How optimal are computer-calculated optimal diets?

Objective: This paper aims at clarifying the meaning of `optimal diets'. Design: Two different optimization approaches are considered. Both depart from people's actual diets which in general do not meet all their nutrient requirements. In the ®rst case (non fuzzy approach) the actual diets are modi®ed such that all nutrient intake recommendations are met and the resulting changes in the persons ' food habits are as little as possible. In the second case (fuzzy approach) the actual diets are modi®ed such that the persons ' food habits undergo little change and the nutrient intake recommendations are met as well as possible. Both approaches are evaluated using data on the actual diets of three randomly chosen participants of the Bavarian Food Intake Survey. Results: With both approaches feasible solutions can be found. The optimal diets computed with the non fuzzy approach satisfy all the persons ' nutrient requirements, but make it necessary to change the persons ' food habits considerably. When compared to the subjects ' actual diets, the optimal diets computed with the fuzzy approach yield improvements for the intake of many nutrients, but in some cases deteriorations also occur (for example, concerning folate, calcium and iodine). So, the process and the results of the fuzzy approach are not `Pareto ef®cient'. But it has the advantage that the corresponding optimal diets deviate from the actual diets by no more than about 3 ± 5 kitchen units per day. Conclusions: The term `optimal ' does not have a general meaning. An `optimal diet ' does not necessarily meet all the nutrient requirements of a person. Optimality only depends on the respective conditions any optimization is based on. Descriptors: nutrition optimization; quadratic programming; fuzzy logic; nutritional requirements; health planning guidelines; Bavarian Food Intake Surve