Table 2.
Overview of the studies using factor analysis or principal component analysis (PCA) to derive dietary patterns (DP)
| Author/study (year) | Diet assessment instrument | Details of diet assessment instrument | Reported DP method | Details of DP method | Label of DP | Variation of DP across regions/countries | Reliability/validity (yes/no) |
|---|---|---|---|---|---|---|---|
| Balder/DIETSCAN project including: ATBC NLCS SMC ORDET (2003)( 13 ) | Country-specific validated dietary instruments | Aggregation of the four FFQ data to 51 common food groups (including also country-specific foods) | Exploratory factor analysis | Sensitivity analyses; decision for extraction: eigenvalue>1 and scree plot; dichotomised variables with>75 % non-users (non-user=0, user=1); no transformations to enhance linearity or normality; no exclusion of outliers, because of intensive data cleaning; factor loadings>0·35 considered | Factors labelled: (salad) vegetables; pork, processed meat, potatoes; cooked vegetables; alcohol; sweet and/or savoury snacks; brown/white bread substitution; others | Vegetables and meat pattern for all NLCS men: cooked vegetables, sweet/savoury snacks, bread substitution NLCS women: sweet/savoury snacks, bread substitution, fat dairy SMC: alcohol, margarine/butter substitution ORDET: cooked vegetables, alcohol | No internal validity via several sensitivity analyses |
| Bamia/ EPIC-Elderly study (2005)( 11 ) | Country-specific dietary self-reported or interviewer-administered questionnaires | 22 condensed energy-adjusted food groups (residual method); validated questionnaires | PCA | Number of PC retained by three criteria: eigenvalue exceeding 1, scree plot, interpretability of each component | Plant-based DP | Greece, Italy, Spain and France highest proportion in the third tertile; Sweden, Denmark low scores; Germany, The Netherlands and UK highest proportion in the second tertile | No |
| Iqbal/INTERHEART study (2008)( 17 ) | 19-item qualitative food group frequency questionnaire (already condensed food items) | Generic questionnaire to be applicable in multiple countries; no portion size, only frequency; standardised in consumption per day | Exploratory factor analysis | Rotated orthogonally; retain factors with eigenvalue>1; scree test, factor interpretability (not the percentage of variance) | Oriental pattern: tofu and soy and other sauces; Western pattern: fried food, salty snacks, and meat intake; prudent pattern: fruit and vegetable intake | Western and central Europe: highest adherence to prudent pattern, followed by western and oriental pattern | No |
| Menotti/Seven Countries Study (1999)( 18 ) | 7-d record (14 of 16) 1 d record (USA) 4 d record (Japan) | 18 food groups classified in all cohorts: bread, cereals, potatoes, vegetables, legumes, fruits, sugar, oils, butter, meat, fish, eggs, margarine + lard, milk, cheese, pastries, alcohol, and ‘other’; some analyses run on combinations of the 18 food groups: ‘vegetables foods’, ‘animal foods’, ‘sweets' | Factor analysis | No information | 1 factor score | Highest factor score and highest risk for CHD: east and west Finland, the Netherlands, Serbia; lowest factor and lowest risk for CHD: Italy and Greece | No |
| Pala/IDEFICS (2013)( 19 ) | CEHQ | Reproducible and validated; completed by parents or other caregivers (asks about consumption in preceding month) | PCA | Kaiser–Meyer–Olkin sampling adequacy was>0·6, which supports use of PCA; criteria for retained DP: eigenvalue, scree plot, factor interpretability), factor loadings>0·2 considered; for stability, also DP from the follow-up sample assessed; simplified DP (food variables with high loadings were standardised and summed) | Component 1 (Snacking) Component 2 (sweet and fat) Component 3 (vegetables and wholemeal) Component 4 (protein and water) | No information | Reliability assessed by generating DP in the subgroup of the follow-up participants and comparing it with the original DP |
DIETSCAN, Dietary Patterns and Cancer; ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study; NLCS, Netherlands Cohort Study; SMC, Swedish Mammography Cohort; ORDET, Hormones and Diet in the Etiology of Breast Cancer Risk; EPIC, European Prospective Investigation into Cancer and Nutrition; IDEFICS, Identification and prevention of Dietary- and lifestyle-induced health EFfects In Children and infantS; CEHQ, Children's Eating Habits Questionnaire.