research paper on diet

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

Defining a Healthy Diet: Evidence for the Role of Contemporary Dietary Patterns in Health and Disease

Hellas cena.

1 Laboratory of Dietetics and Clinical Nutrition, Department of Public Health, Experimental and Forensic Medicine, University of Pavia, 27100 Pavia, Italy

2 Clinical Nutrition and Dietetics Service, Unit of Internal Medicine and Endocrinology, ICS Maugeri IRCCS, 27100 Pavia, Italy

Philip C. Calder

3 Human Development and Health, Faculty of Medicine, University of Southampton, Southampton SO16 6YD, UK

4 NIHR Southampton Biomedical Research Centre, University Hospital Southampton NHS Foundation Trust and University of Southampton, Southampton SO16 6YD, UK

The definition of what constitutes a healthy diet is continually shifting to reflect the evolving understanding of the roles that different foods, essential nutrients, and other food components play in health and disease. A large and growing body of evidence supports that intake of certain types of nutrients, specific food groups, or overarching dietary patterns positively influences health and promotes the prevention of common non-communicable diseases (NCDs). Greater consumption of health-promoting foods and limited intake of unhealthier options are intrinsic to the eating habits of certain regional diets such as the Mediterranean diet or have been constructed as part of dietary patterns designed to reduce disease risk, such as the Dietary Approaches to Stop Hypertension (DASH) or Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diets. In comparison with a more traditional Western diet, these healthier alternatives are higher in plant-based foods, including fresh fruits and vegetables, whole grains, legumes, seeds, and nuts and lower in animal-based foods, particularly fatty and processed meats. To better understand the current concept of a “healthy diet,” this review describes the features and supporting clinical and epidemiologic data for diets that have been shown to prevent disease and/or positively influence health. In total, evidence from epidemiological studies and clinical trials indicates that these types of dietary patterns reduce risks of NCDs including cardiovascular disease and cancer.

1. Introduction

Non-communicable diseases (NCDs) such as cardiovascular disease, cancer, chronic respiratory diseases, diabetes, obesity, and cognitive impairment are among the leading causes of death and disability throughout the world, affecting populations in developed as well as developing countries [ 1 ]. Although there are established genetic and environmental contributors to NCD risk, modifiable lifestyle-related factors play a large role at the individual level [ 2 , 3 , 4 ]. Dietary choices, for example, contribute to the risk for developing hypertension, hypercholesterolemia, overweight/obesity, and inflammation, which in turn increase the risk for diseases that are associated with significant morbidity and mortality, including cardiovascular disease, diabetes, and cancer [ 5 ]. Indeed, the marked rise in chronic NCDs has a causal link to global dietary patterns that are becoming increasingly Westernized [ 6 ], being characterized by high levels of fatty and processed meats, saturated fats, refined grains, salt, and sugars but lacking in fresh fruits and vegetables.

In recognition of the importance of the diet as a determinant of disease risk, the World Health Organization (WHO) Global Action Plan for the Prevention and Control of Noncommunicable Diseases includes strategies for addressing unhealthy diet patterns among its initiatives directed at reducing behavioral risk factors; the other components comprise physical inactivity, tobacco use, and harmful alcohol use [ 1 ]. Dietary changes recommended by WHO include balancing energy intake, limiting saturated and trans fats and shifting toward consumption of unsaturated fats, increasing intake of fruits and vegetables, and limiting the intake of sugar and salt. Many of these dietary targets naturally occur in regional diets such as the Mediterranean diet [ 7 ] or are included as part of evidence-based diets designed to reduce disease risk, such as the Dietary Approaches to Stop Hypertension (DASH) [ 8 ] or Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) [ 9 ] diets. To better understand the current concept of a “healthy diet”, this narrative review describes the features and supporting clinical and epidemiologic data for diets that align with the general WHO guidance and have been shown to prevent disease and/or positively influence health.

2. Components of a Healthy Diet and Their Benefits

A healthy diet is one in which macronutrients are consumed in appropriate proportions to support energetic and physiologic needs without excess intake while also providing sufficient micronutrients and hydration to meet the physiologic needs of the body [ 10 ]. Macronutrients (i.e., carbohydrates, proteins, and fats) provide the energy necessary for the cellular processes required for daily functioning [ 11 ]. Micronutrients (i.e., vitamins and minerals) are required in comparatively small amounts for normal growth, development, metabolism, and physiologic functioning [ 12 , 13 ].

Carbohydrates are the primary source of energy in the diet and are found in the greatest abundance in grains, fruits, legumes, and vegetables [ 14 ]. In terms of deriving a health benefit, whole grains are preferred over processed grains, the latter having been stripped of germ and bran during the milling process, resulting in lower amounts of fiber and micronutrients [ 15 ]. Meta-analyses of prospective cohort studies have linked increased whole-grain intake to a reduced risk of coronary heart disease, stroke, cardiovascular disease, and cancer, as well as to the decreased risk of mortality due to any cause, cardiovascular disease, cancer, respiratory disease, diabetes, and infectious disease [ 15 , 16 , 17 ]. Fresh fruits and vegetables supply energy as well as dietary fiber, which promotes the feeling of satiety and has positive effects on gastrointestinal function, cholesterol levels, and glycemic control [ 18 ]. In addition, fresh fruits and vegetables are key sources of phytochemicals (e.g., polyphenols, phytosterols, carotenoids), which are bioactive compounds believed to confer many of the health benefits associated with fruit and vegetable consumption [ 19 ]. The mechanistic effects of these various phytochemicals are unclear but include their antioxidative properties, as well as their role in regulating nuclear transcription factors, fat metabolism, and inflammatory mediators. For example, flavonoids have been shown to increase insulin secretion and reduce insulin resistance, suggesting that these phytochemicals provide some benefits in obesity and diabetes [ 20 ]. Additionally, polyphenols interact with gastrointestinal microbiota in a bi-directional manner by enhancing gut bacteria and being metabolized by these bacteria to form more bioactive compounds [ 20 ]. Fruit and vegetable intake has been shown to inversely correlate with the risk of NCDs, including hypertension [ 21 ], cardiovascular disease [ 22 , 23 ], chronic obstructive pulmonary disease [ 24 ], lung cancer [ 25 ], and metabolic syndrome [ 26 ].

Dietary proteins provide a source of energy as well as amino acids, including those that the human body requires but cannot produce on its own (i.e., essential amino acids). Dietary proteins are derived from both animal (meat, dairy, fish, and eggs) and plant (legumes, soya products, grains, nuts, and seeds) sources, with the former considered a richer source due to the array of amino acids, high digestibility, and greater bioavailability [ 27 ]. However, animal-based sources of protein contain saturated fatty acids, which have been linked to cardiovascular disease, dyslipidemia, and certain cancers. Although the mechanisms are unclear, red meat, and processed meat in particular, have been associated with an increased risk of colorectal cancer [ 28 , 29 ]. Animal-derived proteins also increase the dietary acid load, tipping the body’s acid-base balance toward acidosis [ 30 , 31 ]. The increased metabolic acid load has been linked to insulin resistance, impaired glucose homeostasis, and the development of urinary calcium stones [ 30 , 31 ].

Adequate dietary protein intake is important for maintaining lean body mass throughout the life span. In older adults, protein plays an important role in preventing age-related loss of skeletal muscle mass [ 32 ], preserving bone mass, and reducing fracture risk [ 33 ]. For older individuals not obtaining adequate protein from their diets, supplementation with amino acids can improve strength and functional status [ 34 ].

Fats (or lipids) are the primary structural components of cellular membranes and are also sources of cellular energy [ 35 ]. Dietary fats fall into 4 categories: monounsaturated fats, polyunsaturated fats, saturated fats, and trans fats. The fat content of food is generally an admixture of these different types [ 35 ]. Unsaturated fats are found in a variety of foods, including fish, many plant-derived oils, nuts, and seeds, whereas animal products (and some plant-derived oils) contribute a larger proportion of saturated fats [ 35 , 36 ]. Trans fats found in foods are predominantly the result of processing vegetable oils but are also present in small quantities in animal products (i.e., ruminant trans fats from cows, sheep, and goats) [ 35 , 36 ]. Among the types of dietary fats, unsaturated fats are associated with reduced cardiovascular and mortality risks, whereas trans fats and, to a lesser degree, saturated fats are associated with negative impacts on health, including increased mortality risk [ 36 , 37 ]. Two families of polyunsaturated fatty acids, omega-3 and omega-6, are described as essential fatty acids, because they are required for normal growth and reproduction but are not produced by the body and, therefore, must be obtained from dietary sources [ 10 ]. Omega-3 fatty acids, in particular, eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA), have been widely studied for their potential health benefits, with evidence suggesting positive effects including cardioprotection, preventing cognitive decline, reducing inflammation, sustaining muscle mass, and improving systemic insulin resistance [ 38 , 39 , 40 ]. Seafood, especially oily fish, provides EPA and DHA, and supplements are widely available for those not meeting recommended intakes with diet alone [ 41 , 42 ]. Nuts and some seeds and plant oils provide alpha-linolenic acid, the major plant omega-3 fatty acid [ 43 ].

Although required in trace amounts compared with macronutrients, micronutrients are necessary for normal growth, metabolism, physiologic functioning, and cellular integrity [ 12 , 13 ]. The shift from whole foods to processed, refined foods has reduced the micronutrient quality of the modern Western diet [ 44 ]. Vitamin and mineral inadequacies have been implicated in cellular aging and late-onset disease, as scarcity drives chronic metabolic disruption. Keeping with these observations, adequate dietary intake of, or supplementation with, micronutrients that have antioxidant properties (e.g., vitamins A, C, and E, copper, zinc, and selenium) has been suggested as a means to reduce the risk for and progression of age-related diseases [ 45 ].

Water is the principal component of the body, constituting the majority of lean body mass and total body weight [ 13 ]. Water not only provides hydration but also carries micronutrients, including trace elements and electrolytes [ 46 , 47 ]. Drinking water may supply as much as 20% of the daily recommended intake of calcium and magnesium [ 47 ]. Our understanding of water requirements and water’s effect on health and disease is limited, although the global increase in intake of high-calorie beverages has refocused attention on the importance of water for maintaining health and preventing disease [ 46 ].

3. Common Health-Promoting Dietary Patterns

Based on our understanding of nutritional requirements and their likely health impacts as described above, healthy dietary patterns can be generally described as those that are rich in health-promoting foods, including plant-based foods, fresh fruits and vegetables, antioxidants, soya, nuts, and sources of omega-3 fatty acids, and low in saturated fats and trans fats, animal-derived proteins, and added/refined sugars [ 48 ]. Patterns such as these are naturally occurring in certain regions of the world and rooted in local/regional tradition and food sources, as is the case for the traditional Mediterranean and Asian diets. Healthy dietary patterns have also been developed based on studies of nutrient intake and subsequent health measures or outcomes (e.g., the DASH [ 8 ] and MIND [ 9 ] diets) that share some common characteristics ( Figure 1 ).

A generalized healthy diet and lifestyle pyramid.

3.1. Mediterranean Diet

The Mediterranean diet is based on components of the traditional dietary patterns of Euro-Mediterranean countries and encompasses not only the types of foods consumed and their relative contributions to daily nutrient intake, but also an approach to eating that is cognizant of how foods are sourced (e.g., sustainability and eco-friendliness), cooked, and eaten, as well as lifestyle considerations such as engaging in regular physical activity, getting adequate rest, and participating in fellowship when preparing and sharing meals [ 7 ]. Within the core framework of the Mediterranean diet, variations based on geography and culture are reflected in the emphasis on the inclusion of traditional and local food products. The primary basis of daily meals in the Mediterranean diet is cereals such as whole-grain bread, pastas, couscous, and other unrefined grains that are rich in fiber and a variety of fruits and vegetables of different colors and textures that are high in micronutrients, fiber, and phytochemicals ( Table 1 ) [ 7 , 9 , 49 , 50 , 51 , 52 ]. Dairy products, preferably low-fat yogurt, cheese, or other fermented dairy products, are recommended daily in moderation as a source of calcium, which is needed for bone and heart health. Olive oil serves as the primary source of dietary lipids and is supplemented with olives, nuts, and seeds. Water (1.5–2.0 L/day or ~8 glasses) is recommended as the main source of hydration, whereas wine and other fermented alcoholic beverages are generally permitted in moderation, to be consumed with meals. Fish, white meat, and eggs are the primary sources of protein; red meat and processed meats are consumed less frequently and in smaller portions. Legumes are also a preferred source of plant-based proteins [ 7 ].

Comparison of nutritional/lifestyle components among different healthy diet options.

a Recommendations shown here are based on a 2000 calorie per day eating plan. b Contribution of total fat and quality of fat from cheese to stay within the recommended daily intake.

The health benefits of the Mediterranean diet were first described in 1975 by Ancel Keys, who observed a reduction in cardiovascular disease risk among populations whose nutritional model was consistent with practices of peoples from the Mediterranean Basin [ 53 ]. Since that time, research has revealed beneficial effects of the Mediterranean diet on a number of NCDs and related health measures, including cardiovascular and cerebrovascular disease [ 54 ], cancer [ 55 ], glycemic control [ 56 ], and cognitive function [ 57 , 58 ]. Although publication of a key intervention study (Prevención con Dieta Mediterránea; PREDIMED) conducted at multiple sites across Spain and evaluating the Mediterranean diet for the primary prevention of cardiovascular disease was retracted due to irregularities in randomization [ 59 ], a subsequent analysis adjusting for these issues reported a consistent positive effect of adhering to a Mediterranean diet supplemented with olive oil or nuts compared with a reduced-fat diet [ 59 ]. Substudies of PREDIMED have also shown that, compared with a low-fat control diet, the Mediterranean diet supplemented with olive oil or nuts is associated with a 30% reduced risk of major cardiovascular risk events [ 59 ] and reductions in systolic blood pressure (SBP) and diastolic blood pressure (DBP) of 5.8–7.3 mmHg and 3.3–3.4 mmHg, respectively [ 60 ]. In addition, cardiovascular factors such as mean internal carotid artery intima-media thickness (−0.084 mm; p < 0.05) and maximum plaque height (−0.091 mm; p < 0.05) are improved with the Mediterranean diet supplemented with nuts [ 61 ]. Greater intake of polyphenols (phytochemicals found in fruits, vegetables, tea, olive oil, and wine) correlated with a 36% reduced risk of hypertension ( p = 0.015) [ 62 ] and improvements in inflammatory biomarkers related to atherosclerosis (i.e., interleukin [IL]-6, tumor necrosis factor-alpha, soluble intercellular adhesion molecule-1, vascular cell adhesion molecule-1, and monocyte chemotactic protein-1; p < 0.05 for each), as well as in high-density lipoprotein cholesterol (HDL-C; p = 0.004) [ 62 , 63 ].

3.2. Dietary Approaches to Stop Hypertension (DASH)

The DASH diet derives its name from the Dietary Approaches to Stop Hypertension study, which evaluated the influence of dietary patterns on blood pressure [ 8 ]. Patients who consumed a diet that was rich in fruits, vegetables, and low-fat dairy and that included a reduced amount of saturated and total fat and cholesterol experienced significantly greater reductions in blood pressure than patients who consumed a control diet that was similar in composition to a typical American diet (difference in SBP/DBP, −5.5/−3.0 mmHg; p < 0.001) or a diet rich in fruits and vegetables with a reduced amount of snacks and sweets (−2.7/−1.9 mmHg; p ≤ 0.002). All 3 diets had a sodium content of 3 g per day. A subsequent study (DASH-Sodium) that explored the DASH diet or a control diet in combination with varying levels of sodium intake (high, intermediate, and low) found that the DASH diet significantly reduced SBP during the high, intermediate, and low sodium intake phases of both diets (high: −5.9 mmHg; p < 0.001; intermediate: −5.0 mmHg; p < 0.001; low: −2.2 mmHg; p < 0.05) [ 64 ]. The DASH diet also significantly reduced DBP versus the control diet during the high (−2.9 mmHg; p < 0.001) and intermediate (−2.5 mmHg; p < 0.01) sodium intake phases but not during the low intake phase (−1.0 mmHg). Although reducing sodium intake also significantly reduced blood pressure in the control diet group ( p < 0.05), the low sodium phase of the DASH diet elicited significant decreases in SBP/DBP of −8.9/−4.5 mmHg ( p < 0.001 for each) compared with high sodium intake phase of the control diet.

Subsequent controlled trials, as a whole, support the results of the DASH and DASH-Sodium studies in terms of blood pressure reduction. Moreover, these studies expanded the positive impacts of the DASH diet to include improvements in other cardiovascular risk factors or comorbidities (e.g., low-density lipoprotein cholesterol [LDL-C], total cholesterol, overweight/obesity, and insulin sensitivity) [ 65 , 66 , 67 , 68 ] and reductions in adverse outcomes such as development of cardiovascular disease, coronary heart disease, stroke, heart failure, metabolic syndrome, and diabetes (including improved pregnancy outcomes in women with gestational diabetes) [ 68 , 69 , 70 , 71 , 72 ]. Meta-analyses of studies using the DASH diet have demonstrated that LDL-C is significantly reduced by −0.1 mmol/L ( p = 0.03) [ 65 , 68 ], total cholesterol by −0.2 mmol/L ( p < 0.001) [ 65 , 68 ], body weight by −1.42 kg ( p < 0.001) [ 66 , 68 ], and fasting insulin by −0.15 μU/mL ( p < 0.001) [ 65 , 66 , 67 , 68 ]. With the DASH diet, the risk of cardiovascular disease is reduced by 20%, stroke by 19%, and heart failure by 29% ( p < 0.001 for each) [ 69 , 71 ]. The overall risk of diabetes is reduced by 18% [ 68 ], and children and adolescents with higher DASH scores (i.e., those whose diets included the highest intakes of fruits, vegetables, nuts, legumes, low-fat dairy, and whole grains) were at 64% lower risk of developing metabolic syndrome than those with the lowest DASH scores ( p = 0.023) [ 71 ]. Furthermore, rates of cesarean section decreased by 47% [ 72 ], incidence of macrosomia (birth weight > 4000 g) decreased from 39% to 4% ( p = 0.002) [ 70 ], and significantly fewer women experienced gestational diabetes that required insulin therapy on the DASH diet (23%) compared with the control diet (73%; p < 0.0001) [ 70 ].

The dietary pattern derived from the DASH study emphasizes the consumption of an array of vegetables (including colorful varieties, legumes, and starchy vegetables), fruits, fat-free or low-fat dairy products, whole grains, and various protein sources (e.g., seafood, lean meats, eggs, legumes, nuts, seeds, and soya) ( Table 1 ) [ 49 ]. Limited consumption of added sugars (< 10% of calories per day), saturated fats (< 10% of calories per day), sodium (< 2300 mg/day), and alcohol (≤ 1 drink per day for women and ≤ 2 drinks per day for men) is suggested. In addition, further reductions in blood pressure may be achievable by further reducing sodium intake, although practical challenges may limit the ability to achieve sodium intake of 1200 mg or less per day [ 49 ].

3.3. Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND)

The MIND diet combines elements of the Mediterranean and DASH diets with the goal of sustaining cognitive health throughout older age [ 9 ]. Both the Mediterranean and DASH diets have been individually linked to positive cognitive outcomes, including the prevention of cognitive decline or impairment and better cognitive performance [ 73 , 74 , 75 ]. Two high-quality cohort studies have reported associations between adherence to the MIND diet and a 53% lower risk for developing Alzheimer’s disease ( p = 0.002 for linear trend) [ 50 ] and slower declines in cognitive functioning, both overall and within specific cognitive domains (e.g., episodic, semantic, and working memory and perceptual speed and organization), such that the highest adherence rates to the MIND diet were associated with cognitive function equivalent to being 7.5 years younger [ 50 , 76 ]. Interestingly, even modest adherence to the MIND diet was associated with a 35% risk reduction for Alzheimer’s disease versus the lowest adherence group ( p = 0.002 for linear trend), whereas high adherence was needed to demonstrate 54% and 39% risk reductions with the Mediterranean and DASH diets, respectively; high adherence to the Mediterranean and DASH diet showed a statistically significant benefit [ 50 ].

The MIND diet focuses on increasing the intake of fresh fruits and vegetables and emphasizes brain-healthy foods such as green leafy vegetables, nuts, berries, beans, whole grains, fish, poultry, olive oil, and wine in moderation ( Table 1 ) [ 9 , 50 ]. Additionally, foods that are thought to be unhealthy for the brain, such as red meats, butter/margarine, cheese, pastries, sweets, and fried or fast food, are limited [ 9 ]. The specificity regarding the types of foods on the healthy and unhealthy lists differentiates MIND from the Mediterranean or DASH diets [ 50 ].

3.4. Nordic Diet

Iterations of a Nordic diet (e.g., the healthy Nordic diet, New Nordic Diet) arose from the desire to translate the Mediterranean, DASH, and other health-promoting diets into a regionally tailored dietary pattern that uses traditional, local Nordic foods and would be attractive to the public, sustainable, and eco-friendly [ 77 , 78 ]. Overarching tenets of the New Nordic Diet are to consume more (1) calories from plant sources and fewer from animal sources, (2) foods from seas and lakes, and (3) foods from the wild countryside [ 78 , 79 ]. A generalized Nordic dietary pattern would include green leafy vegetables, other vegetables, fruits, fish and seafood, potatoes, berries, whole grains (e.g., wheat, rye, oats, barley), nuts, low-fat dairy products, rapeseed, sunflower, and/or soya oils and limited intake of fresh red meat and sugar [ 78 , 80 ]. Specific dietary recommendations based on the NORDIET clinical trial are presented in Table 1 [ 51 ].

The randomized, controlled NORDIET study compared a healthy Nordic diet with a control diet (the participant’s usual Western diet) [ 77 ]. Over 6 weeks, the Nordic diet improved the lipid profile (including a 0.98 mmol/L reduction in total cholesterol [ p < 0.0001] and a 0.83 mmol/L reduction in LDL-C [ p < 0.001]), lowered SBP by 6.6 mmHg ( p = 0.008), and improved insulin sensitivity (homeostatic model assessment-insulin resistance decreased 0.11; p = 0.01) compared with the control diet. Those on the Nordic diet also experienced a 3.0 kg decrease in body weight ( p < 0.001) despite food being available ad libitum.

Results from subsequent studies conducted using Nordic diet variations are consistent with those from studies with the NORDIET study, demonstrating improvements relative to the control diet in blood lipid profile (LDL-C/HDL-C ratio, −0.15; p = 0.046) [ 81 ], inflammation (IL-1 receptor antagonist, −84 ng/L; p < 0.001) [ 81 ], blood pressure (DBP, −4.4 mmHg ( p = 0.001), and mean arterial pressure (−4.2 mmHg; p = 0.006) among patients with metabolic syndrome [ 82 ] and weight loss (−3.22 kg; p < 0.001) [ 83 ] and blood pressure reduction (SBP/DBP, −5.13/−3.24 mmHg; p < 0.05) in individuals with obesity [ 83 ]. Compared with baseline values, one study demonstrated blood pressure reductions of −6.9 mmHg (SBP) and −3.2 mmHg (DBP; p < 0.01) [ 83 , 84 ]. Additionally, a study conducted in children reported an improvement in omega-3 fatty acid status with the Nordic diet that was associated with improvements in school performance ( p < 0.05) [ 85 ]. A systematic review parsing the individual components of the Nordic diet found that evidence supported the protective effects of eating whole grains on type 2 diabetes and cardiovascular disease risk, but that there was insufficient evidence for other foods in the Nordic diet [ 86 ].

3.5. Traditional Asian Diets

Although there is substantial evidence supporting the Mediterranean and other European-based diets, traditional regional dietary patterns from other parts of the world that follow similar principles have less–well-established links to positive health outcomes. A full description of the breadth of regional diets and the associated evidence bases is beyond the scope of this publication, but we consider some Asian-based diets to be particularly relevant to this discussion.

The traditional Korean diet is composed of rice and other whole grains, fermented food, indigenous land and sea vegetables, proteins primarily from legumes and fish as opposed to red meat, medicinal herbs (e.g., garlic, green onions, ginger), and sesame and perilla oils [ 87 ]. Meals typically consist of multiple small-portion dishes are often derived from seasonal food sources and are home-cooked. Unlike the Western diet, the traditional Korean diet does not include many fried foods [ 87 ]. Epidemiologic data suggest a reduced risk of metabolic syndrome (odds ratio [OR]: 0.77; 95% CI: 0.60–0.99), obesity (OR: 0.72; 95% CI: 0.55–0.95), hypertension (OR: 0.74; 95% CI: 0.57–0.98), and hypertriglyceridemia (OR: 0.76; 95% CI: 0.59–0.99) among individuals who follow traditional Korean dietary patterns [ 88 ]. These findings are consistent with a controlled clinical trial that explored the effects of a traditional Korean diet compared with a control diet (“eat as usual”) on cardiovascular risk factors in patients with diabetes and hypertension. In that study, adherence to a traditional Korean diet favorably influenced body composition (body weight, −2.3 kg; body mass index [BMI], −0.83 kg/m 2 ; body fat, −2.2%; p < 0.01), heart rate (−7.1 bpm; p = 0.002), and glycemic control (HbA1c, −0.72%; p = 0.003) [ 89 ].

The traditional Chinese diet features rice or noodles, soups, vegetables, steamed breads or dumplings, fruits and vegetables, soy, seafood, and meat [ 90 , 91 ]. Although higher in carbohydrates and lower in fat compared with a Western diet, the traditional Chinese diet does not appear to promote weight gain in healthy, normal-weight Chinese, suggesting that carbohydrate restriction may not be a universally applicable intervention to combat obesity and cardiometabolic risk [ 92 ]. One 6-week controlled trial demonstrated that 52% of non-Chinese individuals with overweight or obesity who adhered to a traditional Chinese diet had a reduction in BMI while preserving lean body mass compared with 28% of those who followed a Western diet at the 1-year follow-up assessment [ 93 ]. In another trial, BMI decreased by 0.37 kg/m 2 and lean mass by 0.21 kg among subjects who adhered to a traditional Chinese diet for 6 weeks, whereas those who followed a Western diet had 0.26 kg/m 2 and 0.49 kg reductions in BMI and lean body mass, respectively [ 94 ]. Notably, both of these studies restricted caloric intake to 1,200 Kcal for the test and control diet groups.

Similar to the Korean diet, the traditional Japanese diet (known as Washoku) is characterized by small portions of multiple components, primarily including rice, fish (often eaten raw), soups, and pickles [ 95 ]. Fermented soybean paste (dashi) serves as the base of many of the soups that are central to the traditional Japanese diet; other ingredients include seaweed, fruits and vegetables, and mushrooms. The use of chopsticks, alternating between dishes of small portion size throughout a meal, and the base flavor of Japanese food (umami) enhance satiety and help to prevent overeating. Adherence to a traditional Japanese dietary pattern has been associated with favorable effects on blood pressure among apparently healthy Japanese adults [ 96 ]. This is consistent with data from the 2012 Japan National Health and Nutrition Survey demonstrating that adherence to a traditional Japanese diet compared with a Western diet or a meat- and fat-based dietary pattern was associated with a lower prevalence of hypertension in men [ 97 ]. However, in the same study, a traditional Japanese diet was associated with higher DBP in women, as well as higher waist circumference and BMI in men. Further study is needed to elucidate the health impacts of traditional Japanese and other Asian dietary patterns.

4. Additional Factors

While the evidence reviewed here suggests that the described dietary patterns positively influence measures of health and disease risk and outcome because they encourage the intake of foods that individually have beneficial effects and the avoidance of unhealthy options, additional factors combine to create a lifestyle that promotes health. For example, healthy diets include adequate hydration, typically in the form of water or tea/herbal infusions [ 7 , 49 , 51 , 52 ]. In addition to the dietary components, a healthy lifestyle is one that incorporates regular exercise, socialization, and adequate sleep [ 7 , 52 ], and minimizes elements that have a negative effect on health such as tobacco use, excessive alcohol consumption, physical inactivity, large amounts of screen time, and stress.

The importance of non-dietary factors is reflected in their inclusion in modern food pyramids. Built on a base of positive lifestyle factors, the lower tiers indicate daily consumption of adequate hydration and nutrient-rich, plant-based foods, with animal-derived products (meat, fish, and dairy) and sweets comprising higher tiers of the pyramid (i.e., less frequently or infrequently consumed items).

Whereas the goal may be to achieve nutrient requirements through food and water intake alone, there are situations in which food-derived nutrient intake might be inadequate due to increased need, selective eating, or food insecurity/limited access to more nutritious foods [ 98 , 99 , 100 ]. Therefore, for some individuals, dietary supplements may be required, particularly at certain life phases. For example, later in life, the recommended intake of calcium increases to sustain bone mineral density [ 101 ]; hence, supplementation with calcium may be necessary to meet recommended intake levels in older adults. Before initiating supplementation, dietary intake levels should be considered to avoid exceeding the upper tolerability limits and causing adverse events.

There are a number of other traditional regional diets that likely have similar benefits to those that we describe here. However, we made the decision to narrow our focus to those diets with evidence from randomized, controlled trials demonstrating their health benefits. For example, the African Heritage Diet focuses on traditional ingredients that may be beneficial to African American populations who experience disproportionately higher risks for chronic diseases related to their diets [ 102 ]. Future research is warranted to evaluate the impact of the African Heritage Diet and other regional dietary patterns on health.

5. Conclusions

Healthy diets, arising either by tradition or design, share many common features and generally align with the WHO Global Action Plan for the Prevention and Control of Noncommunicable Diseases. In comparison with a Western diet, these healthier alternatives are higher in plant-based foods, including fresh fruits and vegetables, whole grains, legumes, seeds, and nuts and lower in animal-based foods, particularly fatty and processed meats. Evidence from epidemiologic studies and clinical trials indicates that these types of dietary patterns reduce risks of NCDs ranging from cardiovascular disease to cancer. Further endeavors are needed to integrate these healthy dietary and lifestyle choices into daily living in communities throughout the world and to make healthy eating accessible, achievable, and sustainable.

Acknowledgments

Medical writing support was provided by Crystal Murcia, PhD, and Dennis Stancavish, MA, of Peloton Advantage, LLC, an OPEN Health company, and was funded by Pfizer Consumer Healthcare. On 1 August 2019, Pfizer Consumer Healthcare became part of GSK Consumer Healthcare.

Author Contributions

H.C. and P.C.C. contributed to the conception of the work; the acquisition, analysis, and interpretation of data; drafting; and revision of the work. Both have approved the final version for submission and agree to be personally accountable for their contributions and for ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated, resolved, and documented in the literature. All authors have read and agreed to the published version of the manuscript.

Medical writing support was funded by Pfizer Consumer Healthcare; this research received no other external funding. The APC was funded by Pfizer Consumer Healthcare. On 1 August 2019, Pfizer Consumer Healthcare became part of GSK Consumer Healthcare.

Conflicts of Interest

Hellas Cena received travel reimbursement from Pfizer Consumer Healthcare to attend a discussion meeting prior to drafting the manuscript and acts as a consultant to companies that manufacture or market dietary supplements, including Pfizer Consumer Healthcare. Philip C. Calder received travel reimbursement from Pfizer Consumer Healthcare to attend a discussion meeting prior to drafting the manuscript. Pfizer Consumer Healthcare funded this project, but the company had no role in the design, execution, interpretation, or writing of the paper.

• Open access
• Published: 19 April 2021

A review of statistical methods for dietary pattern analysis

• Junkang Zhao 1 ,
• Zhiyao Li 1 ,
• Qian Gao 1 ,
• Haifeng Zhao 2 ,
• Shuting Chen 1 ,
• Lun Huang 1 ,
• Wenjie Wang 1 &
• Tong Wang   ORCID: orcid.org/0000-0002-9403-7167 1  

Nutrition Journal volume  20 , Article number:  37 ( 2021 ) Cite this article

27k Accesses

123 Citations

21 Altmetric

Metrics details

Dietary pattern analysis is a promising approach to understanding the complex relationship between diet and health. While many statistical methods exist, the literature predominantly focuses on classical methods such as dietary quality scores, principal component analysis, factor analysis, clustering analysis, and reduced rank regression. There are some emerging methods that have rarely or never been reviewed or discussed adequately.

This paper presents a landscape review of the existing statistical methods used to derive dietary patterns, especially the finite mixture model, treelet transform, data mining, least absolute shrinkage and selection operator and compositional data analysis, in terms of their underlying concepts, advantages and disadvantages, and available software and packages for implementation.

While all statistical methods for dietary pattern analysis have unique features and serve distinct purposes, emerging methods warrant more attention. However, future research is needed to evaluate these emerging methods’ performance in terms of reproducibility, validity, and ability to predict different outcomes.

Selection of the most appropriate method mainly depends on the research questions. As an evolving subject, there is always scope for deriving dietary patterns through new analytic methodologies.

Peer Review reports

Dietary intake, one of the essential factors that influence health, varies widely among individuals. The changes from the first Dietary Guidelines for Americans in 1980 to those in 2015 show that the focus of nutritional epidemiology has gradually shifted from single nutrients to dietary patterns, focusing on features of the entire diet [ 1 ]. There are several reasons for this shift [ 2 ]. First, each type of food contains multiple nutrients with complex interactions and latent cumulative relationships [ 3 , 4 ]. Hence, it is not feasible to isolate and examine their separate effects on diseases [ 2 ]. Additionally, it is difficult to analyze the role of individual foods because a typical diet is characterized by a mixture of different foods with substitution effects, where an increase in the consumption of some foods will lead to a decrease in the consumption of others [ 5 ]. If we include all collected food items in an analytical model simultaneously, multicollinearity, due to the complex interactions and relationships among them, will make inferences about individual foods difficult [ 6 ]. Due to the growing recognition of the complexity of dietary intake and its interactions with health outcomes, research on the health effects of dietary patterns is necessary alongside that of individual nutrients [ 7 ]. Dietary patterns consider the complex interrelationships between different foods or nutrients as a whole, reflect individuals’ actual dietary habits, and provide more information to indicate when many nutrients are associated with diseases [ 1 , 4 ]. Additionally, dietary patterns are more consistent over time and have a greater effect on health outcomes than individual nutrients [ 6 ]. Hence, dietary pattern analysis is considered a technology complementary to the study of single nutrients or food.

In the past few decades, statistical methods have emerged that make full use of dietary information collected across populations to create dietary patterns [ 2 , 4 , 8 ]. In nutritional epidemiology studies, regardless of the statistical method used for dietary pattern analysis, the goal is to explore the relationship between dietary patterns and health outcomes [ 2 , 3 ]. From this perspective, evaluating a method depends not only on whether the dietary patterns derived by the method comprehensively reflect the dietary preferences but also on whether these patterns can predict diseases more accurately and promote health.

The majority of published reviews divide the statistical methods for dietary pattern analysis into three categories: investigator-driven, data-driven, and hybrid methods widely used in nutritional epidemiology [ 2 , 3 , 8 , 9 , 10 ]. Additionally, several emerging methods have been applied to dietary pattern analyses that are less often or never reviewed adequately. To demonstrate these methods more clearly, we classify the emerging methods based on the existing categories and add a new category.

Since the finite mixture model (FMM) is a model-based clustering method and treelet transform (TT) combines principal component analysis (PCA) and clustering algorithms in a one-step process, they are classified as data-driven methods. Data mining (DM) and least absolute shrinkage and selection operator (LASSO) consider health outcome in identifying dietary patterns and are therefore classified as hybrid methods. Compositional data analysis (CODA)—the latest addition in dietary pattern research—identifies dietary patterns by transforming dietary intake into log-ratios and is thus categorized separately due to the particularity of suitable data.

This paper provides an updated landscape review of these methods based on the underlying concepts, strengths, limitations, and software packages commonly used while paying particular attention to emerging methods. The subsequent content is introduced from the following aspects: (1) investigator-driven methods, containing various dietary scores and dietary indexes; (2) data-driven methods, comprising PCA, factor analysis, traditional cluster analysis (TCA), FMM, and TT; (3) hybrid methods, consisting of reduced rank regression (RRR), DM, and LASSO; (4) compositional data analysis, including compositional principal component coordinates, balance coordinates and principal balances. To conclude, we compare and evaluate these methods, identify the remaining methodological issues, and provide suggestions for future research.

Investigator-driven methods

Investigator-driven methods are also called a priori approaches, and they include dietary scores and dietary indexes (collectively called dietary quality scores). These methods define dietary guidelines aligned with current nutritional knowledge or dietary recommendations that affect health as dietary patterns [ 9 ]. The foods or nutrients consumed by a person are scored based on some quality score (e.g., the Healthy Eating Index (HEI) shown in Table  1 ), and the results are summarized to produce dietary quality scores [ 12 , 13 ]. Dietary quality scores measure the extent to which individuals adhere to dietary guidelines and recommendations to assess the population’s overall dietary quality and predict diseases [ 9 , 13 ]. The classification of these scores is shown in Table  2 .

Recent studies on the relationship between dietary quality scores and health indicate that scores such as the HEI, Alternative Healthy Eating Index (AHEI) [ 15 ], Alternative Mediterranean Diet [ 35 ], and Dietary Approaches to Stop Hypertension (DASH) diet scores [ 27 ] are negatively correlated with the risk of death from cardiovascular disease, cancer, and all-cause mortality [ 36 , 37 , 38 , 39 , 40 ]. The last three dietary patterns were also recommended as easy and practical dietary plans for the public in the 2015 Dietary Guidelines for Americans [ 41 ]. Additionally, plant-based diets are receiving increasing attention because of their benefits to human health and environmental sustainability. Three plant-based diet indexes have been established in recent years: the total Plant-based Diet Index (PDI), Healthy Plant-based Diet Index (hPDI), and Unhealthy Plant-based Diet Index (uPDI) [ 42 , 43 ]. Unlike other dietary quality scores, these plant-based dietary indexes focus on the quality of plant foods in the diet; all animal foods, including those animal foods known to promote health, are negatively scored when calculating the plant-based dietary indexes [ 44 , 45 ]. Research has found that the higher the hPDI score, the lower the risk of coronary heart disease, type 2 diabetes, and all-cause mortality, whereas the uPDI shows the opposite trend [ 44 , 45 , 46 , 47 ].

Each component is individually scored and summed into a total score in the different scoring systems, but the total scores of different dietary quality scores vary greatly. Additionally, the total score can also be dichotomized but is less used [ 48 , 49 ]. No research has been done to establish the preferable scoring system for specific situations [ 12 ]. It is important to consider the research purpose when applying dietary quality scores and that there is not necessarily a single diet plan to follow to achieve a healthy dietary pattern [ 9 , 41 ].

The dietary guidelines and recommendations used to construct dietary quality scores are primarily based on scientific evidence from health and disease prevention studies. These scores can be used to describe overall dietary characteristics and repeat or compare results across populations. Many dietary quality scores have significant associations with disease and mortality outcomes. The total score is easy to understand and use, and the summing process is simpler than in other statistical methods for dietary pattern analysis.

Disadvantages

The construction of the scores, the definition of dietary diversity, and interpretation of the guidelines are generally determined subjectively by the researchers [ 2 ]. Additionally, dietary scores cannot describe overall dietary patterns because they focus only on selected aspects of diet and do not consider the correlation of different dietary components [ 2 , 13 ]. Finally, since a diet is usually multidimensional, the comprehensive dietary scores do not provide specific information on multiple foods, often leading to an unclear interpretation of intermediate scores. Individuals with a middle-range score likely have different nutritional compositions and dietary patterns [ 2 , 9 ].

Commonly available software and packages

No special program or package is required. Mainstream statistical analysis software, such as SAS, R, and STATA, are available.

Data-driven methods

In nutritional epidemiological studies, data-driven methods refer to the dietary intake patterns derived from population data through data dimensionality reduction techniques. These methods use the existing data collected from food frequency questionnaires, 24-h recall questionnaires, or dietary records to obtain dietary patterns instead of defined dietary guidelines [ 2 , 3 , 50 ].

Principal component analysis (PCA) and exploratory factor analysis (EFA)

PCA and EFA are the most commonly used methods in research on dietary patterns and, since they are based on similar mathematical concepts, they are discussed together in this section [ 3 ]. The PCA replaces a set of possibly correlating food groups with a new set of comprehensive indexes (principal components) that are uncorrelated and retain as much of the foods’ variance as possible. When deriving dietary patterns, it is common practice to pre-group food items before calculating principal components through the optimal weighted linear combination of food groups based on their correlation. Among all principal components, only a few that explain the most variation are retained for subsequent analysis. However, when the relationship between dietary patterns and demographic characteristics (e.g., age, income) is the focus, a posteriori exploratory analysis called Focused Principal Component Analysis (FPCA) can be applied [ 51 ]. The dietary patterns derived by FPCA are based on socioeconomic variables of interest and presented as concentric circles, where the center of the circle is a variable of interest. The distribution of different food group variables in the circle represents positive or negative correlations with the socioeconomic variable of interest in different colors or patterns. The smaller the radius, the stronger the correlation. The FPCA visualizes not only the relationship between the diet and a variable of interest but also the correlation between different food groups [ 51 ]. Like PCA, EFA reduces the dimensionality of food groups to a few factors with minimal loss of information. It decomposes each food group into common factors and a special factor: common factors are shared by all food groups, and special factors are unique to each food group. Each common factor represents a dietary pattern.

When determining the number of principal components or factors to be retained, the three selection criteria that are typically used include 1) retaining factors with an eigenvalue greater than one, 2) the scree plot, and 3) the interpretable variance percentage [ 8 ]. The correlation coefficients between the principal component and the food groups are called factor loadings, and they reflect the importance of the food groups. The greater the absolute value of the factor loadings, the stronger is the correlation between the corresponding food groups and the principal components or factors. Therefore, the principal components or factors are named primarily based on the food groups retained by the selection criteria applied to the factor loadings. Owing to the similarity between PCA and EFA [ 10 ], only PCA is shown in Fig.  1 .

The principal component analysis with D food group variables. Each PC is a linear combination of D food groups and corresponds to a dietary pattern

Unlike EFA, confirmatory factor analysis (CFA) is seldom used in nutritional epidemiology [ 52 ]. However, CFA can impose statistical tests on the factor structure and factor loadings of food groups and determine the number of factors and food groups contributing significantly to those factors [ 2 , 8 ]. In the past, CFA was applied as a second step to verify the goodness of fit and reproducibility of the factor structure of dietary patterns after PCA or EFA in the first step [ 9 , 53 , 54 ]. However, it remains uncertain whether the results are better than those obtained only with EFA [ 54 ]. Therefore, several studies have used CFA as a one-step approach to replace PCA or EFA [ 52 , 55 ]. The advantage of CFA is that a latent variable model can be specified and tested, and additional priori knowledge can also be incorporated into the model [ 55 ].

These methods describe the population’s variation in dietary intake and evaluate the overall quality of the diet. The resulting unrelated patterns capture the different dietary traits in the population and can be used directly as covariates to construct statistical models with health outcomes. Thus, they are more interpretable and meaningful than traditional methods that use a single nutrient or food. Moreover, some studies have found that several major dietary patterns derived by these methods show some reproducibility in different populations [ 56 , 57 , 58 , 59 ].

These methods have subjectivity in selecting food groups, determining the number of principal components or factors, selecting which foods have large factor loadings, and the patterns’ nomenclature. In classic PCA and EFA, each principal component or factor is a linear combination of all the food groups, which creates interpretive difficulties. The extracted dietary pattern can only explain part of the total variance of the food groups; therefore, it only represents the optimal model related to the explainable variance. Although other patterns may provide important information, they may not be retained by the selection criteria, and thus this important information is ignored [ 60 ]. In response to the question, “Which dietary patterns have the most predictive capability of a disease?” both PCA and EFA are unable to give an accurate answer. Additionally, FPCA can only determine the correlation between one lifestyle and dietary patterns, but dietary patterns may have strong interactions with many lifestyle characteristics simultaneously, and it is difficult to separate dietary pattern effects from other lifestyle effects [ 61 , 62 ].

The “proc princomp” and “proc factor” commands in SAS. The “survival” and “psych” packages in R. The “pca” and “factor” commands in STATA. SPSS.

Clustering methods

In PCA and EFA, the food items collected are pre-grouped to the extent that they are correlated with one another, and each person receives a score for each dietary pattern. Therefore, these methods can help us understand which foods are eaten simultaneously among the population and the relationships between dietary patterns and health outcomes. Both PCA and EFA are considered methods for “clustering” the food groups [ 10 ]. However, clustering methods can classify individuals into different groups based on their characteristics [ 63 ]. The dietary differences of individuals among different groups can be compared, and the characteristics of dietary patterns can be described by calculating the average intake level of different food groups within each group. Groups can also be compared with a specified control group to explore the risk of disease outcomes in different groups. In the study of dietary patterns, the clustering methods are summarized in the following two categories.

Traditional cluster analysis (TCA)

In nutrition research, TCA is based on the use of individual dietary characteristics to separate people into mutually exclusive clusters. One cluster represents a dietary pattern, with the individuals only belonging to one cluster [ 10 ], which is also called “hard” clustering. Before clustering, all the selected dietary variables (nutrients, food, or both) must be standardized to prevent variables with large variances from disproportionately affecting the clustering results [ 8 ]. The analyst needs to select the measure of similarity in individual dietary intakes, such as the Euclidean distance, Mahalanobis distance, and similarity coefficient, of individual dietary intakes. Clustering algorithms are then used to place similar individuals into the same category, and dissimilar individuals are dispersed as far as possible [ 10 ]. There are many clustering algorithms in TCA; three are commonly applied in dietary pattern analysis: k-means clustering, Ward’s minimum-variance method, and flexible-beta clustering [ 2 , 64 ]. Figure  2 shows the main principles of TCA using k-means clustering as an example for comparison with FMM.

The k-means clustering with n individuals and g clusters. The individuals with similar dietary characteristics are assigned to one cluster

The k-means clustering algorithm is the most commonly used algorithm [ 65 ]. It has the advantages of low computation complexity, fast calculation speeds, and suitability for large samples. However, the k value often needs to be pre-specified by the researcher. Ward’s minimum-variance method is a hierarchical clustering algorithm, and all of the calculations required for the clustering process occur at once [ 10 ]. Even if the number of clusters changes, recalculation is not required. However, the calculation is complex and slow, making this method unsuitable for large samples [ 66 ]. The flexible-beta clustering algorithm is an agglomerative hierarchical clustering algorithm with a specified parameter and robust results [ 64 , 67 ]. This algorithm introduces a new parameter β in the distance formula, for which the selected values are usually − 0.25 and − 0.50 [ 67 ]. However, there are only a few examples applying this method to the analysis of dietary patterns.

There is no singular method for identifying the number of clusters or an appropriate clustering algorithm [ 68 , 69 ]. One approach is to combine several methods, that is, based on factor analysis, the appropriate k value and a reasonable initial cluster center are identified by hierarchical clustering to minimize the influence of subjective judgment on the clustering results [ 68 , 70 ]. The other approach is the optimal clustering method, in which several different k values are tried, and quantitative indicators for these k values are compared to select the optimal value of k [ 8 , 71 ]. The selection of the clustering algorithm mainly depends on the stability of the clusters and their reproducibility, which are often evaluated by the split-half cross-validation method or classifier [ 64 , 72 ]. The most appropriate clustering algorithm is the one with the highest reproducibility and stability.

Distinct subgroups of individuals can be identified according to their dietary characteristics, and everyone belongs only to one specific dietary pattern group. Thus, the relationship between dietary pattern subgroups and health outcomes or other characteristics can be examined, and the subgroup at nutritional risk can also be identified. The results are also highly intuitive, and a dendrogram can be drawn to show the clustering process and results visually.

There are, however, a few drawbacks: first, each individual is assigned a cluster with a probability of 1 or 0, without considering the uncertainty of individual classification [ 73 ]. Second, the researcher is required to make several subjective decisions, such as the selection of the food groupings, clustering algorithms to determine the similarity of individuals, initial values, and the number of clusters. Although some relatively objective methods for selecting clustering algorithms and the number of clusters exist, the reproducibility of results cannot ensure their validity [ 64 ]. Third, there is no convenient method for comparing different clustering criteria [ 74 ]. Finally, the use of a control group and the unequal sample size of different clusters will limit the power of the statistical analysis [ 75 ].

The “proc cluster” command in SAS. The “psych” packages in R. The “cluster”, “clustermat” and “cluster kmeans” commands in STATA. SPSS.

The finite mixture model (FMM)

The FMM is a clustering method based on a latent variable model [ 73 , 76 ]. It measures classification uncertainty by calculating a posterior probability of different clusters based on given data; it is also called “soft” clustering [ 73 , 74 ]. The FMM assumes that the observed dietary data will be decomposed into a mixture distribution representing a finite sum of different food consumption probability distributions. Each distribution represents an unobserved cluster corresponding to a dietary pattern [ 73 ]. Through FMM, each individual’s posterior probability is calculated for each cluster; the individual is then assigned to the cluster with the highest posterior probability (Fig.  3 ). The posterior probability can measure the uncertainty of assigning individuals to different clusters. The process is similar to a k-means algorithm, but the probability of each individual assigned to each cluster is used for classification.

The finite mixture model with n individuals and g clusters. Each individual is only assigned to the cluster with the highest probability

Because FMM has many parameters, large samples are required. Thus, a restricted mixture model is proposed that reduces the number of parameters and is suitable for small- to moderately-sized samples [ 77 ]. The FMM method can also be used to classify the population according to the factor scores from factor analysis, also called a two-step classification, combining the advantages of both [ 76 ]..

The choice of k values or models can be transformed into a statistical model selection problem. The final model is then identified according to the maximum Bayes Information Standard after the FMM is fitted by setting different k values or imposing different restrictions on covariance matrixes [ 78 ]. The FMM is more flexible than TCA as it can account for the within-class correlation between variables [ 63 ], allow the variances of food consumption frequencies to vary within and between clusters, and enable covariate adjustment for food intake (e.g., energy intake and age) simultaneously with the fitting process [ 74 , 77 ].

The observed data may violate the distribution hypothesis, especially when there are many zero values so that the flexibility of the FMM cannot be fully realized. Although there are some common methods for dealing with zero values, the need to deal with zero values increases the model’s complexity, as does the high number of parameters to be estimated [ 63 ]. Its algorithm for estimating parameters still has flaws such as sensitivity to the initial value, convergence to local extremum, and slow convergence speed.

The “flexmix” and “mclust” packages in R. The “proc fmm” and “proc lca” commands in SAS. The “fmm” and “gllamm” commands in STAT A. Latent GOLD. Mplus.

The Treelet transform (TT)

Both PCA and FA are the most popular methods for identifying dietary patterns, but their qualitative interpretation is difficult and requires subjective judgment [ 79 ]. Additionally, cluster analysis fails to give numeric summary variables like factors or components. To overcome these limitations, the TT was developed to simplify the explanation of the factors while at the same time combining the advantages of PCA and the hierarchical clustering algorithm [ 79 , 80 ].

Like PCA, TT produces a set of factors based on the food groups’ covariance or correlation matrix and introduces the sparsity hypothesis into the factor loadings. Consequently, only a few of the factor loadings of the food variable are non-zero, and others are all zero [ 79 , 80 ], simplifying the explanation of factors. In nutrition epidemiology, the sparsity hypothesis holds if some foods are consumed independently of the foods included in the dietary patterns, or there is no variation in the population [ 81 ]. In the first layer of the cluster tree, the method identifies the two variables with the highest correlation among all the food groups and performs a PCA to produce two factors. The first factor is called the sum variable representing the weighted average of the largest variance, and the second factor is called the difference variable representing the orthogonal residual factor. Only the sum variable is retained in the cluster tree to repeat the algorithm above until each food variable is included in the cluster tree (Fig.  4 ).

A cluster tree produced by the treelet transform with five food group variables. As the dashed line goes up, the cutting level moves away from the root, so the factor loadings become more sparse

After the cluster tree is built, it is “cut” at a given level to produce a high variance factor describing the relevant food groups. Unlike PCA, TT requires a researcher to cut the cluster tree at a given level and then extract the factors based on the factor variance at that level. After the retained number of factor k is determined, the optimal cut level is identified by 10 cross-validations [ 79 , 80 ]. When the cutting level increases, the optimal cutting level corresponds to the inflection point when the cross-validation score (i.e., the mean of the k-factor variance sum) is no longer increased [ 79 , 80 ]. Additionally, the TT analysis is repeated at ±3 levels of optimal cut levels to evaluate the sensitivity of different cut levels [ 80 ].

Like PCA, the TT produces a set of factors, but each factor involves only a small percentage of food groups that simplify dietary patterns. When sample sizes are small, and the data are sparse with unknown groupings of correlated or collinear variables, TT is remarkably suitable for dimension reduction and feature selection before regression and classification [ 80 ]. Moreover, TT visualizes the results by constructing a hierarchical clustering tree for all variables, making the final results easily interpretable.

Choosing the cutting level of the cluster tree before extracting factors requires subjective judgment. When the cutting level is close to the root, more variables are contained in the factors, and the difficulty of interpretation also increases. As the cutting level gradually moves away from the root, the factor loadings become sparse, and the factors become easily interpretable; however, the diet’s complexity cannot be reflected by some food groups [ 82 ]. If food groups are all associated in a meaningful way, or the correlation of some foods is too strong, then the sparsity hypothesis may not hold [ 81 ]. Additionally, it remains debatable whether TT is superior to other methods in exploring the relationship between diet and health outcomes [ 79 , 83 ].

The “treelet” package in R. The “tt” commands in STATA.

Hybrid methods

Investigator-driven methods are hypothesis-oriented approaches, which neither reflect the overall dietary patterns nor consider the relevant relational structure of nutrients. In addition, data-driven methods do not consider any priori professional knowledge on health outcomes; therefore, both methods are nonoptimal for identifying which dietary patterns can best predict disease risk [ 84 ]. Hybrid methods combine these two classes of methods to identify dietary patterns.

Reduced rank regression (RRR)

The RRR method considers both the disease-relevant variation in dietary intake and available dietary data in deriving dietary patterns [ 85 , 86 ]. Specifically, RRR selects a set of disease-related variables, known as intermediate response variables, based on priori knowledge, then derives dietary patterns based on the existing dietary data [ 85 ]. Its mathematical foundation and method of deriving dietary patterns are similar to those of PCA. However, unlike PCA, which explains as much variance in food groups as possible, RRR identifies linear combinations of food groups that can explain the maximum variance in intermediate response variables (Fig.  5 ). Both RRR and PCA produce components, which are based on the number of food variables and response variables, respectively. Therefore, RRR can be considered a PCA of intermediate response variables. The key to RRR is the choice of intermediate response variables, which should be related to both the disease of interest and the diet. The commonly used response variables include nutrients, biomarkers, contaminants, and intermediate phenotypes, or a combination of several kinds of them, in which nutrients and biomarkers are the most widely used [ 84 ].

The reduced rank regression with D food group variables and g intermediate response variables (M). Each PC corresponding to a dietary pattern is a linear combination of D food groups which explaining as much variance (V max ) in M as possible. D is larger than g

A method similar to RRR is partial least squares (PLS), a regression model of multiple predictor variables on multiple response variables [ 85 ]. The PLS method uses the covariance matrix of multiple intermediate response variables and multiple food groups to produce the factors; it is regarded as a compromise between PCA and RRR [ 85 , 87 , 88 ]. It not only contains information about intermediate response variables but also enables the discovery of important disease-related dietary intake, in which some nutrients may not be included in the intermediate response variables [ 87 ].

RRR uses both priori information for defining appropriate intermediate response variables and the existing data. Thus, it combines the respective characteristics of investigator- and data-driven methods. This method includes the pathophysiological pathway linking dietary patterns with the disease [ 89 ]; therefore, the correlation between dietary patterns and disease outcomes may be more robust in RRR than in other methods, and the importance of dietary patterns in the etiology of diseases can be better studied [ 9 ]. The effect of dietary patterns on disease risk can be described and explained by changes in biologically important intermediate variables [ 8 ]. The relationship between dietary patterns and diseases of interest can be reproduced across studies [ 50 , 84 ].

The underlying disease development mechanisms need to be identified, as they are the effective intermediate response variables. If the information for disease development is absent, then RRR cannot be used [ 9 , 90 ]. Additionally, there is no best way to choose the most appropriate intermediate response variables, and the commonly used method is based on priori information [ 8 ]. For many chronic diseases, complex interactions in metabolic pathways can link dietary intake to disease, but it is unclear whether the biomarkers of one metabolic pathway used in RRR are more effective than other potential metabolic pathways. Additionally, relying solely on the information of selected intermediate response variables to derive dietary patterns may lead to the omission of those dietary patterns related to nutrients in the disease’s biological pathways but are not included in the intermediate response variables [ 91 ].

The “proc pls” commands in SAS. The “rrr” and “rrpack” packages in R. The “rrr” commands in STATA.

Data mining (DM)

DM can extract hidden information from large databases, allowing researchers to focus on the most important information in the data [ 92 ]. This method uses various data analysis tools to derive dietary patterns and help researchers make decisions [ 93 , 94 ]. As one of the most important classification tools in DM, decision tree induction can be regarded as a clustering algorithm that makes full use of interesting health outcome. There have only been a few studies using this method in nutritional epidemiology until now [ 9 , 94 , 95 , 96 ].

Decision tree induction is also known as a classification and regression tree [ 9 ]. The main idea is to build a decision tree through a set of known training data and then use the established decision tree to predict new data sets. Establishing a decision tree can be regarded as the process of generating data rules, and the most classic algorithm is C4.5 [ 97 ]. This algorithm first pre-processes the selected food group variables by discretizing variables (e.g., expressing them as the frequency of food consumption). The classification result of interest is the health outcome. Then a “best” food group is selected as the root node of the decision tree and split according to its value to produce different subsets (“best” means that as far as possible, all individuals in the subset have the same outcome after splitting the data). The above procedures are then repeated on the subsets until the outcome of all individuals in each subset is the same. Each subset is called a leaf node, which constitutes the final decision tree (Fig.  6 ). A classification rule is a path from the root node to a leaf node associated with health outcomes. In the dietary study, the C4.5 algorithm needs to be run for all the combinations of different numbers of food groups to produce hundreds of classification rules. Repetitive and meaningless rules are deleted. The reserved rules correspond to dietary patterns. The intensity and direction of a food group’s association with diseases can be identified by comparing rules for which the only difference is the food group. Additionally, some other DM methods, such as random forest, artificial neural networks, and Naïve Bayes Classifiers, have also been used to analyze the relationship between dietary patterns and diseases [ 94 , 95 , 98 ], but they are all belongs to clustering algorithms and less common in nutritional epidemiology, so they are not introduced in more detail.

The decision tree generated by the C4.5 algorithm

When there is obvious heterogeneity in the dietary behavior of a population, DM can be used to reveal such heterogeneity and develop personalized preventive measures; the extent to which dietary components or patterns affect the course of the disease can also be identified [ 93 ]. It is also particularly useful in identifying disease risk based on a combination of known food groups and other non-dietary confounders [ 9 ]. Lastly, decision tree analysis can generate new hypotheses without priori assumptions or potential risk factors [ 99 ].

If many classification rules are generated in the DM process, the selection of meaningful rules will require considerable professional knowledge. Rules containing many variables can be long and complex even if they are meaningful, making it difficult to translate them into simple health information. Additionally, one key variable can dominate the model; therefore, misclassification is more likely to occur with DM than with other methods [ 94 ].

The “proc split” and “proc hpsplit” commands in SAS and SAS/EM module. The “RWeka” and “rpart” packages in R. The “chaid” and “crtrees” commands in STATA. WEKA. SPSS.

Least absolute shrinkage and selection operator (LASSO)

The LASSO model is a regression-based method that penalizes the regression coefficients’ absolute value so that the coefficients in the overall regression are shrunk [ 100 ]. Under the constraint that the sum of the absolute values of the regression coefficients is less than a constant, the sum of the squares of the residuals is minimized to obtain a sparse model in which some regression coefficients are shrunk to 0 [ 100 ]. Lasso’s complexity is controlled by the model tuning parameter λ; the greater the λ, the greater is the penalization of the model, resulting in a model with fewer variables. The LASSO model is hence a form of automatic feature selection. While identifying the dietary patterns, LASSO is directly applied to the defined food groups to predict health outcome [ 101 ]. Different λ results in different numbers of food groups with a non-zero coefficient selected into the model (Fig.  7 ). Cross-validation is used to select λ, which forces some coefficients of the food groups to zero and, hence, selecting food groups with non-zero coefficients [ 101 ]. The λ parameter is determined by the rule of minimum mean cross-validation error or one standard error. The selected food groups are then regarded as the dietary pattern.

The least absolute shrinkage and selection operator. The number of points at which the dashed line intersects the curve represents the number of nonzero coefficients D. The smaller λ, the larger D

LASSO considers the outcome variable when deriving dietary patterns, thus achieving higher prediction accuracy. As a shrinkage method, the LASSO model selects for a subset of food groups to predict outcome that result in a more interpretable and relevant set of food groups.

LASSO is still less applied in dietary pattern analysis, so its validity and reproducibility need to be confirmed in future studies. In addition, whether LASSO is superior to other dietary pattern methods in exploring the relationship between diet and health outcomes is yet to be verified.

The “glmnet” package in R. The “lassopack” commands in STATA.

Compositional data analysis (CODA)

Usually, changes in one dietary component are accompanied by compensatory changes in others if the total energy intake is kept constant [ 5 ]. Therefore, dietary data can be regarded as compositional, that is, the data can also be referred to as compositional data [ 102 , 103 ]. Compositional data can be used in analyzing the relative importance of food consumption and have great potential in dietary pattern analysis [ 5 ].

In the case of compositional data, x is a positive vector of D parts ( x  = [ x 1 ,  x 2 , … x D ]) and usually is a closed-form expression (proportions or percentages). Every composition x i represents relative information that describes the parts of the whole. The mathematical difficulties inherent in compositional data have hampered their wider use [ 104 ]. Therefore, a method called compositional data analysis (CODA) [ 104 ] has been proposed; the method uses log-ratio coordinates to transform compositional data into a form that can be analyzed using standard multivariate statistical analysis. Owing to the compositions’ proportional nature, the only valid function of compositional data is composed of the ratio of different parts [ 102 , 104 ]. There are three widely used transformation methods for log-ratio coordinates: additive log-ratio (alr), centered log-ratio (clr), and isometric log-ratio (ilr) transformations. In alr, each of the first D-1 parts is divided by the final part, but the transformation is not orthogonal; therefore, the rationality of statistical analysis cannot be guaranteed. The clr method can solve this limitation by dividing each part by the geometric mean of the D parts [ 105 ]; however, the sum of those clr variables is zero, meaning that perfect collinearity exists [ 106 ], which can be solved by the ilr transformation [ 107 ]. The ilr transformation preserves the original mathematical properties and geometric features; therefore, the rationality of directly applying the classic statistical method is ensured. Compositional data analysis has been applied in health research only recently, and there is less research on the relationship between dietary patterns and health [ 5 ]. There are three approaches to building the ilr transformational variables for dietary pattern analysis: compositional principal component coordinates, balance coordinates, and principal balances (PBs).

Compositional principal component coordinates

Due to the constant sum and possible nonlinearity in compositional data, directly applying the traditional PCA will likely result in many problems [ 5 ]. Thus, Aitchison extended standard PCA to compositional data [ 105 ]. The main idea is that the standard PCA is applied to the clr transformed covariance matrix to extract the principal components called PC coordinates. It can be proved that PC coordinates satisfy all the ilr transformation conditions and are equivalent to ilr coordinates. The first few PC coordinates explaining the most variance in dietary intake can be used for studying the relationship between dietary patterns and health outcomes.

Balance coordinates

The use of PC coordinates can be regarded as a data-driven ilr transformation, but it can also be a priori-driven based on the researcher’s questions or interests. In epidemiology, a priori-driven ilr transformation is calculated mainly by easily explainable balance coordinates [ 103 , 108 ] representing the relationship between different groups of parts. To build balance coordinates, sequential binary partition (SBP) is used to divide the complete composition of D parts into two groups of parts successively in a hierarchical manner: one part for the numerator and the other for the denominator. Similarly, each of the two groups is again split into two new groups to create the new balance coordinate and so on until step D-1, when only a single part is left in each group. Then, D-1 different ilr balance coordinates are produced [ 108 ]. Each set of balance coordinates corresponds to a dietary pattern. Positive coordinates indicate that the numerator has a relatively high weight, and negative coordinates indicate that the denominator has a higher weight.

To enhance the interpretation of the analysis, SBP can be constructed based on the purpose of the study [ 102 , 109 ]. For example, if the research aims to extract dietary patterns, ilr balance coordinates can be constructed according to natural or artificial clustering of different foods or nutrients in groups. Thus, balance coordinates are not data-driven and mainly focus on the research questions, unlike hierarchical clustering analysis. Since the total variance of the complete composition is decomposed into D-1 parts and the balance coordinates are independent of each other, all D-1 balance coordinates must be included as explanatory variables in the model simultaneously [ 5 ]. Balance coordinates can be visualized through a tree diagram, called the CoDa-dendrogram or the balance dendrogram, which is also a tool for describing the whole process of SBP [ 5 , 110 ].

Principal balances

Principal balances are data-driven balance coordinates that can not only concentrate a large proportion of the total variance in a few coordinates but are also convenient for comparing groups of parts in the numerator and the denominator [ 109 ]. The first PB is the balance coordinate that maximizes the explained variance. The kth PB maximizes the remaining variance and is orthogonal to the previous k-1 PBs. All the PBs or the PBs with the highest variance can be used for subsequent analysis. In the CoDa-dendrogram, PBs are ordered by the variance of the balance, which are different from balance coordinates ordered by the sequence of the partitions [ 109 ]. A CoDa-dendrogram of PBs is shown in Fig.  8 .

CoDa-dendrogram of PBs with six food group variables. Each PB corresponds to a dietary pattern. The closer the contact point is to a food, the more of that food is relatively more abundant

The optimal algorithm of PBs is an exhaustive search of all possible SBPs [ 109 ]. If data are highly dimensional, the computer space and time required will be large and long, respectively, so that using PBs on a personal computer becomes difficult. At present, suboptimal but faster algorithms have been proposed to search for PBs, such as the new constrained PCs algorithm and Ward’s cluster method; the proposed methods produce PBs whose variances are slightly lower than those obtained by optimization algorithms and are more applicable for high-dimensional compositional data [ 109 ].

These three coordinates—compositional principal component coordinates, balance coordinates, and PBs—can extract compositional information in dietary patterns for further direct application of classic multivariate statistical methods. The results emphasize that any dietary pattern is a balance between different intakes of food. When the relationship between dietary patterns and health outcomes is studied, the results can be interpreted as the effects of increasing the intake of some foods and reducing the intake of other foods proportionally on health outcomes. Therefore, considering food intake as compositional data is more consistent with the intuitive concept of dietary patterns and the practice of dietary recommendations. The first few PC coordinates and PBs can explain a large proportion of dietary intake variation. Additionally, balance coordinates and PBs are more easily explainable than PC coordinates, and they can be depicted as a CoDa-dendrogram.

Like PCA, each PC coordinate contains all the food groups, which complicates the explanation of the results, and factor loadings need to be recalculated for each application to different data sets [ 5 , 105 ]. The balance coordinates are an investigator-driven method requiring sufficient priori knowledge to provide SBP, meaning that the subjectivity of SBP is inevitable. Especially when D is large enough, the selection of SBP will become difficult, or there will be more than one SBP. Additionally, most of the total variation cannot be explained by a few balance coordinates [ 5 ]. Finally, for a large number of zero values in the compositional data, especially absolute zero values, no method works well [ 111 ].

The “coda.base,” “compositions,” “robCompositions,” and “zCompositions” packages in R. Stand-alone programs such as “SparCC” and “CoDaPack.”

With the development of nutritional epidemiology over the past decades, there is extensive research on dietary patterns describing the features of dietary behavior or habits and explaining the relationship between diet and diseases [ 2 ]. Moreover, there is growing evidence that food-based dietary patterns are a better way of reducing cardiovascular disease, diabetes, and obesity than single dietary components, total fat, and calories [ 112 ]. Previous reviews have already introduced several classic methods of deriving dietary patterns, mainly focusing on dietary quality scores, PCA, FA, TCA, and RRR. However, other methods of identifying dietary patterns are rarely or never reviewed [ 2 , 3 , 7 , 9 , 10 ]. This paper provides an updated overview of the methodological aspects of various methods and briefly introduces their underlying concepts, advantages and disadvantages, and the software available for their implementation. These methods describe and explain potential complex eating behaviors from different perspectives. They aid researchers in studying the relationship between diet and diseases more comprehensively.

Dietary quality scores mainly aim to evaluate the quality of the overall diet and test the validity of dietary guidelines or recommendations [ 9 , 13 ]. While MDS, HEI, AHEI, and DASH are especially recommended to predict disease risk, only the Mediterranean diet has been proven to reduce disease risk in both observational studies and randomized controlled trials [ 6 , 41 ]. Data-driven methods are especially important for identifying the priorities of nutritional interventions and exploring the health effects of different dietary habits [ 9 ]. However, they are often criticized for not considering priori knowledge about diseases, so they are preferred methods for performing an explorative analysis [ 87 ]. Both PCA and FA capture the interrelation between dietary components by creating principal components or factors, but they are not easy to explain. The TT can be regarded as a complementary method to PCA because it produces similar scores, which are easily interpretable as the patterns have no contributions from some foods or food groups. Nevertheless, the assumption of such scenarios is often hard to verify, and sometimes the relationship between TT-derived dietary patterns and the disease is different from that of previous results [ 83 ], probably because not all foods are included in the score calculation, and the patterns fail to reflect the real complexity of diet intake. The main advantage of TCA is that it assigns each individual a specific dietary pattern subgroup, which is difficult for PCA, FA, and TT; thus, individualized dietary advice can be provided.

Another clustering method is FMM, which can calculate the probability of each individual assigned to each category, and the covariate adjustment is considered in the fitting process. However, it is still not as widely used as TCA, probably because of the requirements for distribution, the model’s complexity, and the need for more statistical expertise. Furthermore, FMM does not consistently give much better clustering results than the k-means algorithm at the cost of increasing model complexity [ 63 ]. None of these data-driven methods consider the health outcome when deriving dietary patterns and they are data- and population-specific; therefore, the results do not adequately explain the relationship between diet and diseases and have limited reproducibility.

The RRR method makes full use of a priori knowledge of biological relations to identify the dietary patterns with significant influence in the etiology of disease [ 85 , 113 ] and is particularly useful in deriving dietary patterns related to given diseases and is reproducible across populations [ 50 ]. However, its application is limited to only diseases with adequate priori knowledge of intermediate response variables. Unlike RRR, the DM and LASSO methods use only one outcome variable at a time to identify dietary patterns. However, DM divides individuals into distinct subgroups similar to clustering algorithms to predict outcomes. It can identify which subgroups are at risk of the disease and explore new patterns of various diet and non-diet combinations. The LASSO model uses food groups to predict outcomes directly instead of constructing new underlying variables or dividing individuals into mutually exclusive subgroups. It performs prediction and variable selection simultaneously to build a sparse model.

Dietary intake data can also be regarded as compositional data with varying total diet intake among individuals [ 5 , 114 , 115 ]. Additionally, metabolic dysfunction can be caused not only by a lack of nutrients but often by an imbalance between nutrients [ 114 ]. Although compositional data are not a new concept, they have only recently been applied to nutritional epidemiology [ 5 , 102 , 103 , 114 ]. In addition to being applied for dietary patterns, the CODA methods can also separate the specific effects of macronutrients from the generic effects of total calorie intake simultaneously [ 103 ]. Several new algorithms applying clustering methods (e.g., FMM and k-means clustering) or hybrid methods (e.g., RRR) to compositional data and compositional substitution models which will be possible to investigate specific food substitution have been proposed. However, they have not yet been applied in dietary pattern analysis [ 116 , 117 , 118 ].

Classical methods are useful in nutritional epidemiology, but we should not limit ourselves to them since emerging methods can provide improved results and new ideas to overcome the shortcomings and inapplicable problems of the classic methods under suitable scenarios. Therefore, emerging methods deserve more attention. Among them, CODA methods especially seem to hold great potential and promise for deriving dietary patterns and studying the relationship between diet intake and health outcomes differently. However, future research is needed to evaluate these emerging methods’ performance in terms of reproducibility, validity, and predicting different outcomes.

In summary, all methods of deriving dietary patterns can be used to answer different research questions. Hence, when conducting dietary pattern analysis, the first step is determining the problems to be solved and then selecting the appropriate method. If it is unclear which method is most suitable, combining multiple methods in the same study to produce complementary results and explanations is a good choice. However, there are many other problems that these methods cannot solve well, such as measurement errors (including large proportions of zeros), the interactions between dietary patterns and other non-dietary confounders, and the predictive effect of changes in dietary patterns on disease over time.

Some efforts have been made to address these problems. For example, some measurement error correction methods and new biomarkers of food intake have been developed for the measurement error [ 119 , 120 ]; EPCA, DM, and LASSO can be used to explore the correlations between different diet and any other non-dietary confounders [ 51 , 93 ]; and repeated measures of food intake in cohort studies can assess the changes in dietary patterns and provide stronger causality between food intake and disease [ 6 , 41 ]. Additionally, we may also need to learn methods from other disciplines, including substitution models in behavioral epidemiology, pattern recognition methods in mathematics and computer science, and decision-making and optimization methods in operations research [ 2 , 117 ]. Although increasing attention has been paid to dietary pattern research, it should be noted that dietary pattern research is not meant to replace single-nutrient research; the two types of research should coexist and complement each other.

We hope that this landscape review will help researchers in this field to understand and apply various methods effectively in practice and familiarize interested researchers outside the field with these methods. We also hope that methodological limitations will gain more attention and be improved to simulate new study ideas that may more accurately disclose the relationship between diet and health.

Availability of data and materials

Not applicable.

Kelly OJ, Gilman JC, Ilich JZ. Utilizing dietary micronutrient ratios in nutritional research may be more informative than focusing on single nutrients. Nutrients. 2018;10(1):107. https://doi.org/10.3390/nu10010107 .

Article   CAS   PubMed Central   Google Scholar  

Moeller SM, Reedy J, Millen AE, Dixon LB, Newby PK, Tucker KL, et al. Dietary patterns: challenges and opportunities in dietary patterns research an experimental biology workshop, April 1, 2006. J Am Diet Assoc. 2007;107(7):1233–9. https://doi.org/10.1016/j.jada.2007.03.014 .

Article   PubMed   Google Scholar  

Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev. 2004;62(5):177–203. https://doi.org/10.1111/j.1753-4887.2004.tb00040.x .

Article   CAS   PubMed   Google Scholar  

Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curropinlipidol. 2002;13(1):3–9.

CAS   Google Scholar  

Solans M, Coenders G, Marcos-Gragera R, Castelló A, Gràcia-Lavedan E, Benavente Y, et al. Compositional analysis of dietary patterns. Stat Methods Med Res. 2018;28(9):2834–47. https://doi.org/10.1177/0962280218790110 .

Schulze MB, Martínez-González MA, Fung TT, Lichtenstein AH, Forouhi NG. Food based dietary patterns and chronic disease prevention. Bmj. 2018;361:k2396.

Article   Google Scholar  

Jannasch F, Riordan F, Andersen LF, Schulze MB. Exploratory dietary patterns: a systematic review of methods applied in pan-European studies and of validation studies. Br J Nutr. 2018;120(6):601–11. https://doi.org/10.1017/S0007114518001800 .

Article   CAS   PubMed   PubMed Central   Google Scholar  

Michels KB, Schulze MB. Can dietary patterns help us detect diet-disease associations? Nutr Res Rev. 2005;18(2):241–8. https://doi.org/10.1079/NRR2005107 .

Ocké MC. Evaluation of methodologies for assessing the overall diet: dietary quality scores and dietary pattern analysis. Proc Nutr Soc. 2013;72(2):191–9. https://doi.org/10.1017/S0029665113000013 .

Gleason PM, Boushey CJ, Harris JE, Zoellner J. Publishing nutrition research: a review of multivariate techniques--part 3: data reduction methods. J Acad Nutr Diet. 2015;115(7):1072–82. https://doi.org/10.1016/j.jand.2015.03.011 .

Krebs-Smith SM, Pannucci TE, Subar AF, Kirkpatrick SI, Lerman JL, Tooze JA, et al. Update of the healthy eating index: HEI-2015. J Acad Nutr Diet. 2018;118(9):1591–602. https://doi.org/10.1016/j.jand.2018.05.021 .

Article   PubMed   PubMed Central   Google Scholar  

Trijsburg L, Talsma EF, de Vries JH, Kennedy G, Kuijsten A, Brouwer ID. Diet quality indices for research in low- and middle-income countries: a systematic review. Nutr Rev. 2019;77(8):515–40. https://doi.org/10.1093/nutrit/nuz017 .

Article   PubMed Central   Google Scholar  

Waijers PM, Feskens EJ, Ocke MC. A critical review of predefined diet quality scores. Br J Nutr. 2007;97(2):219–31. https://doi.org/10.1017/S0007114507250421 .

Haines PS, Siega-Riz AM, Popkin BM. The diet quality index revised: a measurement instrument for populations. J Am Diet Assoc. 1999;99(6):697–704. https://doi.org/10.1016/S0002-8223(99)00168-6 .

Chiuve SE, Fung TT, Rimm EB, Hu FB, Mccullough ML, Molin W, et al. Alternative dietary indices both strongly predict risk of chronic disease. J Nutr. 2012;142(6):1009–18. https://doi.org/10.3945/jn.111.157222 .

Kosti RI, Panagiotakos DB, Mariolis A, Zampelas A, Athanasopoulos P, Tountas Y. The Diet-Lifestyle Index evaluating the quality of eating and lifestyle behaviours in relation to the prevalence of overweight/obesity in adolescents. Int J Food Sci Nutr. 2009;60(sup3):34–47.

Kant AK, Schatzkin A, Graubard BI, Schairer C. A prospective study of diet quality and mortality in women. Jama. 2000;283(16):2109–15. https://doi.org/10.1001/jama.283.16.2109 .

Stampfer MJ, Hu FB, Manson JE, Rimm EB, Willett WC. Primary prevention of coronary heart disease in women through diet and lifestyle. N Engl J Med. 2000;343(1):16–22. https://doi.org/10.1056/NEJM200007063430103 .

Hu FB, Manson JE, Stampfer MJ, Colditz G, ., Liu S, ., Solomon CG, et al. Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N Engl J Med 2001;345(11):790–797, DOI: https://doi.org/10.1056/NEJMoa010492 .

Kant A. Indexes of overall diet quality: a review. J Am Diet Assoc. 1996;96(8):785–91. https://doi.org/10.1016/S0002-8223(96)00217-9 .

Murphy SP, Davis MA, Neuhaus JM, Lein D. Dietary quality and survival among middle-aged and older adults in the NHANES I epidemiologic follow-up study. Nutr Res. 1996;16(10):1641–50. https://doi.org/10.1016/0271-5317(96)00183-2 .

Nitin S, Steck SE, Hurley TG, Hussey JR, Hébert JR. Designing and developing a literature-derived, population-based dietary inflammatory index. Public Health Nutr. 2014;17(8):1689–96.

Martínez-González MA, Fernández-Jarne E, Serrano-Martínez M, Marti A, Martinez JA, Martín-Moreno JM. Mediterranean diet and reduction in the risk of a first acute myocardial infarction: an operational healthy dietary score. Eur J Nutr. 2002;41(4):153–60. https://doi.org/10.1007/s00394-002-0370-6 .

Monteagudo C, Mariscal-Arcas M, Rivas A, Lorenzo-Tovar ML, Tur JA, Olea-Serrano F. Proposal of a Mediterranean diet serving score. PLoS One. 2015;10(6):e0128594. https://doi.org/10.1371/journal.pone.0128594 .

Osler M, Heitmann BL, Gerdes LU, Jørgensen LM, Schroll M. Dietary patterns and mortality in Danish men and women: a prospective observational study. Br J Nutr. 2001;85(2):219–25. https://doi.org/10.1079/BJN2000240 .

Patterson RE, Haines PS, Popkin BM. Diet quality index: capturing a multidimensional behavior. J Am Diet Assoc. 1994;94(1):57–64. https://doi.org/10.1016/0002-8223(94)92042-7 .

Fung TT, Chiuve SE, McCullough ML, Rexrode KM, Logroscino G, Hu FB. Adherence to a DASH-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med. 2008;168(7):713–20. https://doi.org/10.1001/archinte.168.7.713 .

Yuan YQ, Li F, Dong RH, Chen JS, He GS, Li SG, et al. The development of a Chinese healthy eating index and its application in the general population. Nutrients. 2017;9(9):977. https://doi.org/10.3390/nu9090977 .

Kuriyama N, Murakami K, Livingstone MBE, Okubo H, Kobayashi S, Suga H, et al. Development of a food-based diet quality score for Japanese: associations of the score with nutrient intakes in young, middle-aged and older Japanese women. J Nutr Sci. 2016;5:e41. https://doi.org/10.1017/jns.2016.36 .

Custodio E, Kayikatire F, Fortin S, Thomas AC, Kameli Y, Nkunzimana T, et al. Minimum dietary diversity among women of reproductive age in urban Burkina Faso. Matern Child Nutr. 2020;16(2):e12897. https://doi.org/10.1111/mcn.12897 .

Miguel MA, Ana R, Celia M, Alicia G, Isabel C, Fatima OS. Proposal of a Mediterranean diet index for pregnant women. Br J Nutr. 2009;102(5):744–9.

Wong JE, Skidmore PML, Williams SM, Parnell WR. Healthy dietary habits score as an indicator of diet quality in New Zealand adolescents. J Nutr. 2014;144(6):937–42. https://doi.org/10.3945/jn.113.188375 .

Bork K, Cames C, Barigou S, Cournil A, Diallo A. A summary index of feeding practices is positively associated with height-for-age, but only marginally with linear growth, in rural Senegalese infants and toddlers. J Nutr. 2012;142(6):1116–22. https://doi.org/10.3945/jn.112.157602 .

De Jonge EA, Kiefte-de Jong JC, De Groot LC, Voortman T, Schoufour JD, Zillikens MC, et al. Development of a food group-based diet score and its association with bone mineral density in the elderly: the Rotterdam study. Nutrients. 2015;7(8):6974–90. https://doi.org/10.3390/nu7085317 .

Fung TT, Rexrode KM, Mantzoros CS, Manson JE, Willett WC, Hu FB. Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation. 2009;119(8):1093–100. https://doi.org/10.1161/CIRCULATIONAHA.108.816736 .

George SM, Ballard-Barbash R, Manson JE, Reedy J, Shikany JM, Subar AF, et al. Comparing indices of diet quality with chronic disease mortality risk in postmenopausal women in the Women's Health Initiative observational study: evidence to inform national dietary guidance. Am J Epidemiol. 2014;180(6):616–25. https://doi.org/10.1093/aje/kwu173 .

Harmon BE, Boushey CJ, Shvetsov YB, Reynolette E, Jill R, Wilkens LR, et al. Associations of key diet-quality indexes with mortality in the multiethnic cohort: the dietary patterns methods project. Am J Clin Nutr. 2015;101(3):587–97. https://doi.org/10.3945/ajcn.114.090688 .

Jill R, Krebs-Smith SM, Miller PE, Liese AD, Kahle LL, Yikyung P, et al. Higher diet quality is associated with decreased risk of all-cause, cardiovascular disease, and cancer mortality among older adults. J Nutr. 2014;144(6):881–9.

Francesco S, Claudio M, Rosanna A, Gian Franco G, Alessandro C. Mediterranean diet and health status: an updated meta-analysis and a proposal for a literature-based adherence score. Public Health Nutr. 2014;17(12):2769–82.

Schwingshackl L, Hoffmann G. Diet Quality as Assessed by the Healthy Eating Index, the Alternate Healthy Eating Index, the Dietary Approaches to Stop Hypertension Score, and Health Outcomes: A Systematic Review and Meta-Analysis of Cohort Studies. J Acad Nutr Dietetics. 2015;115(5):780–800 e5.

Sotos-Prieto M, Bhupathiraju SN, Mattei J, Fung TT, Li Y, Pan A, et al. Association of Changes in diet quality with Total and cause-specific mortality. N Engl J Med. 2017;377(2):143–53. https://doi.org/10.1056/NEJMoa1613502 .

Willett W, Rockström J, Loken B, Springmann M, Lang T, Vermeulen S, et al. Food in the Anthropocene: the EAT–lancet commission on healthy diets from sustainable food systems. Lancet. 2019;393(10170):447–92. https://doi.org/10.1016/S0140-6736(18)31788-4 .

Baden MY, Liu G, Satija A, Li Y, Sun Q, Fung TT, et al. Changes in plant-based diet quality and Total and cause-specific mortality. Circulation. 2019;140(12):979–91. https://doi.org/10.1161/CIRCULATIONAHA.119.041014 .

Satija A, Bhupathiraju SN, Rimm EB, Spiegelman D, Chiuve SE, Borgi L, et al. Plant-based dietary patterns and incidence of type 2 diabetes in US men and women: results from three prospective cohort studies. PLoS Med. 2016;13(6):e1002039. https://doi.org/10.1371/journal.pmed.1002039 .

Satija A, Bhupathiraju SN, Spiegelman D, Chiuve SE, Manson JAE, Willett W, et al. Healthful and unhealthful plant-based diets and the risk of coronary HeartDisease in U.S. adults. J Am Coll Cardiol. 2017;70(4):411–22. https://doi.org/10.1016/j.jacc.2017.05.047 .

Baden MY, Satija A, Hu FB, Huang T. Change in plant-based diet quality is associated with changes in plasma adiposity-associated biomarker concentrations in women. J Nutr. 2019;149(4):676–86. https://doi.org/10.1093/jn/nxy301 .

Kim H, Caulfield LE, Rebholz CM. Healthy plant-based diets are associated with lower risk of all-cause mortality in US adults. J Nutr. 2018;148(4):624–31. https://doi.org/10.1093/jn/nxy019 .

Golley RK, Smithers LG, Mittinty MN, Brazionis L, Emmett P, Northstone K, et al. An index measuring adherence to complementary feeding guidelines has convergent validity as a measure of infant diet quality. J Nutr. 2012;142(5):901–8. https://doi.org/10.3945/jn.111.154971 .

Vadiveloo M, Dixon LB, Mijanovich T, Elbel B, Parekh N. Development and evaluation of the US healthy food diversity index. Br J Nutr. 2014;112(9):1562–74. https://doi.org/10.1017/S0007114514002049 .

Krebs-Smith SM, Subar AF, Reedy J. Examining dietary patterns in relation to chronic disease: table. Circulation. 2015;132(9):790–3. https://doi.org/10.1161/CIRCULATIONAHA.115.018010 .

Canuto R, Camey S, Gigante DP, Menezes AMB, Olinto MTA. Focused principal component analysis: a graphical method for exploring dietary patterns. Cadernos de Saúde Pública. 2010;26(11):2149–56. https://doi.org/10.1590/S0102-311X2010001100016 .

Varraso R, Garcia-Aymerich J, Monier F, Le Moual N, De Batlle J, Miranda G, et al. Assessment of dietary patterns in nutritional epidemiology: principal component analysis compared with confirmatory factor analysis. Am J Clin Nutr. 2012;96(5):1079–92. https://doi.org/10.3945/ajcn.112.038109 .

Ryman TK, Boyer BB, Scarlett H, Jacques P, Diane OB, Kenneth T, et al. Characterising the reproducibility and reliability of dietary patterns among Yup'ik Alaska native people. Br J Nutr. 2015;113(4):634–43. https://doi.org/10.1017/S0007114514003596 .

Newby PK, Weismayer C, Akesson A, Tucker KL, Wolk A. Long-term stability of food patterns identified by use of factor analysis among Swedish women. J Nutr. 2006;136(3):626–33. https://doi.org/10.1093/jn/136.3.626 .

Bédard A, Garcia-Aymerich J, Sanchez M, Le Moual N, Clavel-Chapelon F, Boutron-Ruault M-C, et al. Confirmatory factor analysis compared with principal component analysis to derive dietary patterns: a longitudinal study in adult women. J Nutr. 2015;145(7):1559–68. https://doi.org/10.3945/jn.114.204479 .

Hu F, Rimm E, Sa W, Feskanich D, Stampfer M, Ascherio A, et al. Reproducibility and validity of dietary patterns assessed with a food-frequency questionnaire. Am J Clin Nutr. 1999;69(2):243–9. https://doi.org/10.1093/ajcn/69.2.243 .

Murakami K, Shinozaki N, Fujiwara A, Yuan X, Hashimoto A, Fujihashi H, et al. A systematic review of principal component analysis-derived dietary patterns in Japanese adults: are major dietary patterns reproducible within a country? Adv Nutr. 2019;10(2):237–49. https://doi.org/10.1093/advances/nmy079 .

Hong X, Ye Q, Wang Z, Yang H, Chen X, Zhou H, et al. Reproducibility and validity of dietary patterns identified using factor analysis among Chinese populations. Br J Nutr. 2016;116(5):842–52. https://doi.org/10.1017/S000711451600249X .

Castelló A, Lope V, Vioque J, Santamariña C, Pedraz-Pingarrón C, Abad S, et al. Reproducibility of data-driven dietary patterns in two groups of adult Spanish women from different studies. Br J Nutr. 2016;116(4):734–42. https://doi.org/10.1017/S000711451600252X .

Schulze MB, Kurt H, Anja K, Heiner B. Risk of hypertension among women in the EPIC-Potsdam study: comparison of relative risk estimates for exploratory and hypothesis-oriented dietary patterns. Am J Epidemiol. 2003;158(4):365–73. https://doi.org/10.1093/aje/kwg156 .

Martínez ME, Marshall JR, Sechrest L. Invited commentary: factor analysis and the search for objectivity. Am J Epidemiol. 1998;148(1):17–9. https://doi.org/10.1093/oxfordjournals.aje.a009552 .

Slattery ML, Edwards SL, Boucher KM, Anderson K, Caan BJ. Lifestyle and colon cancer: an assessment of factors associated with risk. Am J Epidemiol. 1999;150(8):869–77. https://doi.org/10.1093/oxfordjournals.aje.a010092 .

Greve B, Pigeot I, Huybrechts I, Pala V, Börnhorst C. A comparison of heuristic and model-based clustering methods for dietary pattern analysis. Public Health Nutr. 2015;19(02):255–64.

Lo Siou G, Yasui Y, Csizmadi I, McGregor SE, Robson PJ. Exploring statistical approaches to diminish subjectivity of cluster analysis to derive dietary patterns: the tomorrow project. Am J Epidemiol. 2011;173(8):956–67. https://doi.org/10.1093/aje/kwq458 .

Devlin UM, McNulty BA, Nugent AP, Gibney MJ. The use of cluster analysis to derive dietary patterns: methodological considerations, reproducibility, validity and the effect of energy mis-reporting. Proc Nutr Soc. 2012;71(4):599–609. https://doi.org/10.1017/S0029665112000729 .

Saxena A, Prasad M, Gupta A, Bharill N, Patel OP, Tiwari A, et al. A review of clustering techniques and developments. Neurocomputing. 2017;267:664–81. https://doi.org/10.1016/j.neucom.2017.06.053 .

Milligan GW. A study of the Beta-flexible clustering method. Multivar Behav Res. 1989;24(2):163–76. https://doi.org/10.1207/s15327906mbr2402_2 .

Article   CAS   Google Scholar  

Xu SH, Qiao N, Huang JJ, Sun CM, Cui Y, Tian SS, et al. Gender differences in dietary patterns and their association with the prevalence of metabolic syndrome among Chinese: a cross-sectional study. Nutrients. 2016;8(4):180. https://doi.org/10.3390/nu8040180 .

Wirfält E, Midthune D, Reedy J, Mitrou P, Flood A, Subar A, et al. Associations between food patterns defined by cluster analysis and colorectal cancer incidence in the NIH-AARP diet and health study. Eur J Clin Nutr. 2008;63:707–17.

He Y, Ma G, Zhai F, Li Y, Hu Y, Feskens EJ, et al. Dietary patterns and glucose tolerance abnormalities in Chinese adults. Diabetes Care. 2009;32(11):1972–6. https://doi.org/10.2337/dc09-0714 .

Wirfalt E, Midthune D, Reedy J, Mitrou P, Flood A, Subar AF, et al. Associations between food patterns defined by cluster analysis and colorectal cancer incidence in the NIH-AARP diet and health study. Eur J Clin Nutr. 2009;63(6):707–17. https://doi.org/10.1038/ejcn.2008.40 .

Sauvageot N, Schritz A, Leite S, Alkerwi A, Stranges S, Zannad F, et al. Stability-based validation of dietary patterns obtained by cluster analysis. Nutr J. 2017;16(1):4. https://doi.org/10.1186/s12937-017-0226-9 .

Fahey MT, Thane CW, Bramwell GD, Coward WA. Conditional Gaussian mixture modelling for dietary pattern analysis. J R Stat Soc Ser A (Statistics in Society). 2007;170(1):149–66. https://doi.org/10.1111/j.1467-985X.2006.00452.x .

Fahey MT, Ferrari P, Slimani N, Vermunt JK, White IR, Hoffmann K, et al. Identifying dietary patterns using a normal mixture model: application to the EPIC study. J Epidemiol Community Health. 2012;66(1):89–94. https://doi.org/10.1136/jech.2009.103408 .

Thorpe MG, Milte CM, Crawford D, McNaughton SA. A comparison of the dietary patterns derived by principal component analysis and cluster analysis in older Australians. Int J Behav Nutr Phys Act. 2016;13(1):1–14.

Sotres-Alvarez D, Herring AH, Siega-Riz AM. Latent class analysis is useful to classify pregnant women into dietary patterns. J Nutr. 2010;140(12):2253–9. https://doi.org/10.3945/jn.110.124909 .

Gaio AR, Costa JP, Santos AC, Ramos E, Lopes C. A restricted mixture model for dietary pattern analysis in small samples. Stat Med. 2012;31(19):2137–50. https://doi.org/10.1002/sim.5336 .

Fraley C, Raftery AE. How many clusters? Which clustering method? Answers via model-based cluster analysis. Comput J. 1998;41(8):578–88. https://doi.org/10.1093/comjnl/41.8.578 .

Gorst-Rasmussen A, Dahm CC, Dethlefsen C, Scheike T, Overvad K. Exploring dietary patterns by using the treelet transform. Am J Epidemiol. 2011;173(10):1097–104. https://doi.org/10.1093/aje/kwr060 .

Lee AB, Nadler B, Wasserman L. Treelets--an adaptive multi-scale basis for sparse unordered data. Ann Appl Stat. 2008;2(2):435–71.

Google Scholar  

Imamura F, Jacques PF. Invited commentary: dietary pattern analysis. Am J Epidemiol. 2011;173(10):1105–10. https://doi.org/10.1093/aje/kwr063 .

Assi N, Moskal A, Slimani N, Viallon V, Chajes V, Freisling H, et al. A treelet transform analysis to relate nutrient patterns to the risk of hormonal receptor-defined breast cancer in the European prospective investigation into Cancer and nutrition (EPIC). Public Health Nutr. 2015;19(02):242–54.

Schoenaker DA, Dobson AJ, Soedamah-Muthu SS, Mishra GD. Factor analysis is more appropriate to identify overall dietary patterns associated with diabetes when compared with Treelet transform analysis. J Nutr. 2013;143(3):392–8. https://doi.org/10.3945/jn.112.169011 .

Weikert C, Schulze MB. Evaluating dietary pattern the role of reduced rank regression. Curr Opin Clin Nutr Metab Care. 2016;19(5):341–6. https://doi.org/10.1097/MCO.0000000000000308 .

Hoffmann K. Application of a new statistical method to derive dietary patterns in nutritional epidemiology. Am J Epidemiol. 2004;159(10):935–44. https://doi.org/10.1093/aje/kwh134 .

Hoffmann K, Zyriax BC, Boeing H, Windler E. A dietary pattern derived to explain biomarker variation is strongly associated with the risk of coronary artery disease. Am J Clin Nutr. 2004;80(3):633–40. https://doi.org/10.1093/ajcn/80.3.633 .

DiBello JR, Kraft P, McGarvey ST, Goldberg R, Campos H, Baylin A. Comparison of 3 methods for identifying dietary patterns associated with risk of disease. Am J Epidemiol. 2008;168(12):1433–43. https://doi.org/10.1093/aje/kwn274 .

Melaku YA, Gill TK, Taylor AW, Adams R, Shi Z. A comparison of principal component analysis, partial least-squares and reduced-rank regressions in the identification of dietary patterns associated with bone mass in ageing Australians. Eur J Nutr. 2018;57(5):1969–83. https://doi.org/10.1007/s00394-017-1478-z .

Van Dam RM. New approaches to the study of dietary patterns. Br J Nutr. 2005;93(05):573.

Kroke A. Re: "application of a new statistical method to derive dietary patterns in nutritional epidemiology". Am J Epidemiol. 2004;160(11):1132–3. https://doi.org/10.1093/aje/kwh329 .

Yang TC, Aucott LS, Duthie GG, Macdonald HM. An application of partial least squares for identifying dietary patterns in bone health. Arch Osteoporos. 2017;12(1):63. https://doi.org/10.1007/s11657-017-0355-y .

Linden A, Yarnold PR. Using data mining techniques to characterize participation in observational studies. J Eval Clin Pract. 2016;22(6):835–43. https://doi.org/10.1111/jep.12625 .

Lazarou C, Karaolis M, Matalas A-L, Panagiotakos DB. Dietary patterns analysis using data mining method. An application to data from the CYKIDS study. Comput Methods Prog Biomed. 2012;108(2):706–14. https://doi.org/10.1016/j.cmpb.2011.12.011 .

Hearty AP, Gibney MJ. Analysis of meal patterns with the use of supervised data mining techniques--artificial neural networks and decision trees. Am J Clin Nutr. 2008;88(6):1632–42. https://doi.org/10.3945/ajcn.2008.26619 .

Easton JF, Roman Sicilia H, Stephens CR. Classification of diagnostic subcategories for obesity and diabetes based on eating patterns. Nutr Dietetics. 2019;76(1):104–9. https://doi.org/10.1111/1747-0080.12495 .

Panaretos D, Koloverou E, Dimopoulos AC, Kouli G-M, Vamvakari M, Tzavelas G, et al. A comparison of statistical and machine-learning techniques in evaluating the association between dietary patterns and 10-year cardiometabolic risk (2002–2012): the ATTICA study. Br J Nutr. 2018;120(03):326–34. https://doi.org/10.1017/S0007114518001150 .

Quinlan JR. C4.5: programs for machine learning. San Francisco: Morgan Kaufmann Publishers Inc; 1993.

Biesbroek S, van der AD BMC, Beulens JW, Verschuren WM, van der Schouw YT, et al. Identifying cardiovascular risk factor-related dietary patterns with reduced rank regression and random forest in the EPIC-NL cohort. Am J Clin Nutr. 2015;102(1):146–54. https://doi.org/10.3945/ajcn.114.092288 .

Ziegler A, Maccluer JW, Almasy L. Brief review of regression-based and machine learning methods in genetic epidemiology: the genetic analysis workshop 17 experience. Genet Epidemiol. 2011;35(S1):S5–S11.

Tibshirani R. Regression shrinkage and selection via the Lasso. J R Stat Soc Ser B Methodol. 1996;58(1):267–88.

Zhang F, Tapera TM, Gou J. Application of a new dietary pattern analysis method in nutritional epidemiology. BMC Med Res Methodol. 2018;18(1):119. https://doi.org/10.1186/s12874-018-0585-8 .

Leite MLC, Prinelli F. A compositional data perspective on studying the associations between macronutrient balances and diseases. Eur J Clin Nutr. 2017;71(12):1365–9. https://doi.org/10.1038/ejcn.2017.126 .

Leite MLC. Applying compositional data methodology to nutritional epidemiology. Stat Methods Med Res. 2016;25(6):3057–65. https://doi.org/10.1177/0962280214560047 .

Aitchison J. The statistical analysis of compositional data. J R Stat Soc Ser B Methodol. 1982;44:139–60.

Aitchison J. Principal component analysis of compositional data. Biometrika. 1983;70(1):57–65. https://doi.org/10.1093/biomet/70.1.57 .

Bruno F, Greco F, Ventrucci M. Spatio-temporal regression on compositional covariates: modeling vegetation in a gypsum outcrop. Environ Ecol Stat. 2015;22(3):445–63. https://doi.org/10.1007/s10651-014-0305-4 .

Egozcue JJ, Pawlowsky-Glahn V, Mateu-Figueras G, Barceló-Vidal C. Isometric Logratio transformations for compositional data analysis. Math Geol. 2003;35(3):279–300. https://doi.org/10.1023/A:1023818214614 .

Egozcue JJ, Pawlowsky-Glahn V. Groups of parts and their balances in compositional data analysis. Math Geol. 2006;37(7):795–828.

Martín-Fernández JA, Pawlowsky-Glahn V, Egozcue JJ, Tolosona-Delgado R. Advances in principal balances for compositional data. Math Geosci. 2017;50(3):273–98.

Pawlowsky-Glahn V, Egozcue JJ. Exploring compositional data with the CoDa-Dendrogram. Aust Stat Soc. 2011;40(1):103–13.

Palarea-Albaladejo J, Martín-Fernández JA. A modified EM alr-algorithm for replacing rounded zeros in compositional data sets. Comput Geosci. 2008;34(8):902–17. https://doi.org/10.1016/j.cageo.2007.09.015 .

Mozaffarian D. Dietary and policy priorities for cardiovascular disease, diabetes, and obesity: a comprehensive review. Circulation. 2016;133(2):187–225. https://doi.org/10.1161/CIRCULATIONAHA.115.018585 .

Schulze MB, Hoffmann K. Methodological approaches to study dietary patterns in relation to risk of coronary heart disease and stroke. Br J Nutr. 2007;95(5):860–9.

Corrêa Leite ML. Compositional data analysis as an alternative paradigm for nutritional studies. Clin Nutr ESPEN. 2019;33:207–12. https://doi.org/10.1016/j.clnesp.2019.05.011 .

Arnold KF, Berrie L, Tennant PWG, Gilthorpe MS. A causal inference perspective on the analysis of compositional data. Int J Epidemiol. 2020;49(4):1307-13.

Godichon-Baggioni A, Maugis-Rabusseau C, Rau A. Clustering transformed compositional data using K-means, with applications in gene expression and bicycle sharing system data. J Appl Stat. 2019;46(1):47–65. https://doi.org/10.1080/02664763.2018.1454894 .

Dumuid D, Pedišić Ž, Stanford TE, Martín-Fernández JA, Hron K, Maher CA, et al. The compositional isotemporal substitution model: a method for estimating changes in a health outcome for reallocation of time between sleep, physical activity and sedentary behaviour. Stat Methods Med Res. 2019;28(3):846–57. https://doi.org/10.1177/0962280217737805 .

Dumuid D, Stanford TE, Martin-Fernández J-A, Pedišić Ž, Maher CA, Lewis LK, et al. Compositional data analysis for physical activity, sedentary time and sleep research. Stat Methods Med Res. 2018;27(12):3726–38. https://doi.org/10.1177/0962280217710835 .

Zhang S, Midthune D, Guenther PM, Krebs-Smith SM, Kipnis V, Dodd KW, et al. A new multivariate measurement error model with zero-inflated dietary data, and its application to dietary assessment. Ann Appl Stat. 2011;5(2B):1456–87. https://doi.org/10.1214/10-AOAS446 .

Brennan L, Hu FB. Metabolomics-based dietary biomarkers in nutritional epidemiology-current status and future opportunities. Mol Nutr Food Res. 2019;63(1):e1701064. https://doi.org/10.1002/mnfr.201701064 .

Download references

Acknowledgments

This present review was supported by the National Natural Science Foundation of China (81072385, 82073674).

Author information

Authors and affiliations.

Department of Health Statistics, School of Public Health, Shanxi Medical University, No.56 Xinjian South Road, Taiyuan, 030001, Shanxi province, China

Junkang Zhao, Zhiyao Li, Qian Gao, Shuting Chen, Lun Huang, Wenjie Wang & Tong Wang

Department of Nutrition & Food Hygiene, School of Public Health, Shanxi Medical University, No.56 Xinjian South Road, Taiyuan, 030001, Shanxi province, China

Haifeng Zhao

You can also search for this author in PubMed   Google Scholar

Contributions

Junkang Zhao, Zhiyao Li, and Qian Gao researched, wrote, and prepared the manuscript. Haifeng Zhao, Shuting Chen, Lun Huang, and Wenjie Wang critically revised the manuscript. Tong Wang provided the idea and did the proofreading and corrections. All authors critically reviewed the manuscript and approved its final version.

Corresponding author

Correspondence to Tong Wang .

Ethics declarations

Ethics approval and consent to participate, consent for publication.

The authors agree to publish in this journal.

Competing interests

The authors declare no conflict of interest.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Zhao, J., Li, Z., Gao, Q. et al. A review of statistical methods for dietary pattern analysis. Nutr J 20 , 37 (2021). https://doi.org/10.1186/s12937-021-00692-7

Download citation

Received : 27 October 2020

Accepted : 01 April 2021

Published : 19 April 2021

DOI : https://doi.org/10.1186/s12937-021-00692-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Dietary patterns
• Dietary quality scores
• Principal component analysis
• Factor analysis
• Clustering analysis
• Treelet transform
• Reduced rank regression
• Data mining
• Least absolute shrinkage and selection operator
• Compositional data analysis

Nutrition Journal

ISSN: 1475-2891

• General enquiries: [email protected]

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Review Article
• Open access
• Published: 12 September 2019

The effects of plant-based diets on the body and the brain: a systematic review

• Evelyn Medawar   ORCID: orcid.org/0000-0001-5011-8275 1 , 2 , 3 ,
• Sebastian Huhn 4 ,
• Arno Villringer 1 , 2 , 3 &
• A. Veronica Witte 1  

Translational Psychiatry volume  9 , Article number:  226 ( 2019 ) Cite this article

307k Accesses

184 Citations

1433 Altmetric

Metrics details

• Human behaviour
• Molecular neuroscience
• Psychiatric disorders

Western societies notice an increasing interest in plant-based eating patterns such as vegetarian and vegan, yet potential effects on the body and brain are a matter of debate. Therefore, we systematically reviewed existing human interventional studies on putative effects of a plant-based diet on the metabolism and cognition, and what is known about the underlying mechanisms. Using the search terms “plant-based OR vegan OR vegetarian AND diet AND intervention” in PubMed filtered for clinical trials in humans retrieved 205 studies out of which 27, plus an additional search extending the selection to another five studies, were eligible for inclusion based on three independent ratings. We found robust evidence for short- to moderate-term beneficial effects of plant-based diets versus conventional diets (duration ≤ 24 months) on weight status, energy metabolism and systemic inflammation in healthy participants, obese and type-2 diabetes patients. Initial experimental studies proposed novel microbiome-related pathways, by which plant-based diets modulate the gut microbiome towards a favorable diversity of bacteria species, yet a functional “bottom up” signaling of plant-based diet-induced microbial changes remains highly speculative. In addition, little is known, based on interventional studies about cognitive effects linked to plant-based diets. Thus, a causal impact of plant-based diets on cognitive functions, mental and neurological health and respective underlying mechanisms has yet to be demonstrated. In sum, the increasing interest for plant-based diets raises the opportunity for developing novel preventive and therapeutic strategies against obesity, eating disorders and related comorbidities. Still, putative effects of plant-based diets on brain health and cognitive functions as well as the underlying mechanisms remain largely unexplored and new studies need to address these questions.

Similar content being viewed by others

Associations of dietary patterns with brain health from behavioral, neuroimaging, biochemical and genetic analyses

Ruohan Zhang, Bei Zhang, … Wei Cheng

Role of diet and its effects on the gut microbiome in the pathophysiology of mental disorders

J. Horn, D. E. Mayer, … E. A. Mayer

What makes a plant-based diet? a review of current concepts and proposal for a standardized plant-based dietary intervention checklist

Maximilian Andreas Storz

Introduction

Western societies notice an increasing interest in plant-based eating patterns such as avoiding meat or fish or fully excluding animal products (vegetarian or vegan, see Fig.  1 ). In 2015, around 0.4−3.4% US adults, 1−2% British adults, and 5−10% of German adults were reported to eat largely plant-based diets 1 , 2 , 3 , 4 , due to various reasons (reviewed in ref. 5 ). Likewise, the number of scientific publications on PubMed (Fig.  2 ) and the public popularity as depicted by Google Trends (Fig.  3 ) underscore the increased interest in plant-based diets. This increasing awareness calls for a better scientific understanding of how plant-based diets affect human health, in particular with regard to potentially relevant effects on mental health and cognitive functions.

From left to right: including all food items (omnivore), including all except for meat (pesco-vegetarian) or meat and fish (ovo-lacto-vegetarian) to including only plant-based items (vegan)

Frequency of publications on PubMed including the search terms “vegan” (in light green), vegetarian (in orange) and plant-based (dark green)—accessed on 19 April 2019

Note indicates technical improvements implemented by Google Trends. Data source: Google Trends . Search performed on 18 April 2019

A potential effect of plant-based diets on mortality rate remains controversial: large epidemiological studies like the Adventist studies ( n  = 22,000−96,000) show a link between plant-based diets, lower all-cause mortality and cardiovascular diseases 6 , 7 , while other studies like the EPIC-Oxford study and the “45 and Up Study” ( n  = 64,000−267,000) show none 8 , 9 . Yet, many, but not all, epidemiological and interventional human studies in the last decades have suggested that plant-based diets exert beneficial health effects with regard to obesity-related metabolic dysfunction, type 2 diabetes mellitus (T2DM) and chronic low-grade inflammation (e.g. refs. 6 , 7 , 10 , 11 , for reviews, see refs. 12 , 13 , 14 , 15 , 16 , 17 , 18 ). However, while a putative link between such metabolic alterations and brain health through pathways which might include diet-related neurotransmitter precursors, inflammatory pathways and the gut microbiome 19 becomes increasingly recognized, the notion that plant-based diets exert influence on mental health and cognitive functions appears less documented and controversial 20 , 21 , 22 , 23 , 24 . We therefore systematically reviewed the current evidence based on available controlled interventional trials, regarded as the gold standard to assess causality, on potential effects of plant-based diets on (a) metabolic factors including the microbiome and (b) neurological or psychiatric health and brain functions. In addition, we aimed to evaluate potential underlying mechanisms and related implications for cognition.

We performed a systematic PubMed search with the following search terms “plant-based OR vegan OR vegetarian AND diet AND intervention” with the filter “clinical trial” and “humans”, preregistered at PROSPERO (CRD42018111856; https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=111856 ) (Suppl. Fig.  1 ). PubMed was used as search engine because it was esteemed to yield the majority of relevant human clinical trials from a medical perspective. Exclusion criteria were insufficient design quality (such as lack of a control group), interventions without a plant-based or vegetarian or vegan diet condition, intervention with multiple factors (such as exercise and diet), and the exclusive report of main outcomes of no interest, such as dietary compliance, nutrient intake (such as vitamins or fiber intake), or nonmetabolic (i.e., not concerning glucose metabolism, lipid profile, gastrointestinal hormones or inflammatory markers) or non-neurological/psychiatric disease outcomes (e.g. cancer, caries).

Studies were independently rated for eligibility into the systematic review by three authors based on reading the abstract and, if needed, methods or other parts of the publication. If opinions differed, a consensus was reached through discussion of the individual study. This yielded 27 eligible out of 205 publications; see Table  1 for details. To increase the search radius for studies dealing with microbial and neurological/psychiatric outcomes, we deleted the search term “intervention”, which increased the number of studies by around one third, and checked for studies with “microbiome/microbiota”, “mental”, “cognitive/cognition” or “psychological/psychology” in the resulting records. Through this, we retrieved another five studies included in Table  1 . Further related studies were reviewed based on additional nonsystematic literature search.

Section I: Effects of plant-based diets on body and brain outcomes

Results based on interventional studies on metabolism, microbiota and brain function.

Overall, the vast majority of studies included in this systematic review reported a short-term beneficial effect of plant-based dietary interventions (study duration 3−24 months) on weight status, glucose, insulin and/or plasma lipids and inflammatory markers, whereas studies investigating whether plant-based diets affect microbial or neurological/psychiatric disease status and other brain functions were scarce and rather inconclusive (Table  1 ).

More specifically, 19 out of 32 studies dealing with T2DM and/or obese subjects and seven out of 32 dealing with healthy subjects observed a more pronounced weight loss and metabolic improvements, such as lowering of glycated hemoglobin (HbA1c)—a long-term marker for glucose levels—decreased serum levels of low-density (LDL) and high-density lipoproteins (HDL) and total cholesterol (TC), after a plant-based diet compared to an omnivore diet. This is largely in line with recent meta-analyses indicating beneficial metabolic changes after a plant-based diet 25 , 26 , 27 .

For example, Lee et al. found a significantly larger reduction of HbA1c and lower waist circumference after vegan compared to conventional dieting 28 . Jenkins et al. found a disease-attenuating effect in hyperlipidemic patients after 6 months adopting a low-carbohydrate plant-based diet compared to a high-carbohydrate lacto-ovo-vegetarian diet 29 , 30 . However, lower energy intake in the vegan dieters might have contributed to these effects. Yet, while a plant-based diet per se might lead to lower caloric intake, other studies observed nonsignificant trends toward higher effect sizes on metabolic parameters after a vegan diet, even when caloric intake was comparable: two studies in T2DM patients 31 , 32 compared calorie-unrestricted vegan or vegetarian to calorie-restricted conventional diets over periods of 6 months and 1.5 years, respectively, in moderate sample sizes ( n  ~ 75−99) with similar caloric intake achieved in both diet groups. Both studies indicated stronger effects of plant-based diets on disease status, such as reduced medication, improved weight status and increased glucose/insulin sensitivity, proposing a diabetes-preventive potential of plant-based diets. Further, a five-arm study comparing four types of plant-based diets (vegan, vegetarian, pesco-vegetarian, semi-vegetarian) to an omnivore diet (total n  = 63) in obese participants found the most pronounced effect on weight loss for a vegan diet (−7.5 ± 4.5% of total body weight) 33 . Here, inflammation markers conceptualized as the dietary inflammatory index were also found to be lower in vegan, vegetarian and pesco-vegetarian compared to semi-vegetarian overweight to obese dieters 33 .

Intriguingly, these results 28 , 29 , 30 , 31 , 32 , 33 cohesively suggest that although caloric intake was similar across groups, participants who had followed a vegan diet showed higher weight loss and improved metabolic status.

As a limitation, all of the reviewed intervention studies were carried out in moderate sample sizes and over a period of less than 2 years, disregarding that long-term success of dietary interventions stabilizes after 2−5 years only 34 . Future studies with larger sample sizes and tight control of dietary intake need to confirm these results.

Through our systematic review we retrieved only one study that added the gut microbiome as novel outcome for clinical trials investigating the effects of animal-based diets compared to plant-based diets. While the sample size was relatively low ( n  = 10, cross-over within subject design), it showed that changing animal- to plant based diet changed gut microbial activity towards a trade-off between carbohydrate and protein fermentation processes within only 5 days 35 . This is in line with another controlled-feeding study where microbial composition changes already occurred 24 h after changing diet (not exclusively plant-based) 36 . However, future studies incorporating larger sample sizes and a uniform analysis approach of microbial features need to further confirm the hypothesis that a plant-based diet ameliorates microbial diversity and health-related bacteria species.

Considering neurological or psychiatric diseases and brain functions, the systematic review yielded in six clinical trials of diverse clinical groups, i.e. migraine, multiple sclerosis, fibromyalgia and rheumatoid arthritis. Here, mild to moderate improvement, e.g. measured by antibody levels, symptom improvement or pain frequency, was reported in five out of six studies, sometimes accompanied by weight loss 37 , 38 , 39 , 40 (Table  1 ). However, given the pilot character of these studies, indicated by small sample sizes ( n  = 32−66), lack of randomization 37 , or that the plant-based diet was additionally free of gluten 40 , the evidence is largely anecdotal. One study in moderately obese women showed no effects on psychological outcomes 41 , two studies with obese and nonobese healthy adults indicated improvements in anxiety, stress and depressive symptom scores 23 , 24 . Taken together, the current evidence based on interventional trials regarding improvements of cognitive and emotional markers and in disease treatment for central nervous system disorders such as multiple sclerosis or fibromyalgia remains considerably fragmentary for plant-based diets.

Among observational studies, a recent large cross-sectional study showed a higher occurrence of depressive symptoms for vegetarian dieters compared to nonvegetarians 20 . Conversely, another observational study with a sample of about 80% women found a beneficial association between a vegan diet and mood disturbance 24 .

Overall, the relationship between mental health (i.e. depression) and restrictive eating patterns has been the focus of recent research 20 , 21 , 22 , 24 , 42 ; however, causal relationships remain uninvestigated due to the observational design.

Underlying mechanisms linking macronutrient intake to metabolic processes

On the one hand, nutrient sources as well as their intake ratios considerably differ between plant-based and omnivore diets (Suppl. Table  1 ), and on the other hand, dietary micro- and macromolecules as well as their metabolic substrates affect a diversity of physiological functions, pointing to complex interdependencies. Thus, it seems difficult to nail down the proposed beneficial effects of a plant-based diet on metabolic status to one specific component or characteristic, and it seems unlikely that the usually low amount of calories in plant-based diets could explain all observed effects. Rather, plant-based diets might act through multiple pathways, including better glycemic control 43 , lower inflammatory activity 44 and altered neurotransmitter metabolism via dietary intake 45 or intestinal activity 46 (Fig.  4 ).

BMI body-mass-index, HbA1c hemoglobin A1c, LDL-cholesterol low-density lipoprotein cholesterol, Trp tryptophan, Tyr tyrosine. Images from commons.wikimedia.org , “Brain human sagittal section” by Lynch 2006 and “Complete GI tract” by Häggström 2008, “Anatomy Figure Vector Clipart” by http://moziru.com

On the macronutrient level, plant-based diets feature different types of fatty acids (mono- and poly-unsaturated versus saturated and trans) and sugars (complex and unrefined versus simple and refined), which might both be important players for mediating beneficial health effects 18 . On the micronutrient level, the EPIC-Oxford study provided the largest sample of vegan dieters worldwide ( n (vegan) = 2396, n (total) = 65,429) and showed on the one hand lower intake of saturated fatty acids (SFA), retinol, vitamin B12 and D, calcium, zinc and protein, and on the other hand higher intake of fiber, magnesium, iron, folic acid, vitamin B1, C and E in vegan compared to omnivore dieters 47 . Other studies confirmed the variance of nutrient intake across dietary groups, i.e. omnivores, vegetarians and vegans, showing the occurrence of critical nutrients for each group 48 , 49 . Not only the amount of SFA but also its source and profile might be important factors regulating metabolic control (reviewed in ref. 14 ), for example through contributing to systemic hyperlipidemia and subsequent cardiovascular risk. Recently, it has been shown in a 4-week intervention trial that short-term dietary changes favoring a diet high in animal-based protein may lead to an increased risk for cardiovascular derangements mediated by higher levels of trimethylamine N-oxide (TMAO), which is a metabolite of gut bacteria-driven metabolic pathways 50 .

Secondly, high fiber intake from legumes, grains, vegetables and fruits is a prominent feature of plant-based diets (Table  1 ), which could induce beneficial metabolic processes like upregulated carbohydrate fermentation and downregulated protein fermentation 35 , improved gut hormonal-driven appetite regulation 51 , 52 , 53 , 54 , 55 , and might prevent chronic diseases such as obesity and T2DM by slowing down digestion and improving lipid control 56 . A comprehensive review including evidence from 185 prospective studies and 58 clinical trials concluded that risk reduction for a myriad of diseases (incl. CVD, T2DM, stroke incidence) was greatest for daily fiber intake between 25 and 29 g 57 . Precise evidence for underlying mechanisms is missing; however, more recently it has been suggested that high fiber intake induces changes on the microbial level leading to lower long-term weight gain 58 , a mechanism discussed below.

The reason for lower systemic inflammation in plant-based dieters could be due to the abundance of antiinflammatory molecule intake and/or avoidance of proinflammatory animal-derived molecules. Assessing systemic inflammation is particularly relevant for medical conditions such as obesity, where it has been proposed to increase the risk for cardiovascular disease 59 , 60 . In addition, higher C-reactive protein (CRP) and interleukin-6 (IL-6) levels have been linked with measures of brain microstructure, such as microstructural integrity and white matter lesions 61 , 62 , 63 and higher risk of dementia 64 , and recent studies point out that a diet-related low inflammatory index might also directly affect healthy brain ageing 65 , 66 .

Interventional studies that focus on plant- versus meat-based proteins or micronutrients and potential effects on the body and brain are lacking. A meta-analysis including seven RCTs and one cross-sectional studies on physical performance and dietary habits concluded that a vegetarian diet did not adversely influence physical performance compared to an omnivore diet 67 . An epidemiological study by Song et al. 11 estimated that statistically replacing 3% of animal protein, especially from red meat or eggs, with plant protein would significantly improve mortality rates. This beneficial effect might however not be explained by the protein source itself, but possibly by detrimental components found in meat (e.g. heme-iron or nitrosamines, antibiotics, see below).

Some studies further hypothesized that health benefits observed in a plant-based diet stem from higher levels of fruits and vegetables providing phytochemicals or vitamin C that might boost immune function and eventually prevent certain types of cancer 68 , 69 , 70 . A meta-analysis on the effect of phytochemical intake concluded a beneficial effect on CVD, cancer, overweight, body composition, glucose tolerance, digestion and mental health 71 . Looking further on the impact of micronutrients and single dietary compounds, there is room for speculation that molecules, that are commonly avoided in plant-based diets, might affect metabolic status and overall health, such as opioid-peptides derived from casein 72 , pre- and probiotics 73 , 74 , carry-over antibiotics found in animal products 75 , 76 or food-related carcinogenic toxins, such as dioxin found in eggs or nitrosamines found in red and processed meat 77 , 78 . Although conclusive evidence is missing, these findings propose indirect beneficial effects on health deriving from plant-based compared to animal-based foods, with a potential role for nonprotein substances in mediating those effects 18 . While data regarding chemical contaminant levels (such as crop pesticides, herbicides or heavy metals) in different food items are fragmentary only, certain potentially harmful compounds may be more (or less) frequently consumed in plant-based diets compared to more animal-based diets 79 . Whether these differences lead to systematic health effects need to be explored.

Taken together, the reviewed studies indicating effects of plant-based diets through macro- and micronutrient intake reveal both the potential of single ingredients or food groups (low SFA, high fiber) and the immense complexity of diet-related mechanisms for metabolic health. As proposed by several authors, benefits on health related to diet can probably not be viewed in isolation for the intake (or nonintake) of specific foods, but rather by additive or even synergistic effects between them (reviewed in refs. 12 , 80 ). Even if it remains a challenging task to design long-term RCTs that control macro- and micronutrient levels across dietary intervention groups, technological advancements such as more fine-tuned diagnostic measurements and automated self-monitoring tools, e.g. automatic food recognition systems 81 and urine-related measures of dietary intake 82 , could help to push the field forward.

Nutrients of particular interest in plant-based diets

As described above, plant-based diets have been shown to convey nutritional benefits 48 , 49 , in particular increased fiber, beta carotene, vitamin K and C, folate, magnesium, and potassium intake and an improved dietary health index 83 . However, a major criticism of plant-based diets is the risk of nutrient deficiencies for specific micronutrients, especially vitamin B12, a mainly animal-derived nutrient, which is missing entirely in vegan diets unless supplemented or provided in B12-fortified products, and which seems detrimental for neurological and cognitive health when intake is low. In the EPIC-Oxford study about 50% of the vegan dieters showed serum levels indicating vitamin B12 deficiency 84 . Along other risk factors such as age 85 , diet, and plant-based diets in particular, seem to be the main risk factor for vitamin B12 deficiency (reviewed in ref. 86 ), and therefore supplementing vitamin B12 for these risk groups is highly recommended 87 . Vitamin B12 is a crucial component involved in early brain development, in maintaining normal central nervous system function 88 and suggested to be neuroprotective, particularly for memory performance and hippocampal microstructure 89 . One hypothesis is that high levels of homocysteine, that is associated with vitamin B12 deficiency, might be harmful to the body. Vitamin B12 is the essential cofactor required for the conversion of homocysteine into nonharmful components and serves as a cofactor in different enzymatic reactions. A person suffering from vitamin B12 insufficiency accumulates homocysteine, lastly promoting the formation of plaques in arteries and thereby increasing atherothrombotic risk 90 , possibly facilitating symptoms in patients of Alzheimer’s disease 91 . A meta-analysis found that vitamin B12 deficiency was associated with stroke, Alzheimer’s disease, vascular dementia, Parkinson’s disease and in even lower concentrations with cognitive impairment 92 , supporting the claim of its high potential for disease prevention when avoided or treated 93 . Further investigations and longitudinal studies are needed, possibly measuring holotranscobalamin (the active form of vitamin B12) as a more specific and sensitive marker for vitamin B12 status 94 , to examine in how far nonsupplementing vegan dieters could be at risk for cardiovascular and cognitive impairment.

Similar health dangers can stem from iron deficiency, another commonly assumed risk for plant-based dieters and other risk groups such as young women. A meta-analysis on 24 studies proposes that although serum ferritin levels were lower in vegetarians on average, it is recommended to sustain an optimal ferritin level (neither too low nor too high), calling for well-monitored supplementation strategies 95 . Iron deficiency is not only dependent on iron intake as such but also on complimentary dietary factors influencing its bioavailability (discussed in ref. 95 ). The picture remains complex: on the one hand iron deficiency may lead to detrimental health effects, such as impairments in early brain development and cognitive functions in adults and in children carried by iron-deficient mothers 96 and a possible role for iron overload in the brain on cognitive impairment on the other hand 97 . One study showed that attention, memory and learning were impaired in iron-deficient compared to iron-sufficient women, which could be restored after a 4-month oral iron supplementation ( n  = 118) 98 . Iron deficiency-related impairments could be attributed to anemia as an underlying cause, possibly leading to fatigue, or an undersupply of blood to the brain or alterations in neurobiological and neuronal systems 99 provoking impaired cognitive functioning.

This leads to the general recommendation to monitor health status by frequent blood tests, to consult a dietician to live healthily on a plant-based diet and to consider supplements to avoid nutrient deficiencies or nutrient-overdose-related toxicity. All in all, organizations such as the Academy of Nutrition and Dietetics 100 and the German Nutrition Society do not judge iron as a major risk factor for plant-based dieters 101 .

Section II: Effects of diet on the gut microbiome

The link between diet and microbial diversity.

Another putative mechanistic pathway of how plant-based diets can affect health may involve the gut microbiome which has increasingly received scientific and popular interest, lastly not only through initiatives such as the Human Microbiome Project 102 . A common measure for characterizing the gut community is enterotyping, which is a way to stratify individuals according to their gut bacterial diversity, by calculating the ratio between bacterial genera, such as Prevotella and Bacteroides 103 . While interventional controlled trials are still scarce, this ratio has been shown to be conclusive for differentiating plant-based from animal-based microbial profiles 36 . Specifically, in a sample of 98 individuals, Wu et al. 36 found that a diet high in protein and animal fats was related to more Bacteroides, whereas a diet high in carbohydrates, representing a plant-based one, was associated with more Prevotella. Moreover, the authors showed that a change in diet to high-fat/low-fiber or to low-fat/high-fiber in ten individuals elicited a change in gut microbial enterotype with a time delay of 24 h only and remained stable over 10 days, however not being able to switch completely to another enterotype 36 . Another strictly controlled 30-day cross-over interventional study showed that a change in diet to either an exclusively animal-based or plant-based diet promoted gut microbiota diversity and genetic expression to change within 5 days 35 . Particularly, in response to adopting an animal-based diet, microbial diversity increased rapidly, even overshadowing individual microbial gene expression. Beyond large shifts in overall diet, already modest dietary modifications such as the daily consumption of 43 g of walnuts, were able to promote probiotic- and butyric acid-producing bacterial species in two RCTs, after 3 and 8 weeks respectively 104 , 105 , highlighting the high adaptability of the gut microbiome to dietary components. The Prevotella to Bacteroides ratio (P/B) has been shown to be involved in the success of dietary interventions targeting weight loss, with larger weight loss in high P/B compared to low P/B in a 6-month whole-grain diet compared to a conventional diet 106 . Only recently, other microbial communities, such as the salivary microbiome, have been shown to be different between omnivores and vegan dieters 107 , opening new avenues for research on adaptable mechanisms related to dietary intake.

A continuum in microbial diversity dependent on diet

Plant-based diets are supposed to be linked to a specific microbial profile, with a vegan profile being most different from an omnivore, but not always different from a vegetarian profile (reviewed in ref. 15 ). Some specifically vegan gut microbial characteristics have also been found in a small sample of six obese subjects after 1 month following a vegetarian diet, namely less pathobionts, more protective bacterial species improving lipid metabolism and a reduced level of intestinal inflammation 108 . Investigating long-term dietary patterns a study found a dose-dependent effect for altered gut microbiota in vegetarians and vegans compared to omnivores depending on the quantity of animal products 109 . The authors showed that gut microbial profiles of plant-based diets feature the same total number but lower counts of Bacteroides, Bifidobacterium, E. coli and Enterobacteriaceae compared to omnivores, with the biggest difference to vegans. Still today it remains unclear, what this shift in bacterial composition means in functional terms, prompting the field to develop more functional analyses.

In a 30-day intervention study, David et al. found that fermentation processes linked to fat and carbohydrate decomposition were related to the abundance of certain microbial species 35 . They found a strong correlation between fiber intake and Prevotella abundance in the microbial gut. More recently, Prevotella has been associated with plant-based diets 110 that are comparable to low-fat/high-fiber diets 111 and might be linked to the increased synthesis of short-chain fatty acids (SCFA) 112 . SCFAs are discussed as putative signaling molecules between the gut microbiome and the receptors, i.e. free fatty acid receptor 2 (FFA2) 51 , found in host cells across different tissues 113 and could therefore be one potential mechanism of microbiome−host communication.

The underlying mechanisms of nutrient decomposition by Prevotella and whether abundant Prevotella populations in the gut are beneficial for overall health remain unknown. Yet it seems possible that an increased fiber intake and therefore higher Prevotella abundance such as associated with plant-based diets is beneficial for regulating glycemic control and keeping inflammatory processes within normal levels, possibly due to reduced appetite and lower energy intake mediated by a higher fiber content 114 . Moreover, it has been brought forward that the microbiome might influence bodily homeostatic control, suggesting a role for the gut microbiota in whole-body control mechanisms on the systemic level. Novel strategies aim to develop gut-microbiota-based therapies to improve bodily states, e.g. glycemic control 115 , based on inducing microbial changes and thereby eliciting higher-level changes in homeostasis. While highly speculative, such strategies could in theory also exert changes on the brain level, which will be discussed next in the light of a bi-directional feedback between the gut and the brain.

Effects on cognition and behavior linking diet and cognition via the microbiome−gut−brain axis

While the number of interventional studies focusing on cognitive and mental health outcomes after adopting plant-based diets overall is very limited (see Section I above), one underlying mechanism of how plant-based diets may affect mood could involve signaling pathways on the microbiome−gut−brain axis 116 , 117 , 118 , 119 . A recent 4-week intervention RCT showed that probiotic administration compared to placebo and no intervention modulated brain activity during emotional decision-making and emotional recognition tasks 117 . In chronic depression it has been proposed that immunoglobulin A and M antibodies are synthesized by the host in response to gut commensals and are linked to depressive symptoms 120 . Whether the identified gram-negative bacteria might also play a role in plant-based diets remains to be explored. A meta-analysis on five studies concluded that probiotics may mediate an alleviating effect on depression symptomatic 121 —however, sample sizes remained rather small ( n  < 100) and no long-term effects were tested (up to 8 weeks).

Currently, several studies aim to identify microbial profiles in relation to disease and how microbial data can be used on a multimodal way to improve functional resolution, e.g. characterizing microbial profiles of individuals suffering from type-1 diabetes 122 . Yet, evidence for specific effects of diet on cognitive functions and behavior through changes in the microbiome remains scarce. A recent study indicated the possibility that our food choices determine the quantity and quality of neurotransmitter-precursor levels that we ingest, which in turn might influence behavior, as shown by lower fairness during a money-redistribution task, called the ultimatum game, after a high-carbohydrate/protein ratio breakfast than after a low-ratio breakfast 123 . Strang et al. found that precursor forms of serotonin and dopamine, measured in blood serum, predicted behavior in this task, and precursor concentrations were dependent on the nutrient profile of the consumed meal before the task. Also on a cross-sectional level tryptophan metabolites from fecal samples have been associated with amygdala-reward network functional connectivity 124 . On top of the dietary composition per se, the microbiota largely contributes to neurotransmitter precursor concentrations; thus, in addition to measuring neurotransmitter precursors in the serum, metabolomics on fecal samples would be helpful to further understand the functional role of the gut microbiota in neurotransmitter biosynthesis and regulation 125 .

Indicating the relevance of gut microbiota for cognition, a first human study assessing cognitive tests and brain imaging could distinguish obese from nonobese individuals using a microbial profile 126 . The authors found a specific microbiotic profile, particularly defined by Actinobacteria phylum abundance, that was associated with microstructural properties in the hypothalamus and in the caudate nucleus. Further, a preclinical study tested whether probiotics could enhance cognitive function in healthy subjects, showing small effects on improved memory performance and reduced stress levels 127 .

A recent study could show that microbial composition influences cerebral amyloidogenesis in a mouse model for Alzheimer’s disease 128 . Health status of the donor mouse seemingly mattered: fecal transplants from transgenic mice had a larger impact on amyloid beta proliferation in the brain compared to wild-type feces. Translational interpretations to humans should be done with caution if at all—yet the results remain elucidative for showing a link between the gut microbiome and brain metabolism.

The evidence for effects of strictly plant-based diets on cognition is very limited. For other plant-based diets such as the Mediterranean diet or DASH diet, there are more available studies that indicate protective effects on cardiovascular and brain health in the aging population (reviewed in refs. 129 , 130 ). Several attempts have been made to clarify potential underlying mechanisms, for example using supplementary plant polyphenols, fish/fish-oil consumption or whole dietary pattern change in RCTs 131 , 132 , 133 , 134 , 135 , 136 , 137 , yet results are not always equivocal and large-scale intervention studies have yet to be completed.

The overall findings of this paragraph add to the evidence that microbial diversity may be associated with brain health, although underlying mechanisms and candidate signaling molecules remain unknown.

Based on this systematic review of randomized clinical trials, there is an overall robust support for beneficial effects of a plant-based diet on metabolic measures in health and disease. However, the evidence for cognitive and mental effects of a plant-based diet is still inconclusive. Also, it is not clear whether putative effects are due to the diet per se, certain nutrients of the diet (or the avoidance of certain animal-based nutrients) or other factors associated with vegetarian/vegan diets. Evolving concepts argue that emotional distress and mental illnesses are linked to the role of microbiota in neurological function and can be potentially treated via microbial intervention strategies 19 . Moreover, it has been claimed that certain diseases, such as obesity, are caused by a specific microbial composition 138 , and that a balanced gut microbiome is related to healthy ageing 111 . In this light, it seems possible that a plant-based diet is able to influence brain function by still unclear underlying mechanisms of an altered microbial status and systemic metabolic alterations. However, to our knowledge there are no studies linking plant-based diets and cognitive abilities on a neural level, which are urgently needed, due to the hidden potential as a dietary therapeutic tool. Also, further studies are needed to disentangle motivational beliefs on a psychological level that lead to a change in diet from causal effects on the body and the brain mediated e.g., by metabolic alterations or a change in the gut microbiome.

GOV.UK. National Diet and Nutrition Survey: headline results from years 1, 2 and 3 (combined) of the rolling programme 2008/09–2010/11. https://www.gov.uk/government/statistics/national-diet-and-nutrition-survey-headline-results-from-years-1-2-and-3-combined-of-the-rolling-programme-200809-201011 (2012).

V. E. B. U. Deutschland & Joy, S. Anzahl der Veganer und Vegetarier in Deutschland. Stand 31 , 2016 (2015).

Mensink, G., Barbosa, C. L. & Brettschneider, A.-K. Verbreitung der vegetarischen Ernährungsweise in Deutschland 1 , (2016).

The Vegetarian Resource Group. How many adults in the U.S. are vegetarian and vegan? http://www.vrg.org/nutshell/Polls/2016_adults_veg.htm (2016).

Rosenfeld, D. L. & Burrow, A. L. Vegetarian on purpose: understanding the motivations of plant-based dieters. Appetite 116 , 456–463 (2017).

Article   PubMed   Google Scholar  

Orlich, M. J. et al. Vegetarian dietary patterns and mortality in Adventist Health Study 2. JAMA Intern. Med. 173 , 1230–1238 (2013).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Le, L. T. & Sabaté, J. Beyond meatless, the health effects of vegan diets: findings from the Adventist cohorts. Nutrients 6 , 2131–2147 (2014).

Article   PubMed   PubMed Central   Google Scholar  

Mihrshahi, S. et al. Vegetarian diet and all-cause mortality: evidence from a large population-based Australian cohort-the 45 and up study. Prev. Med. 97 , 1–7 (2017).

Key, T. J. et al. Mortality in British vegetarians: results from the European Prospective Investigation into Cancer and Nutrition (EPIC-Oxford). Am. J. Clin. Nutr. 89 , 1613S–1619S (2009).

Article   CAS   PubMed   Google Scholar  

Fung, T. T. et al. Low-carbohydrate diets and all-cause and cause-specific mortalitytwo cohort studies. Ann. Intern. Med. 153 , 289–298 (2010).

Song, M. et al. Association of animal and plant protein intake with all-cause and cause-specific mortality. JAMA Intern. Med. 176 , 1453–1463 (2016).

Hu, F. B. Plant-based foods and prevention of cardiovascular disease: an overview. Am. J. Clin. Nutr. 78 , 544S–551S (2003).

Tonstad, S., Butler, T., Yan, R. & Fraser, G. E. Type of vegetarian diet, body weight, and prevalence of type 2 diabetes. Diabetes Care 32 , 791–796 (2009).

McEvoy, C. T., Temple, N. & Woodside, J. V. Vegetarian diets, low-meat diets and health: a review. Public Health Nutr. 15 , 2287–2294 (2012).

Glick-Bauer, M. & Yeh, M.-C. The health advantage of a vegan diet: exploring the gut microbiota connection. Nutrients 6 , 4822–4838 (2014).

Appleby, P. N. & Key, T. J. The long-term health of vegetarians and vegans. Proc. Nutr. Soc. 75 , 287–293 (2016).

Eichelmann, F., Schwingshackl, L., Fedirko, V. & Aleksandrova, K. Effect of plant‐based diets on obesity‐related inflammatory profiles: a systematic review and meta‐analysis of intervention trials. Obes. Rev. 17 , 1067–1079 (2016).

McMacken, M. & Shah, S. A plant-based diet for the prevention and treatment of type 2 diabetes. J. Geriatr. Cardiol. 14 , 342 (2017).

CAS   PubMed   PubMed Central   Google Scholar  

Rogers, G. B. et al. From gut dysbiosis to altered brain function and mental illness: mechanisms and pathways. Mol. Psychiatry 21 , 738–748 (2016).

Hibbeln, J. R., Northstone, K., Evans, J. & Golding, J. Vegetarian diets and depressive symptoms among men. J. Affect Disord. 225 , 13–17 (2018).

Forestell, C. A. & Nezlek, J. B. Vegetarianism, depression, and the five factor model of personality. Ecol. Food Nutr. 57 , 246–259 (2018).

Matta, J. et al. Depressive symptoms and vegetarian diets: results from the constances cohort. Nutrients 10 , 1695 (2018).

Article   PubMed Central   Google Scholar  

Agarwal, U. et al. A multicenter randomized controlled trial of a nutrition intervention program in a multiethnic adult population in the corporate setting reduces depression and anxiety and improves quality of life: the GEICO study. Am. J. Health Promot. 29 , 245–254 (2015).

Beezhold, B., Radnitz, C., Rinne, A. & DiMatteo, J. Vegans report less stress and anxiety than omnivores. Nutr. Neurosci. 18 , 289–296 (2015).

Barnard, N. D., Levin, S. M. & Yokoyama, Y. A systematic review and meta-analysis of changes in body weight in clinical trials of vegetarian diets. J. Acad. Nutr. Diet. 115 , 954–969 (2015).

Huang, R.-Y., Huang, C.-C., Hu, F. B. & Chavarro, J. E. Vegetarian diets and weight reduction: a meta-analysis of randomized controlled trials. J. Gen. Intern. Med. 31 , 109–116 (2016).

Benatar, J. R. & Stewart, R. A. H. Cardiometabolic risk factors in vegans: a meta-analysis of observational studies. PLoS ONE 13 , e0209086 (2018).

Lee, Y.-M. et al. Effect of a brown rice based vegan diet and conventional diabetic diet on glycemic control of patients with type 2 diabetes: a 12-week randomized clinical trial. PLoS ONE 11 , e0155918 (2016).

Article   PubMed   PubMed Central   CAS   Google Scholar  

Jenkins, D. J. A. et al. Effect of a 6-month vegan low-carbohydrate (‘Eco-Atkins’) diet on cardiovascular risk factors and body weight in hyperlipidaemic adults: a randomised controlled trial. BMJ Open 4 , e003505 (2014).

Jenkins, D. J. A. et al. The effect of a plant-based low-carbohydrate (“Eco-Atkins”) diet on body weight and blood lipid concentrations in hyperlipidemic subjects. Arch. Intern. Med. 169 , 1046–1054 (2009).

Barnard, N. D. et al. A low-fat vegan diet and a conventional diabetes diet in the treatment of type 2 diabetes: a randomized, controlled, 74-wk clinical trial. Am. J. Clin. Nutr. https://doi.org/10.3945/ajcn.2009.26736H (2009).

Kahleova, H., Hill M. & Pelikánova, T. Vegetarian vs. conventional diabetic diet—a 1-year follow-up. Cor Vasa 56 . https://doi.org/10.1016/j.crvasa.2013.12.004 (2016).

Article   Google Scholar  

Turner-McGrievy, G. M., Davidson, C. R., Wingard, E. E., Wilcox, S. & Frongillo, E. A. Comparative effectiveness of plant-based diets for weight loss: a randomized controlled trial of five different diets. Nutrition 31 , 350–358 (2015).

Wing, R. R. & Phelan, S. Long-term weight loss maintenance. Am. J. Clin. Nutr. 82 , 222S–225S (2005).

David, L. A. et al. Diet rapidly and reproducibly alters the human gut microbiome. Nature 505 , 559–563 (2014).

Wu, G. D. et al. Linking long-term dietary patterns with gut microbial enterotypes. Science (80-) 334 , 105–108 (2011).

Article   CAS   Google Scholar  

Kaartinen, K. et al. Vegan diet alleviates fibromyalgia symptoms. Scand. J. Rheumatol. 29 , 308–313 (2000).

Yadav, V. et al. Low-fat, plant-based diet in multiple sclerosis: a randomized controlled trial. Mult. Scler. Relat. Disord. 9 , 80–90 (2016).

Rauma, A. L., Nenonen, M., Helve, T. & Hänninen, O. Effect of a strict vegan diet on energy and nutrient intakes by Finnish rheumatoid patients. Eur. J. Clin. Nutr. 47 , 747–749 (1993).

CAS   PubMed   Google Scholar  

Elkan, A.-C. et al. Gluten-free vegan diet induces decreased LDL and oxidized LDL levels and raised atheroprotective natural antibodies against phosphorylcholine in patients with rheumatoid arthritis: a randomized study. Arthritis Res. Ther. 10 , R34 (2008).

Karlsson, J. et al. Predictors and effects of long-term dieting on mental well-being and weight loss in obese women. Appetite 23 , 15–26 (1994).

Beezhold, B. L. & Johnston, C. S. Restriction of meat, fish, and poultry in omnivores improves mood: a pilot randomized controlled trial. Nutr. J. 11 , 9 (2012).

Yokoyama, Y., Barnard, N. D., Levin, S. M. & Watanabe, M. Vegetarian diets and glycemic control in diabetes: a systematic review and meta-analysis. Cardiovasc. Diagn. Ther. 4 , 373–382 (2014).

PubMed   PubMed Central   Google Scholar  

Sutliffe, J. T., Wilson, L. D., de Heer, H. D., Foster, R. L. & Carnot, M. J. C-reactive protein response to a vegan lifestyle intervention. Complement Ther. Med. 23 , 32–37 (2015).

Strasser, B., Gostner, J. M. & Fuchs, D. Mood, food, and cognition: role of tryptophan and serotonin. Curr. Opin. Clin. Nutr. Metab. Care 19 , 55–61 (2016).

O’Mahony, S. M., Clarke, G., Borre, Y. E., Dinan, T. G. & Cryan, J. F. Serotonin, tryptophan metabolism and the brain-gut-microbiome axis. Behav. Brain Res. 277 , 32–48 (2015).

Article   PubMed   CAS   Google Scholar  

Davey, G. K. et al. EPIC–Oxford: lifestyle characteristics and nutrient intakes in a cohort of 33 883 meat-eaters and 31 546 non meat-eaters in the UK. Public Health Nutr. 6 , 259–268 (2003).

Schüpbach, R., Wegmüller, R., Berguerand, C., Bui, M. & Herter-Aeberli, I. Micronutrient status and intake in omnivores, vegetarians and vegans in Switzerland. Eur. J. Nutr. 56 , 283–293 (2017).

Clarys, P. et al. Comparison of nutritional quality of the vegan, vegetarian, semi-vegetarian, pesco-vegetarian and omnivorous diet. Nutrients 6 , 1318–1332 (2014).

Park, J. E., Miller, M., Rhyne, J., Wang, Z. & Hazen, S. L. Differential effect of short-term popular diets on TMAO and other cardio-metabolic risk markers. Nutr. Metab. Cardiovasc. Dis. 29 , 513–517 (2019).

Psichas, A. et al. The short chain fatty acid propionate stimulates GLP-1 and PYY secretion via free fatty acid receptor 2 in rodents. Int J. Obes. 39 , 424 (2015).

Lin, H. V. et al. Butyrate and propionate protect against diet-induced obesity and regulate gut hormones via free fatty acid receptor 3-independent mechanisms. PLoS ONE 7 , e35240 (2012).

Canfora, E. E., Jocken, J. W. & Blaak, E. E. Short-chain fatty acids in control of body weight and insulin sensitivity. Nat. Rev. Endocrinol. 11 , 577 (2015).

Guo, Y. et al. Physiological evidence for the involvement of peptide YY in the regulation of energy homeostasis in humans. Obesity 14 , 1562–1570 (2006).

Holzer, P., Reichmann, F. & Farzi, A. Neuropeptide Y, peptide YY and pancreatic polypeptide in the gut–brain axis. Neuropeptides 46 , 261–274 (2012).

Kendall, C. W. C., Esfahani, A. & Jenkins, D. J. A. The link between dietary fibre and human health. Food Hydrocoll. 24 , 42–48 (2010).

Reynolds, A. et al. Carbohydrate quality and human health: a series of systematic reviews and meta-analyses. Lancet . https://doi.org/10.1016/S0140-6736(18)31809-9 (2019).

Menni, C. et al. Gut microbiome diversity and high-fibre intake are related to lower long-term weight gain. Int. J. Obes. 41 , 1099 (2017).

Van Gaal, L. F., Mertens, I. L. & Christophe, E. Mechanisms linking obesity with cardiovascular disease. Nature 444 , 875 (2006).

Ferreira, C. M. et al. The central role of the gut microbiota in chronic inflammatory diseases. J. Immunol. Res. 2014 , https://doi.org/10.1155/2014/689492 (2014).

Wersching, H. et al. Serum C-reactive protein is linked to cerebral microstructural integrity and cognitive function. Neurology 74 , 1022–1029 (2010).

Gu, Y. et al. Circulating inflammatory biomarkers in relation to brain structural measurements in a non-demented elderly population. Brain Behav. Immun. 65 , 150–160 (2017).

Lampe, L. et al. Visceral obesity relates to deep white matter hyperintensities via inflammation. Ann. Neurol. 85 , 194–203 (2018).

Google Scholar  

Schmidt, R. et al. Early inflammation and dementia: a 25‐year follow‐up of the Honolulu‐Asia Aging Study. Ann. Neurol. 52 , 168–174 (2002).

Rosano, C., Marsland, A. L. & Gianaros, P. J. Maintaining brain health by monitoring inflammatory processes: a mechanism to promote successful aging. Aging Dis. 3 , 16 (2012).

PubMed   Google Scholar  

Tangney, C. C. et al. Relation of DASH-and Mediterranean-like dietary patterns to cognitive decline in older persons. Neurology 83 , 1410–1416 (2014).

Craddock, J. C., Probst, Y. & Peoples, G. Vegetarian nutrition—comparing physical performance of omnivorous and vegetarian athletes. J. Nutr. Intermed. Metab. 4 , 19 (2016).

Liu, R. H. Health benefits of fruit and vegetables are from additive and synergistic combinations of phytochemicals. Am. J. Clin. Nutr. 78 , 517S–520S (2003).

Boffetta, P. et al. Fruit and vegetable intake and overall cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC). J. Natl. Cancer Inst. 102 , 529–537 (2010).

Reczek, C. R. & Chandel, N. S. Revisiting vitamin C and cancer. Science (80-) 350 , 1317–1318 (2015).

Probst, Y. C., Guan, V. X. & Kent, K. Dietary phytochemical intake from foods and health outcomes: a systematic review protocol and preliminary scoping. BMJ Open 7 , e013337 (2017).

Hartmann, R. & Meisel, H. Food-derived peptides with biological activity: from research to food applications. Curr. Opin. Biotechnol. 18 , 163–169 (2007).

Tillisch, K. et al. Consumption of fermented milk product with probiotic modulates brain activity. Gastroenterology 144 , 1394–1401 (2013).

Gibson, G. R. et al. Expert consensus document: The International Scientific Association for Probiotics and Prebiotics (ISAPP) consensus statement on thedefinition and scope of prebiotics. Nat. Rev. Gastroenterol. Hepatol. 14 , 491 (2017).

Nisha, A. R. Antibiotic residues-a global health hazard. Vet. World 1 , 375–377 (2008).

Wang, H. et al. Antibiotic residues in meat, milk and aquatic products in Shanghai and human exposure assessment. Food Control 80 , 217–225 (2017).

Bertazzi, P. A. et al. Health effects of dioxin exposure: a 20-year mortality study. Am. J. Epidemiol. 153 , 1031–1044 (2001).

Bouvard, V. et al. Carcinogenicity of consumption of red and processed meat. Lancet Oncol. 16 , 1599–1600 (2015).

Van Audenhaege, M. et al. Impact of food consumption habits on the pesticide dietary intake: comparison between a French vegetarian and the general population. Food Addit. Contam . 26 , 1372–1388 (2009).

Jacobs, D. R. & Tapsell, L. C. Food synergy: the key to a healthy diet. Proc. Nutr. Soc. 72 , 200–206 (2013).

Kawano, Y. & Yanai, K. Foodcam: a real-time food recognition system on a smartphone. Multimed. Tools Appl. 74 , 5263–5287 (2015).

Garcia-Perez, I. et al. Objective assessment of dietary patterns by use of metabolic phenotyping: a randomised, controlled, crossover trial. Lancet Diabetes Endocrinol. 5 , 184–195 (2017).

Turner-McGrievy, G. M. et al. Changes in nutrient intake and dietary quality among participants with type 2 diabetes following a low-fat vegan diet or a conventional diabetes diet for 22 weeks. J. Am. Diet. Assoc. 108 , 1636–1645 (2008).

Gilsing, A. M. J. et al. Serum concentrations of vitamin B12 and folate in British male omnivores, vegetarians and vegans: results from a cross-sectional analysis of the EPIC-Oxford cohort study. Eur. J. Clin. Nutr. 64 , 933–939 (2010).

Allen, L. H. How common is vitamin B-12 deficiency? Am. J. Clin. Nutr. 89 , 693S–696S (2008).

Pawlak, R., Parrott, S. J., Raj, S., Cullum-Dugan, D. & Lucus, D. How prevalent is vitamin B12 deficiency among vegetarians? Nutr. Rev. 71 , 110–117 (2013).

Rizzo, G. et al. Vitamin B12 among vegetarians: status, assessment and supplementation. Nutrients 8 , 767 (2016).

Article   PubMed Central   CAS   Google Scholar  

Stabler, S. P. Vitamin B12 deficiency. N. Engl. J. Med . 368 , 149–160 (2013).

Köbe, T. et al. Vitamin B-12 concentration, memory performance, and hippocampal structure in patients with mild cognitive impairment, 2. Am. J. Clin. Nutr. 103 , 1045–1054 (2016).

Ganguly, P. & Alam, S. F. Role of homocysteine in the development of cardiovascular disease. Nutr. J. 14 , 6 (2015).

McCaddon, A., Regland, B., Hudson, P. & Davies, G. Functional vitamin B12 deficiency and Alzheimer disease. Neurology 58 , 1395–1399 (2002).

Moore, E. et al. Cognitive impairment and vitamin B12: a review. Int. Psychogeriatr. 24 , 541–556 (2012).

Spence, J. D. Metabolic vitamin B12 deficiency: a missed opportunity to prevent dementia and stroke. Nutr. Res. 36 , 109–116 (2016).

Nexo, E. & Hoffmann-Lücke, E. Holotranscobalamin, a marker of vitamin B-12 status: analytical aspects and clinical utility. Am. J. Clin. Nutr. 94 , 359S–365S (2011).

Haider, L. M., Schwingshackl, L., Hoffmann, G. & Ekmekcioglu, C. The effect of vegetarian diets on iron status in adults: a systematic review and meta-analysis. Crit. Rev. Food Sci. Nutr. 58 , 1359–1374 (2018).

Lozoff, B. & Georgieff, M. K. et al. Iron deficiency and brain development. Semin. Pediatr. Neurol. 13 , 158–165 (2006).

Ayton, S. et al. Brain iron is associated with accelerated cognitive decline in people with Alzheimer pathology. Mol. Psychiatry 1 , https://doi.org/10.1038/s41380-019-0375-7 (2019).

Murray-Kolb, L. E. & Beard, J. L. Iron treatment normalizes cognitive functioning in young women. Am. J. Clin. Nutr. 85 , 778–787 (2007).

Beard, J. Iron deficiency alters brain development and functioning. J. Nutr. 133 , 1468S–1472S (2003).

Melina, V., Craig, W. & Levin, S. Position of the Academy of Nutrition and Dietetics: vegetarian diets. J. Acad. Nutr. Diet. 116 , 1970–1980 (2016).

Richter, M. et al. For the German Nutrition Society (DGE)(2016) Vegan diet. Position of the German Nutrition Society (DGE). Ernaehrungsumschau 63 , 92–102 (2016).

Peterson, J. et al. The NIH human microbiome project. Genome Res. 19 , 2317–2323 (2009).

Arumugam, M. et al. Enterotypes of the human gut microbiome. Nature 473 , 174–180 (2013).

Bamberger, C. et al. A walnut-enriched diet affects gut microbiome in healthy Caucasian subjects: a randomized, controlled trial. Nutrients 10 , 244 (2018).

Holscher, H. D. et al. Walnut consumption alters the gastrointestinal microbiota, microbially derived secondary bile acids, and health markers in healthy adults: a randomized controlled trial. J. Nutr. 148 , 861–867 (2018).

Hjorth, M. F. et al. Pre-treatment microbial Prevotella-to-Bacteroides ratio, determines body fat loss success during a 6-month randomized controlled diet intervention. Int J. Obes. 42 , 580 (2018).

Hansen, T. H. et al. Impact of a vegan diet on the human salivary microbiota. Sci. Rep. 8 , 5847 (2018).

Kim, M., Hwang, S., Park, E. & Bae, J. Strict vegetarian diet improves the risk factors associated with metabolic diseases by modulating gut microbiota and reducing intestinal inflammation. Environ. Microbiol. Rep. 5 , 765–775 (2013).

Zimmer, J. et al. A vegan or vegetarian diet substantially alters the human colonic faecal microbiota. Eur. J. Clin. Nutr. 66 , 53–60 (2012).

De Filippis, F., Pellegrini, N., Laghi, L., Gobbetti, M. & Ercolini, D. Unusual sub-genus associations of faecal Prevotella and Bacteroides with specific dietary patterns. Microbiome 4 , 57 (2016).

Kumar, M., Babaei, P., Ji, B. & Nielsen, J. Human gut microbiota and healthy aging: Recent developments and future prospective. Nutr. Health Aging 4 , 3–16 (2016).

Wu, G. D. et al. Comparative metabolomics in vegans and omnivores reveal constraints on diet-dependent gut microbiota metabolite production. Gut 65 , 63–72 (2014).

Morrison, D. J. & Preston, T. Formation of short chain fatty acids by the gut microbiota and their impact on human metabolism. Gut Microbes 7 , 189–200 (2016).

Wanders, A. J. et al. Effects of dietary fibre on subjective appetite, energy intake and body weight: a systematic review of randomized controlled trials. Obes. Rev. 12 , 724–739 (2011).

Brunkwall, L. & Orho-Melander, M. The gut microbiome as a target for prevention and treatment of hyperglycaemia in type 2 diabetes: from current human evidence to future possibilities. Diabetologia 60 , 943–951 (2017).

Lach, G., Schellekens, H., Dinan, T. G. & Cryan, J. F. Anxiety, depression, and the microbiome: a role for gut peptides. Neurotherapeutics 15 , 36–59 (2018).

Bagga, D. et al. Influence of 4-week multi-strain probiotic administration on resting-state functional connectivity in healthy volunteers. Eur. J. Nutr. 58 , 1821–1827 (2018).

Foster, J. A. & Neufeld, K.-A. M. Gut–brain axis: how the microbiome influences anxiety and depression. Trends Neurosci. 36 , 305–312 (2013).

Saulnier, D. M. et al. The intestinal microbiome, probiotics and prebiotics in neurogastroenterology. Gut Microbes 4 , 17–27 (2013).

Maes, M., Kubera, M., Leunis, J.-C. & Berk, M. Increased IgA and IgM responses against gut commensals in chronic depression: further evidence for increased bacterial translocation or leaky gut. J. Affect Disord. 141 , 55–62 (2012).

Huang, R., Wang, K. & Hu, J. Effect of probiotics on depression: a systematic review and meta-analysis of randomized controlled trials. Nutrients 8 , 483 (2016).

Heintz-Buschart, A. et al. Integrated multi-omics of the human gut microbiome in a case study of familial type 1 diabetes. Nat. Microbiol. 2 , 16180 (2016).

Strang, S. et al. Impact of nutrition on social decision making. Proc. Natl Acad. Sci. 114 , 6510–6514 (2017).

Osadchiy, V. et al. Correlation of tryptophan metabolites with connectivity of extended central reward network in healthy subjects. PLoS ONE 13 , e0201772 (2018).

Franzosa, E. A. et al. Sequencing and beyond: integrating molecular’omics’ for microbial community profiling. Nat. Rev. Microbiol. 13 , 360–372 (2015).

Fernandez-Real, J.-M. et al. Gut microbiota interacts with brain microstructure and function. J. Clin. Endocrinol. Metab. 100 , 4505–4513 (2015).

Allen, A. P. et al. Bifidobacterium longum 1714 as a translational psychobiotic: modulation of stress, electrophysiology and neurocognition in healthy volunteers. Transl. Psychiatry 6 , e939 (2016).

Harach, T. et al. Reduction of Abeta amyloid pathology in APPPS1 transgenic mice in the absence of gut microbiota. Sci. Rep. 7 , 41802 (2017).

Huhn, S., Masouleh, S. K., Stumvoll, M., Villringer, A. & Witte, A. V. Components of a Mediterranean diet and their impact on cognitive functions in aging. Front Aging Neurosci 7 , 132 (2015).

Larsson, S. C., Wallin, A. & Wolk, A. Dietary approaches to stop hypertension diet and incidence of stroke: results from 2 prospective cohorts. Stroke 47 , 986–990 (2016).

van de Rest, O. et al. Effect of fish oil on cognitive performance in older subjects: a randomized, controlled trial. Neurology 71 , 430–438 (2008).

Witte, A. V. et al. Long-chain omega-3 fatty acids improve brain function and structure in older adults. Cereb. Cortex 24 , 3059–3068 (2013).

Witte, A. V., Kerti, L., Margulies, D. S. & Flöel, A. Effects of resveratrol on memory performance, hippocampal functional connectivity, and glucose metabolism in healthy older adults. J. Neurosci. 34 , 7862–7870 (2014).

Brickman, A. M. et al. Enhancing dentate gyrus function with dietary flavanols improves cognition in older adults. Nat. Neurosci. 17 , 1798 (2014).

Martínez-González, M. A. et al. Benefits of the Mediterranean diet: insights from the PREDIMED study. Prog. Cardiovasc. Dis. 58 , 50–60 (2015).

Huhn, S. et al. Effects of resveratrol on memory performance, hippocampus connectivity and microstructure in older adults—a randomized controlled trial. Neuroimage (2018).

Rosenberg, A. et al. Multidomain lifestyle intervention benefits a large elderly population at risk for cognitive decline and dementia regardless of baseline characteristics: The FINGER trial. Alzheimer’s. Dement. 14 , 263–270 (2018).

Turnbaugh, P. J. Microbes and diet-induced obesity: fast, cheap, and out of control. Cell Host Microbe 21 , 278–281 (2017).

Turner-Mc Grievy, G. M., Barnard, N. D. & Scialli, A. R. A two-year randomized weight loss trial comparing a vegan diet to a more moderate low-fat diet*. Obesity 15 , 2276–2281 (2007).

Burke, L. E. et al. A randomized clinical trial of a standard versus vegetarian diet for weight loss: the impact of treatment preference. Int. J. Obes. 32 , 166–176 (2008).

Barnard, N. D. et al. A low-fat vegan diet and a conventional diabetes diet in the treatment of type 2 diabetes: a randomized, controlled, 74-wk clinical trial. Am. J. Clin. Nutr. 89 , 1588S–1596S (2009).

Marniemi, J., Seppänen, A. & Hakala, P. Long-term effects on lipid metabolism of weight reduction on lactovegetarian and mixed diet. Int. J. Obes. 14 , 113–125 (1990).

Acharya, S. D., Brooks, M. M., Evans, R. W., Linkov, F. & Burke, L. E. Weight loss is more important than the diet type in improving adiponectin levels among overweight/obese adults. J. Am. Coll. Nutr. 32 , 264–271 (2013).

Wright, N., Wilson, L., Smith, M., Duncan, B. & McHugh, P. The BROAD study: A randomised controlled trial using a whole food plant-based diet in the community for obesity, ischaemic heart disease or diabetes. Nutr. Diabetes 7 , e256 (2017).

Turner-McGrievy, G. M., Davidson, C. R., Wingard, E. E. & Billings, D. L. Low glycemic index vegan or low-calorie weight loss diets for women with polycystic ovary syndrome: a randomized controlled feasibility study. Nutr. Res. 34 , 552–558 (2014).

Kahleova, H. et al. Vegetarian diet improves insulin resistance and oxidative stress markers more than conventional diet in subjects with Type 2 diabetes. Diabet. Med 28 , 549–559 (2011).

Ferdowsian, H. R. et al. A multicomponent intervention reduces body weight and cardiovascular risk at a GEICO corporate site. Am. J. Heal. Promot 24 , 384–387 (2010).

Mishra, S. et al. A multicenter randomized controlled trial of a plant-based nutrition program to reduce body weight and cardiovascular risk in the corporate setting: the GEICO study. Eur. J. Clin. Nutr. 67 , 718 (2013).

Agarwal, U. et al. A multicenter randomized controlled trial of a nutrition intervention program in a multiethnic adult population in the corporate setting reduces depression and anxiety and improves quality of life: the GEICO study. Am. J. Heal. Promot 29 , 245–254 (2015).

Kahleova, H., Dort, S., Holubkov, R. & Barnard, N. A plant-based high-carbohydrate, low-fat diet in overweight individuals in a 16-week randomized clinical trial: the role of carbohydrates. Nutrients 10 , 1302 (2018).

Barnard, N., Scialli, A. R., Bertron, P., Hurlock, D. & Edmonds, K. Acceptability of a therapeutic low-fat, vegan diet in premenopausal women. J. Nutr. Educ. 32 , 314–319 (2000).

Gardner, C. D. et al. The effect of a plant-based diet on plasma lipids in hypercholesterolemic adults: a randomized trial. Ann. Intern. Med. 142 , 733 (2005).

Macknin, M. et al. Plant-based, no-added-fat or American Heart Association diets: impact on cardiovascular risk in obese children with hypercholesterolemia and their parents. J. Pediatr. 166 , 953–959 (2015).

Sciarrone, S. E. et al. Biochemical and neurohormonal responses to the introduction of a lacto-ovovegetarian diet. J. Hypertens. 11 , 849–860 (1993).

Alleman, R. J., Harvey, I. C., Farney, T. M. & Bloomer, R. J. Both a traditional and modified Daniel Fast improve the cardio-metabolic profile in men and women. Lipids Health Dis. 12 , 114 (2013).

Neacsu, M., Fyfe, C., Horgan, G. & Johnstone, A. M. Appetite control and biomarkers of satiety with vegetarian (soy) and meat-based high-protein diets for weight loss in obese men: a randomized crossover trial–. Am. J. Clin. Nutr. 100 , 548–558 (2014).

Koebnick, C. et al. Double-blind, randomized feedback control fails to improve the hypocholesterolemic effect of a plant-based low-fat diet in patients with moderately elevated total cholesterol levels. Eur. J. Clin. Nutr. 58 , 1402 (2004).

Kjeldsen-Kragh, J., Haugen, M., Førre, Ø., Laache, H. & Malt, U. F. Vegetarian diet for patients with rheumatoid arthritis: can the clinical effects be explained by the psychological characteristics of the patients? Rheumatology 33 , 569–575 (1994).

Bunner, A. E., Agarwal, U., Gonzales, J. F., Valente, F. & Barnard, N. D. Nutrition intervention for migraine: a randomized crossover trial. J. Headache Pain. 15 , 69 (2014).

Kahleova, H., Hrachovinova, T., Hill, M. & Pelikanova, T. Vegetarian diet in type 2 diabetes–improvement in quality of life, mood and eating behaviour. Diabet. Med 30 , 127–129 (2013).

Turner-McGrievy, G. M. et al. Randomization to plant-based dietary approaches leads to larger short-term improvements in Dietary Inflammatory Index scores and macronutrient intake compared with diets that contain meat. Nutr. Res. 35 , 97–106 (2015).

Download references

Acknowledgements

This work was supported by a scholarship (E.M.) by the German Federal Environmental Foundation and by the grants of the German Research Foundation contract grant number CRC 1052 “Obesity mechanisms” Project A1 (AV) and WI 3342/3-1 (A.V.W.).

Author information

Authors and affiliations.

Department of Neurology, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany

Evelyn Medawar, Arno Villringer & A. Veronica Witte

Berlin School of Mind and Brain, Humboldt-Universität zu Berlin, Berlin, Germany

Evelyn Medawar & Arno Villringer

Charité—Universitätsmedizin Berlin, Humboldt-Universität zu Berlin, Berlin, Germany

Helmholtz Centre for Environmental Research GmbH—UFZ, Leipzig, Germany

Sebastian Huhn

You can also search for this author in PubMed   Google Scholar

Contributions

E.M., A.V. and A.V.W. designed research; E.M. conducted research; E.M., S.H. and A.V.W. analyzed data; E.M. and A.V.W. wrote the paper; E.M., A.V. and A.V.W. had primary responsibility for final content. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Evelyn Medawar .

Ethics declarations

Conflict of interest.

The authors declare that they have no conflict of interest.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Suppl. table 1, suppl. figure 1, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Medawar, E., Huhn, S., Villringer, A. et al. The effects of plant-based diets on the body and the brain: a systematic review. Transl Psychiatry 9 , 226 (2019). https://doi.org/10.1038/s41398-019-0552-0

Download citation

Received : 20 February 2019

Revised : 22 June 2019

Accepted : 17 July 2019

Published : 12 September 2019

DOI : https://doi.org/10.1038/s41398-019-0552-0

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Nutrient scoring for the degs1-ffq – from food intake to nutrient intake.

• Ronja Thieleking
• Lennard Schneidewind
• Evelyn Medawar

BMC Nutrition (2023)

Environmental pressures and pesticide exposure associated with an increase in the share of plant-based foods in the diet

• Emmanuelle Kesse-Guyot
• Benjamin Allès
• Julia Baudry

Scientific Reports (2023)

Animal protein intake is directly associated with serum level of pentraxin 3 in hemodialysis patients

• Fatemeh Navab
• Sahar Foshati
• Mohammad Hossein Rouhani

Integrated multiomic wastewater-based epidemiology can elucidate population-level dietary behaviour and inform public health nutrition assessments

• Devin A. Bowes
• Erin M. Driver
• Rolf U. Halden

Nature Food (2023)

Influence of the vegan, vegetarian and omnivore diet on the oral health status in adults: a systematic review and meta-analysis

• Luana Giò Azzola
• Nicolas Fankhauser
• Murali Srinivasan

Evidence-Based Dentistry (2023)

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies