Resistance Training Improves Integrated Health-Related Fitness in University Students: A Phenotype-Based Longitudinal Analysis

Background Health-related fitness comprises both physical fitness and body composition, but these domains are often analyzed separately. Most studies evaluate these components separately or emphasize average changes following exercise interventions. These approaches describe group-level change but do not show how fitness and body composition combine within individuals over time. Therefore, an integrated analysis of body composition and field-based fitness measures may provide a clearer description of health-related fitness assessment in university students. Methods In this secondary analysis, 259 university students completed baseline body composition and health-related fitness assessments. Body composition was assessed using bioelectrical impedance analysis, and field-based fitness tests included grip strength, modified sit-up, 20-m shuttle run, and sit-and-reach. A hierarchical principal component analysis was used to derive a fitness component, a body composition component, and a final integrated health-related fitness score. K-means clustering was then applied to classify participants into Low, Moderate, and High integrated health-related fitness phenotypes. After a 16-week class-based intervention or control period, phenotype transitions were evaluated using a transition matrix and Bowker’s test for symmetry. Cluster validation and bootstrap stability analyses were performed to assess the robustness of the selected phenotype solution. Results The fitness-domain PC1 explained 52.67% of the variance, and the body composition-domain PC1 explained 68.58%. The final PCA integrated these two domain-specific components into an integrated health-related fitness score, which explained 84.02% of the variance. K-means clustering identified three ordered phenotypes: Low (n = 90), Moderate (n = 99), and High (n = 70). The three profiles differed in fitness tests, body composition, and integrated score variables, with higher phenotypes showing more favorable multidimensional profiles. Cluster validation supported the three-cluster solution, and Repeated subsampling demonstrated high clustering stability (mean Adjusted Rand Index, 0.917) with the original cluster assignment. Among 258 participants with complete follow-up data, 230 participants (89.1%) remained stable, 25 (9.7%) improved to a higher phenotype, and 3 (1.2%) declined to a lower phenotype. Bowker’s test indicated significant asymmetry in phenotype transitions from baseline to 16 weeks (χ² = 17.37, df = 2, p < 0.001). Conclusion A hierarchical PCA framework identified three ordered integrated health-related fitness phenotypes in university students. These profiles were used to describe transitions over 16 weeks. Most participants remained stable, whereas a smaller subset showed upward transition, suggesting that integrated HRF phenotypes may reflect relatively stable but modifiable multidimensional fitness profiles. This framework may be useful for descriptive fitness profiling, but external validation is needed before broader application.