Emerging Biomarkers, Tools, and Treatments for Diabetic Polyneuropathy Free

Abstract

Diabetic neuropathy, with its major clinical sequels, notably neuropathic pain, foot ulcers, and autonomic dysfunction, is associated with substantial morbidity, increased risk of mortality, and reduced quality of life. Despite its major clinical impact, diabetic neuropathy remains underdiagnosed and undertreated. Moreover, the evidence supporting a benefit for causal treatment is weak at least in patients with type 2 diabetes, and current pharmacotherapy is largely limited to symptomatic treatment options. Thus, a better understanding of the underlying pathophysiology is mandatory for translation into new diagnostic and treatment approaches. Improved knowledge about pathogenic pathways implicated in the development of diabetic neuropathy could lead to novel diagnostic techniques that have the potential of improving the early detection of neuropathy in diabetes and prediabetes to eventually embark on new treatment strategies. In this review, we first provide an overview on the current clinical aspects and illustrate the pathogenetic concepts of (pre)diabetic neuropathy. We then describe the biomarkers emerging from these concepts and novel diagnostic tools and appraise their utility in the early detection and prediction of predominantly distal sensorimotor polyneuropathy. Finally, we discuss the evidence for and limitations of the current and novel therapy options with particular emphasis on lifestyle modification and pathogenesis-derived treatment approaches. Altogether, recent years have brought forth a multitude of emerging biomarkers reflecting different pathogenic pathways such as oxidative stress and inflammation and diagnostic tools for an early detection and prediction of (pre)diabetic neuropathy. Ultimately, these insights should culminate in improving our therapeutic armamentarium against this common and debilitating or even life-threatening condition.

The 21st century saw the further increase in the prevalence of diabetes and prediabetes (1–3). Diabetes leads to a serious increase in morbidity and mortality due to cardiovascular disease (CVD) (4–6) but also due to its microvascular complications (7, 8). Of note, diabetic foot complications are frequent in diabetes, difficult to heal, with a tendency to recur, sometimes presenting as emergencies, and generally they carry a sinister prognosis (9–11). In this context, prediabetes is receiving more attention, because it is now appreciated as a stage of already incipient vascular damage rather than just a condition of borderline increased glucose values (12). Both macrovascular and microvascular disease may have their beginning in prediabetes and may adversely affect morbidity already at this stage (12).

Classification and epidemiology of diabetic neuropathy

Diabetic neuropathy (DN) has been defined as a demonstrable disorder, either clinically evident or subclinical, that occurs in the setting of diabetes mellitus without other causes for peripheral neuropathy. It includes manifestations in the somatic and/or autonomic parts of the peripheral nervous system (PNS) that are being classified using clinical criteria (13). A recent classification differentiates between typical forms such as diffuse neuropathies including distal sensorimotor polyneuropathy (DSPN) and autonomic neuropathy as well as atypical forms, including mononeuropathy (mononeuritis multiplex) and radiculopathy or polyradiculopathy (14). DSPN represents the most relevant clinical manifestation affecting ~30% of diabetes patients, whereas the incidence of DSPN is ~2% per year (14, 15). DSPN has been defined as a symmetrical, length-dependent sensorimotor polyneuropathy attributable to metabolic and microvessel alterations as a result of chronic hyperglycemia exposure (diabetes) and cardiovascular risk covariates (16). Naturally, the type of neuropathy examined, the population or patient series included, and the diagnostic methodology used has an impact on the prevalence estimates of neuropathy (15). In hospitalized patients, the prevalence of DSPN ranges between 13% and 23% in type 1 diabetes (T1D) and between 18% and 75% in type 2 diabetes (T2D) (15). In primary care or general populations, the prevalence rates of DSPN range between 8% and 63% in T1D and between 13% and 51% in T2D (15). In newly diagnosed diabetes patients, the prevalence of DSPN ranges from 6% to 29% (15).

Typical positive sensory symptoms due to DSPN include pain, paresthesias, dysesthesias, and numbness, but up to 50% of diabetic peripheral neuropathies may be asymptomatic (14). Neuropathic pain has been defined as “pain caused by a lesion or disease of the somatosensory system” (17). Neuropathic pain, including burning, stabbing, lancinating, or shooting (electric shock–like) pain, may be spontaneous or evoked, as an increased response to a painful stimulus (hyperalgesia) or a painful response to a normally nonpainful stimulus (allodynia). The diagnosis of neuropathic pain requires a history of nervous system injury such as DN and a neuroanatomically plausible distribution of the pain. Usually pain is regarded as chronic when it lasts or recurs for >3 to 6 months (14, 18). There is no generally accepted definition of chronic painful DSPN and, hence, definitions and methodology were highly variable across studies. Chronic painful DSPN has been found in 13% to 26% of patients with diabetes in representative cohorts (15). The major risk factors for DSPN include age, poor glycemic control, diabetes duration, hypertension, dyslipidemia, smoking, and (in T1D) height (15). Frequent comorbid conditions are microvascular disease (nephropathy or retinopathy), macrovascular disease (peripheral arterial disease or CVD), and depression (15).

Emerging role for polyneuropathy in prediabetes

There has been growing appreciation that neuropathy may in some patients start very early, even during the stage of prediabetes (19–21). This knowledge has been based on evidence both from patient series in specialized clinics and from epidemiological population studies (20, 22, 23). In brief, 25% to 62% of patients with polyneuropathy of unknown cause may have prediabetes; DSPN may be found in 11% to 25% of persons with prediabetes and neuropathic pain in 13% to 21% of them (20). Likewise, cardiovascular autonomic neuropathy diagnosed by reduced heart rate variability appears to be more prevalent in persons with prediabetes compared with those with normal glucose tolerance (24). Therefore, an expert panel of the American Diabetes Association has recently stated that patients with prediabetes featuring symptoms of neuropathy should be appropriately screened for this complication (14). DSPN in prediabetes is, generally, less severe than in overt diabetes mellitus and mainly affects small fibers (20, 21), but pain may also be present (20, 21). Diagnosis rests on clinical examination as with DN, and the emphasis is on assessment of small fiber function (20, 25, 26).

The main etiological factor of DSPN in prediabetes is probably postprandial glucose excursions (27–29). This is why it is generally more common in subjects with impaired glucose tolerance than in those with impaired fasting glucose (20, 21). Other contributing factors include dyslipidemia, obesity, microvascular abnormalities, and the metabolic syndrome (20, 21). Treatment follows the same principles as in DN (20, 21). Accomplishment of normoglycemia is the cornerstone of management, although lifestyle modifications, management of dyslipidemia and hypertension, weight loss, and smoking cessation may also be helpful (20, 21).

Current concepts in the diagnosis and treatment of DN

DSPN is primarily diagnosed by clinical examination using simple evaluation tests (14, 30, 31). These tests are established, easy to use, inexpensive, and should generally be performed at least annually (14, 30). The examination should include assessment of both small and large nerve fiber function (2, 19, 31). In everyday practice, the former can be carried out by evaluation of pinprick (via sterile test strips) and qualitative temperature sensation (via hot/cold rods) (14, 30). For the latter, the main modality assessed is vibration sensation at the hallux (via a graded 128-Hz tuning fork), and this may be combined with examination of Achilles tendon reflexes (ankle jerks) (2, 19). History of symptoms (e.g., burning/lancinating pain, numb feet) may alert the clinician to the presence of DSPN (9, 14, 31).

Unfortunately, in a population-based survey about three-fourths of older subjects were unaware of having the condition (32). In a recent educational initiative conducted nationwide in Germany, painful and painless DSPN were previously undiagnosed in 57% and 82% of the participants with T2D, respectively, suggesting that attention to diabetic foot prevention practice is inadequate and should be fostered by implementing effective strategies to timely detect DSPN (33). According to the American Diabetes Association 2018 clinical practice guidelines (30), it is best also to include an annual examination of protective sensation by means of a 10-g Semmes Weinstein monofilament, given that inability to feel this precipitously increases the risk of future diabetic foot ulceration and even amputation. However, we should not fail to remember that the 10-g Semmes Weinstein monofilament only detects very severe sensory loss, and so it should not be the only sensory modality evaluated. Complementary tests of small and large fiber function have been described but never gained wide applicability (31).

Autonomic neuropathy is diagnosed by appropriate function testing of the corresponding organ manifestation (14, 30). Depending on the presence and/or nature of autonomic symptoms along with any particular clinical suspicion, the examination may include, for example, cardiovascular reflex tests, gastrointestinal tract motility, genitourinary function, and others (14, 30, 34, 35).

Therapeutic targets include normoglycemia, pain relief, and other symptomatic management (14, 30, 36). Normoglycemia or near normoglycemia is a foremost goal but cannot be achieved in many patients. Moreover, intensive diabetes therapy does not fully prevent the development of DN in T1D, while there is no clear evidence for benefit in T2D as discussed in detail in “Glycemic control” below (14, 30, 36). Multimodal management of neuropathic pain includes symptomatic pharmacotherapy and nonpharmacological options such as electrical stimulation, physical therapy, and psychological support. α₂δ Ligands such as pregabalin and serotonin and norepinephrine reuptake inhibitors such as duloxetine are the two classes of analgesic drugs recommended for the initial symptomatic treatment of neuropathic pain (14, 30), and their combination may also be used (36). Sodium channel blockers, capsaicin 8% patch, and opioids are further treatment options (14, 30, 36). However, the efficacy of each of these compounds is limited, because they provide substantial pain relief in no more than ~50% of the patients and may have significant side effects. A number of novel analgesic compounds are investigated in randomized clinical trials (RCTs), the details of which have recently been reviewed elsewhere (36, 37).

Postural hypotension, gastroparesis, and erectile dysfunction represent the major autonomic symptoms needing management: both pharmacological and nonpharmacological measures have been used, although it is of paramount importance to obtain a detailed pharmacological history and identify any aggravating drugs (e.g., anticholinergics, tricyclic antidepressants) (30).

Pathophysiology

Oxidative stress

Oxidative stress has been suggested to play a pivotal role in the pathogenesis of DN (38, 39). As reviewed in detail in “Systemic biomarkers” below, there is a wide range of mainly systemic biomarkers for oxidative stress. Moreover, there is a variety of small reactive signaling molecules (e.g., O₂⁻, nitric oxide [NO]⁻, NO₂⁻, HS⁻, CO). Of note, they may have opposing effects; for example, both promoting and preventing cell death, inflammation, or aging. Under physiological conditions, the formation of small reactive signaling molecules is balanced by their consumption by antioxidants. A decreased antioxidant protection, an increased production of small, reactive signaling molecules, or a failure of cellular mechanisms to repair oxidative damage can result in oxidative stress and cell death.

Reactive oxygen species (ROS) are linked to the development of microvascular pathology and neuropathy in experimental diabetes. Findings in streptozotocin-injected diabetic rats indicate that oxidative stress contributes to deficits in nerve conduction (40) and nerve blood flow (41). In the same animal model oxidative stress leads to injury in dorsal root ganglia (DRGs) neurons (42).

ROS-induced activation of nuclear poly(ADP-ribose) polymerase-1 (PARP1) is a fundamental mechanism in the pathogenesis of diabetes complications (43, 44) (Fig. 1). DNA damage resulting from free radicals and oxidants leads to an activation of PARP1, which functions as a DNA damage sensor and signaling molecule. In T2D patients, oxidative DNA damage was shown to be higher in patients with DSPN compared with those without DSPN and in control individuals (45). Experimental studies have demonstrated that PARP1 activation with resulting nicotinamide adenine dinucleotide/ATP depletion is implicated in motor and sensory nerve conduction deficits (46, 47) and small fiber neuropathy (SFN) (48). Accordingly, PARP1 inhibition has been shown to counteract diabetes-induced oxidative stress and improve large and small nerve fiber function in rats (49).

Reactive carbonyls, as precursors of advanced glycation end products (AGEs), have also been implicated in the pathogenesis of DN (50, 51). Methylglyoxal is one of the most potent dicarbonyls leading to the formation of AGEs, nonenzymatic protein, and DNA modifications formed under the influence of glycemic and oxidative stress (Fig. 1). A recent in vitro study has demonstrated that formation of methylglyoxal is associated with an increase in markers for oxidative stress, oxidative DNA damage, and PARP1 activation (52).

Oxidative stress and abnormal mitochondrial function are considered to play a primary role in the pathogenesis of neurodegenerative diseases (53). The overproduction of superoxide by the mitochondrial electron transport chain was proposed by Brownlee (54) as the central unifying mechanism contributing to diabetic microvascular complications. There is strong evidence that aberrant mitochondrial function is implicated in DN (55, 56). Superoxide anion (O₂⁻) is the major ROS produced in mitochondria (Fig. 1). Superoxide dismutase (SOD) enzymes convert O₂⁻ to hydrogen peroxide. Mitochondrial SOD (SOD2), the physiologically most important SOD isoform, is expressed exclusively in mitochondria. In contrast to the other two isoforms, the genetic deletion of SOD2 in neonatal mice is lethal (57). It has been shown that SOD2 prevents neuronal injury in cell culture and animal models of DN, possibly by modifying neuronal oxidative defense against hyperglycemia (58).

Collectively, experimental data suggest that reduced antioxidant defense along with increased ROS production contribute to the development of experimental DN substantiated by impaired endoneurial blood flow and hypoxia, nerve conduction deficits, SFN, and axonal atrophy.

Inflammation

Although DSPN is usually classified as a “noninflammatory” neuropathy in contrast to, for example, chronic inflammatory demyelinating polyradiculoneuropathy (59), there are multiple lines of evidence supporting an inflammatory component in its development (Fig. 1). The nervous and immune systems are tightly integrated in their sensing of external and internal danger signals and in their coordination of defense responses (60). However, the redundancy and pleiotropy of the immune system make it difficult to identify independent, complementary, or synergistic effects of specific mediators. The triggers of subclinical inflammation leading to DN are most likely largely overlapping with the proinflammatory risk factors leading to T2D and other diabetic complications. These include obesity, hyperglycemia, dyslipidemia, dietary factors, physical inactivity, smoking, psychosocial stress, and environmental risk factors such as air pollution (61). Several animal models of DN in T1D- or T2D-like conditions are available (62) to study which inflammatory processes may contribute on the one hand to the loss of myelinated and unmyelinated nerve fibers and on the other hand to damage of the blood–nerve barrier and the microvasculature (39, 63). Gene expression analyses of sciatic nerve biopsies of mouse models of DN demonstrated a profound dysregulation of inflammatory and immune-regulated pathways (64). This involved the activation of the Janus kinase–signal transducer and activator of transcription pathway in DN models of both T1D and T2D, which responds to many cytokines, including IL-6 and interferons. Interestingly, these experiments found notable differences in differential gene expression patterns between T1D and T2D (65). At the level of systemic proteins, IL-6 and the proinflammatory cytokines IL-1β and TNFα have consistently been linked with reduced nerve conduction velocity (NCV) and neuropathic pain in animal models of DN (66, 67).

The aforementioned immune mediators and various other cytokines have also been implicated in the progression from painless to painful DN (68, 69). Peripheral sensitization has been described in response to multiple immune mediators, including IL-1β, IL-6, TNFα, interferon γ, chemokines [C-C motif ligand (CCL)2, CX3CL1, C-X-C motif ligand (CXCL)1, CXCL5, CXCL12], and prostaglandin E₂ (60, 70–72). Binding of these immune mediators to their receptors activates signaling pathways, including the IκB kinase subunit β/nuclear factor κB (NF-κB), c-Jun N-terminal kinase, and p38 MAPK pathways, and stimulates the expression of the proinflammatory enzymes cyclooxygenase-2 (Cox-2) and inducible NO synthase (66, 67, 73, 74).

The relevance of inflammatory processes for DN is underscored by observations that genetic deletion of cytokines and pharmacological inhibition or neutralizing antibodies against the proinflammatory cytokines IL-1β (75, 76), IL-6 (77), and TNFα (78–80) result in beneficial effects on nerve conduction deficits and alleviation of pain. The finding from one of these studies that TNFα secreted from bone marrow–derived cells is essential for the development of DN in a mouse model highlights the crucial role of the immune system in this context (79). Genetic knockout or selective inhibition of cyclooxygenase-2 (Cox-2) prevents or delays the development of peripheral neuropathy in diabetic rodents (81) and thus points toward a role of prostaglandins in the pathophysiology of DN.

The studies focusing on proinflammatory mediators are complemented by data indicating DN-protective effects of the anti-inflammatory cytokines IL-4 and IL-10 (70). Moreover, the activation of the adenosine monophosphate protein kinase improved molecular changes in neuronal cells and allodynia that may partially be mediated by anti-inflammatory effects (82).

Among the aforementioned cytokines at least IL-6 appears to have a more complicated role in DSPN. On the one hand, IL-6 injection causes neuropathic pain (77), but on the other hand, treatment with IL-6 improved nerve function and/or morphology in diabetic rodent models (83, 84). IL-6 is a complex cytokine with proinflammatory and anti-inflammatory properties depending on the immunological context. Importantly, note that inflammation is not only involved in nerve degeneration, but also in regenerative processes (85, 86), and IL-6 may indeed have a dual role under different conditions that remain to be elucidated (84, 87).

Despite the advance in knowledge provided by short-lived animal models of DNs, it is important to highlight their limitations, which include differences in the time required to develop DN, different triggers, and thus differences in pathomechanisms between diabetic rodents and humans (68). Most studies assessing nerve structural injury are based on two models of T1D (streptozotocin-induced diabetic rats and BB/Wor rats), whereas only few studies used models of T2D. However, fixed and prepared nerve trunks from rodent models do not reflect fiber loss and demyelination seen in DN in man. Moreover, it will be important to identify common themes across multiple rodent models to rule out that strain-specific issues will be misinterpreted as basic pathologic findings (88). Finally, most studies on neuropathic pain were conducted in nondiabetic animals.

Studies in humans are indispensable to assess the relevance of the contribution of inflammatory processes to DSPN. Histological studies have long established the increased presence of immune cells such as macrophages, mast cells, CD4⁺ T cells, and CD8⁺ T cells in nerves of patients with DSPN (89–92). Multiple studies have demonstrated higher systemic levels of acute-phase proteins, cytokines, chemokines, and soluble adhesion molecules in patients with DSPN (reviewed in detail in “Systemic biomarkers” below; see also Table 1) (93–117). Most of the available studies are cross-sectional, but TNFα and IL-6 have emerged as predictors of incident DSPN, whereas IL-1 receptor antagonist (IL-1RA) and soluble intercellular adhesion molecule-1 (sICAM-1) were related to the progression of DSPN in a large prospective population-based study in older adults with a high proportion of individuals with impaired glucose regulation and T2D (107). Increased levels of inflammation-related biomarkers have also been reported in the cerebrospinal fluid in smaller studies. The macrophage activation marker soluble CD163 has been linked with impaired peripheral nerve function in T2D (118), and cerebrospinal fluid levels of the chemokines CXCL6, CXCL10, CCL8, CCL11, and CCL23 were found positively associated with neuropathic pain (119).

To appreciate the potential clinical relevance of inflammation for DN, note that most risk factors contributing to the multifactorial development of DSPN such as diabetes, impaired glucose regulation, aging, obesity, and dyslipidemia represent established triggers not only of oxidative stress, but also of subclinical inflammation (Fig. 1), so that inflammation may act as a “druggable” mediator between these risk factors and DN (39, 54, 63, 120). However, several important open questions remain to be addressed before pathogenesis-derived therapies targeting inflammation can be designed. Which are the most relevant stimuli of inflammation in DSPN? Which are the cellular sources of inflammatory mediators? To what extent is inflammation the primary driver or rather a mediating/modifying link between risk factors and disease?

Despite preliminary evidence that risk factors for DSPN may differ between diabetes types (121), only very few studies investigated potential differences between T1D and T2D (65, 106). Additionally, no data are available on potential differences between individuals with manifest diabetes compared with prediabetic stages, which would be of interest given the increasing prevalence of these conditions in aging populations.

Microvascular alterations

Structural, functional, and hemodynamic changes of microvessels in the neurovascular system are implicated in the development and progression of DN. Peripheral nerves are vascularized by endoneurial, perineurial, and epineurial vessels that form a network of anastomoses within the nerve and with the regional external vasculature. Various studies have demonstrated structural alterations of endoneurial vasculature in DN, namely proliferation of endothelial cells (ECs) and basement membrane thickening, leading to a reduction in endovascular luminal size and, ultimately, in connection with other vascular factors such as increased blood viscosity, to ischemic conditions in nerve fibers (122). Increased endoneurial capillary density in sural nerve biopsies was demonstrated to be associated with IGT and diabetes, decreased capillary luminal area with deterioration of glucose tolerance, and increased basement membrane area with DSPN (123). Sensory neurons in the DRGs might be particularly susceptible to microvascular alterations, because DRGs are characterized by a higher independently regulated blood flow with lower oxygen tensions compared with peripheral nerve trunks and possess a less efficient neurovascular barrier (124).

At present, the potential to gain further insight in the pathogenesis of DN from studying merely structural microvascular anomalies seems rather exhausted, although extensive efforts are taken to learn from functional and dynamic microvascular interplays. According to a recent hypothesis, tissue hypoxia in neurons and Schwann cells exacerbating oxidative stress and inflammatory processes may not be initially caused by an obvious reduction of endoneurial blood supply but rather by a dysregulation of the capillary flow, resulting in an increased heterogeneity of capillary blood transit time and hence an impaired extraction of diffusible molecules such as oxygen and glucose (125, 126). Evidence from human studies is needed to prove whether this tempting hypothesis is effectively involved in the development of human DN. A crucial link in the microvascular etiology of DN is endothelial dysfunction. ECs regulate arterial tone and blood flow by production of vasoactive substances such as NO, prostacyclin, and endothelium-derived hyperpolarizing factor that induce vasodilatation and vasoconstrictive factors such as endothelin-1 and angiotensin II. Secretion of vasodilators in ECs is stimulated by circulating factors that also promote opposite effects directly on vascular smooth muscles such that the mediation of blood flow by ECs has to be viewed as a homeostatic process. In microvessels, blood flow regulation is mainly driven by non–endothelium-dependent vasodilators, but endothelium-dependent NO-driven vasodilation is important in case of increased oxygen demand. Moreover, ECs express endothelial leukocyte adhesion molecules such as ICAM-1, vascular cell adhesion molecule-1 (VCAM-1), and endothelial leukocyte adhesion molecule-1 (E-selectin) that are involved in the process of inflammation. Additionally, prothrombotic as well as antithrombotic molecules are produced by ECs (127). Although the term microvascular endothelial dysfunction in the context of DN is often used to describe insufficient endothelium-dependent vasodilation of microvessels due to lower secretion of NO, altered homeostasis of EC-adhesive and thrombotic factors also contributes to the definition of microvascular endothelial dysfunction.

Several mechanisms are involved in the pathogenesis of DN, and the question of cause and consequence is seldom trivial. It has been demonstrated, however, that endothelial dysfunction alone is sufficient to induce neuropathy (128). In diabetes or even prediabetes, hyperglycemia affects neurovascular endothelial function via several pathways, leading to decreased bioavailability of both NO and prostacyclin, resulting in a preponderance of vasoconstriction impairing the nerve blood flow, and culminating in ischemic conditions damaging nerve fibers. Likewise, proinflammatory and prothrombotic endothelial factors prevail (54). Via the hexosamine pathway, hyperglycemia induces changes in gene expression (Fig. 1), leading to EC damage caused by increased levels of TGF-β1 and plasminogen activator inhibitor-1 (54). Hyperglycemia induces oxidative stress by an increased polyol pathway flux consuming reduced nicotinamide adenine dinucleotide phosphate, which is necessary to regenerate glutathione needed to decrease the development of ROS (Fig. 1). In ECs, ROS induce uncoupling of endothelial NO synthase, resulting in increased inflammation, impaired vasodilation, platelet aggregation, and thrombus formation (129). Secretion of proinflammatory factors is induced via increased gene expression of NF-κB (130).

AGEs are a heterogeneous group of substances that result from a nonenzymatic reaction between sugars and proteins, nucleic acids, or lipids. The common feature of AGEs is binding to the surface receptor of AGEs, which induces the production of ROS and therefore exposes ECs to the effects of oxidative stress and inflammation (131). Because the formation of AGEs is irreversible, they will persist even under rigorous glycemic control. Hence, their accumulation is part of the glycemic memory in DN (132) (see “Glycemic control” below)

“TNF α and IL-6 have emerged as predictors of incident DSPN.”

The increased gene expression in ECs activated by hyperglycemia (including endothelial NO synthase, NF-κB, and ICAM-1) may induce epigenetic changes in target cells without changes in the underlying DNA sequence (132, 133). Changes persist in mainly noncoding RNAs, which also regulate gene expression and therefore may have long-term effects in pathways related to endothelial dysfunction such as oxidative stress. Ultimately, as epigenetic changes are potentially reversible, it will be of interest as to whether they can be addressed in future therapies (133).

Nerve degeneration and regrowth

Diabetic conditions lead to a plethora of degenerative, remodeling, and regenerative processes in axons, glia cells, and the axon-surrounding microenvironment, ultimately leading to impaired nerve function. Although axon damage may represent the most salient feature of an injured PNS, the role of regeneration and an altered interplay between axons and nonaxonal components involved in the functional whole of neural plasticity has recently gained considerable attention. This section focuses on recent advances in understanding degenerative/regenerative processes in injured axons and surrounding tissues relevant to DN and neuropathic pain in the PNS.

In unmyelinated nerves, small-caliber axons are surrounded by a nonmyelinating Schwann cell forming Remak bundles, whereas myelinated fibers are wrapped individually by a myelinating Schwann cell forming myelin shafts separated only by unmyelinated nodes of Ranvier for fast saltatory axonal conduction. The major pathogenic pathways involved in the development of DN, such as the polyol, pentose phosphate, hexosamine, and protein kinase C pathways, as well as oxidative stress and inflammation, are activated under hyperglycemic conditions in neurons and independently also in Schwann cells, underlining their significance in the pathogenesis of nerve dysfunction in diabetes (126) (Fig. 1).

Early morphologic changes in peripheral nerves under diabetic conditions include signs of both degeneration and compensatory regeneration of unmyelinated nerve fibers characterized by a reduction of axon diameters and concomitant increased axon frequency. In myelinated nerve fibers, segmental demyelination and remyelination as well as abnormalities around the nodes of Ranvier can occur preceding axonal defects or electrophysiological abnormalities, suggesting early occurrence of Schwann cell impairment in DN (134). We previously demonstrated that both small and large nerve fibers appeared to be affected in parallel in the early course of T2D (135).

The process of nerve regeneration in diabetes between nerve injury and functional repair must overcome various obstacles. First, in axotomy-like injuries, axons and myelin sheaths distal to the injury site undergo Wallerian degeneration to allow for regrowth. This process involves activation of Schwann cells and macrophages and was shown to be delayed in animal models of diabetes. Elongation and repair of the remaining nerve fiber stump proximal to injury are fostered by components of the extracellular matrix (ECM) supporting the surrounded cellular components and modulating maintenance and repair mechanisms (136). Therefore, changes of the ECM considerably influence nerve regeneration (137). Matrix metalloproteinases (MMPs) are part of the ECM’s regulatory system and are activated by hyperglycemia, oxidative stress, and inflammatory cytokines. MMP-2 and MMP-9 in particular are stimulated following nerve injury (138). Whereas MMP-9 may play a role in Wallerian degeneration early after injury, MMP-2 upregulation seems to support axonal regrowth. In DRGs of diabetic rats, downregulation of MMP-2 was reported, suggesting an involvement of reduced MMP-2 in impaired axonal regeneration (136). Interestingly, inhibition of MMP-2 and MMP-9, both of which were upregulated after experimental nerve axotomy, attenuated allodynia (138). These findings support the rationale that regenerative processes may also contribute to the development of neuropathic pain.

The actual process of axon elongation within physiological neural plasticity or after nerve injury involves the expression of neurotrophin receptors and regeneration-associated proteins in axonal growth cones dependent on stimulation by the surrounding molecular and cellular components, including Schwann cells and ECM. It is therefore reasonable to suggest that alterations caused by diabetes in various compartments such as neurons, ECM, Schwann cells, and keratinocytes affect nerve regeneration. Based on the role of neurotrophins in axonal support and elongation, various studies examined their therapeutic potential to improve structural and functional features in animal models of diabetes. Although promising results were reported, translational studies and RCTs failed to demonstrate favorable effects in human DN (136).

The glucagon-like peptide-1 (GLP-1) receptor (139) is known for its high expression on pancreatic islet β-cells and is also widely expressed in other tissues, including axons and Schwann cells within the PNS, where it might possess neurotrophic properties. In experimental DN, GLP-1 agonists induced neurite outgrowth and attenuated neuropathic pain, but the role of GLP-1 in the pathogenesis of DN in humans apart from its glycemic effects remains unclear (124, 139).

Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is an intrinsic inhibitor of the phosphoinositide 3-kinase pathway and serves as a tumor-suppressor protein. Moreover, it is expressed in DRGs, axons, and Schwann cells (136). Several in vivo and in vitro studies have demonstrated increased PTEN expression in DRGs in animal models of diabetes. PTEN has inhibitory effects on axonal growth, and its inhibition or knockdown is followed by structural and functional recovery of nerve fibers (140, 141). Moreover, PTEN-deficient mice displayed protection from insulin resistance and T2D without malignant cell growth (142). Remarkably, lower PTEN expression was demonstrated to be associated with enhanced insulin sensitivity in humans, with PTEN gene mutations leading to PTEN insufficiency (143). However, there are no data available on the role of PTEN in human DN. So far, one can only speculate whether tissue-specific PTEN inhibitory substances may have the potential to emerge as novel treatments of diabetes (142) and its complications, including neuropathy.

The mammalian target of rapamycin (mTOR) signaling pathway is related to the aforementioned phosphoinositide 3-kinase pathway. Semaphorin 3A, an inhibitory axonal guidance molecule, is expressed via the mTOR signaling pathway in neurons, Schwann cells, and surrounding tissues, including keratinocytes. In skin diseases, it is involved in regulating epidermal innervation (144). Keratinocytes surround epidermal nerve fibers and have various functions in modulating epidermal innervation by producing nerve repellant as well as elongation factors, but under hyperglycemic conditions, the structure and function of keratinocytes may be altered. A recent translational study reported upregulated semaphorin 3A expression and mTOR signaling, accompanied by lower intraepidermal nerve fiber density (IENFD) in skin biopsies from diabetes patients and under hyperglycemic conditions in an animal model, where rapamycin treatment ameliorated neuropathic signs (144). Therefore, it is tempting to speculate that the mTOR pathway and overexpression of semaphorin 3A in keratinocytes is another piece of the puzzle in elucidating the mechanisms of epidermal nerve fiber loss in diabetes. Conversely, semaphorin 3A perfusion into regenerative neuroma after nerve injury reversed pain (145), showing a similar pattern for semaphorin 3A and MMPs with higher levels increasing nerve regeneration as well as neuropathic pain on the one hand and inhibition resulting in pain reduction but impaired regeneration on the other hand. Collectively, these findings underscore the difficulty of target reinnervation in the process of nerve regeneration after injury and the role of a maladaptive nerve regeneration in the development of neuropathic pain (146).

Emerging Biomarkers for Early Detection and Prediction

Genetic predisposition

Because the development and progression of DN in any single patient cannot be completely predicted by hyperglycemia or other traditional risk factors, the search for genetically determined factors remains of major interest (147). In this context, it has been emphasized that, in general, genetic studies should (1) select genes that encode proteins involved in the known mechanisms of nerve protection or damage in diabetes, (2) identify single-nucleotide polymorphisms (SNPs), (3) document that these gene polymorphisms affect the activity of the encoded proteins, and (4) search for associations between gene polymorphisms and DN. A number of studies have explored candidate genes potentially susceptible to DSPN, but apart from the few exceptions described below, most of these had a small sample size, thus weakening the ability to identify the contributory role of common alleles (147).

The angiotensin-converting enzyme (ACE) converts angiotensin I to angiotensin II, a potent vasoconstrictor, and inhibits the vasodilators bradykinin and kallidin. Consequently, it represents a key component in both the renin–angiotensin–aldosterone system and the kinin–kallikrein system. Angiotensin II mediates the activation of the AT₁ receptor, which not only affects several physiological functions, including vasoconstriction, but also increases the production of ROS and thereby contributes to endothelial dysfunction involved in the pathogenesis of DN (148). A recent systematic review with meta-analysis included five case-control studies from several countries with detailed genotype data aimed to detect associations between ACE gene polymorphisms and DSPN in diabetes. The meta-analysis demonstrated valid associations between the insertion or depletion polymorphism of the ACE gene and the development of DSPN (149).

Moreover, this meta-analysis study included four case-control studies of methylene-tetrahydrofolate reductase (MTHFR), a gene involved in the conversion of homocysteine to methionine. Elevated serum levels of homocysteine are rated among the risk factors of DN, but the precise mechanism is not fully understood. An MTHFR polymorphism with C to T point transformation at nucleotide 677 (MTHFR 677 C>T) was found to be associated with DSPN (149). It is therefore reasonable to assume that ACE insertion or depletion and MTHFR 677 C>T gene polymorphisms might increase the risk of developing DSPN in diabetes patients.

Shunting of glycolytic intermediates into the pentose phosphate pathway (PPP) has been suggested to protect from hyperglycemia-induced microvascular damage (Fig. 1). Transketolase, an enzyme involved in the PPP, is activated by its cofactor thiamine and diverts metabolic intermediates from glycolysis to the PPP. We genotyped nine SNPs of the transketolase gene in patients with recent-onset diabetes from the German Diabetes Study baseline cohort. After adjusting for multiple testing, associations of SNPs with neuropathic symptoms and reduced thermal detection thresholds were found that were primarily driven by male sex and the presence of T2D. These findings suggest a role of pathways metabolizing glycolytic intermediates in early DSPN (150). Numerous other usually smaller studies examined the potential association of DSPN with SNPs in T2D such as AKR1B1, ADRA2B, APOE, GLO1, GPX1, IL4, IL10, IFNG, MIR146A, MIR128-1, NOS1AP, NOS3, TLR4, UCP2, and vascular endothelial growth factor (VEGF).

In the PNS, voltage-gated sodium channels (VGSCs) are the backbone of an intact electric signaling cascade, allowing cells to depolarize via influx of sodium ions. In myelinated axons, VGSCs are primarily densely packed at the nodes of Ranvier, whereas in nonmyelinated axons they are distributed less tightly along the fibers (151–153). Nine α-subunit isoforms have been recognized so far (Na_v1.1 to Na_v1.9). Three of them might be particularly relevant in peripheral neuropathies due to their expression profile in the PNS (DRGs, peripheral axons, and/or Schwann cells). Na_v1.7 is encoded by SCN9A (sodium voltage-gated channel α subunit 9), Na_v1.8 by SCN10A, and Na_v1.9 by SCN11A (151). Several studies reported associations of sodium channel mutations with DN, painful neuropathies, and cardiac autonomic neuropathies. A recent study, however, failed to detect differences in VGSC gene variants between painful and painless DSPN, although many variants were seen in very few patients only (154). Several hyperfunctional variants in sodium channel Na_v1.7 (SCN9A) were associated with neuropathic pain in a cohort of nearly 1000 individuals with diabetes (155). A total of 12 rare Na_v1.7 variants were recently found in 9% of the participants with painful DN, whereas no rare variants could be identified in those with painless DN (156). Ongoing studies such as the European PROPANE project (151) will reveal whether VGSCs are a viable target for future treatment options.

Genome-wide association studies in painful DN showed a nearly significant genome-wide association for glial cell–derived neurotrophic factor family receptor α2, encoded by GFRA2, as well as for a wide region in women on chr1p35.1, gated by zinc-finger and SCAN domain–encoding ZSCAN20 on one end and Toll-like receptor 12 pseudogene TLR12P on the other end, and in men a high-mobility group box 1 pseudogene 46, HMGB1P46, on chr8p23.1 (157). In a recent review, the reported genetic contributors to neuropathic pain of various etiologies were surveyed and submitted for validation in a 150,000-participant sample of the UK Biobank cohort. Associations with a neuropathic pain construct were successfully replicated for two variants in IL10, pointing to a central role of neuroimmune interactions in the pathophysiology of neuropathic pain and supporting the hypothesis that timely treatment targeting the immune system could be helpful in mitigating neuropathic pain (157).

Overall, there is some support from SNP and genome-wide association studies for genetic susceptibility to DSPN, but definite information on genetic biomarkers for DSPN can only be obtained from large-scale studies.

Local biomarkers

Mitochondrial SOD2

Manganese SOD2 is considered one of the most important antioxidant enzymes in human cells. It is responsible for superoxide detoxification in mitochondria and has antiapoptotic properties against oxidative stress and inflammatory cytokines, both of which are implicated in the pathogenesis of DN (158, 159). Overexpression of subepidermal SOD2 was observed in participants with recent-onset T2D from the baseline cohort of the German Diabetes Study (Fig. 2) and was associated with increasing diabetes duration and sympathovagal balance. This points to a compensatory antioxidative defense mechanism leading to upregulated SOD2 in early T2D (159). Whether SOD2 overexpression has a predictive value in the development of DSPN has to be determined in prospective studies.

Langerhans cells

Exploring inflammatory processes in the skin might be another useful tool to gain knowledge about the role of inflammation in the development of DN. Langerhans cells (LCs) are dendritic cells of the epidermis maintaining immune homeostasis in the skin and can be identified using antibodies against CD207 (Langerin) (Fig. 2). In patients with T1D, cutaneous LC number in the forearm was markedly reduced immediately at diabetes onset, but not in those patients with known diabetes duration of 6 months (160). Another study reported that in T1D and T2D patients with a history of neuropathy, more LCs in the epidermis of foot ulcers correlated with healing outcome (161). We described a pronounced reduction of epidermal LCs, independent of IENFD, in patients with recent-onset T2D from the German Diabetes Study baseline cohort, suggesting a shift to a proinflammatory cutaneous environment at an early stage of T2D (162). In contrast, patients with neuropathic pain in the context of SFN, especially those who had diabetes, had a higher epidermal LC count, suggesting a role in the generation or maintenance of neuropathic pain (163). However, owing to the small samples studied, the latter results have to be interpreted with caution. Prospective studies should examine whether the early reduction of cutaneous LCs in T2D could promote a cutaneous immunogenic imbalance toward inflammation predisposing to polyneuropathy and foot ulcers (162).

CD31

The quantification of microvascular damage in diabetes could be helpful to better understand the role of microangiopathy in the pathogenesis of DN. CD31, also known as platelet EC adhesion molecule-1, is expressed on ECs, platelets, and several types of leukocytes. In skin biopsies, the CD31-positive EC area has recently been employed to detect subepidermal microvasculature. In participants from the German Diabetes Study with recent-onset T2D, the EC area was similar to that of healthy individuals (159). Using the capsaicin model to study vessel growth, no microvascular abnormalities in diabetes patients with and without DSPN were detected compared with controls, neither before nor after capsaicin-induced denervation (164). In contrast, delayed blood vessel growth rate was observed using CD31 in long-term diabetes patients with DSPN following experimental intracutaneous axotomy (165). The intracutaneous axotomy model seems to be an appropriate way to study vascular regeneration, because not only nerve fibers but also blood vessels are removed by the intervention. Overall, these findings suggest that cutaneous microvascular alterations are not detectable in the early course of diabetes, but they appear to become relevant with increasing diabetes duration.

Growth-associated protein 43

Growth-associated protein 43 (GAP-43) is used as a marker for regenerating nerve fibers due to its role in the process of nerve regeneration and its expression in peripheral nerve fiber areas with high neural plasticity, such as the epidermis and dermis (Fig. 2). It serves as a major constituent in axonal growth cones during axon elongation after nerve injury (166). Several studies used GAP-43 in comparison with protein gene product 9.5 (PGP9.5) as biomarkers to test the hypothesis whether GAP-43 may improve the early detection of DSPN and that the regenerative capacity of cutaneous nerve fibers may be reduced in painful or painless DSPN. The considerable variability of the results is at least partly due to the aforementioned findings. Additionally, it is difficult to rule out confirmation bias toward reduced GAP-43 against the panaxonal marker PGP9.5, because it is almost impossible to accomplish blinding of the markers in nonautomated morphometric approaches and as yet no study reported investigator blinding in this regard. A recent study found higher GAP-43– than PGP9.5-positive dermal nerve fiber length in skin biopsy sections stained using double immunofluorescence but no reduction in GAP-43–positive vs PGP9.5-positive nerve fibers (167). Likewise, more GAP-43–positive than PGP9.5-positive IENFs and higher GAP-43 expression were found in patients with various neuropathies compared with controls (168). Another study suggested that GAP-43–positive dermal nerve fibers could be detected earlier than intraepidermal PGP9.5-positive nerve fibers after capsaicin-induced skin denervation in healthy volunteers (169). These findings suggest that in areas with high neural plasticity, GAP-43 may not exclusively label regenerating nerve fibers, but it may be able to detect new axonal growth earlier than PGP9.5, which could be explained by faster axonal transport mechanisms operant for GAP-43. Therefore, GAP-43 may be used in areas with high neural plasticity, and the GAP-43/PGP9.5 ratio could be a more useful tool to assess regenerative capacity of cutaneous nerve fibers than PGP9.5 or GAP-43 alone (167).

Transient receptor potential vanilloid type 1

Transient receptor potential vanilloid type 1 (TRPV1), the capsaicin receptor, mediates burning or painful sensations when activated by heat, capsaicin, or acidic conditions and is expressed throughout the nociceptive pathway of the PNS, including epidermal nerve fibers. Because TRPV1 is involved in nociception and possibly in several types of hyperalgesia, there is considerable interest in the potential role of TRPV1 in painful neuropathies, and it could be a future target for analgesic treatment (170). However, when tested as a neuronal marker in both healthy individuals and patients with DSPN, TRPV1 and PGP9.5 showed no differences in nerve fiber densities. Therefore, TRPV1 may be used as a nonspecific neuronal skin biopsy marker rather than a specific nociceptive marker (170–172).

p75 Neurotrophin receptor

The p75 neurotrophin receptor (p75NTR) plays a fundamental role in synaptic plasticity and is expressed on axons, Schwann cells, and immune cells such as macrophages and microglia (173). After peripheral nerve injury p75NTR is upregulated in neurons and Schwann cells. In patients with diabetes, high p75NTR expression has been observed in myelin sheaths around axons susceptible to degeneration, raising interest in a potential involvement in the development of DSPN (126). In two studies (165, 174), p75NTR was used as a marker for Schwann cells, and differential patterns of p75NTR expression and Schwann cell migration were observed in the intracutaneous axotomy model and after nerve transection in patients with chronic neuropathies. Further studies are needed to shed more light on the role of p75NTR and Schwann cell involvement in the development of DSPN in diabetes.

Axonal swellings

The term “axonal swellings” is used to describe nerve fiber enlargements observed in IENFs or, less often, in dermal nerve fibers. Depending on the definition, axonal swellings constitute round structures >1.5 µm wide or more than twice the diameter of the associated nerve fiber. They have been described as a marker preceding nerve fiber degeneration (175, 176). The quantification of the axonal swelling/IENFD ratio was associated with decreased IENFD in painful neuropathies (177). A higher axonal swelling/IENFD ratio was found in painful compared with painless DSPN in diabetes patients (178). Using more selective neuronal markers, axonal swellings showed higher percentages of tropomyosin receptor kinase A and substance P in painful DSPN compared with painless DSPN and were positive for the nociceptor TRPV1 and GAP-43. In contrast, no correlation of neuropathic pain or clinical neuropathy severity was observed with axonal swellings in diabetes patients with DSPN (179). Using electron microscopy, accumulations of mitochondria, vesicular organelles, and neurofilaments in axonal swellings were described in patients with neuropathy of different origin (174), which likely represent predegenerative changes at sites of impaired axonal transport.

Experimental postinjury models

Capsaicin-induced denervation: the capsaicin model.

The capsaicin model, a standardized cutaneous nerve regeneration model, was designed to fulfill several criteria when studying epidermal nerve regeneration: short study duration, prospective in vivo data, high reproducibility (equivalent baseline conditions, reliable techniques, and valid parameters), good tolerability, and safety. Capsaicin, the active ingredient in hot chili peppers, possesses a neurotoxic potential through activation of TRPV1. During chronic activation of TRPV1, desensitization and degeneration of IENFs occurs, followed by regeneration after removal of the noxious agent. This process is thought to mimic a transient SFN (164) and involves several steps: (1) defining an area at the distal lateral thigh, (2) attaching an occlusive bandage containing capsaicin cream for 48 hours, (3) a baseline skin biopsy at the contralateral site, (4) first skin biopsy immediately after the 48-hour denervation period, and (5) nonoverlapping follow-up biopsies within the capsaicin application site in regular intervals up to 100 days (180). A reduced IENF regeneration rate was observed in individuals with diabetes compared with healthy individuals adjusted for baseline IENFD, which was associated with the regeneration rate (180). Likewise, in an 8-week clinical trial, baseline IENFD (in addition to height and race) was associated with the regeneration rate (181). Further studies are required to validate the capsaicin model as an endpoint to assess the rate of nerve regeneration following interventions in RCTs of DSPN.

Intracutaneous axotomy model.

Although capsaicin induces a chemical axotomy to cutaneous nerve fibers, other structures remain intact. Therefore, the capsaicin model was extended by a mechanical intracutaneous axotomy using a 3-mm biopsy punch after capsaicin-induced denervation to study the regeneration of cutaneous structures other than neurons, such as Schwann cells and blood vessels (165). After a predefined interval ranging from 1 to 3 months allowing for wound healing by granulation, a second biopsy of 4-mm diameter is performed right around the scar tissue of the initial smaller biopsy. Consequently, the obtained skin specimen contains two different areas: the inner area affected by both mechanical and chemical transection, and the outer ring analog to the capsaicin model solely affected by chemical denervation. This allows the assessment of blood vessel regrowth, Schwann cell growth, regenerative axonal sprouting (nerve fiber growth from deep to superficial layers), and collateral axonal sprouting (ingrowth from adjacent areas), all of which were reduced in diabetes patients with DSPN. These findings indicate that the local environment in patients with diabetes not only affects neural processes but also vascular repair processes (172).

Systemic biomarkers

The studies on systemic biomarkers for DN reflect our current knowledge of the principal pathomechanisms of the condition. Most of these studies have a cross-sectional design allowing for the characterization of biomarkers that may be involved in the pathophysiology underlying DN. However, prospective studies are required to identify genuine risk factors that may improve the prediction of disease onset over and above established clinical and other risk factors (182). As discussed in “Pathophysiology” above, key processes in the development of DN include oxidative stress, subclinical inflammation, microvascular alterations, and nerve degeneration and regrowth. This section provides an overview of available cross-sectional and prospective biomarker studies for DSPN with a focus on studies with at least moderate sample size (>100 participants) and with multivariable adjustment of the main analyses for relevant confounders (Table 1). We also considered smaller studies when their design was prospective. Because particular emphasis has to be put on the early stages of diabetes, the baseline cohort of the German Diabetes Study with diabetes duration of up to 1 year is of major interest in this context. The German Diabetes Study is a large prospective longitudinal cohort study describing the impact of subphenotypes on the course of diabetes aimed at identifying the prognostic factors and mechanisms underlying the development of related comorbidities (183).

An important methodological issue is the problem of confounding in observational studies. As already mentioned in “Inflammation” above, oxidative stress and subclinical inflammation are triggered by many factors that also have an impact on the risk of DN and most of which importantly represent modifiable risk factors of DN. Briefly, higher age, higher levels for body mass index (BMI), waist circumference, height, leg length, HbA1c, duration of diabetes, total cholesterol, low-density lipoprotein cholesterol, and triglycerides as well as hypertension, smoking, presence of the metabolic syndrome, or other diabetic complications (CVD, retinopathy, peripheral artery disease, renal dysfunction) have been found associated with a higher risk of DN or deterioration of peripheral nerve function in prospective studies in patients with T1D (184–187), patients with T2D (188–191), or in the general population (107, 192–194). Findings from cross-sectional or cohort studies reporting associations between biomarkers of oxidative stress or subclinical inflammation and DN should ideally be adjusted for as many of these variables as possible. Therefore, Table 1 does not only list key findings from observational studies, but it also gives a detailed account of confounders that were considered in the respective statistical analyses.

“Regenerative processes may also contribute to the development of neuropathic pain.”

Biomarkers of oxidative stress

Systemic oxidative stress can be assessed using a range of biomarkers in serum or plasma, although analytical issues such as the relative instability of many of these biomarkers limit their widespread clinical use (195, 196). Potential biomarkers include (1) sources of ROS, (2) ROS levels, (3) compounds that are modified by ROS, and (4) antioxidant enzymes mediating the response to ROS (196).

Myeloperoxidase (MPO), a heme peroxidase that catalyzes the production of HOCl from hydrogen peroxide and chloride ions, represents the most commonly measured pro-oxidant enzyme in the field of cardiometabolic diseases (195). MPO showed a robust positive association with DSPN after adjustment for multiple confounders in the German Cooperative Health Research in the Region of Augsburg (KORA) F4 study, a population-based study comprising a large proportion of individuals with prediabetes and T2D (197). In contrast, a second study observed lower MPO levels in DSPN patients with diabetes based on a minimally adjusted model (94), which appears counterintuitive. Thus, further data are needed to corroborate that increased systemic levels of ROS sources are related to DSPN.

Circulating levels of ROS are challenging to quantify because of their short half-life in serum or plasma. However, one study using indirect measurements observed that higher superoxide generation and lower peroxynitrite lag term were associated with a higher Neuropathy Impairment Score of the lower limbs (NIS-LL) (93). These data were supported by a prospective study indicating that increases in superoxide generation during 6 years were paralleled by decreases in median sensory NCV (SNCV) after multiple adjustments (96).

The interaction between ROS and proteins, lipids, and other molecules leads to compounds such as protein carbonyls and AGEs, the systemic levels of which were increased in diabetes and its complications (195). However, no association of methylglyoxal with quantitative tests and neuropathy questionnaires was found in a large cross-sectional study in patients with T2D (95), whereas higher levels of methylglyoxal were associated with the risk of incident DSPN in an age- and sex-adjusted model in the same cohort (189).

Among the antioxidant enzymes, systemic levels of the extracellular SOD3 were reduced in patients with DSPN in both participants with recently diagnosed T1D and T2D from the baseline cohort of the German Diabetes Study. Additionally, positive associations between SOD3 and several measures of NCV were robust to adjustment for confounders (97). In contrast, higher SOD3 levels were linked with higher risk of DSPN in the KORA F4/FF4 cohort (197). Thus, antioxidative defense may be linked with DSPN, but differences between study populations need further investigation.

Several vitamins have antioxidant effects, but a detailed overview of individual studies on vitamin levels and DSPN are beyond scope of this review. Meta-analyses provided evidence that systemic levels of vitamin B12 (198), folate (198), and vitamin D (199–201) may be lower in diabetes patients with DSPN compared with those without DSPN. However, the pooled effect estimates from these meta-analyses were mainly based on cross-sectional or case-control studies showing a high degree of heterogeneity and lack of adjustment for confounders. The Italian InCHIANTI Study investigated the potential link between vitamin E and NCV in a large population-based sample and reported a positive association between α-tocopherol and NCV (adjusted for multiple confounders) (99).

Taken together, despite the biological plausibility of the role of oxidative stress in the development of DSPN in diabetes, data from relevant biomarker studies with at least moderate sample sizes and minimal levels of adjustment are scarce (Table 1). Additionally, only three cohort studies demonstrated a link between biomarkers of oxidative stress and a reduction in NCV (96) or incident DSPN (190, 197). Thus, there is a need of large-scale prospective studies to identify oxidative stress biomarkers linked to the incidence and/or progression of DSPN.

Biomarkers of subclinical inflammation

Although there is no clear definition of subclinical or low-grade inflammation, as it is commonly seen in individuals with advanced age and/or increased cardiometabolic risk, this condition includes higher white blood cell counts and higher serum or plasma levels of inflammation-related biomarkers compared with younger, healthy individuals.

The KORA F4/FF4 cohort represents the cohort with the most comprehensive phenotyping with respect to biomarkers of subclinical inflammation and DSPN (Table 1). Importantly, most substudies within the KORA cohort were based on a population-based sample aged 61 to 82 years comprising individuals without diabetes, with prediabetes, and with manifest T2D. A joint analysis of these subgroups was performed because neuropathic symptoms and signs were already present in a considerable proportion of the nondiabetic and prediabetic study sample (102). Cross-sectional analyses demonstrated that higher levels of the proinflammatory cytokine IL-6 and the anti-inflammatory IL-1RA and lower omentin levels were associated with DSPN and/or a higher Michigan Neuropathy Screening Instrument (MNSI) score (102, 104). Higher IL-6 levels were also related to neuropathic pain (105). Further cross-sectional studies from Europe, the United States, and Asia have reported positive associations between acute-phase proteins (CRP, fibrinogen), proinflammatory cytokines (IL-6, TNFα), soluble cytokine receptors (soluble IL-6 receptor, soluble TNF receptor 1, soluble TNF receptor 2), adipokines (adiponectin, leptin), other inflammation-related proteins (osteoprotegerin), and white blood cell counts (neutrophil-to-lymphocyte ratio) with higher odds of DSPN or reduced NCV in population-based samples and in patients with T1D or T2D (94, 98–101, 103, 108, 109). Notably, only one study directly compared such associations between T1D and T2D. In patients with recently diagnosed T1D from the German Diabetes Study, higher total and high–molecular weight adiponectin were associated with faster NCV. In contrast, patients with recently diagnosed T2D showed inverse associations of these biomarkers with NCV and positive associations of IL-6 and total and high–molecular weight adiponectin with DSPN (106). Thus, associations between biomarkers of inflammation and DSPN and NCV may differ between both diabetes types.

“Variants in sodium channel Na_v1.7 were associated with neuropathic pain.”

Evidence from prospective studies is limited to the population-based KORA F4/FF4 cohort, in which higher levels of CRP, IL-6, TNFα, and IL-1RA and lower adiponectin levels were related to incident DSPN after adjustment for age and sex. IL-6 and TNFα remained associated with the incidence of DSPN in the fully adjusted model. Adding both cytokines to a clinical risk model containing multiple known DSPN risk factors improved model fit and reclassification. Higher IL-1RA levels were associated with the progression of DSPN (107). An extension of the study comprising a total of 71 biomarkers of inflammation found that higher levels of 26 biomarkers were associated with incident DSPN at P < 0.05, and 6 biomarkers remained significant after adjustment for multiple testing (three chemokines: MCP-3/CCL7, MIG/CXCL9, and IFN-γ–inducible protein 10/CXCL10; three soluble forms of transmembrane receptors: DNER, CD40, and TNFRSF9). When added to a clinical risk model, the C-statistic improved from 0.748 to 0.783 (P = 0.011). Of note, pathway analyses suggested a complex crosstalk between innate and adaptive immunity in the development of DSPN (110).

In conclusion, proinflammatory biomarkers are associated with DSPN and lower NCV in multiple cohorts (Table 1). The association of IL-1RA with DSPN and its progression is intriguing, as it indicates that an upregulation of anti-inflammatory biomarkers in response to metabolic and/or immunological stimuli may also be linked with the development of DSPN. Such an association has previously been linked to incident T2D and cardiovascular events as outcomes (202, 203). However, most data are derived from cross-sectional analyses, and direct comparisons of study samples with T1D and T2D are scarce. Therefore, there is an unmet need for prospective studies involving both patients with T1D and T2D as well as high-risk individuals from the general population to improve our understanding of inflammatory mechanisms in the development of DSPN and its progression, including the onset of neuropathic pain.

Biomarkers of microvascular alterations

Systemic biomarkers that specifically indicate microvascular alterations in different endothelial beds are currently not available. However, circulating levels of soluble cell adhesion molecules are commonly seen to reflect endothelial activation and vascular inflammation and are thus of interest in the context of DSPN.

ICAM-1 is an endothelial- and leukocyte-associated transmembrane protein involved in cell–cell interactions and leukocyte endothelial transmigration. Systemic levels of its sICAM-1 were associated with the presence of DSPN (94, 112), higher Neuropathy Disability Score (NDS) or MNSI score (102), reduced NCV (106), and neuropathic pain (105) in patients with diabetes and population-based samples, although this association was partially explained by confounders in some cases (102, 106). In line with these data, one study also reported positive associations between soluble VCAM-1 and soluble E-selectin with DSPN (94).

These cross-sectional results were corroborated by two prospective studies. Higher ICAM-1 and E-selectin levels at baseline were associated with a stronger decline in peroneal NCV during 5 years in 28 patients with T1D and T2D (111). In the KORA F4/FF4 cohort, higher baseline levels of sICAM-1 were associated with incident DSPN during 6.5 years in the age- and sex-adjusted analysis and with progression of DSPN (i.e., increase in MNSI in patients with DSPN at baseline) in the fully adjusted model (107). Thus, the aforementioned studies indicate that endothelial activation and vascular inflammation may be independently associated with different aspects of DSPN, but the evidence is mainly limited to sICAM-1 as a biomarker (Table 1).

Biomarkers of nerve degeneration and regrowth

Despite the interest in neurotrophic factors as potential therapeutics of DSPN (39), studies on systemic levels of biomarkers of nerve degeneration and regrowth in the context of DSPN are extremely limited (Table 1). Only one study measured nerve growth factor (NGF) in the serum of patients with T2D and reported a positive association between NGF levels and DSPN, but at the same time an inverse association of NGF with NDS (113). Therefore, it is currently unclear to what extent systemic levels of such biomarkers may be relevant for the pathogenesis or prediction of DSPN.

Other biomarkers for DSPN

On the basis of different pathophysiological considerations, additional biomarkers have been related to DSPN. Cross-sectional studies indicated that other growth factors such as EGF, platelet-derived growth factor AB/BB and VEGF (94, 112), neuron-specific enolase (116), and cystatin C (117) may all be positively associated with the presence of DSPN (Table 1). In contrast, results for heat shock protein 27 were conflicting (114, 115). Despite the novelty of these results, their relevance cannot be assessed in the absence of replication studies and without prospective analyses in well-defined cohorts.

Systemic biomarkers: summary

In summary, multiple associations between systemic levels of biomarkers of oxidative stress, inflammation, and vascular alterations with DSPN have been reported, whereas data on neurotrophic factors or biomarkers related to other potentially relevant pathomechanisms are scarce or missing (Table 1). The current evidence is almost exclusively restricted to cross-sectional studies, and only few prospective studies sought to identify biomarkers that are linked with the incidence or progression of DSPN. Moreover, hardly any data on painful DSPN compared with painless DSPN are available. The studies listed in Table 1 were performed in heterogeneous study populations, which limit the comparability of findings given that pathomechanisms may differ between individuals with prediabetes, T2D, and T1D. Also, the sometimes small sample sizes and the heterogeneity of neuropathy outcomes complicate a synthesis of results. It is striking that in contrast to other diabetic complications studies based on hypothesis-free approaches using novel omics technologies such as transcriptomics, proteomics, and metabolomics are still missing for DN. These studies can be expected to yield valuable insights into the mechanisms underlying the development of DSPN and novel hypotheses to be tested in experimental and clinical studies.

Novel Tools for Early Detection and Prediction

Skin biopsy

The gold standard (PGP9.5)

The current gold standard for the diagnosis of small-fiber damage in neuropathies including DN is the assessment of intraepidermal nerve fibers from a small skin specimen (3-mm punch biopsies) taken from the distal calf and subsequent immunohistochemical (bright-field) or immunofluorescent nerve fiber staining with PGP9.5, a member of the ubiquitin hydroxylase system that is widely accepted as a nonspecific panaxonal marker (Fig. 2). Normative reference data for PGP9.5-positive IENFD is available for both bright-field and immunofluorescence techniques, which showed a high level of agreement and comparable results. Manual nerve fiber counting, manual morphometric analyses of digitized slide sections, as well as semiautomatic assessment can be performed and all three options are validated and feasible tools for the evaluation of cutaneous nerve fiber counts, lengths, and densities (204). Although the assessment of IENFD in clinical practice is mainly used (or limited) to diagnose conditions characterized by small–nerve fiber involvement, the current capability of skin biopsies in experimental settings is more widespread, including, among others, their potential role as clinical trial outcome measures.

Evidence has emerged suggesting that IENFD is reduced as early as within the first 12 months after diagnosis of T2D (135, 167), and there is a wide consensus that IENF loss is a hallmark of manifest SFN (205, 206). Several authors tried to shed light on the question as to whether dermal nerve fibers are also reduced in DSPN (167, 207, 208) and whether painful DSPN is characterized by a predominant small-fiber involvement represented by a more severe cutaneous nerve fiber loss as opposed to painless DSPN, but the current evidence seems mixed and study results are difficult to interpret due to small sample sizes, inclusion of groups including neuropathies of origins other than diabetes, use of analgesic drugs that may impede a dichotomous DSPN classification on the basis of neuropathic pain (167), and the ongoing question of whether both DSPN entities generally show similar severity to match groups accordingly. However, recent studies focusing on DN reported no differences in PGP9.5 IENFD between painful and painless DSPN (167, 178, 209). Whether morphometric measures represent a morphological correlate of nerve function tests is difficult to answer, because quantitative changes per se do not assess the functional integrity of remaining nerve fibers which could be normofunctional, hypofunctional, or hyperfunctional (176). Nevertheless, correlations were found between IENFD and both small- and large-fiber quantitative sensory testing (QST) results in a large cohort of diabetes patients with DSPN (209).

Only a few smaller studies aimed to explore the temporal sequence of cutaneous nerve fiber loss in SFN and DSPN. A decrease in IENFD and dermal nerve fiber immunoreactivity was observed in a cohort of 29 patients with T1D or T2D and DSPN after 6 months, whereas clinical symptoms, clinical scores, and several sensory tests remained unchanged, suggesting that IENFs and dermal nerve fibers could emerge as useful indicators of DSPN progression (172). Another study found a rate of IENFD decrease several times higher in patients with T2D of short duration compared with controls and suggested a relative rather than linear rate of IENF loss when taking the baseline IENFD into account (210). In a study using skin biopsies from three sites along the leg from patients with SFN due to IGT or diabetes found similar rates of IENFD decrease after ≥2 years in proximal and distal sites, challenging the assumption of length dependence of axonal degeneration (211). However, this report should be interpreted with caution, because sample size was small and mean IENFD was surprisingly high. Moreover, IENFD and nerve conduction studies (NCS) suggest an exceptional susceptibility to progression in this cohort or some unrecognized bias (212). Altogether, there is evidence that monitoring of IENFD could be useful to determine DSPN progression in diabetes patients, but the validity of existing prospective studies is constrained by small cohort sizes and/or inclusion of SFNs of other origins.

“GAP-43/PGP9.5 ratio could be a useful tool to assess regenerative capacity of cutaneous nerve fibers.”

Novel statistical models for IENF analysis

Seeking to augment the diagnostic potential of IENF studies, novel statistical models have been developed taking into account the spatial structure patterns and tree-like branching characteristics of IENFs. It has been suggested that the spatial distribution of nerve fibers becomes more “clustered” as the neuropathy advances. Estimated second-order properties of the observed data patterns have been employed to describe the degree and scale of fiber clustering (213). More recently, a model was developed including nerve fiber base point (entry into epidermis) and endpoint (termination of nerve fibers) analyses as well as distances between point clusters and epidermal area covered by IENFs (214), and hierarchical models with nerve branching analyses for different segments and levels of branching were added (215). Although these statistical models for IENFs are still at an experimental stage, the direction toward advanced algorithm-based analysis is promising. The future will show whether new concepts based on task-specific algorithms or deep learning neural networks will help to improve the detection of DSPN.

Corneal confocal microscopy

Corneal confocal microscopy (CCM) is a relatively new modality for the morphometric diagnosis of corneal SFN (216, 217). The cornea of the human eye is richly innervated by nerve fibers from the trigeminal nerve, and these fibers are evaluated by CCM (135, 216, 217). Primarily, CCM has been used to assess the subbasal nerve plexus underneath the basal epithelium of the cornea (216, 217). CCM offers two advantages: it assesses small fibers, and it is noninvasive. The technique is sophisticated but not difficult to use with experienced personnel (135, 216, 217). CCM employs a light beam, which is focused by an objective lens into the examined cornea layer while all light coming from other points is appropriately eliminated (216, 217). Depending on the precise technology used, one distinguishes between the tandem scanning CCM, slit scanning CCM, and laser scanning CCM, which offers a higher resolution (216–218). The accuracy and utility of the latter have also been improved by automated software, real-time images, three-dimensional reconstruction, and spatial analysis (135, 216, 217, 219–221).

The applications of CCM have been identified and reviewed elsewhere (217). We will follow a similar, simplified classification in less detail. In the study of corneal nerve morphology, the main parameters used are corneal nerve fiber density (CNFD) and corneal nerve fiber length (CNFL). The former is defined as the total number of major nerves per mm², and the latter is defined as the total length of all nerve fibers and branches (mm/mm²) (216, 217). Of note, these are well reproducible (222), although manual CNFD and automated CNFL have been identified as most accurate as well (223). Three further important parameters include corneal nerve branch density (CNBD), defined as the number of branches emanating from major nerves per mm²; corneal nerve fiber tortuosity (CNFTo), calculated as the variability of nerve fiber directions; and tortuosity-standardized CNFL (217).

Detection of DSPN

The aforementioned CCM parameters show impairments early in the course of DSPN, and these impairments become more pronounced with increasing DSPN severity (224–227). For the diagnosis of DSPN, CNFD (threshold <27.8/mm²) yielded 82% sensitivity and 52% specificity (227). For the diagnosis of at-risk foot, CNFD (threshold <20.8/mm²) yielded 71% sensitivity and a 64% specificity (227). A further work (228) looked at the diagnostic performance of CNFL in more detail and found that CNFL ≤14.0 mm/mm² had 85% sensitivity and 84% specificity, and CNFL ≥15.8 mm/mm² had 91% sensitivity and 93% specificity.

Automated corneal nerve fiber quantification is rendering the technique easier and faster, without compromising accuracy (229, 230). For the diagnosis of DSPN, the area under the receiver operating characteristic (ROC) curve was 0.82 for manual CNFD, 0.80 for automated CNFD: this compared with 0.66 for IENFD (229). In a very recent work in T1D, the sensitivity and specificity of CNFD were 77% and 79%, respectively, and the area under the ROC curve was 0.81 (231). The corresponding values for IENFD were 61%, 80%, and 0.73 (231). A recent innovation is the examination of the inferior whorl, a more distal area that is inferior and nasal to the central cornea (232). Inferior whorl length yielded an area under the ROC curve similar to that of CNFD and CNFL (0.70 to 0.75), whereas the combination of IWL plus CNFD slightly increased the area under the ROC curve to 0.76 (232).

Severity of DSPN

Especially CNFD, CNBD, and CNFL exhibit a progressive reduction as neuropathic deficits get worse (224–227, 233–235). Alternatively, these parameters show a positive albeit modest correlation with NCV (236–238). Moreover, impairments in CCM parameters have been linked with reduced pain/temperature perception and other measures of SFN (226, 236, 237). Taken together, these data suggest that CCM may be used to classify the severity of DSPN. Moreover, CCM may be used to study the exact anatomy of nerve impairments. For example, CNFL may be reduced in T1D earlier (even before the onset of manifest DSPN) than in T2D (234).

Early corneal nerve fiber loss

In participants with recent-onset T2D from the German Diabetes Study baseline cohort, we demonstrated that the extent of corneal nerve fiber loss mirrors the magnitude of IENFD reduction. However, because CNFD and IENFD did not correlate, a patchy manifestation pattern of early SFN in T2D is likely (135). More recently, we reported in the same cohort that the spatial point pattern analysis of the distribution of corneal nerve branching points reveals an increased clustering rather than random distribution of early CNF loss and substantially improves its detection when combined with CNFL. Thus, when combining an individual spatial point pattern analysis parameter with CNFL, one or both of two indices were below the first percentile of controls in 28.6% of patients compared with 2.1% of controls, whereas for the conventional CNFL/CNFD/CNBD combination the corresponding rates were 16.3% vs 2.1% (221).

Natural history

With advancing age, a slight reduction in nerve fibers may occur, and this may be studied noninvasively with CCM. CNFL has been found to be reduced in association with age and T1D duration as well as to predict the onset of DSPN after 4 years (239). CCM has also been used to study neuropathy in prediabetes. Indeed, reductions in CNFD, CNBD, and CNFL, along with other SFN tests, have been demonstrated in individuals with IGT (240). Among individuals with IGT, those who developed T2D exhibited reductions in CNFL, CNFD, and CNBD at baseline, as well as further deterioration of CNFL at 3 years (241). The increasing use of CCM has enabled the generation of normative values in an adequately large sample of healthy individuals (242).

Evoked responses

A promising approach to overcome the inherent limitations of the existing tools to assess SFNs, being either invasive or affected by psychophysical variability, involves evoked responses to external small-fiber stimuli such as heat or pain (243). Currently, there are several different ways to assess evoked responses, characterized by different methods to provoke and record. Evoked responses that are detected and recorded using simultaneous electroencephalography include laser-evoked potentials (LEPs), contact heat-evoked potentials (CHEPs), and pain-related–evoked potentials (PREPs) (204). Both LEPs and CHEPs are induced by rapid heat stimuli and conveyed by small-caliber Aδ- and C-fibers. To elicit PREPs, epidermal nerve fibers are stimulated slightly above the pinprick detection thresholds with a concentric surface electrode. Whereas CHEPs are induced by a contact thermode, LEPs rely on heat pulses by laser beams without physical contact between skin and stimulator (163, 244).

Several studies that investigated CHEPs or LEPs in diabetes patients with DSPN or persons with SFN reported a strong connection between measures of SFN (IENFD, thermal thresholds) and evoked potentials (163, 171, 243, 245). Overall, a good comparability between LEPs and CHEPs was reported (163, 245). Normative values of CHEPs in healthy individuals have been published (246). Two studies should be mentioned, in which evoked potentials were performed in combination with the capsaicin model (see “Skin biopsy” above). Correlations between skin biopsy markers, thermal perception thresholds, and LEPs were observed in a study using the capsaicin model (169). In another recent study comparing CHEPs, LEPs, and PREPs with IENFD before and after topical capsaicin application, both CHEPs and LEPs were completely abolished after capsaicin-induced denervation, whereas PREPs remained unaltered despite an extremely diminished IENFD, suggesting that PREPs may detect nerve fiber sections other than epidermal endings alone (245). Findings that PREPs are strongly correlated with IENFD and are useful to detect early small fiber involvement in diabetes need to be confirmed in larger studies.

A different noninvasive approach to test small-fiber function using evoked responses is the laser Doppler imaging (LDI) flare method, which assesses an axon reflex–triggered neurovascular response. Under physiological conditions, stimulation of C-fibers lead to neurogenic vasodilation, resulting in an increased microvascular blood flow. For the LDI flare, C-fiber stimulation is performed by heating the skin using a contact thermode. The microvascular flow is measured in a predefined skin area via LDI. Finally, the areas of microvascular response and maximal hyperemic response are analyzed (207, 247). However, good sensitivity and specificity for detecting DSPN in a sufficiently large sample were published only by one study (248), whereas a larger study concluded that QST performed better in this regard (249).

Imaging techniques

Although nerve imaging techniques based on magnetic resonance and ultrasound have provided information on several focal and inflammatory neuropathies, their use in diffuse polyneuropathies such as DSPN has been challenging. Recently, some evidence has emerged that advances in imaging technologies could also be used to assess peripheral nerve damage and altered brain structure and activity in patients with diabetes (204, 250).

Brain magnetic resonance imaging

DN has been considered primarily a PNS condition, but evidence has accumulated suggesting that parts of the central nervous system are also affected, underscoring the diagnostic and scientific potential of brain imaging techniques (251, 252). Especially functional magnetic resonance imaging (MRI) has been extensively used to assess brain activity in patients with neuropathic pain (251). Recent work suggests that a descending pain modulatory system dysfunction may reflect a brain-based pain facilitation mechanism contributing to painful DSPN (253). The finding that the neural circuitry subserving pain perception interacts with the cerebral correlates of peripheral nociceptive fibers implicates an indirect role for skin nerves in human pain perception (254). Associations between brain volume and functional connectivity of certain pain-processing regions, particularly in the anterior cingulate cortex and IENFD, were reported in patients with SFN of different origin (255). In a smaller study analyzing brain MRI scans of T1D patients, lower levels of gray matter volume were associated with more severe DSPN (250). Using single-photon emission CT, a recent study reported an increased cerebral blood flow in the anterior cingulate cortex in patients with painful DN that was normalized when pain relief was achieved following a 12-week analgesic treatment with duloxetine (256). Collectively, these cross-sectional studies point to central nervous system involvement in both painful and painless DN, but prospective and multimodal studies are required to ascertain whether MRI could help to develop more specific treatments (251).

“Capsaicin induces a chemical axotomy to cutaneous nerve fibers.”

Magnetic resonance neurography

High-resolution magnetic resonance neurography can be used to assess morphological features of peripheral nerves and identify functionally relevant lesions (204). A recent magnetic resonance neurography study challenged the traditional distal-to-proximal pattern of nerve pathology in DSPN by claiming a proximal-to-distal gradient of sciatic nerve lesions with a predominance of lesions at the thigh level compared with more distal nerve sections in diabetes patients with DSPN (257). However, because the clinical scores used are not suitable to sensitively determine the time course of nerve pathology, these findings should be verified in appropriate longitudinal studies. These authors also reported distinct patterns of sciatic nerve lesions between T1D and T2D patients and associations of such lesions with the severity of clinical DSPN. They hypothesized that these differences might reflect different pathogenic factors involved in the development of DSPN in T1D compared with T2D (258).

Diffusion tensor imaging

Diffusion tensor imaging (DTI) is a magnetic resonance technique that is sensitive to small changes in nerve microstructure and allows for quantitative assessment of nerve integrity (204). A recent study reported close associations of DTI parameters assessed in the sciatic or tibial nerve with the presence and severity of DSPN in T1D patients, whereas no differences in the ratio of tibial and sciatic DTI parameters were observed between T1D patients with and without DSPN and healthy controls (259). Another study reported an excellent interoberserver reliability but moderate diagnostic accuracy of DTI parameters in diabetes patients with DSPN, but only a small sample size was studied (260). It would be of interest to explore possible associations between DTI and other morphometric and functional parameters in a larger cohort of patients with T1D and T2D to evaluate the potential of DTI in the early detection and prediction of DSPN.

Nerve ultrasound

Ultrasound imaging is a widely available and fairly mobile technique, and high-frequency transducers allow for good visualization of peripheral nerves. Several cross-sectional studies have recently reported increased cross-sectional areas of peripheral nerves in diabetes with or without DSPN compared with controls (261, 262). However, the reports on associations between cross-sectional areas and measures of DSPN such as NCS vary considerably, indicating the need for longitudinal studies to explore the predictive value of ultrasound for DSPN in individuals with diabetes.

Bedside tests

NC-stat DPNCheck

NCS are the most objective diagnostic tool to assess peripheral large–nerve fiber function. NC-stat DPNCheck is an automatic NCS point of care (POC) device designed to assess sural SNCV and sensory nerve action potential (SNAP) in an easy to use fashion without the need for extensive electrophysiological training. The device consists of a nerve stimulator and a single-use flexible sensor surface unit placed in a fixed distance from the stimulator, a thermometer, and an LCD screen. It is applied at the distal-lateral thigh with the stimulator aligned to the outer malleolus and the sensor firmly pushed down on the skin. If skin temperature is in an acceptable range between 23 and 30°C, 100 mA of current are distributed and the obtained sural SNCV and SNAP are displayed on an LCD screen or can transferred to a PC. Reference data are used to check for abnormalities. The validity and effectiveness in diabetes patients with DSPN has been confirmed (263). A recent study reporting a sensitivity of 77% and specificity of 68% to detect DSPN proposed its use as a quick large-fiber test in combination with electrochemical skin conductance testing (see “Sudomotor function tests” below) to covering the small fiber component. Using abnormality in either of the results yielded a sensitivity and specificity of 93.2% and 52.8%, respectively (264). Another recent study suggested the usefulness of NC-stat DPNCheck as a diagnostic tool in lower resource settings (265).

Vibratip

The Vibratip is a tiny and lightweight disposable device basically consisting only of a battery, a 128-Hz vibration electric motor, a plastic housing with a round tip at the end, and an on/off switch activating the motor while the device is being squeezed by the investigator between the thumb and index finger. It is used to test whether a patient can discriminate between vibration and no vibration when touched with the round plastic tip, and it seeks to add a simple method besides a monofilament and 64-Hz Rydel-Seiffer tuning fork. It was evaluated against vibration perception thresholds (VPTs) and the clinical NDS and showed very good accuracy to detect severe neuropathy (31). In a recent study from Columbia, a high specificity, albeit low sensitivity, to detect neuropathy was reported (266), as neuropathy was classified by a relatively low score on MNSI of >2. A recent study compared the performance of Vibratip, a 128-Hz tuning fork, and a neurothesiometer in the detection of elevated VPT and showed that the percentage of participants who did not perceive vibrations was highest when using the Vibratip (28.5%), followed by the neurothesiometer (21%) and the 128-Hz tuning fork (12%), suggesting that to avoid misdiagnosis, different modalities should be used in diabetes patients, and when these do not concur, further evaluation should be performed (267). Vibratip was selected by the UK Medical Technologies Advisory Committee to undergo evaluation through the National Institute for Health and Care Excellence. The UK Medical Technologies Advisory Committee considered that the technology showed promise but decided that the case for adoption was not proven (268).

Multifrequency vibrometry

In analogy to audiometry, where hearing thresholds for individual frequencies are determined, multifrequency vibrometry can be performed to obtain VPTs at different vibration frequencies. The underlying principle is that different skin mechanoreceptors respond only within a certain frequency range, with a maximum frequency where they respond best (269). The VibroSense Meter can investigate seven frequencies between 8 and 500 Hz, yielding a vibrogram curve. The device is connected to a PC and measurement is controlled via software. A ratio is then calculated between the area under the curve of the vibrogram curve and an age-matched normal curve, resulting in the Sensibility Index. In patients with diabetes, impaired VPTs at low frequencies (8 to 32 Hz) were reported (270). It would be of interest to determine in further studies whether vibration detection by Meissner’s corpuscles and Pacinian corpuscles is differently affected in DSPN.

NeuroQuick

The NeuroQuick is a handheld POC device with a fan emitting cold air and two crossing laser beams to meet a standardized distance from the skin. The fan velocity can be adjusted in 10 steps, and the patient is asked to report when a cold air sensation is felt on the dorsum of the foot. The NeuroQuick threshold is defined by the lowest fan velocity level at which the airflow is recognized. It has been suggested that the NeuroQuick is a valid and reliable screening tool for quantitative assessment of small-fiber dysfunction and appears to be more sensitive in detecting early DSPN than both elaborate thermal testing and screening tests such as the tuning fork (271). Although currently not commercially available, a recent study tested the NeuroQuick against other economical POC tests for an early diagnosis of neuropathy in leprosy, and a very good accuracy was reported (272).

Smartphone-generated vibrations

According to a recent study, a common smartphone with an inbuilt linear oscillating motor (for vibrating alert) is an accurate screening tool for DSPN. The vibration had ~25 Hz. When tested at the head of the first metatarsal, the phone used with free software showed better accuracy in screening for DSPN than did a Rydel-Seiffer tuning fork or 10-g monofilament (273). This study primarily highlights the future potential of smartphones as medical devices, because these have a user interface, ports or wireless connections, a vibration motor, can even be forced to generate heat (from processing units), and are virtually ubiquitous in many societies.

Skin autofluorescence

The accumulation of AGEs plays a role in the development of diabetic microvascular complications including DN, but evaluating AGE levels from serum samples does not permit drawing conclusions about AGE accumulation in specific tissues. The AGE Reader was designed to assess AGE accumulation in the skin by measuring skin autofluorescence (SAF) in a noninvasive, safe, standardized, and clinically feasible fashion. The principle of autofluorescence is that a material or molecule reflects light with a different wavelength than the light source by which it is illuminated. The skin possesses autofluorescence properties from molecules, such as collagen or the reduced form of nicotinamide adenine dinucleotide, phosphate, and AGEs as well. Provoked by an excitation wavelength of 370 nm, AGEs emit light at 440 nm. The AGE Reader analyzes the wavelength spectra emitted by the skin when illuminated from a UV light source. The intensity of the fluorescence in the wavelengths of AGEs is supposed to correlate with the quantity of AGEs in the skin. Skin autofluorescence correlated with markers of AGE accumulation in skin biopsies, but a formal validation against skin AGEs was not undertaken (274). Various studies have reported positive associations between SAF and diabetes and its complications, but evidence on independent associations with DN, retinopathy, or nephropathy is mixed (275–277). There is a plethora of studies available reporting associations between various symptoms or diseases with SAF, so the specific use in detecting DN should be viewed in a critical light, whereas the potential to predict DN specifically could be further evaluated in prospective studies.

Sudomotor function tests

Sudomotor function refers to the activity of sweat secretory glands (34). Easy-to-apply tests of sudomotor function include the Neuropad indicator test (278) and the Sudoscan (279).

Neuropad.

Neuropad is a visual test, documenting the adequate/inadequate sweat production via a color change from blue to pink (278, 280). This change should occur in <10 minutes: a normal response has been defined as a complete and uniform change from blue to pink in both feet within this time interval (278, 280). Lack of change or an incomplete (blue and pink or “patchy”) response has been defined as abnormal (278, 280). The color change can be explained as follows: water produced by the sweat glands is absorbed by the blue salt anhydrous cobalt(II) chloride, which is contained in the indicator test, and the result of this chemical reaction is pink cobalt(II) chloride salt (278, 280). The reproducibility of results is very good (281). A larger multicenter study including 1010 patients has reported 94.9% sensitivity, 70.2% specificity, and 98.1% negative predictive value (NPV) (282). The corresponding values for small fiber dysfunction were 85.6%, 71.2%, and 93.3% (282). Clearly, the foremost advantage of Neuropad is its very high NPV: thus, it may serve as a screening test primarily to exclude DSPN (280, 282). Conversely, owing to its rather moderate specificity, abnormal results would need conformation by additional testing (280, 282). An 86% average sensitivity and a 65% average specificity have also been confirmed in a meta-analysis including 3470 participants with diabetes (283). The positive likelihood ratio was 2.44 and the negative likelihood ratio was 0.22 (283). The very low negative likelihood ratio corroborates the utility of Neuropad as a screening test to exclude DSPN (283).

“There is a need of large-scale prospective studies to identify oxidative stress biomarkers.”

Moreover, Neuropad has been shown to contribute to the early diagnosis of DSPN in T2D patients (282). Although clinical examination at baseline was normal in all patients, development of DSPN after 5 years was more likely in those who initially had abnormal Neuropad response (284). We examined the performance of Neuropad for DSPN in patients with recent-onset diabetes from the German Diabetes Study baseline cohort and found 87.5% sensitivity and 47.7% specificity in T1D patients as well as 65.1% sensitivity and 48.2% specificity in those with T2D (285). Neuropad may also enable the diagnosis of DSPN in patients with prediabetes (286). This has been evaluated in a population study including elderly subjects: sensitivity, specificity, NPV, and positive predictive value were 57.8%, 33.3%, 76.5% and 17.3%, respectively (286).

More sophisticated methods involve measuring the absolute time to color change of Neuropad (in minutes or seconds) (278, 281, 287–290) and, most recently, reading the percentage of Neuropad surface exhibiting color change (291, 292). The area with color change was lower in the presence (50%) vs absence (90%) of DSPN, and it was inversely associated with severity of DSPN. Again, Neuropad had a high sensitivity and a moderate specificity for all measures of DSPN, including symptoms (292). Then, an automated continuous image analysis was used to quantify the area with color change (291). Thus, high sensitivity and specificity were obtained for corneal nerve fiber density (88% and 78%) and sensory nerve action potential (88% and 83%) (291), suggesting that this new image analysis approach may improve the diagnostic yield of Neuropad. Overall, Neuropad is being increasingly appreciated for the study of DSPN, and especially small-fiber dysfunction (233, 282, 291, 292) and as a screening tool (266, 280), and it may become useful in neuropathy of other causes (272).

Sudoscan.

Sudoscan (279) is another noninvasive test to assess sweat gland secretory function. Similar to the Neuropad, it is based on a simple chemical reaction (279). Again, no expert medical personnel are needed. The principle is the reaction between sodium chloride of sweat and nickel of the electrodes placed on patient’s palms and feet: the latter produce a low-voltage electrical current, which attracts sodium chloride from sweat in palms and foot soles (279, 293, 294). The device measures the electrochemical skin conductance (ESC) in these two areas as a ratio between two currents: the resultant current and the one produced by the device (279). ESC depends on the amount of sodium chloride (i.e., sweat). ESC exhibits an inverse correlation with VPT (293, 295). It is also associated with longer T2D diabetes duration and nephropathy (279, 296, 297). No discomfort is felt during the examination, and results are reproducible (279, 293).

DSPN was associated with lower ESC in feet and hands in diabetes patients (298). ESC in the feet exhibited inverse correlations with neuropathic impairments, QST, and autonomic failure (298). Its reproducibility was very good in controls, and its sensitivity and specificity for DSPN were 78% and 92%, respectively (298). In a more recent work from China (299), sensitivity and specificity were 88.2% and 46.9%, respectively. Especially in China, there is accumulating experience with ESC as a screening tool for DSPN (300, 301). ESC correlated with neuropathic symptoms and signs as well as VPT (301). However, a recent systematic review including 37 studies argued among others that large combined data sets do not support a high sensitivity and specificity, normative ESC values are inconsistent across publications, and there is insufficient evidence supporting the claim that ESC tests sudomotor or sensory nerve fiber function (302).

Quantitative sensory testing

Neuropathic pain can be accompanied by both positive and negative sensory signs, and extensive efforts were undertaken to classify these as specific sensory phenotypes using QST with the ultimate goal to establish a mechanism-based analgesic pharmacotherapy (303, 304). Thermal and mechanical tests are performed to assess sensory loss and mechanical and thermal hyperalgesia, respectively. It has been suggested that two major pain phenotypes may be present in patients with neuropathic pain: preserved small-fiber function with hyperalgesia referred to as “sensory gain” or “irritable nociceptor” and “non-irritable nociceptor” dominated by sensory loss (305, 306). Recently, phenotypic data from patients with peripheral neuropathic pain collected by three large multinational consortia were combined to perform a cluster analysis that identified three different clusters. Cluster 1 represented dominant sensory loss, whereas clusters 2 and 3 were characterized by thermal and mechanical hyperalgesia, respectively (307). It has been suggested that pain phenotypes may predict the individual response to treatment, paving the way to mechanism-based personalized treatment. For example, one RCT reported better pain relief by oxcarbazepine in the irritable nociceptor group (305). However, several studies concluded that differences in sensory phenotypes between painful and painless DSPN are marginal at best, and neither sensory phenotypes nor single QST parameters, besides evoked allodynia, are useful to distinguish between painful and painless neuropathies of similar severity (209, 306, 308, 309), casting doubt about a link between sensory phenotypes and neuropathic pain. In contrast, QST is a powerful instrument for early detection of small-fiber dysfunction in DN (308) and is well established to assess small- and large-fiber sensory function regardless of neuropathic pain (306). Moreover, the occurrence of an “irritable nociceptor” phenotype seems rare in DN, as it was observed only in a minority (6.3% and 14.6%) of participants with painful DN in the PiNS and ncRNAPain studies, respectively (209, 310). Thus, it remains to be seen whether QST will help to pave the way toward mechanism-based personalized analgesic pharmacotherapy in the future, albeit preferentially in painful neuropathies due to conditions other than diabetes.

Causal Treatment

Lifestyle intervention

Lifestyle management is an essential component of diabetes prevention and care and includes diabetes self-management education and support, medical nutrition therapy, physical activity, smoking cessation counseling, and psychosocial care (311). In the Da Qing study, lifestyle intervention for 6 years was associated with a 47% reduction in the incidence of severe, vision-threatening retinopathy during 20 years, primarily due to the reduced incidence of diabetes in the intervention group. However, similar benefits were not seen for nephropathy or neuropathy (312). In the Diabetes Prevention Program (DPP), all participants were offered lifestyle training at the end of the randomized comparison of lifestyle intervention, metformin, or placebo after 3 years. After a mean follow-up of 15 years, diabetes incidence was reduced by 27% in the lifestyle intervention group compared with the placebo group. The prevalence rates at the end of the study of the aggregate microvascular outcome were not different between the treatment groups in the total cohort. However, in a post hoc analysis among participants whose most recent HbA1c was ≥6.5%, representing ~26% of the cohort, the lifestyle intervention group showed reductions compared with placebo and metformin in the aggregate microvascular outcome, retinopathy, and neuropathy assessed by the 10-g monofilament (relative risk, 0.38; 95% CI, 0.19 to 0.75 and 0.39, 0.19 to 0.79), suggesting that in people who convert to diabetes lifestyle intervention, prevalent neuropathy may be reduced (313).

In overweight or obese adults with T2D, however, the large-scale controlled Look AHEAD study reported that an intensive lifestyle intervention (ILI) focusing on weight loss through decreased caloric intake and increased physical activity (intervention group) compared with diabetes support and education (DSE, control group) did not reduce the rate of cardiovascular events after 9.6 years (314). The interventions were terminated 9 to 11 years after randomization, but both groups continued to be followed for both primary and secondary outcomes. Neuropathy evaluations included the MNSI questionnaire completed at baseline and repeated annually thereafter, and the MNSI physical examination and light touch sensation testing was conducted 1 to 2.3 years after discontinuation of the intervention (315). ILI resulted in less prominent increase in neuropathic symptoms, which was associated with the magnitude of weight loss. In both the ILI and DSE groups, changes in the MNSI questionnaire score were also related to changes in HbA1c and lipids. There were no effects of ILI on the MNSI physical examination score, except for light touch sensation, which was better in the ILI group when measurements were combined for both toes (315). However, a limiting factor in the DPP and Look AHEAD studies was that no baseline measurements were available to analyze incident DSPN.

In a small uncontrolled study, diet counseling and exercise counseling for 1 year based on the DPP approach in individuals with IGT were associated with an increase in IENFD in the proximal thigh but not in the calf, indicating that at least IENFs at more proximal sites could be amenable to lifestyle intervention resulting in partial cutaneous reinnervation. However, RCTs are needed to confirm these preliminary findings (316).

Exercise intervention studies indicate that balance training appears to be the most effective type of intervention. Studies focusing exclusively on strength, or a combination of endurance and strength, appear to have a lower impact. Endurance training also plays an important role for DSPN. Further research with high methodological quality needs to be conducted to establish evidence-based clinical recommendations for neuropathic patients (317). It has also been suggested that aerobic exercise produces salutary effects in many of the pathways implicated in the pathogenesis of DN, improves symptoms of neuropathy, and promotes regrowth of cutaneous small-diameter fibers (318). However, the studies on which this notion was based were uncontrolled and small.

“Proinflammatory biomarkers are associated with DSPN and lower NCV in multiple cohorts.”

Bariatric surgery

Bariatric surgery results in pronounced and sustained weight loss, and it is associated with other important health outcomes, among which the effect on remission of T2D is substantial. However, the benefits of weight loss surgery on other obesity-related comorbidities are less clear (319). Only a few uncontrolled studies assessed the effects of bariatric surgery on measures of DSPN. In a large retrospective observational cohort study, patients who experienced T2D remission after bariatric surgery had 29% lower risk of incident microvascular disease compared with patients who never remitted (320). However, the vast majority of incident microvascular complications were attributable to incident retinopathy, whereas incident neuropathy was very low. In a small uncontrolled study the NDS improved after 12 months following Roux-en-Y gastric bypass in relationship to improvements in systemic biomarkers of oxidative, nitrosative, and carbonyl stress (321), but in another study motor and sensory NCV did not change after 12 months following Roux-en-Y gastric bypass (322). Altogether, low-grade evidence indicates that bariatric surgery may improve measures of DSPN, but RCTs are required to confirm these preliminary findings. The possible benefit of bariatric surgery should be weighed against the risk of subacute axonal neuropathy caused by micronutrient deficiencies, particularly low thiamine levels (323) and autonomic dysfunction including orthostatic intolerance (324).

Glycemic control

T1D

The Diabetes Control and Complications Trial (DCCT) and Epidemiology of Diabetes Interventions and Complications (EDIC) Study and other smaller trials demonstrated that intensive insulin therapy (IIT) aimed to achieve near-normal glycemia is essential to prevent, albeit not completely, or delay progression of DSPN in T1D patients. At DCCT closeout, the participants originally assigned to conventional insulin treatment were also encouraged to adopt intensive treatment. Notably, the previously assigned conventional insulin treatment patients continued to develop complications, including neuropathy at higher rates than the previous IIT therapy group, despite nearly similar HbA1c levels during the EDIC Study. This persistence of benefit from early application of IIT therapy has been termed “metabolic memory” (325). Differences in epigenetic DNA methylation during the DCCT persist at certain loci associated with glycemia for several years during the EDIC Study, supporting an epigenetic explanation for metabolic memory (326). Similar to the reported long-term benefits of prior intensive glycemic control on retinopathy, nephropathy, and CVD, the differences in the incidence and prevalence of DSPN reflect differences in glucose control, favoring HbA1c levels that are closer to nondiabetic levels (325). In a prospective observational study, we found that near-normoglycemia maintained from the diagnosis of T1D during 24 years was associated with a complete prevention of the decline in hyperglycemia-related peripheral and autonomic nerve function, as well as development of confirmed clinical DSPN, suggesting that long-term near-normoglycemia might even fully prevent the development of DSPN (187).

There is accumulating evidence from smaller trials to suggest that CCM can document nerve fiber regeneration following stringent therapeutic interventions. Simultaneous pancreas and kidney transplantation in patients with T1D resulted in improvements in CCM parameters at 6 and 12 months (327, 328). Moreover, continuous subcutaneous insulin infusion during 24 months resulted in increases of CNFD, CNBD, and CNFL (241). Similarly, improved cholesterol levels and glycemic control led to improved CCM parameters after 24 months (329). In contrast, prospective studies during several years after simultaneous pancreas and kidney or islet cell transplantation did not demonstrate any improvement in IENFD (132, 330). Although the aforementioned trials were small and uncontrolled, CCM seems to offer the opportunity to study the effects of interventions on DSPN, but this remains to be shown in RCTs.

T2D

In contrast to T1D, there is no convincing evidence in T2D patients to suggest that intensive diabetes therapy has a favorable effect on the development or progression of DSPN (331). Of note, these trials were not specifically designed to evaluate the effects of intensive diabetes treatment on DSPN. Thus, only a minority of the patients enrolled in these studies had symptomatic polyneuropathy at entry. Moreover, in contrast to the DCCT, the trials conducted in patients with T2D have used only a few clinical endpoints of DSPN or VPT rather than an array of quantitative functional measures including NCV or morphometric measures such as IENFD or CCM.

Multifactorial risk intervention

Numerous studies have shown the efficacy of controlling individual cardiovascular risk factors in preventing or slowing CVD in people with diabetes. In a recent Swedish cohort study in T2D patients who had five CVD risk variables within the target ranges appeared to have little or no excess risk of death, myocardial infarction, or stroke, as compared with the general population (332). Therefore, CVD risk factors should be systematically assessed at least annually in all patients with diabetes and be treated according to current guidelines (4). However, only two studies assessed the effect of multifactorial cardiovascular risk intervention (CVRI) on DSPN. In the Steno 2 Study, intensified CVRI including intensive diabetes treatment, angiotensin-converting enzyme (ACE) inhibitors, antioxidants, statins, aspirin, and smoking cessation in patients with microalbuminuria showed no effect on DSPN after 7.8 years (range, 6.9 to 8.8 years) and again at 13.3 years, after the patients were subsequently followed observationally for a mean of 5.5 years, whereas the progression of cardiovascular autonomic neuropathy could be retarded (333). In the ADDITION-Europe Study, neuropathy was present at 5 years in 4.9% and 5.9% of the patients receiving intensive and routine care, respectively (334). However, apart from the lack of baseline assessment, it is conceivable that differences between the groups in DSPN prevalence could not be observed simply due to the fact that both groups achieved similar improvements in CVD risk factors in the ADDITION Study. Thus, not least because CVRI trials may not result in meaningful differential effects of intensive and routine care on CVD risk factors sufficient to potentially discriminate between the groups in DSPN outcomes, there are insufficient data to appraise the evidence supporting a favorable effect of CVRI on DSPN.

Treatment Based on Pathogenetic Concepts

Recent experimental studies suggest a multifactorial pathogenesis of DN. Importantly, from the clinical point of view, note that, based on the various pathogenetic mechanisms, therapeutic approaches could be derived, some of which have been evaluated in RCTs, including the aldose reductase inhibitors (alrestatin, sorbinil, ponalrestat, tolrestat, epalrestat, zopolrestat, zenarestat, fidarestat, ranirestat), the antioxidant α-lipoic acid (thioctic acid), essential fatty acids (γ-linolenic acid), ACE inhibitors (trandolapril), prostacyclin (PGI2) analogs (iloprost, beraprost), prostaglandin derivatives (PGE₁⋅αCD), NGF, protein kinase C (PKC)β inhibitor (ruboxistaurin), C-peptide, VEGF, benfotiamine (vitamin B1 derivative), and Actovegin (36, 335). These drugs have been designed for disease modification, that is, to favorably influence the underlying neuropathy rather than for symptomatic pain treatment. Painful DSPN can be treated with analgesic drugs, but these have no effect on sensory deficits or the pathogenesis underlying DN. Because in the foreseeable future, normoglycemia will not be achievable in most diabetes patients, the advantage of the aforementioned treatment approaches is that they may exert their effects despite prevailing hyperglycemia. For clinical use, only α-lipoic acid, benfotiamine, and Actovegin are licensed and used for treatment of symptomatic DSPN (Fig. 1) in several countries, whereas epalrestat is marketed in Japan and India.

Both T1D and T2D have different, but important, immunological components (336), and the development of all diabetes-related complications is exacerbated by subclinical inflammation (66, 337–339). Anti-inflammatory drugs have not been tested in the prevention and treatment of DN, but they will also be discussed here briefly because of the aforementioned link between subclinical inflammation and DSPN (see “Systemic biomarkers” above) and because of promising data regarding their use in patients with T2D and other complications.

α-Lipoic acid

α-Lipoic acid, a naturally occurring dithiol compound with antioxidant properties (Fig. 1), has long been known as an essential cofactor for mitochondrial bioenergetic enzymes. Several meta-analyses and systematic reviews (evidence class Ia) suggest that α-lipoic acid is an effective and safe drug for the treatment of symptomatic DSPN (340–345). There is evidence to suggest that short-term treatment using 600 mg of α-lipoic acid IV per day for 3 weeks or 600 mg orally for 5 weeks reduces the main neuropathic symptoms such as pain, paresthesias, and numbness as well as deficits (signs, impairments) of DSPN. In the NATHAN 1 trial, a response analysis of clinically meaningful improvement and progression in the NIS and NIS-LL by at least 2 points showed that the rates of clinical responders were higher and those of progressors were lower with α-lipoic acid vs placebo for NIS and NIS-LL after 4 years of treatment, respectively, suggesting that the drug may improve neuropathic deficits in the long term (346). A recent post hoc analysis of the NATHAN 1 study showed that improvement and prevention of progression of NIS-LL with α-lipoic acid vs placebo after 4 years was predicted by higher age, lower BMI, male sex, normal blood pressure, history of CVD, insulin treatment, longer duration of diabetes and neuropathy, and higher neuropathy stage. Thus, better outcome in neuropathic impairments was predicted by normal baseline CVD risk factors and higher burden due to CVD, diabetes, and neuropathy, suggesting that optimal control of CVD risk factors could contribute to improved efficacy of α-lipoic acid in patients with higher disease burden (347). Clinical and postmarketing surveillance studies revealed a favorable safety profile of the drug (348).

“IENFD is reduced as early as within the first 12 months after diagnosis of T2D.”

Actovegin

Actovegin, a deproteinized ultrafiltrate of calf blood composed of >200 active biological substances, has been shown to exhibit a range of pleiotropic effects and to improve experimental DN presumably via antiapoptotic/oxidative stress mechanisms, including poly(ADP-ribose) (349) (Fig. 1). In a multicenter trial, 567 patients with symptomatic DSPN were randomized to receive 20 daily IV infusions of Actovegin (2000 mg/d) followed by three Actovegin tablets daily (1800 mg/d) or placebo for 140 days. Both neuropathic symptoms and VPT improved with Actovegin treatment, and the drug was well tolerated, with an adverse event profile similar to placebo (350). In a recent post hoc analysis of this trial, response to treatment with Actovegin compared with placebo was associated with better odds of response, defined as a clinically meaningful improvement in neuropathic deficits and/or symptoms from baseline to 6 months (351). The results of this RCT should be confirmed by a longer-term pivotal trial. A recent 12-month trial has shown that Actovegin treatment also leads to improved cognitive outcomes in patients with poststroke cognitive impairment (352), supporting the notion that the drug may have neuroprotective potential.

Benfotiamine

The lipid-soluble thiamine derivative benfotiamine inhibits three of the major biochemical pathways implicated in the pathogenesis of hyperglycemia-induced vascular damage (hexosamine pathway, AGE formation pathway, and diacylglycerol–protein PKC pathway) (Fig. 1) by activating transketolase in retinas of diabetic animals and it also prevents experimental diabetic retinopathy (353). The BENDIP study showed that neuropathic symptoms, with Neuropathy Symptom Score as the primary endpoint, were improved after 6 weeks of treatment using a benfotiamine dose of 300 mg twice a day but not 300 mg once a day (354), whereas the BEDIP study showed an improvement in a score combining neuropathic symptoms and signs after 3 weeks of treatment with benfotiamine at 100 mg four times a day (355). The incidence of adverse events did not differ between active and placebo treatment, but there remains a need for a longer-term RCT. Currently underway is a 1-year RCT (BOND study) to assess the effects of treatment with benfotiamine on morphometric, neurophysiological, and clinical measures in T2D patients with mild to moderate symptomatic DSPN. Several transketolase SNPs were recently found to be associated with measures of DSPN in patients recently diagnosed with diabetes (150). As a consequence, it has been suggested to maintain the focus on the therapeutic attempt to target thiamine and transketolase (147). Genetic variations in transketolase enzyme could be useful for the identification of responders/nonresponders to benfotiamine treatment and might open up new perspectives in the pharmacogenomics of this drug in the future.

Novel agents

Owing to overall weak efficacy, the development of recombinant human NGF, acetyl-l-carnitine, γ-linolenic acid, the PKCβ inhibitor ruboxistaurin, the aldose reductase inhibitor ranirestat, and pegylated C-peptide for DSPN was discontinued (36). Pirenzepine, a selective antagonist of the muscarinic acetylcholine receptor M₁, which has been used in the treatment of peptic ulcers for decades, has recently been shown to prevent or reverse indices of peripheral neuropathy, such as depletion of sensory nerve terminals, thermal hypoalgesia, and NCV slowing in diverse rodent models of diabetes. Because a variety of antimuscarinic drugs are approved for clinical use in other conditions, prompt translation of this therapeutic approach to clinical trials appears feasible (356).

Metanx

Metanx is a combination of l-methylfolate, methylcobalamin, and pyridoxal-5-phosphate. In a 24-week double-blind clinical trial, 214 T2D patients with DSPN were randomly assigned to Metanx or placebo. There was no effect on VPT (primary endpoint). However, neuropathic symptoms improved with Metanx vs placebo at weeks 16 and 24. This was accompanied by improvement in NDS at week 16 but not at week 24. The drug was generally well tolerated without increased adverse events (357).

Tocotrienols (vitamin E subtypes)

In a large RCT including 300 diabetic patients with DSPN, oral supplementation of mixed tocotrienols (400 mg/d) for 1 year did not improve overall neuropathic symptoms. However, in post hoc subgroup analyses, tocotrienols reduced lancinating pain among patients with baseline HbA1c levels >8% or those with normohomocysteinemia. The preliminary observations on lancinating pain among subsets of patients require further exploration (358).

Erythropoietin analog (ARA 290)

ARA 290 is an 11–amino acid peptide with a mass of 1257 Da derived from helix B of the erythropoietin molecule. In a small trial, 48 T2D patients were randomized to ARA 290 or placebo (both subcutaneously administered in the anterior thigh) and treated for 28 days. ARA 290 was associated with improvement in neuropathic symptoms and some parameters of quality of life. With ARA 290, there was also a nonsignificant increase in CNFD, which achieved significance in the subset of patients whose initial CNFD had been clearly abnormal. The rates of adverse events did not differ between the groups (359).

PhVEGF165 gene transfer (SB-509)

In a small study, patients with DSPN were randomized to plasmid VEGF (n = 39) or placebo (n = 11). Injections were unilaterally administered close to the sciatic, peroneal, and tibial nerve. After 6 months, improvements in neuropathic symptoms and pin-prick sensation were observed in the treated limbs. There were 10 and 2 patients experiencing serious events in the VEGF and placebo groups, respectively (360).

Human hepatocyte growth factor gene transfer (VM202)

In a phase 2 randomized, double-blind, placebo-controlled study, intramuscular injections with 8 or 16 mg of plasmid (VM202) encoding two human hepatocyte growth factor isoforms or placebo were administered per leg. Divided doses were administered on day 0 and day 14. Eighty-four patients completed the study after 9 months. Patients receiving 8 mg of VM202 per leg improved the most in all efficacy measures, including a reduction in the mean pain score at 3 months but not at 6 and 9 months. Patients not on pregabalin or gabapentin had the largest reductions in pain. There were no significant adverse events attributable to VM202 (361).

Altogether, from the clinical point of view, there is a continuing need for the development of novel drugs tailored to target the pathogenetic mechanisms underlying DSPN. Experimental studies of low-dose combined drug treatment suggest enhanced drug efficacy mediated by facilitatory interactions between drugs. Although considerable improvement in the quality of RCTs has recently been achieved, no major breakthrough in slowing the progression of DN in the long run has been achieved with drugs used on the basis of present pathogenetic concepts. Some of the newer drugs have shown promising results in phase 2 trials that require confirmation with further robust evidence from large phase 3 trials together with an improved understanding of the mechanisms of action of emerging treatments.

Obviously the failure to show treatment benefits in previous trials could have been due to a very slow worsening of DSPN possibly due to a better control of CVD risk factors by concomitant pharmacotherapy or lifestyle modification, a placebo effect for symptoms and signs, and measurement noise. It has been suggested that demonstrating disease progression in controlled trials of DSPN among others is more likely when patients with developing rather than established DSPN are selected, patients are selected who cannot or will not achieve ideal glycemic control, endpoints chosen are known to show monotonic worsening, and restricted numbers of centers and expert examiners (trained, certified, using standard approaches, and reference values and interactive surveillance of tests) are used (362). Furthermore, it has been proposed that future clinical trials in patients with DN should minimally enroll patients with mild or moderate neuropathy, preferentially use surrogate endpoints of small fiber repair, and operate in a regulatory environment that accepts small-fiber repair as a desirable primary endpoint (363). Finally, it is also conceivable that drugs interfering with the pathogenesis of DN may be most effective in terms of prevention, rather than intervention.

Anti-inflammatory drugs

Immune activation and subclinical inflammation characterize T1D and T2D (336, 364). Recently, large studies have been completed or are ongoing to assess the efficacy of anti-inflammatory approaches to treat diabetic complications. The results of the Canakinumab Anti-inflammatory Thrombosis Outcomes Study trial demonstrated that an intervention based on an IL-1β–inhibiting antibody administered subcutaneously every 3 months lowered the risk of the primary cardiovascular endpoint (nonfatal myocardial infarction or stroke, cardiovascular death) by 15% (95% CI, 2%, 26%) at the intermediate dose of 150 mg compared with placebo (365). Importantly, the decrease in cardiovascular risk correlated with the reduction in CRP during the trial (366). Comparable studies are ongoing for methotrexate and colchicine (367).

Both experimental and epidemiological studies have implicated IL-1β and its antagonist IL-1RA in DSPN (Fig. 1). The positive association between IL-1RA and prevalence and progression of DSPN (102, 107) is reminiscent of the increase of systemic levels of this protein before manifestation of T2D and CVD in the general population (203, 368). Thus, IL-1β inhibition may also be an interesting approach in the prevention and treatment of DSPN.

Thus, both diabetes types and their complications have inflammatory components that are increasingly targeted in RCTs. Clinical research on DSPN has lagged behind in this respect, and further studies are timely and warranted to assess the potential and safety of anti-inflammatory drugs to modify the development and progression of this diabetic complication (Fig. 1). Finally, keep in mind that most or all lifestyle approaches to treat diabetic complications (as discussed in “Lifestyle intervention” above) have anti-inflammatory effects, which may partially mediate the association between intervention and improved outcomes (369).

Conclusions

Recent years have witnessed substantial new insights into the multifactorial pathogenic mechanisms involved in the development and progression of DN. Several lines of evidence link oxidative stress and inflammation with the development of DNs. However, experimental data originated from short-lived animal models with known limitations for studying the human pathophysiology, which may explain the major difficulties in transferring the favorable effects of numerous pathogenesis-derived agents from experimental DN into the clinical arena. Thus, there remains an unmet need for translational data from animal models suitable to emulate the neuropathic processes encountered in humans. Novel biomarkers derived from the current pathogenic concepts are promising in predicting the development and progression of DN, with biomarkers of oxidative stress and inflammation being the most promising candidates so far. However, large prospective studies are required to further validate these markers. Moreover, hypothesis-free omics technologies should be employed to identify novel biomarkers and pathways that can potentially be targeted to prevent or treat DN. Novel sophisticated and simple diagnostic tools have been introduced focusing primarily on assessing the morphology and function of small nerve fibers at early stages of diabetes, but it remains to be seen whether these tools may predict the feared clinical endpoints of DN such as neuropathic pain, foot ulcers, amputations, or even mortality and whether they can be successfully applied to monitor the development and progression of DN in its natural course or in RCTs. Some agents derived from the pathogenetic concepts of DN are being used in several countries around the world, but the existing favorable data obtained for new drugs in early clinical trials have to be replicated in phase 3 RCTs. Moreover, there is a clear unmet need not only for the development of new, more efficacious compounds preferentially even addressing multiple mechanistic pathways, but also for further trials addressing nonpharmacological strategies such as lifestyle intervention. Only if in the future these interventions will prove to be successful in preventing and treating DN and neuropathic pain, it will also become realistic to prevent the feared sequels such as foot ulcers and amputations.

Abbreviations

Abbreviations
 
• ACE

  angiotensin converting enzyme

 
• AGE

  advanced glycation end product

 
• BMI

  body mass index

 
• CCL

  C-C motif ligand

 
• CCM

  corneal confocal microscopy

 
• CHEP

  contact heat-evoked potential

 
• CNBD

  corneal nerve branch density

 
• CNFD

  corneal nerve fiber density

 
• CNFL

  corneal nerve fiber length

 
• Cox-2

  cyclooxygenase-2

 
• CVD

  cardiovascular disease

 
• CVRI

  cardiovascular risk intervention

 
• (CXCL)1

  C-X-C motif ligand

 
• DCCT

  Diabetes Control and Complications Trial

 
• DN

  diabetic neuropathy

 
• DPP

  Diabetes Prevention Program

 
• DRG

  dorsal root ganglion

 
• DSPN

  distal sensorimotor polyneuropathy

 
• DTI

  diffusion tensor imaging

 
• EC

  endothelial cell

 
• ECM

  extracellular matrix

 
• EDIC

  Epidemiology of Diabetes Interventions and Complications

 
• ESC

  electrochemical skin conductance

 
• E-selectin

  endothelial leukocyte adhesion molecule-1

 
• GAP-43

  growth-associated protein 43

 
• GLP-1

  glucaon-like peptide-1

 
• ICAM-1

  intercellular adhesion molecule-1

 
• IENF

  intraepidermal nerve fiber

 
• IENFD

  IENF density

 
• IGT

  impaired glucose tolerance

 
• IIT

  intensive insulin therapy

 
• ILI

  intensive lifestyle intervention

 
• IL-1RA

  IL-1 receptor antagonist

 
• KORA

  Cooperative Health Research in the Region of Augsburg

 
• LC

  Langerhans cell

 
• LDI

  laser Doppler imaging

 
• LEP

  laser-evoked potential

 
• MMP

  matrix metalloproteinase

 
• MNSI

  Michigan Neuropathy Screening Instrument

 
• MPO

  myeloperoxidase

 
• MRI

  magnetic resonance imaging

 
• MTHFR

  methylene-tetrahydrofolate reductase

 
• mTOR

  mammalian target of rapamycin

 
• NCS

  nerve conduction studies

 
• NCV

  nerve conduction velocity

 
• NDS

  Neuropathy Disability Score

 
• NF-κB

  nuclear factor κB

 
• NGF

  nerve growth factor

 
• NIS-LL

  Neuropathy Impairment Score of the lower limbs

 
• NLD

  nonlinear dynamics

 
• NO

  nitric oxide

 
• NPV

  negative predictive value

 
• PARP1

  poly(ADP-ribose) polymerase-1

 
• PGP9.5

  protein gene product 9.5

 
• PI3K

  phosphoinositide 3-kinase

 
• PKC

  protein kinase C

 
• PNS

  peripheral nervous system

 
• p75NTR

  p75 neurotrophin receptor

 
• POC

  point of care

 
• PPP

  pentose phosphate pathway

 
• PREP

  pain-related evoked potentials

 
• PTEN

  phosphatase and tensin homolog deleted on chromosome 10

 
• QST

  quantitative sensory testing

 
• RCT

  randomized clinical trial

 
• ROC

  receiver operating characteristic

 
• SAF

  skin autofluorescence

 
• sICAM-1

  soluble ICAM-1

 
• SNAP

  sensory nerve action potential amplitude

 
• SNCV

  sensory NCV

 
• SNP

  single-nucleotide polymorphism

 
• SOD

  superoxide dismutase

 
• SOD2

  mitochondrial SOD

 
• SOD3

  extracellular SOD

 
• SPECT

  single-photon emission computed tomography

 
• T1D

  type 1 diabetes

 
• T2D

  type 2 diabetes

 
• TRPV1

  transient receptor potential vanilloid type 1

 
• VCAM-1

  vascular cell adhesion molecule-1

 
• VEGF

  vascular endothelial growth factor

 
• VGSC

  voltage-gated sodium channel

 
• VPT

  vibration perception threshold

Acknowledgments

Financial Support: This work was supported by the Ministry of Culture and Science of the State of North Rhine–Westphalia and by the German Federal Ministry of Health. This study was supported in part by a grant of the Federal Ministry for Research to the German Center for Diabetes Research and in part by a grant from the German Center for Diabetes Research.

Author Contributions: G.J.B., C.H., D.Z., N.P., and A.S. wrote the article; M.R. reviewed and edited the article. D.Z. is the guarantor of this work and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Disclosure Summary: D.Z. has been a consultant for Mitsubishi Tanabe, Wörwag, Pfizer, TrigoCare, NeuroMetrix, Allergan, Berlin-Chemie, Teva, Astellas, Meda, Novartis, and Takeda; has received speaker honoraria from Wörwag, Pfizer, Eli Lilly, Takeda, Astellas, AstraZeneca, Meda, Berlin-Chemie, and Impeto Medical; and has received research support from Wörwag and Mitsubishi Tanabe. M.R. has been recently on advisory boards of Boehringer-Ingelheim, Genetech, Merck Poxel, and Sanofi and has received speaker honoraria from Boehringer-Ingelheim, Eli Lilly, and Novo Nordisk. N.P. has been an advisory board member of TrigoCare International, Abbott, AstraZeneca, Elpen, MSD, Novartis, Novo Nordisk, Sanofi-Aventis, and Takeda and received speaker honoraria from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Elpen, Galenica, MSD, Mylan, Novartis, Novo Nordisk, Pfizer, Sanofi-Aventis, Takeda, and Vianex. C.H. received speaker honoraria from Eli Lilly and Sanofi-Aventis. The remaining authors have nothing to disclose.

References

Author notes