33] stimulated the left PMC/DLPFC but terminated the study due to a lack of effect. The role of the DLPFC in pain modulation has been substantiated by numerous experimental and neuroimaging investigations. In fact, specific studies employing DLPFC rTMS in healthy individuals have illustrated that such stimulation induces relief from experimental thermal pain [35]. Taylor et al. [36] conducted a study with healthy volunteers, comparing the analgesic effects of DLPFC rTMS with and without naloxone in hot allodynia. The study demonstrated that naloxone pretreatment significantly abolished rTMS-induced analgesia. In addition, in the fMRI analysis, rTMS induced a reduction of BOLD signal response to painful stimuli across pain processing regions, including the midbrain and medulla. These initial findings suggest that left DLPFC rTMS elicits top-down opioidergic analgesia. Andrade et al. [37] investigated the contribution of N-methyl-d-aspartate (NMDA) glutamate receptors to DLPFC and M1 TMS in healthy individuals. The study revealed that ketamine injection, a noncompetitive NMDA antagonist, attenuated the analgesic effects of rTMS in healthy subjects. Notably, this effect was observed with both DLPFC and M1 stimulation, suggesting a shared pathway involving NMDA receptors for both types of stimulation. Furthermore, these findings suggest that an LTP-like phenomenon may underlie the effects of TMS. Despite the findings from studies involving healthy individuals, both the guidelines for the therapeutic use of TMS and the systematic reviews conducted by O’Connell et al. [38] and Jiang et al. [39] did not find evidence supporting the effectiveness of DLPFC rTMS in reducing pain intensity in chronic pain. De Oliveira et al. [33] suggested that disruption of somatosensory integrity in CPSP and other neuropathic pain conditions could potentially reduce rTMS-induced analgesia in this group of patients.

On the other hand, all of the TMS studies included in this review that stimulated the M1 area using HF demonstrated a statistically significant reduction in pain intensity [29–31]. These results are consistent with the guideline for the therapeutic use of TMS, which provided Level A evidence for the treatment of peripheral neuropathic pain using HF M1 TMS. Additionally, they align with the findings of the systematic review conducted by O’Connell et al. [38] and Jiang et al. [39], who reported evidence showing that HF M1 rTMS may have short-term effects on chronic pain. Another critical consideration in NBS is the selection of the stimulation site for rTMS. In cases of peripheral neuropathic pain, it is established practice to apply M1 rTMS on the side contralateral to the pain. However, in post-stroke complications, it is advised to administer HF rTMS on the side of the lesion and low-frequency rTMS on the contralateral side of the lesion [14,40,41]. These localizations are based on the theory of interhemispheric inhibition dysregulation following a stroke. Under normal conditions without lesions, both hemispheres regulate each other’s cortical excitability via the corpus callosum pathway, maintaining a balanced bilateral interhemispheric excitability. However, following a stroke, this balance is disrupted, resulting in imbalanced interhemispheric inhibition. This imbalance leads to reduced cortical excitability in the injured hemisphere due to excessive suppression, while cortical excitability in the uninjured hemisphere is increased [40]. The study conducted by Kim et al. [14] identified the network of brain regions functionally connected to each lesion site and identified connections significantly associated with CPSP. The authors found that CPSP lesions exhibited stronger connectivity with bilateral M1, S1, and the occipital cortex compared to controls. Additionally, they demonstrated that patients with CPSP showed decreased glucose metabolism in the ipsilesional M1, ipsilesional S1, and contralesional occipital cortex. These findings support the notion that ipsilesional M1 rTMS is more closely aligned with the lesion-based pain network than other neuromodulation targets with less evidence of efficacy [14]. Among the articles analyzed, only Zhao et al. [31] clearly delineated that the stimulation targeted the M1 area of the injured hemisphere. Conversely, the remaining studies applied rTMS to the contralateral M1 in relation to the patient’s pain. Given that in CPSP, sensory abnormalities and pain typically manifest contralateral to the lesion in stroke, the chosen stimulation sites in the reviewed articles appear appropriate. However, it is essential to note that in cases of brainstem lesions, pain may affect the ipsilateral face [42]. Therefore, it is recommended that future articles provide a more detailed description of the precise site of rTMS application to enhance clarity and facilitate comparability across studies.

The study conducted by Ojala et al. [32] evaluated S2 rTMS, comparing it with M1 and sham stimulation. The authors observed that S2 rTMS significantly reduced the weekly average pain intensity, as well as the long-term pain intensity at 1 month, compared with baseline, while M1 rTMS and sham did not yield similar effects. However, no significant differences were found between the groups. Although Zhao et al. [31] suggest that S2 may represent a promising new target for rTMS, the data from this study are not conclusive for this new indication. This is because S2 rTMS did not demonstrate superiority over sham stimulation, and the study has limitations, such as its crossover design and a relatively small sample size. On the other hand, in the study conducted by Kim et al. [14], previously cited, it was found that the lesion-based pain network includes not only M1 but also extends beyond it, specifically into S1 and the superior parietal lobe. Consequently, the authors propose that since noxious stimuli undergo parallel processing in both S1 and M1, neuromodulation treatments targeting both regions simultaneously could potentially yield greater benefits than targeting either M1 or S1 alone.

Regarding rTMS protocols, there was no homogeneity between them. The stimulation frequency (10 or 20 Hz), the intensity used (80–90% of the RMT, except De Oliveira et al. [33] that used an intensity of 120% of the RMT), and the number of sessions (ranging from 4 sessions to 20 sessions) varied between studies and seems to not be associated with different efficacy or side effects occurrence. To date, the optimal protocol for rTMS in CPSP has yet to be established. However, based on existing scientific evidence regarding the effectiveness of rTMS in neuropathic pain, it has been observed that variations in the frequency of rTMS (5, 10, or 20 Hz) do not significantly influence the effectiveness of analgesia. Conversely, the number of sessions appears to have an impact on the effectiveness of analgesia and the duration of its effects. It is hypothesized that repeated rTMS sessions may have a cumulative effect on pain relief [15]. A team of French researchers released two articles presenting findings from the extended application of M1 rTMS in patients with central neuropathic pain, including CPSP. The initial phase of the protocol involved a series of four sessions conducted over 2 months. Subsequently, for participants exhibiting a pain relief percentage exceeding 10%, sessions were prolonged and administered at intervals tailored to the duration of the analgesic effect observed in each individual [43,44]. In the study conducted by Pommier et al. [44], it was observed that nine participants did not exhibit a response to the initial sessions, while the remaining 31 patients experienced a cumulative effect across sessions, resulting in an average pain relief of 41% over a span of 15.6 days. Furthermore, a correlation was identified between the degree of pain relief observed in the initial session and the long-term pain alleviation. The study by Quesada et al. [43] demonstrates similar results. Of the patients commencing treatment, 76% exhibited a response, with an average pain relief of 10%. Over the course of 12 months (15 sessions), a cumulative effect was observed, resulting in a notable increase in the percentage of average pain relief (48%) and the duration of symptom relief (20 days). This effect attained statistical significance after four sessions and persisted throughout the 12-month duration. In the studies included in this review that solely evaluated M1 rTMS, Zhao et al. [31] conducted the largest number of sessions (18 sessions), while Khedr et al. [29] and Quesada et al. [30] administered 5 and 4 sessions, respectively. Despite Zhao et al. [31] and Khedr et al. [29] observing a significant difference between real rTMS and sham rTMS in all follow-ups, a slight increase in pain scores was noted from the second follow-up week. Conversely, in the study by Quesada et al. [30] (with a 3-week follow-up), pain scores remained relatively stable. Although it was expected that Zhao et al. [31] results would demonstrate better stability based on the theory of cumulative effects of repeated rTMS sessions on pain relief. However, differences in pain intensity measurement tools, CPSP phase, and other protocol variations render comparisons among these studies of limited relevance. These findings underscore the significance of conducting studies with extended follow-up periods and the necessity for the development of open-label naturalistic studies that incorporate maintenance sessions. Such endeavors are crucial for establishing comprehensive guidelines for clinical practice.

Notably, Zhao et al. [31] was the sole study to include participants with acute CPSP. There is limited information available regarding the use of rTMS in the acute phase of CPSP. However, in the context of other comorbidities associated with stroke, the optimal timing for rTMS application has been extensively discussed, whether in the acute (subacute) or chronic phase. The utilization of rTMS in the acute phase of CPSP warrants further investigation. It is well-established that during the initial phase of stroke recovery, neurological recuperation occurs more rapidly. Therefore, seizing the opportunity for high-intensity interventions during this period could significantly benefit symptom recovery. For instance, the first 6 months post-stroke represent a critical window for patients to regain upper limb function. Failure to capitalize on this timeframe may result in lost opportunities for upper limb recovery, as subsequent rehabilitation efforts may have minimal impact on neurological improvement [40].

Zhao et al. [31] found that the analgesic effect of rTMS in CPSP patients was followed by increased secretion of BDNF and enhanced cortical excitability in the third week. There was a negative correlation between serum BDNF levels and pain scores, similar to previous studies with chronic myofascial pain syndrome and depression [45,46]. BDNF plays a crucial role in brain plasticity, facilitating the survival, differentiation, and maturation of neurons within the nervous system. In addition, it exhibits a neuroprotective effect in challenging conditions, such as neurotoxicity, cerebral ischemia, and hypoglycemia. Moreover, BDNF stimulates and regulates neurogenesis – the growth of new neurons from neural stem cells [47]. A systematic review conducted by Karantali et al. [48] revealed that BDNF levels were significantly lower in patients with stroke in comparison to healthy individuals. It is hypothesized that the level of BDNF could serve as an indirect indicator of the potential for improvement in stroke outcomes [47]. Therefore, the findings presented by Zhao et al. [31] may offer a novel perspective for the treatment of CPSP using rTMS. Further studies with long-term follow-up, particularly focusing on the acute phase of CPSP, are warranted to assess the true benefits of rTMS during the initial stages of stroke recovery.

A quantitative analysis was not feasible, due to the heterogeneity of the design of the studies and low sample size, which precluded random effects analyses.

Only one RCT assessing the effectiveness of tDCS was found; however, the authors chose not to include it in the review because they considered that the data were not properly analyzed [28]. Contrary to TMS, tDCS does not provide sufficient evidence of its efficacy in neuropathic pain and, for that reason, there are not any Level A or B recommendation for its use. A Level C recommendation is proposed for anodal tDCS of the left M1 (or contralateral to pain side, with right orbitofrontal cathode) in chronic lower limb neuropathic pain secondary to spinal cord lesion [49].

Concerning the quality of life and functional limitation, only two studies [32,33] evaluated them and did not show significant changes after treatment. All of the studies assess the occurrence of adverse effects, with no or mild and transient effects reported. Three of the studies [31–33] evaluated the psychological state before and after the treatment, but no difference in anxiety and depression scores was observed. None of the articles evaluated the use of on-demand therapy.

Another noteworthy finding from the included studies is the proportion of individuals categorized as non-responders. In the study by Quesada et al. [30], 53% of the sample did not achieve a decrease in pain intensity of at least 10% compared to baseline, classifying them as non-responders. Similarly, Khedr et al. [29] identified 36% of individuals with a poor response, defined as a reduction in pain score by less than 40%. In the open-label studies conducted by Pommier et al. [44] and Quesada et al. [43], approximately 24% of participants were classified as non-responders, failing to attain an average pain relief of 10%. Currently, the adoption of precision medicine is advocated, emphasizing treatment tailored to individual patient needs, encompassing genetic, molecular, clinical, and psychosocial considerations. According to this approach, it is hypothesized that the challenge in determining treatment effectiveness for chronic pain conditions may stem from the heterogeneity observed among patients sharing the same diagnosis. Consequently, there is a call for the stratification of patient subgroups with similar characteristics (clustering). By assessing these inter-subject variabilities, it becomes possible to identify subgroups more likely to respond favorably to a specific intervention (prediction) [50]. Some fMRI studies have indicated a relationship between the efficacy of rTMS and the integrity of the corticospinal tract and thalamocortical tract. One study demonstrated a significant correlation between the integrity of the superior thalamocortical tract in the ipsilesional hemisphere and the change in VAS score following rTMS. Additionally, fMRI revealed significantly decreased activity in the secondary somatosensory cortex, insula, prefrontal cortex, and putamen among rTMS responders, while non-responders showed no change [51]. Another study found that the effectiveness of rTMS was more strongly associated with lesions of the thalamocortical tract than with lesions of the corticospinal tract, suggesting that severe impairment of the thalamic nuclei and thalamocortical tract may influence hyperexcitability in the thalamus and cortex, or within the rTMS pain relief pathway [52]. Hence, there is a critical need to develop predictive biomarkers to gauge the effectiveness of rTMS. Techniques like fMRI assessment and biomarkers such as BDNF, as demonstrated by Zhao et al. [31], hold promise in this regard.

Despite the increasing evidence regarding the efficacy of NBS in pain management, its mechanisms of action are not completely understood. Some theories have been proposed to explain the mechanism by which NBS of the M1 area exerts their analgesic effects in neuropathic pain. The neural response induced by rTMS is not confined to the stimulated brain area, but can also spread to other cortical regions remote from the stimulated area [53,54].

Activation of top-down control of intracortical horizontal fibers located parallel to the precentral gyrus seems to stimulate inner pathways and structures such as the insula, the cingulate cortex, the thalamus, and portions of the brainstem, which might explain the effects of NBS on the affective dimension of pain and the endogenous opioid system [55]. Lamusuo et al. [56] showed activation of endogenous opioids through TMS applied to M1 or S1, with PET neurotransmitter, thereby reinforcing its likely involvement in the analgesic effect of TMS. Furthermore, Strafella et al. [57] demonstrated a dopamine release following cortical TMS stimulation, emphasizing the crucial role of the dopamine-opioid pathway in pain. Evidence also showed that HF-rTMS can repair malfunctioning intracortical inhibition, associated with chronic neuropathic pain, through the modulation of both inhibitory (GABAergic) and excitatory (glutamatergic) pathways [58,59].

A profound understanding of the neural mechanisms underlying rTMS is essential for developing efficient therapy in the future. Additional placebo-controlled studies and large studies with adequate blinding to the sham procedure, precision for coil positioning, and direct comparisons of cortical targets of stimulations are needed, to better understand the rTMS action mechanisms, namely long-term effects. Considering the aforementioned points, the utilization of M1 rTMS in post-stroke pain management currently presents a low quality of evidence but shows promising results. This review revealed that rTMS applied to the ipsilesional M1 region led to significant improvements compared to placebo for at least 3 weeks post-treatment. Interestingly, variations in stimulation frequency (10 or 20 Hz) and the number of pulses per session (ranging from 1,500 to 2,000) did not appear to influence effectiveness. Moreover, the number of sessions did not impact the efficacy of rTMS based on the included studies. However, data from open-label investigations suggest that tailoring session frequency over time to individual needs may enhance effectiveness. Additionally, assessing treatment response in initial sessions can help avoid prolonged and potentially unnecessary treatments.

6 Conclusion

NBS seems to be effective in reducing pain in short term, in patients with CSPS, when applied in the M1 area, without being associated with severe or permanent adverse effects. Quality of life, functional limitation, and psychological state do not appear to improve with the treatment; however, few studies formally evaluated them. Despite the good outcomes related to pain relief, these results should be taken into account with prudence, as they have some limitations: reduced number of RCTs studying the efficacy of NBS in patients with CSPS, reduced sample size (that can lead to type I and II errors) and heterogeneous protocols. As most of the studies only evaluated the short-term efficacy of noninvasive transcranial brain stimulation in CPSP, there is also a lack of information regarded to long-term efficacy of these treatments or the necessity of a maintenance session. Future studies should be performed using a more rigorous design and efforts should be taken to homogenize the protocols and outcomes assessment tools, to allow a better analysis and inference of treatment effects.