CIDP and other inflammatory neuropathies in diabetes diagnosis and management

Transcription

1 PERIPHERAL NEUROPATHIES CIDP and other inflammatory neuropathies in diabetes diagnosis and management 1 Aston Brain Centre, School of Life & Health Sciences, Aston University, Aston Triangle, Birmingham B4 7ET, UK. 2 Regional Neuromuscular Service, University Hospitals Birmingham, Birmingham B15 2WB, UK. 3 Department of Neurology, University Hospital Essen, Hufelandstrasse 55, D Essen, Germany. 4 Department of Neurology, Medical Faculty, Research Group for Clinical and Experimental Neuroimmunology, Heinrich-Heine University, Moorenstrasse 5, Düsseldorf, Germany. 5 Weill Cornell Medicine-Qatar, Education City, PO Box 24144, Doha, Qatar. 6 Division of Cardiovascular Medicine, University of Manchester, 46 Grafton Street, Manchester M13 9NT, UK. Correspondence to Y.A.R. y.rajabally@aston.ac.uk doi: /nrneurol Published online 15 Sep 2017 Yusuf A. Rajabally 1,2, Mark Stettner 3,4, Bernd C. Kieseier 4, Hans-Peter Hartung 4 and Rayaz A. Malik 5,6 Abstract Distal symmetric polyneuropathy (DSPN) is the most common neuropathy to occur in diabetes mellitus. However, patients with diabetes can also develop inflammatory neuropathies, the most common and most treatable of which is chronic inflammatory demyelinating polyradiculoneuropathy (CIDP). Whether diabetes is a risk factor for CIDP remains under debate. Early studies suggested that patients with diabetes were at increased risk of CIDP, but epidemiological studies failed to confirm the association, and subsequent data have re opened the debate. Inadequate interpretation of investigations and differentials between CIDP and other neuropathies that can occur in diabetes, such as DSPN, diabetic radiculoplexus neuropathies and vasculitic multiple mononeuropathy, might mean that CIDP is under-recognized. Despite a response rate of >80% to first-line therapies for CIDP in patients with or without diabetes, those with diabetes often present with greater disability owing to late referral and axonal pathology attributed to DSPN. The increasing worldwide prevalence of diabetes creates an urgent need to improve identification of potentially treatable neuropathies, such as CIDP. In this Review, we consider the features of CIDP in patients with diabetes, and discuss how these features can be used to differentiate the condition from other neuropathies. We also review the management options for CIDP and other inflammatory neuropathies in patients with diabetes. Diabetic neuropathy is the most prevalent chronic complication of diabetes mellitus (referred to throughout this Review as diabetes). The term diabetic neuropathy is normally used to refer to the forms of neuro pathy that are most prevalent among patients with diabetes, which are distal symmetric polyneuropathy (DSPN) and diabetic autonomic neuropathies 1 4 ; DSPN presents in 10 15% of patients with newly diagnosed type 2 dia betes, and occurs in 36.8% of patients who have had diabetes for >10 years 5,6. However, not all patients with diabetes and neuropathy symptoms have these typical diabetic neuro pathies 1 4,7, and patients with diabetes can develop inflammatory neuropathies. These inflammatory neuropathies include diabetic radiculoplexus neuro pathies and vasculitic multiple mononeuropathies, but the most common and most treatable is chronic inflammatory demyelinating polyneuropathy (CIDP). Therefore, neuropathies that are not typically associated with diabetes and might be treatable should be actively sought and managed in patients with diabetes 8, and when a patient with diabetes presents with neurological deficits that are not typical of DSPN, inflammatory neuropathies must be considered. In this Review, we consider the features of CIDP in patients with diabetes, and discuss how these features can be used to differentiate the condition from other neuropathies. We also review the management options for CIDP and other inflammatory neuropathies in patients with diabetes. Clinical features of CIDP CIDP is a heterogeneous entity caused by an immune-mediated inflammatory process that involves nerve roots, plexuses and peripheral nerve trunks. In its typical form, CIDP produces symmetrical proximal and distal muscle weakness in all four limbs, large-fibre sensory loss, and reduced or absent reflexes 9. CIDP is usually progressive over at least 8 weeks, although it can occur in a relapsing remitting pattern with long periods of remission Atypical CIDP might be as common as typical forms 13, and is diverse, including focal and multifocal asymmetric, distal, pure motor and pure sensory forms 9. NATURE REVIEWS NEUROLOGY VOLUME 13 OCTOBER

2 Key points The main neuropathy that occurs in diabetes is distal symmetric polyneuropathy (DSPN), but inflammatory neuropathies can also occur Chronic inflammatory demyelinating polyradiculoneuropathy (CIDP) is the most common and most treatable inflammatory neuropathy in patients with diabetes, although the extent of its occurrence in association with diabetes is debated Other inflammatory neuropathies that occur in diabetes are diabetic radiculoplexus neuropathies and vasculitic multiple mononeuropathy Diagnosis of CIDP in the presence of diabetes can be made mainly on the basis of clinical characteristics and specific electrophysiological criteria; cerebrospinal fluid analysis, imaging and neuropathology are occasionally helpful The amenability of CIDP to treatment makes its identification important First-line treatment options for CIDP in diabetes include intravenous immunoglobulin and corticosteroids; plasma exchange can be used when these treatments are ineffective, and immunosuppression can occasionally be considered in refractory disease Is diabetes a risk factor for CIDP? The frequency of CIDP among patients with diabetes is uncertain. An association between CIDP and diabetes has been acknowledged for several decades, but the condition remains poorly recognized owing to the concurrent existence of a typical diabetic neuropathy in many patients. A report published in 1986 described a patient with diabetes and a relapsing inflammatory demyelinating polyneuropathy, and is probably one of the earliest published descriptions of CIDP in a patient with diabetes 14. Subsequent studies published in the late 1990s also described the association, and emphasized the importance of its recognition, given the potential for treatment 15,16. Epidemiological data on the association between CIDP and diabetes entered the literature from the early 2000s. A retrospective study from the USA showed that 30 of 87 patients with CIDP had diabetes 17. In a French series of 100 consecutive patients with diabetes (74 of whom had type 2 diabetes) and neuropathy who were referred to a secondary or tertiary centre, nine had CIDP 18. Another North American non-population-based study demonstrated that the risk of CIDP in patients with diabetes was 11 fold higher than in those without diabetes 19. Furthermore, the frequency of CIDP was the same in type 1 and type 2 diabetes. The same study showed that the odds of an individual with CIDP having diabetes were >20 fold higher than in those with myasthenia gravis or amyotrophic lateral sclerosis (ALS). The authors acknowledged referral bias as a potential confounding factor, because patients with typical diabetic neuropathy would not undergo electrophysiological studies under normal circumstances. However, they emphasized the very high frequency of CIDP in patients with diabetes, and felt that the difference in frequency when compared with other neuromuscular conditions could not be explained by selection bias, as the frequencies of diabetes in their patients with myasthenia gravis or ALS were similar to those observed in previous population-based case control epidemiological studies. In an epidemiological study conducted in the UK, seven of 46 (15.2%) patients with CIDP (diagnosed according to the European Federation of Neurological Societies Peripheral Nerve Society (EFNS PNS) criteria) had diabetes 20. In another study from South East England, ten of 101 (10%) people with CIDP had dia betes 21. Both of these studies were uncontrolled, but identi fied a higher rate of diabetes among people with CIDP than had been reported among the general population (5 6%) 22. In a study published in 2016, the prevalence of CIDP and diabetes was investigated by using a health insurance administrative claims database (Phar-Metrics Plus Database, Watertown, Massachusetts, USA) 23 of over 100 million patients across the USA. People aged >65 years were under-represented, and diagnostic criteria for CIDP and diabetes were lacking, but the frequency of CIDP was six per 100,000 patients without diabetes and 54 per 100,000 patients with diabetes, a ninefold increase. Other epidemiological data suggest no associ ation of CIDP with diabetes. A large population-based Italian study of 4,334,225 people showed that among 155 patients with CIDP, only 9% had diabetes, equating to a standardized morbidity ratio of only 1.07 (95% CI, ) for diabetes in CIDP 24. The methods used to diagnose diabetes might have been suboptimal, however, as they relied exclusively on fasting blood glucose levels or the fact that an individual used oral anti diabetic drugs. Epidemiological data from the USA have also challenged the association between CIDP and dia betes 25. Among 260,000 inhabitants of Olmsted County, Minnesota, the prevalence of CIDP was 8.9 per 100,000 people, and only one of 23 patients with CIDP (4.3%) had diabetes, compared with 14 of 115 (12%) age-matched and sex-matched controls. Diabetes was diagnosed on the basis of documented treatment with antidiabetic medication, two or more fasting blood glucose level values >126 mg/dl, two or more non-fasting blood glucose level values >200 mg/dl, or a coded diagnosis in the medical records. The odds ratio for prevalent diabetes among patients with CIDP was 0.3 (95% CI ), and did not change after adjustment for age, sex and previously diagnosed diabetes. The variability in the reported prevalence of coexisting CIDP and diabetes might be attributed to difficulties in accurate identification of individuals with the two conditions, partly owing to different diagnostic criteria for CIDP and diabetes, and variation among the populations studied. Concurrent neuropathic symptoms of DSPN, the rigour with which electrodiagnostic studies are undertaken, and the interpretation of the electrodiagnostic data could contribute to overdiagnosis or underdiagnosis of CIDP. Diagnosis of CIDP in diabetes The same criteria are used to diagnose CIDP in patients with and without diabetes, and diagnostic information is provided by a variety of techniques. Electrophysiology. Diagnosis of CIDP relies heavily on the detection of demyelination by means of electrophysio logical tests. Several electrodiagnostic criteria for CIDP have been proposed in the past 30 years 26. A multicentre European study in which different criteria were evaluated found that the EFNS PNS criteria offer 600 OCTOBER 2017 VOLUME 13

3 the best combination of sensitivity and specificity 27. These criteria have since become the most widely used in clinical research worldwide 28, and their use is advisable in routine clinical practice, as the high sensitivity minimizes the likelihood that the diagnosis of a treatable condition will be missed. Cerebrospinal fluid analysis. Cerebrospinal fluid (CSF) protein levels are increased in most patients with CIDP. However, as protein levels can be normal in >50% of atypical forms 29, the precise global sensitivity of this marker is uncertain, although a figure as high as 95% has been reported for typical forms 30. However, a substantial proportion of patients with diabetes in particular, those with a longer disease duration can have high levels of CSF protein 31. Consequently, making a diagnosis of CIDP as opposed to another neuropathy in a patient with diabetes is challenging on the basis of CSF protein levels alone, although levels >1 g/l seem to be exceptional in patients with diabetes unless they have CIDP 18. Use of this value as a cut-off which, for similar reasons, has been proposed for distinguishing between patients with Charcot Marie Tooth disease and those with Charcot Marie Tooth disease associated with CIDP 32 might be helpful. Imaging. Various imaging techniques have been added to the diagnostic armamentarium for CIDP in the past 15 years. Magnetic resonance neurography (MRN) might aid the diagnosis of CIDP by enabling identification of nerve hypertrophy and/or hyperintensity 33,34. Gadolinium enhancement on MRI, particularly in hypertrophied nerve roots, might also contribute by indicating the presence of an inflammatory process, as occurs in CIDP 35. Whether MRN might be useful for differentiating CIDP from other neuropathies in patients with diabetes is currently unknown, particularly because several studies have shown that MRN detects proximal lesions in patients with DSPN this observation raises doubts about the specificity of changes detected by MRN in patients with CIDP and diabetes, as opposed to diabetes alone. A study published in 2017 has shown that diffusion tensor imaging (DTI) MRN can identify nerve lesions in patients with type 1 diabetes, reflecting proximal and distal nerve fibre pathology 39. However, nerve abnormalities on DTI have also been described in CIDP 40, and further research is needed to determine whether the technique can distinguish between the two conditions. Finally, ultrasound imaging has indicated that crosssectional nerve area enlargement is present in DSPN 41,42 and CIDP 43,44, and an association between nerve size and CIDP disease status has been reported 45. Future studies might determine whether CIDP and DSPN can be differentiated according to sites and patterns of nerve size abnormalities. Neuropathology. Early reports of CIDP emphasized the value of nerve biopsy findings 46, but the relevance of histology is currently a matter of debate among specialists. As nerve biopsy is an invasive procedure, samples from healthy controls are rare, so establishing the normal parameters, which would enable a reliable disease biomarker to be established, is difficult 47. The overwhelming consensus is that detection of demyelination in semi-thin or ultra-thin sections is neither specific nor sensitive for CIDP 48, particularly when attempting to distinguish CIDP from demyelinating hereditary polyneuropathies 49,50. Total T-cell and macrophage counts, which are markers of immune-mediated nerve damage, also have low sensitivity for CIDP diagnosis 51,52, although the presence of perivascular macrophage clusters has been proposed as a valid criterion to differentiate inflammatory neuropathy from other forms of neuropathy with high sensitivity and specificity 53. Some evidence suggests that other inflammatory markers, such as endoneurial oedema or the activity of matrix metallo proteinase 9 (MMP-9), in patients with diabetes and CIDP have higher diagnostic value 54,55, but these studies were exploratory, and the methods used are not widely available 47. For peripheral neuropathology, the quality of the reports is highly dependent more so than for other diagnostic work-ups on the availability of specialized laboratories and expertise. The consensus among neuromuscular physicians is that a nerve biopsy is not needed for the initial diagnosis of CIDP, as it is unlikely to provide a conclusive answer. Nevertheless, the method might prove helpful in cases of atypical CIDP 48, for the exclusion of differential diagnoses, and for further investigation when patients do not respond to treatment. An alternative technique that has the potential to aid diagnosis of CIDP is corneal confocal microscopy (CCM), a noninvasive and rapid ophthalmological imaging technique that can quantify axonal loss in a variety of peripheral neuropathies 56. CCM may be a viable surrogate end point for early diagnosis of DSPN, stratification of patients with diabetes according to the severity of their neuropathy, and assessment of response to treatment 57,58. In a cross-sectional study, patients with CIDP, multifocal motor neuropathy, or neuropathy with MGUS (monoclonal gammopathy of undetermined significance) were compared with healthy controls (in total, 182 patients and controls were included). Reduced corneal nerve fibre density, branch density and length were associated with Langerhans cell infiltration around corneal nerves (FIG. 1) and correlated with the degree of neurological deficits 59. Further longitudinal studies are required to determine the utility of this technique for assessing progression of disease and response to treatment. Distinguishing CIDP The challenge of determining whether a patient has CIDP is greatest in patients with diabetes, because these individuals can have neurological deficits that resemble CIDP phenotypes but are caused by other neuropathies. These neuropathies can include the typical diabetic neuropathy, or the atypical inflammatory neuropathies. Here, we discuss the features of these conditions, and how they can be distinguished from CIDP in patients with diabetes. DSPN. DSPN is characterized by early involvement of small fibres, which leads to pain and dysaesthesias 1,4,60,61, and later involvement of large fibres, which NATURE REVIEWS NEUROLOGY VOLUME 13 OCTOBER

4 a c Figure 1 Corneal confocal microscopy images in health, diabetes and CIDP. Images are of the sub-basal layer of the cornea. a Sub-basal central corneal nerves (arrow) in a healthy individual, without cell infiltrates. b Density of central corneal nerve fibres is decreased in a patient with diabetes, and the number of cell infiltrates is low. c Sub-basal layer in a patient with chronic inflammatory demyelinating polyneuropathy (CIDP). The number of corneal nerve fibres is lower than in a healthy individual, and there is a high density of mainly dendritic cell infiltrates (inset magnification). d Sub-basal layer in a patient with CIDP who has a low number of corneal nerve fibres at the inferior whorl and mainly non-dendritic cell infiltrates (inset magnification). Scale bars 200 μm. Part d modified with permission from Wiley and Sons Stettner, M. C. et al. Neurology 3, (2016). can cause numbness and loss of protective sensation 62. Neurological assessment reveals distal sensory loss to touch and pinprick in both feet 63. Typically, sensory deficits are greater than motor deficits, symptoms are symmetrical, and progression is slow, so if other presentations indicate an alternative aetiology 2,64 66, early referral to a neurologist is recommended 8. The presentation of CIDP contrasts with the predominant small-fibre involvement in DSPN, and the overt motor impairment that is characteristic of advanced DSPN 67, although evidence from the past 2 years suggests that patients with type 2 diabetes or impaired glucose tolerance without clinically evident neuropathy can develop an early subclinical reduction in proximal and distal strength, and an altered gait 68,69. The typically rapid progressive course of CIDP also contrasts with the evolution of DSPN, which progresses slowly over years; consequently, rapid progression of neurological disability in a patient with diabetes is deemed to be a red flag for a neuropathy other than DSPN, and CIDP should be considered. Some forms of CIDP can develop insidiously, however, with pain, fatigue or mild neurological deficits, making the differential diagnosis more difficult 70 73, particularly in patients with diabetes. In this context, regular follow up with neurological evaluation especially of proximal b d strength, proprioceptive function and reflexes is paramount to avoid delays in diagnosis of CIDP. A degree of electrophysiological motor conduction slowing may be observed in DSPN. The slowing seen in DSPN, however, is generally not to the degree that is observed in CIDP. In initial descriptions of the association between CIDP and diabetes, most patients with either disorder met the highly stringent and specific American Academy of Neurology (AAN) diagnostic criteria for CIDP 15,16. A study published in 2013 revealed that although 54% of patients with DSPN had a demyelin ating form either pure or combined with axonal loss the degree of slowing was variable, and mild slowing that did not meet the EFNS PNS diagnostic criteria for CIDP was present in many individuals 74. In patients with type 1 diabetes, this slowing was associated with poor glycaemic control. The same degree of conduction slowing had been suggested in a previous study 75. In a subsequent study, patients with demyelinating DSPN were compared with patients who had CIDP and dia betes 76. Neuropathy scores were worse and demyelin ation was clearly more marked among patients with CIDP and diabetes than among patients with demyelin ating DSPN. Interestingly, patients with CIDP were older, in keeping with the known higher prevalence of the disorder with age 20. Furthermore, these patients had a shorter duration of diabetes and better glycaemic control than patients with DSPN, which seems to confirm the involvement of different pathophysiological mechanisms in concurrent CIDP. In summary, patients with diabetes and CIDP can be reliably differentiated, in most cases, from those with diabetes and DSPN with slow motor conductions, by using adequate electrophysiological parameters and cut-offs. Motor conduction velocities <70% of the lower limit of normal, distal motor latencies >150% of the upper limit of normal, and conduction block >50% are, when detected in multiple nerves, highly specific for CIDP. Distal compound muscle action potential dispersion is also helpful in identifying the demyelination that occurs in CIDP, as this feature is infrequently observed in axonal neuropathies 77,78. F wave prolongation might be less helpful Sural nerve biopsy studies in CIDP can reveal pathology that is similar to that seen in DSPN: demyelination and remyelination (the onion bulb formation) is a charac teristic feature (FIG. 2) 51. Endoneurial oedema has also been observed in CIDP, particularly in acuteonset CIDP 54. In a study of sural nerve biopsy samples, epi neurial perivascular T-cell infiltrates and immunoreactivity for MMP 9 was found in patients with CIDP and diabetes, but these features were absent in patients with DSPN, suggesting that they could represent a biomarker for CIDP 55. A diagnostic tool that combines clinical, electrophysio logical, CSF protein and serological markers of autoimmunity to identify CIDP in patients with diabetes has been proposed, and was validated in a small numbers of patients 12 with CIDP and diabetes, 18 with CIDP only, and 27 with DSPN only 79. The parameters that were used comprised elements that support 602 OCTOBER 2017 VOLUME 13

5 a b c Figure 2 Histopathological features in CIDP. a Staining against CD3 with diaminobenzidine (DAB) reveals perivascular T-cell infiltrates (brown) in the sural nerve of a patient with chronic inflammatory demyelinating polyradiculoneuropathy (CIDP); cell nuclei are counterstained with haematoxylin and appear blue (magnification 200x). b In the endoneurium, large numbers of CD68 positive foamy macrophages can be seen with DAB staining (brown; magnification 200x). c Semi-thin nerve sections in CIDP reveal a large degree of myelin loss as well as thinning of the myelin (toluidine blue staining, magnification 400x). a diagnosis of CIDP, derived from characteristic features of CIDP, and elements that contradict a diagnosis of CIDP, derived from the characteristics of the axonal neuropathy of DSPN (TABLE 1). Diabetic radiculoplexus neuropathies. Diabetic radiculoplexus neuropathy (also known as diabetic amyotrophy, Bruns Garland syndrome or diabetic poly radiculoneuropathy) typically involves the lumbosacral plexus or, less frequently, the cervical plexus, and usually occurs in middle-aged to elderly men with type 2 diabetes 83. The life-long incidence among patients with type 2 diabetes is ~1% 84. Diabetic lumbosacral radiculoplexus neuropathy (DLRPN) presents with extreme unilateral thigh pain with weakness, starting proximally and monolaterally, and often extending distally and contralaterally. Pain and sensory symptoms occur early, and weakness and muscle wasting are long-term features of the condition. Weight loss is common, and can be substantial. Upper-limb involvement can occur. DLRPN can also occur in a painless form, which is considered by some to be a diabetic CIDP, given that its onset is symmetrical and slower than that of DLRPN, and that the symptoms are more widespread and frequently involve the upper limbs 85,86. In a study of 23 patients with slowly progressive, painless, symmetrical, motor-predominant neuropathy, nerve biopsy demonstrated a reduction in myelinated fibre density with epineurial vessel wall inflammation and microvasculitis 87, similar to that seen in the typical painful form 88,89. In both the painless and painful forms, CSF protein levels are raised and electrophysiology can reveal mild conduction slowing. The disease course is monophasic, and patients recover without treatment, although the condition can last for up to 18 months 80,82 and recovery can take up to 3 years 90. Multiple studies have suggested that DLRPN has an autoimmune basis, as indicated by the microvasculitis that has been demonstrated pathologically 88,89. Diabetic cervical radiculoplexus neuropathy (DCRPN) has similar characteristics to DLRPN 82. Onset is acute in most patients, pain is the presenting symptom in >60% of patients, weakness is an almost universal feature, and sensory symptoms and weight loss are common. Panplexopathy is also common, as is contra lateral spreading. Concurrent lumbosacral or thoracic involvement occurs in 25% of cases. CSF protein levels are mildly raised, and electrophysiology indicates axonal rather than demyelinating involvement. Abnormal MRI findings, including increased T2 signals, nerve hypertrophy and, in rare cases, contrast enhancement, are common. As in the case of DLRPN, the disease course is monophasic. Diabetic radiculoplexus neuropathies can be difficult to distinguish from CIDP, particularly given the proximal motor weakness and possible bilateral involvement. However, the acute presentation and the prominence of pain in diabetic radiculoplexus neuropathy is not typical of CIDP, and is sufficient to enable a correct diagnosis. In addition, the conduction slowing that can be observed in diabetic radiculoplexus neuropathy is not sufficient to meet the stringent criteria for CIDP 87. Vasculitic multiple mononeuropathies. Vasculitic multiple mononeuropathy (also called mononeuritis multi plex), which does not resemble the radiculoplexus neuropathies, can occur in diabetes 91. This condition presents with painful mononeuropathies that manifest in succession over days to weeks. The distinction of vasculitic multiple mononeuropathies from CIDP is usually straightforward owing to the mononeuropathic pattern and the acute and painful presentation, which are not features of CIDP. CIDP presenting as acute inflammatory neuropathy. The progressive course of CIDP distinguishes the condition from the acute inflammatory polyneuropathies, or Guillain Barré syndrome, in which the nadir is reached by week 4. However, ~15% of patients with CIDP can experience an acute onset that is indistinguishable from that of Guillain Barré syndrome 92, although the subsequent relapsing or progressive course confirms the diagnosis of CIDP 93. The absence of a preceding infection, facial weakness, dysautonomia or a need for mechanical ventilation indicate CIDP, as does the presence of sensory signs 92,93. NATURE REVIEWS NEUROLOGY VOLUME 13 OCTOBER

6 Management of CIDP in diabetes CIDP is the most common and treatable inflammatory neuropathy that occurs in patients with diabetes. The main first-line therapies are intravenous immunoglobulin (IVIg) or corticosteroids, and their use is based on the knowledge that CIDP results from aberrant immune responses to peripheral nerve antigens (FIG. 3). The inflammation that results from this response is charac terized by nerve oedema, endoneurial infiltration by macrophages, perivascular infiltration by T cells 53,94, and increased expression of cytokines and other inflammatory molecules in the CSF and blood Breakdown of the blood nerve barrier is the critical initial stage that leads to the passage of activated T cells and entry of inflammatory mediators, which further contribute to permeability of the blood nerve barrier and upregulation of the neural inflammatory response. CD4 + and CD8 + T cells and macrophages cluster around endoneurial vessels and mediate demyelination 98. Various hypoth eses have been proposed about the underlying mechan isms that trigger the immune response , but a full understanding remains elusive. In addition to IVIg and corticosteroids 102,103, a third evidence-based option for first-line treatment of CIDP is plasma exchange. However, this approach is used less frequently than the other two owing to its impracticality and the fact that the response is short-term and necessitates repeated treatment 104. The decision whether to use IVIg or steroids depends on the individual patient s comorbidities, the type of CIDP, and the disease severity. In patients with diabetes, corticosteroids are generally avoided, as they can result in deterioration of glycaemic control. However, caution is needed with IVIg in the presence of multiple vascular risk factors, including diabetes 105,106. Similarly, history of a recent thromboembolic event should lead to avoidance of IVIg. Motor CIDP should be treated with IVIg, as corticosteroids can cause worsening of symptoms 107. The latest comparative study of IVIg and corticosteroids showed that the former is also less likely to be stopped as a result of ineffectiveness or adverse effects 108. This finding suggests that IVIg may be preferred for severe disability that requires rapid improvement, although long-term Table 1 Selected features of DSPN, CIDP and CIDP in patients with diabetes Feature DSPN CIDP CIDP in patients with diabetes Clinical presentation Diagnostics Histology Treatment Early small-fibre involvement, pain, dysaesthesia, numbness If autonomic neuropathy: tachycardia, orthostatic hypotension, gastroparesis, constipation, diarrhoea, erectile dysfunction, neurogenic bladder and sudomotor dysfunction 1 4,60,61,157,158 Progression over years Electrophysiology: sensorimotor polyneuropathy, mild slowing of NCV with normal latencies 74,75 CSF protein levels might be slightly elevated 31 Myelinated fibre loss Reduction in unmyelinated axon density and diameter Increase in unassociated Schwann cell profile density Endoneurial microangiopathy (luminal narrowing, endothelial cell hyperplasia and hypertrophy, pericyte cell loss and basement membrane thickening ) Symptomatic Improvement of glycaemic control might limit progression in type 1 diabetes, but has no effect in type 2 diabetes No evidence to support use of immunotherapy Symmetrical proximal and distal muscle weakness, sensory loss to large-fibre modalities (vibration and proprioception) and reduced or absent reflexes 9 Often greater motor than sensory deficits Rapid progression over months (at least 8 weeks) Subgroups: focal and multifocal asymmetric presentation, distal forms, pure motor forms and pure sensory forms Electrophysiology: high-sensitivity and high-specificity EFNS PNS criteria met Increased CSF protein levels are common MRN and ultrasound might be helpful in some cases for detecting heterogeneously thickened and hyperintense nerves Sural nerve biopsy might be helpful in some cases for demonstrating inflammatory features 47 Demyelination and remyelination (onion bulb formation) Endoneurial oedema and epineurial perivascular T cell infiltrates 51,54 Intravenous immunglobulin or corticosteroids as first-line therapy Plasma exchange 118 Cyclophosphamide 131,132 or rituximab 133 in refractory disease Features of CIDP without diabetes Older patients Short duration of diabetes Good glycaemic control 76 Electrophysiology: NCV to <70% of the LLN, distal motor latency >150% of the ULN, and conduction block >50%, all in multiple nerves, highly specific; distal compound muscle action potential dispersion 83,87 89 CSF: not specific to distinguish from DSPN, but levels >1 g/l are unusual in diabetes 32,159 Sural nerve biopsy might be helpful in some cases for demonstrating the inflammatory features of CIDP 47 Epineurial perivascular T cell infiltrates Immunoreactivity for matrix metalloproteinase 9 (REF. 55) Corticosteroids to be avoided Iintravenous immunoglobulin as first-line therapy Plasma exchange 118 Cyclophosphamide 131,132 or rituximab 133 in refractory disease CIDP, chronic inflammatory demyelinating neuropathy; CSF, cerebrospinal fluid; DSPN, distal symmetric peripheral neuropathy; EFNS PNS, European Federation of Neurological Societies Peripheral Nerve Society; LLN, lower limit of normal; MRN, magnetic resonance neurography; NCV, nerve conduction velocity; ULN, upper limit of normal. 604 OCTOBER 2017 VOLUME 13

7 Systemic immune compartment BNB Peripheral nerve Myelin APC Apoptosis T cell Compact myelin T cell Macrophage Schwann cell IL-4 IL-6 Antibodies T cell T H 1 IL-10 TGF-β IFN-γ TNF C5b 9 P0 P2 MBP Non-compact myelin Extracellular B cell T H 2 Intracellular MAG Cx32 GLP Figure 3 Current model of immunopathogenesis in CIDP. The immune response in chronic inflammatory demyelinating polyneuropathy (CIDP) involves cellular and humoral immune responses. Antigen-presenting Nature Reviews cells Neurology (APCs), such as dendritic cells or macrophages, activate autoreactive T cells. These T cells can cross the blood nerve barrier (BNB) to enter the peripheral nerve, a process mediated in part by chemokines, cellular adhesion molecules and metalloproteinases (orange circles). In addition, T cells can activate B lymphocytes via secretion of cytokines, such as IL 4 and IL 6. Within the PNS, T cells are re exposed to the target antigen by local APCs (a macrophage is depicted), thereby reactivating these cells and driving clonal expansion within the peripheral nerve. CD4 + T cells can differentiated into T-helper 1 (T H 1) cells, which, once activated by APCs, release pro-inflammatory cytokines and drive the inflammatory process, or T H 2 cells, which control and mitigate inflammation by secreting anti-inflammatory mediators. Macrophages are predominantly activated by T H 1 cells, and exhibit increased phagocytic activity, produce cytokines and release toxic mediators, such as nitric oxide, that can directly damage Schwann cells and contribute to axonal damage. In addition, pro-inflammatory cytokines, such as tumour necrosis factor (TNF) or IFNγ, are locally produced by B cells, and contribute to demyelination and axonal damage. Antibodies can mediate demyelination by antibody-dependent cellular cytotoxicity, can block epitopes that are functionally relevant to nerve conduction, and can activate the complement system via the classical pathway, yielding pro-inflammatory mediators and the lytic C5b 9 terminal complex. Termination of the inflammatory response is mediated, in part, by macrophages through induction of T-cell apoptosis and release of anti-inflammatory cytokines, such as IL 10 and transforming growth factor (TGF)-β. Myelin proteins, such as P0, P2, myelin basic protein (MBP), myelin-associated glycoprotein (MAG), connexin 32 (Cx32) and glucagon-like peptide (GLP), are putative target antigens in CIDP. follow up of patients in this trial and in a previous retrospective study suggest that corticosteroids induce a higher remission rate and longer remission times than IVIg 109,110. Consequently, cautious use of corticosteroids might be considered, even for patients with diabetes and mild disease. Options for corticosteroid therapy are oral dexametha sone pulse treatment, intravenous methylprednisolone, or a daily oral regime of prednisolone. Current evidence indicates that pulse therapy provides faster onset of benefit and has a better safety profile than the other options 111 : pulse therapy has been associated with significantly lower rates of sleeplessness and moon facies. In a comparison of pulse therapy with daily oral regimens, a significantly lower risk of weight gain >3 kg and a near-significant reduced risk of raised blood sugar were also observed with pulse therapy 111. Long-term treatment with any of these corticosteroids requires regular monitoring. For IVIg treatment, dosage requirements are variable, and should be titrated to the minimal effective dose in each case. Treatment withdrawal trials should be considered at regular intervals 112 ; trials can be attempted yearly, although the frequency depends on individual patients circumstances. Response rates to treatment for CIDP are comparable between patients with and without diabetes 15,113. However, the benefit in patients with diabetes might be limited by underlying DSPN 85. Indeed, more-severe baseline axonal loss predicts a poor treatment response 114. Efforts to identify electrophysiological predictors of treatment response in CIDP initially yielded negative results 115, although a report published in 2015 suggested NATURE REVIEWS NEUROLOGY VOLUME 13 OCTOBER

8 that fulfilment of the AAN or EFNS PNS diagnostic criteria for CIDP is associated with a higher chance of a therapeutic response 116. Another study published in 2015 showed that greater demyelination in patients with CIDP and diabetes was associated with better treatment responses 117. This finding implies that in order to be diagnosed with CIDP, patients with diabetes should fulfil at least two EFNS PNS CIDP criteria for demyelination, and that more-stringent cut-off points are needed to define demyelination. In fact, cut-offs of 30% beyond the normal values for distal latency, motor conduction velocity and minimum F wave latency enabled better prediction of a CIDP treatment response in patients with diabetes than did lower cut-offs, and cut-offs of only 10% produced a comparable level of prediction in patients without diabetes. For CIDP in patients with diabetes that is refractory to initial monotherapy, subsequent treatment should be identical to that for refractory CIDP without diabetes. Plasma exchange and combination therapies (in particular, IVIg or plasma exchange with corticosteroids) should be considered 10,118. No evidence currently supports the use of immunosuppressants in CIDP 119,120 in any context. Randomized controlled trials in patients with CIDP have demonstrated no efficacy of azathioprine, methotrexate or IFNβ1a Furthermore, evidence for efficacy of cyclosporin 125,126, mycophenolate mofetil 127, natalizumab 128 and alemtuzumab 129,130 is limited. A phase III clinical trial of fingolimod (NCT ) was terminated early owing to futility (reported by Hartung et al., American Academy of Neurology, Boston, 2017). Immunosuppressants are probably less justifiable for patients with diabetes and pre-existing axonal loss as a result of DSPN, as the potential for neurological recovery in these patients is lower. Nevertheless, if a patient with diabetes is severely affected by recent-onset CIDP, immunosuppression should be considered. In a series of patients with treatment-refractory CIDP, the responder rate to cyclophosphamide was >70%, and improvements in quality-of life measures were seen; the treatment regimens that were used included high-dose cyclophosphamide without stem-cell rescue 131,132. This treatment requires exceptional consideration after obtaining carefully informed consent; all necessary precautionary measures the use of antibacterial, antifungal and uroprotective agents must be taken and the patient must be closely monitored for adverse effects, particularly infectious, haematological and urological effects. Rituximab might be an alternative therapeutic option, as it has been used successfully in case series of patients with refractory CIDP, including patients with diabetes A few case reports have shown that autologous haemato poietic stem cell transplantation can be effective for refractory CIDP , and in a multicentre case series of 11 Swedish patients with CIDP, this procedure induced remission in eight patients for the 2 year observation period 141. Despite these encouraging data, a careful risk benefit analysis should always be performed before embarking on this aggressive approach, which can cause considerable morbidity and even death. Prognosis Few studies have reported on the outcome of treatment in patients with CIDP and diabetes. Some useful information can be derived from those that have been published, but most were small, had a retrospective design, and used assessment tools that are now considered outdated. In two separate case series from the USA and Italy, all 14 patients with diabetes and CIDP responded to immunomodulatory and/or immunosuppressive treatments 15,16. In the US study, modified Rankin scale (mrs) scores improved by 2 points in six patients, and by 1 point in one other 15. Similarly, in the Italian study, mrs score improved by 2 points in six patients and by 1 point in another, and Neurological Disability Scores also substantially improved in all patients, both overall and in the motor component 16. In a comparison of 14 patients who had CIDP and diabetes with 60 patients who had idiopathic CIDP, the proportion of patients who responded to IVIg, corticosteroids, plasma exchange or cyclophosphamide measured with the mrs was similar in the two groups 85. However, improvements in MRC (Medical Research Council) and mrs scores were significantly greater among the patients with idiopathic CIDP than among patients with CIDP and diabetes 85. Baseline characteristics of the groups were comparable, except that the patients with diabetes were older, and had a higher frequency of gait imbalance and a greater degree of axonal loss. In a series of nine patients with CIDP from a cohort of 100 patients with diabetes, only two of four who were treated with corticosteroids responded, and none of three who received IVIg responded 18. In another comparison of CIDP patients with and without diabetes (n = 6 and n = 8, respectively), all patients responded to corticosteroids alone or in combination with plasma exchange, IVIg or azathioprine. The patient groups were comparable in all respects except age and the degree of axonal loss, both of which were greater in patients with CIDP and diabetes 86. Few prospective studies of the treatment of CIDP in patients with diabetes have been published. In one open-label, 4 week study, 26 patients who met the highly specific AAN diagnostic criteria for CIDP 142 were treated with IVIg at a standard dose of 400 mg/kg daily for 5 days 143. For 21 of the 26 patients (80.8%), the Neuropathy Impairment Score improved by >5 points by week 4 after treatment. For one in six patients, scores improved by the end of the 5 day infusion, and lowerlimb function had improved significantly 4 weeks after treatment. Only the presence of conduction block predicted a treatment response and absence of relapse; age, sex, duration of diabetes, CSF protein levels and glycated haemoglobin did not. Of the five patients who did not respond to IVIg, two improved slightly with plasma exchange and corticosteroids in combination. The global response rate to all treatment, therefore, was 88.5%. Six patients had a relapsing course, but five of these individuals responded to IVIg or plasma exchange. In another prospective study, 16 patients with diabetes and a demyelin ating neuropathy that fulfilled AAN diagnostic criteria for CIDP 144 were followed up for 40 months. Nine had type 1 diabetes, and seven had type 2 diabetes. 606 OCTOBER 2017 VOLUME 13

9 All participants had experienced proxi mal and distal limb weakness for 4 12 months before their initial examination, and all were treated with IVIg. Neuropathy Impairment Scores improved for 14 patients (87.5%), and mean summated conduction velocities improved in the upper and lower limbs, although motor and sensory axonal loss progressed. The largest study to date is a retrospective comparative analysis of 134 patients with CIDP: 67 with diabetes and 67 without diabetes 113. In this analysis, the diagnosis of CIDP was based on expert opinion and use of diagnostic criteria that did not include electrophysiological measures 145, and were previously found to be highly specific but less sensitive than the EFNS PNS criteria for CIDP 27. Assessment of response was based on a combination of physician and patient evaluations, including the Toronto Clinical Neuropathy Score (TCNS). Patients with CIDP and diabetes had significantly higher TCNSs and a greater extent of proximal weakness than those without diabetes, but significantly fewer of these patients received treatment (57% versus 93%, P <0.0001). No overall difference was seen in the responses to treatment with IVIg, corticosteroids, plasma exchange, azathioprine or mycophenolate mofetil, although the patients with diabetes were less likely to respond to IVIg (52%, compared with 87% for patients without diabetes, P <0.0001). In the whole cohort, only a shorter duration of neuropathy had a favourable influence on response; age, the presence or duration of diabetes, the severity of neuropathy, and the degree of axonal loss had no impact. Besides its retrospective design, this study was limited by the fact that validated functional scales for evaluation of CIDP, such as the Overall Neuropathy Limitations Scale and the Inflammatory Rasch-Built Overall Disability Score, were not used, the possibility that cases of DLRPN were misclassified as CIDP because electrophysiology was not mandatory for diagnosis, and the possibility of selection bias owing to the relatively low treatment rate among patients with CIDP and diabetes. Also of note, the response rate among the 67 patients with CIDP without diabetes selected from a cohort of 1,950 people with CIDP to be age-matched and sex-matched to the patients with diabetes was higher than that reported in a previous therapeutic trial 146, but was comparable to the rate reported for treatment-naive patients with CIDP 147. Managing other inflammatory neuropathies Radiculoplexus neuropathies. Treatments for painful DLRPN and DCRPN are mainly symptomatic. No evidence currently indicates a benefit of immunotherapy for these conditions in diabetes 83, and trials might be difficult to conduct owing to the heterogeneity of the presentation. For patients who have extreme pain that is refractory to analgesics, intravenous steroids could be considered in selected cases, but do not seem to hasten recovery of strength 83. Small case series, including fewer than 20 patients in total, have suggested that IVIg is beneficial, but no evidence has been generated in randomized controlled trials 83. Similarly, case series indicate that plasma exchange is beneficial, but these studies also included fewer than 20 patients in total. Comparative analysis has suggested that untreated patients eventually improve to a similar degree to those receiving IVIg or plasma exchange 148. Furthermore, the clinical descriptions of some patients who responded to treatment imply that they had CIDP rather than DLRPN 149. Descriptions of the same treatment being ineffective also contradict the positive findings 150. Owing to these uncertainties, no evidence base supports any treatment for diabetic radiculoplexus neuropathies, particularly considering that corticosteroids impair glycaemic control in this population 151. Distinguishing CIDP from microvasculitis can be difficult 87, but in practice, patients for whom either diagnosis is reasonable should be considered to have CIDP and treated accordingly. A proportion of patients with clinically typical CIDP without diabetes do not fulfil electrophysiological criteria for demyelination 27,145, so denying these patients treatment is inappropriate. Furthermore, the technical expertise required for an adequate diagnostic neuropathology report greatly limits the use of nerve biopsies in this urgent setting, so the clinical and electrophysiological findings are often relied on to make a therapeutic decision. Nevertheless, the possibility of monophasic disease must be kept in mind to avoid unnecessary overtreatment and therapeutic escalation 152. Vasculitic multiple mononeuropathy. If vasculitic multiple mononeuropathy is suspected, an exhaustive diagnostic work up for vasculitic neuropathy is indicated, and if the work up is negative, nerve biopsy preferably to sample a clinically affected sensory nerve should be performed for pathological confirmation of vasculitic multiple mononeuropathy. Patients with and without diabetes should receive identical treatment for proven or probable vasculitic multiple mononeuropathy, which is the same as the treatment for nonsystemic vasculitic neuropathy without diabetes 153. Conclusions and practice implications The most common neuropathy in patient with diabetes is DSPN, but other neuropathies including radiculoplexus neuropathies, vasculitic multiple mononeuropathies and CIDP can occur and must be diagnosed because they require different treatment approaches. DLRPN and DCRPN are generally easy to recognize, although they can be painless and extensive, making their identification more difficult. The course of these conditions is monophasic, and the prognosis is generally favourable. CIDP is the most treatable neuropathy that can occur in diabetes. The frequency of CIDP among patients with diabetes remains controversial, but the rapidly progressive disability caused by CIDP warrants awareness among health-care professionals of the association and the potential for immunomodulatory therapy. The diagnosis of CIDP in diabetes is primarily made clinically, and electrophysiological measurements in CIDP with dia betes are less straightforward than in CIDP without diabetes. Nevertheless, currently recommended cut-offs for demyelination remain highly specific for CIDP in patients with diabetes 9. No diagnostic criteria offer perfect sensitivity, NATURE REVIEWS NEUROLOGY VOLUME 13 OCTOBER