Sponsored Satellite Program: Eating to Protect Our Health and Our Planet
Healthy and Sustainable Diets: What Do We Learn from Modeling Studies? (Sponsored Satellite Program)
Tuesday, June 2, 2020
8:55 AM – 9:15 AM EDT
To design healthy and sustainable diets, linear programming has been used extensively. This technique combines separate foods into a diet that fulfils preset criteria on daily nutrient requirements and environmental indicators such as greenhouse gas emissions (GHGE). However, acceptability to consumers is hard to account for. Data Envelopment Analysis (DEA) is a benchmarking approach that bypasses this shortcoming. It combines existing ‘best practices’ (benchmarks) of whole diets into new diets that are subsequently optimized for e.g. minimal deviation from current diets, nutritional quality, and environmental sustainability. We present results from food- and diet-based models to illustrate their pro’s and con’s.
The food-based Optimeal model was used to generate diets for Dutch adults. Its input data were current average food consumption, food composition data, LCA footprints, nutrient requirements, and GHGE-targets; consumer acceptability was incorporated by a (quadratic) score for deviance from the current average Dutch diet. Results showed that a healthy diet that meets the 2030 and 2050 GHGE-targets consistent with the Paris agreement on Climate Change (1.50C; ca 50% and 75% reduction of current emissions), required a shift from animal- to plant-sourced foods, reduction (2030 target) or elimination (2050 target) of cheese, whereas liquid dairy was reduced by 1-30% (2030 target) or by 70% (2050 target). An increase was observed for Ca and vitB12 enriched soy drinks (x2 and x10 for the 2030 and 2050 targets respectively). The diet shifts were stronger and diets became unrealistic when very strict criteria for GHGE reduction (2050 target) were applied.
The diet-based DEA model was applied to adults from four EU countries (DK, CZ, IT, FR). Input data were averaged daily diets from food consumption surveys, best practices were based on food based dietary guidelines (FBDGs), and optimisation was based on either nutrients, GHGE or a (linear) score for deviance from the initial diet. The model aims to design and compare diets that are SHARP: environmentally Sustainable, nutritionally Healthy, and Affordable, Reliable and Preferable for consumers. When environmental sustainability was used to optimise the diets, the Nutrient Rich Diet score (NRD15.3) became ~9% higher, GHGE was ~21% lower, and ~73% of food intake remained similar. Modelled diets had a similar proportion of animal-sourced foods, and plant-based food increased at the expense of alcoholic and sweet beverages. Protein sources, however, shifted from red and processed meat to either eggs, fish or dairy. Depending on the country, liquid dairy and cheese changed by +28% (5 to 45%) or -8% (-23 to +22%), with lowest and highest values for DK and CZ respectively.
Both modelling approaches suggest that liquid dairy (and to a lower extent cheese) fit into healthy and sustainable diets. This is because of the nutrient richness and lower GHGE footprints of liquid dairy as compared to other animal-sourced products like cheese and especially beef meat. Depending on the nutrient provision in a country, cheese may need reduction when strict GHGE targets are set. As compared to Optimeal, the SHARP-benchmarking model resulted in less extreme dietary changes with smaller improvements in nutritional quality and GHGE-reduction. Optimeal provides advice at the population level and needs explicit criteria for acceptability, whereas the SHARP-model provides individual diet advice that fits within the prevailing food culture.
Describe difference between food-based and diet-based modelling of healthy and sustainable diets.
Explain how a benchmarking model takes prevailing dietary habits into account to arrive at acceptable diet solutions.
Describe differences in results on food groups as a result from the the model used and/or the population studied.
State the need for harmonizing input data and reporting of results from studies on diet modelling.