Two common dietary patterns identified in British adults, which include high intakes of chocolate and confectionary, may be associated with an increased risk of cardiovascular disease and death in middle-age, according to a study published in the open access journal BMC Medicine.
Carmen Piernas, the corresponding author said: “Cardiovascular disease is one of the main causes of death and disability in the UK and poor diet is a major contributor to this. The most common dietary guidelines are based on the nutrients found in foods rather than foods themselves and this can be confusing for the public.
Our findings help identify specific foods and beverages that are commonly eaten in Britain and that may increase the risk of cardiovascular disease and mortality.”
Researchers from the University of Oxford, UK identified two diets that were associated with an increased risk of cardiovascular disease and death in middle-age in Britain. The first was high in chocolate, confectionary, butter and white bread and low in fresh fruit and vegetables.
The second was high in sugar-sweetened beverages, fruit juice, chocolate, confectionary, table sugar and preserves and low in butter and higher-fat cheese.
The researchers found that those whose diet included higher amounts of chocolate, confectionary, butter and white bread, were more likely to be male, younger, experiencing economic deprivation, current smokers, less physically active, living with obesity or have hypertension compared to those whose diet did not include high amounts of these foods.
In this group, individuals who were younger than 60 years old or living with overweight or obesity had a higher risk of cardiovascular disease than individuals who were older than 60 years or not living with overweight or obesity.
Those whose diet was high in sugar-sweetened beverages, fruit juice and preserves were found to have an increased risk for cardiovascular disease and mortality, even though they also tended to be physically active and less likely to be current smokers or living with obesity, hypertension, diabetes or high cholesterol, than those who did not eat this diet.
Women, individuals who were younger than 60 years old or who lived with obesity in particular had a higher risk of cardiovascular disease, if they consumed a diet high in these foods.
To examine the effects of diet on the risk of cardiovascular disease and mortality, the authors analysed data collected from 116,806 adults from England, Scotland and Wales who were recruited to the UK Biobank between 2006 and 2010. Participants were aged between 37 and 73 years old, with an average age of 56 years old. Participants reported the food they ate during the previous 24 hours on between two and five occasions.
The researchers then identified the nutrients and food groups eaten by participants. The incidence of cardiovascular disease and mortality was calculated using hospital admission and death registry records until 2017 and 2020, respectively.
The authors caution that the observational nature of the study does not allow for conclusions about a causal relationship between diet, cardiovascular disease and mortality. Additionally, as dietary data was taken from individual 24 hour assessments rather than a continuous period of time, it may not be representative of participants’ lifetime diets.
Future research could investigate the potential reasons for the associations between the two diets investigated in this study and cardiovascular disease and mortality.
Carmen Piernas said: “Our research suggests that eating less chocolate, confectionery, butter, low-fibre bread, sugar-sweetened beverages, fruit juice, table sugar and preserves could be associated with a lower risk of cardiovascular disease or death during middle-age.
This is consistent with previous research which has suggested that eating foods that contain less sugar and fewer calories may be associated with a lower risk of cardiovascular disease. The findings of this study could be used to create food-based dietary advice that could help people eat more healthily and reduce their risk of cardiovascular disease.”
There is much controversy surrounding the optimal diet for cardiovascular (CV) health. Data relating diet to cardiovascular diseases (CVDs) has predominantly been generated from high-income countries (HIC), but over 80% of CVD deaths occur in low- and middle-income countries (LMIC). Relatively sparse data on diet and CVD exist from these countries though new data sources are rapidly emerging (1,2).
Non-communicable diseases (NCDs) are forecasted to increase substantially in LMIC because of lifestyle transitions associated with increasing urbanization, economic development and globalization. The Global Burden of Disease study cites diet as a major factor behind the rise in hypertension, diabetes, obesity, and other CVD components (3). There are an estimated over 500 million obese (4,5) and close to 2 billion overweight or obese individuals worldwide (6). Furthermore, unhealthy dietary patterns have negative environmental impacts, notably on climate change.
Poor quality diets are high in refined grains and added sugars, salt, unhealthy fats and animal-source foods; and low in whole grains, fruits, vegetables, legumes, fish and nuts. They are often high in processed food products – typically packaged and often ready to consume – and light on whole foods and freshly-prepared dishes. These unhealthy diets are facilitated by modern food environments, a problem that is likely to become more widespread as food environments in LMIC shift to resemble those of HIC (5,7,8).
In this paper, we summarize the evidence relating food to CVD, and the powerful forces that underpin the creation of modern food environments — what we call the global food system — to emphasize the importance of identifying systemic solutions to diet-related health outcomes. We do this in the context of increasing global attention to the importance of improving food systems by the international development and nutrition community (9–11).
While the “food system” may feel remote to a clinician sitting in an office seeing a patient, its impacts on the individuals they are trying to treat are very real. The paper is based on a World Heart Federation international workshop to review the state of knowledge on this topic.
This review of diet, dietary patterns and CVD is not based on new systematic reviews or meta-analyses but represents a careful review of many published meta-analyses, seminal primary studies, and recent research by the scholars who participated in the Consensus conference.
The paper presents: 1) an overview of the development of the modern, globalized food system and its implications for the food supply; 2) a consensus on the evidence relating various macronutrients and foods to CVD and its related comorbidities, and 3) an outline of how changes to the global food system can address current diet-related public health problems, and simultaneously have beneficial impacts on climate change.
THE CHANGING FOOD SYSTEM AND FOOD SUPPLY AND IMPLICATIONS FOR DIETS AND THE ENVIRONMENT
The development of the modern, globalized food system
Food systems were once dominated by local production for local markets, with relatively little processing before foods reached the household (see Supplementary Box 1) (12). In contrast, the modern food system is characterized by a global web of interactions between multiple actors from farm to fork, geared towards maximizing efficiency to reduce costs and increase production (Figure 1). The major actors who control this system have changed dramatically in HIC and LMIC, as described below (13).

Food System Impact on Nutrition-Related Non-Communicable Diseases
The shift to a global food system started in the US and other high income industrialized countries, and was driven initially by government investment and intervention in markets, infrastructure and research intended to raise farm-sector productivity. Building on actions taken in the late 19th century (14), policies on agricultural research and supporting on-farm production introduced in the 1930–60 period in the US (14) and Europe focused on few major crops, particularly grains (e.g., wheat, corn, rice), oilseeds (e.g., soybeans), livestock (e.g., pigs, poultry, and cattle) and critical cash crops, especially sugar cane and other sources of sugar (15–18). State intervention in most LMICs, took a different form, such as policies to subsidize food, taxes on agricultural producers; and systems to control the supply and marketing of key commodities (19–22). The 1960s also saw the start of significant agricultural transformation in LMIC, with the “Green Revolution” in which focused on increasing productivity of corn, rice and wheat.
These investments and changes in production systems were designed to make calories from staples (e.g., wheat, corn, rice) cheaply available, in order to simultaneously address hunger in LMIC and national food insecurity in HICs (23). In addition to vastly increasing the calorie supply, the ensuing productivity boom also provided the basis of cheap feed for livestock and cheap inputs for processed foods, in turn creating incentives for the growth of manufacturers of processed foods (24). This coincided with huge technological innovations in food processing, (24–28), the rise of to mass marketing to persuade consumers to eat more, supermarket retailing and fast food (29) (30), As a result of these changes, the transformation of raw commodities into food and the distribution of consumable food items beyond the farm gate has become far more important (31). Today integration and control of our farm to fork food supply by major agribusinesses, food manufacturers, retailers, food service companies is more the rule that the exception (13). Meanwhile, production of less processed foods such as coarse grains (e.g., millet, sorghum), roots, tubers, and legumes has declined (32,33) while animal source food production grown dramatically (34).
Figure 2 sets out the stages of change involved in leading to this modern food system. This model has spread unevenly to most LMIC (35–37). Many countries retain various forms of state intervention in agriculture and food systems (18,38–41), but policies to liberalize trade and private sector investment have revolutionized the entire sector in many regions (13,42). Retailing has been transformed in LMIC through the growth of supermarkets (18,38–41). While this process originated with companies in industrialized countries looking for growth in foreign markets, companies based in LMICs are now also investing back into HICs.

Stages of Global Agricultural System’s Development
Dietary impacts
The way people eat has changed greatly across the globe; moreover the pace of change in LMIC’s is quickening. Snacking and snack foods have grown in frequency and number (43–48); eating frequency has increased; away-from-home-eating in restaurants, in fast food outlets, and from take-out meals is increasing dramatically in LMIC; both at home and away-from-home-eating increasingly involve fried and processed food (47,49); and the overall proportion of highly processed food in diets has grown (50,51).
These changes in the global food system coupled with these food behavior shifts have enabled some critical changes to the global food supply, all with dietary implications. First is the shift to refined carbohydrates – refined grains and added sugars. Rapidly increasing production of starchy staples combined with processing technologies mean that refined flour is increasingly dominant in diets. White bread, for example, once rarely consumed in Latin America, became widespread after the introduction of high-yield wheat varieties. In Asia, white rice became dominant as a staple over legumes and coarse grains, with a more recent trend being rapidly rising consumption of instant noodles as a staple (52,53).
Since 1964, average total carbohydrate intake in the US has increased from about 375 g/d to 500 g/d (from 2 to 6 kg/year of ready-to-eat cereals), but the percent of carbohydrate that is fiber has not substantially changed over this time, reflecting increased refined carbohydrates and sugar sweetened beverages (SSBs) is high in HIC (54). In the 1985 to 2005 period extensive added sugar intake occurred across HIC (55) but more recently large increases have occurred in LMIC, particularly in consumption of sugar-sweetened beverages and processed foods (56–59). Today in the US packaged and processed food supply over 75% of foods have some form of added sugar (60). With urbanization there is some evidence to show that refined carbohydrate consumption is increasing, whereas consumption of traditional grains (i.e. millet, maize) is decreasing in LMIC (61,62).
A second key change has been the increasing intake of vegetable oils, including processed vegetable oils, and a decline in consumption of animal fats (2,63). This was initially driven by rising production of soybeans in the US, later in Argentina and Brazil, and then palm oil in East Asia. Oilseeds are now among the most widely traded crops, and are also processed to create margarines and vegetable shortenings and into partially hydrogenated fats and bleached deodorized oils for use in processed foods, (see also Supplementary Box 2).
Between 1958 and 1996, a major global shift occurred in the amount and types of available fats, with soybean, palm, and rapeseed/canola oils replacing butter, tallow, and lard. In 1958–1962, soybean, palm, and rapeseed/canola oils represented 20% of the 29 million metric tons of fat produced globally/year, while butter, lard, and tallow represented 37%. By 1996–2001, these oils accounted for 52% of the 103 million metric tons of fat produced globally/year, while butter, lard, and tallow contributed 20% (64).
This has implications for consumption of fatty acids. Palm oil (deodorized) has become an increasing source of saturated fatty acids; and partially hydrogenated fats are the main source of trans-fatty acids Between 1990 and 2010, global saturated fat, dietary cholesterol, and trans-fat intakes remained stable (trans-fats going down in HIC and up in LMIC), while n-6, seafood n-3, and plant n-3 fat intakes each increased (65).
Vegetable oil consumption remains two times higher in HIC than in LMIC. However, trans-fat consumption is very high in many LMIC while decreasing markedly in HIC. In India, for example, vanaspati, a vegetable ghee used in bakery products, fried snacks and foods sold by street vendors is a leading source of trans-fats, as is bakery shortening. In 75 countries representing 61.8% of the world’s adult population, global saturated fat consumption was ≈9% of energy; though considerable variation existed across countries (range: 2 to 28%).
Country-specific omega 6 consumption ranged from ≈1 to 13% (global mean: 6%) of energy; trans-fat consumption ranged from ≈0.2 to 6.5% (global mean: 1%) of energy; dietary cholesterol consumption ranged from ≈97 to 440 mg (global mean: 228 mg) per day; seafood n-3 ranged from ≈35 to 3,886 mg (global mean: 163 mg) per day; and plant n-3 ranged from <≈100 to 5,542 mg (global mean: 1,371 mg) per day (65).
A third key change has been the increasing global consumption of meat, which has been made economically feasible by subsidized production of crops for animal feed – most importantly corn and soybeans (soybean oil is a byproduct of soymeal production for animals) (66–69). At very low levels of intake animal food consumption may not induce harm, providing high quality protein and iron, whereas excess animal food intake in HIC may be linked to adverse health outcomes, particularly from processed meats (70).
Meat consumption has increased considerably worldwide, and there is substantially greater production of meat in HIC than in LMIC (71). North and South America, Europe, and Australia/New Zealand have the highest meat intake, whereas Asia and Africa have the lowest (72,73). Processed meats (which refers to post-butchering modifications of foods such as curing, smoking, or addition of sodium nitrate), account for a modest proportion of all meat consumed in HIC (unpublished data from the PURE study) (71).
Dietary consumption patterns of other protein sources have been mixed. Between 1973 and1997, dairy consumption per capita (kg) increased in LMIC by ~48% and is projected to almost double (93% increase) by 2020. On average, fish consumption is 2 to 3 times higher in HIC than LMIC (74), although with marked heterogeneity within income categories. China has the highest per capita consumption of fish in the world, followed by Oceania, North America, and Europe (74). Globally, although there has been little or no increase in sea fish consumption per capita since the 1960s, catches per year have risen exponentially (75) and freshwater fish intake has intake has increased during this time (71). Eggs are similarly consumed in higher quantities (2–6 times) in HIC relative to LMIC, with a 14% decline in consumption in HIC observed between 1980–2000, and no change was observed in LMIC (76). The consumption of legumes declined in the US from 1960 and into the 1980’s, with reduced consumption patterns observed globally (8). Relatively, HIC such as Canada, US, and Western Europe, tend to consume the lowest quantities of legumes per capita in the world, whereas LMIC within Africa and India consume the greatest quantities of legumes, along with certain South American countries where beef is uncommon, such as Colombia and Peru (77–79). Globally, pulse consumption has decreased since 1961, from ≈9.5 kg/person/year in 1961 to 6.5 kg/person/year in 2006. In LMIC countries pulses contributed ≈4% of energy to the diets, and just 1% of energy to diets of HIC (80). Total production of tree nuts in 2012 was 3.5 million metric tons, a 5.5% increase from 2011. World consumption of tree nuts in 2011 exceeded 3 million metric tons (81).
A fourth key change is the marked growth of purchases of all packaged foods and beverages (all categories of processing). This process is accelerating across all LMIC markets (13,82,83). For example, 58% of calories consumed by Mexicans come from packaged foods and beverages, which is similar throughout the Americas (83) and even with the US (66%) (65,84). The proportion for China is 28.5% and rising rapidly (36,82,83). The component that is “ultra-processed” – ready to eat, of snack, foods – varies depending on the method of measurement but is increasing wherever it is studied at all income levels (50,85,86). The shift to ultra-processed foods has not just affected the food available for consumption but also the way food is consumed (87). The way people eat has changed greatly across the globe and the pace of change is quickening. Snacking and snack foods have grown in frequency and number (43–48); eating frequency has increased; away-from-home-eating in restaurants, in fast food outlets, and from take-out meals is increasing dramatically in LMIC; both at home and away-from-home-eating increasingly involve fried and processed food (47); and the overall proportion of highly processed food in diets has grown (50,51).
A fifth trend noted above in relation to the added sugar change is the shift in the way LMIC are experiencing a marked increase in added sugar in beverages. In the 1985 to 2005 period extensive added sugar intake occurred across HIC (55) but more recently large increases have occurred in LMIC’s, particularly in consumption of sugar-sweetened beverages and ultra-processed foods (56–59). Today in the US packaged and processed food supply over 75% of foods have some form of added sugar (60).
In addition, fruit and vegetable intake has remained inadequate. Fruit and vegetable consumption is substantially higher in HIC compared to LMIC (88). Analysis of 52 LMIC countries taking part in the World Health Survey (2002–2003) (89) found that low fruit and vegetable consumption (i.e., less than 5 fruits and vegetables per day) prevalence ranged from 36.6% (Ghana) to 99.2% (Pakistan) for men and from 38.0% (Ghana) to 99.3% (Pakistan) for women. Overall, 77.6% of men and 78.4% of women consumed less than the minimum recommended five daily servings of fruits and vegetables. In the US, 32.6% of adults consumed fruit two or more times per day and 27.2% ate vegetables three or more times per day (90). In 2012, 40.6% of Canadians aged 12 and older, reported consuming fruit and vegetables five or more times per day (91).
While all of these changes across LMIC display great heterogeneity (92), the global food system has clearly reached all corners of the LMIC urban and rural sector and major shifts in diets appear to be accelerating.
Implications for environmental impacts
The modern food system is a major force in a range of serious environmental problems, including climate change [as a leading source of greenhouse gas (GHG) emissions, including carbon dioxide, methane, and nitrous oxide], the loss of biodiversity, the strain on freshwater resources, and the release of persistent toxins, excess nitrates and phosphates (from fertilizer and concentrated livestock operations, causing widespread problems of eutrophication), and animal pharmaceutical residues into waterways (12,93–95).
The major causes are beef and other large animals for GHG and the metabolic losses associated with shifting the product of nearly one third of the world’s arable land to concentrated animals, which effectively magnifies the resource budgets and pollution loads of industrial monocultures (34). The expansion of low-input agriculture and extensive ranching are also major factors in deforestation, which bear heavily on both climate change (as carbon is released from vegetation and soils and sequestration capacity diminishes) and biodiversity loss.
In addition to being a major force in many environmental problems, world agriculture is also extremely vulnerable to climate change, biodiversity loss, declining freshwater availability, and the inevitable limits of non-renewable resources (e.g., fossil energy, high grade phosphorous) although vulnerability is highly uneven on a world scale (96). Many of the world’s poorest regions are poised to be most adversely affected by rising average temperatures, aridity, and water stress, as well as through increasingly severe extreme weather events like drought or flooding; in fact some feel this is already occurring (97–99).
It is also noteworthy that the FAO estimates that one-third of all food produced for human consumption globally is wasted before it is consumed, which has both social and environmental costs that have been precisely measured in the US (100–102). Waste affects the entire food system from production to post harvest (including inside the home) (102). Hall’s estimate to home food waste of about 1400 kcal/capita for the US adds up to 25–45% waste of total food for some HIC (101).
MACRONUTRIENTS, FOODS, AND CVD RISK FACTORS
Carbohydrates
Refined Carbohydrates
a) CVD Risk Factors
Ecological evidence from the US demonstrates an association of refined carbohydrate (as corn syrup) with type 2 diabetes (T2DM) and obesity (54). Robust data from systematic reviews and high-quality RCTs support a harmful effect of highly-refined, high-glycemic-load carbohydrates. A meta-analysis of observational studies indicated that high–glycemic index (GI) foods are associated with T2DM (103). Proof-of-concept studies have used alpha-glucosidase inhibitors, such as acarbose, to lower the glycemic index of the foods consumed; for example, in the STOP-NIDDM trial, acarbose reduced progression to T2DM by 25% compared with placebo (104). T2DM risk in individuals with the highest glycemic load (GL) and lowest cereal fiber is 2.5-fold that of those with the lowest GL and highest cereal fiber diet (105). A large Danish prospective cohort study of the impact of replacing saturated fats with high-GI carbohydrates found that when high-GI carbohydrates replace saturated fat, myocardial infarction (MI) risk increases 33% (106). A meta-analysis of 10 prospective cohort studies (n=296,849) (107) found increased GL associated with a 27% increased coronary heart disease (CHD) and MI risk (108). In controlled-feeding studies, replacing saturated fat with carbohydrates lowers low-density lipoprotein (LDL) and high-density lipoprotein (HDL) cholesterol and increases triglycerides (109). Dietary interventions which raise HDL-C may not necessarily translate into CVD risk reduction, as serum HDL-c has recently been called into question regarding its role in the causal pathway of CVD (110). However, in a 5-week controlled-feeding trial of 163 generally healthy but overweight adults, low compared with high-glycemic index diets did not improve insulin sensitivity, lipid levels, or systolic blood pressure in the context of a healthy, DASH-like dietary pattern, and a low compared with high-glycemic index diet increased LDL-C when the carbohydrate content of the diet was high (109). This study suggests that the adverse effect on CHD risk may not be mediated by short term effects on classical risk factors, but that postprandial hyperglycemia and hyperinsulinemia or hyperlipidemia may play a mediating role; further work is needed to clarify the effect of the glycemic index on LDL-C and related lipid risk factors.
b) LMIC
Highly refined carbohydrates include polished white rice, cornstarch, and white wheat flour with reduced fiber content. Carbohydrate refinement is common in HICs and increasing in LMIC (111,112). Traditional diets in LMIC, once rich in whole grains and dietary fiber, now include highly refined carbohydrates, such as polished white rice and refined flours. In East Asian countries, white rice consumption is associated with a 55% higher T2DM risk (111,112). In the cross-sectional CURES 57 study, higher refined grain intake was associated with higher waist circumference, systolic and diastolic blood pressure (SBP and DBP), fasting glucose, triglycerides, and insulin resistance and lower HDL-C levels (61). Replacing white with brown rice (50 grams/day) reduced T2DM risk by 16% (113), and substituting beans for white rice reduced the odds of metabolic syndrome by 35% (108). However, a 16-week randomized trial of replacing white with brown rice in middle-aged Chinese men and women with or at high risk for T2DM only improved HDL-C and reduced DBP in the brown rice group. Data from the international Prospective Urban and Rural Epidemiological Study (PURE; 138,926 individuals in 628 communities 17 countries) suggest a need to consider the context and availability of specific foods before making food choice recommendations (114). Increasing whole grain and cereal fiber consumption, while decreasing total and high-GI carbohydrate, are helpful strategies to prevent T2DM and CVD in the general population (115). Furthermore, low-glycemic index diets improve glycemic control and serum lipids in RCTs of participants with T2DM with major implications for CHD risk reduction in this vulnerable segment of the population whose numbers are increasing rapidly globally (116–119).
Sugar-Sweetened Beverages
a) SSB and CV Risk Factors
Sugar-sweetened beverage (SSB) consumption accounts for up to 50% of added sugar in the American diet (120,121). The epidemiological relationships between SSB consumption, overweight, obesity, hypertension, and T2DM are strong (122). In a meta-analysis of prospective studies of SSB and hypertension, CHD, and stroke, the RR for a 1-serving increase in SSB/d was 1.17 (95 % CI 1.10, 1.24) for CHD and 1.08 (95% CI: 1.04 to 1.12) for incident hypertension, but no clear effect was seen for total stroke (RR 1.06, 95 % CI 0.97, 1.15) (123).
A meta-analysis of seven cohort studies and five randomized controlled trials (RCTs) found SSB’s increased weight by 0.12 kg/serving/year (95% confidence interval [CI]: 0.10–0.14) in adults (124). A World Health Organization (WHO) systematic review reported similar positive associations (125). In a pooled analysis of Nurses’ Health (NHS) and Health Professionals Follow-up Study (HPFS) data, a one SSB serving/day increment is associated with a 1-kg weight gain over a four-year period (126), possibly due to incomplete compensation for this energy at other meals. After adjusting for important potential confounders, including body mass index (BMI), ≥ 1/d versus < 1 SSB/mo increased T2DM by 39% (127).
A meta-analysis of 310,819 participants and 15,043 cases of T2DM reported a 26% increased T2DM risk among those consuming 1–2 SSB servings/day compared to non-consumers (128). A meta-analysis of 4 cohort studies reported a linear association between SSB and hypertension risk RR: 1.08 (95 % CI: 1.04, 1.12) per serving/d. The Framingham Offspring Study reported a 22% higher incidence of hypertension among those consuming ≥ 1 SSB serving/day compared with non-consumers (130).
A potential explanation of the SSB-T2DM association is the high content of rapidly-absorbed sugar from corn syrup, which increases blood glucose and insulin and de novo lipogenesis –which may contribute to pancreatic beta cell dysfunction and eventually T2DM (131). In the National Health and Nutrition Examination Survey III (NHANES III), with 831 CVD deaths during 163,039 person-years of follow-up, consumption of ≥ 7 SSB servings/week was associated with a 29% increased CVD mortality risk compared with < 1 serving/week (120), with no increased risk up to 6 drinks/week.
Few studies have investigated the association between SSBs and CVD events. Those that have report increased CHD and stroke risk with SSB consumption. A 2015 meta-analysis reported a relative risk for incident CHD of 1.17 (95 % CI: 1.10, 1.24) per serving/d increase in SSB consumption (129). Both the HPFS and the NHS find a ≈20% increased CHD risk in the highest category of SSB consumption compared with the lowest category (132,133); and, after adjusting for dietary and non-dietary cardiovascular risk factors, a 16% increased stroke risk (134). Similarly, two Swedish prospective cohort studies in women and men reported a relative risk (RR) of 1.19 (95% CI: 1.04 and 1.36) for stroke among those consuming ≥ 2 SSB servings/day (135).
b) LMIC
Facing health concerns, the beverage industry is shifting from full-calorie carbonated soft drinks (CSDs) to lower-calorie CSDs, coffees, and teas. According to the NHANES, between 1999 and 2006 the average US full-calorie CSD intake decreased, while intake of diet CSDs, low-calorie fruit drinks, and other sweetened beverages increased (122). These products will gradually enter the markets in China, Brazil, and other LMICs. However, there are practically no data on the effects of SSBs on health outcomes from LMIC (136,137).
Fats and Oils
a) CVD Risk Factors
Vegetable oils that are primarily comprised of mono- (e.g., olive oil) and polyunsaturated fatty acids appear to reduce CHD risk, and sources of the n-3 polyunsaturated fat alpha-linolenic acid (ALA), such as rapeseed or canola oil, are cardioprotective. Replacing saturated fat with monounsaturated or polyunsaturated fat reduces low-density lipoprotein cholesterol (LDL-C) and preserves HDL-C (109). Further, canola oil, as part of a low-glycemic index diet, improves glycemic control and blood lipids in type 2 diabetes (138). Trans-fatty acids increase CHD risk compared with other macronutrients, with strong evidence of adverse effects of small amounts of trans-fats on lipids (109,139) and CVD risk (140,141).
Though total fat (142), and specifically saturated fats, have generally been considered to be deleterious to insulin sensitivity (143), in large cohort studies, saturated fat is not associated with development of T2DM, after adjustment for BMI, total dietary fiber, or magnesium intake (144–149). Macronutrient exchange generally does not influence markers of glucose homeostasis, though in two relatively large trials, replacing SFA with either MUFA or carbohydrate improved indices of glucose homeostasis (150,151). Associations have been seen between major food sources of saturated fat, such as red and processed meat, and development of T2DM (152,153), though dairy products, notably fermented dairy, may be protective (154,155).
b) CVD
Saturated fats have not been consistently associated with CVD in meta-analyses of cohort studies (odds ratio [OR]: 1.07; 95% CI: 0.96–1.19) of higher compared with lower intakes (156). However, in most of these studies the association of high saturated fat intake largely represents replacing highly refined carbohydrates. Replacing saturated fat with highly refined carbohydrate is not associated with lower CHD risk, whereas replacing saturated fat with polyunsaturated fat reduces CHD risk (157,158). This benefit of polyunsaturated fat includes the primary n-6 polyunsaturated fatty acid, linoleic acid (158,159). Replacing saturated fat with high-GI carbohydrate increases MI risk by 33%, whereas replacing with low- and medium-GI carbohydrates appears neutral (106). Emerging evidence suggests that the effect of a saturated fat on CHD may depend on the type of fatty acid and the specific food source (i.e., dairy vs. meat) (160,161). Consumption of non-hydrogenated vegetable oils appears to be superior to consumption of animal fats (162). Consumption of plant oils in a Mediterranean diet reduced CVD in two RCTs. The Lyon Diet Heart Study found regular consumption of ALA (canola oil) significantly reduced cardiac deaths and nonfatal CHD (163), while the PREDIMED RCT found that a Mediterranean diet (50 grams/d extra virgin olive oil) reduced CVD events by 30% in over a five-year period (164).
Palm oil is the dominant fat globally and is relatively high in saturated fat. Based on controlled-feeding studies examining changes in blood lipids, replacing palm oil with unsaturated fatty acids would be expected to lower CHD risk (109), but palm oil would be preferred to partially hydrogenated oils high in trans-fatty acids. Few studies have directly compared palm oil with other oils for CHD risk. One large case-control study in Costa Rica found that soybean oil consumption was associated with lower acute MI risk compared to palm oil consumption (165).
Summary of Fats and Oils
compared to saturated fat, vegetable oils rich in polyunsaturated fats reduce the TC:HDL-C ratio and CHD incidence; inclusion of n-3 fatty acids (ALA) with the vegetable oils is important for CHD prevention. The effect of replacing saturated fat with carbohydrate on CHD risk appears to depend on the quality of the carbohydrate. Prospective studies consistently indicate adverse effects of trans-fats on CHD. Effects of monounsaturated fat from plant sources require further study; extra virgin olive oil appears to reduce CVD.
Protein sources
Meats
a) Nutrients and CVD Risk Factors
Meat is rich in protein, iron, zinc, and B-vitamins, but can also contain significant amounts of cholesterol and saturated fatty acids, which raise LDL-C and lower triglyceride (157). A high red meat intake (rich in heme iron), increases endogenous formation of N-nitroso compounds in the gastrointestinal tract that are associated with increased epithelial proliferation, oxidative stress, and iron-induced hypoxia signaling (166–168).
In a meta-analysis of 17 cohort studies (16 from Western countries), consumption of red and processed meat increased T2DM and CHD risk; but few included studies examined unprocessed red meat (169). Intake of red or processed meat was not associated with stroke, but only three studies evaluated these relationships (169). In more recent analyses, red meats, particularly processed red meats, were associated with increased CVD, CHD, stroke, and cancer mortality, while poultry was not (170–172).
Potential mechanisms linking unprocessed red meat with CVD include saturated fat, cholesterol, iron, phosphatidylcholine, and carnitine; and cooking methods (e.g., barbecuing), that increase heterocyclic amine content and N-nitroso compounds, also implicated in colorectal cancer (167). The evidence suggests that processed meat consumption increases CHD risk, while unprocessed meat consumption has a small or no association with CHD, mainly when compared to refined starch and sugar. Both unprocessed and processed red meats are associated with greater CVD risk compared to poultry, fish, or vegetable protein sources. Both types of meat are associated with higher T2DM risk, although gram-for-gram the effect size is notably larger for processed meats.
b) LMIC
Data relating meat consumption to CVD risk in LMICs is limited. A recent pooled analysis of data from 296,721 individuals from Asian countries (i.e., Bangladesh, mainland China, Japan, Korea, and Taiwan) found no association between red meat and poultry consumption and CVD, cancer mortality, or all-cause mortality (173). Red meat intake is generally much lower in these areas than in HICs however, and current consumption does not likely reflect long-term patterns.
Dairy
a) CVD Risk Factors
The consumption of dairy products has been associated with weight loss in small studies (174), but the overall literature does not confirm an important effect on body weight. However, increased low-fat dairy consumption is associated with lower LDL-C, triglycerides, plasma insulin, insulin resistance, waist circumference, BMI, possibly blood pressure (BP); and reduced diabetes risk (174–182). In a large meta-analysis of cohort studies (13,000 incident cases), and the EPIC InterAct case-cohort study (12,000 incident cases), fermented dairy (i.e., yogurt, cheese, and thick fermented milk), but not total dairy, was inversely associated with T2DM (155,183). In a meta-analysis of prospective cohort studies, milk consumption was inversely associated with total CVD in a small subset of studies with few cases, but using a larger body of data with more specific endpoints, milk was not associated with CHD or stroke (173). In a prospective cohort study of 53,387 Japanese men and women, higher dairy calcium (173 mg/d vs. none) reduced risk for hemorrhagic stroke, ischemic stroke, and stroke mortality by ≈50% over a 10-year follow-up (184). Collectively, these studies do not suggest a strong or consistent relationship between consumption of dairy products and T2DM and CVD risk.
b) LMIC
Data on dairy consumption and CVD in LMICs are limited. In a prospective cohort of 2,091 middle-aged Chinese men and women monitored for six years, individuals who reported consuming > 1 dairy serving/day were 35% less likely to develop T2DM (RR: 0.65; 95% CI: 0.49 and 0.85) than nonconsumers (185).
Egg
a) CVD Risk Factors
Eggs are a relatively inexpensive and low-calorie source of protein, folate, and B vitamins (186). Eggs are also a source of dietary cholesterol (a medium egg contains ~225 milligrams of cholesterol) (187). A meta-analysis showed that eggs increase TC, HDL-C, and TC:HDL-C,(188) but five RCTs subsequently reported that egg consumption did not significantly alter these parameters (189–191) or endothelial function (192,193). No RCT has tested the effect of egg consumption on CVD events. In a meta-analysis (194) of 16 prospective cohort studies (90,735 participants) (191), egg consumption was not associated with overall CVD or CHD, stroke, or CHD or stroke mortality; but was associated with T2DM. Overall, consumption of eggs in moderation (one egg/day) is likely neutral for CVD. However, relative to other protein-rich foods that lower LDL cholesterol, such as whole grains and nuts, eggs would likely increase CVD risk.
b) LMIC
Unpublished data from three large international studies, PURE, ONTARGET, and INTERHEART, with collectively 200,000 individuals and 22,000 CVD events from regions including China, India, and Africa, show that moderate egg consumption appears to be neutral or protective against CVD. However, significant variations exist across regions, with a benefit of daily egg consumption in China but possible harm in South Asia.
Fish
a) CVD Risk Factors
Fish are a source of protein, vitamin D, multiple B vitamins, essential amino acids, and trace elements; and the long-chain omega-3 (n-3) fatty acids docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) (195,196), though amounts vary over tenfold across seafood species. Fatty fish, such as salmon, sardines, trout, white tuna, anchovies, and herring, have the highest concentrations. In clinical trials and meta-analyses of these trials the long-chain omega-3 polyunsaturated fatty acids in fish reduced multiple CVD risk factors, including vascular resistance, BP, inflammation, serum lipids, and endothelial function (197).
Some prospective cohort studies find an inverse association between fish intake and CVD mortality, whereas others do not (198). Available evidence suggests cardiovascular benefits with fish consumption in secondary prevention, but the evidence is inconsistent regarding primary prevention. Possible explanations include differences in the amounts and types of fish consumed, cooking methods, and background fish consumption. Fifteen of the 16 cohort studies [exception (199)] were conducted in North America and European countries, where deep-frying fish is common.
The Diet and Reinfarction Trials (DART-1), a secondary prevention trial and DART-2, in men with stable angina, are the only randomized trial of fish intake and CVD outcomes. They arrive at opposite conclusions. In DART-1, fish lowered all-cause mortality and trended towards reducing CVD events after 2 y (194). In DART-2 oily fish did not affect all-cause mortality or CVD events after 3–9y, and increased sudden cardiac death, largely confined to the subgroup given fish oil capsules (200). Differential behavioral change or CVD stage may explain the discrepancy (201). Follow-up of DART-1 at 5y also showed increased rates of CVD in the fish/fish oil group that did not persist through the 10y assessment (202). We know of no primary prevention trial on fish intake and CVD outcomes, but a meta-analysis of fish oil supplement RCTs is neutral (203).
b) Low-income Countries and HICs
Most of the data indicating that fish is protective comes from studies in HICs (204–208). (204,205). Unpublished data from three large international studies (PURE, ONTARGET, and INTERHEART) reflect considerable heterogeneity in the association between fish intake and CVD outcomes. In the PURE study fish intake was inversely associated with CVD outcomes in South America, China, North America, and Europe (RR: 0.76–0.84) but positively associated in South Asia (RR: 1.97).
However, no associations between fish consumption and CVD outcomes were observed in a high-risk secondary population in ONTARGET. INTERHEART found fish intake beneficial in North America and Europe (RR: 0.73; 95% CI: 0.62–0.87) but harmful in the Middle East (RR: 1.59; 95% CI: 1.32–1.93). More work is needed in China and India to understand the effects of regional types and preparation methods of fish a.
Nuts
a) CVD Risk Factors
In clinical trials nuts improve serum lipids (209) (210,211). In observational studies, nuts lower CHD (196,212,213) and hypertension risk (213); but not stroke (207, (213). In one meta-analysis of prospective cohorts, nuts were not associated with T2DM (214), but in another, they were protective (207). Despite their high energy density, nuts do not contribute to weight gain, changes in waist circumference, or obesity perhaps owing to their satiating effects and increased fecal energy losses. (215,216). In observational studies (133,217–220) and RCTs (164,221–223) a Mediterranean diet including nuts lowers CVD risk. However, no RCTs have assessed the effects of nut consumption alone on CVD events. Taken together, the published data from observational studies and clinical trials support nuts for lowering CVD risk.
b) LMIC
Unpublished data from the PURE cohort indicate that nut consumption is very low in LMIC (60% of individuals consume ≤ 1 nut serving/week). No associations between nut consumption and CVD outcomes is seen at this low level.
Legumes
a) CVD Risk Factors
In observational studies and RCTs consumption of legumes improves CVD risk factors, such as waist circumference, cholesterol, BP, C-reactive protein, glucose; and is protective against T2DM (224–231). A meta-analysis of 26 RCTs (1,037 participants) finds that 130 grams legumes/d (~1 serving) reduced LDL-C by 0.17 mmol/L (95% CI: −0.25, −0.09) (228). Legumes also reduce SBP (by 2 mmHg) and lowered mean arterial pressure(231) in a meta-analysis of eight RCTs. A legume-rich diet reduced HbA1c (−0.3%; 95% CI: −1.4, −0.1%) in an RCT in participants with diabetes (226).
b) CVD
In a meta-analysis of five observational studies, 100 grams of legumes four times/week is inversely associated with CHD (RR: 0.86; 95% CI: 0.78–0.94) (232). However, that study and another meta-analysis of eight prospective cohort studies found no association between legumes and diabetes (207) or stroke (232,233). A large prospective study of two cohorts of US health professionals found a 45% increased risk of ischemic stroke per daily serving of legumes (RR: 1.45; 95% CI: 1.06–2.00) (234). This suggests that while legumes are valuable to reduce CHD risk, more research is required to understand their impact on total stroke risk.
c) LMIC
In a cohort of Chinese men and women, soy (the primary legume consumed in China) was negatively correlated with TC and LDL-C (235). A population-based cross-sectional study in India found that women consuming legumes once/day were less likely to develop T2DM (OR: 0.55; 95% CI: 0.34–0.88; n=99,574). A similar but non-significant trend was observed in men (OR: 0.70; 95% CI: 0.39–1.26; n=56,742) (236). In a prospective cohort study of middle-aged Shanghai women, a ~50 gram/day increase in legume consumption over 4.6 years reduced T2DM risk by 38% (RR: 0.62; 95% CI: 0.51–0.74) (237). A prospective cohort study of 64,915 Chinese women aged 40 to 70 (n=62 cases; follow-up 2 y) found that individuals consuming ≥ 11.2 grams/day of soy protein were less likely to develop CHD (RR: 0.25; 95% CI: 0.10 to 0.63) than those consuming ≤ 4.5 grams/day (238).
Summary of major protein sources
Reducing red meats, especially processed meats, and increasing fish, nuts, legumes, and possibly fermented dairy products are likely beneficial. Sustainability issues discussed in this paper must also be addressed specifically in relation to meat, fish, and dairy foods (12,34,239–241).
Fruits and Vegetables
a) CVD Risk Factors
In the Dietary Approaches to Stop Hypertension (DASH) RCT, higher intake of fruits and vegetables, either as part of a typical Western diet or the DASH eating plan, reduced BP, TC, LDL-C, and HDL-C without affecting triglycerides (242). A meta-analysis of three prospective cohorts (3,415 cases) found high adherence to the DASH eating reduced T2DM risk by 27% (243). A strong evidence base from observational studies indicates that high consumption of vegetables and fruits reduces CHD and stroke (196).
b) CVD in HIC and LMIC
Large global studies and systematic reviews of prospective cohorts generally support a protective role of fruits and vegetables against CVD (244–246). In the global INTERSTROKE study (3,000 stroke cases and 3,000 controls), compared with fewer than 1 serving per day, 1 serving of fruit was protective against stroke (OR: 0.61, 95% CI: 0.50 to 0.73), but the benefits were not clear with higher consumption of up to 3 servings/d of vegetables was not t (OR: 0·91; 95% CI: 0.75 to 1.10) (244). A meta-analysis of 20 prospective cohort studies (16,981 stroke events) found that fruit and vegetable consumption was associated with decreased stroke risk (RR: 0.79; 95% CI: 0.75–0.84 for highest vs. lowest categories), as were fruits (RR: 0.77; 95% CI: 0.71–0.84) and vegetables (RR: 0.86; 95% CI: 0.79–0.93) separately (245). In unpublished INTERHEART data a one-serving/day increase in fruit decreased MI risk by 12% (RR: 0.88; 95% CI: 0.84–0.92); a one-serving/day increase in vegetables decreased MI risk by 5% (RR: 0.95; 95% CI: 0.92–0.97). Large meta-analyses of observational studies support a ≈5–10% reduction in CVD mortality per serving per day (246), and a protective association of green leafy vegetables with T2DM (RR per 0.2 servings/day: 0.87; 95% CI: 0.81–0.93) (247). The beneficial effects of fruits and vegetables appear consistent across regions of the world. A major issue has been the lack of success in encouraging increased fruit and vegetable consumption by the citizens of western nations (248).
Overall Summary
Increased refined grains, starches, and added sugar (i.e., carbohydrates of lower quality) has paralleled the rise in obesity and T2DM. Collectively, the data support emphasizing low-GI whole grains, legumes, fruits, and vegetables and minimizing high-GI refined grains and foods with added sugars, including SSBs.
Dietary Sodium
Sodium is an essential nutrient required for normal physiological function (249) (250,251). Increasing sodium intake well beyond physiological requirements increases BP (241–243) and CVD mortality (95–97). In LMIC sodium sources include table salt and salt additives and spices; while processed foods are the primary source in HICs. Sodium consumption has declined slightly in LMIC as refrigeration has replaced salted-preserved foods. RCTs have reported reductions in BP with reduced sodium intake to < 1.5 grams/day.
Population- recommendations for low sodium intake (252–254) (< 2.0 grams/day) have been achieved in short-term feeding clinical trials (254), but not sustained in longer-term clinical trials (> 6 months) (255) (256,257). No RCTs have determined whether low sodium intake reduces CVD events or deaths compared with moderate intake (249). Meta-analyses of CVD event trials did not report a significant reduction in CVD with lower sodium intake, but the included trials were underpowered to detect moderate risk reductions (258). Prospective cohort studies suggest a J-shaped association between sodium intake and CVD events, consistent across methods of sodium estimation (259–265).
In a recent Cochrane Review (266) of 23 epidemiological studies (n=274,683), the lowest risk of CVD events and deaths occurs at an intake between 2.7 and 5.0 grams/day. PURE study findings (263) are consistent with this evidence, with both high (> 6 grams/day) and low (< 3 grams/day) sodium excretion associated with higher mortality and CVD events compared to average (4.00–5.99 grams/day) sodium excretion despite a positive association between sodium excretion and BP (251,263). The increased risk of CVD events with higher sodium intake (> 5 grams/day) was most prominent in those with hypertension (263). However, methodological issues make studies of sodium intake particularly challenging (267). Large RCTs are needed for definitive evidence on optimal sodium intake for preventing CVD events (249,268–270).
Alcohol
a) Consumption patterns
Greater amounts of alcohol are consumed per capita by adults in HIC (9.6 litres/y) than in LMIC (4.1 litres/y), or LIC (3.1 litres/y), although this may be an underestimate in LMIC and LIC due to the use of homemade or unsafe alcohol products (e.g., industrial or medical) (271). In the PURE study, 31% of participants identified as current drinkers, with considerable variation across income regions (13–80%). In all regions, men were more likely to be current drinkers than women (272).
b) CVD Risk Factors
In a meta-analysis of randomized controlled intervention studies of 2–8 weeks’ duration (273), moderate alcohol consumption (up to one drink or 15 g/d in women; or up to 2 drinks or 30 g/d in men) compared with no alcohol: raised HDL-C, Apo-A1, and adiponectin and reduced LDL-C and fibrinogen. Total cholesterol, triglycerides, and lp (a) were not significantly affected. Results stratified by beverage type (wine/beer/spirits) were similar to the pooled analyses. The changes in lipid and hemostatic risk factors align with lower cardiovascular risk.
c) CVD
The European Prospective Investigation on Cancer and nutrition (EPIC) study monitored 380,395 participants for 13 years (4,187 CVD deaths) (274). Female never-drinkers were at increased risk of CVD death (RR: 1.31; 95% CI: 1.13 to 1.53) compared with moderate drinkers (0.1 to 4.9 g/d); but the association was not significant for male never-drinkers (RR: 1.20; 95% CI: 0.89 to 1.62). In a dose-response meta-analysis of 24 prospective cohort studies (275), the association of drinking with CHD was J-shaped: non-drinkers were at slightly increased risk; in men the greatest protection was at 31 g/d with a trend for increased risk beginning at 63 g/d; for women, the greatest protection was at 11 g/d with a trend for increased risk beginning at 14 g/d. In a meta-analysis of 27 prospective studies of alcohol and stroke, no association was found between light or moderate alcohol use and stroke, but heavy alcohol use (>45 g/d) was shown to increase total stroke risk (RR: 1.20; 95% CI: 1.01 to 1.43), particularly hemorrhagic stroke (RR: 1.29; 95% CI: 0.98 to 1.71) (275). Patterns of alcohol consumption may be a particularly important determinant of the magnitude of the association—in a meta-analysis of 10 prospective cohort and 4 case-control studies, irregular “binge” drinking (>60 g of pure alcohol or >=5 drinks per occasion at least monthly) increased CHD risk by 45% (RR: 1.45; 95% CI: 1.24 to 1.70) (276). In summary, small and inconsistent cardiovascular benefit is seen in moderate drinkers, but a strong increase in CVD and stroke risk is seen with regular heavy alcohol consumption and binge drinking.
d) LMIC
In HIC/UMIC in the PURE study (272), current drinking was protective against MI, but not in LIC/LMIC (p=0·02 for interaction); a similar trend was seen for stroke, but the between-region difference was not significant (P=0.15 for interaction). In LIC and LMIC, heavy episodic drinking pattern and higher intakes of alcohol were associated with a composite outcome of mortality, MI, stroke, cancer, injury, and hospitalization. Current drinking was associated with an increased risk of alcohol-related cancers.
Summary of alcohol
The CVD risk benefits of moderate alcohol consumption (up to 5 g/d) compared with abstinence are small but consistent within individual study populations. Higher levels of consumption increase CVD risk, beginning at 30 g/d in women and 45 g/d in men. Therefore, recommendations for moderate alcohol consumption for CVD prevention in people who already drink must consider individual preferences, population-specific responses, and risk of comorbid conditions, such as addiction and cancers. It not recommended that non-drinkers take up drinking.
Dietary Patterns
Dietary pattern analysis assesses overall diet quality and is easier to translate into dietary recommendations than recommendations for single foods or nutrients (277). The U.S. Dietary Guidelines Advisory Committee modeled 3 dietary patterns–the Healthy-US type Pattern, the Healthy Mediterranean style Pattern and the Healthy Vegetarian Pattern–for healthfulness and nutritional adequacy. All 3 include components of a dietary pattern associated with health benefits (277).
The DASH diet emphasizes fruits, vegetables, whole grains, low-fat dairy products and discourages sugar-sweetened foods and beverages, red meat, and added fats (254). The DASH diet was associated with lower SBP, DBP, and cholesterol. However, the DASH trials were not designed to evaluate the impact of this diet on CVD clinical events.
Further improvements in cardiovascular risk factors were seen when part of the carbohydrate in DASH was replaced with either monounsaturated fat or protein (278). In prospective cohort studies, a higher adherence to a DASH dietary pattern is associated with reduced stroke, as well as all-cause, CVD, and cancer mortality (279,280).
The Portfolio Diet (281), which includes 30 grams/day of nuts, 20 grams/day of viscous fiber (i.e., oats, barley, legumes), 80 grams/day of vegetable protein (soy, beans, chickpeas, lentils), and 2 grams/day of plant sterols (plant sterol margarine) from plant foods lowers LDL-C by up to 30% within four weeks, similar to statins (281). Reductions of 13–23% are maintained when this diet was extended to 12 months (282,283).
However, no RCT has evaluated the impact of this diet on CVD clinical events. Strong evidence from prospective cohort studies and randomized trials shows reduced CHD with the consumption of a higher-fat Mediterranean diet (196). In the PREDIMED randomized trial a plant-based diet high in nuts or extra virgin olive oil, vegetables, fruits, legumes, fish, and poultry but low in red meats, sweets, and whole-fat dairy was superior to the control group assigned to a low-fat Western diet in preventing CVD (164).
It bears mention that low-fat diets have not been particularly effective for long-term CVD risk reduction, partly due to the difficulty sustaining such a diet long-term (284,285). Table S1 translates the essence of the Mediterranean diet to other regions and cultures, suggesting replacement foods that maintain the major food groups in particular regions. Randomized trials of CVD outcomes are needed in different regions of the world. Supplementary table 1 provides a Mediterranean Style Diet adapted to various regions of the world. Figure 3 provides an overview of all the components of the food system and the summary of the dietary changes that need to come from food system changes.

Global food system changes needed to meet create a sustainable healthy diet
reference link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4597475/
More information: Associations between dietary patterns and the incidence of total and fatal cardiovascular disease and all-cause mortality in 116,806 individuals from the UK Biobank: a prospective cohort study, BMC Medicine (2021). DOI: 10.1186/s12916-021-01958-x