https://orcid.org/0000-0002-7038-4770
Abstract
Trials demonstrate that following the DASH diet lowers blood pressure, which may prevent the development of heart failure (HF). We investigated the association between long-term adherence to the DASH diet and food substitutions within the DASH diet on the risk of HF.
Men and women aged 45–83 years without previous HF, ischaemic heart disease or cancer at baseline in 1998 from the Cohort of Swedish Men (n = 41 118) and the Swedish Mammography Cohort (n = 35 004) were studied. The DASH diet emphasizes intake of fruit, vegetables, whole grains, nuts and legumes, and low-fat dairy and deemphasizes red and processed meat, sugar-sweetened beverages, and sodium. DASH diet scores were calculated based on diet assessed by food frequency questionnaires in late 1997 and 2009. Incidence of HF was ascertained using the Swedish Patient Register. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs). During the median 22 years of follow-up (1998–2019), 12 164 participants developed HF. Those with the greatest adherence to the DASH diet had a lower risk of HF compared to those with the lowest adherence (HR 0.85, 95% CI: 0.80, 0.91 for baseline diet and HR 0.83, 95% CI: 0.78, 0.89 for long-term diet, comparing quintiles). Replacing 1 serving/day of red and processed meat with emphasized DASH diet foods was associated with an 8–12% lower risk of HF.
Long-term adherence to the DASH diet and relevant food substitutions within the DASH diet were associated with a lower risk of HF.
Introduction
Heart failure (HF) is an immense public health problem, with an estimated global incidence of 10–90 cases/10 000 person-years, and is the number one cause of hospitalization among the elderly in high-income countries.1 Hence, prevention of HF has a high priority.
Adherence to healthy dietary patterns, as defined by national dietary guidelines, including the Nordic Nutrition Recommendations used in Sweden, and the Mediterranean diet, has been associated with the lower risk of all-cause mortality, cardiovascular mortality, and incidence of HF.2–5 In randomized controlled trials (RCTs), consuming a healthy diet has also been shown to improve key risk factors for the development of HF including high blood pressure, high body mass index (BMI), high blood glucose, and LDL-cholesterol.6 In particular, strong evidence from RCTs has shown that the Dietary Approach to Stop Hypertension (DASH) diet, recommending higher intakes of fruits, vegetables, whole grains, nuts and legumes, and low-fat dairy and lower intakes of red and processed meat, sugar-sweetened beverages, and sodium, lowered blood pressure.6 While sodium reduction in patients with HF is controversial, there is good evidence for its role in the prevention of HF, particularly in high-risk populations.7
Concerning the role of the DASH diet in the prevention of HF, evidence from US cohorts has been conflicting. A higher HF risk was observed among those with the greatest adherence to the DASH diet in the Cardiovascular Health Study8 whereas a lower risk was reported in the Multi-Ethnic Study of Atherosclerosis9 and the Reasons for Geographic And Racial Differences in Stroke cohort.10 We have previously observed that adherence to the DASH diet was associated with lower HF risk in women11 and men,12 but these studies examined baseline diet only and had a shorter follow-up time. Updating these analyses would improve our understanding of the long-term impact of adhering to the DASH diet on HF risk. In addition, several cohort studies have investigated individual components included in the DASH diet (e.g. vegetables) in relation to HF; however, none has specified food substitutions, which provide a clearer interpretation of the association of specific food items with health outcomes and direct simple actionable dietary shifts.13 One important food group is red and processed meat as it contains detrimental (e.g. sodium in processed meat) but also beneficial nutrients (e.g. U-shaped relation for iron status14) for the development of HF.
Our main hypothesis was that long-term adherence to the DASH diet is associated with a lower risk of HF in middle-aged and elderly adults. We also examined the association between relevant food substitutions among DASH diet foods and risk of HF.
Methods
Ethical approval
The study complies with the Declaration of Helsinki and was approved by the Regional Ethical Review Board at Karolinska Institute, Stockholm, Sweden. Completion and return of the questionnaires was considered consent. The two population-based cohorts from central Sweden, the Cohort of Swedish Men (COSM) and the Swedish Mammography Cohort (SMC), were registered at ClinicalTrials.gov with identifiers NCT01127711 and NCT01127698, respectively, and both cohorts belong to the national research infrastructure SIMPLER (www.simpler4health.se).
Study population
We used self-reported questionnaire-based information on demographics, dietary intake, and lifestyle obtained in late 1997 and in 2009 in both COSM and SMC. The COSM includes 48 850 men born between 1918 and 1952 (49% of invited). The SMC includes 39 984 women born 1914–48 (70% of invited). In 2009, an updated questionnaire was sent to participants in the COSM (n = 29 068 responded, 90% of invited) and SMC (n = 30 134 responded, 84% of invited) who were still alive and residing in the study areas.15 We started follow-up of these cohorts on 1 January 1998 and, at baseline, responders were comparable to the general Swedish population regarding the distribution of age, education, and BMI.15
We excluded participants with missing or incorrect ID or a diagnosis of cancer (n = 5770) and further excluded participants with HF (n = 1268) or ischaemic heart disease (n = 4710) before the baseline. Those with cancer were excluded due to change in health status and likely in health behaviour. Participants with previous ischaemic heart disease were also excluded because of the conditions’ strong relationship with the development of HF and our focus on prevention in the general population. To improve the quality of reported intakes, we additionally excluded implausible total energy intakes (±3 SD of mean total energy intake on the log scale) (n = 964) (Supplementary material online, Figures S1 and S2). A total of 76 122 participants were included in the analysis using the baseline data.
Diet assessments
Dietary intake in late 1997 was assessed using a 96-item food frequency questionnaire (FFQ) that asked participants about their average intake of foods and meals during the last 12 months.15,16 Response categories ranged from ‘never’ to ‘≥3 times/day’ for 88 items and 8 items were open-ended questions. Age-specific portion sizes together with information from the Swedish Food Administration nutrient composition database17 were used to estimate average intake of foods and nutrients for each participant. The follow-up FFQ in 2009 was extended to 132 food items to include new foods on the market and usual foods consumed at the time.15
DASH diet
We used the score proposed by Fung et al.18 to calculate adherence to the DASH diet. The diet score includes eight dietary components: fruits, vegetables, nuts and legumes, low-fat dairy, whole grains, sodium, sweetened beverages, and red and processed meat. Supplementary material online, Table S1 shows which foods were included in each category. For each component, participants are categorized into quintiles according to the amount of intake of the specific item and scored 1–5 points based on the quintile they are in. The total score is the sum of points (maximum 40) across the dietary components. We calculated the score in men and women combined. Correlations between the FFQ-based estimates and four 1-week diet records distributed over 1 year were 0.5–0.7 for fruits,19 0.4–0.6 for vegetables,19 0.5–0.7 for whole grains,20 0.4–0.6 for dairy,21 0.3–0.7 for red and processed meats,22 and 0.6 for sweetened beverages.23
Heart failure ascertainment and follow-up of health status
HF incidence was ascertained using the Swedish Patient Register (inpatient and outpatient specialist care) and defined according to the International Classification of Disease-10 (ICD-10) codes I50 and I11.0, listed as either primary diagnosis or at any diagnosis position. Diagnosis of HF in the Register was previously validated against review of medical records using the European Cardiology Society criteria for HF and had been found that 95% of patients with HF as primary diagnosis code and 76% with HF as secondary diagnosis code were correctly classified.24 Dates of death were obtained from the Swedish Death Register.
Assessment of covariates
Information on education, weight, height, minutes per day walking/cycling, use of aspirin, corticosteroids, dietary supplements, use of hormone replacement therapy (women), alcohol, smoking status, history of hypercholesterolaemia, hypertension, and diabetes, and family history of myocardial infarction was self-reported in the baseline questionnaire. Atrial fibrillation ICD-10 code I48 was ascertained by linkage to the Swedish Patient Register.
Statistical analyses
To estimate the association, hazard ratio (HR) and 95% confidence intervals (CIs), between adherence to the DASH diet and risk of HF, we used Cox proportional hazards regression with age as the underlying time scale. Participants were followed up from January 1, 1998, to the date of diagnosis of HF, death, or end of follow-up December 31, 2019, whichever came first. Adherence to the DASH diet was modelled as quintiles of the DASH diet score in men and women combined because results were similar in analyses when DASH diet scores were sex specific (P for interaction = 0.56). Potential confounders were selected based on a review of the literature on risk factors for the development of HF and the use of directed acyclic graphs to show assumed relationship between variables (Supplementary material online, Figure S3). The basic model was adjusted for age and sex (men, women). The multivariable model was adjusted for potential confounders including education (primary, high school, university), smoking (never; ex-smoker <20, 20–39, or ≥40 pack-years; current <20, 20–39, or ≥40 pack-years), alcohol (g/day, continuous), walking/cycling (<20, 20–40, 40–60, or >60 min/day), corticosteroid use (yes, no), aspirin use (yes, no), dietary supplement use (regularly, sometimes, no), use of hormone replacement therapy in women (yes, no), family history of myocardial infarction (yes, no), and total energy intake (kcal/day, continuous). The model was further adjusted for BMI (≤18.5, 18.5–24.9, 25–29.9, >30 kg/m2), hypertension (yes, no), hypercholesterolaemia (yes, no), diabetes (yes, no), and atrial fibrillation (yes, no). Due to few missing values in the included variables, they were indicated with their own ‘missing’ category in the main analysis. To investigate the association with long-term adherence, we updated the dietary intake to the average intake of the two FFQs among those who had completed the follow-up FFQ in 2009. Baseline intake was used for participants who were censored before 2009 or did not complete the FFQ in 2009. The assumption of proportional hazards was assessed using log–log plots and no evidence for violation of the assumption was detected. The association between adherence to the DASH diet and risk of HF was also modelled using a restricted cubic spline function with four knots of the DASH diet score.
We estimated the association related to food substitutions between foods within the DASH diet. One serving/day of red and processed meat (100 g/day) was replaced with 1 serving/day of fruit (100 g/day), vegetables (100 g/day), nuts and legumes (20 g/day), whole grains (100 g/day), or low-fat dairy (200 g/day). The food substitution was obtained by including all DASH diet foods (1 serving/day, continuous) separately in the statistical model, except for the food to be replaced (here red and processed meat), and a variable that is the sum of all the foods included in the DASH diet score.13 As the total intake of all DASH diet score foods is fixed in the model, a higher intake of each of the included foods must be at the expense of the food not included in the model (here red and processed meat). In other words, participants with a higher intake of fruit and a lower intake of red and processed meat are compared with participants with a lower intake of fruit and a higher intake of red and processed meat. Because the energy density of the substituted foods is not equal, there is a residual energy substitution from foods not included in the model. As a sensitivity analysis, we estimated the food substitutions without adjustment for total energy intake. We also investigated the association between each of the food groups (i.e. sodium not included) in the DASH diet separately, adjusting for the other food groups on risk of HF. Sodium, as the only nutrient, was investigated separately. Each item was modelled using a restricted cubic spline function with four knots.
Potential effect modification by self-reported history of hypertension and history of diabetes was also assessed in stratified analyses. To estimate the public health benefit on the HF risk of adhering to the DASH diet, assuming a causal relationship, we used the additive Cox model25 by estimating multivariable adjusted rate differences and corresponding 95% CIs at 20 years of follow-up.
In sensitivity analyses, we used multiple imputation with chained equations to impute five datasets in which we re-ran the main analysis and subsequently pooled the results because of missing data (<0–11% missing) on some of the included variables. We also further excluded those with a high risk of HF at baseline, i.e. those with hypertension, diabetes, and/or atrial fibrillation, to avoid bias due to underlying illnesses that could have influenced dietary intake. We re-ran the main analysis including only those with HF as the main diagnosis code to investigate potential outcome misclassification. As anti-cancer treatment may induce HF,26 we censored all participants who developed any type of cancer before development of HF. Incidence of cancer was obtained from the Swedish Cancer Register. Lastly, because of the built-in selection bias of the HR,27 we truncated the end of follow-up date at 5, 10, 15, and 20 years of follow-up to investigate the stability of HRs over time.
All analyses were conducted using R (R Core Team, Vienna, Austria) version 3.6.0. The data underlying this article cannot be shared publicly because the data are classified as sensitive data under the European General Data Protection Regulation. The data and underlying code can be shared on reasonable request to the corresponding author through the Swedish National Research Infrastructure SIMPLER (www.simpler4health.se).
Results
Baseline characteristics
Participants with the greatest adherence to the DASH diet (score 30–40 points) generally had healthier characteristics at the baseline than those with the lowest adherence (score 8–21 points) (Table 1). Those with the greatest adherence appear to be older, be women, and have a normal BMI, a longer education, never smoked, a lower intake of alcohol, and a less physical inactivity and to take dietary supplements. The median follow-up time was 22 years and during 6 001 289 person-years of observation 12 164 participants developed HF [as primary diagnosis n = 4609 (2625 men, 1984 women), or at any other diagnosis position n = 7555 (3963 men, 3592 women)]. Of all men, 7% were in the highest and 35% in the lowest category of adherence to the DASH diet, corresponding numbers for women were 26% and 10%.
Baseline characteristics of participants from the Cohort of Swedish Men and Swedish Mammography Cohort 1997 examination according to quintiles of Dietary Approach to Stop Hypertension diet score adherence
| Characteristics | 8–21 points (n = 18 027), median (10th, 90th), n (%) | 22–24 points (n = 17 620), median (10th, 90th), n (%) | 25–26 points (n = 12 541), median (10th, 90th), n (%) | 27–29 points (n = 15 721), median (10th, 90th), n (%) | 30–40 points (n = 12 213), median (10th, 90th), n (%) | Overall (n = 76 122), median (10th, 90th), n (%) |
|---|---|---|---|---|---|---|
| DASH diet score (points) | 19 (16, 21) | 23 (22, 24) | 26 (25, 26) | 28 (27, 29) | 31 (30, 34) | 25 (19, 31) |
| Age (years) | 57 (48, 74) | 59 (49, 75) | 59 (49, 75) | 60 (50, 74) | 60 (50, 74) | 59 (49, 74) |
| Sex (women) | 3629 (20%) | 6667 (38%) | 6069 (48%) | 9352 (59%) | 9287 (76%) | 35004 (46%) |
| BMI (kg/m2) | ||||||
| Normal weight (18.6–24.9) | 7610 (42%) | 8008 (45%) | 5987 (48%) | 7878 (50%) | 6629 (54%) | 36112 (47%) |
| Missing | 908 (5%) | 671 (4%) | 421 (3%) | 406 (3%) | 225 (2%) | 2631 (3%) |
| Education | ||||||
| University | 2451 (14%) | 2718 (15%) | 2318 (18%) | 3191 (20%) | 2972 (24%) | 13650 (18%) |
| Missing | 72 (0%) | 72 (0%) | 46 (0%) | 62 (0%) | 48 (0%) | 300 (0%) |
| Smoking (pack-years) | ||||||
| Never | 6259 (35%) | 7208 (41%) | 5672 (45%) | 7766 (49%) | 6605 (54%) | 33510 (44%) |
| Missing | 1133 (6%) | 985 (6%) | 651 (5%) | 725 (5%) | 506 (4%) | 4000 (5%) |
| Alcohol (g/day) | 6.2 (0.0, 21) | 5.4 (0.0, 19) | 5.1 (0.0, 18) | 4.5 (0.0, 17) | 3.5 (0.0, 14) | 5.0 (0.0, 18) |
| Walking and cycling (min/day) | ||||||
| Never/seldom | 2870 (16%) | 2120 (12%) | 1212 (10%) | 1325 (8%) | 789 (6%) | 8316 (11%) |
| Missing | 2014 (11%) | 1677 (10%) | 1020 (8%) | 1133 (7%) | 768 (6%) | 6612 (9%) |
| Aspirin use (yes) | 10281 (57%) | 9487 (54%) | 6516 (52%) | 7977 (51%) | 5993 (49%) | 40254 (53%) |
| Missing | 1873 (10%) | 1970 (11%) | 1475 (12%) | 1746 (11%) | 1424 (12%) | 8488 (11%) |
| Corticosteroid use (yes) | 1121 (6%) | 1115 (6%) | 857 (7%) | 1171 (7%) | 889 (7%) | 5153 (7%) |
| Missing | 1580 (9%) | 1820 (10%) | 1330 (11%) | 1808 (12%) | 1553 (13%) | 8091 (11%) |
| Dietary supplement use (regularly) | 2062 (11%) | 2675 (15%) | 2292 (18%) | 3367 (21%) | 3318 (27%) | 13714 (18%) |
| Missing | 1320 (7%) | 1323 (8%) | 870 (7%) | 1025 (7%) | 748 (6%) | 5286 (7%) |
| HRT usea (yes) | 1422 (39%) | 2831 (42%) | 2724 (45%) | 4450 (48%) | 4613 (50%) | 16040 (46%) |
| Missing | 167 (5%) | 260 (4%) | 185 (3%) | 294 (3%) | 266 (3%) | 1172 (3%) |
| History of diabetes (yes) | 735 (4%) | 833 (5%) | 619 (5%) | 802 (5%) | 598 (5%) | 3587 (5%) |
| History of hypertension (yes) | 3639 (20%) | 3636 (21%) | 2625 (21%) | 3252 (21%) | 2473 (20%) | 15625 (21%) |
| History of hypercholesterolaemia (yes) | 1786 (10%) | 1732 (10%) | 1165 (9%) | 1545 (10%) | 1167 (10%) | 7395 (10%) |
| History of atrial fibrillation (yes) | 271 (2%) | 254 (1%) | 192 (2%) | 212 (1%) | 162 (1%) | 1091 (1%) |
| Family history of MI (yes) | 497 (3%) | 521 (3%) | 356 (3%) | 520 (3%) | 435 (4%) | 2329 (3%) |
| Total energy intake (kJ/day) | 2200 (1300, 3400) | 2100 (1200, 3400) | 2100 (1300, 3400) | 2000 (1300, 3400) | 2000 (1400, 3200) | 2100 (1300, 3400) |
| Fruits (g/day) | 67 (19, 150) | 100 (33, 220) | 140 (55, 290) | 180 (85, 350) | 260 (130, 460) | 130 (37, 320) |
| Vegetables (g/day) | 82 (28, 160) | 120 (47, 220) | 150 (68, 270) | 180 (92, 320) | 230 (130, 400) | 140 (51, 290) |
| Nuts and legumes (g/day) | 24 (0.0, 65) | 26 (3.2, 77) | 29 (4.8, 83) | 35 (11, 90) | 49 (19, 120) | 28 (3.5, 83) |
| Low-fat dairy products (g/day) | 19 (0.0, 440) | 150 (0.0, 600) | 210 (0.0, 660) | 260 (0.0, 690) | 320 (61, 750) | 190 (0.0, 650) |
| Whole grains (g/day) | 90 (17, 250) | 120 (38, 310) | 140 (45, 330) | 140 (53, 330) | 160 (67, 330) | 130 (39, 320) |
| Sodium (mg/day) | 3600 (2500, 4300) | 3400 (2300, 4200) | 3100 (2200, 4200) | 2800 (2200, 4100) | 2600 (2200, 3800) | 3200 (2200, 4200) |
| Sweetened beverages (g/day) | 140 (0.0, 660) | 31 (0.0, 510) | 0.0 (0.0, 330) | 0.0 (0.0, 260) | 0.0 (0.0, 140) | 0.0 (0.0, 430) |
| Red and processed meat (g/day) | 100 (47, 160) | 84 (35, 150) | 77 (33, 140) | 69 (30, 130) | 56 (20, 99) | 78 (31, 140) |
| Characteristics | 8–21 points (n = 18 027), median (10th, 90th), n (%) | 22–24 points (n = 17 620), median (10th, 90th), n (%) | 25–26 points (n = 12 541), median (10th, 90th), n (%) | 27–29 points (n = 15 721), median (10th, 90th), n (%) | 30–40 points (n = 12 213), median (10th, 90th), n (%) | Overall (n = 76 122), median (10th, 90th), n (%) |
|---|---|---|---|---|---|---|
| DASH diet score (points) | 19 (16, 21) | 23 (22, 24) | 26 (25, 26) | 28 (27, 29) | 31 (30, 34) | 25 (19, 31) |
| Age (years) | 57 (48, 74) | 59 (49, 75) | 59 (49, 75) | 60 (50, 74) | 60 (50, 74) | 59 (49, 74) |
| Sex (women) | 3629 (20%) | 6667 (38%) | 6069 (48%) | 9352 (59%) | 9287 (76%) | 35004 (46%) |
| BMI (kg/m2) | ||||||
| Normal weight (18.6–24.9) | 7610 (42%) | 8008 (45%) | 5987 (48%) | 7878 (50%) | 6629 (54%) | 36112 (47%) |
| Missing | 908 (5%) | 671 (4%) | 421 (3%) | 406 (3%) | 225 (2%) | 2631 (3%) |
| Education | ||||||
| University | 2451 (14%) | 2718 (15%) | 2318 (18%) | 3191 (20%) | 2972 (24%) | 13650 (18%) |
| Missing | 72 (0%) | 72 (0%) | 46 (0%) | 62 (0%) | 48 (0%) | 300 (0%) |
| Smoking (pack-years) | ||||||
| Never | 6259 (35%) | 7208 (41%) | 5672 (45%) | 7766 (49%) | 6605 (54%) | 33510 (44%) |
| Missing | 1133 (6%) | 985 (6%) | 651 (5%) | 725 (5%) | 506 (4%) | 4000 (5%) |
| Alcohol (g/day) | 6.2 (0.0, 21) | 5.4 (0.0, 19) | 5.1 (0.0, 18) | 4.5 (0.0, 17) | 3.5 (0.0, 14) | 5.0 (0.0, 18) |
| Walking and cycling (min/day) | ||||||
| Never/seldom | 2870 (16%) | 2120 (12%) | 1212 (10%) | 1325 (8%) | 789 (6%) | 8316 (11%) |
| Missing | 2014 (11%) | 1677 (10%) | 1020 (8%) | 1133 (7%) | 768 (6%) | 6612 (9%) |
| Aspirin use (yes) | 10281 (57%) | 9487 (54%) | 6516 (52%) | 7977 (51%) | 5993 (49%) | 40254 (53%) |
| Missing | 1873 (10%) | 1970 (11%) | 1475 (12%) | 1746 (11%) | 1424 (12%) | 8488 (11%) |
| Corticosteroid use (yes) | 1121 (6%) | 1115 (6%) | 857 (7%) | 1171 (7%) | 889 (7%) | 5153 (7%) |
| Missing | 1580 (9%) | 1820 (10%) | 1330 (11%) | 1808 (12%) | 1553 (13%) | 8091 (11%) |
| Dietary supplement use (regularly) | 2062 (11%) | 2675 (15%) | 2292 (18%) | 3367 (21%) | 3318 (27%) | 13714 (18%) |
| Missing | 1320 (7%) | 1323 (8%) | 870 (7%) | 1025 (7%) | 748 (6%) | 5286 (7%) |
| HRT usea (yes) | 1422 (39%) | 2831 (42%) | 2724 (45%) | 4450 (48%) | 4613 (50%) | 16040 (46%) |
| Missing | 167 (5%) | 260 (4%) | 185 (3%) | 294 (3%) | 266 (3%) | 1172 (3%) |
| History of diabetes (yes) | 735 (4%) | 833 (5%) | 619 (5%) | 802 (5%) | 598 (5%) | 3587 (5%) |
| History of hypertension (yes) | 3639 (20%) | 3636 (21%) | 2625 (21%) | 3252 (21%) | 2473 (20%) | 15625 (21%) |
| History of hypercholesterolaemia (yes) | 1786 (10%) | 1732 (10%) | 1165 (9%) | 1545 (10%) | 1167 (10%) | 7395 (10%) |
| History of atrial fibrillation (yes) | 271 (2%) | 254 (1%) | 192 (2%) | 212 (1%) | 162 (1%) | 1091 (1%) |
| Family history of MI (yes) | 497 (3%) | 521 (3%) | 356 (3%) | 520 (3%) | 435 (4%) | 2329 (3%) |
| Total energy intake (kJ/day) | 2200 (1300, 3400) | 2100 (1200, 3400) | 2100 (1300, 3400) | 2000 (1300, 3400) | 2000 (1400, 3200) | 2100 (1300, 3400) |
| Fruits (g/day) | 67 (19, 150) | 100 (33, 220) | 140 (55, 290) | 180 (85, 350) | 260 (130, 460) | 130 (37, 320) |
| Vegetables (g/day) | 82 (28, 160) | 120 (47, 220) | 150 (68, 270) | 180 (92, 320) | 230 (130, 400) | 140 (51, 290) |
| Nuts and legumes (g/day) | 24 (0.0, 65) | 26 (3.2, 77) | 29 (4.8, 83) | 35 (11, 90) | 49 (19, 120) | 28 (3.5, 83) |
| Low-fat dairy products (g/day) | 19 (0.0, 440) | 150 (0.0, 600) | 210 (0.0, 660) | 260 (0.0, 690) | 320 (61, 750) | 190 (0.0, 650) |
| Whole grains (g/day) | 90 (17, 250) | 120 (38, 310) | 140 (45, 330) | 140 (53, 330) | 160 (67, 330) | 130 (39, 320) |
| Sodium (mg/day) | 3600 (2500, 4300) | 3400 (2300, 4200) | 3100 (2200, 4200) | 2800 (2200, 4100) | 2600 (2200, 3800) | 3200 (2200, 4200) |
| Sweetened beverages (g/day) | 140 (0.0, 660) | 31 (0.0, 510) | 0.0 (0.0, 330) | 0.0 (0.0, 260) | 0.0 (0.0, 140) | 0.0 (0.0, 430) |
| Red and processed meat (g/day) | 100 (47, 160) | 84 (35, 150) | 77 (33, 140) | 69 (30, 130) | 56 (20, 99) | 78 (31, 140) |
BMI, body mass index; DASH, Dietary Approach to Stop Hypertension; HRT, hormone replacement therapy; MI, myocardial infarction.
In women only (n = 35 004).
Baseline characteristics of participants from the Cohort of Swedish Men and Swedish Mammography Cohort 1997 examination according to quintiles of Dietary Approach to Stop Hypertension diet score adherence
| Characteristics | 8–21 points (n = 18 027), median (10th, 90th), n (%) | 22–24 points (n = 17 620), median (10th, 90th), n (%) | 25–26 points (n = 12 541), median (10th, 90th), n (%) | 27–29 points (n = 15 721), median (10th, 90th), n (%) | 30–40 points (n = 12 213), median (10th, 90th), n (%) | Overall (n = 76 122), median (10th, 90th), n (%) |
|---|---|---|---|---|---|---|
| DASH diet score (points) | 19 (16, 21) | 23 (22, 24) | 26 (25, 26) | 28 (27, 29) | 31 (30, 34) | 25 (19, 31) |
| Age (years) | 57 (48, 74) | 59 (49, 75) | 59 (49, 75) | 60 (50, 74) | 60 (50, 74) | 59 (49, 74) |
| Sex (women) | 3629 (20%) | 6667 (38%) | 6069 (48%) | 9352 (59%) | 9287 (76%) | 35004 (46%) |
| BMI (kg/m2) | ||||||
| Normal weight (18.6–24.9) | 7610 (42%) | 8008 (45%) | 5987 (48%) | 7878 (50%) | 6629 (54%) | 36112 (47%) |
| Missing | 908 (5%) | 671 (4%) | 421 (3%) | 406 (3%) | 225 (2%) | 2631 (3%) |
| Education | ||||||
| University | 2451 (14%) | 2718 (15%) | 2318 (18%) | 3191 (20%) | 2972 (24%) | 13650 (18%) |
| Missing | 72 (0%) | 72 (0%) | 46 (0%) | 62 (0%) | 48 (0%) | 300 (0%) |
| Smoking (pack-years) | ||||||
| Never | 6259 (35%) | 7208 (41%) | 5672 (45%) | 7766 (49%) | 6605 (54%) | 33510 (44%) |
| Missing | 1133 (6%) | 985 (6%) | 651 (5%) | 725 (5%) | 506 (4%) | 4000 (5%) |
| Alcohol (g/day) | 6.2 (0.0, 21) | 5.4 (0.0, 19) | 5.1 (0.0, 18) | 4.5 (0.0, 17) | 3.5 (0.0, 14) | 5.0 (0.0, 18) |
| Walking and cycling (min/day) | ||||||
| Never/seldom | 2870 (16%) | 2120 (12%) | 1212 (10%) | 1325 (8%) | 789 (6%) | 8316 (11%) |
| Missing | 2014 (11%) | 1677 (10%) | 1020 (8%) | 1133 (7%) | 768 (6%) | 6612 (9%) |
| Aspirin use (yes) | 10281 (57%) | 9487 (54%) | 6516 (52%) | 7977 (51%) | 5993 (49%) | 40254 (53%) |
| Missing | 1873 (10%) | 1970 (11%) | 1475 (12%) | 1746 (11%) | 1424 (12%) | 8488 (11%) |
| Corticosteroid use (yes) | 1121 (6%) | 1115 (6%) | 857 (7%) | 1171 (7%) | 889 (7%) | 5153 (7%) |
| Missing | 1580 (9%) | 1820 (10%) | 1330 (11%) | 1808 (12%) | 1553 (13%) | 8091 (11%) |
| Dietary supplement use (regularly) | 2062 (11%) | 2675 (15%) | 2292 (18%) | 3367 (21%) | 3318 (27%) | 13714 (18%) |
| Missing | 1320 (7%) | 1323 (8%) | 870 (7%) | 1025 (7%) | 748 (6%) | 5286 (7%) |
| HRT usea (yes) | 1422 (39%) | 2831 (42%) | 2724 (45%) | 4450 (48%) | 4613 (50%) | 16040 (46%) |
| Missing | 167 (5%) | 260 (4%) | 185 (3%) | 294 (3%) | 266 (3%) | 1172 (3%) |
| History of diabetes (yes) | 735 (4%) | 833 (5%) | 619 (5%) | 802 (5%) | 598 (5%) | 3587 (5%) |
| History of hypertension (yes) | 3639 (20%) | 3636 (21%) | 2625 (21%) | 3252 (21%) | 2473 (20%) | 15625 (21%) |
| History of hypercholesterolaemia (yes) | 1786 (10%) | 1732 (10%) | 1165 (9%) | 1545 (10%) | 1167 (10%) | 7395 (10%) |
| History of atrial fibrillation (yes) | 271 (2%) | 254 (1%) | 192 (2%) | 212 (1%) | 162 (1%) | 1091 (1%) |
| Family history of MI (yes) | 497 (3%) | 521 (3%) | 356 (3%) | 520 (3%) | 435 (4%) | 2329 (3%) |
| Total energy intake (kJ/day) | 2200 (1300, 3400) | 2100 (1200, 3400) | 2100 (1300, 3400) | 2000 (1300, 3400) | 2000 (1400, 3200) | 2100 (1300, 3400) |
| Fruits (g/day) | 67 (19, 150) | 100 (33, 220) | 140 (55, 290) | 180 (85, 350) | 260 (130, 460) | 130 (37, 320) |
| Vegetables (g/day) | 82 (28, 160) | 120 (47, 220) | 150 (68, 270) | 180 (92, 320) | 230 (130, 400) | 140 (51, 290) |
| Nuts and legumes (g/day) | 24 (0.0, 65) | 26 (3.2, 77) | 29 (4.8, 83) | 35 (11, 90) | 49 (19, 120) | 28 (3.5, 83) |
| Low-fat dairy products (g/day) | 19 (0.0, 440) | 150 (0.0, 600) | 210 (0.0, 660) | 260 (0.0, 690) | 320 (61, 750) | 190 (0.0, 650) |
| Whole grains (g/day) | 90 (17, 250) | 120 (38, 310) | 140 (45, 330) | 140 (53, 330) | 160 (67, 330) | 130 (39, 320) |
| Sodium (mg/day) | 3600 (2500, 4300) | 3400 (2300, 4200) | 3100 (2200, 4200) | 2800 (2200, 4100) | 2600 (2200, 3800) | 3200 (2200, 4200) |
| Sweetened beverages (g/day) | 140 (0.0, 660) | 31 (0.0, 510) | 0.0 (0.0, 330) | 0.0 (0.0, 260) | 0.0 (0.0, 140) | 0.0 (0.0, 430) |
| Red and processed meat (g/day) | 100 (47, 160) | 84 (35, 150) | 77 (33, 140) | 69 (30, 130) | 56 (20, 99) | 78 (31, 140) |
| Characteristics | 8–21 points (n = 18 027), median (10th, 90th), n (%) | 22–24 points (n = 17 620), median (10th, 90th), n (%) | 25–26 points (n = 12 541), median (10th, 90th), n (%) | 27–29 points (n = 15 721), median (10th, 90th), n (%) | 30–40 points (n = 12 213), median (10th, 90th), n (%) | Overall (n = 76 122), median (10th, 90th), n (%) |
|---|---|---|---|---|---|---|
| DASH diet score (points) | 19 (16, 21) | 23 (22, 24) | 26 (25, 26) | 28 (27, 29) | 31 (30, 34) | 25 (19, 31) |
| Age (years) | 57 (48, 74) | 59 (49, 75) | 59 (49, 75) | 60 (50, 74) | 60 (50, 74) | 59 (49, 74) |
| Sex (women) | 3629 (20%) | 6667 (38%) | 6069 (48%) | 9352 (59%) | 9287 (76%) | 35004 (46%) |
| BMI (kg/m2) | ||||||
| Normal weight (18.6–24.9) | 7610 (42%) | 8008 (45%) | 5987 (48%) | 7878 (50%) | 6629 (54%) | 36112 (47%) |
| Missing | 908 (5%) | 671 (4%) | 421 (3%) | 406 (3%) | 225 (2%) | 2631 (3%) |
| Education | ||||||
| University | 2451 (14%) | 2718 (15%) | 2318 (18%) | 3191 (20%) | 2972 (24%) | 13650 (18%) |
| Missing | 72 (0%) | 72 (0%) | 46 (0%) | 62 (0%) | 48 (0%) | 300 (0%) |
| Smoking (pack-years) | ||||||
| Never | 6259 (35%) | 7208 (41%) | 5672 (45%) | 7766 (49%) | 6605 (54%) | 33510 (44%) |
| Missing | 1133 (6%) | 985 (6%) | 651 (5%) | 725 (5%) | 506 (4%) | 4000 (5%) |
| Alcohol (g/day) | 6.2 (0.0, 21) | 5.4 (0.0, 19) | 5.1 (0.0, 18) | 4.5 (0.0, 17) | 3.5 (0.0, 14) | 5.0 (0.0, 18) |
| Walking and cycling (min/day) | ||||||
| Never/seldom | 2870 (16%) | 2120 (12%) | 1212 (10%) | 1325 (8%) | 789 (6%) | 8316 (11%) |
| Missing | 2014 (11%) | 1677 (10%) | 1020 (8%) | 1133 (7%) | 768 (6%) | 6612 (9%) |
| Aspirin use (yes) | 10281 (57%) | 9487 (54%) | 6516 (52%) | 7977 (51%) | 5993 (49%) | 40254 (53%) |
| Missing | 1873 (10%) | 1970 (11%) | 1475 (12%) | 1746 (11%) | 1424 (12%) | 8488 (11%) |
| Corticosteroid use (yes) | 1121 (6%) | 1115 (6%) | 857 (7%) | 1171 (7%) | 889 (7%) | 5153 (7%) |
| Missing | 1580 (9%) | 1820 (10%) | 1330 (11%) | 1808 (12%) | 1553 (13%) | 8091 (11%) |
| Dietary supplement use (regularly) | 2062 (11%) | 2675 (15%) | 2292 (18%) | 3367 (21%) | 3318 (27%) | 13714 (18%) |
| Missing | 1320 (7%) | 1323 (8%) | 870 (7%) | 1025 (7%) | 748 (6%) | 5286 (7%) |
| HRT usea (yes) | 1422 (39%) | 2831 (42%) | 2724 (45%) | 4450 (48%) | 4613 (50%) | 16040 (46%) |
| Missing | 167 (5%) | 260 (4%) | 185 (3%) | 294 (3%) | 266 (3%) | 1172 (3%) |
| History of diabetes (yes) | 735 (4%) | 833 (5%) | 619 (5%) | 802 (5%) | 598 (5%) | 3587 (5%) |
| History of hypertension (yes) | 3639 (20%) | 3636 (21%) | 2625 (21%) | 3252 (21%) | 2473 (20%) | 15625 (21%) |
| History of hypercholesterolaemia (yes) | 1786 (10%) | 1732 (10%) | 1165 (9%) | 1545 (10%) | 1167 (10%) | 7395 (10%) |
| History of atrial fibrillation (yes) | 271 (2%) | 254 (1%) | 192 (2%) | 212 (1%) | 162 (1%) | 1091 (1%) |
| Family history of MI (yes) | 497 (3%) | 521 (3%) | 356 (3%) | 520 (3%) | 435 (4%) | 2329 (3%) |
| Total energy intake (kJ/day) | 2200 (1300, 3400) | 2100 (1200, 3400) | 2100 (1300, 3400) | 2000 (1300, 3400) | 2000 (1400, 3200) | 2100 (1300, 3400) |
| Fruits (g/day) | 67 (19, 150) | 100 (33, 220) | 140 (55, 290) | 180 (85, 350) | 260 (130, 460) | 130 (37, 320) |
| Vegetables (g/day) | 82 (28, 160) | 120 (47, 220) | 150 (68, 270) | 180 (92, 320) | 230 (130, 400) | 140 (51, 290) |
| Nuts and legumes (g/day) | 24 (0.0, 65) | 26 (3.2, 77) | 29 (4.8, 83) | 35 (11, 90) | 49 (19, 120) | 28 (3.5, 83) |
| Low-fat dairy products (g/day) | 19 (0.0, 440) | 150 (0.0, 600) | 210 (0.0, 660) | 260 (0.0, 690) | 320 (61, 750) | 190 (0.0, 650) |
| Whole grains (g/day) | 90 (17, 250) | 120 (38, 310) | 140 (45, 330) | 140 (53, 330) | 160 (67, 330) | 130 (39, 320) |
| Sodium (mg/day) | 3600 (2500, 4300) | 3400 (2300, 4200) | 3100 (2200, 4200) | 2800 (2200, 4100) | 2600 (2200, 3800) | 3200 (2200, 4200) |
| Sweetened beverages (g/day) | 140 (0.0, 660) | 31 (0.0, 510) | 0.0 (0.0, 330) | 0.0 (0.0, 260) | 0.0 (0.0, 140) | 0.0 (0.0, 430) |
| Red and processed meat (g/day) | 100 (47, 160) | 84 (35, 150) | 77 (33, 140) | 69 (30, 130) | 56 (20, 99) | 78 (31, 140) |
BMI, body mass index; DASH, Dietary Approach to Stop Hypertension; HRT, hormone replacement therapy; MI, myocardial infarction.
In women only (n = 35 004).
Adherence to the DASH diet and risk of HF
After multivariable adjustment (Table 2, multivariable adjusted 1), those with the greatest adherence to the DASH diet at baseline (HR 0.85, 95% CI: 0.80, 0.91) and long-term adherence (HR 0.83, 95% CI: 0.78, 0.89) had a lower risk of HF compared to those with the lowest adherence. Results were similar after further adjustment for BMI, hypertension, hypercholesterolaemia, diabetes, and atrial fibrillation at baseline. When adherence to the DASH diet was modelled as restricted cubic splines, a clear linear inverse association was observed with greater adherence to the DASH diet (Figure 1). For each 2 standard deviations higher DASH diet score (9.1 points), a lower risk of HF was observed (HR 0.89, 95% CI: 0.86, 0.93). We estimated that, under the assumption of a causal effect, about 19 cases (95% CI: 13, 25) of HF could have been prevented for every 10 000 person-years if all had the greatest adherence to the DASH diet (Table 3). In stratified analyses, there was not any clear effect modification by previous diabetes or hypertension (Supplementary material online, Figures S4 and S5).
Association between adherence to the Dietary Approach to Stop Hypertension diet score and risk of heart failure. The Dietary Approach to Stop Hypertension diet score was modelled as a restrictive cubic spline with four knots (reference is set at 20 points). Multivariable adjusted: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, use of hormone replacement therapy in women, family history of myocardial infarction and total energy intake.
Hazard ratios and 95% confidence intervals for the association between adherence to the Dietary Approach to Stop Hypertension diet score and risk of heart failure (n = 76 122, n HF cases = 12 164, person-years = 6 001 289)
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2881 | 2908 | 2089 | 2482 | 1804 |
| Person-years | 1 391 820 | 1 384 579 | 992 371 | 1 252 345 | 980 174 |
| HR (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 1.00 (ref) | 0.93 (0.88, 0.98) | 0.90 (0.85, 0.95) | 0.82 (0.78, 0.87) | 0.76 (0.71, 0.81) |
| Multivariable adjusted 1b | 1.00 (ref) | 0.96 (0.91, 1.01) | 0.96 (0.91, 1.02) | 0.90 (0.85, 0.95) | 0.85 (0.80, 0.91) |
| Multivariable adjusted 2c | 1.00 (ref) | 0.94 (0.90, 0.99) | 0.94 (0.89, 1.00) | 0.87 (0.82, 0.92) | 0.83 (0.78, 0.88) |
| Long-term diet | |||||
| Age and sex adjusted | 1.00 (ref) | 0.87 (0.83, 0.92) | 0.81 (0.77, 0.86) | 0.79 (0.74, 0.83) | 0.74 (0.69, 0.78) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.91 (0.87, 0.96) | 0.87 (0.82, 0.92) | 0.86 (0.81, 0.91) | 0.83 (0.78, 0.89) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.90 (0.86, 0.95) | 0.86 (0.81, 0.91) | 0.84 (0.80, 0.89) | 0.82 (0.77, 0.87) |
| Primary HF diagnosis onlyd | |||||
| Cases | 1124 | 1116 | 756 | 925 | 688 |
| Person-years | 1 254 117 | 1 241 648 | 885 081 | 1 126 044 | 889 236 |
| Age and sex adjusted | 1.00 (ref) | 0.93 (0.86, 1.01) | 0.87 (0.79, 0.95) | 0.82 (0.75, 0.90) | 0.77 (0.70, 0.86) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.97 (0.89, 1.05) | 0.93 (0.85, 1.02) | 0.90 (0.82, 0.98) | 0.88 (0.79, 0.98) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.95 (0.87, 1.03) | 0.91 (0.83, 1.00) | 0.86 (0.79, 0.95) | 0.85 (0.77, 0.95) |
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2881 | 2908 | 2089 | 2482 | 1804 |
| Person-years | 1 391 820 | 1 384 579 | 992 371 | 1 252 345 | 980 174 |
| HR (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 1.00 (ref) | 0.93 (0.88, 0.98) | 0.90 (0.85, 0.95) | 0.82 (0.78, 0.87) | 0.76 (0.71, 0.81) |
| Multivariable adjusted 1b | 1.00 (ref) | 0.96 (0.91, 1.01) | 0.96 (0.91, 1.02) | 0.90 (0.85, 0.95) | 0.85 (0.80, 0.91) |
| Multivariable adjusted 2c | 1.00 (ref) | 0.94 (0.90, 0.99) | 0.94 (0.89, 1.00) | 0.87 (0.82, 0.92) | 0.83 (0.78, 0.88) |
| Long-term diet | |||||
| Age and sex adjusted | 1.00 (ref) | 0.87 (0.83, 0.92) | 0.81 (0.77, 0.86) | 0.79 (0.74, 0.83) | 0.74 (0.69, 0.78) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.91 (0.87, 0.96) | 0.87 (0.82, 0.92) | 0.86 (0.81, 0.91) | 0.83 (0.78, 0.89) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.90 (0.86, 0.95) | 0.86 (0.81, 0.91) | 0.84 (0.80, 0.89) | 0.82 (0.77, 0.87) |
| Primary HF diagnosis onlyd | |||||
| Cases | 1124 | 1116 | 756 | 925 | 688 |
| Person-years | 1 254 117 | 1 241 648 | 885 081 | 1 126 044 | 889 236 |
| Age and sex adjusted | 1.00 (ref) | 0.93 (0.86, 1.01) | 0.87 (0.79, 0.95) | 0.82 (0.75, 0.90) | 0.77 (0.70, 0.86) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.97 (0.89, 1.05) | 0.93 (0.85, 1.02) | 0.90 (0.82, 0.98) | 0.88 (0.79, 0.98) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.95 (0.87, 1.03) | 0.91 (0.83, 1.00) | 0.86 (0.79, 0.95) | 0.85 (0.77, 0.95) |
BMI, body mass index; CI, confidence interval; DASH, Dietary Approach to Stop Hypertension; HF, heart failure; HR, hazard ratio.
Age and sex adjusted: age, sex.
Multivariable adjusted 1: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, use of hormone replacement therapy in women, family history of myocardial infarction, total energy intake.
Multivariable adjusted 2: multivariable adjusted 1, BMI, history of hypertension, hypercholesterolaemia, diabetes and atrial fibrillation at baseline.
n = 68 567, n HF cases = 4609, baseline diet.
Hazard ratios and 95% confidence intervals for the association between adherence to the Dietary Approach to Stop Hypertension diet score and risk of heart failure (n = 76 122, n HF cases = 12 164, person-years = 6 001 289)
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2881 | 2908 | 2089 | 2482 | 1804 |
| Person-years | 1 391 820 | 1 384 579 | 992 371 | 1 252 345 | 980 174 |
| HR (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 1.00 (ref) | 0.93 (0.88, 0.98) | 0.90 (0.85, 0.95) | 0.82 (0.78, 0.87) | 0.76 (0.71, 0.81) |
| Multivariable adjusted 1b | 1.00 (ref) | 0.96 (0.91, 1.01) | 0.96 (0.91, 1.02) | 0.90 (0.85, 0.95) | 0.85 (0.80, 0.91) |
| Multivariable adjusted 2c | 1.00 (ref) | 0.94 (0.90, 0.99) | 0.94 (0.89, 1.00) | 0.87 (0.82, 0.92) | 0.83 (0.78, 0.88) |
| Long-term diet | |||||
| Age and sex adjusted | 1.00 (ref) | 0.87 (0.83, 0.92) | 0.81 (0.77, 0.86) | 0.79 (0.74, 0.83) | 0.74 (0.69, 0.78) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.91 (0.87, 0.96) | 0.87 (0.82, 0.92) | 0.86 (0.81, 0.91) | 0.83 (0.78, 0.89) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.90 (0.86, 0.95) | 0.86 (0.81, 0.91) | 0.84 (0.80, 0.89) | 0.82 (0.77, 0.87) |
| Primary HF diagnosis onlyd | |||||
| Cases | 1124 | 1116 | 756 | 925 | 688 |
| Person-years | 1 254 117 | 1 241 648 | 885 081 | 1 126 044 | 889 236 |
| Age and sex adjusted | 1.00 (ref) | 0.93 (0.86, 1.01) | 0.87 (0.79, 0.95) | 0.82 (0.75, 0.90) | 0.77 (0.70, 0.86) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.97 (0.89, 1.05) | 0.93 (0.85, 1.02) | 0.90 (0.82, 0.98) | 0.88 (0.79, 0.98) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.95 (0.87, 1.03) | 0.91 (0.83, 1.00) | 0.86 (0.79, 0.95) | 0.85 (0.77, 0.95) |
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2881 | 2908 | 2089 | 2482 | 1804 |
| Person-years | 1 391 820 | 1 384 579 | 992 371 | 1 252 345 | 980 174 |
| HR (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 1.00 (ref) | 0.93 (0.88, 0.98) | 0.90 (0.85, 0.95) | 0.82 (0.78, 0.87) | 0.76 (0.71, 0.81) |
| Multivariable adjusted 1b | 1.00 (ref) | 0.96 (0.91, 1.01) | 0.96 (0.91, 1.02) | 0.90 (0.85, 0.95) | 0.85 (0.80, 0.91) |
| Multivariable adjusted 2c | 1.00 (ref) | 0.94 (0.90, 0.99) | 0.94 (0.89, 1.00) | 0.87 (0.82, 0.92) | 0.83 (0.78, 0.88) |
| Long-term diet | |||||
| Age and sex adjusted | 1.00 (ref) | 0.87 (0.83, 0.92) | 0.81 (0.77, 0.86) | 0.79 (0.74, 0.83) | 0.74 (0.69, 0.78) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.91 (0.87, 0.96) | 0.87 (0.82, 0.92) | 0.86 (0.81, 0.91) | 0.83 (0.78, 0.89) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.90 (0.86, 0.95) | 0.86 (0.81, 0.91) | 0.84 (0.80, 0.89) | 0.82 (0.77, 0.87) |
| Primary HF diagnosis onlyd | |||||
| Cases | 1124 | 1116 | 756 | 925 | 688 |
| Person-years | 1 254 117 | 1 241 648 | 885 081 | 1 126 044 | 889 236 |
| Age and sex adjusted | 1.00 (ref) | 0.93 (0.86, 1.01) | 0.87 (0.79, 0.95) | 0.82 (0.75, 0.90) | 0.77 (0.70, 0.86) |
| Multivariable adjusted 1 | 1.00 (ref) | 0.97 (0.89, 1.05) | 0.93 (0.85, 1.02) | 0.90 (0.82, 0.98) | 0.88 (0.79, 0.98) |
| Multivariable adjusted 2 | 1.00 (ref) | 0.95 (0.87, 1.03) | 0.91 (0.83, 1.00) | 0.86 (0.79, 0.95) | 0.85 (0.77, 0.95) |
BMI, body mass index; CI, confidence interval; DASH, Dietary Approach to Stop Hypertension; HF, heart failure; HR, hazard ratio.
Age and sex adjusted: age, sex.
Multivariable adjusted 1: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, use of hormone replacement therapy in women, family history of myocardial infarction, total energy intake.
Multivariable adjusted 2: multivariable adjusted 1, BMI, history of hypertension, hypercholesterolaemia, diabetes and atrial fibrillation at baseline.
n = 68 567, n HF cases = 4609, baseline diet.
Rate differences and 95% confidence intervals for the association between adherence to the Dietary Approach to Stop Hypertension diet score and risk of heart failure after 20 years of follow-up (n = 76 122, n cases = 10 634, person-years = 5 912 940)
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2526 | 2580 | 1807 | 2167 | 1554 |
| Person-years | 1 371 560 | 1 364 929 | 977 732 | 1 233 664 | 965 055 |
| Rate difference per 10 000 person-years (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 0 (ref) | −8.1 (−13.0, −3.3) | −14.3 (−19.6, −9.0) | −21.8 (−26.9, −16.7) | −30.9 (−36.3, −25.5) |
| Multivariable adjusted 1b | 0 (ref) | −3.5 (−8.3, 1.39) | −6.2 (−11.3, −0.8) | −11.4 (−16.5, −6.3) | −17.3 (−22.9, −11.7) |
| Multivariable adjusted 2c | 0 (ref) | −4.1 (−8.9, 0.8) | −6.9 (−12.2, −1.5) | −12.6 (−17.7, −7.5) | −18.9 (−24.5, −13.3) |
| Long-term diet | |||||
| Age and sex adjusted | 0 (ref) | −14.1 (−19.1, −9.1) | −22.2 (−27.6, −16.8) | −25.8 (−30.9, −20.7) | −33.1 (−38.6, −27.6) |
| Multivariable adjusted 1 | 0 (ref) | −8.7 (−13.7, −3.7) | −13.8 (−19.2, −8.4) | −14.7 (−19.9, −9.5) | −19.0 (−24.7, −13.3) |
| Multivariable adjusted 2 | 0 (ref) | −9.0 (−14.0, −4.0) | −14.6 (−20.0, −9.2) | −15.2 (−20.4, −10.0) | −19.8 (−25.5, −14.1) |
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2526 | 2580 | 1807 | 2167 | 1554 |
| Person-years | 1 371 560 | 1 364 929 | 977 732 | 1 233 664 | 965 055 |
| Rate difference per 10 000 person-years (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 0 (ref) | −8.1 (−13.0, −3.3) | −14.3 (−19.6, −9.0) | −21.8 (−26.9, −16.7) | −30.9 (−36.3, −25.5) |
| Multivariable adjusted 1b | 0 (ref) | −3.5 (−8.3, 1.39) | −6.2 (−11.3, −0.8) | −11.4 (−16.5, −6.3) | −17.3 (−22.9, −11.7) |
| Multivariable adjusted 2c | 0 (ref) | −4.1 (−8.9, 0.8) | −6.9 (−12.2, −1.5) | −12.6 (−17.7, −7.5) | −18.9 (−24.5, −13.3) |
| Long-term diet | |||||
| Age and sex adjusted | 0 (ref) | −14.1 (−19.1, −9.1) | −22.2 (−27.6, −16.8) | −25.8 (−30.9, −20.7) | −33.1 (−38.6, −27.6) |
| Multivariable adjusted 1 | 0 (ref) | −8.7 (−13.7, −3.7) | −13.8 (−19.2, −8.4) | −14.7 (−19.9, −9.5) | −19.0 (−24.7, −13.3) |
| Multivariable adjusted 2 | 0 (ref) | −9.0 (−14.0, −4.0) | −14.6 (−20.0, −9.2) | −15.2 (−20.4, −10.0) | −19.8 (−25.5, −14.1) |
BMI, body mass index; CI, confidence interval; DASH, Dietary Approach to Stop Hypertension.
Age and sex adjusted: age, sex.
Multivariable adjusted 1: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, family history of myocardial infarction.
Multivariable adjusted 2: multivariable adjusted 1, BMI, history of hypertension, hypercholesterolaemia, diabetes and atrial fibrillation at baseline.
Rate differences and 95% confidence intervals for the association between adherence to the Dietary Approach to Stop Hypertension diet score and risk of heart failure after 20 years of follow-up (n = 76 122, n cases = 10 634, person-years = 5 912 940)
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2526 | 2580 | 1807 | 2167 | 1554 |
| Person-years | 1 371 560 | 1 364 929 | 977 732 | 1 233 664 | 965 055 |
| Rate difference per 10 000 person-years (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 0 (ref) | −8.1 (−13.0, −3.3) | −14.3 (−19.6, −9.0) | −21.8 (−26.9, −16.7) | −30.9 (−36.3, −25.5) |
| Multivariable adjusted 1b | 0 (ref) | −3.5 (−8.3, 1.39) | −6.2 (−11.3, −0.8) | −11.4 (−16.5, −6.3) | −17.3 (−22.9, −11.7) |
| Multivariable adjusted 2c | 0 (ref) | −4.1 (−8.9, 0.8) | −6.9 (−12.2, −1.5) | −12.6 (−17.7, −7.5) | −18.9 (−24.5, −13.3) |
| Long-term diet | |||||
| Age and sex adjusted | 0 (ref) | −14.1 (−19.1, −9.1) | −22.2 (−27.6, −16.8) | −25.8 (−30.9, −20.7) | −33.1 (−38.6, −27.6) |
| Multivariable adjusted 1 | 0 (ref) | −8.7 (−13.7, −3.7) | −13.8 (−19.2, −8.4) | −14.7 (−19.9, −9.5) | −19.0 (−24.7, −13.3) |
| Multivariable adjusted 2 | 0 (ref) | −9.0 (−14.0, −4.0) | −14.6 (−20.0, −9.2) | −15.2 (−20.4, −10.0) | −19.8 (−25.5, −14.1) |
| DASH diet score | |||||
|---|---|---|---|---|---|
| Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | |
| Range | 8–21 | 22–24 | 25–26 | 27–29 | 30–40 |
| Cases | 2526 | 2580 | 1807 | 2167 | 1554 |
| Person-years | 1 371 560 | 1 364 929 | 977 732 | 1 233 664 | 965 055 |
| Rate difference per 10 000 person-years (95% CI) | |||||
| Baseline diet | |||||
| Age and sex adjusteda | 0 (ref) | −8.1 (−13.0, −3.3) | −14.3 (−19.6, −9.0) | −21.8 (−26.9, −16.7) | −30.9 (−36.3, −25.5) |
| Multivariable adjusted 1b | 0 (ref) | −3.5 (−8.3, 1.39) | −6.2 (−11.3, −0.8) | −11.4 (−16.5, −6.3) | −17.3 (−22.9, −11.7) |
| Multivariable adjusted 2c | 0 (ref) | −4.1 (−8.9, 0.8) | −6.9 (−12.2, −1.5) | −12.6 (−17.7, −7.5) | −18.9 (−24.5, −13.3) |
| Long-term diet | |||||
| Age and sex adjusted | 0 (ref) | −14.1 (−19.1, −9.1) | −22.2 (−27.6, −16.8) | −25.8 (−30.9, −20.7) | −33.1 (−38.6, −27.6) |
| Multivariable adjusted 1 | 0 (ref) | −8.7 (−13.7, −3.7) | −13.8 (−19.2, −8.4) | −14.7 (−19.9, −9.5) | −19.0 (−24.7, −13.3) |
| Multivariable adjusted 2 | 0 (ref) | −9.0 (−14.0, −4.0) | −14.6 (−20.0, −9.2) | −15.2 (−20.4, −10.0) | −19.8 (−25.5, −14.1) |
BMI, body mass index; CI, confidence interval; DASH, Dietary Approach to Stop Hypertension.
Age and sex adjusted: age, sex.
Multivariable adjusted 1: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, family history of myocardial infarction.
Multivariable adjusted 2: multivariable adjusted 1, BMI, history of hypertension, hypercholesterolaemia, diabetes and atrial fibrillation at baseline.
Food substitutions and risk of HF
Replacing 1 serving/day of red and processed meat with 1 serving/day of fruits, vegetables, nuts and legumes, low-fat dairy, or whole grains was associated with ∼8–12% lower risk of HF after multivariable adjustment (Figure 2). Similar associations were observed after adjustment for BMI, hypertension, hypercholesterolaemia, diabetes, and atrial fibrillation at baseline (Supplementary material online, Figure S6) and when the model was not adjusted for total energy intake (Supplementary material online, Figures S7 and S8).
Replacement of red and processed meat with other Dietary Approach to Stop Hypertension diet foods on risk of heart failure. Serving sizes were 100 g/day for red and processed meat, 100 g/day for fruit, vegetables, and whole grains, 200 g/day for low-fat dairy and 20 g/day for nuts and legumes. Adjusted for each of the included foods, except for red and processed meat, a total-variable all substitution foods as well as: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, use of hormone replacement therapy in women, family history of myocardial infarction and total energy intake.
Individual DASH diet components and risk of HF
Non-linear associations were observed for the intake of low-fat dairy and red and processed meat (Figure 3). For low-fat dairy, the lowest risk of HF, compared with no intake, was observed at around 150 g/day (pnon-linear trend = 0.002). For red and processed meat, the lowest risk of HF, compared with no intake, was observed around 50 g/day (Pnon-linear trend < 0.001). In contrast, a linear relation was observed for sweetened beverages (Pnon-linear trend = 0.40) and for sodium (Pnon-linear trend = 0.51) (Figure 3). Linear relationships were also observed for remaining component (Supplementary material online, Figure S9).
Association between adherence to each Dietary Approach to Stop Hypertension diet score component and risk of heart failure. Red and processed meat, low-fat dairy, sweetened beverages and sodium was modelled as restrictive cubic splines with four knots. Adjusted for: age, sex, education, smoking, alcohol, walking/cycling, corticosteroid use, aspirin use, dietary supplements use, use of hormone replacement therapy in women, family history of myocardial infarction, total energy intake and mutually adjusted for each of the other foods. Sodium was not included in models with foods.
Sensitivity analyses
The pattern of associations was similar to the main analysis after including only those with a primary diagnosis code for HF (Table 2), using multiple imputation for missing variables (HR 0.86, 95% CI: 0.80, 0.91 for baseline comparing the highest to lowest adherence group) or excluding those at high risk of HF at baseline (HR 0.80, 95% CI: 0.73, 0.87 for baseline comparing the highest to lowest adherence group). In total, 2121 participants developed cancer before HF. Associations were similar after censoring participants who developed cancer before HF [HR (95% CI), baseline diet; for multivariable model 2: 0.84 (0.78, 0.90); multivariable model 3: 0.81 (0.77, 0.87); comparing the highest to lowest adherence group]. During the follow-up period, HRs were attenuated over time when comparing those with the highest to the lowest adherence to the DASH diet using multivariate model 1 and stopping the follow-up time after 5, 10, 15, and 20 years of follow-up [corresponding HRs (95% CIs) were: 0.77 (0.63, 0.93), 0.77 (0.68, 0.87), 0.83 (0.78, 0.90), 0.85 (0.80, 0.92)].
Discussion
In this cohort of Swedish middle-aged and elderly adults, we found that greater adherence to the DASH diet was associated with a lower risk of HF. Replacing 1 serving/day of red meat with 1 serving/day of fruits, vegetables, nuts, and legumes, low-fat dairy, or whole grains was also associated with an 8–12% lower risk of HF, although investigation into individual foods suggested that low intakes of low-fat dairy (around 150 g/day) and red meat (around 50 g/day) may be associated with a lower risk whereas higher intakes were associated with a higher risk of HF, compared with no intake.
Strengths and limitations
Strengths of this study include the prospective and population-based design, nearly complete follow-up via linkage to national health registries with high validity of the HF diagnoses and assessment of long-term dietary intakes. Dietary intake was measured using self-reported FFQs and, although the FFQ showed reasonable correlations compared with four 1-week diet records distributed over a 1-year period, measurement error is present. Due to the design, this measurement error is, however, unlikely to be related to the outcome and, therefore, most likely to have attenuated our results. Indeed, when we updated the dietary information among those who answered the second FFQ in 2009, using a better measure of long-term dietary intake, the associations were slightly stronger indicating that bias towards the null was present. The sensitivity analyses did not indicate substantial bias from misclassification of HF diagnosis, reverse causation, missing data or by HF induced by anti-cancer treatments, as results were in the same direction as the main analysis. HRs were attenuated over the follow-up period, indicating an effect of the built-in selection bias of the HR; likely in combination of the misclassification of dietary intake over time. The characteristics of those who adhered to the DASH diet, compared with those who did not, suggested an overall healthier lifestyle. Hence, we cannot exclude residual confounding from a healthy lifestyle in domains that were not measured that would bias our results towards greater benefit of adherence. We also did not have information on the use of cardiovascular medications beside aspirin intake and we did not have information on subtypes of HF from the register. However, in a recent study, there was not observed a difference in the association of DASH diet with incident HF with reduced ejection fraction compared with HF with preserved ejection fractions.10
Comparison to other studies
Only few cohort studies have looked at adherence to the DASH diet and incidence for HF and results have been conflicting. A lower risk was observed in the Multi-Ethnic Study of Atherosclerosis9 whereas a higher risk was reported in the Cardiovascular Health Study; neither being statistically significant,8 which could be because both cohorts included fewer cases than in our study or because the study population was older in the Cardiovascular Health Study, which has been found to modify the association in other cohorts.10 More in line with our results, a 27% (95% CI: 8, 42%) lower risk was observed in a diverse US population.10 The stronger association observed, compared with our study, may be due to shorter follow-up or differences in the underlying dietary patterns such as lower intake of sugar-sweetened beverages in the Swedish compared with the USA cohort. In our two previous publications, we observed that adherence to the DASH diet was associated with a lower risk of HF in men and women.11,12 In those studies, the follow-up time was shorter and with fewer HF cases than in the present study. The point estimates were stronger in those two previous studies but had wider CIs compared to this study. The stronger point estimates may be due to the FFQs being more representative of usual diet over the shorter time period, as our sensitivity analyses with updated dietary intake and different follow-up times indicated. More broadly, adherence to the DASH has in multiple other cohorts been associated with a lower risk of cardiovascular disease and all-cause mortality.5,28
The initial DASH diet trial reported a 5.5-mmHg lower systolic blood pressure with the DASH diet compared with a standard American diet over 8 weeks29 and RCTs using pharmacological interventions have demonstrated that every 5 mmHg lower systolic blood pressure lowers the risk of HF by 24%,30 which is comparable to our estimate of 18% (95% CI: 13–23%) lower risk in the highest quintile of the DASH diet.
Potential mechanisms explaining results
Several trials of the DASH diet (0.5–12 months long) have investigated its effect on key risk markers for the development of HF. An umbrella review of meta-analyses of the RCTs concluded that consuming a DASH diet reduced blood pressure, body weight, improved glucose regulation, and lowered LDL-cholesterol compared with different control diets,6 in line with our finding of a lower long-term risk. Also, adherence to a DASH diet lowered levels of biomarkers for cardiac strain and cardiac injury in a controlled feeding study over 4 weeks.31
When investigating individual food components of the DASH diet, we found that replacing red and processed meat with other DASH diet foods was associated with a lower risk. This is also consistent with results from a meta-analysis of RCTs specifying control diets to interventions of red meat; intake of red meat increased LDL-cholesterol compared with plant-protein sources such as legumes and nuts.32 Our substitution analysis assumed a linear relationship; however, when we investigated the individual components using a flexible model, low intakes of red and processed meat (around 50 g/day) and low-fat dairy (around 200 g/day) were associated with a lower risk, whereas higher intakes were associated with a higher risk compared with no intake. This shows that red meat, an important source of haem iron, may at low intake levels prevent iron deficiency and lower risk of HF.14 Iron deficiency is recognized in 35–55% of patients with HF, which worsens their prognosis.14 However, higher intake, especially of processed red meat (41% of total red and processed meat consumed in our study population) containing high amounts of sodium, is related to a higher risk,33 as suggested by a strong dose–response relation between total sodium intake and risk of HF in this study. A similar association was observed for low-fat dairy where an intake of about 150 g/day was associated with a lower HF risk, an intake comparable to the DASH diet trials, and intakes beyond this were associated with a higher risk. Very high intake of milk has been hypothesized to induce oxidative stress,34 part of which could be due to the displacement of other foods like fruits and vegetables in the diet, and thereby be associated with a higher risk of HF. Intake of sugar-sweetened beverages has shown to increase body weight in RCTs and cohort studies,35 and this may explain why we found a strong linear increased risk, whereas higher intake of vegetables, which has been shown to lower body weight in ad libitum settings,36 was associated with a lower risk.
Public health implications
Based on our long-term prospective study with robust findings across sensitivity analyses and findings in accordance with short-term RCTs we, under the assumption of a causal relation, estimated that between 13 and 25 cases per 10 000 person-years could have been prevented if the study population aged 45–85 years had adhered to the DASH diet. Viewed in the light of everyone being ‘exposed’, this would be an important contribution to population-based prevention efforts.
Although there are biological arguments for the investigation of effect modification by the history of diabetes or hypertension,1 we found little evidence for effect modification, suggesting particular subgroups that would benefit from adhering to the DASH diet by any of these factors on either scale. One explanation is lack of power, as these subgroups included few participants. As we used self-reported prevalence of both history of hypertension and diabetes at the baseline, measurement error could also explain why we did not observe any statistically significant differences.
Conclusion
Long-term adherence to the DASH diet was associated with a lower risk of HF in a cohort of middle-aged and elderly Swedish adults. Replacing 1 serving/day of red and processed meat with 1 serving/day of other DASH diet foods, such as fruits, vegetables, low-fat dairy, nuts and legumes, or whole grains, was also associated with about 10% lower risk of HF. Our findings, together with the evidence from relatively short RCTs, underline the importance of supporting the population in adhering to a healthy diet, like the DASH diet, for the prevention of HF in the general population.
Supplementary material
Supplementary material is available at European Heart Journal – Cardiovascular Pharmacotherapy online.
Acknowledgements
We acknowledge the National Research Infrastructure SIMPLER (www.simpler4health.se), supported by the Swedish Research Council (grant number 2017-00644), for use of the databases from the Swedish Mammography Cohort and the Cohort of Swedish Men.
Data availability
The data underlying this article cannot be shared publicly because the data are classified as sensitive data under the European General Data Protection Regulation. The data and underlying code can be shared on reasonable request to the corresponding author through the Swedish National Research Infrastructure SIMPLER (www.simpler4health.se).
Funding
This study was supported by the Swedish Research Council (VR grant 2015—02302) and by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS grant 2016—00308).
Conflict of interest: E.B.L. reports research funding from Amgen, Inc, and consulting fees for a scientific project funded by Novartis, unrelated to the current research. Remaining authors declare no conflicts of interest.
Comments