Effect of Physiotherapeutic Interventions on Biomarkers of Neuropathic Pain: A Systematic Review of Preclinical Literature

The purpose of this systematic review was to evaluate the effects of physiotherapeutic interventions on biomarkers of neuropathic pain in preclinical models of peripheral neuropathic pain (PNP). The search was performed in Pubmed, Web of Science, EMBASE, Cochrane, Cinhal, Psycinfo, Scopus, Medline, and Science Direct. Studies evaluating any type of physiotherapy intervention for PNP (systemic or traumatic) were included. Eighty-one articles were included in this review. The most common PNP model was chronic constriction injury, and the most frequently studied biomarkers were related to neuro-immune processes. Exercise therapy and Electro-acupuncture were the 2 most frequently studied physiotherapy interventions while acupuncture and joint mobilization were less frequently examined. Most physiotherapeutic interventions modulated the expression of biomarkers related to neuropathic pain. Whereas the results seem promising; they have to be considered with caution due to the high risk of bias of included studies and high heterogeneity of the type and anatomical localization of biomarkers reported. The review protocol is registered on PROSPERO (CRD42019142878).


Introduction
Neuropathic pain (NP) is defined as pain caused by a lesion or a disease of the somatosensory system 1 and is estimated to affect between 6.9 and 10% of the general population. 2,3 Peripheral neuropathic pain is becoming more prevalent due to an aging world population, the rising impact of diabetes mellitus as well as higher survival rates of cancer and the implications of chemotherapy. 4 Management of NP remains challenging, as many patients do not experience adequate pain relief. [5][6][7][8] Treatment of neuropathic pain usually focuses on symptom management. 9 Nonsurgical interventions are recommended as first-line treatments for patients with neuropathic pain. 10 Among the nonsurgical interventions, the Neuropathic Pain Special Interest Group of the International Association for the Study of Pain recommends pharmacology as first-line treatment. 5,11 However, efficacy is limited 12 with often unacceptable side effects. [12][13][14] Over the past decade, the role of Physiotherapy and physical activity has gained increasing interest in the treatment of neuropathic pain. 15 Several studies have been published on the efficacy of physiotherapy on peripheral neuropathic pain resulting from systemic 16 or focal nerve damage. 17,15 In addition several guidelines propose active exercise as a treatment option for neuropathic pain. 18, 19 Although some studies suggest that physiotherapy provides significant improvements in pain, quality of life and disability in patients with peripheral neuropathies and neuropathic pain, 20,21 other studies did not report similar findings 15 and the mixed quality of studies prevents firm conclusions. 15 Whereas human studies evaluating physiotherapy for neuropathic pain focus on improving pain, function and quality of life, the mechanisms by which physiotherapy interventions work remains poorly understood. A better understanding of the mechanisms of action of physiotherapy would help the selection of the most promising disease modulating physiotherapy interventions for future clinical trials.
The body of literature exploring the mechanisms of action of physiotherapeutic interventions using preclinical models has grown substantially over the past years. The main objective of this systematic review is therefore to summarize this literature by assessing the effect of physiotherapeutic interventions on biomarkers of neuropathic pain in pre-clinical models.

Methods
This systematic review was conducted following the guidelines of the Systematic Review Center for Laboratory Animal Experimentation (SYRCLE), the Cochrane Handbook for Systematic Review of Intervention, 22 the original guide "Preferred Reporting Items for Systematic Reviews, PRISMA" and the most recent update from 2021. 23 The protocol has been prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO, CRD42019142878).

Literature Search
A systematic search was developed following the step-by-step guide suggested by Leenaars et al. 24 The following databases were searched from inception to 13th January 2020 and updated in February 2022: MED-LINE EMBASE, CINAHL, SCOPUS, Web of Science, PubMed, Cochrane library and PsycINFO. The search strategy is described in Appendix 1.

Types of Studies
Original animal studies reporting the effect of physiotherapeutic interventions compared to a control group on peripheral neuropathic pain were included. Case studies, cross-over studies, and studies without a separate control group were excluded. Letters, reports, or abstracts from congresses were not included. Only articles with access to the full-text in English and Spanish language were included.

Animal Models
In-vivo animal models of neuropathic pain induced by both systemic (eg, diabetic or chemotherapy induced neuropathy) and focal nerve injury (eg, nerve ligation, crushing or transection) were included. We excluded studies where due to the model or validation tests (eg, sensory thresholds), we could not ascertain that the animals had developed neuropathic pain. We also excluded studies with animals with co-morbidities (eg, pre-ischemic physiologic conditions such as ischemic injury) and studies that evaluated the prevention rather than the treatment of already existing neuropathic pain.

Study Selection
Before carrying out the article selection procedure, a search for duplicates was carried out with MENDELEY. In a first phase, 2 independent reviewers (L.M and A.A.) assessed the eligibility of the studies based on information from title, abstract and keywords. During the second phase, the full text was independently reviewed by both reviewers for eligibility. A third reviewer (C.G.) acted as a mediator when there were differences of opinion between the 2 reviewers, with the 3 reviewers reaching consensus. 25

Data Extraction and Management
Data of included studies were extracted by 2 independent reviewers (L.M and A.A.). This involved registered bibliographic data, such as first author and year of publication, animal characteristics (species, age, weight, and gender), neuropathic pain model, treatment groups and intervention characteristics (physiotherapeutic intervention, timing of intervention, number of treatment sessions, duration, dose and location). We also extracted the type of biomarkers including in which tissue they were measured. We attempted to extract means, standard deviations, and P values for all biomakers. If available, we recorded behavioral test outcomes to confirm the presence of neuropathic pain. Finally, both authors reached consensus on each item of extracted data. In case of disagreement between the authors, a third author (C.G.) made the final decision.

Risk of Bias Assessment
The risk of bias of each study was assessed using SYRCLE's risk of bias tool 26 scored by 2 independent reviewers (Y.G and E.C.). The tool provides 10 items. These categories are related to selection bias, performance bias, detection bias, attrition bias, information bias, and other biases. Half of these items match those in the tool developed by Cochrane. If there was any disagreement or discrepancy, it was resolved by a third reviewer (J.F.C.). As the tool does not include a specific cut-off, we considered studies to have low risk of bias if they were rated as high bias on less than half of the scoring criteria (<5 out of 10).

Reporting Quality
To evaluate the reporting quality of the studies we used the "Animals in research: reporting in vivo experiments" (ARRIVE) guidelines. 27 The scale has 20 items. Each item refers to a specific section of an article (eg, title, abstract), and other items refer to specific elements of preclinical research (eg, allocation of the animals, housing and husbandry). The score was assessed by 2 independent reviewers (Y.G and E. C.). Any discrepancies were resolved by consensus with a third reviewer (F.C.M). Each ARRIVE item was graduated into 3 descriptive levels: complete (green) when all sub-items in the topic have been described; partial (yellow) when one or more of the sub-items have been described; and incomplete (red) when none of the sub-items have been described. As the tool does not include a cut-off, we considered articles to have good reporting quality if they reported at least 60% of items completely.

Qualitative Analysis
For the description of the results, the studies were grouped by type of intervention (eg, exercise, electroacupuncture) as well as type and location of reported biomarkers.
Due to the heterogeneity of reported biomarkers, anatomical measurement sites and measurement methods (eg, gene expression, immunohistochemistry, protein level), and the missing summary statistics in many studies, a meta-analysis could not be carried out.
Instead, we report these findings with heat maps for each intervention and at each location (eg, spinal cord, dorsal root ganglia): color coding was assigned according to the frequency of studies reporting any change on individual biomarker expression (eg, increase, decrease or no change) after the intervention.

Selection of the Studies
The database search retrieved a total of 5,038 articles. After reviewing the titles and abstracts, 179 studies were assessed for eligibility. Of those, 94 were excluded because they did not satisfy the eligibility criteria. This resulted in the inclusion of 85 full-text articles. The flow diagram is shown in Fig 1. The country that produced the most eligible studies is China (38.8%), followed by Brazil (20%) and Taiwan (16.4%). Italy, the United States and Japan contributed with 4.7% each, while Spain, South Korea and Turkey produced 3.5% of included studies. After the selection process, all articles were written in English. No articles in Spanish were found.

Risk of Bias Analysis
Only 2 of the 85 papers had a low risk of bias, obtaining a 5 per 10 score on the SYRCLE tool. The remaining articles had a high risk of bias (Table 1).

ARTICLE IN PRESS
Reporting Quality According to ARRIVE Fifty-eight (71.6%) out of 85 articles were rated as 60% or more "complete" according to the ARRIVE guidelines. Twenty-one (80.8%) of the 26 articles exploring the effect of exercise are of good quality. Thirty-three percent (1 out of 3) of the acupuncture and joint mobilization articles have low quality. Of the reports on electroacupuncture, 24.14% (7 of the 29) have low methodological quality. All articles on neural mobilization showed good methodological quality (5 out of 5). Of studies including physical agents, 57.9 % (11 out of 19) were of good quality (Supplementary Table 1).

Characteristics of the Studies
Characteristics of the included articles, such as details of animal species, neuropathic pain models and treatment groups and interventions are shown in supplementary Table 2.
The most widely used model of neuropathic pain was traumatic nerve injury (78.9%), with chronic constriction injury being the most studied model (55.8%) followed by sciatic nerve cut (13%). Other models reported were diabetic neuropathy, complex regional pain and chemotherapy induced neuropathy. 82.72% of the articles confirmed the presence of NeuP with behavioral tests before treatment started.
Rats were the most prevalent species studied (85.2%) followed by mice (14.8%). Only 1 report with rabbits was included. Whereas 92.5% of studies included only male animals, 7.4 % of studies studied female animals. None of the studies included both sexes.

Qualitative Analysis
Supplementary

Exercise
Two types of exercises were investigated in the studies, swimming, and treadmill running.

Matesanz-Garc ıa et al The Journal of Pain 13
Swimming was one of the two activities studied by 4 out of 26 studies (15.4%). The dose for swimming exercise ranged from 40 to 60 minutes and was performed on 5 days per week. Swimming reduced the concentration of proinflammatory cytokines in the injured nerve tissue, 28 as well as the concentration of neurotrophins in spinal cord, dorsal root ganglia, and peripheral nerve tissue in the medium term. 29,30 Only 1 article found no post-treatment differences in BDFN concentrations. 31 One paper found an increase of GAP-43 in the peripheral nerve. 31 Treadmill aerobic training was the most used by the studies (23 out of 26 studies, 88.5%), both in isolation and using it against other therapies. The dose of treadmill running ranged from 60 minutes to exhaustion and was performed between 3 and 5 days per week over a period of 3 to 8 weeks. Treadmill running was able to reduce proinflammatory cytokines and increase antiinflammatory cytokines mainly in peripheral nerves, 32 −35 with changes in DRG and spinal cord also reported. 36−39,33,40,41 Only one article found increased proinflammatory cytokines in nerve and dorsal horn of the spinal cord. 39 Only 1 study found no difference in the sub-group "other inflammatory markers" of the immune system 42 The concentration of neurotrophins was lowered after treadmill exercise. 9,43,44,30 One study reported increased expression of at least one of these biomarkers when treadmill running was combined with electrical stimulation. 9 Treadmill running was also effective in reducing the activation of glial cells in DRG and spinal cord. 39,45,46,42,43 Only 1 article did not find changes in the spinal cord after intervention. 47 In that experiment, the animals ran until exhaustion, 47 while in the others it was of a fixed duration. 39,45,47,42,43 Studies reported a direct relationship between increased expression of inhibitory neurotransmitters, such as serotonin in the brain and spinal cord and exposure to treadmill running. 48,49,44 Only 1 study found a decrease in neurotrophin expression in the peripheral nerve. 32 In contrast, the effect on excitatory neurotransmitters was only evaluated in 2 articles, with mixed results, however different neurotransmitters were measured (GABA and Substance P). 50,51 Two articles reported a decline in the expression of inflammatory markers in the dorsal horn. 47,41

Neural Mobilization
Five articles studied neural mobilization. The most frequently reported dose was 20 oscillations per minute for 2 minutes and 25 seconds of rest, for 10 minutes for a total of 10 sessions. Only 1 showed no difference in posttreatment biomarkers of neuropathic pain. 52 Whereas Giardini et al 52 evaluated changes in the thalamus, midbrain and PAG, the other studies examined biomarkers in SCDH, DRG, and sciatic nerve. Neural mobilization consistently reduced the concentration of neurotrophic factors and the expression of substance P, TRPV1, and MOR 53,54 in the spinal cord. One article reported an increased concentration of NGF in the sciatic nerve. 55 Whereas most studies used the chronic constriction model, one used a diabetic neuropathy model 56 and reported a decrease in intraneural proinflammatory cytokines on the treated side.

Joint Mobilization
Two studies evaluated the effect of joint mobilization on biomarkers of neuropathic pain. The dose for joint mobilization ranged from 1 series of 10 repetitions to 3 minutes series with 30 seconds' rest. The frequency ranged from every 2 days to 5 consecutive days for a total of 12 to 15 days. Joint mobilization consistently reduced activation of the immune system (glial cells mainly) in the SCDH. 57 Their effect on cytokine expression revealed controversial results; while the concentration of cytokines in the DRG remained the same after treatment, only anti-inflammatory cytokines increased their expression in the spinal cord. 58 One of the 2 studies used rhythmic mobilization techniques 57 and the other high-speed manipulations. 58 The place of application was different as well as the dose, so the results must be interpreted with caution.

Physical Agents
Nineteen studies investigated a range of physical agents including laser, therapeutic ultrasound, and transcranial direct current stimulation. The dose for ultrasound most frequently reported was 1 MHz 0.5 to 1 w/cm 2 during 5 minutes.
Therapeutic ultrasound reduced the expression of substance P in both studies. 59,60 Further, a reduction of cytokines (tumor necrosis factor [TNF] and interleukin-6 [IL-6]) 59 and TRPV1 expression 60 was apparent at sciatic nerve and dorsal root ganglia respectively.
Of the 5 articles including laser therapy, only 1 measured the changes generated on enkephalines 61 with no changes after treatment. Three papers report a decrease of cytokine concentration. 62 Two studies investigated tDCS. tDCS increased TNF-a concentrations in the brain and spinal cord, whereas IL-1b and IL-10 only changed significantly in the spinal cord, with a decreasing concentration of both cytokines. 66 tDCS also reduced the activation of glial cells in spinal cord dorsal horn 67 and decreased BDNF concentrations both in the central nervous system and in blood serum. 68 Three studies reported on the effect of TENS therapy. TENS could not reduce proinflammatory cytokines (TNF-a) in the sciatic nerve, 69 in fact 1 study reported an increase in that biomarker. 70 However, TENS did reduce the concentration of proinflammatory cytokines in the spinal cord. 71 The glial activity in the spinal cord was reduced after the application of TENS, and the expression of opioid receptors increased in the same location. 71 Contradictory results were reported regarding the presence of excitatory neurotransmitters in the spinal cord. 72 The pulse electromagnetic field was consistent in modulating the cytokine concentrations, in both the spinal cord and the peripheral nerve tissue that caused the injury. 73,74 Electro-Acupuncture Electroacupuncture reduced the concentrations of proinflammatory cykines. The doses reported ranged from 1 to 2 mA, fluctuating between 2 and 100 Hz, 1.05 to 2.85 milli seconds for 30 minutes. Most of the changes seem to occur in the dorsal horn [75][76][77][78][79][80] although changes in the nerve, 81,82 blood, 83 and DRG 84 were also reported. In contrast, four articles did not find changes in cytokine concentrations following electroacupuncture. 81,83,85,76 The effect of electroacupuncture reported on neurotrophins has been mixed. Articles reported decreased concentrations of nerve growth factors (NGF and BDNF) in dorsal root ganglia and spinal cord dorsal horn 86,87,76,88 while others obtained significant increases in the same anatomical sites for NGF, 84 BDNF, 89 and GDNF. 90 These differences may be due to the starting times and duration of treatment. It seems that most of the articles that reported a decreased concentration 86,87,76,88 had a treatment duration greater or equal to 2 weeks. In contrast those that increased pain markers expression only treated the animals for 1 week. 89,84 Acupuncture The three acupuncture articles included were very heterogeneous. Wang et al 91 91 Chang et al started the intervention 24 days after surgery, during a period of 5 days. 93 The location of biomarker measurement were different; Wang et al measured cytokines in the blood meanwhile Tang et al measured in the sciatic nerve, Chang et al measured Cdc2 and P-vim in the sciatic nerve and DRG with no difference after treatment. 93 Tang performed the treatment for 20 minutes in contrast to the others two articles, that did the same 30-minute daily dose was applied, but the duration of treatment varied between 1 and 2 weeks.

Discussion
This systematic review summarizes the results of 85 studies that report the influence of different types of physiotherapy modalities on biomarkers of peripheral neuropathic pain in pre-clinical models. The 2 most studied interventions were electro-acupuncture and exercise, with neural mobilization, joint mobilization and physical agents being less commonly studied. The most frequently measured biomarker group was related to the neuro-immune system, specifically cytokines. The dorsal horn is the anatomical site where biomarkers were measured most frequently. Most studies, despite their heterogeneous nature, report significant postintervention changes of the biomarkers of neuropathic pain. Our findings indicate that physiotherapy interventions downregulate the expression of pronociceptive (eg. immune system or neurotrophins) markers and upregulate the expression of markers that dampen neuropathic pain (eg. opioid system). However, risk of bias was high in 97.5% of studies.
Our findings about the most common model is similar to previous reviews about preclinical models of NP were traumatic injury (78.9%) is the most commun. 94 Although neuropathic pain induced by chemotherapy 95 or diabetic painful neuropathy are growing problems, 96 the models of neuropathic pain induced by chemotherapy and diabetic neuropathy have not been used very often in preclinical physiotherapy studies (2.5% and 11.1%, respectively).

Effects of Physiotherapy
Exercise was one of the main interventions studied, specifically swimming and running (treadmill). It is well established that aerobic exercise induces analgesic effects in preclinical models. 97 Our results demonstrate that aerobic exercise has promising effects on biomarker modulation in neuropathic pain. There seems to be a consistent effect of aerobic exercise on the modulation of markers of neuro-inflammation in the peripheral and central nervous system. Other biomarkers, such as neurotrophins and neurotransmitters are also modulated by exercise. Of note, studies which did not demonstrate an effect on biomarkers used exercise duration of less than 40 minutes, 29,31 perhaps insufficient time to generate changes. In contrast, studies showing an effect on biomarkers included sessions with a duration between 60 and 90 minutes. 28,30 For treadmill running, only 1 article did not find changes after intervention. 46 In this experiment the animals ran until exhaustion, 46 while in the others it was of a fixed duration. 39,45,46,42,43 It could thus be speculated that reaching exhaustion may counteract the positive effects of physical activity in regulating glial cell activity.
Neural Mobilizations have shown efficacy in human trials of patients with referred leg or arm pain of neural origin, 98 however their exact mechanisms of action remain speculative. In line with findings in animal models, 54,56 neural mobilizations improve mechanical hyperalgesia in patients after neural mobilization intervention. 99 Our findings indicate that neural mobilizations may exert their beneficial effect through modulating neuroinflammation, opioid system, and neurotrophins. The ability of neural mobilization to disperse fluids has been reported with cadaveric models. 100 In patients, there is also some indication that neuroinflammation may be a target. Schmid et al reported a reduction of intreanueral edema after 1 week of neural mobilization in patients with carpal tunnel syndrome. 101 Although Joint mobilization techniques are often used, they seem to have only short term analgesic effects in humans. 102,103 In addition they are not usually used for neuropathic pain, but for nociceptive pain. 104, 105 Both preclinical studies included in our systematic review reported a decrease of mechanical hyperalgesia after the interventions. 57,58 Similarly, Krouwel et al reported an increase on the pain pressure thresholds in humans after a lumbar joint mobilization. 106,103 Interestingly, our data indicate that joint mobilization may exert their beneficial effects through modulation of glial cells and cytokines. However, only two articles were included, both using different techniques which make it difficult to draw firm conclusions.
Physical agents are often used clinically as analgesic treatments. However, their clinical benefit remains contradictory. For instance, a Cochrane review about the use of TENS in adults with neuropathic pain could not draw firm conclusions whether TENS is effective for pain control due to the very low quality of the evidence. 107 Another review from Akyuz et al conclude that physical modalities such as ultrasound or laser are not effective for the treatment of neuropathic pain when applied alone. 108 Our data suggest that physical agents mainly seems to modulate neuropathic pain through regulation of neuroinflammation, such as a downregulation of TNF and IL-1b which are associated with the maintenance of neuropathic pain after peripheral injury. 109 Nevertheless, physical agents could also modulate other biomarkers, for instance neurotrophins or neurotransmitters.
Electroacupuncture has shown some evidence in reducing pain in patients with osteoarthritis mediated by b-endorphins. 110 Human evidence for the effect of electroacupuncture on neuropathic pain remains controversial. Penza et al did not find pain improvements following electroacupuncture treatment in patients with neuropathic pain 111 whereas Galantino et al reported some improvement in patients with human immunodeficiency virus-related peripheral neuropathy. 112 In both reports the number of patients included was small, so these results remain preliminary. Our findings indicate that electroacupuncture may exert beneficial effects through modulating neuroinflammation, regulating neurotrophins and neurotransmitters as well as decreasing ATP and ion channels, such as TRPV1. 113-115, 85,76,116,84, 117, 79,118 Another possible mechanism is that this type of electrical stimulation may be activating the endogenous opioid system by the release of enkephalins and b-endorphins. 119 As we only identified three articles about acupuncture, it is difficult to hypothesize about its mechanisms of action. Preliminary data suggest that similar to electro-acupuncture this technique might modulate the activation of the neuro-immune system, 93,92,91 but further research is needed. In line with our preclinical findings, a Cochrane review about the use of acupuncture in humans with any type of neuropathic pain reports limited evidence. 120 Another review about acupuncture and its effect on pain could also not establish a clear relationship between the technique and the analgesics effects in humans. 121

Implications for Humans
The importance of specific biomarkers to maintain neuropathic pain is not only clear in preclinical models, 122 but also in humans. 123 Our findings suggest that Physiotherapy can modulate biomarkers related to neuropathic pain in preclinical models. Although the most studied biomarkers related to the immune system and neurotrophins, this review identified other targets, such as neurotransmitters or the opioid system. In recent years, several publications have reported the possible relationship between the presence of neuropathic pain and some of the reported biomarkers of humans. For instance, neuroinflammation is thought to play a crucial role in the generation and maintenance of neuropathic pain in preclinical models 124 Similarly, there is a growing body of evidence confirming the importance of neuroinflammation in neuropathic pain in humans. Inflammation in the pathophysiology of neuropathic pain 123 This is apparent both in patients with focal nerve injuries, 65 but also in patients with polyneuropathies. 125,126 As such, our findings indicate that physiotherapy can modulate biomarkers that are relevant in patients with neuropathic pain.
In addition to the neuroimmune system, other systems may influence the presence of NP. For example, neurotrophins have been implicated with neuropathic pain. For Instance, NGF acts as a pathogenic pain mediator 127 and also in humans, high levels of NGF have been associated with pain. 128 BDNF shows similar hyperalgesic effects and its presence in the dorsal root ganglia and the spinal cord correlate with neuropathic pain behaviour. 129 The dysfunction of the opioid system has been described in preclinical 130 and in humans with NP. 131 And other indirect measure from the opioid system is the conditioned pain modulation which is mediated by the endogenous opioid system. 132 This type of alteration has been reported in patients with different types of NP, such as complex regional pain syndrome 133 or carpal tunnel syndrome. 134 These 2 systems look like a promising target which required further investigation in human trials.
So far, pharmacological management has been the first line of treatment for NP in humans. Tricyclic antidepressants (eg, amitriptyline), and serotoninnoradrenaline reuptake inhibitors (eg, duloxetine) or anticonvulsants (eg, pregabalin) have been use as first line option. 4 Also opioids, like tramadol have been use to target the opioid system. 5 Even Combination therapy have been used in these kind of patients, for instance the use mixed of morphine and gabapentin provided better pain relief together but that gain was also modest. 135 Despite of this evidence, some trials have report controversial results 136,137 in addition of the concerns about side effects reported of long term used 138 advises on looking for new, safer treatment options.
Future targets to investigate are the endogenous cannabinoids, such as CB2 receptor which recently have been shown to increase hypersensitivity in models of neuropathic pain 74 and we have not found this to have been evaluated in physiotherapy studies.
Whereas the results of this study seem to suggest promising effects of biomarker modulation of physiotherapy interventions for peripheral neuropathic pain, these findings cannot be directly translated to understand the mechanism of these therapies in humans. Nevertheless, these findings can provide guidance on the type and design of future physiotherapy interventions in clinical trials.
One of the most recommended treatment option for the treatment of neuropathic pain, a part of pharmacology, is exercise. 18,19 In humans is well establish that the hypoalgesic effects are correlated with the intensity or the prescribed dose. [139][140][141] Only three articles analyzed in this review reported the intensity of the intervention. 37,38,39 The 3 reports used low intensity prescription and they reported changes in biomarkers concentrations in both, locally and remotely. This is intriguing since, in humans, has been reported central activation mechanisms only with high intensity. 141 Future research taking the intensity into account should be done.

Limitations
We have identified some limitations in our review. As we have not extracted the data from behavioral assessments, we cannot classify the interventions and the posterior analysis by the potential neuropathic pain mechanisms. Only studies written in English were included after the selection process. The heterogeneity of the measurement methods as well as the large number of different biomarkers analyzed challenges the interpretation. Of note, 92.5% of studies only included male rats. It is well established that pain behavior and underlying mechanisms differ according to sex, 142 thus limiting the generalizability of our findings. Importantly, risk of bias was high and reporting according to the ARRIVE guidelines was poor in the majority of studies. The inconsistent reporting of summary statistics prevented a meta-analysis. Poor reporting and methodological quality have been identified as major challenges in preclinical research including in the pain field. 143,144 With the recent publication of the ARRIVE guidelines, it is hoped that the quality of preclinical studies and their reporting will improve, thus facilitating future systematic reviews. 27

Conclusion
Our results suggest that exercises, electro-acupuncture, neural mobilization, and physical agents modulate biomarkers of neuropathic pain in preclinical models.
Only few studies were available for joint mobilization and acupuncture, thus preventing firm conclusions. Physiotherapy interventions seem to regulate the expression of a range of biomarkers particularly associated with the neuro-immune system, opioid system, neurotransmitters, neurotrophins, and receptors. The high risk of bias and poor reporting quality however prevents firm conclusions. Nevertheless, our findings may be used to inform the design of future human studies. Future preclinical studies need to follow higher standards of methodological quality and reporting to advance this promising field.