Diagnostic Challenges in Chronic Inflammatory Demyelinating Polyradiculoneuropathy

Filip Eftimov; Ilse M. Lucke; Luis A. Querol; Yusuf A. Rajabally; Camiel Verhamme


Brain. 2020;143(11):3214-3224. 

In This Article

Diagnostic Tests

Nerve Conduction Studies

The diagnosis of CIDP relies heavily on identification of demyelinating features on motor NCS. The electrophysiological demyelinating features are not equivalent to classical demyelination as found in nerve biopsy, but rather are markers for functional disruption or slowing of the saltatory conduction of the myelinated axons. Based on the amount and certainty of demyelinating features, this will lead to a definite, probable or possible electrodiagnosis according to the EFNS/PNS 2010 criteria (Van den Bergh et al., 2010). A definite electrodiagnostic diagnosis requires at least two demyelinating features in two different nerves. A probable diagnosis requires two probable blocks or a probable block and one other demyelinating feature in a different nerve, while a possible diagnosis requires one demyelinating feature in one nerve. Recent studies highlighted the importance of correct interpretation of NCS and electrodiagnostic criteria, as this often led to misdiagnosis (Allen and Lewis, 2015; Allen et al., 2018). A pitfall is that the electrophysiological criteria are sensitive to diagnose a demyelinating neuropathy, but they are also fulfilled in other diseases (Table 2).

Another frequent pitfall is severe axonal loss that can have a profound influence on nerve conduction velocity if the largest, fastest conducting axons are involved. For this reason, several criteria have additional fulfilments for the compound muscle action potential (CMAP) amplitude (Van den Bergh et al., 2010). In general, demyelinating features should be identified with caution in measurements with CMAP amplitudes <1 mV, with particular focus on the clinical context (Van Asseldonk et al., 2005). Recordings of more proximal muscles, such as the flexor carpi radial muscle to test the median nerve, can sometimes be of added value in case of profound distal axonal loss.

Demyelinating features in segments prone to compression or entrapment should not be considered as supportive for CIDP. The most important reason is that compression or entrapment itself can give demyelinating features. More severe demyelination was not observed at entrapment sites as compared with that observed at other nerve segments in CIDP (Padua et al., 2004; Rajabally and Narasimhan, 2011b).

Interpretation of CMAP amplitude reductions as conduction blocks should be done with care in trajectories of nerves that are adjacent to each other, as co-stimulation and co-registration may occur. Moreover, physiological anastomoses, such as the Martin-Gruber anastomosis, have to be considered. These reasons should be ruled out prior to concluding that (apparent) CMAP reductions are due to conduction blocks, especially in the lower arm segment of the median or the ulnar nerve.

Both severely lengthened minimal F-wave latencies and absent F-waves may be classified as demyelinating features, but these findings are not specific, especially in case of the fibular nerve (Puksa et al., 2003; Argyriou et al., 2006; Pastore-Olmedo et al., 2009). However, F-wave analysis may be of diagnostic utility when more distal NCS are found normal in presence of clinical features suggestive of the diagnosis (Rajabally and Varanasi, 2013).

The distal CMAP duration criterion in the guideline has been debated as it does not take differences in filter settings between centres into account (Isose et al., 2009), which can greatly influence the duration (Rajabally et al., 2012; Mitsuma et al., 2015). However, when cut-off values are adjusted based on the filter settings used in the individual centre, it is a useful criterion that can aid diagnosis with limited testing.

Sensory NCS are not included in the electrodiagnostic criteria but can support sensory involvement and may be of value as one of the supportive criteria in the ENFS/PNS guidelines. Currently, a normal sural with abnormal median (excluding median neuropathy at the wrist from carpal tunnel syndrome) or radial sensory nerve action potential (SNAP) and/or a conduction velocity of <80% of lower limit of normal (<70% if SNAP amplitude <80% of lower limit of normal) are included as supportive criteria (Van den Bergh et al., 2010). Additional review of the literature does not justify a more prominent position for sensory NCS in the guidelines and suggests that the evidence for the currently included supportive criteria, is limited with particularly low sensitivity levels (Kimura et al., 1986; Kincaid et al., 1988; Tamura et al., 2005; Rajabally and Narasimhan, 2007; Bragg and Benatar, 2008; Rajabally and Samarasekera, 2010). Patients with clinically pure sensory involvement often also have motor abnormalities at NCS that can lead to the diagnosis of CIDP. However, an uncertain proportion of these patients with pure sensory involvement do not fulfil the electrophysiological criteria for at least possible CIDP, as these are based on motor NCS (Van den Bergh et al., 2010). Some studies have proposed alternative diagnostic criteria for this specific CIDP phenotype (Ayrignac et al., 2013).

In clinical practice the order and extensiveness of NCS vary widely. If too few nerve segments are tested, this may lead to underdiagnosis. The guideline advises to first test the median and ulnar nerve at one forearm and the fibular and tibial nerve of one lower leg (Van den Bergh et al., 2010). However, in case of CIDP suspicion, there are arguments to initiate the study with the median and the ulnar nerves up to Erb's point, including F-waves, as demyelinating features may more often be found in the arms, including the more proximal (above) elbow to axilla and axilla to Erb's point segments, than in the legs (Rajabally et al., 2005; Rajabally and Narasimhan, 2011a; Lucke et al., 2019b). CMAP amplitudes of distal leg muscles are often too low for proper interpretation, while the proximal parts of the leg nerves are not accessible for NCS. Moreover, proximal CMAP amplitude reductions and temporal dispersion in the legs should be interpreted more stringently, taking into account physiological phenomena, especially in the tibial nerve. One study demonstrated that conduction blocks at the axilla and Erb's point were highly specific for CIDP and that proximal investigations improved the sensitivity of the diagnostic criteria (Rajabally and Jacob, 2006). This study defined a block at Erb's point as an amplitude reduction of >50% between wrist and Erb's point (Rajabally and Jacob, 2006). However, determining a block at Erb's point may be challenging, with a higher risk of submaximal stimulation at Erb's point due to the depth of the nerves, and no more proximal sites to verify CMAP amplitude reduction. Optimal stimulation at Erb's point is especially of importance, as CMAP amplitude reductions in normal control subjects may be substantial over longer arm nerve trajectories due to physiological temporal dispersion (Johnsen et al., 2006).

CSF Examination

Elevated protein in the CSF with normal leucocytes is found in up to 90% of patients with typical CIDP and is thought to be one of the hallmark features of the disease (Dyck et al., 1975; Prineas and McLeod, 1976; McCombe et al., 1987). In atypical CIDP variants such as the asymmetric subtype, protein elevation might be less pronounced, or absent (Rajabally and Chavada, 2009). Elevation of the CSF protein is not specific for CIDP and especially patients with diabetes mellitus or hereditary demyelinating neuropathies (CMT1) can have slightly elevated protein levels (Bouche et al., 1983; Kobessho et al., 2008). To prevent misdiagnosis, it was recently suggested to increase the cut-off value for CSF protein to 0.6 g/l in patients older than 50 (Breiner et al., 2019). If an elevated leucocyte count (>10/mm3) is found, infections or malignancies should be considered. However, slightly elevated leucocyte counts (>10 mm3) have been reported in up to 11% of CIDP patients (van Doorn et al., 1991; Press et al., 2003; Lucke et al., 2018), meaning that this does not automatically exclude the diagnosis. One study found that 8 of 14 (57%) patients with elevated leucocytes (>10 mm3) had a (sub)acute onset of disease and that leucocytes spontaneously decreased over time (Lucke et al., 2018).


MRI and nerve ultrasound can be a valuable addition in the diagnostic work-up, as proximal segments such as the proximal part of the brachial plexus and the lumbosacral plexus can be assessed, while NCS cannot study these regions. In both techniques, one of the main parameters is nerve hypertrophy. It is noteworthy that this is not an exclusive phenomenon for acquired inflammatory neuropathies such as CIDP and may be seen in other relatively prevalent diseases such as diabetes mellitus, hereditary demyelinating neuropathies and neuralgic amyotrophy (Breiner et al., 2017; Padua et al., 2018; van Rosmalen et al., 2019).

Multiple studies were carried out to evaluate MRI in CIDP with widely varying results: nerve hypertrophy was found in 37–100% of cases (Tazawa et al., 2008; Sinclair et al., 2011; Lozeron et al., 2016; Goedee et al., 2017a; Jongbloed et al., 2017), hyperintensity ranged from 56–100% (Adachi et al., 2011; Sinclair et al., 2011; Shibuya et al., 2015; Goedee et al., 2017a; Jongbloed et al., 2017) and enhancement after gadolinium administration was reported in 0–69% (Midroni et al., 1999; Adachi et al., 2011; Goedee et al., 2017a). Most cohorts had small sample sizes and consisted of prevalent cases with a typical presentation, who met the electrophysiological criteria for definite CIDP. Only few studies included a clinically relevant control group, leading to less generalizable results for the daily practice where distinguishing CIDP from its differential diagnoses is often difficult. A recent study has evaluated the diagnostic performance of MRI in differentiating CIDP and multifocal motor neuropathy from disease controls (segmental spinal muscular atrophy) and healthy control subjects (Oudeman et al., 2020). In that study, intra- and inter-observer agreement for qualitative assessment of nerve hypertrophy and hyperintensity as scored on STIR varied widely. This was also the case for qualitative scoring on magnetic resonance neurography, a newer sequence technique.

Several studies showed changes in diffusion tensor imaging in CIDP, a technique that enables quantitative measurements of water diffusivity within nerve tissue, but the diagnostic contrast seems relatively small and may not be useful in clinical practice (Kakuda et al., 2011; Markvardsen et al., 2016; Kronlage et al., 2017; Oudeman et al., 2020). Given the rapid development of MRI techniques, the aim may be to develop other preferably quantitative measures suited to peripheral nerve tissue in health and disease.

Increases in nerve cross-sectional area on nerve ultrasound have been reported in several studies. Nerve enlargement was found in 69–100% of CIDP patients (Matsuoka et al., 2004; Zaidman et al., 2009; Sugimoto et al., 2013). Studies showed that the CSA of the brachial plexus and the median nerve were the most adequate measurements to distinguish between CIDP and axonal neuropathy (Grimm et al., 2014; Goedee et al., 2017b). A recent single centre study provided cut-off values based on the upper limits for axonal neuropathies and found a 100% specificity for enlargement of one or more segments of the median nerve or the cervical trunci/roots (Goedee et al., 2017b). However, also in this study, only patient controls with a clear-cut diagnosis of axonal neuropathy or amyotrophic lateral sclerosis were studied. A more recent single centre study that included 100 patients clinically suspected of an acquired inflammatory neuropathy found a high sensitivity (97%) for nerve ultrasound, but the specificity was lower than previously reported (69%) (Herraets et al., 2020). In addition, ultrasound identified patients that responded to treatment but did not meet the electrodiagnostic criteria (Herraets et al., 2020).

A good inter-observer variability of the nerve ultrasound was shown in a multicentre study, including acquired inflammatory neuropathies (Telleman et al., 2019). An advantage of nerve ultrasound over MRI is that nerve ultrasound is a relatively easy, quick, patient-friendly tool.

Evoked Potentials

SSEP and triple stimulation technique (TST) may be of help in diagnosing CIDP. SSEPs are used to assess the functioning of the whole sensory pathway, including the nerve roots. The evidence that supports the use of SSEP in diagnosing CIDP is limited. Proximal sensory nerve involvement as investigated with SSEP was found in 38–100% of CIDP cases (Pineda et al., 2007; Yiannikas and Vucic, 2008; Tsukamoto et al., 2010; Salhi et al., 2014). Sample sizes of these studies were often small. None of the studies investigated the diagnostic utility of SSEP in treatment-naïve patients or with the use of SSEP as a primary investigation. The role of the SSEP in diagnosing CIDP is clearly limited if the electrodiagnostic criteria are met. However, studies showed that in up to 100% of patients with a pure sensory presentation, who did not meet the electrophysiological criteria, SSEP showed involvement of the nerve roots (Sinnreich et al., 2004; Ayrignac et al., 2013). In clinical practice, SSEP should be considered in patients with predominant sensory ataxia and areflexia, if the electrodiagnostic criteria are not fulfilled.

Another potential method to overcome the challenges of assessing demyelination in proximal parts might be motor evoked potentials. Several explorative studies using motor evoked potentials in CIDP showed (very) prolonged peripheral conduction times and, less often, changes in central conduction times, but formal diagnostic accuracy studies have not been performed (Takada and Ravnborg, 2000; Pineda et al., 2007). TST is a specialized diagnostic tool to examine the nerve roots proximal of Erb's point, and includes the use of motor evoked potentials. It may demonstrate proximal motor conduction blocks, even if patients do not meet the electrodiagnostic criteria (Attarian et al., 2015; Cao et al., 2018), indicating that TST can probably increase sensitivity. However, specificity is yet unknown and there is very limited experience with this technique.

Nerve Biopsy

Whether nerve biopsy has additional value in diagnosing CIDP has long been a matter of debate. Usually, the nerve selected for biopsy is the sural nerve, as it is easily accessible. The pathological findings are segmental demyelination and remyelination, onion bulb formation and inflammatory infiltrates (Krendel et al., 1989). These findings were reported in 48–71% of the biopsies in patients that met diagnostic criteria (Barohn et al., 1989; Krendel et al., 1989; Molenaar et al., 1998; Bouchard et al., 1999). However, sample sizes of the studies were small and few studies compared biopsy results from CIDP patients with disease controls. One study suggested that none of the biopsy findings were specific for CIDP, as these findings were also found in vasculitis, axonal and demyelinating hereditary neuropathies and monoclonal gammopathies (Krendel et al., 1989). Other studies also showed that nerve biopsies failed to differentiate between CIDP and axonal neuropathies or diabetic neuropathies (Molenaar et al., 1998; Uncini et al., 1999; Bosboom et al., 2001). Some studies suggested that there might be some value in atypical CIDP cases (Vallat et al., 2003; Ayrignac et al., 2013). The recent discovery of specific nodal and paranodal abnormalities in the ultrastructural analysis of the nerve in patients with antibodies against node of Ranvier cell adhesion molecules may have diagnostic utility in specific patients (Koike et al., 2017; Vallat et al., 2017; Uncini and Vallat, 2018). Nerve biopsy is considered invasive and will lead to persisting sensory loss in most patients, while persisting pain, infections and dysaesthesias have also been reported in a minority of patients (Gabriel et al., 2000; Ruth et al., 2005).


Autoantibody search has been an important topic of research in CIDP in the last decades, but just recently, subgroups of CIDP patients with antibodies targeting the nodes of Ranvier and paranodal regions have been described (Querol et al., 2017a; Vural et al., 2018). The discovery of these antibodies, associating with antibody-specific clinical features, boosted interest in the role of antibodies as diagnostic and prognostic biomarkers. Up to 25% of patients with CIDP show evidence of circulating autoantibodies targeting antigens of the peripheral nerve structures, including nodal and paranodal regions (Querol et al., 2017b; Vural et al., 2018; Broers et al., 2019). Autoantibodies specifically targeting the nodal and paranodal regions are found in ~10% of CIDP patients. Antibodies to neurofascin 155 (NF155) are the most frequent, while antibodies to neurofascin 140 (NF140) and neurofascin 186 (NF186), contactin-1 (CNTN1) and contactin-associated protein 1 (CASPR1) are less common (Querol et al., 2017a; Vural et al., 2018; Bunschoten et al., 2019). Most are of the IgG4 isotype. A systematic review reported a very low sensitivity of these autoantibodies but specificities of 100% in CIDP (Hu et al., 2018). Specific clinical phenotypes have been described in patients with autoantibodies to the nodes and paranodes (Table 3). In general, antibodies to nodal and paranodal proteins, regardless of the autoantibody, associate with a subacute onset and more progressive CIDP phenotypes, initially often classified as Guillain-Barré syndrome, and poorer responses to immunoglobulins than patients without these autoantibodies (Hu et al., 2018; Vural et al., 2018). The clinical significance of antibodies other than those targeting nodal and paranodal proteins is unclear.

The optimal technical approach to test for nodal/paranodal antibodies is currently under investigation. However, most published studies use cell-based assays (either immunocytochemistry or flow-cytometry) performed with HEK293 cells transfected with the human recombinant protein target of the antibodies, ELISAs using human recombinant proteins as the protein substrate or immunohistochemistry on teased-nerve fibres to detect the typical nodal or paranodal staining. Considering that the detection of these autoantibodies usually leads to the use of therapeutic algorithms that includes therapies (e.g. rituximab) that are not first line therapies in CIDP, it seems reasonable to try to confirm the detected antibodies with at least two different techniques to increase diagnostic specificity (Martín-Aguilar et al., 2020).

Two independent series, in which in a total of 113 patients were included, described six patients fulfilling the diagnostic criteria for CIDP, in which anti-MAG antibodies were detected in the absence of IgM paraproteinaemia. These patients presented and progressed similar to anti-MAG-positive patients with IgM paraproteinaemia (Sakamoto et al., 2017; Pascual-Goñi et al., 2019) and two of them developed detectable paraproteinaemia years after disease-onset, suggesting that a subset of patients classified as distal CIDP could indeed be patients with early anti-MAG-positive monoclonal gammopathy of undetermined significance related polyneuropathy (MGUSP) (Allen and Lewis, 2015). It is important to take this into account to avoid misdiagnosis, especially in patients with the distal phenotype.

Testing for antibodies should be considered in treatment unresponsive patients, especially in the presence of atypical symptoms such as a subacute onset, severe ataxia, pain or a tremor. In case of the distal phenotype and treatment unresponsiveness, M-protein reanalysis and anti-MAG antibodies may be considered. International, multicentric studies are currently underway, which focus on standardization of measuring autoantibodies, but also on providing better estimates of the diagnostic accuracy of autoantibody testing in patients suspected of CIDP. If autoantibody testing is shown to be widely reliable and reproducible, it can be introduced for standard clinical practice.