Differential effects of changes in cardiorespiratory fitness on worst- and best- school subjects

The epidemic of physical inactivity and the corresponding reduction in physical fitness during childhood has become a global public health problem. A large body of cross-sectional studies has consistently found a positive association between physical fitness, including cardiorespiratory, muscular, and motor fitness, and academic performance measured by grade points (GP) or standardized test scores. Several longitudinal studies, including observational studies and randomized controlled trials (RCTs), have also supported a positive causal link between improvements in physical fitness via regular physical activity and academic performance. However, recent large-scale cluster RCTs have reported no effect of physical activity interventions on academic performance, casting doubt on causality. Potential moderators include participant characteristics and characteristics of the physical activity. This study focuses on intra-individual variability in GP as a potential moderator. The association between fitness and academic performance may differ by subject (e.g., math vs. reading), with mixed findings across studies. Prior work indicates baseline performance might moderate intervention effects: individuals with lower baseline cognitive performance show greater gains in executive function from physical activity interventions, and subgroup analyses suggest children with lower baseline academic performance benefit more from interventions. Thus, inter-individual differences in baseline performance may moderate effects on academic outcomes. From an intra-individual perspective, variability in a student’s GPs might affect the longitudinal association between improved fitness and academic performance. Evidence suggests an individual’s worst subjects may be more sensitive to fitness improvements than their best subjects, and importantly, physical activity interventions do not appear to harm performance in high-performing students. Recent work has also improved control for confounders such as socioeconomic status (SES), but out-of-school learning activities (homework, cram schools, tutoring) are often unmeasured despite their direct impact on academic performance. Therefore, it is essential to test associations after controlling for out-of-school learning time. This study reanalyzed data from a two-year longitudinal study to examine whether changes in cardiorespiratory fitness differentially relate to changes in a student’s worst vs. best GPs, while rigorously controlling for SES and out-of-school learning time. The hypothesis was that improvements in fitness would be associated with improvements in the lowest GP, with weak or no association for the highest GP.

Design and participants: Two-year longitudinal reanalysis of data from public junior high schools in and near Sapporo, Hokkaido, Japan. Twenty schools were approached; 11 declined participation and 3 declined SES questionnaires. Two additional schools lacked follow-up. Inclusion required parental/guardian informed consent (consent rate = 74%). Final analytic sample: 4 schools, N = 469 students (214 girls, 255 boys), who completed the two-year period. Institutional approvals were obtained from Hokkaido University of Education IRB and participating school principals. Timeline: Japanese academic year runs April–March. Academic performance was collected at end of 7th and 9th grades (two years apart). Cardiorespiratory fitness and out-of-school learning time were assessed in October of 7th and 9th grades. SES was assessed in October of 7th grade only. Measures:

• Academic performance: Grade points (GP) for five national curriculum subjects (Japanese, mathematics, social studies, science, English), each on a 1–5 scale (higher = better). For each student, the lowest and highest subject GPs were identified at end of 7th and 9th grades and used as outcomes.
• Cardiorespiratory fitness: Schools administered either a 20-m shuttle run or endurance run (1500 m boys; 1000 m girls) during PE classes, per Japan’s Ministry of Education standard. Performance was converted to an age- and sex-normed score from 1 to 10 (higher = better cardiorespiratory fitness).
• Out-of-school learning time: Self-report, five-point scales for weekdays and weekends (from “<30 min/day” to “>3 h/day”); summed score analyzed. Learning time data were available for 451 at baseline and 466 at follow-up due to missing values.
• Body mass index (BMI): Calculated from school-measured height and weight (kg/m²); available for 468 at baseline and 356 at follow-up (one school’s follow-up BMI missing).
• Socioeconomic status (SES): Parent/guardian questionnaire (baseline only) on household income (1 = <2 million yen to 5 = >8 million yen) and maternal educational attainment (1 = junior high school to 5 = bachelor’s degree). SES data obtained for household income (N=354) and maternal education (N=361). Participant flow (Fig. 2): 20 schools approached; 14 excluded (11 declined participation; 3 refused SES questionnaire). Of 6 schools (N=946), 2 schools lost to follow-up (N=370), and additional missing data excluded (e.g., GP, fitness). Final included: 4 schools (N=469). Statistical analysis: Pearson’s correlations tested associations between changes (9th − 7th grade) in cardiorespiratory fitness and out-of-school learning time with changes in lowest and highest GPs. Multiple regression analyses using full-information maximum likelihood estimated the unique association of change in cardiorespiratory fitness with changes in lowest and highest GPs, controlling for household income, maternal education, sex, change in BMI, and change in learning time. Alpha = 0.05; analyses conducted in RStudio v1.1.463. Post-hoc power analyses (G*Power 3.1.2) characterized effect sizes using Cohen’s f (small = 0.02, medium = 0.15, large = 0.35).

• Correlations (Table 1):
  • Change in cardiorespiratory fitness positively correlated with change in lowest GP (r = 0.15, p < 0.05) but not with change in highest GP (r = 0.09, ns).
  • Change in out-of-school learning time correlated positively with both change in lowest GP (r = 0.17, p < 0.05) and change in highest GP (r = 0.21, p < 0.05).
  • Change in cardiorespiratory fitness was not associated with changes in learning time (r = 0.05, ns).
• Multiple regression (Table 2), controlling for household income, maternal education, sex, change in BMI, and change in learning time:
  • Lowest GP: Change in cardiorespiratory fitness β = 0.15, SE = 0.05, 95% CI [0.06, 0.25], p = 0.002; adjusted R² = 0.02. Change in learning time β = 0.16, SE = 0.05, 95% CI [0.07, 0.25], p < 0.001.
  • Highest GP: Change in cardiorespiratory fitness β = 0.06, SE = 0.05, 95% CI [-0.04, 0.16], p = 0.25; adjusted R² = 0.003. Change in learning time β = 0.22, SE = 0.05, 95% CI [0.13, 0.31], p < 0.001.
  • Other covariates (household income, maternal education, sex, change in BMI) were not significant predictors at α = 0.05.
• Effect size and power: For lowest GP, Cohen’s f ≈ 0.02 (small), achieved power 1–β = 0.87; for highest GP, f ≈ 0.003, 1–β = 0.22. Overall: Increases in cardiorespiratory fitness were selectively associated with improvements in students’ worst subject grades, not their best, independent of SES, sex, BMI change, and out-of-school learning time. Increased out-of-school learning time was associated with improvements in both lowest and highest GPs.

The findings support the hypothesis that improvements in cardiorespiratory fitness selectively benefit students’ lowest-performing subjects, with no detectable longitudinal association for their best subjects. This pattern aligns with prior evidence that children with lower baseline cognitive or academic performance derive greater benefits from physical activity interventions. By focusing on intra-individual variability (worst vs. best subjects) rather than only inter-individual baseline differences, this study shows that fitness gains can raise the floor of performance even among students with high overall baseline achievement. The null association with highest GP suggests that increasing fitness does not detract from academic performance in students’ best subjects, supporting guidance that physical activity does not harm academic outcomes. The independent contribution of fitness to the lowest GP, beyond out-of-school learning time, points to potential mechanisms within classroom behavior—such as enhanced executive function, sustained attention, and academic motivation—through which fitness gains may preferentially support weaker subjects. The study also offers a perspective on prior null RCT findings: when overall test performance is relatively high (ceiling effects), benefits may be obscured unless analyses consider intra-individual variability or lower-performing subdomains.

This two-year longitudinal reanalysis demonstrates that improvements in cardiorespiratory fitness are positively associated with improvements in students’ worst subject grades, while showing no adverse or detectable benefits for best subject grades. The results underscore the role of childhood fitness in supporting healthy brain development and academic success and suggest that analyses focusing on intra-individual variability can reveal selective benefits. Future research should examine mechanisms (e.g., executive function and classroom behaviors), differentiate fitness components and physical activity types, and test generalizability across grading systems and populations.

• Observational design cannot rule out catch-up growth effects; however, intra-individual analyses suggest the lowest GP benefits are linked to fitness improvements rather than solely developmental catch-up.
• Cardiorespiratory fitness was used as a proxy for regular physical activity; the relative contributions of developmental changes versus activity-induced fitness gains to academic outcomes remain unclear. Future studies should jointly assess fitness and physical activity levels.
• Focus limited to cardiorespiratory fitness; other fitness components (e.g., coordinative or resistance training) may have different cognitive/academic effects.
• Generalizability: The GP system differs across countries; data are from Japanese junior high schools with teacher-assessed GPs across multiple criteria.
• Lack of assessment of participants’ mental health, learning disabilities, or neurological disorders; potential neuropsychological confounding cannot be excluded.
• Potential selection bias due to missing informed consent and missing data (approximately one-fourth not included), and missing follow-up BMI data from one school.