Article Text

Download PDFPDF

Original research
Chronic inflammatory demyelinating polyneuropathy associated with diabetes: a European multicentre comparative reappraisal
  1. Yusuf A Rajabally1,2,
  2. Stojan Peric3,
  3. Mina Cobeljic3,
  4. Saadia Afzal1,
  5. Ivo Bozovic3,
  6. Aleksa Palibrk3,
  7. Ivana Basta3
  1. 1 Inflammatory Neuropathy Clinic, University Hospitals of Birmingham, Birmingham, UK
  2. 2 Aston Medical School, Aston University, Birmingham, United Kingdom
  3. 3 Neurology Clinic, Clinical Center of Serbia, School of Medicine, University of Belgrade, Belgrade, Serbia
  1. Correspondence to Prof. Yusuf A Rajabally, Inflammatory Neuropathy Clinic, University Hospitals of Birmingham, Birmingham B4 7ET, UK; y.rajabally{at}


Introduction The association between chronic inflammatory demyelinating polyneuropathy (CIDP) and diabetes is uncertain despite important diagnostic and management implications.

Methods We retrospectively analysed two European cohorts, totaling 257 patients with ‘definite’ or ‘probable’ CIDP, from Serbia and Birmingham, UK.

Results Diabetes was present at CIDP diagnosis in 25/139 (18%) subjects in the Serbian cohort and in 23/118 (19.5%) in the UK cohort. In both cohorts, diabetes prevalence was higher than local general population prevalence rates (RR: 2.09; 95% CI 1.39 to 2.95 and RR: 2.22; 95% CI 1.46 to 3.17, respectively). Considering typical CIDP only, diabetes prevalence was greater than expected in both cohorts (RR: 2.58; 95% CI 1.60 to 3.82 and RR: 2.68; 95% CI 1.71 to 3.87, respectively). CIDP with diabetes occurred later in life than CIDP without diabetes (58.96 years, SD: 11.09 vs 51.71 years, SD: 16.02; p=0.003) and presented more frequently in the typical form than in patients without diabetes (79.2% vs 61.2%; p=0.02). Baseline Inflammatory Neuropathy Cause and Treatment disability scores were similar in patients with and without diabetes (p=0.90). Proportions of treatment responders were similar in both groups (70% vs 74.9%; p=0.65), as were response amplitudes (p=0.87).

Discussion Our results, both for all CIDP and typical CIDP presentations, support a twofold increased relative risk of diabetes compared with the general population. CIDP with diabetes appears to present older and more frequently in the typical form, as compared with CIDP without diabetes. CIDP with diabetes appears similar to CIDP without diabetes in disability levels at diagnosis and probability, as well as amplitude of treatment response.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The association between diabetes and chronic inflammatory demyelinating polyneuropathy (CIDP) remains to date, controversial. Although earlier non-population-based studies had suggested an up to 11-fold increased risk of CIDP in individuals with diabetes,1 2 subsequent epidemiological studies from the USA3 and Italy4 challenged these findings, by finding no association between the 2 disorders. More recently, results from a study of a large healthcare database in the USA found a ninefold higher prevalence of CIDP in subjects with diabetes, compared with those without diabetes.5 A number of reasons explain the difficulties to be confident, or not, of an association between the two disorders. First, CIDP is a heterogeneous clinical entity and misdiagnosis rates are known to be high. The ‘typical’ form presents with characteristic weakness of proximal and distal weakness of the four limbs and propioceptive sensory loss, most often straightforward to differentiate from common diabetic peripheral nerve manifestations. However, ‘atypical’ forms may present with distal and focal involvement which are clinically difficult to distinguish from diabetic polyneuropathic or radiculoplexoneuropathic presentations.6 It is possible therefore that atypical forms, known to be not infrequent,7 may have contributed to CIDP misdiagnosis and subsequent overdiagnosis of the association. Diabetic polyneuropathy may furthermore present with a degree of electrophysiological conduction slowing which may erroneously lead to a CIDP diagnosis.8 Although adherence to strict existing diagnostic criteria should, to a large extent resolve this problem, severe diabetes as well as renal impairment, which may furthermore coexist, could lead to genuine demyelinating range slowing, compounding diagnostic difficulty. Cerebrospinal fluid (CSF) levels are known to be often elevated in patients with diabetes and cannot be relied on for CIDP diagnosis, and can be a further cause of overdiagnosis. The usefulness of nerve imaging and pathology, in addition, currently remain very much uncertain with recent reports of abnormalities found in diabetic subjects similar to those of inflammatory neuropathy.9

Doubts about the frequency of the association persist and with them, uncertainties in clinical practice, about the importance of considering a CIDP diagnosis in the context of diabetes and of justifying an attempt of immunomodulatory treatment. It is otherwise not currently known whether there may be potential useful clinical pointers which may be of help to increase the index of suspicion, irrespective of the frequency of the association.

We here aimed to study the association of diabetes in patients with CIDP in our patient cohorts, from two different European regions, in relation to local diabetes prevalence rates, age of CIDP onset and CIDP subtype, disability scores at onset, treatment response rates and amplitudes of amelioration.


We retrospectively studied the multinational Serbian CIDP database as well as the electronic hospital records of patients with a diagnosis of at least ‘possible CIDP’, attending the Inflammatory Neuropathy Clinic of the Regional Neuromuscular Service, University Hospitals Birmingham, UK.

The Serbian database includes all patients with CIDP diagnosed from tertiary healthcare centres from Serbia, Republic of Srpska (Bosnia and Herzegovina) and Montenegro. Data on all incident and prevalent cases of CIDP are collected by treating neurologists. The Birmingham Inflammatory Neuropathy Clinic is a UK tertiary regional referral centre for dysimmune neuropathies. Clinic records are systematically populated for diagnostic, treatment and monitoring data. Diagnosis in each individual case was made by a neurologist specialised in neuromuscular disease. Each case was clinically characterised in detail for pattern of motor weakness, disease symmetry, reflexes and sensory dysfunction, so as to establish the CIDP diagnosis and the subtype classification, as per the European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) of 2010.6 Electrophysiology had been performed in each case by a qualified and experienced neurophysiologist, as per routine procedures and fulfilment of electrodiagnostic subtype category, using each laboratory’s normative values, had been established and verified in each patient.

Of all patients aged 18 years or over, only those meeting requirements for a diagnosis of ‘definite’ or ‘probable’ CIDP as per the EFNS/PNS Guidelines were included. This was decided in view of the previously demonstrated poor specificity of the ‘possible’ CIDP subcategory, justifying its exclusion from the current analysis, particularly given the degree of motor slowing that may be observed in diabetic neuropathy.10 We assessed demographics, diagnostic subcategory, CIDP subtype, Inflammatory Neuropathy Cause and Treatment (INCAT) score,11 CSF protein levels and eventual effective treatment(s) administered. INCAT disability scores were determined at baseline and post treatment. We, in combination, and separately, conducted analyses considering all forms of CIDP and the typical form of CIDP, only. We defined the association of CIDP and diabetes, as a pre-existing or concurrent diagnosis of diabetes at the time of the CIDP diagnosis. The diagnosis of diabetes was made as per current international WHO guidelines. In Serbia, the diagnosis was established through serum fasting glucose level, 2 hours glucose tolerance testing (GTT) and/or glycosylated haemoglobin (HbA1C) readings >48 mmol/mol, and in the UK, either through GTT or by two HbA1C readings >48 mmol/mol.

Epidemiological data for diabetes prevalence in Serbia were obtained from available WHO figures.12 The WHO methods for estimating prevalence are derived from data from random national, subnational or community populations with clearly described survey methods and clearly defined diagnostic ascertainment, using statistical modelling.13 Those for England, which we used instead of that of the city of Birmingham given the wide geographical distribution of residence of attending patients, were based on figures provided by Public Health England, UK.14 These use a model using data from Health Surveys for England which create estimates from resident populations (themselves calculated from the Office for National Statistics population projections), and General Practice-registered populations.15

Relative risks (RR) for diabetes and 95% CIs were established for our respective two cohorts. Comparison of proportions were performed by Fisher exact tests, comparison of means by t-tests.


We included a total of 257 patients with a diagnosis of definite or probable CIDP as per EFNS/PNS Guidelines.6 Of those, 139 were from the Serbian CIDP database and 118 from the Birmingham Inflammatory Neuropathy Clinic. Six (2.3%) had ‘probable CIDP’ and 251 (97.7%) had definite CIDP. A majority (166/257; 64.6%) had typical CIDP, and the remainder, atypical forms. Main cohort demographics are detailed in table 1. There were 81 women and 176 men (ratio of 1:2.17). Mean age at onset was 53.09 years (SD: 15.47). There were no differences in the age of onset of the Serbian and UK cohorts (52.49 years (SD: 15.24) vs 53.81 years (SD: 15.77); p=0.50), or INCAT disability score at time of diagnosis (3.76 (SD: 2.16) vs 4.06 (SD: 2.39); p=0.30). However, there were significantly more men in the UK cohort (77.1% vs 61.1%; p=0.007). A significantly larger proportion of the UK cohort had typical CIDP (68.6% vs 58.3%; p=0.026).

Table 1

Characteristics of Serbian and Birmingham, UK, cohorts of patients with definite or probable chronic inflammatory demyelinating polyneuropathy (CIDP)

Of the whole combined series of 257 subjects, 48 (18.7%) had concurrent diabetes at the time of CIDP diagnosis. The majority had type 2 diabetes (43/48; 89.6%). Twenty-five of the 139 patients from the Serbian database (18%) and 23/118 (19.5%) from the Birmingham had diabetes. All patients with diabetes and CIDP fulfilled criteria for definite CIDP. There was no significant difference in rates of diabetes between the two cohorts (p=0.87). The estimated general diabetes prevalence rates were 8.6% in adults in Serbia12 and 8.8% in England.14 The rates of diabetes found in our two CIDP cohorts were hence greater than expected for both (RR: 2.09 (95% CI 1.39 to 2.95) and 2.22 (95% CI 1.46 to 3.17), for Serbian and Birmingham cohorts, respectively). Considering exclusively the typical form of CIDP, the rates of diabetes were also significantly higher in both cohorts (RR: 2.58 (95% CI 1.60 to 3.82) and RR: 2.68 (95% CI 1.71 to 3.87), for Serbian and Birmingham cohorts, respectively).

We compared the 48 patients with CIDP and diabetes, with the 209 patients with CIDP without diabetes. The results are summarised in table 2. The CIDP populations with and without diabetes had similar gender distributions (p=0.73). The proportion of typical CIDP cases was greater in patients with than without diabetes (79.2% vs 61.2%; p=0.02). There was no difference in the proportion of typical CIDP comparing subjects with type 1 and type 2 diabetes (100% vs 83.33%; p=1). Atypical CIDP subtypes were diagnosed in patients with and without diabetes in comparable proportions (Lewis Sumner syndrome (LSS): 6.2% vs 16.3%; p=0.072, pure motor CIDP: 4.2% vs 6.7%; p=0.74, pure sensory CIDP: 2.1% vs 8.1%; p=0.21, distal acquired demyelinating sensory and motor variant (DADS): 8.33% vs 7.65%; p=0.77). Patients with diabetes were older at CIDP onset than those without diabetes (59.13 years (SD: 11.03) vs 51.71 years (SD: 16.02) p=0.003). Thus, only 7.8% of subjects with CIDP aged <50 years at disease onset had concurrent diabetes as compared with 27.5% of those aged >50 years (p=0.0002). The baseline INCAT disability scores were similar in both groups (3.94 (SD: 1.90) vs 3.89 (SD: 2.35); p=0.9). Within the whole combined cohort, 52.1% and 54.5%, of patients with and without diabetes, had undergone a CSF examination. Using a cut-off >0.50 g/L to define elevated CSF protein in the tested subjects, the rates of abnormal protein levels was similar (88% vs 78.1%; p=0.41). Using instead a cut-off >1 g/L, the rates of abnormal levels were also comparable (59.1% vs 51%; p=0.64). The mean CSF protein levels were similar in patients with and without diabetes (1.43 g/L (SD: 0.93) vs 1.31 g/L (SD: 2.04); p=0.77). Follow-up data regarding treatment response were available for 35/48 patients (72.9%) with CIDP and diabetes and 168/209 patients (80.4%) with CIDP without diabetes (p=0.25). The proportion of treatment responders (defined by a 1-point improvement of the INCAT disability score), to any of the therapies administered (immunoglobulins, oral or pulse oral or intravenous corticosteroids, plasma exchanges, immunosuppressants, alone or in combination), was comparable in patients with and without diabetes (70% vs 74.9%, respectively; p=0.65). The amplitudes of INCAT disability score improvement post treatment were also similar in patients with and without diabetes (2.61 (SD: 1.83) vs 2.68 (SD: 1.89), respectively; p=0.87). Corticosteroids were used more frequently in Serbia with 119/139 (85.6%) subjects being treated by pulse or daily oral therapy, versus 32/118 (27.1%) in the UK patients (p<0.001). On the other hand, immunoglobulins were more commonly used in the UK, with 70/118 patients (59.3%) receiving intravenous or subcutaneous treatment, versus 59/139 (42.4%), in Serbia (p=0.0086).

Table 2

Comparison of patients with and without diabetes from the combined Serbian and Birmingham chronic inflammatory demyelinating polyneuropathy (CIDP) cohorts

We in addition, conducted separately, comparative analyses of the diabetic and non-diabetic components of the Serbian and UK CIDP cohorts. No differences were observed between the Serbian and UK diabetic cohorts for age of onset (p=0.94), gender distribution (p=1), proportion of typical cases of CIDP (p=0.29), proportions of LSS presentations (p=0.24), of pure motor forms (p=0.49), of pure sensory forms (p=1), of the DADS variant (p=0.11). There were also no differences of CSF protein levels (p=0.78), INCAT disability score at onset (p=0.50), amplitude of INCAT score amelioration with treatment (p=0.95), or responder status based on INCAT disability score improvement of at least one point (p=0.12). However, none of the subjects with diabetes from Serbia had type 1 diabetes, compared with 5/23 of those from Birmingham (p=0.0197).

No differences were observed between the two non-diabetic cohorts with regards to age of onset (p=0.52) or proportion of typical cases of CIDP (p=0.064). There were however more men in the UK non-diabetic group (77.9% vs 59.6%; p=0.007). Although the frequency LSS and pure sensory forms were comparable (p=0.094 and p=0.21, respectively), pure motor forms and DADS presentations were commoner in the Serbian non-diabetic cohort (p=0.023 and p=0.013, respectively). CSF protein levels were similar (p=0.16). Although INCAT disability scores at onset and amplitudes of INCAT disability score amelioration post treatment were similar in the two non-diabetic groups (p=0.39 and p=0.58, respectively), the proportion of responders, as defined by a one point or more INCAT disability score amelioration, was greater in the UK patients without diabetes than those from Serbia (86.4% vs 67.4%; p=0.004).


The association of diabetes and CIDP has remained elusive with discrepant findings previously reported in the literature. This has however implications from the point of view of diagnosis and treatment, CIDP representing the most common treatable inflammatory neuropathy worldwide and the prevalence of diabetes increasing rapidly.

We here studied two large CIDP cohorts from two distinct European populations, with regards to this association. Our principal finding is a similar twofold increased risk of diabetes in CIDP in both our cohorts. This provides support for an increased index of suspicion of CIDP in patients with diabetes, in clinical practice. The RR levels found did not reach the ninefold figure from the most recent study of a US healthcare database,5 or the 11-fold figure from earlier US non-population-based studies,1 which, in each case involved methodological issues, including possible CIDP overdiagnosis for the former and referral bias for the second. The two previously reported epidemiological studies reporting lack of an association were, for their part, performed in Italy4 and the USA.3 The first relied exclusively on fasting glucose levels or use of antidiabetic medication for diagnosis, with possible underdiagnosis of diabetes. The second was a case-control single-centre study, using institution-registered controls. Both studies found substantially higher CIDP prevalence rates than reported elsewhere. The Italian study described a prevalence of 3.58 per 100 000, which was obtained using the American Academy of Neurology diagnostic criteria. The use of these criteria produced prevalence rates of half or less in other studies. The US study described a threefold higher CIDP than the mean global world prevalence rate, in a small population of about 250 000. This analysis used different clinical and electrodiagnostic criteria for CIDP to those of the currently most widely used EFNS/PNS Guidelines.6 Whether these elements may have played a role in the findings of both these studies is possible. Our results were acquired from study of a large combined cohort from two different patient populations, meeting clinical and electrophysiological criteria for definite or probable CIDP only, as per EFNS/PNS Guidelines, both categories having been previously found highly specific for CIDP versus other neuropathies, including diabetic polyneuropathy.10 16 Local, recent diabetes prevalence general population rates were furthermore used separately for analysis, with similar results.

We also found a twofold increased RR of diabetes, specifically in subjects with typical CIDP. To our knowledge, such an association has not been described previously. This result is of importance as excludes distal and focal forms of CIDP, which may be confused with known diabetic complications such as diabetic polyneuropathy, diabetic amyotrophy or diabetic cervical or lumbosacral radiculoplexus neuropathy. These findings hence confirm that focal and distal CIDP presentations, for which diagnostic uncertainty is greater, do not explain the increased RR of diabetes found for CIDP as a whole. The independent association between the commoner and classical form of CIDP, with diabetes, also indicates the importance, in practice, of adequate detailed clinical evaluation in patients with diabetes and any degree of functional change which may suggest new proximal and distal motor weakness and propioceptive sensory involvement.

Previous epidemiological studies have established increased CIDP prevalence rates with age.17 18 The mean age of onset in our individual and combined cohorts was similar to that described in the literature. We however found that CIDP with diabetes presented on average over 7 years later than CIDP without diabetes, in our cohort. We are not aware of previous similar reports. The association was most frequent in the 55–74 years age group for CIDP onset (representing 68% and 56% of Serbian and UK diabetic cohorts, respectively), with lower frequencies in other age groups, including the older, over 75 years group. In view of small patient numbers involved, specific age-group RR calculations were not performed.

All subjects in the current analysis with diabetes and CIDP fulfilled the EFNS/PNS criteria for definite CIDP, which demonstrates that this most stringent subcategory is not of lower electrodiagnostic sensitivity in patients with diabetes than in non-diabetic individuals. We found similar rates of CSF protein elevation, using traditional (0.50 g/L) and modified (1 g/L) abnormal cut-off values and similar mean raw level values in patients with and without diabetes. In the absence of controls, this is of uncertain value for diagnostic purposes. High CSF protein is well described in diabetes,19 our findings not suggesting a diagnostically helpful, cumulative effect, of diabetes and CIDP on CSF protein level elevation.

We found similar baseline disability in subjects with and without diabetes which suggests that the functional deficits are predominantly due to CIDP in affected patients, rather than in addition to that of diabetes, as has been postulated previously. This contradicts the assumption that part of the functional disability observed in such cases is due to diabetic polyneuropathy, which may lead in some cases to decide to withhold or delay immunomodulatory treatment, considered as of less benefit in this setting.20 Such treatment itself was as frequently effective in our cohorts, with equivalent degrees of INCAT disability score amelioration, in subjects with and without diabetes, highlighting the importance of prompt diagnosis and identical therapeutic management in CIDP with diabetes as would be considered in the absence of diabetes. These findings support previously published experience of comparable responder rates in diabetics.21 22 Although the diabetic CIDP cohort from Serbia had a lower responder rate than that from Birmingham, this was not significant, possibly due to numbers studied. The same findings for the non-diabetic CIDP cohorts reached, on the other hand, statistical significance. These differences may be due to ease of access to treatments, in particular immunoglobulins, as illustrated by intercohort treatment rate differences, differences in effectiveness of treatments used, as well as variations in treatment strategies, all beyond the scope of this study.

The mechanisms promoting the onset of CIDP in patients with diabetes are poorly understood and potentially diverse. Microangiopathy has been long described in diabetic nerves,23 as well as antibody-mediated junctional disruption in endoneurial endothelium in dysimmune neuropathy.24 Similar blood-nerve-barrier alterations were described in familial amyloid polyneuropathy.25 It is possible that modifications of barrier permeability as a result of chronic hyperglycaemia may, in susceptible individuals, facilitate entry of proinflammatory cells and mediators implicated in the immunopathogenesis of CIDP.

Our study has a number of limitations. First, this was a retrospective analysis which used local general prevalence rates, without age, gender and body mass index-matched controls. The numbers of patients studied, although relatively elevated in view of general worldwide CIDP prevalence,26 were limited and may have impacted on our findings. The duration of diabetes pre-CIDP diagnosis was not considered, in the absence of reliable data for a large number of subjects and the known rates of delayed diagnosis of diabetes, in subjects not regularly tested throughout their adult lives.27 CSF analysis was performed at different times in the disease course in only a proportion of patients, reflecting clinical practice. Also, treatments administered varied, and even more so doses and frequencies of administration, as well as combination regimes, related to different practices. We were unfortunately unable to perform, for those reasons, further meaningful analyses in relation to the dose and frequency of administration of first-line agents or to the frequency of immunosuppressive treatment, in diabetic and non-diabetic cohorts. Despite these drawbacks, we believe our analysis provides support for an increased RR of diabetes in CIDP, including specifically, of typical forms, particularly in patients with older onset CIDP. Levels of functional disability, probability and amplitude of treatment response appear however similar to those in patients with CIDP without diabetes.

Further multicentre comparative studies in large cohorts are needed to confirm our findings in CIDP subjects from other parts of the world. Our results, at the current time, emphasise the importance of maintaining a high index of suspicion for the association of CIDP and diabetes in day-to-day clinical practice, as well as the justification of considering all available therapies and appropriate treatment escalation, for management of affected subjects.


The authors are grateful to Dr. Peter Nightingale, Statistician, University Hospitals Birmingham, UK, for his help with statistical analysis.



  • Contributors YAR, SP and IBa: study conception, data collection, design, analysis and writing of the first draft. MC, SA, IBo and AP: data collection, analysis and review of the first draft for important intellectual content.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests YAR has received speaker/consultancy honoraria from CSL Behring, LFB, Grifols, BPL, Octapharma and Kedrion, has received educational sponsorships from LFB, CSL Behring and Baxter and has obtained research grants from CSL Behring and LFB. SP has received consultancy honoraria from Argenx, Pfizer, Mylan, Salveo and ADOC, speaker honoraria from Pfizer, Actavis, Berlin Chemie Menarini, Mylan, Krka, Wowag Pharma and ADOC and has obtained research grants from Kedrion and Octapharma. IBa has received speaker honoraria from Pfizer, Actavis, Berlin Chemie Menarini, Mylan, Krka and ADOC and has obtained research grants from Kedrion and Octapharma.

  • Patient consent for publication Not required.

  • Ethics approval The study was approved and registered by our both respective relevant institutional boards. This was by the Ethical Board of the Neurology Clinic, Clinical Centre of Serbia and the University Hospitals of Birmingham Clinical Audit Office (CARMS no. 15747, December 2019).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available upon reasonable request.

Linked Articles

  • Editorial commentary
    Satoshi Kuwabara Atsuko Tsuneyama Sonoko Misawa
  • Neuromuscular
    Pietro Emiliano Doneddu Dario Cocito Fiore Manganelli Raffaella Fazio Chiara Briani Massimiliano Filosto Luana Benedetti Elisa Bianchi Stefano Jann Anna Mazzeo Giovanni Antonini Giuseppe Cosentino Girolama Alessandra Marfia Andrea Cortese Angelo Maurizio Clerici Marinella Carpo Angelo Schenone Gabriele Siciliano Marco Luigetti Giuseppe Lauria Tiziana Rosso Guido Cavaletti Ettore Beghi Giuseppe Liberatore Lucio Santoro Emanuele Spina Erdita Peci Stefano Tronci Marta Ruiz Stefano Cotti Piccinelli Elena Pinuccia Verrengia Luca Gentile Luca Leonardi Giorgia Mataluni Laura Piccolo Eduardo Nobile-Orazio Italian CIDP Database study group Francesca Gallia Antonietta Topa Alice Todeschini Ilaria Callegari Angela Zuppa Corrado Cabona Erika Schirinzi Patrizia Dacci Mario Sabatelli