Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 16

Training & Testing

Validity of a Repeated-Sprint Test for Football

Authors F. M. Impellizzeri1, 2, E. Rampinini2, C. Castagna3, D. Bishop4, D. Ferrari Bravo2, A. Tibaudi2, U. Wisloff5


Affiliations The affiliations are listed at the end of the article
Key Abstract and compared according to competitive level
words l" ! or playing position. Standard error of
reliability Three studies involving 108 football players measurement values expressed as coefficient
l" construct were conducted to examine the reliability of of variation for RSSA mean time and best
validity l" a re-peated-shuttle-sprint ability (RSSA) test time were 0.8 and 1.3 % (short-term
soccer and its ability to differentiate between reliability) and 0.9 and 1.2% (long-term
l" playing players of various competitive levels and reliability), respectively. The smallest
positions l" playing positions. Study 1: Short-term worthwhile changes were 0.5 % for both
competitive reliability was determined in 22 professional mean and best time. Professional players
level players completing the RSSA test (6 × 40-m showed bet-ter RSSA performance than
sprints with 20 s of recovery between amateur players, and defenders displayed
sprints) on two separate occasions. Study 2: the lowest RSSA perfor-mance. In conclusion,
Long-term reliability (seasonal changes) was the RSSA test showed ad-equate construct
ex-amined in 31 professional players validity but only RSSA mean time showed
completing the RSSA test four times (during sufficient reliability to detect large training-
the preseason peri-od, at the start, middle induced changes but not small impor-tant
and end of the competi-tive season). Study differences.
3: 108 players were divided
Fax: + 41(0)4 43 85 75 90
franco.impellizzeri@kws.ch
Introduction
!
Association football (soccer) is a complex
sport requiring the repetition of many
different activ-ities such as jogging, sprinting
and jumping [6, 18, 26, 31]. Players are often
accepted after revision required to repeat-edly produce maximal or
February 19, 2008 near maximal sprints of short duration (1 – 7
Bibliography s) with brief recovery pe-riods [6, 31].
DOI 10.1055/s-2008- Therefore, the ability to repeat mul-tiple
1038491 Published sprints at high speed is important for soccer
online 2008 physical performance [5, 32]. The use of
Int J Sports Med © tests consisting of several sprints
Georg Thie-me Verlag interspersed with brief recovery periods,
KG Stuttgart • instead of a single sprint, should ensure
New York • ISSN 0172- physiological responses similar to those
4622
occurring during intense periods of play in
Correspondence actual matches [24, 27, 32]. For these
Dr. Franco M. reasons, the use of repeated-sprint ability
Impellizzeri exercises for the training and testing of
Neuromuscular
soccer players is increas-ing [19, 23, 24, 27].
Research
Laboratory
Schulthess
The validity of most currently used repeated-
Clinic
sprint ability tests is based predominantly on
Lengghalde
their intrinsic characteristics (logical validity).
2 8008
Zurich The use of these tests often assumes that
Switzerland they ac-tually measure match-related
Phone: + 41(0)4 43 85 physical perfor-
75 87
repeated-shut-tle-sprint ability (RSSA) test taxed during the high-intensity phases of the
for soccer players [21]. Rampinini et al. [21] game. Given that several factors that can
mance (construct have shown, in profes-sional soccer players, influence actual physical-match per-
validity) [14]. However, that there are moderate but significant formance, these correlations support the
to the authors’ correlations between sprinting (r = con-struct validity of the investigated RSSA
knowledge, only one – 0.65) and high-intensity running (r = – 0.60) test. How-ever, the strength of the
study examined the completed during official match-play and the correlations does not support the predictive
construct validity, as mean performance during an RSSA shuttle- validity of the test, for which r values above
indicated by match-re- run-ning test (six 40-m shuttle sprints 0.90 are usually necessary. Although
lated physical interspersed with 20 s of passive recovery). Rampinini et al. [21] have provided pre-
performance, of a These relation-ships confirmed the liminary convergent evidence to the
involvement of physical ca-pacities actually construct
Impellizzeri FM et al. Validity of a …Int J
Sports Med
Training & Testing tapering. In the preseason pe-riod, tests were performed
during a tapering week, after two to three days of reduced
training load.
validity of the RSSA test, several studies are required to
build up a body of evidence to support the validity of a test
[2]. The con-struct validity of a test can also be examined Study 1
by comparing the test scores of populations assumed to Short-term reliability. Reliability was determined in 22
differ in physical require-ments during a match (know- profes-sional football players (age 22 ± 1 years, body mass
group difference technique) and hence assumed to be 73 ± 5 kg, and height 177 ± 4 cm) from the same team who
characterized by different physical capaci-ties [28]. Some completed the RSSA test twice within one week; at least 48
studies, indeed, have examined the ability of field tests to hours separated the tri-als. The RSSA tests were completed
differentiate between soccer players of different compet- at the same time of the day after a 10-min warm-up of
itive levels or playing positions [9,15]. To be valid and low-intensity running and striding, followed by three
applicable, a test must also be reliable. The knowledge of submaximal 40-m shuttle sprints (20 + 20 m). The RSSA test
the reliability of a test allows one to understand if the was completed within 5 min after the warm-up. Players
instrument can be used to detect changes as a were already familiar with this test that was part of their
consequence of interventions (signal-to-noise ratio) and to routine assessment. Therefore, no familiarization trials
better interpret the test results. Therefore, both short- and were necessary.
long-term reliability should be calculated to supply
information to researchers and practitioners interested in
examining the efficacy of specific interventions over Study 2
different time frames. For example, short-term reliability Long-term reliability (seasonal changes). In the second
can be used to calculate the individual minimum study, we examined the long-term reliability in the RSSA
detectable change, while long-term reliability allows the test in 30 pro-
estimation of sample sizes nec-essary to detect meaningful
changes in intervention studies and/ or to estimate the
individual differences in response to treat-ments [11].

Therefore, three studies were conducted to determine 1)


the short- and 2) long-term reliability, and 3) to examine
the ability of the RSSA test proposed by Rampinini et al.
[21] to differentiate between football players of different
competitive levels and playing positions (construct
validity).

Methods
!
Subjects and study design
The present investigation consisted of three separate
studies in-volving a total of 108 male soccer players
participating in one or more of these studies. The study
was approved by the Indepen-dent Institutional Review
Board according to the Guidelines and Recommendations
for European Ethics Committees by the Euro-pean Forum
for Good Clinical Practice. Data collection started in 2005
and finished in 2007. Before each testing session, subjects
were instructed not to eat for at least three hours before
testing and to maintain the normal dietary habits in the
two days before testing. Tests were always completed at
least 48 hours after the match and after one to two days of
fessional soccer players (age 25 ± 5 years, body mass 78 ± Study 3
8 kg, and height 181 ± 5 cm) recruited from three teams Differences between competitive level and playing
position.
competing in their national league and in international and
In this study, 108 football players (age 24 ± 4 years, body
national cup com-petitions. Data were collected from 2006
mass 75 ± 7 kg, and height 179 ± 5 cm) tested from three
and 2007 and only players with field tests completed in
to five times during the season were selected among
four selected periods of the season (every three months)
professional and ama-teur players. For each player, the
were involved in the study: the first week of the preseason
best test was selected to obtain a more representative test
training (PRE); within the first 4 weeks from the start of the
score of the examined physical ca-pacity. Players were
competitive season (EARLY); within the 4 weeks in the
classified in three groups according to their level: 1) top-
middle of the competitive season (MID); and within 4
professional (players of teams participating in the first
weeks before the end of the season (END). During the Early
division championship of their nation); mid-professional
to Mid period, teams played one match a week (national
(players of teams participating in the second and third
championship) and one in-week match (national and
division championships of their nation); and 3) amateur
interna-tional competitions).
(nonprofession-al). Soccer players were also assigned to
one of four playing posi-tion groups based on the
The generic training plan completed by the soccer players indication of their coaches: 1) defend-ers; 2) fullbacks; 3)
in-volved in this study was supplied by the fitness coaches midfielders; and 4) forwards.
of the teams. The soccer players completed RSSA-based
training ses-sions two to three times a week during the RSSA test protocol
pre-competition training period, and once a week during To measure RSSA, we used a test consisting of six 40-m (20
the competitive season. In the remaining part of the + 20-m sprints with 1808 turns) shuttle sprints separated
physical training (twice a week dur-ing the preseason and by 20 s of passive recovery [21]. This test was designed to
once a week during the competitive sea-son), players measure both re-peated-sprint and change in direction
completed aerobic high-intensity interval training sessions abilities. The athletes started from a line, sprinted for 20
using both specific (small-sided games and soccer-spe-cific m, touched a line with a foot and came back to the starting
circuits) and generic (running) exercises [13]. Training also line as fast as possible. After 20 s of passive recovery, the
included sessions of plyometric training, but no resistance soccer player restarted again. Immediately after the warm-
train-ing sessions were completed with the exception of up, each player completed a preliminary single shuttle-
injured play-ers who performed resistance training during sprint test using a photocells system (Microgate, Bolza-no,
the rehabilitation (these players were excluded from the Italy). This trial was used as the criterion score during the
study). Each set (from 2 to 4) of the RSSA training consisted subsequent 6 × 40-m shuttle sprint test [21]. After the first
of 5 to 10 sprints (5, 10, 20 and 40 m), for a maximal total pre-liminary single shuttle-sprint, subjects rested for 5 min
distance of 40 m for each sprint. The recovery between before the start of the RSSA test. If performance in the first
sprints ranged from 20 s to 1 min, and the pas-sive sprint of an
recovery duration between sets ranged from 2 to 4 min.
Impellizzeri FM et al. Validity of a …Int J
Sports Med
present study, the SWC (expressed as percentage) was
calcu-lated as a proportion of the effect size which
represents the mag-nitude of improvement in a variable as
RSSA test was worse than the criterion score (i.e., an
increase in time greater than 2.5 %), the test was a function of the be-tween-subjects standard deviation of
terminated immediately and subjects were required to the investigated popula-tion (i.e., 0.2 times the between-
repeat the RSSA test with maxi-mum effort after a further subject SD of top- and mid-pro-fessional football players)
5-min rest. Five seconds before the start of each sprint,
[11,12].
subjects assumed the ready position and waited for the
acoustic start signal (with 5 s of countdown). Best time in a
single trial (RSSAbest), mean time (RSSAmean) and dec-
rement (RSSAdecrement) were determined according to
Rampinini et al. [21]. Specifically, the RSSAdecrement was The differences between competitive levels and playing
calculated as RSSAmean/RSSAbest and expressed as
posi-tions were examined using a two-way ANOVA (3 × 4
percent. All tests were com-pleted outdoors on natural
grass surface. design). The independent variables included a between-
subject factor “competitive level” with three levels (top-
professional, mid-pro-fessional, and amateur), and a
Statistical analyses
between-subject factor “playing position” with four levels
Unless otherwise noted, all data are presented as mean ±
(defender, fullback, midfielder, and forward). When a
stan-dard deviation (SD). Relative reliability concerns the
significant F-value was found (p £ 0.05), the Bonferroni
degree to which individuals maintain their position in a
post hoc test was applied. All statistical analyses were
sample with re-peated measurements [4]. We assessed
performed with SPSS 13.0.
this type of reliability using the intraclass correlation
coefficient [ICC(2,1), a two-way random effects model with
single measure]. We considered an ICC over 0.90 as high,
between 0.80 and 0.90 as moderate and below 0.80 as low
[30]. Absolute reliability is the degree to which repeated
measurements vary for individuals [4] and we expressed
this type of reliability with the standard error of
measurement expressed in absolute terms (SEM) or as
coeffi-cient of variation (CV) [11]. Short-term relative and
absolute re-liability were determined using the tests
completed twice in one week (Study 1). The percent
change was calculated using the change scores of log
transformed data of the two trials. The ef-fect size of the
difference (d) was determined as: (mean value trial 2 –
mean value of trial 1)/pooled SD. The modified scale by
Hopkins (www.sportsci.org/resource/stats/2002) was used
for the interpretation of d: trivial, < 0.2; small, 0.2 – 0.6;
moderate, 0.6 – 1.2; and large, > 1.2. Long-term absolute
and relative reli-ability were examined using the four RSSA
test sessions (Early, Start, Mid, and End). Percent
differences in means (seasonal changes) and the
corresponding 90 % confidence intervals were calculated
from log transformed data. We also calculated the
likelihood that the true values of estimated difference in
RSSA parameters were substantial (i.e., larger than the
smallest worthwhile change, SWC). Threshold for assigning
qualitative terms to chances of substantial differences
were as follows: < 1 %, almost certainly not; < 5 %, very
unlikely; < 25 %, unlikely or probably not; > 50 %, possibly;
> 75 %, likely or probable; > 95 %, very likely; > 99 %,
almost certain [16]. Different methods can be used to
determine the small worthwhile change [3,12]. In the
Training & Testing RSSA mean time decreased by a moderate and substantial
– 2.2 % from Pre to Early (l Fig. 1 A). This improvement was
followed by a small but likely 0.8 % worsening in
Results RSSAmean in the Mid session, that persisted until the end
! of the competitive season. A similar change pattern was
Study 1 found for RSSAbest, with a small (– 1 %) but likely
Short-term reliability. Absolute and relative SEM were 0.06 substantial improvement between Pre and Early, followed
s (CI 90% 0.04 – 0.07 s) and 0.8 % (CI 90% 0.6 – 1.0 %) for by a small (0.9 %) and likely worsening in RSSAbest, and a
RSSAmean, 0.09 s (CI 90% 0.07 – 0.12 s) and 1.3 % (CI 90% trivial change (– 0.2 %) between Mid and End (l Fig. 1 B).
1.0 – 1.7 %) for RSSA-best, and 1.2 s (CI 90% 0.9 – 1.6 s) There was a small and very likely 22.7 % decrease in
and 30.2 % (CI 90% 23.6 – 42.7%) for RSSAdecrement. RSSAdecrement, from Pre to Early followed by trivial
Trivial to small differences were found be-tween trial 1 and changes (0 to – 6.2 %) (l Fig. 1 C).
trial 2 for RSSAmean (7.20 ± 0.11 s vs. 7.19 ± 0.14 s,
respectively: effect size 0.09 [trivial]), for RSSAbest, (6.90 ±
Study 3
0.09 s vs. 6.92 ± 0.10 s, respectively; effect size = 0.24
[small]), and for RSSAdecrement (4.3 ± 1.2 % vs. 3.8 ± 1.4 Differences between competitive level and playing
%, respectively; ef-fect size = 0.36 [small]). ICC values were position.
0.81 (CI 90% 0.64 – 0.90) for RSSAmean, for 0.15 (– 0.21 – Detailed results of the two-way ANOVAs are presented in l
0.48) for RSSAbest, and 0.17 (CI 90% Ta-ble 2. No “competitive level” × “playing position”
– 0.18 – 0.49) for RSSAdecrement.
interactions were found, while main effects were
Study 2 significant. Professional players showed higher RSSA
performance (RSSAmean, RSSAbest and RSSAdecrement)
Long-term reliability (seasonal changes). Descriptive data
than amateur players. Defenders were players with the
of the four testing sessions and the SEM values expressed
lower RSSA performance (RSSAmean and RSSAbest) com-
as CV cal-culated from two consecutive sessions are
pared to the other playing positions.
presented in l Table 1. The CV calculated increasing the
time between tests (i.e., Pre vs. Mid and Pre vs. End) was
similar to Pre vs. Early (data not shown). The mean CV
calculated from all the four testing ses-sions (Pre, Early, Discussion
Mid and End) was 0.9 % (CI 90% 0.8 – 1.1 %) for !
RSSAmean, 1.2 % (CI 90% 1.1 – 1.4 %) for RSSAbest, and
The differences found between players of different
29.8 % (CI 90% 26.1 – 34.7 %) for RSSAdecrement. The
SWC values for RSSAmean RSSAbest and RSSAdecrement, competitive levels and playing positions further support
were 0.5, 0.5 and 8.4 %, respectively. The ICC values were the construct valid-ity of this test for measuring the ability
0.58 (CI 90% 0.38 – 0.74) for RSSAmean, 0.63 (CI 90% 0.49 to repeat shuttle-sprints in football players. However,
– 0.75) for RSSAbest, and 0.49 (CI 90% 0.33 – 0.65) for among the various RSSA parame-ters, only the mean time
RSSAdecrement.
showed sufficient reliability to detect large training-
induced changes, but not smallest important dif-ferences.
Impellizzeri FM et al. Validity of a …Int J
Sports Med
Training & Testing 0. 0.
nt S 1 1
of D 4 5 6
(2 m
4. ea
(0 (0 0 su
(C .8 .8 – re
I – – 3 m
90 1. 1. 9. en
%) 2) 3) 6) t 7. 6.
En ex M1 8
(n d pr ea 6 7 4.
= – 2 es n ± ± 3
30 Mi C 0. 1. 9. se ± 0. 0. ±
) d V 9 0 8 d Ea S 1 1 1.
re (1 as rly D 5 7 7
pe 7. err ± 0. 0.
at (0 (0 5 or S 1 1 2.
ed (C .7 .9 – of D 3 5 2
- I – – 2 st 7. 6.
sh 90 1. 1. 8. an M 3 9 5.
utt %) 1) 4) 5) da Pr ea 2 4 4
le Mi rd e n ± ± ±
sp d ch
rin – 2 an
t Ea C 0. 1. 1. ge
ab rly V 8 1 6 s
ilit an (
y d %
(R T )
SS a
A) bl P
te e a
st 1 r
pa S a
ra e m
m a e
et s t R R R
er 3 o e S S S
s 6. n r S S S
cal 7 al s A A A
cul 1. 1. (2 d
at C 0 6 9. e
ed Ea V (0 (1 4 c
fro rly (C .7 .3 – r
m – I – – 4 e
th pr 90 1. 2. 9. mb m
e e %) 1) 1) 2) e e e
for a s n
th
n t t
e
(s (s
th ± 0. 0.
) )
re S 1 1 1.
e D 3 5 7
va
ria
tio 7. 6.
n M 2 9 4.
(C En ea 0 2 0
V) d n ± ± ±
co Mi M 7. 6. 4.
e d ea 2 9 2
ffi n 2 3 ±
cie ± ± ± 1.
Fig. 1 A to C Percentage changes between testing
sessions (bars indicate uncertainty in the true mean
change; 90 % confidence limits). Trivial area was
calculated from the smallest worthwhile change
determined as 0.2 times the between-athlete variation.

The results of the short-term reliability study (Study 1)


showed that RSSAmean is the parameter with the greatest
absolute reli-ability, while the percent decrement is the
least reliable. The CV values found in the present study for
RSSAmean is similar to the values showed by Fitzsimons et
al. [10] who reported a CV of 0.8 % for the total time of a
repeated-sprint ability running test consisting of 6 × 40-m
sprints with 30 s of recovery. In the present study, the test
included shuttle-sprints. A greater CV (1.8 %) has been
reported by Wragg et al. [32] for the repeated-sprint ability
test proposed by Bangsbo [5] which included 7 × 34.2-m
sprints with direction changes. Consistent with pre-vious
investigations using both running and cycling sprints [10,
Impellizzeri FM et al. Validity of a …Int J
Sports Med
Training & Testing

Table 2 Differences in repeated-shuttle-sprint ability (RSSA) between competitive levels


and playing positions
RSSA best (s) RSSA mean (s) RSSA dec (%)
Mean ± SD Mean ± SD Mean ± SD
Competitive level
Top-pro (n = 30) 6.88 ± 0.19 7.12 ± 0.17 3.3 ± 1.5
Mid-pro (n = 45) 6.83 ± 0.18 7.20 ± 0.19 5.1 ± 1.8
Amateur (n = 33) 7.08 ± 0.23 7.55 ± 0.25 6.1 ± 2.0
Main factor p < 0001 p < 0001 p < 0.001
Post hoc analysis TP = P < AM* TP = P < AM* TP < MP = A
Playing position
Defenders (n = 34) 7.01 ± 0.23 7.40 ± 0.28 5.2 ± 2.4
Fullbacks (n = 20) 6.83 ± 0.22 7.18 ± 0.27 4.8 ± 2.2
Midfielders (n = 33) 6.90 ± 0.21 7.25 ± 0.27 4.8 ± 1.9
Forwards (n = 21) 6.91 ± 0.23 7.26 ± 0.21 4.8 ± 1.7
Main factor p = 0.012 p = 0.004 p = 0.930
Post hoc analysis (D > FU = M) = FO* D > FU = M = FO*
Role X level interaction p = 0.900 p = 0.927 p = 0.714
* p < 0.05; TP: top-professional players; P: mid-professional players; AM: amateur players; D: defenders; FU: fullbacks; M:
midfielders; FO: forwards
17], the least reliable parameters calculated from the RSSA sion a week during the competitive season, it is possible
tests is the percent decrement. Indeed, in the present that this low specific volume was not sufficient to maintain
study, the CV found for the decrement in performance this physical ability even with the additional in-week
during the RSSA is sim-ilar to the 31.2 % reported by
matches for interna-tional and national competitions
McGawley and Bishop et al. [17] during 6-s cycling sprints,
(cups) played in the Early to Mid period. Alternatively, the
and greater that 18.5 % reported by Fitzsimons et al. [10]
during sprint running. Relative reliability showed very low fatigue accumulated in the first part of the competitive
ICC for RSSAbest and RSSAdecrement, while mod-erate ICC season may have reduced this physical ability. However,
was found for RSSAmean. Therefore, the only parameter further studies are necessary to confirm these
showing an absolute and relative reliability acceptable for speculations.
mon-itoring football players is RSSAmean. The
RSSAdecrement consis-tently showed the poorest The absolute and relative SEM values obtained in the long-
reliability (absolute and relative) indi-cating that this term reliability study can supply useful information to
parameter should not be used to evaluate foot-ball researchers and practitioners interested in examining the
players. Indeed, the minimal detectable change [11], which efficacy of specific interventions (training, nutritional, etc.).
is the individual minimum difference that can be Indeed, the SEM can be used to calculate the sample size
interpreted as “real” with an acceptable probability level needed to detect changes over time frames similar to the
(84 %), resulted in 60 % for the RSSAdecrement, while the ones used in the present study. Fur-thermore, these results
minimum detectable changes for RSSA mean and best can assist to better interpret the results of intervention
were 1.6 and 2.6, respectively. Trivial to small differences studies, for example, by calculation of the indi-vidual
were found between tests. However, all the subjects responses to interventions [11]. The long-term reliability
involved in this study were already familiar with the results can also be used to estimate the sensitivity of RSSA
specific test protocol since it has been used for their pa-rameters to training interventions as expressed by the
routine test-ing and/or training exercises. With players not signal-to-noise ratio (intervention-induced changes/typical
previously famil-iarized, one or two familiarization trials error of measurement) [1]. In a previous study [7], we
may be necessary to prevent a learning effect in repeated- found a change in RSSAmean of 2.1 % after 7 weeks of
sprint-based tests [25].
repeated-sprint based training completed at the beginning
of the competitive season. The CV obtained in the present
The CV determined from Pre to Early in the long-term study from Early to Mid was about one-third of the
reliability study were slightly greater than those obtained changes induced on RSSAmean by the specific sprint
training. This suggests an acceptable sensitivity of this
in the short-term reliability study and the other seasonal
parame-ter to detect changes determined by specific
phases. This was expected since, in the short-term
training interven-tions even during the competitive season.
reliability, it can be assumed that there is no true change in
On the other hand, the RSSA best and decrement scores
individuals’ measurements be-tween trials, and during the changed by 1 and 10 % re-spectively, corresponding to
competitive season physical capaci-ties tend to remain values similar or lower than the CV found for these
relative stable at least at group level [29]. On the other parameters in the reliability studies. This indi-cates a low
hand, during the preseason period with players resum-ing sensitivity of these parameters, especially if com-pared to
training after a relative long rest period (three to four the RSSAmean. Therefore, only RSSAmean can be useful to
weeks), substantial changes in physical capacity can be quantify large changes induced by specific training
more easily de-tected. Therefore, the CV obtained during strategies. However, as the reliability of RSSAmean was
the preseason training period was probably inflated by twice the small worthwhile change, it is unlikely that it can
be used to detect smaller but important differences or
individual differences in train-ing response [11], which can
individual changes. Fur-thermore, low ICCs were found for
increase the CV.
all the RSSA parameters. Therefore, other strategies such
as the average of repeated tests should be used to improve
The RSSA mean and best time showed a likely substantial
the reliability of RSSA parameters in-cluding RSSAmean.
but small increase between the start and middle of the
competitive season. This worsening of RSSA performance
after the first part of the competitive season persisted until
the end of the season. As the training programs included
only one RSSA training ses-
Impellizzeri FM et al. Validity of a …Int J
Sports Med

You might also like