- Email: [email protected]
Change in excitability of motor axons modifies statistical MUNE results
Motor Unit Number Estimation (MUNE) and Quantitative EMG (Supplements to Clinical Neurophysiology, Vol. 60) Editors: M.B. Bromberg # 2009 Elsevier B.V. All rights reserved
Chapter 2
Change in excitability of motor axons modifies statistical MUNE results Lora A. Majora, K. Ming Chanb, Hugh Bostockc and Kelvin E. Jonesa,* a
Faculty of Physical Education and Recreation, E-488 Van Vliet Centre, University of Alberta, Edmonton, AB T6G 2H9, Canada b
Division of Physical Medicine and Rehabilitation, University of Alberta, Edmonton, AB T6G 2V2, Canada c Sobell Department, Institute of Neurology, Queen Square, London WC1N 3BG, UK
1. Introduction Statistical motor unit number estimation (MUNE) is one of the more widely used methods for estimating the number of functional connections between motor neurons and the muscle fibres they innervate (Daube, 1995). The popularity of the method is largely due to its ease of use and availability within a commercial EMG hardware-software system (VIASYS NeuroCare, Madison, WI, USA). Statistical MUNE estimates the average size of surface motor unit action potentials (S-MUAP) by indirect measurements, and the underlying calculations rely on several intricate and inconspicuous assumptions. Therefore, the method remains somewhat mysterious to the average user and is perhaps the
*
Correspondence to: Kelvin E. Jones, Ph.D., Faculty of Physical Education and Recreation, E-488 Van Vliet Centre, University of Alberta, Edmonton, AB T6G 2H9, Canada. Tel.: (780) 492-0650; Fax: (780) 492-8259; E-mail: [email protected]
least-understood MUNE technique. It is clear that the method uses the statistical variability resulting from peripheral nerve stimulation to perform the calculations that generate the MUNE value. It is not clear, however, whether these calculations are susceptible to physiological changes in axon excitability. Nerve excitability testing focuses on submaximal stimuli (Bostock et al., 1998; Kiernan et al., 2000; Burke et al., 2001; Nodera and Kaji, 2006). Axonal excitability studies have demonstrated that changes in voltage-gated ion channels and the electrogenic Naþ/Kþ pump can be indirectly assessed with the right combination of hardware and software. In recent years axon excitability testing has been successfully applied to a wide range of disorders including diabetes, carpal tunnel syndrome, renal failure and motor neuron disease (Krishnan et al., 2006; Kuwabara et al., 2006; Misawa et al., 2006). The recent work on motor neuron disease, or amyotrophic lateral sclerosis (ALS), prompted the present study. Fasciculations have been interpreted as evidence of a hyperexcitable membrane in motor
28 axons or axon terminals in ALS (Bostock et al., 1995; Mogyoros et al., 1998). Further experiments on axon excitability in ALS suggest that there might be a change in the voltage-gated Na þ or K þ channels (Priori et al., 2002; Kanai et al., 2006; Vucic and Kiernan, 2006). Interpretation of the observed excitability changes was supported by complimentary computer models (Bostock and Rothwell, 1997; Hales et al., 2004). In particular, a stochastic model of the effects of sodium channel dynamics predicted that changes in single motor axon threshold, resulting from elevated levels of persistent sodium current (Hales et al., 2004), may underlie some of the changes in excitability measured in ALS patients. Since motor axon excitability varies as part of the neurodegenerative process of ALS, any diagnostic method used to track the number of remaining motor units must not be influenced by this variation. The statistical MUNE method uses the probabilistic response of motor axons to a constant stimulus to calculate the number of motor
units. We hypothesized that changes in motor axon excitability may alter the MUNE calculated with the statistical method even though the number of functional motor units remained the same. This hypothesis was first tested using a computer model that mimicked the statistical MUNE paradigm. We found that two properties of axon thresholds were important for the validity of the statistical MUNE method: the relative spread (RS) of single axon threshold, and the distribution of the mean axon thresholds within a nerve (Fig. 1.2). Both of these properties are amenable to experimental manipulation using polarization of the axon membrane prior to stimulation with a test pulse (Kiernan and Bostock, 2000; Hales et al., 2004). If the axon membrane is depolarized prior to the test stimulus, the RS tends to increase (Fig. 2.1A) and the distribution of mean axon thresholds becomes more spread out (Fig. 2.1B). The effects are the opposite if the axon membrane is hyperpolarized prior to the test stimulus. We applied depolarizing and
Single Axon Activation Functions
Populations of Mean Axon Thresholds
0.75
Count
Probability of Activation
1
0.5
0.25 S D
A
shift the thresholds
expand the distribution
RS = SD/mean
Mean Thresholds
B
shift the thresholds
original threshold distribution
narrow the dist’n
depolarized distribution hyperpolarized distribution mean mean distribution mean
Fig. 2.1 Excitability properties of motor axon populations: relative spread (RS) of single axon activation curves and distribution of thresholds across the axon population. A: activation curves of two single axons. Probability of activation increases monotonically with stimulus intensity (abscissa). Mean threshold is defined as the stimulus current at which the units have a 50/50 chance of firing. RS quantifies the width of an activation curve normalized to the mean threshold of that axon. B: excitability changes in a population of motor axons may cause expansion, contraction, or a shift in the distribution of mean axon thresholds.
29 hyperpolarizing conditioning pulses at the same time as performing a statistical MUNE test to determine if the changes in axon excitability would change the estimated number of motor units. 2. Materials and methods 2.1. Subjects Eight healthy subjects (4 males) aged 22–38 years were tested. Subjects were pre-screened for neuromuscular dysfunction and gave informed consent to the study, which was approved by the Health Research Ethics Board at the University of Alberta. The MUNE data collected from one female subject were later deemed unreliable due to technical complications and therefore excluded from that analysis. 2.2. Experimental protocol 2.2.1. Standard statistical MUNE We used the Nicolet VikingSelect Master Software version 7.3 (VIASYS NeuroCare, Madison, WI, USA) to perform a statistical MUNE test on each subject. Each test consisted of repeated stimulation to collect compound muscle action potential (CMAP) responses within four target windows: 10–20%, 25–35%, 40–50%, and 55–65% (LomenHoerth and Olney, 2000, 2001). Collection of responses from one target window was called a “run”. Each run consisted of 4–10 groups of 30 stimuli, per the usual statistical MUNE protocol. The Viking software automatically calculated the “program-determined” MUNE (pdMUNE, seen on the Viking readout as “tested þ untested”) and the “number-weighted” modification (nwMUNE) (Daube, 1995; Shefner et al., 1999; Simmons et al., 2001). MUNE results recorded and calculated by the Viking system will hereafter referred to as VSM (Viking Statistical MUNE). The following experimental specifications were applied to all statistical MUNE protocols used for this study. Stimulus pulses of 100 ms width were
applied to the median nerve at the wrist through a “Medi-Trace Mini” (Kendall/Tyco, USA) snaptype, pre-gelled, foam, Ag/AgCl cathode. Stimuli were delivered at 2 pulses/s. The return electrode (anode, same type as the cathode) was placed 100–150 mm proximal and slightly lateral to avoid activation of ulnar or superficial radial nerves. All EMG data were recorded with the EA-2 amplifier (Nicolet Biomedical), which is part of the Viking system. The data were filtered at a bandpass of 10 Hz to 3 kHz. Recording electrodes were “Q-Trace Gold” tab-type, pre-gelled, Ag/AgCl (Kendall/Tyco, USA), cut in half lengthwise and placed on the thenar eminence in a belly-tendon configuration. Position of the recording electrodes was optimized for largest CMAP amplitude and fastest CMAP rise time. The recording ground (same type of snap electrode as for stimulation) was placed on the dorsum of the hand. Subjects’ thumbs were taped adducted to the palm to eliminate movement artefact in the recorded CMAP waveforms. An adhesive temperature strip was placed near the recording site and skin temperature was maintained over 32 C throughout the experiments. 2.2.2. Validation of statistical data post processing During VSM, the analogue output of the EA-2 amplifier was digitized using Micro1401 mk II data acquisition unit and Spike2 (version 5) software by CED (Cambridge Electronic Design, Cambridge, UK). Spike2 software was also used in offline analysis to measure the negative peak amplitudes and areas of every CMAP response, and to exclude those responses that were rejected internally by the VSM protocol (Fig. 2.2A). CMAP size values were then exported to MATLAB, using algorithms designed to match the S-MUAP and MUNE calculations of the proprietary Viking software for calculation of pdMUNE and nwMUNE. 2.2.3. Statistical MUNE using QTrac A statistical MUNE protocol was programmed in QTrac# (Institute of Neurology, London, UK) to allow simultaneous testing with superimposed conditioning stimuli. QTrac was interfaced with
30 VSM Setup
A QSM setup
B Fig. 2.2 Schematic representation of the experiments. A: a standard MUNE examination on the Nicolet Viking system while digitizing the EMG signal with separate hardware. Re-analysis of the EMG data validates our implementation of the Poisson Statistical MUNE post-recording algorithm. B: Poisson Statistical MUNE examinations implemented in QTrac with or without DC polarization.
a DS5 stimulator (Digitimer Ltd, Welwyn Garden City, Hertfordshire, UK) via the Micro1401s digital-to-analogue output (Fig. 2.2B). CMAP waveforms were digitized at 10 kHz. Our implementation of the statistical MUNE protocol will hereafter be referred to as QSM (QTrac Statistical MUNE). The QSM protocol, like VSM, started with a stimulus–response (SR) curve (also called “muscle scan”). Unlike VSM, which gives 30 stimuli equally spaced between subthreshold and maximal intensity, QSM begins at zero and increments the stimulus by 2.5% of the supramaximal intensity until the user halts stimulation.
In VSM and QSM the target response windows are based on the negative peak area of the maximum CMAP acquired prior to the SR curve. The target response windows were set at: 10–20%, 25–35%, 40–50%, 55–65% (Fig. 2.3). Unlike VSM where four groups and runs are collected sequentially, in QSM the four runs were recorded concurrently, alternating between the four different stimulus intensities. Between each of the four submaximal stimuli, a supramaximal stimulus was delivered to measure the current maximum CMAP. The motivation behind acquiring runs concurrently rather than sequentially was to circumvent any changes in
Avg. CMAPs (mV)
Stimulus (mA)
31 supramax
20
Run 4 Run 3 Run 2 Run 1
10 0
depolarizing condition hyperpolarizing condition
10 0
20
30
40
Sweep (ms)
NP Area (mV.ms)
Stimulus (mA)
fixed supramax stim ~22mA initial stimulus tracking (not shown)
11
Run 4 Run 3 Run 2 Run 1
40
max
13 12
fixed stimuli
stimulus correction fixed stimuli
…etc…
20 0
SR 46
record Group 2
record Group 1 47
48
49
50
…etc.…
~60% ~45% ~30% ~15%
51
Elapsed time (min)
Fig. 2.3 Screenshot of QTrac MUNE data. All arrows, brackets, and italicized text have been added for this illustration.
axon excitability or CMAP that may occur during the course of an experiment. In a concurrent protocol these changes would affect all runs equally. Between groups, the QTrac# software automatically checked that the response was within the target window and corrected the stimulus amplitude if needed. This correction was designed to account for any drift in axon thresholds during a MUNE procedure. Stimulus intensities were fixed while collecting a group of 30 responses. A total of 10 groups were acquired. All data processing for QSM was done offline, following acquisition of all required EMG responses from the subject. The maximum CMAP negative peak amplitude, the numerator in the MUNE equation (N ¼ max CMAP/mean S-MUAP), was calculated from the average size of all maximum responses recorded during the
10 groups. S-MUAP estimates were calculated for each group using the Viking algorithm [S-MUAP ¼ variance/(mean – min)]. If the standard error of the S-MUAP estimate was less than 10% of the running average S-MUAP, after a minimum of four groups, the remaining groups for that run were discarded. One S-MUAP estimate was generated for each of the four runs and used in the calculations for the two statistical MUNE variants: pdMUNE and nwMUNE. Unlike the response-targeting portion of the protocol, which measured negative peak area, all S-MUAP size estimates were based on the negative peak amplitude of responses. Each subject was tested with VSM and QSM in consecutive sessions without moving the stimulation or recording electrodes so that QSM MUNE outcomes could be validated in a pair-wise comparison against VSM results.
32 2.2.4. Manipulation of axon excitability by polarization To test for the effect of changes in axonal excitability on MUNE, we carried out an additional statistical MUNE protocol, identical to QSM, but with application of 20 ms of 1 mA constant current preceding each stimulus pulse. The polarizing current was maintained during the stimulus pulse and for 30 ms afterward – a total of 50 ms constant current conditioning per stimulus. In this QTrac protocol the two MUNE tests were run concurrently, alternating between stimuli using positive (depolarizing) and negative (hyperpolarizing) bias conditions. This interleaved polarization strategy was implemented to circumvent any cumulative effects of the polarizing current on axonal membrane excitability. In the absence of cumulative polarization effects, the effect of each polarization treatment on axon excitability was assumed to be independent of conditioning effects induced previously within the same polarized MUNE test. SR curves were recorded with both positive and negative polarization prior to starting the concurrent polarized MUNE tests. 2.2.5. Experimental run time Not including the time required for setup and subject preparation, the average times required for each statistical MUNE test were: VSM 23 min, QSM 23 min, and QSM with polarization 46 min. The total stimulation and recording time per subject ranged from 1 h and 21 min to 1 h and 48 min. Tests were run in the same order on each subject: QSM with interleaved polarization, QSM, VSM. 2.2.6. Statistical analysis The large inter-individual variance in MUNE would tend to reduce the power of a between subjects design. Therefore, we chose a within subjects experimental design (repeated measures) to reduce the probability of Type II errors. That is, we used one sample of healthy participants (n ¼ 7) that were tested under all conditions. Three null hypotheses were tested in the following order:
1. There is no difference between MUNE results produced by the Viking system and those calculated from the same EMG sampled simultaneously by the Spike2 system (Fig. 2.2A). 2. There is no difference between MUNE results produced by the VSM and QSM setups (Fig. 2.2). 3. There is no effect of DC polarization on the MUNE results obtained with the QSM setup. If the null hypothesis of either of the first two tests is rejected (P < 0.05), then the third cannot be determined. Statistical analyses were performed using Microsoft Excel 2003 and MATLAB R2006a (Statistics Toolbox version 5.2). Values are reported as average standard deviation unless indicated otherwise. 3. Results 3.1. Reprocessing of Viking EMG data To test the validity of our MUNE calculations, we compared the values reported by the Viking system (VSM) to the values calculated from the simultaneously sampled EMG data (Fig. 2.2A). There was no difference between the results for the two reported MUNE values (paired t-test: pdMUNE P ¼ 0.96, nwMUNE P ¼ 0.94). The small differences between VSM and recalculated MUNEs for each subject (pdMUNE: 12 13; nwMUNE: 3.6 2.0) were attributed to the precision of the measurements from the EMG data. This result confirmed the accuracy of our algorithms for calculating a statistical MUNE result. 3.2. Comparing the statistical MUNE by the Viking and QTrac systems The next step was to compare the MUNE results generated by the new QTrac implementation (Fig. 2.2B) with the standard Viking system. There was no significant difference in the MUNE using the VSM and QSM systems (n ¼ 7 healthy subjects, pdMUNE P ¼ 0.31, nwMUNE P ¼ 0.10).
33 250 pdMUNE r = 0.79
QTrac MUNE
200
nwMUNE r = 0.85
150
100
50
0 0
50
100 150 Viking M UNE
200
250
Fig. 2.4 Data from 7 subjects showing MUNE as estimated by the QTrac system (QSM) versus the expected MUNE based on back-to-back tests with the Viking system (VSM). Both program-determined (pdMUNE) and number-weighted (nwMUNE) estimates are displayed. The solid diagonal line is the line of identity, where the data would lie if Viking and QTrac systems were identical and 100% repeatable.
The MUNE values produced by the two systems were well correlated (pdMUNE r ¼ 0.79, nwMUNE r ¼ 0.85). Figure 2.4 shows that the MUNE generated with the QSM system tended to be lower than the VSM value because most points fell below the line of identity. Despite the tendency for this bias, the null hypothesis could not be rejected, i.e., VSM and QSM give the same MUNE value. We concluded on the basis of this result that the new QSM methodology was an accurate representation of the statistical MUNE performed with the proprietary Viking system. 3.3. Changing axon excitability using DC polarization 3.3.1. Effect on axon thresholds If DC polarization has an effect on axon excitability, there should be a change in the stimulus–response (SR) curves (Fig. 2.1). Based on others’ work (Kiernan and Bostock, 2000; Hales et al., 2004) we expected that depolarization of the axon membranes would cause a shift to the left and a decreased slope of the SR curve.
The opposite effect was expected for a hyperpolarizing DC bias current. To measure the effect of DC polarization we fit a cumulative gaussian function, by the Nelder– Mead simplex direct search, to the SR curve for each individual (n ¼ 8) in each of three conditions: control, depolarized and hyperpolarized. In every case, the curves were well fit (r2 0.995) and yielded two parameters: mean and standard deviation. A decrease in the mean is equivalent to a left shift and an increase in the coefficient of variation (CV ¼ SD/mean) indicated a decreased slope of the SR curve. As expected, depolarization resulted in a decrease of 36.8 14.2% while hyperpolarization resulted in an increase of 30.8 6.5% of the mean (control 18.1 3.6 mA, depolarization 11.7 4.1 mA, hyperpolarization, 23.6 4.1 mA). SR curves for every subject followed this trend and an example is shown in Fig. 2.5A. The coefficient of variation of the SR curves was 13.0 4.4% without polarization, which widened to 23.3 8.8% with depolarization and narrowed to 11.3 5.4% during hyperpolarization. SR curves
.
34 Normalized Stimulus Response Curves
Stimulus Response Curves 100
14
90
No DC −1 mA +1 mA
12
80 70
NP Amp (%)
NP Amp (mV)
10 8 6
60 50 40 30
4
20 2 0
SD 6
8
10
12
14
16
18
20
0 0.5 0.6 0.7 0.8 0.9
22
Stimulus Current (mA)
A
CV
10
1
1.1 1.2 1.3 1.4 1.5
Stimulus (fraction of SR mean)
mean of best-fit normal threshold distribution
B
Fig. 2.5 Representative subject MspA01 where SR curves were well fit by cumulative gaussian functions having CV ¼ 17.4%, 11.0%, and 8.1%, respectively, for depolarization, no polarization, and hyperpolarization treatments.
from every subject but one followed this trend (Fig. 2.5B). These results, summarized in Table 2.1, confirm that DC polarization had an effect on axon excitability. Therefore we could use DC polarization to test the effect of changes in axon excitability on statistical MUNE. From observations reported in previous studies on single axon threshold excitability (Hales et al., 2004), we predicted that the polarization would affect both the mean and the relative spread of the axon thresholds (Fig. 2.1A). Post hoc analysis of the MUNE results combined with refinement of the original computer simulations indicated that the direct current polarization had little
overall effect on relative spreads throughout the pool of motor axons. The details of this conclusion are explained in the Discussion. 3.3.2. Effect on statistical MUNE The final hypothesis to be tested (see Section 2.2.6) was whether DC polarization, via its effect on axon excitability, would change the MUNE generated by the statistical method. MUNE comparisons between treatment conditions were tested using the number-weighted MUNE calculation (Shefner et al., 1999). The number-weighted MUNE values estimated by QSM were chosen because they had a stronger correlation to VSM,
TABLE 2.1 EFFECT OF POLARIZATION ON STIMULUS–RESPONSE CURVES (n ¼ 8) AND QSM (n ¼ 7) Experimental condition No polarization Depol (þ1 mA) Hyperpol (1 mA)
Threshold dist mean (mA)
D in threshold dist mean (%)
CV (%)
nwMUNE by QSM
18.1 3.6 11.7 4.1 23.6 4.1
36.8 14.2 30.8 6.5
13.0 4.4 23.3 8.8 11.3 5.4
109 31 114 28 99 18
35 exhibited less inter-subject variability, and were normally distributed across subjects. Also, the number-weighted calculation has been more widely used in recent clinical trials (Shefner et al., 2004). We found a substantial difference between MUNE values when comparing the depolarized versus hyperpolarized conditions, an average increase 14 15. In the control condition, without DC polarization, the average number of motor units was 109, therefore the effect of axon excitability represents a 13% change in the MUNE. Repeated measures ANOVA showed a significant difference between the three polarization conditions [F (2,12) ¼ 3.93; P < 0.05], However, pairwise comparison (with Bonferroni adjustment for three planned comparisons) revealed that the MUNE differences between any pair of conditions did not reach significance. 4. Discussion MUNE has the potential to be a sensitive, quantitative statistic of the progression of lower motor neuron loss in diseases like ALS (Shefner and Gooch, 2003). MUNE could also be used to gauge the effectiveness of new medical interventions in slowing disease progression. To be a valid measure, the method used to perform a MUNE must be sensitive to the number of intact motor units and insensitive to physiological changes in the neuromuscular system that do not alter the number of motor units. The purpose of this study was to determine if the statistical MUNE method was susceptible to changes in motor axon excitability when the number of motor units remained unchanged.
subject increased on average by 14 motor units (13% of control MUNE) when comparing superimposed depolarization to hyperpolarization. The observed change is similar in magnitude to MUNE decrements that have been attributed to pathological motor unit loss in ALS patients (Aggarwal and Nicholson, 2001). In light of the confounding effects of axon membrane excitability on statistical MUNE, within-patient changes of less than 20% should be interpreted with caution. It is clear from the analysis of the stimulus– response curves that conditioning with direct current polarization induced changes in motor axon excitability (Fig. 2.5). However, these imposed changes in otherwise healthy motor axons are not the same as the changes in axon excitability produced by a disease process. In fact, we suspect that the effects described here would be exacerbated in the pathologic nerve because the efficacy of the DC polarization is a function of distance from the stimulating electrodes on the skin to the axons. Therefore, the change in axon excitability induced by polarization is larger for the lowest-threshold axons that are near the stimulating electrodes. In contrast, pathological excitability changes would not be biased to the lowest-threshold axons. If the disease process affected the excitability of all motor axons in a nerve, the impact on the statistical MUNE would be greater than the effects seen in this study. With that said, the reader should bear in mind the small sample size (n ¼ 7) used in this preliminary study and the bias in QSM toward lower MUNE values when compared to the VSM method (Fig. 2.4). While there was no significant difference in the MUNE from the same subject using QSM and VSM, the difference did approach significance (P ¼ 0.10). A larger study with more participants is needed to resolve this issue.
4.1. Experimental findings 4.2. Comparison to simulated results The results demonstrate that changes in axon excitability produced by DC polarization have a confounding effect on the statistical MUNE method (Daube, 1995). The MUNE from the same
The hypothesis that changes in axon excitability would affect statistical MUNE results arose from computer simulations. The experimental results
36 qualitatively agree with the original predictions of the computer model. However, to quantitatively match the predictions of the model to the experimental results required further post-hoc analysis. In the original computer simulations we assumed that DC polarization would have an effect on all motor axons in the nerve. In addition we assumed that the polarization would affect both the mean and the relative spread of the axon threshold (Fig. 1A). Instead it seems more likely that the DC polarization caused a shift in the mean with little effect on the relative spread. This conclusion was deduced by exploring the following scenarios with the computer model: 1. polarization affects the relative spread of all axon threshold curves, but does not affect the distribution of mean thresholds; 2. polarization shifts and compresses (or expands) the mean threshold distribution, plus affects the relative spread of every axon threshold curve; 3. polarization shifts and compresses the mean threshold distribution, but does not affect relative spread. Only in the third case did the outcomes of the computer simulations match the experimental data. From this analysis we concluded that our original modelling efforts overestimated the effect of polarization. We conclude with this word of caution. Since changes in axonal excitability have been reported for ALS (Priori et al., 2002; Kanai et al., 2006; Vucic and Kiernan, 2006), it is likely that statistical MUNE results in these patients are confounded. In fact, any pathophysiological process or pharmacological intervention that alters motor axon excitability would preclude the use of the statistical method for monitoring the number of motor units innervating a muscle. This result is especially worrying for longitudinal studies that use statistical MUNE as an outcome measure of treatment efficacy.
Summary Motor unit number estimation (MUNE) techniques – whether they reflect a true motor unit count or some related index – should not be confounded by changes in the neuromuscular system other than a decline in the number of functional motor units. In neurodegenerative conditions such as amyotrophic lateral sclerosis (ALS), there is evidence of changes in the excitability of motor axons. If changes in axon excitability confound a particular MUNE technique, this would influence the use of that technique in ALS patients. We hypothesized on the basis of computational models that changes in axon membrane excitability would change the outcome of a statistical MUNE test, even though the true number of motor units remained unchanged. To test the validity of the model predictions we induced changes in axon excitability of healthy control subjects by applying a polarizing current while simultaneously carrying out a statistical MUNE test. In a group of 7 subjects we found a significant difference in MUNE as a result of the change in axon excitability produced by the polarizing current (paired t-test, P < 0.05). We conclude that the statistical MUNE method is confounded by changes in axon excitability. Since increasing evidence shows that axon excitability is altered as part of the pathophysiological process underlying ALS, clinical researchers should be cautious when using statistical MUNE with this patient population. Acknowledgements Supported by grants from the Alberta Heritage Foundation for Medical Research (AHFMR), The Whitaker Foundation, the Canadian Institutes of Health Research, and the Natural Sciences and Engineering Research Council of Canada. Kelvin Jones is an AHFMR Scholar.
37 References Aggarwal, A. and Nicholson, G. (2001) Normal complement of motor units in asymptomatic familial (SOD1 mutation) amyotrophic lateral sclerosis carriers. J. Neurol. Neurosurg. Psychiatry, 71: 478–481. Bostock, H. and Rothwell, J. C. (1997) Latent addition in motor and sensory fibres of human peripheral nerve. J. Physiol. (Lond.), 498: 277–294. Bostock, H., Sharief, M. K., Reid, G. and Murray, N. M. (1995) Axonal ion channel dysfunction in amyotrophic lateral sclerosis. Brain, 118: 217–225. Bostock, H., Cikurel, K. and Burke, D. (1998) Threshold tracking techniques in the study of human peripheral nerve. Muscle Nerve, 21: 137–158. Burke, D., Kiernan, M. C. and Bostock, H. (2001) Excitability of human axons. Clin. Neurophysiol., 112: 1575–1585. Daube, J. R. (1995) Estimating the number of motor units in a muscle. J. Clin. Neurophysiol., 12: 585–594. Hales, J. P., Lin, C. S. and Bostock, H. (2004) Variations in excitability of single human motor axons, related to stochastic properties of nodal sodium channels. J. Physiol. (Lond.), 559: 953–964. Kanai, K., Kuwabara, S., Misawa, S., Tamura, N., Ogawara, K., Nakata, M., Sawai, S., Hattori, T. and Bostock, H. (2006) Altered axonal excitability properties in amyotrophic lateral sclerosis: impaired potassium channel function related to disease stage. Brain, vol: page–page. Kiernan, M. C. and Bostock, H. (2000) Effects of membrane polarization and ischaemia on the excitability properties of human motor axons. Brain, 123: 2542–2551. Kiernan, M. C., Burke, D., Andersen, K. V. and Bostock, H. (2000) Multiple measures of axonal excitability: a new approach in clinical testing. Muscle Nerve, 23: 399–409. Krishnan, A. V., Phoon, R. K., Pussell, B. A., Charlesworth, J. A., Bostock, H. and Kiernan, M. C. (2006) Ischaemia induces paradoxical changes in axonal excitability in endstage kidney disease. Brain, 129: 1585–1592. Kuwabara, S., Misawa, S., Tamura, N., Nakata, M., Kanai, K., Sawai, S., Ogawara, K. and Hattori, T. (2006) Latent addition in human motor and sensory axons: different
site-dependent changes across the carpal tunnel related to persistent Na þ currents. Clin. Neurophysiol., 117: 810–814. Lomen-Hoerth, C. and Olney, R. K. (2000) Comparison of multiple point and statistical motor unit number estimation. Muscle Nerve, 23: 1525–1533. Lomen-Hoerth, C. and Olney, R. K. (2001) Effect of recording window and stimulation variables on the statistical technique of motor unit number estimation. Muscle Nerve, 24: 1659–1664. Misawa, S., Kuwabara, S., Kanai, K., Tamura, N., Nakata, M., Ogawara, K., Yagui, K. and Hattori, T. (2006) Nodal persistent Na þ currents in human diabetic nerves estimated by the technique of latent addition. Clin. Neurophysiol., 117: 815–820. Mogyoros, I., Kiernan, M. C., Burke, D. and Bostock, H. (1998) Strength–duration properties of sensory and motor axons in amyotrophic lateral sclerosis. Brain, 121: 851–859. Nodera, H. and Kaji, R. (2006) Nerve excitability testing and its clinical application to neuromuscular diseases. Clin. Neurophysiol., 117: 1902–1916. Priori, A., Cinnante, C., Pesenti, A., Carpo, M., Cappellari, A., Nobile-Orazio, E., Scarlato, G. and Barbieri, S. (2002) Distinctive abnormalities of motor axonal strength– duration properties in multifocal motor neuropathy and in motor neurone disease. Brain, 125: 2481–2490. Shefner, J. M. and Gooch, C. L. (2003) Motor unit number estimation. Phys. Med. Rehab. Clin. N. Am., 14: 243–260. Shefner, J. M., Jillapalli, D. and Bradshaw, D. Y. (1999) Reducing intersubject variability in motor unit number estimation. Muscle Nerve, 22: 1457–1460. Shefner, J. M., Cudkowicz, M. E., Zhang, H., Schoenfeld, D. and Jillapalli, D. (2004) The use of statistical MUNE in a multicenter clinical trial. Muscle Nerve, 30: 463–469. Simmons, Z., Epstein, D. K., Borg, B., Mauger, D. T., Kothari, M. J. and Shefner, J. M. (2001) Reproducibility of motor unit number estimation in individual subjects. Muscle Nerve, 24: 467–473. Vucic, S. and Kiernan, M. C. (2006) Axonal excitability properties in amyotrophic lateral sclerosis. Clin. Neurophysiol., 117: 1458–1466.
Chapter 2
Change in excitability of motor axons modifies statistical MUNE results Lora A. Majora, K. Ming Chanb, Hugh Bostockc and Kelvin E. Jonesa,* a
Faculty of Physical Education and Recreation, E-488 Van Vliet Centre, University of Alberta, Edmonton, AB T6G 2H9, Canada b
Division of Physical Medicine and Rehabilitation, University of Alberta, Edmonton, AB T6G 2V2, Canada c Sobell Department, Institute of Neurology, Queen Square, London WC1N 3BG, UK
1. Introduction Statistical motor unit number estimation (MUNE) is one of the more widely used methods for estimating the number of functional connections between motor neurons and the muscle fibres they innervate (Daube, 1995). The popularity of the method is largely due to its ease of use and availability within a commercial EMG hardware-software system (VIASYS NeuroCare, Madison, WI, USA). Statistical MUNE estimates the average size of surface motor unit action potentials (S-MUAP) by indirect measurements, and the underlying calculations rely on several intricate and inconspicuous assumptions. Therefore, the method remains somewhat mysterious to the average user and is perhaps the
*
Correspondence to: Kelvin E. Jones, Ph.D., Faculty of Physical Education and Recreation, E-488 Van Vliet Centre, University of Alberta, Edmonton, AB T6G 2H9, Canada. Tel.: (780) 492-0650; Fax: (780) 492-8259; E-mail: [email protected]
least-understood MUNE technique. It is clear that the method uses the statistical variability resulting from peripheral nerve stimulation to perform the calculations that generate the MUNE value. It is not clear, however, whether these calculations are susceptible to physiological changes in axon excitability. Nerve excitability testing focuses on submaximal stimuli (Bostock et al., 1998; Kiernan et al., 2000; Burke et al., 2001; Nodera and Kaji, 2006). Axonal excitability studies have demonstrated that changes in voltage-gated ion channels and the electrogenic Naþ/Kþ pump can be indirectly assessed with the right combination of hardware and software. In recent years axon excitability testing has been successfully applied to a wide range of disorders including diabetes, carpal tunnel syndrome, renal failure and motor neuron disease (Krishnan et al., 2006; Kuwabara et al., 2006; Misawa et al., 2006). The recent work on motor neuron disease, or amyotrophic lateral sclerosis (ALS), prompted the present study. Fasciculations have been interpreted as evidence of a hyperexcitable membrane in motor
28 axons or axon terminals in ALS (Bostock et al., 1995; Mogyoros et al., 1998). Further experiments on axon excitability in ALS suggest that there might be a change in the voltage-gated Na þ or K þ channels (Priori et al., 2002; Kanai et al., 2006; Vucic and Kiernan, 2006). Interpretation of the observed excitability changes was supported by complimentary computer models (Bostock and Rothwell, 1997; Hales et al., 2004). In particular, a stochastic model of the effects of sodium channel dynamics predicted that changes in single motor axon threshold, resulting from elevated levels of persistent sodium current (Hales et al., 2004), may underlie some of the changes in excitability measured in ALS patients. Since motor axon excitability varies as part of the neurodegenerative process of ALS, any diagnostic method used to track the number of remaining motor units must not be influenced by this variation. The statistical MUNE method uses the probabilistic response of motor axons to a constant stimulus to calculate the number of motor
units. We hypothesized that changes in motor axon excitability may alter the MUNE calculated with the statistical method even though the number of functional motor units remained the same. This hypothesis was first tested using a computer model that mimicked the statistical MUNE paradigm. We found that two properties of axon thresholds were important for the validity of the statistical MUNE method: the relative spread (RS) of single axon threshold, and the distribution of the mean axon thresholds within a nerve (Fig. 1.2). Both of these properties are amenable to experimental manipulation using polarization of the axon membrane prior to stimulation with a test pulse (Kiernan and Bostock, 2000; Hales et al., 2004). If the axon membrane is depolarized prior to the test stimulus, the RS tends to increase (Fig. 2.1A) and the distribution of mean axon thresholds becomes more spread out (Fig. 2.1B). The effects are the opposite if the axon membrane is hyperpolarized prior to the test stimulus. We applied depolarizing and
Single Axon Activation Functions
Populations of Mean Axon Thresholds
0.75
Count
Probability of Activation
1
0.5
0.25 S D
A
shift the thresholds
expand the distribution
RS = SD/mean
Mean Thresholds
B
shift the thresholds
original threshold distribution
narrow the dist’n
depolarized distribution hyperpolarized distribution mean mean distribution mean
Fig. 2.1 Excitability properties of motor axon populations: relative spread (RS) of single axon activation curves and distribution of thresholds across the axon population. A: activation curves of two single axons. Probability of activation increases monotonically with stimulus intensity (abscissa). Mean threshold is defined as the stimulus current at which the units have a 50/50 chance of firing. RS quantifies the width of an activation curve normalized to the mean threshold of that axon. B: excitability changes in a population of motor axons may cause expansion, contraction, or a shift in the distribution of mean axon thresholds.
29 hyperpolarizing conditioning pulses at the same time as performing a statistical MUNE test to determine if the changes in axon excitability would change the estimated number of motor units. 2. Materials and methods 2.1. Subjects Eight healthy subjects (4 males) aged 22–38 years were tested. Subjects were pre-screened for neuromuscular dysfunction and gave informed consent to the study, which was approved by the Health Research Ethics Board at the University of Alberta. The MUNE data collected from one female subject were later deemed unreliable due to technical complications and therefore excluded from that analysis. 2.2. Experimental protocol 2.2.1. Standard statistical MUNE We used the Nicolet VikingSelect Master Software version 7.3 (VIASYS NeuroCare, Madison, WI, USA) to perform a statistical MUNE test on each subject. Each test consisted of repeated stimulation to collect compound muscle action potential (CMAP) responses within four target windows: 10–20%, 25–35%, 40–50%, and 55–65% (LomenHoerth and Olney, 2000, 2001). Collection of responses from one target window was called a “run”. Each run consisted of 4–10 groups of 30 stimuli, per the usual statistical MUNE protocol. The Viking software automatically calculated the “program-determined” MUNE (pdMUNE, seen on the Viking readout as “tested þ untested”) and the “number-weighted” modification (nwMUNE) (Daube, 1995; Shefner et al., 1999; Simmons et al., 2001). MUNE results recorded and calculated by the Viking system will hereafter referred to as VSM (Viking Statistical MUNE). The following experimental specifications were applied to all statistical MUNE protocols used for this study. Stimulus pulses of 100 ms width were
applied to the median nerve at the wrist through a “Medi-Trace Mini” (Kendall/Tyco, USA) snaptype, pre-gelled, foam, Ag/AgCl cathode. Stimuli were delivered at 2 pulses/s. The return electrode (anode, same type as the cathode) was placed 100–150 mm proximal and slightly lateral to avoid activation of ulnar or superficial radial nerves. All EMG data were recorded with the EA-2 amplifier (Nicolet Biomedical), which is part of the Viking system. The data were filtered at a bandpass of 10 Hz to 3 kHz. Recording electrodes were “Q-Trace Gold” tab-type, pre-gelled, Ag/AgCl (Kendall/Tyco, USA), cut in half lengthwise and placed on the thenar eminence in a belly-tendon configuration. Position of the recording electrodes was optimized for largest CMAP amplitude and fastest CMAP rise time. The recording ground (same type of snap electrode as for stimulation) was placed on the dorsum of the hand. Subjects’ thumbs were taped adducted to the palm to eliminate movement artefact in the recorded CMAP waveforms. An adhesive temperature strip was placed near the recording site and skin temperature was maintained over 32 C throughout the experiments. 2.2.2. Validation of statistical data post processing During VSM, the analogue output of the EA-2 amplifier was digitized using Micro1401 mk II data acquisition unit and Spike2 (version 5) software by CED (Cambridge Electronic Design, Cambridge, UK). Spike2 software was also used in offline analysis to measure the negative peak amplitudes and areas of every CMAP response, and to exclude those responses that were rejected internally by the VSM protocol (Fig. 2.2A). CMAP size values were then exported to MATLAB, using algorithms designed to match the S-MUAP and MUNE calculations of the proprietary Viking software for calculation of pdMUNE and nwMUNE. 2.2.3. Statistical MUNE using QTrac A statistical MUNE protocol was programmed in QTrac# (Institute of Neurology, London, UK) to allow simultaneous testing with superimposed conditioning stimuli. QTrac was interfaced with
30 VSM Setup
A QSM setup
B Fig. 2.2 Schematic representation of the experiments. A: a standard MUNE examination on the Nicolet Viking system while digitizing the EMG signal with separate hardware. Re-analysis of the EMG data validates our implementation of the Poisson Statistical MUNE post-recording algorithm. B: Poisson Statistical MUNE examinations implemented in QTrac with or without DC polarization.
a DS5 stimulator (Digitimer Ltd, Welwyn Garden City, Hertfordshire, UK) via the Micro1401s digital-to-analogue output (Fig. 2.2B). CMAP waveforms were digitized at 10 kHz. Our implementation of the statistical MUNE protocol will hereafter be referred to as QSM (QTrac Statistical MUNE). The QSM protocol, like VSM, started with a stimulus–response (SR) curve (also called “muscle scan”). Unlike VSM, which gives 30 stimuli equally spaced between subthreshold and maximal intensity, QSM begins at zero and increments the stimulus by 2.5% of the supramaximal intensity until the user halts stimulation.
In VSM and QSM the target response windows are based on the negative peak area of the maximum CMAP acquired prior to the SR curve. The target response windows were set at: 10–20%, 25–35%, 40–50%, 55–65% (Fig. 2.3). Unlike VSM where four groups and runs are collected sequentially, in QSM the four runs were recorded concurrently, alternating between the four different stimulus intensities. Between each of the four submaximal stimuli, a supramaximal stimulus was delivered to measure the current maximum CMAP. The motivation behind acquiring runs concurrently rather than sequentially was to circumvent any changes in
Avg. CMAPs (mV)
Stimulus (mA)
31 supramax
20
Run 4 Run 3 Run 2 Run 1
10 0
depolarizing condition hyperpolarizing condition
10 0
20
30
40
Sweep (ms)
NP Area (mV.ms)
Stimulus (mA)
fixed supramax stim ~22mA initial stimulus tracking (not shown)
11
Run 4 Run 3 Run 2 Run 1
40
max
13 12
fixed stimuli
stimulus correction fixed stimuli
…etc…
20 0
SR 46
record Group 2
record Group 1 47
48
49
50
…etc.…
~60% ~45% ~30% ~15%
51
Elapsed time (min)
Fig. 2.3 Screenshot of QTrac MUNE data. All arrows, brackets, and italicized text have been added for this illustration.
axon excitability or CMAP that may occur during the course of an experiment. In a concurrent protocol these changes would affect all runs equally. Between groups, the QTrac# software automatically checked that the response was within the target window and corrected the stimulus amplitude if needed. This correction was designed to account for any drift in axon thresholds during a MUNE procedure. Stimulus intensities were fixed while collecting a group of 30 responses. A total of 10 groups were acquired. All data processing for QSM was done offline, following acquisition of all required EMG responses from the subject. The maximum CMAP negative peak amplitude, the numerator in the MUNE equation (N ¼ max CMAP/mean S-MUAP), was calculated from the average size of all maximum responses recorded during the
10 groups. S-MUAP estimates were calculated for each group using the Viking algorithm [S-MUAP ¼ variance/(mean – min)]. If the standard error of the S-MUAP estimate was less than 10% of the running average S-MUAP, after a minimum of four groups, the remaining groups for that run were discarded. One S-MUAP estimate was generated for each of the four runs and used in the calculations for the two statistical MUNE variants: pdMUNE and nwMUNE. Unlike the response-targeting portion of the protocol, which measured negative peak area, all S-MUAP size estimates were based on the negative peak amplitude of responses. Each subject was tested with VSM and QSM in consecutive sessions without moving the stimulation or recording electrodes so that QSM MUNE outcomes could be validated in a pair-wise comparison against VSM results.
32 2.2.4. Manipulation of axon excitability by polarization To test for the effect of changes in axonal excitability on MUNE, we carried out an additional statistical MUNE protocol, identical to QSM, but with application of 20 ms of 1 mA constant current preceding each stimulus pulse. The polarizing current was maintained during the stimulus pulse and for 30 ms afterward – a total of 50 ms constant current conditioning per stimulus. In this QTrac protocol the two MUNE tests were run concurrently, alternating between stimuli using positive (depolarizing) and negative (hyperpolarizing) bias conditions. This interleaved polarization strategy was implemented to circumvent any cumulative effects of the polarizing current on axonal membrane excitability. In the absence of cumulative polarization effects, the effect of each polarization treatment on axon excitability was assumed to be independent of conditioning effects induced previously within the same polarized MUNE test. SR curves were recorded with both positive and negative polarization prior to starting the concurrent polarized MUNE tests. 2.2.5. Experimental run time Not including the time required for setup and subject preparation, the average times required for each statistical MUNE test were: VSM 23 min, QSM 23 min, and QSM with polarization 46 min. The total stimulation and recording time per subject ranged from 1 h and 21 min to 1 h and 48 min. Tests were run in the same order on each subject: QSM with interleaved polarization, QSM, VSM. 2.2.6. Statistical analysis The large inter-individual variance in MUNE would tend to reduce the power of a between subjects design. Therefore, we chose a within subjects experimental design (repeated measures) to reduce the probability of Type II errors. That is, we used one sample of healthy participants (n ¼ 7) that were tested under all conditions. Three null hypotheses were tested in the following order:
1. There is no difference between MUNE results produced by the Viking system and those calculated from the same EMG sampled simultaneously by the Spike2 system (Fig. 2.2A). 2. There is no difference between MUNE results produced by the VSM and QSM setups (Fig. 2.2). 3. There is no effect of DC polarization on the MUNE results obtained with the QSM setup. If the null hypothesis of either of the first two tests is rejected (P < 0.05), then the third cannot be determined. Statistical analyses were performed using Microsoft Excel 2003 and MATLAB R2006a (Statistics Toolbox version 5.2). Values are reported as average standard deviation unless indicated otherwise. 3. Results 3.1. Reprocessing of Viking EMG data To test the validity of our MUNE calculations, we compared the values reported by the Viking system (VSM) to the values calculated from the simultaneously sampled EMG data (Fig. 2.2A). There was no difference between the results for the two reported MUNE values (paired t-test: pdMUNE P ¼ 0.96, nwMUNE P ¼ 0.94). The small differences between VSM and recalculated MUNEs for each subject (pdMUNE: 12 13; nwMUNE: 3.6 2.0) were attributed to the precision of the measurements from the EMG data. This result confirmed the accuracy of our algorithms for calculating a statistical MUNE result. 3.2. Comparing the statistical MUNE by the Viking and QTrac systems The next step was to compare the MUNE results generated by the new QTrac implementation (Fig. 2.2B) with the standard Viking system. There was no significant difference in the MUNE using the VSM and QSM systems (n ¼ 7 healthy subjects, pdMUNE P ¼ 0.31, nwMUNE P ¼ 0.10).
33 250 pdMUNE r = 0.79
QTrac MUNE
200
nwMUNE r = 0.85
150
100
50
0 0
50
100 150 Viking M UNE
200
250
Fig. 2.4 Data from 7 subjects showing MUNE as estimated by the QTrac system (QSM) versus the expected MUNE based on back-to-back tests with the Viking system (VSM). Both program-determined (pdMUNE) and number-weighted (nwMUNE) estimates are displayed. The solid diagonal line is the line of identity, where the data would lie if Viking and QTrac systems were identical and 100% repeatable.
The MUNE values produced by the two systems were well correlated (pdMUNE r ¼ 0.79, nwMUNE r ¼ 0.85). Figure 2.4 shows that the MUNE generated with the QSM system tended to be lower than the VSM value because most points fell below the line of identity. Despite the tendency for this bias, the null hypothesis could not be rejected, i.e., VSM and QSM give the same MUNE value. We concluded on the basis of this result that the new QSM methodology was an accurate representation of the statistical MUNE performed with the proprietary Viking system. 3.3. Changing axon excitability using DC polarization 3.3.1. Effect on axon thresholds If DC polarization has an effect on axon excitability, there should be a change in the stimulus–response (SR) curves (Fig. 2.1). Based on others’ work (Kiernan and Bostock, 2000; Hales et al., 2004) we expected that depolarization of the axon membranes would cause a shift to the left and a decreased slope of the SR curve.
The opposite effect was expected for a hyperpolarizing DC bias current. To measure the effect of DC polarization we fit a cumulative gaussian function, by the Nelder– Mead simplex direct search, to the SR curve for each individual (n ¼ 8) in each of three conditions: control, depolarized and hyperpolarized. In every case, the curves were well fit (r2 0.995) and yielded two parameters: mean and standard deviation. A decrease in the mean is equivalent to a left shift and an increase in the coefficient of variation (CV ¼ SD/mean) indicated a decreased slope of the SR curve. As expected, depolarization resulted in a decrease of 36.8 14.2% while hyperpolarization resulted in an increase of 30.8 6.5% of the mean (control 18.1 3.6 mA, depolarization 11.7 4.1 mA, hyperpolarization, 23.6 4.1 mA). SR curves for every subject followed this trend and an example is shown in Fig. 2.5A. The coefficient of variation of the SR curves was 13.0 4.4% without polarization, which widened to 23.3 8.8% with depolarization and narrowed to 11.3 5.4% during hyperpolarization. SR curves
.
34 Normalized Stimulus Response Curves
Stimulus Response Curves 100
14
90
No DC −1 mA +1 mA
12
80 70
NP Amp (%)
NP Amp (mV)
10 8 6
60 50 40 30
4
20 2 0
SD 6
8
10
12
14
16
18
20
0 0.5 0.6 0.7 0.8 0.9
22
Stimulus Current (mA)
A
CV
10
1
1.1 1.2 1.3 1.4 1.5
Stimulus (fraction of SR mean)
mean of best-fit normal threshold distribution
B
Fig. 2.5 Representative subject MspA01 where SR curves were well fit by cumulative gaussian functions having CV ¼ 17.4%, 11.0%, and 8.1%, respectively, for depolarization, no polarization, and hyperpolarization treatments.
from every subject but one followed this trend (Fig. 2.5B). These results, summarized in Table 2.1, confirm that DC polarization had an effect on axon excitability. Therefore we could use DC polarization to test the effect of changes in axon excitability on statistical MUNE. From observations reported in previous studies on single axon threshold excitability (Hales et al., 2004), we predicted that the polarization would affect both the mean and the relative spread of the axon thresholds (Fig. 2.1A). Post hoc analysis of the MUNE results combined with refinement of the original computer simulations indicated that the direct current polarization had little
overall effect on relative spreads throughout the pool of motor axons. The details of this conclusion are explained in the Discussion. 3.3.2. Effect on statistical MUNE The final hypothesis to be tested (see Section 2.2.6) was whether DC polarization, via its effect on axon excitability, would change the MUNE generated by the statistical method. MUNE comparisons between treatment conditions were tested using the number-weighted MUNE calculation (Shefner et al., 1999). The number-weighted MUNE values estimated by QSM were chosen because they had a stronger correlation to VSM,
TABLE 2.1 EFFECT OF POLARIZATION ON STIMULUS–RESPONSE CURVES (n ¼ 8) AND QSM (n ¼ 7) Experimental condition No polarization Depol (þ1 mA) Hyperpol (1 mA)
Threshold dist mean (mA)
D in threshold dist mean (%)
CV (%)
nwMUNE by QSM
18.1 3.6 11.7 4.1 23.6 4.1
36.8 14.2 30.8 6.5
13.0 4.4 23.3 8.8 11.3 5.4
109 31 114 28 99 18
35 exhibited less inter-subject variability, and were normally distributed across subjects. Also, the number-weighted calculation has been more widely used in recent clinical trials (Shefner et al., 2004). We found a substantial difference between MUNE values when comparing the depolarized versus hyperpolarized conditions, an average increase 14 15. In the control condition, without DC polarization, the average number of motor units was 109, therefore the effect of axon excitability represents a 13% change in the MUNE. Repeated measures ANOVA showed a significant difference between the three polarization conditions [F (2,12) ¼ 3.93; P < 0.05], However, pairwise comparison (with Bonferroni adjustment for three planned comparisons) revealed that the MUNE differences between any pair of conditions did not reach significance. 4. Discussion MUNE has the potential to be a sensitive, quantitative statistic of the progression of lower motor neuron loss in diseases like ALS (Shefner and Gooch, 2003). MUNE could also be used to gauge the effectiveness of new medical interventions in slowing disease progression. To be a valid measure, the method used to perform a MUNE must be sensitive to the number of intact motor units and insensitive to physiological changes in the neuromuscular system that do not alter the number of motor units. The purpose of this study was to determine if the statistical MUNE method was susceptible to changes in motor axon excitability when the number of motor units remained unchanged.
subject increased on average by 14 motor units (13% of control MUNE) when comparing superimposed depolarization to hyperpolarization. The observed change is similar in magnitude to MUNE decrements that have been attributed to pathological motor unit loss in ALS patients (Aggarwal and Nicholson, 2001). In light of the confounding effects of axon membrane excitability on statistical MUNE, within-patient changes of less than 20% should be interpreted with caution. It is clear from the analysis of the stimulus– response curves that conditioning with direct current polarization induced changes in motor axon excitability (Fig. 2.5). However, these imposed changes in otherwise healthy motor axons are not the same as the changes in axon excitability produced by a disease process. In fact, we suspect that the effects described here would be exacerbated in the pathologic nerve because the efficacy of the DC polarization is a function of distance from the stimulating electrodes on the skin to the axons. Therefore, the change in axon excitability induced by polarization is larger for the lowest-threshold axons that are near the stimulating electrodes. In contrast, pathological excitability changes would not be biased to the lowest-threshold axons. If the disease process affected the excitability of all motor axons in a nerve, the impact on the statistical MUNE would be greater than the effects seen in this study. With that said, the reader should bear in mind the small sample size (n ¼ 7) used in this preliminary study and the bias in QSM toward lower MUNE values when compared to the VSM method (Fig. 2.4). While there was no significant difference in the MUNE from the same subject using QSM and VSM, the difference did approach significance (P ¼ 0.10). A larger study with more participants is needed to resolve this issue.
4.1. Experimental findings 4.2. Comparison to simulated results The results demonstrate that changes in axon excitability produced by DC polarization have a confounding effect on the statistical MUNE method (Daube, 1995). The MUNE from the same
The hypothesis that changes in axon excitability would affect statistical MUNE results arose from computer simulations. The experimental results
36 qualitatively agree with the original predictions of the computer model. However, to quantitatively match the predictions of the model to the experimental results required further post-hoc analysis. In the original computer simulations we assumed that DC polarization would have an effect on all motor axons in the nerve. In addition we assumed that the polarization would affect both the mean and the relative spread of the axon threshold (Fig. 1A). Instead it seems more likely that the DC polarization caused a shift in the mean with little effect on the relative spread. This conclusion was deduced by exploring the following scenarios with the computer model: 1. polarization affects the relative spread of all axon threshold curves, but does not affect the distribution of mean thresholds; 2. polarization shifts and compresses (or expands) the mean threshold distribution, plus affects the relative spread of every axon threshold curve; 3. polarization shifts and compresses the mean threshold distribution, but does not affect relative spread. Only in the third case did the outcomes of the computer simulations match the experimental data. From this analysis we concluded that our original modelling efforts overestimated the effect of polarization. We conclude with this word of caution. Since changes in axonal excitability have been reported for ALS (Priori et al., 2002; Kanai et al., 2006; Vucic and Kiernan, 2006), it is likely that statistical MUNE results in these patients are confounded. In fact, any pathophysiological process or pharmacological intervention that alters motor axon excitability would preclude the use of the statistical method for monitoring the number of motor units innervating a muscle. This result is especially worrying for longitudinal studies that use statistical MUNE as an outcome measure of treatment efficacy.
Summary Motor unit number estimation (MUNE) techniques – whether they reflect a true motor unit count or some related index – should not be confounded by changes in the neuromuscular system other than a decline in the number of functional motor units. In neurodegenerative conditions such as amyotrophic lateral sclerosis (ALS), there is evidence of changes in the excitability of motor axons. If changes in axon excitability confound a particular MUNE technique, this would influence the use of that technique in ALS patients. We hypothesized on the basis of computational models that changes in axon membrane excitability would change the outcome of a statistical MUNE test, even though the true number of motor units remained unchanged. To test the validity of the model predictions we induced changes in axon excitability of healthy control subjects by applying a polarizing current while simultaneously carrying out a statistical MUNE test. In a group of 7 subjects we found a significant difference in MUNE as a result of the change in axon excitability produced by the polarizing current (paired t-test, P < 0.05). We conclude that the statistical MUNE method is confounded by changes in axon excitability. Since increasing evidence shows that axon excitability is altered as part of the pathophysiological process underlying ALS, clinical researchers should be cautious when using statistical MUNE with this patient population. Acknowledgements Supported by grants from the Alberta Heritage Foundation for Medical Research (AHFMR), The Whitaker Foundation, the Canadian Institutes of Health Research, and the Natural Sciences and Engineering Research Council of Canada. Kelvin Jones is an AHFMR Scholar.
37 References Aggarwal, A. and Nicholson, G. (2001) Normal complement of motor units in asymptomatic familial (SOD1 mutation) amyotrophic lateral sclerosis carriers. J. Neurol. Neurosurg. Psychiatry, 71: 478–481. Bostock, H. and Rothwell, J. C. (1997) Latent addition in motor and sensory fibres of human peripheral nerve. J. Physiol. (Lond.), 498: 277–294. Bostock, H., Sharief, M. K., Reid, G. and Murray, N. M. (1995) Axonal ion channel dysfunction in amyotrophic lateral sclerosis. Brain, 118: 217–225. Bostock, H., Cikurel, K. and Burke, D. (1998) Threshold tracking techniques in the study of human peripheral nerve. Muscle Nerve, 21: 137–158. Burke, D., Kiernan, M. C. and Bostock, H. (2001) Excitability of human axons. Clin. Neurophysiol., 112: 1575–1585. Daube, J. R. (1995) Estimating the number of motor units in a muscle. J. Clin. Neurophysiol., 12: 585–594. Hales, J. P., Lin, C. S. and Bostock, H. (2004) Variations in excitability of single human motor axons, related to stochastic properties of nodal sodium channels. J. Physiol. (Lond.), 559: 953–964. Kanai, K., Kuwabara, S., Misawa, S., Tamura, N., Ogawara, K., Nakata, M., Sawai, S., Hattori, T. and Bostock, H. (2006) Altered axonal excitability properties in amyotrophic lateral sclerosis: impaired potassium channel function related to disease stage. Brain, vol: page–page. Kiernan, M. C. and Bostock, H. (2000) Effects of membrane polarization and ischaemia on the excitability properties of human motor axons. Brain, 123: 2542–2551. Kiernan, M. C., Burke, D., Andersen, K. V. and Bostock, H. (2000) Multiple measures of axonal excitability: a new approach in clinical testing. Muscle Nerve, 23: 399–409. Krishnan, A. V., Phoon, R. K., Pussell, B. A., Charlesworth, J. A., Bostock, H. and Kiernan, M. C. (2006) Ischaemia induces paradoxical changes in axonal excitability in endstage kidney disease. Brain, 129: 1585–1592. Kuwabara, S., Misawa, S., Tamura, N., Nakata, M., Kanai, K., Sawai, S., Ogawara, K. and Hattori, T. (2006) Latent addition in human motor and sensory axons: different
site-dependent changes across the carpal tunnel related to persistent Na þ currents. Clin. Neurophysiol., 117: 810–814. Lomen-Hoerth, C. and Olney, R. K. (2000) Comparison of multiple point and statistical motor unit number estimation. Muscle Nerve, 23: 1525–1533. Lomen-Hoerth, C. and Olney, R. K. (2001) Effect of recording window and stimulation variables on the statistical technique of motor unit number estimation. Muscle Nerve, 24: 1659–1664. Misawa, S., Kuwabara, S., Kanai, K., Tamura, N., Nakata, M., Ogawara, K., Yagui, K. and Hattori, T. (2006) Nodal persistent Na þ currents in human diabetic nerves estimated by the technique of latent addition. Clin. Neurophysiol., 117: 815–820. Mogyoros, I., Kiernan, M. C., Burke, D. and Bostock, H. (1998) Strength–duration properties of sensory and motor axons in amyotrophic lateral sclerosis. Brain, 121: 851–859. Nodera, H. and Kaji, R. (2006) Nerve excitability testing and its clinical application to neuromuscular diseases. Clin. Neurophysiol., 117: 1902–1916. Priori, A., Cinnante, C., Pesenti, A., Carpo, M., Cappellari, A., Nobile-Orazio, E., Scarlato, G. and Barbieri, S. (2002) Distinctive abnormalities of motor axonal strength– duration properties in multifocal motor neuropathy and in motor neurone disease. Brain, 125: 2481–2490. Shefner, J. M. and Gooch, C. L. (2003) Motor unit number estimation. Phys. Med. Rehab. Clin. N. Am., 14: 243–260. Shefner, J. M., Jillapalli, D. and Bradshaw, D. Y. (1999) Reducing intersubject variability in motor unit number estimation. Muscle Nerve, 22: 1457–1460. Shefner, J. M., Cudkowicz, M. E., Zhang, H., Schoenfeld, D. and Jillapalli, D. (2004) The use of statistical MUNE in a multicenter clinical trial. Muscle Nerve, 30: 463–469. Simmons, Z., Epstein, D. K., Borg, B., Mauger, D. T., Kothari, M. J. and Shefner, J. M. (2001) Reproducibility of motor unit number estimation in individual subjects. Muscle Nerve, 24: 467–473. Vucic, S. and Kiernan, M. C. (2006) Axonal excitability properties in amyotrophic lateral sclerosis. Clin. Neurophysiol., 117: 1458–1466.