A predictive model for longitudinal cognitive subtypes in Parkinson's disease.

BACKGROUND

Longitudinal cognitive changes in Parkinson's disease (PD) exhibit considerable heterogeneity.Predicting cognitive trajectories in early PD patients can improve prognostic counseling and guide clinical trials.

METHODS

This study included 337 early PD patients with 6-year follow-up in the Parkinson's Progression Markers Initiative (PPMI) database.Cognitive function was assessed using the Montreal Cognitive Assessment (MoCA) to identify subtypes of longitudinal cognitive trajectories, and a nomogram predictive model was constructed using baseline clinical variables.

RESULTS

The 337 PD patients had a mean age of 61.0 years, mean disease duration of 0.55 years, and mean MoCA score of 27.1 points. Latent class mixed models (LCMM) identified two longitudinal cognitive subtypes: cognitive stable (276 cases, 81.9%) and cognitivel deteriorating (61 cases, 18.1%). The cognitively deteriorating subtype presented poorer baseline cognition, older age, and more severe motor and non-motor symptoms. On biomarkers, the cognitively deteriorating subtype revealed higher serum NFL levels and lower mean striatum DAT uptake. Six baseline clinical variables (age, Letter Number Sequencing score, Symbol Digit Modalities Test score, Benton Judgment of Line Orientation Test score, Hopkins Verbal Learning Test-Revised score, and REM Sleep Behavior Disorder) were selected to construct the nomogram predictive model which achieved an AUC of 0.92.The calibration curve demonstrated high consistency between predicted and observed probabilities.The predictive model has potential utility in disease-modifying clinical trials by pre-screening patients at high risk for cognitive deterioration.

CONCLUSION

This study identified two longitudinal cognitive subtypes: cognitive stable and cognitive deterioration within 6-year follow-up, and eighteen percent of early PD patients shared the cognitive deterioration subtype The predictive model, incorporating six baseline variables could estimate the risk of longitudinal cognitive deterioration in PD.