Harmonizing the preclinical Alzheimer cognitive composite for multicohort studies

Area of Science:

• Neuroscience
• Psychometrics
• Biostatistics

Background:

• The preclinical Alzheimer cognitive composite (PACC) is increasingly used across diverse research cohorts.
• Harmonizing PACC data is crucial for robust multi-cohort analyses in Alzheimer's disease research.

Purpose of the Study:

• To develop a harmonized PACC using modern psychometric approaches.
• To create a latent PACC (lPACC) that facilitates more accurate cross-cohort comparisons and analyses.

Main Methods:

• Utilized longitudinal data from ADNI, HABS, and AIBL cohorts (n=2,712) and A4 Study data (n=4,492).
• Employed confirmatory factor analysis (CFA) with overlapping tests as anchors for harmonization.
• Generated latent PACC (lPACC) scores and validated against the common standardized PACC (zPACC).

Main Results:

• The harmonized lPACC revealed baseline differences across cohorts, unlike the zPACC.
• Baseline lPACC demonstrated a slightly superior ability to predict dementia progression compared to zPACC.
• Longitudinal lPACC showed less variability and stronger association with baseline β-amyloid status than zPACC.

Conclusions:

• The proposed harmonization procedures for PACC make fewer assumptions than zPACC, enabling robust multi-cohort studies.
• This item response theory-based implementation is adaptable for future harmonized composite measures in Alzheimer's research.