|Home | About | Journals | Submit | Contact Us | Français|
Guidelines to prevent and treat hypertension advocate the Dietary Approaches to Stop Hypertension (DASH) diet.
We studied whether a greater concordance with the DASH diet is associated with reduced incidence of hypertension (self-reported) and mortality from cardiovascular disease in 20,993 women initially aged 55–69. We created a DASH diet concordance score using food frequency data in 1986 and followed the women for events through 2002.
No woman had perfect concordance with the DASH diet. Adjusted for age and energy intake, incidence of hypertension was inversely associated with the degree of concordance with the DASH diet, with hazard ratios across quintiles of 1.0, 0.91, 0.95, 0.99, and 0.87 (p trend = 0.02). There also were inverse, but not monotonic, associations between better DASH diet concordance and mortality from coronary heart disease, stroke, and all CVD. However, after adjustment for other risk factors, there was little evidence that any endpoint was associated with the DASH diet score.
Our results suggest that greater concordance with DASH guidelines did not have an independent long-term association with hypertension or cardiovascular mortality in this cohort. This implies that very high concordance, as achieved in the DASH trials, may be necessary to achieve the benefits of the DASH diet.
Current guidelines on the prevention and treatment of hypertension prominently emphasize lifestyle modifications in addition to pharmacologic therapy.1 In particular, the National High Blood Pressure Education Program and the American Heart Association have broadly recommended the Dietary Approaches to Stop Hypertension (DASH) diet, plus sodium reduction, to reduce high blood pressure.2,3
The DASH diet emphasizes increased consumption of fruits, vegetables, and low-fat dairy products; includes whole grains, poultry, fish and nuts; and is reduced in fats, red meat, sweets, and sugar containing beverages. In controlled feeding studies of people with above optimal blood pressure, the DASH diet for 2 months substantially lowered blood pressure (systolic by 5.5 mm Hg) compared with a control diet low in micronutrients,4 and when combined with sodium reduction lowered blood pressure even further (systolic by 8.9 mm Hg).5,6 The DASH diet also beneficially lowered total and LDL cholesterol (−14 and −11 mg/dL, respectively), but it lowered HDL cholesterol (−4 mg/dL) as well.7 To test the generalizability of the DASH diet findings, the PREMIER clinical trial tested behavioral interventions to encourage adoption of the DASH diet to lower above optimal blood pressure in outpatients.8 The PREMIER intervention for six months was able to stimulate increases in fruit, vegetable and dairy consumption and thereby reduced blood pressure (systolic by 4.3 mm Hg) among those randomized to follow DASH diet plus other established nonpharmacologic anti-hypertension guidelines versus advice-only controls; however the systolic blood pressure reduction was only 0.6 mm Hg less than in those following the established guidelines without the DASH diet.8,9
A reduction in blood pressure and LDL cholesterol in the DASH feeding studies is presumed to translate to lower cardiovascular disease incidence. To our knowledge, there have been no large longitudinal studies of whether adherence, complete or partial, to the DASH guidelines ultimately reduces occurrence of hypertension or cardiovascular events. Beneficial results from clinical trials sometimes do not translate to benefit at the population level because of different population characteristics or levels of dietary compliance. We therefore examined whether a diet with greater concordance with DASH guidelines is associated with reduced incidence of hypertension and mortality from cardiovascular disease in initially non-hypertensive participants in the Iowa Women’s Health Study (IWHS).
In January 1986, an explanatory cover letter and questionnaire were mailed to 98,826 women aged 55–69 years randomly selected from the Iowa state driver’s license list.10 A total of 41,386 women completed the questionnaire, which included questions regarding diet, smoking habits, physical activity, weight, height, reproductive history, history of cardiovascular disease, hypertension, diabetes, and demographic information. Subsequently, women were resurveyed by mail in 1987 (91% response), 1989 (90%), 1992 (83%), and 1997 (79%), to obtain vital status and certain disease endpoints. As approved by the Institutional Review Boards of the University of Minnesota and University of Iowa, a returned questionnaire implied informed consent.
With the baseline questionnaire, a paper tape measure was enclosed for a friend to measure the participant’s waist (umbilical level) and hips (maximum) so that waist/hip ratio could be derived. Reported body weight and height were used to calculate body mass index (in kg/m2). Participants were asked to report how often they took part in moderate and vigorous physical activities, and a physical activity index was derived.11
Usual dietary and alcohol intake was assessed with a semi-quantitative food-frequency questionnaire used in the Nurses’ Health Study.12 Participants were asked their average consumption, over the past year, of 127 food items. Daily intakes of nutrients were calculated from the food frequency by multiplying the frequency of consumption of the specified unit of each food by the nutrient contents of that unit of food. The food-frequency questionnaire used in the present study was found to account for 93% of total energy intake in a validation study with 194 female nurses.12 In a validation study in the Iowa population, the energy adjusted correlation between total fat, saturated fat, and carbohydrate intake estimates derived from the food frequency questionnaire compared with average of five 24-hour dietary recalls were 0.62, 0.59, and 0.45 respectively.13 The median correlation between energy-adjusted intake by the two methods was r = 0.45 for all macronutrients and r = 0.64 for micronutrients (including supplements).11
Although indices to assess dietary concordance with the Dietary Guidelines for Americans14 have been developed and evaluated,15,16 an index for the DASH diet has not been derived previously. Thus, using the general approach applied in developing other dietary indices, we constructed an index to measure the degree of concordance with the DASH low sodium diet,2,17 hereafter called the “DASH diet.” The components of the index, outlined in Table 1, were weighted and summed to calculate a single score to rank each participant according to dietary concordance. Wherever possible, quantitative criteria provided as part of the DASH diet were used in establishing cut points for scoring index items. More specifically, daily servings recommended for individuals needing 2,000 calories a day were used in establishing cut-point criteria for food groups that could be estimated from our food frequency questionnaire. For example, the DASH diet guidelines recommend 4–5 daily servings of vegetables for those needing 2,000 calories a day. Thus, a full point was assigned to those consuming 4 or more servings of vegetables daily, with a partial point (0.5) given to those with vegetable intake approaching the recommended level (2–3 servings/day) and no points (0) given to those consuming far less than the recommended level (fewer than 2 servings/day). Reverse scoring was applied for food groups where less intake is recommended. For example, a full point was assigned to those consuming 5 or less servings of sweets per week, with a partial point (0.5) given to those consuming 6–7 servings per week, and no points (0) given to those consuming 8 or more servings of sweets per week. Serving size specifications for the index were matched to the DASH Diet plan (2) serving size recommendations to the extent possible. Due to the way in which the study food frequency questionnaire was designed, it was not possible to make the serving sizes equivalent for all foods (Table 2). For example, the food frequency questionnaire asked participants to report how frequently they drank orange juice, with “small glass” provided as the referent portion on the questionnaire. In contrast, the DASH eating plan specifies “1/2 cup” as the serving size for fruit juices. Nutrient intake estimates (percentage of calories from fat and percentage of calories from saturated fatty acids) were used to evaluate concordance with the fat-related components of the DASH diet because fats and oils are spread across various food groupings making this element of the diet difficult to account for in the scoring of individual food groupings. Furthermore, although the DASH diet emphasizes low-fat dairy and low-fat meats, it is difficult to classify individual dairy foods and meats as ‘low-fat’ (e.g., 2% milk could be considered either a low-fat or regular-fat dairy food). Estimated sodium intake (mg/day) was used a criterion because the DASH diet guidelines include quantitative criteria for this nutrient.
The scoring system was based on the premise that each of the major DASH diet guidelines should contribute equally to the total index score. The maximum score for each component was 1, with a maximum total of 11, which would indicate total concordance with the DASH diet guidelines. A score of 0 would represent total non-concordance.
Information on physician diagnoses of hypertension was collected by self-report in each follow-up mailed survey through 1997. The accuracy of self-report of hypertension was verified previously in a sub-sample of the IWHS cohort.18 Physicians of 45 women who self-reported hypertension were contacted, and they verified the diagnosis in 72%. Physicians of 42 women who reported no hypertension verified that 100% were not hypertensive. Incident hypertension was defined as occurring at the midpoint between the questionnaire reporting hypertension and the preceding questionnaire reporting no hypertension. Person-years of follow-up for hypertension otherwise went until death, loss to follow-up or December 31, 1997, whichever occurred first.
Deaths were found through the State Health Registry of Iowa, follow-up mail surveys, and for survey nonrespondents, via the National Death Index. Underlying causes of deaths were coded by state nosologists according to the International Classification of Diseases 9th or 10th revisions. Person-years of follow-up for mortality went until death or else December 31, 2002.
To avoid bias related to pre-existing disease impacting diet, we excluded from all analyses women who, at baseline, self-reported a history of physician-diagnosed hypertension (n = 15,390), heart attack (n = 498), angina or other heart disease (n = 1,234), or diabetes (n = 887). In addition, we also excluded those with more than 29 blank items on the food-frequency questionnaire or with unusual energy intake of <500 or ≥5000 kcal (n = 1,626) and those with missing covariates (n = 1,208). A total of 20,993 women at risk remained.
The DASH diet score was divided into five ascending categories on an ordinal scale. Means or prevalences of baseline characteristics were computed for each category. The associations of the DASH diet index with incidence rates of hypertension and mortality from cardiovascular disease were examined using Cox proportional hazards regression. Trends in hazard ratios across categories, designated by their median values, were tested by a chi-squared statistic. Analyses were performed with SAS version 8.2 (SAS Institute, Inc., Cary, NC).
Table 1 shows that at baseline at least 30 percent of women met the DASH diet guidelines for whole grains, vegetables, dairy foods, and meats. However, less than 25 percent of women achieved the highest score for total grain, fruits, fats, or sodium. The mean (SD) energy intake estimated from the food frequency questionnaire was 1812 (608) kcal/d.
No woman complied with all 11 DASH diet guidelines. The observed range of concordance score was from 0.5 to 10.0 (Table 3). Women in the higher categories of the DASH score tended to have a lower mean BMI, lower smoking prevalence, and higher levels of physical activity and multivitamin use. As expected, women in the higher versus lower DASH score categories also tended to have greater intake of healthful foods and nutrients. Unexpectedly, sodium and energy intakes increased modestly across DASH score categories.
Over the follow-up period, 19 percent of women reported a new physician diagnosis of hypertension, 2 percent died from coronary heart disease, 1 percent died from stroke, and 4 percent died from all cardiovascular diseases. As shown in Table 4, in the model adjusted for age and energy intake, incidence of hypertension was inversely associated with the degree of concordance with the DASH diet, with hazard ratios across categories of 1.0, 0.91, 0.95, 0.99, and 0.87 (p trend = 0.02). There also were inverse, but not monotonic, associations between better DASH diet concordance and mortality from coronary heart disease, stroke, and all CVD. However, after adjustment for other risk factors, there was little evidence that the incidence of hypertension, or mortality from coronary heart disease, stroke or total CVD, was associated with the DASH diet score. Hazard ratios in Model 2 (HR2) for all four endpoints were close to 1.0 for most categories of DASH diet score (Table 4).
We conducted several supplemental analyses. Firstly, we divided the upper category of the DASH score and found no enhanced benefit for being in the upper 12.5 percentile of greatest concordance with the diet. Secondly, we reran Model 2 after excluding BMI, and our findings were unchanged. Thirdly, we examined the relation of every individual component in Table 1 with incident hypertension. Although many of those components showed some age-adjusted associations with hypertension incidence in the expected direction, after multivariate analysis, only dairy intake (inversely) and saturated fat (inversely – opposite expectation) were associated statistically (p trend < 0.05) significantly with hypertension incidence. The hazard ratios of incident hypertension across five ascending categories of DASH concordance were 1.0, 1.01, 0.96, 0.94, and 0.88 (p trend = 0.008) for dairy and 1.0, 0.96, 0.93, 0.92, and 0.87 (p trend = 0.006) for saturated fat. The adjusted hazard ratios of incident hypertension across ascending sodium intake quintiles were 1.0, 0.89, 1.04, 0.97, and 1.03 (p trend = 0.42). In a final supplemental analysis, we added back for mortality analyses the women who reported hypertension at baseline in 1986. Results based on this larger number of women (n = 33,082) were similar to those in Table 4.
Although randomized trials of one to two months duration have shown that the DASH diet reduces blood pressure in those with above optimal blood pressure levels,4–6 the six-month PREMIER trial did not show a clear benefit of the DASH diet beyond other lifestyle interventions.7 In this prospective study of women living in Iowa, greater concordance with DASH diet guidelines was associated with somewhat lower incidence of hypertension and mortality from coronary heart disease, stroke or total cardiovascular disease. However, these associations were eliminated after adjustment for other risk factors. To our knowledge, this is the first prospective population-based study of long-term hypertension incidence and cardiovascular mortality in relation to the degree of concordance with the DASH diet.
Three clear and potentially important differences between our study and the DASH diet clinical trials need mention. Firstly, the two-month DASH trial4,5 provided food, with high dietary compliance, and the six-month PREMIER trial7 promoted the DASH diet, with moderate compliance. In contrast, our observational study assessed how well women’s usual diets agreed with DASH guidelines. Yet, even women in the highest 12.5 percent on the DASH score did not have reduced risk of study endpoints over the more than a decade of follow-up. Secondly, many of the IWHS participants studied would have had initially normal, not above optimal, blood pressure levels. It is likely that the long-term effect of the DASH diet differs by initial level of blood pressure.4,5 Blood pressure was not measured in the IWHS. Thirdly, the IWHS cohort was almost completely Caucasian, whereas the DASH trials over-represented African-Americans.
In this cohort study, unlike in a randomized trial, dietary intake also was self-selected. We adjusted the association between diet and endpoints for many baseline characteristics, but residual confounding by unmeasured variables related to diet choice is likely. Nevertheless, such uncontrolled confounding probably would have yielded an inverse association between diet concordance and cardiovascular endpoints, rather than no association, because better DASH diet concordance tended to correlate with healthier lifestyles (Table 3).
Diet was assessed in this cohort by a semiquantative food frequency questionnaire. Although the validity of this questionnaire has been documented,12,13 diet assessment is typically imprecise and energy intake is often underestimated. Under-reporting could affect the absolute DASH diet score and estimated concordance with it, but women’s concordance level should have been ranked appropriately.
The DASH Diet Index developed for this study has limitations. The DASH diet guidelines were not easy to operationalize and assumptions were made in developing the scoring system. For example, although the number of food groups recommended for an individual following the DASH diet are based on daily calorie needs, the index was based on recommendations for one calorie level (2,000 calorie plan). This approach was required because the daily calorie needs of participants in our study were not known. Although DASH scores generally corresponded with food and nutrient intakes in expected directions (Table 3), sodium intake unexpectedly increased across DASH score categories. This finding suggests that in this population of older women consumption of a diet that corresponds with most components of the DASH diet tended to be higher in sodium. Consequently, we could not evaluate the DASH diet fully, because the low-sodium component of the diet was not represented in the context of the other DASH diet components.
Random measurement error in DASH diet classification, or diet changes over time, likely would bias hazard ratios toward the null and thus could have obscured relations between DASH diet concordance and lower incident hypertension and cardiovascular mortality endpoints. Few women in this general population sample met a high level of concordance with DASH diet guidelines, but still with our large sample size statistical power was adequate to detect moderate associations between the concordance score and study endpoints. Because of the nature of our study population, our findings may not be generalizable to men or younger women.
When we examined individual components of the DASH diet, only greater intakes of dairy and saturated fat were statistically significantly associated with incident hypertension in multivariable models. Thus, on the whole, diet composition was not strongly related to hypertension occurrence in this cohort. However, the precision of a single food frequency questionnaire is low enough that some real but modest associations may have been missed.
Another shortcoming of our study was the reliance on self-reports for hypertension incidence. Previous evidence suggests hypertension reporting in this cohort is moderately accurate (see Methods) and associated positively with BMI, waist/hip ratio, and physical inactivity,18 but error in classifying hypertension could have masked a real association between the degree of DASH diet concordance and hypertension. Reliance on mortality for coronary heart disease, stroke, and total cardiovascular disease endpoints is only a modest shortcoming, as associations of risk factors with cardiovascular disease tend to be similar for mortality and incidence.
In summary, this report shows little evidence that greater concordance with DASH diet guidelines had an independent long-term impact on incidence of hypertension or mortality from cardiovascular disease in a nonhypertensive sample of older women. It may be true that the DASH diet really does not offer long term cardiovascular benefit. However, unlike the trials,4,5,7 few IWHS women achieved high concordance. Therefore, alternatively, it may be that very high concordance is necessary to achieve the benefits of the DASH diet. A long-term clinical trial would be needed to verify this, but may not be feasible.
Sources of Support: This work was supported by National Cancer Institute Grant CA39742.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.