Over the past two decades, research in the domain of talent identification and development in youth soccer has grown exponentially. Anthropometric, motor coordination and physical performance measures (i.e., explosivity, speed and endurance) have shown to discriminate between successful and less successful youth soccer players (Figueiredo et al., 2009; Vaeyens et al., 2006), and are thought to be predictive for future adult soccer success (Gonaus and Muller, 2012; Le Gall et al., 2010). However, biological maturation confounds these identification and selection processes as late maturing players are systematically excluded as age and sports specialization increases (Malina et al., 2000).
Longitudinal designs are necessary in defining pathways to excellence and maturational status should be considered when evaluating young athletes (Malina et al., 2000; 2004; Vaeyens et al., 2008). Philippaerts et al. (2006) showed that the average age at peak height velocity (13.8 [+ or -] 0.8 years) in 33 male youth soccer players was slightly earlier compared to the general population (between 13.8 and 14.2 years). For example, corresponding data for peak oxygen uptake indicated that maximal gains occur at the time of peak height velocity, with continued improvements during late adolescence (Mirwald and Bailey, 1986). Thus, it would seem that around the age of 14 years, maturational status has a critical impact on the development of physiological characteristics in pubertal athletes, and therefore has strong implications for talent identification and development programs (Baxter-Jones et al., 1993).
The Yo-Yo Intermittent Recovery Test Level 1 (YYIR1) is a field test that measures the ability to (quickly) recover between repeated intensive efforts (e.g., sprinting, tackling, jumping) and that maximally stresses the aerobic energy system through intermittent exertion (Krustrup et al., 2003). Previous studies in youth and adult soccer have shown that the Yo-Yo IR1 performance has an adequate to high level of reproducibility (Deprez et al., 2014; Krustrup et al., 2003) and is a valid measure of prolonged, high intensity intermittent running capacity (Sirotic and Coutts, 2007).
When predicting future success at young age, it is important to know whether anthropometrical and physical performances measures are stable in the long-term. This refers to the consistency of the position or rank of individuals in the group relative to others. A review by Beunen and Malina (1988) showed, that in the general population, the stability of physical fitness was moderate (Maia et al., 2003) to good (Maia et al., 2001) throughout adolescence. They also reported that individuals who performed well for their maturity level during adolescence still had a good possibility of performing above average at the age of 30 (Lefevre et al., 1990). In contrast however, within a general sporting population, the best performing players at young age might not remain the best over one year, accounting for poor long-term stability (Abbott and Collins, 2002). Recently, a longitudinal study in 80 pubertal soccer players showed high stability (ICC's: 0.91 to 0.96) for anthropometric measures, moderate stability (ICC's: 0.66 to 0.71) for sprint, speed and explosive leg power and high stability for maximal aerobic speed (ICC: 0.83) (Buchheit and Mendez-Villanueva, 2013).
However, to date, no such data are available in youth soccer for the intermittent-endurance performance. Therefore, the aim of the present study is to examine the changes in body dimensions and YYIR1 performance in high-level pubertal youth soccer players over two-to-four years. More precisely, we examined whether the baseline values could influence the magnitude of improvement, and whether this improvement is related to maturational status.
Subjects and study design
A longitudinal study design was conducted over a twoand four-year-period. Subjects were 42 young high-level pubertal soccer players aged between 11 and 16 years from two Belgian professional soccer clubs. All players participated in a high-level training program with minimal 7.5 training hours and 1 game (on Saturday) per week. The two-year follow-up subsample included 21 soccer players, aged 13.2 [+ or -] 0.3 y at baseline, who were assessed annually, each time at the end of August (a total of three test moments). In addition, the four-year follow-up subsample included 21 players, aged 12.2 [+ or -] 0.3 y at baseline, who were assessed every second year, each time at the end of August (a total of three test moments).
All subjects and their parents or legal representatives were fully informed about the aim and the procedures of the study before giving their written informed consent. The study was carried out in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the University Hospital.
Stature (0.1 cm, Harpenden Portable Stadiometer, Holtain, UK), sitting height (0.1 cm, Harpenden sitting height table, Holtain, UK) and body mass (0.1 kg, total body composition analyzer, TANITA BC-420SMA, Japan) were assessed according to manufacturer guidelines. Leg length was calculated by subtracting sitting height from stature. All anthropometric measures were taken by the same investigator to ensure test accuracy and reliability. For height, the intra-class correlation coefficient for test-retest reliability and technical error of measurement (test-retest period of one hour) in 40 adolescents were 1.00 (p
An estimation of maturity status was calculated using equation 3 from Mirwald et al. (2002) for boys. This noninvasive method predicts years from peak height velocity as the maturity offset, based on anthropometric variables (stature, sitting height, weight, leg length). Subsequently, the age at peak height velocity (APHV) is determined as the difference between the chronological age and the maturity offset. According to Mirwald et al. (2002), this equation accurately estimates the APHV within an error of [+ or -] 1.14 years in 95% of the cases in boys, derived from three longitudinal studies on children who were four years from and three years after peak height velocity (i.e., 13.8 years). Accordingly, the age range for which the equation confidently can be used is between 9.8 and 16.8 years which matches the present age range (11.7-16.7 y).
High intensity intermittent running performance
High intensity intermittent running performance was investigated using the YYIR1. This test was conducted according to the methods described in Krustrup et al. (2003). Participants were instructed to refrain from strenuous exercise for at least 48 hours before the test sessions and to consume their normal pre-training diet before the test session. All tests were conducted on the same indoor venue with standardized environmental conditions. Players completed the YYIR1 test with running shoes and followed a standardized warm-up. To investigate the effect of baseline high intensity intermittent running performance on its changes over the years, players in each subsample were divided into three performance groups according to their YYIR1 performance at baseline: players which YYIR1 performance was below percentile 33 (P33) were classified as 'low ', between P33 and P66, as 'average' and above P66, as 'high'.
The YYIR1 test has shown good test-retest reliability in 13 adult male experienced soccer players (CV of 4.9%) and in 16 recreational adults (CV of 8.7%), respectively (Krustrup et al., 2002; Thomas et al., 2006). Recently, in a non-elite youth soccer population, Deprez and colleagues (2014) reported a CV of 17.3%, 16.7% and 7.9% for the YYIR1 test in under-13 (n = 35), under-15 (n = 32) and under-17 (n = 11) age groups, respectively, showing adequate to good reliability. However, of importance in interpreting differences between measures, it is not the CV of a measure that matters, but the magnitude of this 'noise' compared with (1) the usually observed changes (signal) and (2) the changes that may have a practical effect (smallest worthwhile difference) (Hopkins, 2004). A measure showing a large CV, but which responds largely to training can actually be more sensitive and useful than a measure with a low CV but poorly responsive to training. The greater the signal-to-noise ratio, the likely greater is the sensitivity of the measure.
All statistical analyses were completed using SPSS for windows (version 20.0). First, for each of the two sub-samples (two- and four-year follow-up, respectively) differences between the three performance groups (low, average and high) were investigated using multivariate analysis of variance (MANOVA) with performance group as independent and age, maturity offset, stature, body mass and YYIR1 as dependent variables. After running normality tests (Shapiro-Wilk) for all dependent variables in each performance group (in both two- and four-year subsamples), the data passed the assumption of normality (p-values between 0.058 and 0.855) (except for MatOffSet (p = 0.019) in the low performance, four-year subsample group). Since MANOVA revealed a significant main effect (Wilks' Lambda) in both the two- (F = 15.517; p
For the two-...