Sensitivity comparison of longitudinal cognitive function indicators of Alzheimer's disease after mild cognitive...

Accurate prediction of progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD) is critical for early intervention. Many existing models lack the ability to capture the nonlinear nature of cognitive decline and individual heterogeneity. This study employed a semi‑parametric joint model to analyze longitudinal cognitive trajectories and identify robust predictors of conversion. Data from 596 participants (184 AD converters, 412 stable MCI) were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Longitudinal assessments included ADAS‑Cog13, ADAS‑Cog11, CDR‑SB, FAQ, RAVLT‑IR, RAVLT‑L, and MMSE. A semi‑parametric joint model combining B‑splines for the longitudinal process with a Cox survival submodel was fitted for each cognitive measure. Model performance was evaluated using AIC, BIC, intraclass correlation coefficient (ICC), time‑dependent C‑index, dynamic AUC, and calibration curves. Subgroup analyses were conducted by APOE‑ε4 carrier status. In multivariable joint models, APOE‑ε4 carriage was a consistent risk factor (HR range: 1.38-1.77). Higher scores on ADAS‑Cog13 (HR = 3.71 per SD), ADAS‑Cog11 (HR = 2.71), CDR‑SB (HR = 3.79), and FAQ (HR = 2.85) increased the hazard of conversion, whereas higher scores on RAVLT‑IR (HR = 0.23), RAVLT‑L (HR = 0.14), and MMSE (HR = 0.53) were protective. All models showed high ICCs (0.94-0.98) and moderate‑to‑good predictive accuracy over 2, 5, and 8 year horizons (C‑index: 0.585-0.668). CDR‑SB and FAQ exhibited the strongest effect sizes and clearest dose‑dependent trajectories across APOE‑ε4 subgroups. Calibration curves demonstrated good agreement between predicted and observed survival. The semi‑parametric joint model effectively captures nonlinear cognitive‑functional decline and provides validated predictions of AD risk. APOE‑ε4 genotype combined with longitudinal monitoring of CDR‑SB and FAQ offers a robust framework for stratifying progression risk in clinical MCI management.