base estimates of chronic toxic effects such as cancer. For acute toxic effects, peak exposures over shorter periods are more appropriate for risk assessment.
The development of food consumption data for evaluating chronic toxicity requires careful consideration. In general, food consumption surveys yield data on the consumption of that food over all days for which data are available. The average daily consumption for children within a given age class is then obtained by averaging across all the individuals in the age class.
Estimating the average daily consumption of a particular food within a given class warrants some discussion. Since some foods will not be consumed at all by some individuals, estimates of average daily consumption based on all individuals in the sample will underestimate average consumption for the subpopulation of individuals who consume the food in question. For this reason, separate estimates of average daily consumption for "all children" and for "eaters only" are considered when estimating exposure. Average consumption levels for "eaters only'' are typically 2 to 3 times higher than those for "all children."
Because food consumption data are available for only a few days each year, the proportion of children falling into the eaters-only group is underestimated. This problem is accentuated if only a 24-hour recall or 24-hour food record is used. If food consumption data were available for every day of the year, more children who consume the food of interest on an infrequent basis would be included in the eaters-only group. Thus, since the eaters-only group omits some individuals whose consumption levels are low, the average food consumption for "eaters only" calculated in this way actually overestimates the average consumption for this group. This bias does not occur when information on food consumption is obtained through food frequency questionnaires rather than 24-hour recalls or 24-hour food diaries, since food frequency tables in principle accurately identify those individuals who consume the food at any time during a given year.
Scientists working with food consumption data have long recognized that consumption by a "typical" individual will not be representative of consumption by people who eat large amounts of a particular food. This has stimulated interest in examining the distribution of average daily consumption levels across individuals in order to estimate consumption by individuals who consistently consume greater quantities of the food of interest than the average. This distribution of average daily consumption across individuals can be used to estimate upper quantiles of consumption,