Task dependency in back muscle fatigue – Correlations between two test methods

1. Introduction 

According to the task, different physiological mechanisms are involved in fatiguing contractions. This is referred to as task dependency of muscle fatigue. Influencing variables can be, e.g., subject motivation, central command, intensity and duration of the activity, speed and type of contraction, and intermittent or sustained activities (Enoka, 1995). Electromyography (EMG) is one way of studying muscle fatigue that has been extensively used during the last decades. The median (or mean) frequency of the power spectrum will decrease during an isometric muscle contraction as a sign of myoelectric muscle fatigue (De Luca, 1997). A faster decline for patients compared to healthy persons have been found (Mayer et al., 1989, Roy et al., 1989), though more recent studies have shown the opposite results (Elfving et al., 2003, Kramer et al., 2005, Larivière et al., 2003, Peach and McGill, 1998) or no difference (Crossman et al., 2004).

Various test positions have been used in studies of isometric back muscle fatigue. Most common is the Sørensen test, i.e., prone unsupported trunk in horizontal position, e.g., (Kankaanpää et al., 1998, Koumantakis et al., 2001, Mannion et al., 1997, Ng et al., 1997) or modified with the hips flexed 40° (Dedering et al., 1999). Other test positions used are standing upward pull (Arnall et al., 2002, Humphrey et al., 2005, Mannion et al., 1997) back extension in standing position (Koumantakis et al., 2001, Roy et al., 1989), semistanding (Peach and McGill, 1998) or seated position (Elfving et al., 1999, van Dieën et al., 1998) with simultaneous monitoring of force feedback. Contraction forces between 40 and 80% of maximal voluntary contraction (MVC) have been used in these studies. Ratings of muscle fatigue is an important complementary measure to EMG. Perceived effort increases non-linearly with increasing force level (Dedering et al., 2002, Enoka and Stuart, 1992).

Few studies have been made that compare different methods of testing back muscle fatigue. The Sørensen test has been compared with standing isometric backward press at 60% of MVC (Crossman et al., 2004, Koumantakis et al., 2001) and with standing isometric upward pull at 60% of MVC (Mannion et al., 1997). In these studies mean differences between the methods were reported and in the study by Koumantakis et al. (2001) the reliability of the two methods was compared. However, little investigation has been made about correlations between test methods. An important question would be if an individual fatigues himself equally in different test situations, or if the measured fatigue indices depend on the type of muscle fatigue test. For example, Bonato et al. (2003) found low correlation between a dynamic lifting task and a static submaximal task at 80% of MVC as regards decline of the median frequency.

Followingly, task dependency for tests of back muscle fatigue is not thoroughly investigated and knowledge is still lacking. The hypothesis would be that the choice of test method may influence the recorded back muscle EMG fatigability of an individual. Our research group has used the Sørensen test (Dedering et al., 1999, Dedering et al., 2000) and a seated test position in a David Back Extension Device (Elfving et al., 2000, Elfving et al., 2003) in studies of EMG fatigue and ratings of back muscle fatigue. In the present study we have, using healthy subjects, compared the two test methods with the aim of studying a possible task dependency in lumbar muscle fatigue. The following questions were addressed: what differences in EMG spectral variables could be seen between the two test methods? Were EMG spectral variables and ratings of fatigue correlated between the two test methods? Were there any correlations between EMG spectral variables and rated muscle fatigue within each test?

2. Methods 

2.4. Electromyography 

EMG was recorded during the fatigue tests from four sites on the lower back muscles. Four pairs of electrodes (Blue Sensor N-00-S, Medicotest A/S, Denmark) were placed at 3cm to the left and right of the spinous processes of the 1st and 5th lumbar vertebrae. The centre-to-centre distance of the electrodes was 2cm. A ground electrode was placed on the left malleolus (Blus Sensor VL-00-S, Medicotest A/S, Denmark). A telemetric EMG system was used (Telemyo 16, Noraxon, Scottsdale, Arizona, USA). The bandwidth was 10–500Hz with a high-pass filter of −12dB/oct and a low-pass filter of −30dB/oct. The sampling frequency was 1000Hz. EMG analysis was done with software from Noraxon (MyoResearch 97, Noraxon, USA). The median frequency (f) of the power spectrum was calculated for consecutive 1-s intervals of the recorded signal. A Hanning window was used prior to Fast-Fourier transformation.

To obtain the EMG variables, linear regression was used with f as a function of time (s). Time was 45s in the seated test and total contraction time until exhaustion in the prone test.

Outcome EMG variables were:

•Initial median frequency fi (Hz). Seated test: intercept of the regression line. Prone test: mean of the first 5s.

•End median frequency fe (Hz). Seated test: fi + slope of regression line∗45s. Prone test: mean of the last 5s.

•Normalised slope of median frequency (%/s): both tests: slope of the regression line/fi∗100.

•Median frequency decrease for the whole contraction time (%): both tests:

•Recovery half-time (s): non-linear regression of f as a function of time (min) by fit of recorded f to the expression

The non-linear regression analysis is described in detail in Elfving et al. (2002a).

3. Results 

Significant differences between the mean values obtained in the seated and the prone test were found for the EMG variables initial median frequency, slope and decrease, and for the ratings (Table 1). No significant differences were found for end median frequency and recovery half-time.

Table 1. Variables from the seated and the prone test

	Test position			P-value
		Seated	Prone
		Mean (SD)	Mean (SD)
	Time (s)	45	299 (153)
	MVC (Nm/kg)	2.57 (0.49)
	EMG
	fi (Hz)	55.2 (8.2)	70.9 (7.5)	P<0.001
	fe (Hz)	42.0 (6.4)	42.0 (9.4)	0.968
	Slope (%/s)	−0.505 (0.263)	−0.152 (0.072)	P<0.001
	Decrease (%)	−22.7 (11.8)	−39.3 (13.3)	P<0.001
		Median (Q1–Q3)	Median (Q1–Q3)
	Recovery half-time (s)	40.2 (20.9–71.1)	56.2 (45.7–84.3)	0.249
	Ratings
	Borg	8 (5–10)	10 (10–12)	P<0.001

For the seated position the time was set to 45s (at 80% of MVC). P-values for differences between the seated and the prone test are shown. MVC, maximal voluntary contraction, fi, initial median frequency, fe, median frequency at the end of contraction, slope, decline of median frequency during contraction, decrease, (fi−fe)∗100/fi.

Moderate and significant correlations between the seated and the prone test were found for the initial median frequency (r=0.69, P<0.05) (Fig. 1a) and for the ratings of muscle fatigue (rs=0.51, P<0.05) (Fig. 1b). Low correlation was found for the end median frequency (r=0.45) (P<0.05) (Fig. 1c) and the decrease (r=0.42) (Fig. 1d). No correlation was found for the slope (r=−0.08) (Fig. 1e) and the recovery half-time (rs=−0.05) (Fig. 1f).

View Large Image Download to PowerPoint

Fig. 1. Correlation between the seated and the prone test for (a) the initial median frequency (Hz) r=0.69; P<0.05, (b) the Borg ratings of back muscle fatigue (Borg scale 0–10), rs=0.51; P<0.05, (c) the end median frequency (Hz) r=0.45; P<0.05, (d) the median frequency decrease (%) r= 0.42, (e) the slope of median frequency (%/s) r=−0.08, and (f) the median frequency recovery half-time (s) rs=−0.05.

Neither for the seated test nor for the prone test was any correlation found between the Borg ratings of back muscle fatigue and the EMG variables indicating muscle fatigue (slope and decrease) (rs<0.19).

4. Discussion 

The most notable findings were that the EMG fatigue variables showed low correlations between the two tests of back muscle fatigue, while the ratings of back muscle fatigue correlated to a higher degree. The myoelectric manifestations of fatigue therefore seem to be more task dependent than the subjective experience of fatigue. It may also be noted that the initial median frequency showed a nearly high correlation between the tests, but this variable is a measure of the non-fatigued muscle at the start of the contraction.

Important similarities between the two tests were the isometric muscle contraction and the EMG method. However, some factors differed between the seated and the prone test such as muscle force, contraction time, coordination, muscle length, and verbal encouragement. The initial median frequency and the slope were, as expected (Arnall et al., 2002, Dedering et al., 2002, Elfving et al., 2002b, Mannion and Dolan, 1996), influenced by contraction force. Since the contraction force as a percentage of MVC in the prone test was much lower than the 80% of MVC in the seated test, the average initial median frequency was also significantly higher in the prone test, and the average slope was significantly less (Table 1). In the seated test the lordosis was kept in position by a pad behind the lower back, resulting in a somewhat shorter length of the erector spinae muscle fibers than in the prone test. A posture with shorter muscle length has shown to have higher median frequency compared to a posture with longer muscle length (Mannion and Dolan, 1996, Rosenburg and Seidel, 1989). In our study, however, the opposite was found, i.e., a lower initial median frequency in the seated position with shorter muscle length, probably due to the higher muscle force applied in this position.

A steeper slope in the seated test did not entail a steeper slope for the same individuals in the prone test. Anthropometric measures may be a reason, but no correlations were found between body mass index and the slope. Another reason might be the low reliability for the slope previously shown (Elfving et al., 1999). It may be noted that the median frequency decrease showed somewhat higher correlation between the two tests than the slope. Contraction time, slope and decrease are linked (decrease=slope∗contraction time). Since the slope can be regarded as approximately constant (i.e., linear regression is used) one might argue that the slope cannot be a measure of total (accumulated) muscular fatigue obtained during a test. However, the decrease may be proposed to be such a measure. It is interesting to note that in both tests the median frequency decreased to approximately the same end value (fe Table 1), although starting at a higher initial value in the prone test.

Ratings of fatigue during the contraction in the present study showed a moderate correlation between the seated and the prone test. However, the ratings and the EMG variables showed very low correlations within both tests. This agrees with previous results (Dedering et al., 2002, Elfving et al., 2000). For low force and long duration contractions no significant median frequency decrease was found over time in spite of high subjective fatigue (Farina et al., 2003). This lack of correlation raises interesting questions on validity. The construct to be measured is muscle fatigue. Muscular fatigue is, however, a component of both peripheral and central factors (Gandevia, 2001). From the current results it seems that individuals can experience muscle fatigue without any measurable decrease in median frequency.

Several studies using EMG frequency spectrum variables (Elfving et al., 2003, Humphrey et al., 2005, Klein et al., 1991, Mayer et al., 1989) have supported the ability of these variables to discriminate between patients with low back pain and healthy subjects using various test methods. Most tests previously used (Klein et al., 1991, Peach and McGill, 1998, Kramer et al., 2005, Elfving et al., 2003) were of less than a minute’s duration with relatively high contraction forces, although some tests were performed in positions more typical in physical work (lifting) (Bonato et al., 2003, Humphrey et al., 2005, Mannion et al., 1997). Testing in typical work situations would seem preferable in a clinical perspective. During a working day the postural back muscles contract with very low force for hours, alternately with higher exertion levels depending on the work task. To our knowledge, the only authors who have used low force (about 13% of MVC) and long duration (30min) contractions of the low back muscles are Farina et al. (2003) in a methodological paper. However, they conclude that, for these types of contraction the EMG spectral variables could not be recommended, since the non-stable motor unit pool has a major influence on the EMG variables. The presently found low correlation of median frequency decrease between the two studied tests is interesting from the point of validity. It may be that the tests measure different aspects of muscle fatigue. However, neither test measure in a position appropriate for activities of daily living.

5. Conclusion 

The presently found low correlation between the two test methods indicates that EMG variables assessed in one type of task in a fatiguing test may not be valid for other types of fatiguing tasks, for example in daily life work situations. Thus task dependency has to be considered when using surface EMG in determining lumbar muscle fatigue. Ratings of fatigue seem to be less task dependent than the EMG variables. Physiologic fatigue and subjective fatigue showed no association.