Joint longitudinal trajectories of depressive symptoms and cognitive function in relation to new-onset stroke among adults aged ≥ 50 years with diabetes or prediabetes: a multi-cohort study

Adults aged ≥ 50 years with diabetes or prediabetes frequently experience depressive symptoms and cognitive impairment, both of which may be associated with cerebrovascular vulnerability. However, little is known about whether their joint longitudinal patterns are associated with subsequent stroke risk. We analyzed data from three nationally representative aging cohorts: the Health and Retirement Study, the English Longitudinal Study of Ageing, and the China Health and Retirement Longitudinal Study.

Participants with diabetes or prediabetes and without stroke during the trajectory assessment period were included. Group-based trajectory modeling was used to identify joint longitudinal profiles of depressive symptoms and a composite cognitive score comprising total word recall, temporal orientation, and Serial 7s. Associations with new-onset stroke were evaluated using Cox proportional hazards models, with competing-risk and discrete-time logistic models as sensitivity analyses.

Machine learning models were developed to evaluate the importance of trajectory membership alongside other predictors. Distinct joint trajectory profiles were identified within each cohort. Relative to the most favorable profile, higher stroke risks were observed for the low-declining cognition/moderate-stable CES-D profile (HR = 1.52, 95% CI: 1.25–1.85) and the moderate-declining cognition/high-increasing CES-D profile (HR = 1.37, 95% CI: 1.24–1.51) in HRS; the low-declining cognition/high-increasing CES-D profile in ELSA (HR = 1.64, 95% CI: 1.39–1.94); and the low-declining cognition/high-decreasing CES-D profile in CHARLS (HR = 1.15, 95% CI: 1.01–1.31).

The findings were consistent in sensitivity analyses. Trajectory membership contributed to model predictions but showed relatively low global SHAP importance. In adults aged ≥ 50 years with diabetes or prediabetes, adverse joint trajectories of depressive symptoms and cognitive function are associated with increased stroke risk.

Further studies should determine whether trajectory information provides incremental predictive value beyond baseline cognitive and depressive symptom levels. We are grateful to the CHARLS, ELSA, and HRS study, which provided data for this research. We also thank all the workers, volunteers, and respondents devoted to the CHARLS, ELSA, and HRS project.

This work was supported by Clinical Medicine Project of Nantong University (2023JZ026) and Yancheng Science and Technology Bureau (YCBE202530). These authors contributed equally to this work: Lijian Han and Shufang Wang. Department of Neurology, Yancheng Third People’s Hospital (Affiliated Hospital 6 of Nantong University, The Yancheng School of Clinical Medicine of Nanjing Medical University), Xindu Road 2#, Yancheng, 224008, Jiangsu, China Lijian Han, Shufang Wang, Jianping Liu, Yuanying Song, Gendi Wang & Yuan Shen School of Pharmacy, Nanjing Medical University, Nanjing, 211166, Jiangsu, China Correspondence to Gendi Wang or Yuan Shen.