The following is a summary of “Prediction of Future Parkinson Disease Using Plasma Proteins Combined With Clinical-Demographic Measures,” published in the July 2024 issue of Neurology by You et al.
Researchers conducted a retrospective study to identify a combination of blood proteins and common clinical data that could predict Parkinson’s disease (PD) risk.
They utilized data from the UK Biobank (UKB), a nationwide participant recruitment initiative in the UK, to uncover incident PD predictors. Participants without PD at baseline and with baseline plasma proteins were selected. Employing machine learning (ML) techniques, key predictors were identified from a dataset containing 1,463 plasma proteins and 93 clinical-demographic variables. Predictors underwent external validation using the Parkinson’s Progression Marker Initiative (PPMI) cohort. Furthermore, a nested case-control study within the UKB explored temporal predictor trends.
The results showed 52,503 participants, with a median age of 58 and 54% female, were free of PD. During a median follow-up of 14.0 years, 751 individuals developed PD, with a median age of 65 and 37% female. A panel of 22 plasma proteins was selected using a forward selection method for optimal prediction. Utilizing the Light Gradient Boosting Machine (LightGBM) algorithm, the model achieved an AUC of 0.800 (95% CI 0.785–0.815). Integrating plasma proteins and clinical-demographic variables improved predictive accuracy, yielding an AUC of 0.832 (95% CI 0.815–0.849). Key predictors included age, education years, traumatic brain injury history, and serum creatinine. Incorporating 11 plasma proteins boosted predictive accuracy. External validation in the PPMI cohort confirmed reliability, achieving an AUC of 0.810 (95% CI 0.740–0.873). Changes in these predictors were detectable years before PD diagnosis.
Investigators concluded that integrating clinical data with plasma proteins in an ML model holds promise for identifying high-risk individuals for PD in the general population. Still, validation in a more diverse group is warranted.