The assessment of back muscle capacity using intermittent static contractions. Part I – Validity and reliability of electromyographic indices of fatigue

Available online at www.sciencedirect.com

Journal of Electromyography and Kinesiology 18 (2008) 1006–1019 www.elsevier.com/locate/jelekin

Review

The assessment of back muscle capacity using intermittent static contractions. Part I – Validity and reliability of electromyographic indices of fatigue Christian Larivie`re

a,*

, Denis Gagnon b, Denis Gravel c, A. Bertrand Arsenault

c

a Occupational Health and Safety Research Institute Robert-Sauve´, Montreal, Quebec, Canada H3A 3C2 Department of Kinanthropology, University of Sherbrooke, 2500 Boul. Universite´, Sherbrooke, Quebec, Canada J1K 2R1 School of Rehabilitation, University of Montreal, C.P. 6128, Succursale Centre-Ville, Montreal, Quebec, Canada H3C 3J7

b c

Received 30 August 2006; received in revised form 6 February 2007; accepted 12 March 2007

Abstract Introduction: Back muscle capacity is impaired in chronic low back pain patients but no motivation-free test exists to measure it. The aims of this study were to assess the reliability and criterion validity of electromyographic indices of muscle fatigue during an intermittent absolute endurance test. Methods: Healthy subjects (44 males and 29 females; age: 20–55 yrs) performed three maximal voluntary contractions (MVC) and a fatigue test while standing in a static dynamometer. Surface EMG signals were collected from four pairs of back muscles (multifidus at the L5 level, iliocostalis lumborum at L3, and longissimus at L1 and T10). The fatigue test, assessing absolute endurance (90-Nm torque), consisted in performing an intermittent extension task to exhaustion. Strength was defined as the peak MVC whereas our endurance criterion was defined as the time to reach exhaustion (Tend) during the fatigue test. From the first five min (females) or ten min (males) of EMG data, frequency and time-frequency domain analyses were applied to compute various spectral indices of muscle fatigue. Results: The EMG indices were more reliable when computed from the time-frequency domain than when computed from the frequency domain, but showed comparable correlation results (criterion validity) with Tend and Strength. Some EMG indices reached moderate to good correlation (range: 0.64–0.69) with Tend, lower correlations (range: 0.39–0.55) with Strength, and good to excellent between-day test-retest reliability results (intra-class correlation range: 0.75–0.83). The quantification of the spectral content more locally in different frequency bands of the power spectrum was less valid and reliable than the indices computed from the entire power spectrum. Differences observed among muscles were interpreted in light of specific neuromuscular activation levels that were observed during the endurance test. These findings supported the use of an intermittent and time-limited (5–10 min) absolute endurance test, that is a practical way to assess the back capacity of chronic low back pain subjects, to assess absolute endurance as well as strength with the use of electromyographic indices of muscle fatigue.  2007 Elsevier Ltd. All rights reserved. Keywords: Low back pain; Back muscles; Endurance; Strength; Impairment; Reliability; Validity; Dynamometry; Electromyography; Spectral analysis

Contents 1. 2.

*

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Schedule of assessments and tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Corresponding author. Tel.: +1 514 288 1551x217; fax: +1 514 288 6097. E-mail address: [email protected] (C. Larivie`re).

1050-6411/$ - see front matter  2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jelekin.2007.03.012

1007 1008 1008 1008

3.

4.

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

1007

2.3. Measurement techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Data processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Overall description of subjects performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. How many cycles are required to compute EMG indices without requiring a too long endurance test? . . . . . . . . . 3.3. Criterion validity and reliability of EMG indices of back muscle fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Neuromuscular activation level of back muscles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. SFFT vs wavelets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Global vs local analyses of the EMG spectral content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Criterion validity of EMG fatigue indices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4. Reliability of EMG spectral indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5. Neuromuscular activation level of back muscles during the endurance test . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1009 1010 1010 1012 1012 1012 1013 1014 1014 1014 1015 1015 1015 1016 1016 1017

1. Introduction The deterioration of the trunk extensor muscles following a first episode of low back pain (LBP), which has been substantiated by a decrease of muscle mass (Danneels et al., 2000; Hides et al., 1994) and a progressive alteration of muscle fiber characteristics (Mannion et al., 2000), is recognized as a potential cause of the recurrent nature of LBP (Hides et al., 1996; Mannion, 1999). This has been attributed to the so-called deconditioning syndrome, a concept that is operationally defined, through functional measures, by a decrease of the strength and endurance of the erector spinae muscle (Hultman et al., 1993), to name only a few out of the deconditioning-related functional changes (Verbunt et al., 2003). The role of strength in the etiology of low back pain is not straightforward because it is emphasized only when strength is expressed in relation to job demands (Dempsey et al., 1997). The role of back muscle endurance, on the other hand, is clearly identified as one predictor of the first occurrence of low back pain (Biering-Sorensen, 1984; Luoto et al., 1995) and also as a predictor of longterm back-related disability when assessed four weeks post-injury (Enthoven et al., 2003). This probably explains why a considerable amount of research has been performed on this topic in the last decades (Moreau et al., 2001). However, the validity of strength and endurance tests to assess physical fitness in persons with chronic LBP (CLBP) is questioned because of the detrimental influence of psychological factors on these measures (Verbunt et al., 2003). The development of psychological-free measurements is proposed to better define the relationship between the various components of physical fitness and CLBP (Verbunt et al., 2003). The use of surface electromyography (EMG) to quantify muscle fatigue during a submaximal endurance test represents a good example. It is a motivation-free alternative to the usual back muscle endurance tests requiring the subject to sustain an effort until exhaustion (De Luca, 1993). The capacity of EMG frequency analyses to predict muscle fatigue has been demonstrated

for back muscles (Kankaanpaa et al., 1997; Mannion and Dolan, 1994; Mannion et al., 1997a; Sparto et al., 1999) as well as for other muscle groups (Maisetti et al., 2002; Merletti and Roy, 1996), to give a few examples. Most EMG-based back muscle fatigue assessments generally involves a sustained static effort at a high level of strength (Lariviere et al., 2003a; Roy et al., 1989). Although this allows the phenomenon of muscle fatigue to be quickly revealed through processed EMG, it is clear that muscle endurance is evaluated under conditions that do not correspond to tasks performed in the workplace (intermittent contractions at a low to moderate level of strength). Therefore, the endurance of muscle fibers used at high levels of strength would be evaluated (60–80% of the MVC) with complete occlusion of intramuscular blood flow instead of evaluating the endurance of the muscle fibers used at low to moderate levels of strength (25–60% of the MVC) with partial and intermittent occlusion of blood flow, two situations involving completely different mechanisms leading to fatigue (Enoka and Stuart, 1992). In fact, the physiological mechanisms involved are very different because the energy pathways are not involved in the same proportions and the metabolic byproducts are not generated and eliminated at the same rate (Christmass et al., 1999b; Christmass et al., 1999a). Also, the recruitment patterns of motor units are altered by brief variations of contraction amplitude (Westad et al., 2003). Nevertheless, changes in EMG occur less rapidly during intermittent contractions (Christensen and FuglsangFrederiksen, 1988; Hagberg, 1981), which corresponds to the direct observations (mechanical criteria) made under these specific muscle contraction conditions (Bjorksten and Jonsson, 1977;Pitcher and Miles, 1997). There is growing interest to apply EMG measurements to quantify muscle fatigue during more complex tasks performed intermittently and at low to moderate intensity to better mimic occupational tasks (Nussbaum, 2001). Such assessment protocols would help to make inference to muscle endurance in relation to work.

1008

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

An important problem in endurance testing is the choice of the load to be sustained. In theory, endurance testing should be done with a load proportional to the maximal strength considering the relationship between endurance time and relative loading (Rohmert, 1960). However, except for Roy et al. (Roy et al., 1989) where the subjects were tested during their pain remission, evidence is accumulating that using a relative load (% of maximal strength) generally leads to counter-intuitive results, that is CLBP patients shows more back muscle endurance than healthy controls (Capodaglio et al., 1995; Elfving et al., 2003; Kramer et al., 2005; Lariviere et al., 2003a; Oddsson and De Luca, 2003). A number of researchers attributes these contradictory results to the production of underestimated maximal voluntary contractions (MVC) which are used to define the relative load (Biedermann et al., 1991; Kramer et al., 2005; Lariviere et al., 2003a). This would be explained by the fear of pain and the fear of injury behaviors that is extensively observed in CLBP patients (Verbunt et al., 2003; Vlaeyen and Linton, 2000). The present study presents a new fatigue task, the functional endurance test (FET), to deal with the limitations of previous EMG-based endurance tests of back muscles. Briefly, the new fatigue task involves repeated (cyclic) intermittent back extension efforts at a predefined absolute L5/ S1 extension torque level of 90 Nm. This is accessible to practically all individuals (males and females, weak and strong subjects, back pain and healthy subjects). Obviously, the test assesses absolute endurance which is partly dependent on the strength as well as the relative endurance of individuals. Thus, it would be more appropriate to say that the test assess back muscle capacity as a whole and not endurance per se, considering that the ‘‘pure’’ definition of endurance (relative endurance) requires to exclude the influence of strength. However, practically it is justified to test this form of endurance (absolute endurance) because the same load is imposed to all individuals in most work environment. Furthermore, considering the relation between strength and absolute endurance as well as the relation between several EMG variables and the force level, it is likely that the variables presumably sensitive to muscle fatigue would also be sensitive to strength per se. Consequently, strength could be possibly predicted, in addition to absolute endurance, from the same submaximal test. The aim of this study was fourfold: to assess (1) how many cycles are necessary to predict endurance without requiring a too long endurance test (2) the criterion validity and (3) the between-day reliability of EMG indices conventionally used to assess muscle fatigue and (4) to quantify the muscle activation level of different back muscles during the FET. The first element refer to the general idea of the endurance test that is to collect surface EMG during a time-limited task to predict back muscle capacity without the need to produce a maximal performance to reach exhaustion. The criterion validity was evaluated through correlation analyses with an endurance criterion as well as with a strength criterion. In part 2 of this 2-part series,

other correlates of back muscle strength and endurance were tested. The overall objective of this duo is to retain the most promising correlates of the back muscle capacity. The general idea is to eventually consider them collectively, using stepwise multiple linear regression procedures, to predict back muscle strength and absolute endurance. 2. Methods 2.1. Subjects Seventy-three subjects (44 males and 29 females) aged between 20 and 55 yrs were recruited on a voluntary basis from the general population (Table 1). Exclusion criteria were the following: back pain in the previous year or back pain lasting longer than one week in previous years, surgery on the musculoskeletal system of the trunk, known congenital malformation of the spine or scoliosis, leg length discrepancy (>1.5 cm), body mass index >30 kg/m2 (obesity criterion, (Garrow and Webster, 1985)), systemic – neurological – degenerative disease, history of stroke, pregnancy, one positive response to the Physical Activity Readiness Questionnaire (Thomas et al., 1992), abnormal blood pressure, family history of heart attack, medication for cholesterol or triglyceride control and involvement in a new training program. The subjects were informed of the experimental protocol and potential risks and gave written consent prior to their participation. The study and consent form were approved by the ethics committee of the Centre de Recherche Interdisciplinaire en Re´adaptation du Montre´al me´tropolitain. 2.2. Schedule of assessments and tasks Each subject (n = 73) was assessed on two different days and a sub-sample of subject performed a third session. All sessions were approximately at the same hour of the day to control for the effect of circadian rhythms on muscle strength measures (Gauthier et al., 2001; Martin et al., 1999). Anthropometric measures and questionnaires were collected to document the general characteristics of the subjects as well as their percentage of body fat, using skinfold measures (Durnin and Womersley, 1974), and their physical activity level, using the questionnaire developed by (Baecke et al., 1982). The first session (session 1) was used to familiarize the subject with maximal voluntary contractions (MVC), which were be used in session 2 to define our criterion of Strength. This familiarization was necessary to control for the confounding learning effect Table 1 Demographic characteristics and strength of males and females (n = 73) who volunteered for the validation study Variable

Age (yr) Mass (kg) Height (m) Strength (Nm) 90 Nm/Strength

Males (n = 44)

Females (n = 29)

Mean

(SD)

Mean

(SD)

33 75 1.74 343 0.28

(10) (9) (0.06) (75) (0.07)

35 61 1.63 208 0.45

(11) (8) (0.05) (36) (0.08)

Pa

0.430 0.000 0.000 0.000 0.000

Strength: peak value of the maximal voluntary contractions (session 2) in Nm (L5/S1 extension moment); 90 Nm/Strength: ratio to determine the relative load induced by the 90 Nm load during the fatigue task. a The statistical test (t-test, Aspin-Welch, Wilcoxon) was determined after verification of normality and equality of variance assumptions.

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

1009

Table 2 Demographic characteristics and strength of males and females (n = 30) who volunteered for the reliability study

Port Washington NY 11050, USA) to ensure that the maximal value (220 minus age) of each subject was not exceeded.

Variable

2.3. Measurement techniques

Age (yr) Mass (kg) Height (m)

Males (n = 19)

Females (n = 11)

Mean

(SD)

Mean

(SD)

29 75 1.76

(9) (7) (0.05)

41 63 1.63

(10) (8) (0.05)

t-Test P

0.003 0.000 0.000

in the generation of a MVC (Lariviere et al., 2003b; Newton and Waddell, 1993). The increase of back muscle strength due to motor learning should be stabilized by the second session (Graves et al., 1990; Lariviere et al., 2003b). The dataset of the second session was used for the validation study. A sub-sample of 30 subjects (n = 19 males + 11 females, see Table 2) came for a third session two weeks later so that the datasets of sessions 1 and 3 (at least 2.5 weeks apart) could be used to assess reliability. The reliability was assessed with the dataset of session 1 to have subjects who are not familiarized to the FET, which should represent clinical practice. Sessions 2 and 3 were two weeks apart to ensure that the two days of evaluation for the reliability study (sessions 1 and 3) are sufficiently far apart to eliminate the initial task learning effect (Kroll, 1963) and possible muscle soreness. This should also represent clinical practice where pre- and postintervention evaluations are at least four weeks apart. In each session, three MVCs separated by a minimum of 2 min rest were followed, after 5 min rest, by the FET. Back strength (Strength) was defined as the peak L5/S1 extension moment among the three MVCs of session 2. The FET consisted in repeated 8-s cycles subdivided into three segments: 1.5 s of progressive rise to reach a 90 Nm absolute torque (L5/S1 extension moment), 5 s to sustain this torque (plateau), and 1.5 s of rest (Fig. 1d). The FET was performed during 10 min in sessions 1 and 3 (reliability study) whereas in session 2 (validation study), it was performed until exhaustion, or until the 60 min time-limit was reached, to determine the mechanical criterion of fatigue (time to exhaustion or Tend = 8 s · number of cycles). In each session, the subjects were given strong verbal encouragement during MVCs as well as during the FET to delay exhaustion (session 2), defined as the inability to generate or maintain the 90 Nm plateau for three consecutive cycles. Heart rate was monitored (Polar Electro Inc.,

The experimental setup is depicted in Fig. 1. The subject was positioned in a trunk dynamometer in a semi-seated position to minimize the contribution of the hip extensors and to avoid fatigue in the lower limbs (Fig. 1b). This dynamometer consists of a triaxial force platform (Advanced Mechanical Technology Incorporated, model MC6-6-1000) mounted on a steel frame that allows stabilization of the feet, knees and pelvis (details in Lariviere et al., 2001). Trunk extension was generated against a padded bar fixed on the surface of the force platform and adjusted at the T4 level. During each extension effort, the extension moment at L5/S1 was displayed in real time as visual feedback on a monitor positioned in front of the subject (Fig. 1c). The visual feedback consisted of a vertically moving square target with lower and upper bounds corresponding to a tolerance limit of ±10% of the prescribed extension moment, and lateral bounds corresponding to a tolerance limit of 4.5 Nm (value at 40% MVC, Lariviere et al., 2001) to control for unwanted axial rotation moments. This target was designed to help the subject to respect the parameters of the functional fatigue test without generating asymmetric efforts (close to pure sagittal extension efforts). The torque data were collected at a sampling rate of 128 Hz. The EMG signals from four pairs of back muscles were collected (bandpass filter: 20–450 Hz; preamplification gain: 1000; sampling rate: 2048 Hz) with active surface electrodes (Model DE-2.3, DelSys Inc., Wellesley, MA). After the skin at the electrode sites was shaved and abraded with alcohol, the electrodes were positioned bilaterally over the multifidus at the L5 level (MU-L5, 3 cm from the midline of the back), iliocostalis lumborum at L3 (IL-L3, 5–6 cm from midline) and longissimus at L1 (LO-L1, 3 cm from midline) following the recommendations of (Defoa et al., 1989) with regards to muscle fiber direction (Fig. 1a). Two additional electrodes were positioned over the belly of the longissimus at the T10 level (LO-T10,  4–5 cm from midline). To summarize, electrode sites were determined not entirely, but mostly with the use of bony landmarks, thus without considering the position of innervation zones. A reference silver– silver chloride electrode was positioned over the T8 spinous

Fig. 1. Experimental setup (details in the methods – measurement techniques). (a) Position of the surface electrodes at the different vertebral levels; (b) The triaxial dynamometer; (c) Visual feedback displayed to the subject. For the purpose of this picture, the subject, represented by a small square, is positioned outside of the target; (d) Description of one work-rest cycle (8 s) of the functional endurance test (only 2 cycles are represented).

1010

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

process. For each subject, a template was produced during session 1, by copying electrodes locations along with natural skin blemishes on acetate. This help to ensure the same placement of the EMG surface electrodes in sessions 2 and 3. The difficulty in capturing the multifidus muscle with surface electrodes (Stokes et al., 2003) is acknowledged and therefore the validity of the electromyographic signal was assigned to the landmarked location rather than to the multifidus muscle itself. Given that the relative position of the targeted muscles relative to the anatomical landmarks may vary between subjects and that it is impossible to rule out the possibility of cross-talk, this way of reporting the results was adopted for all electrode sites. 2.4. Data processing The L5/S1 extension moments were first inspected visually to correctly identify the time-windows corresponding to the effort (plateau) portions of each cycle. Only the EMG signals [considering a 130-ms electromechanical delay, (van Dieen et al., 1991)] corresponding to these events were further processed to obtain the EMG indices described in the following paragraphs. From the EMG signals corresponding to each plateau (3–5 s of data), different EMG indices based on spectral analyses were computed to assess muscle fatigue. In addition to the usual short time fast Fourier transform (SFFT), time-frequency analyses of the EMG signals, such as wavelet analyses, were also investigated because they were shown to decrease the variability of spectral estimates (Karlsson et al., 1999). Temporal and spectral analyses were applied on successive time-windows (80% overlapped) of 500 ms (1024 points) to obtain the corresponding root mean square [RMS(t,c)] and median frequency [MF(t,c)] values that were afterward averaged to obtain one RMS and one MF value per cycle, RMS(c) and MF(c), where c represents the cycle number. Intensity patterns [WINTn(t,c), n representing the wavelet number] calculated from the wavelet transform were also performed using a set of 11 wavelets (Wn) adapted to physiological functions of the EMG signal (von Tscharner, 2000), a method that was implemented in Matlab (The MathWorks Inc., Natick, MA, release 13). Briefly, this method allows the computation of the time-series of the intensity in the EMG signal (analogous to the amplitude of the EMG signal) that represent the power of the EMG signal contained within the frequency band covered by each respective wavelet. The frequency band associated with each wavelet (Wn) [(W1): 2–12; (W2): 12–27; (W3): 27–49; (W4): 48–76; (W5): 75–110; (W6): 108–149; (W7): 147–194; (W8): 192–244; (W9): 242–301; (W10): 297–364; (W11): 359–432 Hz] can be obtained by calculating the FFT of the wavelet. Intensities WINT1(t,c) and WINT2(t,c) were eliminated because our surface electrodes reject frequencies below 20 Hz. WINT11(t,c) was also excluded because the power of the EMG signal corresponding to back muscles is negligible above 300 Hz (Dolan et al., 1995). Hereafter, the more generic term WINTn(t,c) will be used to represent WINT3(t,c) to WINT10(t,c), n representing the wavelet number that ranged from 3 to 10 and c the cycle. The term WINTn(t,c) is only for notation convenience, it refers to each wavelet intensity pattern taken separately and do not represent an averaged intensity value of several frequency band. The instantaneous mean frequency [IMNF(t,c)], that is the time-frequency counterpart of the MF (computed from SFFT), was also calculated from the entire set of wavelets (n = 11) according to another formulation (Karlsson and Gerdle, 2001). Finally, since there were several esti-

mates of the different variables [WINT3(t,c) to WINT10(t,c) and IMNF(t,c)] within each plateau, these estimates were time-averaged within each cycle to give one value per cycle [WINTn(c) and IMNF(c)]. No normalization method was applied on raw EMG signals nor wavelets transforms. Then, linear regression was applied to each time-series [RMS(c), MF(c), IMNF(c), WINTn(c) values across cycles] to get their slope (RMSslp, MFslp, IMNFslp, and WINTnslp) and intercept (RMSi, MFi, IMNFi, and WINTni) values. Finally, all the EMG fatigue indices (RMSslp, MFslp, IMNFslp, and WINTnslp) were divided by their corresponding intercept value (RMSi, MFi, IMNFi, and WINTni) and multiplied by 100 to get normalized indices (NRMSslp, NMFslp, NIMNFslp, and NWINTnslp). All EMG fatigue indices (NRMSslp, NMFslp, NIMNFslp, NWINTnslp) as well as RMSi were averaged bilaterally to reduce data and to increase reliability (Larivie`re et al., 2002). Global EMG indices across all muscles were also calculated such as their average as well as their extreme values because EMG indices derived from several muscles usually show more valid and reliable results (Larivie`re et al., 2002; Mannion and Dolan, 1994; van Dieen et al., 1998). The global EMG indices were computed for 8 muscles, 6 muscles (T10 electrodes eliminated) and 4 muscles (T10 and L3 electrodes eliminated) to see if reducing the number of sampled muscles, to simplify the whole procedure, would lead to equivalent validity and reliability results. Electrodes at T10 and L3 were eliminated based on past (Larivie`re et al., 2002) and present reliability findings. To have an estimate of the level of activation at the different electrodes sites, RMSi amplitude values (a fatigue-free index of muscle activation) were divided by the maximal activation level (EMGmax) and multiplied by 100 to get a normalized value (NRMSi). EMGmax was the maximal RMS value computed across the successive (80% overlapped) 500 ms time-windows obtained from the three MVCs.

2.5. Statistics How many cycles are necessary to predict endurance without requiring a too long endurance test? In the protocol (sessions 1 and 3), it was decided to ask to all subjects to perform at least 75 cycles (10 min.), which was defined as an upper limit in a practical point of view. It must now be determined whether it is possible to reduce further the duration of the FET. In fact, the duration of the test has to optimally meet three conditions: (1) most of the subjects do not reach exhaustion before the end of the test, (2) enough EMG data (number of cycles) is collected to have enough data points to produce stable regression lines in terms of intercept and slope (parameters used to calculate EMG indices) and (3) the duration is long enough to predict endurance time. With reference to the third condition, non-linear IMNF time-series were observed in some subjects (Fig. 3, left plots) so that the corresponding slope index (NIMNFslp) regressed to zero as more EMG was retained to compute it (Fig. 2). Therefore, to help determine an appropriate duration to compute EMG indices, NIMNFslp was computed from the EMG signals (electrodes at L5 and L1 bilaterally) corresponding to different portions of the entire endurance test (adding 10% Tend of EMG data at a time). Using an ANOVA for repeated measures (n = 60, 10 repeated measures), the test duration, where NIMNFslp would not be different from the best criterion at hand (NIMNFslp computed from the entire signal) was found. Likewise, Pearson correlations were also

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

1011

Fig. 2. NIMNFslp calculated from the EMG signals collected during different proportions (percent of Tend) of the entire endurance test for four electrode sites (upper plot) and the corresponding Pearson’s correlation with Tend (lower plot). Standard deviations are not presented for clarity. Note that even though NIMNFslp non-transformed values are presented, transformed values were used to obtain normal distributions (for NIMNFslp and Tend) and run parametric statistical analyses (ANOVA results and Pearson correlations; please see Section 2.5).

Fig. 3. Time courses of IMNF spectral estimates (EMG collected with the left-L5 electrode) of the 60 subjects who reached exhaustion before 60 min. To clarify the general behavior of each curve, a regression polynomial (degree 3) was fitted to the time series but the EMG indices were computed from the original values (without polynomial). Plots at your left display the entire time-series until exhaustion whereas plots on your right displays the data used (5 min for females and 10 min for males) for linear regression analyses and to obtain intercept (IMNFi) and slope (IMNFslp) indices.

carried out between NIMNFslp computed at each time-interval and NIMNFslp computed from the entire test. Criterion validity. To assess validity, Pearson correlation coefficients were computed between each EMG index and our gold standard measures of strength (Strength) and fatigue (Tend). Based on the Wilk–Shapiro test, several variables were trans-

formed to get a normal distribution. These transformations were necessary to carry out parametric statistics such as Pearson correlations but also the multiple linear regression analyses that will be presented in a subsequent paper. Tend values required a logarithmic transformation whereas all NWINTnslp, NMFslp and NIMNFslp indices required an inverse hyperbolic sine transfor-

1012

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

Table 3 Physical, demographic and strength characteristics of males who reached exhaustion before the 60-min limit (Tend < 60 min) and males that reached the 60-min limit (Tend = 60 min) Variable

Age (yr) Mass (kg) Height (m) Body fat (%) FFM (kg) BMI Level physical activity Strength (Nm) 90 Nm/Strength

Tend <60 min (n = 32)

Males (n = 12)

t-Test P

Mean

(SD)

Mean

(SD)

35 75 1.73 19.8 60 24.9 8.0 327 0.28

(10) (9) (0.06) (5.9) (7) (2.7) (1.6) (64) (0.07)

27 75 1.76 17.9 62 24.3 9.6 387 0.24

(9) (7) (0.05) (6.1) (5) (2.4) (1.2) (87) (0.06)

0.014a 0.873 0.194 0.361 0.460 0.513 0.003 0.017 0.031a

a

In some cases, a Mann–Whitney test was used because the normality assumption was rejected. FFM: fat free mass; BMI: body mass index; Strength: peak value of the maximal voluntary contractions (session 2) in Nm (L5/S1 extension moment); 90 Nm/Strength: ratio to determine the relative load induced by the 90 Nm load during the fatigue task.

mation. No transformation was efficient to correct for the abnormal distribution of NRMSslp so Spearman correlations were applied for this variable. Test-retest reliability. To address the between-day reliability, the datasets from sessions 1 and 3 (10 min fatigue task at 90 Nm at least 2.5 weeks apart) were used. The intra-class correlation coefficients (ICC, type 2,1) and the standard error of measurement (SEM) corresponding to all EMG indices computed from the first 5 min and from the whole FET (10 min) were calculated. ICCs and SEMs were calculated using the different sources of variances (Subject, Day, Subject · Day) computed from a oneway ANOVA with repeated measures on the Day factor. The following equations were used: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r2S ICC ¼ 2 and SEM ¼ r2D þ r2SD ; rS þ r2D þ r2SD where r2S , r2D and r2SD are Subject, Day and Subject · Day variances, respectively. ICCs were interpreted as follows: <0.40 – poor, 0.40– 0.75 – moderate, and >0.75 – excellent (Fleiss, 1986). The SEM was expressed as a percentage of the grand mean (across days). To help in the interpretation of the different results (criterion validity and reliability of EMG indices computed from different electrode sites), a one-way ANOVA was applied (4 levels: L5, L3, L1 and T10) on NIMNFslp and NRMSi whereas its non-parametric equivalent (Kruskal–Wallis one-way ANOVA) was used for non-normal distributed NRMSslp values.

3. Results 3.1. Overall description of subjects performance Among the 73 volunteers who performed the fatigue task until exhaustion in session 2, 13 subjects (12 males and 1 female) have reached the upper limit of 60 min without being exhausted. The female (age: 35 yrs) was relatively heavy and tall (mass: 73 kg; height: 1.70 m) and was the strongest of the group (275 Nm versus 153–262 Nm for the others). Considering that this subgroup was mainly composed of males (n = 12), they were compared to the other males (n = 32) who did reach exhaustion before

60 min (Table 3). The physical characteristics (mass, height, percent body fat, fat free mass and BMI) of both groups were comparable. However, the males who reached the 60-min limit were significantly younger, more physically active and stronger (higher MVCs) than the other males. The relative load sustained by these stronger males during the 90 Nm FET was easier for them because of their lower 90 Nm/Strength ratio (Table 3). Considering that our mechanical criterion of fatigue (Tend) required the subject to reach exhaustion, the Pearson correlations with the different EMG indices (reported in the next sections) were computed only for the 60 subjects meeting this criterion, thus excluding the 13 subjects not exhausted after 60 min. 3.2. How many cycles are required to compute EMG indices without requiring a too long endurance test? It was first observed that within the 60 subjects that reached exhaustion, the Tend values were significantly longer for males (22 ± 15 min; median: 17 min; n = 32) than for females (14 ± 11 min; median: 10 min; n = 28). It was thus appropriate to compute the various EMG indices for different durations of the endurance task to respect this gender difference. The ANOVAs that were carried out to address this question were significant (P = 0.000) for all the muscles, revealing that the number of cycles considered to compute EMG indices is of importance. Pairwise comparisons (Tukey-Kramer) showed that NIMNFslp values were no more significantly different than the criterion value (NIMNFslp computed from 100% of the signal) when reaching 40% and 30% of Tend for the left and right L5 electrodes, respectively, whereas it required 60 and 50% of Tend for the left and right L1 electrodes. It was thus decided that about 50% of Tend is required to compute NIMNFslp, on average. These results concur with previous findings on back muscles (van Dieen et al., 1998) where it was concluded that the duration of the task should be at least 50% of Tend to be able to predict Tend with such EMG indices of fatigue. Furthermore, Pearson correlations

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

(n = 60 subjects) were computed between NIMNFslp (at each 10% Tend) and Tend and it was observed that the correlation results leveled off approximately at 50–60% of Tend (Fig. 2 – lower plot), which represents a good compromise to predict endurance without requiring a too long endurance test. Consequently, considering the Tend results of males and females reported above, it was decided, as mentioned in the methods section, to compute the EMG indices from the first 5 min. for females and from the first 10 min. for males. Overall, six males out of 32 and seven females out of 28 reached exhaustion before 5 min. and 10 min, respectively.

3.3. Criterion validity and reliability of EMG indices of back muscle fatigue The EMG fatigue indices based on EMG amplitude values (NRMSslp) showed low non-significant correlations with Tend (electrodes at L5, L3 and L1) except for electrodes at T10 either with the first 5 min (r = 0.45;

1013

P = 0.000) or 10 min of EMG data (r = 0.44; P = 0.000). However, the reliability results were poor with maximum ICCs of 0.58 (again at T10) and all SEMs above 198%. Fig. 3 depicts the time-series of IMNF spectral estimates (EMG collected with the left-L5 electrode) of the 60 subjects who reached exhaustion before 60 min. It can be observed from the right plots of this figure (5 and 10 min of data used to compute EMG indices in females and males, respectively) that the time-series were slightly nonlinear for many subjects. For the EMG indices based on spectral analyses, as expected the results demonstrated that the position of the electrodes, that is to say the muscles investigated, was of importance. Correlations with Tend were always smaller (0.05–0.10 lower) with L3 and T10 electrodes than with L5 and L1 electrodes and the reliability results were also better for EMG indices for L5 and L1 electrodes (results not shown). The most promising EMG frequency-domain indices of muscle fatigue (with electrodes at L5 and L1, without wavelets 5–6–7) are identified in Table 4.

Table 4 Validity and reliability results of the EMG spectral indices computed from the first 5 min (females) and 10 min (males) of the fatigue task EMG indices

Musclesa

Median (min, max) (n = 73)b

Criterion validity: Pearson rc

Reliability (n = 30)

Tend (n = 60)

ICC

Strength (n = 73)

SEM (%)

NMFslp (short fast Fourier transform)

X-L5 X-L1 X4 M4

1.4 0.9 1.2 1.9

(15.6, (11.0, (13.3, (17.7,

5.9) 2.2) 2.0) 0.2)

0.60 0.61 0.66 0.67

0.33 0.45 0.39 0.53

0.68 0.57 0.70 0.54

67 77 61 64

NIMNFslp (wavelet transform)

X-L5 X-L1 X4 M4

1.2 1.0 1.1 1.7

(14.4, (11.3, (12.9, (16.6,

2.4) 1.0) 0.3) 0.2)

0.66 0.64 0.69 0.68

0.39 0.49 0.46 0.55

0.75 0.81 0.83 0.78

63 47 46 42

NWINT3slp (27–49 Hz)

X-L5 X-L1 X4 M4

3.6 3.4 3.4 6.9

(4.7, 103.3) (2.0, 53.8) (2.3, 69.5) (0.7, 169.1)

0.44 0.37 0.49 0.52

0.16 0.19 0.20 0.26

0.67 0.65 0.67 0.53

151 121 135 163

NWINT4slp (48–76 Hz)

X-L5 X-L1 X4 M4

2.5 1.6 2.2 5.2

(10.7, 48.0) (5.0, 43.2) (5.6, 38.0) (3.3, 69.6)

0.30 0.24 0.33 0.37

0.06 0.09 0.11 0.18

0.69 0.63 0.69 0.77

95 119 89 51

NWINT8slp (192–244 Hz)

X-L5 X-L1 X4 M4

2.6 2.0 2.2 4.1

(18.6, (26.4, (22.5, (27.2,

6.8) 45.0) 25.9) 5.7)

0.42 0.43 0.43 0.47

0.31 0.35 0.33 0.38

0.39 0.53 0.41 0.64

226 129 158 61

NWINT9slp (242–301 Hz)

X-L5 X-L1 X4 M4

3.3 2.4 3.0 5.2

(20.6, (27.6, (24.1, (30.5,

8.3) 35.5) 20.2) 2.6)

0.42 0.48 0.47 0.50

0.29 0.39 0.36 0.41

0.41 0.40 0.34 0.55

190 133 146 71

NWINT10slp (297–364 Hz)

X-L5 X-L1 X4 M4

3.8 2.9 3.9 5.6

(21.9, (28.3, (24.3, (32.9,

7.6) 25.2) 14.9) 2.6)

0.46 0.47 0.49 0.50

0.32 0.39 0.36 0.41

0.54 0.62 0.57 0.66

113 80 86 51

a

X and M: mean and maximal value computed from 4 electrodes (L5 and L1 bilaterally). Considering that several indices showed non normal distribution, the median and extreme values (minimum, maximum) were reported to better describe the data. However, the reader is reminded that before computing Pearson correlation coefficients, these indices (as well as Tend) were transformed (inverse hyperbolic sine) to obtain a normal distribution. c Statistically significant Pearson correlation coefficients (p 6 0.05) are identified by bold characters. b

1014

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

Fig. 4. NRMSi (upper plot), NIMNFslp (middle plot) and NRMSslp (lower plot) averaged values for each muscle pair (non-transformed values). Thick horizontal bars identify pairwise significant differences between muscles (post hoc analyses). Note that even though non-transformed values are presented, ANOVA results on NIMNFslp are based on transformed values to obtain a normal distribution (see Section 2).

EMG indices based on wavelets 5, 6 and 7 (NWINT5slp, NWINT6slp, NWINT7slp), that cover the middle frequency bands of the EMG signals (75–194 Hz), were not significantly correlated to Tend and were thus eliminated (results not shown). Among the EMG indices of Table 4, NWINT3p and NWINT4p showed low to moderate correlation coefficients with Tend (r range: 0.24 to -0.52; P < 0.05) and no correlation at all with Strength. Better results were obtained for the other EMG indices with significant correlation coefficients ranging between 0.42 and 0.69 (P < 0.000) with Tend and ranging between 0.29 and 0.55 with Strength. Clearly, the best correlation results (with Tend or Strength) were obtained with EMG indices computed from the entire power spectrum (NMFslp, NIMNFslp), in comparison with EMG indices computed from individual wavelets (NWINT3slp, NWINT4slp, NWINT8slp, NWINT9slp, NWINT10slp). The reliability results showed quite different trends, with the highest ICCs (range: 0.53–0.83) for NWINT3slp and NIMNFslp. Interestingly, much better reliability results, in terms of ICC and SEM values, were observed for NIMNFslp (wavelets transform) than NMFslp (SFFT). Overall, considering both the correlation and reliability results, NIMNFslp was the best EMG index. 3.4. Neuromuscular activation level of back muscles The analysis of NRMSi amplitude values revealed a significantly (P = 0.000) lower activation level at L3 and T10 than at L5 and L1 (Fig. 4 – upper plot). This partially reflect the behavior of NIMNFslp because muscles at L5 showed more negative slopes (more fatigue) than at L3 and T10 (Fig. 4 – middle plot). Moreover, the Pearson correlation between these two EMG indices (activation level

and NIMNFslp pooled across the four electrode levels) was significant (r = 0.40; P = 0.000). Interestingly, at the opposite, NRMSslp showed higher values at T10 than at the three other spinal levels (Fig. 4 – lower plot). 4. Discussion The present study was designed to address different questions, the main idea being the development of a submaximal endurance test, combining dynamometry and surface EMG, to predict back muscle capacity. First of all, it was necessary to know how long the FET should be to allow EMG indices to predict Tend. The results led to the conclusion that the test duration should be 5 and 10 min for females and males, respectively. The second finding was that these EMG indices were more reliable when computed in the time-frequency domain (NIMNFslp computed from wavelets transform) than when computed in the frequency domain (NMFslp computed from FFT), but showed comparable criterion validity results. NIMNFslp showed moderate to good correlation (range: 0.64–0.69) with Tend, lower correlations (range: 0.39–0.55) with Strength, and good to excellent reliability results (ICC range: 0.75–0.83). It was also found that measures of spectral content in different frequency bands (NWINTnslp indices) were less valid and reliable than indices computed from the entire power spectrum (NMFslp or NIMNFslp). Finally, specific neuromuscular activation patterns seem to occur during the endurance test and must be taken into account in the interpretation of the various EMG fatigue indices. 4.1. SFFT vs wavelets EMG indices of muscle fatigue based on time-frequency analyses were similarly correlated with Tend or Strength

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

than indices based on the usual SFFT. Interestingly, the use of time-frequency analyses led to more reliable fatigue indices, thus supporting previous work (Karlsson et al., 1999) showing lower variability in spectral estimates using wavelets and this, even during static contractions in which the EMG signal is considered quasi-stationary. However, torque variations increased as fatigue progressed, as revealed by the steadiness results (ST–SDslp) in the part-II article. Therefore, the stationarity of the EMG signals might have been affected, thus justifying the use of wavelets. 4.2. Global vs local analyses of the EMG spectral content In general, NIMNFslp, computed from the entire power spectrum, was better correlated with Tend or Strength and was more reliable (considering ICCs and SEMs jointly) than the selected EMG indices computed from individual wavelets (NWINT3slp, NWINT4slp, NWINT8slp, NWINT9slp, NWINT10slp). From a reliability perspective, these results were not surprising because the random variations that is usually present in EMG signals and that should be reflected in specific bands of the power spectrum are expected to have more influence on NWINTnslp indices than on the more global NIMNFslp index. The same phenomenon would also affect correlation results because random variations decrease correlations. However, these results only apply to indices of muscle fatigue which are essentially a measure of the trend of the data over time (NWINTn/time slope). The analysis of the power content in the different wavelets is particularly efficient to reveal specific fiber-type recruitment patterns during much smaller time-intervals of dynamic muscle contractions (von Tscharner and Goepfert, 2003; Wakeling et al., 2002). 4.3. Criterion validity of EMG fatigue indices Before discussing the criterion validity results, the large inter-subject variations in Tend and Strength values as well as in the EMG fatigue indices (depicted in Fig. 3) has to be discussed. In fact, this variability probably reflects the ability of the FET to highlight the differences in back muscle capacity between individuals, which depends on the strength and endurance of the muscles, but maybe more importantly on the capacity of the muscles to take advantage of intermittent contractions to recover from cycle to cycle. On the other hand, such variable results might also compromise the diagnosis value of the test in CLBP patients. Comparisons with CLBP patients have to be carried out to answer this question and to substantiate the construct validity of the FET. In fact, the FET may highlight muscle impairments of CLBP patients especially if they are weaker because of muscle atrophy (Danneels et al., 2000), are more fatigable because of altered back muscle composition (Mannion et al., 1997b) and/or have lower capacity to produce ATP via aerobic energy pathways (not substantiated in the literature).

1015

Correlations between NIMNFslp and Tend ranged between 0.64 and 0.69. These values are in the mid-range of values obtained, also for back muscles, during static continuous contractions [range: 0.52 to 0.85: (Dedering et al., 1999; Kankaanpaa et al., 1998a; Mannion and Dolan, 1994; Mannion et al., 1997a)] and flexion-extension dynamic contractions [range: 0.50 to 0.89: (Kankaanpaa et al., 1997, 1998b; Sparto et al., 1999)]. Considering that the subjects were not specifically trained to fatigue back muscles to exhaustion and, consequently, that the fatigue criterion (Tend) has probably been affected, the correlations obtained were considered as more than satisfactory. NIMNFslp was also positively correlated with Strength, that is to say that stronger subjects were more fatigue resistant (NIMNFslp closer to zero, thus less negative slope) during the FET. Effectively, stronger subjects perform the FET at a lower relative load thus recruiting proportionally more fatigue-resistant type-I muscle fibers. The finding that NIMNFslp was more correlated with Tend than with Strength might be attributed to the sensitivity of EMG spectral indices of fatigue to additional determinants of muscle fatigue such as muscle perfusion (Merletti et al., 1984), which was probably reflected in Tend, especially during intermittent contractions (Christensen and Fuglsang-Frederiksen, 1988). 4.4. Reliability of EMG spectral indices The lower between-day reliability results obtained from EMG indices corresponding to L3 and T10 electrodes concur with previous findings showing lower reliability of EMG fatigue indices computed for more laterally located back muscles (Larivie`re et al., 2002; Ng and Richardson, 1996). This might be explained by the fact that lateral muscles are not solicited as consistently as medial muscles during back extension efforts (Sparto et al., 1997; van Dieen et al., 1998). Hence, subtle efforts are observed in the frontal and transverse planes during extension tasks (Parnianpour et al., 1991; Tan et al., 1993) and they are preferentially produced by the lateral muscles (Jonsson, 1970) because of their mechanical advantage relative to medial muscles. Moreover, in the present study, the more laterally located back muscles (electrodes L3 and T10) were shown to be significantly less activated than the medially located muscles (electrodes at L5 and L1). Given that a rather weak (r = 0.40) correlation was obtained between the activation level and muscle fatigue and that NIMNFslp were significantly lower (more fatigue) at L5, it is probable that muscles at L3 and T10 were not enough fatigued to get reproducible results. Consequently, their criterion validity may have been impaired relative to electrodes located at L5 and L1. As long as electrodes at L5 and L1 are concerned, the reliability of NMFslp indices (SFFT) was lower but the reliability of NIMNFslp indices was similar to our previous work (Larivie`re et al., 2002). The same measurement protocol (electrode position, computation of the NMFslp index) was used but the task was different (30 s continuous static contraction at 75% MVC). More specifically, in comparison

1016

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

with Larivie`re et al. (2002), the reliability of NIMNFslp indices (for electrodes at L5 and L1) was similar in terms of ICCs (current study: 0.75–0.83; previous study: 0.69–0.82) but to some extent worse in terms of SEMs (current study: 44–63%; previous study: 21–34%). The good to excellent ICC values indicate that these variables are well suited to distinguish between subjects (diagnosis). On the other hand, the somewhat more disappointing SEMs, which represent the precision of the measurement (experimental noise), point to a poor capacity to detect changes over time (ex: effect of a rehabilitation program). However, the larger SEMs can be partly explained by the lower task intensity used in the current study (intermittent contractions at a lower relative load), which led to NIMNFslp slope values closer to zero and artificially inflated the relative SEM values. Unfortunately, no reliability results have been reported for EMG fatigue indices during intermittent contractions. 4.5. Neuromuscular activation level of back muscles during the endurance test NRMSslp showed not only poor reliability results, as observed in previous studies (Larivie`re et al., 2002; Nargol et al., 1999), but also showed low correlations with Tend.Interestingly, the best correlation results were observed for electrodes at T10 (r = 0.44). Still, the NRMSslp index is not a good candidate to predict Tend because specific neuromuscular activation patterns appears to intervene, adding variability to this index, as discussed in the next paragraph. NRMSi results showed that more laterally located back muscles (electrodes L3 and T10) were significantly less activated than the medially located muscles (electrodes at L5 and L1). This concurs with previous findings on back muscles using surface EMG (Vink et al., 1988) and muscle functional magnetic resonance imaging (Mayer et al., 2005). In these studies it was observed that laterally located back muscles are less activated than medially located muscles but only at low to moderate force levels. Effectively, the opposite phenomenon occurs at higher force levels (Mayer et al., 2005; Vink et al., 1988). Consequently, NRMSi results suggest that medial back muscles should fatigue faster, which was partly substantiated with (1) NIMNFslp because muscles at L5 fatigued faster than lateral muscles (electrodes at L3 and T10) and (2) a statistically significant correlation (r = 0.40) was obtained between the muscle activation level (NRMSi) and muscle fatigue (NIMNFslp). Furthermore, Vink’s and Mayer’s results would also predict that if lumbar spinal muscles decrease their capacity to produce force (fatigue), more laterally located back muscles would compensate by increasing their relative mechanical contribution as fatigue progress. This is partly what was observed with higher NRMSslp values at the T10-electrode than for the three other electrode positions. This is also in accordance with other studies showing that thoracic erector spinae (equivalent to EMG collected at T10) appear to compensate for

the presumed decrease of lumbar stability in chronic low back pain subjects (van Dieen et al., 2003). Interestingly, the relative contribution of the thoracic erector spinae might also be increased relative to the lumbar erector spinae, possibly because of their mechanical advantage (longer lever arms), to help in the support of the lumbar moment while decreasing the compressive cost on the lumbar spine (Potvin et al., 1991). Clearly, further research is needed to better understand what is the role of medial vs lateral and of lumbar vs thoracic back muscles during different tasks challenging the mechanical stability as well as the compression and shear loading of the spinal column. It appears that RMS-based EMG indices are too much influenced by variable neuromuscular patterns as fatigue progress. This explains why the NRMSslp index is not well suited to document muscle fatigue. In the part-II article (Larivie`re et al., 2008), more relevant EMG amplitude indices based on the variations of neuromuscular activation patterns between muscle synergists are proposed. The main limitation of the present study is that the results (validity and reliability of EMG fatigue indices) cannot be generalized to CLBP patients. However, testing criterion validity would represent a problem in this population because CLBP patients are known to be reluctant to produce maximal performances (Verbunt et al., 2003). Unfortunately, maximal performances are required to obtain the muscle fatigue (time to exhaustion) and muscle strength (maximal voluntary contractions) criteria. However, other elements of validity (construct and predictive validity) could be tested in CLBP patients in the future. According to previous findings (Larivie`re et al., 2002), the reliability of the EMG fatigue indices should be similar in CLBP patients. However, the fact that the fatigue task was quite different than the FET in that study (30 s continuous contraction at 75% MVC) may also produce some variations in the results. To conclude, the present findings support the use of the FET, an intermittent and time-limited absolute endurance test, to predict the capacity of back muscles using surface EMG. One advantage of the FET is to eliminate the use of MVCs. The EMG indices assessed in the present study are conventionally used to assess fatigue and were consequently more correlated to absolute endurance than to strength. However, in the part-II article (Larivie`re et al., 2008), other types of EMG indices are much more correlated to strength. Acknowledgements The present research project was funded by the Occupational Health and Safety Research Institute Robert-Sauve´ (IRSST) of Quebec (Canada) and the Canadian Institutes of Health Research (CIHR). Special thanks to Nathaly Gaudreault and Rubens Alexandre Da Silva Junior (research assistants) who provided high quality data even though the procedures were numerous and very demanding for them and the subjects.

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

References Baecke JAH, Burema J, Frijters JER. A short questionnaire for the measurement of habitual physical activity in epidemiological studies. Am J Clin Nutr 1982;36:936–42. Biedermann HJ, Shanks GL, Forrest WJ, Inglis J. Power spectrum analyses of electromyographic activity. Discriminators in the differential assessment of patients with chronic low-back pain. Spine 1991;16:1179–84. Biering-Sorensen F. Physical measurements as risk indicators for low-back trouble over a one-year period. Spine 1984;9:106–19. Bjorksten M, Jonsson B. Endurance limit of force in long-term intermittent static contractions. Scand J Work Environ Health 1977;3:23–7. Capodaglio P, Nilsson J, Jurisic DH. Changes in paravertebral EMG spectrum parallel to strength increases after rehabilitation in chronic low back pain patients. Clin Rehabil 1995;9:354–62. Christensen H, Fuglsang-Frederiksen A. Quantitative surface EMG during sustained and intermittent submaximal contractions. Electroencephalogr Clin Neurophysiol 1988;70:239–47. Christmass MA, Dawson B, Arthur PG. Effect of work and recovery duration on skeletal muscle oxygenation and fuel use during sustained intermittent exercise. Eur J Appl Physiol 1999a;80:436–47. Christmass MA, Dawson B, Passeretto P, Arthur PG. A comparison of skeletal muscle oxygenation and fuel use in sustained continuous and intermittent exercise. Eur J Appl Physiol 1999b;80:423–35. Danneels LA, Vanderstraeten GG, Cambier DC, Witvrouw EE, De Cuyper HJ. CT imaging of trunk muscles in chronic low back pain patients and healthy control subjects. Eur Spine J 2000;9:266–72. De Luca CJ. Use of the surface EMG signal for performance evaluation of back muscles. Muscle Nerve 1993;16:210–6. Dedering A, Ne´meth G, Harms-Ringdahl K. Correlation between electromyographic spectral changes and subjective assessment of lumbar muscle fatigue in subjects without pain from the lower back. Clin Biomech 1999;14:103–11. Defoa JL, Forrest W, Biedermann HJ. Muscle fibre direction of longissimus, iliocostalis and multifidus: landmark-derived reference line. J Anat 1989;163:243–7. Dempsey PG, Burdorf A, Webster BS. The influence of personal variables on work-related low-back disorders and implications for future research. J Occup Environ Med 1997;39:748–59. Dolan P, Mannion AF, Adams MA. Fatigue of the erector spinae muscles. A quantitative assessment using ‘‘frequency banding’’ of the surface electromyography signal. Spine 1995;20:149–59. Durnin JVGA, Womersley J. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16 to 72 years. Brit J Nutr 1974;32:77–97. Elfving B, Dedering A, Nemeth G. Lumbar muscle fatigue and recovery in patients with long-term low-back trouble-electromyography and health-related factors. Clin Biomech 2003;18:619–30. Enoka RM, Stuart DG. Neurobiology of muscle fatigue. J Appl Physiol 1992;72:1631–48. Enthoven P, Skargren E, Kjellman G, Oberg B. Course of back pain in primary care: a prospective study of physical measures. J Rehabil Med 2003;35:168–73. Fleiss RL. The design and analysis of clinical experiments. New York: John Wiley and Sons; 1986. Garrow JS, Webster J. Quetelet’s index (W/H2) as a measure of fatness. Int J Obesity 1985;9:147–53. Gauthier A, Davenne D, Martin A, Van Hoecke J. Time of day effects on isometric and isokinetic torque developed during elbow flexion in humans. Eur J Appl Physiol 2001;84:249–52. Graves JE, Pollock ML, Carpenter DM, Leggett SH, Jones A, MacMillan M, Fulton M. Quantitative assessment of full range-of-motion isometric lumbar extension strength. Spine 1990;15:289–94. Hagberg M. Work load and fatigue in repetitive arm elevations. Ergonomics 1981;24:543–55.

1017

Hides JA, Richardson CA, Jull GA. Multifidus muscle recovery is not automatic after resolution of acute, first-episode low back pain. Spine 1996;21:2763–9. Hides JA, Stokes MJ, Saide M, Jull GA, Cooper DH. Evidence of lumbar multifidus muscle wasting ipsilateral to symptoms in patients with acute/subacute low back pain. Spine 1994;19:165–72. Hultman G, Nordin M, Saraste H, Ohlsen H. Body composition, endurance, strength, cross-sectional area, and density of MM erector spinae in men with and without low back pain. J Spinal Disord 1993;6:114–23. Jonsson B. The functions of individual muscles in the lumbar part of the erector spinae muscle. Electromyography 1970;10:5–21. Kankaanpaa M, Taimela S, Laaksonen D, Hanninen S, Airaksinen O. Back and hip extensor fatigability in chronic low back pain patients and controls. Arch Phys Med Rehabil 1998b;79:412–7. Kankaanpaa M, Taimela S, Webber CL, Airaksinen O, Hanninen O. Lumbar paraspinal muscle fatigability in repetitive isoinertial loading: EMG spectral indices, Borg scale and endurance time. Eur J Appl Physiol 1997;76:236–42. Kankaanpaa M, Laaksonen D, Taimela S, Kokko SM, Airaksinen O, Hanninen O. Age, sex, and body mass index as determinants of back and hip extensor fatigue in the isometric Sorensen back endurance test. Arch Phys Med Rehabil 1998a;79:1069–75. Karlsson S, Gerdle B. Mean frequency and signal amplitude of the surface EMG of the quadriceps muscles increase with increasing torque – a study using the continuous wavelet transform. J Electromyogr Kinesiol 2001;11:131–40. Karlsson S, Yu J, Akay M. Enhancement of spectral analysis of myoelectric signals during static contractions using wavelet methods. IEEE Trans Biomed Eng 1999;46:670–84. Kramer M, Ebert V, Kinzl L, Dehner C, Elbel M, Hartwig E. Surface electromyography of the paravertebral muscles in patients with chronic low back pain. Arch Phys Med Rehabil 2005;86:31–6. Kroll W. A reliable method of assessing isometric strength. Res Quart 1963;34:350–5. Larivie`re C, Gravel D, Gagnon D, Arsenault AB. The assessment of back muscle capacity using intermittent static contractions. Part II –Validity and reliability of biomechanical correlates of muscle fatigue. J Electromyogr Kinesiol 2008;18:1020–31. Larivie`re C, Arsenault AB, Gravel D, Gagnon D, Loisel P. Evaluation of measurement strategies to increase the reliability of EMG indices to assess back muscle fatigue and recovery. J Electromyogr Kinesiol 2002;12:135–46. Lariviere C, Arsenault AB, Gravel D, Gagnon D, Loisel P. Surface electromyography assessment of back muscle intrinsic properties. J Electromyogr Kinesiol 2003a;13:305–18. Lariviere C, Gravel D, Gagnon D, Arsenault AB, Loisel P, Lepage Y. Back strength cannot be predicted accurately from anthropometric measures in subjects with and without chronic low back pain. Clin Biomech 2003b;18:473–9. Lariviere C, Gagnon D, Gravel D, Arsenault AB, Dumas J, Goyette M, Loisel P. A triaxial dynamometer to monitor lateral bending and axial rotation moments during static trunk extension efforts. Clin Biomech 2001;16:80–3. Luoto S, Heliovaara M, Hurri H, Alaranta H. Static back endurance and the risk of low-back pain. Clin Biomech 1995;10:323–4. Maisetti O, Guevel A, Legros P, Hogrel JY. Prediction of endurance capacity of quadriceps muscles in humans using surface electromyogram spectrum analysis during submaximal voluntary isometric contractions. Eur J Appl Physiol 2002;87:509–19. Mannion AF. Fibre type characteristics and function of the human paraspinal muscles: normal values and changes in association with low back pain. J Electromyogr Kinesiol 1999;9:363–77. Mannion AF, Dolan P. Electromyographic median frequency changes during isometric contraction of the back extensors to fatigue. Spine 1994;19:1223–9. Mannion AF, Connolly B, Wood K, Dolan P. The use of surface EMG power spectral analysis in the evaluation of back muscle function. J Rehabil Res Develop 1997a;34:427–39.

1018

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019

Mannion AF, Weber BR, Dvorak J, Grob D, Muntener M. Fibre type characteristics of the lumbar paraspinal muscles in normal healthy subjects and in patients with low back pain. J Orthop Res 1997b;15:881–7. Mannion AF, Kaser L, Weber E, Rhyner A, Dvorak J, Muntener M. Influence of age and duration of symptoms on fibre type distribution and size of the back muscles in chronic low back pain patients. Eur Spine J 2000;9:273–81. Martin A, Carpentier A, Guissard N, Van Hoecke J, Duchateau J. Effect of time of day on force variation in a human muscle. Muscle Nerve 1999;22:1380–7. Mayer JM, Graves JE, Clark BC, Formikell M, Ploutz-Snyder LL. The use of magnetic resonance imaging to evaluate lumbar muscle activity during trunk extension exercise at varying intensities. Spine 2005;30:2556–63. Merletti R, Roy S. and mechanical manifestations of muscle fatigue in voluntary contractions. JOSPT 1996;24:342–53. Merletti R, Sabbahi MA, De Luca CJ. Median frequency of the myoelectric signal. Effects of muscle ischemia and cooling. Eur J Appl Physiol 1984;52:258–65. Moreau CE, Green BN, Johnson CD, Moreau SR. Isometric back extension endurance tests: a review of the literature. J Manip Physiol Ther 2001;24:110–22. Nargol AVF, Jones APC, Kelly PJ, Greenough CG. Factors in the reproducibility of electromyographic power spectrum analysis of lumbar paraspinal muscle fatigue. Spine 1999;24:883–8. Newton M, Waddell G. Trunk strength testing with iso-machines. Part 1: review of a decade of scientific evidence. Spine 1993;18:801–11. Ng J, Richardson CA. Reliability of electromyographic power spectral analysis of back muscle endurance in healthy subjects. Arch Phys Med Rehabil 1996;77:259–64. Nussbaum MA. Static and dynamic myoelectric measures of shoulder muscle fatigue during intermittent dynamic exertions of low to moderate intensity. Eur J Appl Physiol 2001;85:299–309. Oddsson LI, De Luca CJ. Activation imbalances in lumbar spine muscles in the presence of chronic low back pain. J Appl Physiol 2003;94:1410–20. Parnianpour M, Campello M, Sheikhzadeh A. The effect of posture on triaxial trunk strength in different directions: its biomechanical consideration with respect to incidence of low-back problem in construction industry. Int J Indust Ergonomics 1991;8: 279–87. Pitcher JB, Miles TS. Influence of muscle blood flow on fatigue during intermittent human hand-grip exercise and recovery. Clin Exp Pharmacol Physiol 1997;24:471–6. Potvin JR, Norman RW, McGill SM. Reduction in anterior shear forces on the L4/L5 disc by the lumbar musculature. Clin Biomech 1991;6:88–96. Rohmert W. Ermittlung von Erholungspausen fur statische Arbeit des Menschen. Int Z angew Physiol einschl Arbeitsphysiol 1960;18: 123–64. Roy SH, De Luca CJ, Casavant DA. Lumbar muscle fatigue and chronic lower back pain. Spine 1989;14:992–1001. Sparto PJ, Parnianpour M, Barria EA, Jagadeesh JM. Wavelet analysis of electromyography for back muscle fatigue detection during isokinetic constant-torque exertions. Spine 1999;24:1791–8. Sparto PJ, Parnianpour M, Reinsel TE, Simon S. Spectral and temporal responses of trunk extensor electromyography to an isometric endurance test. Spine 1997;22:418–26. Stokes IA, Henry SM, Single RM. Surface EMG electrodes do not accurately record from lumbar multifidus muscles. Clin Biomech 2003;18:9–13. Tan JC, Parnianpour M, Nordin M, Hofer H, Willems B. Isometric maximal and submaximal trunk extension at different flexed positions in standing. Triaxial torque output and EMG. Spine 1993;18: 2480–90.

Thomas S, Reading J, Shephard RJ. Revision of the Physical Activity Readiness Questionnaire (PAR-Q). Can J Sport Sci 1992;17:338–45. van Dieen JH, Cholewicki J, Radebold A. Trunk muscle recruitment patterns in patients with low back pain enhance the stability of the lumbar spine. Spine 2003;28:834–41. van Dieen JH, Heijblom P, Bunkens H. Extrapolation of time series of EMG power spectrum parameters in isometric endurance tests of trunk extensor muscles. J Electromyogr Kinesiol 1998;8:35–44. van Dieen JH, Thissen CEAM, van de Ven AJGM, Toussaint HM. The electro-mechanical delay of the erector spinae muscle: influence of rate of force development, fatigue and electrode location. Eur J Appl Physiol 1991;63:216–22. Verbunt JA, Seelen HA, Vlaeyen JW, van de Heijden GJ, Heuts PH, Pons K, Knottnerus JA. Disuse and deconditioning in chronic low back pain: concepts and hypotheses on contributing mechanisms. Eur J Pain 2003;7:9–21. Vink P, van der Velde EA, Verbout AJ. A functional subdivision of the lumbar extensor musculature. Recruitment patterns and force-RAEMG relationships under isometric conditions. Electromyogr Clin Neurophysiol 1988;27:517–25. Vlaeyen JWS, Linton SJ. Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art. Pain 2000;85:317–32. von Tscharner V. Intensity analysis in time-frequency space of surface myoelectric signals by wavelets of specified resolution. J Electromyogr Kinesiol 2000;10:433–45. von Tscharner V, Goepfert B. Gender dependent EMGs of runners resolved by time/frequency and principal pattern analysis. J Electromyogr Kinesiol 2003;13:253–72. Wakeling JM, Kaya M, Temple GK, Johnston IA, Herzog W. Determining patterns of motor recruitment during locomotion. J Exp Biol 2002;205:359–69. Westad C, Westgaard RH, De Luca CJ. Motor unit recruitment and derecruitment induced by brief increase in contraction amplitude of the human trapezius muscle. J Physiol 2003;552:645–56. Christian Larivie`re received a B.sc. in physical education, a M.sc. in Kinanthropology and a Ph.D. in clinical sciences from the university of Sherbrooke, Sherbrooke, Canada, in 1992, 1994 and 1999, respectively. from 1999 to 2000, he was a post-doctoral fellow in biomedical sciences at the university of Montreal, Quebec, Canada. He is currently researcher at the occupational health and safety research institute robert-sauve´, Montreal, Quebec, Canada. His research interest is focused on the quantification of lumbar impairments through kinematic, kinetic and electromyographic measures. Denis Gagnon received a B.Sc. in physical education (1980) and M.sc. in kinanthropology (1985) from University of Sherbrooke, and a Ph.D. in biomechanics in 1990 from university of Montreal. He is a professor at the department of Kinanthropology of the university of Sherbrooke and Director of the Occupational Biomechanics Laboratory. His research interests focus on the study of trunk muscle coactivity strategies during dynamic lifting and on the investigation of back muscle fatigue during static effort in healthy and low back pain individuals.

C. Larivie`re et al. / Journal of Electromyography and Kinesiology 18 (2008) 1006–1019 Denis Gravel is Professor at the school of rehabilitation of the University of montreal and researcher at the research center of the montreal rehabilitation institute. after he received his B. Sc. in Physical Therapy in 1970, he practiced physical therapy for two years. then, he completed his M.Sc. degree at the department of anatomy of the University of Montreal. from 1976 to 1983, he acted as a professor at the school of rehabilitation of the university of montreal. From 1984 to 1991, he completed his Ph.D. degree in neurobiology at Laval University. from 1992 to 1996, he was granted from the FRSQ (Fonds de Recherche en Sante´ du Que´bec) as a clinical researcher. His research interest focus on the evaluation of normal and pathologic motor function using electromyography, biomechanics and dynamometry techniques.

1019

A. Bertrand Arsenault received a B.Sc. (Physical Therapy), M.Sc. (Kinesiology) and Ph.D. (Kinesiology) from University of Montreal, Simon Fraser University and University of Waterloo respectively. He practiced as a therapist for several years before joining, in 1980, the School of Rehabilitation of the University of Montreal and the Research Centre of the Montreal Rehabilitation Institute (MRI). Since 1980, he has acted as a professor, director of graduate studies, director of the physical therapy program and director of that School at University of Montreal and as a researcher and director of the MRI Research Centre. He has been involved in research activities focussing on the evaluation of the musculoskeletal system of stroke patients as well as of subjects suffering from back and neck pain.