See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/367379209

Football movement profile analysis and creatine kinase relationships in youth

national team players

Article in Physiology International · January 2023

DOI: 10.1556/2060.2023.00160

CITATIONS READS

0 134

8 authors, including:

Gyorgy Szigeti Péter Revisnyei

Hungarian Football Federation Budapest University of Technology and Economics
4 PUBLICATIONS 7 CITATIONS 10 PUBLICATIONS 30 CITATIONS

SEE PROFILE SEE PROFILE

Alija Pašić Tim Gabbett

Budapest University of Technology and Economics Gabbett Performance Solutions
35 PUBLICATIONS 273 CITATIONS 354 PUBLICATIONS 24,469 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

COST CA15127 Action "Resilient Communication Services Protecting End-user Applications from Disaster-based Failures" (RECODIS) View project

Biodesp Institute View project

All content following this page was uploaded by Tim Gabbett on 20 February 2023.

The user has requested enhancement of the downloaded file.

Physiology International
DOI: 10.1556/2060.2023.00160

Football movement profile analysis and creatine

kinase relationships in youth national team
players

GYÖRGY SZIGETI1,2p , GÁBOR SCHUTH1,2, TAMÁS KOVÁCS1,

PÉTER REVISNYEI3, ALIJA PASIC3, ÁDÁM SZILAS1, TIM GABBETT4,5,6 and
GÁBOR PAVLIK2

1
Department of Sport Medicine and Sport Science, Hungarian Football Federation, Budapest, Hungary
2
Department of Health Sciences and Sport Medicine, University of Physical Education, Budapest,
Hungary
3
MTA-BME Information Systems Research Group, Budapest University of Technology and Economics
(BME), Budapest, Hungary
4
Gabbett Performance Solutions, Brisbane, QLD, Australia
5
Centre for Health Research, University of Southern Queensland, Ipswich, QLD, Australia
6
Health Innovation and Transformation Centre, Federation University, Ballarat, VIC, Australia

Received: October 7, 2022 • Revised manuscript received: December 22, 2022 • Accepted: January 2, 2023

© 2023 Akadémiai Kiadó, Budapest

ABSTRACT
Objective: Creatine kinase (CK) is widely used as a monitoring tool to make inferences on fatigue and
readiness in elite soccer. Previous studies have examined the relationship between CK and GPS pa-
rameters, however these metrics may not accurately describe the players’ load during soccer-specific
movements. Football Movement Profile (FMP) monitoring is a viable option for such purposes,
providing solely inertial sensor-based data and categorizing movements according to intensity (very
low, low, medium, high) and movement type (running-linear locomotive, dynamic – change of di-
rection or speed). Methods: We investigated the relationship between the FMP distribution of youth
(U16–U21) national team soccer players and the absolute day-to-day change in CK. We applied
Spearman’s correlations, principal component analysis and K-means clustering to classify players’ CK

p
Corresponding author. Department of Sport Medicine and Sport Science, Hungarian Football Federation, Szajkó
utca 39, 2089 Telki, Budapest, Hungary; Tel.: þ36 1 577 9500. E-mail: [email protected]

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

2 Physiology International

responses according to their specific FMP. Results: Moderate to large negative associations were found
between very low intensity FMP parameters and CK change (r 5 0.43 ± 0.12) while large positive
associations were identified between CK change and other FMP metrics (r 5 0.62 ± 0.12). Best fitting
clustering methods were used to group players depending on their CK sensitivity to FMP values.
Principal component analysis explained 83.0% of the variation with a Silhouette score of 0.61 for the 4
clusters. Conclusions: Our results suggest that soccer players can be clustered based on the relationship
between FMP measures and the CK change. These findings can help to plan soccer training or recovery
sessions according to the desired load on skeletal muscle, as FMP monitoring might bridge the lim-
itations of GPS telemetry.

KEYWORDS
soccer, elite, GPS, load, monitoring, FMP

INTRODUCTION
The physical demands of soccer have increased considerably in recent years, with players
required to perform more high-intensity efforts and sprints, and explosive actions during
Premier League matches [1]. Global Positioning System (GPS) technology is widely used for
recording and analysing match and training load of players [2].
Despite the general popularity of GPS, it also has its limitations. Closed stadiums, indoor
training sessions, sampling frequency, and other external factors can affect the usability and
accuracy of traditional GPS parameters, such as high-intensity actions, sprint running dis-
tance and total distance covered [3, 4]. Since most modern monitoring units are also
equipped with microsensors (accelerometer, gyroscope, magnetometer), they are able to
provide Inertial Measurement Unit (IMU) data and certain external load parameters without
the use of GPS function. Overcoming the limitations of GPS systems to precisely monitor
soccer-specific movements, athlete monitoring company Catapult Sports has developed the
Football Movement Profile (FMP) algorithm, which relies solely on IMU data. FMP assigns
all soccer movements into categories based on their intensity (very low, low, medium, high)
and movement type (running – linear locomotive; dynamic – change of direction or
speed) [5].
Creatine Kinase (CK) is commonly used as a blood marker to monitor the muscle damage
and impact of soccer matches and training load on players [6]. Serum CK level is elevated
after extensive mechanical load, indicating damage in the skeletal muscle [7]. Previous studies
have demonstrated that CK is highest 24–48 h after matches and returns to baseline after 72 h
or even later in special circumstances, such as congested schedules [8, 9]. The CK response to
external load also shows large inter-individual variation [10]. Several authors have investi-
gated the relationship between external load (recorded via GPS) and subsequent CK response
in soccer [11–17]. Total number of accelerations and decelerations, as well as sprint distance
is positively associated with elevated CK activity after soccer matches [11, 18]. In the most
recent studies, relationships between high-intensity (>5.5 m s1) running distance and CK
activity have also been found [12, 13]. However, many studies investigated the GPS volume
parameters [12–14, 17, 18], with few investigating intensity parameters [11, 15]. In the

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

Physiology International 3

aforementioned studies, significant correlations between high intensity GPS parameters

(distance/min, high intensity distance/min, number of sprints/min) and CK response were
reported, but only in the 24 h post-match [11, 15]. In our recent work with youth players, we
also found differences among playing positions and activity type (training, compensation,
match) for the external load parameters influencing CK change [16]. As technological de-
velopments provide opportunities to more accurately monitor and quantify soccer-specific
movements, understanding the relationship between soccer-specific load and the internal
load response (measured as CK increase in blood) represents an important pursuit for
practitioners working with football players.
Therefore, the first aim of this study was to examine the relationship between Football
Movement Profile parameters and CK change in the subsequent day in youth national team
soccer players. Secondly, we aimed to find a best fit clustering method in order to group players
based on their CK sensitivity to specific FMP parameters recorded during matches and
training sessions. Based on the published studies and our previous results we hypothesized that
day-to-day CK change will be influenced mainly by medium and high intensity FMP
parameters. Furthermore, we hypothesized that players’ CK change would demonstrate indi-
vidual dependent responsivity. Our results may assist practitioners to better plan and evaluate
their training load according to the expected individual CK response.

MATERIALS AND METHODS

Experimental approach to the problem
National team settings represent a special congested schedule for players, having to compete in
intense international matches, occasionally every 2–3 days. In such a demanding environment
load (internal and external), fatigue and ‘readiness’ monitoring is paramount. For this reason,
training and match monitoring and CK measurements are an important part of the daily routine
in national team training camps.
In this longitudinal retrospective study, we monitored the match and training physical data
of youth national team players. External load and blood marker analysis was part of the stan-
dardized daily national team routine. Although CK response to many external load parameters
has been measured, in our study we investigated the relationship between day-to-day CK change
and the different FMP parameters.

Subjects
Training and match data of 31 elite youth (U-16–U-21) national team soccer players (age
18.3 ± 1.1 years; weight 72.6 ± 5.4 kg; height 181.1 ± 6.7 cm) were collected between July
2019 and March 2021. The large variance of participating players’ age was due to the
intention of creating a larger data set with more individual datapoints. Players’ data were
only included in the analysis if they had at least 20 data pairs (FMP and CK). The football
specific position distribution was as follows: 3 goalkeepers, 4 central defenders, 7 full backs,
9 central midfielders, 4 wide midfielders, 4 strikers. All participants were elite soccer players
having at least five years of football experience and performed 5–7 training sessions and
1–3 matches per week during the sampling periods. The study was approved by the ethical

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

4 Physiology International

review board of the University of Physical Education (Budapest) and conformed to the Code
of Ethics of the World Medical Association (Declaration of Helsinki) (TE-KEB/No9S/2020).
Players and their parents/guardians provided written consent for the use of physical and
medical data.

Procedures

Football movement profile (FMP) monitoring. The physical activity of players was recorded
during training sessions and international matches (friendly and competitive) with 10 Hz
GPS units (Catapult Sports, Melbourne, Australia; S7–field players; G7–goalkeepers), worn in
a custom-made vest between the shoulder blades. Each player wore the same unit during the
study period, to avoid inter-unit differences. Devices were equipped with 100 Hz micro-
sensors (accelerometer, gyroscope, magnetometer) providing non-GPS derived data for the
Inertial Movement Analysis (IMA). IMA analysis includes data from accelerations, de-
celerations, jumps and changes of directions; good reliability and validity has been shown for
this analysis [19]. The Football Movement Profile (FMP) was recently developed by Catapult
Sports, for the purpose of quantifying the intensity of soccer-specific activities [5]. The FMP
algorithm is based on the IMA measurements excluding GPS data and assigns all soccer
movements into one of six movement categories, defined by a combination of intensity (very
low, low, medium, high) and movement type (running-linear locomotive, dynamic–change of
direction or speed):
 Very Low Duration % (Standing like movements)
 Low Duration % (Walking like movements)
 Running Medium Duration % (Steady state linear and curved jogging and running move-
ments at medium intensity)
 Running High Duration % (Steady state linear and curved running movements at high
intensity)
 Dynamic Medium Duration % (Medium-intensity multi-directional movements, linear ac-
celeration and deceleration movements)
 Dynamic High Duration % (High-intensity multidirectional movements, linear acceleration
and deceleration movements)
The relative FMP distribution of the entire session can also be described by taking the ab-
solute variables and expressing them relative to the duration in each activity zone. The intensity
thresholds of the FMP algorithm are based on Player LoadTM values, a parameter derived from
accelerometer data, which has previously been shown to be a valid and reliable metric [20].
Differentiation between running and dynamic movements is based on variance in data in the
medio-lateral and posterior-anterior planes. We used the following parameters: 1. Duration % of
activity in each FMP zone; 2. Duration (minutes) of activity in each FMP zone; 3. Aggregated
parameters (Running medium þ high duration %; Dynamic medium þ high duration %;
Running þ Dynamic duration %; Running þ dynamic high duration %).

Creatine kinase (CK). Capillary blood samples were taken from fingertips after overnight
fasting, every morning as part of the normal protocol at a standardized time in order to measure

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

Physiology International 5

CK levels (CK-MM isoenzyme). Blood samples were collected and analysed using a Reflotron
Plus Clinical Chemistry Analyzer (Roche, Mannheim, Germany) according to the manufac-
turer’s instructions. The validity of this method has been determined in previous studies, with
very large correlations reported between capillary and venous samples (P < 0.01, r 5 0.99 and
r ≥ 0.99 respectively) [14]. CK response was defined as a day-to-day change in absolute values
(ΔCKABS 5 abs [CKday1CKday2]) and as a percentage (ΔCK% 5 CKday1/CKday2). Our pre-
liminary analysis found superior correlation coefficients for ΔCKABS, and we subsequently
excluded the ΔCK% values from our statistical procedures.

Statistical analysis
The main goal of the statistical procedures was to separate the players into groups based on
the relationship between their daily FMP parameters and CK levels. Spearman’s correlation
test was used to describe the relationship, after which Principal Component Analysis (PCA)
and K-Means clustering were applied to find the most suitable cluster format to group the
players.
The filtered dataset was tested for normality and as our data was not normally distrib-
uted, multiple Spearman’s correlation tests were performed. For each player’s ΔCKABS,
values were paired with every FMP parameter’s value and a correlation test was performed.
The p-values from these tests were adjusted using the Benjamini-Hochberg procedure [21].
In the further steps only the significant correlation coefficients (P ≤ 0.05) were kept, and the
non-significant ones were excluded from the clustering process. Therefore, the zero cluster
includes players with no significant correlation between ΔCKABS and FMP parameters.
Players with one or more significant correlations of the examined parameters were clustered
based on their correlation coefficient. We interpreted the coefficient values as trivial
(0.0–0.1), small (0.1–0.3), moderate (0.3–0.5), large (0.5–0.7), very large (0.7–0.9), and
nearly perfect (0.9–1.0) [22].
PCA was applied to the data for the purpose of creating two principal components from the
primary correlation results [23]. Thereafter, the first and second principal components were
used for clustering with the K-Means algorithm [24]. Multiple sets of clusters were created with
the algorithm and in each set the players were divided into different numbers of clusters. The
Cluster set with the highest silhouette score was selected in order to group players most
accurately [25].

RESULTS
Correlation of ΔCKABS and FMP parameters
The total dataset consisted of 757 (24.4 ± 3.8 per player) individual observations (CK and FMP).
The average of individual correlation coefficients for the FMP Very Low percentage and
duration was negative and moderate to large (0.43 ± 0.12; range: 0.36 to 0.56), but all other
examined parameters showed a large positive correlation (0.62 ± 0.12; range: 0.53–0.69).
ΔCKABS showed higher correlation values with the FMP Duration, than FMP % parameters. The
correlation values and the number of significant correlations of all participants (n 5 31) are
illustrated in Table 1.

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

6 Physiology International

Table 1. Average correlation distribution of CK change and FMP parameters. In the second column the
number (N) of significant correlations (P ≤ 0.05) are displayed. A clear difference is visible between the
Very low intensity and all other FMP parameters regarding the orientation of the relationship. While Very
low intensity FMP shows a moderate to large negative correlation with CK change, suggesting a possible
regenerative effect, the other FMP metrics display a large positive correlation. Interestingly Duration
values have a stronger relationship with CK change, indicating the importance of the actual length of
the soccer-specific activity
FMP parameter N5 r values (mean ± SD)
FMP Very Low Duration % 16 0.564 ± 0.116
FMP Low Duration % 17 0.556 ± 0.104
FMP Running Medium Duration % 9 0.562 ± 0.076
FMP Running High Duration % 15 0.558 ± 0.131
FMP Dynamic Medium Duration % 15 0.534 ± 0.106
FMP Dynamic High Duration % 20 0.601 ± 0.107
FMP Very Low Duration 1 0.357
FMP Low Duration 20 0.689 ± 0.108
FMP Running Medium Duration 21 0.613 ± 0.102
FMP Running High Duration 22 0.619 ± 0.121
FMP Dynamic Medium Duration 26 0.638 ± 0.134
FMP Dynamic High Duration 26 0.652 ± 0.140
FMP Running Medium þ High 21 0.627 ± 0.106
FMP Dynamic Medium þ High 26 0.650 ± 0.141
FMP Running þ Dynamic Medium 23 0.656 ± 0.121
FMP Running þ Dynamic High 26 0.647 ± 0.134
N: number of significant correlations

Formation of clusters based on the relationship between ΔCKABS and FMP

On an individual level the correlation coefficients were trivial to very large (Table 2).
After the correlation results were obtained, a PCA was performed first, and player clusters
were created. The correlation coefficients of the selected FMP parameters and CK values were
utilized for the PCA. The first principal component (PC1) obtained explained 63.8% of the
variance in the data, and the second principal component (PC2) explained 19.2%, the two
together accounted for a total of 83.0%.
The clustering was performed for several cluster numbers, and the silhouette score was used
to determine the proper cluster number. Based on the results, three clusters were created to find
the most accurate fit with the lowest possible number of clusters (the Silhouette score for 4
clusters was 0.61).
Figure 1 represents the cluster distribution of the results of the highest Silhouette scores as a
result of the principal component analysis. Each player in our dataset was categorized into one
of the clusters based on their individual CK responses of the previous day’s FMP load
distribution.
The average values of the individual correlation coefficients in each cluster are presented in
Table 3.

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

Physiology International 7

Table 2. Correlation and standard error values for each participant which forms the basis of the clustering
process (p P ≤ 0.05; pp P ≤ 0.01). Large inter-individual differences reinforce the need for clustering and an
individual approach towards CK response to external match and training load
FMP
FMP Running Dynamic
FMP Low Medium FMP Running Medium FMP Dynamic
Name Duration Duration High Duration Duration High Duration Clusters
pp pp p pp pp
Player1 0.60 ± 0.01 0.57 ± 0.00 0.51 ± 0.00 0.54 ± 0.00 0.58 ± 0.00
Player2 0.9 ± 0.01pp 0.64 ± 0.01pp 0.62 ± 0.00p 0.87 ± 0.01pp 0.90 ± 0.00pp
Player3 0.70 ± 0.01pp 0.71 ± 0.00pp 0.56 ± 0.00p 0.75 ± 0.00pp 0.75 ± 0.00pp
Player4 0.63 ± 0.00pp 0.48 ± 0.00p 0.61 ± 0.00pp 0.71 ± 0.00pp 0.70 ± 0.00pp
Player5 0.55 ± 0.01pp 0.52 ± 0.00pp 0.52 ± 0.00pp 0.68 ± 0.00pp 0.73 ± 0.00pp
Player6 0.78 ± 0.00pp 0.80 ± 0.00pp 0.68 ± 0.00pp 0.79 ± 0.00pp 0.8 ± 0.00pp
Player7 0.79 ± 0.01pp 0.61 ± 0.00pp 0.82 ± 0.00pp 0.79 ± 0.00pp 0.82 ± 0.00pp
Player8 0.65 ± 0.02pp 0.63 ± 0.01p 0.57 ± 0.00p 0.69 ± 0.01pp 0.69 ± 0.01pp All
Player9 0.69 ± 0.01pp 0.63 ± 0.01pp 0.77 ± 0.00pp 0.78 ± 0.01pp 0.79 ± 0.00pp responders
Player10 0.72 ± 0.01pp 0.57 ± 0.01pp 0.71 ± 0.00pp 0.67 ± 0.01pp 0.70 ± 0.00pp
Player11 0.84 ± 0.01pp 0.74 ± 0.00pp 0.59 ± 0.00pp 0.77 ± 0.00pp 0.78 ± 0.00pp
Player12 0.83 ± 0.01pp 0.72 ± 0.00pp 0.87 ± 0.00pp 0.81 ± 0.00pp 0.86 ± 0.00pp
Player13 0.78 ± 0.01pp 0.67 ± 0.00pp 0.79 ± 0.00pp 0.76 ± 0.00pp 0.75 ± 0.00pp
Player14 0.66 ± 0.00pp 0.60 ± 0.00pp 0.73 ± 0.00pp 0.63 ± 0.00pp 0.71 ± 0.00pp
Player15 0.66 ± 0.01pp 0.52 ± 0.00p 0.60 ± 0.00p 0.62 ± 0.01p 0.57 ± 0.00p
Player16 0.67 ± 0.01pp 0.76 ± 0.00pp 0.68 ± 0.00pp 0.73 ± 0.00pp 0.67 ± 0.00pp
Player17 0.37 ± 0.01 0.59 ± 0.01pp 0.29 ± 0.00 0.55 ± 0.01p 0.48 ± 0.01p Running
Player18 0.31 ± 0.01 0.34 ± 0.01 0.50 ± 0.00p 0.46 ± 0.01 0.55 ± 0.00p and
Player19 0.28 ± 0.01 0.69 ± 0.00pp 0.56 ± 0.00p 0.39 ± 0.01 0.42 ± 0.00 Dynamic
Player20 0.40 ± 0.01 0.45 ± 0.00 0.52 ± 0.00p 0.35 ± 0.01 0.54 ± 0.00p high
responders
Player21 0.62 ± 0.01p 0.44 ± 0.01 0.52 ± 0.00p 0.58 ± 0.01p 0.57 ± 0.00p Low
Player22 0.70 ± 0.01pp 0.38 ± 0.00 0.42 ± 0.00 0.58 ± 0.00pp 0.63 ± 0.00pp Intensity,
Player23 0.56 ± 0.01p 0.53 ± 0.00p 0.46 ± 0.00 0.48 ± 0.00p 0.58 ± 0.00p Dynamic
Player24 0.45 ± 0.01p 0.31 ± 0.01 0.33 ± 0.00 0.54 ± 0.00pp 0.58 ± 0.00pp medium
and high
responders
Player25 0.38 ± 0.01 0.39 ± 0.01 0.23 ± 0.00 0.45 ± 0.01 0.40 ± 0.00
Player26 0.10 ± 0.02 0.32 ± 0.00 0.05 ± 0.00 0.12 ± 0.00 0.06 ± 0.00
Player27 0.25 ± 0.01 0.16 ± 0.01 0.34 ± 0.00 0.44 ± 0.01 0.42 ± 0.00
Player28 0.07 ± 0.01 0.04 ± 0.00 0.34 ± 0.00 0.14 ± 0.00 0.19 ± 0.00 Non-
Player29 0.39 ± 0.02 0.23 ± 0.01 0.16 ± 0.00 0.48 ± 0.01 0.39 ± 0.01 responders
Player30 0.07 ± 0.03 0.24 ± 0.00 0.16 ± 0.00 0.19 ± 0.01 0.39 ± 0.00
Player31 0.38 ± 0.02 0.10 ± 0.01 0.24 ± 0.00 0.43 ± 0.01 0.38 ± 0.01

DISCUSSION
This study examined the relationship between inertial sensor derived Football Movement Profile
data and the day-to-day Creatine Kinase change in elite youth national team soccer players. We

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

8 Physiology International

Fig. 1. Distribution of the correlation in the created 4 clusters. While the clustering process resulted in a 4-
cluster structure, there are apparent differences within clusters. Even if players may not be perfectly
clustered into demarcated groups, trends of individual responsiveness can be identified. The four resultant
clusters are characterised as the following: C1, All responders; C2, Running and dynamic high responders;
C3, Low intensity, Dynamic medium and high responders. The cluster of Non responders is not presented
in the figure as they show no significant correlation values.

Table 3. Average correlation values of CK change and FMP Duration parameters in each created cluster.
The participating players (n 5 31) were clustered into 4 groups based on their CK responsiveness to
intensity categories (Low intensity, Medium intensity running, High intensity running, Medium intensity
dynamic, High intensity dynamic FMP) created taking into account individual correlation values
FMP Running FMP Dynamic
FMP Low Medium FMP Running Medium FMP Dynamic
Cluster N5 Duration Duration High Duration Duration High Duration
1 16 0.71 ± 0.09 0.64 ± 0.09 0.66 ± 0.11 0.73 ± 0.08 0.74 ± 0.09
2 4 0.0 ± 0.0 0.32 ± 0.37 0.39 ± 0.26 0.14 ± 0.27 0.39 ± 0.26
3 4 0.58 ± 0.10 0.13 ± 0.27 0.13 ± 0.26 0.54 ± 0.05 0.59 ± 0.03
4 7 0.23 ± 0.15 0.06 ± 0.26 0.17 ± 0.18 0.32 ± 0.16 0.30 ± 0.18
N: Number of players.

also investigated the feasibility of clustering players based on the individual CK sensitivity to
FMP parameter distribution. For this purpose, we used the FMP and CK data of youth national
team soccer players, who play regularly at international level. We created 4 clusters representing
the FMP-CK response. Our results demonstrate the importance of monitoring the external load
and daily CK changes on an individual basis and highlight the utility of new technology for
measuring football-specific movements.
In elite sports, creatine kinase is commonly used as an indirect marker of muscle damage and
fatigue [6]. Previous studies have described the relationship between soccer load and CK change
after training and matches [6, 12, 15, 17]. CK recovery has shown a prolonged time course after
demanding soccer matches, while limiting the performance of players [8]. It has recently been
found that elevated CK levels corresponded with a reduction in total high-speed distance, very

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

Physiology International 9

high-speed distance, accelerations, decelerations, explosive distance, and maximal velocity

during training sessions [26]. Players participating in more than 60 min of a soccer game have
demonstrated larger increase in CK, hence most studies use this as a cut off threshold when
examining the impact of match play on biomarkers, because this might extend the time of
recovery, especially in congested schedules [27]. In most CK related studies the data from senior
players has been presented, with only a few studies examining the CK characteristics of youth
athletes [12, 16, 28]. Therefore, our results provide support for practitioners working with this
age group. These findings should be used with caution for adult players, as CK kinetics show age
dependency [29].
The relationship between GPS-measured external load parameters and CK has been
extensively investigated in the literature [9, 11–16]. Most studies investigated associations
between CK change and the locomotor (total distance, high intensity distance, sprint distance)
and mechanical (high intensity accelerations and decelerations, changes of directions) load of
soccer player [6, 11, 13, 15, 18, 26, 27]. Volume metrics have been the main focus of studies
investigating the load-CK relationship [12–14, 17], however intensity parameters have
received comparatively less attention [11, 15]. Intensity of soccer-specific movements play a
crucial role in players’ overall load and can be monitored live, creating opportunity for
intervention during training sessions. Therefore, volume and intensity parameters, which are
mainly calculated by normalizing volume for one minute, are equally important when
defining soccer load. Even though the FMP displays load as a volume measure, it is able to
detect and classify soccer-specific movements into intensity categories based on inertial
unit data.
In our research we have found moderate to large negative associations between the day-to-
day ΔCKABS and the duration spent in Very Low Intensity (r 5 0.36; P < 0.05) and Very Low
Intensity % FMP parameter (r 5 0.56; P < 0.05). From this result we can conclude that the
distribution of the very low intensity component of movement during training sessions might
have a close relationship with recovery, which is represented in the decrease of the next day’s CK
levels. In national team settings matches are played every 2–3 days, therefore the day after a
game is normally a recovery session of very light intensity for the players. In this respect,
training sessions are designed to exclude high-intensity movements, promoting the regeneration
of players. Although a significant correlation does not demonstrate causation, it is possible that
the greater proportion of low-intensity movements (and lower proportion of high-intensity
movements) might explain, at least in part, the lower ΔCKABS. Given the correlation values, it
seems likely that the reduction of CK is related more to the large percentage of very low intensity
activity, than the duration of the session. Therefore, for recovery benefits, reducing the intensity
of recovery training may be more important than altering the duration of the session.
While we investigated the relationship between the basic FMP % distribution and other
aggregated parameters, they did not provide added value; ΔCKABS correlated best with the FMP
Duration metrics in our dataset. Large correlations were found between ΔCKABS and the Low
(r 5 0.69 ± 0.11; n 5 20), Medium (r 5 0.61 ± 0.10; n 5 21) and High Intensity Running
(r 5 0.62 ± 0.12; n 5 22), and Medium (r 5 0.64 ± 0.13; n 5 26) and High Intensity Dynamic
FMP parameters (r 5 0.65 ± 0.14; n 5 26). This finding partially confirms our hypothesis and
supports the trend of previous studies, that high intensity locomotor and mechanical GPS
parameters are closely associated with the CK response [11–14]. Our use of inertial sensor-based
metrics is of particularly importance as GPS telemetry is largely dependent on satellite

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

10 Physiology International

availability, which is influenced by closed stadiums and sessions performed in indoor facilities
[30]. Practitioners are increasingly relying on inertial measurement sensor data as it shows
smaller measurement error than GPS [4]. FMP represents a practical tool for monitoring the
load of soccer-specific movements. FMP parameters also offer an easier method of interpreting
soccer-specific load, as it does not use multiple different variables, as is the case with GPS
parameters.
Most studies have examined the CK response to external training and match load [11, 13, 14],
however the load-CK response appears to be individual-specific. In our previous study with youth
soccer players, we found a significant soccer-specific position and training type dependency on the
subsequent CK change [16]. Although in our dataset ΔCKABS showed large correlations with both
medium and high intensity dynamic FMP parameters, certain players displayed nearly perfect
relationships (r 5 0.90). These results also show large inter-individual differences, further sug-
gesting an individual responsivity of CK to soccer-specific activity (Table 3).
The large inter-individual variance in the correlation analysis of FMP variables and ΔCKABS
supports the clustering of players based on their individual responsiveness. The players in our
dataset were grouped into 4 distinct clusters, depending on their individual correlations of FMP
parameters and ΔCKABS. While the clustering process is not perfect, we aimed to create the
lowest possible number of groups into which players could be classified. A 0.61 Silhouette score
shows a reasonable structure for this method, supporting the feasibility and importance of such
grouping of players. The created clusters are as follows:
1. All responders (n 5 16): Players’ ΔCKABS have a large correlation (r 5 0.64–0.74) with all
presented FMP parameters and these values cannot be definitely distinguished. These players
seem to be responsive to almost equally any type of load characterized by FMP parameters.
2. Running and dynamic high responders (n 5 4): these players show the highest, although
statistically moderate correlation values with Running parameters and the Dynamic-high
parameters (r 5 0.32–0.39).
3. Low intensity, dynamic-medium and high responders (n 5 4): The players in this cluster
showed large correlations among ΔCKABS and Low intensity (r 5 0.58), Dynamic medium
(r 5 0.54) and high (r 5 0.59) parameters. The CK response appears to occur as a result of a
more dynamic load in these players.
4. Non-responders (n 5 7): The ΔCKABS of these players does not show statistically significant
associations with FMP values. Further investigations into the type of load that elicits a
ΔCKABS is warranted in these players (Tables 2 and 3).
We conclude that the resultant clusters in our dataset are highly population dependent and
even though players can be grouped based on their CK sensitivity to FMP, there are large inter-
individual differences which indicate the need for an individual-specific approach. However, our
results may help practitioners to better understand training and match load response and
manage potential fatigue of players on an individual basis.
Our study has several limitations that warrant discussion. First, in order to increase our
sample size, we included 16–21-year-old players; as such, the age of participants has a large
variance. Second, as the Football Movement Profile is a specific algorithm developed by a
commercial GPS supplier, the exact algorithm is not publicly available. Finally, the CK response
of players may be affected by factors other than training and match load, including direct muscle
contusions, resistance training and individual neuroendocrine characteristics. In conclusion,

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

Physiology International 11

we found a relationship between FMP parameters and the day-to-day CK change but large inter-
individual differences exist in the players’ response. Closely monitoring the intensity distribu-
tion, in addition to traditional volume parameters, may provide a better understanding of the
individual CK response to soccer-specific load.

PRACTICAL APPLICATION
Creatine Kinase is commonly used in elite soccer as an objective marker of muscle damage in
soccer players, as its elevation is a sign of extensive load on the soft tissues resulting from high-
intensity locomotive and mechanical load. Even though it has been found that CK change shows
a relationship between the overall volume of several GPS parameters, such as high intensity and
sprint distance, as well as accelerations and decelerations, we also suggest tracking the intensity
of soccer-specific movements. For this purpose, FMP is a viable option as it classifies activities
into specific zones, based on their locomotor orientations and intensity. In addition, as FMP
relies solely on inertial sensors, and it is not affected by GPS signal, it can be monitored more
accurately. FMP may help coaches and practitioners to better understand load of soccer-specific
movements and could potentially decrease the number of parameters used to monitor players’
activities. Individual and team physical preparation can be based upon FMP by defining specific
target values or thresholds. We can also forecast the players’ next day CK response based on
their cluster classification and response to different types of soccer-specific load. Consequently,
more precise planning of recovery strategies is possible, which is necessary during the congested
schedules of national team settings. Despite the positive properties of FMP monitoring, we
suggest using it in conjunction with the traditionally used GPS metrics to obtain a more accurate
individual profile of CK response to the load of soccer-specific movements.

Conflict of interest: The authors declare no conflict of interest and they have no professional
relationships with any company or manufacturer who would benefit from the results of the
current study.

ACKNOWLEDGEMENTS
The authors have no conflicts of interest to declare. Special thanks are owed to the youth
national team soccer players and coaches who participated in the study and the Hungarian
Football Federation for supporting this study.

REFERENCES

1. Barnes C, Archer DT, Hogg B, Bush M, Bradley PS. The evolution of physical and technical performance
parameters in the English Premier League. Int J Sports Med 2014; 35(13): 1095–100. https://doi.org/10.1055/
s-0034-1375695.
2. Teixeira JE, Forte P, Ferraz R, Leal M, Ribeiro J, Silva AJ, et al. Monitoring accumulated training and match
load in football: a systematic review. Int J Environ Res Public Health 2021; 18(8): 3906. https://doi.org/10.
3390/ijerph18083906.

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

12 Physiology International

3. Hoppe WM, Baumgart C, Polglaze T, Freiwald J. Validity and reliability of GPS and LPS for measuring
distances covered and sprint mechanical properties in team sports. PLoS One 2018; 13(2): e0192708. https://
doi.org/10.1371/journal.pone.0192708.
4. Pillitteri G, Thomas E, Battaglia G, Navarra GA, Scardina A, Gammino V, et al. Validity and reliability of an
inertial sensor device for specific running patterns in soccer. Sensors (Basel) 2021; 21(21): 7255. https://doi.
org/10.3390/s21217255.
5. Szigeti G, Schuth G, Kovacs T, Pavlik G, Barnes C. The football movement profile of youth national team
players. Sport Perform & Sci Rep 2021; 1(Nov): 145.
6. Silva JR, Rumpf MC, Hertzog M, Castagna C, Farooq A, Girard O, et al. Acute and residual soccer match-
related fatigue: a systematic review and meta-analysis. Sports Med 2018; 48(3): 539–83. https://doi.org/10.
1007/s40279-017-0798-8.
7. Baird MF, Graham SM, Baker JS, Bickerstaff GF. Creatine-kinase- and exercise-related muscle damage im-
plications for muscle performance and recovery. J Nutr Metab 2012; 2012: 960363. https://doi.org/10.1155/
2012/960363.
8. Doeven SH, Brink MS, Kosse SJ, Lemmink KAPM. Postmatch recovery of physical performance and
biochemical markers in team ball sports: a systematic review. BMJ Open Sport Exerc Med 2018; 4(1):
e000264. https://doi.org/10.1136/bmjsem-2017-000264.
9. Wiig H, Raastad T, Luteberget LS, Ims I, Spencer M. External load variables affect recovery markers up to 72 h
after semiprofessional football matches. Front Physiol 2019; 10: 689. https://doi.org/10.3389/fphys.2019.00689.
10. Becker M, Sperlich B, Zinner C, Achtzehn S. Intra-individual and seasonal variation of selected biomarkers
for internal load monitoring in U-19 soccer players. Front Physiol 2020; 11: 838. https://doi.org/10.3389/
fphys.2020.00838.
11. Csala D, Kovács BM, Bali P, Reha G, Pánics G. The influence of external load variables on creatine kinase
change during preseason training period. Physiol Int 2021; 108: 371–82. https://doi.org/10.1556/2060.2021.
30019.
12. de Hoyo M, Cohen DD, Sañudo B, Carrasco L, Álvarez-Mesa A, del Ojo JJ, et al. Influence of football match
time-motion parameters on recovery time course of muscle damage and jump ability. J Sports Sci 2016;
34(14): 1363–70. https://doi.org/10.1080/02640414.2016.1150603.
13. Hader K, Rumpf MC, Hertzog M, Kilduff LP, Girard O, Silva JR. Monitoring the athlete match response: can
external load variables predict post-match acute and residual fatigue in soccer? A Systematic Review with
Meta-analysis. Sports Med Open 2019; 5(1): 48. https://doi.org/10.1186/s40798-019-0219-7.
14. Hecksteden A, Meyer T. Blood-borne fatigue markers during major international football tournaments – a
retrospective analysis of data from the FIFA World Championships and UEFA European Championships
2006–2016. Sci Med Footb 2020; 4: 135–41. https://doi.org/10.1080/24733938.2019.1692144.
15. Russell M, Sparkes W, Northeast J, Cook CJ, Bracken RM, Kilduff LP. Relationships between match activities
and peak power output and creatine kinase responses to professional reserve team soccer match-play. Hum
Mov Sci 2016; 45: 96–101. https://doi.org/10.1016/j.humov.2015.11.011.
16. Schuth G, Szigeti G, Dobreff G, Revisnyei P, Pasic A, Toka L, et al. Factors influencing creatine kinase
response in youth national team soccer players. Sports Health 2021; 13(4): 332–40. https://doi.org/10.1177/
1941738121999387.
17. Thorpe R, Sunderland C. Muscle damage, endocrine, and immune marker response to a soccer match.
J Strength Cond Res 2012; 26(10): 2783–90. https://doi.org/10.1519/JSC.0b013e318241e174.
18. Varley I, Lewin R, Needham R, Thorpe RT, Burbeary R. Association between match activity variables,
measures of fatigue and neuromuscular performance capacity following elite competitive soccer matches.
J Hum Kinet 2017; 60: 93–9. https://doi.org/10.1515/hukin-2017-0093.

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

Physiology International 13

19. Luteberget LS, Holme BR, Spencer M. Reliability of wearable inertial measurement units to measure physical
activity in team handball. Int J Sports Physiol Perform 2018; 13(4): 467–73. https://doi.org/10.1123/ijspp.
2017-0036.
20. Barrett S. Monitoring elite soccer players’ external loads using real-time data. Int J Sports Physiol Perform
2017; 12(10): 1285–7. https://doi.org/10.1123/ijspp.2016-0516.
21. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple
hypothesis testing. J R Stat Soc Ser B Methodol 1995; 57: 289–300. https://doi.org/10.1111/j.2517-6161.1995.
tb02031.x.
22. Hopkins WG, Marshall SW, Batterham AM, Hanin J. Progressive statistics for studies in sports medicine and
exercise science. Med Sci Sports Exerc 2009; 41(1): 3–12. https://doi.org/10.1249/MSS.0b013e31818cb278.
23. Pearson KFRS. LIII. On lines and planes of closest fit to systems of points in space. The Lond Edinb Dublin
Philosophical Mag and J Sci 1901; 2(6): 559–72. https://doi.org/10.1080/14786440109462720.
24. Lloyd S. Least squares quantization in PCM. IEEE Trans Inf Theor 1982; 28(2): 129–37. https://doi.org/10.
1109/TIT.1982.1056489.
25. Rousseeuw PJ. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J Comput
Appl Math 1987; 20: 53–65. https://doi.org/10.1016/0377-0427(87)90125-7.
26. Malone S, Mendes B, Hughes B, Roe M, Devenney S, Collins K, et al. Decrements in neuromuscular per-
formance and increases in creatine kinase impact training outputs in elite soccer players. J Strength Cond Res
2018; 32(5): 1342–51. https://doi.org/10.1519/JSC.0000000000001997.
27. García-Romero-Pérez Á, Ordonez FJ, Reyes-Gil F, Rodríguez-López ES, Oliva-Pascual-Vaca Á. Muscle
damage biomarkers in congestion weeks in English Premier League soccer players: a prospective study for two
consecutive seasons. Int J Environ Res Public Health 2021; 18(15): 7960. https://doi.org/10.3390/
ijerph18157960.
28. De Ste Croix M, Lehnert M, Maixnerova E, Zaatar A, Svoboda Z, Botek M, et al. Does maturation influence
neuromuscular performance and muscle damage after competitive match-play in youth male soccer players?
Eur J Sport Sci 2019; 19(8): 1130–9. https://doi.org/10.1080/17461391.2019.1575913.
29. Neil RC, Ferdinand KC, Ycas J, Miller E. Relationship of ethnic origin, gender, and age to blood creatine
kinase levels. Am J Med 2009; 122(1): 73–8. https://doi.org/10.1016/j.amjmed.2008.08.033.
30. Ravé G, Granacher U, Boullosa D, Hackney AC, Zouhal H. How to use Global Positioning Systems (GPS)
data to monitor training load in the “Real World” of elite soccer. Front Physiol 2020; 11: 944. https://doi.org/
10.3389/fphys.2020.00944.

Authenticated gyorgyszigeti87/ Author's copy | Downloaded 01/26/23 10:54 AM UTC

View publication stats