This cross-sectional study followed the guidelines and the Declaration of Helsinki, and the protocol was approved by the university's institutional review board (QU-IRB 1542-FBA/21) and the Ministry of Education and Higher Education Qatar (REF: 18/2021). Independent from the generally accepted criteria for obesity and overweight and population levels, participants were divided into three groups depending on BMI age-adjusted values (Harrington et al., 2013; Hermassi et al., 2021a), namely, lowest (n = 21; BMI: 18.0–20.9 kg/m²), middle (n = 23; BMI: 21.0–22.9 kg/m²), and highest (n = 25, BMI: > 23.0 kg/m²). A total of 69 male schoolchildren (age: 12.4 ± 0.7 years of age, range: 11–14 years; body mass: 58.5 ± 7.2 kg; stature: 1.62 ± 0.09 m; and BMI: 22.4 ± 3.3 kg/m²) were analyzed. Of them, 85% (n = 59) of the children aged 12 or 13 years. Only 6 (9%) and 4 (6%) were younger (i.e., 11 years) or older (i.e., 14 years). Children were recruited by convenience sampling from one school in a Doha Community (Qatar). Participants were excluded if they were: (1) diagnosed with a psychological disorder (e.g., anxiety, depression, or attention-deficit disorder) or (2) on medication (e.g., antidepressants or medication affecting the nervous system). Moreover, participants that failed to present a signed informed consent form by their legal guardians were further excluded. Study objectives were communicated to the parents, school, and the management team. Parents or legal guardians of the participants were provided with an information sheet explaining the purpose of the experiment and a consent form. Of these participants, 65% returned a signed consent form and were eligible to participate. Participants and their respective guardians and teachers were debriefed about the experimental procedure, right to withdraw from the experiment, and provided with a consent form to complete for participation in the experiment, in accordance with the Helsinki Declaration.

Data collection took place between 08:00 h and 10:00 h every day in an indoor sport court with consistent environmental conditions (temperature 24.5°C ± 0.5°C and relative humidity 65% ± 5%). Participants were asked to maintain regular food and drinks habits but to abstain from drinking caffeinated products, performing vigorous PA, and eating for 24, 4, and 2 h before testing, respectively. Testing was performed in a fixed order over a period of 4 days to ensure consistent fatigue levels and learning effects between participants. The tests for this experiment were selected based on the suitability for middle-school students and to evaluate these dimensions:

After a general warm-up consisting of 5 min low-intensity running, 3 × 15 m progressive accelerations and a maximal 20 m sprint, interspersed with 3 min periods of passive recovery were performed. In addition to the low-intensity running, submaximal dynamic stretches and throws were performed, consistent with our previous work. Day 1 consisted of anthropometric testing and the Yo-Yo IR1. On day 2, the stork test of static balance was performed. Day 3 consisted of sprint performance and handgrip force. On day 4, testing consisted of the agility T-half test and medicine ball throw. Test-retest reliability was assessed by repeating fitness tests from day 1 to 3 after 2 weeks of the initial testing period. The second set of test scores generated for the test-retest reliability purposes was included in final analyses. Anthropometric and Yo-Yo IR1 test assessments were performed once at the initial testing period, for players' limited time commitments and schedules.

Anthropometric measurements incorporated body height (Holtain stadiometer, Crosswell, Crymych, Pembrokeshire, United Kingdom) and body mass (model TBF 105; Tanita Corporation of America, Inc, Arlington Heights, Illinois) measured to the nearest 0.1 cm and 0.1 kg. BMI was calculated as the ratio between body mass (kg) and body height squared (m²).

The Yo-Yo IR1 was conducted in line with Krustrup et al. (2003). A standardized warm-up consisted of 5 min of low-intensity running. Next, 20 m shuttle runs were performed at increasing velocities until exhaustion, with 10 s intervals of active recovery (2 × 5 m of jogging) between runs. The test was terminated if it met objective criteria (with the participant twice failing to reach the front line in time) and/or subjective criteria (with the participant feeling unable to complete another shuttle at the required speed). The total distance covered was considered as the test “score.”

To assess static balance, we utilized the stork balance test (Miller, 2002). Participants stood with their opposite foot against the inside of the supporting knee with both hands on the hips. On the “go” signal, subjects raised the heel from the floor and held this position for as long as possible. The test was terminated when the heel of the supporting leg touched the ground, or the foot moved away from the kneecap. The test was timed using a stopwatch.

A 15-min warm-up with 10 min running, change of direction activities, and dynamic stretching was performed. Next, participants sprinted 15 m from a standing position 0.2 m behind the first photocell beam. The 15 and 30 m sprint times were recorded by paired photocells (Racetime 2 SF, Microgate, Italy) located 1 m above the ground at the start and finish. Three trials were separated by 6–8 min of recovery, and the fastest trial was retained for further analyses.

A standard adjustable digital hand-grip dynamometer (T.K.K. 5401, Tokyo, Japan) was employed to measure grip strength of the dominant hand, with a sensitivity of 10 N. The anthropometric apparatus and hand-grip dynamometer were calibrated before use. Participants were tested 3 min after a separate warm-up and before a throwing velocity test, following the anthropometry assessment. This test was conducted with the arm extended parallel to the body. Participants were not allowed to display movements of the arm or wrist. Peak force was recorded. Players performed two repetitions of maximum intensity with 3 min break between each repetition to minimize fatigue. The final analyses included the best trial only.

Prior to each test, a 10 min warm-up consisting of jogging, jumping, lateral displacements, and dynamic stretching was performed. Electronic timing sensors (photocells, Kit Racetime 2 SF, Microgate, Italy) were employed to record T-half tests (Sassi et al., 2009) data. The electronic timing sensors were set 0.75 m above the floor, 3 m apart, and facing each other at the starting line A. Participants started each trial with their front foot 0.2 m behind line A. Next, participants sprinted forward to cone B and touched its base with their right hand. Without crossing their feet and facing forward, participants shuffled to the left of cone C and touched the base of this cone with their left hand. Thereafter, participants shuffled right to cone D, touched its base with their right hand, then ran back to the left of the cone, and touched its base. At last, participants ran backward, returning to line A as quickly as possible. Participants had to repeat a trial if they crossed one foot in front of another, failed to touch the base of the cone, and/or failed to face forward throughout the trial. Participants repeated until they completed two successful trials; there was a 3 min break between the trials, and only the best trial was included in the final analyses.

Prior to this test, participants performed a 5 min warm-up that included a 3 min run and dynamic activities. Throws were performed using rubber medicine balls with a diameter of 21.5 cm and a weight of 2 kg. Participants were debriefed about the optimal technique, determining an optimal release angle for obtaining the maximum distance (Negrete et al., 2010). The sitting participant clasped the medicine ball using both hands and pushed the ball forcefully from their chest when given a signal. The score was determined based on the place the ball landed from the sitting line's front. Each participant performed three trials with 1 min break between, and the best result was reported to 0.01 m.

Academic performance was evaluated through school records. Academic performance consisted of actual grade point average (GPA) and score (0–100) as endorsed in the Qatar State in mathematics and science from the second semester of the academic year 2020–2021. The reason for only including two academic subjects was due to our interest in science-related courses. We focused on mathematics and science due to the greater association with fitness.

The Statistical Package for the Social Sciences (SPSS) version 28.0 (IBM, Armonk, NY, USA) software was utilized to perform all analyses. Retest reliability was described using the intraclass correlation coefficient (ICC) and the coefficient of variation (CV) (Schrama et al., 2014). Previously reported guidelines helped to inform reliability (Shrout and Fleiss, 1979; Hopkins, 2000; Portney and Watkins, 2009; Hopker et al., 2010). A power calculation (nQuery Advisor 4.0; Statistical Solutions, Saugus, MA, USA) showed with n = 7 in each group; we would have been able to detect a mean difference of 0.4 s in 10 m sprint using a two-sided t-test with the α level of 0.05 and a pooled standard deviation (SD) of 0.3 s, with a statistical power of 0.8 (Bortz, 1999). Data were analyzed for homogeneity of variance (Levene's test), and normality (Shapiro-Wilk test). Based on the sample size of n = 69 and the symmetrical sample distribution (n = 21, n = 23, and n = 25), it was possible and legitimate to use parametric tests (Weiß, 2002), such as one-way analysis of variance (ANOVA), to examine the anthropometric and performance differences between BMI groups. Differences were considered meaningful if p < 0.05 and partial eta-squared (ηp2) > 0.15 (Bortz, 1999; Richardson, 2011), to refrain from overestimating the differences. T-tests with Bonferroni correction were conducted for pairwise comparison purposes (0.05/17 or p = 0.003) to protect against type I error (Bortz, 1999). Pearson's product moment correlation coefficients (r) were conducted to determine relationships between anthropometric, physical performance, and academic performance variables and interpreted employing Cohen's thresholds (Cohen, 1988). Magnitudes of correlations (r) were thus interpreted as follows: r < 0.1; the correlation was trivial; 0.1–0.3, it was small; in the range of 0.3–0.5, moderate; 0.5–0.7, large; 0.7–0.9, very large, and 0.9–1.0, almost perfect (Cohen, 1988).

All six performance tests showed excellent relative reliability (ICC ≥ 0.75). ICC ranged from 0.87 (stork balance test) to 1.00 (all other tests with the omission of medicine ball throw). Apart from the stork balance test (CV = 13.9%), all variables had excellent absolute reliability, with CV < 5% (Table 1).

Only two variables (i.e., sprint 15 m: p = 0.285; medicine ball throw: p = 0.059) were normally distributed. Only the parameters, such as 10 m sprint (p = 0.001), Yo-Yo IR1 (p = 0.001), and science (p = 0.007), were heterogeneous in variance. Otherwise, all p-values were >0.110.

All three groups were not different concerning age (ηp2 = 0.03) but showed differences regarding anthropometric parameters (Table 2).

The largest difference was observed for the parameter BMI (ηp2 = 0.69). The BMI of the lowest group (19.1 ± 1.5 kg/m²) was markedly lower than for the middle group (25.5 ± 2.2 kg/m²) and the highest group (25.6 ± 2.3 kg/m²). Differences in body height (ηp2 = 0.30) and body mass (ηp2 = 0.35) were of a smaller magnitude.

Regarding physical performance and academic performance parameters, we found three significant group differences (i.e., 10 m sprint, mathematics, and science) (Table 3). The largest difference was calculated for 10 m sprint (ηp2 = 0.26).

Significant effects were observed only between the middle and highest groups (i.e., medicine ball throw, sprint 10 m, mathematics, and science) or between lowest and middle participants (i.e., sprint 10 m, Arabic, mathematics, and science). The largest difference (p < 0.001) was observed between overweight (2.25 ± 0.22 s) and obese (2.68 ± 0.34 s) subjects for the 10 m sprint. The largest number of pairwise differences (5) was calculated for academic performance parameters. The differences ranged from p = 0.002 (mathematics: middle vs. highest) to p = 0.035 (science: middle vs. highest).

In contrast to the academic performance (all maxima in the middle group), the performance maxima concerning physical performance are distributed among the three groups (Table 3). Most of the maxima (5) were found for the middle group (parameters: Yo-Yo IR1, sprint 10 m, sprint 15 m, storck balance test, and agility T-half). The subjects of the highest group showed the highest performance for the medicine ball throw (5.16 ± 0.61 m) and handgrip force (30.1 ± 11.9 N).

We did not find any relevant (r > 0.5) correlation between parameters of different dimensions (e.g., academic vs. physical performance; anthropometric vs. physical performance). The parameters medicine ball throw and handgrip force showed the highest interaction (r = 0.470) with each other. Regarding the differentiating parameter BMI, all product moment correlations were below r = 0.303 (95% CI: 0.071–0.504, medicine ball throw).

During school physical education classes, short sprints, changes of direction, accelerations, and decelerations are fundamental abilities developed and utilized (Hermassi et al., 2021c). It is well-known that obesity is negatively associated with physical fitness. However, it has been reported that BMI is a poor surrogate for obesity (which is purportedly a disease of excessive fatness, although not if measured by BMI) in adolescents (Shang et al., 2010; Kruschitz et al., 2013; Hermassi et al., 2021d). We found significant group differences regarding 10 m sprint performance, and the largest difference was observed between the middle and highest BMI group (p < 0.001). This has implications for school student's success as they commonly sprint for ~10–30 m in game-defining epochs during physical education classes.

Most of the maximal values (5) were exhibited by the middle BMI group (parameters: Yo-Yo IR1, 10 m sprint, 15 m sprint, stock balance test, and agility T-half). The highest BMI group exhibited the greatest performance for the medicine ball throw and handgrip force. This is relatively unsurprising, as the highest BMI group would exhibit the greatest absolute volume of lean mass, which is a key determinant of muscle force (Bandini et al., 1990; Esmaeilzadeh and Ebadollahzadeh, 2012; Esmaeilzadeh and Kalantari, 2013). In this context, Esmaeilzadeh and Kalantari (2013) controlled for body fat mass and observed that in some physical fitness tests (such as sit and reach, standing long jump, and run speed tests), overweightness was associated with improved performance compared with underweight counterparts, indicating fat mass as a limiting factor for test performance. Results of Esmaeilzadeh and Kalantari (2013) indicated that underweight adolescents possessed the highest cardiorespiratory fitness among the weight statuses, but being underweight was related to poorer performance of some physical fitness tests (primarily related to muscle strength). It is possible that their poorer performance resulted from lower fat-free mass compared with that in normal and overweight counterparts, which was supported by these results when controlling for fat-free mass.

We found relevant group differences regarding the academic performance parameters mathematics (ηp2 = 0.19) and science (ηp2 = 0.15). For Arabic language (ηp2 = 0.10), the group difference did not reach the level of relevance (p < 0.05 and ηp2 ≥ 0.15). Furthermore, the largest number of pairwise differences (5) was calculated for academic performance parameters. The differences ranged from p = 0.002 (mathematics: middle vs. highest) to p = 0.035 (science: middle vs. highest). These data corroborate previous cross-sectional studies that observed fitter students exhibited greater academic attainment (Davis and Cooper, 2011; Hraste et al., 2018). To date, some investigations reported an association between body composition (not part of our study) and academic performance (Sabia, 2007; Sigfúsdóttir et al., 2007; Shore et al., 2008). For example, our previous work demonstrated that a nonobese adolescent group had a higher mean academic performance than the obese group (Hermassi et al., 2021d). The observed difference between groups was markedly greater for mathematics (ηp2 = 0.367) than for science (ηp2 = 0.195).

It is well-known that adiposity has negative effects on cognition, learning, and memory (Li et al., 2008; Yu et al., 2009). However, weight-related bias or discrimination may further influence self-esteem and behavioral problems, mediating or moderating obese children's academic performance (Fowler-Brown et al., 2010; Griffiths et al., 2010). Uncertainty remains whether excess adiposity itself affects academic performance or if this effect is mediated by other factors observed in individuals with obesity. A systematic review noted a relationship between obesity, cognitive function, and academic attainment in adolescents (Kamijo et al., 2012). In this study, the highest group had poorer academic achievement, suggesting that being obese during adolescence has profound academic consequences. Paradoxically, other studies have observed no effect of obesity on academic achievement in children (Kwak et al., 2009; Van Dusen et al., 2011). Reasons for result divergence remain unclear, but differences in methods for quantifying obesity (i.e., body fat or BMI), participant characteristics, growth stages, or analysis procedures for academic attainment (Kwak et al., 2009; Ruiz et al., 2010) may explicate some ambiguity. Furthermore, if previous studies are interpreting results solely on the alpha level and using an arbitrary threshold of p < 0.05 with small sample sizes, it is possible these studies fell afoul of type II error. In this context, based on Cohen's thresholds (Cohen, 1988), the present magnitudes were classified as “small,” and thus the highest BMI group may only experience a small negative effect on academic performance, which may be missed by underpowered research studies. If studies use different analysis procedures for academic attainment, this may also explain differences in results as obesity is not associated with lower test scores (Datar and Sturm, 2006) but does associate with lower GPA (Li et al., 2008). This is puzzling, as one may assume lower test scores would result in lower GPA; however, it does emphasize (a) the importance of specifying how academic performance was quantified and (b) the multifactorial nature of how obesity associates with GPA, as it is clearly not simply a case of poorer test performance.

Limitations of this investigation include not assessing sexual maturity (i.e., biological maturation) as growth spurt timing differences could influence BMI and relationships between BMI and physical fitness. A further limitation is that since the study was cross-sectional, it is difficult to attribute causality. For a wider interpretation and scope, it is necessary to involve female subjects. In addition, the parameter body fat should be assessed to distinguish between fat and lean mass. Finally, we did not use the generally accepted criteria for obesity and overweight. Instead of this, we divided the sample into tertiles based on BMI.