Behm, D. G., D. M. M. St-Pierre, and D. Perez. Muscle inactivation: assessment of interpolated twitch technique.J. Appl. Physiol. 81(5): 2267–2273, 1996.—The validity, reliability, and protocol for the interpolated twitch technique (ITT) were investigated with isometric plantar flexor and leg extension contractions. Estimates of muscle inactivation were attempted by comparing a variety of superimposed with potentiated evoked torques with submaximal and maximal voluntary contraction (MVC) torques or forces. The use of nerve and surface stimulation to elicit ITT was reliable, except for problems in maintaining maximal stimulation with nerve stimulation at 20° plantar flexion and during leg extension. The interpolated twitch ratio-force relationship was best described by a shallow hyperbolic curve resulting in insignificant MVC prediction errors with second-order polynomials (1.1–6.9%). The prediction error under 40% MVC was approximately double that over 60% MVC, contributing to poor estimations of MVC in non-weight-bearing postimmobilized ankle fracture patients. There was no significant difference in the ITT sensitivity when twitches, doublets, or quintuplets were used. The ITT was valid and reliable when high-intensity contractions were analyzed with a second-order polynomial.
- electrical stimulation
- muscle activation
the interpolated twitch technique (ITT) was first used by Merton (13) to observe possible muscle inactivation with a fatiguing protocol of the adductor pollicis. Merton superimposed an evoked stimulation on a voluntary contraction to detect the presence of muscle fibers not activated by the voluntary contraction. A number of researchers have used ITT to demonstrate that full activation of the dorsiflexors (3) and quadriceps (Quads) (4, 5, 8) is possible in untrained individuals. Others have reported that not all of their subjects could fully activate their plantar flexors (PF) (3) or elbow flexors (1, 11). Dowling et al. (9) reported that none of their subjects could fully activate their elbow flexors. Could the effective detection of muscle inactivation be related to the sensitivity of their techniques? Some studies have superimposed twitches (3, 5, 8) on a voluntary contraction, and investigators have visually inspected the contraction for evoked increases in force. The noise of a high-intensity voluntary contraction may obscure the superimposed evoked torque, diminishing the sensitivity of the measurement. This may be particularly true in large muscles such as the Quads. Gandevia and McKenzie (10) used multiple stimuli (2-4) to increase the signal-to-noise ratio and inspected the superimposed torque after it was further amplified (10 times) and offset by a DC clamp amplifier. Dowling et al. (9) achieved a very high superimposed signal-to-noise ratio using a triggered averaging technique with an amplified (10 times) offset system. There have been very few studies (7, 12,16) that have extensively investigated the ITT methodology. Do attempts to increase the signal-to-noise ratio with the ITT make a significant difference in the ability to detect muscle inactivation?
Rather than just detecting muscle inactivation, is the measurement of superimposed torque a valid tool for quantifying the extent of muscle activation? Although some studies have reported a linear relationship between the interpolated twitch (IT) ratio (superimposed torque/potentiated torque) and voluntary force (8, 15), others have shown the relationship to be nonlinear (3, 9, 11, 14, 16). Some researchers have assumed a linear relationship and attempted to estimate the extent of muscle activation by using a single datum point (1, 5, 6, 15). A nonlinear relationship would not allow a simple extrapolation of muscle activation from a single IT ratio. Bulow et al. (7) reported that the curvilinear relationship between superimposed twitch torque size and Quads voluntary force was more linear when force levels >25% of the maximum voluntary contraction (MVC) were used. Loring and Hershenson (12) studied the adductor pollicis and found that the curvilinear relationship became more linear with a noncompliant loading device. Does the IT ratio-force relationship allow for a quantitative and sensitive measure of muscle inactivation?
Many of the previously cited studies have indicated that the subjects needed a number of trials to achieve full activation. Once the subject is accustomed to the experimental setup, can full activation be achieved repeatedly and reliably? Because there have not been any studies to systematically measure the reliability of the ITT with the PF or Quads, another objective of this study was to investigate the reliability of the ITT with different modes of stimulation and ankle joint angles.
Subjects. Table1 summarizes the series of experiments and number of subjects. Subjects were recruited from McGill University students and staff. All subjects were fully informed of the procedures and signed a consent form before experimentation. The study was approved by McGill University’s Ethics Committee.
Experimental setup. For all voluntary and evoked contractile properties, the subjects were seated in a straight back chair with their hips and knees flexed at 90°. For measurements taken at the knee, ankles were secured in a padded strap attached to a high-tension wire clamped to a strain gauge perpendicular to the line of pull of the lower limb. PF subjects had their leg secured in a modified boot apparatus (3). Testing was done with the ankle in either a 90° neutral position or 20° of plantar flexion or dorsiflexion. Tibial nerve surface-stimulating electrodes were placed on the popliteal space and distal portion of the triceps surae. Femoral nerve- stimulating electrodes were placed on the inguinal triangle and buttocks. Electrodes were shifted during initial stimulation to determine the optimal position for the greatest peak torque. During bipolar stimulation, electrodes were secured to the superior and distal aspects of the triceps surae or Quads muscle groups. Polarity was reversed to determine the best arrangement for twitch torque.
Compound muscle action potentials (M waves) were monitored with surface electromyographic (EMG) electrodes (Medi-Trace) placed 3–5 cm apart on the soleus or vastus lateralis distal to the stimulating electrodes. A ground electrode was secured superficially to the head of the tibia. Thorough skin preparation for all electrodes included sanding of the skin around the designated areas followed by cleansing with an isopropyl alcohol swab. To ensure that the intensity of stimulation remained constant throughout the experiment, M waves were monitored during data collection for all knee measurements (nerve and bipolar stimulation) and with the ankle at 90°. M waves were amplified (Isolated Head Stage amplifier 830, Biomedical amplifier 830 EMG), filtered (10–1,000 Hz), and monitored on an oscilloscope (model 2220, Tektronix). The EMG signal was recorded at a sampling rate of 2,000 Hz.
Evoked and voluntary torque. Stimulating electrodes were connected to a high-voltage stimulator (model DS7, Digitimer Stimulator). The amperage (10–100 mA) and duration (500–2,000 μs) of a 400-V rectangular pulse was progressively increased in an attempt to obtain a plateau in the twitch torque. This was achieved in all PF subjects but only in 9 of 16 Quads subjects. Doublets were elicited by two twitches with an interval of 10 ms (100 Hz). Supramaximum PF tetanic stimulation was evoked through the tibial nerve in a separate group of eight subjects for 1.5 s at 100 Hz. All evoked and voluntary torques were detected by a force transducer (PF: custom design; Quads: model 3SB, BLH Electronics), amplified (recording amplifier and AC-DC differential amplifiers from model NL900A, Neurolog Systems), and monitored on an oscilloscope. All data were stored on computer (model ASI 9000 486DX, Seanix) after being directed through an analog-to-digital board (Lab Master) (2,000 Hz). Data were recorded and analyzed with a custom-designed software program (Distributions Physiomonitor, Actran).
ITT. Three doublets interspersed at 900-ms intervals were evoked and superimposed on a series of 3-s-duration submaximal (20, 40, 60, 80% of MVC) and three MVCs to estimate an average superimposed signal. Superimposed doublets rather than twitches were used to increase the signal-to-noise ratio. In addition, doublets were recorded at 1-s intervals after the voluntary contractions. Superimposed and potentiated twitches and quintuplets (5 stimulations at 10-ms intervals: 100 Hz) were also used with ITT to determine possible changes in sensitivity. Torque signals were sent through both a low- and high-gain amplifier. The resident software program offset the high-gained superimposed signal 100 ms be- fore each stimulation for improved resolution (Fig.1). A ratio was calculated that compared the amplitudes of the superimposed stimulation with the potentiated stimulation (IT ratio) to estimate the extent of inactivation during a voluntary contraction. Because the potentiated evoked stimulation represents full muscle activation, the superimposed torque using the same intensity of stimulation would activate those fibers left inactivated by the voluntary contraction. All maximal and submaximal (100, 80, 60, 40, and 20% of MVC) forces were correlated with their respective IT ratios to generate linear or second-order polynomial equations for all subjects.
Statistical analyses. Linear and second-order regression equations were used to determine the line of best fit and validity of the data in predicting MVC. Differences between the actual and predicted MVC, mode of stimulation (twitch vs. doublet vs. quintuplet), and tetanus torque vs. MVC were analyzed separately using one-way analysis of variance (ANOVA) with repeated measures. A randomized one-way ANOVA was used to compare group differences of the y-intercept (a), slope (bx), and curvature of the slope (cx 2) as derived from the polynomial equations (a +bx +cx 2). Test-retest reliability was determined with the intraclass correlation coefficient applied to the repeated measures ANOVA for different ankle angles and forms of stimulation (17).F-ratios were considered significant at P < 0.05. If significant interactions were present, a Tukey’s post hoc test was conducted. Descriptive statistics include means ± SD. Data in Figs. 1, 2, 3, 4, 5, 6 are presented as means ± SE.
Description of IT ratio-force relationship. Increases in contraction intensity were correlated with decreases in the IT ratio. There were no significant changes in the M-wave amplitude at any contraction intensity with tibial nerve stimulation. The plotting of the five ratios (20, 40, 60, 80, 100% MVC) produced high linear regression values (r 2) for all angles: neutral, 0.90; 20° dorsiflexion, 0.94; 20° plantar flexion, 0.90. Higher values were found when the data were subjected to a second-order polynomial. Ther 2 values for the various angles were: neutral, 0.99; 20° dorsiflexion, 0.98; 20° plantar flexion, 0.92. There was no significant difference in the IT ratio-force relationship of tibial nerve and bipolar PF stimulation (Fig. 2). Thus with data collapsed across all PF groups, the IT ratio-force relationship was best fit by a shallow hyperbolic curve (Fig. 3). Visual inspection of the ratios shows greater linearity at the lower contraction intensities with a tendency for the ratios to plateau at ∼60–80% of MVC.
The IT ratio-force relationship plateau could represent synergists not activated with evoked stimulation. To verify this hypothesis, maximal PF tetanic and voluntary PF torque were compared. Tetanic PF torque (65.6 ± 19.1 N ⋅ m) was 18.7% less (P < 0.01) than PF MVC (80.7 ± 27.6 N ⋅ m).
The Quads IT ratios presented a similar relationship. The linear regression coefficients for bipolar (r 2 = 0.96)- and femoral nerve (r 2= 0.95)-stimulated Quads IT ratios were slightly higher than PF IT ratios. The fit of a second-order polynomial to the Quads was similar to the PF with r 2values of 0.99 (Fig. 4) and 0.96, respectively. Difficulties were encountered with maintaining the intensity of femoral nerve stimulation. M-wave amplitudes elicited during rest (7 ± 3.1 mV) were significantly (P > 0.001) larger than M waves elicited during MVC (6.2 ± 2.9 mV). There was no significant difference in bipolar-stimulated M-wave amplitudes between rest and voluntary contractions. The plateau of the PF IT ratios at higher contraction intensities was not as evident with the Quads.
A comparison of the values representing they-intercept, slope of the line, and curvature of the line from the second-order polynomial equation demonstrated no significant difference between PF and Quads.
Validity. If the IT ratios can be used to measure the extent of muscle inactivation, then this information should allow a prediction of an individual’s true MVC when not fully activated. Table 2 illustrates the poor ability of the PF and Quads IT ratio to predict the MVC of fully activated individuals when a single IT ratio is used. The significant errors in predicting the MVC with single IT ratios and linear equations were eliminated with second-order polynomial equations (Table 2).
Further analysis of the ratios indicated the prediction of MVC to be less accurate when low contraction intensities were used with the second-order polynomial. Table 3demonstrates the improved prediction of MVC when contraction intensities of >40% of MVC were included (5.8–16.3% of MVC). The exclusion of ratios of >40% of MVC resulted in more than double the prediction error (33.3% of MVC). When tested at 20% of MVC, a number of subjects exhibited IT ratios >1, suggesting >100% inactivation. Anomolous MVC predictions were found with two non-weight-bearing previously immobilized ankle fracture subjects. The prediction of one patient overestimated the contralateral MVC by 71.1%. Even when bilateral limb differences are considered, this MVC prediction would be improbable due to the effects of disuse atrophy.
Sensitivity. To determine whether the frequency of the superimposed signal influences the sensitivity of the ITT to detect the lack of full activation, single, doublet, and quintuplet stimulation were compared at the neutral angle. The larger summated torques of the doublet and quintuplet should increase the signal-to-noise ratio and possibly improve the sensitivity and thus the predictability of ITT. With use of the best fit equation (second-order polynomial), quintuplet stimulation showed a slightly better (3.3 ± 2.3%) but statistically insignificant improvement in MVC prediction than with either doublet (5.5 ± 2.9%) or single stimulation (5.4 ± 3.2%).
All the protocols in this study compared the amplitude of the superimposed torque with the potentiated evoked torque (single, doublet, quintuplet) immediately after the voluntary contraction. To further determine the best exponential fit, further analysis examined the effectiveness of comparing the superimposed doublet with an unpotentiated doublet. There was no significant difference in the prediction of the observed MVC with a potentiated doublet or an unpotentiated doublet.
Reliability. Table4 illustrates the high to very high reliability of all variables at all ankle angles with tibial nerve and bipolar stimulation except for the IT ratio with tibial nerve stimulation at 20° plantar flexion. Quads also had high to very high reliability of all variables with femoral nerve- and bipolar-stimulated ITT, except for the bipolar-stimulated twitch. In addition, eight healthy subjects had both limbs subjected to ITT. In all cases, MVC force (92.7 vs. 98.5 N ⋅ m), the extent of muscle activation (97.6 vs. 98.7%), and the regression values for the IT ratio-force relationship (0.99 vs. 0.95) were similar for both limbs.
ITT validity. One of the most important findings of this study was the inability of the ITT to predict the MVC from a single submaximal IT ratio. A perfectly linear relationship between the IT ratio and force would allow a single IT ratio to accurately predict an individual’s MVC. For example, in a perfectly linear relationship, a superimposed twitch-to-potentiated twitch ratio of 0.25 would result in a muscle force equal to 75% of MVC. Linear relationships between superimposed twitch and force levels have been reported with the adductor pollicis (12) and Quads (8). This study, however, found a better fit with a shallow hyperbolic curve for both the PF (Fig. 3) and Quads (Fig. 4). Other researchers have reported that the decrease in superimposed twitch force with increasing voluntary force results in a shallow hyperbolic relationship with the PF and dorsiflexors (3), Quads (7, 14, 16), and elbow flexors (9). A hyperbolic curve best predicts thex-intercept (MVC) with a second-order polynomial equation that is an exponential rather than a linear function. The nonlinearity of the slope, however, does not permit an accurate prediction of the MVC from a single IT ratio. Although the latter could be used as a general indication of muscle inactivation. Researchers, however, must be cautious in extrapolating quantitative measures from a single datum point.
Similar caution should be exercised when the ITT depends on submaximal contractions to predict the MVC. This study has demonstrated that using contraction intensities of <40% of MVC resulted in an unacceptable 33.3% error. The use of submaximal contractions under 80% of MVC still resulted in a 13–16% error (Table 3). This is similar to the findings of Bulow et al. (7), who stated that an intensity of 75% of muscle force is necessary to achieve a sufficiently accurate prediction. Maximal or near-maximal contractions should be used to obtain the most accurate prediction of MVC.
Another interesting finding was the disproportionately large superimposed doublets at low contraction intensities (20% of MVC). In many subjects contracting at 20% of MVC, the amplitude of the superimposed doublet was equal to or greater than the potentiated doublet. To produce a contraction, a portion of the muscle must have been activated and thus a smaller superimposed torque than the potentiated evoked torque would be expected. The disproportionately large doublets would create a gross overestimation of the muscle inactivation at that contraction intensity as well as contribute to the nonlinearity of the IT ratio-force relationship. Bulow et al. (7) reported a nonlinear twitch-voluntary force relationship at contraction intensities of <25% of MVC, resulting in resting twitches as well as superimposed twitches at 10% of MVC to be smaller than the superimposed twitches on 20–25% of MVC. They attributed the smaller resting and low-intensity twitches to viscoelastic force loss, suggesting that force is dissipated when it is transferred from the stimulated to the nonstimulated portion of the muscle. In addition, there would be additional force loss in the attempt to transfer force through subcutaneous fat and connective tissue. Belanger and McComas (3) hypothesized that the reduction in the slack of the series elastic component with weak contractions would contribute to a larger superimposed than resting twitch. The inability to produce maximal or near-maximal contractions would limit the applicability of the ITT to predict MVC in some patients who are unable to produce strong contractions because of pain, swelling, or apprehension.
In a study that investigated the effects of muscle fatigue after ankle fractures, we attempted to use ITT to predict MVC to compare patients working at the same intensity (relative to predicted MVC). This was not possible in patients tested within 1–2 wk of cast removal because subjects were either unable or unwilling to generate high-intensity contractions. The predicted MVC of one of the subjects exceeded the contralateral MVC by 71.1%, which, given the presence of atrophy, is highly unlikely. An inspection of the non-weight-bearing IT ratio-force curve (Fig. 5) illustrates a severe hyperbola with little distribution of the data points. The inability to accurately predict MVC may be related to the large error experienced by normal subjects when only low-intensity contractions are used. Therefore, the ITT may only be useful in estimating muscle inactivation in patients who can generate relatively strong contractions.
Mechanisms of IT ratio-force relationship. The plateau of the IT ratio-force relationship at high contraction intensities has been found by other researchers (3, 9, 12, 14, 16). The high-intensity plateau may be related to the contribution of synergists to the total force output. Although the majority of PF torque is produced by the gastrocnemius and soleus innervated by the tibial nerve, there is a contribution by the peroneus longus and brevis innervated by the peroneal branch of the lateral popliteal nerve (2). The plateau of the IT ratio-force relationship at high contraction intensities may signify nearly full activation of the triceps surae while extra force is contributed by the peronei in which activation levels remain undetected by ITT (3). To verify this assumption, we compared the maximum PF tetanic force with MVC. Tibial nerve-stimulated tetanic torque was 18.7% less than MVC. This disparity is very similar to the contraction intensities at which the plateau effect commences in the PF of this study. This could also be an adequate explanation for the plateau in the data of the study of Dowling et al. (9), which examined the ITT of the bicep brachii. Their plateau may reflect the extra contribution of the brachioradialis to elbow flexion. The Quads, however, do not have any synergists contributing to leg-extension torque. The Quads IT ratio-force relationship is also more linear (0.96) than the PF (0.9–0.94). Thus synergistic activity may have a significant contribution to the nonlinearity of the PF IT ratio-force relationship.
Another possibility was offered by Loring and Hershenson (12), who investigated the effect of series compliance on the superimposed twitch of the adductor pollicis. The relationship of superimposed twitch forces when attached to a compliant link resembled the shallow hyperbolic curve illustrated in this and other studies. A noncompliant link, however, resulted in a more linear superimposed twitch-voluntary force relationship. The compliance associated with the relatively long Achilles and patellar tendons may contribute to the shallow hyperbolas in this study. The placement of the ankle at a neutral testing angle as well as 20° PF and dorsiflexor is expected to have some effect on the compliance of the achilles tendon. Although placing the Achilles tendon under greater stretch with dorsiflexion did result in a slightly more linear relationship (0.98) than with plantar flexion (0.92), there was no significant difference in the prediction of MVC. Even though it is not significantly different, the greater linearity of the stretched dorsiflexor position does provide some corroborating evidence to the proposal of Loring and Hershenson.
Protocol sensitivity. It may also be hypothesized that the high-intensity plateau effect may be related to a low signal-to-noise ratio. Difficulty in ascertaining the true amplitude of the superimposed doublet among the noise and fluctuations of a high-intensity contraction may contribute to the plateau effect. A larger signal-to-noise ratio may be achieved by increasing the amplitude of the superimposed stimulation. Single, doublet, and quintuplet stimulations were compared to determine whether the greater summated torques of doublets and quintuplets would improve the signal-to-noise ratio. Although the quintuplet showed a slightly more linear relationship (r 2 = 0.96) than the single (r 2 = 0.94) or doublet (r 2 = 0.92) stimulation, the data were still best fit by a second-order polynomial (quintuplet, r 2 = 0.99; doublet, r 2= 0.97; single,r 2 = 0.98). There were no significant differences in the prediction of MVC with the three forms of stimulation.
A similar lack of difference was found when investigating the effectiveness of comparing the superimposed doublet to either a potentiated doublet or unpotentiated doublet. The most important factor seems to be the resolution of the superimposed torque. Gandevia and McKenzie (10) improved their resolution with an offset DC clamp amplifier. Dowling et al. (9) achieved high resolution of their superimposed doublet by using triggered averaging with an amplified offset to enhance the signal-to-noise ratio. Dual amplification (low- and high-gain amplifiers, computer software) of the signal as well as precise temporal measurement of the offset signal allowed for high resolution in this study.
Reliability. A technique must not only be valid but reliable as well. Test-retest reliability measures were very highly correlated for both nerve and bipolar stimulation of the PF (Table 4). A moderate level of reliability was found with tibial nerve stimulation of the ankle plantar flexed at 20°. The lower level of reliability at 20° plantar flexion could be attributed to a movement of the stimulating electrode away from the optimal position on the popliteal surface. Subjects had a tendency to involve the hamstrings when contracting at 20° plantar flexion, pushing the electrode from its original position, decreasing the intensity of the stimulation.
The very high reliability of the Quads IT ratio with bipolar stimulation (0.96) contrasted with the less stringent but still highly correlated reliability with femoral nerve stimulation (0.78) (Table 4). This may also be related to electrode displacement. Quads M-wave amplitudes elicited during rest were significantly (P < 0.0001) larger than M waves elicited during MVC. The contraction of the Quads during MVC probably displaced the stimulating electrode from its optimal position over the femoral nerve in the inguinal space. Although Rutherford et al. (16) did not find a significant difference in the Quads IT ratio-force relationship with femoral nerve or percutaneous stimulation, they surmised that the effectiveness of the ITT was not dependent on full evoked activation. To improve reliability, bipolar muscle stimulation or improved placement of nerve-stimulating electrodes is necessary to ensure that the same proportion of muscle is stimulated with all contractions.
The moderate reliability of bipolar stimulation twitch did not affect the very high reliability of the bipolar PF and Quads IT ratio (Table4). Lower twitch reliability may be attributed to inaccurate placement of electrodes from test to test. Smaller interelectrode distance could result in less muscle mass stimulated, affecting the amplitude of the stimulated torque. This would not severely affect the ITT, since the IT ratio is derived from the superimposed and potentiated torque stimulated from the same proportion of muscle. Although the absolute amount of muscle stimulated from test to test may differ, the ratio should not be affected.
Summary. MVCs estimated by using second-order polynomial equations derived from IT ratio-force relationships were not statistically different from the observed MVCs of subjects able to maximally activate, thus providing an acceptable estimation of muscle activation. The exclusion of MVCs or near-maximal voluntary contractions in the polynomial equations resulted in significant MVC prediction errors. The use of single IT ratios or linear equations resulted in MVC prediction errors of 21.5 and 10.9%, respectively. The shallow hyperbolic curve of the IT ratio-voluntary force relationship was not significantly altered when using either single, doublet, or quintuplet stimulation for the superimposed and potentiated torque. The nonlinearity of the relationship may be attributed to synergistic muscle contribution and/or compliance of the system. The technique was shown to be reliable in both the PF and Quads.
Address for reprint requests: D. G. Behm, School of Physical Education and Athletics, Memorial Univ. of Newfoundland, St. John’s, Newfoundland A1C 5S7, Canada.
- Copyright © 1996 the American Physiological Society