Article Text

Download PDFPDF

Validity of diagnostic criteria for chronic inflammatory demyelinating polyneuropathy: a multicentre European study
  1. Y A Rajabally1,
  2. G Nicolas2,3,
  3. F Piéret4,
  4. P Bouche3,
  5. P Y K Van den Bergh4
  1. 1
    Neuromuscular Clinic, Department of Neurology, University Hospitals of Leicester, Leicester, UK
  2. 2
    Centre de Référence Maladies Neuromusculaires de l’Enfant et de l’Adulte Nantes-Angers, Centre Hospitalier Universitaire d’Angers, Angers, France
  3. 3
    Fédération de Neurophysiologie Clinique, Hôpital Salpêtrière, Paris, France
  4. 4
    Centre de Référence Neuromusculaire, Service de Neurologie, Cliniques Universitaires St-Luc, Université Catholique de Louvain, Brussels, Belgium
  1. Correspondence to Dr Y A Rajabally, Neuromuscular Clinic, Department of Neurology, University Hospitals of Leicester, Leicester General Hospital, Leicester LE5 4PW, UK; yusuf.rajabally{at}


Background: Diagnostic criteria for chronic inflammatory demyelinating polyneuropathy (CIDP) have variable sensitivity and specificity. Newly published criteria by Koski et al combine clinical and electrophysiological components, either of which suffices to establish the diagnosis. European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) criteria require mandatory electrophysiology, as do other sets of criteria.

Methods: The value of the two above-mentioned sets of criteria, on 151 patients with CIDP, and 162 controls with axonal neuropathy, from four European centres was assessed. Results were compared with Van den Bergh and Piéret’s criteria and those of the American Academy of Neurology (AAN). The utility of more extensive nerve-conduction studies was ascertained.

Results: Koski et al’s criteria had a sensitivity of 63% and specificity of 99.3%. With unilateral, right-sided, forearm/foreleg, four-nerve studies, EFNS/PNS criteria offered a sensitivity of 81.3% and specificity of 96.2% for “definite/probable” CIDP. Van den Bergh and Piéret’s criteria had a sensitivity of 79.5% and specificity of 96.9%. AAN criteria were poorly sensitive (45.7%) but highly specific (100%). “Possible” electrophysiological CIDP as per EFNS/PNS criteria were poorly specific (69.2%). More extensive studies increased the diagnostic sensitivity of EFNS/PNS criteria (96.7%) but reduced the specificity (79.3%).

Conclusions: In our patient populations, the EFNS/PNS criteria were the most sensitive and allowed identification of a highly significantly greater number of patients than Koski et al’s criteria. The latter were comparable in specificity with the “definite/probable” EFNS/PNS electrodiagnostic subcategories. More extensive nerve-conduction studies improved diagnostic yield but resulted in loss of specificity.

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.

Chronic inflammatory demyelinating polyneuropathy (CIDP) is a clinically heterogeneous disorder. In clinical practice, diagnosis relies heavily on electrophysiology, but electrodiagnostic criteria lack sensitivity.1 2 3 4 This may result in misdiagnosis, potentially denying patients effective treatment.

A new set of criteria comprising clinical and electrophysiological components has been reported and was validated in 48 patients by Koski et al.5 These diagnostic criteria represent the only criteria reported so far for which either clinical or electrophysiological requirements are sufficient to establish a diagnosis of CIDP. Their reliability has not so far been ascertained in other larger CIDP patient populations. Guidelines published in 2006 from the European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) had also defined clinical and electrodiagnostic criteria.1 The electrophysiological component within these criteria, essential for establishing the diagnosis, was based on a set originally described by Van den Bergh and Piéret.2 Distal compound muscle action potential duration prolongation, found in one study to be a highly sensitive independent marker of CIDP,6 was used as an additional criterion to those proposed by Van den Bergh and Piéret, to produce the EFNS/PNS electrodiagnostic criteria. A further electrodiagnostic subcategory (“possible CIDP”) was also added, encompassing cases with demyelinating features in a single nerve instead of in two nerves. Subsequently, a high diagnostic sensitivity and high negative and positive predictive values of the EFNS/PNS criteria were demonstrated in one population of CIDP patients.7 This has not been confirmed however more widely, in different populations. The American Academy of Neurology (AAN) criteria8 also require mandatory electrophysiology for the diagnosis.

We evaluate here the sensitivity and specificity of Koski et al’s criteria5 (table 1) and the EFNS/PNS criteria1 (table 2), in a study involving CIDP patients from four European Neurology/Neurophysiology centres. Comparison of nerve-conduction studies of varying extensiveness was also made, as was comparison with the sensitivity and specificity of the electrophysiological part of the American Academy of Neurology (AAN) criteria,8 also essential for the diagnosis (table 3). We also aimed to evaluate the diagnostic value of distal compound muscle action potential (CMAP) duration prolongation6 by comparing the EFNS/PNS electrophysiological criteria with the original criteria of Van den Bergh and Piéret,2 which did not include this parameter.

Table 1

Koski et al’s criteria for chronic inflammatory demyelinating polyneuropathy (CIDP)5

Table 2

EFNS/PNS electrodiagnostic criteria for chronic inflammatory demyelinating polyneuropathy (CIDP)1

Table 3

American Academy of Neurology electrodiagnostic criteria for chronic inflammatory demyelinating polyneuropathy8

Materials and methods

From the four participating centres, Leicester (UK), Paris (France), Angers (France) and Brussels (Belgium), 151 patients with CIDP were included, and their clinical and electrophysiological records were reviewed. Inclusion criteria were progressive, stepwise or relapsing proximal and distal weakness of more than one limb developing over at least 2 months. In addition, we included patients with Lewis–Sumner syndrome, predominantly distal weakness, focal involvement and those with associated concomitant diseases: diabetes mellitus, IgG and IgA monoclonal gammopathy, and IgM monoclonal gammopathy without anti-MAG activity. Exclusion criteria were a history of drug or toxin exposure, genetic neuropathy and the presence of a sensory level or sphincter disturbance. The study was reviewed by our relevant Institutional Boards and did not require Ethics Commitee approval.

In view of the retrospective design, electrophysiological studies had been performed in a non-standardised manner, the number of nerves studied having varied from four to eight. Studies had been performed in the majority of patients pretreatment, although in some cases, due to technical issues and lack of sufficiently extensive earlier evaluations (less than four motor nerves adequately tested), a subsequent, post-treatment study was considered for the current analysis. The extensiveness of the study of arm nerves varied from the distal forearm segment only, to a full-length study up to Erb's point. The CMAPs were evoked from the median nerve (stimulating at wrist, elbow and, in some, axilla, Erb's point and recording at the abductor pollicis brevis muscle), ulnar nerve (stimulating at wrist, below elbow and, in some, above elbow, axilla, Erb's point and recording at the abductor digiti minimi muscle), common peroneal nerve (stimulating at ankle and fibular neck and recording at the extensor digitorum brevis muscle) and tibial nerve (stimulating at ankle and popliteal fossa and recording at the abductor hallucis muscle). All nerve conductions were performed at an ankle temperature of ⩾30°C and palm temperature of ⩾33°C.1 Results were analysed with each laboratory’s range of normal values, and presence of demyelinating range values determined for each relevant parameter. Fulfilment of each set of electrodiagnostic criteria for CIDP was ascertained in each case. For consideration with proximal studies, the EFNS/PNS “probable CIDP” subcategory was modified, redefining conduction block as at least 50% amplitude reduction in all cases (rather than 30%), at Erb's point, for proximal upper limb studies, in accordance with previous evidence.9 10

For the criteria described by Koski et al, patients with Lewis–Sumner Syndrome or monoclonal gammopathy were excluded from analysis, as per these authors’ definitions.5 Clinical records and histories were reviewed in detail to establish fulfilment of the clinical component of these criteria. Electrophysiological criteria fulfilment were established in each case, using available distal (foream and foreleg) studies with AAN criteria cut-offs.5 8

Sensitivity and specificity were ascertained with four different nerve-conduction study protocols (“Protocols I–IV”) for EFNS/PNS criteria. “Protocol I” corresponded to a unilateral right-sided distal-segment four-nerve studies, including cases of “probable” and “definite” CIDP only.2 “Protocol II” considered the same unilateral four-nerve study but also included cases of “possible CIDP.”2 “Protocol III” related to nerve-conduction studies of five to eight motor nerves, with proximal stimulations for at least two, excluding the cases of “possible CIDP.” Finally, “Protocol IV” corresponded to similar more extensive nerve conductions, but including the cases of “possible CIDP.” Fulfilment of Van den Bergh and Piéret’s criteria2 was ascertained using Protocol I. AAN criteria8 were evaluated with all available nerve-conduction results (irrespective of side of study).

To determine the specificity of the different criteria, we utilised the electrophysiological data from controls suffering from distal sensory or sensori-motor axonal neuropathy, from three of our four centres (Leicester, Paris and Brussels). None fulfilled EFNS/PNS clinical criteria for typical CIDP. All had been investigated with nerve-conduction studies evaluating at least four motor nerves in the forearm and foreleg segments. A proportion had been investigated more extensively, with five to eight nerves studied, and at least two upper-limb nerves tested in their proximal segments.

The comparative diagnostic gain in sensitivity, achieved with the use of the more sensitive set of criteria, was calculated using a McNemar test.


We included 54 patients from Leicester, UK, 40 patients from Paris, France, 29 patients from Angers, France and 28 patients from Brussels, Belgium, totalling 151 patients. Patients with CIDP had a mean age at onset of 52.5 years (mean for Leicester: 56.5; for Paris: 49.0; for Angers: 51.8; for Brussels: 50.4). There were 95 males and 56 females. Seven of our 151 patients had not been treated with immunomodulatory therapy, due to mild disease. Of the 144 who were treated, 136 (94.4%) had responded to treatment. Of the 108 CIDP patients who had a documented lumbar puncture result, 83 (76.9%) had a raised cerebrospinal fluid protein level. We utilised a total of 162 controls with chronic distal sensori-motor axonal neuropathy. Thirty-five were recruited from Paris and had been investigated for five to eight nerves with proximal stimulations for at least two. Ninety-five controls were recruited from Leicester. Of those, 47 had been investigated for five to eight nerves, with proximal stimulations for at least two, and 48 had only undergone the basic unilateral right-sided four-nerve distal segment study. Thirty-two controls were recruited from Brussels, all having undergone distal segment four-nerve studies. All controls were used to determine the specificity of AAN8 and Van den Bergh and Piéret’s2 criteria. Controls from Brussels were not included in the calculation of specificity of EFNS/PNS criteria, as their data did not include distal CMAP duration. Only 22 of the controls from Brussels were used to determine the specificity of Koski et al’s criteria, as detailed clinical records were required for that purpose and were unavailable for 10. Controls were older than CIDP patients (mean age: 61.2 years) and consisted of 93 males and 69 females.

Each of our 151 patients fulfilled clinical EFNS/PNS criteria for typical or atypical CIDP. The results for sensitivity are summarised in table 3. Those relating to specifity are detailed in table 4. Koski et al’s criteria5 offered a sensitivity of 63% and specificity of 99.3% versus controls. The total number of CIDP patients evaluated for these criteria was 127, excluding as per their definition cases of Lewis–Sumner syndrome and those associated with monoclonal gammopathy. Of the cases meeting Koski et al’s criteria, 30 fulfilled both clinical and electrophysiological components, 25 fulfilled only clinical criteria, and 25 met only the electrophysiological requirements. The criteria were highly specific (99.3%).

Table 4

Sensitivity of Koski et al’s criteria,5 European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) electrodiagnostic criteria,1 Van den Bergh and Piéret’s electrodiagnostic criteria2 and American Academy of Neurology (AAN) electrodiagnostic criteria8 in patients with a clinical diagnosis of chronic inflammatory demyelinating polyneuropathy (CIDP)

Only CIDP patients from Leicester, Paris and Angers were evaluated for distal CMAP duration prolongation, and had more extensive nerve-conduction studies. Hence, the sensitivity of the EFNS/PNS electrophysiological criteria for CIDP1 in those three centres combined was 81.3%, and specificity was 96.2% using Protocol I. With Protocol II, which also included the “possible CIDP” cases, sensitivity improved to 96.7%. However, this also resulted in increased misdiagnosis of controls as having CIDP, thereby reducing specificity to 69.2%. Sensitivity rose to 96.7% using Protocol III, specificity being of 79.3%. Finally, Protocol IV offered a sensitivity of 99.2%, but the specificity was only 61%. The sensitivity of Van den Bergh and Piéret’s criteria2 was 79.5% for the four centres combined. The combined specificity was 96.9%. There were no significant differences in the sensitivity rates between Leicester, Paris and Angers for the EFNS/PNS criteria, with any protocol. The sensitivity and specificity of Van den Bergh and Piéret’s criteria were also equivalent in all four centres, and comparable with those of the EFNS/PNS with Protocol I, indicating that distal CMAP duration prolongation as an additional parameter to the electrophysiological criteria did not have much impact. American Academy of Neurology (AAN) electrophysiological criteria for CIDP8 showed a sensitivity of 45.7% in our 151 patients, and specificity of 100% versus controls. Again, the figures were comparable in all four centres.

The EFNS/PNS criteria, used with Protocol I, offered a significant improvement in diagnostic sensitivity compared with Koski et al’s criteria. We evaluated both these criteria on 101 of our patients, excluding those with serum paraprotein and with a clinical diagnosis of Lewis–Sumner syndrome, which Koski et al distinguished from CIDP,5 and excluding the patients from Brussels, for whom distal CMAP duration data were not available. Considering these 101 patients, 65 (25/39 (64.1%) from Leicester, 26/40 (65%) from Paris and 14/22 (63.6%) from Angers) fulfilled Koski et al’s criteria, in comparison with 85 (33/39 (84.6%) from Leicester, 33/40 (82.5%) from Paris and 19/22 (86.4%) from Angers), who met EFNS/PNS criteria with Protocol I (McNemar test; p = 0.000002). Forty-one of the 101 patients fulfilled Koski et al’s clinical criteria, and 51 fulfilled the electrophysiological criteria. All 101 cases met EFNS/PNS clinical criteria for typical or atypical CIDP, and 85 fulfilled the electrodiagnostic requirements. The 20 additional diagnoses achieved with the EFNS/PNS criteria were those which did not meet either clinical or electrophysiological criteria described by Koski et al but met the EFNS/PNS clinical criteria for atypical CIDP and its electrophysiological requirements.

Compared with AAN criteria which only identified 56 of 123 cases from the above-mentioned three centres, EFNS/PNS criteria allowed diagnosis of 100 of 123 these patients (McNemar test: p<0.000001). With Protocol I, EFNS/PNS criteria were of comparable specificity with those of Koski et al (p = 0.11).


The diagnostic criteria by Koski et al5 offer the potential convenience and advantage of combining clinical and electrophysiological components, either of which suffices for a positive diagnosis of CIDP. These criteria were derived from classification, performed through generation of an expert consensus diagnosis (the “gold standard”) and regression tree analysis of 150 patients, among which 58 had CIDP, and later validated in 48 other CIDP subjects. However, they exclude patients with Lewis–Sumner syndrome, as well as those with associated monoclonal gammopathy.

Although shown to be of high sensitivity and specificity in one CIDP patient population,7 the electrophysiological criteria for CIDP defined by the EFNS/PNS Guidelines1 have not been evaluated more widely in different patient populations. Importantly, the issue of their specificity, when including cases of “possible CIDP,” has remained uncertain. Furthermore, although utilised within other criteria,9 11 and shown to be reliable with appropriate cut-offs for conduction block and temporal dispersion,10 the usefulness of proximal studies has not been previously assessed with the EFNS/PNS criteria.

The present study demonstrates the high sensitivity of the EFNS/PNS criteria with unilateral four-nerve distal segment studies. Our findings show the added benefit of these criteria in comparison with those of Koski et al (McNemar test; p = 0.000002). The specificity was high overall with Protocol I, and individually in the two centres where controls were recruited. As compared with the AAN criteria, the EFNS/PNS criteria allowed a significant additional number of patients to be diagnosed (McNemar test; p<0.00001, for both Protocols I and III).

We found that more extensive nerve-conduction studies (Protocol III) improved the diagnostic sensitivity from 80.6% to 98.7%. However, this was achieved with a decline in specificity from 96.1% to 79.3%. Use of the “possible CIDP” subcategory improved the sensitivity of unilateral four-nerve distal studies (Protocol II) by 15.4% but unacceptably worsened specificity by 24.6%. With more extensive nerve conductions and inclusion of cases of “possible CIDP” (Protocol IV), additional gain in sensitivity was minimal (2.5%), but loss of specificity considerable (further decline by 18.3%).

The recommended unilateral distal segment four-nerve study protocol allowed in our patients the identification of over 80% of cases with the EFNS/PNS criteria, with a very high specificity, comparable with that of the AAN criteria. Likely reasons for this high sensitivity are the need for abnormality of a single parameter to be present in two nerves, in contrast to the AAN criteria, which require abnormalities of three different parameters, notwithstanding the fact that the EFNS/PNS criteria set higher abnormality thresholds (eg, for motor nerve conduction velocity, distal motor latency and conduction block), thereby allowing preservation of specificity. The usefulness of eight nerve distal segment studies and proximal upper limb stimulations has been demonstrated previously with other electrodiagnostic criteria for CIDP.3 9 High specificity of the proximal component of upper-limb nerve conductions has also been shown,10 at previously advised cut-offs.9 12 Our present study confirms the utility and reliability of these procedures with the EFNS/PNS criteria. Although reduced, specificity remains high at around 80% with such more extensive studies, using Protocol III. One reason for this relative reduction in specificity despite improved sensitivity with increase in number of tested nerves was the severity of axonal loss in some controls, accompanied by a significant motor conduction velocity reduction, distal motor latency prolongation or temporal dispersion, these parameters having standardised cut-off values independent of CMAP. Furthermore, the specificity of temporal dispersion >30% at axilla and Erb's point is suboptimal, as previously demonstrated,10 and this also contributed to false positives.

To our knowledge, a single previous study demonstrated the high diagnostic value of the EFNS/PNS criteria, reporting a sensitivity of 97%.7 The extensiveness of the electrophysiology was uncertain as described in this analysis, in particular regarding uni- or bilaterality of the studies. This could explain the higher sensitivity to that we have found with Protocol I. Furthermore, only symmetrical presentations were included in that study, and this may also have have increased the sensitivity of the EFNS/PNS criteria, as the majority of patients with Lewis–Sumner syndrome do not have diffuse electrophysiological demyelination.13 Finally, this study also excluded cases of CIDP with concurrent illnesses such as diabetes, which may have impacted on the results.

Van den Bergh and Piéret’s original criteria2 independently offered a sensitivity comparable with those of the EFNS/PNS with Protocol I, in our patients. These results suggest that the use of distal CMAP duration prolongation may not contribute to substantially increasing diagnostic yield within the criteria. Considering only cases from Leicester, Paris and Angers, a distal CMAP duration prolongation of >9 ms in any motor nerve contributed to only one additional diagnosis (100 vs 99) from a total of 123 patients, as all the other patients positive for this criterion also showed demyelinating range changes for other parameters. The high diagnostic sensitivity for Van den Bergh and Piéret’s criteria was initially demonstrated in the original series (75%).2 This was also shown in a study of 32 patients from a single centre (71.8%),14 which compared these criteria with those of the Inflammatory Neuropathy Cause and Treatment group criteria (sensitivity of 68.7%)9 and those of the AAN (sensitivity of 48.6%).6 The current findings are otherwise also in keeping with those of another previous analysis, from one of our four centres, which found distal CMAP duration prolongation in only six of 20 (30%) CIDP patients.10

The comparative values of diagnostic criteria for CIDP have rarely been studied in a large number of patients from different centres. Previous studies have highlighted the poor sensitivity of older criteria such as those of the AAN.1 2 3 4 8 Our findings are limited by the difficulties in comparing a set of criteria requiring fulfilment of either clinical or electrophysiological components (Koski et al’s),5 and of another requiring mandatory electrophysiology (EFNS/PNS).1 The methodology was further complicated by the different clinical definitions for “CIDP” as used within the two different sets. However, comparison of the 101 patients from our study, acceptable for evaluation according to the definitions of both sets (excluding cases of Lewis–Sumner syndrome and with associated monoclonal gammopathy), suggests that the EFNS/PNS criteria are more sensitive than those of Koski et al while offering a similar high specificity. In our current analysis, only 51 of the 101 patients met Koski et al’s electrophysiological criteria, while 85 met those of the EFNS/PNS. Koski et al’s clinical criteria fail to pick up atypical cases such as those with distal or asymmetrical involvement or patients without four-limb motor weakness. Considering that diagnostic difficulty may precisely arise in such patients in practice, this is a limitation to their utility. Illustrating this point, only 41 of the 101 patients met Koski et al’s clinical criteria, while they all fulfilled EFNS/PNS criteria for typical or atypical CIDP. In addition, Koski et al’s electrophysiological criteria require at least 75% of excitable motor nerves and >50% of these exhibiting either abnormal distal latency, conduction velocity or F latency. This can be problematic in cases with severe secondary axonal degeneration, or in patients tested for an even number of nerves. Furthermore, here we compared EFNS/PNS criteria with Protocol I, that is using four studied nerves, with Koski et al’s criteria, using all available nerves, which artificially increased the sensitivity of the latter, as all our patients were evaluated for four nerves or more. Also, as proximal weakness in at least one limb was mandatory for fulfilment of Koski et al’s clinical criteria, the use of controls with distal sensory or sensori-motor neuropathy clearly increased the specificity of these criteria. Finally, the present results also demonstrate the possibility of usefully and reliably utilising the EFNS/PNS criteria with more extensive nerve-conduction studies, including proximal stimulations for upper-limb nerves. The poor specificity of the “possible CIDP” electrodiagnostic subcategory found here, however, suggests it should be used cautiously.

Table 5

Specificity of Koski et al’s criteria, European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) criteria with various diagnostic subcategories, Van den Bergh and Piéret’s criteria and American Academy of Neurology (AAN) criteria: controls with a clinical diagnosis of chronic distal sensory or sensori-motor neuropathy



  • Competing interests None.

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