129x Filetype PDF File size 1.47 MB Source: www.thelancet.com
Articles Health and environmental impacts of plant-rich dietary patterns: a US prospective cohort study Aviva A Musicus*, Dong D Wang*, Marie Janiszewski, Gidon Eshel, Stacy A Blondin, Walter Willett, Meir J Stampfer Summary Background Diets that are rich in animal-based foods threaten planetary and human health, but plant-rich diets have Lancet Planet Health 2022; varied health and environmental effects. We aimed to characterise a healthy dietary index and three plant-based 6: e892–900 indices by their environmental impacts and associations with risk of cardiovascular disease. *Joint first authors Department of Nutrition Methods In this prospective cohort study, we used data from a food-frequency questionnaire in the US-based Nurses’ (A A Musicus ScD, D D Wang MD, Health Study II. Participants were categorised by quintiles of four dietary indices, including the alternative healthy S A Blondin PhD, eating index-2010 (AHEI), plant-based diet index (PDI), unhealthy PDI, and healthy PDI. We calculated environmental Prof W Willett MD DrPH, Prof M J Stampfer MD DrPH), impacts (greenhouse gas emissions and irrigation water, nitrogenous fertiliser, and high-quality cropland needs), and Department of Epidemiology relative risks (RRs) of cardiovascular disease from 1991–2017, comparing quintiles. (Prof W Willett, Prof M J Stampfer), Harvard Findings T H Chan School of Public We included 90 884 participants in the health-impact analysis and 65 625 participants in the environmental- Health, Boston, MA, USA; impact analysis. Comparing the top and bottom quintiles, higher AHEI scores were associated with a decreased Channing Division for Network cardiovascular disease risk (relative risk 0·77 [95% CI 0·66–0·89]); 30% lower greenhouse gas emissions Medicine, Department of (Q5 2·6 kg CO equivalent vs Q1 3·7 kg CO equivalent); and lower fertiliser, cropland, and water needs (all Medicine, Brigham and 2 2 Women’s Hospital and Harvard ptrends<0·0001). Similarly, the highest healthy PDI and PDI quintiles were associated with a decreased cardiovascular Medical School, Boston, MA, disease risk (healthy PDI 0·71 [0·60–0·83] and PDI 0·74 [0·63–0·85]) and lower environmental impacts (PDI water USA (D D Wang, needs p =0·0014; all other p <0·0001). Conversely, the highest unhealthy PDI quintile had a higher cardiovascular M Janiszewski BFA, trend trends Prof W Willett, disease risk compared with the lowest unhealthy PDI quintile (1·15 [1·00–1·33]; p =0·023) and required more trend Prof M J Stampfer); Bard cropland (p <0·0001) and fertiliser (p =0·0008). College, Annandale-on- trend trend Hudson, NY, USA Interpretation Dietary patterns that are associated with better health had lower greenhouse gas emissions and (Prof G Eshel PhD); nitrogenous fertiliser, cropland, and irrigation water needs. Not all plant-based diets conferred the same health and Correspondence to: environmental benefits. US dietary guidelines should include nuanced consideration of environmental Dr Aviva A Musicus, Department of Nutrition, Harvard T H Chan sustainability. School of Public Health, Boston, MA 02115, USA Funding US National Institutes of Health. aam231@mail.harvard.edu Copyright © 2022 Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license. Introduction varied associations with health. Dietary indices can be The global food system has enormous impacts on used to differentiate plant-based dietary patterns and freshwater and land use, nitrogen cycles, and climate quantify their associations with health across graded change.1 differences in diets. For instance, the overall plant-based The food-supply chain is responsible for approximately 25% of all human-generated greenhouse diet index (PDI) emphasises consumption of all plant- gas emissions, and uses approximately 70% of the planet’s based foods regardless of nutritional value, and higher consumptive freshwater withdrawals and nearly 40% of PDI scores have been associated with lower risks of global land.2 several chronic diseases, including coronary heart Food systems’ detrimental impacts are not 3,4 limited to the environment. An increase in unhealthy disease. The unhealthy PDI emphasises consumption diets that are rich in heavily processed and animal-based of plant-based foods that are rich in refined grains and foods (eg, red meat) is threatening both planetary and added sugars; diets with higher unhealthy PDI scores human health, contributing to increased rates of obesity, are associated with a higher risk of coronary heart type 2 diabetes, and other non-communicable diseases disease compared with plant-based diets that are rich in worldwide. Because adverse associations between diets whole grains, legumes, nuts, fruits, and vegetables, 4 that are rich in animal-based foods and human and which have higher scores on the healthy PDI. Another environmental health are well established,1 widespread diet index that emphasises plant-based foods is the adoption of healthier plant-rich diets has the potential to alternative healthy eating index (AHEI), which provides reduce disease risk and environmental degradation. higher scores for healthy plant-based foods, and also for Dietary patterns that are rich in plant-based foods some animal-sourced foods such as fish. Diets with differ in the types of foods they include and the extent to higher AHEI scores are associated with a lower risk of 5,6 which they exclude animal-based foods, and thus have major chronic disease. www.thelancet.com/planetary-health Vol 6 November 2022 e892 Articles Research in context Evidence before this study environmental impact and health effects with multiple diet We did not do a formal literature review before undertaking this indices in the same longitudinal cohort. We found that research. Previous research has documented heterogeneous participants in the highest AHEI and healthy PDI score health effects of different plant-based dietary patterns. quintiles had a reduced risk of cardiovascular disease; reduced Prospective cohort studies have found that the alternative greenhouse gas emissions; and reduced use of cropland, healthy eating index-2010 (AHEI), plant-based diet index (PDI), irrigation water, and nitrogenous fertiliser. Participants in the and healthy PDI are all associated with a reduced risk of chronic highest unhealthy PDI score quintile had an increased risk of disease, whereas the unhealthy PDI is associated with an cardiovascular disease, and their diets required more cropland increased risk of chronic disease. Other studies have modelled and fertiliser, compared with those in the lowest unhealthy the environmental impacts of various dietary patterns that are PDI score quintile. rich in plant-based foods, such as vegetarian diets, but research Implications of all the available evidence that simultaneously examines environmental and health Plant-based dietary patterns that are associated with better impacts of plant-rich diets is scarce. human health are also associated with better environmental Added value of this study health. Future US dietary guidelines should include We characterised the health and environmental impacts consideration of environmental sustainability and recognise associated with high versus low scores on various plant-rich the human and environmental health co-benefits of more dietary indices in a US cohort. To our knowledge, this is the sustainable diets, but also that not all plant-based diets confer first study to directly link multiple indicators of the same health and environmental benefits. Although the health effects of different plant-based Boards of the Brigham and Women’s Hospital and the dietary patterns have been widely studied, less research Harvard T H Chan School of Public Health. has addressed the environmental impacts of these Starting in 1991, diet was assessed every 4 years using dietary patterns. Previous studies have modelled a previously validated semi-quantitative food-frequency the environmental impacts of various dietary patterns questionnaire.11–13 For each of approximately 150 listed 7 that are rich in plant-based foods (eg, vegetarian diets), food items, participants reported their usual intake of and some work has addressed specific protective a standard portion of each food item during the previous or deleterious dietary components, such as red meat.8 year. These responses were translated to daily nutrient But, to our knowledge, no study has directly linked intake using the Harvard Food Composition Database multiple metrics of environmental impacts of multiple derived from US Department of Agriculture (USDA) dietary patterns with long-term health outcomes in nutrient data. The reproducibility and validity of nutrient, the same cohort of participants. We aimed to char- food, and dietary pattern measurements using the food- acterise the health and environmental impacts of frequency questionnaire in the Nurses’ Health Study 14 various plant-rich dietary patterns within a longitudinal have been described previously in detail. We did the US cohort. For enviromental outcomes, we primarily health-impact analysis using longitudinally collected data measured greenhouse gas emissions, but also mea- from the baseline (1991) until 2017 in the Nurses’ Health sured irrigation water use, nitrogenous fertiliser use, Study II, based on our previously published work.6 We and high-quality cropland use, all of which have excluded participants with a cancer diagnosis at baseline 9 substantial environmental impacts but are more and cardiovascular disease, and those who reported dependent on local circumstances than greenhouse gas implausible calorie intakes of below 500 kcal/day or emissions are. greater than 3500 kcal/day. The environmental-impact analysis was further restricted to participants within the Methods health-impact-analysis population who completed the Study population and dietary assessment 2011 food-frequency questionnaire, which we used In this prospective cohort study, we used data from the because it was the most detailed food-frequency Nurses’ Health Study II, which began in 1989 when questionnaire and showed recent dietary information 116 430 US female nurses aged 25–42 years completed available from that cohort. a posted questionnaire regarding their medical history The AHEI-2010 was developed as a measure of diet 10 quality, with higher index scores associated with lower risk and lifestyle practices. This cohort has been followed of major chronic diseases.5 up via self-administered questionnaires to update This index scores individuals information on lifestyle and medical history and on the basis of their intake of 11 food groups related to ascertain clinical outcomes every 2 years, with health outcomes, with scores ranging from 0 to 10 for each a follow-up rate of approximately 90% per cycle. The item. We excluded alcoholic beverages from the indices study protocol was approved by the Institutional Review because epidemiological evidence has found alcohol e893 www.thelancet.com/planetary-health Vol 6 November 2022 Articles consumption to be associated both positively and assessments that represent key producing areas. For negatively with several health outcomes, depending on water use when no lifecycle assessment results were intake level. Higher intake of vegetables, fruits, whole available, we used the USDA and other sources, including For the USDA data see https:// grains, nuts and legumes, long chain n-3 fats, and supplementary text and data files from Eshel and quickstats.nass.usda.gov/ polyunsaturated fats are scored positively, whereas sugary colleagues.8,20,21 Additionally, when environmental impact beverages, red and processed meat, trans fat, and sodium estimates were unavailable for a particular food, we used are scored negatively. The total score for an individual data for similar food items with similar agricultural could range from 0 (worst diet) to 100 (best diet). resource use. For example, “other carbonated beverage” The overall PDI and its healthy and unhealthy was equated to Coca-Cola, and parsley to spinach; we also sub-indices were calculated using methods described used the same values for regular and decaffeinated 3,4 previously. Briefly, the PDI scores dietary patterns using coffee. Additionally, the environmental impacts for some 18 food groups on the basis of nutrient and culinary foods were calculated on the basis of the conversion of similarities. These groups are seven healthy plant food one constituent ingredient to another (eg, corn oil, corn groups (whole grains, fruits, vegetables, nuts, legumes, syrup, or popcorn were derived from the environ- vegetable oils, and tea and coffee), five unhealthy plant mental attributes of corn, accounting for corn mass used food groups (fruit juices, sugar-sweetened beverages, derived from fat content, augmented for intermediate refined grains, potatoes, and sweets and desserts), and processing). six animal food groups (animal fats, dairy, eggs, fish and The food-frequency questionnaire included 88 com- seafood, meat including poultry and red and processed pound foods made of two or more food items meat, and miscellaneous other animal-based foods). We (eg, sweetened yogurt or mixed dried fruit). Environmental summed intakes of these 18 food groups, scored as attributes of each compound food were estimated on the quintiles, to obtain overall scores, with a possible range basis of each food’s individual constituent components, of 18–90. For the overall PDI, intake of healthy and with the relative amounts of the components of the unhealthy plant foods increases the index score, whereas compound foods weighted on the basis of US intake of animal foods reduces the score. For the healthy consumption patterns at that time. Nutrient profiles and PDI, only the intake of healthy plant foods increases the recipes for food-frequency questionnaire food items were score, whereas for the unhealthy PDI, only the intake of derived from USDA data and product label information. unhealthy plant foods increases the score; intake of the other two food groups in each index reduces each score. Statistical analysis For all three plant-based diet indices, higher scores We calculated the environmental impact per person for reflect lower animal-based food intake. We excluded each food-frequency questionnaire item by multiplying margarine because of the changes in trans-fat content each individual’s serving intake level by the over time. However, we adjusted for alcoholic beverages environmental impact associated with a one-serving and margarine consumption as covariates in the increment in the intake of that food-frequency analyses. questionnaire item. We then summed those environ- mental impacts across all food-frequency questionnaire Assessment of environmental impact items for each individual, obtaining cropland, reactive We estimated greenhouse gas emissions and use of high- nitrogen, and irrigation water use, and greenhouse gas quality cropland (as distinct from rangeland or grassland), emissions per person per day. To calculate the reactive nitrogen (from fertiliser) and irrigation water environmental impacts attributable to a specific food from field to farm gate (impact of food production group, we summed the environmental impacts within measured until the food is ready for consumption, but each food group and calculated the percentage before it is transported or consumed) for each of the contribution to each environmental impact metric by 156 items included in the 2011 Nurses’ Health Study II each food group. To quantify the associations of diet food-frequency questionnaire. Our analysis excluded healthiness with environmental attributes, we first post-farm-gate environmental impacts (eg, transportation categorised participants by index score quintiles for each and waste). Our full food-frequency questionnaire envi- of the four dietary indices (AHEI, PDI, unhealthy PDI, ronmental database provides detailed background, data, and healthy PDI). We then applied general linear models assumptions, and citations regarding all food items’ with environmental attributes as dependent variables 15 and quintiles of dietary indices and total energy intake as environmental impacts. Environmental impacts were primarily derived from independent variables to calculate the energy-adjusted previously reported values that were based on lifecycle mean environmental impact in each quintile of dietary 8 indices and standardised the mean values to a total assessment studies and supplemented by other 16,17 energy intake of 2000 kcal daily. published values. For the five key livestock categories— beef, dairy, poultry, pork, and eggs—we relied on analyses We estimated relative risks (RRs) for incidence of 18,19 cardiovascular disease (the leading cause of death in the of national consumption and production statistics. For 6 most other items, we mostly relied on several lifecycle USA) across quintiles of dietary indices using Cox www.thelancet.com/planetary-health Vol 6 November 2022 e894 Articles proportional hazards models adjusted for age, race and cohorts have been described in detail.4,6,22,23 To estimate ethnicity, marriage status, living status (living alone or the uncertainty (95% CI) of the RRs, we used Monte- not), family history of myocardial infarction, meno- Carlo simulations to take 1000 draws from the pausal status, oral contraceptive use, multivitamin use, distribution of differences in dietary index and the RRs aspirin use, total energy intake, smoking status, alcohol simultaneously, propagating the uncertainty in the drinking, physical activity, and body-mass index (BMI) dietary index and RRs into the final estimates. We did all in the Nurses’ Health Study II (1991–2017). We calculated analyses at a two-tailed α of 0·05, using SAS (version 9.4) cumulative means to the start of each 2-year follow-up and R (version 4.0.3). interval to best represent long-term dietary patterns and reduce within-person variation; we then included the Role of the funding source cumulative means as main exposures in the Cox The funder of the study had no role in study design, data proportional hazards models. We calculated person- collection, data analysis, data interpretation, or writing of years of follow-up from baseline to the earliest of time of the report. death, cardiovascular disease, loss to or unavailability for follow-up, or the end of follow-up. In secondary analyses, Results we estimated RRs for incidence of and mortality from The health outcomes analysis included 90 884 partic- cancer, coronary heart disease (non-fatal myocardial ipants, and the environmental impact analysis included infarction plus death due to coronary heart disease), 65 625 participants. Participants with diets in the highest stroke, type 2 diabetes, cardiovascular disease (excluding quintiles of greenhouse gas emissions, land use, fertiliser coronary heart disease), respiratory and neurodegener- inputs, and water use generally had higher BMIs ative diseases, and total mortality. We calculated RRs by (28·5–29·4 kg/m²) and lower levels of physical activity (16·4–18· multiplying across-quintile differences in AHEI by 4 metabolic equivalents (METs) × h per week) multivariable-adjusted RRs associated with a one-unit than those in the lowest quintiles (BMI 25·4–26·8 kg/m²; increment in the dietary index. These multivariable- physical activity 21·2–24·7 METs × h per week; table 1). adjusted RRs were estimated from the Nurses’ Health Furthermore, those in the highest quintiles of all Study and Health Professionals Follow-Up Study and four environmental attributes were more likely to be 4,22,23 previously published. The methods for follow-up and current smokers (7·5–9·4%) and have hypertension docu mentation of disease and mortality outcomes in the (33·8–35·6%) and type 2 diabetes (8·4–10·2%) compared Quintiles of greenhouse gas Quintiles of cropland needs Quintiles of fertiliser needs Quintiles of irrigation water needs emission Q1 Q3 Q5 Q1 Q3 Q5 Q1 Q3 Q5 Q1 Q3 Q5 Environmental attribute* 1·7 2·5 3·9 9·7 19·6 40·5 51·5 69·2 97·6 0·4 0·6 0·9 Age, years† 56·7 56·5 56·5 56·9 56·5 56·4 56·8 56·4 56·4 56·8 56·5 56·3 BMI, kg/m² 25·9 27·6 28·9 25·4 27·6 29·4 26·1 27·5 29·0 26·8 27·4 28·5 Ethnicity‡ White 12 367 12 550 12 456 12 319 12 599 12 474 12 552 12 514 12 337 12 536 12 532 12 335 (94·2%) (95·6%) (94·9%) (93·9%) (96·0%) (95·0%) (95·6%) (95·3%) (94·0%) (95·5%) (95·5%) (94·0%) Married 8886 9648 9394 8780 9691 9519 9010 9626 9349 8682 9609 9569 (67·7%) (73·5%) (71·6%) (66·9%) (73·8%) (72·5%) (68·6%) (73·3%) (71·2%) (66·2%) (73·2%) (72·9%) Current smoker 446 638 1094 367 697 1048 357 610 1230 590 632 984 (3·4%) (4·9%) (8·3%) (2·8%) (5·3%) (8·0%) (2·7%) (4·6%) (9·4%) (4·5%) (4·8%) (7·5%) Physical activity, metabolic 22·9 20·1 17·7 24·7 19·7 16·4 23·2 20·4 16·8 21·2 20·6 18·4 equivalents × h per week Energy intake, kcal/day 1718 1720 1740 1705 1707 1751 1717 1710 1728 1698 1722 1721 Multivitamin use 5450 5361 5698 5204 5380 5927 5321 5400 5823 5308 5442 5759 (41·5%) (40·8%) (43·4%) (39·7%) (41·0%) (45·2%) (40·5%) (41·1%) (44·4%) (40·4%) (41·5%) (43·9%) Hypertension 3062 3936 4655 2894 3895 4669 3127 3857 4585 3372 3840 4442 (23·3%) (30·0%) (35·5%) (22·0%) (29·7%) (35·6%) (23·8%) (29·4%) (34·9%) (25·7%) (29·3%) (33·8%) Hypercholesterolaemia 4534 4857 5066 4257 4784 5213 4398 4845 5202 4617 4821 5015 (34·5%) (37·0%) (38 ·6%) (32·4%) (36·4%) (39·7%) (33·5%) (36·9%) (39·6%) (35·2%) (36·7%) (38·2%) Type 2 diabetes 585 801 1186 471 796 1336 593 803 1186 708 824 1100 (4·5%) (6·1%) (9·0%) (3·6%) (6·1%) (10·2%) (4·5%) (6·1%) (9·0%) (5·4%) (6·3%) (8·4%) Data are means for continuous variables and n (%) for categorical variables. Quintiles were calculated on the basis of energy residuals of environmental variables. *Values are mean daily CO equivalent of 2 greenhouse gas emitted (kg), high-quality cropland required (m²), nitrogenous fertiliser (g), and irrigation water (m³). †Not age-adjusted values. ‡Further disaggregated race and ethnicity data were not available. Table 1: Age-adjusted characteristics of study population included in the environmental impact analysis (n=65 625) e895 www.thelancet.com/planetary-health Vol 6 November 2022
no reviews yet
Please Login to review.