Clinical Risk Scores and Alzheimer’s Disease Endophenotypes

This repository contains the full data processing and analysis pipeline used in:

“Associations of dementia polyexposure scores to Alzheimer’s disease endophenotypes in a diverse population”

The project evaluates how clinical risk scores (CRS)—mCAIDE, WHICAP, LIBRA, and CogDRisk—relate to:

• Alzheimer’s disease (AD) endophenotypes (plasma biomarkers, neuroimaging, cognition)
• pTau217/Aβ42 positivity (proxy for amyloid PET positivity)
• Cognitive impairment (MCI, dementia)

---

Repository Structure

The workflow is organized into three sequential stages:

1. Data Standardization (Preprocessing)
2. Imputation (Handling Missingness)
3. Analysis (Statistical Modeling and Results)

---

1. Data Standardization

File: crs_standardization.qmd

Performs initial preprocessing and harmonization of raw HABS-HD data.

Key Operations

• Biomarker processing

  • Log-transformation of plasma biomarkers
  • Z-score normalization (mean = 0, SD = 1)
  • Construction of ratios (e.g., Aβ42/Aβ40, pTau217/Aβ42)
  • Outlier removal using IQR-based filtering

• Neuroimaging variables

  • Cortical thickness aggregation
  • Hippocampal volume averaging

• Cognitive measures

  • Z-score standardization
  • Composite domains:
    • Memory
    • Verbal ability
    • Executive function

• Clinical covariates

  • BMI, hypertension, diabetes, dyslipidemia
  • Depression, smoking, alcohol use, physical activity
  • Alignment of CRS-specific variables

---

2. Imputation

File: imputation.qmd

Handles missing data across CRS variables and covariates.

Key Operations

• missForest imputation

  • Supports mixed data types
  • Non-parametric (random forest–based)

• Feature inclusion

  • Demographics
  • Clinical variables
  • Biomarkers
  • CRS predictors

• Validation

  • Comparison with complete-case analysis
  • Bias assessment

---

3. Analysis - Endophenotypes and Cognitive Status

File: crs_analysis.qmd

Core statistical analysis linking CRS to outcomes and biomarkers.

---

3.1 CRS Construction

• mCAIDE
• WHICAP
• LIBRA
• CogDRisk
• CRS - demographic covariates

---

3.2 Cognitive Impairment Analysis

• Logistic regression:

  • MCI
  • Dementia
  • Combined outcome

• Adjustments:

  • APOE genotype
  • Demographics (where applicable)

• Metrics:

  • AUC
  • Nagelkerke R²
  • DeLong test

---

3.3 AD Endophenotype Analysis

• Linear regression for:
  • Plasma biomarkers
  • Neuroimaging measures
  • Cognitive performance

---

3.4 Stratified Analysis

• Groups:

  • Non-Hispanic White (NHW)
  • Hispanic/Latinx (LA)
  • Black/African American (AA)

• Pairwise z-tests for coefficient differences

---

3.5 Sensitivity Analysis

• Removes demographic components from CRS
• Reintroduces them as covariates in each regression
• Evaluates incremental predictive contribution

---

---

4. Analysis - pTau217/Aβ42 Positivity

File: cutoff_analysis.qmd

Derives and validates the pTau217/Aβ42 cutoff.

Role

• Youden index cutoff derivation
• Logistic regression
• Model evaluation:
  • Odds ratios
  • AUC
  • R²

---