Saccadic network dysfunction underlies eye-movement variability in early Parkinson’s disease

Abstract Parkinson’s disease is a neurodegenerative disorder characterized by altered subcortical-cortical connectivity, leading to motor and cognitive deficits. Saccadic eye movement abnormalities are among the earliest and most persistent markers of the disease, serving as sensitive indicators of disrupted oculomotor and cognitive control. While changes in average saccadic behavioral measures are widely reported in Parkinson’s disease, intra-individual variability in saccadic performance and the underlying network dysfunction remain poorly understood. Here, we quantified intra-individual saccadic variability in early-stage Parkinson’s disease and identified its functional correlates within the resting-state saccadic network. We analyzed data from 91 patients with early-stage Parkinson’s disease and 46 healthy controls who underwent oculomotor testing and resting-state functional magnetic resonance imaging. First, at the behavioral level, we assessed intra-individual variability in saccadic performance within each group using the coefficient of variation of reaction time and normalized peak velocity across trials. Second, at the brain level, we assessed group differences in static and dynamic functional connectivity across 39 regions comprising the saccadic network. We examined the relationship between network connectivity and saccadic variability using partial least squares analysis. We also quantified network organization using graph-theory measures of global efficiency and clustering coefficient. Finally, we tested the relevance of intra-individual saccadic variability and network features in distinguishing patients from healthy controls using a multivariate Mahalanobis classifier. Our results showed that, compared to HC, patients had higher intra-individual variability in saccadic behavioral measures of reaction time (p Holm = 0.012) and normalized peak velocity (p Holm = 0.012). Interestingly, patients had (i) reduced static and dynamic connectivity within the saccadic network and (ii) weaker associations between network connectivity and intra-individual saccadic variability of specifically normalized peak velocity. These results were particularly driven by regions within the cognitive saccade network. The classifier that integrated intra-individual saccadic variability and network connectivity achieved 95.62% accuracy in distinguishing patients from healthy controls, with dynamic connectivity contributing more strongly to classification performance than static connectivity. We demonstrate that elevated intra-individual variability of saccades, underpinned by altered functional connectivity within cognitive-oculomotor networks, is a prominent feature of early-stage Parkinson’s disease. Emerging as a promising physiological marker of disease pathophysiology, intra-individual variability offers a powerful new framework to refine early diagnosis and improve the tracking of neurodegeneration before severe clinical milestones are reached.