Assessment of the reliability of central and peripheral fatigue after sustained maximal voluntary contraction of the quadriceps muscle

ABSTRACT: The aim of the present study was to further confirm thevalidity of measurements for characterizing neuromuscular alterations byestablishing their reliability both before and after fatigue. Thirteen men (28 �5 years) volunteered to participate in two separate identical sessions requir-ing the performance of a sustained maximal voluntary contraction (MVC)with the quadriceps muscle for 2 min. MVC and transcutaneous electricalstimulations were used before and immediately after the fatiguing contrac-tion to investigate the reliability of MVC torque, central activation, andperipheral variables (M-wave properties, peak twitch, peak doublet) withinand between sessions. Based on previous and present results, we advisethe use of (1) voluntary activation level with potentiated doublet as a refer-ence to describe central fatigue, (2) electromyographic activity of vastuslateralis muscle as a surrogate for quadriceps for both voluntary and evokedcontraction, and (3) potentiated peak doublet amplitude to investigate con-tractile fatigue. These findings can be useful in the choice of the parametersdescribing central and peripheral fatigue of the quadriceps muscle in futurestudies.

Muscle Nerve 35: 486–495, 2007

ASSESSMENT OF THE RELIABILITY OF CENTRALAND PERIPHERAL FATIGUE AFTER SUSTAINEDMAXIMAL VOLUNTARY CONTRACTIONOF THE QUADRICEPS MUSCLE

NICOLAS PLACE, PhD,1 NICOLA A. MAFFIULETTI, PhD,2 ALAIN MARTIN, PhD,1

and ROMUALD LEPERS, PhD 1

1 INSERM ERM 207 Motricite Plasticite, Faculte des Sciences du Sport, Universite de Bourgogne,BP 27877, 21078 Dijon, France

2 Neuromuscular Research Laboratory, Schulthess Clinic, Zurich, Switzerland

Accepted 30 October 2006

Muscle fatigue can be defined as an exercise-induced reduction in maximal voluntary muscleforce.16 Although the underlying mechanisms havenot been fully described, it is accepted that centralfatigue (including alteration in processes locatedabove the neuromuscular junction) can be distin-guished from peripheral fatigue.17 The use of volun-tary contraction and noninvasive transcutaneouselectrical stimulation combined with surface electro-

myography allows exploration of the sites of neuro-muscular fatigue in humans.8,15,16,27,33

Data from recent studies suggest that voluntaryactivation level with reference to the potentiateddoublet (VALp)4 is the most valid method based onthe twitch interpolation technique26 to assess thelevel of muscle activation14 and therefore centralfatigue.10 However, other authors have used the cen-tral activation ratio (CAR)20,22 or the voluntary acti-vation level with reference to the resting doublet(VALr).28,36 Furthermore, central fatigue can also beinvestigated as the change in the electromyographic(EMG) signal recorded during the maximal volun-tary contraction (MVC) normalized to the com-pound muscle action potential (M wave).27 A previ-ous study1 showed that EMG activity of the vastuslateralis (VL) muscle could be considered as a sur-rogate of quadriceps muscle EMG activity, but itremains to be determined whether this finding isconsistent under fatigue for both voluntary andevoked contractions (M wave). Kufel et al.23 sug-gested that the potentiated twitch is better for de-

Abbreviations: CAR, central activation ratio; CV, coefficient of variation;EMG, electromyographic; ICC, intraclass correlation coefficient; LOA ratio,limits of agreement ratio; MVC, maximal voluntary contraction; Pt, peaktwitch; RF, rectus femoris; RMS, root mean square; RMS.M�1 ratio, maximalRMS normalized to the amplitude of the M wave; VALp, maximal voluntaryactivation level with reference to the potentiated doublet; VALr, maximal vol-untary activation level with reference to the resting doublet; VL, vastus latera-lis; VM, vastus medialisKey words: central fatigue; coefficient of variation; fatigue; intraclass corre-lation coefficient; muscle activation; M wave; peripheral fatigue; twitchCorrespondence to: N. Place; e-mail: [email protected]

© 2007 Wiley Periodicals, Inc.Published online 12 January 2007 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mus.20714

486 Neuromuscular Fatigue and Reliability MUSCLE & NERVE April 2007

tecting peripheral fatigue than the unpotentiatedtwitch. Nevertheless, the amplitude of the unpoten-tiated mechanical response induced by a single stim-ulus (twitch)28,32 or paired stimuli (doublet)21,32 isstill considered in a number of studies to describecontractile fatigue. The reliability of these measuresduring the day and between days has never beenexamined in the fatigued condition. Therefore thepurpose of the present study was to further confirmthe validity of the measurements used to character-ize neuromuscular alterations by establishing theirreliability both before and after fatigue.

MATERIALS AND METHODS

Subjects. A convenience sample of 13 men (27.8 �4.9 years, 177.9 � 6.1 cm, 72.8 � 7.4 kg, mean � SD)without any known neuromuscular disorder volun-teered to participate in this study. All subjects werewell familiarized with voluntary and evoked contrac-tion of the quadriceps, having participated in severalexperiments on this muscle in the laboratory duringthe past 3 years. They were asked to refrain fromstrenuous physical activity for 24 h prior to eachtesting session and to maintain their normal physicalactivity level during the experimental period. Theprocedures were conducted according to the Decla-ration of Helsinki. Prior to the study, each subjectgave written consent and the University of BurgundyCommittee on Human Research approved the studyprotocol.

Apparatus. A Biodex isokinetic dynamometer(Shirley, New York) was adjusted to position theknee joint at 75° of flexion (0° � leg fully extended)and the hip joint at an angle of 90°. Experimentswere performed on the dominant (right for all sub-jects) lower limb. Quadriceps muscle was chosen as(1) it has been extensively used in previous fatiguestudies, and (2) EMG activity and muscular torqueinduced by electrically evoked contractions are easyto record. The axis of the dynamometer was alignedwith the knee extension axis, and the lever arm wasattached to the shank with a strap. Extraneous move-ment of the upper body was limited by two crossovershoulder harnesses and a belt across the abdomen.Subjects were asked to cross their arms during thetesting procedure. Gravity correction was performedto account for the weight of the limb.

Protocol. The experimental procedure consisted oftwo sessions, separated by 3–5 days. The following

design was strictly repeated for the two differentdays.

Warm-Up. Before any measurement, subjectsperformed a warm-up that included �10 brief andnon-fatiguing submaximal contractions of the kneeextensor muscles. This was followed by a 2-min restperiod before starting the experimental procedurein order to standardize the protocol.

Prefatigue and Postfatigue Tests. A graphic over-view of the protocol for one session is provided inFigure 1. A series of neuromuscular tests were ap-plied to each subject. This sequence lasted 33 s andincluded: (1) three electrically evoked twitches sep-arated by 3 s (total duration of 9 s); (2) one 5-s kneeextensors MVC torque assessment with doublet de-livered 3 s before (resting doublet), over the isomet-ric plateau (superimposed doublet), and �2 s afterthe MVC (potentiated doublet) (total duration of15 s); and (3) three electrically evoked twitches sep-arated by 3 s (total duration of 9 s).

This sequence was performed twice before fa-tigue with 1 min rest to allow for recovery after the

FIGURE 1. Graphic overview of the procedures used for one ofthe two sessions. (A, B) Sequences pre 1 and pre 2; (C) 2-minsustained MVC; (D, E) sequences post 1 and post 2. Small arrow,single stimulation; large arrow, paired stimulation.

Neuromuscular Fatigue and Reliability MUSCLE & NERVE April 2007 487

first MVC and twice after fatigue with a rest period ofabout 10 s to limit the recovery process. The greatestlevel of torque achieved by the subjects before thefatiguing exercise was taken as the MVC torque.Before fatigue, the two trials for MVC were within5% for every subject. The procedures were identicalfor the second session.

Fatigue Test. After a recovery time of 3 min,subjects had to perform a MVC for 2 min37 with thequadriceps femoris muscle. Strong verbal encour-agement was provided by the investigators through-out the duration of the sustained MVC.

Electrical Stimulation. Transcutaneous electricalstimulation was induced using a high-voltage (maxi-mal voltage 400 V) constant-current stimulator(model DS7, Digitimer, Welwyn Garden City, UnitedKingdom). The femoral nerve was stimulated using amonopolar cathode ball electrode (0.5-cm diame-ter) pressed into the femoral triangle by the sameinvestigator (N. P.) during all experiments. The siteof stimulation, i.e., the position giving the greatestvisible contraction of the whole quadriceps musclegroup, was marked on the skin so that it could beutilized again after the sustained contraction. Theanode was a 50-cm2 (10 � 5 cm) rectangular elec-trode (Compex SA, Ecublens, Switzerland) locatedin the gluteal fold opposite the cathode. The optimalintensity of stimulation (i.e., that which recruited allknee-extensor motor units) was considered to bereached when an increase in the stimulation inten-sity did not induce a further increase in the ampli-tude of the twitch force and the peak-to-peak ampli-tude of knee-extensor M waves. This intensity wasfurther increased by 25% (to ensure that it was su-pramaximal) and then maintained for single andpaired stimulations. The stimulus duration was 1 msand the interval between stimuli in the doublet(paired stimulations) was 10 ms. The supramaximalintensities ranged from 35 to 87 mA.

Electromyography. Surface EMG activity of the vas-tus lateralis (VL), vastus medialis (VM), and rectusfemoris (RF) muscles was recorded with pairs ofsilver chloride, circular (recording diameter, 10mm) surface electrodes (Controle Graphique Medi-cal, Brie-Comte-Robert, France) positioned length-wise over the middle of the muscle belly with aninterelectrode (center-to-center) distance of 20 mm.These recording sites were determined in pilot test-ing at the beginning of the first test session by elic-iting the largest M wave for each muscle via femoralnerve stimulation and marked so that they could bereutilized for the second test session. Low resistance

between the two electrodes (� 5 k�) was obtainedby abrading the skin, and oil and dirt were removedfrom the skin using alcohol. The reference electrodewas fixed to the patella of the opposite leg. Myoelec-trical signals were amplified with a bandwidth fre-quency ranging from 15 Hz to 5 kHz (commonmode rejection ratio, 90 dB; impedance input, 100M�; gain, 1,000), digitized on-line at a samplingfrequency of 2 kHz using a computer (IPC 486;Newton PC, Dijon, France) and stored for analysiswith commercially available software (Tida; Hekaelektronik, Lambrecht/Pfalz, Germany).

Data Analysis. The three trials for the mechanicaltwitch response were averaged and further analyzed.Peak twitch (Pt), resting peak doublet, and potenti-ated peak doublet were analyzed. Peak-to-peak am-plitude and duration of the M wave were analyzedfor VL, VM, and RF muscles with the average of thethree trials used for analysis.

Global central activation was determined withthe three different calculations most often usedwhen assessing central fatigue.

Central activation ratio (CAR) is equal to peakMVC divided by the sum of peak MVC and superim-posed doublet amplitude.20 Maximal voluntary acti-vation level with reference to the resting doublet(VALr) is calculated as [1 � (superimposed doubletamplitude/resting doublet amplitude)] � 100.28

Maximal voluntary activation level with reference tothe potentiated doublet (VALp) is [1 � (superim-posed doublet amplitude/potentiated doublet am-plitude)] � 100.4

A correction was consistently applied to the orig-inal equation when the superimposed doublet waselicited slightly before or after the real maximalvoluntary torque.40 In these cases, VAL was calcu-lated as [1 � (superimposed doublet amplitude �voluntary torque level just before the superimposeddoublet/maximal voluntary torque)/potentiateddoublet amplitude] � 100.

Furthermore, EMG activity of the VL, VM, andRF muscles collected during the knee-extensor MVCwas quantified as the root mean square (RMS) for a0.5-s interval at peak force (250-ms interval eitherside of the peak force). Maximal RMS values for VL,VM, and RF muscles were then normalized to theamplitude of the M wave for the respective muscles,in order to obtain RMS.M�1 ratio. This normaliza-tion procedure accounted for peripheral influencesincluding neuromuscular propagation failure andchanges in impedance from the EMG recordings.


Statistical Analysis. Statistical tests were conductedfor the following variables: Pt, resting peak doublet,potentiated peak doublet, M-wave amplitude andduration of VL, VM, and RF muscles, MVC torque,torque during the sustained MVC, CAR, VALr, VALp,and RMS.M�1 ratio for VL, VM, and RF muscles. AKolmogorov–Smirnov test revealed that all thesevariables were normally distributed. Data are re-ported as means � standard deviation (SD) withinthe text and tables.

Separated three-way ANOVAs [sequence (e.g.,pre 1 vs. pre 2) � time (pre- vs. post-fatigue) �session (1 vs. 2)] were used to compare dependentvariables when appropriate. Post-hoc analyses(Tukey’s honestly significant difference test) wereused to test for differences among pairs of meanswhen appropriate. A significance level of P � 0.05was used to identify statistical significance.

Relative reliability concerns the degree to whichindividuals maintain their position in a sample withrepeated measurements.5 We assessed this type ofreliability with the intraclass correlation coefficient(ICC) (2,1), a two-way random effects model withsingle measure reliability in which variance over therepeated session is considered.38 The ICC indicatesthe error in measurements as a proportion of thetotal variance in scores. ICC was not calculated forCAR, VALr, and VALp due to the ceiling effect (poorbetween-subject variability) associated with thesemeasurements.41

Absolute reliability is the degree to which re-peated measurements vary for individuals5; this typeof reliability is expressed either in the actual units ofmeasurement (typical error) or as a proportion ofthe measured values (coefficient of variation, CV).Both typical error and CV refer to intrasubject vari-ation between two measurements. Typical error (orstandard error of measurement) was calculated asstandard deviation of the differences between thetwo measurements/�2. CV was calculated for eachsubject as standard deviation of the two measure-ments/mean of these two measurements �100.Mean CV corresponding to the 13 subjects was thenreported.

Bland and Altman’s 95% limits of agreement,11

another measure of absolute reliability, was used tofurther explore reproducibility of the central activa-tion parameters for which no ICC was calculated(i.e., CAR, VALr and VALp). This was done in thefollowing manner: (1) the difference between thetwo measurements for each subject was plottedagainst their average value of the two trials, and (2)the plots were examined to see whether any datapoints were beyond 1.96 � SD of the differences

(95% confidence limits) above and below the meanof the two trials.11 As recommended by Atkinson andNevill,5 we also determined from the Bland andAltman plot the presence of heteroscedasticity(when the differences depend on the magnitude ofthe mean) by using the same criterion as the recentstudy of Clark et al.13 A correlation (R2) between theabsolute differences and the mean values between 0and 0.1 is considered homoscedastic (no relationbetween error and the size of the measured vari-able), and R2 values greater than 0.1 are heterosce-dastic.13 In this last case, limits of agreement ratio(LOA ratio) was calculated as (1.96 � SD diff/grandmean) � 100, where “SD diff” represents standarddeviation of the differences between the two testsand “grand mean” is (mean of test 1 � mean of test2)/2.5 In the case of homoscedasticity, systematicbias (mean of the differences between test 1 and test2) and random error (SD of the differences � 1.96)were determined.5

Statistical analyses were performed using Statis-tica software for Windows (Statsoft, version 6.1; Sta-tistica, Tulsa, Oklahoma). ICC and typical error anal-yses were done using a downloadable spreadsheet.19

RESULTS

Fatigue. The torque produced during the 2-minsustained MVC decreased significantly for the twosessions (on average from 263 � 42 N.m at thebeginning of the contraction to 59 � 19 N.m at theend of the 2 min (P � 0.001), i.e., a loss of 77 � 8%).No significant difference in torque at start, middle,and end of the fatiguing contraction was found be-tween the two sessions, indicating a comparable ef-fort from the subjects during the fatiguing task of thetwo sessions.

To explore whether the model we chose inducedneuromuscular fatigue of the quadriceps muscle andto determine the potential sites of fatigue, three-wayANOVAs with repeated measures was consistentlyadopted for each dependent variable. ANOVAshowed a decrease for all measurements (P � 0.05),except for M-wave amplitude and duration, whichwere not affected by the fatiguing task (Table 2).Figure 2 illustrates the decrease in MVC torque andthe reduction in peak doublet amplitude after thesustained contraction.

From the different parameters depressed follow-ing the sustained MVC, only Pt, peak doublet andpotentiated peak doublet started to recover betweenpost 1 and post 2 sequences, and this recovery wasfound whatever the session (on average by 28% �22%, 16% � 12%, and 14% � 9% for Pt, resting


peak doublet and potentiated peak doublet, respec-tively, P � 0.05). A one-way ANOVA showed that Ptrecovery was faster than for doublets (P � 0.05).

Reliability. When comparing a measurement be-tween the two sessions and for the same sequence(i.e., pre 1 vs. pre 1, pre 2 vs. pre 2, post 1 vs. post 1,and post 2 vs. post 2), ANOVA failed to show any

significant differences for all parameters (no “ses-sion” effect, P 0.05), except for the RMS.M�1 ratioof RF muscle (P � 0.035).

We chose to present intraday reliability (withinthe same test session) during the first session forevery subject and for measurements performed be-fore fatigue only (Table 1); some parameters par-tially or totally recovered between the two trials after

FIGURE 2. Representative recording of a superimposed maximal voluntary contraction with potentiated doublet for: (A) session 1sequence pre 1; (B) session 1 sequence post 1; (C) session 2 sequence pre 1; (D) session 2 sequence post 1. Note the decreased MVCtorque and peak doublet amplitude after the fatiguing task.

Table 1. Intraday reliability of neuromuscular function measures before fatigue (session 1 sequence pre 1 vs. session 1 sequence pre 2).

Dependent variable

Intraday reliability before fatigue

Mean value (SD) ICC (95% CI)Mean CV

(SD) Typ. err.

MVC (N.m) 272 (46) vs. 276 (51) 0.98 (0.94–0.99) 2.2 (1.2) 6.6Central parameters

Muscle activationCAR (%) 97.2 (1.1) vs. 97.6 (1.2) Not calculated 0.7 (0.5) 0.8VALr (%) 90.8 (3.9) vs. 92.2 (4.1) Not calculated 2.5 (1.7) 2.6VALp (%) 92.9 (2.8) vs. 93.7 (2.0) Not calculated 1.7 (1.2) 1.9RMS.M�1 ratio VL 0.054 (0.025) vs. 0.051 (0.026) 0.96 (0.86–0.99) 11.2 (10.0) 0.005RMS.M�1 ratio VM 0.067 (0.018) vs. 0.064 (0.023) 0.75 (0.32–0.93) 15.3 (12.0) 0.010RMS.M�1 ratio RF 0.108 (0.051) vs. 0.099 (0.040) 0.85 (0.56–0.95) 10.5 (9.9) 0.018

Peripheral parametersM-wave properties

Amplitude VL (mV) 12.6 (4.6) vs. 12.6 (5.0) 0.99 (0.96–1) 2.7 (2.8) 0.5Amplitude VM (mV) 9.8 (3.0) vs. 9.3 (3.3) 0.86 (0.59–0.96) 6.1 (12.2) 1.2Amplitude RF (mV) 5.0 (2.2) vs. 5.0 (2.1) 0.99 (0.96–1) 3.4 (3.7) 0.2Duration VL (ms) 7.9 (3.3) vs. 7.9 (3.4) 0.99 (0.98–1) 2.6 (3.2) 0.2Duration VM (ms) 6.0 (1.9) vs. 6.1 (2.1) 0.98 (0.94–0.99) 3.4 (2.9) 0.3Duration RF (ms) 8.6 (2.0) vs. 8.7 (2.1) 0.97 (0.91–0.99) 3.1 (2.8) 0.3

Contractile propertiesPeak twitch (N.m) 45.5 (9.9) vs. 45.2 (10.6) 0.98 (0.95–1) 2.1 (1.9) 1.3Rest. Peak doublet (N.m) 84.5 (15.2) vs. 87.3 (16.6) 0.98 (0.93–0.99) 2.7 (2.0) 2.4Pot. Peak doublet (N.m) 103.7 (17.2) vs. 102.2 (19.0) 0.94 (0.82–0.98) 3.0 (3.2) 4.3

CAR, central activation ratio; CV, coefficient of variation; ICC, intraclass correlation coefficient; MVC, maximal voluntary contraction; 95% CI, lower and upperconfidence intervals; Pot. peak doublet, potentiated peak doublet; Rest. peak doublet, resting peak doublet; RF, rectus femoris; RMS.M�1 ratio, maximal RMSnormalized to the amplitude of the M-wave; Typ. err., typical error; VALp, maximal voluntary activation level with reference to the potentiated doublet; VALr,maximal voluntary activation level with reference to the resting doublet; VL, vastus lateralis; VM, vastus medialis.


fatigue and the estimation of the reliability in theseconditions would be pointless. For reasons of clarity,interday reliability (between the two testing sessions)is presented before exercise for the first sequenceonly (Table 2), as ANOVA showed no differencebetween the two sequences conducted before fatiguefor any measurements (P 0.05). Finally, interdayreliability after fatigue is presented only for the firstsequence immediately after the sustained MVC tolimit the number of data (Table 2). To compareindividual values of muscle activation between thetwo sessions, Bland and Altman’s 95% limits ofagreement for CAR, VALr, and VALp are presentedbefore (sequence pre 1) and after fatigue (sequencepost 1) in Figure 2. Reliability of these parametersusing the limits of agreement measures are detailedin Table 3 for both intraday and interday compari-sons.

From Tables 1 and 2, it appears that both intra-day and interday reliability measured before fatiguewere high, with ICC often above 0.8, CV often �5%for intraday measurements and often �15% for in-terday parameters. Low typical errors were found,but always lower for intraday than interday measures.ICC remained high after fatigue on average (over 0.8for 9 of the 13 variables investigated) and CV was

slightly higher than before fatigue (8.9% vs. 10.8%on average). Typical errors remained in the samerange compared with before fatigue.

Table 2. Interday reliability of neuromuscular function measures before (session 1 sequence pre 1 vs. session 2 sequence pre 1) and after(session 1 sequence post 1 vs. session 2 sequence post 1) fatigue.*

Dependent variable

Interday reliability before fatigue Interday reliability after fatigue

Mean value (SD) ICC (95% CI)

Mean CV

(SD)

Typ.

Err. Mean value (SD) ICC (95% CI)

Mean CV

(SD)

Typ.

err.

MVC (N.m) 272 (46) vs. 278 (45) 0.90 (0.75; 0.97) 3.5 (4.2) 14.0 180 (54) vs. 193 (49) 0.91 (0.77; 0.97) 7.4 (7.4) 15.0

Central parameters

Muscle activation

CAR (%) 97.2 (1.1) vs. 97.9 (1.5) Not calculated 1.2 (1.0) 1.5 93.9 (3.5) vs. 93.7 (4.7) Not calculated 2.5 (2.4) 3.1

VALr (%) 90.8 (3.9) vs. 92.8 (4.9) Not calculated 4.3 (3.4) 4.9 83.4 (7.6) vs. 82.5 (10.5) Not calculated 7.8 (7.2) 8.3

VALp (%) 92.9 (2.8) vs. 94.1 (4.0) Not calculated 3.1 (2.7) 3.9 86.4 (6.1) vs. 85.9 (8.2) Not calculated 6.0 (5.4) 6.7

RMS.M�1 ratio VL 0.054 (0.025) vs. 0.052 (0.027) 0.78 (0.41; 0.93) 12.1 (10.2) 0.012 0.043 (0.032) vs. 0.046 (0.025) 0.71 (0.24; 0.91) 15.6 (13.7) 0.015

RMS.M�1 ratio VM 0.067 (0.018) vs. 0.061 (0.011) 0.45 (�0.17; 0.81) 13.1 (10.1) 0.011 0.052 (0.025) vs. 0.047 (0.016) 0.87 (0.59; 0.96) 15.6 (10.2) 0.008

RMS.M�1 ratio RF 0.108 (0.051) vs. 0.122 (0.051) 0.90 (0.70; 0.97) 13.4 (11.8) 0.016 0.083 (0.043) vs. 0.094 (0.043) 0.91 (0.72; 0.97) 15.5 (8.7) 0.013

Peripheral parameters

M-wave properties

Amplitude VL (mV) 12.6 (4.6) vs. 11.0 (3.7) 0.71 (0.27; 0.91) 14.6 (13.5) 2.3 12.2 (4.3) vs. 10.8 (3.3) 0.70 (0.21; 0.91) 17.0 (13.8) 2.1

Amplitude VM (mV) 9.8 (3.0) vs. 9.3 (2.8) 0.28 (�0.32; 0.72) 18.8 (18.2) 2.5 9.8 (2.8) vs. 9.7 (2.6) 0.60 (0.04; 0.87) 13.9 (10.8) 1.7

Amplitude RF (mV) 5.0 (2.2) vs. 4.2 (1.5) 0.83 (0.52; 0.95) 14.7 (10.0) 0.8 4.8 (2.2) vs. 4.3 (1.7) 0.87 (0.59; 0.96) 13.0 (7.7) 0.7

Duration VL (ms) 7.9 (3.3) vs. 7.7 (3.3) 0.97 (0.89; 0.99) 6.9 (4.0) 0.6 8.2 (3.6) vs. 8.3 (3.2) 0.95 (0.84; 0.99) 5.8 (4.8) 0.7

Duration VM (ms) 6.0 (1.9) vs. 6.6 (2.8) 0.86 (0.57; 0.96) 11.0 (7.0) 0.9 6.2 (1.8) vs. 6.3 (2.2) 0.91 (0.68; 0.98) 10.7 (14.3) 0.6

Duration RF (ms) 8.6 (2.0) vs. 8.8 (2.9) 0.82 (0.46; 0.95) 8.4 (8.5) 1.0 9.0 (1.9) vs. 8.4 (1.8) 0.38 (�0.33; 0.81) 16.5 (18.2) 1.5

Contractile properties

Peak twitch (N.m) 45.5 (9.9) vs. 44.4 (10.3) 0.92 (0.74; 0.98) 5.6 (5.0) 2.9 30.8 (13.2) vs. 33.2 (14.7) 0.94 (0.82; 0.98) 9.3 (10.2) 3.3

Rest. peak doublet

(N.m)

84.5 (15.2) vs. 83.5 (18.5) 0.85 (0.55; 0.95) 5.9 (6.2) 6.7 67.4 (25.5) vs. 70.1 (25.3) 0.94 (0.81; 0.98) 7.8 (9.8) 6.1

Pot. peak doublet

(N.m)

103.7 (17.2) vs. 98.9 (21.4) 0.93 (0.78; 0.98) 5.2 (4.5) 5.1 75.2 (26.0) vs. 80.8 (27.5) 0.92 (0.75; 0.98) 8.3 (10.1) 7.6

*Abbreviations are the same as in Table 1.

Table 3. Limits of agreement measures of central activationmeasured within the same session (intraday) before fatigue andbetween the two sessions (interday) before and after fatigue.*

Dependentvariable Heteroscedasticity

Systematicbias

Randomerror

LOAratio

Intraday. preCAR (%) No 0.38 2.14VALr (%) No 1.49 7.08VALp (%) No 0.77 5.27

Interday. preCAR (%) No 0.62 4.08VALr (%) No 2.09 13.72VALp (%) Yes 11.32

Interday. post ICAR (%) No �0.19 8.72VALr (%) No �0.91 22.93VALp (%) No �0.54 18.42

CAR, central activation ratio; LOA ratio, limits of agreement ratio; post 1,sequence post 1; post 2, sequence post 2; pre, sequence pre 1; VALp,maximal voluntary activation level with reference to the potentiated doublet;VALr, maximal voluntary activation level with reference to the resting doublet.*Systematic bias was not significantly different from 0 (P � 0.05, testedwith a paired t-test).


Figure 3 and Table 3 showed that CAR was themost reproducible measurement of central activa-tion, with the lowest dispersion of the individual dataand the lowest random errors. It also appears thatrandom error and absolute difference dispersion isconsistently higher for VALr compared to VALp (Ta-bles 1 and 2).

DISCUSSION

The main objective of this study was to confirm thevalidity of measures classically used to describe fatigueof the neuromuscular system by establishing their reli-ability both before and especially after a fatiguing task.We found a high level of reproducibility for centralparameters and evoked force, and moderate to highrepeatability for measurements involving surface EMG(M wave and RMS.M�1 ratio). By combining previousresults on validity and present findings on reliability,the discussion will aim to provide methodological rec-ommendations in the choice of variables that shouldbe used in further studies to describe fatigue along theneuromuscular system.

Origin of Fatigue. The 2-min sustained MVC waschosen in the present study as it induces both centraland peripheral fatigue.37 We observed a significantdrop in MVC after the fatiguing contraction(�33%). Furthermore, significant central fatigue(�4.1% for CAR, �11.6% for VALr, �9.0% for

VALp, and �22.4% for RMS.M�1 ratio on averagefor the three knee extensor muscles investigated)was reported during the MVC performed immedi-ately after the sustained isometric contraction, sug-gesting that inability to voluntarily recruit all themotor units of quadriceps femoris muscle, or a sub-optimal discharge rate of the motor units pool,41

accounts for the loss in voluntary force. It might bethat central fatigue was underestimated to a similarextent for the two sessions, as superimposed MVCswere conducted a few seconds after the sustainedMVC; however, subjects experienced a large tremornear the end of the 2-min contraction, and the su-perimposition at that time would have led to biasedcalculations.

We also found an excitation–contraction couplingfailure, as shown by the decrease in force evoked byelectrical stimulation after fatigue (�30.0% for peaktwitch, �19.1% for resting peak doublet, and �36.5%for potentiated peak doublet). The unchanged M waveafter the sustained MVC indicated no failure to driveaction potentials to the muscle fibers, as previouslyobserved after sustained MVC of hand muscles,9 andsuggests that alterations in calcium kinetics beyond themuscle cell-membrane2 might have occurred. Howeverthe main aim of the present work was to determine thelevel of reliability for classic neuromuscular measure-ments after fatigue, in order to assess whether the samefatiguing exercise induced comparable alterations.

FIGURE 3. (A–F) Bland-Altman plots of differences (session 2 - session 1) vs. mean of the two sessions for CAR, VALr and VALp for the13 subjects. The lowest and highest horizontal lines represent the 95% limits of agreements, whereas the middle dashed line symbolizesthe mean of the differences between test 1 and test 2 (systematic bias). Note the difference in scale on y-axis between the three variables.Heteroscedasticity was only found for VALp before fatigue (R2 � 0.12, Fig. 2C). CAR, central activation ratio; VALr, maximal voluntaryactivation level with reference to the resting doublet; VALp, maximal voluntary activation level with reference to the potentiated doublet;pre, sequence pre 1; post, sequence post 1.


Which Variables Should We Consider When Assessing

Neuromuscular Fatigue? We assessed both intradayand interday (3–5 days interval) reliability of selectedparameters before fatigue. All variables had a lowerlevel of variability within the same session than be-tween sessions, as confirmed by CV and typical errorvalues. This observation was the same for relativereliability (ICC), which describes the agreement be-tween the repeated measures. We have not providedthreshold values for ICC, as no consensus really ex-ists; as suggested by Morrow and Jackson,29 thereader has to determine the “practical significance”of the obtained reliability.

MVC, the “gold standard” of muscle fatigue, washighly reproducible, with ICC 0.90 and CV � 8%;MVC has already been reported to be highly repro-ducible without fatigue for quadriceps muscle, withCV �7% within a testing session and equal to 10%across sessions separated by 2–7 days.24,41 Todd etal.41 observed an ICC of 0.79 between sessions,slightly below our 0.90, which could be explained bythe larger interval between the two testing sessions intheir study (1 week). Fatigue did not alter variabilityof MVC torque, as typical error after the 2-min sus-tained MVC was equivalent to that calculated beforefatigue (�15 N.m). Furthermore, CV was in line witha previous study which assessed MVC reproducibilityafter a sustained 4-min isometric MVC of the dorsi-flexor muscles.20 Altogether, these findings suggestthat performing a MVC after a fatiguing exercise is areliable measure to detect muscle fatigue.

The validity of three methods based on interpo-lated twitch technique has been investigated in pre-vious studies. For example, the CAR method hasbeen criticized,39 as the calculation compares a su-perimposed force with a voluntary contraction inwhich additional synergistic muscles not recruited bythe stimulation could contribute to torque produc-tion.7 We observed that this technique led to anoverestimation of central activation compared to theVAL method, where the superimposed evokedtorque is compared with a resting evoked torque.Furthermore, Bilodeau10 reported that the VALmethod is more sensitive to fatigue than CAR. Itappears also that the use of a potentiated doublet asa reference is more appropriate, as the superim-posed response is also under the influence of poten-tiation during the MVC.14 Our results showed thatthe three indexes of central activation based ontwitch interpolation were highly consistent beforefatigue, with CV and typical error lower than 5%,confirming previous studies.3,6,41 Fatigue resulted inadditional but acceptable variability, with CV � 8%.It is noteworthy that CAR was more reliable than

VAL, both before and after fatigue. Furthermore,VALp was slightly but consistently more reproduciblethan VALr. The analysis of limits of agreement re-vealed that random error (i.e., biological or mechan-ical variation) could account for the vast majority ofthe observed variance for central activation parame-ters, as systematic bias, referring to a general trendfor measurements to be different in a particulardirection between repeated tests, was not signifi-cantly different from 0.5 As for CV and typical error,random error was consistently lower for CAR thanfor VAL measures and slightly lower for VALp thanVALr. As VALp before fatigue and between sessionswas heteroscedastic, the LOA ratio was calculatedand it can be said that “any two tests will differ due tomeasurement error by no more than 11% either in apositive or negative direction.”5 Finally, Figure 2showed that data dispersion was limited for all mea-sures of central activation, as all the test–retest dif-ferences (except one for VALr after fatigue) werewithin the 95% limits of agreement that were calcu-lated for those differences. However, the results con-cerning LOA ratio have to be considered cautiouslybecause of potential variability from sample to sam-ple (based on the relatively small sample size). Ourresults showed a high level of reliability for the threemethods to determine muscle inactivation after fa-tigue; within the context of validity discussed above,we propose that VALp should be used to describe thecentral component of fatigue.

Although less used in the literature than thetwitch interpolation technique, RMS.M�1 ratio is an-other technique to detect neural alterations due tofatigue.27 This method is based on surface electro-myography and thus allows changes in descendingdrive between the synergists of a muscle group to bedistinguished. Few studies have compared the reli-ability of EMG measurements across the knee exten-sor muscles. For example, Pincivero et al.31 reporteda greater reliance on EMG of the VL than VM andRF muscles to estimate muscle torque at submaximallevels. Mathur et al.25 reported comparable ICC andstandard errors of EMG measurements between VL,VM, and RF during sustained isometric contractionsat 20% and 80% MVC. Finally, Rainoldi et al.34 ob-served a slightly higher level of reliability of EMGmeasurements for VL than VM during sustained iso-metric contractions at 50% MVC and suggested thatthis was due to a more critical electrode reposition-ing pertaining to a lower cross-sectional area18 of thelatter compared to the former. We obtained quitevariable measurements (CV 10%) with ICC 0.70,as already noticed by Gondin et al.,18 except for theVM muscle before fatigue between sessions (ICC �


0.45). Fatigue added only minor variability in themeasurements. Besides, ANOVA revealed a sessioneffect for RF muscle, suggesting that care should betaken when considering EMG activity of RF muscle.Nonetheless the main advantage of this technique isto compare neural alterations between different syn-ergists investigated after a fatiguing task so that sev-eral muscles could be considered; thus our resultssuggest that studies using RMS.M�1 ratio and moregenerally surface EMG as an index of voluntary driveshould be interpreted with caution.

Although it appeared that neuromuscular trans-mission or propagation was unchanged after fatigue,we assessed repeatability of peak-to-peak amplitudeand duration of M wave for the three knee extensormuscles investigated. Reliability was very good withinthe same testing day for amplitude and duration withCV � 7% and ICC 0.85, as previously observed forVL and RF muscles.32 However, when assessed be-tween different test days, M-wave variability in-creased in the present study (CV often 10% andtypical errors three time higher than within the sametesting session), whereas M-wave properties are reli-able on separate days.12,13,18 As fatigue did not in-duce further variance, our finding suggests that M-wave parameters are as reliable with as withoutfatigue and thus that M-wave characteristics could beused to detect failure in neuromuscular propaga-tion. Nevertheless, we observed that ICC were rela-tively low for M-wave amplitude of the VM muscle,confirming the RMS.M�1 ratio results.

It is appropriate to consider which muscle headcan be used as a surrogate for the quadriceps femo-ris. Previous works showed that the VL muscle isrepresentative of the quadriceps muscle.1,43 The cur-rent results confirm this assertion, as VL presentedthe most reproducible data. Besides, the relativelylow levels of reliability for VM suggest that this mus-cle is of secondary importance when consideringEMG of the quadriceps. Finally, the intermediatereliability found for the RF and its unique anatomicfunction within the quadriceps (both knee extensorand hip flexor) indicates that it would be interestingto consider this muscle to investigate muscular syn-ergies within the quadriceps (e.g., the interaction ofmono- and bi-articular muscles35).

Contractile failure was assessed with the ampli-tude of the electrically evoked responses. In a recentstudy, Kufel et al.23 observed that potentiated evokedresponses are more sensitive for detecting early fa-tigue than non-potentiated ones. Furthermore, theyreported that potentiated responses were as reliableas non-potentiated ones, despite the increased vari-ability associated with potentiated twitches reported

by Morton et al.30 In the present study, electricallyevoked torques were found to be highly reliablebefore fatigue with ICC 0.85 and CV � 10%, inagreement with previous studies both for intraday32

and interday41,42 comparisons; as for the MVC, aslightly higher variability, as revealed by greater typ-ical errors, was found between sessions compared towithin sessions. Reliability of Pt, resting peak dou-blet, and potentiated peak doublet was not influ-enced by the fatiguing contraction and was very closefor each of the three parameters investigated. Inter-estingly, we observed a fast recovery of evokedtorque, but it was twofold less for potentiated peakdoublet than for Pt (14% vs. 28%, P � 0.05). Asreliability was comparable between the three indexesof peripheral fatigue considered here, the most validmeasurement should be adopted. It appears that (1)potentiated peak doublet recovery was slower com-pared to unpotentiated responses immediately afterfatigue, and (2) the influence of a preceding con-traction on possible potentiation can be discardedwhen considering potentiated peak doublet, leadingto a more sensitive quantification of peripheral fa-tigue.23 Consequently, these arguments suggest thatpotentiated peak doublet should be adopted to char-acterize both contractile impairment and central fa-tigue (with VALp).

The authors thank Dr. Julie Erskine for her helpful comments.

REFERENCES

1. Alkner BA, Tesch PA, Berg HE. Quadriceps EMG/force rela-tionship in knee extension and leg press. Med Sci SportsExerc 2000;32:459–463.

2. Allen DG, Westerblad H. Role of phosphate and calciumstores in muscle fatigue. J Physiol (Lond) 2001;536:657–665.

3. Allen GM, Gandevia SC, McKenzie DK. Reliability of measure-ments of muscle strength and voluntary activation usingtwitch interpolation. Muscle Nerve 1995;18:593–600.

4. Allen GM, McKenzie DK, Gandevia SC. Twitch interpolationof the elbow flexor muscles at high forces. Muscle Nerve1998;21:318–328.

5. Atkinson G, Nevill AM. Statistical methods for assessing mea-surement error (reliability) in variables relevant to sportsmedicine. Sports Med 1998;26:217–238.

6. Behm D, Power K, Drinkwater E. Comparison of interpolationand central activation ratios as measures of muscle inactiva-tion. Muscle Nerve 2001;24:925–934.

7. Behm DG, St-Pierre DM, Perez D. Muscle inactivation: assess-ment of interpolated twitch technique. J Appl Physiol 1996;81:2267–2273.

8. Bigland-Ritchie B, Furbush F, Woods JJ. Fatigue of intermit-tent submaximal voluntary contractions: central and periph-eral factors. J Appl Physiol 1986;61:421–429.

9. Bigland-Ritchie B, Kukulka CG, Lippold OC, Woods JJ. Theabsence of neuromuscular transmission failure in sustainedmaximal voluntary contractions. J Physiol (Lond) 1982;330:265–278.

10. Bilodeau M. Central fatigue in continuous and intermittentcontractions of triceps brachii. Muscle Nerve 2006;34:205–213.


11. Bland JM, Altman DG. Statistical methods for assessing agree-ment between two methods of clinical measurement. Lancet1986;1:307–310.

12. Calder KM, Hall LA, Lester SM, Inglis JG, Gabriel DA. Reli-ability of the biceps brachii M-wave. J Neuroengineering Re-habil 2005;2:33.

13. Clark BC, Cook SB, Ploutz-Snyder LL. Reliability of tech-niques to assess human neuromuscular function in vivo. JElectromyogr Kinesiol 2007;17:90–101.

14. Folland JP, Williams AG. Methodological issues with the in-terpolated twitch technique. J Electromyogr Kinesiol 2007, inpress.

15. Fuglevand AJ, Zackowski KM, Huey KA, Enoka RM. Impair-ment of neuromuscular propagation during human fatiguingcontractions at submaximal forces. J Physiol (Lond) 1993;460:549–572.

16. Gandevia SC. Spinal and supraspinal factors in human musclefatigue. Physiol Rev 2001;81:1725–1789.

17. Gandevia SC, Enoka RM, McComas AJ, Stuart DG, ThomasCK. Neurobiology of muscle fatigue. Advances and issues. AdvExp Med Biol 1995;384:515–525.

18. Gondin J, Guette M, Ballay Y, Martin A. Electromyostimula-tion training effects on neural drive and muscle architecture.Med Sci Sports Exerc 2005;37:1291–1299.

19. Hopkins WG. Reliability from consecutive pairs of trials (Ex-cel spreadsheet). In: A new view of statistics. sportsci.org:Internet Society for Sport Science, http://www.sportsci.org/resource/stats/xrely.xls, 2000.

20. Kent-Braun JA, Le Blanc R. Quantitation of central activationfailure during maximal voluntary contractions in humans.Muscle Nerve 1996;19:861–869.

21. Klass M, Guissard N, Duchateau J. Limiting mechanisms offorce production after repetitive dynamic contractions in hu-man triceps surae. J Appl Physiol 2004;96:1516–1521.

22. Kooistra RD, de Ruiter CJ, de Haan A. Muscle activation andblood flow do not explain the muscle length-dependent vari-ation in quadriceps isometric endurance. J Appl Physiol 2005;98:810–816.

23. Kufel TJ, Pineda LA, Mador MJ. Comparison of potentiatedand unpotentiated twitches as an index of muscle fatigue.Muscle Nerve 2002;25:438–444.

24. Madsen OR. Torque, total work, power, torque accelerationenergy and acceleration time assessed on a dynamometer:reliability of knee and elbow extensor and flexor strengthmeasurements. Eur J Appl Physiol Occup Physiol 1996;74:206–210.

25. Mathur S, Eng JJ, MacIntyre DL. Reliability of surface EMGduring sustained contractions of the quadriceps. J Electro-myogr Kinesiol 2005;15:102–110.

26. Merton PA. Voluntary strength and fatigue. J Physiol 1954;123:553–564.

27. Millet GY, Lepers R. Alterations of neuromuscular functionafter prolonged running, cycling and skiing exercises. SportsMed 2004;34:105–116.

28. Millet GY, Lepers R, Maffiuletti NA, Babault N, Martin V,Lattier G. Alterations of neuromuscular function after anultramarathon. J Appl Physiol 2002;92:486–492.

29. Morrow JR Jr., Jackson AW. How “significant” is your reliabil-ity? Res Q Exerc Sport 1993;64:352–355.

30. Morton JP, Atkinson G, MacLaren DP, Cable NT, Gilbert G,Broome C, et al. Reliability of maximal muscle force andvoluntary activation as markers of exercise-induced muscledamage. Eur J Appl Physiol 2005;94:541–548.

31. Pincivero DM, Coelho AJ, Campy RM, Salfetnikov Y, Suter E.Knee extensor torque and quadriceps femoris EMG duringperceptually-guided isometric contractions. J ElectromyogrKinesiol 2003;13:159–167.

32. Place N, Lepers R, Deley G, Millet GY. Time course of neu-romuscular alterations during a prolonged running exercise.Med Sci Sports Exerc 2004;36:1347–1356.

33. Place N, Maffiuletti NA, Ballay Y, Lepers R. Twitch potentia-tion is greater after a fatiguing submaximal isometric contrac-tion performed at short vs. long quadriceps muscle length.J Appl Physiol 2005;98:429–436.

34. Rainoldi A, Bullock-Saxton JE, Cavarretta F, Hogan N. Re-peatability of maximal voluntary force and of surface EMGvariables during voluntary isometric contraction of quadri-ceps muscles in healthy subjects. J Electromyogr Kinesiol2001;11:425–438.

35. Rochette L, Hunter SK, Place N, Lepers R. Activation variesamong the knee extensor muscles during a submaximal fa-tiguing contraction in the seated and supine postures. J ApplPhysiol 2003;95:1515–1522.

36. Scaglioni G, Ferri A, Minetti AE, Martin A, Van Hoecke J,Capodaglio P, et al. Plantar flexor activation capacity and Hreflex in older adults: adaptations to strength training. J ApplPhysiol 2002;92:2292–2302.

37. Schillings ML, Hoefsloot W, Stegeman DF, Zwarts MJ. Relativecontributions of central and peripheral factors to fatigueduring a maximal sustained effort. Eur J Appl Physiol 2003;90:562–568.

38. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessingrater reliability. Psychol Bull 1979;86:420–428.

39. Stackhouse SK, Dean JC, Lee SC, Binder-MacLeod SA. Mea-surement of central activation failure of the quadriceps fem-oris in healthy adults. Muscle Nerve 2000;23:1706–1712.

40. Strojnik V, Komi PV. Neuromuscular fatigue after maximalstretch-shortening cycle exercise. J Appl Physiol 1998;84:344–350.

41. Todd G, Gorman RB, Gandevia SC. Measurement and repro-ducibility of strength and voluntary activation of lower-limbmuscles. Muscle Nerve 2004;29:834–842.

42. Todd G, Taylor JL, Gandevia SC. Reproducible measurementof voluntary activation of human elbow flexors with motorcortical stimulation. J Appl Physiol 2004;97:236–242.

43. Trappe TA, Lindquist DM, Carrithers JA. Muscle-specific at-rophy of the quadriceps femoris with aging. J Appl Physiol2001;90:2070–2074.


Documents

Assessment of the reliability of central and peripheral fatigue after sustained maximal voluntary contraction of the quadriceps muscle