doi: /brain/awh375 Brain (2005), 128,

Transcription

1 doi: /brain/awh375 Brain (2005), 128, Criteria for demyelination based on the maximum slowing due to axonal degeneration, determined after warming in water at 37 C: diagnostic yield in chronic inflammatory demyelinating polyneuropathy J. T. H. Van Asseldonk, 1 L. H. Van den Berg, 2 S. Kalmijn, 2,3 J. H. J. Wokke 2 and H. Franssen 1 1 Neuromuscular Research Group, Rudolf Magnus Institute of Neuroscience, Department of Clinical Neurophysiology, 2 Department of Neurology and 3 Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, The Netherlands Summary The diagnosis of chronic inflammatory demyelinating polyneuropathy (CIDP) is based on clinical and laboratory results and on features of demyelination found in nerve conduction studies. The criteria that are currently used to reveal demyelinative slowing in CIDP have several limitations. These criteria were only determined in lower arm and lower leg nerve segments, were not defined with respect to nerve temperature, and the relationship with distal compound muscle action potential (CMAP) amplitudes is unclear. The aim of our study was to determine criteria for demyelinative slowing for lower arm and leg segments as well as for upper arm and shoulder segments at a temperature of 37 C, and to assess whether criteria have to be modified when the distal CMAP is decreased. Included were 73 patients with lower motor neuron disease (LMND), 45 patients with CIDP and 36 healthy controls. The arms and legs were warmed in water at 37 C for at least 30 min prior to an investigation and thereafter kept warm with infrared heaters. The proposed criteria for demyelinative slowing were based on the maximum conduction slowing that may occur as a consequence of axonal degeneration and consisted of the upper boundary (99%) or the lower boundary (1%) of conduction values in Correspondence to: H. Franssen, MD, PhD, Department of Clinical Neurophysiology, University Medical Centre Utrecht, P.O. Box 85500, 3508 GA Utrecht, The Netherlands h.franssen@neuro.azu.nl LMND. In LMND, the maximum conduction slowing was different for arm and leg nerves and for segments within the arm nerves. Moreover, distal motor latency and motor conduction velocity were slower in nerves with distal CMAP amplitudes below 1 mv than in nerves with distal CMAP amplitudes above 1 mv. For these reasons, separate criteria were proposed for arm nerves, for leg nerves and for different segments within arm nerves, and more stringent criteria were proposed for distal motor latency and motor conduction velocity when the distal CMAP amplitude was below 1 mv. The diagnostic yield in CIDP was assessed using the nerve, and not the patient, as the unit of measurement. Thus, whether demyelinative slowing was present was determined for each nerve. Compared with other criteria, our criteria increased the specificity without affecting sensitivity. We conclude that the present criteria, based on the maximum slowing that may occur as a result of axonal degeneration, allow more accurate detection of demyelinative slowing in CIDP compared with other criteria. It should be emphasized that the proposed criteria can only be applied if the method of warming in water at 37 C for at least 30 min is adopted. Keywords: axonal degeneration; chronic inflammatory demyelinating polyneuropathy; criteria; demyelinative slowing; lower motor neuron disease Abbreviations: CI = confidence interval; CIDP = chronic inflammatory demyelinating polyneuropathy; CMAP = compound muscle action potential; DML = distal motor latency; duration prolongation P/D = duration prolongation on proximal versus distal stimulation; LMND = lower motor neuron disease; LR = likelihood ratio; MCV = motor conduction velocity; OR = odds ratio; PV = negative predictive value; PV + = positive predictive value Received June 24, Revised August 27, Accepted November 30, Advance Access publication February 2, 2005 # The Author (2005). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please journals.permissions@oupjournals.org

2 Introduction Chronic inflammatory demyelinating polyneuropathy (CIDP) is a progressive or relapsing disorder, characterized by multifocal demyelination in peripheral nerves and spinal roots (Dyck et al., 1975; Barohn et al., 1989; Said, 2002). Differentiation from other polyneuropathies is important as CIDP improves with immunological treatment (Dyck et al., 1982, 1986; Van Doorn et al., 1990; Vermeulen et al., 1993; Hahn et al., 1996; Hadden et al., 1999; Hughes et al., 2001; Mendell et al., 2001). The diagnosis of CIDP is based on clinical and laboratory results and on features of demyelination found on nerve conduction studies (American Academy of Neurology, 1991). For the diagnosis of CIDP, criteria for nerve conduction slowing consistent with demyelination (demyelinative slowing) and conduction block have been grouped into sets. Criteria are cut-off values that can be used to determine whether a value in a given patient is compatible with demyelination. Sets consist of several of these criteria for different variables that are required to be fulfilled in order to make a diagnosis of CIDP. Most sets require that criteria for different variables, reflecting demyelinative slowing or conduction block, are fulfilled in a number of nerves (American Academy of Neurology, 1991; Albers and Kelly, 1989; Barohn et al., 1989; Logigian et al., 1994; Uncini et al., 1999; Hughes et al., 2001; Saperstein et al., 2001; Nicolas et al., 2002). The application of these sets in the diagnosis of CIDP has been criticized because the sets were shown to have a moderate sensitivity, which may lead to under-diagnosis of a potentially severe but treatable neuropathy (Bromberg, 1991; Logigian et al., 1994; Latov, 2002; Molenaar et al., 2002; Magda et al., 2003; Van den Bergh and Pieret, 2004). Other limitations are that several of the criteria for demyelinative slowing are not based on empirical studies and that the criteria were not defined with respect to nerve temperature. Because not all slowing reflects demyelination, criteria for demyelinative slowing require that nerve conduction slowing exceeds values that may occur as a consequence of axonal degeneration. For conduction velocity and temporal dispersion, criteria based on the slowing that may result from axonal degeneration were determined for lower arm and lower leg nerve segments but not for proximal nerve segments (Buchthal and Behse, 1977; Logigian et al., 1994; Cappellari et al., 1997). Despite this, many sets allow the application of these criteria to upper arm and shoulder segments. To account for additional slowing due to severe axonal degeneration, some sets defined more stringent criteria for demyelinative slowing if the amplitude of the compound muscle action potential on distal stimulation is decreased (American Academy of Neurology, 1991; Uncini et al., 1999; Saperstein et al., 2001). It is not certain, however, whether these more stringent criteria are justified, as a relation between amplitude and slowing was found to be either slight for all variables or non-existent for most variables (Cornblath et al., 1992; Logigian et al., 1994). Criteria for demyelinative slowing in CIDP 881 Criteria for demyelinative slowing were determined from nerve conduction studies carried out at uncontrolled temperature or after warming the limbs with infrared heaters (Albers and Kelly, 1989; Barohn et al., 1989; American Academy of Neurology, 1991; Logigian et al., 1994; Uncini et al., 1999; Saperstein et al., 2001; Nicolas et al., 2002). However, other studies have shown that warming with infrared heaters is not suitable for attaining a nerve temperature of 37 C and that this value will only be reached by warming in water at 37 C for a sufficiently long period (Geerlings and Mechelse, 1985; Franssen and Wieneke, 1994). Furthermore, the decrease in conduction velocity with temperature appeared to be more pronounced in normal nerves than in nerves affected by axonal degeneration or demyelination (Notermans et al., 1994; Franssen et al., 1999). For these reasons, accurate assessment of demyelinative slowing needs to be performed in nerves warmed to 37 C. The objective of our present study was to determine criteria for demyelinative slowing in lower arm and leg segments as well as upper arm and shoulder segments, and to assess whether criteria have to be modified when the distal compound muscle action potential is decreased, in nerves investigated at a temperature of 37 C. The criteria were derived from the maximal conduction slowing found in 73 patients with sporadic lower motor neuron disease (LMND), a disorder characterized by axonal degeneration. We preferred to determine the maximum slowing that may occur as a consequence of axonal degeneration in LMND rather than in axonal polyneuropathies. In LMND, axonal degeneration occurs frequently in arm as well as in leg nerves (McLeod and Prineas, 1971; Van den Berg-Vos et al., 2003), whereas in axonal polyneuropathies the axonal degeneration is often restricted to leg nerves. It is, therefore, more likely that the maximum slowing that may occur due to axonal degeneration in arm and in leg nerves will be found in LMND than in axonal polyneuropathies. Subsequently, the diagnostic yield for demyelinative slowing of our criteria and that of the different criteria that are currently used in several newer and older sets was determined in 45 patients with CIDP. Patients and methods Patients Included were 73 patients with adult-onset sporadic LMND, 45 patients with CIDP and 36 healthy controls (21 men) with a median age of 51 years (range years). All patients visited the neuromuscular outpatient clinics of the University Medical Centre Utrecht, a tertiary referral centre for patients with neuromuscular disease in The Netherlands. Healthy controls were employees of the University Medical Centre Utrecht or relatives of patients without a history of motor or sensory deficits in the arms or legs. All participants gave informed consent to participation in the study, which was approved by the Medical Ethical Committee of the University Medical Centre Utrecht.

3 882 J. T. H. Van Asseldonk et al. Patients with sporadic LMND were included on the basis of a slowly progressive lower motor neuron syndrome with adult onset, evidence of lower motor neuron involvement on neurological examination (weakness, atrophy and fasciculations) and electrophysiological evidence of lower motor neuron involvement on needle electromyography (Van den Berg-Vos et al., 2003). Exclusion criteria were a family history of LMND, a deletion in the survival motor neuron 1 gene, more than 40 CAG repeats in the androgen receptor gene, clinical signs of upper motor neuron involvement, structural lesions that could account for the clinical findings on MRI or myelography of the spinal cord, abnormalities of sensory conduction or evidence of motor conduction block on extensive standardized nerve conduction studies (Van den Berg-Vos et al., 2000, 2003). With respect to verification bias, the use of conduction block as an exclusion criterion precluded determination and proposal of criteria for conduction block in the present study. Patients with CIDP were selected on the basis of a clinical consensus diagnosis. For all patients who were suspected of CIDP on the basis of their clinical presentation, we presented the medical history, physical examination, disease course, treatment, response to treatment and the results of ancillary investigations, except those of the electrophysiological examination, to two neurologists specializing in neuromuscular disorders (L.H.V.D.B. and J.H.J.W.). These neurologists independently categorized the patients as having definite CIDP, probable CIDP or a disorder other than CIDP. The weighted k for the level of agreement between the two neurologists was The patients categorized as having definite CIDP by both neurologists were included and the patients categorized as having a disorder other than CIDP by both neurologists were excluded. The remaining patients were categorized as having CIDP or as a disorder other than CIDP during a second consensus meeting between the two neurologists. The characteristics of the selected patients are shown in Table 1. The percentage of men and the age at the time of examination were within similar ranges for patients with LMND and for patients with CIDP. In the majority of patients with LMND and CIDP, motor deficits were found in distal arm and/or leg muscles. Characteristics of patients with CIDP were similar to those reported in previous series: motor and sensory deficits were frequently found in distal parts of limbs and motor deficits were frequently found in proximal parts of limbs (Said, 2002). Of all patients with CIDP, 82% were treated with immune-modulating drugs, to which 81% responded positively. Electrophysiological studies Nerve conduction was studied bilaterally in all patients with LMND and in 30 patients with CIDP, and unilaterally in 15 patients with CIDP and in healthy controls. All nerve conduction studies were performed by the same investigator (H.F.) according to a standardized protocol using Ag AgCl surface electrodes with a diameter of 9 mm (Van den Berg et al., 1998). Motor nerve conduction was investigated in the following nerves: median (stimulation: wrist, elbow, axilla and Erb s point; recording: m. abductor pollicis brevis), ulnar (stimulation: wrist, 5 cm below the elbow, 5 cm above the elbow, axilla and Erb s point; recording: m. abductor digiti V), peroneal (stimulation: ankle, 5 cm below the fibular head, popliteal fossa; recording: m. extensor digitorum brevis) and tibial (stimulation: ankle, popliteal fossa; recording: m. abductor hallucis). Responses were only scored if supramaximal stimulation was possible [at least 20% above the strength yielding a maximal compound muscle action potential (CMAP); for Erb s Table 1 Patient characteristics LMND CIDP Number Male: n (%) 55 (75) 30 (67) Age at onset: median; range (years) 51; ; Age at examination: median; 56; ; range (years) Neurological examination (%) Symmetrical Sensory only 0 4 Motor only Sensory and motor 0 82 Motor Leg only 4 16 Arm only 36 4 Arm + leg Distal only Proximal only 7 2 Distal + proximal Sensory Leg only 0 18 Arm only 0 4 Arm + leg 0 64 Distal only 0 80 Proximal only 0 0 Distal + proximal 0 7 Laboratory features CSF protein: median; range* 0.3; ; Abnormal MRI brachial plexus: n (%) y 0 (0) 11 (73) Motor = motor deficits; sensory = sensory deficits; CSF = cerebrospinal fluid; abnormal MRI brachial plexus = swelling and/or increased signal intensity on MRI. *Investigated in 17 patients with LMND and 27 patients with CIDP. y Investigated in 17 patients with LMND and 15 patients with CIDP. point at least 30%]. F waves were recorded after 20 distal stimuli to nerves in which the distal CMAP amplitude was above 0.5 mv. If necessary, a collision technique was used to detect effects of costimulation (Kimura, 1989). Prior to an investigation, the arms up to the axilla and legs up to the knee were continuously warmed in water at 37 C for at least 30 min; thereafter they were kept warm with infrared heaters (Kimura, 1989; Franssen and Wieneke, 1994). The following conduction variables were determined: distal amplitude (amplitude of the negative part of the distal CMAP); distal motor latency (DML); shortest F M latency, distal duration (duration of the negative part of the distal CMAP); motor conduction velocity (MCV) per segment and duration prolongation P/D per segment [(proximal CMAP duration distal CMAP duration) 3 100%/ distal CMAP duration]. For waveforms that were polyphasic, CMAP duration was measured to the first baseline crossing. Variables for conduction slowing also included distal duration and duration prolongation P/D, since their values are influenced by unequal slowing among the axons of a nerve (Brown and Snow, 1991; Oh et al., 1994; Cappellari et al., 1997). Proposed criteria for demyelinative slowing The criteria for demyelinative slowing were derived from cut-off values of conduction variables in patients with LMND. The term demyelinative slowing was used for DML, shortest F M latency, distal duration and duration prolongation P/D as well as for MCV

4 consistent with demyelination. The upper boundary (99%) was used to determine the criteria for DML, shortest F M latency, distal duration and duration prolongation P/D. The lower boundary (1%) was used to determine the criteria for MCV. Conduction variables that were normally distributed were expressed as a percentage of the normal mean. For these variables, mean values were compared between homologous segments, i.e. between the lower arm segment of the median nerve and the lower arm segment of the ulnar nerve, between the upper arm segment of the median nerve and the upper arm segment of the ulnar nerve, and between the lower leg segment of the peroneal nerve and the lower leg segment of the tibial nerve. If no significant differences were found, variables of homologous segments were pooled for the median and ulnar nerves and for the peroneal and tibial nerves. The proposed criteria derived from these pooled variables were expressed as a percentage of the normal mean and rounded off to a multiple of 5. In addition, criteria consisting of absolute values were recalculated separately for the median, ulnar, peroneal and tibial nerves. For this purpose, criteria expressed as a percentage of the normal mean were multiplied by their normal mean value. For conduction variables that were not normally distributed, criteria were only expressed as absolute values; they were determined separately for the median, ulnar, peroneal and tibial nerves and were rounded off to values that are of practical use. Diagnostic yield of criteria for demyelinative slowing The diagnostic yield for demyelinative slowing of the proposed criteria was compared with that of other published criteria. The nerve, and not the patient, was used as the unit of measurement. Whether demyelinative slowing was present was determined for each nerve. To determine the diagnostic yield, the percentage of nerves with demyelinative slowing was assessed for each patient group. In patients with LMND, nerves with demyelinative slowing were scored as false positive and nerves without demyelinative slowing as true negative. In patients with CIDP, nerves with demyelinative slowing were scored as true positive and nerves without demyelinative slowing as false negative. For each variable and for combinations of variables, we calculated the specificity (number of true negative nerves/number of nerves in LMND), sensitivity (number of true positive nerves/number of nerves in CIDP), positive predictive value (PV +; number of true positive nerves/number of true and false positive nerves), negative predictive value (PV ; number of true negative nerves/number of true and false negative nerves) and the positive likelihood ratio (LR). The LR was calculated as [sensitivity/(1 specificity)] and represents the odds of finding demyelinative slowing in nerves of patients with CIDP compared with nerves of patients with LMND. The other published criteria for CIDP that were used for comparison included the criteria for DML, shortest F M latency and MCV described in the sets of the American Academy of Neurology (American Academy of Neurology, 1991) and Albers and Kelly (Albers and Kelly, 1989), the criteria for distal CMAP duration described by Thaisetthawatkul and colleagues (Thaisetthawatkul et al., 2002) and by Oh and colleagues (Oh et al., 1994), and the criteria for duration prolongation P/D reported by Brown and Feasby (Brown and Feasby, 1984) and by Oh and colleagues (Oh et al., 1994). The criteria for DML, shortest F M latency and MCV that are used in other sets (Hughes et al., 2001; Saperstein et al., 2001; Nicolas et al., 2002; Berger et al., 2003; Magda et al., 2003) are Criteria for demyelinative slowing in CIDP 883 similar to the criteria described in the criteria-set of the American Academy of Neurology (American Academy of Neurology, 1991). Statistical analysis Differences in continuous variables between two groups were tested using the non-parametric Mann Whitney U test. Whether or not variables were normally distributed in healthy controls was tested with the Shapiro Wilk test and checked by visual inspection. Whether demyelinative slowing in CIDP occurred more often in nerves with decreased distal amplitudes than in nerves with normal distal amplitudes was calculated using logistic regression analysis and was expressed as the odds ratio (OR) and 95% confidence interval (CI). To determine whether the diagnostic yield of the proposed criteria differed from that of other published criteria, we calculated the 95% CI for specificity, sensitivity, positive predictive value and negative predictive value. A significant difference was assumed when confidence intervals were not overlapping. In other circumstances, P < 0.05 was considered to be statistically significant. Results Conduction variables in healthy controls, patients with LMND and patients with CIDP Figures 1 and 2 show the distribution of the different conduction variables in healthy controls, patients with LMND and patients with CIDP. For most nerves, conduction variables were slower in LMND than in healthy controls and slower in CIDP than in LMND. Relation between conduction slowing and distal amplitude To determine whether conduction slowing is related to distal amplitude, scatter plots were examined (Fig. 3). The distal amplitude was plotted against each of the conduction variables, except for the MCV and duration prolongation P/D of the upper arm and shoulder segments. The results for the median and ulnar nerves and those for the peroneal and tibial nerves were pooled. In LMND, DML and MCV were slower in leg nerves (and to a lesser extent in arm nerves) with a distal amplitude below 1 mv than in nerves with a distal amplitude above 1 mv. This relationship was not apparent for the other variables. A distal CMAP amplitude below 1 mv was found in 8% of arm and in 17% of leg nerves. The median values of DML and MCV for arm and leg nerves and the median values of the shortest F M latency for leg nerves were significantly slower in nerves with a distal amplitude below 1 mv than in nerves with a distal amplitude above 1 mv. In CIDP, the DML, distal duration and MCV were slower in leg nerves with distal amplitudes below 1 mv than in leg nerves with distal amplitudes above 1 mv (Fig. 3). This relationship was not apparent for the other variables in leg nerves or for variables in the arm nerves. A distal CMAP amplitude below 1 mv was found in 2% of arm and in 31% of leg nerves. The median values of DML and MCV were significantly

5 884 J. T. H. Van Asseldonk et al. slower for leg nerves with a distal amplitude below 1 mv than for leg nerves with a distal amplitude above 1 mv. In arm nerves, median values were not compared as a distal amplitude below 1 mv hardly ever occurred. Proposed criteria for demyelinative slowing The values of DML, MCV, shortest F M latency and distal duration were normally distributed in healthy controls and could, therefore, be expressed as a percentage of the normal mean. In LMND, median values of DML and shortest F M latency were significantly slower in arm than in leg nerves; median values of MCV were significantly slower in lower arm than in lower leg nerve segments. In LMND, systematic differences between conduction values in homologous segments of the median and ulnar nerves and between conduction values in homologous segments of the peroneal and tibial nerves were not found (Figs 1 and 2). For these reasons the conduction values in homologous segments of the median and ulnar nerves (lower arm, upper arm, shoulder) were pooled, as were those in the lower leg segments of the peroneal and tibial nerves. From the pooled conduction values, criteria, expressed as a percentage of the normal mean, were derived for DML, MCV, shortest F M latency and distal duration (Table 2). As the median values of DML and MCV were slower for nerves with distal amplitudes below 1 mv than for nerves with distal amplitudes above 1 mv, separate criteria for nerves with distal amplitudes below 1 mv were derived for DML and MCV (Table 2). The criteria shown in Table 2 were used to calculate absolute cut-off values for the median, ulnar, peroneal and tibial nerves (Table 3). Duration prolongation P/D was not normally distributed in healthy controls; hence these criteria were expressed in absolute values and determined for the median, ulnar, peroneal and tibial nerves (Table 3). In LMND, median values of duration prolongation P/D were significantly slower in lower leg than in lower arm nerve segments. The most stringent criterion for duration prolongation P/D was proposed for homologous segments of arm and leg nerves, if there was a difference (Table 3). Relation between demyelinative slowing and distal amplitude in CIDP We determined whether demyelinative slowing in CIDP occurred more often in nerves with a decreased distal amplitude than in nerves with a normal distal amplitude. Demyelinative slowing was defined according to the criteria presented in Table 3. A decreased distal amplitude, defined as a distal CMAP amplitude below the lower limit of normal, Fig. 1 Conduction values of DML, shortest F M latency and distal duration. Values are shown for healthy controls (&), patients with LMND (&) and patients with CIDP (&). DML = distal motor latency; median = median nerve; ulnar = ulnar nerve; peroneal = peroneal nerve; tibial = tibial nerve. All conduction values are expressed as a percentage of the normal mean.

6 Criteria for demyelinative slowing in CIDP 885 Fig. 2 Conduction values of MCV and duration prolongation P/D. Values are shown for healthy controls (&), patients with LMND (&) and patients with CIDP (&). MCV = motor conduction velocity; median = median nerve; ulnar = ulnar nerve; peroneal = peroneal nerve; tibial = tibial nerve. The values of MCV are expressed as a percentage of the normal mean and the values of duration prolongation P/D are expressed as absolute values. was found in 46% of arm and in 39% of leg nerves. For each conduction variable, with the exception of duration prolongation P/D, the chance of finding demyelinative slowing was significantly higher for nerves with a decreased distal amplitude than for nerves with a normal distal amplitude (Table 4). The chance of finding at least one conduction variable that fulfilled the criteria for demyelinative slowing was also significantly higher for nerves with a decreased distal amplitude (Table 4). Comparison of diagnostic yield of proposed and other criteria The diagnostic yield of the proposed and other published criteria is shown for all variables in arm and leg nerves in Table 5. In general, when compared with those of other published criteria, the positive predictive values of the Table 2 Proposed criteria for demyelinative slowing (percentage of normal mean) Nerve Arm Leg Distal amplitude <1 mv >1 mv <1 mv >1 mv DML Shortest F M latency Distal duration MCV lower arm/leg MCV upper arm MCV shoulder DML = distal motor latency; MCV = motor conduction velocity. proposed criteria were higher and the negative predictive values similar. For arm nerves, the positive predictive values of the proposed criteria for DML, distal duration, MCV and duration prolongation P/D were significantly higher than

7 886 J. T. H. Van Asseldonk et al. those of other criteria. For leg nerves, positive predictive values of the proposed criteria for DML, F M latency, MCV and duration prolongation P/D were significantly higher than those of other criteria. Negative predictive values of the proposed criteria for arm and leg nerves, were not significantly different from those of the other criteria. The likelihood ratios of the proposed criteria were higher than those of other criteria, except for distal duration in the leg nerves. For example, the likelihood ratio of the proposed criteria was 50 for DML in arm nerves, meaning that a DML value fulfilling the proposed criteria is 50 times more likely to be found in CIDP than in LMND. The diagnostic yield of the proposed and other published criteria for MCV and duration prolongation P/D in different arm nerve segments are shown in Table 6. The positive predictive values of the proposed criteria for MCV and duration prolongation P/D were significantly higher than those of other criteria in the lower arm, upper arm and shoulder segment, except for MCV in the lower arm. Negative predictive values of the proposed criteria were not significantly different from those of the other criteria. The likelihood ratios of the proposed criteria were higher than those of other criteria. The diagnostic yield of a set consisting of the proposed criteria was compared with that of other published sets (Fig. 4). To express the diagnostic yield of each set, the posttest probability of a positive test (presence of demyelinative slowing) and that of a negative test (absence of demyelinative slowing) was calculated for any given pre-test probability (prevalence of demyelinative slowing). The variables of a set only included DML, shortest F M latency and MCV since other sets do not include distal duration and only the set of the American Academy of Neurology included duration prolongation P/D. Demyelinative slowing was assumed within a nerve when at least one of the criteria for DML, shortest F M latency or MCV was fulfilled. For low pre-test probabilities, the proposed set will lead to higher post-test probabilities than the other sets (Fig. 4). The proposed set increased the post-test probabilities of the presence of demyelinative slowing by up to 26% compared with the set of the American Academy of Neurology and up to 37% compared with the set published by Albers and Kelly (Fig. 4). For all pre-test probabilities, the post-test probabilities of the absence of demyelinative slowing were similar for the proposed set and the other sets. The specificity for demyelinative slowing in arm or leg nerves was 99% for the proposed set, 96% for the set Fig. 3 Relationships between conduction variables and distal CMAP amplitude. Relationships are shown for patients with LMND (n) and patients with CIDP (&). CMAP = compound muscle action potential; DML = distal motor latency; MCV = motor conduction velocity. The values of DML, shortest F M latency, distal CMAP duration and MCV are expressed as a percentage of the normal mean and the values of duration prolongation P/D are expressed as absolute values. The vertical line represents a distal CMAP amplitude of 1 mv.

8 Table 3 Proposed criteria for demyelinative slowing (absolute values) Criteria for demyelinative slowing in CIDP 887 Nerve Median Ulnar Peroneal Tibial Distal amplitude <1 mv >1 mv <1 mv >1 mv <1 mv >1 mv <1 mv >1 mv Variable DML (ms) Shortest F-M latency (ms) Distal duration (ms) MCV lower arm/leg (m/s) MCV upper arm (m/s) MCV shoulder (m/s) Duration prolongation P/D lower arm/leg (%) Duration prolongation P/D upper arm (%) Duration prolongation P/D shoulder (%) DML = distal motor latency; MCV = motor conduction velocity; Duration prolongation P/D = (proximal CMAP duration distal CMAP duration) 3 100%/distal CMAP duration. Table 4 Relationship between demyelinative slowing and distal amplitude in CIDP Percentage of nerves with demyelinative slowing Distal amplitude <LLN >LLN OR (95% CI) DML ( )* Shortest F M latency ( )* Distal duration ( )* MCV y ( )* Duration prolongation P/D y ( ) Any variable ( )* LLN = lower limit of normal; DML = distal motor latency; MCV = motor conduction velocity; duration prolongation P/D = (proximal CMAP duration distal CMAP duration) 3 100%/distal CMAP duration; any variable = at least one conduction variable fulfilling criteria for demyelinative slowing; OR = odds ratio (expresses the risk of demyelinative slowing in nerves with distal amplitudes below the LLN, compared with nerves with distal amplitudes above the LLN). y In at least one segment of a nerve. *P < of the American Academy of Neurology and 93% for the set of Albers and Kelly. These differences were significant. The sensitivity for demyelinative slowing was 56% for the proposed set, 58% for the set of the American Academy of Neurology and 59% for the set of Albers and Kelly. These differences were not significant. Overall, the proposed set increased the diagnostic yield for demyelinative slowing in CIDP compared with other sets as a result of increased specificity without decreasing sensitivity. Table 5 Diagnostic yield of proposed and other criteria in arm and leg nerves Arm Leg PV + PV LR PV + PV LR DML Proposed AAN 0.91* * Albers and Kelly 0.85* * Shortest F M latency Proposed AAN * Albers and Kelly * Distal duration Proposed Thaisetthawatkul * Oh/Doo 0.75* * MCV y Proposed AAN * Albers and Kelly 0.93* * Duration prolongation P/D y Proposed Oh/Doo 0.69* * Brown/Feasby 0.68* * AAN = American Academy of Neurology; DML = distal motor latency; MCV = motor conduction velocity; duration prolongation P/D = (proximal CMAP duration distal CMAP duration) 3 100%/distal CMAP duration; PV + = positive predictive value; PV = negative predictive value; LR = likelihood ratio. y In at least one segment of a nerve. *PV+ of criterion proposed by others significantly different from PV+ of criterion proposed in present study. PV of criteria proposed by others was not significantly different from PV of criteria proposed in the present study. Discussion The proposed criteria for demyelinative slowing in CIDP in our study were based on the maximum conduction slowing that occurred in LMND. In LMND, the maximum conduction slowing was different for arm and leg nerves and for segments within arm nerves. Moreover, DML and MCV were slower in nerves with distal CMAP amplitudes below 1 mv than in nerves with distal CMAP amplitudes above 1 mv. For these reasons, separate criteria for all conduction variables were proposed for arm nerves, for leg nerves and for segments within arm nerves, and more stringent criteria were proposed

9 888 J. T. H. Van Asseldonk et al. Table 6 Diagnostic yield of proposed and other criteria in arm nerve segments Lower arm Upper arm Shoulder PV + PV LR PV + PV LR PV + PV LR MCV Proposed AAN Albers and Kelly * * Duration prolongation P/D Proposed Oh/Doo 0.76* * Brown/Feasby 0.80* * * AAN = American Academy of Neurology; MCV = motor conduction velocity; duration prolongation P/D = (proximal CMAP duration distal CMAP duration) 3 100%/distal CMAP duration; PV + = positive predictive value; PV = negative predictive value; LR = likelihood ratio. *PV+ of criterion proposed by others significantly different from PV+ of criterion proposed in present study. PV of criteria proposed by others was not significantly different from PV of criteria proposed in present study. Fig. 4 Diagnostic yield for demyelinative slowing of a proposed set of criteria compared with sets proposed by others. The solid lines above the diagonal line represent, for any pre-test probability, the post-test probabilities for the presence of demyelinative slowing. The solid lines below the diagonal line represent, for any pre-test probability, the post-test probabilities for the absence of demyelinative slowing. The broken lines represent, for any pre-test probability, the increased post-test probability of the proposed set compared with other sets. for DML and MCV when the distal CMAP amplitude was below 1 mv. In CIDP, the chance of finding demyelinative slowing was significantly higher for nerves with decreased distal amplitudes than for nerves with normal distal amplitudes. Compared with other criteria, our criteria increased the diagnostic yield for demyelinative slowing in CIDP by increasing the specificity without affecting sensitivity. Buchtal and Behse investigated median nerve DML, peroneal nerve MCV for the lower leg, and distal median and sural nerve sensory conduction velocity in patients with Charcot Marie Tooth polyneuropathy, which was classified as demyelinating or axonal on the basis of sural nerve histology (Buchthal and Behse, 1977). In the demyelinating forms, values were slower, and in the axonal forms the values were faster than a criterion corresponding to 60% of the mean value in normal subjects (for DML the reciprocal value was calculated). This criterion was used in the set of criteria reported by Kelly (Kelly, 1983) to assess whether MCVs in patients with monoclonal gammopathy were consistent with demyelination. From these findings, criteria for demyelinative slowing were derived, which required nerve conduction to be slower than a given percentage beyond the limit of normal (Albers and Kelly, 1989; Cornblath, 1990; American Academy of Neurology, 1991; Saperstein et al., 2001). These criteria were subsequently shown to be in agreement with investigations carried out in the lower arm or lower leg segments of the median, ulnar, peroneal and tibial nerves in patients with polyneuropathy of various aetiologies (Logigian et al., 1994). In polyneuropathies that were classified as axonal on the basis of sural nerve histology, MCV was faster than 75% of the lower limit of normal, and DML or F wave latency was faster than 130% of the upper limit of normal. In patients with amyotrophic lateral sclerosis (ALS), findings were roughly similar (Cornblath et al., 1992). For duration prolongation P/D, the maximal amount of slowing that may result from axonal degeneration has been determined in patients with ALS (Cappellari et al., 1997). Contrary to the findings in hereditary demyelinating polyneuropathies, nerve conduction in CIDP may range from normal to markedly slowed, so that criteria for demyelinative slowing cannot be derived from studies performed in CIDP patients (Bromberg, 1991). For distal

10 duration, a criterion for demyelinative slowing was obtained by optimizing cut-off values that were designed to distinguish CIDP from ALS or diabetic polyneuropathy (Thaisetthawatkul et al., 2002). In the present study, criteria for demyelinative slowing were determined on the basis of the maximum slowing as a consequence of axonal degeneration in lower arm and leg segments as well as in upper arm and shoulder segments. The maximum slowing was determined in slowly progressive LMND, and not in ALS, because disease progression of CIDP is also usually slowly progressive. The accuracy of criteria proposed for the shoulder segment may have been restricted by the selective use of collision techniques since this was only performed if stimulation at Erb s point, compared with stimulation in the axilla, yielded a CMAP with greater size or a different shape. To allow application to conduction values obtained in electromyographic laboratories elsewhere, the proposed criteria were expressed as a percentage of the normal mean. Previously, criteria proposed by others were expressed as a percentage of the lower limit of normal because the lower limit of normal would be more widely available than the normal mean (Cornblath, 1990). We preferred to express criteria as a percentage of the normal mean as this method is the least dependent on the range of conduction values and requires a smaller normal population sample. The application of our proposed criteria to conduction values obtained elsewhere will be most accurate if the normal mean is determined in healthy subjects whose age and sex are similar to those of patients suspected to have CIDP. Although the proposed criteria were more stringent for nerves with distal amplitudes below 1 mv, their application in CIDP revealed more demyelinative slowing in nerves with decreased distal amplitudes than in nerves with normal distal amplitudes. A decreased distal amplitude may result from demyelination distal to the most distal site of stimulation or from axonal degeneration. Demyelination may lead to a decrease in distal amplitude due to conduction block or increased temporal dispersion (Albers et al., 1985; Rhee et al., 1990; Cappellari et al., 1996). Temporal dispersion distal to the most distal stimulation site cannot be determined, but was suggested to be reflected by prolongation of the distal duration (Thaisetthawatkul et al., 2002). The strong relationship between decreased distal amplitudes and prolonged distal duration, as found in the present study, suggests that increased temporal dispersion due to demyelination distal to the most distal site of stimulation may have contributed to decreased distal amplitudes in CIDP. Axonal degeneration, on the other hand, may also have contributed to decreased distal amplitudes in our patients with CIDP, since in CIDP the number of spinal motor neurons was found to be decreased in spinal cords at necropsy (Nagamatsu et al., 1999). Regardless of the mechanisms underlying decreased distal amplitudes in CIDP, our results suggest that conduction studies of nerves with low distal amplitudes may increase the sensitivity for demyelinative slowing in CIDP. Criteria for demyelinative slowing in CIDP 889 The purpose of this study was to define new criteria for demyelinative slowing that may in future be used in criteria sets for the diagnosis of CIDP rather than to develop a new set of criteria for CIDP. For these reasons we only determined whether demyelinative slowing was present within a nerve, not whether it was present within a patient. A set consisting of part of the proposed criteria increased the diagnostic yield for demyelinative slowing in nerves of patients with CIDP compared with parts of older sets (Albers and Kelly, 1989; American Academy of Neurology, 1991). We did not compare this for newer sets (Hughes et al., 2001; Saperstein et al., 2001; Nicolas et al., 2002; Berger et al., 2003; Magda et al., 2003). Future validation studies are required to determine whether sets consisting of a combination of the proposed criteria improve identification of patients with CIDP compared with older as well as newer sets. The sensitivity of the proposed and other criteria for demyelinative slowing in CIDP is restricted (Bromberg, 1991; Logigian et al., 1994; Latov, 2002; Molenaar et al., 2002; Magda et al., 2003; Van den Bergh and Pieret, 2004). One reason is that demyelination in CIDP occurs in a multifocal pattern. Pathological, MRI and nerve conduction studies have shown that demyelination is present in some but not in all nerves and nerve segments (McLeod et al., 1973; Lewis and Sumner, 1982; Bromberg and Albers, 1993; Raynor et al., 1995; Van Es et al., 1997; Wilson et al., 1998; Haq et al., 2000; Bosboom et al., 2001). Furthermore, conduction values in CIDP overlap with those in axonal polyneuropathies (Bromberg, 1991). This may indicate that demyelination will not always slow conduction below values resulting from axonal degeneration. Thus, all criteria that are based on the principle that demyelinative slowing is only considered to be present when conduction is slower than in axonal degeneration will have a limited sensitivity. The finding that the proposed criteria increased specificity compared with other criteria for demyelinative slowing should be interpreted with caution. The derivation set, in which the proposed criteria were determined, and the part of the validation set in which the specificity of the proposed criteria was assessed, consisted of the same nerves of patients with LMND. A specificity of 99%, as found for the proposed criteria in the present study, was expected since the proposed criteria were based on an upper boundary of 99% and a lower boundary of 1% of the conduction values in LMND. For reliable assessment of the specificity in CIDP, a novel validation set consisting of conduction values in nerves of patients suspected of CIDP is required. Most probably, the proposed criteria will increase specificity compared with other criteria in such a novel validation set as a result of the proposal of separate criteria for arm and leg nerves, separate criteria for distal and proximal arm nerve segments and separate criteria for nerves with distal amplitudes below 1 mv. Although proposed for CIDP, the determination of the criteria in LMND occurred independently of conduction values in CIDP, implying that the proposed criteria can also be used to reveal demyelinative slowing in patients suspected of any other polyneuropathy.

11 890 J. T. H. Van Asseldonk et al. We conclude that the proposed criteria based on the maximum slowing as a result of axonal degeneration in a large number of arm and leg nerves at a standardized temperature may allow more accurate detection of demyelinative slowing in CIDP compared with other criteria, because separate criteria are proposed for arm and leg nerves, for distal and proximal arm nerve segments and for nerves with low distal amplitudes. It should be emphasized that the proposed criteria can only be applied if the method of warming the arms up to the axilla and legs up to the knee in water at 37 C for at least 30 minutes is adopted. Acknowledgements This work was supported by a grant from the Prinses Beatrix Fonds. The research of L.H.V.d.B. was supported by a research grant from The Netherlands Organisation for Health Research and Development. References American Academy of Neurology (AAN). Research criteria for diagnosis of chronic inflammatory demyelinating polyneuropathy (CIDP). Report from an ad hoc subcommittee of the American Academy of Neurology AIDS Task Force. Neurology 1991; 41: Albers JW, Kelly JJ Jr. Acquired inflammatory demyelinating polyneuropathies: clinical and electrodiagnostic features. Muscle Nerve 1989; 12: Albers JW, Donofrio PD, McGonagle TK. Sequential electrodiagnostic abnormalities in acute inflammatory demyelinating polyradiculoneuropathy. Muscle Nerve 1985; 8: Barohn RJ, Kissel JT, Warmolts JR, Mendell JR. Chronic inflammatory demyelinating polyradiculoneuropathy. Clinical characteristics, course, and recommendations for diagnostic criteria. Arch Neurol 1989; 46: Berger AR, Bradley WG, Brannagan TH, et al. Guidelines for the diagnosis and treatment of chronic inflammatory demyelinating polyneuropathy. J Peripher Nerv Syst 2003; 8: Bosboom WM, Van denberg LH, Franssen H, et al. Diagnostic value of sural nerve demyelination in chronic inflammatory demyelinating polyneuropathy. Brain 2001; 124: Bromberg MB. Comparison of electrodiagnostic criteria for primary demyelination in chronic polyneuropathy. Muscle Nerve 1991; 14: Bromberg MB, Albers JW. Patterns of sensory nerve conduction abnormalities in demyelinating and axonal peripheral nerve disorders. Muscle Nerve 1993; 16: Brown WF, Feasby FE. Conduction block and denervation in Guillain-Barre polyneuropathy. Brain 1984; 107: Brown WF, Snow R. Patterns and severity of conduction abnormalities in Guillain-Barre syndrome. J Neurol Neurosurg Psychiatry 1991; 54: Buchthal F, Behse F. Peroneal muscular atrophy (PMA) and related disorders. I. Clinical manifestations as related to biopsy findings, nerve conduction and electromyography. Brain 1977; 100: Cappellari A, Nobile-Orazio E, Meucci N, Scarlato G, Barbieri S. Multifocal motor neuropathy: a source of error in the serial evaluation of conduction block. Muscle Nerve 1996; 19: Cappellari A, Nobile-Orazio E, Meucci N, Levi MG, Scarlato G, Barbieri S. Criteria for early detection of conduction block in multifocal motor neuropathy (MMN): a study based on control populations and follow-up of MMN patients. J Neurol 1997; 244: Cornblath DR. Electrophysiology in Guillain-Barre syndrome. Ann Neurol 1990; 27 Suppl: S Cornblath DR, Kuncl RW, Mellits ED, et al. Nerve conduction studies in amyotrophic lateral sclerosis. Muscle Nerve 1992; 15: Dyck PJ, Lais AC, Ohta M, Bastron JA, Okazaki H, Groover RV. Chronic inflammatory polyradiculoneuropathy. Mayo Clin Proc 1975; 50: Dyck PJ, O Brien PC, Oviatt KF, et al. Prednisone improves chronic inflammatory demyelinating polyradiculoneuropathy more than no treatment. Ann Neurol 1982; 11: Dyck PJ, Daube J, O Brien P, et al. Plasma exchange in chronic inflammatory demyelinating polyradiculoneuropathy. N Engl J Med 1986; 314: Franssen H, Wieneke GH. Nerve conduction and temperature: necessary warming time. Muscle Nerve 1994; 17: Franssen H, Notermans NC, Wieneke GH. The influence of temperature on nerve conduction in patients with chronic axonal polyneuropathy. Clin Neurophysiol 1999; 110: Geerlings AH, Mechelse K. Temperature and nerve conduction velocity, some practical problems. Electromyogr Clin Neurophysiol 1985; 25: Hadden RD, Sharrack B, Bensa S, Soudain SE, Hughes RA. Randomized trial of interferon beta-1a in chronic inflammatory demyelinating polyradiculoneuropathy. Neurology 1999; 53: Hahn AF, Bolton CF, Pillay N, Chalk C, Benstead T, Bril V, et al. Plasmaexchange therapy in chronic inflammatory demyelinating polyneuropathy. A double-blind, sham-controlled, cross-over study. Brain 1996; 119: Haq RU, Fries TJ, Pendlebury WW, Kenny MJ, Badger GJ, Tandan R. Chronic inflammatory demyelinating polyradiculoneuropathy: a study of proposed electrodiagnostic and histologic criteria. Arch Neurol 2000; 57: Hughes R, Bensa S, Willison H, Van den Bergh P, Comi G, Illa I, Nobile-Orazio E, et al. Randomized controlled trial of intravenous immunoglobulin versus oral prednisolone in chronic inflammatory demyelinating polyradiculoneuropathy. Ann Neurol 2001; 50: Kelly, JJ. The electrodiagnostic findings in peripheral neuropathy associated with monoclonal gammopathy. Muscle Nerve 1983; 6: Kimura J. Electrodiagnosis in disease of nerve and muscle: principles and practice. 2nd edn. Philadelphia: F.A. Davis; p Latov N. Diagnosis of CIDP. Neurology 2002; 59: S2 6. Lewis RA, Sumner AJ. The electrodiagnostic distinctions between chronic familial and acquired demyelinative neuropathies. Neurology 1982; 32: Logigian EL, Kelly JJ, Adelman LS. Nerve conduction and biopsy correlation in over 100 consecutive patients with suspected polyneuropathy. Muscle Nerve 1994; 17: Magda P, Latov N, Brannagan TH III, Weimer LH, Chin RL, Sander HW. Comparison of electrodiagnostic abnormalities and criteria in a cohort of patients with chronic inflammatory demyelinating polyneuropathy. Arch Neurol 2003; 60: McLeod JG, Prineas JW. Distal type of chronic spinal muscular atrophy. Clinical, electrophysiological and pathological studies. Brain 1971; 94: McLeod JG, Prineas JW, Walsh JC. The relationship of conduction velocity to pathology in peripheral nerves. A study of the sural nerve in 90 patients. In: New Developments in Electromyography and Clinical Neurophysiology, Vol. 2. Basel: Karger; p Mendell JR, Barohn RJ, Freimer ML, et al. Randomized controlled trial of IVIg in untreated chronic inflammatory demyelinating polyradiculoneuropathy. Neurology 2001; 56: Molenaar DS, Vermeulen M, de Haan RJ. Comparison of electrodiagnostic criteria for demyelination in patients with chronic inflammatory demyelinating polyneuropathy (CIDP). J Neurol 2002; 249: Nagamatsu M, Terao S, Misu K, et al. Axonal and perikaryal involvement in chronic inflammatory demyelinating polyneuropathy. J Neurol Neurosurg Psychiatry 1999; 66: Nicolas G, Maisonobe T, Le Forestier N, Leger JM, Bouche P. Proposed revised electrophysiological criteria for chronic inflammatory demyelinating polyradiculoneuropathy. Muscle Nerve 2002; 25: Notermans NC, Franssen H, Wieneke GH, Wokke JH. Temperature dependence of nerve conduction and EMG in neuropathy associated with gammopathy. Muscle Nerve 1994; 17: