Evaluation of Dietary Patterns and All-Cause Mortality

1Every 5 years, the US Department of Agriculture and Department of Health and Human Services convene the Dietary Guidelines Advisory Committee to review existing evidence on diet and health to inform the Dietary Guidelines for Americans. Over time, the committee’s focus has shifted away from single nutrients or foods and toward overall dietary patterns. Nutrient and food analyses cannot account for the interactions among or the degree of independent variation of coingested nutrients and food components. To advance nutrition research and inform the 2020-2025 Dietary Guidelines for Americans, understanding the role of dietary patterns in optimizing health and reducing the risk of chronic disease across the lifespan is a high priority.

2Dietary patterns are the quantities, proportions, variety, or combination of different foods, beverages, and nutrients in diets as well as the frequency with which they are habitually consumed., Different approaches can be used to study dietary patterns. A priori methods are based on scientific consensus or evidence-based approaches and use scores that reflect the degree of adherence.,, A posteriori methods identify which factors explain the variation in patterns or aggregate individuals into groups with nonoverlapping patterns. Other approaches can include hybrid methods, clinical trials that assign consumption to a specific pattern, and/or observational studies on food avoidance (eg, vegetarian diets).

3The 2015 committee reviewed the associations between dietary patterns and multiple outcomes, including cause-specific mortality from cardiovascular disease (CVD) and dementia. Other literature reviews have examined specific diets or foods and cause-specific mortality.,,, In line with these studies, the 2020 committee, with support from the Nutrition Evidence Systematic Review (NESR) team of the US Department of Agriculture, conducted a systematic review of the literature to ascertain the association between dietary patterns consumed and all-cause mortality (ACM).

4The systematic review question was, “what is the relationship between dietary patterns consumed and ACM?” It also addressed the topic of diets on the basis of macronutrient distribution. Presented in full detail elsewhere, the review used NESR systematic review methods, which are summarized herein. We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.

5The committee developed a protocol, including an analytical framework (Figure 1) and inclusion and exclusion criteria (Table). The protocol was made available for public comment and was discussed by the committee in public meetings before evidence synthesis.

6The NESR librarians developed and implemented literature searches in PubMed, the Cochrane Central Register of Controlled Trials, and Embase to identify articles that were published from January 1, 2000, to October 4, 2019. These studies evaluated dietary patterns and ACM in participants aged 2 years and older. Two NESR analysts independently screened the search results to identify articles that met the predetermined criteria (Table), and then they manually searched the reference lists of the included articles. The NESR staff extracted data from and completed 2 independent risk-of-bias assessments of each included article.

7The 2020 committee synthesized the evidence qualitatively. The evidence was synthesized not by diet type or label but rather by similarities and differences in the foods and beverages that made up the dietary patterns that were examined. This synthesis approach was taken because the results of this systematic review were intended to inform dietary guidance and to allow for conclusions to be drawn regarding which dietary pattern components were most associated with risk of ACM.

8The results of the systematic review were organized according to a dietary pattern approach (a priori, a posteriori, and other). Based on the synthesized evidence, the committee developed conclusion statements to answer the systematic review question and graded the strength of evidence for the conclusions as follows: strong, moderate, limited, or grade not assignable. Grades were assigned using the NESR’s predefined criteria for grading elements (risks of bias, consistency, directness, precision, and generalizability) and approach, which takes study design into consideration. Future research recommendations for strengthening the body of evidence were based on the gaps and limitations that were identified during the systematic review process.

9After dual screening, 11 547 relevant studies were identified. The body of evidence included 153 articles (involving 6 550 664 individuals), of which 1 was from a randomized clinical trial (RCT) and 152 were from observational studies,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,(Figure 2). Multiple articles used data from the same study but used different methods or represented unique subsamples or dietary patterns. Studies enrolled adults and older adults (aged 17-84 years at baseline) from 28 countries with a high or very high Human Development Index classification (Australia, Belgium, Canada, Czech Republic, Croatia, Denmark, Finland, France, Germany, Greece, Hong Kong, Hungary, Iran, Italy, Japan, Korea, Norway, Netherlands, Poland, Portugal, Russia, Serbia, Singapore, Spain, Sweden, Switzerland, United Kingdom, and United States). Fifty-three studies,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, originated from the US. Some studies exclusively enrolled female,,,,,,,,,,,,,, or male,,,,,,,,,,,,,,, participants.

10The approaches for assessing dietary patterns included an RCT; a priori methods, such as index or score analysis,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,; a posteriori methods, such as factor or cluster analysis,,,,,,,,,,,,,,,,,,,,,,,,; and other methods,,,,,,,,,,, such as reduced rank regression and comparisons by animal product avoidance or ultraprocessed food consumption. eTable 1 in the Supplement describes the 185 dietary patterns that were examined. The ACM data were collected through validated methods (population-level registries, electronic databases, or medical or historical records) or active follow-up.

11In the RCT, participants at high-risk for CVD were randomized to a Mediterranean diet with extra-virgin olive oil or mixed nuts or to a control diet. Consumption of either intervention diet was associated with reduced ACM risk after a median follow-up of 4.8 years. Results were similar in subanalyses, which removed participants with protocol deviations and randomization issues. Most included studies of dietary patterns (110 of 153 [72%]),,,,,,,,,,,,,,,,,,,, used index or score analysis. Many different measures were reported, including variations of Mediterranean diet (n = 31), DASH (Dietary Approaches to Stop Hypertension) diet scores (n = 1), country-specific indices (n = 16), Healthy Eating Index or Dietary Guidelines for Americans scores (n = 7), and other indices or scales (n = 24) (eTable 1 in the Supplement). A complete description of these components and the scoring procedures has been published elsewhere.

12Findings were consistent across the studies, despite the variety of indices or scores used (eTable 2 in the Supplement). Nutrient-dense dietary patterns, regardless of pattern label or name, were associated with significantly lower ACM risk. For example, all 12 comparisons that examined DASH diet scores suggested that higher adherence was associated with lower ACM risk,,,,,,,,,,,, and 54 of the 63 comparisons that were reported in 44 articles that examined a Mediterranean-type index or score suggested that higher adherence was associated with lower ACM risk.,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,

13A few articles reported nonsignificant results with similar patterns.,,,,,,,,, Common elements of these dietary indices or scores were (1) relatively higher intake of vegetables (with or without potatoes); legumes; fruit; nuts; either whole grains specifically, cereals unspecified, or nonrefined grains; fish and/or seafood; lean meat or poultry, when included (Dietary Guidelines for Americans or Healthy Eating Index scores reflect total protein foods and/or ratio of white to red meat); and unsaturated fats vs saturated fats; and (2) relatively lower intake of red and processed meat and/or meat and meat products; refined grains; added sugars and/or sugar-sweetened beverages; solid fats, saturated fats, and/or trans fat; and excessive sodium.

14Many of the indices considered alcoholic beverage intake in low to moderate amounts or within a threshold (eg, 10-25 g/d; 0.5-1.5 drinks/d for women and 0.5-2.0 drinks/d for men) as a positive component within the context of the entire dietary pattern. However, measurement and scoring procedures varied between indices.

15The associations reported between a priori–derived dietary patterns and ACM remained in sensitivity or subgroup analyses after combining the dietary patterns with anthropometry, physical activity, and/or smoking status,,,,,; performing stratification or additional adjustment for anthropometry, sex, age, educational level, race/ethnicity, and/or smoking status,,,,,,,,,,,,,,,,,,,,,; excluding early deaths or the first few years of follow-up results,,,,,,,,,,,,,,,,,; accounting for chronic disease status (eg, diabetes or CVD,,,,,,,,,,,,,); or underreporting or misreporting chronic disease.,,, When adherence to dietary patterns was combined with other healthier lifestyle factors (eg, not smoking and meeting recommended physical activity levels), stronger associations were typically observed. For example, participants who adhered to a Mediterranean diet met recommended physical activity recommendations, were long-term nonsmokers, and had a lower mortality risk (relative risk [RR], 0.65; 95% CI, 0.63-0.68) compared to those that only followed the Mediterranean diet (RR, 0.86; 95% CI, 0.83-0.88).

16Most of the 25 articles (19 [76%]) that examined dietary patterns by using factor or cluster analysis reported significantly lower ACM risk when comparing higher with lower adherence to the same derived pattern,,,,,,,,,,,,,,,, or to different patterns, (eTable 2 in the Supplement). For example, participants with higher vs lower adherence to the Prudent pattern had lower mortality risk (Quintile 1: reference; Q2: RR, 0.85 [95% CI, 0.78-0.92; Q3: RR, 0.84 [95% CI, 0.78-0.91]; Q4: RR, 0.81 [95% CI, 0.74-0.88]; Q5: RR, 0.83 [95% CI, 0.76-0.90]; P for trend <.001). Participants with higher vs lower adherence to the Western pattern had higher mortality risk (Q1: reference; Q2: RR, 1.00 [95% CI, 0.92-1.08]; Q3: RR, 1.10 [95% CI, 1.02-1.20]; Q4: RR, 1.16 [95% CI, 1.06-1.26]; Q5: RR, 1.21 [95% CI, 1.12-1.32]; P for trend <.001). Dietary patterns were given different labels, such as healthy, prudent, and Mediterraneanlike, but were similarly characterized by higher intake of vegetables, fruits, fish or other seafood, legumes and/or whole grains, nuts, vegetable or olive oils, and/or poultry (eg, white meat, such as chicken or turkey).

17Several articles (5 of 25 [20%]) reported that the dietary patterns associated with significantly higher ACM risk (and/or shorter survival,,,,) emphasized the following commonalities: higher intake of (1) meat and meat products such as beef, pork, sausage; red meat and meat products; red meat and processed meats; fresh and processed meats and seafood; (2) high-fat dairy products such as ice cream, cheese, and whole milk; and/or (3) refined grains, or flour-based foods such as pastries and/or sweets and desserts,, such as cake, cookies, chocolate, and candy; as well as lower intake of (4) low-fat dairy products, rice and pasta, fruits, fish and other seafood, and dark green vegetables.

18Results were similar whether stratified by country, after excluding participants with CVD or cancer at baseline, or in models that accounted for co-twin pairs. Stratification by sex was inconsistent, with no significant associations in women only but significant associations in pooled analyses. For example, Krieger et al reported lower risk of ACM at mean follow-up of 25 years in overall analyses, with men and women pooled and when comparing fish (hazard ratio [HR], 0.87; 95% CI, 0.78-0.97) and traditional (HR, 0.89; 95% CI, 0.80-0.98) dietary patterns to the sausage and vegetables dietary pattern (reference; HR, 1.0). However, in the Krieger et al analyses, which were stratified by sex, the results in women only were not statistically significant: sausage and vegetables pattern (reference; HR, 1.0) was compared with meat and salad (HR, 0.93; 95% CI, 0.80-1.08), fish (HR, 0.98; 95% CI, 0.83-1.15), traditional (HR, 1.02; 95% CI, 0.87-1.19), and high-fiber foods (HR, 0.91; 95% CI, 0.79-1.05) patterns.

19Nonsignificant associations were reported between a posteriori–derived dietary patterns and ACM, although the direction generally aligned with the significant results we have described, and may be attributed to smaller sample sizes, the sample evaluated, or the gradient between exposure comparisons.

20Dietary patterns that were derived from reduced rank regression showed inconsistent results (eTable 2 in the Supplement).,, Five articles examined the dietary patterns based on avoidance of animal-based products (eg, vegetarian diets).,,,, Vegetarian or plant-based patterns were associated with significantly lower ACM risk in 2 studies,, but no significant associations were found in other exposure groups (eg, vegetarian vs pescovegetarian, and meat eaters vs vegetarian or vegan).,,

21Three studies reported that the dietary pattern of higher vs lower consumption of ultraprocessed foods was associated with higher ACM risk,, as defined by the fourth level of the NOVA Food Classification System. These patterns differed by the foods comprising the pattern (eg, some of which included highly palatable foods, such as ice cream and processed meats, and others included artificial flavors and texturizing agents), reducing the generalizability of these patterns and thus the results.

22Most studies were conducted with rigorous methods and at low or moderate risks of bias across domains. Most studies accounted for potential confounders except for race/ethnicity, which was often homogenous or not reported. In addition, most studies assessed dietary intake once and were, therefore, at risk of bias because of potential changes in dietary patterns among participants over time. However, these studies used validated methods, and their results aligned with findings of studies that assessed diet over time. Several studies did not account for missing data, primarily diet or exposure data at baseline. Participants with implausible energy intake, incomplete dietary data, and/or a history or presence of chronic diseases or medical conditions at baseline were typically removed from the studies or analyses.

23Dietary patterns that were examined with different approaches showed consistent direction and magnitude of associations with ACM. Precision was indicated by relatively narrow CIs between studies. Most studies did not report power analyses or sample size calculations but reported analytic sample sizes that were large enough to investigate the association, ranging from 161 participants to 451 256 participants. Total number of deaths ranged from 53 with approximately 4 years of total follow-up to 51 702 with approximately 13 to 18 years of follow-up. Although the incident number of deaths differed between studies, the number of events reported within groups confirmed precision across the body of evidence. The evidence base had directness and was generalizable to the US population. Results may be less generalizable to younger or less healthy populations. Results are likely generalizable to adults of various underreported racial/ethnic backgrounds, although it is difficult to ascertain exactly how race and/or ethnicity was involved in the association between dietary patterns and ACM because many studies did not report that information.

24The 2020 committee used multiple databases to obtain publications from a large, comprehensive search. Most of the studies identified had large prospective cohorts, but smaller sample sizes were also included. Therefore, risk of publication bias was low. Details regarding how the NESR accounted for publication bias have been described elsewhere. Primary and secondary analyses were included from the PREDIMED (Prevención con Dieta Mediterránea) trial, which reported randomization issues. However, PREDIMED investigators republished the data after reanalysis and reconfirmed the initial findings.

25Despite the different approaches, study designs, dietary assessment methods, geographical regions, and dietary pattern labels, the evidence demonstrated that dietary patterns associated with lower ACM risk were consistently characterized by higher intake of vegetables; legumes; fruits; nuts; either whole grains, cereals, or nonrefined grains; fish; and unsaturated vegetable oils. These patterns were also characterized by lower or no consumption of animal products (red and processed meat, meat and meat products, and high-fat dairy products), refined grains, and sweets (ie, higher in added sugars). Labels that were assigned to the dietary patterns varied widely (eg, Mediterranean, prudent, Healthy Eating Index, DASH, and plant-based), highlighting that high-quality diets with nutrient-dense foods are associated with better health, regardless of diet type or dietary pattern name. Although we believe this systematic review is the most comprehensive examination of US dietary patterns and ACM that is currently available, its findings align with the results of previous meta-analyses that focused on select diet types and/or cause-specific mortality.,,,,

26This systematic review included the most suitable study designs available to answer the question. Given the nature of the outcome (ACM), expense, duration, and power necessary for an experimental study design to explore this question, it is unlikely that many RCTs or non-RCTs would be available (funded, conducted, or published) for consideration. Studies that were included for this review were well designed and conducted using rigorous methods. Despite the preponderance of evidence from observational studies, this review presented a conclusion statement that was assigned a grade of strong, which was in accordance with studies at generally low or moderate risks of bias and with high consistency in direction and magnitude of findings, precision, directness, and generalizability. The methods used for grading the evidence underlying this conclusion statement align with other grading approaches,, and ensure that strengths and weaknesses in study design as well as each grading element were considered.

27The findings of this review are further supported by a central tenet behind other works of the 2020 committee: a nutrient-dense dietary pattern can minimize the risk of multiple diet-related chronic diseases, such as CVD, obesity, diabetes, and some cancers, and can support bone and neurocognitive health., For women who are pregnant, a similar healthful dietary pattern has been associated with lower risk of poor maternal-fetal outcomes. Achieving a healthy dietary pattern at each life stage should also support health in subsequent life stages. The core elements of these dietary patterns across the committee’s reviews reflect higher diet quality, are appropriate to consume across the lifespan, and have the potential to substantially minimize chronic disease risk and mortality risk.

28Studies in this review suggested that dietary patterns containing nutrient-dense foods and alcoholic beverages that were within a given threshold or at low to moderate levels were associated with lower risk of ACM. However, the preponderance of this evidence varied in the methodological assessment of alcohol intake and scoring procedures within the context of the dietary patterns that were assessed. The committee conducted a separate systematic review to answer the question, what is the association between alcoholic beverages and ACM? Briefly, the focused analysis on alcoholic beverage consumption suggested that alcohol may increase ACM risk, but only minimally at low levels of intake. The committee concluded that moderating alcohol consumption to lower levels is recommended to better protect health. Other recommendations that were incorporated into the Dietary Guidelines for Americans, 2020-2025 were for individuals to not start drinking, avoid heavy drinking, and proceed with caution regarding alcoholic beverage consumption because it can contribute to increased rates of liver disease and excess energy intake, which is not advisable.

29Dietary patterns provide a meaningful and interpretable database on foods and food groups. This review found that multiple food choices can be made toward a healthy diet that promotes beneficial outcomes. Dietary pattern approaches are advantageous because they place an emphasis on the combination of foods and beverages that meet total energy needs and are associated with health, instead of the selection of specific nutrients or foods alone, which may be more difficult for consumers to translate into a total diet. We believe such an approach provides flexibility, allowing consumers the freedom to tailor food and food group combinations that were identified as healthful to their preferences (eg, cultural acceptability and taste) and needs (eg, cost).

30This study has several limitations. Observing study participants over time provides valuable insights regarding risk for mortality, although many included studies examined diet at only 1 time point (ie, baseline), which may or may not reflect usual dietary patterns. Most of the available evidence on dietary patterns was derived from studies that were conducted in adults. Data on dietary patterns earlier in the life course and ACM were unavailable. However, this lack of data was likely associated with the challenge and expense of undertaking long-term RCTs and/or studies with repeated measures over time, particularly starting in childhood, that examine mortality as an outcome. As a result, longitudinal data, particularly information on dietary patterns from childhood and continuing throughout the life course, are still greatly needed. Further exploration is also needed on the factors that modify or mediate the association between dietary patterns and ACM as well as the role that selected food groups, such as meat, plays in this association. In addition, the available evidence does not identify the inadequate or excessive intake or quantify the types and amounts of foods and food groups consumed in the context of dietary patterns.

31Active ongoing cohorts informed most of the findings reported herein, which showed consistency across the US and more than 20 other middle- to high-income countries. Results indicated good external validity, but future work is needed within population subgroups that could not be analyzed in this review because of the lack of reported details on individuals from racial/ethnic minority groups. The available evidence was not sufficient to assess how race/ethnicity affects the association between dietary patterns and ACM. In addition, this systematic review was conducted to identify the nature and direction of the association between dietary patterns and ACM, with an intention to consider dietary patterns regardless of label, name, or type. Because of this comprehensive approach, the methodological heterogeneity of the dietary patterns included in this review was best suited for qualitative synthesis of quantitative data. However, in the future, it may be worth exploring whether and how the magnitude of these findings could be quantified through a meta-analysis.

32Evidence from this systematic review by the 2020 Dietary Guidelines Advisory Committee demonstrated that nutrient-dense dietary patterns, which were characterized by higher intake of vegetables, fruits, legumes, nuts, whole grains, unsaturated vegetable oils, fish, and lean meat or poultry, when included, were associated with decreased ACM risk in broadly generalizable populations of adults and older adults. These dietary patterns included relatively lower intake of red and processed meat, high-fat dairy, and refined carbohydrates or sweets. Some of these dietary patterns also included moderate intake of alcoholic beverages. Results based on additional analyses with confounding factors generally confirmed the robustness of the main findings.