Introduction: Plasma P-tau217, a promising biomarker for Alzheimer's disease(AD) biomarker, facilitates early detection and monitoring. Predicting elevated P-tau217 in cognitively unimpaired individuals could enhance diagnostics and trial enrichment strategies.
Objective: We developed five-year prognostic models for estimating risk of elevated plasma P-tau217.
Methods: This analysis utilised the A4 and LEARN studies involving cognitively unimpaired adults aged 65-85 years with amyloid positron emission tomography results at baseline. Participants with baseline plasma P-tau217 below the highest tertile were analyzed for time to elevation. Random survival forests identified predictors using Cox models. Discrimination and calibration were assessed using Harrell’s C-index, AUC and IBS. Kaplan–Meier analyses evaluated the ability of the models to meaningfully separate risk groups over five years.
Results: Of 1,707 participants, 881 were included (A4:636, LEARN:245). Mean age was 70.6 years, 62.0% were female, and 47.1% carried at least one APOE ε4 allele. Over a median follow-up of 4.2 years, 22.0% developed elevated plasma P-tau217. Baseline plasma P-tau217 was the most important predictor. The model with baseline plasma P-tau217, age, and education performed better (C-Index=0.822; AUC=0.759; IBS=0.069) than models without P-tau217 data.
Conclusion: Prognostic models can stratify five-year risk of elevated P-tau217, guiding diagnostics and AD trial enrichment strategies. The model included P-tau217 yielded the best performance, compared to models without P-tau217.