*3.3. Food Pattern Analysis*

We also applied PCA to the 971 food items to confirm the nutrient patterns we found. Seven food patterns had both an eigenvalue greater than 1 and accounted for at least 1% of the variation in the dataset (Figure S2). The seven FPs explained 97.4% of the total variation in the data (Table 5). The maximum absolute loading for components ranged from 0.07 to 0.16, hence, we interpreted the components using absolute loadings that were greater than 0.05. The patterns found when analysing the food items reflected the patterns found when analysing the nutrients thus, confirming the presence of nutrient patterns and validating our results. Both analyses identified a pattern grouping together wheat products, leafy vegetables, and legumes (NP 1 and FP 2) as well as patterns for milk and milk products (NP 5 and FP 3), and eggs and food items using eggs (NP 7 and FP 4). However, applying PCA to the food items enabled better discrimination of fruit and vegetables by vitamin C (FP 1) and beta-carotene (FP 6) content. Orange-coloured fruit and vegetables were identified with FP 6 in the food item analysis. In addition, greater discrimination was apparent among dark leafy greens, which were split between FP 2 and FP 6 compared to being grouped together under the nutrient analysis. Composite dishes using distinctive ingredients were also able to be identified and grouped with the raw versions of the ingredient. For example, carrot cake grouped with carrots and pastries made using eggs grouped with eggs. Rusks made with wholewheat flour (FP 2) and rusks made with white flour (FP 5) were also able to be identified and separated. Processed meat such as luncheon meat and sausages were separated from meat and meat products, and instead grouped with processed cheese (FP 5) and processed fish. Sodium scored high on this food pattern. The last pattern in the food item analysis (FP 7) separated soft maize meal from the stiff and crumbly versions based on its higher moisture content, similar to the results of the nutrient analysis. Soft maize meal was grouped together with other moisture rich food items such as beverages, cabbage and brinjal.


**Table 5.** Characterisation of food patterns (FP).

The patterns found supported the South African FBDGs [4], as shown in Table 6. Guideline 1 aims to facilitate balanced nutrient intake by encouraging the consumption of a variety of foods. As the nutrient patterns obtained differ in nutritional composition, consuming foods from different patterns supports this guideline. Starchy foods, as described in Guideline 2, such as bread, rice, cereals, and pasta were associated with NP 6. NP 6 also contained products high in sugar content such as cakes, cookies, and sweets and reflects Guideline 10. Similarly, other nutrient patterns were able to be matched to the South African FBDGs.

Guideline 11 was best captured by FP 5 and reflected categories targeted by the national sodium regulation [6], as highlighted in Table 7. Foods affected by the regulation were all found within FP 5.

**South African Food-Based Dietary Guidelines [4] Corresponding Pattern** 1. Enjoy a variety of foods. The nutrient patterns obtained differ in nutritional composition. 2. Be active! Not applicable 3. Make starchy foods part of most meals. NP 6 4. Eat plenty of vegetables and fruit every day. NP 8 5. Eat dry beans, split peas, lentils, and soya regularly. NP 1 6. Have milk, maas, or yoghurt every day. NP 5 7. Fish, chicken, lean meat, or eggs can be eaten daily. NP 3, NP 7 8. Drink lots of clean, safe water. Not applicable 9. Use fats sparingly. Choose vegetable oils, rather than hard fats. NP 4 10. Use sugar and foods and drinks high in sugar sparingly. NP 6 11. Use salt and food high in salt sparingly. FP 5

**Table 6.** Comparison between the South African food-based dietary guidelines and principal component analyses.

**Table 7.** Comparison between the food categories targeted by the national sodium regulation and foods associated with FP 5.

