Joint cognitive–depression trajectories during aging and risk of new-onset cancer: cross-national evidence from three population-based cohorts

Abstract Background Cognitive decline and depressive symptoms commonly co-occur during aging and may jointly influence cancer risk through chronic inflammation, neuroendocrine dysregulation, and adverse behavioral changes. However, longitudinal evidence on synchronized cognitive–depression trajectories and new-onset cancer remains limited across diverse populations. This study investigated the association between joint cognitive–depression trajectories and new-onset cancer risk using three nationally representative cohorts, while further exploring the consistency of predictor importance using complementary machine learning approaches. Methods Adults aged ≥ 65 years from HRS ( n = 7,077), ELSA ( n = 3,879), and CHARLS ( n = 3,549) who were cancer-free at baseline and had at least three repeated assessments of cognition and depressive symptoms were included. Group-based trajectory modeling identified synchronized cognitive–depression trajectories. Cox proportional hazards models examined associations with new-onset cancer, with Fine–Gray competing risk models, discrete-time logistic regression, and subgroup analyses assessing robustness. Seven machine learning models were developed following LASSO and Boruta feature selection, with SHAP, calibration curves, and decision curve analysis used for interpretation and model evaluation. Results Over median follow-up periods of 10, 6, and 8 years, 1,187, 310, and 100 new-onset cancer cases were identified in HRS, ELSA, and CHARLS, respectively. Two trajectory groups were consistently identified: High Cognition and Low CESD, and Low Cognition and High CESD. Compared with the favorable trajectory, the adverse trajectory was associated with significantly increased new-onset cancer risk in HRS (HR = 1.23, 95% CI: 1.05–1.45), ELSA (HR = 2.05, 95% CI: 1.57–2.67), and CHARLS (HR = 1.61, 95% CI: 1.03–2.52). Findings were robust across multiple models. CatBoost demonstrated the best predictive performance, and cognitive–depression trajectory was consistently identified as a key predictor. Conclusions Synchronized cognitive decline and worsening depressive symptoms were significantly associated with elevated new-onset cancer risk across three countries. Joint cognitive–depression trajectories may represent a potential behavioral–psychological indicator associated with cancer risk among older adults; however, their clinical utility for individual-level risk stratification or cancer prevention requires further prospective validation.