Factors in fatigue during intermittent electrical stimulation of human skeletal muscle

David W. Russ, Krista Vandenborne, Stuart A. Binder-Macleod


During an electrically elicited isometric contraction, the metabolic cost of attaining is greater than of maintaining force. Thus fatigue produced during such stimulation may not simply be a function of the force-time integral (FTI), as previously suggested. The goal of the present study was to evaluate fatigue produced in human medial gastrocnemius by intermittent, isometric electrical stimulation with trains of different frequencies (20, 40, or 80 Hz) and durations (300, 600, or 1,200 ms) that produced different peak forces and FTIs. Each subject (n = 10) participated in a total of six sessions. During each session, subjects received a pre- and postfatigue testing protocol and a different, 150-train fatiguing protocol. Each fatiguing protocol used only a single frequency and duration. The fatigue produced by the different protocols was correlated to the initial peak force of the fatiguing protocols (r 2= 0.74–0.85) but not to the initial or total FTI. All of the protocols tested produced a proportionately greater impairment of force in response to low- vs. high-frequency stimulation (i.e., low-frequency fatigue). There was no effect of protocol on low-frequency fatigue, suggesting that all the protocols produced comparable levels of impairment in excitation-contraction coupling. These results suggest that, for brief stimulated contractions, peak force is a better predictor of fatigue than FTI, possibly because of the different metabolic demands of attaining and maintaining force.

  • force-time integral
  • excitation-contraction coupling

muscle fatigue is defined as a transient decrease in the force-generating ability of muscle, resulting from recent contraction (46). Despite this relatively simple description, fatigue is a complex, multifactorial phenomenon and has been associated with impairments at a number of sites, ranging from central activation to myofilament interaction (for recent reviews, see Refs. 20,43). Further complication arises from the fact that fatigue is task specific, in that for a given task one particular site or mechanism may be more or less responsible for the decline in muscle performance (16). One example of the task dependency of fatigue is the phenomenon known as low-frequency fatigue (LFF). LFF, first described by Edwards et al. (14), is characterized by a proportionately greater loss of force in response to low- vs. high-frequency muscle stimulation and is considered to be indicative of impairments in excitation-contraction coupling, in particular a decrease in the calcium released during depolarization (12, 27,45). Studies investigating LFF have induced it by using voluntary contractions (29) and both high- (100 Hz) (11, 12, 45) and low-frequency (10–40 Hz) (22,26) electrical stimulation. Although the mechanism that produces LFF is unknown, both metabolite buildup and elevation in intracellular Ca2+ concentration have been suggested to play a role in the development of LFF (11, 12).

Because muscle contraction increases the demand for ATP by an order of magnitude (34) and fatigue is a consequence of muscle contraction, the metabolic cost of contraction is believed to be a primary factor in fatigue (34). Despite the lack of consensus regarding the relationship of specific metabolic by-products to fatigue and the mechanisms by which they act in vivo, several studies have demonstrated that voluntary exercise and stimulation protocols that produce the greatest metabolic changes also produce the greatest fatigue (2, 10, 25, 35, 37), supporting the theory that fatigue is related to metabolic cost, although factors other than metabolites contribute to the reduction in force (2, 11, 20). Total ATP consumption by skeletal muscle increases with increased work and power (1, 18, 19, 39). In the case of isometric contractions, no physical work is performed, and so the area under the force-time curve, or force-time integral (FTI), is often used as a measure of isometric work (8, 10,25). This approach is apparently justified by the observation that the FTI and work produced by a series of twitches relate to ATP consumption in a similar manner (23). However, a number of studies have demonstrated that during intermittent, isometric, electrical stimulation with trains of pulses rather than twitches, shorter duration contractions produce greater fatigue than longer contractions when force, total contraction time, FTI, and number of stimulation pulses are controlled (4, 10, 25, 42). Protocols using short-duration contractions produced greater ATP turnover as well as greater fatigue. Finally, pilot work from our laboratory demonstrated that repetitive stimulation with 12-pulse, 50-Hz trains produced more fatigue than stimulation with 12-pulse, 10-Hz trains, despite the fact that the number of pulses and contractions were equal, and the 10-Hz trains produced higher FTIs (37). However, 50-Hz trains did produce higher initial peak forces and greater metabolic changes.

The purpose of this experiment was to examine the fatigue produced by isometric stimulation protocols that consisted of equal numbers of brief, repetitive stimulation trains and to relate that fatigue to the peak forces and FTIs produced by the different trains. We chose to examine brief trains, as they are often used in clinical applications such as functional electrical stimulation, which uses electrical activation of skeletal muscle to aid individuals with central nervous system impairments in performing functional movements (32, 40). In addition, motor units often fire in brief bursts of activity during volitional activation (24, 31). We varied the peak forces and FTIs generated by the trains in the different fatiguing protocols by using different stimulation frequencies and train durations. The effects of the fatiguing protocols were evaluated in two ways: 1) by examining the percent decline in force during the fatiguing protocol and 2) by comparing the force responses to a standard set of low- and high-frequency trains delivered before and after the fatiguing protocols to control for impairments of excitation-contraction coupling that might affect trains of different frequencies to different extents (i.e., LFF). Because of this added level of control, we felt that the second method was more accurate.

We hypothesized that the fatigue produced by the protocols tested here would relate more closely to the peak force produced by the trains than to the FTI, on the basis of our laboratory's previous work demonstrating that force generation was more metabolically costly than force maintenance (38). For the brief (≤1.2 s) trains examined here, the greater metabolic cost of the force-generation phase, which would be related to the peak force achieved, would constitute a greater percentage of the total metabolic cost of each train than it would during a longer, sustained contraction of several minutes. In an earlier study (8), our laboratory found that the degree of LFF produced by a given protocol was related to the mean FTI, but not to the peak force, produced by the trains in the protocol, at least when tested immediately after the fatiguing stimulation. For this reason, we hypothesized that greater LFF would be observed after the protocols using trains that produced the greatest FTIs.



Twelve healthy subjects (6 men) ranging in age from 20 to 33 yr of age (mean 24.9 ± 4.29 yr) with no history of muscle or joint problems participated in this study. All subjects were informed of the purpose and procedures of the study and gave written, informed consent. Experimental protocols were approved by the Human Subjects Review Boards of the Universities of Delaware and Pennsylvania.

Experimental Procedures

Training session.

Each subject participated in one training session. During this session, subjects were trained to perform maximum, volitional, isometric contractions (MVICs) of the plantar flexors and to relax their muscles during electrical stimulation. In addition, one of the investigators (D. W. Russ) located the motor point of the medial gastrocnemius muscle (MG) on each subject by using a probe stimulator connected to a Grass S8800 stimulator with a SIU8T stimulus isolation unit (Astro-Med, West Warwick, RI) to deliver single pulses. Using a standard clinical text of muscle motor points (36) as a guide to initial placement, the investigator increased the stimulus intensity until a visible contraction of the MG was attained. The investigator then moved the probe over the muscle, while palpating the soleus muscle distal to the heads of the gastrocnemius. The probe location that produced the greatest force without any palpable soleus contraction was accepted as the location of the motor point, and its position relative the posterior fibular head, the medial tibial plateau, and the tibial tubercle was recorded. These positional measurements were used to assist the investigator in finding the motor point during subsequent testing sessions.

During the training session, the subject's plantar flexion MVIC was assessed. We confirmed that the subject was performing a true MVIC with a burst-superimposition technique, previously described for the quadriceps muscle (41). The criterion for acceptance of an effort as an MVIC was that the superimposed burst increased the volitional force by ≤5%. All subjects recruited were able to reach this criterion for acceptance within the training session. The same stimulator used to find the motor point was used in the burst-superimposition test, but instead of the point stimulator attachment, self-adhesive, 5 × 5-cm electrodes were used. One electrode was placed directly over the MG motor point, and the other was placed over the distal muscle belly. The subject was positioned supine, with the foot at a 90° angle to the shank and the knee at 0° of flexion, in a KinCom III isokinetic dynamometer (Chattecx, Chattanooga, TN) that was set up in the manufacturer's recommended plantar flexion-testing configuration. This device was calibrated per the manufacturer's guidelines and has been shown to be reliable (17, 33). The subject was stabilized by using inelastic nylon straps with velcro closures that were pulled tightly across the dorsum of the foot, the front of the ankle joint, the shin, the thigh, and the hips. Additional stabilization was provided by a custom-made shoulder harness that prevented subjects from sliding in a cephalad direction during the plantar flexion contractions.

Testing sessions.

Each subject participated in six sessions. These sessions were separated by ≥48 h. Patient positioning and setup was the same as that described for the training session. The subject's plantarflexion MVIC was determined as described for the training session, including the burst superimposition. If the subject could not produce an MVIC within ±5% of that achieved in the training session after three attempts, the session was cancelled and another session was scheduled. Each of the sessions consisted of a prefatigue test, a fatiguing protocol, and a postfatigue test.

If the session did proceed, the next step was to set the stimulation intensity to produce 10% of the plantar flexion MVIC by using a 200-ms, 100-Hz stimulation train. Pilot work determined that this force level was well tolerated by subjects and produced consistent force responses (37). This level of intensity was chosen as it provided a compromise between the issues of keeping the stimulation low enough to avoid activating other triceps surae muscles (especially the soleus), as determined by palpation, but high enough to provide adequate resolution with our dynamometer system. Before setting the stimulation intensity, electrode placement was modified slightly. The electrode over the distal muscle belly was removed, and a 12.7 × 7.6-cm self-adhesive electrode was placed over the anterior knee, with care taken to avoid contact with any visible muscle mass. The electrode over the MG motor point was left in its original position. This monopolar arrangement reduced subject complaints of discomfort during testing.

The medial gastrocnemius has been estimated to comprise ∼30% of the cross-sectional area of the human plantar flexor muscle mass (15,21). On the basis of the relationship between muscle cross-sectional area and force (9), if only the MG were recruited, an electrically elicited force equal to 10% MVIC would correspond to ∼33% of the medial gastrocnemius MVIC force. However, evidence exists that as much as 50% of the MG may have been recruited on the basis of the subject positioning used in the present study (13). In either case, comparable stimulation protocols in the human quadriceps femoris have demonstrated that for forces up to 80% MVIC, surface stimulation recruits a random population of motor units (7). Thus, although it is unlikely that we recruited the same population of motor units across sessions, the percentage of each motor unit type recruited was randomly distributed and thus similar across sessions. Moreover, in an attempt to recruit a consistent population of motor units within a session, the stimulation intensity and electrode placements were not changed during a given session.

After the stimulation intensity was set, the muscle was potentiated with ten 200-ms, 100-Hz trains, delivered at a rate of 1 train/10 s. Work in our laboratory has suggested that potentiation with electrical stimulation is a function of the number of pulses delivered and that 200–300 pulses are necessary to achieve full potentiation (6). Because we wanted to avoid fatigue before the start of the fatiguing protocols, we chose the low end of this range. Thus a consistent, high level of potentiation was produced in all subjects, although we did not test to see whether all subjects were fully potentiated. Within 5 s of completion of the potentiating stimulation, the prefatigue test was started, during which trains were delivered at a rate of 1 train/10 s to avoid fatigue. The prefatigue test consisted of four different stimulation trains, each delivered twice for a total of eight trains. The order in which the trains were delivered was randomly determined for each subject. Each train lasted 1 s and had a frequency of 1 (twitch), 10, 50, and 100 Hz. These frequencies were chosen so that we could evaluate the presence of LFF by examining changes in two force ratios: 1:100 Hz force (a typical twitch-tetanus ratio), and 10:50 Hz force. This second ratio was used because we felt it encompassed a more functional range of motor unit discharge frequencies (3, 8). Because the MG demonstrates slower contractile properties than the quadriceps (7, 44), we chose to compare 10-Hz stimulation to 50-Hz stimulation, rather than using the 20:50 Hz force ratio previously used to assess LFF in the quadriceps (11, 14, 27).

After completion of the prefatigue test, the subject rested for 5 min. The muscle was then potentiated as before, and the fatiguing protocol commenced within 5 s of the end of potentiation. Each fatiguing protocol consisted of 150-stimulation trains with an intertrain interval of 1.2 s. It was believed that maintaining a constant intertrain interval was of critical importance to this study, as we were attempting to evaluate the fatigue produced by different stimulation trains. The use of a constant intertrain interval was the best way to isolate the effect of trains. In an earlier study (8), we evaluated similar protocols but delivered the trains at a constant train period. As a result, the longer duration trains had less recovery time within the fatiguing protocols, confounding the effect of the train itself. For similar reasons, we chose not to use a constant duty cycle.

The different trains used during the fatiguing protocols are listed in Table 1, and only one train was used per fatiguing protocol. The postfatigue test began 1.2 s after the end of the fatiguing protocol and consisted of the same eight trains delivered in the prefatigue test. These trains were delivered in the same order as in the prefatigue test, but they were delivered at a higher rate (1.2-s intertrain interval), and two of the trains used in that session's fatiguing protocol were interposed between each postfatigue testing train to ensure consistent levels of fatigue and activation history, as has been done previously (Fig.1) (8).

View this table:
Table 1.

Fatiguing protocol trains

Fig. 1.

Schematic of a single, hypothetical experimental session. Each subject participated in 6 such sessions, with the only difference among sessions for a given subject being the different fatiguing protocols. The only differences across subjects were the order in which the trains in the pre- and postfatigue testing sequences were delivered and the order in which the fatiguing protocols were administered. * See Table 1. ITI, intertrain interval.

The specific trains used in the fatiguing protocols were chosen on the basis of previous work in our laboratory that demonstrated that the metabolic cost (ATP turnover per contraction) to attain a level of force during a contraction was greater than that needed to maintain that force (38). We chose only to evaluate a brief 80-Hz train in an effort to avoid electrical fatigue that might occur during long, high-frequency trains. We chose the different durations at 20 and 40 Hz to allow us to compare trains that reached different force levels and to compare trains that spent different amounts of time attaining vs. maintaining force (Fig. 2). In addition, the choices of frequency and duration allowed us to make comparisons of trains that held the number of pulses constant (i.e., 25 pulses for the 80-Hz 300-ms, 40-Hz 600-ms, and 20-Hz 1,200-ms protocols; and 13 pulses for the 40-Hz 300-ms and 20-Hz 600-ms protocols). We wanted to compare trains with identical numbers of pulses because it has been suggested that the number of pulses is a major factor in fatigue (31).

Fig. 2.

Raw force responses to the trains used in the 7 fatiguing protocols from an individual subject. Note that neither the 80-Hz, 300-ms train nor the 40-Hz, 300-ms train attained a force plateau; the 40-Hz, 600-ms and 20-Hz, 300-ms trains nearly attained a force plateau; and the 20-Hz, 600-ms and 20-Hz, 1,200-ms trains all reached a force plateau. Vertical lines from left to right indicate 300-, 600-, and 1,200-ms time points, respectively.

Initially, we attempted to test also the effects of the 40-Hz 1,200-ms fatiguing protocol. In over half of the subjects, however, marked fluctuations in the force level occurred during the fatigue test, suggesting the presence of electrical fatigue. Motor units that experience electrical fatigue do not depolarize and thus contribute to the loss of force because they are not activated. This loss of depolarization recovers rapidly, however (5, 27, 28), and when subsequent trains are again able to activate affected motor units, force rises markedly and suddenly. These fluctuations in force may be further exacerbated by metabolic recovery in the idle fibers of the resting units. We observed these large, rapid changes in force with the 40-Hz 1,200-ms fatiguing protocol only. For this reason, the data related to the 40-Hz 1,200-ms fatiguing protocol are not included in the results presented. We did not observe this type of fatigue in any of the other protocols that used trains of shorter duration and/or lower frequency.

Data Acquisition and Analysis

Force responses to each train in the fatiguing protocols were collected and digitized on-line at a sampling rate of 200 Hz by using customized software (LabView 4.0, National Instruments, Austin, TX). Custom-written software (LabView 4.0) was also used to calculate the peak force and FTI responses to each stimulation train. FTI was calculated as the area under the force-time curve during muscle contraction.

The percent decline in peak force and FTI during each fatigue protocol was calculated by subtracting the mean of the final six fatiguing protocol responses from the response to the first train of the fatiguing protocol and dividing by the first train's response. Differences in percent decline in peak force across fatiguing protocols were evaluated by using a one-way, repeated-measures ANOVA. The peak force and FTI data were plotted as functions of contraction number rather than time. This study was designed to examine the fatigue induced by different stimulation trains, and it was believed that presenting the data in terms of contractions (with rest time constant across protocols) more accurately captured the effect of specific trains. Total FTI for each protocol was calculated by summing FTI responses to all 150 trains.

A nonlinear curve-fitting routine (SigmaPlot, Jandel Scientific) was used to fit the peak force responses of each subject to each protocol with a four-parameter Hill equation to compare the rates of fatiguey=y0+[(axb)(cb+xb)] Equation 1where y is force, x is contraction number,y 0 is minimum peak force produced by a given train, a is scaling factor, c is contraction at which 50% of the force decline is achieved, and b is a measure of the steepness of the slope. Both c andb were compared across protocols to provide information regarding the rates of decline. Separate one-way, repeated-measures ANOVAs were performed to determine the effect of fatiguing protocol on both b and c.

The pre- and postfatigue testing train responses were calculated from the means of the two trains of each frequency delivered before and after the fatiguing protocols, respectively. These mean responses were used for all subsequent analyses of the testing train data. One-way, repeated-measures ANOVAs were performed on the prefatigue mean peak force and FTI responses. Percent declines for each variable (testing train peak force and FTI) were calculated by subtracting the postfatigue value from the prefatigue value and dividing this difference by the prefatigue value. A two-way, repeated-measures ANOVA was used to test for significant main effects of fatiguing protocol and testing train (6 × 4) on the percent decline in peak force. If significant main effects were detected, post hoc testing was performed by using paired-sample t-tests, corrected for multiple comparisons by using Holm's sequentially rejective correction (30). Rather than make every possible post hoc comparison, we decided a priori to make certain specific comparisons. We chose to compare the 80-Hz 300-ms, 40-Hz 600-ms, and 20-Hz 1,200-ms protocols to each other and 40-Hz 300-ms and 20-Hz 600-ms protocols to each other, as these protocols contained the same number of pulses per stimulation train but were expected to produce markedly different peak forces and FTIs. LFF was evaluated by calculating the 1:100-Hz peak force and 10:50-Hz peak force ratios in the pre- and postfatigue conditions. Two-factor (fatigue state × protocol; 2 × 6), repeated-measures ANOVAs were performed on both of these force ratios. The postfatigue ratio was then divided by the prefatigue ratio, such that an equal decline in high- and low-frequency force would result in a value of 1, but the presence of LFF would result in a value of <1. Differences in the presence of LFF were evaluated by using a one-way, repeated-measures ANOVA to test for the effect of fatiguing protocol on these pre-/postfatigue ratios for both the 1:100-Hz peak force and 10:50-Hz peak force ratios.


Fatigue Protocols

Two of the subjects demonstrated an inconsistent pattern of force decline in response to several of the stimulation protocols similar to that described for the 40-Hz 1,200-ms train described inmethods. We took this as an indication that we did not recruit a consistent population of motor units during their tests, most likely due to electrical fatigue. Thus their results were excluded, and all data presented here relate to the remaining 10 subjects (5 men). The generally smooth and regular declines in peak force and FTI in the remaining subjects suggest that electrical fatigue did not affect force production during their fatigue protocols (Fig.3, A and B).

Fig. 3.

Mean (±SE, n = 10) peak force (A) and force-time integral (FTI; B) responses during the different fatiguing protocols. For purposes of clarity, only every third response is plotted, and error bars have been included only for every 30th contraction. C: mean percent decline (±SE) in peak force during each fatiguing protocol. Note that these values should not be confused with the declines in the testing trains presented in Fig. 5.D: mean (±SE) contraction number at which 50% fatigue occurred for each fatiguing protocol, as determined by fitting data inA with Eq. 1 (see methods). * Significantly different from 20-Hz, 300-ms train (P ≤ 0.05). †Significantly different from 40-Hz, 300-ms train (P ≤ 0.05). ¶Significantly different from 20-Hz, 600-ms train (P ≤ 0.05).

The greatest percent declines in peak force were seen during the 40-Hz 600-ms and 80-Hz 300-ms protocols (67.99 and 67.66%, respectively). The smallest decline in peak force was seen during the 20-Hz 300-ms protocol (43.28%), which was significantly different from all other protocols except the 40-Hz 300-ms protocol (Fig. 3 C). There was an effect of protocol on the rate of fatigue observed, as indicated by contraction at which 50% of the force decline was achieved (calculated as the c values in Eq. 1 ) (F = 7.34, P < 0.01). The 20-Hz 300-ms protocol exhibited the slowest rate of fatigue, which was significantly less than the 40-Hz 600-ms protocol and the 20-Hz 600- and 1,200-ms protocols (Fig. 3 D). No effect of protocol was observed for the other parameter calculated from Eq. 1 , namely theb values, which were representative of the steepness of the slope of the linear portion of force decline during the fatigue test. Interestingly, there were no significant differences among the peak forces produced at the end of the six fatiguing protocols, despite fairly large differences observed in the initial peak forces (Fig.3 A).

The percent declines in peak force during the fatiguing protocols were more strongly correlated to the initial peak force than to the initial FTI (Fig. 4, A andB). The contraction at which 50% of fatigue occurred, however, was more closely correlated to FTI than to peak force (Fig. 4,C and D).

Fig. 4.

Mean percent decline in the fatiguing protocols vs. mean initial peak force (A) and mean initial FTI (B). Mean rate of decline (contraction number at 50% fatigue) vs. mean initial peak force (C) and mean initial FTI (D). Regression lines, R 2 values, and Pvalues are presented in each panel.

Testing Trains

The prefatigue values of the peak forces and FTIs of the four testing trains were not significantly different across protocols. Similarly, the prefatigue values of the 1:100-Hz and 10:50-Hz force ratios across protocols were not significantly different. The peak force and FTI responses to the testing trains followed similar patterns of decline after the fatiguing protocols. For this reason, only the peak force responses are discussed here. Statistical analysis (two-way, repeated-measures ANOVA) of the declines in peak force revealed significant main effects for both testing train (F = 19.18, P ≤ 0.001) and fatiguing protocol (F = 25.21, P ≤ 0.001) but no significant interaction. Thus the different protocols produced different degrees of force decline in the four testing trains, but the pattern of changes in the testing trains, relative to each other, was similar across all protocols (Fig. 5).

Fig. 5.

Mean (± SE, n = 10) percent declines in the peak force responses to the testing trains after each of the fatiguing protocols. These values represent relative changes in testing trains delivered before and after the fatiguing protocols and should not be confused with the values in Fig. 3 C, which reflect the declines in force that occurred during the fatiguing protocols themselves. Note the consistent pattern of greater attenuation of low-frequency (1 and 10 Hz) vs. high-frequency (50 and 100 Hz) testing trains across all protocols.

We chose a priori to compare the attenuation of the testing train forces produced by the 20-Hz 1,200-ms, 40-Hz 600-ms, and 80-Hz 300-ms protocols to each other and also the attenuation produced by the 20-Hz 600-ms and 40-Hz 300-ms protocols (see Statistical Analysis). The data from these protocols (Table2) show that, with the number of pulses held constant, protocols that produced the greatest peak forces produced the greatest fatigue in the testing trains, whereas the protocols with the largest FTIs did not. In general, the differences observed were not as great among the 13-pulse trains as they were among the 25-pulse trains, showing only strong trends toward significance for the 50- and 100-Hz testing trains. However, the absolute difference in initial peak force between the two 13-pulse trains, although statistically significant, was not very large (∼11 N). The overall relationship of initial peak force of the fatiguing protocols to fatigue of the testing trains is reinforced by Fig.6, which plots the mean percent decline in each of the testing trains against the initial peak force. In each case, linear fitting of the data produced R 2values of >0.70, all of which were significant. Similar regression analyses were performed for the decline in testing train forces vs. initial FTI (data not shown). R 2 values for these plots ranged from 0.19 to 0.413, none of which was statistically significant, as determined by regression analysis. Regression analyses of percent decline in testing train force vs. total FTI (data not shown) produced R 2 values that ranged from 0.374 to 0.425, and again none was statistically significant. Fitting the plots of the initial and total FTI vs. decline in testing train force with a sigmoidal function (four-parameter Hill equation, SigmaPlot, Jandel Scientific) improved R 2 values slightly but still did not find any statistically significant correlations.

View this table:
Table 2.

Comparisons of fatigue for trains with the same number of pulses

Fig. 6.

Scatter plots for each testing train of the mean percent decline produced by each fatiguing protocol vs. mean initial peak force produced by each fatiguing protocol (means ± SE,n = 10 subjects). A: 1-Hz testing train;B: 10-Hz testing train; C: 50-Hz testing train;D: 100-Hz testing train. Regression lines withR 2 and P values are presented on each graph.

Every protocol produced greater impairment of peak force production by low-frequency trains than by high-frequency trains (see Figs. 5 and7). There was a significant effect of fatigue state (pre- vs. postfatigue) for both force ratios (1:100 Hz,F = 11.75, P = 0.008; 10:50 Hz,F = 20.65, P = 0.001). There was, however, no significant effect of protocol and no fatigue state-by-protocol interaction on either force ratio. In addition, there were no significant main effects of fatiguing protocol on the percent changes in either the 1:100-Hz (F = 0.65,P = 0.662) or the 10:50-Hz (F = 1.09,P = 0.382) force ratios. Together, these findings indicate that the degree of LFF was similar across protocols.

Fig. 7.

Mean (±SE, n = 10) pre- and postfatigue values for twitch/tetanus ratio (A) and 10:50-Hz force ratio (B). * Significant effect of fatigue as determined by 2 × 6 (fatigue state × protocol) repeated-measures ANOVA (A: F = 11.75, P < 0.01;B: F = 20.65, P > 0.01). There was no effect of protocol and no significant interaction for either ratio.


The results of this study indicate that, consistent with our hypothesis, fatigue was more closely correlated to peak force than to initial FTI, total FTI, or number of pulses for brief, intermittent stimulation trains, when rest time between trains was held constant. This is not to say that FTI can have no effect on fatigue, however. Both peak force and FTI no doubt relate to fatigue; but, for the short trains tested here, peak force appeared to be the more potent factor. In addition, all of the fatiguing protocols tested produced comparable degrees of LFF. Because all of the trains did produce LFF, use of the rate or amount of force decline during the protocols might have overestimated the fatigue produced by the lower frequency fatiguing trains. This finding supported our choice to evaluate fatigue through the percentage decline in the testing trains.

There were no significant differences among the forces produced at the end of each fatiguing protocol. This was due, at least in part, to using 150 contractions in all the protocols. Had each test contained only 60 contractions, differences in force would have been present (Fig. 3). However, we felt that because we were evaluating the response to multiple testing trains, it was important to deliver them during a stable level of fatigue, necessitating the greater number of contractions. Despite the similarity in the forces produced at the end of the fatiguing protocols, there were clear differences in the fatigue produced, as evidenced by the different percentage declines in the testing trains (Fig. 6). The changes in testing train force demonstrated that the initial peak force associated with the fatiguing protocol was a better predictor of fatigue than FTI, a quantity commonly used as a measure of work during isometric contraction (8, 10, 25), or the number of stimulation pulses delivered, which has been suggested as a major factor in fatigue (31). These results conflict with those of a previous study from our laboratory (8) that showed that initial FTI was a better predictor of fatigue than peak force. In this earlier study, however, the stimulation trains all contained the same number of pulses and were delivered at a constant train period, not at a constant intertrain interval. As a result, the protocols that produced the highest peak forces had much greater recovery times between trains than those that produced the highest FTIs. On the basis of the present findings, we would expect that our earlier results would have been different had a constant intertrain interval been used.

We based our hypothesis that the fatigue produced during a repetitive stimulation protocol would be more strongly related to the peak force than FTI on a study from our laboratory that demonstrated that the metabolic demand of attaining a given level of force was greater than that of maintaining it and the well-established relationship between metabolic demand and fatigue (2, 10, 25, 35,37). Although both the force-generation phase (related to peak force) and force-maintenance phase (related to FTI) contributed to the total metabolic cost, for the brief trains used in the present study, the effect of the force-generation phase would predominate. We did not collect any metabolic data during any of the fatiguing protocols and cannot say for certain that the observed differences in fatigue were the result of changes in metabolites. However, the results did support our hypothesis, which was based on previous metabolic findings (37, 38).

The present results, however, may not apply to all types of exercise. Meyer and Foley (34) divide muscle stimulation into a low-rate nonfatiguing domain, a high-rate fatiguing domain, and a less well-defined transitional domain. The fatiguing domain is characterized by twitch stimulation at frequencies of ≥5 Hz or repetitive tetani with duty cycles of >5–10%. The lowest duty cycle in any of the protocols used here was 20% (for the 300-ms protocols), thus all of the protocols were operating well within the fatiguing domain. Factors other than metabolic cost per contraction, and thus peak force, may be more of a factor during less intense exercise. These other factors may account for the present study's results with regard to LFF.

An unexpected finding of the present study was that comparable degrees of LFF occurred during all of the fatiguing protocols (Fig. 7). This result conflicts with our previous findings that LFF immediately after a fatigue protocol was related to the FTI (8). As noted above, the differences between that study and the present one may be due to the differences in rest times between trains. In addition, the earlier study used much briefer trains (all six pulses) and were thus probably less metabolically demanding. However, factors other than metabolites, such as elevated calcium levels, have been implicated in the production of LFF (11, 12). It is worth noting that all of the protocols in the present study contained the same number of trains, an observation that is consistent with the suggestion that LFF may be related to the number of tetani a muscle receives (12).

The results of the present study demonstrate that during brief, high-intensity, repetitive muscle contractions, the FTI produced by a stimulation train is a poor indicator of the fatigue that train will generate when a consistent rest time between trains is maintained. Initial peak force produced by a given stimulation, however, shows a strong correlation to the fatigue it produces during repetitive activation. These findings may be related to the observation that the metabolic cost of generating force is greater than that of maintaining it (38). However, further research using electromyogram and metabolic measures (magnetic resonance spectroscopy, biopsy) would be helpful in determining the mechanisms underlying the present findings. On the basis of these results, clinicians and researchers should exercise caution when using FTI as a predictor of fatigue during brief contractions, such as those commonly used during functional electrical stimulation.


Partial funding for this study was provided by the University of Delaware Office of Graduate Studies and the Foundation for Physical Therapy (to D. W. Russ), and National Institute of Child Health and Human Development Grants HD-33738 (to K. Vandenborne) and HD-42164 (to S. A. Binder-Macleod).


  • Address for reprint requests and other correspondence: S. A. Binder-Macleod, 323 McKinly Laboratory, Dept. of Physical Therapy, Univ. of Delaware, Newark, DE 19716 (E-mail:sbinder{at}udel.edu).

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

  • April 19, 2002;10.1152/japplphysiol.01010.2001


View Abstract