*3.2. Nutrient Pattern Analysis*

Eight nutrient patterns had an eigenvalue greater than 1 (Figure S1) that explained 73.4% of the total nutritional variation in the data. The rotated loadings are presented in Table A1 in Appendix A. A characterisation of the patterns, using the nutrients that loaded highly (absolute loadings >0.4) on each, is shown in Table 3, along with the supporting nutrients that had absolute loadings between 0.3 and 0.4. At least two nutrients per pattern had high loadings. PC scores were calculated for each food item and the highest score determined pattern membership. NP 1 was characteristic of food items high in plant protein, total fibre, magnesium, potassium, and manganese. Iron also featured on NP 1 but had a loading of 0.27—less than our threshold of 0.3. Wheat products, dark leafy greens, legumes, nuts, and seeds scored highest on this pattern. NP 2 was found to be high in vitamin A, copper, riboflavin, and vitamin B12 and linked with foods such as kidney, liver, mussels, and oysters. Meat, meat products, crab, and oily fish scored high on NP 3 as they shared high levels of animal protein, niacin, and vitamin B6. Fortified bread was also included due to its increased vitamin B6 content. Saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids, and vitamin E characterised NP 4 and were found to be highest in fats and oils, avocados, some nuts (almonds, pecans, walnuts, macadamias, and coconuts), and sauces. Foods made or fried with oil or margarine also scored highly on this pattern, as well as chicken skin and processed meats. NP 5 identified

foods high in calcium, phosphorous and sodium such as milk, milk products, canned vegetables, biltong, and shrimp. Foods made with milk and cheese were also found to be associated with NP 5. While most nutrients had positive loadings, moisture and zinc had a negative loading on NPs 6 and 8, respectively. NP 6 had positive loadings of available carbohydrate and thiamin, correlating with baked items, dried fruit, jams, as well as sugar and sweets, while NP 8 had positive loadings of vitamin A and vitamin C. Fruits and vegetables related mostly to NP 8, as well as soft maize meal. High cholesterol and vitamin D content characterised NP 7. Foods associated with this pattern were eggs, composite dishes using eggs, fish, offal, and tripe. Fortified milk powder and breastmilk substitutes also scored high on this nutrient pattern. Pantothenic acid featured on NP 2 and NP 7 but, like iron, had loadings below the absolute value threshold of 0.3.

**Figure 2.** Spearman's rank correlations between the nutrients.


## **Table 3.** Characterisation of nutrient patterns (NP).

Nutrients with absolute loadings >0.4 are indicated in italic.

Table 4 compares the food categories in the SAFCDB to the groupings found by the PCA. PCA groupings consisted of food items across the SAFCDB groupings, and the grouping structures were significantly associated (*p* < 0.001). Vegetables and legumes contributed 42% and 23% of NP 1, respectively. Meat and seafood together accounted for 91.8% of NP 3. All food items in the category 'Eggs' were grouped under NP 7, together with composite dishes from 'Cereals and cereal products' that were made with eggs. Most of the food items within 'Legume and legume products', 'Milk and milk products', 'Fats and oils' and 'Sugar, syrups and sweets' remained together under the PCA groupings.

**Table 4.** Food group percentage for each nutrient pattern (NP); *n* (%).


<sup>a</sup> Blank cells represent 0 (0).

Table A2 in Appendix A reports the median (IQR) nutrient values for each principal component grouping. Median nutrient values for each NP agreed with the characterisation of the patterns and are graphically represented in Figure 3. Table A2 in Appendix A and Figure 3 represent the expected nutritional composition of an average food item from each pattern. A randomly selected food item from NP 3 will, on average, contain the highest amount of animal protein and niacin than a food item from any of the other NPs. Sodium content can be expected to be lowest in foods from NP 8, and highest in foods from NP 5 and NP 7. NP 2 and NP 8 had the greatest cumulative quantity of vitamins, with vitamin C contributing the largest proportion of the composition. Niacin was also a large contributor of vitamin content in NP 2. The first three patterns had the highest quantity of minerals made up largely from phosphorous and potassium.

**Figure 3.** *Cont.*

**Figure 3.** Median nutrient values per nutrient pattern (NP) for (**a**) macronutrients; (**b**) minerals and (**c**) vitamins.
