Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects

Clinical Neurophysiology xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Clinical Neurophysiology journal homepage: www.elsevier.com/locate/clinph

Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects Sally M. Rosengren ⇑ Neurology Department, Royal Prince Alfred Hospital, Sydney, NSW 2050, Australia Central Clinical School, University of Sydney, Sydney, NSW 2006, Australia

a r t i c l e

i n f o

Article history: Accepted 29 December 2014 Available online xxxx Keywords: Sternocleidomastoid muscle EMG VEMP Stimulus intensity Otolith

h i g h l i g h t s  The effect of sternocleidomastoid (SCM) contraction on cVEMP amplitude is strong and linear.  Small muscle contraction differences have the same effect on cVEMP amplitude as 5–9 dB stimulus

intensity changes.  Minimum contraction levels are required for accurate interpretation of cVEMPs.

a b s t r a c t Objective: Cervical vestibular evoked myogenic potentials (cVEMPs) are vestibular-dependent muscle reflexes recorded from the sternocleidomastoid (SCM) muscles in humans. cVEMP amplitude is modulated by stimulus intensity and SCM muscle contraction strength, but the effect of muscle contraction is less well-documented. The effects of intensity and contraction were therefore compared in 25 normal subjects over a wide range of contractions. Methods: cVEMPs were recorded at different contraction levels while holding stimulus intensity constant and at different intensities while holding SCM contraction constant. Results: The effect of muscle contraction on cVEMP amplitude was linear for most of the range of muscle contractions in the majority of subjects (mean R2 = 0.93), although there were some nonlinearities when the contraction was either very weak or very strong. Very weak contractions were associated with absent responses, incomplete morphology and prolonged p13 latencies. Normalization of amplitudes, by dividing the p13–n23 amplitude by the muscle contraction estimate, reduced the effect of muscle contraction, but tended to underestimate the amplitude with weak contractions. Conclusions: Minimum contraction levels are required for accurate interpretation of cVEMPs. Significance: These data highlight the importance of measuring SCM contraction strength when recording cVEMPs. Ó 2015 Published by Elsevier Ireland Ltd. on behalf of International Federation of Clinical Neurophysiology.

1. Introduction Cervical vestibular evoked myogenic potentials (cVEMPs) are muscle reflexes produced by stimulation of the vestibular system with bursts of sound, vibration or galvanic stimulation. They are a type of vestibulo-collic reflex: mediated by the vestibular organs, vestibular nerve and nucleus, vestibulospinal tract, and accessory nucleus and nerve (Uchino and Kushiro, 2011). They are most commonly recorded from the sternocleidomastoid (SCM) neck muscles ⇑ Address: Royal Prince Alfred Hospital, Neurology Department, Level 8, Missenden Rd, Camperdown, NSW 2050, Australia. Tel.: +61 295157565. E-mail address: [email protected]

in response to stimulation with loud bursts of air-conducted (AC) sound (see Rosengren et al., 2010 for review). The reflex can be recorded from an active surface electrode placed near the middle of the SCM muscle belly and a reference over the medial clavicle, and appears as a short-latency, biphasic positive–negative potential with peak latencies of approximately 13 and 23 ms (i.e. p13– n23). cVEMPs evoked by AC sound are thought to originate predominantly in the ipsilateral saccule and are therefore used in neuro-otology settings as a test of saccular function (Rosengren et al., 2010). Left–right amplitude symmetry is typically the most important response metric, however amplitude is affected by two main factors in addition to the integrity of the sacculo-collic pathway: stimulus intensity and SCM contraction strength.

http://dx.doi.org/10.1016/j.clinph.2014.12.027 1388-2457/Ó 2015 Published by Elsevier Ireland Ltd. on behalf of International Federation of Clinical Neurophysiology.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

2

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

Colebatch and Rothwell (2004) demonstrated that the p13–n23 potential is produced by an inhibition of the SCM muscle. In order to detect a reduction in muscle activity the muscle needs to be tonically active. cVEMP amplitude is greater during stronger muscle contractions as more tonically active units can be inhibited by the vestibular stimulus. The effect of contraction strength is common to many reflexes and is thought to be a general property of the motor unit pool, allowing reflexes to scale with contractions in order to maintain appropriate sensitivity (Matthews, 1986). The effect of muscle contraction on the cVEMP was noted in the first detailed report of the reflex (Colebatch et al., 1994), but has been systematically studied only rarely. Several studies have shown that the effect is mostly linear (Akin et al., 2004; Akin and Murnane, 2001; Colebatch et al., 1994; Lim et al., 1995; Watson and Colebatch, 1998), although there are some data suggesting a possible saturation effect with strong contractions (Colebatch et al., 1994; McCaslin et al., 2014). It has also recently been suggested that the relationship might be significantly nonlinear in some subjects (Bogle et al., 2013). However, there has been little detailed analysis of very strong or very weak contractions to date. As the muscle contraction has a significant effect on cVEMP amplitude, electromyogram (EMG) monitoring and measurement are important. The most important goal is to ensure that the contraction is sufficient, so that a response is not missed due to a weak contraction. Previous authors have proposed minimum levels of contraction (e.g. 40–50 lV, Rosengren et al., 2010), though these values are typically based on clinical experience rather than experimental data. The second goal of monitoring the muscle contraction is to ensure a fair comparison across trials or between the left and right sides. If the SCM contraction is asymmetric, a patient may erroneously appear to have an asymmetric cVEMP. Measurement of EMG allows the contraction strength to be matched across trials or sides, or to be used to correct for any contraction asymmetry. In the latter scenario, the EMG estimate is used to express the raw p13–n23 cVEMP amplitude as a ratio (or proportion) of the background muscle contraction, thus normalizing the amplitude measure (sometimes called a ‘corrected value’). If the muscle contraction effect is indeed linear, normalization procedures should cancel it (Colebatch et al., 1994; Lim et al., 1995). This was recently demonstrated at a group level for moderate to strong contractions (McCaslin et al., 2014). Data from normal subjects have also shown that normalized amplitudes are less variable (van Tilburg et al., 2014) and more symmetric (McCaslin et al., 2013; Welgampola and Colebatch, 2001). However, it is not clear if normalization is successful across a wide range of muscle contraction strengths in individual subjects. For example, if cVEMP amplitude saturates during very strong contractions the ratio would be expected to be underestimated for those trials (e.g. McCaslin et al., 2014). Different forms of amplitude correction have been used for some time (Basta et al., 2005; Brantberg et al., 2003; Tseng et al., 2013; Vanspauwen et al., 2009; Welgampola and Colebatch, 2001). However, it is not known if the normalization technique is always successful. One problem is that individual measures or ranges of muscle contraction strength are often not reported in the literature, even in studies specifically examining the effects of muscle contraction. Data are also sometimes averaged over multiple trials or sides, obscuring the natural variability of muscle contractions, and it is important to know the maximum effect that muscle contraction strength or asymmetry can have in a single patient or trial, as individuals are the focus of clinical testing. The primary objective of the current study was therefore to investigate the effect of muscle contraction on cVEMP parameters over a wide range of SCM contraction strengths. Particular emphasis was placed on the range of values obtained, to record the extremes that can occur under different test conditions.

The second objective was to compare the magnitude of the muscle contraction and stimulus intensity effects. The question was: how much is cVEMP amplitude altered by small changes in muscle contraction and what degree of stimulus intensity change would produce the same effect? This enabled the muscle contraction effect to be expressed in terms of equivalent sound intensity units (dB), which are typically more familiar to clinicians than EMG units (e.g. mean rectified EMG in lV), to highlight the magnitude of the background contraction effect.

2. Methods 2.1. Subjects Twenty-five normal volunteers with no history of vestibular dysfunction or neurological disease were tested (mean age 35 years, range 22–62 years; 6 males, 19 females). The participants gave informed written consent according to the Declaration of Helsinki and the study was approved by the local ethics committee (X13-0270 & HREC/13/RPAH/354). 2.2. Stimulation and recording parameters Subjects were stimulated in one ear (12 right, 13 left, selected pseudo randomly). As the effects of two experimental factors were investigated, muscle contraction and stimulus intensity, only one ear was tested to minimise the duration of the experiment. The stimulus was an AC tone burst (500 Hz, 2 ms plateau, 0 ms rise/fall) delivered using headphones and a custom amplifier (TDH 39, Telephonics Corp., Farmingdale, USA). The stimuli were generated with Signal software and a laboratory interface (micro1401, both from Cambridge Electronic Design Ltd [CED], Cambridge, United Kingdom) and delivered at a rate of 7.5 Hz for 200 repetitions per trial. Stimulus intensities ranged from 112 to 136 dB peak sound pressure level (SPL, 84–108 dB LAeq, integrated A-weighted intensity at equivalent sound level). SCM muscle activity was recorded bilaterally (in 21 subjects and ipsilaterally only in 4) from surface electrodes (Clear trace, Conmed Corp., Utica New York, USA) placed over the SCM muscle belly (active, inverting) and medial clavicle (reference). An earth electrode was placed on the sternum. EMG was sampled at 10 kHz from 20 ms before to 80 ms after stimulus onset, amplified and bandpass filtered (5 Hz–2 kHz), using 1902 amplifiers (CED) and the same micro1401 data acquisition interface and Signal software as described above. Negative potentials at the active electrodes were displayed as upward deflections. 2.3. Procedure The effect of stimulus intensity was tested while keeping the SCM muscle contraction relatively constant and the effect of muscle contraction was tested while keeping the stimulus intensity constant. Subjects first reclined to approx. 20° above horizontal, lifted their head off the bed, faced forwards and held their head against gravity. In this position, the test ear was stimulated at maximal intensity (136 dB peak SPL) and EMG from both sides of the neck was measured to assess contraction symmetry (N = 21 subjects). The intensity was then systematically decreased in steps of 3 dB until threshold was reached (N = 25). Trials were repeated near threshold to confirm the presence of a response. The experimenter aimed to keep the muscle contraction on each trial as close as possible to the contraction measured on the first trial. To achieve this, rectified EMG was monitored in real time and subjects were instructed to turn their head slightly during the head lift when required.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

To test the effect of muscle contraction, stimuli were set to 133 dB peak SPL and a combination of turning and lifting actions were performed. This was done to allow a very wide range of contractions to be tested, and it was assumed that the altered head positions would have only a small impact on the measured EMG. Although turning the head can change the relative location of the recording electrode over the muscle, and this may have systematically altered the estimates of EMG, the bipolar montage is relatively resistant to position effects. Subjects were first asked to sit upright, face straight ahead and make no voluntary contraction of the SCM muscle (‘at rest’ condition, N = 22). Recordings were then made with subjects turning their heads about 45° (N = 22) and 90° (N = 16) away from the stimulated ear. There was no instruction about the expected strength of this contraction, only the approximate position of the head, which was controlled visually by the experimenter. The strength of the SCM contraction was then systematically increased over successive trials (N = 25 subjects). This was achieved by first asking subjects to turn their head away from the stimulated ear. Trials were repeated with gradually increasing angle and effort. To further increase the contraction, subjects reclined and lifted their head, then lifted and turned the head away from the stimulus to varying degrees. The bed was then lowered and the lift and turn procedure repeated until the strongest possible contraction that could be maintained over the recording interval was reached. Finally, subjects were asked to use their hand to add resistance to the head turn (N = 21). Breaks were given when required. The total number of trials varied across subjects, as the primary goal was to achieve the largest possible set of trials with graded muscle contractions. Trials were ordered from weak to strong contraction in all subjects to avoid the fatiguing effects of strong contractions carrying over to trials with weak contractions.

3

of stimulus intensity change would produce the same effect. The side differences in contraction levels recorded on the first trial were considered typical of the range of contraction differences likely to be encountered in general. An additional comparison was made in which the effect of changing the contraction by a fixed 20 lV was tested, to demonstrate the likely effect of a larger change in muscle contraction. The 20 lV figure was larger than the median side difference, but within the range of values recorded. 3. Results All subjects had cVEMPs at a minimum of three stimulus intensities, enabling the effects of both stimulus intensity and muscle contraction to be measured. None of the subjects reported any adverse events, such as tinnitus, hearing or neck pain/discomfort, either during the experiment or the following day. 3.1. Effects of muscle contraction 3.1.1. Linearity of muscle contraction effect cVEMP amplitude increased with increasing muscle contraction in all subjects and the effect was linear over most of the range of contractions for nearly all subjects (Figs. 1 and 2). Over all trials, the mean slope was 2.1 (SD 1.0, range 0.6–4.0), the constant was 13.8 (41, 101.5 to 90.1) and the R2 value was 0.93 (0.12, 0.41– 0.99; N = 25). There were nonlinearities in some subjects at very low levels of muscle contraction and at high levels of contraction, particularly when subjects were asked to use their hand to provide additional contraction. For weak contractions, small changes in

2.4. Data analysis The EMG signal from each SCM was split into two channels. The first channel remained unrectified, while the second was full-wave rectified in real time. After averaging 200 repetitions, the cVEMP was measured from the unrectified channel and muscle contraction strength was measured from the rectified channel. To quantify the cVEMP, amplitudes and latencies were measured at the p13 and n23 response peaks. Amplitudes were expressed as the difference of the p13 and n23 amplitudes in lV (raw amplitude). Latencies were adjusted by 0.5 ms to correct for a fixed delay in the recording system (1902 digital filtering). For trials with weak muscle contraction, repetitions including heartbeat (electrocardiogram, ECG) were removed by hand. To quantify the muscle contraction, the mean level of rectified EMG was measured over the 20 ms pre-stimulus period (‘‘mean rectified EMG’’, e.g. Welgampola and Colebatch, 2001). In the text, means (standard deviation [SD], and range) are reported unless otherwise specified. Muscle contraction asymmetry was calculated using the mean rectified EMG recorded on the first trial of the experiment, when subjects lifted their heads and faced straight ahead, using the formula: 100  ((larger EMGsmaller EMG)/(larger EMG + smaller EMG)). Linear regression equations were calculated for each stimulus intensity-cVEMP amplitude function and each muscle contraction-cVEMP amplitude function for all subjects. Differences between selected conditions were tested using paired t-tests and associations between variables were tested with correlations. The significance level was set at a = 0.05. The regression equations were used to compare the effects of muscle contraction and stimulus intensity. First, the side differences in muscle contraction were calculated for each subject. These values were used to calculate how much cVEMP amplitude is altered by small differences in muscle contraction and what degree

Fig. 1. cVEMPs recorded from a single subject at different strengths of SCM muscle contraction. The mean levels of full-wave rectified EMG for each recording are shown on the left. The actions performed to produce these EMG levels were, in order top to bottom: turn approx. 45°, turn approx. 90°, lift and look straight, lift and turn approx. 45°, and lift and turn approx. 90° with strong effort. The lowermost traces are depicted in grey as they have a different scale. At the weakest contraction, a broad p13 response could be seen, followed by a late, non-vestibular peak, but no n23 response was present. As the muscle contraction increased, the n23 response appeared and became larger, and the p13–n23 peak-to-peak amplitude increased.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

4

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

was 466.1 lV (223.2, 149.5–993.6 lV, N = 21) and with the strongest non-resistance contraction was 396.1 lV (195.8, 135.6– 871.8 lV). In two subjects the resistance task produced a weaker rectified EMG level than the previous contraction (Fig. 2D). Due to these obvious effects of very weak and very strong muscle contractions, the regression equations were recalculated after removing all trials where the mean rectified EMG was below an arbitrary cutoff of 50 lV and all trials where subjects used their hands to increase the SCM contraction (Fig. 2). The new values for the slope (mean 2.2, SD 1.0, range 0.8–4.1) and constant (mean 23.8, SD 42.8, range 122.3 to 61.2) were similar to the previous regression, while the range of R2 values improved (mean 0.93, SD 0.07, range 0.76–1.0). These equations were used for all further analyses. The effect was best fit by a linear function in 21/25 subjects, while the remaining 4 subjects showed some decrease in slope with increasing contraction that could be better modelled with a curvilinear function, suggesting saturation of the reflex (e.g. Fig. 2D).

Fig. 2. The effect of muscle contraction on cVEMP amplitude in four representative subjects. Muscle contraction was measured by averaging the rectified EMG over the pre-stimulus interval for each subject. Linear regression equations were calculated for each subject based on the trials in which muscle contractions (performed by lifting and/or turning the head) were stronger than 50 lV (black diamonds). The grey diamonds represent muscle contractions below 50 lV. The white diamonds show trials recorded with resistance (i.e. when subjects pressed their hands to their face to increase the SCM muscle contraction). For the subject in Part A, there was a linear relationship between muscle contraction and cVEMP amplitude over the entire range of contractions tested. This pattern was seen in most subjects. In Part B, the effect was nonlinear at weak contractions and showed some saturation with strong contraction. Part C shows that contractions produced by resistance can have a large impact on the relationship between muscle contraction and cVEMP amplitude. In Part D, the effect of muscle contraction was nonlinear over the entire range of muscle contractions.

rectified EMG sometimes had little effect on cVEMP amplitude (Fig. 2B). For strong contractions, resistance against the hand nearly always increased the level of rectified EMG (in 19/21 subjects), but often led to cVEMPs with smaller amplitude than predicted by the linear regression (in 12/19 subjects; e.g. Fig. 2B and C). The strongest muscle contraction with resistance across subjects was 272.8 lV (106, 109.6–542.7 lV) and without resistance was 198.9 lV (61.8, 64.0–296.3 lV). The resistance task led to a mean 46% increase in rectified EMG compared to the previous trial, but only a 29% increase in cVEMP amplitude (N = 19). The mean cVEMP amplitude with the strongest contraction against resistance

3.1.2. cVEMPs at rest or with head rotation The mean rectified EMG measured at rest (sitting, facing straight ahead) was on average 15.3 lV (4.4, 9.8–24.7 lV) before and 8.3 lV (2.4, 6.1–15.9 lV) after repetitions contaminated by ECG were removed. ECG therefore added a mean of 7 lV (2.9, 2.8–13.3 lV) to the recordings. Only 1/22 subjects had a small cVEMP in this head position. Turning the head by approx. 45° significantly increased the level of muscle contraction, though not to a level that would commonly be accepted as sufficient (mean 10.7 lV, SD 5.2 lV, range 4.1–29.3 lV, t(21) = 2.4, P = 0.024). In this position the number of subjects with cVEMPs rose to 3/22. Turning the head to 90° further increased the contraction strength (mean 29.4 lV, SD 19.1 lV, range 9.1–76.5 lV, t(15) = 4.4, P < 0.001 compared to 45°) and the proportion of subjects with responses rose to 12/16. In this position, 5 subjects had EMG levels that would have previously been considered sufficient (near or over 50 lV) and very clear cVEMPs. For others, there was little increase in EMG compared to the straight ahead condition and the responses were very small, with broad p13 responses and virtually no n23 peak (e.g. top trace in Fig. 1). The mean raw amplitude of the cVEMPs at 90° head rotation was 43.5 lV (30.3, 10.5–96.6 lV, N = 12). Irrespective of head position, the mean smallest level of rectified EMG required for a cVEMP to be present was 24.8 lV (11.4, 6.6–61.5 lV, N = 22). 3.1.3. Muscle contraction asymmetry The muscle contractions measured on each side of the neck on the first trial of the experiment, when subjects were asked to lift their heads off the bed and look straight ahead, were relatively closely matched in most subjects (Fig. 3). The median absolute sideto-side difference was 10.2 lV (N = 21) and values ranged from 1.2 up to 60.7 lV, showing that some subjects had quite asymmetric muscle contractions. When these side differences were expressed as asymmetry ratios, the median value was again quite low (8.8%), but the range extended up to 27.5% (Fig. 3). 3.2. Effects of stimulus intensity While testing the effect of intensity, the average level of mean rectified EMG was 67.0 lV (33.1–100.3 lV over all trials). For 23/ 25 subjects the muscle contraction strength across all intensity trials fell within 10% of the mean contraction, and was within 15% for the remaining two subjects. cVEMP amplitude decreased with decreasing stimulus intensity and the mean threshold was 120.3 dB peak SPL (SD 5.3, range 112– 130 dB peak SPL; equivalent to mean 92.3 LAeq). The relationship between stimulus intensity and raw cVEMP amplitude was

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

5

Fig. 3. Comparison of muscle contraction strength in the ipsilateral and contralateral SCM muscles. On the first trial of the experiment subjects lifted their heads and faced straight ahead, allowing comparison of the typical difference in EMG across the sides (Part A) and calculation of asymmetry values (Part B). EMG was wellmatched in most subjects (median 10.2 lV), but was near or above 20 lV in several subjects. Asymmetry was also generally small (median 8.8%), but exceeded 20% in several subjects. The subject with a side difference of 10.2 lV and asymmetry of 8.4% shown by the white diamonds is the same subject whose data is illustrated in Fig. 4.

well-fit with a linear regression in most subjects. The mean slope was 5.4 (2.0, 2.5–9.9) and the mean constant was 614.8 (243.3, 248.5 to 1163.2). The regression fit (R2) was above 0.80 in all but 3 subjects (mean R2 = 0.88, SD 0.14, range 0.36–0.99). In 15 subjects the relationship was marginally better described by a power function over the intensities tested (mean R2 = 0.88, SD 0.13, range 0.44–0.99), as reported by Dennis et al. (2014). However the difference between the fit of these functions was not significant (t(24) = 0.23, P = 0.820), and therefore for simplicity the linear regressions were used for all further analyses. 3.3. Relationships between variables There was a significant correlation between subject age and cVEMP threshold (r = 0.40, P = 0.048), whereby older subjects had higher cVEMP thresholds. The relationships between the other variables were therefore tested by partial correlations, after removing the contribution of age. This analysis showed a significant correlation between the slope of the muscle contraction effect and cVEMP threshold (r = 0.613, P = 0.004). Subjects with lower cVEMP thresholds, i.e. who were stimulated at higher intensities relative to their thresholds, were more sensitive to the effects of changing muscle contraction. 3.4. Comparing muscle contraction and stimulus intensity The linear regressions described above were used to express the muscle contraction effect in terms of sound intensity units (see Fig. 4). Subjects in whom the effect of muscle contraction was linear were used for this analysis (N = 21). To test how much small differences in muscle contraction affected cVEMP amplitude, the side differences measured on the first trial of the experiment (shown in Fig. 3, Part A) were taken as examples of small contraction differences. The muscle contraction regression equations were

Fig. 4. Comparison of the effects of muscle contraction and stimulus intensity. Part A shows the effect of muscle contraction in a single representative subject stimulated at 133 dB peak SPL, while Part B shows the effect of intensity measured at a fixed muscle contraction of 65 lV in the same subject. This subject had a side difference in muscle contraction of 10.2 lV measured on the first trial of the experiment (see Fig. 3). The regression line shows that increasing the muscle contraction by 10.2 lV increased cVEMP amplitude by 24.8 lV (black dotted lines, Part A). The same cVEMP amplitude increase was achieved by increasing stimulus intensity by 6.3 dB (black dotted lines, Part B). A similar calculation was performed after changing the contraction by 20 lV, to test the effect of a larger side difference (grey dotted lines). A 20 lV increase in muscle contraction increased cVEMP amplitude by 48.6 lV, which was equivalent to a 12.4 dB increase in stimulus intensity. Small differences in muscle contraction have effects on cVEMP amplitude that are comparable to considerable differences in stimulus intensity.

used to calculate the increase in cVEMP amplitude caused by changing the muscle contraction by the side difference (e.g. amplitude at 100 lV rectified EMG versus amplitude at 100+ side difference lV rectified EMG). The mean increase in cVEMP amplitude produced by increasing the muscle contraction was 22 lV (14, 3– 51 lV). The degree of stimulus intensity change required to produce the same effect on cVEMP amplitude was then calculated using the intensity regression equations. The same increase in cVEMP amplitude could be produced by increasing the stimulus intensity by a mean of 4.8 dB (SD 3.2 dB, ranging from 0.7 up to 12.5 dB). A second calculation was performed using a fixed increase in muscle contraction of 20 lV for each subject. This value was chosen to represent a larger difference in muscle contraction level. The mean cVEMP increase produced by increasing the muscle contraction by 20 lV would be 44 lV (21, 17–83 lV). The same increase in cVEMP amplitude could be produced by increasing the stimulus intensity by a mean of 9.2 dB (SD 5.5 dB, ranging from 2.3 up to 21.6 dB). That is, changing the contraction by 20 lV is equivalent to changing the stimulus intensity by 9.2 dB. The relationship was also calculated in reverse. To keep the muscle contraction effect at or below the equivalent of a 5 dB intensity difference, rectified EMG levels would need to remain on average within 16 lV (12, 5–44 lV) of each other.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

6

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

3.5. Correcting for muscle contraction Fig. 5 shows the effect of expressing cVEMP amplitude as a ratio of raw amplitude to background contraction, i.e. normalizing the amplitude. Normalization reduced the effect of muscle contraction considerably. The mean smallest raw amplitude across all subjects was 28.8 lV and the largest was 488.5, while the mean smallest ratio was 0.98 and the largest was 2.33. Thus the ratio measure reduced a 17-fold effect to a 2.4-fold effect. The ratio was constant over a wide range of muscle contractions, effectively cancelling the effect of muscle contraction, though it was most effective when contractions were moderately strong. In trials where the cVEMP seemed to saturate with strong contractions the ratio was smaller than expected (e.g. Fig. 5B). In addition, at weak contractions the ratio tended to systematically underestimate the size of the reflex. The latter effect is due to the constant (or error term) in the function determining cVEMP raw amplitude: y = ax + b (where y is cVEMP amplitude, a is the slope (or gain) of the muscle contraction effect, x is the SCM muscle activity, and b is the constant). When cVEMP amplitude is expressed as a ratio of raw amplitude to measured background contraction the relationship becomes: ratio y/x = a + b/x. The aim of the ratio measure in the clinic is to reveal the value of a, without needing to take multiple measures at different muscle contractions. When there is no constant (b = 0) the ratio is a true indication of a. When there is a constant, its relative contribution will be greater at increasingly smaller muscle contractions (x). In the clinic, the value of b will not be known, and can be either positive or negative. In the current sample most of the values were negative (18/25 subjects), suggesting that a systematic error might be present. One obvious concern is

that this is caused by the measure of muscle contraction being inflated by electrical activity from sources other than the SCM muscle (e.g. from other nearby muscles or ECG). The effect of a negative constant can be seen in Fig. 5A, where the regression lines do not pass through zero. In the ratio measure this leads to a systematic underestimation of the ratio at weak contractions, shown in Fig. 5B. Across the sample, as the contraction became weaker, the measured ratio (y/x) fell below 80% of the slope (a) on two consecutive trials in 18/25 subjects, while in the remaining subjects there was no sustained decrease or increase. The smallest muscle contraction required for the error term to be minimised can be estimated using the current experimental data. For example, the rectified EMG level required to ensure that the ratio (y/x) was within 80% of a was calculated for each subject. The median required contraction strength was 48 lV and the 3rd quartile was 97 lV, while the maximum value (269 lV) was influenced by outliers and was too high to be practically useful. A stricter criterion, to keep the ratio within 90% of a, required median and 3rd quartile EMG levels of 96 and 194 lV, while a more lenient criterion of 70% required median and 3rd quartile EMG levels of 32 and 65 lV. 3.6. Effects on p13 latency Latency was constant across most muscle contraction strengths, but often became prolonged with weaker contractions (Fig. 6). The mean latency at the weakest contraction was 13.7 ms (1.6, 11.2– 17.0 ms) and at the strongest contraction was 12.7 ms (0.7, 11.3– 13.7 ms; t(24) = 3.5, P = 0.002). The mean latency at the 3 strongest contractions was calculated and the latency from each trial was compared to this value. In 11/25 subjects, latency became prolonged by more than 1 ms with weak contractions. In 3 subjects the delay was above 3 ms and the largest delay was 4.1 ms. The mean contraction below which latency exceeded a 1 ms delay was 33.7 lV (12.7, 14.4–60.9 lV, N = 11). There was also a significant effect of stimulus intensity on p13 latency, whereby latencies were on average 0.6 ms longer following the loudest stimulus compared to the softest stimulus (t(24) = 4.5, P < 0.001). The mean latency for the loudest stimulus was 12.8 ms (0.7, 11.2–14.1 ms) and for the softest was 12.2 ms (0.8, 10.9–13.8 ms), while the largest latency difference between these conditions was 1.9 ms. 4. Discussion The strength of the tonic SCM muscle contraction is an important predictor of cVEMP amplitude. The data reported here confirm this effect and show that muscle contraction also has significant effects on cVEMP morphology, latency, and normalized amplitude. In support of previous studies, the data showed that the effect of

Fig. 5. The effect of normalization on cVEMP amplitude in 3 representative subjects. Raw p13–n23 amplitudes (A) and ratios (raw amplitude divided by background muscle contraction) (B) are plotted against muscle contraction. As in Fig. 2, the regression lines in Part A were calculated for the contractions above 50 lV and without use of resistance against the hand to increase the contraction. Part A shows that p13–n23 amplitude varied greatly over the range of contractions studied in all 3 subjects, while in Part B the range was significantly reduced when ratios were calculated. The ratio approximated the slope of the intensity function from Part A for much of the range of contractions. In one subject the ratio provided a good estimate of the slope over the whole range of contractions (grey diamonds), while for the other subjects the slope was underestimated by the ratio at weak contractions. In one subject (grey circles, same subject as in Fig. 2B) the ratio was also smaller than expected at the strongest contraction, showing evidence of response saturation.

Fig. 6. Effect of muscle contraction on p13 latency in 3 individual subjects. The black and grey diamonds show data from 2 subjects in whom p13 latency increased with weaker contractions. The white diamonds show data from a subject in whom latency was relatively constant across the range of contractions.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

SCM muscle contraction on cVEMP amplitude is linear in the majority of subjects over a very wide range of contractions (Akin et al., 2004; Colebatch et al., 1994; Lim et al., 1995). This result is in contrast to a recent report by Bogle et al. (2013), which showed a significantly nonlinear relationship between muscle contraction and cVEMP amplitude. However, their result was likely affected by their method of EMG measurement, which involved measuring the EMG of single repetitions and re-grouping the repetitions in order of contraction strength. EMG measures taken from short periods of single repetitions can vary widely, while the average of many contiguous repetitions will likely better approximate the strength of the underlying muscle contraction during a recording. There was some evidence of saturation of the reflex with very strong contractions, as 4 of 25 subjects showed slowed amplitude growth at stronger contractions. Colebatch et al. (1994) also found saturation in one of their subjects and McCaslin et al. (2014) found evidence of saturation at their 400 lV EMG target level, although it was unclear how many subjects were affected. On the other hand, the majority of subjects showed no saturation, even at maximal contractions up to 500 lV, suggesting that saturation is not an inevitable or widespread consequence of strong SCM contractions. Saturation might be the result of a disproportionate contribution of nearby neck muscles as the contraction increases, altering the relationship between SCM muscle contraction and cVEMP amplitude. This is also likely to be the cause of nonlinearities seen in trials where subjects used their hand increase the muscle contraction. Although the EMG can be assumed to originate predominantly in the SCM as it is a large superficial neck muscle, contributions from other muscles can be expected using the belly-tendon montage and may change with the method of contraction. This might also explain the trials in which the contraction appeared to decrease with resistance, as some subjects may have inadvertently activated other neck muscles, such as trapezius, to steady their head against their hand. Order effects caused by fatigue and electrode position errors caused by varying degrees of head rotation are assumed to be small but cannot be ruled out. At the opposite end of the contraction range, it was possible to record cVEMPs with extremely weak muscle contractions in some subjects, as low as 7 lV (mean 25 lV). However, while it is theoretically possible to record a cVEMP as soon as there are active muscle units available to be inhibited, there are several reasons why weak contractions should be avoided. First, some of the subjects required stronger contractions before a response was seen on the surface (up to 60 lV). There may be a minimum number of active units required to record a synchronous response at the surface. As the rectified EMG is sometimes overestimated at weak contractions, it is difficult to determine the actual minimum levels required, however as these data were collected on a relatively young sample, in whom stimulus intensities were relatively high above cVEMP threshold, the lower EMG limits required to record a response in clinical samples might be higher. Second, during the weakest contractions the morphology of the reflex was altered. The n23 component was sometimes absent in the earliest recordings and gradually appeared as the contraction increased in strength. Third, there was also a systematic effect of weak contraction on p13 latency, whereby latencies were prolonged at the weakest contractions. This may be due to the recruitment properties of the motor unit pool, whereby small motor units with slow conduction velocities are recruited first, followed by progressively larger units with faster conduction velocities (Henneman et al., 1965). During weak contractions the mean size and conduction velocity of the active units will be lower than at stronger contractions, leading to longer latencies. An effect of muscle contraction on latency has not been found previously (Akin et al., 2004; Lim et al., 1995; McCaslin et al., 2014), but the range of contractions extended lower in the current study.

7

The magnitude of the muscle contraction effect differed between subjects and there was a systematic relationship between the slope of the contraction effect and cVEMP threshold. Subjects with low thresholds, in whom the standard high-intensity stimulus would have been relatively strong, were especially sensitive to the effects of muscle contraction. A similar finding was reported by Lim et al. (1995), who tested the effect of muscle contraction at three stimulus intensities and found steeper muscle contraction slopes with louder stimuli. This interaction between intensity and muscle contraction is probably due to each intensity stimulating a fixed proportion of tonically active motor units. At higher intensities the number of motor units inhibited at each incremental EMG level will be proportionately greater than at lower intensities. This suggests that SCM contraction differences will have a disproportionate effect in patients with large reflexes or low thresholds. A main aim of the study was to compare the effects of SCM contraction and stimulus intensity, and to express the muscle contraction effect in equivalent sound intensity units. This comparison showed that small changes in muscle contraction have the same impact on cVEMP amplitude as quite significant intensity differences. Muscle contraction differences of 10–20 lV produced on average the same cVEMP changes as intensity differences of 4.8– 9.2 dB (up to 12.5–21.6 dB). This analysis demonstrated that when a muscle contraction changes by 10–20 lV between trials, or if there are side differences of this magnitude, the effect on cVEMP amplitude is the same as stimulating the ear with quite different stimulus intensities. As audiometers have a tolerance of ±3–5 dB and errors of around 5 dB are commonly accepted in audiology (Haughton, 2002), a muscle contraction difference equivalent to a 5 dB intensity difference might be a reasonable goal. To achieve this, EMG would need to be matched to within about 16 lV on average using the methods described here. Only half of the subjects satisfied this criterion, as no attempt was made to match EMG levels and symmetric posture alone did not guarantee a symmetric contraction. In fact, expressing the side difference in muscle contraction as an asymmetry ratio also served to highlight the potential impact of muscle contraction strength. The asymmetry value approached or exceeded 20% in four subjects and in one subject was as high as 27.5%. This is significant, as a contraction asymmetry will sum with any vestibular asymmetry and potentially change the outcome of the test. The effect of muscle contraction should therefore be controlled whenever possible. The current data showed that normalization of cVEMP amplitude, by expressing the raw amplitude as a ratio of the background contraction, successfully reduced the variability of amplitudes, and was particularly effective when the contraction was moderately strong. Early studies suggested that a ratio measure would theoretically even out the differences caused by unequal muscle contraction (e.g. Colebatch et al., 1994; Lim et al., 1995), and the current study shows that ratios are indeed constant in individual subjects over a large range of muscle contractions. This finding is supported by data from McCaslin et al. (2014), who found no mean difference between normalized amplitudes at target EMG levels between 100 and 400 lV. However, the current data additionally showed that the ratio was underestimated in the subjects who showed saturation at strong contractions and became systematically smaller when the contraction was very weak. The latter effect is because the cVEMP contains a degree of error, shown by the regression constant. Although a constant can either increase or decrease the ratio estimate, the results showed that at weak contractions the ratio was either at or consistently below the level expected. This is likely due to an overestimation of the background contraction. Estimates of muscle activity based on EMG recorded using a belly-tendon montage often include

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

8

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx

elements of unwanted electrical activity, as the electrodes record a sum of activity from all underlying structures and the spatial separation of electrodes renders them susceptible to crosstalk (Chowdhury et al., 2013). The extra signal can be due to biological interference (e.g. ECG, EEG, or EMG from nearby muscles) or sometimes noise due to technical factors (e.g. 50 Hz). For example, the data showed that ECG can add up to 13 lV to the rectified EMG measure. During weak SCM contractions, the proportion of total electrical activity due to extraneous factors will be larger than during strong contractions, producing smaller ratios. This effect is independent of the method of EMG measurement (e.g. rectified versus root mean square [RMS]) and is a potential problem even when the EMG estimate is used only to match the contractions on two sides, as matching will be based on measures of EMG that may be inflated by different degrees. The constant can be reduced by careful skin preparation, symmetric electrode placement, optimal recording conditions and (when appropriate) filters to remove irrelevant biological signals (Chowdhury et al., 2013). However, a pure estimate of SCM activity is probably not possible using this montage. An alternate montage for recording the background contraction has also been suggested, in which the EMG is recorded from a pair of electrodes placed close together over the muscle belly (e.g. Akin and Murnane, 2001; Akin et al., 2011). EMG estimates will be smaller using this montage, as signals common to both electrodes will be cancelled. However, while this method reduces crosstalk, it is sensitive to small changes in electrode position, especially if the electrodes are placed over the motor point (Merletti et al., 2001; Merletti and Hermens, 2004; Mesin et al., 2009). This could result in significant errors in EMG estimates due to small differences in electrode placement or anatomy. While both montages have their pitfalls, the advantage of the belly-tendon montage for estimation of SCM activity is that it is not as sensitive to electrode position and the problem of crosstalk can be minimised by avoiding small muscle contractions. The data suggest that to ensure ratios are within 20% of the actual value for the majority of the sample, the minimum level of mean rectified EMG required is approx. 100 lV. The cut-off value will vary when different techniques for recording muscle activity are used, e.g. estimates of muscle contraction using the same electrode configuration, but based on RMS instead of rectified p EMG, would be higher (100 lV mean rectified EMG = 100 (p/ 2) = 125 lV RMS EMG; De Luca, 1979; Lim et al., 1995). The results clearly show that a minimum muscle contraction is required to ensure the accuracy of both the cVEMP recording and the muscle contraction estimate. Very weak contractions degraded the cVEMP by causing absent responses, delayed latencies and missing n23 peaks. A contraction of at least 60 lV mean rectified EMG was required in this sample to ensure that all subjects had recordable cVEMPs, however the minimum level might be higher when older subjects are tested. A stronger contraction of 100 lV was required to minimize the effect of crosstalk on the muscle contraction estimate. cVEMPs will nearly always be present with contractions below 100 lV and often below 60 lV: there is simply an elevated risk of a type I error, i.e. that responses may be missed, or erroneously appear to be asymmetric or delayed, when they are actually normal. Likewise, EMG matching and normalization of responses at EMG levels below these levels will still reduce the impact of the muscle contraction effect: just not to the same degree as stronger contractions. A minimum value of 100 lV is higher than that proposed previously (e.g. Rosengren et al., 2010), but should protect against all of the abovementioned problems with weak contractions. However, any minimum level obviously needs to be balanced out by practical concerns, as some patients will not be able to produce this level of contraction. A minimum of around 80 lV might be a more reasonable goal in the clinic. The current study did not compare different

methods of achieving an adequate muscle contraction, but showed that simple rotations of the head by about 45° or 90° away from the stimulated ear were not sufficient in most subjects. On the other hand, the data suggest that extremely strong contractions are also not ideal. Saturation can occur in a minority of subjects and very strong contractions using the hand to create resistance also disturbed the linear relationship between muscle contraction and reflex. For these reasons a moderate contraction is preferable, above about 80 lV, but below maximal. This should ensure that attempts to match the EMG on the left and right sides are accurate and the normalization method will cancel out any residual asymmetries in muscle contraction. Overall, the data suggest that if muscle contraction cannot be controlled and/or measured, or a patient cannot achieve a sufficient level of contraction, the results should be interpreted with due caution. Acknowledgements I thank Prof. James Colebatch for his helpful comments on the manuscript. Dr. Sally Rosengren was supported by the National Health and Medical Research Council of Australia (GNT1058056). Conflict of interest statement: There are no potential conflicts of interest to be disclosed. References Akin FW, Murnane OD. Vestibular evoked myogenic potentials: preliminary report. J Am Acad Audiol 2001;12:445–52. Akin FW, Murnane OD, Panus PC, Caruthers SK, Wilkinson AE, Proffitt TM. The influence of voluntary tonic EMG level on the vestibular-evoked myogenic potential. J Rehabil Res Dev 2004;41:473–80. Akin FW, Murnane OD, Tampas JW, Clinard CG. The effect of age on the vestibular evoked myogenic potential and sternocleidomastoid muscle tonic electromyogram level. Ear Hear 2011;32:617–22. Basta D, Todt I, Ernst A. Normative data for P1/N1-latencies of vestibular evoked myogenic potentials induced by air- or bone-conducted tone bursts. Clin Neurophysiol 2005;116:2216–9. Bogle JM, Zapala DA, Criter R, Burkard R. The effect of muscle contraction level on the cervical vestibular evoked myogenic potential (cVEMP): usefulness of amplitude normalization. J Am Acad Audiol 2013;24:77–88. Brantberg K, Tribukait A, Fransson P-A. Vestibular evoked myogenic potentials in response to skull taps for patients with vestibular neuritis. J Vestib Res 2003;13:121–30. Chowdhury RH, Reaz MB, Ali MA, Bakar AA, Chellappan K, Chang TG. Surface electromyography signal processing and classification techniques. Sensors 2013;13:12431–66. Colebatch JG, Halmagyi GM, Skuse NF. Myogenic potentials generated by a clickevoked vestibulocollic reflex. J Neurol Neurosurg Psychiatry 1994;57:190–7. Colebatch JG, Rothwell JC. Motor unit excitability changes mediating vestibulocollic reflexes in the sternocleidomastoid muscle. Clin Neurophysiol 2004;115:2567–73. De Luca CJ. Physiology and mathematics of myoelectric signals. IEEE Trans Biomed Eng 1979;26:313–25. Dennis DL, Govender S, Chen P, Todd NP, Colebatch JG. Differing response properties of cervical and ocular vestibular evoked myogenic potentials evoked by airconducted stimulation. Clin Neurophysiol 2014;125:1238–47. Haughton P. Acoustics for audiologists. London: Academic Press; 2002. Henneman E, Somjen G, Carpenter DO. Excitability and inhibitability of motoneurons of different sizes. J Neurophysiol 1965;28:599–620. Lim CL, Clouston P, Sheean G, Yiannikas C. The influence of voluntary EMG activity and click intensity on the vestibular click evoked myogenic potential. Muscle Nerve 1995;18:1210–3. Matthews PB. Observations on the automatic compensation of reflex gain on varying the pre-existing level of motor discharge in man. J Physiol 1986;374:73–90. McCaslin DL, Fowler A, Jacobson GP. Amplitude normalization reduces cervical vestibular evoked myogenic potential (cVEMP) amplitude asymmetries in normal subjects: proof of concept. J Am Acad Audiol 2014;25:268–77. McCaslin DL, Jacobson GP, Hatton K, Fowler AP, DeLong AP. The effects of amplitude normalization and EMG targets on cVEMP interaural amplitude asymmetry. Ear Hear 2013;34:482–90. Merletti R, Hermens HJ. Detection and conditioning of the surface EMG signal. In: Merletti R, Parker P, editors. Electromyography: physiology, engineering, and noninvasive applications. Hoboken, New Jersey, US: IEEE Press, John Wiley & Sons Inc.; 2004. p. 107–31.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027

S.M. Rosengren / Clinical Neurophysiology xxx (2015) xxx–xxx Merletti R, Rainoldi A, Farina D. Surface electromyography for noninvasive characterization of muscle. Exerc Sport Sci Rev 2001;29:20–5. Mesin L, Merletti R, Rainoldi A. Surface EMG: the issue of electrode location. J Electromyogr Kinesiol 2009;19:719–26. Rosengren SM, Welgampola MS, Colebatch JG. Vestibular evoked myogenic potentials: past, present and future. Clin Neurophysiol 2010;121:636–51. Tseng CC, Wang SJ, Young YH. Comparison of head elevation versus rotation methods for eliciting cervical vestibular-evoked myogenic potentials via boneconducted vibration. Int J Audiol 2013;52:200–6. Uchino Y, Kushiro K. Differences between otolith- and semicircular canal-activated neural circuitry in the vestibular system. Neurosci Res 2011;71:315–27.

9

van Tilburg MJ, Herrmann BS, Guinan Jr JJ, Rauch SD. Normalization reduces intersubject variability in cervical vestibular evoked myogenic potentials. Otol Neurotol 2014;35:e222–227. Vanspauwen R, Wuyts FL, Van de Heyning PH. Vestibular evoked myogenic potentials: test-retest reliability and normative values obtained with a feedback method for the sternocleidomastoid muscle contraction. J Vestib Res 2009;19:127–35. Watson SRD, Colebatch JG. Vestibulocollic reflexes evoked by short-duration galvanic stimulation in man. J Physiol 1998;513:587–97. Welgampola MS, Colebatch JG. Vestibulocollic reflexes: normal values and the effects of age. Clin Neurophysiol 2001;112:1971–9.

Please cite this article in press as: Rosengren SM. Effects of muscle contraction on cervical vestibular evoked myogenic potentials in normal subjects. Clin Neurophysiol (2015), http://dx.doi.org/10.1016/j.clinph.2014.12.027