How Americans eat red and processed meat: an analysis of the contribution of thirteen different food groups

Objective: Dietary patterns characterised by high intake of red and processed meat are associated with detrimental health and environmental outcomes. To better understand how Americans consume red and processed meat, this study examined the food groups that are the greatest contributors to red and processed meat intake in US diets. Design: Cross-sectional analysis of total red and processed meat, unprocessed red meat and processed meat using data from the National Health and Nutrition Examination Survey (2015–2016 and 2017–2018). Items containing red or processed meat were classified into thirteen mutually exclusive food groups. For highly consumed food groups (≥10 % of meat intake), contribution to meat intake was further assessed by source, sex, income and education. Setting: Nationally representative sample of the US population. Participants: Teens (aged 12–19 years) and adults (aged ≥20 years) who reported meat consumption (n 8178). Results: Meat mixed dishes (18·6 % (95 % CI 16·2, 20·9)), burgers (17·3 % (95 % CI 15·3, 19·3)) and beef excluding ground (17·0 % (95 % CI 13·8, 20·1)) were the top contributors to unprocessed red meat intake. For processed meat, four food groups made up about four-fifths of total intake: cold cuts and cured meats (37·7 % (95 % CI 34·6, 40·8)), sausages and frankfurters (20·3 % (95 % CI 18·6, 22·0)), bacon (14·0 % (95 % CI 12·3, 15·6)) and pizza (10·1 % (95 % CI 8·7, 11·5)). Fast-food restaurants were the top source for burgers and pizza, whereas stores were the top source for all other highly consumed food groups. Few differences were seen in patterns of intake by sociodemographic characteristics. Conclusions: No single food group accounts for a majority of meat intake in the USA. Many behaviour change opportunities for healthier, more sustainable substitutions exist.

The 2020 US Dietary Guidelines Advisory Committee report concluded that there is 'strong' evidence that dietary patterns characterised by a higher intake of red or processed meat are associated with an increased risk of all-cause mortality and CVD, and 'moderate' evidence of increased risk of overweight and obesity, type 2 diabetes and colorectal cancer (1) . Unprocessed red meat is a good source of protein and micronutrients such as Fe and vitamin B 12 (2) . However, in high-income contexts such as the USA, most of the population (92 %) has adequate intakes of vitamin B 12 (3) , intake of protein far surpasses minimum requirements (4) , and the prevalence of Fe deficiency anaemiathe most common nutrient deficiency in the USAis 10·4 % among females and 5·2 % among males (5) , much lower than that seen in lower-income contexts. At the same time, meta-analyses suggest that high intake of unprocessed meat is harmful for health (6)(7)(8) , associated with a 16 % higher risk of CHD (9) and 19 % higher risk of type 2 diabetes (10) . The harmful effects of high intakes of processed meat are even greater (6)(7)(8) .
In addition to the adverse health effects associated with red and processed meat consumption, there are important environmental consequences of intensive meat production, with beef production being particularly harmful (11) . Specifically, US beef production requires twenty-eight times more land, eleven times more irrigation water and results in five times more greenhouse gases, compared to the average of other livestock categories (dairy, poultry, pork and eggs) (12) . Meats (beef, pork, chicken and other meats) account for 57 % of total greenhouse gas emissions in US diets, with 81 % of those meat-related emissions coming from beef alone (13) . In order to optimise both human nutrition and planetary health, the EAT-Lancet commission recommends consuming no more than 98 g/week of red meat and especially low intakes of processed meat (14) . Similarly, the American Heart Association 2020 Strategic Impact Goals for diet include reducing processed meat to none or ≤2 servings/week or about 100 g/week (15) . However, mean intake in the USA is much higher than these targets, at 284 g/week for unprocessed red meat and 187 g/week for processed meat (16) .
Few previous studies have evaluated which food groups contribute to unprocessed red meat and processed meat intake in the USA. An analysis of national dietary intake data from 1999-2016 found that lunch meats, sausages, hot dogs and ham account for nearly 90 % of processed meat intake but did not evaluate food groups contributing to unprocessed red meat intake (16) . Another recent analysis used the What We Eat in America (WWEIA) food categories to evaluate the American diet with respect to the Global Burden of Disease recommendations but did not disaggregate dietary sources of red and processed meat (17) . All of these studies are outdatednone used the latest round (2017-2018) of national dietary intake data.
In addition, few studies have examined the association of sociodemographic characteristics with meat intake (16,18,19) . The latest study to explore these associations found that income is not significantly related to unprocessed red meat or processed meat intake, but that unprocessed red meat intake is higher in individuals aged 35 to 64 years (compared to younger and older individuals) and in individuals with a high school education or less (16) . These differences are important given that dietary choices are influenced by income and educational background, among other factors (20) . Indeed, in a national survey, cost was identified as the most important reason for reducing meat intake among lower-income individuals (21) . Yet, how food groups contributing to meat intake differ according to sociodemographic characteristics has not been explored.
In order to identify opportunities to intervene on consumer behaviour, it is important to understand the way in which meat is prepared and consumed. Americans frequently eat meat in mixed dishes such as tacos, pizza or pasta, but the meat eaten in these food groups may be overlooked relative to foods such as steak or pork chops that only contain red meat (17) . The WWEIA Food Categories, which classify all foods reported in the National Health and Nutrition Examination Survey (NHANES) into mutually exclusive food groups based on usage and nutrition (22,23) , can be leveraged to better understand not only how much meat Americans are eating, but also the myriad ways in which meat is consumed. Previous studies have evaluated the contribution of different food groups to specific nutrients such as protein (19) , saturated fat (24,25) and Na (25)(26)(27) . None has comprehensively quantified the contribution of different food groups to red and processed meat intake.
The primary objective of this study was to identify the food groups that contribute the most to red and processed meat intake in American diets using the most recently available nationally representative data. Secondary objectives included evaluating differences in the contribution of food groups to red and processed meat intake by (1) sociodemographic characteristics and (2) source (i.e. where meat is purchased). Results will inform the development of strategies to reduce red and processed meat consumption.

Data sources
This study used data from a single-day dietary recall in the 2015-2016 and 2017-2018 waves of NHANES. NHANES is a repeated cross-sectional survey that uses multistage probability design to sample the civilian, noninstitutionalised population residing in all fifty states and the District of Columbia (28) . Trained interviewers used the US Department of Agriculture (USDA) Automated Multiple Pass Method to collect data on individual food intake (29) . Participants were asked to recall all foods and beverages they consumed the previous day (29) . Measuring guides were used to assist with approximating the portion sizes of consumed foods (29) .
Eligible participants for this analysis were teens (12-19 years) (30) and adults (20 years and older) (31) who had 1 d of valid dietary intake data and were neither pregnant nor lactating.
Derivation of food groups and contributions to meat intake NHANES 24-h dietary recall data were merged to the USDA Food Patterns Equivalent Database (FPED) to obtain grams of meat consumed per d. FPED is a tool that compliments each cycle of the NHANES survey and uses standard recipes to disaggregate mixed dishes into their component ingredients, such that the weight of cooked meat was disaggregated from the weight of the remaining dish ingredients (32,33) . Using FPED, unprocessed red meat intake was defined as follows: (1) FPED 'Meat' component, including mammalian muscle meat from beef, veal, pork, lamb and/or game meat and (2) FPED 'Organ Meat' component, including organ meat from beef, veal, pork, lamb, game and/or poultry (32) . Processed meat intake was defined using the FPED 'Cured Meat' component, which includes bacon, frankfurters, hot dogs, sausages, pepperoni, jerky and luncheon meats that are made from beef, pork or poultry (32) . Total red and processed meat intake was the total of unprocessed red meat intake þ processed meat intake, that is, the total of FPED Meat, FPED Organ Meat and FPED Cured Meat. For dishes in which there were both unprocessed red meat and processed meat components, the FPED classification scheme enabled the disaggregation of the weight of unprocessed red meat and processed meat so that they could be analysed separately.
Ounce equivalent weights were converted from FPED to grams using a conversion factor of 1 ounce = 28·35 g (17) . Individuals with zero reported intake of both unprocessed red and processed meat were retained in the analytic dataset for survey-weighting purposes but excluded from further analyses (see online Supplemental Fig. 1).
The WWEIA Food Categories are released as a tool with each cycle of NHANES and classify all NHANES into 167 mutually exclusive categories by grouping similar foods together based on usage and nutrient content (22,23) . NHANES data, with grams of red and processed meat derived from FPED, were merged to WWEIA by food code. Thirteen food groups were derived from the twenty-eight WWEIA Food Categories containing red or processed meat (Table 1) (22) . WWEIA categories that contributed less than 1 % of total meat intake by weight were collapsed into 'Other dishes'.

Source and sociodemographic characteristics
The source of the food group was based on the answer to the question, 'Where did you get this or most of the ingredients for this?' (29) . Source was classified into four categories: stores (grocery stores, supermarkets, convenience stores and storesno additional info); fast-food restaurants; full-service restaurants and others (including cafeterias, dining halls, community meal programmes, gifts and others) (34) .
For age, participants were classified as teens (aged 12-19 years) (30) or adults (aged 20 years and older) (31) . Teens were included in this analysis to capture a period of development during which individuals are heavily targeted by marketing, susceptible to peer influence, and form their food preferences (35,36) . For income, the Poverty Income Ratio (PIR), a measure of family income relative to the Federal Poverty Level, was used to create income categories. Family income was categorised as low (PIR 0-185 %), middle (PIR 186-400 %) and high (PIR >400 %) (37) . For educational attainment, adult participants were classified as low (less than high school degree), middle (high school graduate/GED or equivalent or some college/associate degree) and high (college graduate or above) (28,31) . Educational attainment of the household reference person was used to classify teens into low (less than high school degree), middle (high school/GED or equivalent or some college/associate degree) and high (college graduate or above) in accordance with the NHANES Sample Design and Estimation Procedures Manual (28) and specifications in the NHANES Data Documentation Codebook (31) .

Statistical analyses
Descriptive statistics were used to quantify the per cent contribution (95 % CI) of each food group to total red and processed meat, unprocessed red meat, and processed meat intake. Results were obtained overall and, for food groups that accounted for ≥10 % of either unprocessed red or processed meat intake, further by source, age, sex, income and educational attainment. The relative CI widths of the proportional measures were examined to assess the reliability of estimates in accordance with recommendations of the National Center of Health Statistics (38) . Adjusted Wald tests were used to test for significant differences between groups. Mean meat intakes (95 % CI) were calculated for descriptive purposes. Due to differences in energy needs between teens and adults, mean daily intakes are presented separately for these two groups. All analyses were conducted in Stata SE 16 (39) , using Stata's svy and subpop commands with the NHANES survey weights to account for complex survey design.

Discussion
This study provides the most up-to-date information on how Americans are consuming red and processed meat. Overall, and consistent with prior literature, average meat  intake among consumers -105 g/d for adults and 94 g/d for teensexceeds recommendations outlined by the EAT-Lancet Commission on Food, Planet, Health and the American Heart Association that cite a target of about 100 g of total meat intake per week (40,41) . Three of the thirteen food groups evaluatedburgers, beef-excludes ground and meat mixed dishesaccounted for 17-19 % of unprocessed red meat intake each. Mexican mixed dishes were also significant contributors, accounting for 14 % of unprocessed red meat intake. For processed meat, four food groups made up about four-fifths of total intake: cold cuts and cured meats, sausages and frankfurters, bacon, and pizza. Cold cuts and cured meats alone accounted for over a third of all processed meat intake in the US diet, almost double the contribution of sausages and frankfurters, the second-highest contributor. Nonetheless, no one food group stood out as being the main contributor to red meat intake, and so interventions could target a variety of food groups, such as sandwich meats, burgers, sausages, traditional US meat mixed dishes such as meatloaf. Replacing even one of these food groups with healthier, more sustainable foods could have a meaningful impact on total meat intake in the USA. Changing dietary behaviour is challenging, particularly for processed meat, which has remained unchanged over the past 18 years, and unprocessed red meat, which has declined very little over the same period (16) . However, a recent systematic review found that several randomised controlled trials showed promising effects on meat reduction, especially reducing meat portion sizes and providing meat alternatives (42) . Moreover, an estimated 16 % of Americans were considered 'potential changers' should the Dietary Guidelines for Americans include information and suggestions for sustainable diets (43) and 55 % of Americans reported reducing the amount of processed meat they consumed in the past 3 years and 41 % reduced red meat (21) . This previous literature supports the possibility of future declines in meat intake, particularly with targeted, evidence-based interventions.
Teens as compared to adults ate less total red and processed meat and were more likely than adults to get their meat from burgers and pizza, which were primarily sourced from fast-food restaurants and accounted for 28·5 % of teens' unprocessed red meat intake and 17·6 % of their processed meat intake, respectively. Previous studies have shown that teens are more motivated to adopt plant-based diets for ethical reasons including environmental concern as compared to older adults who are more motivated to adopt similar diets for health reasons (44) . Emphasising the environmental benefits of meat reduction or choosing plant-based alternatives may be especially impactful for this demographic group, particularly because they already eat less meat relative to American adults in this study. The increasing availability of biomimicry burgers in the USA (e.g. the Beyond Burger and Impossible Burger), including at major fast-food chains (45,46) , represents one promising opportunity for replacement of red meat (in particular, beef) with more sustainable plant-based burgers. Additionally, pizza is an emerging market for biomimicry products, as supported, for example, by a national pizza chain launching a Beyond Pepperoni pizza in 2021 (47,48) . Additionally, many popular pizzas are already meat-free, providing an opportunity to shift behaviour away from meat towards toppings such as vegetables, mushrooms or plain cheese (49) .
With regard to processed meat, cold cuts and cured meats made up more than one-third of intake. Sausages and frankfurters made up another 20 %, while bacon and pizza accounted for about 14 % and 10 %, respectively. These findings are consistent with a recent analysis of NHANES data from 2015 to 2016 that reported the following gram-weight proportional contributions to processed meat: 39·3 % from lunch meat, 24·4 % from sausage, 9·4 % from hot dogs, 9·4 % from ham and 4·6 % from bacon (16) . It is clear from these analyses that cold cuts and cured meats are a major contributor to processed meat intake in the USA and may be consumed as part of a broader unhealthy dietary pattern (50) . Thus, they should be a focus for food policy, including school-based interventions.
Considering health and environmental impacts, it is important not to treat all meat as homogenous. For example, processed meat has stronger adverse effects on health than unprocessed meat (51,52) , and beef has a much higher carbon footprint than pork (53) . In this study, burgers and beef-excludes ground accounted for about 35 % of all unprocessed red meat consumption by Americans, whereas cold cuts and cured meatswhich have relatively more pork than beef productsaccounted for most processed meat consumption. Future research should explore how public health interventions or new products targeting specific food groupsfor example, plant-based burgers, sausages and pepperonimight have differential environmental impacts and identify which interventions and products optimise both health and environmental outcomes.
There are several strengths and limitations of this study. We used data from the two most recent waves of NHANES and applied weights such that estimates are nationally representative. These are best practices for obtaining reliable estimates of population-level red and processed meat intake (54) . However, no food grouping system is perfect. This meant that some foods in the WWEIA Food Categoriesand therefore the derived analytic food groupscould have fallen under multiple classifications. For example, 'Meat and hominy soup, Mexican style' was classified as a 'Soup' but could have reasonably been classified as 'Mexican mixed dish'. Additionally, the FPED food groups do not allow for analysing beef separately from pork, and poultry is included in processed foods and organ meats. Given that the environmental impact of beef, pork and poultry varies substantiallyfor example, beef has approximately eight times the carbon footprint of chicken (32·8 CO 2 eq/kg v. 4·2 CO 2 eq/kg per kg of edible boneless weight) (13) our ability to conclusively comment on the environmental impact of observed changes is limited. While we may have missed some combination foods, using the WWEIA Food Categories facilitates comparisons to previous studies. Related to these two limitations, the 2020 US Dietary Guidelines Advisory Committee report highlighted the need for future work on eating patterns to more clearly differentiate lean meat, red meat and processed meat (1) . Findings of this study may be used to refine these definitions and also to improve translation into practice.
This study of how Americans eat red and processed meat indicates that burgers, beef, mixed dishes, cold cuts, sausages, frankfurters, pizza and bacon contributed more than 10 % to either unprocessed red or processed meat intake and are major routes of exposure to meat. Further efforts are needed to promote reduction in top contributors such as cured meats, cold cuts, beef and pizza, either through the development of biomimicry products such as plant-based lunch meats and pepperoni, or other plant-based protein choices such as beans or nut-based spreads (e.g. peanut butter) that are already enjoyed by millions of Americans (19,55) . However, it will be important to closely monitor trends in protein choices to ensure that environmental gains are not traded at the cost of health (e.g. ensuring nut-based spreads are not high in added sugars). Additionally, it is important that culturally appropriate protein alternatives are available, accessible and affordable. Nonetheless, no one food group stood out as being the main contributor to red meat intake, and so interventions could target a variety of food groups, such as sandwich meats, burgers, sausages and traditional US meat mixed dishes such as meatloaf. Replacing even one of these food groups with healthier, more sustainable foods could have a meaningful impact on total meat intake in the USA in communities throughout the USA. Overall, the pattern of meat-eating in the USA is quite complex, and so a one-size-fits-all approach will not be sufficient for meeting meat reduction targets.