Low-Frequency Rectangular Pulse Is Superior to Middle Frequency Alternating Current Stimulation in Cycling of People With Spinal Cord Injury

Participants 

Eleven otherwise healthy people with chronic SCI (8 with American Spinal Injury Association [ASIA] grade A, 3 with ASIA grade B) and low levels of muscle spasm (Modified Ashworth Scale score range, 0–2) participated in this study (table 1). The muscle fiber composition of their paralyzed muscles was stable.5 For this study, subjects who had limited experience in FES cycling training (average of 6–28mo for 0.6 training sessions per week) were selected. Therefore, their strength condition corresponded to the typical initial situation of SCI patients joining the outpatient clinic. The University of Munich ethics committee approved the study and the subjects gave their written informed consent prior to their participation.

Table 1. Characteristics of Study Participants

	Subject No.	Age (y)	Sex	Age at Injury (y)	SCI Level	ASIA Classification
	1	32	M	25	T12	A
	2	38	M	19	C5	A
	3	44	M	18	T4	A
	4	41	M	23	T1	B
	5	15	M	3	T7	B
	6	37	F	29	T9	A
	7	42	F	37	T6	A
	8	33	F	29	T2	B
	9	47	M	44	T6	A
	10	30	M	20	T5	A
	11	32	M	27	T7	A

Abbreviations: F, female; M, male.

Study Design 

Each subject underwent 3 different experimental sessions: (1) isometric measurements using LFRP and MFAC stimulation, (2) ergometry using LFRP stimulation, and (3) ergometry using MFAC stimulation. Session order was randomized and each session was performed on a different day within 6 weeks. Subject 8 did not participate in the isometric measurement.

Stimulation 

The quadriceps, hamstrings, and glutei muscle groups were electrically stimulated for ergometer cycling.16 Pairs of auto-adhesive gel electrodes (Flextrode)a (size, 4.5×9.5cm2) were placed on the skin over the proximal and distal fourth of each muscle bulk. A constant current 8-channel stimulator (Motionstim)a provided the LFRP current (rectangular, biphasic, charged balanced pulses; frequency, 20Hz; maximum pulse amplitude, 127mA; constant pulse width, 500μs) (fig 1A). For MFAC stimulation a middle frequency constant current 6-channel stimulatorb provided the modulated MFAC (4kHz sinusoidal modulated with 50Hz on-off rectangles; duty cycle, 1:1) (fig 1B). We selected a carrier frequency of 4kHz because it had previously produced a pronounced reduction of fatigue rate18 in able-bodied subjects.

View Large Image Download to PowerPoint

Fig 1. (A) LFRP and (B) MFAC waveforms.

The middle frequency stimulator could provide maximally 140V peak-to-peak. Assuming a skin resistance of 1kΩ, at 140V the current density on the electrode area (1.14mA RMS/cm2, where RMS is root mean square) did not exceed the safety limit23 for alternating current (ie, 2mA RMS/cm2). Both stimulators had the capacity to reach the saturation region of the recruitment curves for all participants. During the isometric measurements and the ergometer cycling the stimulators were controlled from a laptop computer by serial communication (fig 2).

View Large Image Download to PowerPoint

Fig 2. Isometric and ergometer measurement setups. Because patients’ geometric sitting position and the stimulator setup were identical, a combined draft of isometric and ergometer experiments is presented. The crank angle was set by manually turning the torque transducer axle (TTA) by a lever and fixing it with a screw during isometric measurement. Resistance torque and crank angle provided by TTA and decoder (C), respectively, were collected. Electrode leads were connected at a given crank angle alternately to the LFRP and MFAC stimulator by a switch. During ergometric measurements pedaling was motor braked. Tangential forces provided by built-in sensors in the crank arm and crank angle were collected. Either LFRP or MFAC stimulation was used. Legend: 1, 2, 3: stimulation of the quad, hamstrings, and gluteus muscle groups, respectively.

During ergometer cycling, the laptop directed the muscle stimulator to induce muscle contractions at the appropriate crank angles (table 2) to produce pedaling. Because muscle stimulation angles were statically defined,16 angular compensations were necessary for the dynamic application. Assuming a muscle force rise time of 140ms, it was calculated that at 35rpm muscle stimulation should occur 28° earlier.

Table 2. The LFRP Stimulation Angles for Patient 10 and for All Study Participants

	Muscle Group	Start Patient 10 (deg)	Stop Patient 10 (deg)	Start All 11 Subjects (deg)	Stop All 11 Subjects (deg)
	Left quadriceps	349	171	7±22	183±11
	Left hamstrings	87	272	94±31	259±41
	Left gluteus	87	272	94±31	259±41
	Right quadriceps	177	0	191±28	23±25
	Right hamstrings	320	85	287±46	83±23
	Right gluteus	320	85	287±46	83±23

NOTE. Values are mean ± standard deviation (SD) unless otherwise indicated. These stimulation angle ranges contain an angular compensation of 28°. Hamstring and glutei muscle ranges were set equal due to the stimulator’s technical limitations. Zero degrees refers to the backward-pointing left crank arm.

Isometric Torque Measurements 

A stationary tricyclec with its front wheel replaced by a torque transducerd served as the test bed for isometric torque measurements (see fig 2). An 8-bit incremental encoder, synchronized to turn with the crankshaft, determined the actual position of the crank. The ankle joint was immobilized at 90° and leg movement was restricted to the sagittal plane by using shank and foot orthoses. Shanks and feet were fixed to the orthoses by self-adhesive (Velcro) straps. A lever manually moved the crank into 18 equiangular positions. First, the maximal torque-producing crank position was determined for each muscle group; the optimal crank angle (eg, for LFRP stimulation of the left quadriceps of subject 10 the optimal crank angle was 75°, with corresponding left hip angle of 58° and knee angle of 93°) (see fig 2), maximal torque, and maximal current were noted. Stimulation at 40% to 80% of the maximal current (depending on the patient) was then sequentially applied to the 6 muscle groups. The stimulation phases lasted 1.5 seconds and were followed by breaks of 3 seconds duration.

Moments were evoked at the given crank angle by consecutively using LFRP and MFAC stimulation (electrode leads were connected alternately to the low- and the middle-frequency stimulators by means of a switch) and were subsequently recorded on the laptop. Torque recordings of the 6 muscle groups provided the isometric moment versus angle characteristics.

We attempted to include only active muscle contributions to the torque recordings by eliminating offline the bias of isometric torque due to passive gravitational and elastic components. Furthermore, the isometric moments were extrapolated to maximal torque (100%) values taken at optimal crank angles. Conforming to common FES cycling practice, only the positive (acting in the drive direction) half-waves of the moment versus angle characteristics24 were used to calculate the sum of these half-waves (see Results and figs 3A, 3B). The sum was integrated over 0° to 360° and divided by 360° to give the mean maximal isometric torque, which was used in further processing. We used the mean maximal isometric torque in addition to maximal quadriceps torque because in FES cycling the mean maximal torque, including quadriceps, hamstrings, and glutei positive half-wave contributions, must overcome the drive resistance.24

View Large Image Download to PowerPoint

Fig 3. Isometric torques of left quadriceps muscle (red) and the sum of all muscle torques together (blue) produced when subject 10 pedaled in positive drive direction with (A) MFAC and (B) LFRP stimulation. Torques were actually measured at 40% to 80% of maximal stimulation and subsequently extrapolated to 100% (arrows). Zero degrees refer to the backward-pointing left crank arm. The dashed line indicates mean maximal isometric torque.

Ergometric Experiments 

Because untrained SCI subjects are quite weak,21, 25 we used an ergometer (see fig 2) with a motor-powered brake and drive.e This permitted a low minimal braking torque of 1Nm on the crank (in the first gear). We measured the tangential forces applied by the rider’s right and left legs using strain-gauge instrumented cranksf that could measure up to 1kN, and had an accuracy of ±1N. The participant sat on the ergometer chair in the same geometrical position (hip − crank axis distance and tilt) as when isometric measurements were taken on the cycle.

All participants had to complete 20 minutes (considered in a previous study26 to be relevant for SCI cycling) of continuous pedaling. After a warm-up phase of several minutes, during which the legs were passively turned by the ergometer motor, the stimulation intensity was gradually increased over an average 5±1.7 minutes to the maximum current (100%). Ergometer cycling resistance was increased with stimulation intensity to maintain pedaling cadence in the range of 35 to 55rpm. Stimulation was fixed at 100% for the remainder of the pedaling. As a subject’s muscles fatigued and cadence dropped below 36rpm the braking resistance was manually reduced (in steps of 0.7Nm) to restore pedaling cadence. The crank angle position (10-bit resolution) and the tangential forces collected by the instrumented crank arm were transmitted serially every 50ms to the laptop for recording.

Cadence was computed from the change of crank position over time, which was digitally filtered with a second-order Butterworth filter with a cutoff frequency of 4Hz. The resultant 2-legged crank torque was calculated by adding the left and right tangential crank forces, multiplying this sum by the crank arm length (.15m), and smoothing it by a second-order, zero-phase low-pass filter with a time constant of 1 second. Pedaling work was computed as the integral of cadence by crank torque over the 20 minutes of stimulation (20-min work, see Results and figs 4A, 4B), which included the stimulation startup period (increasing stimulation) and maximal stimulation period. We did not consider the warm-up period of passive cycling for integration.

View Large Image Download to PowerPoint

Fig 4. Power course (upper graph) of (A) subject 10 and (B) subject 9 generated during 20 minutes of FES ergometry using LFRP (blue) and MFAC (red) stimulation. Twenty-minute pedaling work was defined as the area under the corresponding curves. Stimulation course (lower graph) is expressed in percentage of maximal stimulation.

During ergometer cycling at maximal stimulation the visual analog scale (VAS) was recorded to subjectively quantify the discomfort of each stimulation pattern.

Statistical Analysis 

The sample size of 11 was chosen according to a required minimal detectable difference27 of 3600J (estimated in preliminary experiments) between averaged 20-minute work generated under LFRP and MFAC stimulation, requiring a significance α less than .05 and power 1 − β greater than 0.7.

After the data were collected and processed, we deduced descriptive statistics for mean maximal isometric torque and 20-minute pedaling work. Because normal distributions for either isometric torque or pedaling work could not be assumed, we applied the nonparametric paired-sample Wilcoxon signed-rank test and the Spearman rank correlation test to compare and investigate the linear dependency of these variables under LFRP and MFAC conditions. Descriptive statistics are presented as mean ± standard deviation (SD) and unless otherwise mentioned, statistical tests were considered significant at P less than .05. All analyses were performed with the Statistics Toolbox in Matlab.g

Torque and Work 

The major finding of this study is that LFRP stimulation is superior to MFAC in producing isometric torque and dynamic cycling work by paralyzed skeletal leg muscle of SCI subjects. We expected that LFRP isometric torque would be superior to MFAC torque. This finding is similar to findings in studies with able-bodied subjects13, 28 (we know of no studies on MFAC stimulation-evoked torque or work in SCI subjects).

A second important finding is that pedaling work produced during the first 20 minutes of cycling is higher during maximal stimulation with LFRP than with MFAC. A comparison of LFRP and MFAC conditions showed that 20 minutes of pedaling work diminished more (down to average of 59% [4716J/8445J]) than maximal isometric torque (down to an average of 86% [14.2Nm/16.6Nm]). This means that the reduction of the maximal torque that can be evoked during MFAC is accompanied by an increased fatigue rate in SCI patients, which is in contrast to MFAC stimulation in able-bodied subjects,18 where there was a decreased fatigue rate at increasing kilohertz frequencies.

Explanation of the Increased Dynamic Fatigue Rate in MFAC Stimulation 

Any interpretation of these findings must consider that the rate of fatigue during voluntary dynamic exercise (shortening contractions) in able-bodied persons (no similar data are available for SCI29) is greater than during voluntary isometric contraction, that is, the dynamic fatigue rate is generally higher than the static fatigue rate.30 Presumably the reasons for the greater fatigue rate in dynamic contractions are: (1) the energy requirements are higher for dynamic compared with isometric exercise, and (2) shortening contractions are more sensitive to the metabolic changes that occur in exercising muscle and they alter the force-velocity relationship through the selective fatigue of fast-twitch fibers. This effectively transforms the muscle into a slower type with reduced power output. Moreover, this could be explained by the complete exhaustion of the fast-twitch fibers during dynamic contractions, which has been observed in cat muscles31 at higher contraction velocities.

Our data suggest that in chronically paralyzed muscle, MFAC stimulation induces more dynamic fatigue (vs isometric fatigue) than LFRP stimulation, thus enhancing the effects of the 2 general causes of dynamic fatigue by using a common physiologic pathway. Two mechanisms may be involved in accentuating dynamic fatigue: (1) high-frequency stimulation, particularly MFAC, is energetically more demanding than LFRP stimulation,32 and (2) the dropout of fast-fatigue IIb fibers in MFAC stimulation actually signifies a “conversion” into a slow-fatigue IIa type of muscle with a reduced power output. This is similar to the fast-fiber dropout during MFAC stimulation of able-bodied subjects.18 Further work is necessary to determine whether a selective dropout mechanism that depends on the fatigue properties of fibers is actually responsible for differing fatigue rates during MFAC and LFRP stimulation of SCI muscle, or whether a nonselective mechanism such as higher energy demand in the case of MFAC must be considered.

Advantage of LFRP Over MFAC Stimulation is Strength Condition Dependent 

This study has proven that the difference between LFRP and MFAC work correlates highly significantly with the absolute amount of work produced by an individual subject. Therefore, as regards induced functional movement, the advantage of LFRP stimulation over MFAC stimulation becomes more important as an SCI subject’s strength increases. We assume that strength is indicated by the absolute amount of work performed.

From a practical viewpoint, our results show that the force contribution produced by fast-fatigue IIb fibers (or simply by more fibers in case of nonselective dropout) cannot be ignored when optimizing FES cycling in people with SCI. LFRP has the most advantage over MFAC stimulation in strong SCI subjects, who probably have more fast-fatigue IIb fibers (or simply more fibers if nonselective dropout applies) than do weaker subjects. We expect that the well-trained FES cyclist that we mentioned earlier,17 whose performance was outstanding, could probably drive even better with LFRP than with MFAC stimulation.

Pain 

We measured pain in SCI subjects during the application of relevant maximal stimulation. In contrast to studies of the able-bodied,14, 33 there was a nonsignificant trend for more pain and discomfort to be perceived when MFAC stimulation was used. It was remarkable that all participants who still had pain sensations (3 of 4 had sensory incomplete SCI) felt abdominal tugging during MFAC stimulation. Because MFAC stimulation penetrates more deeply into the quadriceps and glutei muscles, it may affect deeper branches of the iliohypogastric nerve. High current densities that cause thermal burns (blisters) are also more likely to occur with MFAC stimulation than with standard LFRP stimulation.1

Further Work 

Comparison of the functional outcome achieved in SCI subjects with 4kHz MFAC and LFRP stimulation raises the question of whether there is a carrier frequency in the range of 1 to 4kHz that would ensure a function outcome superior to that obtained from LFRP. Based on our results and a comparison with isometric torque performed in able-bodied subjects,13, 14, 28 it can be assumed that the isometric torque in SCI subjects is higher and the dynamic fatigue rate is lower at 2.5kHz MFAC than at 4kHz MFAC. It therefore seems expedient to determine in further investigations whether 2.5kHz MFAC is more effective than LFRP stimulation in regard to torque, work, and dynamic fatigue rate.

The correlation between the interconditional work difference LFRP-MFAC and subject strength that we found in this study suggests that LFRP stimulation-based training of subjects with SCI increases the advantage of LFRP over MFAC stimulation in regard to work performed. Stefanovska and Vodovnik13 found that MFAC stimulation-based training of the able-bodied led to less isometric force gains than LFRP stimulation-based training. It must still be determined whether MFAC stimulation-based training of SCI subjects can efficiently strengthen the IIa fibers (slow-fatigue) through the presumed selective dropout of the IIb fibers (fast-fatigue). In our opinion, this might explain the outstanding performance described in our SCI case study.17