Compilation of the food composition table

In order to compute fat-soluble vitamin intake, a food composition table (FCT) needed to be compiled. For each food item of the FFQ, values were retrieved for vitamin A, β-carotene, vitamin D, E and K. A protocol was developed to ensure that a uniform method was applied. Figure 1 presents an overview of the successive steps followed in the compilation of the food composition table. The Belgian food composition table Nubel was used as basis [23]. Since corresponding food codes in the table often lacked data on fat-soluble vitamins, five other databases were used in the following order: 1) NEVO, the Dutch food composition database, 2) McCance and Widdowson’s composition of foods integrated dataset, the UK food composition table, 3) Clical, the French composition table, 4) the Danish food composition table, and 5) the American food composition table [24–28]. When data were missing for the first source, the next source was consulted until the appropriate values were found.

Successive steps in the compilation of the food composition table, VITADEK study, 2015–2016

All nutritive values were derived from the tables except for the values of the fortified nutrients. Those values were taken from the ingredient list of the food item. Fortified values were marked so that contribution of fortified foods in total vitamin intake could be derived afterwards.

For all prepared foods such as meat, meat substitutes, eggs, fish and vegetables, cooking fats were not taken into consideration when compiling the FCT. Fat uptake was calculated afterwards following Bognar et al. [22]. Vitamin values were thus obtained from food items that were prepared fat-free. This implies the use of preparation methods such as boiling, preparation in the microwave and in the steamer. Because the FFQ does not take the preparation or cooking method of foods into account, the latter was determined using a data-based approach. Data of the BFCS 2014 specifies the number of people that consumed a certain food item raw or prepared (baked, steamed, cooked etc.). Depending on the method that was most frequently used, the vitamin values of either the prepared food item or the raw food item were included in the FCT. When only the raw food item was available in the composition table, this was marked. Thus, the nutrient composition of the cooked food could be calculated later on from the raw food by applying the nutrient retention factor (calculated by the True retention method (%TR)). The nutrient retention factor accounts for the additional loss of vitamins due to the cooking method [29].

After consulting these six composition tables (Nubel and five international tables) a substantial number of food items still contained some missing values. The products with missing values for vitamin A (retinol and carotenoids) and E were analysed in the nutriFOODchem lab of Ghent University, while missing values for vitamin D and vitamin K were derived from ingredient or recipe-based calculations, through similarities or as a median value.

Using the method of Bolton-Smith et al., missing values for vitamin K and D were calculated based on the vitamin K or D values of their components [20]. The ingredient list was consulted in order to extract the proportion of each of the main components. If no such list was available, a cookbook was consulted to derive the proportion of ingredients. Ingredient-based calculations were applied for vitamin K and vitamin D values in cookies and spreads and for vitamin K values in sauces, bread, pancakes, waffles, pastry and some milk substitutes. In this process, the following assumptions of Bolton-Smith were applied. When the type of oil or margarine was not stated or different oils/margarines were used depending on the brand of the food, blended vegetable oil or margarine was chosen. The contribution of spices, dried herbs and seasonings were assumed negligible [20].

Missing values for vitamin K in some cheese and fish were derived from products with similar fat and/or water content [20, 28]. Finally, lacking values for foods that were part of a group of similar food items (for example cereals), were determined based on median values. This was applied for vitamin K values in cereals.