A linear programming based method for designing menus for controlled feeding trials

Background: Controlled feeding trials are an important method to determine cause-effect relationships between dietary intake and metabolic parameters, risk factors, or health outcomes. Participants of a controlled feeding trial receive full-day menus during a prespecified period of time. The menus have to comply with the nutritional and operational standards of the trial. Levels of nutrients under investigation should differ sufficiently between intervention groups, and be as similar as possible for all energy levels within intervention groups. Levels of other key nutrients should be as similar as possible for all participants. All menus have to be varied and manageable. Designing these menus is both a nutritional and a computational challenge that relies largely on the expertise of the research dietician. The process is very time consuming, and last-minute disruptions are very hard to manage.

Objective: This paper demonstrates a mixed integer linear programming model to support the design of menus for controlled feeding trials.

Methods: The model is demonstrated for a trial that involved consumption of individualized, isoenergetic menus with either a low or a high protein content.

Results: All menus generated by the model comply with all standards of the trial. The model allows for including tight ranges on nutrient composition, and complex design features. The model is very helpful in managing contrast and similarity of key nutrient intake levels between groups and energy levels, and in coping with many energy levels and nutrients. The model helps to propose several alternative menus and to manage last-minute disruptions. The model is flexible; it can easily be adapted to suit trials with other components or different nutritional requirements.

Conclusions: The model helps to design menus in a fast, objective, transparent, and reproducible way. It greatly facilitates the design procedure for menus in controlled feeding trials and lowers development costs.