Body composition, anxiety, and fitness test performance in Chinese college students: biopsychosocial association patterns across gender-differentiated testing contexts.

BACKGROUND: Physical and psychological factors jointly relate to physical performance, yet few studies have examined the indirect associations through which body composition and exercise behavior relate to performance via psychological factors. This study examined whether biopsychosocial association patterns with fitness test performance differ across gender-differentiated testing contexts that impose different physical demands, using China's mandatory national fitness testing system as a naturalistic paradigm.

METHODS: A cross-sectional survey was conducted with 1,916 Chinese college students (492 males, 1,424 females) recruited from universities in Shanghai during the 2025 autumn fitness testing period. Participants completed measures of fitness test anxiety (FTAS), mental toughness (MTI), and exercise behavior (PARS-3). Body mass index (BMI) and official composite fitness test scores were obtained from university records. Structural equation models were estimated separately for each testing context using maximum likelihood with robust standard errors, supplemented by path-by-path constraint tests within a multi-group framework. Sensitivity analyses re-estimated the models using a non-BMI reweighted composite score, a log-transformed exercise measure, and a specification excluding training experience; additional shared-item and raw-score sensitivity analyses evaluated whether the BMI-related contrast was also present in test components completed by both sexes. Network analysis (EBICglasso) with bridge centrality and a Network Comparison Test were conducted on 24-node networks.

RESULTS: BMI was negatively associated with scores in both testing contexts, more strongly in the male context (β = -0.519 vs. -0.242); this contrast persisted in the full-battery sensitivity analyses (non-BMI composite and log-transformed exercise measure) but was attenuated when the outcome was restricted to shared test components and was no longer detectable when the four shared components were represented as raw, directionally aligned z scores. Exercise showed a significant direct association with scores in the male context and a weaker, less stable association in the female context. Path-by-path constraint tests indicated that context-dependent differences were concentrated in the physical-to-performance interface (BMI → score: Δχ² = 34.29, p < .001), although the exercise-related contrasts were sensitive to the scaling of the exercise measure. The psychological-to-performance interface did not differ across contexts (anxiety → score: p = .338; mental toughness → score: p = .592). Network analysis identified a specific cognitive anxiety item concerning failure to meet the passing standard as the strongest psychological bridge node across both contexts.

CONCLUSIONS: The findings indicate context-specific biopsychosocial association patterns within a gender-differentiated national fitness testing system. The BMI-related contrast appears localized to the gender-differentiated and algorithmically scored components of the official testing battery rather than being a general feature of shared raw performance; exercise-related contrasts should be interpreted cautiously given their sensitivity to measurement scaling. Because gender and task structure are institutionally confounded in this design, the findings cannot separate task demands from gender or biological sex. The cross-context stability of the psychological-to-performance paths, which was robust across all sensitivity specifications including shared-item analyses, suggests that institutional reforms reducing evaluative threat may benefit students across testing contexts.