The human cost of current and recommended diets in the U.S.

The human cost of current and recommended diets in the U.S. | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The human cost of current and recommended diets in the U.S. Jessica Decker Sparks, Edgar Rodríguez-Huerta, Brooke Bell, Kyra Battaglia, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4999594/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 08 Oct, 2025 Read the published version in Nature Food → Version 1 posted You are reading this latest preprint version Abstract Modeling research on sustainable diets has primarily focused on human and planetary health, while neglecting the social dimension of sustainability, despite the agricultural and fishing sector’s significant global employment and high forced labor rates. To address this gap, our prior work developed a forced labor risk scoring method and applied it to the U.S. food supply. Expanding on this, we assess forced labor risk in current U.S. diets, three U.S.-specific recommended dietary patterns, and the EAT-Lancet Planetary Health Diet. We find that the forced labor risk is highest in the Mediterranean-Style and the U.S.-Style recommended patterns and is lowest in the Planetary Health Diet pattern, with the biggest differences driven by intake of fruit, dairy, and red meat. These results highlight synergies and tradeoffs between human health, environmental sustainability, and social well-being that should be considered in dialogue and action on sustainable diets. Earth and environmental sciences/Environmental social sciences/Sustainability Scientific community and society/Agriculture Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction A great transformation of food systems is required to sustainably and equitably meet food needs into the future. Integral to this transformation is dietary change 1 . While many have assessed the combined dangers posed by sub-optimal diets for human health and environmental sustainability 1–3 , and their economic implications 4–9 , research on the social sustainability of diets remains scarce 10,11 . Specifically, no modeling studies have analyzed the labor conditions embedded upstream in different dietary patterns, leaving the implications of current consumption patterns, recommended diets, and dietary transitions on food systems workers largely unknown. This is a yawning gap, given that agri-food supply chains employ 1.23 billion people globally 12 . Truly sustainable diets cannot be actualized without eliminating forced labor in food supply chains. Forced labor is defined by the International Labor Organization as “all work or service which is exacted from any person under the threat of penalty” which can include violence or intimidation, debt, retention of identity documents, or threats 13,14 . Though the prevalence of forced labor has not been estimated for full agrifood supply chains, which encompass multiple sectors, the agriculture, forestry and fishing sector has one of the highest rates of forced labor 15 . Numerous scholars have documented exploitative labor conditions in the U.S. food system 16–19 . Our prior work focused on assessing risk of forced labor in fruits and vegetables 20 and the land-based U.S. food supply 21 , including domestic production and trade. We found the majority of forced labor risk in the U.S. food supply was domestically sourced and stemmed from a small number of food groups 22 . In this assessment, we expand on our previous work by incorporating the U.S. seafood supply and livestock feed to our datasets on forced labor risk. This approach gives a more holistic picture of risk that encompasses all major food groups, including feed, food production, and food processing. We map the risk of current U.S. consumption using nationally-representative food intake data from the National Health and Nutrition Examination Survey (NHANES) and compare this to the risk embedded in recommended dietary patterns from the EAT-Lancet Commission (Planetary Health Diet) and U.S. government (Healthy U.S.-Style, Healthy Mediterranean, and Healthy Vegetarian patterns in the 2020-2025 Dietary Guidelines for Americans). Risk is quantified in the unit medium risk hours-equivalent (mrh-eq), where a score of one equals an hour worked at “medium” risk of forced labor 20–22 . Scores include qualitative risk levels (e.g., low, medium) for each commodity and country of origin that have been quantitatively characterized, (e.g., low = 0.01, medium = 1 mrh-eq), and then multiplied by corresponding labor intensity values 20,22 . Results The forced labor risk scores for over 200 food commodities were used to calculate weighted averages (see Methods ) for six food groups and 18 food subgroups corresponding to current and recommended diets for the U.S. ( Table 1 ). Among the dietary patterns, two of the recommended diets had higher forced labor risk than the current U.S. (CURRENT) dietary pattern (0.610 mrh-eq/capita/day): the Healthy Mediterranean-Style (MED) dietary pattern (0.824 mrh-eq/capita/day) and the Healthy U.S.-Style (HUS) dietary pattern (0.773 mrh-eq/capita/day). Two of the recommended diets had lower forced labor risk than CURRENT: the Healthy Vegetarian (VEG) dietary pattern (0.568 mrh-eq/capita/day) and the Planetary Health Diet (PHD) (0.546 mrh-eq/capita/day) ( Table 2; Figure 1a ). For the MED, HUS, and CURRENT patterns, the protein, dairy, and fruit food groups were major drivers of risk ( Figure 1a ). In the MED pattern, protein foods were responsible for 43.1% of total risk (0.355 mrh-eq) ( Figure 1b ). The MED pattern has the same recommended amounts for protein foods as the HUS pattern, plus an additional 29 grams of seafood per 2000 kcal per day; it compensates for this increased intake in protein elsewhere in the pattern (i.e., less dairy). Indeed, in the MED pattern, seafood stood out among protein foods, contributing the greatest proportion of total pattern risk among all food subgroups at 18.9% ( Figure 1b ). Contrastingly, in the CURRENT pattern, red meat had the highest food subgroup contribution at 27.1%. The MED pattern also included the greatest fruit intake of all patterns analyzed, which led to the highest absolute risk for fruit among all patterns (0.196 mrh-eq) and a substantial fraction of MED pattern risk at 23.7%. Fruit was also a top contributor to risk for the VEG pattern, reflecting high risk in fruit but also the higher serving size amounts recommended in all of the Dietary Guidelines for Americans patterns, compared to the PHD and current consumption ( Table 1 ). In the HUS pattern, the dairy food subgroup was the highest contributor to overall risk at 23.7%. The dairy subgroup was also the top contributor to the VEG pattern at 32.3% and a notable contributor to risk in the PHD at 15.0%, despite a much lower recommendation. For the PHD, the protein food group was the leading contributor to risk at a combined 42.6%. In addition to protein foods, the vegetables food group was a major contributor to risk for the PHD and VEG and PHD patterns, at 14.3% and 13.0%, respectively. Across all patterns, the grains food group made the smallest contribution to total forced labor risk (0.022 mrh-eq/capita/day for CURRENT, HUS, and MED; 0.024 mrh-eq/capita/day for VEG; and 0.029 mrh-eq/capita/day for PHD) ( Figure 1a ). Added fats and sugar (AFS) also made a minor contribution to risk across all patterns, with the exception of the CURRENT diet (0.095 mrh-eq) at 15.6% of the total risk of forced labor in that pattern. The protein food group accounted for nearly half of the risk in all patterns, except for VEG. Figure 2 shows the intake distribution in grams of the six protein subgroups (eggs, poultry, red meat, seafood, nuts and seeds, and legumes) compared to the total forced labor risk distribution of the six protein subgroups for all five dietary patterns. Comparing the relative amounts consumed (or recommended) against risk allows us to examine where risk is disproportionately high in certain patterns. This highlights what is driving the resulting risk: amounts consumed or recommended, high embedded risk, or both. In the PHD and VEG patterns, the nuts and seeds forced labor risk contribution is around two to five times larger than the nuts and seeds intake contribution, indicating that the per unit forced labor risk of nuts and seeds is driving that risk hotspot ( Figure 2 ). For the CURRENT, HUS, MED, and PHD patterns, the red meat risk contribution is over 1.5 times greater than red meat intake contribution, indicating disproportionately high forced labor risk compared to intake, but less stark than that of nuts and seeds. Commodities Driving Risk in the Food Subgroups Figure 3 shows the percentage contribution to risk from the commodities included in each food subgroup. The consumption-weighted average scores for the 18 food subgroups are shown in the final column of Table 1 . Total risk for each food subgroup is a function of commodity-level consumption, inedible amount, wasted amount, and risk level, and can be primarily driven by one—or multiple—of these variables. Supplementary Figure 1 shows the distribution of NHANES participants’ daily food commodity intake by food subgroup. Comparing the values in Figure 3 and Supplementary Figure 1 allows us to examine what factor or factors are primarily driving risk in each food subgroup. For seven of the 18 (38.9%) food subgroups, the forced labor risk from only one commodity contributed to more than half of the subgroup-level risk. For example, asparagus contributed to 54.6% of the subgroup-level dark green vegetable risk, despite accounting for only 6.8% of intake ( Supplementary Figure 1 ). Similarly, cashews contributed to 73.2% of the total risk in nuts and seeds but only 10.8% of intake ( Supplementary Figure 1 ). Other food subgroups had a more uniform distribution of risk from commodities but did show hotspots. Subgroup-level risk for whole fruit and other vegetables did not have a commodity that contributed to more than one quarter (25.0%) of the subgroup-level risk. However, avocados represented only 4.0% of whole fruit intake, and contributed 22.1% of whole fruit’s total risk. Sensitivity Analyses To assess the robustness of our overall results, we performed several sensitivity analyses and assessed whether the relationships between the patterns changed using ranks. The total amount of forced labor risk for the five patterns was ranked from 1 (lowest total risk) to 5 (highest total risk). In our main analyses (i.e., the baseline scenario), the PHD pattern had the lowest total forced labor risk (0.546 mrh-eq/capita/day) and was assigned Rank 1, and the MED pattern had the highest total forced labor risk (0.824 mrh-eq/capita/day) and was assigned Rank 5 ( Figure 4 ). Because commodity risk scores vary widely within food subgroups, we replaced the weighted average subgroup-level risk scores with the lowest and highest corresponding commodity-level risk scores ( Supplementary Table 1) , rerunning the original analysis, and recalculating the ranks (see Methods ). Overall, approximately 23 of the total 36 scenarios (63.9%) resulted in the same pattern ranking as the baseline scenario. There were no scenarios where the rank for all five patterns changed. Additional sensitivity analyses were run with and without risk in feed incorporated in the patterns; see Supplementary Information . Discussion We presented, for the first time, an estimation of the risk of forced labor embedded in dietary patterns. Focusing on the United States, we found that healthy diets could have higher or lower risk of forced labor compared to current consumption, depending on how those healthy diets are operationalized. Notably, two of the three patterns included in the 2020-2025 Dietary Guidelines for Americans—the Healthy Mediterranean-Style (MED) and Healthy U.S.-Style (HUS) pattern—had higher risk of forced labor than current U.S. intake, findings which were robust even when the risk embedded in animal feed was removed entirely from the analysis. While attention has been drawn previously to the potential environmental impacts of these patterns 23 , here we highlight potential social consequences of healthy diets, focusing on extreme forms of labor exploitation. It is important to underscore that the MED pattern in the Dietary Guidelines for Americans likely diverges from other Mediterranean diet archetypes, where meat, poultry, eggs, and dairy are de-emphasized relative to seafood 24 . The PHD, by contrast, had the lowest risk at baseline and in the majority of sensitivity analyses. This pattern was developed as a global archetype to promote human health within several planetary boundaries 1 . This pattern may present a win-win-win opportunity for health, ecosystems, and labor in the U.S. context, potentially reducing risk of forced labor relative to current consumption. At the same time, there were examples of increased forced labor risk relative to current and recommended consumption in the sensitivity analyses. For instance, changing the nuts and seeds risk score to the highest risk commodity—shelled cashews—resulted in forced labor risk much greater than current intake and all other recommended patterns. This underscores the importance of food choice in the context of healthy diets, and more importantly, the imperative to reduce risk upstream in the supply chains that bring healthy foods to the table. Our analysis focused on U.S. food consumption, which is underpinned by complex food supply chains that rely on domestic production and imports 22 . The forced labor risk embedded in these commodities, food groups, and diets would not be the same for other countries; the magnitude and distribution of forced labor risk in other countries’ food supplies is to date unknown and an important area of future research. Additionally, our data are cross-sectional and give an assessment of risk at a point in time, whereas dietary changes are likely to occur over long periods of time. Longitudinal data monitoring systems are needed to continuously assess evolving and shifting risk and working conditions, as well as knock-on effects that may occur. For instance, the social impacts of increased avocado production in Mexico for global consumption (characterized in our work as “very high risk’’) 22 have been widely documented 25–29 . Moreover, Magrach & Sanz (2020) exposed environmental and social consequences of increased demand for 'superfoods', such as cacao, coconuts, avocado, quinoa, almonds, and açai, which have led to changes from traditional production methods to monoculture, affecting the livelihoods of local communities 29 . Macro-economic benefits are sometimes coupled with negative social consequences such as increased inequity, the growing involvement of criminal organizations, and the use of forced labor in farming 26 . Understanding the multi-factorial social implications warrants further study to inform food systems transformation efforts. Our analysis is not without limitations. Part of forced labor risk estimation relies on secondary data, which is based on assumptions to fill missing data, increasing uncertainties in the results; these uncertainties have been exposed in data quality assessment by Blackstone et al. (2023) 22 and should be considered when interpreting the results presented here. Likewise, despite our efforts to map the global feed supply chain through multiple datasets, the complexity and lack of data necessitated a streamlined approach (see Methods ), which generated uncertainties in the analysis. Similarly, incorporating risk in seafood in this analysis marks a significant advancement, which was made possible by overcoming previous data limitations (see Methods ). However, the absence of granular data on gear type, which is directly related to working hours and the forced labor indicator of excessive overtime, means seafood risk scores should be interpreted with caution. Finally, while using NHANES to estimate the CURRENT pattern means the most representative data on U.S. food intake available were used, this also led to a limitation. The dataset available to map forced labor risk scores to NHANES, the Food Commodity Intake Database 30 , separates complex foods into its constituent basic commodities (e.g., dairy products are separated into milk fat, milk non-fat solids, and milk water). Thus, risk embedded in processing was sometimes excluded. This underestimation is likely small, however, as we previously found, 85% of risk in the U.S. food supply is attributable to agriculture 22 . Typically, modeling focused on the sustainability implications of dietary patterns point to the promise of shifting country-level food-based dietary guidelines to reduce impacts (i.e., by recommending less meat intake) 23,31 . In the U.S., such changes have proved challenging to date. Political will aside, for the phenomenon of forced labor, changing recommendations for food groups and subgroups will not solve the underlying structural and governance problems that perpetuate forced labor and other forms of labor exploitation in food supply chains. One promising area of demand-side solutions lies in changing public and institutional food procurement policies. The Dietary Guidelines for Americans shape federal procurement and feeding programs, the largest examples of which are the National School Lunch and Breakfast programs. By law, the nutrition standards outlined in the Guidelines need to be upheld in these programs to promote healthy lifestyles amongst school-age children (i.e. limits to added sugar and sodium in foods provided, availability of fat-free milk, frequency of whole grains served throughout the week) 32,33 . Though nascent, there is also movement towards integrating environmental considerations in public procurement. For example, 16 cities globally, including New York City and Los Angeles, have committed to adopting the PHD in their food policies, public procurement, and school meal programs 34 . While this analysis suggests that the PHD may also mitigate some forced labor risks, public and institutional procurement should also require proactive efforts to identify, mitigate, remedy, and ultimately eliminate and prevent a range of labor and human rights abuses, in food supply chains 35 . These steps would also prepare entities to align their procurement practices in advance with globally proliferating legally and financially binding human rights due diligence directives, which are expected to impact more than 10,000 U.S. businesses – a number that is likely to continue to increase 36 . However, to do so will require companies and institutions to have meaningful, proactive, continuous, and direct worker engagement throughout their supply chains 37 . This could be done by moving beyond respecting the right to freedom of association to creating an enabling environment for unionization efforts or engagement with other evidence-based legally binding worker-driven solutions, such as the worker-driven social responsibility model 38 . At the same time, a critical aspect of procurement and intervention policy is cost-effectiveness; future research is needed to understand the economic implications of such programs, alongside the labor, environmental and health implications. In the past several years, there has been tremendous momentum in developing evidence to support transitions toward healthy diets from sustainable food systems. We have, for the first time, quantified the human cost of bringing these diets to the table; it is steep indeed. Eliminating forced labor in food supply chains must be a starting point, but it cannot be the end. Ensuring decent work for and in collaboration with the “hands that feed us” is necessary to achieve truly sustainable diets 39,40 . Methods Study Overview This cross-sectional study quantitatively assessed the risk of forced labor embedded in (i) current U.S. diets, (ii) three dietary patterns recommended by the Dietary Guidelines for Americans, and (iii) the Planetary Health Diet recommended by the EAT-Lancet Commission. All data were managed and analyzed in R (v.4.4.0), Microsoft Excel (v.16.83), TableauPrep (v.2024.1), and TableauDesktop (v.2023.2.0). Data Risk of Forced Labor Forced labor risk per ton of food product for 147 food products in the U.S. land-based food supply was retrieved from Blackstone et al. (2023) 22 . The risk scores were calculated as a function of characterized risk and worker hours. In summary, we integrated several datasets (supply, prices, characterization risk, and working hours) to estimate the risk associated with each commodity-country, multiplying the characterization risk of forced labor by labor intensity and the supply share at the country level (imported or domestically produced). For the characterization risk process, data for Step 1 (commodity-country risk) and Step 2 (sector-country risk) were updated with new governmental sources 41–43 for country-commodity and country-sector risks using the 2023 report following the protocols established in Blackstone et al. (2023) 22 . Additionally, we applied the same methodology described above to calculate forced labor risk scores for 48 food products in the U.S. sea-based food supply (i.e., seafood), except we used food balance sheets of fish and fishery products as the main data source for estimating the U.S. supply via FishStatJ software (Global Fish Trade Statistics v.2022.1.0). We also incorporated livestock and seafood feed data to more accurately represent the embedded risk for animal products, including cow’s milk, chicken eggs, sheep meat, cattle meat, chicken meat, pig meat, and aquaculture. For livestock, we collected feed requirements from GLEAM, including feed commodities and feed conversion rates (FCR) by animal, region and system (i.e., feedlot, grassland-based). Additional data processing was necessary to generate risk scores for feed items that were not in our original risk database (e.g., byproducts) (see Supplementary Information ). Next, we assigned the risk of forced labor to each feed item and multiplied it by the amount required to obtain one unit of animal product. Forced labor risk scores were obtained from Blackstone et al. (2023) 22 considering a global average risk for feed coming from outside the U.S., and U.S. forced labor risk for domestic production. For aquaculture, we integrated feed requirements from multiple sources 44–46 , standardized each feed item into primary commodities weights, and assigned forced labor risk similar to the livestock method ( Supplementary Tables 2.1 and 2.2 ). Lastly, the additional risk attributable to animal feed was added to the original risk scores for the 58 animal products and byproducts to create new scores that incorporate the risk from both the food product and their corresponding animal feed. A detailed description of the methodology used to calculate the forced labor risk scores is provided in the Supplementary Information . The final scores utilized in the analysis are available in the final column of Table 1 . Current and Recommended Dietary Patterns Dietary intake of 18 food subgroups[1] (i.e., the CURRENT pattern) was estimated using nationally representative data from two recent waves of the National Health and Nutrition Examination Survey (NHANES) (2015-2016 and 2017-2018) 47,48 , accounting for complex survey design and sampling weights to be representative of the U.S. population aged 20 years or older. Per capita daily average intake was estimated by averaging up to two days of 24-hour dietary recalls from each participant, and intake was adjusted for energy intake using the residual method to reduce measurement error. Four recommended dietary patterns at the 2000 kcal/d level were selected to compare to the current U.S. adult dietary pattern. These include three patterns from the 2020–2025 Dietary Guidelines for Americans: the Healthy U.S.-Style Pattern (HUS), the Healthy Vegetarian Pattern (VEG), and the Healthy Mediterranean-Style Pattern (MED) 49 . The development of the three DGA patterns were informed by food pattern modeling and evidence on associations between dietary patterns and health outcomes 50 . The Planetary Health Diet (PHD), a global reference diet developed by the EAT-Lancet Commission on Sustainable Food Systems to meet nutritional needs within environmental limits, was also included 1 . The PHD pattern provides intake values for a 2500 kcal/d pattern, therefore a 2000 kcal/d pattern was derived by decreasing all recommended intake values by 20%. The intake values for the CURRENT and recommended DGA patterns were converted to grams using the conversion factors published in Blackstone & Conrad (2020) 43 . The final intake values (grams per 2000 kcal) for the 18 food subgroups across the five patterns are shown in Table 1 . Food Subgroups The food groups and subgroups from the DGA patterns were modified in order to reconcile differences with the CURRENT and/or PHD patterns. First, the DGA patterns provide one recommended value for each of the following categories: (i) meats, poultry, and eggs, (ii) nuts, seeds, and soy products, and (iii) whole fruit and 100% fruit juice, whereas the other patterns provide separate values for each of these food items. Therefore, the DGA recommended values were disaggregated into the more granular food items to enable comparison across all patterns. We used the NHANES 2015-2018 data to calculate the intake distribution of these individual food items and then applied that proportion to the aggregated values to obtain individual recommended values. Second, in all three DGA patterns, legumes (i.e., beans, peas, lentils) are classified as a vegetable subgroup, while the VEG pattern has an additional legume recommended value as a protein subgroup as well. To reconcile this, we created an overall legumes food subgroup that combined the vegetable and protein legumes values for the DGA patterns. Additionally, the three DGA patterns provide a calorie limit for 'Other' uses, such as added sugars, saturated fats, and alcohol. For this analysis, we assume these 'Other' calories are allocated to added sugars and saturated fats equally. For example, in the 2000 kcal HUS pattern, the 'Other' calories are capped at 240 kcal, so we attribute 120 kcal (equivalent to 7.14 tsp) to added sugars and 120 kcal (equivalent to 13.33 g) to saturated fats in our calculations. Lastly, we constructed a food group called ‘Added Fats and Sugars (AFS)’ that combines the recommended values for all added fats, including unsaturated and saturated, and added sugars across all patterns to enable a consistent comparison when evaluating the forced labor risk. Food Waste and Inedible Portions The values presented for the current and recommended dietary patterns in Table 1 only refer to the amounts of food that are consumed, and do not include the inedible and wasted portions associated with consumed food. However, the forced labor risk scores correspond to both consumed and inedible portions of food, therefore, we needed to adjust the consumption values for all five patterns to additionally incorporate inedible amounts of food. Moreover, we wanted to additionally calculate the risk for food that is wasted at the consumer level, so we also needed to adjust the consumption values to incorporate wasted amounts of food as well. To do this, we utilized the same methodology used in Conrad et al., 2021 51 and the corresponding food waste and inedible coefficients from that study. In summary, we first broke down each NHANES dish into its individual ingredients using the Food Commodity Intake Database 52 which contains data on the weight of nearly 500 ingredients in each NHANES dish. We then used the Conrad dataset to assign wasted and inedible coefficients to each ingredient, which allowed us to calculate the total amounts of wasted and inedible food for each dish. Next, we used our data crosswalk from FCID (i.e., ingredient) codes to 18 distinct food subgroups to determine the total amounts of consumed, inedible, and wasted food, by food subgroup, for each NHANES participant. Finally, we calculated the average amounts of consumed, inedible, and wasted food, by food subgroup, accounting for the complex survey design of NHANES. The following equations were used to calculate the food subgroup-level wasted and inedible coefficients: Wasted Food Coefficient = Wasted Food Amount / Consumed Food Amount Inedible Food Coefficient = Inedible Food Amount / Consumed Food Amount These coefficients were then applied to the consumed food amounts in each dietary pattern to estimate the total amounts of consumed, wasted, and inedible food ( Supplementary Table 3 ). Data Processing Data Mappings The following data mappings (i.e., crosswalks) were manually constructed by the research team in order to connect the datasets described above: 1. Mapping from FCID commodity codes to food subgroups The FCID commodity codes, which represent food commodities rather than foods as consumed (e.g., wheat flour and whole egg vs. noodles), were assigned to 18 food subgroups based on the What We Eat in America (WWEIA) Food Categories 2017-2018 53 . After reviewing these classifications, the research team decided to reclassify 3.9% (16/410) of the FCID codes (not including baby food and water) to food subgroups that we deemed were a better fit. The final mapping is provided in Supplementary Table 4. 2. Mapping from FNDDS food codes to food subgroups Similarly, the FNDDS food codes, which uniquely identify each food or beverage item in FNDDS, were assigned to 18 food subgroups based on the food classification scheme provided by the first two to four digits of the FNDDS food code 54 and the four USDA food categories 55 . The mapping is provided in Supplementary Tables 5.1 and 5.2 . 3. Mapping from FCID codes to forced labor risk scores (land-based) The risk scores for 147 land-based food products ( Supplementary Tables 6.1 and 6.2 ) were manually mapped to the relevant 401 FCID ingredients in the NHANES data by two members of the research team independently, and any disagreements in mappings were resolved by a third member. This protocol is further described in Supplementary Table 7 . Additionally, weight conversion factors were used to adjust from the weight basis as defined by the FAO and the weight basis utilized by FCID. Weight conversion factors were retrieved from a number of sources, including (i) USDA’s Food Intakes Converted to Retail Commodities Database (FICRCD), 2007-2008 56 , (ii) FAO’s Technical Conversion Factors 57 , (iii) USDA’s Conversion Factors and Weights and Measures for Agricultural Commodities and Their Products 58 , and (iv) USDA’s National Nutrient Database for Standard Reference (NDSR), Legacy Release 59 . Conversion factors were selected following a detailed protocol, provided in Supplementary Table 8. 4. Mapping from FNDDS codes to forced labor risk scores (sea-based) For seafood, an adapted approach was taken. There were only six seafood-related FCID commodities available for mapping: freshwater finfish, freshwater finfish (farm raised), saltwater finfish (tuna), saltwater finfish (other), shellfish (crustacean), and shellfish (mollusk); our new dataset, however, included forced labor risk scores for 48 seafood products. Rather than using a weighted average approach to represent each of the six FCID seafood commodities, following the same mapping process described above, two team members matched the 22 condensed seafood risk scores ( Supplementary Table 6.3 ) to seafood dishes in NHANES (at the FNDDS-level) based on the dish descriptions (which typically described the type of seafood consumed, e.g., salmon, catfish, oysters, etc.). If the FNDDS dish description did not specify the seafood product (e.g., fish sandwich, seafood dip), then we created a weighted average risk score for these unspecific seafood commodities based on the global production volumes of all seafood products 60 . The detailed protocol for this mapping process is available in Supplementary Table 9 . In our analysis, we also utilized a crosswalk from FCID codes to FNDDS codes published by Conrad et al. (2022) 30 . Impact Factors First, the forced labor risk scores for each food commodity in the U.S. food supply were obtained from prior work, which developed a method to assess the risk of forced labor in food value chains 22 . These scores, originally in the unit of medium risk hours-equivalent (mrh-eq) per ton of food produced, were divided by 1,000,000 to convert to mrh-eq per gram of food produced. We also created average risk scores for tropical fruit, seeds, grains, beans, flours, and poultry by averaging the risk scores of their related food commodities. These average scores were applied to FCID commodities that did not have an available match in the forced labor risk dataset. Then, we multiplied the risk scores by the corresponding consumed amount in the NHANES dataset to get the total forced labor risk per food item, per day, per NHANES participant. If participants had two days of diet recall, then Day 1’s and Day 2’s amounts of forced labor risk and consumed and inedible amounts were used to get a daily average. If participants had 1 day of diet recall, then Day 1’s values were used as the daily average. Lastly, accounting for NHANES sampling weights and survey design parameters, we calculated (i) average daily amount of forced labor risk, by food subgroup, and (ii) average daily consumed and inedible amount, by food subgroup. The final forced labor risk impact factors for each food subgroup were calculated: Forced labor risk per 1 gram of food(i) = Average daily amount of forced labor risk(i) / Average daily consumed and inedible amounts(i) for food subgroup i ( Table 1 ). Statistical Analysis The total forced labor risk for each pattern was calculated by multiplying the food subgroup-specific risk impact factors by the corresponding intake amounts and summing these values across all food subgroups. For each dietary pattern, the percent contribution of each food subgroup was calculated as the ratio of its forced labor risk to the total forced labor risk for that dietary pattern. We also calculated each protein food subgroup’s percentage contribution to (i) total protein intake and (ii) total forced labor risk of the overall dietary pattern. This allowed quantifying the relative impact of different protein sources on the forced labor risk of each dietary pattern. Sensitivity Analyses Two different sensitivity analyses were conducted to assess the robustness of the main results. The first focused on the risk scores for each of the 18 food subgroups, replacing the baseline scores with the minimum and maximum FCID-level scores in separate scenarios to understand how that would impact the overall diet rankings in terms of total risk. For example, the food commodities with the lowest and highest risk scores contributing to the whole fruit subgroup’s weighted score (0.016 mrh-eq/100g) were pomelo (0.004 mrh-eq/100g) and avocado (0.156 mrh-eq), respectively. In the minimum scenario, the pomelo score was used in replacement of the whole fruit subgroup’s score. Likewise, in the maximum scenario, the avocado was used. The second sensitivity analysis evaluated the impact of including or excluding the feed-related risk scores for animal-based food subgroups, comparing the baseline scenario that included feed scores to a scenario that did not. [1] Dark green vegetables, red and orange vegetables, starchy vegetables, other vegetables, whole fruits (excluding juice), 100% fruit juice, whole grains, refined grains, dairy, eggs, poultry, red meat, seafood, nuts and seeds, legumes, unsaturated fat (oil), saturated fat, and added sugar Abbreviations Dietary Guidelines for Americans (DGA); Food and Agriculture Organization (FAO); Food and Nutrient Database for Dietary Studies (FNDDS), Food Commodity Intake Database (FCID); Healthy Mediterranean-Style (MED); Healthy U.S.-Style (HUS); Healthy Vegetarian (VEG); Medium risk hours-equivalent (mrh-eq); National Health and Nutrition Examination Survey (NHANES); Planetary Health Diet (PHD); United States Department of Agriculture (USDA); What We Eat in America (WWEIA) Declarations Data Availability The detailed results and background data files are available for download at [insert permanent DOI from Dataverse upon manuscript acceptance]. Code Availability Data processing and analysis were performed using R (v.4.4.0), Microsoft Excel (v.16.83), TableauPrep (v.2024.1), and TableauDesktop (v.2023.2.0). R scripts and TFL (Tableau Prep) files are available from the corresponding author upon reasonable request. Acknowledgements This research was supported by the Interdisciplinary Research Innovation Fund (RAFINS) at the Friedman School of Nutrition Science and Policy at Tufts University. We gratefully acknowledge Becket Harney for administrative support provided during the development of this manuscript. Author Contributions Author contributions are listed alphabetically within each category. JLDS and NTB conceptualized the study. BMB, BJ, ERH, JLDS, and NTB designed the methodology. CBN and ZC provided expert input on the methodology. BMB and ERH developed the code. ASM, BMB, BJ, ERH, JLDS, and KB collected and analyzed the data. BMB, BJ, ERH, JLDS and NTB wrote the original draft. All authors edited, reviewed, and approved the final manuscript. JLDS, BJ, and NTB supervised and administered the project. Both BMB and ERH contributed equally and have the right to list their name first on their CV. Both BJ and NTB contributed equally and have the right to list their name last on their CV. Tables Tables 1 to 2 are available in the Supplementary Files section Additional Declarations Yes there is potential Competing Interest. C.B.N. declares that she was a Research Scientist in Social Responsibility with Amazon, Inc. for part of the time this research was in progress and began a role with Target as Senior Social Sustainability Manager shortly after this project began. C.B.N. is also co-owner of NewEarth B and the Social Hotspots Database project. Data from the Social Hotspots Database were provided free of charge for academic use in this research. The remaining authors declare no competing interests. Supplementary Files SupplementaryTables08262024.xlsx Supplementary Tables SUPPLEMENTRiskofforcedlaborinU.S.diets08262024.docx Supplementary Information Tables.docx Cite Share Download PDF Status: Published Journal Publication published 08 Oct, 2025 Read the published version in Nature Food → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4999594","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":359277455,"identity":"b3d7e4d3-5258-439a-8313-796340399f1c","order_by":0,"name":"Jessica Decker Sparks","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABB0lEQVRIiWNgGAWjYBACxgYgkYDg2DAwMB8AsZmJ1pImwcCWjF8Luv7DhLUwt7c//vCgxo5Bfnbzs48z/pyvMzjGf/gDQ4V1YgMug3vOmEkkHEtmMLhzzHjmxrbbEgbHmNkkGM6k49YyI4eNIbGBmcFAIsGY8WEDUMv9ZjYGxrbDuLXMf/74Q2JDPYP8jPTPjA/+nAPZwvyB8R8eLTOA5ic2HGZguJFjzLiB7QBIC4MEYwMeLT05IL8c5zG4c6aYcWZbsuTMY8wgkXRjXFoM248//vijplpOfnb7ZsaeP3b8fMcYH3/4UGMti1MLVIKHQQJZOAGHchCQh7Mk8KgaBaNgFIyCkQ0ArGlatE4dwjQAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0003-0123-0310","institution":"Friedman School of Nutrition Science and Policy, Tufts University","correspondingAuthor":true,"prefix":"","firstName":"Jessica","middleName":"Decker","lastName":"Sparks","suffix":""},{"id":359277456,"identity":"9f52a1f0-1314-4314-9f59-de95a09831c3","order_by":1,"name":"Edgar Rodríguez-Huerta","email":"","orcid":"https://orcid.org/0000-0002-6887-0040","institution":"Rights Lab, University of Nottingham","correspondingAuthor":false,"prefix":"","firstName":"Edgar","middleName":"","lastName":"Rodríguez-Huerta","suffix":""},{"id":359277457,"identity":"a31f265e-1778-479b-80bf-824f59e98452","order_by":2,"name":"Brooke Bell","email":"","orcid":"","institution":"Friedman School of Nutrition Science and Policy, Tufts University","correspondingAuthor":false,"prefix":"","firstName":"Brooke","middleName":"","lastName":"Bell","suffix":""},{"id":359277458,"identity":"faa8f4b3-0c9f-4711-ad8e-ff745ace405a","order_by":3,"name":"Kyra Battaglia","email":"","orcid":"https://orcid.org/0000-0001-8195-1454","institution":"Friedman School of Nutrition Science and Policy, Tufts University","correspondingAuthor":false,"prefix":"","firstName":"Kyra","middleName":"","lastName":"Battaglia","suffix":""},{"id":359277459,"identity":"e3820f7b-dbd6-46e3-bd7e-ef61646c45ea","order_by":4,"name":"Alejandra Sofia Marquez","email":"","orcid":"","institution":"Friedman School of Nutrition Science and Policy, Tufts University","correspondingAuthor":false,"prefix":"","firstName":"Alejandra","middleName":"Sofia","lastName":"Marquez","suffix":""},{"id":359277460,"identity":"61a5cf5e-3ac1-4962-bac6-989081a79a94","order_by":5,"name":"Julia Matteson","email":"","orcid":"","institution":"Friedman School of Nutrition Science and Policy, Tufts University","correspondingAuthor":false,"prefix":"","firstName":"Julia","middleName":"","lastName":"Matteson","suffix":""},{"id":359277461,"identity":"90bc9376-f0cc-47f0-b80e-cbe1cefce74f","order_by":6,"name":"Catherine Benoît Norris","email":"","orcid":"","institution":"NewEarth B","correspondingAuthor":false,"prefix":"","firstName":"Catherine","middleName":"Benoît","lastName":"Norris","suffix":""},{"id":359277462,"identity":"3bf08f78-edcd-4732-84bf-bcd3f904e640","order_by":7,"name":"Zach Conrad","email":"","orcid":"","institution":"William \u0026 Mary","correspondingAuthor":false,"prefix":"","firstName":"Zach","middleName":"","lastName":"Conrad","suffix":""},{"id":359277463,"identity":"64b08d18-6a4d-4d62-bcd6-03feb1d010da","order_by":8,"name":"Bethany Jackson","email":"","orcid":"https://orcid.org/0000-0002-9729-7634","institution":"University of Nottingham","correspondingAuthor":false,"prefix":"","firstName":"Bethany","middleName":"","lastName":"Jackson","suffix":""},{"id":359277464,"identity":"6ccd12a6-28af-4f4d-84ce-00783a2007f8","order_by":9,"name":"Nicole Tichenor Blackstone","email":"","orcid":"https://orcid.org/0000-0001-8343-3650","institution":"Friedman School of Nutrition Science and Policy, Tufts University","correspondingAuthor":false,"prefix":"","firstName":"Nicole","middleName":"Tichenor","lastName":"Blackstone","suffix":""}],"badges":[],"createdAt":"2024-08-29 18:40:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4999594/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4999594/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s43016-025-01242-8","type":"published","date":"2025-10-08T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":65437527,"identity":"132b21c2-bc27-4694-adc1-5f11b7713084","added_by":"auto","created_at":"2024-09-27 12:15:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":195122,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea.\u003c/strong\u003e Total forced labor risk in each dietary pattern (mrh-eq/capita/day), by food group\u003c/p\u003e\n\u003cp\u003eAbbreviations: Added fats and sugar (AFS)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eb.\u003c/strong\u003e Total forced labor risk in each dietary pattern (%), by food subgroup\u003c/p\u003e\n\u003cp\u003eAbbreviations: Added fats and sugar (AFS)\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/a7a845c80b25a5738aef8f2c.png"},{"id":65437478,"identity":"11793438-eb08-46a0-bba7-bac6ca53fdf6","added_by":"auto","created_at":"2024-09-27 12:15:01","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":121118,"visible":true,"origin":"","legend":"\u003cp\u003eProtein subgroup contributions to intake\u003csup\u003e1\u003c/sup\u003e (g/capita/day) and total forced labor risk (mrh-eq/capita/day) across all dietary patterns\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003csup\u003e1\u003c/sup\u003e Intake refers to both consumed amount and plate waste amount (i.e., purchased amount)\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/15d1be017f7eeb9f291ff40a.png"},{"id":65437526,"identity":"80e3df89-c122-47aa-b81e-cb716a531712","added_by":"auto","created_at":"2024-09-27 12:15:15","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":233324,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of commodity-level risk (mrh-eq/capita/day), by food subgroup\u003c/p\u003e\n\u003cp\u003eAbbreviations: Added fats and sugar (AFS)\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/43805d8e305047c6f00eaf13.png"},{"id":65437462,"identity":"3976574d-2b19-4c92-92eb-56c48a110dc1","added_by":"auto","created_at":"2024-09-27 12:15:00","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":238002,"visible":true,"origin":"","legend":"\u003cp\u003eSensitivity analyses results for the minimum and maximum scenarios\u003c/p\u003e\n\u003cp\u003eAbbreviations: Added fats and sugar (AFS)\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/81856e13b7620c3e6f6983f5.png"},{"id":93107923,"identity":"9d2aa719-328b-4230-ab5a-ca49e5b5a488","added_by":"auto","created_at":"2025-10-09 07:08:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1659222,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/b4e6537e-c599-481e-85dd-192767afb77b.pdf"},{"id":65437530,"identity":"3e2e1ab7-84e8-44a3-a563-5ccae9f9c59f","added_by":"auto","created_at":"2024-09-27 12:15:21","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":93474,"visible":true,"origin":"","legend":"Supplementary Tables","description":"","filename":"SupplementaryTables08262024.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/fba99d8ae7f74d5fb5f874a9.xlsx"},{"id":65437518,"identity":"1c46653d-cc54-4639-882e-acb8db7c47da","added_by":"auto","created_at":"2024-09-27 12:15:06","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":1012763,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SUPPLEMENTRiskofforcedlaborinU.S.diets08262024.docx","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/5ad69173c4ccaf9357fd1bc4.docx"},{"id":65437516,"identity":"83973245-2d97-4af5-87fa-8f0c648eebc1","added_by":"auto","created_at":"2024-09-27 12:15:04","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":28131,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.docx","url":"https://assets-eu.researchsquare.com/files/rs-4999594/v1/33dcf17ba43fe20e7c04fad7.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nC.B.N. declares that she was a Research Scientist in Social Responsibility with Amazon, Inc. for part of the time this research was in progress and began a role with Target as Senior Social Sustainability\r\nManager shortly after this project began. C.B.N. is also co-owner of NewEarth B and the Social Hotspots Database project. Data from the Social Hotspots Database were provided free of charge\r\nfor academic use in this research. The remaining authors declare no competing interests.","formattedTitle":"The human cost of current and recommended diets in the U.S.","fulltext":[{"header":"Introduction","content":"\u003cp\u003eA great transformation of food systems is required to sustainably and equitably meet food needs into the future. Integral to this transformation is dietary change\u003csup\u003e1\u003c/sup\u003e. While many have assessed the combined dangers posed by sub-optimal diets for human health and environmental sustainability\u003csup\u003e1–3\u003c/sup\u003e, and their economic implications\u003csup\u003e4–9\u003c/sup\u003e, research on the social sustainability of diets remains scarce\u003csup\u003e10,11\u003c/sup\u003e. Specifically, no modeling studies have analyzed the labor conditions embedded upstream in different dietary patterns, leaving the implications of current consumption patterns, recommended diets, and dietary transitions on food systems workers largely unknown. This is a yawning gap, given that agri-food supply chains employ 1.23 billion people globally\u003csup\u003e12\u003c/sup\u003e. \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTruly sustainable diets cannot be actualized without eliminating forced labor in food supply chains. Forced labor is defined by the International Labor Organization as “all work or service which is exacted from any person under the threat of penalty” which can include violence or intimidation, debt, retention of identity documents, or threats\u003csup\u003e13,14\u003c/sup\u003e. Though the prevalence of forced labor has not been estimated for full agrifood supply chains, which encompass multiple sectors, the agriculture, forestry and fishing sector has one of the highest rates of forced labor\u003csup\u003e15\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNumerous scholars have documented exploitative labor conditions in the U.S. food system\u003csup\u003e16–19\u003c/sup\u003e. Our prior work focused on assessing risk of forced labor in fruits and vegetables\u003csup\u003e20\u003c/sup\u003e and the land-based U.S. food supply\u003csup\u003e21\u003c/sup\u003e, including domestic production and trade. We found the majority of forced labor risk in the U.S. food supply was domestically sourced and stemmed from a small number of food groups\u003csup\u003e22\u003c/sup\u003e. In this assessment, we expand on our previous work by incorporating the U.S. seafood supply and livestock feed to our datasets on forced labor risk. This approach gives a more holistic picture of risk that encompasses all major food groups, including feed, food production, and food processing. We map the risk of current U.S. consumption using nationally-representative food intake data from the National Health and Nutrition Examination Survey (NHANES) and compare this to the risk embedded in recommended dietary patterns from the EAT-Lancet Commission (Planetary Health Diet) and U.S. government (Healthy U.S.-Style, Healthy Mediterranean, and Healthy Vegetarian patterns in the 2020-2025 Dietary Guidelines for Americans). Risk is quantified in the unit medium risk hours-equivalent (mrh-eq), where a score of one equals an hour worked at “medium” risk of forced labor\u003csup\u003e20–22\u003c/sup\u003e. Scores include qualitative risk levels (e.g., low, medium) for each commodity and country of origin that have been quantitatively characterized, (e.g., low = 0.01, medium = 1 mrh-eq), and then multiplied by corresponding labor intensity values\u003csup\u003e20,22\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe forced labor risk scores for over 200 food commodities were used to calculate weighted averages (see \u003cstrong\u003eMethods\u003c/strong\u003e) for six food groups and 18 food subgroups corresponding to current and recommended diets for the U.S. (\u003cstrong\u003eTable 1\u003c/strong\u003e). Among the dietary patterns, two of the recommended diets had higher forced labor risk than the current U.S. (CURRENT) dietary pattern (0.610 mrh-eq/capita/day): the Healthy Mediterranean-Style (MED) dietary pattern (0.824 mrh-eq/capita/day) and the Healthy U.S.-Style (HUS) dietary pattern (0.773 mrh-eq/capita/day). Two of the recommended diets had lower forced labor risk than CURRENT: the Healthy Vegetarian (VEG) dietary pattern (0.568 mrh-eq/capita/day) and the Planetary Health Diet (PHD) (0.546 mrh-eq/capita/day) (\u003cstrong\u003eTable 2; Figure 1a\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor the MED, HUS, and CURRENT patterns, the protein, dairy, and fruit food groups were major drivers of risk (\u003cstrong\u003eFigure 1a\u003c/strong\u003e). In the MED pattern, protein foods were responsible for 43.1% of total risk (0.355 mrh-eq) (\u003cstrong\u003eFigure 1b\u003c/strong\u003e). The MED pattern has the same recommended amounts for protein foods as the HUS pattern, plus an additional 29 grams of seafood per 2000 kcal per day; it compensates for this increased intake in protein elsewhere in the pattern (i.e., less dairy). Indeed, in the MED pattern, seafood stood out among protein foods, contributing the greatest proportion of total pattern risk among all food subgroups at 18.9% (\u003cstrong\u003eFigure 1b\u003c/strong\u003e). Contrastingly, in the CURRENT pattern, red meat had the highest food subgroup contribution at 27.1%. The MED pattern also included the greatest fruit intake of all patterns analyzed, which led to the highest absolute risk for fruit among all patterns (0.196 mrh-eq) and a substantial fraction of MED pattern risk at 23.7%. Fruit was also a top contributor to risk for the VEG pattern, reflecting high risk in fruit but also the higher serving size amounts recommended in all of the Dietary Guidelines for Americans patterns, compared to the PHD and current consumption (\u003cstrong\u003eTable 1\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the HUS pattern, the dairy food subgroup was the highest contributor to overall risk at 23.7%. The dairy subgroup was also the top contributor to the VEG pattern at 32.3% and a notable contributor to risk in the PHD at 15.0%, despite a much lower recommendation. For the PHD, the protein food group was the leading contributor to risk at a combined 42.6%. In addition to protein foods, the vegetables food group was a major contributor to risk for the PHD and VEG and PHD patterns, at 14.3% and 13.0%, respectively. Across all patterns, the grains food group made the smallest contribution to total forced labor risk (0.022 mrh-eq/capita/day for CURRENT, HUS, and MED; 0.024 mrh-eq/capita/day for VEG; and 0.029 mrh-eq/capita/day for PHD) (\u003cstrong\u003eFigure 1a\u003c/strong\u003e). Added fats and sugar (AFS) also made a minor contribution to risk across all patterns, with the exception of the CURRENT diet (0.095 mrh-eq) at 15.6% of the total risk of forced labor in that pattern.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The protein food group accounted for nearly half of the risk in all patterns, except for VEG. \u003cstrong\u003eFigure 2\u003c/strong\u003e shows the intake distribution in grams of the six protein subgroups (eggs, poultry, red meat, seafood, nuts and seeds, and legumes) compared to the total forced labor risk distribution of the six protein subgroups for all five dietary patterns. Comparing the relative amounts consumed (or recommended) against risk allows us to examine where risk is disproportionately high in certain patterns. This highlights what is driving the resulting risk: amounts consumed or recommended, high embedded risk, or both. In the PHD and VEG patterns, the nuts and seeds forced labor risk contribution is around two to five times larger than the nuts and seeds intake contribution, indicating that the per unit forced labor risk of nuts and seeds is driving that risk hotspot (\u003cstrong\u003eFigure 2\u003c/strong\u003e). For the CURRENT, HUS, MED, and PHD patterns, the red meat risk contribution is over 1.5 times greater than red meat intake contribution, indicating disproportionately high forced labor risk compared to intake, but less stark than that of nuts and seeds.\u003c/p\u003e\n\u003ch4\u003eCommodities Driving Risk in the Food Subgroups\u003c/h4\u003e\n\u003cp\u003e\u003cstrong\u003eFigure 3\u003c/strong\u003e shows the percentage contribution to risk from the commodities included in each food subgroup. The consumption-weighted average scores for the 18 food subgroups are shown in the final column of \u003cstrong\u003eTable 1\u003c/strong\u003e. Total risk for each food subgroup is a function of commodity-level consumption, inedible amount, wasted amount, and risk level, and can be primarily driven by one—or multiple—of these variables. \u003cstrong\u003eSupplementary Figure 1\u003c/strong\u003e shows the distribution of NHANES participants’ daily food commodity intake by food subgroup. Comparing the values in \u003cstrong\u003eFigure 3\u003c/strong\u003e and \u003cstrong\u003eSupplementary Figure 1\u0026nbsp;\u003c/strong\u003eallows us to examine what factor or factors are primarily driving risk in each food subgroup. For seven of the 18 (38.9%) food subgroups, the forced labor risk from only one commodity contributed to more than half of the subgroup-level risk. For example, asparagus contributed to 54.6% of the subgroup-level dark green vegetable risk, despite accounting for only 6.8% of intake (\u003cstrong\u003eSupplementary Figure 1\u003c/strong\u003e). Similarly, cashews contributed to 73.2% of the total risk in nuts and seeds but only 10.8% of intake (\u003cstrong\u003eSupplementary Figure 1\u003c/strong\u003e). Other food subgroups had a more uniform distribution of risk from commodities but did show hotspots. Subgroup-level risk for whole fruit and other vegetables did not have a commodity that contributed to more than one quarter (25.0%) of the subgroup-level risk. However, avocados represented only 4.0% of whole fruit intake, and contributed 22.1% of whole fruit’s total risk.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eSensitivity Analyses\u003c/h3\u003e\n\u003cp\u003eTo assess the robustness of our overall results, we performed several sensitivity analyses and assessed whether the relationships between the patterns changed using ranks. The total amount of forced labor risk for the five patterns was ranked from 1 (lowest total risk) to 5 (highest total risk). In our main analyses (i.e., the baseline scenario), the PHD pattern had the lowest total forced labor risk (0.546 mrh-eq/capita/day) and was assigned Rank 1, and the MED pattern had the highest total forced labor risk (0.824 mrh-eq/capita/day) and was assigned Rank 5 (\u003cstrong\u003eFigure 4\u003c/strong\u003e). Because commodity risk scores vary widely within food subgroups, we replaced the weighted average subgroup-level risk scores with the lowest and highest corresponding commodity-level risk scores (\u003cstrong\u003eSupplementary Table 1)\u003c/strong\u003e, rerunning the original analysis, and recalculating the ranks (see \u003cstrong\u003eMethods\u003c/strong\u003e). Overall, approximately 23 of the total 36 scenarios (63.9%) resulted in the same pattern ranking as the baseline scenario. There were no scenarios where the rank for all five patterns changed. Additional sensitivity analyses were run with and without risk in feed incorporated in the patterns; see \u003cstrong\u003eSupplementary Information\u003c/strong\u003e.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe presented, for the first time, an estimation of the risk of forced labor embedded in dietary patterns. Focusing on the United States, we found that healthy diets could have higher or lower risk of forced labor compared to current consumption, depending on how those healthy diets are operationalized. Notably, two of the three patterns included in the 2020-2025 Dietary Guidelines for Americans—the Healthy Mediterranean-Style (MED) and Healthy U.S.-Style (HUS) pattern—had higher risk of forced labor than current U.S. intake, findings which were robust even when the risk embedded in animal feed was removed entirely from the analysis. While attention has been drawn previously to the potential environmental impacts of these patterns\u003csup\u003e23\u003c/sup\u003e, here we highlight potential social consequences of healthy diets, focusing on extreme forms of labor exploitation. It is important to underscore that the MED pattern in the Dietary Guidelines for Americans likely diverges from other Mediterranean diet archetypes, where meat, poultry, eggs, and dairy are de-emphasized relative to seafood\u003csup\u003e24\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe PHD, by contrast, had the lowest risk at baseline and in the majority of sensitivity analyses. This pattern was developed as a global archetype to promote human health within several planetary boundaries\u003csup\u003e1\u003c/sup\u003e. This pattern may present a win-win-win opportunity for health, ecosystems, and labor in the U.S. context, potentially reducing risk of forced labor relative to current consumption. At the same time, there were examples of increased forced labor risk relative to current and recommended consumption in the sensitivity analyses. For instance, changing the nuts and seeds risk score to the highest risk commodity—shelled cashews—resulted in forced labor risk much greater than current intake and all other recommended patterns. This underscores the importance of food choice in the context of healthy diets, and more importantly, the imperative to reduce risk upstream in the supply chains that bring healthy foods to the table.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur analysis focused on U.S. food consumption, which is underpinned by complex food supply chains that rely on domestic production and imports\u003csup\u003e22\u003c/sup\u003e. \u0026nbsp;The forced labor risk embedded in these commodities, food groups, and diets would not be the same for other countries; the magnitude and distribution of forced labor risk in other countries’ food supplies is to date unknown and an important area of future research.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditionally, our data are cross-sectional and give an assessment of risk at a point in time, whereas dietary changes are likely to occur over long periods of time. Longitudinal data monitoring systems are needed to continuously assess evolving and shifting risk and working conditions, as well as knock-on effects that may occur. For instance, the social impacts of increased avocado production in Mexico for global consumption (characterized in our work as “very high risk’’)\u003csup\u003e22\u003c/sup\u003e have been widely documented\u003csup\u003e25–29\u003c/sup\u003e. Moreover, Magrach \u0026amp; Sanz (2020) exposed environmental and social consequences of increased demand for 'superfoods', such as cacao, coconuts, avocado, quinoa, almonds, and açai, which have led to changes from traditional production methods to monoculture, affecting the livelihoods of local communities\u003csup\u003e29\u003c/sup\u003e. Macro-economic benefits are sometimes coupled with negative social consequences such as increased inequity, the growing involvement of criminal organizations, and the use of forced labor in farming\u003csup\u003e26\u003c/sup\u003e.\u0026nbsp;Understanding the multi-factorial social implications warrants further study to inform food systems transformation efforts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur analysis is not without limitations. Part of forced labor risk estimation relies on secondary data, which is based on assumptions to fill missing data, increasing uncertainties in the results; these uncertainties have been exposed in data quality assessment by Blackstone et al. (2023)\u003csup\u003e22\u003c/sup\u003e and should be considered when interpreting the results presented here. Likewise, despite our efforts to map the global feed supply chain through multiple datasets, the complexity and lack of data necessitated a streamlined approach (see \u003cstrong\u003eMethods\u003c/strong\u003e), which generated uncertainties in the analysis. Similarly, incorporating risk in seafood in this analysis marks a significant advancement, which was made possible by overcoming previous data limitations (see \u003cstrong\u003eMethods\u003c/strong\u003e). However, the absence of granular data on gear type, which is directly related to working hours and the forced labor indicator of excessive overtime, means seafood risk scores should be interpreted with caution. Finally, while using NHANES to estimate the CURRENT pattern means the most representative data on U.S. food intake available were used, this also led to a limitation. The dataset available to map forced labor risk scores to NHANES, the Food Commodity Intake Database\u003csup\u003e30\u003c/sup\u003e, separates complex foods into its constituent basic commodities (e.g., dairy products are separated into milk fat, milk non-fat solids, and milk water). Thus, risk embedded in processing was sometimes excluded. This underestimation is likely small, however, as we previously found, 85% of risk in the U.S. food supply is attributable to agriculture\u003csup\u003e22\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTypically, modeling focused on the sustainability implications of dietary patterns point to the promise of shifting country-level food-based dietary guidelines to reduce impacts (i.e., by recommending less meat intake)\u003csup\u003e23,31\u003c/sup\u003e. In the U.S., such changes have proved challenging to date. Political will aside, for the phenomenon of forced labor, changing recommendations for food groups and subgroups will not solve the underlying structural and governance problems that perpetuate forced labor and other forms of labor exploitation in food supply chains.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOne promising area of demand-side solutions lies in changing public and institutional food procurement policies. The Dietary Guidelines for Americans shape federal procurement and feeding programs, the largest examples of which are the National School Lunch and Breakfast programs. By law, the nutrition standards outlined in the Guidelines need to be upheld in these programs to promote healthy lifestyles amongst school-age children (i.e. limits to added sugar and sodium in foods provided, availability of fat-free milk, frequency of whole grains served throughout the week)\u003csup\u003e32,33\u003c/sup\u003e. Though nascent, there is also movement towards integrating environmental considerations in public procurement. For example, 16 cities globally, including New York City and Los Angeles, have committed to adopting the PHD in their food policies, public procurement, and school meal programs\u003csup\u003e34\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eWhile this analysis suggests that the PHD may also mitigate some forced labor risks, public and institutional procurement should also require proactive efforts to identify, mitigate, remedy, and ultimately eliminate and prevent a range of labor and human rights abuses, in food supply chains\u003csup\u003e35\u003c/sup\u003e. These steps would also prepare entities to align their procurement practices in advance with globally proliferating legally and financially binding human rights due diligence directives, which are expected to impact more than 10,000 U.S. businesses – a number that is likely to continue to increase\u003csup\u003e36\u003c/sup\u003e. However, to do so will require companies and institutions to have meaningful, proactive, continuous, and direct worker engagement throughout their supply chains\u003csup\u003e37\u003c/sup\u003e. This could be done by moving beyond respecting the right to freedom of association to creating an enabling environment for unionization efforts or engagement with other evidence-based legally binding worker-driven solutions, such as the worker-driven social responsibility model\u003csup\u003e38\u003c/sup\u003e. At the same time, a critical aspect of procurement and intervention policy is cost-effectiveness; future research is needed to understand the economic implications of such programs, alongside the labor, environmental and health implications.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the past several years, there has been tremendous momentum in developing evidence to support transitions toward healthy diets from sustainable food systems. We have, for the first time, quantified the human cost of bringing these diets to the table; it is steep indeed. Eliminating forced labor in food supply chains must be a starting point, but it cannot be the end. Ensuring decent work for and in collaboration with the “hands that feed us” is necessary to achieve truly sustainable diets\u003csup\u003e39,40\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003ch3\u003eStudy Overview\u003c/h3\u003e\n\u003cp\u003eThis cross-sectional study quantitatively assessed the risk of forced labor embedded in (i) current U.S. diets, (ii) three dietary patterns recommended by the Dietary Guidelines for Americans, and (iii) the Planetary Health Diet recommended by the EAT-Lancet Commission. All data were managed and analyzed in R (v.4.4.0), Microsoft Excel (v.16.83), TableauPrep (v.2024.1), and TableauDesktop (v.2023.2.0).\u003c/p\u003e\n\u003ch3\u003eData\u003c/h3\u003e\n\u003ch4\u003eRisk of Forced Labor\u003c/h4\u003e\n\u003cp\u003eForced labor risk per ton of food product for 147 food products in the U.S. land-based food supply was retrieved from Blackstone et al. (2023)\u003csup\u003e22\u003c/sup\u003e. The risk scores were calculated as a function of characterized risk and worker hours. In summary, we integrated several datasets (supply, prices, characterization risk, and working hours) to estimate the risk associated with each commodity-country, multiplying the characterization risk of forced labor by labor intensity and the supply share at the country level (imported or domestically produced). For the characterization risk process, data for Step 1 (commodity-country risk) and Step 2 (sector-country risk) were updated with new governmental sources\u003csup\u003e41\u0026ndash;43\u003c/sup\u003e for country-commodity and country-sector risks using the 2023 report following the protocols established in Blackstone et al. (2023)\u003csup\u003e22\u003c/sup\u003e. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditionally, we applied the same methodology described above to calculate forced labor risk scores for 48 food products in the U.S. sea-based food supply (i.e., seafood), except we used food balance sheets of fish and fishery products as the main data source for estimating the U.S. supply via FishStatJ software (Global Fish Trade Statistics v.2022.1.0).\u003c/p\u003e\n\u003cp\u003eWe also incorporated livestock and seafood feed data to more accurately represent the embedded risk for animal products, including cow\u0026rsquo;s milk, chicken eggs, sheep meat, cattle meat, chicken meat, pig meat, and aquaculture. For livestock, we collected feed requirements from GLEAM, including feed commodities and feed conversion rates (FCR) by animal, region and system (i.e., feedlot, grassland-based). Additional data processing was necessary to generate risk scores for feed items that were not in our original risk database (e.g., byproducts) (see \u003cstrong\u003eSupplementary Information\u003c/strong\u003e). Next, we assigned the risk of forced labor to each feed item and multiplied it by the amount required to obtain one unit of animal product. Forced labor risk scores were obtained from Blackstone et al. (2023)\u003csup\u003e22\u003c/sup\u003e considering a global average risk for feed coming from outside the U.S., and U.S. forced labor risk for domestic production. For aquaculture, we integrated feed requirements from multiple sources\u003csup\u003e44\u0026ndash;46\u003c/sup\u003e, standardized each feed item into primary commodities weights, and assigned forced labor risk similar to the livestock method (\u003cstrong\u003eSupplementary Tables 2.1 and 2.2\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eLastly, the additional risk attributable to animal feed was added to the original risk scores for the 58 animal products and byproducts to create new scores that incorporate the risk from both the food product and their corresponding animal feed. A detailed description of the methodology used to calculate the forced labor risk scores is provided in the \u003cstrong\u003eSupplementary Information\u003c/strong\u003e. The final scores utilized in the analysis are available in the final column of \u003cstrong\u003eTable 1\u003c/strong\u003e.\u003c/p\u003e\n\u003ch4\u003eCurrent and Recommended Dietary Patterns\u003c/h4\u003e\n\u003cp\u003eDietary intake of 18 food subgroups[1] (i.e., the CURRENT pattern) was estimated using nationally representative data from two recent waves of the National Health and Nutrition Examination Survey (NHANES) (2015-2016 and 2017-2018)\u003csup\u003e47,48\u003c/sup\u003e, accounting for complex survey design and sampling weights to be representative of the U.S. population aged 20 years or older. Per capita daily average intake was estimated by averaging up to two days of 24-hour dietary recalls from each participant, and intake was adjusted for energy intake using the residual method to reduce measurement error.\u003c/p\u003e\n\u003cp\u003eFour recommended dietary patterns at the 2000 kcal/d level were selected to compare to the current U.S. adult dietary pattern. These include three patterns from the 2020\u0026ndash;2025 Dietary Guidelines for Americans: the Healthy U.S.-Style Pattern (HUS), the Healthy Vegetarian Pattern (VEG), and the Healthy Mediterranean-Style Pattern (MED)\u003csup\u003e49\u003c/sup\u003e. The development of the three DGA patterns were informed by food pattern modeling and evidence on associations between dietary patterns and health outcomes\u003csup\u003e50\u003c/sup\u003e. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe Planetary Health Diet (PHD), a global reference diet developed by the EAT-Lancet Commission on Sustainable Food Systems to meet nutritional needs within environmental limits, was also included\u003csup\u003e1\u003c/sup\u003e. The PHD pattern provides intake values for a 2500 kcal/d pattern, therefore a 2000 kcal/d pattern was derived by decreasing all recommended intake values by 20%.\u003c/p\u003e\n\u003cp\u003eThe intake values for the CURRENT and recommended DGA patterns were converted to grams using the conversion factors published in Blackstone \u0026amp; Conrad (2020)\u003csup\u003e43\u003c/sup\u003e. The final intake values (grams per 2000 kcal) for the 18 food subgroups across the five patterns are shown in \u003cstrong\u003eTable 1\u003c/strong\u003e.\u003c/p\u003e\n\u003ch5\u003eFood Subgroups\u003c/h5\u003e\n\u003cp\u003eThe food groups and subgroups from the DGA patterns were modified in order to reconcile differences with the CURRENT and/or PHD patterns.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFirst, the DGA patterns provide one recommended value for each of the following categories: (i) meats, poultry, and eggs, (ii) nuts, seeds, and soy products, and (iii) whole fruit and 100% fruit juice, whereas the other patterns provide separate values for each of these food items. Therefore, the DGA recommended values were disaggregated into the more granular food items to enable comparison across all patterns. We used the NHANES 2015-2018 data to calculate the intake distribution of these individual food items and then applied that proportion to the aggregated values to obtain individual recommended values.\u003c/p\u003e\n\u003cp\u003eSecond, in all three DGA patterns, legumes (i.e., beans, peas, lentils) are classified as a vegetable subgroup, while the VEG pattern has an additional legume recommended value as a protein subgroup as well. To reconcile this, we created an overall legumes food subgroup that combined the vegetable and protein legumes values for the DGA patterns.\u003c/p\u003e\n\u003cp\u003eAdditionally, the three DGA patterns provide a calorie limit for \u0026apos;Other\u0026apos; uses, such as added sugars, saturated fats, and alcohol. For this analysis, we assume these \u0026apos;Other\u0026apos; calories are allocated to added sugars and saturated fats equally. For example, in the 2000 kcal HUS pattern, the \u0026apos;Other\u0026apos; calories are capped at 240 kcal, so we attribute 120 kcal (equivalent to 7.14 tsp) to added sugars and 120 kcal (equivalent to 13.33 g) to saturated fats in our calculations.\u003c/p\u003e\n\u003cp\u003eLastly, we constructed a food group called \u0026lsquo;Added Fats and Sugars (AFS)\u0026rsquo; that combines the recommended values for all added fats, including unsaturated and saturated, and added sugars across all patterns to enable a consistent comparison when evaluating the forced labor risk.\u003c/p\u003e\n\u003ch4\u003eFood Waste and Inedible Portions\u003c/h4\u003e\n\u003cp\u003eThe values presented for the current and recommended dietary patterns in \u003cstrong\u003eTable 1\u003c/strong\u003e only refer to the amounts of food that are consumed, and do not include the inedible and wasted portions associated with consumed food. However, the forced labor risk scores correspond to both consumed and inedible portions of food, therefore, we needed to adjust the consumption values for all five patterns to additionally incorporate inedible amounts of food. Moreover, we wanted to additionally calculate the risk for food that is wasted at the consumer level, so we also needed to adjust the consumption values to incorporate wasted amounts of food as well.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo do this, we utilized the same methodology used in Conrad et al., 2021\u003csup\u003e51\u003c/sup\u003e and the corresponding food waste and inedible coefficients from that study. In summary, we first broke down each NHANES dish into its individual ingredients using the Food Commodity Intake Database\u003csup\u003e52\u003c/sup\u003e which contains data on the weight of nearly 500 ingredients in each NHANES dish. We then used the Conrad dataset to assign wasted and inedible coefficients to each ingredient, which allowed us to calculate the total amounts of wasted and inedible food for each dish. Next, we used our data crosswalk from FCID (i.e., ingredient) codes to 18 distinct food subgroups to determine the total amounts of consumed, inedible, and wasted food, by food subgroup, for each NHANES participant. Finally, we calculated the average amounts of consumed, inedible, and wasted food, by food subgroup, accounting for the complex survey design of NHANES.\u003c/p\u003e\n\u003cp\u003eThe following equations were used to calculate the food subgroup-level wasted and inedible coefficients:\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eWasted Food Coefficient = Wasted Food Amount / Consumed Food Amount\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eInedible Food Coefficient = Inedible Food Amount / Consumed Food Amount\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThese coefficients were then applied to the consumed food amounts in each dietary pattern to estimate the total amounts of consumed, wasted, and inedible food (\u003cstrong\u003eSupplementary Table 3\u003c/strong\u003e).\u003c/p\u003e\n\u003ch3\u003eData Processing\u003c/h3\u003e\n\u003ch4\u003eData Mappings\u003c/h4\u003e\n\u003cp\u003eThe following data mappings (i.e., crosswalks) were manually constructed by the research team in order to connect the datasets described above:\u003c/p\u003e\n\u003ch5\u003e1. Mapping from FCID commodity codes to food subgroups\u003c/h5\u003e\n\u003cp\u003eThe FCID commodity codes, which represent food commodities rather than foods as consumed (e.g., wheat flour and whole egg vs. noodles), were assigned to 18 food subgroups based on the What We Eat in America (WWEIA) Food Categories 2017-2018\u003csup\u003e53\u003c/sup\u003e. After reviewing these classifications, the research team decided to reclassify 3.9% (16/410) of the FCID codes (not including baby food and water) to food subgroups that we deemed were a better fit. The final mapping is provided in \u003cstrong\u003eSupplementary Table 4.\u003c/strong\u003e\u003c/p\u003e\n\u003ch5\u003e2. Mapping from FNDDS food codes to food subgroups\u003c/h5\u003e\n\u003cp\u003eSimilarly, the FNDDS food codes, which uniquely identify each food or beverage item in FNDDS, were assigned to 18 food subgroups based on the food classification scheme provided by the first two to four digits of the FNDDS food code\u003csup\u003e54\u003c/sup\u003e and the four USDA food categories\u003csup\u003e55\u003c/sup\u003e. The mapping is provided in \u003cstrong\u003eSupplementary Tables 5.1 and 5.2\u003c/strong\u003e.\u003c/p\u003e\n\u003ch5\u003e3. Mapping from FCID codes to forced labor risk scores (land-based)\u003c/h5\u003e\n\u003cp\u003eThe risk scores for 147 land-based food products (\u003cstrong\u003eSupplementary Tables 6.1 and 6.2\u003c/strong\u003e) were manually mapped to the relevant 401 FCID ingredients in the NHANES data by two members of the research team independently, and any disagreements in mappings were resolved by a third member. This protocol is further described in \u003cstrong\u003eSupplementary Table 7\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eAdditionally, weight conversion factors were used to adjust from the weight basis as defined by the FAO and the weight basis utilized by FCID. Weight conversion factors were retrieved from a number of sources, including (i) USDA\u0026rsquo;s Food Intakes Converted to Retail Commodities Database (FICRCD), 2007-2008\u003csup\u003e56\u003c/sup\u003e, (ii) FAO\u0026rsquo;s Technical Conversion Factors\u003csup\u003e57\u003c/sup\u003e, (iii) USDA\u0026rsquo;s Conversion Factors and Weights and Measures for Agricultural Commodities and Their Products\u003csup\u003e58\u003c/sup\u003e, and (iv) USDA\u0026rsquo;s National Nutrient Database for Standard Reference (NDSR), Legacy Release\u003csup\u003e59\u003c/sup\u003e. Conversion factors were selected following a detailed protocol, provided in \u003cstrong\u003eSupplementary Table 8.\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003ch5\u003e4. Mapping from FNDDS codes to forced labor risk scores (sea-based)\u003c/h5\u003e\n\u003cp\u003eFor seafood, an adapted approach was taken. There were only six seafood-related FCID commodities available for mapping: freshwater finfish, freshwater finfish (farm raised), saltwater finfish (tuna), saltwater finfish (other), shellfish (crustacean), and shellfish (mollusk); our new dataset, however, included forced labor risk scores for 48 seafood products. Rather than using a weighted average approach to represent each of the six FCID seafood commodities, following the same mapping process described above, two team members matched the 22 condensed seafood risk scores (\u003cstrong\u003eSupplementary Table 6.3\u003c/strong\u003e) to seafood dishes in NHANES (at the FNDDS-level) based on the dish descriptions (which typically described the type of seafood consumed, e.g., salmon, catfish, oysters, etc.). If the FNDDS dish description did not specify the seafood product (e.g., fish sandwich, seafood dip), then we created a weighted average risk score for these unspecific seafood commodities based on the global production volumes of all seafood products\u003csup\u003e60\u003c/sup\u003e. The detailed protocol for this mapping process is available in \u003cstrong\u003eSupplementary Table 9\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eIn our analysis, we also utilized a crosswalk from FCID codes to FNDDS codes published by Conrad et al. (2022)\u003csup\u003e30\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003ch4\u003eImpact Factors\u003c/h4\u003e\n\u003cp\u003eFirst, the forced labor risk scores for each food commodity in the U.S. food supply were obtained from prior work, which developed a method to assess the risk of forced labor in food value chains\u003csup\u003e22\u003c/sup\u003e. These scores, originally in the unit of medium risk hours-equivalent (mrh-eq) per ton of food produced, were divided by 1,000,000 to convert to mrh-eq per gram of food produced. We also created average risk scores for tropical fruit, seeds, grains, beans, flours, and poultry by averaging the risk scores of their related food commodities. These average scores were applied to FCID commodities that did not have an available match in the forced labor risk dataset.\u003c/p\u003e\n\u003cp\u003eThen, we multiplied the risk scores by the corresponding consumed amount in the NHANES dataset to get the total forced labor risk per food item, per day, per NHANES participant. If participants had two days of diet recall, then Day 1\u0026rsquo;s and Day 2\u0026rsquo;s amounts of forced labor risk and consumed and inedible amounts were used to get a daily average. If participants had 1 day of diet recall, then Day 1\u0026rsquo;s values were used as the daily average.\u003c/p\u003e\n\u003cp\u003eLastly, accounting for NHANES sampling weights and survey design parameters, we calculated (i) average daily amount of forced labor risk, by food subgroup, and (ii) average daily consumed and inedible amount, by food subgroup. The final forced labor risk impact factors for each food subgroup were calculated:\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eForced labor risk per 1 gram of food(i) =\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAverage daily amount of forced labor risk(i) / Average daily consumed and inedible amounts(i)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003efor food subgroup \u003cem\u003ei\u003c/em\u003e (\u003cstrong\u003eTable 1\u003c/strong\u003e).\u003c/p\u003e\n\u003ch3\u003eStatistical Analysis\u003c/h3\u003e\n\u003cp\u003eThe total forced labor risk for each pattern was calculated by multiplying the food subgroup-specific risk impact factors by the corresponding intake amounts and summing these values across all food subgroups. For each dietary pattern, the percent contribution of each food subgroup was calculated as the ratio of its forced labor risk to the total forced labor risk for that dietary pattern.\u003c/p\u003e\n\u003cp\u003eWe also calculated each protein food subgroup\u0026rsquo;s percentage contribution to (i) total protein intake and (ii) total forced labor risk of the overall dietary pattern. This allowed quantifying the relative impact of different protein sources on the forced labor risk of each dietary pattern.\u003c/p\u003e\n\u003ch4\u003eSensitivity Analyses\u003c/h4\u003e\n\u003cp\u003eTwo different sensitivity analyses were conducted to assess the robustness of the main results. The first focused on the risk scores for each of the 18 food subgroups, replacing the baseline scores with the minimum and maximum FCID-level scores in separate scenarios to understand how that would impact the overall diet rankings in terms of total risk. For example, the food commodities with the lowest and highest risk scores contributing to the whole fruit subgroup\u0026rsquo;s weighted score (0.016 mrh-eq/100g) were pomelo (0.004 mrh-eq/100g) and avocado (0.156 mrh-eq), respectively. In the minimum scenario, the pomelo score was used in replacement of the whole fruit subgroup\u0026rsquo;s score. Likewise, in the maximum scenario, the avocado was used.\u003c/p\u003e\n\u003cp\u003eThe second sensitivity analysis evaluated the impact of including or excluding the feed-related risk scores for animal-based food subgroups, comparing the baseline scenario that included feed scores to a scenario that did not.\u003c/p\u003e\n\u003cdiv id=\"ftn1\"\u003e\n \u003cp\u003e[1] Dark green vegetables, red and orange vegetables, starchy vegetables, other vegetables, whole fruits (excluding juice), 100% fruit juice, whole grains, refined grains, dairy, eggs, poultry, red meat, seafood, nuts and seeds, legumes, unsaturated fat (oil), saturated fat, and added sugar\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eDietary Guidelines for Americans (DGA); Food and Agriculture Organization (FAO); Food and Nutrient Database for Dietary Studies (FNDDS), Food Commodity Intake Database (FCID); Healthy Mediterranean-Style (MED); Healthy U.S.-Style (HUS); Healthy Vegetarian (VEG); Medium risk hours-equivalent (mrh-eq); National Health and Nutrition Examination Survey (NHANES); Planetary Health Diet (PHD); United States Department of Agriculture (USDA); What We Eat in America (WWEIA)\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eData Availability\u003c/p\u003e\n\u003cp\u003eThe detailed results and background data files are available for download at [insert permanent DOI from Dataverse upon manuscript acceptance].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCode Availability\u003c/p\u003e\n\u003cp\u003eData processing and analysis were performed using R (v.4.4.0), Microsoft Excel (v.16.83), TableauPrep (v.2024.1), and TableauDesktop (v.2023.2.0). R scripts and TFL (Tableau Prep) files are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThis research was supported by the Interdisciplinary Research Innovation Fund (RAFINS) at the Friedman School of Nutrition Science and Policy at Tufts University. We gratefully acknowledge Becket Harney for administrative support provided during the development of this manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAuthor Contributions\u003c/p\u003e\n\u003cp\u003eAuthor contributions are listed alphabetically within each category. JLDS and NTB conceptualized the study. BMB, BJ, ERH, JLDS, and NTB designed the methodology. CBN and ZC provided expert input on the methodology. BMB and ERH developed the code. ASM, BMB, BJ, ERH, JLDS, and KB collected and analyzed the data. BMB, BJ, ERH, JLDS and NTB wrote the original draft. All authors edited, reviewed, and approved the final manuscript. JLDS, BJ, and NTB supervised and administered the project. Both BMB and ERH contributed equally and have the right to list their name first on their CV. Both BJ and NTB contributed equally and have the right to list their name last on their CV.\u003c/p\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 2 are available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-4999594/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4999594/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Modeling research on sustainable diets has primarily focused on human and planetary health, while neglecting the social dimension of sustainability, despite the agricultural and fishing sector’s significant global employment and high forced labor rates. To address this gap, our prior work developed a forced labor risk scoring method and applied it to the U.S. food supply. Expanding on this, we assess forced labor risk in current U.S. diets, three U.S.-specific recommended dietary patterns, and the EAT-Lancet Planetary Health Diet. We find that the forced labor risk is highest in the Mediterranean-Style and the U.S.-Style recommended patterns and is lowest in the Planetary Health Diet pattern, with the biggest differences driven by intake of fruit, dairy, and red meat. These results highlight synergies and tradeoffs between human health, environmental sustainability, and social well-being that should be considered in dialogue and action on sustainable diets.","manuscriptTitle":"The human cost of current and recommended diets in the U.S.","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-09-27 12:12:25","doi":"10.21203/rs.3.rs-4999594/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-food","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"natfood","sideBox":"Learn more about [Nature Food](http://www.nature.com/natfood/)","snPcode":"","submissionUrl":"","title":"Nature Food","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"da6a75c7-194d-4915-9136-88bd7ea06ba4","owner":[],"postedDate":"September 27th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":38228957,"name":"Earth and environmental sciences/Environmental social sciences/Sustainability"},{"id":38228958,"name":"Scientific community and society/Agriculture"}],"tags":[],"updatedAt":"2025-10-09T07:08:07+00:00","versionOfRecord":{"articleIdentity":"rs-4999594","link":"https://doi.org/10.1038/s43016-025-01242-8","journal":{"identity":"nature-food","isVorOnly":false,"title":"Nature Food"},"publishedOn":"2025-10-08 04:00:00","publishedOnDateReadable":"October 8th, 2025"},"versionCreatedAt":"2024-09-27 12:12:25","video":"","vorDoi":"10.1038/s43016-025-01242-8","vorDoiUrl":"https://doi.org/10.1038/s43016-025-01242-8","workflowStages":[]},"version":"v1","identity":"rs-4999594","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4999594","identity":"rs-4999594","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}