The Baltimore Longitudinal Study of Aging (BLSA), which began in 1958, is designed to look at the normal functions of the body and cognitive aspects of aging and disease (Shock et al., 1984
). The current study includes all individuals who have participated in the dietary assessment portion of the study since its onset (n
= 1516). Dietary data were collected with 7-d food records. Earlier reports of dietary intake in the BLSA population and a full description of the dietary data collection methods are published elsewhere (Hallfrisch et al., 1990
; McGandy et al., 1966
). In brief, BLSA participants were trained to record their food intake by dietitians during their examination visit and ambiguous or incomplete records were clarified by telephone interview.
Throughout the study, dietary data were coded and analyzed for nutrient content at different locations. Dietary data collected from 1961 to 1965 were sent to the Heart Disease Control Branch of the NIH for nutrient analysis; from 1968 to 1975, to Washington University; and from 1984 to 1991 they were coded into a nutrient database maintained by the BLSA. Since 1993, dietary data have been entered into the Minnesota Nutrient Data System (NDS) (Program 2.9, Food Database version 11A, Nutrient Database version 26, University of Minnesota Nutrition Coordinating Center) at the Jean Mayer USDA Human Nutrition Research Center on Aging at Tufts University. Dietary data from previous years were recoded into NDS, and nutrient intakes from earlier years were back-adjusted to correct for changes in the food supply using previous USDA food composition tables (U.S. Department of Agriculture, Agricultural Research Service, 2007
), specifically changes in the composition of pork, beef, breakfast cereals, and other fortified foods using data from the USDA for appropriate time intervals (Watt and Merrill, 1963
; U.S. Department of Agriculture, Agricultural Research Service, 1999
For this study, all foods containing grains and mixed dishes with grains, either whole or refined, were identified (). Each of these foods was then assigned the best-matched pyramid code from the Pyramid Servings Database (U.S. Department of Agriculture, Agricultural Research Service, 1997
), a reference database of servings for 30 food groups, including three grain groups (total grain, whole grain, and non-whole grain).
In most cases, an exact match for the foods consumed by the study participants was available in the Pyramid Servings Database. However, there were some instances where foods seldom consumed by participants were not found in the Pyramid Servings Database (e.g., rice flour, rye flour, potato flour, quinoa, and triticale). In these cases, we matched foods to the best possible match in the Pyramid Servings Database on the basis of whole grain content. For example, rye flour was matched to the Pyramid Servings Database food code for whole-wheat flour. When questions remained about whether a product was made with whole or refined grains (such as the whole grain content of corn-based salty snacks, for example), food manufacturers were contacted to obtain information about ingredients and whole grain content.
An important goal of building the whole grains database was to use the dietary data from the 7-d dietary records to estimate grain intakes in gram weights per day rather than servings per day. This process involved several steps because of changes in the nutrient database over time (e.g., categorizing foods vs. ingredients) and challenges in quantifying how many grams of whole grains are contained in a given serving of grain food.
Many of the foods in the whole grains database are recipes or mixed dishes, which are composed of more than one ingredient such as sandwiches, soups, and casseroles. To quantify the whole grain content and the percent contribution of grains (whole grain and non-whole grain) from these mixed dishes and recipes, we merged the whole grains database with the CSFII 1994–96 recipe database to disaggregate whole foods into ingredients and to obtain gram weights of individual ingredients. The grain components of the ingredients were first calculated on the basis of 100 g of the total recipe and then remerged with the Pyramid Servings Database to obtain grain servings. For example, in a cookie with 11 ingredients and 10 g of white flour per 100 g of the total cookie, the flour gram weight was merged to the Pyramid Servings Database to obtain 0.63 servings of a non-whole grain (i.e., refined grain).
In the whole grains database, some foods cannot be disaggregated to the ingredient level. In these cases, the reference gram amount of grains was determined by multiplying by the number representing the pyramid serving. Specifically, a grain serving of 1 slice of bread contains 16 g of grain, while a grain serving of 1/2 cup of cereal contains 28 g of grain. Once the whole grains database was completed, we calculated the absolute amount of grain consumed by each individual by multiplying the gram amount eaten by the reference values/100 for both servings and grams.
Using these methods, we were able to estimate precise quantities of grains at both the ingredient and food levels. Some unexpected foods were identified as containing grain. This is mostly because many foods contain ingredients such as corn starch or other stabilizers and fillers, which are made from grains. For example, imitation mayonnaise contributes a very small amount of non-whole grain due to corn starch. Because the dietary data are obtained from food records, we were able to count these foods as contributing to grain consumption despite their limited contribution, thus improving the accuracy of our estimates.