Sustainable healthy diet modeling for a plant-based dietary transitioning in the United States

The study examines how transitioning U.S. diets toward plant-based patterns affects both nutritional quality and environmental impacts when diets are considered in realistic composite forms. While plant-based diets (PBDs) are associated with health and environmental benefits, definitions vary and adoption is growing in Western countries. Food systems contribute substantially to greenhouse gas emissions, with animal-based foods disproportionately impactful. Prior research (e.g., Food4HealthyLife) modeled life expectancy gains of optimal/feasible diets but did not assess environmental impacts, and other work emphasized single cooked items rather than uncooked composite diets. The research question is how composite dietary scenarios that partially replace animal meats with legumes perform across nutrition and environmental metrics, and what trade-offs identify an optimal sustainable pattern. The purpose is to design and evaluate composite diet scenarios bridging current to optimal patterns and determine an optimal trade-off scenario for sustainable dietary transition in the U.S.

Background literature indicates: (1) Plant-based diets confer health and environmental benefits but are diversely defined across contexts (WHO, Craig et al., Ostfeld). (2) Food production contributes 17–32% of global GHGs; animal-based foods contribute about twice those of plant-based foods (Xu et al.). (3) Western adoption of PBDs is increasing, with growth in vegetarian prevalence, plant-based product markets, and institutional initiatives (e.g., Meatless Monday). (4) Health literature links higher intake of fruits, vegetables, legumes, nuts, and whole grains to lower risks of NCDs including CVD, cancer, and type 2 diabetes. (5) Methodological precedents include the Food4HealthyLife calculator (Fadnes et al.) that modeled life expectancy impacts but did not quantify environmental effects, and Stylianou et al. who assessed nutritional-environmental impacts mostly for single foods, not composite diets. This evidences a gap in assessing composite diets’ joint health and environmental performance and trade-offs.

The study followed four steps: (1) Consumption model and scenario design, (2) Nutritional quality assessment, (3) Environmental impact estimation, and (4) Nutrition–environment trade-off and statistical analysis. 1) Consumption model and scenario design: Using the Food4HealthyLife calculator definitions of current (S1), feasible (S5), and optimal (S10) diets, three consumption models were created to partially replace meats with legumes: M1 (replace red meat), M2 (replace red and white meat), M3 (replace red, white, and processed meat). For each model, 7 additional scenarios were generated via arithmetic equations based on current/feasible/optimal weights of targeted food groups, yielding 21 alternatives plus 3 defaults (S1, S5, S10) for 24 total scenarios across M1–M3. Diets comprised 14 food groups (whole grains, refined grains, nuts, legumes, fruits, fish/seafood, red meat, processed meats, dairy, eggs, white meat, added oils, sugar-sweetened beverages). Each daily diet totaled 1.8 kg (Food4HealthyLife baseline). Representative foods per group were selected using USDA availability/consumption data. Cooking/storage losses were excluded. 2) Nutritional quality assessment: Nutrient contents per 100 kcal for each scenario were computed using FoodStruct (macros; 9 minerals; 12 vitamins; specific lipids including cholesterol, ALA, EPA, DHA). Diets were scored by Food Compass (Mozaffarian et al.) with adaptations: 46 attributes across 7 domains; processing and phytochemical domains excluded (uncooked models; limited data); iodine, trans-fats, medium-chain fatty acids, total flavonoids, and carotenoids not used. Red and processed meats were scored as separate attributes. Volume-based cutoffs were converted to grams. Final Food Compass Scores (FCS) were scaled to 0–100 via FCS = 100 − (26.1 − original score)*99. Health Nutritional Index (HENI) scores were calculated per 100 kcal following Stylianou et al., using GBD 2017 dietary risk factors (9 food groups + 6 nutrients) with their HENI factors; DALYs were converted to minutes via HENI_diet = −0.53 Σ(HENI factor × diet component). 3) Environmental impact estimation: For each scenario, environmental impacts were estimated by mapping ingredients to WWEIA representative foods and applying mean midpoint impacts from Stylianou et al. per RACC. Eighteen indicators were included: short- and long-term global warming, water use, ionizing radiation, mineral resources, freshwater ecotoxicity, ozone layer depletion, fine particulate matter formation, freshwater acidification, fossil energy use, marine eutrophication, land occupation, freshwater eutrophication, terrestrial acidification, human toxicity (cancer and non-cancer), total ecosystem quality damage, and total human health damage. Ingredient impacts (factor per gram × ingredient weight) were summed to scenario totals. 4) Trade-off and statistics: Dual-scale charts compared nutrition (FCS, HENI) with environmental indicators emphasizing short-term global warming (correlated with most indicators), ionizing radiation (human health-relevant magnitude), and freshwater eutrophication (pattern distinct, uncorrelated). Pearson correlations assessed relationships among calories, nutrition scores, and environmental indicators. Kruskal–Wallis ranked scenarios by FCS, HENI, and total human health damage. Analyses used Excel, SPSS v25, and Python 3.10.5/Jupyter.

- Nutritional quality: FCS increased from 65.46 (current S1) to 81.73 (optimal S10); 22/24 scenarios had FCS ≥ 71. HENI ranged from 60.21 (S1) to 320.85 min/100 kcal (S10). FCS and HENI were strongly positively correlated (~93.3%), while HENI correlated negatively with calorie density (~−91.8%). Rankings differed between FCS and HENI for intermediate scenarios (e.g., S7M1 ranked 2nd by HENI but 5th by FCS; S8M1 ranked 2nd by FCS but 23rd by HENI). S1 consistently ranked lowest, S10 highest. - FCS domain contributions: On average, vitamins contributed most (31.41%), specific lipids least (0.051%). As scenarios approached optimal, contributions of nutrient ratios and food ingredients rose, while minerals and vitamins fell; protein/fiber remained relatively stable (~7%). In S1, minerals and vitamins contributed ~80% (40.15% and 40.24%); in S10, these dropped to 24.18% and 14.11%, while food ingredients rose to 41.18%. - HENI risk factor contributions: Major average contributors were vegetables (19.61%), fruits (18.09%), sugar-sweetened beverages (15.16%), milk/dairy (15.11%), whole grains (8.28%), and seafood (7.53%). Beneficial factors increased and harmful factors decreased as diets approached optimal. From S1 to S10, contributions increased for vegetables, fruits, seafood, nuts/seeds, legumes, fiber, and whole grains; SSBs dropped from 31.5% to 0%. - Environmental impacts: Across indicators, impacts generally decreased as scenarios moved toward S9M3 (10% legumes; 0.11% red meat; 0.28% processed meat; 2.81% white meat), then increased at S10 for several categories (e.g., freshwater eutrophication, water use). Transitioning from S1 to S9M3 reduced short-term global warming by 54.72%; S9M2 reduced it by 46.22%. Early transitions (S2M1/M2/M3) increased global warming by ~7.0–7.6% relative to S1, reflecting higher quantities of some healthy but environmentally intensive foods before larger meat reductions. Global warming correlated strongly (r^2 > 0.80, α = 0.01) with most indicators except water use and freshwater eutrophication. - Trade-offs: Environmental impact decreased with increasing FCS and HENI for most scenarios; however, at S10 both nutrition scores and certain environmental impacts rose due to higher fruits/vegetables/cereals/legumes despite zero red/processed meats. S9M3 achieved the most favorable balance: FCS 74.13, HENI 169.21 min/100 kcal, and ~55% reduction in global warming. Least sustainable scenarios by trade-offs were typically S2M2 or S2M3 (low nutrition scores with high impacts). Overall, no consistent correlation between nutrition quality and environmental emissions was confirmed.

The analysis shows the current diet (S1) is nutritionally inferior due to high proportions of red/processed meats and sugar-sweetened beverages and low amounts of fruits, vegetables, whole grains, legumes, nuts, and seafood. Incremental introduction of healthier plant-based components improves Food Compass scores, while HENI rankings are sensitive to calorie density. Replacing animal meats with legumes reduces vitamin and mineral contributions to FCS, consistent with potential micronutrient shortfalls in plant-based substitutes (e.g., calcium, vitamin D, B12, iron). Environmentally, small initial shifts toward healthier foods can increase impacts if not accompanied by sufficient meat reduction, highlighting nonlinearity and the need for carefully modeled substitutions. Larger partial replacement of animal meats with legumes (approximately 76% replacement of animal-based meats in S9M3) delivers substantial environmental gains with high nutritional quality. The absence of a consistent relationship between nutritional quality and environmental emissions underscores that dietary recommendations must consider both domains and quantities consumed.

The study presents an easy-to-use composite diet modeling approach that captures non-linear trade-offs between nutrition and environmental impacts during plant-based dietary transitions. Optimal sustainable outcomes require targeted substitution strategies rather than simply increasing plant foods or decreasing animal foods. A composite diet with 10% legumes, 0.11% red meat, 0.28% processed meat, and 2.81% white meat (S9M3) offers a practical optimum, cutting short-term global warming by about 55% while attaining strong nutritional quality (FCS 74.13; HENI ~169 min/100 kcal). Findings reaffirm that nutritional quality and environmental impact are not inherently aligned; both must be jointly optimized. Future research should refine composite diet modeling, incorporate broader food sources and processing states, and explore real-world cooking effects to guide policy and consumer decisions.

- Environmental impacts were derived from secondary midpoint data (WWEIA foods), which may limit generalizability beyond U.S. dietary patterns and representative commodities. - Phytochemical and carotenoid data were unavailable; the Food Compass scoring excluded these and other attributes (e.g., processing), and bioavailability was not considered. - Cooking and storage losses were not modeled; diet scenarios used uncooked ingredients, and household/commercial cooking variability was not simulated. - Regional differences and alternative plant-based options may require model adaptation for other contexts.