A Multimodal Nomogram Predicting the Risk of Cognitive Impairment in Parkinson Disease
Ahmed NegidaAbstract
Background
Cognitive impairment is the most consequential non‐motor complication of Parkinson's disease (PD), yet no validated tool integrates multimodal data to predict individualized risk trajectories.
Objectives
The aim is to develop and internally validate a time‐to‐event nomogram for predicting 5‐year risk of cognitive impairment in PD using the Parkinson's Precision Markers Initiative (PPMI).
Methods
We analyzed 1188 PD patients with normal cognition at their first cognitive assessment and up to 13 years of follow‐up. Twenty‐seven candidate predictors across seven domains were evaluated. Least absolute shrinkage and selection operator (LASSO)‐penalized Cox regression performed variable selection. Performance was assessed via nested cross‐validation, bootstrap optimism correction, calibration, and time‐dependent area under the curve (AUC).
Results
During median follow‐up of 3.0 years (interquartile range [IQR], 2.0–5.0), 457 patients (38.5%) developed cognitive impairment. LASSO selected 14 predictors. The Cox model achieved a concordance index (C‐index) of 0.747 (optimism‐corrected: 0.733; nested CV: 0.736 ± 0.042), with apparent time‐dependent AUC of 0.77, 0.79, and 0.83 at 3, 5, and 8 years (held‐out: 0.75, 0.78, and 0.81). A companion model without genetic predictors, for unknown GBA status, performed similarly (optimism‐corrected C‐index, 0.729). The strongest predictors were age (hazard ratio [HR], 1.05/year; P < 0.001), GBA mutation (HR, 2.05; P < 0.001), cognitive complaints (HR, 1.67; P < 0.001), and Symbol Digit Modalities Test (HR, 0.98/point; P = 0.001).
Conclusions
A multimodal nomogram integrating clinical, cognitive, genetic, and neuropsychiatric predictors provides individualized estimates of 5‐year risk of cognitive impairment in PD; after external validation, it may support prognostic counseling and trial enrichment.