Predicting CSF α-Synuclein Seed Amplification Assay Status From Demographics and Clinical Data.
This study builds and externally validates logistic models using easily obtainable clinical variables (UPSIT smell percentiles, constipation, sex, and LRRK2/GBA status) to accurately predict CSF alpha-synuclein SAA positivity in PD-enriched cohorts (AUROC 0.92–0.98).
What the AI sees
This study builds and externally validates logistic models using easily obtainable clinical variables (UPSIT smell percentiles, constipation, sex, and LRRK2/GBA status) to accurately predict CSF alpha-synuclein SAA positivity in PD-enriched cohorts (AUROC 0.92–0.98).
Research significance
Provides a non-invasive, pragmatic approach to infer alpha-synuclein pathology that can help enrich and triage participants for biomarker-driven trials and reduce reliance on CSF sampling, aiding therapeutic development and patient stratification.
Source abstract
BACKGROUND AND OBJECTIVE: The alpha-synuclein (α-syn) cerebrospinal fluid seed amplification assay (CSF SAA) presents a promising diagnostic for Parkinson's disease (PD) and other synucleinopathies. The objective of this study was to develop and externally validate models to predict probabilities of α-syn positive or negative status in vivo in a mixture of people with and without PD using easily accessible clinical predictors. METHODS: Uni- and multi-variable logistic regression models were developed in a cohort of participants from the Parkinson Progression Marker Initiative (PPMI) study to predict CSF α-syn status as measured by SAA. Models were externally validated in a cohort of participants from the Systemic Synuclein Sampling Study (S4) that had also measured CSF α-syn status using SAA. RESULTS: The PPMI model training/testing cohort included 1260 participants, 37% of whom were female, and a mean (± standard deviation) age of 62.4 (±10.0) years. Among them, 76% had manifest PD with a mean disease duration of 1.2 (±1.6) years. Overall, 68.7% of the overall PPMI cohort (and 88.0% with PD of those with manifest PD) had positive CSF α-syn SAA status results. Variables from the full multivariable model to predict CSF α-syn SAA status included age- and sex-specific University of Pennsylvania Smell Identification Test (UPSIT) percentile values, sex, self-reported frequency of constipation problems, leucine-rich repeat kinase 2 (LRRK2) genetic status and pathogenic variant, and GBA status. Internal performance of the model on PPMI data to predict CSF α-syn SAA status had an area under the receiver operating characteristic curve (AUROC) of 0.921, and sensitivity/specificity of 0.858/0.868. This model was applied to the external S4 cohort, which included 71 participants, 39% of whom were female, with a mean age of 63.0 (±8.0) years, and included 70.4% with manifest PD (for a mean 5.1 (±4.8) years). The model performed well, achieving an AUROC of 0.978, and sensitivity/specificity of 0.958/0.870. CONCLUSION: Data-driven models using non-invasive clinical features can accurately predict CSF α-syn SAA positive and negative status in cohorts enriched for people living with PD. Scores from the UPSIT were highly significant in predicting α-syn SAA status.