Evaluation of ATNPD Framework and Biofluid Markers to Predict Cognitive Decline in Early Parkinson Disease

Abstract

Background and objectives: In Parkinson disease (PD), Alzheimer disease (AD) copathology is common and clinically relevant. However, the longitudinal progression of AD CSF biomarkers-β-amyloid 1-42 (Aβ₄₂), phosphorylated tau 181 (p-tau₁₈₁), and total tau (t-tau)-in PD is poorly understood and may be distinct from clinical AD. Moreover, it is unclear whether CSF p-tau₁₈₁ and serum neurofilament light (NfL) have added prognostic utility in PD, when combined with CSF Aβ₄₂. First, we describe longitudinal trajectories of biofluid markers in PD. Second, we modified the AD β-amyloid/tau/neurodegeneration (ATN) framework for application in PD (ATN_PD) using CSF Aβ₄₂ (A), p-tau₁₈₁ (T), and serum NfL (N) and tested ATN_PD prediction of longitudinal cognitive decline in PD.

Methods: Participants were selected from the Parkinson's Progression Markers Initiative cohort, clinically diagnosed with sporadic PD or as controls, and followed up annually for 5 years. Linear mixed-effects models (LMEMs) tested the interaction of diagnosis with longitudinal trajectories of analytes (log transformed, false discovery rate [FDR] corrected). In patients with PD, LMEMs tested how baseline ATN_PD status (AD [A+T+N±] vs not) predicted clinical outcomes, including Montreal Cognitive Assessment (MoCA; rank transformed, FDR corrected).

Results: Participants were 364 patients with PD and 168 controls, with comparable baseline mean (±SD) age (patients with PD = 62 ± 10 years; controls = 61 ± 11 years]; Mann-Whitney Wilcoxon: p = 0.4) and sex distribution (patients with PD = 231 male individuals [63%]; controls = 107 male individuals [64%]; χ²: p = 1). Patients with PD had overall lower CSF p-tau₁₈₁ (β = -0.16, 95% CI -0.23 to -0.092, p = 2.2e-05) and t-tau than controls (β = -0.13, 95% CI -0.19 to -0.065, p = 4e-04), but not Aβ₄₂ (p = 0.061) or NfL (p = 0.32). Over time, patients with PD had greater increases in serum NfL than controls (β = 0.035, 95% CI 0.022 to 0.048, p = 9.8e-07); slopes of patients with PD did not differ from those of controls for CSF Aβ₄₂ (p = 0.18), p-tau₁₈₁ (p = 1), or t-tau (p = 0.96). Using ATN_PD, PD classified as A+T+N± (n = 32; 9%) had worse cognitive decline on global MoCA (β = -73, 95% CI -110 to -37, p = 0.00077) than all other ATN_PD statuses including A+ alone (A+T-N-; n = 75; 21%).

Discussion: In patients with early PD, CSF p-tau₁₈₁ and t-tau were low compared with those in controls and did not increase over 5 years of follow-up. Our study shows that classification using modified ATN_PD (incorporating CSF Aβ₄₂, CSF p-tau₁₈₁, and serum NfL) can identify biologically relevant subgroups of PD to improve prediction of cognitive decline in early PD.

Figure 1. Analyte Trajectories Over Time

Least squares regression lines and standard error are plotted for patients with PD (blue) and controls (gray) for each analyte by visit year (0–5).

Figure 2. Changes in Analytes Over Time by Baseline CSF Aβ42

Least squares regression lines with standard error show analyte concentrations over time for patients with PD (left panels) and controls (right panels). For visualization, baseline CSF Aβ₄₂ was dichotomized; color represents low (red; <683) or high (blue; >683) CSF Aβ₄₂ at baseline.

Figure 3. Classifications of PD Across Biomarker Strategies

Classification proportions of PD over time by traditional ATN, A status, and modified ATN_PD. Sample sizes for each classification are reported. Color indicates ATN and ATN_PD interpretation: normal (white; A-T-N-), amyloid (coral; A+T-N-), AD (A+T+N-, A+T+N+), A+SNAP (purple; A+T-N+), and SNAP (blue; A-T+N-, A-T-N+, A-T+N+). For A status, interpretation is either A- (white) or A+ (coral).

Figure 4. Modified ATN_PD vs A Status Associations With Cognitive Decline

Least squares regression lines and standard error intervals plot performance over time for each cognitive metric. Color indicates classification by A status (left panels: A- = gray; A+ = coral) and ATN_PD (right panels: Normal [A-T-N-] = gray; amyloid [A+T-N-] = coral; AD [A+T+N-, A+T+N+] = red; A+SNAP [A+T-N+] = purple; SNAP [A-T+N-, A-T-N+, A-T+N+] = blue). Asterisks indicate nominal p < 0.05.

Figure 5. Modified ATN_PD vs A Status Associations With Motor and Autonomic Decline

Least squares regression lines and standard error intervals plot performance over time for each motor/autonomic metric. Color indicates classification by A status (left panels: A- = gray; A+ = coral) and ATN_PD (right panels: normal = gray; amyloid = coral; AD = red; A+SNAP = purple; SNAP = blue). Asterisks indicate nominal p < 0.05.