Acceleromyography for Use in Scientific and Clinical Practice: A Systematic Review of the Evidence

ACCELEROMYOGRAPHY for clinical use in anesthesia was introduced in 1988.1,2Evidence indicated that postoperative residual curarization (PORC) was a problem3and that there was a need for a simple and user-friendly method of neuromuscular monitoring for use in the clinical setting. In contrast to the more cumbersome methods of electromyography and mechanomyography, acceleromyography might fulfill these criteria. Contrary to mechanomyography, which is based on isometric measurements, and electromyography, which is based on measurement of the compound action potential, acceleromyography in its original form was based on isotonic measurements (freely moving thumb). The theory behind acceleromyography is based on Newton’s second law of motion, force = mass × acceleration. When mass is constant, acceleration is directly proportional to force. For measurement of acceleration, an acceleration transducer is normally used, consisting of a piezoelectric ceramic wafer embedded within a suitable housing (fig. 1). Whenever the piezoelectric wafer is moved (accelerates), a voltage is generated, and if the transducer is fixed to a digit or muscle, any movement generates an electric signal. The signal is subsequently conditioned, analyzed, and recorded in a monitoring unit. The first prototype used a modified Myograph 2000® (Biometer International A/S, Odense, Denmark) as the recording unit,1but it was soon replaced by a commercially available acceleromyograph, the Accelograph® (Biometer International A/S).2Later came the Mini-Accelograph® in combination with Myotest® (Biometer International A/S)4and the TOF-Guard® (Biometer International A/S).5,6Commercially available today are TOF-Watch®, TOF-Watch® S, TOF-Watch® SX (Biometer International A/S), and Infinity® Trident NMT Pod (Dräger Medical AG & Co. KGaA, Lübeck, Germany).

The piezoelectric transducer element is identical in all acceleromyographs, but the electronics have been upgraded over the years. Therefore, the latest models (the TOF-Watch® series) are less sensitive to artifacts, e.g. , accidental movements of the thumb, and the stimulation current circuitry has been improved, allowing constant current stimulation at a higher skin resistance (increased from 3.5 to 5 kΩ). The upgrades do not exclude comparison of measurement obtained with various models, because the accelerometric measurements are performed in an identical manner, and constant current stimulation including stimulation current monitoring has been present in all models. However, the TOF-Watch® and TOF-Watch® S are not intended for use in research. They automatically change the way the train-of-four (TOF) ratio is calculated, ensuring that a TOF value greater than 100% is never displayed.7The TOF-Watch® SX displays the unmodified TOF value and has an optional computer interface for recording stimulus parameters, evoked response data, and other relevant information.

Ideally, neuromuscular function during anesthesia should be monitored objectively, i.e. , using a device that can measure and display the TOF fade ratio in real time.8,9However, there are clinicians who question the necessity and benefits of this practice.10–13 

Furthermore, to our knowledge, in many countries there are no official guidelines recommending routine neuromuscular monitoring.

Neuromuscular function may also be evaluated using subjective clinical tests such as head lift and grip strength, but these tests are often unreliable, require patient cooperation, and may not rule out clinically significant residual curarization.3,14–17Visual or tactile evaluation of the response to nerve stimulation is often used in daily clinical practice, but these tests are relatively insensitive. Even if no fade is felt or seen in response to TOF, double-burst, or 50-Hz tetanic stimulation, residual neuromuscular blockade cannot be excluded.18–20 

Available methods for objective neuromuscular monitoring are mechanomyography, electromyography, kinemyography,21phonomyography,22and acceleromyography. Although all five methods have advantages and disadvantages, acceleromyography is probably the most widely distributed method for objective monitoring of neuromuscular function during clinical anesthesia. In addition, acceleromyography is increasingly being used for research purposes.23–29Acceleromyography has, however, never been evaluated systematically for this purpose30—neither have electromyography and mechanomyography—and it is uncertain to what extent results obtained using acceleromyography can be used interchangeably with results obtained using these two more established methods.

The main purpose of this systematic review is to evaluate the current evidence of the relation between results obtained using acceleromyography and more established methods (mechanomyography and electromyography), and to evaluate whether acceleromyography can be used to exclude clinically significant residual neuromuscular block. Specifically, we aimed at answering the following key questions:

1. Does the use of acceleromyography produce results that differ significantly from those obtained using mechanomyography and electromyography for establishing dose–response relations and for evaluation of neuromuscular block during clinical anesthesia as well as in research?

2. What is the relation between the acceleromyographic TOF response and signs, symptoms, and clinical tests of residual neuromuscular block?

To answer these two questions, we also evaluated methodologic issues connected with the use of acceleromyography.

A comprehensive literature search was performed without time limits until November 2007 in the Cochrane Library, PubMed, BIOSIS, and Embase. We set our searching strategy deliberately broad without language restrictions using the combined search of: #1 (Neuromuscular) AND #2 (Acceleromyographically OR Acceleration transducer OR Acceleromyograph OR Accelerograph OR TOF-Guard OR TOF-Watch OR Mini-Accelograph OR Infinity Trident OR Acceleromyography OR Accelography OR Acceleromyographic monitor OR Accelerometry OR Accelerography). In addition, we studied the reference lists of all articles retrieved in the search and of other relevant articles known to the authors.

The inclusion criterion was acceleromyography used for neuromuscular monitoring. Abstracts of all relevant articles were examined, and articles that were clearly not relevant to the key questions and did not evaluate acceleromyography were excluded. Animal studies were also excluded.

To answer the key questions, the remaining articles were divided into five groups: In group 1, we included studies comparing acceleromyography with mechanomyography or electromyography for construction of dose–response curves. Groups 2 and 3 included pharmacodynamic studies in which acceleromyography was compared with mechanomyography or electromyography, respectively. Group 4 included clinical studies where acceleromyography was compared with signs, symptoms, and tests of PORC (with or without a comparison with mechanomyography or electromyography). Finally, in group 5 were studies primarily dealing with basic methodologic problems comparing acceleromyography with mechanomyography and electromyography, such as preload, normalization (i.e. , referring TOF values during recovery to the baseline value) precision, baseline drift, and stability of the response.

We evaluated the quality of the scientific evidence using the Method for Evaluating Research Guideline Evidence developed by New South Wales Department of Health31and the Scottish Intercollegiate Guidelines Network (SIGN)32,33( appendix 1). However, because the actual influence of potential sources of bias may differ between types of studies, we sought to critically appraise bias control in the individual studies.31,34Accordingly, the quality of each article was evaluated independently by the authors using a checklist ( appendix 2),31–33including salient methodologic issues relative to the outcome measures in question.35–37 

When quality rating the dose–response studies  comparing the use of acceleromyography with mechanomyography or electromyography (group 1), the method used for this comparison was carefully evaluated. Preferably, the cumulative method should only be used for long-acting drugs, and when using the single-bolus method, the patients should be randomly assigned to at least three different doses, which again should surround the anticipated ED values. Finally, handling of 0% and 100% responses should be described in detail.37 

In pharmacodynamic studies  comparing acceleromyography with mechanomyography or electromyography (groups 2 and 3), ideally, the two methods compared should be randomly allocated to the dominant and nondominant arm. For comparison of the results obtained using the different recording methods (mechanomyography, electromyography, and acceleromyography), researchers often use correlation or regression analysis or differences in means. However, as pointed out by Bland and Altman, none of these methods of analysis are suitable for such a comparison.38,39Instead, Bland and Altman have suggested that the precision of the new method, as well as the bias and limits of agreement in relation to the gold standard, should be established. For studies comparing acceleromyography with mechanomyography or electromyography, we therefore evaluated the method(s) used for the comparison. When the Bland–Altman method was used, we sought to establish whether it was used correctly.38,39If the precision (within-subject repeatability) for one of the methods is poor, the agreement between the methods will be poor as well. We therefore sought articles evaluating the within-subject repeatability, including an evaluation of whether or not the repeatability was dependent on the degree of block. According to Bland and Altman, in studies comparing two methods, the data should be plotted as differences between measurements against means of measurements with the two comparison methods.38The mean of these differences is the relative bias, and the hypothesis of zero bias can be examined by a paired t  test. However, the bias may change with the values, i.e. , increase during recovery. Therefore, we examined whether investigators had taken this into account. Although the bias may be insignificant, there may be a clinically significant lack of agreement between individual measurements. For this reason, Bland and Altman suggest studying the limits of agreement (±2 SDs) between the measurements. The confidence intervals for bias as well as for limits of agreement should be given. Finally, if repeated observations are made on each subject, the interdependence between these should be taken into account when constructing limits of agreement.38,39 

In studies comparing signs, symptoms, and tests of PORC with the acceleromyographic response  (group 4), emphasis was put on an evaluation of whether the evaluator was blind as to the acceleromyographic response.

In studies evaluating the effect of applying a preload to acceleromyography  (group 5), we considered it important that the characteristics of the preload arrangement were clearly reported, making the setup reproducible for other investigators.

Based on the type of study and the quality assessment ( appendix 3), each article was allocated a level of evidence ( appendix 4),31and levels of evidence tables were created for each key question, including data for the different outcome measures defining the key question (tables 1–7).33 

For each outcome measure, the total body of evidence was summarized, and the key questions were answered using the best evidence available. In this process, the generalizability (i.e. , the effectiveness as well as the efficacy) and the applicability (i.e. , influence of, for example, age, study setting, and population investigated) of the findings were also evaluated. The summarized evidence was then used to grade the strength of evidence according to a four-category grading system ( appendix 5),32,33unless evidence was lacking or insufficient (table 8).

Most of the studies found in our comprehensive literature search did not evaluate the use of acceleromyography. In these studies, acceleromyography was used for different purposes: in pharmacodynamic studies of neuromuscular blocking agents, to describe the frequency of PORC, or to monitor specific groups of patients (e.g. , children, elderly, patients with specific illness). In the majority of articles, acceleromyography was used without a comparison with mechanomyography or electromyography. Some articles described the use of acceleromyography monitoring sites other than the ulnar nerve/adductor pollicis muscle, i.e. , at the abductor hallucis muscle, the orbicularis oculi, or corrugator supercilii muscles. Only studies using ulnar nerve stimulation were evaluated. When the sites of neuromuscular monitoring differed (i.e. , mechanomyography monitoring of the adductor pollicis and acceleromyography monitoring of the orbicularis oculi), it was not possible to decide whether the reported differences in results were due to the different monitoring techniques or to the different monitoring sites. Therefore, these studies were also excluded. Accordingly, 55 articles evaluating the use of acceleromyography were left for further analysis.

We found three articles comparing acceleromyography to mechanomyography40or electromyography41,42for construction of dose–response curves (table 1). The study by McCluskey et al.  40was stated to be randomized, but the concealed allocation was not described. It was therefore rated as a nonrandomized study, as were the other two. The study by Meretoja et al.  41was judged to be methodologically somewhat weak (table 1), and a significant bias could not be excluded. It was therefore classified as level III− ( appendix 4). The two other studies were classified as level III+. The study of McCluskey et al.  40indicated ED⁵⁰to be 36% higher when measured using acceleromyography than with mechanomyography, and a significant difference in slope was found. However, there was no difference in ED⁹⁵. In contrast, Kopman et al.  42found no differences in ED⁵⁰, ED⁹⁵, or maximum block between acceleromyography and electromyography.

Although two studies were assigned level III+ evidence, the external validity35,36(generalizability) was low. The study of McCluskey et al.  40was performed in pediatric patients, and although the study was otherwise methodologically sound, it was performed in only 15 patients (without a power analysis). The study of Kopman et al.  42was performed in adults with acceleromyography and electromyography ipsilaterally and with a preload applied. However, a novel nonvalidated method was used to construct the dose–response curve (Hill equation). A prerequisite for the validity of this method is that the slope of the dose–response relation is the same for acceleromyography and electromyography, and this has not convincingly been documented to be the case.

There is insufficient evidence to confirm or deny that acceleromyography can be used interchangeably with mechanomyography and electromyography for construction of dose–response relations and for establishing the potency of neuromuscular blocking agents.

In 15 articles, acceleromyography was compared with mechanomyography with respect to different pharmacodynamic variables (table 2). Two of the 15 studies were excluded; in one, the great toe was used for monitoring,43and in the other, a prototype of a mechanomyograph was sought validated using acceleromyography as the gold standard.44 

The remaining 13 studies were all performed in adults (table 2).1,4–6,45–53In 11 of these, the primary aim was to compare acceleromyography with mechanomyography,1,4–6,46–49,51–53and in 2, it was to compare the performance of acceleromyographic TOF with other tests for PORC.45,50In one of the studies, a few intensive care patients were included,52and one study was performed in volunteers.47Otherwise, all studies included patients undergoing surgical procedures.

In none of the studies was a sample size analysis or a concealed randomization to dominant and nondominant arm performed as prescribed in the Consolidated Standards of Reporting Trials (CONSORT) Statement.35,36Only three studies described whether there were any dropouts.45,50,53All studies compared acceleromyography with mechanomyography in the contralateral arm, but only two studies took into account possible differences in response between the two arms.49,52 

In five studies, a Bland–Altman analysis was used to compare the two methods.4,6,45,47,49In none of these studies was it possible to decide whether the analysis was performed correctly, and it often seemed that it was performed incorrectly. In only one study were acceptable limits of agreement between the two methods defined.49 

Four studies compared the onset time found with acceleromyography and mechanomyography.4,6,48,49Two of these studies6,48found no difference in onset time, but the studies were assigned level III− evidence (table 2). In the two other studies, the onset time was found to be slightly longer when measured using acceleromyography.4,49It is uncertain whether this difference was statistically significant in one study, assigned level III+ evidence.4However, the last study, also assigned level III+, found the mean onset time of atracurium to be 23% longer with acceleromyography (160 vs.  130 s).49 

Three studies compared posttetanic count values obtained with acceleromyography to those obtained with mechanomyography during deep/intense neuromuscular block.5,51,52In all three studies, regression analysis was used and a high correlation was found. However, this analysis is inadequate for this purpose,5,51and all three studies were classified as level III−.

Ten studies compared acceleromyography and mechanomyography obtained TOF values during recovery.1,4–6,45–47,49,51,52Five of these studies were methodologically sound (level III+), and all concluded that acceleromyography and mechanomyography cannot be used interchangeably.1,4,45,49,52Three studies1,49,52showed that the bias between the two methods increases during recovery and that it becomes significant at a mechanomyographic TOF ratio of 0.6–0.7 or greater. The limits of agreement between the two methods during recovery are wide, being up to ±0.3 at a mechanomyographic TOF ratio of 0.7.4,45,49 

There is fair evidence (grade C) that acceleromyography and mechanomyography cannot be used interchangeably in pharmacodynamic studies measuring onset time or recovery using TOF stimulation. However, there is insufficient evidence to confirm or deny that the two methods can be used interchangeably for monitoring deep/intense neuromuscular block with posttetanic count stimulation.

In seven studies, acceleromyography was compared with electromyography for recording pharmacodynamic variables during a surgical procedure.54–60We excluded one study because the great toe was used for monitoring.54Accordingly, six studies55–60were analyzed (table 3).

The primary aim of the studies varied. In three studies,56,59,60it was to compare acceleromyography with electromyography; in one,55it was to compare the use of acceleromyography with clinical evaluation of recovery; and in one,58it was to compare onset times at the laryngeal and adductor pollicis muscles using electromyography with those of the adductor pollicis muscle using acceleromyography. In the last study,57the primary aim was to compare the sensitivity of acceleromyography and electromyography with changes in the degree of neuromuscular block and with manipulations of the hand (see Group 5: Methodologic Issues Using Acceleromyography, Stability [Influence of External Disturbances]). Five of the six studies were performed in adults,42,56–58,60and one was performed in pediatric patients.55Four studies compared acceleromyography and electromyography contralaterally,55,56,58,60and two compared the two methods at the same arm.57,59 

In only one of the six studies was a sample size analysis performed,59and in only two were possible dropouts described.55,57In the four studies, where the two methods were used contralaterally, possible differences between the arms were not taken into account, and the two methods were not randomized to dominant and nondominant hand.55,56,58,60In none of the six studies was the nerve stimulations synchronized. Although a Bland–Altman analysis was performed in four studies, acceptable limits of agreement were not defined, and it was not possible to decide whether the analyses were performed correctly.55,56,59,60 

Two studies56,58(both classified as level III+ evidence) stated that onset time does not differ between acceleromyography or electromyography (table 3).

We found no studies comparing acceleromyography with electromyography to monitor deep/intense neuromuscular block.

Three methodologically sound studies compared the TOF during recovery.55,56,59In one study55(level III+), bias between the two methods did not change during recovery; in two studies,56,59the bias did change, but in different directions. Dahaba et al.  56(level III+) found the mean acceleromyographic TOF ratio to be approximately 0.05 higher than the corresponding electromyographic TOF ratio. However, at an electromyographic TOF ratio of 0.5 or greater, the bias was not significant. In contrast, Kopman et al.  59(level III++) found the electromyographic TOF ratio to be 0.6 and 0.85 when the acceleromyographic TOF was 0.7 and 0.9, respectively. All three studies55,56,59found wide limits of agreement (i.e. , 0.15–0.30) between the two methods during recovery and concluded that the methods cannot be used interchangeably.

There is fair evidence (grade C) that acceleromyography and electromyography can be used interchangeably for measuring onset times, but also fair evidence (grade C) that the two methods cannot be used interchangeably in pharmacodynamic studies using TOF stimulation. However, there is no evidence to confirm or deny that acceleromyography and electromyography can be used interchangeably to monitor deep or intense block with posttetanic count stimulation.

In 16 articles, clinical signs and symptoms of residual block, different lung function tests, or visual or tactile evaluation of the response to nerve stimulation were compared with acceleromyographic TOF response (table 4).17,45,47,50,55,61–71In 6 of these 16 studies, either the mechanomyographic or the electromyographic TOF response was used for comparison with the acceleromyographic TOF response or for defining a threshold value (e.g. , mechanomyographic TOF ratio ≥0.9) for excluding PORC.45,47,50,55,68,69In 9 studies, the primary aim was to compare the acceleromyographic TOF response with other tests, including tests of respiratory function,45,47,50,55,62,63,65,67,71and in 4, it was to determine the incidence of PORC after routine use of different neuromuscular blocking agents.17,61,64,66The last 3 studies evaluated the significance of perioperative use of acceleromyography for PORC.68–70All but 1 study55were performed in adults. Except for 3 studies performed in volunteers,47,62,71all were performed in surgical patients.17,45,50,55,61,63–70 

In two studies (level II+), the patients were randomly assigned to be monitored with or without acceleromyography perioperatively.68,69Both studies concluded that perioperative use of acceleromyography prevents PORC (i.e. , mechanomyographic TOF ratio ≥0.7) and is superior to clinical tests. Seven studies compared visual or tactile fade in response to TOF, double-burst, and tetanic stimulation with acceleromyographic TOF monitoring.17,45,50,55,62,65,67All seven studies (level III++ to IV) showed that acceleromyographic TOF was superior to visual and tactile evaluation of fade in excluding PORC. Visual and tactile fade were absent at TOF ratios as low as less than 0.4.17,45,50,55,65Even if the block was reversed, acceleromyography was superior to clinical tests and tactile fade in excluding PORC.61,65Seven studies (level II+ to IV) consistently found acceleromyography to be superior to the “reliable” clinical tests8(e.g. , 5-s head lift).17,61,62,64,67,68,70Four studies examined the relation between acceleromyography and respiratory function, swallowing, and upper airway function.47,62,63,71Three studies (level III+ to III++) indicated that an acceleromyographic TOF ratio of 0.9–1.0 could be used in clinical practice to exclude PORC.47,62,63One study47(level III++) found acceleromyography to be as valid as mechanomyography to predict PORC (i.e. , recovery of pulmonary function). However, a recent, very well-performed and well-documented study (level III++) indicated that full recovery (after rocuronium) is only guaranteed 15 min after acceleromyographic TOF 1.0 is reached.71 

There is good evidence (grade A) that acceleromyography is more sensitive in diagnosing PORC than both of the usually applied clinical tests, and good evidence (grade B) that acceleromyography is more sensitive than subjective (visual or tactile) evaluation of the evoked response to TOF, double-burst, or 50-Hz tetanic stimulation. Also, there is good evidence (grade A) that perioperative monitoring with acceleromyography improves detection of PORC, and that acceleromyography is as useful as mechanomyography in this respect (grade B). However, the evidence is insufficient to decide whether the uncorrected (not normalized) acceleromyographic TOF ratio should be 0.9, 1.0, or even higher to exclude clinically significant PORC.

We found five studies (table 5) evaluating the effect of using a preload.46,59,72–74One of the studies73was excluded, because acceleromyography with a prototype of a preload (TOF tube) was validated using acceleromyography with TOF-Watch® arm board with an insufficiently described rubber band as the comparison method (gold standard). The four other studies46,59,72,74tested different preloads in a research setting. However, the characteristics of the preloads were insufficiently described, and all four studies were assigned level III− (table 6).

There is insufficient evidence to confirm or deny the benefit of using a preload when acceleromyography is used.

Most studies (level III− to III++) have found that the control TOF ratio typically is higher than unity when acceleromyography is used,1,4–6,40,41,49,52,53,56,59,73–75but with large individual differences (0.92 to 1.47). Six studies, each including a control group monitored with either mechanomyography or electromyography (level III++), have documented that the control acceleromyographic TOF ratio is higher than unity when a preload is not used (mean values 1.08–1.16),1,4,41,49,56,59and significantly higher than both control mechanomyography TOF ratio (0.98–1.01)1,4,49and control electromyographic TOF ratio (1.01)41,56,59(tables 2 and 3).

It is uncertain how a preload will affect the control TOF (table 5). Probably because of different preload installations, the same research group found conflicting results in two studies: In one59(level III−), there was no significant difference in control acceleromyographic TOF ratio when using a preload; in another74(level III−), preload decreased the control TOF significantly.

There is good evidence that the control acceleromyographic TOF ratio without a preload most often is higher than unity (grade B) and significantly higher than both control mechanomyographic and electromyographic TOF ratios (grade B). However, there is also evidence that the control acceleromyographic TOF does not always exceed unity (grade B). There is insufficient evidence to confirm or deny that the use of a preload will influence the control acceleromyographic TOF ratio.

Three studies53,75,76examined the effect of normalizing TOF values (table 6). The two studies (level III+) comparing acceleromyography with mechanomyography53or electromyography76showed an improved agreement between acceleromyography and the comparison method when the acceleromyographic TOF response was normalized. However, there were still wide individual differences. In the third study75(level III++), recovery to TOF 0.9 was compared for normalized and raw TOF values, without a comparison method. The time to TOF 0.9 after 0.1 mg/kg vecuronium was significantly longer when acceleromyographic TOF response was normalized as compared with the raw values (mean, 10.0 min; range, 3.0–26.8 min).

There is fair evidence (grade C) that it is beneficial to normalize acceleromyographic TOF values if the aim is to ensure a mechanomyographic TOF ratio of 0.90, but consequently, the duration of time to TOF 0.9 will be prolonged (grade B).

However, there is also fair evidence that because of wide individual differences even when acceleromyographic TOF values are normalized, acceleromyography cannot be used interchangeably with mechanomyography and electromyography (grade C).

Five studies46,47,77–79dealt with the precision (the repeatability or variability) of acceleromyography (table 7), and in only three46,47,77was a control group (i.e. , mechanomyography) included. Two of these46,77were assigned level III−. In the study by Eikermann et al.  47(level III+), the precision was defined as the variance in 20 consecutive TOF measurements, and the variability of acceleromyographic TOF exceeded mechanomyographic TOF. However, the study was performed in awake, partially paralyzed (TOF 0.5–0.8) volunteers, and it is uncertain whether the variability would be the same at all levels of block and in anesthetized patients. In the two studies78,79(level III+) without a control group, the repeatability between two succeeding acceleromyographic TOF responses was evaluated. The study by Baillard et al.  78was performed in awake patients in the postoperative care unit stimulated submaximally, known to decrease the precision.80,81The study by Dubois et al.  79was reported as a letter to the editor in response to the study by Baillard et al.  78Dubois et al.  79measured the response to nerve stimulation (supramaximal stimulation not ensured) in patients before emergence from anesthesia. Not surprisingly, Dubois et al.  79found a somewhat better precision compared with the study by Baillard et al.  78when the assessment was performed during anesthesia. However, it is not possible to draw any conclusions regarding the precision of acceleromyography in general from only two consecutive measurements.

There is insufficient evidence to confirm or deny that the precision of acceleromyography differs from that of mechanomyography and electromyography, or whether the application of a preload will increase the precision of acceleromyography.

In only one study4(classified as level III+) was the magnitude of drift in T1 compared when using mechanomyography and acceleromyography (table 2). The drift was significantly more pronounced with acceleromyography than with mechanomyography: The mean final acceleromyographic T1 was 20.6% lower than control acceleromyographic T1 (range, −54% to 0), as opposed to only 5.7% (range, −37% to +12.5%) with mechanomyography. In another study,82the magnitude of drift in acceleromyography was only −7% (range, −18% to +8%), but the study did not compare acceleromyography with mechanomyography.

There is fair evidence (grade C) that baseline drift in twitch height is more pronounced with acceleromyography than with mechanomyography, the final value often being lower than the control value.

We found only one study57(level III+) comparing the stability of acceleromyographic twitch height (without a preload) and electromyography when the infusion rate of the neuromuscular blocking agent was changed and the hand turned 90° (table 3). The study showed that acceleromyography was significantly more sensitive to hand movements than electromyography. The mean acceleromyographic T1 decreased 10.01% as compared with only 0.26% with electromyographic T1. However, the results may not apply to monitoring using TOF stimulation.

There is fair evidence that acceleromyographic twitch height without preload is more sensitive to external disturbances than electromyography (grade C).

The current evidence for using acceleromyography for monitoring neuromuscular block and to exclude PORC is summarized in table 8.

The three main findings of this systematic review are as follows: First, there is insufficient evidence to confirm or deny that acceleromyography can be used interchangeably with mechanomyography or electromyography for constructing dose–response relation. Second, there is good evidence that acceleromyography cannot be used interchangeably with mechanomyography or electromyography in pharmacodynamic studies. Third, there is good evidence that perioperative monitoring with acceleromyography improves detection of PORC and in this respect is more sensitive than any of the usually applied clinical tests and than subjective visual or tactile evaluation of the response to nerve stimulation.

We have strived to find and evaluate available evidence about the use of acceleromyography for monitoring neuromuscular block. To achieve this goal, two key questions were formulated: Does the use of acceleromyography produce results that differ from those obtained using mechanomyography and electromyography, and what is the relation between the acceleromyographic TOF response and signs, symptoms, and tests of residual block? However, we soon realized that we were facing problems in evaluating relevant studies with respect to these questions. Not only were the studies extremely heterogeneous with respect to aims, methods, and quality, but also we could not rely solely on known quality rating systems designed to evaluate randomized controlled trials, such as Jadad et al.  83The Jadad scale is one of the most cited and validated scales to access the quality of randomized controlled trials. However, the scale consists of only three items directly related to control the bias: randomization, blinding, and withdrawals and dropouts. Obviously, the scale gives more weight to the reporting than the methodologic quality. Actually, the scale does not allow division of trials into “high”- and “low”-quality studies.84The methodologic problems connected with the use of the three different recording systems (acceleromyography, mechanomyography, and electromyography) and not least with comparisons of the systems are quite extensive, and handled differently and often apparently incorrectly in the studies. Based on and inspired by MERGE,31SIGN 50,32,33Good Clinical Research Practice in pharmacodynamic studies of neuromuscular blocking agents,37the CONSORT Statement,35,36and our own experiences, we therefore designed checklists for evaluation of the quality of the studies ( appendix 2). We then used these checklists to quality rate each article and summarize the evidence for the use of acceleromyography. We recognize that these checklists have not been validated. However, we constructed them using both the CONSORT Statement,35,36which is an evidence-based but more comprehensive approach of reporting randomized controlled studies than the Jadad scale,83and Good Clinical Research Practice in pharmacodynamic studies of neuromuscular blocking agents,37which consists of guidelines made to improve the methodologic quality in neuromuscular research. Our assessment approach involves a degree of subjective judgment, and although we based our quality rating on MERGE and SIGN 50, these methods are more comprehensive than used in this study, including a multidisciplinary guideline development group of 15–25 members. These limitations may have introduced bias and influenced our conclusions. Nevertheless, it is our hope that the checklists and the level of evidence tables makes it clear for the reader how we reached our conclusions. Because we were familiar with most of the articles before starting the systematic review, it was not possible to blind the evaluation of the studies.

To minimize bias, we made a comprehensive search strategy not limited to English-language articles. From the title and abstract, we discovered nine articles85–93not written in English (i.e. , seven other languages) that seemed to evaluate acceleromyography. Although these nine articles were not translated, we found nothing in the abstracts indicating that the results would change our conclusions of this review. We therefore decided not to have them translated and further evaluated. Of course, theoretically, this could lead to a language bias. On the other hand, lower quality of trials not published in English may also introduce bias.94 

At first glance, our finding that acceleromyography cannot be used interchangeably with mechanomyography or electromyography in pharmacodynamic studies may seem surprising. According to Newton’s second law of motion, stating that force equals mass times acceleration, acceleromyography should be interchangeable with mechanomyography if the mass (in this case the mass of the thumb) is constant. In theory, electromyography should also be in agreement with acceleromyography, because it measures the compound action potential from many motor units.56However, in contrast to mechanomyography and electromyography, the isotonic contractions during acceleromyography monitoring involve a three-dimensional movement involving three joints, frictional forces, and deformation of tissues, which may at least in part explain the differences.52 

Of the 19 studies comparing acceleromyography with mechanomyography (table 2) or electromyography (table 3), 11 used the method described by Bland and Altman.4,6,40–42,45,47,49,55,56,60However, in none of these studies was the method used according to the original suggestions of Bland and Altman.38,39This is not only a problem when comparing acceleromyography with mechanomyography or electromyography. It is also commonly seen when other measurement methods are compared.95Therefore, the latest version of Good Clinical Research Practice in pharmacodynamic studies of neuromuscular blocking agents now includes suggestions for statistical evaluation when comparing different measurement techinques.37 

It is a prerequisite for acceleromyography that the thumb is allowed to move and for isometric mechanomyography that a preload of 200–300 g is applied.37It is therefore not possible to compare acceleromyography with isometric mechanomyography at the same arm. Accordingly, in all studies, acceleromyography and mechanomyography were tested on contralateral arms. Also, in the majority of studies comparing acceleromyography with electromyography, the two techniques were tested on contralateral arms. Surprisingly, only two studies49,52took into account possible differences between the two arms. Furthermore, the stimulation frequency, the stabilization period, the electrical charge delivered (i.e. , supramaximal stimulation), and the peripheral temperatures were often insufficiently documented, or performed differently on the two arms.

When acceleromyography was first introduced, it was considered a prerequisite that the thumb could move freely.1However, it is not always possible to avoid the thumb touching the palm of the hand or the drapes during monitoring, and the thumb may be displaced to a new position during the stimulation.96It was therefore suggested to use a preload, and in two of the early articles on acceleromyography, an elastic band between the thumb and the index finger was used.48,52Since then, five other studies46,47,77–79have evaluated the use of a preload. However, different preloads were used in the different studies, and the characteristics of the preload were most often insufficiently described, making it difficult to generalize the findings. The manufacturer of the commercially available TOF-Watch®, Organon, also produces a commercially available and simple preload (Hand Adapter), which is now being used also in research.50,53,97–99However, it should be kept in mind that the Hand Adapter has never been sufficiently validated, and as shown in this review, there is insufficient evidence that a preload applied to acceleromyography will improve agreement with mechanomyography (or electromyography) or increase the precision.

Evaluation of precision of acceleromyography was performed differently in the five studies46,47,77–79dealing with precision (table 7). Because the degree of neuromuscular block changes during recovery (even in two consecutive measurements), it is a challenge to establish the precision of the measurements. This is most probably the reason why different approaches for evaluating the precision were chosen in the studies and why there is insufficient evidence to state which method (acceleromyography, electromyography, or mechanomyography) is the most precise method.

The control acceleromyographic TOF value, in contrast to mechanomyographic and electromyographic TOF, is most often higher than unity. To reduce the bias between the TOF ratios measured using acceleromyography, mechanomyography, or electromyography, it has therefore been suggested to refer all acceleromyographic TOF values to the baseline control value. If, for example, the acceleromyographic TOF ratio is 1.20 before injection of a neuromuscular blocking agent, a displayed TOF value of 0.90 during recovery corresponds to a “normalized” TOF of only 0.75 (90/120).74,100When normalized in this way, the mean acceleromyographic TOF values are comparable to those obtained using mechanomyography or electromyography. Therefore, if at the end of a study using acceleromyography the aim is to ensure a mechanomyographic TOF ratio of 0.9, it seems reasonable to “normalize” the acceleromyographic TOF to exclude PORC (using the aforementioned example, acceleromyographic TOF should be 90% of 1.20 = 1.08). Because the acceleromyographic control TOF is most often higher than unity, the time to TOF 0.9 will of course be longer,75and even with normalization the individual differences between acceleromyography and mechanomyography/electromyography are large.53,76So far, there is no consensus on whether to normalize acceleromyographic TOF values,37but studies with only normalized TOF data have been published.101 

The majority of articles where acceleromyography was compared with signs, symptoms, and tests of PORC (table 4) were judged to have a low or very low risk of bias ( appendix 3). Accordingly, the evidence was comparatively strong (grade A or B) for the statements regarding this part of our review (table 8). However, at least one of our statements is at variance with the finding of a recent meta-analysis of the effect of perioperative monitoring of neuromuscular function on the incidence of PORC.13We found strong evidence (grade A) that acceleromyography improves detection of PORC (table 8). In contrast, the authors of the meta-analysis “could not demonstrate that the use of an intraoperative neuromuscular function monitor decreased the incidence of PORC.”13This apparent discrepancy between the findings of our broad systematic review of acceleromyography for use in scientific and clinical practice and the more focused meta-analysis of the significance of neuromuscular monitoring for PORC may be explained by the differences in methodologies. The meta-analysis by Naguib et al.  13included both comparative and noncomparative studies and did not—at least in the original publication—distinguish between objective and subjective monitoring. It is to be expected, however, that the incidence of PORC will depend on whether the monitoring is objective or subjective,102and our review is only concerned with the effect of using acceleromyography. Accordingly, we included and meticulously evaluated the quality of only prospective comparative studies, where acceleromyography was used for this purpose. Of note, of 24 studies included in the meta-analysis of Naguib et al. , only five used objective monitoring, and all five concluded that objective monitoring improves the detection of PORC.68,69,103–105 

Where do the findings of this review leave us with respect to the use of acceleromyography in research and in daily practice?

First, it is important to realize that absence of evidence or insufficient evidence for a given claim does not necessarily indicate that the claim is not true. Evidence may lack because of lack of studies or because of insufficient design of studies actually performed.

Second, we have sought rigorously and systematically to evaluate acceleromyography for use in research as well as in the clinical setting, when possible based on studies comparing acceleromyography with the more established methods, mechanomyography and electromyography. However, neither mechanomyography nor electromyography has been validated systematically in the same way, nor has the precision of the two methods been established with certainty. And as stressed by Bland and Altman,39if the precision of a comparison method (e.g. , mechanomyography) is poor, the agreement between the two methods will be poor as well.

The most important consequence of finding insufficient or no evidence for use of acceleromyography interchangeably with mechanomyography or electromyography for measuring a given variable is of course that results obtained using acceleromyography cannot directly be compared with those obtained using one of the other methods. This implies that practically all results obtained so far using acceleromyography in dose-finding studies and pharmacodynamic studies measuring onset times, duration of action, recovery times, etc.  cannot and should not be compared directly with previous studies performed using mechanomyography or electromyography (with the exception of onset times measured using electromyography, where the evidence is fair for using the methods interchangeably). It is not possible to make any general statement about the significance of these differences in results obtained using acceleromyography, mechanomyography, and electromyography. The magnitude of differences—and thus the clinical significance—depends on several factors, e.g. , the neuromuscular blocking agent and the outcome measurements in question. When investigating a long-acting neuromuscular blocking agent, the difference in time to TOF 0.9 most probably will be both statistically and clinically highly significant. In contrast, when measuring, for example, onset time or time to reappearance of the first twitch when using a rapid-onset and ultrashort-acting agent, the differences between the methods are less pronounced and therefore of less clinical significance.

The new reversal agent sugammadex has gone through phase 1 and 2 studies using acceleromyography to evaluate the dose–response relation.27–29Apparently, acceleromyography was chosen because electromyography and mechanomyography monitors were no longer manufactured, and the simpler method of acceleromyography was widely used in the clinical setting.27,28Therefore, acceleromyography could be used in a large number of test sites with little previous neuromuscular expertise. Another argument was that the slope of the recovery curve after sugammadex reversal is very steep, and accordingly, the differences between the various techniques would be a matter of seconds rather than minutes.28Though not based on evidence, the new Good Clinical Research Practice in pharmacodynamic studies of neuromuscular blocking agents guidelines37do allow acceleromyography to be used in phase 1 and 2 studies. However, again it should be remembered that results obtained using acceleromyography varies from those obtained using mechanomyography or electromyography.

Judging from the increasing number of publications in recent years, acceleromyography is increasingly being used in the clinical setting for titrating muscle relaxants and their antagonists. This review documents that the evidence for this is good. There is good evidence that acceleromyography is better than usually applied clinical tests and subjective evaluation of evoked responses in preventing PORC.

An important question remains: What acceleromyography TOF ratio is necessary to exclude clinically significant PORC? Though with insufficient evidence, three studies1,49,52convincingly indicate that the bias between acceleromyography and mechanomyography increases during recovery and that it becomes significant at a mechanomyographic TOF ratio of 0.70 or greater. The current generally accepted threshold for exclusion of PORC is a mechanomyographic TOF 0.9.8Samet et al.  50showed that the mean time interval from an acceleromyographic TOF 0.9 to a mechanomyographic TOF 0.9 was 4 min during recovery from cisatracurium, but two studies indicate insufficient recovery with an acceleromyographic TOF ratio of 0.9–1.0.17,62However, when acceleromyographic and mechanomyographic responses are related to pulmonary function, both methods predict sufficient recovery equally at TOF 0.9–1.0,47and the negative predictive value of one acceleromyographic TOF ratio of 0.9 for absence of PORC-induced upper airway obstruction is 97%.63Under the assumption that acceleromyographic TOF is approximately 10% higher than mechanomyography, an acceleromyographic TOF value of 1.0 should be aimed at. However, one study71indicated that sufficient recovery after rocuronium is only guaranteed approximately 15 min after acceleromyographic TOF 1.0 is reached. The problem is probably that there is a great individual variation in control acceleromyographic TOF. In accord with other investigators, Suzuki et al.  75found control TOF to be 0.95–1.47. If baseline control TOF is below 1.0, it may be impossible to reach 1.0 during recovery. Furthermore, sufficient recovery may not be reached even 15 min after TOF 1.0 if the control TOF was approximately 1.4. Normalization of TOF values may be the solution to improve the detection of PORC. However, clinicians may not always know the baseline control TOF. In addition, the simplicity of the automatic calculated TOF ratio is lost, and the applicability of the method is more difficult.

An alternative approach is used in two acceleromyograph models (TOF-Watch® and TOF-Watch® S) intended for use in the daily clinic.7These monitors automatically change the way the TOF ratio is calculated, ensuring that the displayed TOF value never exceeds 100%. By definition, the TOF ratio is the height of the fourth twitch divided by the height of the first twitch in the TOF response. However, when neuromuscular recovery is nearly complete, the second and often subsequent acceleromyographic responses may exceed the first (T1). When this occurs, the TOF-Watch® (S) monitors display the T4/T2 rather than the T4/T1 ratio. Further, if this ratio is above 1.0, the monitor will limit the display to 100%.7Because T2 rarely exceeds T1 until the uncorrected TOF ratio is 0.90 or greater, these units will most likely not suggest adequate recovery more falsely than TOF Watch® SX.7Although this algorithm has not been validated for use in the research setting, it seems to be a sensible approach in the clinical setting.

This systematic review documents that the evidence for clinical use of acceleromyography is good, because acceleromyography is better in detecting PORC than usually applied clinical tests and subjective evaluation of evoked responses. Acceleromyography is now also being used not only in phase 3 and 4 studies but also in early phase 1 and 2 studies, and for constructing dose–response relation.23,28However, the current evidence is insufficient to support the use of acceleromyography interchangeably with mechanomyography or electromyography for these purposes.

Although the evidence is insufficient, studies do indicate that it may be beneficial to use a preload to increase the precision of acceleromyography. However, there is currently insufficient evidence to support routine use of a preload and only fair evidence for the use of normalization of the TOF ratio whenever acceleromyography is used.

Finally, it seems from this systematic review that there is a need for well-designed, sufficiently powered, randomized controlled trials comparing acceleromyography with mechanomyography and electromyography with respect to applicability, precision, and accuracy (bias and limits of agreement), and for studies evaluating which of the methods is more applicable, precise, and accurate to predict clinically relevant endpoints.

The authors thank Aaron F. Kopman, M.D. (Professor of Anesthesiology, New York Medical College, at Saint Vincent Catholic Medical Centers, Manhattan, New York), and Søren Larsen, M.Sc.E.E. (Managing Director, Biometer International A/S, Odense, Denmark), for reviewing the manuscript.

• Step 1: Formulation of key questions

• Step 2: Search Strategy to identify possible relevant studies

• Step 3: Inclusion and exclusion criteria to select studies to be included

• Step 4: Dividing of the studies into five groups, according to relevance for the key questions

• Step 5: Criteria used to assess the quality of the included studies ( appendices 2 and 3)

• Step 6: Evidence tables based on study type and quality assessment (tables 1–7)

• Step 7: Considered judgment/summary statement about level of evidence ( appendix 4)

• Step 8: Strength of evidence ( appendix 5; table 8)

• 1.1. Does the study address an appropriate and clearly focused question (i.e. , hypothesis, primary and secondary aims)?

• 1.2. Are relevant outcome measures collected in a standardized, valid, and reliable way?

• 1.3. Is the only relevant difference between groups the recording method (acceleromyography, electromyography, or mechanomyography)?

• 1.4. Are the statistical methods used for the data analyses appropriate and correctly and sufficiently reported?

• 1.5. Is the number of patients included sufficient? (Ideally, was the necessary sample size estimated beforehand, or was a power analysis performed post hoc ?)

• 1.6. Are numbers and reasons for dropouts and/or missing data described?

• 1.7. When relevant:

• 1.7.1. Is the method for randomization adequate (adequate allocation concealment)?

• 1.7.2. Is the method of data collection blind?

• 2.1. Were the electrodes used and the setup procedure appropriate and sufficiently reported?

• 2.2. Was supramaximal stimulation ensured and sufficiently reported?

• 2.3. Was the initial signal stabilization sufficient?

• 2.4. Was the twitch height (T1) referred to a final value at the end of the procedure?

• 2.5. Were the stimulations applied simultaneously and with the same frequency with the two methods?

• 2.6. Were the peripheral and central temperature kept constant and above 32° and 36°C, respectively?