Serving size issues in estimating dietary exposure to food substances.

The intake of cola beverages from two national surveys (NFCS 1977-78 and NHANES II) was used to examine certain assumptions important to a mathematical model employed to estimate dietary exposure to substances in food. The assumptions center on the relationship between frequency of consumption and serving size and on the constancy of serving size within age groups. Results showed the frequency of cola consumption was associated with 55% to 83% of the variability in cola intake and that mean cola serving size was not linearly related to frequency of consumption. However, while upper-level (90th + percentile) cola consumers consumed more frequently, they also had larger than average serving sizes. Therefore, because the model incorporates a standard serving size, it significantly underestimated (p less than .01) intake for the two groups of upper-level consumers examined. Furthermore, comparisons for cola serving sizes reported in the surveys showed differences by sex as well as systematic differences between the surveys in that the average difference between age and sex groups within each survey was similar, but NFCS 1977-78 had a larger average size for each group than did NHANES II. Attention should be given to the nature of the food consumption databases used for estimating dietary exposure because systematic bias may cause considerable differences in estimation.