Comparison of dietary intake by 2 computerized analysis systems

It is unknown how different calculation procedures for quantitating foods can influence variation in nutrient values. Precision of calculation and transcription, plus a double-checking procedure, is standard Bogalusa Heart Study protocol. These procedures likely would minimize human error. Nevertheless, converting gram weight values for food (BHS) into the closest household measure (NCC) could influence analysis. If calculation procedures are similar, different nutrient values in 2 food composition tables would then account for most variation in results. Consistently lower values for sucrose by the NCC likely reflect the program's carbohydrate rationale, i.e., the classification of total carbohydrate into 3 categories--sucrose, starch, and other. Sucrose may then be underestimated, as the other carbohydrate value "indicates combinations of sugars and/or sugars and starch," such as those found in commercial products. The ETNV total sucrose value is a sum of the naturally occurring sucrose and sucrose added in commercial preparations. For this small sample, the group means of selected nutrients by 2 dietary computer analyses systems are quite similar, with the exception of sucrose. With fairly large standard errors for most components, a large sample might create a statistically significant difference. Periodic comparability checks are essential to document quality and similarity of data bases. Quality control procedures in dietary methodology should be analogous to techniques for physiological and biochemical measures, e.g., the Centers for Disease Control and LRC methods and standards. Low-level correlations observed between dietary intakes and risk factor variables imply no association. Comparison and improvements of data bases are timely and should increase the probability of identifying associations between dietary components and indicators of diseases.