[Progress on prognostic assessment methods for mild cognitive impairment]

Mild cognitive impairment (MCI), a key prodromal stage of dementia, requires precise prognostic assessment to delay disease progression. Given the high heterogeneity of MCI, any single indicator has limited predictive efficacy. Different clinical subtypes of MCI, such as amnestic MCI, non-amnestic MCI, and subjective cognitive decline, exhibit fundamental differences in pathological mechanisms and outcomes, forming the basis for stratified prognostic assessment. Abnormal sleep duration, physical functional decline, and psychiatric symptoms (depression, anxiety, apathy) are indicative of cognitive decline risk to varying degrees and can serve as clinical observational indicators for evaluating MCI prognosis. Neuropsychological assessment scales and instrumental activities of daily living assessments can characterize the features of cognitive impairment, but their results are susceptible to educational and cultural influences. Core cerebrospinal fluid (CSF) biomarkers, including amyloid β-protein (Aβ)₄₂, phosphorylated tau protein (P-tau)181, and total tau protein (T-tau), are highly correlated with the core pathology of Alzheimer's disease and can accurately predict MCI prognosis. Plasma biomarkers such as P-tau217, neurofilament light chain, glial fibrillary acidic protein (GFAP), and the Aβ₄₂/Aβ₄₀ ratio are suitable for screening and follow-up; combining CSF and plasma biomarkers enhances predictive performance. In contrast, serum markers like Klotho and insulin-like growth factor-1 lack specificity, and their independent predictive value requires further validation. Multimodal neuroimaging, including structural magnetic resonance imaging (MRI), functional MRI (revealing neural network compensation and decompensation), and positron emission tomography (PET) showing molecular pathological changes (Aβ deposition, tau protein aggregation), can form a complete chain of evidence linking molecular events to clinical phenotypes. Intelligent prediction models, ranging from basic risk stratification and static integrated models to longitudinal dynamic prediction, significantly improve the fusion predictive performance of multimodal data. Consequently, the prognostic assessment of MCI is moving from a single modality toward a stepwise integrated approach: primary screening adopts the combination of clinical information (including MCI subtype characteristics)+core neuropsychological assessment (focusing on delayed recall and executive function subtests of the Montreal Cognitive Assessment)+blood biomarkers (plasma P-tau217 and GFAP testing); the precise diagnostic phase adds structural MRI to assess hippocampal atrophy; for difficult cases and research settings, CSF testing, Aβ-PET, and τ-PET are further introduced. Future research should focus on constructing dynamic monitoring frameworks and deepening mechanistic exploration of modifiable risk factors, thereby advancing individualized prognostic management and early intervention strategies for MCI.