MONITORING muscular responses to electrical stimulation delivered as single twitches, as train-of four (TOF), or in short bursts, is still the mainstay in the evaluation of the action of muscle relaxants. 1
Although several methods have been described that can be used for this purpose, recording force output from muscle remains the gold standard in monitoring the state of neuromuscular block (NMB). Unfortunately, recording of muscle force output is too complicated for routine clinical use. Furthermore, the application of this method is largely limited to the adductor pollicis muscle (AP), whose response to muscle relaxant is not typical of those muscles subserving respiration or maintaining airway patency. 1
Skeletal muscles vibrate laterally during the build-up of longitudinal forces, producing sound waves or acoustic myography (AMG) that can be detected by microphones or accelerometers fixed to the skin overlying the muscle. 2–5
In normal awake volunteers, the amplitude of AMG has been shown to be proportional to force output during twitch contractions of the hypothenar muscles, 6
the AP, 7,8
and the diaphragm 9
in response to graded transcutaneous nerve stimulation and fatigue. Recent reports from this 10
and one other laboratory 11
also indicated a good relation between acoustic and force responses of small hand muscles during competitive NMB, suggesting the possibility of using AMG to monitor the time course of muscular relaxation or recovery. However, the temporal relation between the acoustic and the force responses during competitive NMB has not been investigated. Temporal relations are more relevant to the pharmacodynamics of muscle relaxants. In the current study we report substantial temporal differences between acoustic and force responses at the AP during rocuronium NMB, suggesting that these two methods, although reflecting similar mechanisms, are not equivalent.
Materials and Methods
Patients and Techniques of Anesthesia
After obtaining approval of the institutional ethics committee and written informed consent, 13 patients with American Society of Anesthesiologists physical status I were enrolled in the study. Age varied from 25 to 55 yr (mean, 38.5 yr). Body mass index varied from 23.5 to 31.5 kg/m2 (mean, 26.7 kg/m2). Six subjects were women. All were undergoing short surgical procedures leaving free access to one arm and requiring only one dose of muscular relaxant. No premedication was given. General anesthesia was induced with propofol 1–2 mg/kg preceded by fentanyl 2–3.8 μg/kg or alfentanil 26–35 μg/kg. After loss of consciousness, rocuronium 0.6 mg/kg was injected as a bolus dose. Anesthesia was maintained with a propofol infusion (eight patients) or a halogenated anesthetic (sevoflurane; five patients). In all patients, blood pressure remained stable during the procedures, and no other complications were noted.
The basic experimental setup used in this study has been described in detail previously 8
and is schematically illustrated in figure 1
. The position of the arm and hand was secured by a brace. The isometric force was measured with a strain gauge connected by a strap to the proximal phalanx of the thumb, amplified and powered by a Hewlett Packard 8805 amplifier (Palo Alto, CA). The length of the strap was adjusted in such a way as to produce a passive tension less than 1 kg.
The AMG signal was recorded by a miniature condenser microphone (Monacor, Farum, Denmark), 6 mm in diameter, embedded in a Plexiglas capsule that was fixed with an adhesive ring to the palm surface of the hand overlying the adductor pollicis. The volume of the air chamber in front of the microphone was 0.09 c3. The AMG signal was amplified and bandpass filtered between 0.5 and 500 Hz with a multichannel signal conditioner (MP100 amplifier; Biopac Systems Inc., Santa Barbara, CA).
Muscle twitches were produced by stimulating the ulnar nerve transcutaneously at the wrist with 0.2-ms square-wave pulses generated by a constant current stimulator (Innervator; Fisher and Paykel Healthcare, Laguna Hills, CA) and a current intensity set at 60 mA.
The neurostimulator and recordings of twitch force, AMG, and synchronizing pulses were started with the injection of rocuronium and continued for 45–50 min. In all patients, the stimulator was set at a twitch rate of 0.1 Hz. In seven patients, TOF stimulation was also used intermittently to monitor the recovery of residual block, starting at approximately 20% recovery of the single twitch.
Signal Processing and Analysis
After amplification, all signals were continuously sampled at a rate of 1,000 Hz using a commercially available software and hardware system (Acknowledge; Biopac Systems Inc.) and stored online on a microcomputer.
The twitch force was measured as the peak amplitude of force above baseline (Pt). The AMG signal was measured as peak-to-peak (AMGpp). The total area above and below the isoelectric line of the AMG signal over a period of 300 ms was also measured. However, both analyses provided very similar results (AMG area % = AMGpp% · 0.98 + 1.07; r2 = 0.93; n= 2,466), such that only the peak-to-peak determinations were retained in the final analysis. Pt and AMGpp were each expressed as a fraction of the mean control value (average of 5–10 measurements) recorded just before the injection of rocuronium. For the responses to TOF stimulation, the amplitudes of the fourth twitch (T4) was expressed as percent of the first twitch (T1).
Peak amplitude of force and AMGpp were each plotted against time after a bolus injection of rocuronium and the time to 90%, 75%, 50%, and 25% blockade determined during onset and during recovery of NMB. The maximum level of block and time to maximum blockade were defined as the minimal twitch height as determined by the computer and the time at which this was found.
The onset and recovery times were compared using a paired t test. Linear and curvilinear regression techniques were used to compare Pt and AMGpp. For the latter relations, the data during the plateau phase of relaxation were omitted to obtain a balanced representation of the various levels of relaxation.
The average bias and limits of agreement between the two methods were determined using the method of Bland and Altman. 12
The bias was determined as the difference between AMGpp
and related to the mean of the two measures of muscular relaxation. The upper and lower limits of agreement between these two methods were determined as ± 2 SD of the differences between AMGpp
. All statistical computations were performed using commercially available software (SPSS Advanced Statistics, version 6.1; SPSS Inc., Chicago, IL).
Time Course of Relaxation and Recovery with Use of Mechanomyography or Acoustic Myography
Representative tracings of the force and sound signals during the onset and recovery of muscular paralysis in one patient are illustrated in figure 2
. The sound signal is a multiphasic oscillating signal. The onset and duration of these oscillations approximate those of the force change.
The average time course of muscular relaxation and recovery obtained by ensemble averaging of Pt
from 13 patients is illustrated in figure 3A
. Pharmacodynamic measures of relaxation for the group are given in table 1
. The maximum depression and the time to maximum depression were not found to be significantly different for these two measures of relaxation. However, AMGpp
showed a slower onset and a faster recovery than Pt
. Indeed, the time to 50% and 75% depression were significantly longer, and the time to 25%, 50%, 75%, and 90% recovery were significantly shorter for AMGpp
than for Pt
Relation between Pt and AMGpp
The relation between AMGpp
was best fitted by a second-order polynomial function. The average relation for the group is shown in figure 3B
. For the 13 patients studied, the average regression coefficient (r2
) for the polynomial function was 0.93 ± 0.07. By comparison, linear and power functions yielded r2
values for the group of 0.84 ± 0.09 and 0.87 ± 0.07, respectively. The linear (1.82 ± 0.56) and the quadratic terms (−0.0078 ± 0.0056) of these polynomials for the group were significantly different from zero, but the intercept (−5.8 ± 12.0) was not.
Representative tracings of Pt
during TOF stimulation are shown in figure 2
. The relation between the fade response of AMGpp
and of Pt
during TOF stimulation is shown in figure 4
and was best fitted by a curvilinear function.
Bias and Limits of Agreement
The bias and limits of agreement between AMGpp
were calculated for 3,311 paired observations during single twitches and for 237 paired observations during TOF stimulation. Group results during twitches and during TOF stimulation are illustrated in figures 5A and 5B
, respectively. For single twitches and for TOF stimulation, there was a significant positive bias of 9.68 ± 13.67% and 17.56 ± 13.42%, respectively, indicating that the AMG tends to underestimate the level of neuromuscular blockade as measured by Pt
. For these calculations, the mean of Pt
averaged 52% for single twitches and 67.5% for TOF stimulation. The lower and upper limits of agreement were, respectively, −17.7% and 37.1% for single twitches and −9.3% and 44.4% for TOF stimulation. For single twitches as well as for TOF stimulation, the bias was greatest at intermediate and low levels of block but virtually nil at high levels of block (i.e.
, at a mean of Pt
As these results show, the onset of NMB was associated with a progressive attenuation of the muscular sound that was progressively restored during recovery. During the recovery period, the muscular sound also displayed a characteristic fade response to TOF stimulation. These findings are in accordance with earlier reports from this 10
and another laboratory, 11
suggesting that acoustic signals from muscles can be used to monitor the onset and recovery from NMB. The maximum level of depression was also comparable for these two methods. However, the acoustic response showed a slower onset and a faster recovery than the twitch force. Thus, although these two methods clearly reflect the onset and the recovery of muscular paralysis at AP, they are not interchangeable.
The curvilinear relation between AMGpp
explains the different time course of AMGpp
and of Pt
that we observed. This relation was best described by a second-order polynomial function. A recent study by Dascalu et al.11
reported a linear relation between the acoustic response recorded over the thenar muscle group and the adduction force at the thumb during NMB achieved with various agents. No definite explanation can be offered for these different results. Their data showed considerably more scatter, and their recording technique differed markedly from ours. They used a large microphone positioned over the thenar muscle, and twitch force appears to have been recorded from the distal phalanx of the thumb. Whether these technical differences can account for the different results cannot be determined from the current study. In our preliminary report, we suggested that a linear relation could provide a reasonably good fit to the acoustic versus
However, this suggestion was based on a limited data set obtained mostly during onset and early recovery. Over the range of NMB investigated then, a linear relation indeed provided a reasonably good fit. Similarly, in the current study, a linear fit accounted for 84% of the variance. However, when the full range of NMB is considered, the relation clearly becomes curvilinear. The cause of this curvilinearity is unclear.
Based on current acoustic models for vibrating muscles, 4,5
the amplitude of the lateral vibrations for a given active axial tension should be determined by the length of the muscle fibers, the elastic modulus of the muscle, and its mass. When these factors are controlled or kept constant, as in the current study, the amplitude of the lateral vibrations (and of the acoustic signal) should be directly related to the active tension. In line with these predictions, previous studies in awake volunteers using the hypothenar muscle group, 6
the AP, 7,8
or the dia-phragm 9
as models, all reported linear relations between the acoustic and the force responses during twitch contractions over a wide range of force. In these studies, twitch force was varied by fatiguing the muscle 6–9
or by varying the intensity of nerve stimulation. 8,9
By contrast, in the current study twitch force was manipulated pharmacologically. This opens the possibility that the curvilinear relation observed here could be related to the pharmacologic properties of muscle relaxants.
Several hand muscles other than the AP are activated by ulnar stimulation. Furthermore, there is evidence in the literature suggesting that the relative sensitivity of different hand muscles to muscle relaxants differ. The first dorsal interosseous muscle, which contributes significantly to the adduction force of the thumb, has been shown in electromyography studies to be more sensitive than AP to nondepolarizing muscle relaxants. 13
Conceivably, the AMG recorded over the palmar surface of the hand may be less affected by the cocontraction of first dorsal interosseous muscle than Pt
. This would cause Pt
to decline more rapidly during onset and to recover more slowly during recovery than AMG. It is of interest that comparable differences have been noted previously when comparing muscular relaxation measured by electromyography and mechanomyography. 14,15
The underlying mechanism may also be comparable. Alternatively, the AMG recorded over AP could be contaminated by the activity of other surrounding muscles having a greater resistance to NMB, causing a slower decay and a faster recovery of AMG. Additional studies will be required to test these possibilities.
Whatever the exact mechanism(s) involved, the curvilinearity of the AMG versus
force relation in our study had significant effects on the parameters usually used to describe the pharmacodynamic properties of muscle relaxants. According to this relation, Pt
would be more sensitive than AMGpp
at low levels of NMB but less sensitive at high levels of NMB. The curvilinearity also explains the significant positive bias in favor of AMG. However, the bias was not distributed uniformly across all levels of NMB. At intermediate and low levels of blockade, both methods differed significantly, but at high levels of NMB, both methods were equivalent. Thus, both methods would appear to be equally valid when assessing conditions that require maximum blockade, such as when assessing intubation conditions, but both would differ when evaluating submaximal blockade or residual block. Indeed, as predicted by the curvilinear relation, the fade response of the AMGpp
recovered faster than the fade response of Pt
. A force TOF fade ratio of 0.7 corresponded an AMGpp
TOF fade ratio of 0.9. In view of the now-recommended recovery of the TOF ratio to values as high as 0.9 at AP, 16
these differences between AMGpp
are likely to be of clinical significance. It will thus be important in future studies to determine whether this difference is inherent to the methods of measurement or whether it is a reflection of a differential sensitivity of hand muscles to competitive muscle relaxants.
In summary, the current study has shown that AMG, although showing an equivalent depression as mechanomyography during maximum blockade at AP, has a lower sensitivity than mechanomyography at intermediate and low levels of blockade. As a result of this difference, AMG displayed a slower onset and a faster recovery than mechanomyography. The two methods, therefore, are not interchangeable at submaximal levels of NMB.
1. Donati F: Neuromuscular monitoring: Useless, optional or mandatory? Can J Anesth 1998; 45: R106–11
2. Wollaston WH: On the duration of muscle action. Philos Trans R Soc London 1810; 100: 1–5
3. Oster G, Jaffe JS: Low frequency sounds from sustained contraction of human skeletal muscle. Biophys J 1980; 30: 119–27
4. Frangioni JV, Kwan-Gett TS, Dobrunz LE, McMahon TA: The mechanism of low-frequency sound production in muscle. Biophys J 1987; 51: 775–83
5. Barry DT, Cole NM: Muscle sounds are emitted at the resonant frequencies of skeletal muscle. IEEE Biomed Eng 1990; 37: 525–31
6. Barry DT: Muscle sounds from evoked twitches in the hand. Arch Phys Med Rehabil 1991; 72: 573–5
7. Stokes MJ, Coopper RG: Muscle sounds during voluntary and stimulated contractions of the human adductor pollicis muscle. J Appl Physiol 1992; 72: 1908–13
8. Chen D, Durand LG, Bellemare F: Time and frequency domain analysis of muscular sounds from human muscles. Muscle Nerve 1997; 20: 991–1001
9. Petitjean M, Bellemare F: Phonomyogram of the diaphragm during unilateral and bilateral phrenic nerve stimulation and changes with fatigue. Muscle Nerve 1994; 17: 1201–9
10. Bellemare F, Donati F, Couture J: The relation of phonomyography to force production in the evaluation of muscle relaxants (abstract). Can J Anaesth 1998; 45: A36b
11. Dascalu A, Geller E, Moalem Y, Manoah M, Enav S, Rudick Z: Acoustic monitoring of intraoperative neuromuscular block. Br J Anaesth 1999; 83: 405–9
12. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 8: 307–13
13. Engbaek J: Monitoring of neuromuscular transmission by electromyography during anaesthesia. Danish Med Bull 1996; 43: 301–16
14. Epstein RA, Epstein RM: The electromyogram and the mechanical response of indirectly stimulated muscle in anesthetized man following curarization. A nesthesiology 1973; 38: 212–23
15. Kopman AF: The relationship of evoked electromyographic and mechanical responses following atracurium in humans. A nesthesiology 1985; 63: 208–11
16. Kopman AF, Yee PS, Neuman GG: Relationship of the train-of-four fade ratio to clinical signs and symptoms of residual paralysis in awake volunteers. A nesthesiology 1997; 86: 765–71
© 2000 American Society of Anesthesiologists, Inc.