Optimal dietary patterns for prevention of chronic disease

Abstract

Introduction

Chronic diseases account for more than half of all premature deaths and more than 90% of yearly healthcare spending in the United States^1,2. With 11 million deaths and 255 million disability-adjusted life-years globally attributable to poor nutrition, following a healthy diet can be a potentially cost-effective strategy for lowering the risk of chronic diseases^3,4. Much existing diet-related research, however, focuses on specific foods, and hence may not provide clear knowledge about the ideal diet for overall health. Dietary patterns, characterizing a variety of foods, nutrients, and beverages, may serve as useful tools to represent the overall effects of diet on the risk of health outcomes.

Dietary patterns that emphasize high-quality foods, adherence to dietary recommendations, and a focus on plant-based foods have been demonstrated to reduce the risk of cardiovascular disease (CVD), type 2 diabetes, cancer, and all-cause mortality⁵. Based on this evidence, the 2015-2020 Dietary Guidelines for Americans propose Healthy U.S.-Style, Mediterranean-style, vegetarian diets, and Dietary Approaches to Stop Hypertension as examples of healthy diets⁶. Despite this, only a few studies have explicitly examined these diets in the same context for their potential health impact.

The purpose of this study is to compare the relative effectiveness of dietary patterns in improving general health by focusing on risk reduction of chronic diseases that contribute substantially to mortality in the United States, including CVD, cancer, and diabetes. Using data from three US cohorts (Health Professionals Follow-up Study [HPFS], Nurses’ Health Study [NHS], and NHS II) with up to 32 years of follow-up, we compared several dietary patterns that were promoted in the Dietary Guidelines for Americans and previously created for predicting major chronic diseases or important biological pathways for their associations with major chronic diseases (Fig.1).

Open in a separate window

Fig.1

Detailed components of dietary patterns. Higher points or positive weights were assigned to higher intakes of components in green. Lower points or negative weights were assigned to higher intakes of components in red. Higher points were assigned to moderate intakes of components in blue. AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Results

Dietary patterns and major chronic diseases.

Table 2 shows the hazard ratios (HR) for major chronic disease (the composite of major cardiovascular disease, type 2 diabetes, and total cancer) comparing the 90^th (high adherence) to 10^th (low adherence) percentile scores for each dietary pattern. The multivariable-adjusted associations were strongest for the rEDIH (HR 0.58, 95% confidence interval [CI] 0.57, 0.60), rEDIP (HR 0.61, 95% CI 0.60, 0.63), and DRRD (HR 0.70, 95% CI 0.69, 0.72). Inverse associations with major chronic disease were also observed for AHEI-2010, AMED, hPDI, and DASH, with HRs between 0.76 and 0.80. In contrast, the WCRF/AICR dietary score was not associated with major chronic disease. Although the spline analysis demonstrated statistically significant nonlinearity for the AMED, WCRF/AICR dietary score, and rEDIP, the curves were largely monotonic for the rEDIP and AMED (Extended Data Fig.3) and the relationship between the WCRF/AICR dietary score and major chronic disease also appeared monotonic after the alcohol component was removed from the score (data not shown). Similar findings were observed when using pattern scores in quintiles as the exposure (Supplementary Table 3).

Table 2

Associations of cumulative average dietary patterns (comparing the 90^th to 10^th percentile) with major chronic disease and its major components in the pooled data

Outcome		Major Chronic Disease	Major Cardiovascular Disease	Type 2 Diabetes	Total Cancer
Cases		44,975	12,962	18,615	17,909
Person-year		4,852,894	5,148,378	4,630,725	5,138,951
Pattern	Model	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
AHEI-2010	Age	0.64 (0.62, 0.66)	0.63 (0.60, 0.66)	0.49 (0.47, 0.51)	0.81 (0.78, 0.84)
	MVa	0.76 (0.74, 0.78)	0.77 (0.73, 0.81)	0.62 (0.59, 0.64)	0.94 (0.90, 0.98)
	MV + nSESb	0.78 (0.76, 0.80)	0.79 (0.75, 0.83)	0.66 (0.63, 0.69)	0.93 (0.89, 0.97)
	MV + BMIc	0.83 (0.80, 0.85)	0.79 (0.75, 0.83)	0.73 (0.70, 0.77)	0.95 (0.91, 0.99)
AMED	Age	0.65 (0.64, 0.67)	0.64 (0.61, 0.67)	0.56 (0.53, 0.58)	0.78 (0.75, 0.81)
	MVa	0.79 (0.77, 0.81)	0.80 (0.76, 0.84)	0.71 (0.68, 0.74)	0.92 (0.89, 0.96)
	MV + nSESb	0.81 (0.78, 0.83)	0.82 (0.78, 0.86)	0.74 (0.71, 0.77)	0.92 (0.88, 0.96)
	MV + BMIc	0.86 (0.84, 0.88)	0.83 (0.79, 0.87)	0.85 (0.82, 0.89)	0.94 (0.90, 0.98)
hPDI	Age	0.70 (0.69, 0.72)	0.72 (0.69, 0.76)	0.57 (0.55, 0.59)	0.86 (0.83, 0.90)
	MVa	0.80 (0.78, 0.82)	0.84 (0.80, 0.88)	0.70 (0.67, 0.72)	0.94 (0.90, 0.98)
	MV + nSESb	0.80 (0.78, 0.83)	0.84 (0.80, 0.88)	0.71 (0.68, 0.73)	0.94 (0.90, 0.98)
	MV + BMIc	0.84 (0.82, 0.87)	0.85 (0.81, 0.90)	0.78 (0.75, 0.81)	0.95 (0.91, 0.99)
DASH	Age	0.63 (0.61, 0.64)	0.62 (0.59, 0.65)	0.52 (0.50, 0.54)	0.76 (0.73, 0.79)
	MVa	0.78 (0.76, 0.80)	0.81 (0.77, 0.85)	0.66 (0.64, 0.69)	0.94 (0.91, 0.99)
	MV + nSESb	0.79 (0.77, 0.81)	0.83 (0.79, 0.87)	0.69 (0.66, 0.72)	0.94 (0.90, 0.98)
	MV + BMIc	0.83 (0.81, 0.86)	0.83 (0.79, 0.87)	0.77 (0.74, 0.81)	0.96 (0.92, 1.00)
DRRD	Age	0.61 (0.59, 0.62)	0.62 (0.59, 0.65)	0.44 (0.43, 0.46)	0.83 (0.80, 0.87)
	MVa	0.70 (0.69, 0.72)	0.73 (0.69, 0.76)	0.56 (0.54, 0.58)	0.92 (0.88, 0.96)
	MV + nSESb	0.72 (0.70, 0.73)	0.74 (0.71, 0.78)	0.58 (0.56, 0.60)	0.92 (0.88, 0.96)
	MV + BMIc	0.76 (0.74, 0.78)	0.75 (0.71, 0.79)	0.66 (0.63, 0.69)	0.94 (0.90, 0.98)
WCRF/AICR	Age	0.84 (0.82, 0.86)	0.80 (0.77, 0.84)	0.86 (0.83, 0.90)	0.85 (0.81, 0.88)
	MVa	1.00 (0.98, 1.03)	1.00 (0.96, 1.05)	1.05 (1.01, 1.10)	1.01 (0.97, 1.05)
	MV + nSESb	1.01 (0.98, 1.04)	1.01 (0.96, 1.06)	1.07 (1.02, 1.11)	1.01 (0.97, 1.05)
	MV + BMIc	0.96 (0.94, 0.99)	0.99 (0.94, 1.03)	0.95 (0.91, 0.99)	1.00 (0.96, 1.04)
rEDIH	Age	0.53 (0.52, 0.54)	0.61 (0.59, 0.64)	0.30 (0.29, 0.31)	0.85 (0.82, 0.88)
	MVa	0.58 (0.57, 0.60)	0.68 (0.65, 0.71)	0.35 (0.34, 0.36)	0.90 (0.87, 0.94)
	MV + nSESb	0.59 (0.58, 0.61)	0.70 (0.66, 0.73)	0.36 (0.35, 0.37)	0.90 (0.86, 0.93)
	MV + BMIc	0.75 (0.73, 0.77)	0.76 (0.72, 0.79)	0.57 (0.54, 0.59)	0.95 (0.91, 0.99)
rEDIP	Age	0.59 (0.57, 0.60)	0.67 (0.64, 0.70)	0.33 (0.32, 0.34)	0.93 (0.90, 0.97)
	MVa	0.61 (0.60, 0.63)	0.69 (0.66, 0.72)	0.38 (0.36, 0.39)	0.90 (0.87, 0.94)
	MV + nSESb	0.62 (0.60, 0.63)	0.70 (0.67, 0.73)	0.38 (0.37, 0.40)	0.90 (0.87, 0.94)
	MV + BMIc	0.75 (0.73, 0.77)	0.75 (0.72, 0.79)	0.57 (0.55, 0.59)	0.94 (0.91, 0.98)

Open in a separate window

AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; BMI, body mass index; CI, Confidence interval; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; HR, Hazard ratio; MV, multivariable; nSES, neighborhood socioeconomic status; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

^aAnalyses were stratified by age (in month), calendar year, and cohort. MV model: adjusted for physical activity (<3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, 27.0-41.9, or ≥42 MET-h/week), cigarette smoking status (never, former quitting ≥10 y, former quitting <10 y, current), cigarette smoking pack-years (0, 1-4, 5-14, 15-24, or ≥25 pack-years), multivitamin use (yes or no), regular aspirin use (yes or no), regular non-steroidal anti-inflammatory drugs use (yes or no), postmenopausal hormone use (premenopausal, never, former, or current use) for women, and total energy intake (quintiles). Alcohol consumption (<5.0, 5.0-14.9, or ≥15.0 g/d) was adjusted for DASH, hPDI, and DRRD. For major chronic disease, the model was additionally adjusted for family history of diabetes (yes or no), family history of cancer (yes or no), and family history of cardiovascular disease (yes or no). For type 2 diabetes, the model was additionally adjusted for family history of diabetes (yes or no). For major cardiovascular disease, the model was additionally adjusted for family history of cardiovascular disease (yes or no). For total cancer, the model was additionally adjusted for family history of cancer (yes or no) and height (continuous).

^bAdjusted for nSES (continuous).

^cAdjusted for BMI (continuous).

We next investigated the associations between dietary pattern scores and each individual component of major chronic disease as presented in Fig.2, Table 2, and Extended Data Table 1. In general, the strongest relationships among the various outcomes (cardiovascular diseases, cancer, and type 2 diabetes) were those between dietary patterns and type 2 diabetes. Among the dietary patterns, rEDIH, rEDIP, and DRRD score had the strongest associations with type 2 diabetes (HR: 0.35–0.56). These three pattern scores were also among the top three patterns that were strongly associated with CVD-related outcomes, including major CVD, coronary heart disease, and stroke (HR: 0.63–0.77). The estimates for AHEI-2010 were comparable with DRRD score for coronary heart disease. The estimates for cancer-related outcomes were closer to null compared with type 2 diabetes or cardiovascular diseases, for all dietary patterns. Participants with higher rEDIH, rEDIP, or DRRD scores had a decreased risk of total cancer, obesity-related cancer, and smoking-related cancer (HR: 0.83–0.92). For total cancer and smoking-related cancer, participants with higher AHEI-2010, AMED, and DASH presented comparable or lower risk than those with higher rEDIH, rEDIP, or DRRD. The WCRF/AICR dietary score was not associated with secondary outcomes except for a positive association with type 2 diabetes (HR 1.05; 95% CI: 1.01–1.10). As in the primary analysis, dietary patterns were similarly associated with an alternatively defined major chronic disease, which includes non-traumatic death in addition to major chronic disease incidence.

Open in a separate window

Fig.2

Multivariable-adjusted associations of cumulative average dietary patterns (comparing the 90^th to 10^th percentile) with major chronic disease and secondary outcomes in the pooled data of three cohorts (n = 205,852 participants). The Analyses details and corresponding estimates are provided in Table 2 and Extended Data Table 1. The hazard ratios are indicated by the circles and the 95% confidence intervals are reflected by the error bars. AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Fig.3 shows the associations between individual food groups with dietary patterns and major chronic diseases. The estimates for food groups were largely similar across major chronic diseases, although the associations were slightly stronger for type 2 diabetes. Coffee, whole grains, wine, desserts were inversely associated with major chronic disease, whereas processed meats, low-energy drinks, red meats, French fries, high-energy drinks, and eggs were positively associated with major chronic disease. Compared with other patterns, the rEDIH, rEDIP, and DRRD appeared to be more or at least similarly correlated with food groups that were associated with major chronic diseases (particularly type 2 diabetes), such as coffee, wine, processed meats, red meats, French fries, and sugar-sweetened beverages. The rEDIH and rEDIP had relatively lower correlations with fruit and vegetables compared to other dietary patterns except for leafy green vegetables.

Open in a separate window

Fig.3

Baseline Spearman correlations between energy-adjusted cumulative average dietary patterns and food groups in the pooled data of three cohorts (n = 205,852 participants). Spearman correlation coefficients are shown and highlighted in color. Food groups are ordered based on the hazard ratios of their associations with major chronic disease (Chr). The associations of food groups (comparing the 90^th to 10^th percentile) with major chronic disease (Chr), major cardiovascular disease (CVD), type 2 diabetes (T2D), or total cancer (CA) are indicated on the left of the figure. We reported unadjusted P values based on two-sided statistical tests. Significant associations (P<0.05) are highlighted in color according to the magnitude of hazard ratios. AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Subgroup and latency analyses.

The associations between patterns and major chronic disease persisted in subgroups defined by age, BMI, sex, smoking status, alcohol intake, neighborhood socioeconomic status (nSES), and race/ethnicity(Fig.4, Extended Data Table 2). The inverse associations between dietary patterns and major chronic disease were generally stronger in participants who were younger, overweight or have obesity, women, and Hispanic. Although WCRF/AICR exhibited an inverse association with major chronic disease in participants who drank more alcohol, the rEDIH and rEDIP showed stronger associations in participants who never smoked or drank less. Comparable associations with major chronic disease were observed in subgroups defined by nSES. With longer latency periods, the associations between dietary patterns and major chronic disease were slightly attenuated (Extended Data Fig.4, Extended Data Table 3). A similar trend was observed for major cardiovascular disease and type 2 diabetes, but not for total cancer.

Open in a separate window

Fig.4

Multivariable-adjusted associations between cumulative average dietary patterns (comparing the 90^th to 10^th percentile) and major chronic disease in subgroups. Analyses details and corresponding estimates are provided in Extended Data Table 2. The hazard ratios are indicated by the circles and the 95% confidence intervals are reflected by the error bars. AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; BMI, body mass index; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; nSES, neighborhood socioeconomic status; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Discussion

The current understanding of dietary patterns on general health is based mostly on the synthesis of the literature for individual diseases rather than more composite outcomes of chronic diseases. Furthermore, the relationships between patterns and major chronic diseases have rarely been compared in the same study utilizing the same data collection and statistical analysis approach. In this study, we compared two mechanism-based dietary patterns with six dietary patterns reflecting general or disease-specific dietary guidelines in 205,852 participants over the course of more than two decades. In general, adherence to a healthy diet was associated with a decreased risk of major chronic diseases. Participants who reported high adherence to low insulinemic, low inflammatory, and diabetes risk-reducing dietary patterns displayed a decreased risk for major chronic diseases when examined as individual or composite outcome.

The rEDIH and rEDIP represent dietary patterns empirically constructed based on foods that are either positively or inversely associated with biomarkers of two important and related biological pathways for chronic disease—hyperinsulinemia and chronic inflammation^7,8. Assuming causal associations, the dietary patterns can be interpreted as individuals with high adherence to the rEDIH and rEDIP tending to have lower markers of insulin and inflammation, though these biomarkers could have other determinants. Therefore, strong associations were observed between these two patterns with chronic disease risk, which can be largely shaped by their connections to pathophysiologic underpinnings of insulin resistance and type 2 diabetes. As shown in other studies, participants with low rEDIH or rEDIP scores had a higher risk of developing type 2 diabetes^9,10. Despite this, associations were observed between rEDIH or rEDIP and other elements of chronic diseases, which may be mediated by the same mechanistic pathways. In line with our findings, prior investigations in the NHS and HPFS showed an inverse association between rEDIP and CVD, as well as rEDIH and total mortality¹¹. Additionally, when compared to other patterns, the rEDIH and rEDIP showed greater associations with both composite and specific chronic diseases. Previous studies showed that these two diets are predictive of a diverse range of biomarkers for atherosclerosis, hyperlipidemia, hyperinsulinemia, and systemic inflammation^12,13, supporting the hypothesis that rEDIH and rEDIP address fundamental biological pathways that are shared by multiple chronic diseases. The weighting of foods by a measured biological response may also account for differential measurement errors among specific foods.

Although not as strong as that for rEDIH or rEDIP, we observed an inverse association between DRRD and major chronic disease. Consisting of dietary components with sufficient evidence for type 2 diabetes, the DRRD has been shown to be associated with risk of type 2 diabetes across diverse racial and ethnic populations¹⁴. Participants with higher adherence to DRRD were associated with a lower risk of death from all causes, cardiovascular disease, and cancer in a US-based study¹⁵, which is compatible with our findings, indicating that the preventive role of DRRD may extend beyond type 2 diabetes to cardiovascular disease, cancer, and death. One possible explanation might be that a diet targeting type 2 diabetes captures dietary insulinemic potential that is etiologically crucial for developing a wide spectrum of chronic diseases.

While their relationships with specific outcomes vary somewhat, the AHEI-2010, AMED, DASH, and hPDI all showed an inverse association with major chronic disease. This aligns with numerous studies supporting the protective associations of AHEI-2010, AMED, and DASH with cardiovascular disease, type 2 diabetes, and cancer^16-18. Because we accounted for smoking intensity and quit time, the inverse relationships of AHEI-2010, AMED, and DASH with smoking-related cancer may not be entirely explained by residual confounding by smoking. Although dietary evidence for smoking-related cancer is sparse, a meta-analysis found that the inverse association between the Mediterranean diet and lung cancer was greater in former smokers¹⁹, suggesting potential effect modification by smoking status. Previous research suggested that individuals who have increased their adherence to healthy plant-based diets had decreased risks of type 2 diabetes, overall cardiovascular disease, and overall mortality^20-22. We found that, in agreement with findings from a meta-analysis on plant-based diets²⁰, the inverse relationship between hPDI and cardiovascular disease was more pronounced for coronary heart disease than for stroke. One possibility is that the hPDI may have distinct relationships with ischemic and hemorrhagic stroke.

Despite moderate correlations with other dietary patterns, the WCRF/AICR dietary score includes the fewest items. The WCRF/AICR dietary score was not specifically customized to prevent diseases other than cancer may account for the null association with major chronic disease. However, we found an unexpected positive association for type 2 diabetes, and generally null associations for cancer-related outcomes. Because the 2018 WCRF/AICR recommendations included behavioral components beyond nutritional guidance²³, this score based solely on dietary recommendations may not represent the optimal diet for cancer prevention. Prior studies showed that the inverse associations between the WCRF/AICR lifestyle score and cancer incidence and all-cause mortality, as well as the improvement of biomarker profile, were predominantly driven by physical activity and body weight components, as opposed to the dietary components^24-26.

Due in part to the heterogeneity of cancer, the associations between dietary patterns and cancer-related outcomes were not as strong as those for other outcomes. Total cancer is a diverse constellation of unique diseases with a variety of causes, such as cigarette smoking, obesity, physical inactivity, and infections²⁷. Even within one type of cancer, there are multiple subtypes. Some cancers, such as colorectal cancer, are diet-related while others are not²⁷. A ten percent reduction from dietary patterns alone (ignoring the potential effect of diet on weight control) would be considered relatively large based on current understanding. Although more directly related to cardiometabolic diseases, the rEDIH, rEDIP, and DRRD displayed that a portion of cancers can be prevented by diet. Additionally, following these three diets may be equally or even more beneficial for cancer prevention compared to other healthy diets, including the one that is specially designed for cancer.

The individual food components and scoring methods of dietary patterns may partially explain the differential associations between dietary patterns and chronic disease. The WCRF/AICR score considers alcohol drinking as harmful due to strong evidence supporting its carcinogenic effect²⁷, which might obscure the potential benefits of moderate drinking for reducing coronary heart disease, ischemic stroke, type 2 diabetes, and all-cause mortality^28-30. The AHEI-2010 and AMED discourage low or heavy drinking while supporting moderate drinking. On the other hand, alcoholic beverages are given positive weights in the rEDIP and rEDIH because drinking alcohol is associated with lower levels of insulinemia and inflammation. Given the complex effects of alcohol on health, recommendations on alcohol intake should be personalized based on each person’s risk profile. Nonetheless, the associations of these four scores with risk of subsequent chronic diseases only changed modestly when the alcohol component was eliminated, suggesting that health improvement may still be achieved without alcohol intake. Coffee drinking, which is considered in the DRRD, rEDIH and rEDIP, has been associated with a lower risk of cardiometabolic disease³¹. The lower level of coffee intake among participants who scored high for WCRF/AICR might also contribute to the slightly positive association between WCRF/AICR and type 2 diabetes. Dietary guidelines emphasizing foods contributing to caloric intake would not account for dietary factors such as coffee consumption.

Our results also highlight that dietary patterns could reflect the overall effects of diet beyond the sum of individual foods. From a reductionist perspective, specific dietary components, such as coffee or moderate alcohol intake, may contribute to the benefits of maintaining a healthy diet. The relatively weak correlations of rEDIH and rEDIP with some frequently recommended foods, such as cruciferous vegetables and legumes, suggest that the metabolic effects of diet may be cumulative and not dominated by a few components. The unexpected protective association of dessert intake with chronic diseases may be explained by its inverse correlations with foods that appeared unfavorable for metabolic diseases (such as poultry, tomatoes, and eggs). The flavonoids in the chocolate may also contribute to the inverse associations between dessert and chronic diseases^32,33. Given the varying degrees of health benefits associated with dietary patterns, future research on the subtle difference between patterns may be important for maximizing the effectiveness of dietary interventions.

After adjusting for BMI, the relationships with chronic disease for rEDIH and rEDIP were attenuated most because these two scores had the strongest association with BMI, possibly because of the residual confounding (BMI) or mediation effect (long-term weight gain)³⁴. Our results remained consistent across several sensitivity tests. The stronger associations found in the subgroup analysis imply that dietary modification may be more advantageous for women, younger people, and people with higher BMI. The associations with rEDIH, rEDIP and DRRD were particularly strong for those with high BMI. This result would be expected as the influence of diet related to insulin resistance and inflammation is likely to be greater in individuals at risk for insulin resistance resulting from a high BMI. Yet, even in those with normal BMI, these scores still had the strongest inverse association with chronic disease. The strong associations between dietary patterns and major chronic disease in both ever- and never-smokers underscore the potential of dietary modification.

The goal of dietary guidelines for adults is to provide advice on food and beverage choices to meet nutritional needs and to help prevent diet-related chronic diseases. The pathophysiologic processes underlying major chronic diseases include, among others, lipids, blood pressure, glycemia, insulinemia, and inflammation. From an overall health perspective, the most important dietary components will likely be those that substantially affect the pathophysiologic processes that affect sizable numbers of multiple chronic diseases. Our findings for rEDIH and rEDIP may reflect that hyperinsulinemia and inflammation are strongly influenced by diet, have overlapping factors, and are quantitatively important for multiple diseases. The considerably stronger associations for rEDIH and rEDIP than other recommended dietary patterns may indicate that when recommendations are formulated, greater emphasis may need to be given to specific dietary components that influence inflammation and insulinemia, especially in the context of rising trends in obesity.

The strengths of this study include large sample size, long follow-up period, repeated assessments of dietary intake using validated instruments, detailed collection of lifestyle and medical data allowing for adjustment for potential confounders, and comprehensive comparisons of multiple dietary patterns with major chronic diseases within the same analytical framework.

Study limitations should be acknowledged. The composite outcome of major chronic disease did not include cognitive outcomes and other functional impairments. We included incident cases for diseases that are leading causes of death in the US and are well ascertained in the cohorts. To capture other chronic conditions severe enough to increase mortality, non-traumatic death was included in the alternative definition of major chronic disease. However, the possibility for reverse causation could increase after including mortality endpoints. We attempted to address this by conducting latency analyses and observed consistent strong associations for both individual and composite outcomes. With longer latency between diet and outcome assessment, the analysis may be conservative in not accounting for recent diet, but largely exclude the effect of reverse causation. Our study also implies that diet has more immediate effects on cardiometabolic diseases, but a longer latency period is required for cancer. The Four dietary patterns—AMED, DASH, hPDI, and DRRD—were determined based on the distribution of the research population using either the median or quintiles as the cut-off, which may not represent populations with more diverse racial and ethnic compositions. Yet, we found comparable associations across different ethnic groups, supporting the generalizability of the diet and disease associations.

In conclusion, we found that maintaining a healthy diet was generally associated with a lower risk of developing major chronic diseases. Among the eight dietary patterns examined in this study, those reflecting low insulinemic, low inflammatory, and diabetes risk reduction diet may confer the largest risk reduction for various chronic diseases. The rEDIH and rEDIP were developed empirically based on associations between specific foods and biomarkers of insulinemia and inflammation, which may contribute to their strong predictive capacity. Future research is needed to explore more specific biological mechanisms underlying the relationship between diet and overall health.

Methods

Extended Data

Extended Data Fig.1

Open in a separate window

Flowchart of participants included in the main analysis.

BMI, body mass index; FFQ, food frequency questionnaires; Nurses’ Health Study (NHS); Health Professionals Follow-up Study (HPFS).

Extended Data Fig.2

Open in a separate window

Baseline Spearman correlations between energy-adjusted cumulative average dietary patterns in (A) all cohorts, (B) the Health Professionals Follow-up Study, (C) the Nurses’ Health Study, and (D) Nurses’ Health Study II.

P values based on the two-sided tests were <0.0001 for all correlations (not adjusted for multiple comparisons). AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Extended Data Fig.3

Open in a separate window

Multivariable-adjusted spline analysis of dietary pattern scores with risk of major chronic disease.

P values for nonlinearity based on the two-sided tests were statistically significant for AMED (P = 0.03), WCRF/AICR (P = 0.007), and rEDIP (P = 0.008) (not adjusted for multiple comparisons). The hazard ratios (black line) and the 95% confidence intervals (grey bands) are shown. The models were adjusted for the same list of covariates as in Table 2. AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; HR, Hazard ratio; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Extended Data Fig.4

Open in a separate window

Multivariable-adjusted associations between cumulative average dietary patterns (comparing the 90^th to 10^th percentile) and major chronic disease and major components in the pooled data of three cohorts (n = 205,852 participants) with different lags.

Analyses details and corresponding estimates are provided in Extended Data Table 3. The hazard ratios are indicated by the circles and the 95% confidence intervals are reflected by the error bars. AHEI-2010, Alternative Healthy Eating Index-2010; AMED, Alternate Mediterranean Diet score; DASH, Dietary Approaches to Stop Hypertension score; DRRD, Diabetes Risk Reduction Diet; hPDI, Healthful plant-based diet index; rEDIH, reversed Empirical dietary index for hyperinsulinemia; rEDIP, reversed Empirical dietary inflammation pattern; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) dietary score.

Supplementary Material

Acknowledgements

Footnotes

References

Methods only references