An approach to construct simplified measures of dietary patterns from exploratory factor analysis

Area of Science:

• Nutritional Epidemiology
• Statistical Modeling in Public Health

Background:

• Exploratory factor analysis (EFA) identifies dietary patterns but results are population-specific, limiting cross-study comparisons.
• This population-dependency hinders the reproducibility of disease risk associations derived from EFA-based dietary patterns.

Purpose of the Study:

• To develop and validate a method for creating less population-dependent dietary pattern variables.
• To assess if these novel 'simplified pattern variables' can accurately represent EFA-derived patterns and their associations.

Main Methods:

• Proposed deriving 'simplified pattern variables' by summing unweighted, standardized food variables with high factor loadings.
• Utilized data from the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study for validation.
• Compared simplified pattern variables with EFA-derived factor scores and assessed cross-population generalizability.

Main Results:

• A simplified pattern variable (using 6 food variables) strongly correlated (>0.95) with an EFA factor score (47 food variables).
• Simplified pattern variables demonstrated similar factor loadings and nutrient intake associations compared to a US-derived 'western' pattern.
• This suggests simplified pattern variables can approximate EFA patterns across different study populations.

Conclusions:

• Simplified pattern variables offer a robust, reproducible alternative to EFA-based dietary patterns in epidemiological research.
• This method facilitates cross-centre and multi-population studies investigating diet-disease associations.
• It addresses a key limitation of EFA, enabling more comparable risk estimates in nutritional epidemiology.