Network analysis of associations between anthropometry, physical fitness, and sport-specific performance in young canoe sprint athletes: The role of age and sex

INTRODUCTION: Anthropometric and physical fitness data can predict sport-specific performance (e.g., canoe sprint race time) in young athletes. Of note, inter-item correlations (i.e., multicollinearity) may exist between tests assessing similar physical qualities. However, multicollinearity among tests may change across age and/or sex due to age-/sex-specific non-linear development of test performances. Therefore, the present study aimed at analyzing inter-item correlations between anthropometric, physical fitness, and sport-specific performance data as a function of age and sex in young canoe sprint athletes. METHODS: Anthropometric, physical fitness, and sport-specific performance data of 618 male and 297 female young canoe sprint athletes (discipline: male/female kayak, male canoe) were recorded during a national talent identification program between 1992 and 2019. For each discipline, a correlation matrix (i.e., network analysis) was calculated for age category (U13, U14, U15, U16) and sex including anthropometrics (e.g., standing body height, body mass), physical fitness (e.g., cardiorespiratory endurance, muscle power), and sport-specific performance (i.e., 250 and 2,000-m on-water canoe sprint time). Network plots were used to explore the correlation patterns by visual inspection. Further, trimmed means (μ(trimmed)) of inter-item Pearson's correlations coefficients were calculated for each discipline, age category, and sex. Effects of age and sex were analyzed using one-way ANOVAs. RESULTS: Visual inspection revealed consistent associations among anthropometric measures across age categories, irrespective of sex. Further, associations between physical fitness and sport-specific performance were lower with increasing age, particularly in males. In this sense, statistically significant differences for μ(trimmed) were observed in male canoeists (p < 0.01, ξ = 0.36) and male kayakers (p < 0.01, ξ = 0.38) with lower μ(trimmed) in older compared with younger athletes (i.e., ≥U15). For female kayakers, no statistically significant effect of age on μ(trimmed) was observed (p = 0.34, ξ = 0.14). DISCUSSION: Our study revealed that inter-item correlation patterns (i.e., multicollinearity) of anthropometric, physical fitness, and sport-specific performance measures were lower in older (U15, U16) versus younger (U13, U14) male canoe sprint athletes but not in females. Thus, age and sex should be considered to identify predictors for sport-specific performance and design effective testing batteries for talent identification programs in canoe sprint athletes.