VESTIBULAR EVOKED MYOGENIC POTENTIALS: TEST-RETEST RELIABILITY

Transcription

1 VESTIBULAR EVOKED MYOGENIC POTENTIALS: TEST-RETEST RELIABILITY Maurizio Versino, Silvia Colnaghi, Roberto Callieco, Vittorio Cosi Department of Neurological Sciences, University of Pavia and IRCCS C. Mondino Institute of Neurology, Pavia, Italy Reprint requests to: Dr Maurizio Versino, Dipartimento Scienze Neurologiche Università di Pavia, IRCCS Fondazione Istituto Neurologico C.Mondino, Pavia, Via Palestro 3, Pavia, Italy. Accepted for publication: September 14, 2001 Vestibular evoked myogenic potentials (VEMPs) are myogenic responses induced by stimulation of the saccular macula by intense sound stimuli. The responses are recordable from the sternocleidomastoid (SCM) muscles. We recorded VEMPs from normal subjects (up to three times in each subject) to identify: i) the best recording procedures, ii) the reliability, and iii) the normal limits for both individual point and test-retest evaluation. We adopted a recording setting in which the subjects were asked to simultaneously activate both SCM muscles by pushing their forehead against a load cell during a bilateral acoustic stimulation. This system enabled subjects to monitor their intensity of SCM activation and to keep intensity constant; us to record VEMPs from both sides simultaneously, and thus to minimize the duration of the recording session. For each subject we considered the mean and the difference (divided by the mean) of the values derived from the two SCM muscles of the latency of the P13 and N23 components and of the P13-N23 peak-to-peak amplitude. Reliability was evaluated by estimate of the intraclass correlation coefficient, and was good or excellent for all parameters, with the exception of the P13-N23 amplitude side-difference. To take advantage of all the data available, we computed the normal limits for both individual point and test-retest evaluation by means of the variability indices used for the evaluation of reliability. In this system, VEMP recording is simple, inexpensive and rapid. It is well tolerated by subjects, and easily implemented in laboratories equipped for evoked potential recording. KEY WORDS: Click-evoked response, reliability, vestibulocollic reflex. FUNCT NEUROL 2001;16: INTRODUCTION The clinical and instrumental tools used to diagnose vestibular disorders that derive from canal dysfunction are more numerous than those available for disorders pertaining to otolith dysfunction. More specifically, in clinical settings, both caloric and rotatory tests have been widely used to test rotational vestibular ocular reflex (VOR), which is a canal function. In contrast, the evaluation of the otolith function needs very sophisticated equipment, FUNCTIONAL NEUROLOGY (16)

2 M. Versino et al. such as that used to test the linear VOR. Two exceptions to this rule concern the evaluation of: i) subjective visual vertical, namely the ability to judge the orientation of the gravity vector without allocentric visual cues, and ii) vestibular evoked myogenic potentials (VEMPs). VEMPs are obtained through the delivery of acoustic (or galvanic) stimuli that are able to trigger a vestibulospinal response (1) by stimulating the saccular macula (2,3). The response may be recorded from surface electrodes placed over various neck or limb muscles, and is made of 4 peaks. On the basis of their polarity and their latency, these peaks are labeled P13, N23, N34, P44. The first two peaks depend on the integrity of vestibular afferents and are almost invariably detectable, whereas the last two peaks are likely to be a cochlear response and may even be missing in normal subjects (4). In the past, several authors have investigated the features of the VEMPs obtained by means of acoustic stimulation. Their aims were to elucidate the anatomic pathways involved (5), the stimulation and recording features that determine the response (4,6-9), the reliability of the response (10), and the applicability of VEMPs in vestibular dysfunction ( ). H o w e v e r, we still lack a standard for VEMP recording, and the aim of the present study was to identify a method that is easily applicable in any clinically oriented vestibular laboratory, and that obtains a reliable VEMP from the sternocleidomastoid (SCM) muscles. MATERIALS AND METHODS The study was organized into two series of experiments. The first series aimed to investigate the difference between unilateral and bilateral SCM activation and the difference between monaural and binaural stimulation. The second series was based on bilateral and feedback-controlled SCM activation, and aimed to determine normal individual limits both for an individual point and for a test-retest evaluation. Both series of experiments included a reliability study. The features of the acoustic stimuli and of the recording procedures were the same in both experiments, and consisted of 145 db rarefaction clicks lasting 100 µs. Delivery was either monaural or binaural through a TDH 39 headphone, and the repetition rate was 5 Hz. The surface EMG signal was recorded, during a 50 ms period following each click, by an electrode positioned in the upper half of the SCM, with a reference electrode positioned at the mid-point of the ipsilateral clavicle, and with a ground electrode positioned on the upper part of the sternum. The band-pass filter was Hz. We compared two averages of 250 artifact-free responses, to check their reproducibility. We then performed a grand-average of all the 500 responses. The grand-average was then analyzed for P13, N23 latency and P13-N23 peakto-peak amplitude measurements. In addition, for each subject and for each experimental condition (see below), we computed the side-diff e r- ence, defined as the ratio between the diff e r e n c e of the corresponding parameters measured from the right and the left SCM, and the mean of these two values. For instance, for the latency of P13 the side-difference is. The side-difference values were considered for the reliability study and for the computation of normal limits. We used a Nicolet Pathfinder II Plus for all the stimulation, recording and analysis procedures. In the first series of experiments, the VEMPs were recorded from 13 healthy subjects (9 females, 4 males; age range: years) both while they were lying supine and raising their head from the bed by about 45 degrees against gravity (bilateral SCM activation) and while they were seated on a chair with their head turned rightwards or leftwards by about 80 degrees (unilateral SCM activation). All subjects underwent 300 FUNCTIONAL NEUROLOGY (16)4 2001

3 VEMP reliability binaural and monaural stimulation. At 30 minutes after the end of the first recording session, all the subjects underwent a second recording session following the same procedures. The data were analyzed by means of two sets of repeated measure analyses of variance. First, we considered the different SCM activation modes (bilateral/bed and unilateral/chair) separately, and we tested the possible effect of: the retest factor (2 levels: first and second recordings); the kind of stimulation factor (2 levels: mono and binaural); the recording side factor (2 levels: right and left SCM); the interactions between these factors. Then we considered the different kinds of stimulation separately, and we tested the possible effect of the retest factor (2 levels: first and second recordings); the kind of activation factor (2 levels: uni vs bilateral); the recording side factor (2 levels: right and left SCM); the interactions between these factors. In the second series of experiments, VEMPs were recorded in 18 healthy subjects (9 females, 9 males; age range: years) who were asked to sit and push their forehead against a load cell while looking at a display that showed the measured force. Before starting the recording sessions, the subjects were asked to push as hard as they could against the load cell (maximal level of bilateral SCM activation); during the recording sessions, the subjects were asked to maintain SCM activation at about 80% of their maximal level. VEMPs were recorded simultaneously from both right and left SCMs, and all the subjects underwent both monaural and binaural stimulation. All the subjects underwent two recording sessions separated by a 30-minute interval, and 15 subjects underwent an additional retest session 7 days later. The data were analyzed by means of repeated measures analysis of variance to test for a possible effect of the following factors and of their interactions: the retest factor (either 2 or 3 levels, see next paragraph); the kind of stimulation (2 levels: mono or binaural), and the recording side (two levels: right or left). Since not all the subjects underwent all the recording sessions, and in order to keep the data from the 3 subjects u n d e rgoing only one retest, we performed the analysis of variance twice, and considered the respective retest as 2- and 3-level factors. The re l i a b i l i t y of VEMP parameters was evaluated by means of the intraclass correlation coefficient of reliability (R). For a detailed explanation of R, the reader should refer to Fleiss (18). For the purposes of this study, we summarize herein the theoretical basis of R. If we measure a variable in a subject, the score we obtain (X) is made up of two components: the true value of the variable (T: the error-free score) and the random error e, which will vary around a mean of zero with a variance equal to σ 2 e, and which should be constant if we assume that the distribution of errors is independent of the value of T. The existence of the random error can be appreciated as the difference between the X scores when we repeat the same measurement in the same subject and in the same conditions. Therefore, if we repeat the measurement of a variable in a population of subjects several times, the variance of X is σ 2 x = σ 2 e+σ 2 T, in which σ 2 T is the variance of T. R is defined as the ratio between σ 2 T and σ 2 e+σ 2 T. It represents the portion of variability that can be attributed to between-subject variability, and will increase as the ratio σ 2 T/σ 2 e i n c r e a s e s ; in other words, the smaller the variability due to random error the greater the reliability of the measure will be. R ranges from 0 to 1, and it is ranked as follows: values below 0.4 may be taken to represent poor reliability, values between 0.4 and 0.75 may be taken to represent fair to good reliability and values above 0.75 may be taken to represent excellent reliability. Rˆ is the estimate of R, and it can be computed, by measuring a variable in a sample of subjects several times (the number of measurements is not necessarily the same for all subjects), as: FUNCTIONAL NEUROLOGY (16)

4 M. Versino et al. BMS - WMS Rˆ = BMS + 0 WMS BMS is the between-subject mean square, and is computed as: i x (X i - X ) 2 Σ N - 1 in which κ i and X i are respectively the number of measurements performed and the mean value computed for subject i, X is the overall mean of all the subjects and N is the number of subjects. WMS is the within-subject mean square, and is computed as: ( i - 1) s i 2 Σ K - N in which s 2 i is the variance of subject i, and K is the total number of measurements (i.e., 0 i s computed as 0 = where is the mean mean and S 2 κ the variance of k i. It follows that the larger the BMS is with respect to WMS, the higher Rˆ will be. For the convenience of the reader, the computation of normal limits both for individual point and for test-retest will be described in the Results section. RESULTS First series of experiments The first point concerns the occurrence of the evoked responses that were always detectable only ipsilaterally with respect to the stimulation side and regardless of whether SCM activation was uni- or bilateral; for instance, with bilateral SCM activation and monaural stimulation the VEMPs were detectable only from the ipsilateral and not from the contralateral SCM. The P13 and N23 components were invariably detectable in all subjects, and the mean values for their latency and for P13-N23 amplitude are shown in Table 1. The first repeated measures analysis of variance, which separately considered the data respectively obtained by unilateral and bilateral SCM activation, showed that the recording session factor and the interactions with the other two factors did not significantly modify the mean values of any of the parameters either in unilateral or bilateral activations. In unilateral activation, the only significant effect concerned the latency of the N23 component, which was longer for monaural than for binaural stimulation vs ; F 1,12 = 17.02, p = For bilateral activation, the only significant effect concerned the latency of the P13 component, which was longer for monaural than for binaural stimulation, vs ; F 1,12 = 21.05, p = The second repeated measures analysis of variance, which separately considered the data respectively obtained by monaural and binaural stimulation, showed that neither recording session nor recording side factors (or their interaction) influenced VEMP parameters. In contrast, the kind of activation proved to a ffect significantly the VEMPs. The response obtained by bilateral SCM activation showed shorter P13 and N23 latencies and a larger P13-N23 amplitude. More specifically, for monaural stimulation P13 latency: 11.5 vs 12.3; F 1, 1 2 = 10.2, p = 0.008; N23 latency: 19.3 vs 21.8; F 1, 1 2 = 18.1, p = 0.001; P13-N23: vs 79.9, F 1, 1 2 = 16.8, p = 0.001, and for binaural stimulation P13 latency: 11.3 vs 12.1; F 1, 1 2 = 11.7, p = 0.006; N23 latency: 19.3 vs 20.9; F 1, 1 2 = 10.4, p = 0.008; P13-N23 amplitude: vs 96.8; F 1, 1 2 = 15.7, p = The Rˆ values of the VEMP parameters obtained in this first series of experiments are reported in Table FUNCTIONAL NEUROLOGY (16)4 2001

5 VEMP reliability Table 1 - Mean values of P13 and N23 latencies and of P13-N23 peak-to-peak amplitude (P13-N23 amp) obtained during the first series of experiments. Parameter Stimulation Activation Left SCM-test Left SCM-retest Right SCM-test Right SCM-retest P13 (ms) binaural unilateral P13 (ms) binaural bilateral P13 (ms) monaural unilateral P13 (ms) monaural bilateral N23 (ms) binaural unilateral N23 (ms) binaural bilateral N23 (ms) monaural unilateral N23 (ms) monaural bilateral P13-N23 amp (µv) binaural unilateral P13-N23 amp (µv) binaural bilateral P13-N23 amp (µv) monaural unilateral P13-N23 amp (µv) monaural bilateral Abbreviations: SCM = sternocleidomastoid muscle. Table 2 - Intraclass correlation coefficients, Rˆ, of P13 and N23 latencies and of P13-N23 peak-to-peak amplitude (P13- N23 amp) obtained during the first series of experiments Recording Unilateral SCM Bilateral SCM Stimulus side activation activation P13 Binaural Left N23 Binaural Left P13-N23 amp Binaural Left P13 Binaural Right N23 Binaural Right P13-N23 amp Binaural Right P13 Monaural Left N23 Monaural Left P13-N23 amp Monaural Left P13 Monaural Right N23 Monaural Right P13-N23 amp Monaural Right P13 Binaural Left-Right N23 Binaural Left-Right P13-N23 amp Binaural Left-Right P13 Monaural Left-Right N23 Monaural Left-Right P13-N23 amp Monaural Left-Right Abbreviations: SCM = sternocleidomastoid muscle. Rˆ FUNCTIONAL NEUROLOGY (16)

6 M. Versino et al. In summary, the first series of experiments showed that VEMPs are detectable only from the activated SCM ipsilateral to the stimulated e a r. Therefore a simultaneous recording from both SCM muscles requires activation and stimulation to be bilateral. A bilateral stimulation combined with a unilateral activation or a unilateral stimulation combined with a bilateral activation would enable the detection of the response only from the activated or the stimulated side. In addition, as compared to monaural stimulation, binaural stimulation only slightly modifies the quantitative features of the response. Bilateral SCM activation induces shorter P13 and N23 latencies and a larger P13-N23 amplitude, than unilateral SCM activation does. At least as far as the P13-N23 amplitude is concerned, this can probably be explained by the fact that SCM activation is larger in our bilateral than it is in our unilateral paradigm. There was no significant difference between the test and retest response, regardless of activation and stimulation modes. The reliability indices were similar for both unilateral and bilateral SCM activation, and proved to be very low for the amplitude side-difference. Second series of experiments None of the parameters was significantly correlated with the force expressed by the different subjects during the experiment, which ranged from 3.5 to 11 kg. The repeated measures analysis of variance was performed twice to include all the subjects (2-level retest factor) or the 15 subjects who underwent all three recording sessions (3-level retest factor). For both analyses, none of the main factors (retest, kind of stimulation and recording side) and none of their interactions proved to significantly affect the parameters of the response, with the sole exception of longer P13 latency for monaural than for binaural stimulation (11.38 b vs m; F 1, 1 7 = , p = 0.006). This exception was valid only on the basis of a 2 retest scenario. The values are reported in Table 3. The Rˆ values were excellent for most of the parameters. Rˆ was slightly larger in the case of binaural stimulation and the poorer values were those detectable for the side-diff e r e n c e amplitude values. Normal limits We computed the normal limits both for an individual point and for a test-retest evaluation. We considered only binaural stimulation, since it presents no differences vs monaural stimulation (as previously reported, the factor kind of stimulation is never significant) and less time is needed to perform the test. Moreover, since there are no significant differences between the VEMPs recorded from the two SCM muscles, i.e., in normal subjects the side-difference for a given parameter is not statistically diff e r e n t from 0 (the factor recording side is never significant), we computed the normal limits both for the mean and for the difference of the corresponding values recorded from different sides. For individual point evaluation, we wanted to take into account all the data available (including those from the same subject but from d i fferent recording sessions). Accordingly, we computed the normal limits for the mean by considering the mean value of the mean values of all the subjects weighted by the number of measurements (weighted mean) and the square root of BMS (SQBMS) as a variability index. The 95% confidence interval to define normal limits was therefore computed as weighted mean 2 ± SQBMS. The N13-P23 mean amplitude values presented with a right-tailed distribution, and we set the 5 th and 95 th percentile as the lower and upper normal limits of that distribution. For the computation of the normal limits for both latency and amplitude side-differences, we considered the mean of these differences as 304 FUNCTIONAL NEUROLOGY (16)4 2001

7 VEMP reliability Table 3 - Intraclass correlation coefficients, Rˆ, of P13 and N23 latencies and of P13-N23 peak-to-peak amplitude (P13-N23 amp) obtained during the second series of experiments. Stimulus Recording side Rˆ Rating P13 Binaural Left 0.82 Excellent N23 Binaural Left 0.93 Excellent P13-N23 amp Binaural Left 0.83 Excellent P13 Binaural Right 0.59 Good N23 Binaural Right 0.89 Excellent P13-N23 amp Binaural Right 0.91 Excellent P13 Monaural Left 0.69 Good N23 Monaural Left 0.89 Excellent P13-N23 amp Monaural Left 0.83 Excellent P13 Monaural Right 0.74 Excellent N23 Monaural Right 0.86 Excellent P13-N23 amp Monaural Right 0.76 Excellent P13 Binaural Left-Right 0.7 Good N23 Binaural Left-Right 0.71 Good P13-N23 amp Binaural Left-Right 0.49 Poor P13 Monaural Left-Right 0.52 Good N23 Monaural Left-Right 0.69 Good P13-N23 amp Monaural Left-Right 0.45 Poor equal to 0 (as suggested by the analysis of variance, see above) and the square root of BMS (SQBMS) as variability index. Accordingly, the 95% confidence interval to define normal limits was computed as ± 2 SQBMS. For test-retest evaluation, the maximal retest d i fference (D i f M a x) was computed with WMS as the variability index, thus for parameter z, D i f M a x z = 2 x F ; W M S z ( 1 9 ). Table 4 reports all the normal limits (see o v e r ). DISCUSSION The main features of the VEMPs that we recorded from the SCM in response to intense acoustic stimuli can be resumed in a few points: the response requires activation of the SCM muscle; it consists of three to four components but only the first two (P13 and N23) are detectable in all subjects; it does not change according to whether the acoustic stimulation is unilateral or bilateral; it is recordable only ipsilaterally to the stimulated side. Most of these findings are in keeping with those already reported in the literature, with the sole exception of the last point, which is still open to debate. This point is probably related to the choice of the reference electrode, as proposed in a recent paper by Li et al (6), whose findings support the hypothesis that the responses are purely ipsilateral. We suggest binaural stimulation and bilateral SCM activation, and the use of activation intensity monitoring by means of a load cell. The benefits are: constant SCM activation, short FUNCTIONAL NEUROLOGY (16)

8 M. Versino et al. Table 4 - Normative data from the VEMP individual point and test-retest evaluations of 18 normal subjects P13 N23 P13-N23 AMP P13 (ms) N23 (ms) P13-N23 AMP (µv) (ms) (ms) (µv) left-right left-right left-right N Mean SD Lowest value Highest value BMS WMS Percentile Percentile Lower limit Upper limit Difmax P13, N23 and P13-N23 AMP values = the mean of the latency and amplitude values measured from the two SCM muscles of each subject; P13 left-right, N23 left-right, and P13-N23 AMP left-right = the difference between the latency and amplitude values measured from the two SCM muscles of each subject divided by the corresponding mean values. Abbreviations: N = no. of patients; SD = standard deviation; Lowest value = lowest observed value; Highest value = highest observed value; BMS = between subject mean square. WMS: within subject mean square; Lower limit = lower normal limit for individual point evaluation (for P13-N23 amplitude this corresponds to Percentile 5 - see the text). Upper limit = upper normal limit for individual point evaluation (for P13-N23 amplitude this corresponds to Percentile 95 - see the text); DifMax = maximal normal test-retest difference. test duration and a comfortable setting for the subject. The need for constant activation derives from the relationship between the P13-N23 amplitude and SCM activation. This relationship is very strong within the same subject, but variable between subjects (7). These two findings may explain why we did not find a similar relationship when we pooled the data from all of our subjects. Moreover, although the values from d i fferent subjects varied greatly, they were still comparable since they all corresponded to 80% of their respective maximal intensity levels. Fin a l l y, the display connected to the load cell provided a visual feedback control to help the subject keep activation constant throughout the recording sessions, and thus to improve the reliability of the response. An alternative method to monitor SCM activation would be to measure the average background rectified EMG amplitude (20), and this value should be used to adjust the P13-N23 peak-to-peak amplitude in order to minimize inter-individual variability (note that we did this by asking all the subjects to express 80% of their own maximal force). We measured the latency of P13 and N23 and the P13-N23 peak-to-peak amplitude. Since the parameters of the VEMPs obtained simultaneously from the two SCM muscles of any given subject proved not to differ statistically from those of normal subjects, we suggest consideration be based on both the mean of and the difference (divided by the mean) between the values from the two sides. We evaluated the reliability of all these parameters by means of the estimate of the intraclass correlation of relia b i l i t y, Rˆ. Most of the parameters showed ex- 306 FUNCTIONAL NEUROLOGY (16)4 2001

9 VEMP reliability cellent or good values, and the side-difference values, especially those involving the P13-N23 amplitudes, showed the lowest coeff i c i e n t s. This finding is explicable since Rˆ evaluates reliability by comparing the intraindividual (WMS - Within Subject Mean Square) with the interindividual (BMS - Between Subject Mean Square) variability, and since for each subject we considered the ratio of the side-diff e r e n c e to the corresponding mean value (rather then the side-difference per se), this transformation is very likely to reduce the BMS. Similarly to us, Ferber- Viart et al. (10) retested the VEMPs in the same subjects, and concluded that VEMP parameters were reliable, since there was no difference between the mean values computable for different recording sessions as evaluated by analysis of variance. However, their results suggested only that the amount of variability between different recording sessions for the whole group of subjects did not differ from zero, but this does not imply good reliability. Indeed, this result can be obtained in spite of large intraindividual variations, i.e., in poor reliability con- Left SCM Right SCM A Left ear stimulation Right ear stimulation B Fig. 1 - A shows the vestibular evoked myogenic potentials recorded in a patient with multiple sclerosis. The lower tracing is the average of the two recordings made on each side and shown in the uppermost part of the figure. This patient showed a slightly increased averaged P13 latency (14.6ms). The side-difference was abnormal both for P13 latency (0.48ms) and for N23 latency (0.29) due to a delay from left SCM. B shows the brainstem auditory evoked potentials (BAEPs) recorded from the same patient, that proved to be markedly abnormal for monaural left stimulus suggesting a brainstem dysfunction: peak I was normal, peak III was absent and peak V had minimal amplitude and normal latency with normal inter-peak I-V interval duration. The BAEPs for monaural right stimulus were normal. FUNCTIONAL NEUROLOGY (16)

10 M. Versino et al. Left SCM Right SCM Fig. 2 - Vestibular evoked myogenic potentials recorded in a patient with multiple sclerosis. The lower tracing is the average of the two recordings made on each side and shown in the uppermost part of the figure. This patient showed an increased averaged latency of P13 and an increased side-difference both of P13 latency (0.58) and N23 latency (0.42) due to a delay from left SCM. In this patient the brainstem auditory evoked potential was normal from both sides. ditions, provided that the given variations have d i fferent signs in different subjects (for instance, a l a rge increase in one subject counterbalanced by a large decrease in another subject). Since we wanted to take advantage of all the data available (from the subject group as a whole, and from each individual subject) the computation of normal limits stemmed from our reliability study. We used the BMS to compute the limits for individual point evaluation (with the single exception of P13-N23 amplitude), and the WMS to compute the limits for test-retest evaluation. The side-difference normal limits for individual point evaluation proved to be large; they were comparable to, albeit larger than, those reported by Heide et al. (14) for the VEMPs, and similar to the limits used for caloric testing to define canal paresis. VEMP recording can therefore be simple, inexpensive and fast, using a method that is well tolerated by subjects, and easily implemented in laboratories equipped for recording evoked potentials. The method seems suitable for the evaluation of dizzy patients, in addition to the other vestibular tests, and provides reliable information about otolith function. This may be important, since the symptoms induced by otolith dysfunction may be so vague that it points to a diagnosis of psychogenic vertigo. F i n a l l y, VEMPs may be used for brainstem neurophysiological evaluation in patients not necessarily complaining of balance disturbance. This may be the case of patients suffering from multiple sclerosis (MS), in whom VEMPs may be used in addition to brainstem auditory evoked potentials (BAEPs). Figure 1 shows a patient in whom both VEMPs and BAEPs were abnormal. In contrast, figure 2 shows a patient in whom BAEPs were normal whereas VEMPs showed increased P13 latency and an abnormal P13 and N23 side-difference. I n t e r e s t i n g l y, and similarly to the report by Shimizu et al. (21), MS patients showed a latency delay rather than the reduced amplitude response that is usually reported for vestibular dysfunction. 308 FUNCTIONAL NEUROLOGY (16)4 2001

11 VEMP reliability REFERENCES 11. Bickford RG, Jacobson JL, Cody DTR. Nature of averaged evoked potentials to sound and other stimuli in man. Ann NY Acad Sci ; 11 2 : Murofushi T, Curthoys IS, Gilchrist DP. Response of guinea pig vestibular nucleus neurons to clicks. Exp Brain Res 1996; 111: Murofushi T, Curthoys IS, Topple AN, Colebatch JG, Halmagyi GM. Responses of guinea pig primary vestibular neurons to clicks. Exp Brain Res 1995;103: Colebatch JG, Halmagyi GM, Skuse NF. Myogenic potentials generated by a clickevoked vestibulocollic reflex. J Neurol Neurosurg Psychiatry 1994a;57: Kushiro K, Zakir M, Ogawa Y, Sato H. Saccular and utricular inputs to sternocleidomastoid motoneurons of decerebrate cats. Exp Brain Res 1999;126: Li MW, Houlden D, Tomlinson RD. Click evoked EMG responses in sternocleidomastoid muscles: characteristics in normal subjects. J Vestib Res 1999;9: 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: Murofushi T, Matsuzaki M, Wu CH. Short tone burst-evoked myogenic potentials on the sternocleidomastoid muscle: are these potentials also of vestibular origin? Arch Otolaryngol Head Neck Surg 1999;125: Wu CH, Murofushi T. The effect of click repetition rate on vestibular evoked myogenic potential. Acta Otolaryngol (Stockh) 1999;119: F e r b e r- Viart C, Duclaux R, Colleaux B, Dubreuil C. Myogenic vestibular- e v o k e d potentials in normal subjects: a comparison between responses obtained from sternomastoid and trapezius muscles. Acta Otolaryngol (Stockh) 1997;11 7 : Colebatch JG, Hamalgyi GM. Ve s t i b u l a r evoked potentials in human neck muscles before and after unilateral vestibular deafferentation. Neurology 1992;42: Colebatch JG, Rothwell JC, Bronstein A, Hudman H. Click-evoked vestibular activation in the Tullio phenomenon. J Neurol Neurosurg Psychiatry 1994;57: de Waele C, Huy PT, Diard JP, Freyss G, Vidal PP. Saccular dysfunction in Meniere s disease. Am J Otol 1999;20: Heide G, Freitag S, Wo l l e m b e rg I, Iro H, Schimrigk K, Dillmann U. Click evoked potentials in the differential diagnosis of acute vertigo. J Neurol Neurosurg Psychiatry 1999;66: Matsuzaki M, Murofushi T, Mizuno M. Vestibular evoked myogenic potentials in acoustic tumor patients with normal auditory brainstem responses. Eur Arch Otolaryngol 1999;256: Murofushi T, Halmagyi GM, Yavor RA, Colebatch JC. Absent vestibular evoked myogenic potentials in vestibular neurolabyrinthitis. An indicator of inferior vestibular nerve involvment? Arch Otolaryngol Head Neck Surg 1996;122: Murofushi T, Matsuzaki M, Mizuno M. Vestibular evoked myogenic potentials in patients with acoustic neuromas. Arch Otolaryngol Head Neck Surg 1998;124: Fleiss JL. Reliability of measurement. In: The Design and Analysis of Clinical Experiments. New York; John Wiley & Sons 1985: Versino M, Castelonovo G, Berg a m a s c h i R et al. Quantitative analysis of saccadic and smooth pursuit eye movements. Is it reliable? Invest Ophthalmol Vis Sci 1993; 34: Colebatch JG, DiLazzaro V, Quartarone A, Rothwell JC, Gresty M. Click-evoked vestibulocollic reflexes in torticollis. Mov Disord 1995;10: Shimizu K, Murofushi T, Sakurai M, Halmagyi M. Vestibular evoked myogenic potentials in multiple sclerosis. J Neurol Neurosurg Psychiatry 2000;69: FUNCTIONAL NEUROLOGY (16)