This research project, conducted under the supervision of Professor Qihong Lu at the Department of Neuroscience, City University of Hong Kong, validates the “anosognosia crossover” pattern in Subjective Cognitive Decline (SCD) across three independent cohorts. The crossover pattern describes a phenomenon where self-reported cognitive concerns exceed informant reports at low disease severity (self > informant), but fall below informant reports at higher severity (informant > self) — reflecting a progressive loss of self-awareness (anosognosia) as pathology advances.

Study Design

The study analyses data from three independent cohorts with different self/informant-report instruments, after demographic-only harmonisation:

• ADNI (Alzheimer’s Disease Neuroimaging Initiative): n = 867 (with CDR), using ECog participant-report (EcogPT) and study-partner-report (EcogSP)
• AIBL (Australian Imaging, Biomarkers and Lifestyle): n = 1,193, using MAC-Q (self-report) and Short-IQCODE (informant-report)
• A4 (Anti-Amyloid Treatment in Asymptomatic Alzheimer’s Disease): n = 4,447, using CFI (self-report) and CFS (informant-report)

Key Findings

Primary Finding — Crossover Confirmed:

• ADNI: Crossover confirmed across all 5 statistical tests (Wilcoxon, Jonckheere-Terpstra, Kruskal-Wallis, OLS interaction, MMSE regression). Gap shifts from −0.213 (CDR 0) to +1.117 (CDR >1).
• AIBL: Confirmed by continuous MMSE regression (β = −0.055, 74% of ADNI’s slope), though CDR 1 cell underpowered (n = 39).
• A4: Serves as preclinical anchor — uniformly negative gap (self > informant), consistent with hypernosognosia in preclinical AD.

Biomarker Stratification:

• Crossover present in both amyloid PET-positive and APOE ε4 carrier strata, with stronger effect in biomarker-positive participants.

Longitudinal Validation:

• Within-subject crossover confirmed at MCI→dementia transition in ADNI converters (Δ gap = +0.543, p < .001).

Methodological Highlights

• Five harmonisation methods compared; demographic-only residualisation chosen as primary (severity-mapped method shown to introduce circular reasoning)
• Statistical tests: Wilcoxon signed-rank, Jonckheere-Terpstra, Kruskal-Wallis, OLS with cluster-robust SE, bootstrap CIs (10,000 resamples)
• Effect sizes: Cohen’s d_z, rank-biserial correlation, Cliff’s delta
• Sensitivity analyses: biomarker stratification, floor/ceiling quantification, measurement invariance, longitudinal trajectories, power analysis

Current Status

Manuscript and supplementary materials generated. Targeting Journal of Alzheimer’s Disease for submission.

Outputs

• Manuscript: 7 figures, 7 tables, 40 references
• Supplementary materials: 13 tables, 4 figures (including CONSORT-style flow diagram)
• Statistical outputs: 12 CSV files
• All figures at 300 dpi