Reliability of Intermittent Sprint Cycling

Title:

Peak power output provides the most reliable measure of performance in prolonged intermittent sprint cycling.

Running Title

Reliability of intermittent sprint cycling

Keywords

Repeated sprint ability, team sport players, reliability.

Authors

Mark Hayesa, Drew Smitha, Paul C. Castlebc, Peter Watta, Emma Z. Rossa, Neil S. Maxwella.

Institutional Affiliation

aSchool of Sport and Service Management, University of Brighton, Hillbrow, Denton Road, Eastbourne, BN20 7SR.

bGlaxoSmithKline, GSK House, 980 Great West Road, Brentford, Middlesex, TW8 9GS, United Kingdom.

cDepartment of Sport and Exercise Science, University of Bedfordshire, Bedford, United

Kingdom.

Corresponding author:

Mark Hayes

School of Sport and Service Management, University of Brighton, Hillbrow, Denton Road, Eastbourne, BN20 7SR.

01273 644730

Abstract

The aims of this study were to determine the reliability of an intermittent-sprint cycling protocol and to determine the efficacy of one practice session on main trials. Eleven men, moderately trained team-sport athletes, completed three visits to the laboratory involving a graded-exercise test and practice session and two trials of a Cycling Intermittent-Sprint Protocol separated by three days. Data for practice and main trials were analysed using typical error of measurement, intra-class correlation and least-products regression to determine reliability. Typical error of measurement (expressed as a coefficient of variation) and intra-class correlation for peak power output from all twenty sprints for trial 1 and trial 2 were 2.9 ±12.8% (95% confidence interval: 2.0 – 5.0%) and 0.96 (95% confidence interval: 0.85 – 0.99), respectively. Typical errors of measurement and intra-class correlation for mean power output for all twenty sprints for trials 1 and 2 were 4.2 ± 11.9% (95% confidence interval: 2.9 – 7.4%) and 0.90 (95% confidence interval: 0.66 – 0.97), respectively. The results suggest that peak power output provides a more reliable measure than mean power output. The Cycling Intermittent-Sprint Protocol provides reliable measures of intermittent-sprint performance.

Introduction

Team-sports such as rugby union, hockey and football are characterised by periods of intermittent high-intensity activity, interspersed with longer spells of recovery and lower intensity activity (Williams, 1990). In such sports, athletes typically complete twenty to sixty sprints over 0 – 20 m per match with mean sprint durations of less than 3 s and mean recovery times between sprints of 2 min (DiSalvo, Gregson, Atkinson, Tordoff & Drust, 2009; Roberts, Trewartha, Higgitt, El-Abd & Stokes, 2008; Spencer, Bishop, Dawson & Goodman, 2005; Spencer, Lawrence, Rechichi, Bishop, Dawson & Goodman, 2004). The contribution of sprinting to the total activity profile in field-based sports is therefore small. Nevertheless, sprinting frequently precedes decisive moments in play and hence, can be critical to the outcome of a match (Reilly, 1997; Spencer et al., 2004). Consequently, several performance-based protocols have been devised to examine the physiological and metabolic demands of sprinting in field-based sports, including the Bangsbo Sprint Test (Bangsbo, 1994), the Loughborough Intermittent Shuttle Test (Nicholas, Nutall & Williams, 2000) and the Soccer-Specific Test of Prolonged Repeated-Sprint Ability (Oliver, Armstrong & Williams, 2007).

The majority of these protocols are running-based and use durations, intensities, and exercise-rest ratios developed from time-motion analyses of team-sports. Typically, in running-based protocols, athletes are required to complete 6 – 21 sprints of 20 – 40 m interspersed with 15 – 100 s of recovery (Gabbett, 2010; Nicholas et al., 2000; Oliver et al., 2007). Such protocols have good reliability, construct validity and specificity. However, they are not conducive to sensitive assessments of complex physiological and metabolic variables necessary to develop understanding of the demands of high-intensity activity (McGawley & Bishop, 2006). Cycling- based protocols that focus on assessment of repeated-sprint ability (McGawley & Bishop, 2006; Mendez-Villanueva, Harmer & Bishop, 2008; Spencer et al, 2005) provide valid measures of sprint-performance in match-play (Bishop, Spencer & Duffield, 2001). However, such protocols typically comprise 5 - 10, 6-s sprints interspersed with 24 – 30 s of recovery (Bishop et al., 2001; McGawley & Bishop, 2006; Mendez-Villanueva et al., 2008) and therefore do not replicate intermittent-sprint activity over durations that occur in many team-sports.

Accurate assessment of reliability should consider systematic factors such as learning effects (Phillips, Batterham, Valenzuela & Burkett, 2004). In multiple sprint-cycling based activity, these effects occur (Capriotti, Sherman & Lamb, 1999; McGawley & Bishop, 2006). In contrast to existing literature that identifies the need for two practice sessions when repeated-sprint exercise is performed on a cycle ergometer (Capriotti et al., 1999; McGawley Bishop, 2006), learning effects are minimised during the Cycling Intermittent-Sprint Protocol by completion of one practice session that comprises a quarter of the protocol before the first main-trial (Castle, MacKenzie, Maxwell, Webborn, & Watt, 2011). The Cycling Intermittent-Sprint Protocol examines physiological responses to intermittent-sprint exercise. However, whether a single practice is sufficient to negate learning is unknown.

The Cycling Intermittent-Sprint Protocol has been used to examine effects of pre-cooling on intermittent-sprint performance and has demonstrated 4% improvements in peak power output compared with a no-cooling control (P<0.05) (Castle, MacDonald, Philp, Webborn, Watt & Maxwell, 2006). However, the reliability of the protocol has not been extensively determined (Castle, 2011). Running-based tests designed to assess prolonged repeated-sprint ability have reported typical errors of measurement for peak power output of up to 7.9% (95% confidence interval: 5.8 – 14.4%) (Oliver et al., 2007) In addition, cycle-ergometer based protocols have random errors in repeated trials of 2 – 4% (Paton & Hopkins, 2001). As such it is difficult to determine if changes reported in previous studies using the Cycling Intermittent-Sprint Protocol represent worthwhile interventional change or variation inherent in the test.

Therefore, the primary purpose of this study was to determine the reliability of the Cycling Intermittent-Sprint Protocol using a group of moderately trained team-sport athletes. A secondary purpose was to assess the usefulness of a five-sprint practice session on the main protocol.

Methods

Participants

Eleven men, moderately trained team-sport athletes (mean ± s: age 23 ± 2.4 years, stature 178.5 ± 5.9 cm, body mass 82.3 ± 8.4 kg, sum of skinfolds, 36.7 ± 11.2 mm, peak oxygen consumption [], 42.6 ± 4.8 ml.kg.-1min-1) who competed in team-sports three-to-five times per week were recruited. The study was approved by the University Research Ethics and Governance Committee and conducted in accordance to the Declaration of Helsinki. Before testing, participants completed a medical questionnaire and refrained from alcohol, caffeine and prolonged strenuous activity for 24 hours. Participants replicated dietary and fluid intake in the 12 hours leading up to each session.

Experimental Design

During the first visit to the laboratory body fat was determined using the method of Durnin and Womersley (1974) and participants completed an incremental lactate-threshold-to-test on a modified cycle ergometer (Monark 620 Ergomedic, Varberg, Sweden) fitted with power cranks (SRM: Scientific model, Julick, Germany). After a warm-up of 5 min at 95 W, intensity was increased every 3 min by 24 W and blood was obtained in the final 30 s of each stage from arterialised fingertip capillary samples collected into lithium-heparin coated microvette tubes. Analysis of samples was performed with an automated analyser (YSI 2300 Plus, Yellow Springs Instruments, Ohio, USA) calibrated immediately before each session using the manufacturers 5-mM standard, set to self-calibrate every twenty five minutes and verified after each session using the same manufacturer standard (YSI 2427; coefficient of variation (CV) = 5.5%). When lactate threshold was achieved, determined as the first intensity at which there was a sustained increase in blood lactate concentration above rest (Bourdon, 2000), no further blood samples were taken and intensity was increased by 24 W every 1-min until volitional exhaustion. Oxygen uptake was recorded using open-circuit spirometry and expired air was collected for approximately 45 s during the final minute of each stage. Heart rate (Polar sports tester, Polar Electro, Kempele, Finland) was also recorded every minute throughout the test. Expired oxygen, carbon dioxide, gas temperature and volume of expired air were analysed using a Servomex 4100 xentra gas analyser (Crowborough, UK; CV, = 1.5%; = 1.9%) with a two-point calibration against nitrogen and a gas mixture of known O2 and CO2 concentration (BOC, UK) prior to each test.

After 25 ± 5-min recovery a Cycling Intermittent Sprint Protocol practice was completed. Briefly, the protocol comprised a standardised warm-up (5 min at 95 W and two 30-s bouts at 120 W with 30-s rest in between), followed by twenty consecutive 2-min bouts of activity, with 10 s of passive rest, 5 s of maximal sprinting against a resistance of 7.5% of body weight and 105 s of active recovery at 35% of cycling Active recovery required participants to cycle at 80 r.min-1 and was monitored by a second experimenter in the laboratory who observed the participants at all times. During practice of the Cycling Intermittent Sprint Protocol participants completed the standardised warm-up and five 2-min blocks of the protocol.

On the second and third visits, participants completed the Cycling Intermittent Sprint Protocol (Castle et al., 2006), hereafter referred to as trials 1 and 2. After the standardised warm-up, participants performed the protocol on a modified cycle ergometer (Monark 620 Ergomedic, Varberg, Sweden) fitted with power cranks (SRM, Scientific model, Julick, Germany). Each 5-s sprint was initiated from a stationary start. Power output was determined via the SRM cranks at a rate of 0.5 Hz and stored to a powermeter (SRM Powercontrol V). Data were analysed using SRM software (version 6.42.06). Peak power output (W) was the highest recorded power output during each sprint. Mean power output (W) was the highest 3-s power output from the middle 3 s of the 5-s sprint to counter frictional and other factors experienced when starting sprints from a stationary position (Winter and Fowler, 2009). External work done (J) for each sprint was calculated as mean power output multiplied by duration (3 s) (Castle et al., 2011). To determine the effectiveness of one practice session before main trials to counteract learning effects, results from the five practice sprints were compared with those from the first five sprints of trials one and two.

Statistical Analysis

Data were assessed for normality and sphericity and adjusted where necessary using the Huynh-Feldt method. Peak power output and mean power output were compared using a fully within-groups factorial ANOVA (condition vs. time). Data were assessed for heteroscedasticity using plots of log transformed data and reliability measures. Typical errors of measurement, calculated from the standard deviation of the mean difference for each pair of trials using the formula typical error of measurement = SD(diff)/√2 (Hopkins, 2000) and expressed as a mean coefficient of variation, and intra-class correlation were calculated from log transformation. Intra-class correlations were determined using a reliability spreadsheet (available at newstats.org/xrely.xls) as used by Laursen, Francis Abbiss, Newton & Nosaka (2007). 95% confidence limits were determined for typical errors of measurement and intra-class correlations using the methods described by Hopkins (2007). Least-products regression, where both outcome measures were compared with an assumption of error in each, determined fixed and proportional bias in peak power and mean power output between trials (Ludbrook, 1997). Using this method, estimates of the intercept (a’) and slope (b’) of the least-products regression line (E = a’ + b’x) were calculated and evaluated by 95% confidence intervals (determined from bootstrapping) for a’ and b’ to assess bias. Fixed bias was deemed present when the 95% confidence interval for a’ did not span zero. Proportional bias was deemed present when the 95% confidence interval for b’ did not include one. Sprints were also grouped into four phases, each phase being the mean of five sprints, to improve visual and analytical clarification. All data were analysed using SPSS (version 18.0) and are reported as mean ± standard deviation. Statistical significance was accepted as P≤0.05. Effect sizes were estimated using partial Eta squared (ηp2) where 0.2 represented a ‘small’ effect size, 0.5 a ‘medium’ effect size and 0.8 a ‘large’ effect size (Nakagawa and Cuthill, 2007).

Results

There were no differences between trials 1 and 2 for peak power output across all 20 sprints (p=0.396, ηp2= 0.073). Similarly, there were no differences in peak power output between trials 1 and 2 when the sprints were grouped into four phases (p=0.406, ηp2= 0.070). Mean power output across all 20 sprints did not differ between trials 1 and 2 (p=0.820, ηp2= 0.005). Grouping the sprints into four phases also showed no difference between trials 1 and 2 (p=0.820, ηp2= 0.005). Peak power output and mean power output for the practice period did not differ from the first five sprints of trials 1 or 2 (Table 1, p>0.05, ηp20.058). Typical errors of measurement and intra-class correlations for practice versus trial 1 were 2.6 ± 9.6% (1.8 – 4.7%) and 0.94 and for practice versus trial 2, 2.5 ± 7.2% (1.7 – 4.7%) and 0.95 respectively. Peak power output demonstrated less variability than mean power output throughout the protocol (mean of sprints 1 – 20), despite both showing strong correlations (Table 2).

****Table 1 near here****

****Table 2 near here****

For grouped sprints (mean of five sprints), peak power output in phases 1 and 4 had smaller typical errors of measurement (3.4 ± 10.9% and 3.3 ± 14.1%, respectively) than phases 2 and 3 (3.5 ± 12.8% and 3.9 ± 14.1%, respectively), (Table 3). This also occurred for mean power output. However, typical errors of measurement were greater for mean power output than peak power output in phases two, three and four (Table 3). Intra-class correlation for peak power output in grouped sprints between trials 1 and 2 remained above r = 0.92 throughout each phase, however this was not the case for mean power output in phase 2 and phase 3 (r = 0.85).

****Table 3 near here****

Peak power output and mean power output were normally distributed for trials 1 and 2. Although peak power output tended to be higher in trial 2 than trial 1 (figure 1), there was no fixed bias (a’ = -158.2 W; 95% confidence interval (CI) = -449.6 – 133.2 W). Similarly, there was no proportional bias between trials 1 and 2 for peak power output (b’ = 1.15; 95% CI = 0.90 – 1.39). Conversely, there was fixed bias between trials one and trial two (figure 2) (a’ = -295.0 W; 95% CI = -534.9 – -55.12 W) and proportional bias of ~ 20% (b’ = 1.26; 95% CI = 1.05 – 1.48).

****Figure 1 near here****

****Figure 2 near here****