Development of an individualized prediction model for dynamic adaptations in performance and immune function associated with dietary patterns in endurance athletes using machine learning.

BACKGROUND

Psychological resilience significantly influences immune function and health outcomes in high-stress populations, yet mechanisms underlying nutrition-psychology-immunity interactions remain poorly understood. This study developed an individualized prediction model integrating dietary patterns with psychological and immune adaptations to inform personalized therapeutic approaches.

METHODS

A retrospective cohort analysis examined 200 endurance athletes over 12 months using integrated datasets from NHANES athletic subcohort, UK Biobank, and training monitoring databases. Athletes were categorized into three dietary pattern groups (high-carbohydrate, high-protein, balanced micronutrient) based on their naturalistic dietary intake. This observational design examined associations between dietary patterns and health outcomes without manipulating participant diets. A hybrid LSTM-XGBoost machine learning architecture with SHAP analysis predicted individual responses based on psychological variables, immune markers (IL-6, TNF-α, CRP, IgA), and performance metrics. Statistical analyses controlled for multiple comparisons using Bonferroni correction. Non-normally distributed variables were log-transformed or analyzed using non-parametric methods. Mediation analyses examined psychological pathways linking dietary patterns to immune outcomes.

RESULTS

Psychological resilience emerged as the primary predictor of dietary pattern response (SHAP importance = 0.342), with psychological improvements consistently preceding immune function recovery by 1-2 months. Three distinct resilience-based subgroups demonstrated different response trajectories: high resilience athletes achieved superior improvement rates (0.43 vs. 0.10 points/month) and reached plateau phases earlier (6.8 vs. 11.2 months) compared to low resilience individuals. The predictive model achieved exceptional performance metrics (91.2% sensitivity, 87.6% specificity) for identifying non-responders to dietary patterns. Mediation analysis revealed that 42.4% of the associations between dietary patterns and immune function operated through psychological pathways, with cortisol reduction serving as a critical mechanism.

CONCLUSIONS

Psychological resilience predicts responsiveness to dietary patterns through psychoneuroimmunological pathways. Baseline psychological assessment should guide personalized nutrition strategies in clinical populations experiencing chronic stress and immune dysfunction.