Gait & Posture 46 (2016) 57–62

Contents lists available at ScienceDirect

Gait & Posture

journal homepage: www.e lsev ier .com/ locate /ga i tpost

Trunk muscles activation during pole walking vs. walking performed

at different speeds and grades

Luca Zoffoli a,*, Francesco Lucertini a, Ario Federici a, Massimiliano Ditroilo b

a Department of Biomolecular Sciences—Division of Exercise and Health Sciences, University of Urbino Carlo Bo, Urbino, Italyb Department of Sport, Health and Exercise Science, University of Hull, Hull, United Kingdom

A R T I C L E I N F O

Article history:

Received 9 September 2015

Received in revised form 17 February 2016

Accepted 22 February 2016

Keywords:

EMG

Abdominal muscles

Coactivation

Spinal stability

[2_TD$DIFF]Human locomotion

A B S T R A C T

Given their functional role and importance, the activity of several trunk muscles was assessed (via

surface electromyography—EMG) during Walking (W) and Pole Walking (PW) in 21 healthy adults. EMG

data was collected from the external oblique (EO), the erector spinae longissimus (ES), the multifidus

(MU), and the rectus abdominis (RA) while performing W and PW on a motorized treadmill at different

speeds (60, 80, and 100% of the highest speed at which the participants still walked naturally; PTS60,

PTS80 and PTS100, respectively) and grades (0 and 7%; GRADE0 and GRADE7, respectively). Stride length,

EMG area under the curve (AUC), muscles activity duration (ACT), and percentage of coactivation (CO-ACT)

of ES, MU and RA, were calculated from the averaged stride for each of the tested combinations.

Compared to W, PW significantly increased the stride length, EOAUC, RAAUC and the activation time of

all the investigated muscles, to different extents depending on treadmill speeds and grades. In addition,

MUAUC was higher in PW than in W at GRADE0 only (all speeds, p < 0.01), while ESAUC during W and PW

was similar at all the speeds and grades. These changes resulted in longer CO-ACT in PW than W, at

GRADE0-PTS100 (p < 0.01) and GRADE7 (all speeds, p < 0.01). In conclusion, when compared to W, PW

requires a greater engagement of the abdominal muscles and, in turn, a higher control of the trunk

muscles. These two factors taken together may suggest an elevated spinal stability while walking

with poles.

� 2016 Elsevier B.V. All rights reserved.

1. Introduction

Pole walking (PW) is a walking-based physical activity thatimplies the use of a pair of poles in opposition to the lower limblocomotion [1]. This activity has proved to be effective inmaintenance/improvement of the cardiovascular system function[1] and, when compared to walking (W), to increase both heart rateand oxygen consumption to a higher extent [2]. Nevertheless, themuscular responses to PW, which could help understand thepotential benefits and/or drawbacks of this exercise mode, haveonly been partially investigated. When W and PW have beencompared, the analysis of the muscle activity revealed that theupper limb muscles are generally more active during uphill PW,while uphill W appears to activate more the lower limb muscles[3]. However, only one study focused on the differences betweenW and PW at the trunk level, and found the same activation

* Corresponding author. Tel.: +39 0722 30 4611.

E-mail address: luca.zoffoli@uniurb.it (L. Zoffoli).

http://dx.doi.org/10.1016/j.gaitpost.2016.02.015

0966-6362/� 2016 Elsevier B.V. All rights reserved.

amplitude of the erector spinae longissimus between uphill W andPW while carrying a backpack [4].

The trunk muscles are fundamental for the balance of the wholebody, and it is thought that the neuromuscular system acts throughtheir coactivation to provide adequate spinal stability in differentconditions [5]. They also assist the movement of the arms and legsduring locomotion and other physical activities [6], and modulatetheir activity and function according to a specific task (e.g.changing the W speed) [7]. For instance, while W the erector spinaemuscles preserve the body balance perturbed by arm swing [8] andanticipate and support the pelvis movements [9]. Conversely, theexternal oblique muscles switch their activation pattern from tonicto phasic in a speed-dependent way, reflecting both theirstabilizing and mobilizing role during W [7].

Given the multiple functions and overall importance of thismuscle groups, it is pivotal to examine the role of trunk musclesduring PW and how this compares to W. Elucidating the effect ofspeed and grade on the activity of trunk muscles will provideadditional insights into their neuromuscular response to these twomodes of human locomotion. Accordingly, this study aimed toconcurrently measure and describe the activity of several trunk

L. Zoffoli et al. / Gait & Posture 46 (2016) 57–6258

muscles in a healthy population whilst performing both PW and Wat different speeds and grades.

2. Methods

2.1. Participants

Ten healthy males (age: 28.5 � 5.6 y, mass: 78.3 � 9.5 kg, height:1.77 � 0.06 m) and eleven females (age: 33.0 � 10.1 y, mass:66.2 � 7.5 kg, height: 1.68 � 0.09 m) were recruited. The study wasapproved by the Ethical Research Committee of the Sports, Health andExercise Science Department of the University of Hull (UK). Theparticipants signed a written informed consent before their inclusionin the study and had to fill in a pre-exercise medical questionnaire. Allparticipants were free from chronic low-back pain and were asked torest the day before each testing session.

2.2. Experimental procedure

Participants attended a minimum of two testing sessions, atleast 24 h apart. During the first visit, the preferred transitionspeeds (PTS), i.e. the highest speed at which the participants stillwalk naturally, (at both 0% (GRADE0) and 7% (GRADE7) grades weredetermined (GRADE0 mean � SD: 1.96 � 0.16 m/s; GRADE7

mean � SD: 1.84 � 0.14 m/s) on a motorized treadmill (Pulsar—h/p/cosmos Sports & Medical, Nussdorf-Traunstein, Germany). The PTSwas identified using a modified version of the Hreljac’s protocol [10]:each stage duration was set at 20 s and the speed increment/decrement during each trial was set at 0.2 km/h. In the same testingsession, the participants were familiarized to PW, but additionalfamiliarization sessions were planned if required to meet thefollowing criteria: walk fluently while looking forward; keepingthe poles inclined backwards with the elbows slightly flexed;extending the arms behind the body at the end of the pushingphase. Because different PW techniques exist [1], these criteria werechosen as they are those mainly met by nordic walkers [11], thusallowing a similar PW technique across the participants.

During the last visit, surface EMG data was collected, on thedominant side (defined by asking the participant which foot theywould use to kick a ball [12]), from the external oblique (EO), theerector spinae longissimus (ES), the multifidus (MU) and the rectusabdominis (RA). After the equipment setup, the baseline EMGactivity was collected with the participants standing still for 30 s.Then, a 5-min warm-up (PW-GRADE0-60% PTS) was performedprior to four randomized tests combining either W or PW withGRADE0 and GRADE7. Each test required three 1-min bouts ofexercise at 60, 80 and 100% of the PTS (PTS60, PTS80 and PTS100,respectively). Pilot tests revealed that the PTS60 and PTS80 trialswere generally well tolerated by the participants (6–14 range ofthe Borg’s 6–20 rate of perceived exertion scale; RPE [13]).Conversely, the PTS100 trials were more challenging (13–17 RPErange). Therefore, to reduce the effect of fatigue, 1-min rest wasallowed after the PTS60 and PTS80 trials, whereas, at the end of thePTS100 trials, the participants sat until the heart rate dropped to theresting value (measured for 5 min while sitting before the warm-up). The recovery was assumed to be completed when the heartrate was steadily within the resting value � 5 bpm for at least 1 min.

2.3. Equipment setup

A heart rate monitor (RS800CX—Polar Electro Oy, Kempele,Finland) was worn by the participants during all the testingsessions.

The study was conducted using a pair of trekking telescopicaluminium poles (Forclaz 500—Quechua, Passy, France) withadjustable wrist straps and hard rubber covers at the distal ends,

specifically made to allow to incline the poles backwards. The polelength was adjusted to each participant’s body size [14].

After the skin was shaved, slightly abraded and cleaned with analcohol swab, EMG electrodes (BlueSensor N—Ambu, Copenaghen,Denmark) were placed and secured parallel to the muscles fibres(with 2 cm inter-electrode distance) as follows: EO, 3 cm anteriorto the mid-point of the line between the lateral pelvic crest and thelateral lower ribcage margin [15]; ES, 2 cm apart of the spinalprocess of L1 [16]; MU, about 2 cm apart from the back midline atL5 level [16]; RA, 86% of a line parallel to the linea alba(approximately 2 cm apart) starting from the xiphoid processand ending at level of the superior anterior iliac spine [17]. A 24 Gtri-axial accelerometer was placed and secured on the dominanttibia mid-way of the line between its medial condyle and thelateral malleolus.

2.4. Data collection and processing

All data was collected synchronously (sampling rate: 1500 Hz;input impedance: >100 MV; CMRR: >100 dB; baseline noise:<1 mV RMS; base gain: 200; final gain: 500) and stored on acomputer using a 16 bit resolution wireless system (Desktop DTS—Noraxon USA Inc., Scottsdale, Arizona, USA).

Raw EMG data was processed firstly applying a 2nd order,phase-corrected, band-pass Butterworth filter with bandwidthcut-off of 10–500 Hz. Secondly, the heart beats artefacts wereremoved by a 2nd order, phase-corrected, high-pass Butterworthfilter with cut-off of 30 Hz [18]. Thirdly, the Teager–Kaiser energyoperator was applied in order to improve the muscles activationonset detection during the subsequent analysis [19] (see below).Finally, the signal was full-wave rectified and the linear envelopewas obtained through a 2nd order, phase-corrected, low-passButterworth filter with cut-off frequency of 10 Hz [20].

The static accelerometer tilt was corrected as described byKavanagh [21], then the anterior-posterior accelerations of thetibia were used for strides detection [21] (see the appendix insupplementary material for the MATLAB code).

For each trial, the central 30 consecutive strides were selectedand time-normalized to 101 points prior to the calculation of theirpoint-by-point average. For each participant, the 12 averagestrides obtained (resulting from the combination of two locomo-tion types, two treadmill grades, and three speeds) werenormalized to the peak of the average stride representing thePW condition at GRADE0 and PTS100.

The average stride length was calculated as the product of thetreadmill speed with the average stride duration. The area underthe curve of the EMG signal of the normalized average stride wascomputed, using the trapezoid method, for each muscle (EOAUC,ESAUC, MUAUC, RAAUC) as measure of their EMG amplitude. The timeat which each muscle was active during each stride (EOACT, ESACT,MUACT, RAACT) has been calculated as the percentage of the strideduration at which the normalized signal was higher than thebaseline mean value plus 7 standard deviations [19]. Finally, thecoactivation time (CO-ACT) of flexors and extensors muscles of thespine was calculated as the percentage of the average strideduration at which at least one of the trunk extensors (ES, MU) andRA were active at the same time [20].

2.5. Statistical analysis

Generalized estimating equations were used to test the effectsof locomotion type, treadmill grade and speed on the calculatedparameters, as this approach does not require distributionalassumptions of the data [22]. For each dependent variable, robustsandwich standard errors were calculated and the model’sdistribution family, link function and working correlation matrix

L. Zoffoli et al. / Gait & Posture 46 (2016) 57–62 59

were chosen according to the nature of each variable and as thecombination minimizing the quasi-likelihood information criteri-on [22]. In addition, the effects of the different factors combina-tions were tested by pairwise contrasts with Holm correction formultiple comparisons. The significance level was set atp < 0.05. The data has been processed using MATLAB (ver. 8.5—MathWorks Inc., Natick, Massachusetts, USA), while the statisticalanalysis has been carried out using R (ver. 3.2.1—R Core Team,Vienna, Austria) and the geepack package [23].

3. Results

3.1. Walking vs. pole walking

Compared to W, PW increased EOAUC and EOACT (Fig. 1), as wellas RAAUC and RAACT (Fig. 2) at most grades and speeds. PW alsoincreased MUAUC at GRADE0 (all speeds), and MUACT at PTS80 andPTS100 (Fig. 3). Similar ESAUC were observed between W and PW,while PW resulted in longer ESACT than W, especially at PTS100

(Fig. 4). PW also increased CO-ACT (Fig. 5) at GRADE0 (PTS100 only)and at GRADE7 (all speeds) and resulted in longer strides than W ateach grade and speed combination (Table 1 in the Supplementarymaterial).

3.2. Effects of the treadmill grade

Increasing the treadmill grade did not change EOAUC, EOACT,RAAUC, RAACT and CO-ACT during neither PW nor W at any PTSpercentage (Table 2 in the Supplementary material). Higher gradesincreased both ESAUC and MUAUC during W (p < 0.01 at each speed)and PW (p < 0.01 at PTS60 only for MUAUC). Similarly, higher gradeswere associated with longer ESACT both during W and PW at PTS60

only (p < 0.01 and p < 0.05, respectively), and longer MUACT duringW (p < 0.01 for all speeds) and PW (p < 0.05 at PTS60 only).Additionally, higher grades reduced the stride length during W(p < 0.01 at PTS60 and p < 0.05 at both PTS80 and PTS100).

3.3. Effects of the treadmill speed

Compared to PTS60, PTS80 increased EOAUC, ESAUC, MUAUC andthe stride length both during W and PW at both grades(p < 0.01 for all the comparisons—Table 3 in the Supplementarymaterial). The same comparison increased also RAAUC during W(p < 0.05 at both grades), and RAACT at GRADE0 during PW

[(Fig._1)TD$FIG]

Fig. 1. Normalized EMG amplitude (EOAUC) and activity time (EOACT) of the external obliqu

Boxplots indicate the first and the third quartile of each sample, while its median value i

boxes margins for 1.5 times the inter-quartile range. Abbreviations: PTS60-80-100 = Perce

different from Walking[4_TD$DIFF]: [5_TD$DIFF]* = p < 0.05; ** [6_TD$DIFF]= p < 0.01.

(p < 0.05). In addition, EOACT exhibited greater values at PTS80 thanat PTS60 at both GRADE0 (p < 0.01 for both W and PW) and GRADE7

(p < 0.05 only during W).Elevated stride length, EOACT, RAACT, CO-ACT, EOAUC, ESAUC and

MUAUC were observed at PTS100, when compared to the otherspeeds, during both W and PW and at both grades (all p < 0.01).During PW, PTS100 increased also ESACT (p < 0.05 for the PTS60 vs.PTS100 comparisons at GRADE7 and p < 0.01 for the others), andMUACT at GRADE0 (p < 0.05 and p < 0.01 for the PTS60 vs. PTS100

and the PTS80 vs. PTS100 comparisons, respectively). RAAUC waslonger at PTS100 when compared to the other speeds at GRADE0

both during W and PW (p < 0.01 for all the comparisons), and atGRADE7 during PW (p < 0.05 for the PTS60 vs. PTS100 comparison).

4. Discussion

The aim of this study was to observe and compare the activity ofthe trunk muscles during W and PW at different treadmill speedsand grades. The main findings were: (a) the longer activation timeof the trunk muscles during PW as compared to W, which alsoresulted in longer CO-ACT; (b) the different trunk musclesresponse to PW at different speeds and grades.

4.1. Trunk muscle function in walking and pole walking

The spine extensors ES and MU showed similar amplitudesbetween W and PW, but were active for longer portions of the gaitcycle during the latter. This produced longer CO-ACT during PWeven without differences in the overall activation amplitude of thespine extensors. In line with previous findings [4,24], PW wasassociated with longer strides than W, lending support to thehigher hip joint angle found by Stief et al. [25] during nordicwalking. It can also be postulated that during PW, the alternatingpushing action of the poles, which are kept beside the body andinclined backwards, generates greater torque about the longitu-dinal axis of the body. This might assist the ipsilateral swinging ofthe lower limb through the action of the abdominal muscles.Indeed, in a previous study, Aruin and Latash [8] revealed that armswinging requires the intervention of several postural muscles tokeep the balance of the body. Accordingly, the greater activity ofthese muscles during PW would be necessary to keep the bodybalance and direction by counterbalancing the torque generated bythe poles that would otherwise rotate the upper trunk about thelongitudinal axis of the body. Furthermore, elevated CO-ACT has

e muscle during Walking and Pole walking at different treadmill speeds and grades.

s represented by the inner bold line. The upper and lower whiskers extend from the

ntage of the preferred transition [3_TD$DIFF]speed; GRADE0-7 = Treadmill grade. Significantly

[(Fig._2)TD$FIG]

Fig. 2. Normalized EMG amplitude (RAAUC) and activity time (RAACT) of the rectus abdominis muscle during Walking and Pole walking at different treadmill speeds and

grades. Boxplots indicate the first and the third quartile of each sample, while its median value is represented by the inner bold line. The upper and lower whiskers extend

from the boxes margins for 1.5 times the inter-quartile range.

Abbreviations: PTS60-80-100 = Percentage of the preferred transition speed [7_TD$DIFF]; GRADE0-7 = Treadmill grade. Significantly different from Walking [4_TD$DIFF]: [5_TD$DIFF]* = p < 0.05; [8_TD$DIFF]** = p < 0.01.

L. Zoffoli et al. / Gait & Posture 46 (2016) 57–6260

been previously related to higher spinal stability [5]. Thus, thelonger CO-ACT associated with PW, especially at the highestspeeds, would provide additional support to the hypothesis that,when compared to W, PW requires higher control of the trunkregion in order to offset the pole forces and the increased stridelength.

4.2. Walking and pole walking at different speeds and grades

Both EO and RA were mainly unresponsive to treadmill grademodifications, while ES and MU (the spine extensors) weresignificantly more active in uphill, rather than in level W and PW.In addition, and in line with previous findings [7,26,27], higherspeeds resulted in elevated trunk muscle activity both during Wand PW. However, while EOACT and RAACT increased markedly withthe speed increment, only small changes were observed in ESACT

and MUACT, especially at GRADE0. This resulted in similar CO-ACTvalues up to PTS80 between level W and PW, and in higher CO-ACTvalues in uphill PW than in uphill W at each PTS percentage.Furthermore, while ESAUC was not different between both level oruphill W and PW, MUAUC was higher in level PW as compared tolevel W, but not between uphill W and PW. Previous studies [28]suggested that the increased trunk tilt during uphill W helps the

[(Fig._3)TD$FIG]

Fig. 3. Normalized EMG amplitude (MUAUC) and activity time (MUACT) of the multifidus

Boxplots indicate the first and the third quartile of each sample, while its median value i

boxes margins for 1.5 times the inter-quartile range. Abbreviations: PTS60-80-100 = Perce

different from Walking [4_TD$DIFF]: [5_TD$DIFF]* = p < 0.05; [8_TD$DIFF]** = p < 0.01.

lower limbs in generating more momentum to counteract thegravity acceleration. Thus, uphill W would increase ESAUC andMUAUC, when compared to level W, to generate the higher forcesrequired to maintain the trunk frontally inclined. In addition,higher treadmill grades resulted in higher pole forces production[29]. ESAUC remained unchanged between level and uphill PW andthis could be explained by either the increased support offered bythe poles to the upper body, or by a less tilted trunk due to the useof the poles. This however requires further investigation.

4.3. Limitations

The use of the 30 Hz high-pass Butterworth filter during thedata processing is an established processing technique, howeverthe noise generated by the heart beat is not necessarily completelyremoved from the EMG signal [18]. Therefore, a robust thresholdwas set for the muscle activity onset detection, which could havereduced the activation time for the muscles with the weakestactivity. However this potential bias was consistent across allconditions.

Another possible limitation may be due to the use of amotorized treadmill. The limited size of the belt (200 � 75 cm)could have somehow restricted the technique of PW when

muscle during Walking and Pole walking at different treadmill speeds and grades.

s represented by the inner bold line. The upper and lower whiskers extend from the

ntage of the preferred transition speed [7_TD$DIFF]; GRADE0-7 = Treadmill grade. Significantly

[(Fig._4)TD$FIG]

Fig. 4. Normalized EMG amplitude (ESAUC) and activity time (ESACT) of the erector spinae longissimus muscle during Walking and Pole walking at different treadmill speeds

and grades. Boxplots indicate the first and the third quartile of each sample, while its median value is represented by the inner bold line. The upper and lower whiskers extend

from the boxes margins for 1.5 times the inter-quartile range. Abbreviations: PTS60-80-100 = Percentage of the preferred transition speed [7_TD$DIFF]; GRADE0-7 = Treadmill grade.

Significantly different from Walking[4_TD$DIFF]: [5_TD$DIFF]* = p < 0.05; [8_TD$DIFF]** = p < 0.01.

[(Fig._5)TD$FIG]

Fig. 5. Spine flexor (rectus abdominis) and extensors (erector spinae longissimus and multifidus) coactivation time (CO-ACT) during Walking and Pole walking at

different treadmill speeds and grades. Boxplots indicate the first and the third quartile of each sample, while its median value is represented by the inner bold line. The

upper and lower whiskers extend from the boxes margins for 1.5 times the inter-quartile range. Abbreviations: PTS60-80-100 = Percentage of the preferred transition speed [7_TD$DIFF];

GRADE0-7 = Treadmill grade. Significantly different from Walking [4_TD$DIFF]: [5_TD$DIFF]* = p < 0.05; [8_TD$DIFF]** = p < 0.01.

L. Zoffoli et al. / Gait & Posture 46 (2016) 57–62 61

compared to overground PW. Although this was unavoidable inorder to be able to control for speed and grade, the thoroughfamiliarization sessions proved that it was still adequate toperform the tasks requested in this study.

The choice of selecting individualized treadmill speeds for eachgrade as a percentage of the PTS resulted in slightly lower absolutespeeds at GRADE7 than GRADE0, which might have partiallyreduced the effect of the treadmill grade on some of theinvestigated parameters. However, this design was necessary toreplicate real-life conditions at which W and PW are performedwith grade and speed affecting each other.

5. Conclusions

The present findings suggest that the trunk muscle activity isresponsive to an increase in treadmill speed both in W and PW.Interestingly though, PW yields higher EMG amplitude than W inthe abdominal muscles at both GRADE0 and GRADE7 for about allthe speeds tested. Conversely, at the spine extensors level, PW isassociated with higher MU activation ampitude than W only atGRADE0 for each tested speed. PW is associated with longer strides,greater activation time of all trunk muscles and longer coactivation ofthe spine flexors and extensors suggesting that PW requires greatermuscle engagement and control of the trunk region than W. Thepresent investigation clarified the differences in the trunk muscleactivity between W and PW. Further studies could be undertaken inorder to gain an insight into the pattern of muscle activation withinthe gait cycle during PW. This would assist in understanding thepotential benefits related to the use of poles during W.

Acknowledgments

The authors wish to thank Will Evans, Gallin Montgomery,Stephen Hayes, James Bray (University of Hull) and Eugenio Grassi(University of Urbino Carlo Bo) for their technical assistance.

Conflict of interest statement

The authors declare that they have no conflict of interest anddid not receive any financial support.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.gaitpost.2016.02.015.

References

[1] Fritschi JO, Brown WJ, Laukkanen R, Uffelen JGZ. The effects of pole walking onhealth in adults: a systematic review. Scand J Med Sci Sports 2012;22:e70–8.

[2] Porcari JP, Hendrickson TL, Walter PR, Terry L, Walsko G. The physiologicalresponses to walking with and without power PolesTM on treadmill exercise.Res Q Exerc Sport 1997;68:161–6.

[3] Foissac MJ, Berthollet R, Seux J, Belli A, Millet GY. Effects of hiking pole inertiaon energy and muscular costs during uphill walking. Med Sci Sports Exerc2008;40:1117–25.

[4] Knight CA, Caldwell GE. Muscular and metabolic costs of uphill backpacking:are hiking poles beneficial. Med Sci Sports Exerc 2000;32:2093–101.

L. Zoffoli et al. / Gait & Posture 46 (2016) 57–6262

[5] Granata KP, Orishimo KF. Response of trunk muscle coactivation to changes inspinal stability. J Biomech 2001;34:1117–23.

[6] Bressel E, Dolny DG, Gibbons M. Trunk muscle activity during exercisesperformed on land and in water. Med Sci Sports Exerc 2011;43:1927–32.

[7] Anders C, Wagner H, Puta C, Grassme R, Petrovitch A, Scholle H-C. Trunkmuscle activation patterns during walking at different speeds. J ElectromyogrKinesiol 2007;17:245–52.

[8] Aruin AS, Latash M. Directional specificity of postural muscles in feed-forwardpostural reactions during fast voluntary arm movements. Exp Brain Res1995;103:323–32.

[9] Ceccato J-C, de Seze M, Azevedo C, Cazalets J-R. Comparison of trunk activityduring gait initiation and walking in humans. PLoS ONE 2009;4:e8193.

[10] Hreljac A. Effects of physical characteristics on the gait transition speed duringhuman locomotion. Hum Movement Sci 1995;14:205–16.

[11] Figard-Fabre H, Fabre N, Leonardi A, Schena F. Physiological and perceptualresponses to nordic walking in obese middle-aged women in comparison withthe normal walk. Eur J Appl Physiol 2010;108:1141–51.

[12] Sadeghi H, Allard P, Prince F, Labelle H. Symmetry and limb dominance in able-bodied gait: a review. Gait Posture 2000;12:34–45.

[13] Borg GAV. Psychophysical bases of perceived exertion. Med Sci Sports Exerc1982;14:377–81.

[14] Hansen EA, Smith G. Energy expenditure and comfort during nordic walkingwith different pole lengths. J Strength Cond Res 2009;23:1187–94.

[15] Vera-Garcia FJ, Moreside JM, McGill SM. MVC techniques to normalize trunkmuscle EMG in healthy women. J Electromyogr Kinesiol 2010;20:10–6.

[16] Hermens HJ, Freriks B, Merletti R, Stegeman D, Blok J, Rau G, et al. SENIAM8. European recommendation for surface electromyography. Enschede: Roes-singh Research and Development; 1999.

[17] Barbero M, Merletti R, Rainoldi A. Atlas of muscle innervation zones: under-standing surface electromyography and its applications. Milan: Springer; 2012.

[18] Drake JDM, Callaghan JP. Elimination of electrocardiogram contaminationfrom electromyogram signals: an evaluation of currently used removal tech-niques. J Electromyogr Kinesiol 2006;16:175–87.

[19] Li X, Zhou P, Aruin AS. Teager–Kaiser energy operation of surfaceEMG improves muscle activity onset detection. Ann Biomed Eng 2007;35:1532–8.

[20] Prosser LA, Lee SCK, VanSant AF, Barbe MF, Lauer RT. Trunk and hip muscleactivation patterns are different during walking in young children with andwithout cerebral palsy. Phys Ther 2010;90:986–97.

[21] Kavanagh JJ. Lower trunk motion and speed-dependence during walking. JNeuroEng Rehabil 2009;6:9.

[22] Cui J. QIC program and model selection in GEE analyses. Stata J 2007;7:209–20.[23] Halekoh U, Højsgaard S, Yan J. The R Package geepack for generalized estimat-

ing equations. J Stat Softw 2006;15:1–11.[24] Willson J, Torry MR, Decker MJ, Kernozek T, Steadman JR. Effects of walking

poles on lower extremity gait mechanics. Med Sci Sports Exerc 2001;33:142–7.

[25] Stief F, Kleindienst FI, Wiemeyer J, Wedel F, Campe S, Krabbe B. Inversedynamic analysis of the lower extremities during nordic walking, walking,and running. J Appl Biomech 2008;24:351–9.

[26] Saunders SW, Rath D, Hodges PW. Postural and respiratory activation of thetrunk muscles changes with mode and speed of locomotion. Gait Posture2004;20:280–90.

[27] Hof AL, Elzinga H, Grimmius W, Halbertsma JPK. Speed dependence of aver-aged EMG profiles in walking. Gait Posture 2002;16:78–86.

[28] Leroux A, Fung J, Barbeau H. Postural adaptation to walking on inclinedsurfaces: I. Normal strategies. Gait Posture 2002;15:64–74.

[29] Svoboda Z, Stejskal P, Jakubec A, Krecı J. Kinematical analysis, pole forces andenergy cost of nordic walking–slope influence. Acta Univ Palacki Olomuc FacMed 2011;41:27–34.