Longitudinal network analysis of 19 cognitive tests in 355 people with Parkinson's over four yearly assessments identified five dynamic cognitive dimensions that reorganize over time and differ from baseline cross-sectional domains.

The finding that cognitive domains in PD are temporally dynamic highlights the need to rethink diagnostic criteria, endpoint selection, and patient stratification in trials—potentially improving biomarker and clinical outcome design—though it offers little direct mechanistic or therapeutic targets.

Cognitive domains are central to diagnosing cognitive impairment in people with Parkinson's disease (PwPD), yet are defined by expert consensus and cross-sectional data that assume temporal stability. This study examined whether cognitive domains remain stable or reorganize dynamically over time in PwPD. Using dynamic exploratory graph analysis, we analyzed 19 cognitive test scores from 355 PwPD across four yearly assessments (24,372 data points) from the DEMPARK/LANDSCAPE-study to identify dynamic organization of cognitive dimensions. Panel graphical vector autoregression models assessed dynamic couplings among dimensions. Five dynamic cognitive dimensions emerged, diverging substantially from theoretical domains and cross-sectional dimensions at baseline. Dynamic coupling revealed a temporal separation between card-sorting/cognitive flexibility and visuoconstruction. Cognitive domains in PwPD reorganize over time rather than remaining stable, challenging the static assumption of diagnostic criteria. Using PD as an exemplar, findings demonstrate the need for dynamic frameworks potentially revealing condition-specific temporal architectures when applied to other neurodegenerative disorders.