Non-cognitive intrinsic capacity profile and cognitive ageing across three international ageing cohorts: a harmonised individual-participant data analysis.

OBJECTIVE: To examine whether a four-domain non-cognitive intrinsic capacity (IC) profile, excluding cognition to avoid exposure-outcome circularity, is associated with cognitive ageing across three international cohorts. METHODS: We analysed harmonised CHARLS, ELSA, and HRS data from 26,022 adults aged ≥60 years (132,986 person-waves). Model A assessed longitudinal global cognition (24,741 participants), Model B incident cognitive impairment (22,717 estimable participants; 5682 events), and Model C a sustained low-cognitive-performance indicator requiring two consecutive follow-up assessments with cognition z ≤ -1.5 (18,671 estimable participants; 1783 events). Primary inference used one-stage individual-participant-data models with cohort-by-profile interaction and marginal standardisation. RESULTS: Each additional impaired domain was associated with lower longitudinal cognition (β = -0.123, 95% CI -0.133 to -0.114), incident cognitive impairment (OR = 1.229, 95% CI 1.198-1.261), and the sustained indicator (OR = 1.273, 95% CI 1.217-1.331). Cohort-by-profile interaction was detectable for Models A and B but not Model C (P = 0.099). Adjustment for baseline cognition attenuated Models B and C by 36.3% and 37.0%, respectively. Common episodic-memory analyses showed consistent associations with memory level (β = -0.099, 95% CI -0.108 to -0.090) and incident low memory performance (OR = 1.181, 95% CI 1.148-1.214). CONCLUSIONS: Greater impairment in the non-cognitive IC profile was associated with cognitive vulnerability across cohorts, with directionally similar but heterogeneous effects. The profile is a partial IC construct and should not be interpreted as a diagnostic tool or evidence that modifying its domains prevents cognitive impairment. Its current relevance is to community-based surveillance and hypothesis generation.