Recent advances in understanding chemotherapy-induced peripheral neuropathy

Chemotherapy-induced peripheral neuropathy (CIPN) is a common cause of pain and poor quality of life for those undergoing treatment for cancer and those surviving cancer. Many advances have been made in the pre-clinical science; despite this, these findings have not been translated into novel preventative measures and treatments for CIPN. This review aims to give an update on the pre-clinical science, preventative measures, assessment and treatment of CIPN.


Introduction
The last decade has heralded improvements in cancer survival 1 . However, persistent effects following the treatment of cancer can lead to pain and an impaired quality of life long after treatment has finished or cancer has been cured 2 . Chemotherapyinduced peripheral neuropathy (CIPN) is one of those effects that can lead to a continuing symptom burden after treatment 3 . CIPN is characterised by the classic "glove and stocking" distribution of symptoms. After chemotherapy, 68% of patients have painful neuropathy at 6 months, improving to 33% at 1 year 4 . Although different chemotherapies have variable characteristics, symptoms tend to be predominantly sensory. Sensory toxicity is the predominant feature as dorsal root ganglion (DRG), containing the sensory cell bodies, have a fenestrated endothelium that is more permeable than that found in the spinal cord, where the motor cell bodies lie. Sensory features are characterised by so-called "positive" and "negative" symptoms. "Negative" symptoms include numbness, loss of vibration sense, proprioception and deep tendon reflexes, whereas paraesthesia, dysaesthesia, and cold and mechanical hypersensitivity are referred to as "positive" symptoms.
The development of pain is also a common reason for dose reduction 4,5 , which may have implications for oncological outcome 6 . The situation is further complicated by the effect of "coasting", whereby the development of pain is delayed until after stopping the chemotherapy.
Despite advances in cancer treatment and survival, we still have much to learn about CIPN. It is important to recognise that CIPN is a heterogeneous population; it may be acute, such as the neuropathy commonly experienced with oxaliplatin, or chronic, lasting well beyond the end of treatment. Although there may be some overlap in features, it is likely that the underlying pathophysiology, clinical features and therefore its management differ substantially. Furthermore, not all CIPN is considered painful. This review will focus on the mechanisms but also deliberate on clinical features and treatment of chronic painful CIPN.

Animal models
Animal models of CIPN have increased understanding of the pathophysiology of CIPN, yet a recent meta-analysis highlights problems with the current models and may help deliver more robust and valid models 7 . For example, how do studies assessing short-term pain behaviours in animals without tumour burden model chronic CIPN? Pre-clinical studies often focus on the gain-of-function symptoms rather than the loss of function (for example, numbness) more common with chronic CIPN. Misrepresentation of the sexes is evident; 83% of animals used were male. Newer models have addressed some of these criticisms. Griffiths et al. describe a paclitaxel model of CIPN for 28 days with ethologically relevant behavioural tests that better mirror the clinical picture 8 . Non-human primate models may be more similar to the human condition 9 but, owing to ethical and pragmatic issues, are not a feasible alternative to rodents.

Mechanisms
The main classes of chemotherapeutics that cause neuropathy include the platinum-based anti-cancer therapies (oxaliplatin and cisplatin), vinca alkaloids (vincristine and vinblastine), taxanes (paclitaxel and docetaxel), proteasome inhibitors (bortezomib) and immunomodulatory drugs (thalidomide). These classes have differing anti-neoplastic mechanisms and likely different mechanisms for neuropathy. Evidence suggests that a number of mechanisms are shared between classes of chemotherapeutics, and most studies investigate the taxanes and the platinums. Currently, these mechanisms can be broadly separated into mitochondrial dysfunction and oxidative stress, microtubule disruption, neuroinflammation and immunological processes, and ion channel dysregulation.
Mitochondrial and oxidative stress Bioenergetic pathways, predominantly via the oxidation of glucose through the Krebs cycle within the mitochondria, are responsible for the generation of ATP. Chemotherapeutics commonly target nucleolar DNA and may also affect mitochondrial DNA. Indeed, targeting mitochondrial DNA as a principal therapy is an area of ongoing research 10 . Whereas nucleolar DNA has well-established repair mechanisms, mitochondria do not. Flatters and Bennett showed that paclitaxel treatment in rats led to swollen vacuolated mitochondria that followed the course of pain-like behaviours for almost 3 months 11 . Mitochondrial dysfunction within sensory neurones has also been demonstrated by other chemotherapeutics 12-15 . Krukowski et al. found that cisplatin-induced mechanical allodynia is associated with mitochondrial damage in DRG but that the loss of intra-epidermal nerve fibres (IENFs), seen in patients with CIPN, is related to bioenergetic deficits in peripheral nerves 16 . Gregg et al. found that post-mortem platinum concentrations in patients who received platinum chemotherapy were highest in DRGs and demonstrated a linear relationship between DRG levels and cumulative dose 17 , and levels were higher in patients with neuropathy. Animal data suggest a dose-dependent accumulation within the mitochondria of DRG neurones 18 . Recently, gene expression analysis further supported mitochondrial dysfunction in patients who develop CIPN. Kober et al. 19 found that breast cancer patients who develop neuropathy after paclitaxel demonstrate differential expression in a number of pathways implicated in mitochondrial dysfunction, including oxidative stress 20 . Additionally, genetic polymorphisms in anti-oxidant pathways have been associated with an increased incidence of CIPN 21 .
Numerous animal studies indicate that chemotherapy worsens oxidative stress 22,23 . Furthermore, anti-oxidants prevent the development of mitochondrial dysfunction, IENF loss and pain-like behaviours in animal models 24,25 .
The anti-oxidant alpha-lipoic acid reduces neuropathy in patients with diabetes and also animal models of CIPN 26 . Concurrent administration of alpha-lipoic acid reduces neuropathic symptoms secondary to bortezomib with less alteration to chemotherapy regimen secondary to adverse events 27 . However, despite the neuroprotective effects of anti-oxidants in vitro studies 28 , there is little clinical evidence for other nutraceutical anti-oxidants in the prevention of CIPN 29 . Recently, however, a phase I trial showed that calmangafodipir, a manganese superoxide dismutase mimic that aids reactive oxygen species (ROS) degradation, reduces acute and chronic CIPN after oxaliplatin in patients 30 without affecting response to chemotherapy and life expectancy. Metformin can also reduce neuropathic behaviours via a reduction in oxidative stress 13,31,32 . Metformin treatment in 40 patients receiving oxaliplatin reduced National Cancer Institute-Common Terminology Criteria for Adverse Events (NCI-CTCAE) grade 2 and 3 neuropathy with a moderate reduction in neurotoxicity score and a modest reduction in pain 33 .
Mitochondria play a key role not only in ROS regulation but in numerous other cellular processes, including calcium buffering, apoptosis and energy production via oxidative phosphorylation. Duggett et al. have shown that whereas basal respiration and ATP turnover were unaffected in DRG mitochondria of paclitaxel treated rats, maximal respiration and spare reserve capacity were greatly reduced at peak pain behaviour 12 . This indicates a reduced ability of these neurones to respond to stress, and the authors postulated that a switch to glycolysis could be an adaptive mechanism to reduce harmful ROS production.
Schwann cells play a crucial role in the regrowth of peripheral axons after injury; however, Nishida et al. found that accumulation of platinum compounds within Schwann cells was much lower than that in peripheral nerves and DRG 18 . Conversely, Imai et al. suggested that in vitro platinum compounds cause mitochondrial dysfunction in Schwann cells at drug concentrations lower than those required to induce neurotoxicity 34 , suggesting a greater role for mitochondrial dysfunction in Schwann cells in CIPN.
In animal models, treatment with pifithrin-μ, a molecule that suppresses mitochondrial damage, improves mitochondrial morphology, bioenergetics and IENF density while reducing pain behaviours 14,35 . Combined with evidence that it may act synergistically with the anti-cancer mechanisms of chemotherapeutics 35,36 , pifithrin-μ represents an exciting prospect in cancer care.

Glia and neuroinflammation
Glia are key in maintaining homeostasis and immunity in the central nervous system in both health and disease. In models of non-chemotherapy-induced neuropathy, microglia have been found to play an integral role in the development of the pain state 37,38 . Oxaliplatin-treated rats displayed persistent mechanical allodynia, sensory deficits and decreased density of IENFs 39 . Hu et al. showed a persistent activation of spinal cord microglia through strengthening of triggering receptor expressed on myeloid cells 2 (TREM2) signalling and demonstrated that either inhibiting microglia with minocycline or interrupting TREM2 signalling improved pain-like behaviours and IENF density 40 . Furthermore, an agonist at the CB2 cannabinoid receptor, colocalised with spinal microglia, inhibited microgliosis and pain behaviours in an animal model of paclitaxel-induced neuropathy 41 .
Despite these findings, astrogliosis rather than microgliosis is thought to be of greater importance to the development of CIPN 42,43 , while in some models, astrocyte inhibition with minocycline prevented the development of pain-like behaviours. But how would astrocyte activation lead to the development of CIPN? One proposed mechanism in a rat model of oxaliplatin-induced painful neuropathy is dysregulation of spinal adenosine kinase expression in astrocytes 44 . This may lead to activation of NRLP3/interleukin 1 beta (NRLP3/IL1β) pathway, promoting dorsal horn neuronal excitability with concurrent suppression of the anti-inflammatory IL-10 system, leading to central sensitisation and pain behaviours 44,45 . Importantly, restoration of adenosine signalling with an A3AR adenosine receptor agonist prevents the development of both astrocytosis and pain behaviours 45 . Another mechanism proposed in rodent models is through the alteration of sphingolipid signalling within astrocytes in the superficial layers of the dorsal horn of the spinal cord, an area concerned with nociceptive transmission 46,47 . Maladapted sphingolipid metabolism, through direct bortezomib effects and increased IL-1β, may increase glutamatergic transmission and consequently nociceptive transmission and pain behaviours 47 .
Reasons for the discrepancies in the role glia play in CIPN remain unclear but the discrepancies may be due to variations in chemotherapy, species, time point and sex studied. Pain phenotype differs greatly between male and female patients, and the pathophysiology in animal models is also sex-dependent 48 . In animal models of bortezomib-induced peripheral neuropathy, modulation of sphingolipid signalling attenuates pain behaviours in male but not female rodents 47 . Additional examples of sexual dimorphism are found in paclitaxel-induced peripheral neuropathy, and Toll-like receptor 9 (TLR9) expression in macrophages infiltrating DRG plays a role in the development of pathophysiological changes and behaviours in male mice but not females 49 . Macrophage infiltration into DRG and peripheral nerves has been seen in a number of animal models of CIPN, and as with other models of neuropathic pain, activation of TLR4 seems to be crucial 50-52 .
Clinically, minocycline treatment reduced only the acute pain syndrome associated with paclitaxel infusion but not the development of chronic CIPN 53 . Additionally, in another phase 2 trial, minocycline failed to prevent oxaliplatin-induced peripheral neuropathy 54 . Despite previous pre-clinical trials indicating minocycline's efficacy at inhibiting astrocyte activation and pain behaviours, its actions have been ascribed predominantly to inhibition of microglia and not astrocytes 55,56 . Given the differential role that microglia may have in CIPN, minocycline's lack of clinical efficacy may be of no surprise and neuroinflammation still represents a worthy area for continued research in the prevention of CIPN. Fingolimod, a drug used in the treatment of multiple sclerosis, downregulates the S1PR1 receptor found on astrocytes. Antagonism of this receptor has been shown to reverse immunochemical and behavioural changes in rodent models 47 . This presents the exciting prospect of a potentially new mechanistic target with a readily available therapeutic agent; however, additional trials are required to assess both its effects on CIPN and importantly tumour activity.

Ion channels
Pre-clinical studies have highlighted many chemotherapy-induced changes in ion channel expression, possibly driving behavioural changes in other neuropathic pain states 57 .
Changes in sodium channel expression and their sensitisation increase spontaneous neuronal firing and decrease activation threshold 58 , mechanisms possibly analogous to the allodynia, hyperalgesia and paroxysmal sensations of CIPN. In patients, sodium channel dysfunction is found in acute oxaliplatin toxicity 59 , and sodium channel polymorphisms may have a causal role in the development of acute and possibly chronic CIPN 60 . Furthermore, Na v 1.7 channel has been found to be similarly upregulated in nociceptive neurones in both a rat model and patients with chronic paclitaxel-induced peripheral neuropathy 61 . Although dysregulation of other sodium channels is seen in pre-clinical studies of CIPN 62 , the clinical efficacy of sodium channel blockers has been disappointing 63 .
Potassium channel dysregulation is present in animal models of CIPN 64 . Acutely, oxaliplatin leads to the down-regulation of potassium channels in animal models 62 , and Poupon et al. found that treatment with a riluzole (a potassium channel activator) prevents the development of persistent CIPN in mice 65 . A phase 2 randomised controlled trial (RCT) investigating the efficacy of riluzole in the prevention of CIPN is under way 66 . Transient receptor potential (TRP) channels are critical in temperature transduction. Oxaliplatin treatment leads to an increased expression of TRPA1, TRPV1 and TRPM8 in sensory neurones 67 . Interestingly, suppression of TREK-1 and TRAAK potassium channels (and an increase in pro-excitatory Na v 1.8 and HCN ion channels) is found on neurones expressing TRPM8, a receptor responsive to cold 62 . This may present a mechanism through which menthol provides symptomatic relief and oxaliplatin produces cold hypersensitivity acutely.
Although calcium channel modulation has shown promise in animal models of CIPN 68,69 , no direct calcium channel blockers are in clinical use for neuropathic pain. Cisplatin causes an increase in the calcium channel alpha-2-delta subunit, the target of gabapentinoids 70 , and both topical and systemic treatment with gabapentinoids have been found to be beneficial in rat models of CIPN 71,72 . Despite this, treatment with pregabalin for 3 days before and after each cycle of oxaliplatin failed to prevent CIPN in patients 73 .

Risk factors
There are many potential predictors in the development of CIPN, including patient-related factors, such as increased age, pre-existing neuropathy, smoking status, and impaired renal function, and chemotherapy-related factors, such as type of chemotherapy, cumulative chemotherapy dose, concurrent chemotherapy treatment, and duration of infusion 74-76 . Certain cancers may cause a subclinical neuropathy which may predispose patients to CIPN and worsen outcomes 77 .
Genetic markers have been implicated in chemotherapy-related toxicity, and a number of genome-wide association studies have looked at polymorphisms associated with CIPN. A number of polymorphisms have been identified, none of which (at present) has sufficient prognostic value to be of use in the clinical context 78 . Argyriou et al. 78 called for improved methodology and more standardised diagnostic and severity grading to better inform future studies.

Assessment of CIPN
Despite challenges in prevention and treatment, assessment for CIPN should occur before, during and after chemotherapy. Assessment should include (1) diagnosis (including possible differential diagnoses), (2) severity (including functional impairment) and (3) time course of symptoms and relationship to chemotherapy. (Table 1). Within the history, it is important to determine

Investigations
There has been a great deal of interest in phenotyping CIPN by using minimally invasive tools such as NCSs, quantitative sensory testing (QST) and IENF density. It seems sensible that underlying mechanisms may translate to differing patterns of neuronal loss and therefore differences in functional deficits, yet in practice this theory is not robust. Traditionally, CIPN has been characterised as a predominant sensory neuropathy effecting large myelinated fibre function, and nerve biopsies from patients with cisplatin-and paclitaxel-induced neuropathy show a loss of large fibres with axonal atrophy and secondary demyelination 95,96 .
Platinum chemotherapeutics cause neuronal cycle arrest within the DRG and therefore likely cause a neuronopathy (also referred to as ganglionopathy) and anterograde neuronal degeneration. On NCS, this would manifest as non-lengthdependent neuropathy affecting both the proximal and distal neurone. In contrast, chemotherapeutics interfering with mitochondrial or microtubule function impair axonal transportation giving a length-dependent axonal polyneuropathy, leading to a die back of intraepidermal nerve fibres. However, owing to poor correlation with clinical symptoms, NCSs cannot be routinely recommended. Furthermore, NCSs assess predominantly large-fibre function, missing small-fibre changes that may occur with painful CIPN.
Owing to its ability to assess large-and small-fibre types, QST may be of use in the phenotyping of neuropathic pain 97 and therefore has been proposed as a useful tool in CIPN 98 . In patients with paclitaxel-induced peripheral neuropathy, the reduction in light touch and vibration detection thresholds seen in hands and feet supports the mechanism of paclitaxel causing a distal neuropathy predominantly effecting the large, non-nociceptive neurones 99 . Additionally, some report that thermal detection thresholds and pinprick detection are minimally affected, indicating that small-fibre function is preserved in this group. This is in contrast to findings in patients with vincristine and bortezomib-induced neuropathy, with some studies reporting changes in pinprick perception and warm detection thresholds suggesting small, nociceptive fibre dysfunction in this group of patients 100,101 .
Pre-existing QST sensory deficits increase the risk of developing CIPN 102 ; in some cases, cancer itself may be responsible for QST changes 103 . Although the QST sensory profile may differ between agents 104,105 , QST profiles for painful and painless CIPN may be similar 106 and changes in QST may occur later than symptoms develop 107 . Furthermore, QST requires expertise and time and consequently is not commonly used in routine clinical practice for the evaluation of CIPN.
Skin punch biopsy can inform the diagnosis of small-fibre neuropathies. In CIPN, similar to other small-fibre neuropathies, IENF loss is observed 108 . Taking comparative distal thigh and distal leg punch biopsies can help differentiate between a length-dependent neuropathy or a neuronopathy; however, evaluating CIPN using IENF densities has been found to be unreliable; there is a large overlap between different chemotherapeutics, and results conflict with other assessment tools 109-111 . Furthermore, although punch biopsy can be repeated, it is time-consuming and invasive and IENF density has been found to be a poor correlate of pain 112 .
Other techniques for assessing neuropathy have yet to be fully validated. Nevertheless, simple bedside measures such as vibration sense, light touch and pinprick have good validity in the measurement of neuropathy 113 .

Prevention
Reducing regional perfusion (cryotherapy) may reduce CIPN; cooling gloves and stockings have been shown to reduce the risk of desquamation and nail changes associated with chemotherapy. Of the three published trials, only one showed benefit 114 . Owing to poorly tolerated treatment or a greaterthan-expected control group response, the other studies were negative 115-117 .
In 2014, the American Society of Clinical Oncology evaluated 42 studies while developing guidelines on the prevention of CIPN 63 . Owing to a lack of high-quality data, they were unable to make any recommendations and encouraged additional research. Topical treatments are an attractive option for the management of CIPN. A small non-randomised study of topical menthol in 52 patients showed improved BPI scores 126 , and combination therapy with baclofen, amitriptyline and ketamine showed an improvement on some of the EORTC QLQ-CIPN-20 measures 127 . Topical 8% capsaicin patch application following CIPN has been shown to improve continuous pain, neuropathic pain symptoms, and patient global impression of change 128 . This treatment has also been found to improve IENF density, suggesting underlying disease modification 128 .

Treatment
There is increasing enthusiasm for the use of cannabinoids in the treatment of many chronic pain states. Agonism at CB1 and CB2 receptors has shown analgesia in rodent models of CIPN 129-133 but these findings have not translated into evidence of clinical efficacy. One published pilot study of nabiximols (THC:CBD mix) in 15 patients with CIPN 134 showed no significant improvement in pain, but a 2-point decrease over placebo was seen in five patients classified as "responders" 134 .
Without specific evidence for CIPN, clinicians extrapolate treatments from other neuropathic pain states 135 . Interestingly, strong opioids have some of the best "numbers needed to treat" (NNTs) for neuropathic pain (NNT 4.3, 95% confidence interval 3.4-5.8) 135 . Some clinicians may advocate the use of opioids in CIPN, however with increasing survivorship amongst patients with cancer, the possible benefits of opioids should be continually weighed up against the risk of long-term opioid therapy 2 .

Non-pharmacological
Neuromodulation has shown promise in various neuropathic pain states 136 . A number of case reports indicate that neuromodulation may help refractory CIPN 137,138 , but RCT data are lacking.
A recent study found that the use of wireless transcutaneous electrical nerve stimulation significantly improved some measures of CIPN, including pain, numbness and tingling 139 . Furthermore, scrambler therapy (a novel transcutaneous neurostimulation technique) has been postulated as a potential treatment 140 but was no more effective than sham therapy in a recent RCT 141 .

Acupuncture
A Cochrane Review of the efficacy of acupuncture in the treatment of cancer pain showed insufficient evidence of its efficacy 142 . Since then, a number of trials of acupuncture in CIPN have demonstrated improvements in several domains 143-145 . A systematic review concluded that there was insufficient evidence to recommend acupuncture for the treatment of CIPN 146 , although low risk of harm and possible benefit may allow its pragmatic use in painful CIPN.

Physical therapy
Exercise has been shown to improve a number of facets that contribute to morbidity associated with CIPN, including balance and strength 147,148 , numbness, tingling, and hot and cold sensations 149 . One study found that, on analysis of the quality-of-life data, exercise had a moderate effect on pain in patients undergoing chemotherapy; however, this was not limited to CIPN 150 .

Psychological therapy
Psychological factors have been shown to play a role in both the initiation and maintenance of a number of chronic pain states 151 . The activity of duloxetine, via enhancing descending inhibitory pathways, suggests that alteration of mood may play a role. In favour of this viewpoint, a study of 111 patients who received treatment for breast cancer found that pre-existing anxiety and pre-therapy numbness were the only factors to predict CIPN eight months later 152 . Knoerl et al. found that an eight-week web-based cognitive behavioural programme led to modest improvements in worst pain with no differences in mean pain 153 . It was hypothesised that this would be due to improvements in fatigue, anxiety, sleep-related factors, or depression; however, a follow-up analysis was unable to substantiate these findings 154 .

Future directions
Pre-clinical studies have shown that antagonism of the sigma 1 receptor (present on mitochondrial endoplasmic reticulum) is able to reduce mitochondrial structural changes and pain behaviours that occur in CIPN. A phase II clinical trial found that sigma 1 antagonist treatment during FOLFOX chemotherapy diminished cold hypersensitivity, reduced the dropout rate and allowed a higher cumulative dose of oxaliplatin 155 . Although the long-term pain outcomes are not known, this highlights a pathway for potential therapeutics that could improve CIPN.

Summary
Despite an ever-expanding body of literature behind the pathophysiology and treatment of CIPN, new treatment options are still limited, and a proportion of patients continue to have difficulty controlling symptoms causing a significant impact on quality of life. Guided by the pre-clinical literature, novel targets that may help prevent CIPN are beginning to emerge. However, with continual advancements in chemotherapeutic agents with novel mechanisms, it is important that ongoing development of treatments for CIPN continue.