Development of Food Group Composites and Nutrient Profiles for the MyPyramid Food Guidance System

Step 1. Identification of Food Group(s) Assignments for All Food Items Consumed 

We used the USDA Agricultural Research Service’s (ARS) Pyramid Servings Database (PSD), version 1.08 to identify the Pyramid food groups and subgroups contained in each food reported in NHANES and the amount in servings of each group in the food. The PSD consists of NHANES survey food codes, their descriptions, and numbers of food group servings per 100 grams of food. For example, the PSD identifies that each 100 grams of cooked carrots contains 1.37 servings of orange vegetables, and that each 100 grams of beef stew contains 0.13 servings of orange vegetables. In the same way, the PSD identifies the number of servings of whole and refined grains in 100 grams of a mixed-grain bread product such as oatmeal bread or part-whole wheat bread.

This breakdown allows the calculation of an amount consumed from each food group for a specified intake (in grams) of different food items. The PSD categories include not only the major Pyramid food groups (grains, vegetables, fruits, milk, and meat and beans) and subgroups, but also additional subcategories and components such as “discretionary” fat, added sugars, and alcohol. PSD data for the grains, vegetables, fruits, and milk food groups are reported in servings. Since MyPyramid no longer uses the nomenclature of “servings,” data for food groups from the PSD were translated into cup or ounce equivalents. One serving of fruit or vegetable is equal to one-half cup equivalent; one serving of milk is one cup equivalent; and one serving of grains is one ounce equivalent. For fruit juices, vegetable juices, and nuts and seeds, we updated the equivalencies from PSD version 1.0. PSD data for the meat and beans food group are reported in ounces of cooked lean meat equivalents. Discretionary fat is reported in grams, added sugars in teaspoons, and alcohol in number of drinks.

Step 2. Identification of Specific Foods Within Each Group, Including Disaggregation of Mixed Foods into Component Ingredients (Figure 1) 

The PSD can be used to identify total food group and subgroup consumption for each food reported in NHANES, but it does not indicate which specific item within the group the survey food contains. For example, the PSD shows that vegetable lasagna contains a certain amount of dark-green vegetables, but it is unknown what specific ingredient—such as spinach or broccoli—is responsible for this amount of dark-green vegetables. To determine the consumption-weighted intakes of foods in each food group and subgroup, more specificity is needed.

View Large Image Download to PowerPoint

Figure 1. Example of Disaggregation of a Food (NHANES Food Code 27311410) into Item Clusters

Therefore, in the development of the food patterns for the original Pyramid, food “item clusters” were created that identified specific foods within each Pyramid group.2 An item cluster was originally created for all foods whose consumption was more than 1% of the total number of servings consumed from each group or subgroup. Examples of item clusters are cooked broccoli, apples, and white rice.

For the current research project, we reviewed each food reported in NHANES within a particular food group or subgroup and assigned each food to an item cluster within that food group, based on the ingredients it contained. For example, we reviewed all survey foods containing dark-green vegetables, identified the vegetable they contained, such as cooked spinach or cooked broccoli, and assigned the food to the appropriate item cluster (cooked spinach or cooked broccoli). Each food item reported in NHANES may be a single ingredient food, such as an apple, or may contain multiple ingredients, such as a beef stew. One food as eaten may contribute to multiple food groups, subgroups, or item clusters. For example, a stew can be disaggregated into multiple food item clusters, as shown in Figure 1. We used an ARS food survey recipe file (Continuing Survey of Food Intakes by Individuals [CSFII] 1994-96, 1998)9 when necessary to identify specific ingredients in a mixed food. If the food were a new food in NHANES, we identified the ingredients from the food description or a market check.

Step 3. Aggregation of Foods and Ingredients in Each Item Cluster (Figure 2) 

After disaggregating mixed foods into their component ingredients, these component ingredients were then aggregated into food item clusters with the same ingredient from other foods, to identify total consumption. For example, amounts of tomatoes eaten as part of pizza, chili, spaghetti, and many other foods were identified and grouped together to determine total cooked tomato consumption.

View Large Image Download to PowerPoint

Figure 2. Example of Aggregation of Food Ingredients into Item Clusters

To determine total consumption of specific foods, we summed the amounts, in cup or ounce equivalents, of each survey food that had been assigned to the item cluster. These amounts represented only the fraction of the food that the PSD identified as being in that item cluster. The amounts of other ingredients in a mixed food were counted in other item clusters. For instance, we summed the total amount of cooked carrots from all survey foods containing orange vegetables that we had identified as cooked carrots, to determine the total amount consumed from the cooked carrots item cluster (see Figure 2).

Step 4. Calculation of Consumption from Each Food Group or Subgroup 

We then summed the total group intake from all item clusters within each group and calculated the percentage of total consumption contributed by each item cluster. For instance, we calculated the total amount of orange vegetables consumed (in cup equivalents) and divided the amount of cooked carrots consumed (in cup equivalents) by this total. The resulting percentages are the composite for the group and represent the probability of a food being eaten in comparison to other foods in the group. Note that this probability represents the proportion of all orange vegetables that are cooked carrots consumed by the population, not the frequency with which a person might select cooked carrots in comparison to their selection of other orange vegetables.

As these calculations proceeded, item clusters were reassessed and updated as needed to reflect current food consumption of Americans. Food item clusters were retained for all foods whose intake represented more than 1% of the total intake of the food group or subgroup. Most foods accounting for less than 1% of intake were grouped together or with an existing item cluster. Similarity in nutrient composition and use in meals were used as guides for how foods should be combined. For example, since strawberries are the most widely consumed berry, other berries were grouped with strawberries in a single item cluster. Some exceptions were made to the 1% rule, such as liver, which was less than 1% of meat group consumption but was left as an item cluster, because it differs substantially in nutrient content from other meats. New item clusters were created for several reasons. In some cases, consumption of a food item had been relatively low when the original item clusters were established in the 1980s but had increased to more than 1% of consumption within the food group. One example is popcorn in the whole grains group. Another reason for creating new item clusters was that for some items, such as fish and nuts, a more detailed representation of their consumption was desired for related research.10

Step 5. Selection of Representative Food for Each Item Cluster 

Once item clusters were established, we selected a food from SR17 to serve as the representative food for the cluster. The selected food was in a nutrient-dense form, that is, in a low-fat and no added sugar form. For all food groups, the standard is to select a food to represent each cluster that is in its leanest or lowest fat form and that is prepared without the addition of fat, oil, or sugar. For some groups, the selection of a nutrient-dense form mainly considers preparation, such as “cooked broccoli without fat” representing the cooked broccoli item cluster, and “raw strawberries” representing all berries in that item cluster. For other groups, the selection considers the item with the lowest fat content of all items in the cluster, such as using 95% lean/5% fat ground beef to represent all ground beef.

In some cases, the representative food was not the most highly consumed food within the cluster. For example, boiled potatoes were used as the food to represent all boiled and fried potatoes. French fries and potato chips are highly consumed but contain added fat or oil and therefore are not in the most nutrient-dense form. The fat or oil within foods such as French fries or in higher fat meats was represented within the discretionary fat composite described in step 7. The nutrient values of the food items selected to represent the item clusters were used to calculate the food group or subgroup nutrient profiles (step 6).

Nutrient-dense forms of the food were selected to represent each item cluster, in keeping with one of the guiding principles for the original Pyramid and this revision, which is flexibility. This principle states that “the food guide should allow maximum flexibility for consumers to eat in a way that suits their taste and lifestyle while meeting nutritional criteria.”3 Using these nutrient profiles, food patterns can meet all nutrient needs at relatively low calorie levels. Once these needs are met, the balance of calories to meet energy needs can be selected by consumers from foods they prefer, as long as other nutritional criteria are met.

Step 6. Calculation of Nutrient Profile for Each Food Group and Subgroup (Figure 3) 

We established nutrient profiles for each food group and subgroup for energy plus 27 nutrients. In calculating these nutrient profiles, we assigned a weight to the nutrients from each representative food that corresponded to the percent consumption of its item cluster. We then summed the nutrient values for all item clusters to calculate the total nutrient profile of each food group composite (see Figure 3). Nutrient profiles are based on the nutrient content of a half-cup equivalent for the fruits and vegetables groups, of a 1-ounce equivalent for the grains and meat and beans groups, of 1 cup of milk, of 1 teaspoon of added sugars, and of 10 grams of solid fats and oils.

View Large Image Download to PowerPoint

Figure 3. Calculation of Nutrient Profiles

There were two exceptions to the use of item clusters and composites for the calculation of nutrient profiles. We did not use the item cluster approach to calculate a nutrient profile for added sugars or the milk group. For all added sugars, the nutrients in granulated white sugar were used for the nutrient profile. The nutrient profile for the milk group was represented by the nutrients in 1 cup of fat-free fluid milk, with an exception made for vitamin A. For vitamin A, we used the amount found in whole milk (68.6 μg RAE per cup), instead of the amount found in fortified fat-free milk (149 μg RAE per cup). This exception was made to avoid overestimation of vitamin A for those who consume nonfortified milk products.

Step 7. Calculation of Nutrient Profile for Discretionary Fats 

In addition to the food group composites and nutrient profiles, we developed composites and nutrient profiles for “discretionary fats.” Discretionary fats are fats in the diet above what would be found in lean meats and other low-fat foods used to represent item clusters. These fats may be contained within higher-fat forms of foods, or added in processing, cooking, or at the table. Because solid fats and oils have differing fatty acid profiles and effects on health, we separated solid fats from oils and constructed separate composites and nutrient profiles for each.

To summarize the method for deriving the solid fats and oils composites, we first calculated the amount of fat that was intrinsic in each of the food groups and subgroups. We needed to use the food supply to determine the different fats and oils, because the survey does not distinguish what the specific fats and oils are in many of the food items. We used percentages of different animal and vegetable fats in the food supply for 199611 and an in-house food supply database12 to calculate the total amount of fat in a sample food intake pattern. (The 1996 food supply was used in order to match the CSFII 1994-96 food consumption data that were used for preliminary work on the food composites.) By subtracting the intrinsic fat from the total fat, we calculated the amount of extrinsic fat, or discretionary fat. Then, fat amounts in the food supply were used to estimate discretionary fat sources. These fat amounts reflect fat contributed from major food groups beyond the amounts that are intrinsic in lean choices within each food group composite, fat used in food processing or food preparation, and fat consumed by itself or with other foods. The types of hard and soft margarine chosen for the solid fats and oils composites, respectively, were obtained from data in USDA’s CSFII 1994-96.13

Food group composites and their nutrient profiles were used to calculate nutrient levels in the MyPyramid food intake patterns as they were being iteratively developed. These nutrient levels in the patterns were then compared to nutritional standards to determine whether nutrient recommendations had been met by the food intake patterns. This process is described in detail in an accompanying article.4

SAS statistical software, version 9.1, was used to analyze weighted estimates for the representative population and to develop composites and nutrient profiles.

Composites 

Table 1 shows the consumption of each of the 20 item clusters in the fruit group as a percentage of total cup equivalents. Forty-seven percent of fruits were consumed as juice, while raw and cooked fruit represented about 46% and 7% of total fruit consumption, respectively. In terms of total cup equivalents, orange juice was the most consumed item in the fruit group, and bananas were the next most commonly consumed.

For development of the new food group composites, nutrient profiles, and food intake patterns, the amount of fruit juice considered to be equivalent to 1/2 cup of fruit was changed from 3/4 cup to 1/2 cup. This change was based on differences in the nutrient content of fruits and juices that were largely due to the larger portion size for juices, and a desire to make the cup equivalencies easier for consumers to understand. This change resulted in a more accurate representation of the proportion of overall fruit intake that comes from fruit juice. The reason for this change is further described in an accompanying article.10

Table 2 presents the consumption of item clusters within each vegetable subgroup as a percentage of total cup equivalents. The composites do not directly identify the proportion of total vegetable intake from each subgroup, but these proportions were also calculated. Of all vegetables consumed, 6% were dark-green vegetables, 4% were orange vegetables, 7% were dry beans and peas, 37% were starchy vegetables, and 46% were other vegetables. Broccoli, raw and cooked, was the most commonly consumed dark-green vegetable; carrots were the most commonly consumed orange vegetable; pinto beans the most consumed dry bean; white potatoes (in any form) the most-consumed starchy vegetable; and tomatoes the most-consumed other vegetable. As with fruit juices, the amounts of vegetable juices considered equivalent to 1/2 cup of vegetables were changed from 3/4 cup to 1/2 cup.

The consumption for the numerous item clusters in the meat, poultry, fish, dry beans, eggs, and nuts group is reported in Table 3 as a percentage of total ounce equivalents. Meats (beef, pork, lamb, and organ meats) made up 53% of overall consumption in this group; poultry (chicken and turkey) accounted for about 23% of the group consumption; all fish and shellfish (high and low in eicosapentaenoic acid [EPA] and docosahexaenoic acid [DHA]) were 8% of the food group consumption; eggs were about 7% of overall consumption for the group; and nuts and seeds accounted for about 8% of total consumption.

Table 4 shows the percentage of consumption of the 12 item clusters in the refined grains and 11 item clusters in the whole grains subgroups. These results include the disaggregation of products that are a mix of whole and refined grains into appropriate item clusters in the two grain subgroups. Although not represented in the subgroup analysis, the proportion of all grains consumed that was whole or refined was also calculated. Whole grains made up 13% of all grain consumption, and refined grains made up the other 87%. Whole-grain snack products (corn chips, popcorn, and crackers) (38%) and cereals (36%) made up the majority of the consumption of whole grains, whereas whole-grain breads accounted for about 24% of all whole grains consumed. In contrast, refined breads and crackers (46%) and desserts and other baked products (25%) made up the majority of refined-grain consumption, whereas pasta and rice accounted for 18%, and cereals for only 4% of all refined-grain consumption.

As noted in the methods section, the “discretionary” fat composite from the original Pyramid was divided into separate composites for solid fats and oils/soft margarines. The solid fats composite is made up of butterfat (44%), shortening (21%), hard margarines (16%), lard (11%), beef fat (4%), chicken fat (2%), and pork fat (2%). Solid fats represent fats that may be consumed as part of higher-fat food selections within a food group, as well as those that might be added in food processing, cooking, or at the table. For example, the butterfat in whole milk and cheese is considered part of the solid-fat composite, as is the beef, pork, and chicken fat contained in higher-fat cuts of these foods. Solid fats used in the production of foods are also considered, such as shortening in a cracker or doughnut, and butter in a cake or cookie. The oils and soft margarines composite reflects those used in processing, in cooking, and at the table. It is composed of soybean oil (77%), soft margarines (13%), corn oil (6%), and cottonseed oil (4%). An allowance for oils was included in the final food intake patterns. The solid-fat composite was used in calculations to develop the discretionary calorie allowance for the food intake patterns and to determine the amounts of fatty acids in the resulting patterns.

Nutrient Profiles 

Table 5 lists the amounts of energy and 27 nutrients per reference amount of each food group and subgroup. Reference amounts are 1/2 cup for the fruit and vegetable groups, 1 ounce equivalent for the grains and meat and beans groups, 1 cup for the milk group, 1 teaspoon for sugar, and 10 grams for solid fats and oils. The nutrients shown in Table 5 were those used in determining the nutrient adequacy and moderation of the food intake patterns. (See the accompanying article on the food intake patterns for additional details.)4

Each food group provides a wide array of nutrients for the food patterns in varying amounts. The major sources of several nutrients of concern will be summarized here. These nutrients are those that the Dietary Guidelines identified as “nutrients of concern” for Americans.14