Predicting Progression from Mild Cognitive Impairment to Alzheimer’s Disease using Blood-Based Biomarkers and Amyloid-Beta PET: A Survival Analysis Approach

Abstract

Early identification of mild cognitive impairment (MCI) patients at high risk of progressing to Alzheimer’s disease (AD) is critical for therapeutic planning, clinical trial enrollment, and prognostic assessment. Although amyloid-beta positron emission tomography (PET) is a gold-standard for detecting amyloid plaques, a hallmark of AD neuropathology, its cost, invasiveness, and limited accessibility constrain widespread use. Blood-based biomarkers offer a more practical and scalable alternative, with phosphorylated tau 217 (p-Tau217) emerging as a promising marker of amyloid-related pathology. However, its utility for predicting the timing of progression from MCI to AD dementia remains incompletely characterized. In this study, we compared the prognostic performance of p-Tau217 with amyloid-beta PET and the p-Tau217/amyloid-β42 blood ratio test, which is currently the only FDA-approved diagnostic test for AD. We also assessed whether adding complementary AD-related biomarkers, including amyloid-β42/40, neurofilament light chain (NfL), and glial fibrillary acidic protein (GFAP), improves prediction. We analyzed adults with MCI from the Alzheimer’s Disease Neuroimaging Initiative, classified as stable MCI (sMCI) or progressive MCI (pMCI) based on conversion to AD dementia during follow-up. Random survival forest and Cox proportional hazards models were used to evaluate the prognostic value of amyloid-beta PET and blood-based biomarkers for predicting progression from MCI to AD. Models were benchmarked against a baseline model including age, sex, education, and APOE ε4 status. The cohort included 157 sMCI participants (mean age 74.4 years, 61.4% female) and 44 pMCI participants (mean age 75.3 years, 61.4% female), with a mean follow-up of 4.34 years. At four years, all biomarker-based models outperformed the demographics + APOE ε4 model, including the combined blood biomarker panel (AUROC = 0.818, 95% CI [0.748-0.888]; C-index = 0.819 ± 0.093), p-Tau217 alone (0.787 [0.705-0.868]; 0.808 ± 0.097), and amyloid-beta PET imaging (0.758 [0.669-0.848]; 0.798 ± 0.114). p-Tau217 and amyloid-beta PET performed comparably for predicting MCI-to-AD progression, and there was no evidence that the multi-marker panel or p-Tau217/amyloid-β42 ratio improved upon p-Tau217 alone. Partial dependence analysis showed that as p-Tau217 increased from 0.08 to 0.56 pg/mL, the predicted 4-year probability of conversion rose from about 10% to 45%, supporting p-Tau217 as a meaningful and minimally invasive prognostic marker in prodromal AD.