Article Review #1

945

www.IJSPP-Journal.com ORIGINAL INVESTIGATION

International Journal of Sports Physiology and Performance, 2014, 9, 945-952 http://dx.doi.org/10.1123/ijspp.2013-0418 © 2014 Human Kinetics, Inc.

Barrett is with the Dept of Sport, Health and Exercise Science, University of Hull, Kingston upon Hull, UK. Midgley is with the Dept of Sport and Physical Activity, Edge Hill University, Lancashire, UK. Lovell is with the School of Science and Health, University of Western Sydney, Penrith, Australia. Address author correspondence to Ric Lovell at R.Lovell@ uws.edu.au.

PlayerLoad™: Reliability, Convergent Validity, and Influence of Unit Position During Treadmill Running

Steve Barrett, Adrian Midgley, and Ric Lovell

Purpose: The study aimed to establish the test–retest reliability and convergent validity of PlayerLoad™ (triaxial-accelerometer data) during a standardized bout of treadmill running. Methods: Forty-four team-sport players performed 2 standardized incre- mental treadmill running tests (7–16 km/h) 7 d apart. Players’ oxygen uptake (VO2; n = 20), heart rate (n = 44), and triaxial- accelerometer data (PlayerLoad; n = 44) measured at both the scapulae and at the center of mass (COM), were recorded. Accel- erometer data from the individual component planes of PlayerLoad (anteroposterior [PLAP], mediolateral [PLML], and vertical [PLV]) were also examined. Results: Moderate to high test–retest reliability was observed for PlayerLoad and its individual planes (ICC .80–.97, CV 4.2–14.8%) at both unit locations. PlayerLoad was significantly higher at COM vs scapulae (223.4 ± 42.6 vs 185.5 ± 26.3 arbitrary units; P = .001). The percentage contributions of individual planes to PlayerLoad were higher for PLML at the COM (scapulae 20.4% ± 3.8%, COM 26.5% ± 4.9%; P = .001) but lower for PLV (scapulae 55.7% ± 5.3%, COM 49.5% ± 6.9%; P = .001). Between-subjects correlations between PlayerLoad and VO2, and between PlayerLoad and heart rate were trivial to moderate (r = –.43 to .33), whereas within-subject correlations were nearly perfect (r = .92–.98). Conclusions: PlayerLoad had a moderate to high degree of test–retest reliability and demonstrated convergent validity with measures of exercise intensity on an individual basis. However, caution should be applied in making between-athletes contrasts in loading and when using recordings from the scapulae to identify lower-limb movement patterns.

Keywords: accelerometry, heart rate, gas analysis, interindividual variation, treadmill exercise

External load refers to the totality of mechanical or locomotive stress generated by an individual when undertaking a bout of activ- ity. Quantification of external load in the sport and exercise-science domain is commonly undertaken using tracking technologies to prescribe and monitor the “output” of an individual’s training and competition schedules.1,2 Typically, external load has been deter- mined by measuring the distances covered in a variety of locomotor classifications3 or speed zones.4 This approach, however, is insensi- tive to the totality of mechanical stresses common to team-sport players, such as abrupt changes in running velocity, direction, and impacts.5,6 For this reason, manufacturers of global positioning systems (GPS) have incorporated high-resolution (100-Hz) triaxial accelerometers in their devices, which may be considered a more sensitive measure of the totality of mechanical stresses and external loading on the body.

Triaxial accelerometers have been commonly used in physical activity research as a proxy for energy expenditure in free-living conditions.7,8 Yet, the application of accelerometer data in the elite sporting domain is in its infancy and is limited by the technology’s inability to quantify the net external work performed in cycling, swimming, gradient running, or load-bearing activities. Neverthe- less, accelerometers have already been employed in a variety of team sports including Australian Rules Football,1,2,9,10 netball,11 bas- ketball,12 and soccer.13–15 The most common accelerometer-derived

variable adopted by sports practitioners and researchers is a vector magnitude termed PlayerLoad™. PlayerLoad is an arbitrary unit that is derived from 3-dimensional measures of the instantaneous rate of change of acceleration. Its utility as a training-load marker has been established against criterion measures of both external load (distances covered15) and internal load (heart rate, ratings of perceived exertion2,13,15) in training settings. However, to date there have been few controlled laboratory studies examining the test–retest reliability or convergent validity of PlayerLoad using concurrent recordings of internal load as criterion measures of exercise intensity.16

Boyd et al17 have demonstrated the within- (0.91–1.05% coef- ficient of variation [CV]) and between- (1.02–1.10% CV) units reliability of PlayerLoad using a mechanical shaker and determined its between-units reliability in Australian Rules Football matches (1.90% CV). In field settings it is considered good practice to assign the same unit to a player for repeated external-load assessments, to avoid any between-units bias. However the test–retest reliability of PlayerLoad has not been determined during human locomotion, and the absence of this information makes it difficult for scientists and coaches to interpret meaningful differences in players’ external loading between training stimuli. Nonetheless, some recent studies have examined triaxial-accelerometer data in more detail, by distin- guishing the individual planar contributions to PlayerLoad recorded during competitive team-sport matches.9,11 For example, reduced loading in the vertical loading plane has been reported in elite Aus- tralian Football League players with neuromuscular fatigue9 and in netball players competing at a lower participation standard.11 Given the high coefficient of variation in match-to-match distances covered by team-sport players, particularly at higher running speeds,4,18,19 a laboratory controlled study examining the test–retest reliability of PlayerLoad and its individual planes may be considered overdue.

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

946 Barrett, Midgley, and Lovell

The contribution of loading from individual planes is also likely to be influenced by the anatomical position of the accelerometer. While it is generally accepted that the center of mass is the optimal anatomical location for these devices,7,20 there are often exceptions in the literature.1,2,9,11,13,15,21 With regard to team sports and PlayerLoad, the accelerometers are encompassed in GPS units that are harnessed between the scapulae to enhance its positioning signal. However, to date there is no research available examining the effect of unit location on PlayerLoad or its 3 contributing planes. Information of this nature is likely to have implications for practitioners using PlayerLoad as an athlete-monitoring tool or for those using this technology to identify alterations in movement patterns, intensity, and efficiency. It is also currently unclear if the anatomical location of the accelerometer influences the relationship between PlayerLoad and internal-load measures. Therefore, the aims of this study were to establish the test–retest reliability of triaxial-accelerometer data during a standardized bout of treadmill running, examine the effect of accelerometer location on PlayerLoad data, and investigate the convergent validity of PlayerLoad using heart rate and rate of pulmo- nary oxygen uptake (VO2) as criterion measures of exercise intensity.

Method

Subjects

Forty-four semiprofessional (n = 12) and university-level (n = 32) team-sport players volunteered to participate in the study (age 22 ± 3 y, height 1.80 ± 0.06 m, body mass 78.9 ± 8.6 kg). Atkinson and Nevill22 deemed this sample size adequate for test–retest reliability variable estimation. The study gained ethical approval from the departmental ethics committee before its commencement and conformed to the World Medical Association’s ethical code (Declaration of Helsinki). Participants were informed of the risks and discomforts associated with maximal testing and provided written informed consent.

Experimental Design

Participants were required to visit the laboratory on 3 separate occa- sions. During each visit, the same incremental treadmill test was administered. All tests were separated by 7 days and performed at the same time on each day to attenuate circadian variation. The first laboratory visit was deemed a familiarization trial and consisted of participants running up to a speed of 14 km/h to familiarize them with the experimental equipment and procedures. The test was restricted to 14 km/h to reduce the burden of participation with respect to time constraints and the need to repeatedly perform maximal exercise tests. All participants were instructed to follow the same 6-day exercise regimen before both subsequent visits, with replication of dietary intake in the 24 hours before testing. After a 5-minute warm-up at a self-selected pace on an exercise bike (Wattbike, Wattbike Ltd, Nottingham, UK), participants then performed an incremental treadmill test, starting at a speed of 7 km/h, on an incline of 1% gradient.23 After 4 minutes of running at 7 km/h to accustom the participant with the experimental configura- tion, the treadmill speed was increased by 0.1 km/h every 6 seconds (equivalent to 1-km/h increments each minute) until participants reached their limit of exercise tolerance. During the test participants’ expired gases (VO2; Cortex gas analyzer, Cortex Biophysic, Leipzig, Germany), heart rate (Polar T31, Polar Electro, Oulu, Finland), and PlayerLoad data (MinimaxX S4, Catapult Sports, Melbourne, Australia) were collected throughout and averaged over 1-minute periods. The initial 4 minutes of data were discarded and classified

as the warm-up period that preceded the incremental part of the test protocol. Data from 1-minute exercise stages that were not com- pleted were also discarded. Due to the within- and between-subjects variance in the maximal treadmill running speed, we discarded data above a running speed of 16 km/h. This process also ensured that sufficient statistical power was attained for reliability analyses (n ≥ 20;22) on triaxial-accelerometer data.

PlayerLoad and MinimaxX Accelerometer

The MinimaxX (Catapult Innovations, Scoresby, Victoria) contains a triaxial piezoelectric linear accelerometer (Kionix: KXP94) that samples at a frequency of 100 Hz, as part of other inertial sensors in the micromechanical system. The accelerometer was positioned between the scapulae, as recommended by the manufacturer, in a neoprene undergarment that accommodates the unit in an integrated pouch. A pouch with the same specifications was also fitted at the bottom of the garment. To approximate the center of mass (COM), the pouch was positioned at the intersection of the axial and sagittal planes in line with the iliac crest on the posterior side of the body.20 For the purposes of our study, we assumed the COM unit location to be the criterion for triaxial-accelerometer recordings. Combined triaxial-accelerometer data were presented as PlayerLoad, which is a modified vector magnitude expressed as the square root of the sum of the squared instantaneous rates of change in acceleration in each of the 3 planes divided by 100.17 The vector magnitude of Play- erLoad (PLV-MAG) and individual component planes of PlayerLoad (anteroposterior [PLAP], mediolateral [PLML], and vertical [PLV]) were recorded. Expressed in arbitrary units (au), PlayerLoad data were recorded throughout the treadmill test using Catapult Sprint software (Version 5.0.9.2; Firmware 6.75) and retrieved posttesting.

Gas Analysis and Heart Rate

In a subset of participants (n = 20), we also examined the convergent validity of triaxial-accelerometer data by measuring participants’ cardiovascular and metabolic response to the treadmill protocol. The gas-analysis system was calibrated according to the manufacturer’s instructions using ambient air and a standard gas mixture (16.97% O2, 5.05% CO2, balanced with N2). The turbine flowmeter to determine minute ventilation was calibrated with a 3-L syringe. Breath-by-breath VO2 data were time averaged over 1-minute periods. Participants’ relative VO2 values were allometrically scaled (body mass raised to the power of 0.75) to minimize the bias according to body size. Heart rate was recorded throughout each trial via the Catapult Sprint software (Sprint 5.0.9, Catapult Innovations, Melbourne, Australia), averaged over 1-minute periods, and expressed as a percentage of the individual’s maximum heart rate attained during the protocol (HRAVG).

Statistical Analysis

Test–retest reliability of PlayerLoad variables at the 2 different anatomical locations (COM and scapulae) were reported as the intraclass correlation coefficient (ICC; 2-way random model for absolute agreement) and the within-subject coefficient of variation. We adopted the criteria of Vincent24 to interpret the ICC coefficients, in which values greater than .90 were considered high, from .80 to .89 moderate, and below .80 questionable. Coefficient of variation was calculated by dividing the standard deviation of the between- trials differences by the square root of 2, and subsequently dividing this result by the grand mean derived from both trials. A 2-way repeated-measures ANOVA was used to assess differences in PLV- MAG, PLAP, PLML, and PLV according to increasing treadmill speed

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

PlayerLoad™ During Treadmill Running 947

and unit location. When the sphericity assumption was violated a Huynh-Feldt correction was applied to the degrees of freedom. Post hoc pairwise comparisons, with Sidak-adjusted P values, were con- ducted in the event of a statistically significant F ratio. Relationships between HRAVG and allometrically scaled VO2 versus PLV-MAG were assessed using Pearson product–moment correlation. Correlations were calculated to assess these relationships between subjects, whereas minute-to-minute data were used to derive within-subject correlations as a measure of convergent validity. Magnitude of the correlation coefficients was deemed trivial for r2 < .1, small for .1 < r2 < .3, moderate for .3 < r2 < .5, large for .5 < r2 < .7, very large for .7 < r2 < .9, nearly perfect for r2 > .9, and perfect for r2 = 1.25 Analyses were completed using IBM SPSS Statistics for Windows software (release 20; SPSS Inc, Chicago, IL, USA). Two-tailed statistical significance was accepted as P < .05.

Results PlayerLoad Test–Retest Reliability Trial total values for PLV-MAG, PLAP, PLML, and PLV are shown in Table 1. PLV-MAG, PLAP, and PLV demonstrated high reliability between repeated trials when recorded at both the scapulae and the COM (see Table 2). Moderate test–retest reliability was observed at both anatomi- cal locations for PLML. The PLV-MAG data, expressed on a minute-by- minute basis with increasing treadmill speeds, are presented in Figure 1, and reliability data for each treadmill speed are shown in Table 2.

Unit Location

PLV-MAG at the scapulae was 15.7% ± 9.7% less than the PLV-MAG observed at the COM (P = .001). Absolute loadings in PLAP, PLML, and PLV were 14.7% ± 22.2%, 35.0% ± 20.3%, and 7.9% ± 14.6% higher at the COM, respectively (P = .001). PLV-MAG, PLAP, PLML, and PLV as measured at the scapulae increased each minute with increasing treadmill speed (see Figure 2). In contrast, COM-derived recordings of PLV-MAG and PLML did not increase statistically at treadmill speeds above 13 km/h, with significant increases in PLV at the COM not observed beyond 11 km/h.

The percentage loading contributions from PLAP (scapulae 22.9% ± 3.8%, COM 23.1% ± 4.2%), PLML (scapulae 20.4% ± 3.8%,

COM 26.5% ± 4.9%) and PLV (scapulae 55.7% ± 5.3%, COM 49.5% ± 6.9%) did not change with increasing running speed; however, total protocol values were higher and lower at the COM versus the scapulae in PLML and PLV, respectively (P = .001; see Figure 3).

Convergent Validity

Table 3 contains the between- and within-subject correlation coeffi- cients for PlayerLoad and heart rate, as well as PlayerLoad and VO2. Between-subjects correlations were trivial to moderate, whereas within-subject correlations were nearly perfect.

Discussion In the current study we sought to determine the test–retest reliability of triaxial-accelerometer data during incremental treadmill running. As secondary aims, we also examined the effect of unit placement on the accrued accelerometer data and their reliability, together with relationships to physiological indices of exercise intensity. Several key findings were observed: (1) The PlayerLoad vector magnitude and its individual planar components demonstrated moderate to high test–retest reliability at both the scapulae (ICC .80–.93, CV 5.3–14.8%) and the COM (ICC .87–.97, CV 4.2–11.5%); (2) moderate to high test–retest reliability was demonstrated across treadmill speeds, with the exception of PLML as measured at the scapulae, which was questionable (Table 2); (3) a moderate to large degree of interindividual variation was observed in all accelerom- eter variables, irrespective of unit location; (4) placement at the scapulae underestimated PlayerLoad by 15.7% ± 9.7% and also increased the vertical-plane percentage contribution to loading; (5) scapulae-placed units did not detect the expected changes in vertical and mediolateral loading with increasing running speeds; (6) and within-subject correlations between PlayerLoad and heart rate/VO2 were nearly perfect at both the scapulae and the COM, whereas (7) trivial to moderate associations were observed for the between-subjects correlations of the same variables.

In terms of reliability, our data extend the existing literature in which high between-units reliability, both in a mechanical shaker and also when 2 units were applied to players during Aus- tralian Football League matches, was demonstrated.17 While the

Table 1 Mean ± SD and Between-Trials Statistics for Triaxial-Accelerometer Data Collected at the Scapulae and the Center of Mass During a Repeated Incremental Treadmill Running Test at Speeds of 7 to 16 km/h

Anatomical location Trial 1 Trial 2 Xdiff SDdiff P

Scapulae PLV-MAG 184.7 ± 25.2 186.4 ± 27.6 0.76 4.05 .943

PLAP 59.2 ± 13.0 60.1 ± 16.5 0.82 4.63 .333

PLML 52.9 ± 12.0 53.1 ± 14.7 0.02 5.04 .615

PLV 143.6 ± 21.7 145.0 ± 23.2 0.53 6.44 .824

Center of mass PLV-MAG 224.1 ± 43.8 222.8 ± 41.9 1.27 5.79 .921

PLAP 72.3 ± 16.0 71.8 ± 16.1 0.49 3.50 .459

PLML 85.0 ± 18.5 84.5 ± 19.4 0.49 6.93 .734

PLV 161.7 ± 39.6 159.9 ± 35.9 1.78 8.38 .529

Abbreviations: PLV-MAG, PlayerLoad™ accumulated (vector magnitude drawn from triaxial-accelerometer data); PLAP, PlayerLoad in anteroposterior plane; PLML, PlayerLoad in mediolateral plane; PLV, PlayerLoad in vertical plane; Xdiff, mean difference between trials; SDdiff, standard deviation of the differences for repeated measures.

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

948

Table 2 Test–Retest Reliability Statistics for Triaxial-Accelerometer Data Collected at the Scapulae and the Center of Mass During a Repeated Incremental Treadmill Running Test

Speed, km/h

Reliability statistic Anatomical location 7 8 9 10 11 12 13 14 15 16 Total

ICC Scapulae PLV-MAG .93 .87 .86 .91 .92 .90 .89 .91 .93 .97 .93

PLAP .87 .87 .86 .92 .92 .92 .93 .93 .88 .78 .92

PLML .72 .72 .72 .74 .79 .69 .72 .69 .64 .60 .80

PLV .93 .87 .81 .85 .88 .82 .82 .83 .87 .83 .93

Center of mass PLV-MAG .94 .95 .96 .97 .96 .96 .97 .97 .97 .96 .97

PLAP .84 .88 .92 .88 .89 .88 .91 .92 .96 .96 .94

PLML .92 .87 .87 .85 .88 .82 .80 .82 .65 .83 .87

PLV .95 .93 .94 .95 .95 .93 .94 .93 .90 .94 .95

CV (%) Scapulae PLV-MAG 12.6 13.1 10.1 8.4 5.9 5.2 5.7 5.2 4.8 4.6 5.9

PLAP 17.0 18.2 15.5 15.1 11.4 11.2 10.4 9.5 8.6 9.6 9.1

PLML 17.5 16.3 13.0 17.3 15.0 15.1 14.2 10.0 3.0 3.5 12.0

PLV 14.1 13.7 11.7 10.0 7.9 6.9 7.7 7.4 7.0 6.9 6.3

Center of mass PLV-MAG 8.3 9.4 7.9 5.9 5.6 5.2 5.3 4.6 4.5 3.6 5.2

PLAP 12.4 14.7 12.6 9.9 11.4 10.2 10.4 9.1 8.8 8.6 7.5

PLML 12.2 14.9 14.5 14.8 12.0 14.4 13.3 12.6 13.0 14.6 11.4

PLV 11.3 10.2 8.6 7.5 7.3 8.5 7.9 8.2 8.2 10.9 7.3

Abbreviations: PLV-MAG, PlayerLoad™ accumulated (vector magnitude drawn from triaxial-accelerometer data); PLAP, PlayerLoad in anteroposterior plane; PLML, Play- erLoad in mediolateral plane; PLV, PlayerLoad in vertical plane; CV, coefficient of variation; ICC, intraclass correlation coefficient.

Figure 1 — Vector magnitude of PlayerLoad™ at the scapulae and center of mass (COM) recorded during a repeated incremental treadmill running test at speeds of 7 to 16 km/h. Differences between treadmill speeds are significant unless otherwise stated. nsNonsignificant difference from 16 km/h (P > .05). *Significant difference between COM and scapulae (P < .01). N = 44 at all treadmill speeds unless otherwise stated.

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

PlayerLoad™ During Treadmill Running 949

Figure 2 — Individual component planes for PlayerLoad™ as an absolute value at the scapulae and the center of mass (COM). The graphs represent (A) PlayerLoad in anteroposterior plane, (B) PlayerLoad in mediolateral plane, and (C) PlayerLoad in vertical plane. Differences between treadmill speeds are significant unless otherwise stated. nsNonsignificant difference compared with 16 km/h (P > .05), ~Nonsignificant difference compared with 15 km/h (P > .05), #Nonsignificant difference compared with 14 km/h (P > .05). *Significant difference between COM and scapulae. N = 44 at all treadmill speeds unless otherwise stated.

between-units reliability may be high, in applied practice athletes routinely wear the same GPS devices to avoid any bias, except in some circumstances where units are not available for each athlete. Hence, for the first time, we examined the test–retest reliability of triaxial-accelerometer variables collected from the same units in team-sport players undertaking a repeated incremental treadmill test. We reasoned that information of this nature was necessary to facilitate interpretations of external-load data derived from high- resolution accelerometer devices. We observed high ICCs and low typical errors (as a % CV) for PLV-MAG at both the scapulae and at the COM. High levels of agreement were also demonstrated in the anteroposterior and vertical accelerometer planes, with moderate

agreement observed in mediolateral loading, when taken as a test- total measurement. However, there was moderate to large dispersion of the individual participant values; albeit this tended to be lower at the scapulae for both the PLV-MAG (SD 27.6 vs 41.9 au) and PLV (SD 23.2 vs 35.9 au). The typical error values we report for PLV-MAG are marginally higher than those reported recently by Portas et al16 in a case-study design. Taken together, the results of these 2 studies suggest that the test–retest reliability of PlayerLoad is acceptable. Smallest worthwhile changes (between-subjects SD multiplied by 0.2;26) in PLV-MAG were 2.8% and 3.8% for scapulae and COM positions, respectively. However, given the large interindividual differences in the magnitude of loading accrued during treadmill

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

950 Barrett, Midgley, and Lovell

Figure 3 — Individual component planes for PlayerLoad™ as a percentage contribution to PlayerLoad accumulated (vector magnitude drawn from triaxial accelerometer data) at the scapulae and center of mass (COM). The graphs represent (A) PlayerLoad in anteroposterior plane, (B) PlayerLoad in mediolateral plane, and (C) PlayerLoad in vertical plane. No differences were observed with increasing treadmill speeds. *Significant difference between COM and scapulae. N = 44 at all treadmill speeds unless otherwise stated.

running, we would urge caution when contrasting PlayerLoad data recorded from different athletes and recommend that practitioners perform their own reliability analyses on their athletes on an indi- vidual basis to detect meaningful changes in loading.

Placing the GPS device at the scapulae is recommended by device manufacturers to enhance the positioning signal triangulated by satellites orbiting the earth. Although the use of accelerometers in athletic settings is comparatively embryonic, in physical activity settings the COM is considered the criterion location for this mea- surement.7 This is because vector magnitudes representing overall dynamic body acceleration, such as PlayerLoad, are intended to be a measure of acceleration around the COM. With incremental treadmill running, PLV-MAG was systematically lower when recorded

at the scapulae versus the COM. Moreover, a change in the orien- tation of the device is likely to alter the accelerations recorded in the individual axes and may add a degree of error to estimates of energy expenditure.7 In this study, we observed differences in the relative contributions of PLML and PLV to PLV-MAG as a result of unit positioning, with recordings at the scapulae underestimating PLML and overestimating PLV contributions. In absolute terms, decreased accelerations in PLML (35.0% ± 20.3%) explained most of the underestimation in PLV-MAG when the unit was placed at the scapulae, which may reflect the influence of hip rotation when units are positioned at the COM. The overestimation of vertical-plane loading at the scapulae may be caused by the vertical component of trunk rotation or shoulder-girdle movement with arm swing or due

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

PlayerLoad™ During Treadmill Running 951

Table 3 Between- and Within-Subject Pearson Correlation Coefficients for PlayerLoad™ and HRAVG, and PlayerLoad and VO2, Recorded During a Repeated Incremental Treadmill Running Test at Speeds of 7 to 16 km/h

Between-Subjects Correlations Within-Subject Correlations

HRAVG VO2 HRAVG VO2

Anatomical location Trial 1 Trial 2 Trial 1 Trial 2 Trial 1 Trial 2 Trial 1 Trial2

Scapulae PLV-MAG –.32 –.20 .12 .31 .98 .98 .96 .96

PLAP –.22 –.43 .14 .33 .94 .94 .93 .93

PLML .16 .03 .29 .29 .93 .93 .93 .93

PLV –.38 –.17 –.02 .24 .93 .93 .92 .92

Center of mass PLV-MAG –.20 .09 –.03 .09 .98 .98 .96 .96

PLAP –.28 –.20 –.19 –.02 .97 .97 .97 .97

PLML –.22 .05 –.02 .13 .94 .94 .95 .95

PLV –.11 .16 .02 .00 .93 .93 .94 .94

Abbreviations: PLV–MAG, PlayerLoad accumulated (vector magnitude drawn from triaxial-accelerometer data); PLAP, PlayerLoad in anteroposterior plane; PLML, Player- Load in mediolateral plane; PLV, PlayerLoad in vertical plane; HRAVG, heart rate expressed as a percentage of the individual’s maximum; VO2, rate of pulmonary oxygen uptake allometrically scaled to body mass.

Note: Within-subject correlation coefficients are reported as the group mean.

to greater vertical displacement associated with the trunk flexion during running. Further research is warranted to examine the impact of upper-body running kinematics on PlayerLoad as recorded at the scapulae, the effect of which may be exacerbated at higher running speeds than those examined in this study.

Individual planar contributions to PlayerLoad may also be expected to vary with increasing treadmill speed. Vertical accelera- tions are the largest contribution during straight-line running activi- ties as a result of the ground-reaction forces,27 and these increase in a linear fashion with running speed until approximately 14 to 16 km/h, where a plateau has been observed in young athletic male populations.20,21 Above these speeds greater contributions have been identified from the anteroposterior axis due to biomechanical altera- tions in kinematics21 and kinetics.28,29 We observed patterns in the 3-dimensional accelerometer profile with increasing running speed similar to those reported in the literature20,21 when the unit was worn at the COM. At this location, significant increases in PLV-MAG and PLML with increasing running speed were not present above 13 km/h and above 11 km/h at PLV. Alternatively, when placed at the scapulae, the PlayerLoad indices continued to increase in a linear fashion as a function of increasing running speed. The reasons for this disparity in individual planar-loading patterns to the different unit locations may be due to the comparatively larger degree of between-subjects variance observed at the COM or an artifact due to unit movement within the undergarment. Alternatively, it may reflect the influence of arm swing and trunk rotation on PLML and PLV when measured at the scapulae, which may render this unit location insensitive to the typical adjustments in lower-limb kinematics and kinetics with increasing running speeds. Furthermore, the reliability of PLML on a minute-to-minute basis with increasing treadmill speeds was questionable when measured at the scapulae (Table 2). Hence, we would advise caution when inferring alterations in lower-limb run- ning technique or movement strategy from group-based changes in PlayerLoad as measured at the scapulae during match-play settings.

The utility of PlayerLoad as a tool to monitor training load in athletic populations has been examined in a number of studies.2,15,16 The convergent validity of PlayerLoad has been demonstrated previ-

ously with strong between-subjects relationships reported between global measures of internal (heart rate, ratings of perceived exertion) and external loading (distances covered) collected during a series of team-sport training sessions.2,13 In contrast, we observed only trivial to moderate between-subjects correlations between PlayerLoad versus heart rate and VO2 during treadmill running. Thus, while accelerom- eters may improve our estimation of internal load in comparison with traditional athlete-tracking technologies, they should not be used as a surrogate measure for internal load, particularly for between-subjects comparisons. In contrast, very strong to nearly perfect within-individ- ual correlations for PlayerLoad and physiological indices of exercise intensity were observed in accordance with previous literature.16 Since concomitant monitoring of internal or external load is recommended to detect the subtle variations in individual athletes’ training status or efficiency,2 rather than for between-athletes purposes, this within- subjects method of establishing the convergent validity of PlayerLoad is more appropriate. This is because individual running kinematics such as stride rate have a profound influence on PlayerLoad due to the large contribution of ground-reaction forces to triaxial-accelerometer data.27 Hence, we would recommend that PlayerLoad be used exclu- sively as an external-load measure and that data be treated in an indi- vidual-specific manner to account for differences in athlete running gait.

Practical Applications PlayerLoad measured from triaxial accelerometers in GPS devices can be used reliably to examine differences in individual athletes’ external loading. However, PlayerLoad data collected by a unit positioned between the shoulder blades may be influenced by the individual’s upper-body movements during locomotion and makes comparison between athletes difficult. Based on the current data, we would recommend that practitioners use PlayerLoad data to track the intensity and volume of external loading of individual athletes on a longitudinal basis and avoid comparisons between athletes due the influence of running style. Further laboratory controlled research is necessary to establish the determinants of PlayerLoad during human

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC

952 Barrett, Midgley, and Lovell

locomotion to assist field-based practitioners in their interpretation of triaxial-accelerometer data collected in the sporting domain.

Conclusions In summary, we found PlayerLoad to have a moderate to high degree of test–retest reliability, which was not affected by unit loca- tion. Wearing the unit at the scapulae underestimated PlayerLoad and was not sensitive to the subtle alterations expected in running kinematics and kinetics with increasing running speed. Nonethe- less, the convergent validity of PlayerLoad was demonstrated here with nearly perfect within-subject correlations with measures of internal load, and this too was not influenced by unit positioning. Hence, wearing the unit between the scapulae does not affect the test–retest reliability and convergent validity of PlayerLoad, so this unit placement is appropriate for use in exercise settings. However, the large variance we observed in PlayerLoad variables suggests that caution should be used in making between-athletes contrasts in loading and that any external differences may not reflect between- subjects differences in internal load.

Acknowledgments

The authors would like to thank Chris Towlson, Marcus Godbold, and Matthew Eastwood for their assistance with data collection and extend thanks to Dr Peter Clothier for his insightful comments on our manuscript.

References 1. Boyd LJ, Aughey R, Ball K. Quantifying external load in Australian

Football matches and training using accelerometers. Int J Sports Physiol Perform. 2013;8:44–51. PubMed

2. Scott TJ, Black CR, Quinn J, Coutts AJ. Validity and reliability of the session-RPE method for quantifying training in Australian football: a comparison of the CR10 and CR100 scales. J Strength Cond Res. 2013b;27(1):270–276. PubMed doi:10.1519/JSC.0b013e3182541d2e

3. Reilly T, Thomas V. A motion analysis of work-rate in different positional roles in professional football match-play. J Hum Mov. 1976;2:87–97.

4. Rampinini E, Coutts AJ, Castagna C, Sassi R, Impellizzeri FM. Variation in top level soccer match performance. Int J Sports Med. 2007;28(12):1018–1024. PubMed doi:10.1055/s-2007-965158

5. Osgnach C, Poser S, Bernardini R, Rinaldo R, di Prampero PE. Energy cost and metabolic power in elite soccer: a new match analy- sis approach. Med Sci Sports Exerc. 2010;42(1):170–178. PubMed doi:10.1249/MSS.0b013e3181ae5cfd

6. Varley MC, Aughey R. Acceleration profiles in elite Australian soccer. Int J Sports Med. 2013;34(1):34–39. PubMed

7. Halsey LG, Shepard EL, Wilson RP. Assessing the development and application of the accelerometry technique for estimating energy expenditure. Comp Biochem Physiol A Mol Integr Physiol. 2011;158(3):305–314. PubMed doi:10.1016/j.cbpa.2010.09.002

8. Howe CA, Staudenmayer JW, Freedson PS. Accelerometer prediction of energy expenditure: vector magnitude versus vertical axis. Med Sci Sports Exerc. 2009;41(12):2199–2206. PubMed doi:10.1249/ MSS.0b013e3181aa3a0e

9. Cormack SJ, Mooney MM, Morgan W, McGuigan MR. Influence of neuromuscular fatigue on accelerometer load in elite Australian football players. Int J Sports Physiol Perform. 2013a;8(4):373–378. PubMed

10. Mooney MG, Cormack S, O’Brien BJ, Morgan WM, McGuigan M. Impact of neuromuscular fatigue on match exercise intensity and performance in elite Australian football. J Strength Cond Res. 2013;27(1):166–173. PubMed doi:10.1519/JSC.0b013e3182514683

11. Cormack SJ, Smith RL, Mooney MM, Young WB, O’Brien BJ. Accelerometer load as a measure of physical activity profile in dif- ferent standards of netball match play. Int J Sports Physiol Perform. 2014;9(2):283–291 press.

12. Montgomery PG, Pyne DP, Minahan CL. The physical and physi- ological demands of basketball training and competition. Int J Sports Physiol Perform. 2010;5:75–86. PubMed

13. Casamichana D, Castellano J, Calleja-Gonzalez J, San Roma J, Castagna C. Relationship between indicators of training load within soccer players. J Strength Cond Res. 2013;27(2):369–374. PubMed doi:10.1519/JSC.0b013e3182548af1

14. Castellano J, Casamichana D, Dellal A. Influence of game format and number of players on heart rate responses and physical demands in small sided soccer games. J Strength Cond Res. 2013;27(5):1295– 1303. PubMed doi:10.1519/JSC.0b013e318267a5d1

15. Scott BR, Lockie RG, Knight TJ, Clark AC, Janse de Jonge AK. A comparison of methods to quantify the in-season training load of pro- fessional soccer players. Int J Sports Physiol Perform. 2013;8(2):195– 202. PubMed

16. Portas M, Taylor J, Weston M. Relationship between accelerometry derived external load and internal load during soccer-specific activity. In: World Congress of Science and Soccer Abstract Book. 2012:133.

17. Boyd LJ, Ball K, Aughey RJ. The reliability of MinimaxX acceler- ometers for measuring physical activity in Australian football. Int J Sports Physiol Perform. 2011;6(3):311–321.

18. Gregson W, Drust B, Atkinson G, Di Salvo V. Match–match variability of high-speed activities in premier league soccer. Int J Sports Med. 2010;31(4):237–242. PubMed doi:10.1055/s-0030-1247546

19. Kempton T, Sirotic AC, Coutts AJ. Between match variation in profes- sional rugby league competition. J Sci Med Sport. 2014;17(4):404– 407. PubMed doi:10.1016/j.jsams.2013.05.006

20. McGregor SJ, Busa MJ, Yaggie JA, Bolt EM. High resolution MEMS accelerometers to estimate VO2 and compare running mechanics between highly trained inter-collegiate and untrained runners. PLoS ONE. 2009;4(10):e7355. PubMed doi:10.1371/journal. pone.0007355

21. Fudge BW, Wilson J, Easton C, et al. Estimation of oxygen uptake during fast running using accelerometry and heart rate. Med Sci Sports Exerc. 2007;39:192–198. PubMed

22. Atkinson G, Nevill AM. Statistical methods for assessing measurement error (reliability) in variables related to sports medicine. Sports Med. 1998;26(4):217–238. PubMed doi:10.2165/00007256-199826040- 00002

23. Jones AM, Doust JH. A 1% treadmill grade most accurately reflects the energetic cost of outdoor running. J Sports Sci. 1996;14(4):321–327. PubMed doi:10.1080/02640419608727717

24. Vincent WJ. Statistics in Kinesiology. 3rd ed. Champaign, IL: Human Kinetics; 1993.

25. Hopkins WG, Marshall SW, Batterham AM, Hanin J. Progres- sive statistics for studies in sports medicine and exercise science. Med Sci Sports Exerc. 2009;41:3–13. PubMed doi:10.1249/ MSS.0b013e31818cb278

26. Hopkins WG. How to interpret changes in an athletic performance test. SportScience. 2004;8(1):1–7.

27. Nigg BM. The role of impact forces and foot pronation: a new paradigm. Clin J Sport Med. 2001;11(1):2–9. PubMed doi:10.1097/00042752- 200101000-00002

28. Brage S, Brage N, Franks PW, et al. Branched equation modeling of simultaneous accelerometry and heart rate monitoring improves estimate of directly measured physical activity energy expenditure. J Appl Physiol. 2004;96(1):343–351. PubMed doi:10.1152/jap- plphysiol.00703.2003

29. Cavagna GA, Franzetti P, Fuchimoto T. The sources of external work in level walking and running. J Physiol. 1976;262(3):639–657. PubMed

Brought to you by CENTRAL MICHIGAN UNIVERSITY | Unauthenticated | Downloaded 09/02/21 06:45 PM UTC