Comparison of a Modern Digital Mechanomyograph to a Mechanomyograph Utilizing an Archival Grass Force Transducer

A Grass Instruments FT-10 force transducer (Grass Instruments, USA) and a Grass model 7400 paper strip chart recorder (Astro-Med, Inc., USA) were obtained from storage (fig. 1E). The Grass force transducer was originally developed in the 1970s and operates on the principle of Hooke’s law, which relates force to the deformation of a spring. The system was connected to the physiologic recorder with a pixelated display showing results; alternatively, results could be printed in real-time on graph paper or saved to a 3.5-inch disk. An analog output channel was also available and made functional after wire splicing, shielding, and digitization via an analog-to-digital converter, NI DAQ USB 6009 (National Instruments, USA).

A complex arm holder was needed to suspend the force transducer above the hand of the patient, as shown in figure 1, A to D. Essential considerations in the building process included structural integrity to ensure the transducer stayed in place during high force loads and maintaining a perpendicular angle between the thumb and the transducer spring to maximize the output signal. An archival arm holder device used in earlier studies was obtained. However, it did not include a mechanism for suspending the force transducer, which was built using set screws and a beaker clamp to allow for adjustment (fig. 1A). The thumb was held in place through a commercially available disposable finger splint suspended by Kevlar thread (Superior Threads, USA).

Electronic components were selected based on measurement range, precision, and size. They included a force transducer (C9B 100N, HBM, Germany), amplifier (EMBSGB200 version 2.3; Tacuna Systems, USA), and analog-to-digital signal converter (NI DAQ USB 6009 National Instruments; fig. 2D). The force transducer was secured to the thumb via a custom 3D-printed force transducer insert (fig. 2B). A second 3D-printed part was created to hold the arm in place and was adjustable to account for differences in hand, finger size, and flexibility (fig. 2A). Computer-aided drawings of 3D printed parts are available upon request. Wrist immobilization was provided by a commercial wrist splint device (DonJoy, USA; fig. 2C). Once set to a patient’s hand, the system was completely rigid for measurement of isometric force generated by the thumb.

A linear response to an applied force for each mechanomyograph system was established before use in clinical testing. Each system was calibrated using a series of weights to assess the system’s accuracy, precision, sensitivity, and resolution. For these experiments, a set of calibration weights (Ohaus, USA) up to 2 g were tested on both systems, by averaging 1,000 samples per weight. Data analysis was performed using MATLAB (MathWorks, USA). The calibration process differed based on the structures of the two systems; the weights were hung from the archival mechanomyograph or applied to a plastic plate on top of the force transducer for the new mechanomyograph. The calibration curves were used to calculate the voltage output range for 200 to 300 g of preload. Voltage output was assessed during device attachment to ensure adequate preload before clinical measurements.

Precision was estimated by calculating the mean SD in the voltage signal at each calibration weight. Resolution was calculated from the voltage range for each device divided by the number of bits supported by the analog-to-digital converter. Sensitivity was determined by multiplying the mean SD by the resolution.

This study was approved by the institutional review board, and written informed consent was obtained from patients. The study was registered with Clinicaltrials.gov (NCT05006807, Principal Investigator: Dr. Andrew Bowdle) on August 10, 2021. Patient eligibility criteria included age over 18 yr, requiring general anesthesia with administration of neuromuscular blocking drug, and having access to both arms. Patient recruitment occurred between December 2021 and March 2022. Anesthesia care was at the discretion of the anesthesia team. No constraints were placed on anesthesia care regarding the selection of neuromuscular blocking drugs and timing of reversal. The patients were outfitted with a mechanomyograph device on each arm. The ulnar nerve was stimulated with a 60-mA current supplied by a TwitchView monitor (Blink Device Company, USA). Figure 3 shows the study device setup.

Preload was set to 200 to 300 g and checked every 15 to 30 min. The patient’s thumb was repositioned as needed to maintain adequate preload. Twitch measurements were taken every 20 s to 1 min when one or more twitches were present. During periods of rapid change in train-of-four ratio, frequent measurements were made to maximize measurements across the full range of train-of-four ratios. The paired measurements from each arm were taken within 2 s of each other.

A LABView (National Instruments) program was written to collect the data and save all recordings at a sampling rate of 1,000 measurements/s. A 10-s interval was recorded for each data point comparison and included both mechanomyograph devices. Postprocessing of the data was performed in MATLAB for both mechanomyographs. After removing low signal data and electrocautery noise, a smoothing filter was applied before calculation of the train-of-four ratio. The train-of-four ratio was calculated in two different ways: by measuring the peak amplitude and the area under the curve of each peak. T4 was divided by T1 to determine train-of-four ratio.

The descriptive statistics are presented as number (%), median (range), and mean (SD) as appropriate. A simple regression model was used to compare the correlation of voltage with calibration weight for each device, as well as the correlation of the train-of-four ratio between the new and archival mechanomyography devices.

Precision is an analysis of the noise in a signal, specifically variations in measured voltage when the voltage output is expected to be constant.¹⁶  To calculate precision, 1,000 measurements were made with each device at a constant voltage. These 1,000 measurements were repeated at several different voltage outputs in response to 17 calibration weights ranging from 0 to 1.6 kg in 100-g increments since the amount of noise can vary in a circuit depending on the voltage. For each voltage, a SD of each set of 1,000 measurements was calculated. The mean standard deviations across all voltages were compared using chi-square tests between the new and archival mechanomyography systems.

A convenience sample was used since the number of patients enrolled was limited by the eligibility criteria and willingness of the surgical and anesthesia teams to tolerate the extensive and time-consuming set-up of the archival mechanomyograph. A Bland–Altman plot was used to describe the relationship between the difference and the average of the train-of-four ratio of the two mechanomyographs. To account for assessment of the agreement between methods over a range of repeated measurements in individual participants, we incorporated a modified Bland–Altman method based on ANOVA that has been reported by Olofsen et al.¹⁷  specifying the participant as the unit of clustering. All other statistical comparisons were performed using STATA version 15.0 (StataCorp LP, USA).¹⁸  Statistical significance was defined as a P value of < 0.05.

Calibration testing results are shown in Table A1 of the appendix. The new mechanomyography system had improved precision, resolution, and sensitivity compared to the archival mechanomyography (P < 0.001). The calibration measurements for both devices demonstrated positive association between voltage output and calibration weights with a linear correlation coefficient of 1.00 (fig. 4).

Demographic characteristics and intraoperative variables for the eight patients who participated in the study are shown in Table A2 of the appendix. Sample digital signals collected in the LABView program during data collection are shown in figure 5A. The raw voltage output waveforms before and after the application of a smoothing filter were similar; smoothing removed small variations due to electrical noise (fig. 5B). Train-of-four ratios during spontaneous recovery from neuromuscular blockade after a maintenance dose in a single patient for both mechanomyography devices are shown in figure 5C.

With the archival and new mechanomyography systems, 767 paired train-of-four ratio data points were collected (see Table A3 of the appendix). There was a linear relationship of train-of-four ratio measurements using peak heights obtained with the archival and new mechanomyograph systems with an R² value of 0.94 (P < 0.001; fig. 6A). Bland–Altman analysis of the train-of-four ratio showed a bias of 3.8 ± 2.0% (95% CI, −1.1 to 8.6) with higher values for the new mechanomyograph as compared to the archival mechanomyograph and limits of agreement of −13.4% and 20.9% (95% CI lower limits of agreement, −23.5 to −8.7; 95% CI upper limits of agreement, 16.3 to 31.0; fig. 6B). The train-of-four ratios determined from peak height were equivalent to those determined from the area under the curve (R² = 1.0) for both the new and archival mechanomyographs (fig. 6C; only data for new mechanomyograph shown).

An archival mechanomyograph system utilizing an FT-10 Grass force transducer was modified for digital data acquisition, and train-of-four ratio measurements obtained with the archival system were compared to those obtained with a new mechanomyograph on a contralateral arm. Despite a span of nearly 50 yr separating the designs of these mechanomyographs, the results were remarkably similar, with a bias of 3.8% and limits of agreement of −13% and 21%. Train-of-four ratios calculated from peak amplitude and area under the curve of peaks were similar. The new mechanomyograph had improved resolution, sensitivity, and precision, reflecting technological progress over the years. The similarity of the results from these instruments instills confidence that data obtained in studies utilizing mechanomyography over nearly five decades are likely to be comparable. Mechanomyography is a robust approach to measuring neuromuscular blockade that should continue to be regarded as a gold standard measurement and should be used to validate electromyograph and acceleromyograph neuromuscular blockade monitors.

Mechanomyography is the gold standard for measuring the twitch response to ulnar nerve stimulation for several reasons. First, it directly measures the force of isometric muscle contraction using a sensor (force transducer), the performance of which can be tested and verified ex vivo at the benchtop in the laboratory. Second, the signal generated by the transducer is robust and not easily subject to interference from electrical noise that is ubiquitous in an operating room. Third, because the measurement is direct and robust, it should be highly reproducible, as shown by this study. By contrast, acceleromyography and electromyography have significant drawbacks. Acceleromyography does not measure force directly but derives it from a measurement of acceleration. Acceleromyography is complicated by a high prevalence of “overshoot” in the train-of-four ratio, such that the baseline, unparalyzed train-of-four ratio is often greater than 1.^11,19  Electromyography is an attractive alternative because it directly measures the muscle action potential, and it does not commonly manifest “overshoot” in the train-of-four ratio. However, the electromyogram is a small electrical signal and is highly susceptible to electrical artifacts that can interfere with interpretation. Therefore, the reliability of electromyography will be strongly influenced by the methods used to manage noise, which vary from one device to another. Thus, mechanomyography has been recommended as the gold standard of neuromuscular blockade measurement.¹  Our results reinforce that recommendation.

There are several limitations to this study. A single archival mechanomyography system was studied as that was all that was available to us. Although systems utilizing Grass force transducers were widely utilized for research purposes, other mechanomyography systems have been used, such as the Relaxometer.²⁰  As these other mechanomyography systems were not tested, these results may not be generalizable beyond the archival or the new mechanomyographs described in this article. The paired measurements were not synchronized but were obtained within 2 s; it is possible that minor differences between paired measurements could be due to a lack of perfect synchronization. The paired measurements were obtained from different arms. Arm-to-arm variation in neuromuscular blockade has been shown previously by Claudius et al.²¹  not to be clinically significant. Our experience is that substantial variation in arm-to-arm train-of-four ratio may occur in an individual subject, but that in a large data set, arm-to-arm variation is not evident, in agreement with Claudius et al.²¹  This potential source of variation was unavoidable in our study because it is not possible to place two mechanomyographs on the same arm. Additionally, our study did not employ a stabilization period before administration of neuromuscular blocking drugs. This was not done because the train of four ratio should not require stabilization, as might be required when looking at T1 amplitude,²²  and the additional time needed to obtain these data would have further delayed surgery start beyond the extra time needed to set up two mechanomyographs.

The patient sample size was small, but many paired measurements were obtained. There were more train-of-four ratio measurements at the extremes of train-of-four ratios than mid-range train-of-four ratios. Since the limitations of the study would generally contribute additional variability to measurements, the similarity of the results from the mechanomyographs is perhaps all the more remarkable.

In conclusion, an archival mechanomyograph utilizing a Grass force transducer and a new mechanomyograph built in our laboratory obtained essentially interchangeable train-of-four ratio measurements in anesthetized patients, demonstrating the feasibility of using mechanomyography to validate the performance of electromyographic and acceleromyographic neuromuscular blockade monitors.

The authors acknowledge Edward Pavlin, M.D., Department of Anesthesiology and Pain Medicine, University of Washington, Seattle, Washington, for donating the archival mechanomyograph; John Brock-Utne, M.D., Department of Anesthesiology, Perioperative, and Pain Medicine, Stanford University, Stanford, California, for donating the arm holding apparatus that was modified to fit the archival mechanomyograph; and Aaron Kopman, M.D., Department of Anesthesiology, Weill Cornell Medical College, New York, New York, for providing valuable perspective on the history of mechanomyography and suggestions for the article.

Supported by the Foundation of Anesthesia Education and Research and the Washington Research (Schaumburg, Illinois) Foundation (Seattle, Washington; to Dr. Michaelsen) and in part by the Laura Cheney Professorship in Anesthesia Patient Safety (Seattle, Washington; to Dr. Bowdle). Support was, otherwise, provided from departmental sources.

The authors declare no competing interests.