Effect of model parameters on the development of food-based recommendations using linear programming

Optifood is a tool developing food-based recommendations (FBR) by linear programming. Model parameters are based on dietary pattern, comprising common foods, portion sizes and consumption frequency. The effect of differences in model parameters between studies is unknown. Using repeated 24-h recalls of 62 Kenyan children (4-6y) median portion sizes were calculated. Minimum and maximum consumption frequencies were defined as resp. 5th and 95th percentile of the distribution. Optifood was used to develop FBR and identify problem nutrients. Model results using first recalls and foods consumed by >3% of the children were compared with repeated recalls and foods consumed by >10%. Using a single recall including foods consumed by >3% of the children, the FBR consist of starchy plant foods, vitamin A rich fruit, fortified milk, oil, and small fish each for 7 times/week. The food list contained 48 foods; problem nutrients were vitamin A and zinc. Using foods consumed by >10%, the food list reduced to 26 foods and additional problem nutrients were folate and fat. (Sub)food groups bread, starchy plant foods, fortified milk, vitamin C rich fruit and fish were not included. Using repeated recalls, the food list increased to 57 foods. No additional problem nutrients were identified. The subfood groups nuts and eggs were added. Maximum frequencies of (sub)food groups bread, starchy plant foods, fortified milk, vegetables, fruit and fish & eggs decreased. Compared to single recalls, portion sizes decreased for 19 foods and increased for 18 foods. Differences in defining model parameters affected food based recommendations identified through linear programming and should be further standardised.