Compiling Data for Food Composition Data Bases (UNU, 1991, 68 pages)
 Part II Gathering the data
 6. Calculations for multi-ingredient foods
 (introductory text...) Types of recipes Recipe calculation guidelines Limitations

### Types of recipes

It is conceptually useful to distinguish between recipes for simple combinations which require only mixing of ingredients and then correction for weight or volume, and recipes which require more extensive modifications of the data through estimation of losses or gains of water, fat, vitamins, and minerals. A third class of recipes includes those which require estimation of the amounts of the ingredients, and occasionally the nature of the ingredients themselves.

Simple Combinations

Some recipes require only the addition of the nutrients of the specified quantities of ingredients, followed by adjustment of the weight or volume of the final food, in order to express the nutrient levels per standard amount of the food (such as 100 g or household measures). In these cases the nutrient values should be for the ingredients in an "as consumed" form (i.e., cooked if the ingredients are cooked and containing no refuse). Examples include:

• coffee prepared from instant powder with added sugar and cream;
• baked potato with sour cream and chives;
• tomato soup prepared from condensed soup and whole milk. (Note that heating may modify the nutrient content in this situation.)

The nutrient levels of multi-ingredient foods that are prepared with the addition of water (such as milk reconstituted from powder) are also of this type, and can be estimated quite well if water (e.g., tap water, well water, bottled spring water, distilled water, mineral water) and its associated nutrient levels are included in the data base. If a data base does not include water as a food item, the nutrient contribution from water can be considered zero when calculating the nutrient content of a multi-ingredient food; however, this must be noted, since in some areas and for some nutrients the contribution from water is important. The weight contribution of water must, of course, be considered when expressing the nutrient levels per unit of multi-ingredient food.

A special aspect of these recipes is that, while the weight of ingredients add directly, volume often does not, and special care must be taken to ensure that results expressed on a per volume basis are correct. For example, one cup of powdered milk plus three cups of water yields significantly less than four cups of fluid milk; likewise, 1/2 cup mayonnaise added to one cup chopped apples, Y. cup raisins, and Y. cup chopped walnuts yields about 1.3 cups salad.

Occasionally, simple combinations may be used to calculate the nutrient content of a food from its component parts. In this case, the components are treated as "ingredients". For example, the nutrient composition of a chicken leg may be calculated from the nutrient composition and proportions of meat, skin, and separable fat.

Recipes Which Involve Nutrient Changes

Many recipes specify the combination of "as consumed" ingredients, but then require additional cooking which modifies the levels or densities of the nutrients of the foods involved. Here it is necessary to apply the various factors to correct for these changes due to water or fat loss or gain. Examples of these multi-ingredient items include:

• refried beans, made from boiled beans, lard, and spices which are fried in fat; and
• shepherd's pie, made from cooked minced (ground) beef, boiled onion, boiled potato, water, milk, margarine, salt, and pepper which is baked for 25 minutes to brown.

Recipes become more complicated when starting with the raw ingredients. These recipes require adjustment for refuse and yield during preparation and cooking as well as nutrient losses and gains. Consider, for example, the above shepherd's pie when data are available only for raw meat and vegetables.

Recipes for Which Ingredient Amounts Must Be Estimated

Occasionally, the amounts of ingredients are unknown and must be estimated. If partial nutritional data are available for the recipe, these may be used to estimate ingredient proportions, which in turn may be used to estimate the missing nutrient data. This procedure may be necessary in the case of proprietary food mixtures for which only some data on content (e.g., from the package label) are available, in order to obtain estimates for the other nutrients. These estimates are usually less certain than the usual recipe estimates.

In the following example the zinc content of a product estimated on the basis of information available about proximate nutrients. A more complex example (for a chocolatecoated ice cream bar) is given by Posati [68].

EXAMPLE: A compiler wishes to estimate the zinc content of canned corned beef hash and has only the manufacturer's information for proximate nutrients. The two main ingredients are potatoes and corned beef, but the ratio is not known. However, it is known that the product contains 11 g carbohydrate and 8 g protein per 100 g edible portion. Assuming all the carbohydrate is from the potatoes, and knowing that cooked potatoes have approximately 15 g carbohydrate per 100 g, a proportion of 11/15 or 73% potatoes (27% corned beef) can be assumed. This assumption can be checked by calculating the protein content. The corned beef component contains 25 g protein/100 g, which would contribute about 7 g/100 g to the hash (0.27 x 25). The protein in potatoes (about 2 g/100 g) would contribute about 1 g, giving the correct total of 8 g/100 g for the hash. Given this information, the zinc content of the hash can be estimated if the zinc content of corned beef and of cooked potatoes is known.

Another illustration of the procedure is the use of known ratios to estimate amounts of individual fatty acids and amino acids. Well-defined patterns of fatty acids for different classes of foods can be used to calculate specific fatty acid levels based on known total lipid content, and, similarly, patterns of amino acids can be used to calculate specific amino acid levels based on known total nitrogen content.

EXAMPLE: The amino acid pattern of milk can be used to estimate the individual amino acids of cream given the nitrogen content of cream and milk, and the assumption that cream has the same amino acid pattern as does milk. From Paul and Southgate [59], whole milk has 510 mg of methionine per gram of nitrogen and single cream (21% fat) has 0.376 g of nitrogen per 100 g. From this it is estimated that the methionine content of cream is 0.376 x 510 = 192 mg per 100 g.