The Diagnostic Accuracy of the Second Revision of the European Academy of Neurology/Peripheral Nerve Society Guideline for Diagnosing Chronic Inflammatory Demyelinating Polyradiculoneuropathy

ABSTRACT

Background

Diagnosing chronic inflammatory demyelinating polyneuropathy (CIDP) relies on clinical and electrodiagnostic features. In 2021, the European Academy of Neurology/Peripheral Nerve Society (EAN/PNS) revised the CIDP guideline, superseding the 2010 European Federation of Neurological Societies (EFNS)/PNS guideline. This study compares the diagnostic accuracy of both guidelines and evaluates the impact of specific criterion changes and nerve conduction study (NCS) extensiveness.

Methods

Patients with CIDP were included if they fulfilled the 2010 EFNS/PNS electrodiagnostic criteria for at least possible CIDP or had a presumptive clinical CIDP diagnosis with objective treatment response. Matched controls were patients with suspected chronic immune‐mediated neuropathy who received a diagnosis other than CIDP. Diagnostic accuracy was evaluated using two thresholds: one including possible CIDP and one restricted to (probable/definite) CIDP. The impact of electrodiagnostic criteria changes and NCS extensiveness on diagnostic performance was assessed.

Results

We included 339 patients and 339 matched controls. Sensitivity of the 2010 EFNS/PNS guideline was 92% when including possible CIDP and 91% for probable/definite CIDP, with specificities of 67% and 78%, respectively. For the 2021 EAN/PNS guideline, sensitivity was 94% when including possible CIDP and 83% for CIDP, with specificities of 66% and 86%, respectively. Differences were due to sensory criteria and changes in conduction block criteria. Unilateral NCS would miss 77 possible CIDP diagnoses.

Conclusion

Sensitivity and specificity were comparable when including possible CIDP, but the 2021 EAN/PNS guideline showed lower sensitivity and higher specificity for CIDP alone. Modifying criteria, including removal of sensory abnormalities, and more extensive NCS enhanced accuracy.

1. Introduction

The diagnosis of chronic inflammatory demyelinating polyneuropathy (CIDP) relies on clinical, electrodiagnostic, and sometimes imaging or laboratory features. The first revision of the European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) guideline for CIDP [1], published in 2010, was widely accepted and used due to its high diagnostic accuracy [2, 3]. However, misdiagnosis of CIDP remains common, leading to over‐ or undertreatment, especially in patients with a CIDP variant [4, 5, 6]. In 2021, the European Academy of Neurology/Peripheral Nerve Society (EAN/PNS) guideline (second revision) replaced the 2010 EFNS/PNS guideline, aiming to improve specificity and clinical applicability [7]. Key changes included reducing diagnostic categories from three (definite, probable, and possible CIDP) to two (CIDP and possible CIDP). Clinical definitions for CIDP variants were added and modifications of the electrodiagnostic criteria were done, including adding variant‐specific criteria and sensory abnormalities. These revisions aimed to improve specificity and simplify the guideline but were partially based on expert opinion rather than empirical evidence, and the effects of individual changes remain uninvestigated.

The diagnostic performance of the 2021 EAN/PNS guideline has been evaluated in previous studies [8, 9, 10]. Two studies reported higher specificity but lower sensitivity [8, 10], while one found similar diagnostic accuracy compared to the 2010 EFNS/PNS guideline [9]. However, none of these studies included patient controls with disorders that frequently mimic or resemble CIDP in clinical practice [8, 9, 10]. To properly evaluate diagnostic accuracy, a well‐defined cohort of clinically representative controls is essential.

Limited nerve conduction study (NCS) protocols are commonly used when polyneuropathy is suspected. Both guidelines recommend extending NCS if electrodiagnostic criteria are insufficiently met, but this is likely inconsistently applied in practice due to time constraints. While limited NCSs likely reduce sensitivity [8], the true extent of missed diagnoses has not been systematically studied.

The main objective of this study was to compare the diagnostic accuracy of the 2021 EAN/PNS and 2010 EFNS/PNS guidelines in a large cohort of CIDP patients and patient controls initially suspected of having a chronic immune‐mediated neuropathy. We also evaluated the impact of specific electrodiagnostic criteria changes and the effect of limited NCS protocols on diagnostic yield.

2. Methods

2.1. Screening and Selection

2.1.1. Study Population

Patients and patient controls were eligible if they were included in the International CIDP Outcome Study (ICOS) [11], the machine learning‐CIDP electrodiagnostic (ML‐CIDP EDX) database, or two Dutch prospective inflammatory neuropathy cohort studies [12, 13]. As of October 2022, the ICOS cohort included 302 patients from Erasmus Medical Center (Erasmus MC), Amsterdam University Medical Centers (Amsterdam UMC), and University Medical Center Utrecht (UMC Utrecht). The ML‐CIDP EDX database, consisting of NCS performed at Amsterdam UMC between 2009 and 2021 for suspected immune‐mediated neuropathies, contributed 55 CIDP patients and 934 non‐CIDP controls after exclusion of overlapping cases from the ICOS cohort. Two additional prospective cohort studies (2013–2018) enrolled CIDP patients from UMC Utrecht, Amsterdam UMC, and Elisabeth‐Tweesteden Ziekenhuis, with 95 patients consenting to reuse of their data. ICOS participants who were initially misdiagnosed with CIDP were reassigned to the control group. Data were collected prospectively for all studies except for the ML‐CIDP EDX database, which was retrospective.

2.1.2. Patients With CIDP

The CIDP reference standard included two groups of patients: (1) those who met the 2010 EFNS/PNS electrodiagnostic criteria for possible, probable, or definite CIDP, and (2) those who did not meet these criteria but had a presumptive CIDP diagnosis based on clinical evaluation, at least 6 months of follow‐up, and a documented treatment response as defined in the 2021 EAN/PNS guideline. Patients with fewer than two sensory measurements (or fewer than four for the pure motor variant) were excluded to avoid underestimating the sensitivity of the 2021 guideline, as sensory studies were not always routinely performed under the 2010 EFNS/PNS criteria. Patients with Neurofascin (NF)‐155, Contactin (CNTN)‐1, Contactin‐associated protein (CASPR)‐1 antibody positivity were excluded from this study. Antibody testing was performed in all patients in the ICOS cohort [14] and on clinical indication in the other cohorts.

2.1.3. Patient Controls

Controls were patients with suspected chronic immune‐mediated neuropathy who underwent extensive NCS and were diagnosed with conditions other than CIDP during the initial work‐up or after a minimum of 6 months of follow‐up. Demyelinating neuropathies that may meet CIDP electrodiagnostic criteria were excluded, as their distinction from CIDP is primarily based on clinical course or specific laboratory findings and not the guideline criteria. These excluded conditions include Guillain‐Barré syndrome (acute onset and monophasic course) [15], genetically confirmed Charcot–Marie–Tooth disease (family history, prolonged course, foot deformities, and genetic testing) [16], multifocal motor neuropathy (distal, multifocal motor involvement) [17], and autoimmune nodopathies (paranodal antibody presence) [18, 19]. Each CIDP patient was matched 1:1 to an eligible control without replacement using a composite score based on age, height, and sex. The score incorporated absolute differences in age and height, with an additional penalty for sex mismatch. The control with the lowest composite score was selected. To assess variant‐specific criteria, controls were assigned the CIDP phenotype they most closely resembled, based on recorded neurological examination findings.

2.2. Nerve Conduction Studies and Supportive Criteria

2.2.1. Nerve Conduction Study Analysis

For each patient and control, one NCS from the available study data was selected after warm‐up. If multiple NCSs were available for CIDP patients, the most extensive pre‐treatment NCS was selected. If more than one full pre‐treatment NCS was available, the NCS closest to treatment initiation was chosen. For patient controls, who were typically not treated, the most extensive NCS was selected; when studies were equally extensive, the earliest study was chosen. Reflecting real‐world practice, NCSs were performed as part of standard clinical care, using protocols that were standardized within centers but not across centers. Electrodiagnostic criteria and cut‐off values were applied uniformly according to guideline definitions. Clinical neurophysiologists systematically re‐evaluated tabular NCS data, including waveforms if available, using the 2010 EFNS/PNS and 2021 EAN/PNS guideline criteria. Signs supportive of demyelination were scored for each nerve segment (Table S1 in appendix). Distal CMAP duration thresholds were applied according to the low‐frequency filter setting used, based on the 2021 EAN/PNS guideline, and were applied uniformly for both guidelines. A distal negative CMAP amplitude of $\ge$ 1.0 mV was required to consider the NCS abnormalities supportive of demyelination.

2.2.2. Supportive Criteria

Supportive criteria were applied in a standardized manner across cohorts using predefined definitions and cut‐off values. These were strictly based on the guidelines and are listed in Table 1. Treatment responses to first‐line treatments initiated at diagnosis were evaluated using the 2021 EAN/PNS criteria, as the 2010 EFNS/PNS guidelines lacked well‐defined cut‐offs. If multiple treatments were administered, the most effective first‐line treatment response within the first 6 months was evaluated. For patients or controls not enrolled in prospective studies, or those treated long before enrollment, the Modified Rankin Scale (MRS) was used, with improvement defined as a decrease of at least one point [20]. Since the 2021 EAN/PNS guideline allows flexibility in scoring nerve ultrasound, we rescored the median nerve and brachial plexus bilaterally, considering ultrasound supportive if focal enlargement was present in two different sites (Table 1) [21, 22]. The brachial plexus was considered abnormal if at least two enlargements were present at the interscalene trunks or nerve roots on the same side. An exploratory analysis also evaluated a threshold of one enlargement.

TABLE 1.

Overview of supportive criteria.

Supportive criterion	Criteria or cut‐off values	2010 EFNS/PNS1	2021 EAN/PNS7
Treatment response	Improvement on at least one disability scale (I‐RODS, INCAT‐DS) and one impairment scale (MRC‐ss, mISS, grip strength), or at least one point decrease on the MRS.	Yes	Yes
MRI	Visual enlargement and/or increased signal intensity of nerve root(s) on T2 weighted sequences	Yes	Yes
Nerve ultrasound	Nerve enlargement of at least two sites, including median nerve in forearm (> 10 mm2), in upper arm (> 13 mm2), individual nerve elements at level of interscalene trunks (> 9 mm2) or nerve roots (> 12 mm2) a	No	Yes
CSF protein	Under 50 years: ≥ 0.5 g/L Over 50 years: ≥ 0.6 g/L	Yes	Yes
Abnormal sensory conduction	One of the following in at least one nerve: Sural sparing Conduction velocity < 80% of LLN; < 70% if SNAP amplitude < 80% of LLN	Yes	No b
Nerve biopsy	Presence of thinly myelinated axons and small onion bulbs, thinly myelinated or demyelinated internodes in teased fibers, perivascular macrophage clusters, supportive features of demyelination on electron microscopy	Yes	Yes
SSEP	Delayed somatosensory evoked potentials without central nervous system disease	Yes c	No

Abbreviations: CSF, cerebrospinal fluid; INCAT‐DS, Inflammatory Neuropathy Cause and Treatment disability scale; I‐RODS, Inflammatory Rasch‐built Overall Disability Scale; LLN, lower limit of normal; mISS, Modified INCAT Sensory Sum scale; MRC‐ss, Medical Research Council sum score; MRI, magnetic resonance imaging; MRS, modified ranking scale; SSEP, somatosensory evoked potential.

^a Supportive if at least two enlargements of individual elements at the level of interscalene trunks or at least two nerve roots.

^b Part of sensory conduction criteria for sensory CIDP instead of supportive criteria.

^c Not used in this study.

2.3. Effect of Electrodiagnostic Criteria Modifications

An exploratory analysis assessed whether modifying electrodiagnostic criteria could improve diagnostic accuracy compared to the 2021 EAN/PNS electrodiagnostic criteria. Three predefined sets were tested. Set A excluded all sensory criteria except for sensory conduction criteria required for sensory CIDP. Set B reintroduced partial motor conduction block (≥ 50% reduction of proximal relative to distal negative peak CMAP amplitude) from the 2010 EFNS/PNS as the minimum requirement for weakly supportive of demyelination instead of a conduction block (≥ 30% CMAP amplitude reduction, ≥ 50% for axilla‐Erb segment). Set C included tibial nerve conduction block (≥ 50% CMAP amplitude reduction).

2.4. Effect of Extensiveness of NCS on Diagnostic Accuracy

In another exploratory analysis we evaluated how the number of NCS measurements (NCS extensiveness) affected diagnostic accuracy. We analyzed two datasets: one with all available NCS measurements and one with limited NCS measurements (Table 2). The limited NCSs followed the first steps of 2021 EAN/PNS guideline advice and only included four motor and four sensory nerves with only distal segments on one side. These first steps of 2021 EAN/PNS guideline advice were followed to create a limited set of NCSs, which is commonly used in clinical practice. NCS data were censored to include only four motor and four sensory nerves with distal segments on one side. This allowed us to assess the number of missed diagnoses if testing had stopped.

TABLE 2.

Nerves and nerve segments included in the evaluation of NCS extensiveness.

	Motor nerves	Sensory nerves
All NCS	Median (APB): wrist‐elbow, elbow‐axilla, axilla‐Erb; Median (FCR): elbow‐axilla, axilla‐Erb Ulnar (ADQ): wrist‐distal medial epicondyle, proximal medial epicondyle‐axilla, axilla‐Erb Radial (EIP): forearm‐upper arm Musculocutaneous (BB): axilla‐Erb Peroneal (EDB): ankle‐distal fibular head Tibial (AH): ankle‐knee	Median (dig II): wrist, wrist‐elbow Ulnar (dig V): wrist, wrist‐elbow Radial (over tendon EPL): wrist Sural (lateral malleolus): proximal to lateral malleolus
Limited NCS a , b	Median: wrist‐elbow Ulnar: wrist‐elbow Peroneal: ankle‐distal fibular head Tibial: ankle‐knee	Median: wrist Ulnar: wrist Radial: wrist Sural: proximal to lateral malleolus
Extended NCS a	Median (APB): wrist‐elbow, elbow‐axilla, axilla‐Erb; Median (FCR): elbow‐axilla, axilla‐Erb Ulnar: wrist‐distal medial epicondyle, proximal medial epicondyle‐axilla, axilla‐Erb Peroneal: ankle‐distal fibular head Tibial: ankle‐knee	Median: wrist, wrist‐elbow Ulnar: wrist, wrist‐elbow Radial: wrist Sural: proximal to lateral malleolus

Abbreviations: ADQ, abductor digiti quinti; AH, abductor hallucis; APB, adductor pollicis brevis; BB, biceps brachii; dig II, second digit; dig V, fifth digit; EDB, extensor digitorum brevis; EIP, extensor indicis proprius; EPL, extensor pollicis longus; FCR, flexor carpi radialis; NCS, nerve conduction study.

^a Recording sites in limited and extended NCS as in all NCS.

^b Left side used, unless the nerve measurement was only done on the right side.

Lastly, we performed an explorative subgroup analysis including only patients and patient controls who at least underwent extensive NCSs, meaning at least eight motor and eight sensory nerves, including proximal segments for the arm nerves. This enabled comparison across three levels of NCS extensiveness: all available, limited, and extensive NCS.

2.5. Statistical Analysis

Sensitivity and specificity were calculated for both sets of guidelines. The diagnostic categories were partitioned in two ways: the first distinguishing between no CIDP and a combination of possible, probable, and definite CIDP (i.e., no CIDP versus CIDP, including possible CIDP), and the second distinguishing between no CIDP or possible CIDP and (probable/definite) CIDP. The changes in sensitivity and specificity between the guidelines were analyzed using the McNemar test. The diagnostic accuracy was calculated by taking the number of correctly classified participants (true positive + true negative) divided by the number of all subjects (true positive + true negative + false positive + false negative). Additionally, the positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), and negative likelihood ratio (LR‐) were calculated to further assess diagnostic performance of both guidelines. All statistical tests were conducted using a two‐sided p‐value threshold of < 0.05 for significance. All statistical analyses were carried out using IBM SPSS Statistics, version 28.0.1.1 (IBM Corp, Armonk, NY, USA).

3. Results

3.1. Screening and Selection

Of 428 CIDP patients screened, 89 were excluded, and 339 CIDP patients were included (Figure 1). Of 958 potential patient controls, 267 met exclusion criteria, leaving 691 eligible controls. After 1:1 matching, 339 were selected (Figure 1, Table 3). The control group included patients with axonal neuropathies (30%), motor neuron disease (27%), and a diverse spectrum of demyelinating neuropathies, plexopathies, (poly)radiculopathies, and myopathies (Table 4). Sixteen patient controls were initially misdiagnosed with CIDP and received an alternative diagnosis during follow‐up.

FIGURE 1.

Flowchart of screening and selection of patients and patient controls. † This includes 16 patients who were initially misdiagnosed with CIDP and received an alternative diagnosis during follow‐up. ML‐CIDP EDX, machine learning‐CIDP electrodiagnostic; NCS, nerve conduction study.

TABLE 3.

Characteristics of patients with CIDP and patient controls.

	CIDP patients (n = 339)	Patient controls (n = 339)
Sex, male, n (%)	215 (63%)	215 (63%)
Age at diagnosis, years, median (IQR)	60.0 (49.8–68.2)	61.0 (51.7–70.5)
Known diabetes mellitus (%)	37 (11%)	48 (14%)
Tested for autoimmune nodopathy and excluded, n (%) a	190 (56%)	10 (3%)
Treatment response, n responders/treated (%)	239/302 (79%)	10/36 (28%) b
Nerve ultrasound supportive, n positive/tested (%)	116/171 (68%)	30/99 (30%)
MRI supportive, n positive/tested (%)	33/124 (27%)	1/55 (2%)
CSF protein elevated, n (%)	157/254 (62%)	30/93 (32%)
Median (IQR) in group	0.74 (0.5–1.1)	0.48 (0.3–0.7)
Median (IQR) in subgroup patients with CSF protein elevation	0.92 (0.7–1.4)	0.79 (0.7–1.0)
Nerve biopsy, n positive/tested (%)	5/12 (42%)	0/7 (0%)
At least two supportive criteria 2010 EFNS/PNS evaluated, n (%)	336 (99%)	169 (50%)
Two or more supportive criteria met, n (%)	143 (43%) c	7 (4%)
One supportive criterion met, n (%)	150 (45%) c	35 (21%)
None met, n (%)	43 (13%) c	127 (75%)
At least two 2021 EAN/PNS supportive criteria evaluated, n (%)	304 (90%)	115 (34%)
Two or more supportive criteria met, n (%)	179 (59%)	11 (10%) c
One supportive criterion met, n (%)	103 (34%)	39 (34%) c
None met, n (%)	22 (7%)	65 (57%) c
Number of tested motor nerves, median (IQR)	7 (5–12)	12 (12–12)
Number of tested sensory nerves, median (IQR)	6 (3–8)	8 (8–8)

^a Neurofascin (NF)‐155, contactin (CNTN)‐1, contactin associated protein (CASPR)‐1.

^b All treatment responses occurred in patient controls who were treated with corticosteroids or intravenous immunoglobulins combined with corticosteroids.

^c Percentages have been rounded and may not total 100%.

TABLE 4.

Final diagnosis in patient controls.

Diagnosis	N of patient controls (%) (n = 339)
Axonal neuropathy	102 (30%)
Motor neuron disease	93 (27%)
Compression neuropathy	23 (7%)
Plexopathy	22 (7%)
(Poly)radiculopathy	22 (7%)
Myopathy	20 (6%)
Other demyelinating neuropathy a	15 (4%)
Multiple mononeuropathy	13 (4%)
Small‐fiber neuropathy	8 (2%)
Myelopathy	5 (2%)
Lumbar or cervical stenosis	5 (2%)
Other b	11 (3%)

^a Morbus Waldenström, anti‐myelin‐associated glycoprotein (MAG) neuropathy, or monoclonal gammopathy of unknown significance (MGUS) neuropathy.

^b Myasthenia gravis, post‐polio syndrome, intraneural perineurioma, non‐neurological diagnosis.

3.2. Guideline Comparison

3.2.1. Electrodiagnostic and Guideline Criteria Accuracy

In our cohort, 288 (85%) CIDP patients met the 2010 EFNS/PNS criteria for definite CIDP, 18 (5%) for probable CIDP, 4 (1%) for possible CIDP, and 29 (9%) did not meet these criteria but were included based on clinical evaluation and a documented treatment response. For the 2021 EAN/PNS criteria, 280 (83%) met the criteria for CIDP, 39 (12%) for possible CIDP, and 20 (6%) did not meet the criteria. Compared to the 2010 criteria, the 2021 criteria reclassified more patients from (probable/definite) CIDP to possible CIDP (Table 5).

TABLE 5.

Comparison of the sensitivity and specificity between the 2010 EFNS/PNS and 2021 EAN/PNS criteria.

		CIDP a including possible CIDP versus no CIDP			CIDP a versus no and possible CIDP
		2010 EFNS/PNS	2021 EAN/PNS	p *	2010 EFNS/PNS	2021 EAN/PNS	p *
Sensitivity
All NCS (n = 339)	Electrodiagnostic criteria	91%	89%	NS	84%	76%	< 0.001
	Diagnostic criteria b	91%	94%	NS	90%	83%	< 0.001
Limited NCS (n = 339)	Electrodiagnostic criteria	80%	66%	< 0.001	59%	32%	< 0.001
	Diagnostic criteria b	80%	78%	NS	77%	51%	< 0.001
Specificity
All NCS (n = 339)	Electrodiagnostic criteria	67%	66%	NS	81%	86%	< 0.001
	Diagnostic criteria b	67%	66%	NS	79%	86%	< 0.001
Limited NCS (n = 339)	Electrodiagnostic criteria	83%	83%	NS	96%	98%	NS
	Diagnostic criteria b	83%	82%	NS	94%	97%	NS

Note: Comparison of the sensitivity and specificity of the 2010 EFNS/PNS and 2021 EAN/PNS criteria for CIDP, using both electrodiagnostic and guideline criteria including the supportive criteria. All analyses were conducted in the same cohort of 339 CIDP patients and 339 matched patient controls, censoring any additional measurements in the limited NCS dataset beyond unilateral, distal measurements of four motor and four sensory nerves. For all NCS, the full dataset was used.

Abbreviations: NCS, nerve conduction studies; NS, not significant.

^a CIDP: includes the probable and definite categories for the 2010 EFNS/PNS criteria.

^b After application of supportive criteria.

^* p‐value if significant (< 0.05); McNemar test.

3.2.2. Guideline Sensitivity and Specificity

For the category of CIDP including possible CIDP, sensitivity remained similar between the 2010 EFNS/PNS and 2021 EAN/PNS guidelines, as most reclassifications occurred between possible and (probable/definite) CIDP. For (probable/definite) CIDP, sensitivity decreased from 90% with the 2010 EFNS/PNS criteria to 83% with the 2021 EAN/PNS criteria (p < 0.001). This was mainly due to changes in the electrodiagnostic criteria (Table S2, appendix), which were not fully compensated by the updated supportive criteria (Table 5). Specificity for CIDP including possible CIDP remained unchanged, but for (probable/definite) CIDP specificity increased from 79% to 86% (p < 0.001). Diagnostic accuracy was similar between guidelines: 79% with the 2010 EFNS/PNS and 80% with the 2021 EAN/PNS criteria for CIDP including possible CIDP (Table S3, appendix). For (probable/definite) CIDP, accuracy remained unchanged at 85%. Thus, although sensitivity and specificity varied, overall diagnostic accuracy remained similar.

3.2.3. Shifts in Electrodiagnostic Classification

In total, 26 (8%) CIDP patients and 68 (20%) patient controls were reclassified to a different diagnostic category due to changes in the electrodiagnostic criteria. The new sensory criteria led to the reclassification of 15 CIDP patients and 8 control patients to a lower diagnostic category, decreasing sensitivity and increasing specificity. Excluding tibial nerve conduction blocks reclassified 7 CIDP patients and 23 control patients into a lower diagnostic category. Replacing partial motor conduction blocks (≥ 50% CMAP amplitude reduction) with conduction blocks (≥ 30% CMAP amplitude reduction, ≥ 50% between axilla and Erb) improved sensitivity but negatively impacted specificity. A finding previously considered a probable conduction block became sufficient for weakly supportive of demyelination in four patients, or strongly supportive in two patients when combined with two absent F‐waves. However, this change also led to 28 additional patient controls being classified as weakly supportive due to a single conduction block, and 1 as strongly supportive due to a conduction block with two absent F‐waves. The stricter criterion for temporal dispersion in the tibial nerve (≥ 100% proximal‐to‐distal CMAP duration increase) caused two CIDP patients to no longer meet the criteria without affecting controls. One patient control with a sensory phenotype no longer met criteria due to new variant‐specific sensory conduction criteria. Finally, two patient controls without upper limb motor abnormalities were reclassified from strongly to weakly supportive of demyelination based on the new variant‐specific rules.

3.2.4. Supportive Criteria

Under the 2021 EAN/PNS guideline, 59% of CIDP patients met two or more supportive criteria, compared to 43% under the 2010 EFNS/PNS guideline. This increase was primarily due to the addition of nerve ultrasound, which was supportive in 68% of CIDP patients and 30% of patient controls. Considering one enlargement in the brachial plexus as supportive as opposed to two sites within the plexus, would classify 80% of ultrasounds in CIDP patients and 48% in controls as supportive. However, this did not significantly affect guideline sensitivity (94% for CIDP including possible CIDP; 83% for CIDP only) or specificity (86% for CIDP including possible CIDP and 65% for CIDP only).

3.2.5. CIDP Variants

For typical sensorimotor CIDP, sensitivity decreased from 92% to 86% (p = 0.007) under the 2021 EAN/PNS guideline, while specificity remained unchanged (Table 6). Multifocal CIDP showed decreased sensitivity (83% to 67%; p = 0.012) but increased specificity (70%–88%; p < 0.001). For the other CIDP variants, diagnostic accuracy did not change when possible CIDP was included. The 2021 EAN/PNS guideline allows typical sensorimotor CIDP patients to be classified as possible CIDP when there is an objective treatment response to a first‐line treatment and one other criterion, even if the NCS is unsupportive. Consequently, 16 (31%) of 52 patients with unsupportive NCSs met the additional requirements and were reclassified as possible CIDP. This rule does not apply to CIDP variants, but its application in multifocal CIDP would in theory reclassify four (57%) of seven patients with unsupportive NCSs as possible multifocal CIDP. In the patient control group, one (1%) of 110 patient controls would have upgraded to possible CIDP. This patient control had multiple mononeuropathy due to systemic vasculitis and responded well to corticosteroids.

TABLE 6.

Sensitivity and specificity of 2010 EFNS/PNS and 2021 EAN/PNS criteria per CIDP phenotype.

		CIDP a including possible CIDP versus no CIDP			CIDP a versus no and possible CIDP
		2010 EFNS/PNS	2021 EAN/PNS	p *	2010 EFNS/PNS	2021 EAN/PNS	p *
Sensitivity per phenotype
Typical (n = 241)	Electrodiagnostic criteria	224 (93%)	219 (91%)	NS	210 (87%)	196 (81%)	< 0.001
	Diagnostic criteria b	224 (93%)	235 (98%)	NS	222 (92%)	207 (86%)	0.007
Multifocal c (n = 54)	Electrodiagnostic criteria	47 (87%)	46 (85%)	NS	39 (72%)	31 (57%)	0.008
	Diagnostic criteria b	47 (87%)	46 (85%)	NS	45 (83%)	36 (67%)	0.012
Distal (n = 9)	Electrodiagnostic criteria	8 (89%)	8 (89%)	NS	8 (89%)	8 (89%)	NS
	Diagnostic criteria b	8 (89%)	8 (89%)	NS	8 (89%)	8 (89%)	NS
Motor (n = 17)	Electrodiagnostic criteria	14 (82%)	12 (71%)	NS	13 (77%)	11 (65%)	NS
	Diagnostic criteria b	14 (82%)	12 (71%)	NS	14 (82%)	12 (71%)	NS
Sensory (n = 18)	Electrodiagnostic criteria	17 (94%)	18 (100%)	NS	15 (83%)	13 (72%)	NS
	Diagnostic criteria b	17 (94%)	18 (100%)	NS	17 (94%)	17 (94%)	NS
Specificity per phenotype
Typical (n = 45)	Electrodiagnostic criteria	28 (62%)	30 (67%)	NS	35 (78%)	37 (82%)	NS
	Diagnostic criteria b	28 (62%)	29 (64%)	NS	33 (73%)	37 (82%)	NS
Multifocal c (n = 172)	Electrodiagnostic criteria	112 (65%)	114 (66%)	NS	142 (83%)	151 (88%)	0.004
	Diagnostic criteria b	112 (65%)	114 (66%)	NS	138 (80%)	151 (88%)	< 0.001
Distal (n = 33)	Electrodiagnostic criteria	22 (67%)	20 (61%)	NS	25 (76%)	29 (88%)	NS
	Diagnostic criteria b	22 (67%)	20 (61%)	NS	25 (76%)	29 (88%)	NS
Motor (n = 38)	Electrodiagnostic criteria	24 (63%)	20 (53%)	NS	30 (79%)	30 (79%)	NS
	Diagnostic criteria b	24 (63%)	20 (53%)	NS	28 (74%)	30 (79%)	NS
Sensory (n = 51)	Electrodiagnostic criteria	41 (80%)	39 (77%)	NS	44 (86%)	46 (90%)	NS
	Diagnostic criteria b	41 (80%)	39 (77%)	NS	44 (86%)	46 (90%)	NS

Abbreviations: NCS, nerve conduction study; NS, not significant.

^a CIDP: includes the probable and definite categories for the 2010 EFNS/PNS criteria.

^b After application of supportive criteria.

^c Including focal CIDP.

^* p‐value if significant (< 0.05); McNemar test.

3.3. Effect of Electrodiagnostic Criteria Modifications

To evaluate whether specific modifications improved diagnostic accuracy, we explored several adapted electrodiagnostic criteria sets (Table 7) and their effect on accuracy for possible CIDP and CIDP.

TABLE 7.

Diagnostic accuracy metrics of 2021 EAN/PNS criteria with and without adaptations.

	Sensitivity	Specificity	PPV	NPV	LR+	LR‐	Accuracy
2021 EAN/PNS without adaptations
Electrodiagnostic criteria for weakly supportive of demyelination	89%	66%	72%	86%	2.62	0.17	78%
Diagnostic criteria a for CIDP including possible CIDP	94%	66%	73%	92%	2.76	0.09	80%
Electrodiagnostic criteria for strongly supportive of demyelination	76%	86%	84%	78%	5.43	0.28	81%
Diagnostic criteria a for CIDP	83%	86%	86%	83%	5.93	0.20	85%
Set A
Electrodiagnostic criteria for weakly supportive of demyelination	92%*	62%*	71%	89%	2.42	0.13	77%
Diagnostic criteria a for CIDP including possible CIDP	96%	62%*	72%	94%	2.53	0.06	79%
Electrodiagnostic criteria for strongly supportive of demyelination	81%*	85%	84%	82%	5.40	0.22	83%
Diagnostic criteria a for CIDP	87%*	85%	85%	87%	5.80	0.15	86%
Set B
Electrodiagnostic criteria for weakly supportive of demyelination	87%*	76%*	78%	85%	3.62	0.17	82%
Diagnostic criteria a for CIDP including possible CIDP	92%*	75%*	79%	90%	3.68	0.11	83%
Electrodiagnostic criteria for strongly supportive of demyelination	76%	86%	84%	78%	5.43	0.28	81%
Diagnostic criteria a for CIDP	81%	86%	85%	82%	5.79	0.22	84%
Set C
Electrodiagnostic criteria for weakly supportive of demyelination	90%	60%*	69%	86%	2.25	0.17	75%
Diagnostic criteria a for CIDP including possible CIDP	94%	60%*	70%	91%	2.35	0.01	77%
Electrodiagnostic criteria for strongly supportive of demyelination	79%*	84%*	83%	80%	4.94	0.25	82%
Diagnostic criteria a for CIDP	84%*	84%*	84%	84%	5.25	0.19	82%
Set A + B
Electrodiagnostic criteria for weakly supportive of demyelination	89%	74%*	77%	87%	3.42	0.15	82%
Diagnostic criteria a for CIDP including possible CIDP	93%	73%*	77%	91%	3.44	0.10	83%
Electrodiagnostic criteria for strongly supportive of demyelination	81%*	85%	84%	82%	5.40	0.22	83%
Diagnostic criteria a for CIDP	86%*	85%	85%	86%	5.73	0.16	86%

Note: Including all available NCS measurements of 339 CIDP patients and 339 patient controls. Set A: Excluding sensory criteria, except for sensory conduction criteria required for sensory CIDP. Set B: Increasing the threshold for weakly supportive of demyelination from one conduction block (> 30% CMAP amplitude reduction or > 50% in the axilla to Erb segment) to at least one partial motor conduction block (> 50% CMAP amplitude reduction, excluding axilla‐Erb). Set C: Including tibial nerve blocks (> 50% amplitude reduction).

Abbreviations: LR−, negative likelihood ratio; LR+, positive likelihood ratio; NCS, nerve conduction study; NPV, negative predictive value; PPV, positive predictive value.

^a After application of supportive criteria.

^* p‐value if significant (< 0.05) when compared to the respective sensitivity or specificity of EAN/PNS 2021 criteria; McNemar test.

Excluding sensory abnormalities from the criteria (Set A), except for the sensory conduction criteria for sensory CIDP and sensory criteria for motor CIDP, still resulted in high sensitivity for CIDP including possible CIDP (96% vs. 94%; not significant), but at the cost of reduced specificity (62% vs. 66%; p < 0.001), while accuracy remained similar (79% vs. 80%) compared to the EAN/PNS 2021 guideline. For CIDP only, sensitivity increased from 83% to 87% (p < 0.001), with similar specificity (85% vs. 86%), resulting in comparable accuracy (86% vs. 85%).

Reintroducing partial motor conduction block as a minimum requirement for weakly supportive of demyelination (Set B) primarily affected patients classified as weakly supportive based on a single conduction block. For CIDP including possible CIDP, sensitivity decreased from 94% to 92% (p < 0.05), while specificity increased from 66% to 75% (p < 0.001), leading to improved diagnostic accuracy (80%–83%). For CIDP only, sensitivity (81%), specificity (86%), and accuracy (84%) remained similar.

For Set C, which included tibial nerve blocks, sensitivity remained high for CIDP including possible CIDP (94%), but specificity dropped significantly from 66% to 60% (p < 0.001), negatively affecting accuracy (77%). For CIDP only, sensitivity increased from 83% to 84% (p < 0.05), while specificity decreased from 86% to 84% (p < 0.05), with an accuracy of 82%.

Finally, combining Set A and Set B (excluding sensory abnormalities while reintroducing partial motor conduction blocks as a minimum for weakly supportive of demyelination) resulted in higher specificity (73%; p < 0.001) with similar sensitivity (93%) for CIDP including possible CIDP. This improved accuracy from 80% to 83%. For CIDP only, these adaptations increased sensitivity to 86% (p < 0.001), while maintaining specificity at 85%, resulting in an accuracy of 86%.

3.4. Effect of Extensiveness of NCSs on Diagnostic Accuracy

Compared to using all available NCS measurements, the use of a limited amount of NCS measurements (limited NCS) led to a strongly reduced sensitivity and slightly increased specificity under the 2021 EAN/PNS guideline for possible CIDP and CIDP (Table 5). Consequently, 77 patients (23%) no longer met the electrodiagnostic criteria for weakly supportive of demyelination, leading to the underdiagnosis of 55 patients (16%) with CIDP including possible CIDP, who would be classified as no CIDP.

An exploratory analysis was conducted on a subset of 271 CIDP patients and 335 patient controls who at least underwent extensive NCSs (Table S4 in appendix). The dataset with all available measurements included musculocutaneous and radial motor nerve measurements when tested. Nevertheless, sensitivity and specificity were similar compared to the extensive NCS set.

4. Discussion

Our study demonstrates that the sensitivity and specificity of the 2021 EAN/PNS diagnostic criteria for CIDP including possible CIDP are comparable to the 2010 EFNS/PNS, as most reclassifications occurred between possible and (probable/definite) CIDP. However, for (probable/definite) CIDP without possible CIDP, the 2021 EAN/PNS criteria had a significantly lower sensitivity and a significantly higher specificity than the previous version.

This shift was primarily due to changes in the electrodiagnostic criteria. The addition of sensory abnormalities as a requirement (except for variant‐specific sensory conduction criteria) had the most impact, causing a 4% decrease in sensitivity and a 2% increase in specificity (Table S2, appendix). This modest effect may partly reflect that sensory involvement can be clinically present despite normal sensory nerve conduction studies, particularly in proximal or polyradicular disease. This criterion was based on expert opinion rather than empirical evidence, and our findings align with previous studies that demonstrated the limited diagnostic value of sensory criteria and their potential to lower overall performance [8, 9, 10]. Reintroducing partial motor conduction blocks as the minimum requirement for weakly supportive of demyelination mainly increased specificity. Combining this modification with removal of the sensory abnormalities requirement improved diagnostic accuracy in this cohort, but these findings are exploratory and should be validated in independent cohorts. However, the use of two definitions for conduction blocks adds complexity, which may increase the risk of misinterpretation and misdiagnosis [5, 23]. As many NCS abnormalities suggestive of demyelination overlap between CIDP and other immune‐mediated neuropathies, future guidelines might benefit from shared electrodiagnostic criteria for immune‐mediated neuropathies [7, 17, 19, 24, 25].

More patients met two supportive criteria under the 2021 guideline compared to the 2010 guideline, primarily due to the addition of nerve ultrasound. However, requiring two supportive criteria to move from possible CIDP to CIDP (versus one criterion under the 2010 EFNS/PNS criteria) raised the bar for a higher diagnostic classification. For nerve ultrasound, requiring two enlarged roots/trunks within the brachial plexus did not affect the overall diagnostic accuracy compared to requiring only one enlargement within the plexus.

Limited, unilateral NCSs (four motor and four sensory nerves) with only forearm measurements would miss a considerable proportion of CIDP patients, reducing sensitivity to 66% for strongly or weakly supportive of demyelination, and to 32% for strongly supportive of demyelination only. Including extensive arm measurements up to Erb's point bilaterally improved detection rates compared to unilateral NCS. This aligns with an earlier study recommending more extensive testing [8]. Extending NCSs may increase diagnostic yield in patients with suspected CIDP when initial measurements do not yet fulfill electrodiagnostic criteria. Proximal stimulation, particularly at Erb's point, may be technically challenging and normative values are not universally available; however, guideline‐based interpretation relies on relative changes across segments, allowing application when feasible.

In clinical practice, patients classified as possible CIDP often receive trial treatments like intravenous immunoglobulins or corticosteroids, despite not meeting the criteria for (probable/definite) CIDP. This contrasts with research settings that require more stringent criteria for clinical trial enrollment. While the increased specificity of the 2021 EAN/PNS criteria for CIDP only reduces false positives and unnecessary treatments, the decreased sensitivity may lead to undertreatment. Additional diagnostic evaluations, such as nerve imaging and cerebrospinal fluid protein analysis, can help increase certainty in patients with possible CIDP. Somatosensory evoked potentials may be useful in selected patients with pure sensory symptoms and signs, without weakness, and normal nerve conduction studies when chronic immune sensory polyradiculopathy (CISP) is suspected, but this entity is no longer included within the CIDP spectrum. Standardized assessments of treatment response, as already recommended in the 2021 EAN/PNS guideline, are especially important when clinical but not electrodiagnostic criteria are met and no alternative diagnosis is apparent [4, 7]. However, response to immunomodulatory treatments is not exclusive to CIDP and occurs in other autoimmune (neuromuscular) diseases [26, 27, 28].

This study has several limitations. While most CIDP data were collected prospectively, patient control data were retrospective and based on a clinically relevant convenience sample. Selecting the most extensive available NCS may have increased sensitivity, but this applied to both guideline sets and therefore is unlikely to have biased the comparison between guidelines. To assess the effect of NCS extensiveness, we also evaluated diagnostic performance using a predefined limited NCS set. Controls were matched for age, sex, and height, but not phenotype, which might have influenced results; however, CIDP variants—typically more diagnostically challenging—were well represented among controls. Matching reduced the number of controls included in the final analysis, but was based on demographic and anthropometric variables rather than electrodiagnostic findings or diagnostic category. Therefore, selective exclusion of borderline electrodiagnostic cases is unlikely, although analyses using the full eligible control cohort with statistical adjustment were not performed. Patients with multifocal motor neuropathy, autoimmune nodopathies, Guillain–Barré syndrome, and hereditary neuropathies were excluded as controls. Because these diagnostically challenging demyelinating conditions may fulfill electrodiagnostic or supportive criteria despite representing another diagnosis, their exclusion may have overestimated specificity, but is unlikely to have substantially affected the comparison between guidelines. Electrodiagnostic criteria identify demyelinating features compatible with CIDP, but do not by themselves distinguish CIDP from all demyelinating mimics; alternative diagnoses should therefore still be considered based on clinical phenotype, disease course, and ancillary testing. Finally, the lack of a definitive biomarker for CIDP necessitated reliance on a reference standard based on existing guidelines and expert opinion, which has inherent limitations [24]. Because fulfillment of the 2010 EFNS/PNS criteria was part of this reference standard, incorporation bias may have favored the 2010 EFNS/PNS guideline when comparing diagnostic performance with the 2021 EAN/PNS guideline.

In conclusion, the 2021 EAN/PNS guideline is more specific but slightly less sensitive than the 2010 EFNS/PNS guideline for diagnosing (probable/definite) CIDP without the possible CIDP category. This reduced sensitivity, especially when a limited number of NCS measurements were done, underscores the need for careful application in clinical practice. Additional testing, such as nerve ultrasounds, CSF, and a trial treatment, should be considered in patients with a sensorimotor phenotype or any patient meeting the criteria for weakly supportive of demyelination to increase diagnostic certainty. Exploratory modifications, such as removing the sensory abnormalities criterion, were associated with improved diagnostic accuracy in this cohort, but require validation in independent cohorts.

Author Contributions

Bart C. Jacobs: methodology, validation, writing – review and editing, conceptualization. Ivo N. van Schaik: methodology, validation, writing – review and editing. Pieter A. van Doorn: methodology, validation, writing – review and editing, conceptualization. Krista Kuitwaard: methodology, validation, writing – review and editing. W. Ludo van der Pol: methodology, validation, writing – review and editing. Judith Drenthen: methodology, data curation, supervision, validation, writing – review and editing, conceptualization. Adája E. Baars: data curation, methodology, conceptualization, investigation, writing – original draft, formal analysis. Filip Eftimov: methodology, writing – review and editing, conceptualization, validation. H. Stephan Goedee: methodology, data curation, validation, writing – review and editing, conceptualization, supervision. Iris N. van Doorn: conceptualization, investigation, methodology, visualization, writing – original draft, formal analysis, data curation. Luuk Wieske: methodology, validation, writing – review and editing. Camiel Verhamme: methodology, supervision, writing – review and editing, conceptualization, data curation, validation.

Funding

The authors have nothing to report.

Conflicts of Interest

A.E.B., L.W., I.N.v.S., and J.D. have nothing to report. I.N.v.S. received a personal travel grant from the Dutch Society of Clinical Neurophysiology to attend and present parts of this work at the 2023 Annual Meeting of the Peripheral Nerve Society in Copenhagen, Denmark. B.C.J. reports research grants to the institution from Prinses Beatrix Spierfonds, GBS‐CIDP Foundation International, Hansa Biopharma, Annexon, Roche, CSL Behring, Grifols and Octapharma; consulting fees and Data Safety Monitoring Board payments from Hansa and Annexon, paid to the institution; and unpaid roles on the GBS‐CIDP Foundation International Global Medical Advisory Board, INCbase steering committee, IGOS consortium and ICOS consortium. K.K. reports research grants to the institution from the For Wis(h)dom Foundation, Takeda and the Albert Schweitzer Hospital Stipendium; consulting fees from Takeda; speaker fees from Octapharma and Grifols; and travel support from the GBS‐CIDP Foundation International. W.L.v.d.P. reports ad hoc consultancy payments from Argenx, paid to the institution. P.A.v.D. reports a research grant from Prinses Beatrix Spierfonds; consulting fees for clinical studies from Annexon, Argenx, Hansa, Octapharma, and Sanofi; and honoraria from Grifols and Argenx, all paid to the institution. F.E. reports research grants to the institution from Prinses Beatrix Spierfonds, ZonMw, Dianthus, CSL Behring, Takeda, Kedrion, Terumo BCT, and the GBS‐CIDP Foundation International; consulting fees from Dianthus and Sanofi; honoraria from Takeda, CSL Behring and Grifols; and serves as chair of the INCbase steering committee. H.S.G. reports research grants from Argenx and the Dutch ALS Foundation, paid to the institution; consulting fees from Takeda; honoraria from Takeda and Argenx; and unpaid roles as president of the Dutch Society for Clinical Neurophysiology, chair of the Dutch Neuromuscular Working Group and secretary of the Dutch–Belgian Neuromuscular Study Club. C.V. reports research grants from Prinses Beatrix Spierfonds and the GBS‐CIDP Foundation International, paid to the institution; honoraria from the Dutch Society of Clinical Neurophysiology and Sonoskills; and unpaid roles as board member of ERN EURO‐NMD and medical advisor for Spierziekten Nederland. Unless otherwise stated, disclosures are outside the submitted work.

Supporting information

Table S1: Motor and sensory nerve conduction criteria.

Table S2: Effects of the updated 2021 EAN/PNS electrodiagnostic criteria.

Table S3: Diagnostic accuracy of 2010 EFNS/PNS and 2021 EAN/PNS criteria using all available NCS measurements.

Table S4: Sensitivity and specificity of the 2010 EFNS/PNS and 2021 EAN/PNS guidelines related to extensiveness of NCS.

Data Availability Statement

Data are available upon reasonable request. Anonymized data used for this study are available upon reasonable request from the corresponding author.